xref: /freebsd/sys/netinet6/ip6_input.c (revision 71099ec5097cff9b4a566e5474b7f214bd539e8a)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/sdt.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #include <net/pfil.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_kdtrace.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/in_systm.h>
100 #include <net/if_llatbl.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/nd6.h>
113 
114 #ifdef IPSEC
115 #include <netipsec/ipsec.h>
116 #include <netinet6/ip6_ipsec.h>
117 #include <netipsec/ipsec6.h>
118 #endif /* IPSEC */
119 
120 #include <netinet6/ip6protosw.h>
121 
122 extern struct domain inet6domain;
123 
124 u_char ip6_protox[IPPROTO_MAX];
125 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
126 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
127 VNET_DEFINE(u_long, in6_ifaddrhmask);
128 
129 static struct netisr_handler ip6_nh = {
130 	.nh_name = "ip6",
131 	.nh_handler = ip6_input,
132 	.nh_proto = NETISR_IPV6,
133 	.nh_policy = NETISR_POLICY_FLOW,
134 };
135 
136 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
137 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
138 
139 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
140 
141 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
142 VNET_PCPUSTAT_SYSINIT(ip6stat);
143 #ifdef VIMAGE
144 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
145 #endif /* VIMAGE */
146 
147 struct rwlock in6_ifaddr_lock;
148 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
149 
150 static void ip6_init2(void *);
151 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
152 static struct ip6aux *ip6_addaux(struct mbuf *);
153 static struct ip6aux *ip6_findaux(struct mbuf *m);
154 static void ip6_delaux (struct mbuf *);
155 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
156 #ifdef PULLDOWN_TEST
157 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
158 #endif
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
165 ip6_init(void)
166 {
167 	struct ip6protosw *pr;
168 	int i;
169 
170 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
171 	    &V_ip6_auto_linklocal);
172 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
173 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
174 
175 	TAILQ_INIT(&V_in6_ifaddrhead);
176 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
177 	    &V_in6_ifaddrhmask);
178 
179 	/* Initialize packet filter hooks. */
180 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
181 	V_inet6_pfil_hook.ph_af = AF_INET6;
182 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
183 		printf("%s: WARNING: unable to register pfil hook, "
184 			"error %d\n", __func__, i);
185 
186 	scope6_init();
187 	addrsel_policy_init();
188 	nd6_init();
189 	frag6_init();
190 
191 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
192 
193 	/* Skip global initialization stuff for non-default instances. */
194 	if (!IS_DEFAULT_VNET(curvnet))
195 		return;
196 
197 #ifdef DIAGNOSTIC
198 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
199 		panic("sizeof(protosw) != sizeof(ip6protosw)");
200 #endif
201 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
202 	if (pr == NULL)
203 		panic("ip6_init");
204 
205 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
206 	for (i = 0; i < IPPROTO_MAX; i++)
207 		ip6_protox[i] = pr - inet6sw;
208 	/*
209 	 * Cycle through IP protocols and put them into the appropriate place
210 	 * in ip6_protox[].
211 	 */
212 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
213 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
214 		if (pr->pr_domain->dom_family == PF_INET6 &&
215 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
216 			/* Be careful to only index valid IP protocols. */
217 			if (pr->pr_protocol < IPPROTO_MAX)
218 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
219 		}
220 
221 	netisr_register(&ip6_nh);
222 }
223 
224 /*
225  * The protocol to be inserted into ip6_protox[] must be already registered
226  * in inet6sw[], either statically or through pf_proto_register().
227  */
228 int
229 ip6proto_register(short ip6proto)
230 {
231 	struct ip6protosw *pr;
232 
233 	/* Sanity checks. */
234 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
235 		return (EPROTONOSUPPORT);
236 
237 	/*
238 	 * The protocol slot must not be occupied by another protocol
239 	 * already.  An index pointing to IPPROTO_RAW is unused.
240 	 */
241 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
242 	if (pr == NULL)
243 		return (EPFNOSUPPORT);
244 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
245 		return (EEXIST);
246 
247 	/*
248 	 * Find the protocol position in inet6sw[] and set the index.
249 	 */
250 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
251 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
252 		if (pr->pr_domain->dom_family == PF_INET6 &&
253 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
254 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
255 			return (0);
256 		}
257 	}
258 	return (EPROTONOSUPPORT);
259 }
260 
261 int
262 ip6proto_unregister(short ip6proto)
263 {
264 	struct ip6protosw *pr;
265 
266 	/* Sanity checks. */
267 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
268 		return (EPROTONOSUPPORT);
269 
270 	/* Check if the protocol was indeed registered. */
271 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
272 	if (pr == NULL)
273 		return (EPFNOSUPPORT);
274 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
275 		return (ENOENT);
276 
277 	/* Reset the protocol slot to IPPROTO_RAW. */
278 	ip6_protox[ip6proto] = pr - inet6sw;
279 	return (0);
280 }
281 
282 #ifdef VIMAGE
283 void
284 ip6_destroy()
285 {
286 	int i;
287 
288 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
289 		printf("%s: WARNING: unable to unregister pfil hook, "
290 		    "error %d\n", __func__, i);
291 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
292 	nd6_destroy();
293 	callout_drain(&V_in6_tmpaddrtimer_ch);
294 }
295 #endif
296 
297 static int
298 ip6_init2_vnet(const void *unused __unused)
299 {
300 
301 	/* nd6_timer_init */
302 	callout_init(&V_nd6_timer_ch, 0);
303 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
304 
305 	/* timer for regeneranation of temporary addresses randomize ID */
306 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
307 	callout_reset(&V_in6_tmpaddrtimer_ch,
308 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
309 		       V_ip6_temp_regen_advance) * hz,
310 		      in6_tmpaddrtimer, curvnet);
311 
312 	return (0);
313 }
314 
315 static void
316 ip6_init2(void *dummy)
317 {
318 
319 	ip6_init2_vnet(NULL);
320 }
321 
322 /* cheat */
323 /* This must be after route_init(), which is now SI_ORDER_THIRD */
324 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
325 
326 static int
327 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
328     int *nxt, int *ours)
329 {
330 	struct ip6_hdr *ip6;
331 	struct ip6_hbh *hbh;
332 
333 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
334 #if 0	/*touches NULL pointer*/
335 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
336 #endif
337 		goto out;	/* m have already been freed */
338 	}
339 
340 	/* adjust pointer */
341 	ip6 = mtod(m, struct ip6_hdr *);
342 
343 	/*
344 	 * if the payload length field is 0 and the next header field
345 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
346 	 * option MUST be included.
347 	 */
348 	if (ip6->ip6_plen == 0 && *plen == 0) {
349 		/*
350 		 * Note that if a valid jumbo payload option is
351 		 * contained, ip6_hopopts_input() must set a valid
352 		 * (non-zero) payload length to the variable plen.
353 		 */
354 		IP6STAT_INC(ip6s_badoptions);
355 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
356 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
357 		icmp6_error(m, ICMP6_PARAM_PROB,
358 			    ICMP6_PARAMPROB_HEADER,
359 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
360 		goto out;
361 	}
362 #ifndef PULLDOWN_TEST
363 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
364 	hbh = (struct ip6_hbh *)(ip6 + 1);
365 #else
366 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
367 		sizeof(struct ip6_hbh));
368 	if (hbh == NULL) {
369 		IP6STAT_INC(ip6s_tooshort);
370 		goto out;
371 	}
372 #endif
373 	*nxt = hbh->ip6h_nxt;
374 
375 	/*
376 	 * If we are acting as a router and the packet contains a
377 	 * router alert option, see if we know the option value.
378 	 * Currently, we only support the option value for MLD, in which
379 	 * case we should pass the packet to the multicast routing
380 	 * daemon.
381 	 */
382 	if (*rtalert != ~0) {
383 		switch (*rtalert) {
384 		case IP6OPT_RTALERT_MLD:
385 			if (V_ip6_forwarding)
386 				*ours = 1;
387 			break;
388 		default:
389 			/*
390 			 * RFC2711 requires unrecognized values must be
391 			 * silently ignored.
392 			 */
393 			break;
394 		}
395 	}
396 
397 	return (0);
398 
399 out:
400 	return (1);
401 }
402 
403 void
404 ip6_input(struct mbuf *m)
405 {
406 	struct ip6_hdr *ip6;
407 	int off = sizeof(struct ip6_hdr), nest;
408 	u_int32_t plen;
409 	u_int32_t rtalert = ~0;
410 	int nxt, ours = 0;
411 	struct ifnet *deliverifp = NULL, *ifp = NULL;
412 	struct in6_addr odst;
413 	struct route_in6 rin6;
414 	int srcrt = 0;
415 	struct llentry *lle = NULL;
416 	struct sockaddr_in6 dst6, *dst;
417 
418 	bzero(&rin6, sizeof(struct route_in6));
419 #ifdef IPSEC
420 	/*
421 	 * should the inner packet be considered authentic?
422 	 * see comment in ah4_input().
423 	 * NB: m cannot be NULL when passed to the input routine
424 	 */
425 
426 	m->m_flags &= ~M_AUTHIPHDR;
427 	m->m_flags &= ~M_AUTHIPDGM;
428 
429 #endif /* IPSEC */
430 
431 	/*
432 	 * make sure we don't have onion peering information into m_tag.
433 	 */
434 	ip6_delaux(m);
435 
436 	if (m->m_flags & M_FASTFWD_OURS) {
437 		/*
438 		 * Firewall changed destination to local.
439 		 */
440 		m->m_flags &= ~M_FASTFWD_OURS;
441 		ours = 1;
442 		deliverifp = m->m_pkthdr.rcvif;
443 		ip6 = mtod(m, struct ip6_hdr *);
444 		goto hbhcheck;
445 	}
446 
447 	/*
448 	 * mbuf statistics
449 	 */
450 	if (m->m_flags & M_EXT) {
451 		if (m->m_next)
452 			IP6STAT_INC(ip6s_mext2m);
453 		else
454 			IP6STAT_INC(ip6s_mext1);
455 	} else {
456 		if (m->m_next) {
457 			if (m->m_flags & M_LOOP) {
458 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
459 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
460 				IP6STAT_INC(
461 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
462 			else
463 				IP6STAT_INC(ip6s_m2m[0]);
464 		} else
465 			IP6STAT_INC(ip6s_m1);
466 	}
467 
468 	/* drop the packet if IPv6 operation is disabled on the IF */
469 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
470 		m_freem(m);
471 		return;
472 	}
473 
474 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
475 	IP6STAT_INC(ip6s_total);
476 
477 #ifndef PULLDOWN_TEST
478 	/*
479 	 * L2 bridge code and some other code can return mbuf chain
480 	 * that does not conform to KAME requirement.  too bad.
481 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
482 	 */
483 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
484 		struct mbuf *n;
485 
486 		if (m->m_pkthdr.len > MHLEN)
487 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
488 		else
489 			n = m_gethdr(M_NOWAIT, MT_DATA);
490 		if (n == NULL) {
491 			m_freem(m);
492 			return;	/* ENOBUFS */
493 		}
494 
495 		m_move_pkthdr(n, m);
496 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
497 		n->m_len = n->m_pkthdr.len;
498 		m_freem(m);
499 		m = n;
500 	}
501 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
502 #endif
503 
504 	if (m->m_len < sizeof(struct ip6_hdr)) {
505 		struct ifnet *inifp;
506 		inifp = m->m_pkthdr.rcvif;
507 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
508 			IP6STAT_INC(ip6s_toosmall);
509 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
510 			return;
511 		}
512 	}
513 
514 	ip6 = mtod(m, struct ip6_hdr *);
515 
516 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
517 		IP6STAT_INC(ip6s_badvers);
518 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
519 		goto bad;
520 	}
521 
522 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
523 
524 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
525 
526 	/*
527 	 * Check against address spoofing/corruption.
528 	 */
529 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
530 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
531 		/*
532 		 * XXX: "badscope" is not very suitable for a multicast source.
533 		 */
534 		IP6STAT_INC(ip6s_badscope);
535 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
536 		goto bad;
537 	}
538 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
539 	    !(m->m_flags & M_LOOP)) {
540 		/*
541 		 * In this case, the packet should come from the loopback
542 		 * interface.  However, we cannot just check the if_flags,
543 		 * because ip6_mloopback() passes the "actual" interface
544 		 * as the outgoing/incoming interface.
545 		 */
546 		IP6STAT_INC(ip6s_badscope);
547 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
548 		goto bad;
549 	}
550 
551 #ifdef ALTQ
552 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
553 		/* packet is dropped by traffic conditioner */
554 		return;
555 	}
556 #endif
557 	/*
558 	 * The following check is not documented in specs.  A malicious
559 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
560 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
561 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
562 	 *
563 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
564 	 * support IPv4-less kernel compilation, we cannot support SIIT
565 	 * environment at all.  So, it makes more sense for us to reject any
566 	 * malicious packets for non-SIIT environment, than try to do a
567 	 * partial support for SIIT environment.
568 	 */
569 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
570 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
571 		IP6STAT_INC(ip6s_badscope);
572 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
573 		goto bad;
574 	}
575 #if 0
576 	/*
577 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
578 	 *
579 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
580 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
581 	 * is revised to forbid relaying case.
582 	 */
583 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
584 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
585 		IP6STAT_INC(ip6s_badscope);
586 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
587 		goto bad;
588 	}
589 #endif
590 #ifdef IPSEC
591 	/*
592 	 * Bypass packet filtering for packets previously handled by IPsec.
593 	 */
594 	if (ip6_ipsec_filtertunnel(m))
595 		goto passin;
596 #endif /* IPSEC */
597 
598 	/*
599 	 * Run through list of hooks for input packets.
600 	 *
601 	 * NB: Beware of the destination address changing
602 	 *     (e.g. by NAT rewriting).  When this happens,
603 	 *     tell ip6_forward to do the right thing.
604 	 */
605 	odst = ip6->ip6_dst;
606 
607 	/* Jump over all PFIL processing if hooks are not active. */
608 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
609 		goto passin;
610 
611 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
612 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
613 		return;
614 	if (m == NULL)			/* consumed by filter */
615 		return;
616 	ip6 = mtod(m, struct ip6_hdr *);
617 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
618 
619 	if (m->m_flags & M_FASTFWD_OURS) {
620 		m->m_flags &= ~M_FASTFWD_OURS;
621 		ours = 1;
622 		deliverifp = m->m_pkthdr.rcvif;
623 		goto hbhcheck;
624 	}
625 	if ((m->m_flags & M_IP6_NEXTHOP) &&
626 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
627 		/*
628 		 * Directly ship the packet on.  This allows forwarding
629 		 * packets originally destined to us to some other directly
630 		 * connected host.
631 		 */
632 		ip6_forward(m, 1);
633 		goto out;
634 	}
635 
636 passin:
637 	/*
638 	 * Disambiguate address scope zones (if there is ambiguity).
639 	 * We first make sure that the original source or destination address
640 	 * is not in our internal form for scoped addresses.  Such addresses
641 	 * are not necessarily invalid spec-wise, but we cannot accept them due
642 	 * to the usage conflict.
643 	 * in6_setscope() then also checks and rejects the cases where src or
644 	 * dst are the loopback address and the receiving interface
645 	 * is not loopback.
646 	 */
647 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
648 		IP6STAT_INC(ip6s_badscope); /* XXX */
649 		goto bad;
650 	}
651 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
652 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
653 		IP6STAT_INC(ip6s_badscope);
654 		goto bad;
655 	}
656 
657 	/*
658 	 * Multicast check. Assume packet is for us to avoid
659 	 * prematurely taking locks.
660 	 */
661 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
662 		ours = 1;
663 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
664 		deliverifp = m->m_pkthdr.rcvif;
665 		goto hbhcheck;
666 	}
667 
668 	/*
669 	 *  Unicast check
670 	 */
671 
672 	bzero(&dst6, sizeof(dst6));
673 	dst6.sin6_family = AF_INET6;
674 	dst6.sin6_len = sizeof(struct sockaddr_in6);
675 	dst6.sin6_addr = ip6->ip6_dst;
676 	ifp = m->m_pkthdr.rcvif;
677 	IF_AFDATA_RLOCK(ifp);
678 	lle = lla_lookup(LLTABLE6(ifp), 0,
679 	     (struct sockaddr *)&dst6);
680 	IF_AFDATA_RUNLOCK(ifp);
681 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
682 		struct ifaddr *ifa;
683 		struct in6_ifaddr *ia6;
684 		int bad;
685 
686 		bad = 1;
687 #define	sa_equal(a1, a2)						\
688 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
689 		IF_ADDR_RLOCK(ifp);
690 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
691 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
692 				continue;
693 			if (sa_equal(&dst6, ifa->ifa_addr))
694 				break;
695 		}
696 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
697 		    __func__, lle));
698 #undef sa_equal
699 
700 		ia6 = (struct in6_ifaddr *)ifa;
701 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
702 			/* Count the packet in the ip address stats */
703 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
704 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
705 			    m->m_pkthdr.len);
706 
707 			/*
708 			 * record address information into m_tag.
709 			 */
710 			(void)ip6_setdstifaddr(m, ia6);
711 
712 			bad = 0;
713 		} else {
714 			char ip6bufs[INET6_ADDRSTRLEN];
715 			char ip6bufd[INET6_ADDRSTRLEN];
716 			/* address is not ready, so discard the packet. */
717 			nd6log((LOG_INFO,
718 			    "ip6_input: packet to an unready address %s->%s\n",
719 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
720 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
721 		}
722 		IF_ADDR_RUNLOCK(ifp);
723 		LLE_RUNLOCK(lle);
724 		if (bad)
725 			goto bad;
726 		else {
727 			ours = 1;
728 			deliverifp = ifp;
729 			goto hbhcheck;
730 		}
731 	}
732 	if (lle != NULL)
733 		LLE_RUNLOCK(lle);
734 
735 	dst = &rin6.ro_dst;
736 	dst->sin6_len = sizeof(struct sockaddr_in6);
737 	dst->sin6_family = AF_INET6;
738 	dst->sin6_addr = ip6->ip6_dst;
739 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
740 	if (rin6.ro_rt)
741 		RT_UNLOCK(rin6.ro_rt);
742 
743 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
744 
745 	/*
746 	 * Accept the packet if the forwarding interface to the destination
747 	 * according to the routing table is the loopback interface,
748 	 * unless the associated route has a gateway.
749 	 * Note that this approach causes to accept a packet if there is a
750 	 * route to the loopback interface for the destination of the packet.
751 	 * But we think it's even useful in some situations, e.g. when using
752 	 * a special daemon which wants to intercept the packet.
753 	 *
754 	 * XXX: some OSes automatically make a cloned route for the destination
755 	 * of an outgoing packet.  If the outgoing interface of the packet
756 	 * is a loopback one, the kernel would consider the packet to be
757 	 * accepted, even if we have no such address assinged on the interface.
758 	 * We check the cloned flag of the route entry to reject such cases,
759 	 * assuming that route entries for our own addresses are not made by
760 	 * cloning (it should be true because in6_addloop explicitly installs
761 	 * the host route).  However, we might have to do an explicit check
762 	 * while it would be less efficient.  Or, should we rather install a
763 	 * reject route for such a case?
764 	 */
765 	if (rin6.ro_rt &&
766 	    (rin6.ro_rt->rt_flags &
767 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
768 #ifdef RTF_WASCLONED
769 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
770 #endif
771 #ifdef RTF_CLONED
772 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
773 #endif
774 #if 0
775 	    /*
776 	     * The check below is redundant since the comparison of
777 	     * the destination and the key of the rtentry has
778 	     * already done through looking up the routing table.
779 	     */
780 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
781 	    &rt6_key(rin6.ro_rt)->sin6_addr)
782 #endif
783 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
784 		int free_ia6 = 0;
785 		struct in6_ifaddr *ia6;
786 
787 		/*
788 		 * found the loopback route to the interface address
789 		 */
790 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
791 			struct sockaddr_in6 dest6;
792 
793 			bzero(&dest6, sizeof(dest6));
794 			dest6.sin6_family = AF_INET6;
795 			dest6.sin6_len = sizeof(dest6);
796 			dest6.sin6_addr = ip6->ip6_dst;
797 			ia6 = (struct in6_ifaddr *)
798 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
799 			if (ia6 == NULL)
800 				goto bad;
801 			free_ia6 = 1;
802 		}
803 		else
804 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
805 
806 		/*
807 		 * record address information into m_tag.
808 		 */
809 		(void)ip6_setdstifaddr(m, ia6);
810 
811 		/*
812 		 * packets to a tentative, duplicated, or somehow invalid
813 		 * address must not be accepted.
814 		 */
815 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
816 			/* this address is ready */
817 			ours = 1;
818 			deliverifp = ia6->ia_ifp;	/* correct? */
819 			/* Count the packet in the ip address stats */
820 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
821 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
822 			    m->m_pkthdr.len);
823 			if (free_ia6)
824 				ifa_free(&ia6->ia_ifa);
825 			goto hbhcheck;
826 		} else {
827 			char ip6bufs[INET6_ADDRSTRLEN];
828 			char ip6bufd[INET6_ADDRSTRLEN];
829 			/* address is not ready, so discard the packet. */
830 			nd6log((LOG_INFO,
831 			    "ip6_input: packet to an unready address %s->%s\n",
832 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
833 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
834 
835 			if (free_ia6)
836 				ifa_free(&ia6->ia_ifa);
837 			goto bad;
838 		}
839 	}
840 
841 	/*
842 	 * FAITH (Firewall Aided Internet Translator)
843 	 */
844 	if (V_ip6_keepfaith) {
845 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
846 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
847 			/* XXX do we need more sanity checks? */
848 			ours = 1;
849 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
850 			goto hbhcheck;
851 		}
852 	}
853 
854 	/*
855 	 * Now there is no reason to process the packet if it's not our own
856 	 * and we're not a router.
857 	 */
858 	if (!V_ip6_forwarding) {
859 		IP6STAT_INC(ip6s_cantforward);
860 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
861 		goto bad;
862 	}
863 
864   hbhcheck:
865 	/*
866 	 * record address information into m_tag, if we don't have one yet.
867 	 * note that we are unable to record it, if the address is not listed
868 	 * as our interface address (e.g. multicast addresses, addresses
869 	 * within FAITH prefixes and such).
870 	 */
871 	if (deliverifp) {
872 		struct in6_ifaddr *ia6;
873 
874  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
875 			ifa_free(&ia6->ia_ifa);
876 		} else {
877 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
878 			if (ia6) {
879 				if (!ip6_setdstifaddr(m, ia6)) {
880 					/*
881 					 * XXX maybe we should drop the packet here,
882 					 * as we could not provide enough information
883 					 * to the upper layers.
884 					 */
885 				}
886 				ifa_free(&ia6->ia_ifa);
887 			}
888 		}
889 	}
890 
891 	/*
892 	 * Process Hop-by-Hop options header if it's contained.
893 	 * m may be modified in ip6_hopopts_input().
894 	 * If a JumboPayload option is included, plen will also be modified.
895 	 */
896 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
897 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
898 		int error;
899 
900 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
901 		if (error != 0)
902 			goto out;
903 	} else
904 		nxt = ip6->ip6_nxt;
905 
906 	/*
907 	 * Check that the amount of data in the buffers
908 	 * is as at least much as the IPv6 header would have us expect.
909 	 * Trim mbufs if longer than we expect.
910 	 * Drop packet if shorter than we expect.
911 	 */
912 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
913 		IP6STAT_INC(ip6s_tooshort);
914 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
915 		goto bad;
916 	}
917 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
918 		if (m->m_len == m->m_pkthdr.len) {
919 			m->m_len = sizeof(struct ip6_hdr) + plen;
920 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
921 		} else
922 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
923 	}
924 
925 	/*
926 	 * Forward if desirable.
927 	 */
928 	if (V_ip6_mrouter &&
929 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
930 		/*
931 		 * If we are acting as a multicast router, all
932 		 * incoming multicast packets are passed to the
933 		 * kernel-level multicast forwarding function.
934 		 * The packet is returned (relatively) intact; if
935 		 * ip6_mforward() returns a non-zero value, the packet
936 		 * must be discarded, else it may be accepted below.
937 		 *
938 		 * XXX TODO: Check hlim and multicast scope here to avoid
939 		 * unnecessarily calling into ip6_mforward().
940 		 */
941 		if (ip6_mforward &&
942 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
943 			IP6STAT_INC(ip6s_cantforward);
944 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
945 			goto bad;
946 		}
947 	} else if (!ours) {
948 		ip6_forward(m, srcrt);
949 		goto out;
950 	}
951 
952 	ip6 = mtod(m, struct ip6_hdr *);
953 
954 	/*
955 	 * Malicious party may be able to use IPv4 mapped addr to confuse
956 	 * tcp/udp stack and bypass security checks (act as if it was from
957 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
958 	 *
959 	 * For SIIT end node behavior, you may want to disable the check.
960 	 * However, you will  become vulnerable to attacks using IPv4 mapped
961 	 * source.
962 	 */
963 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
964 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
965 		IP6STAT_INC(ip6s_badscope);
966 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
967 		goto bad;
968 	}
969 
970 	/*
971 	 * Tell launch routine the next header
972 	 */
973 	IP6STAT_INC(ip6s_delivered);
974 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
975 	nest = 0;
976 
977 	while (nxt != IPPROTO_DONE) {
978 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
979 			IP6STAT_INC(ip6s_toomanyhdr);
980 			goto bad;
981 		}
982 
983 		/*
984 		 * protection against faulty packet - there should be
985 		 * more sanity checks in header chain processing.
986 		 */
987 		if (m->m_pkthdr.len < off) {
988 			IP6STAT_INC(ip6s_tooshort);
989 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
990 			goto bad;
991 		}
992 
993 #ifdef IPSEC
994 		/*
995 		 * enforce IPsec policy checking if we are seeing last header.
996 		 * note that we do not visit this with protocols with pcb layer
997 		 * code - like udp/tcp/raw ip.
998 		 */
999 		if (ip6_ipsec_input(m, nxt))
1000 			goto bad;
1001 #endif /* IPSEC */
1002 
1003 		/*
1004 		 * Use mbuf flags to propagate Router Alert option to
1005 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1006 		 */
1007 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1008 			m->m_flags |= M_RTALERT_MLD;
1009 
1010 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1011 	}
1012 	goto out;
1013 bad:
1014 	m_freem(m);
1015 out:
1016 	if (rin6.ro_rt)
1017 		RTFREE(rin6.ro_rt);
1018 }
1019 
1020 /*
1021  * set/grab in6_ifaddr correspond to IPv6 destination address.
1022  * XXX backward compatibility wrapper
1023  *
1024  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1025  * and then bump it when the tag is copied, and release it when the tag is
1026  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1027  * we just bump the ia refcount when we receive it.  This should be fixed.
1028  */
1029 static struct ip6aux *
1030 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1031 {
1032 	struct ip6aux *ip6a;
1033 
1034 	ip6a = ip6_addaux(m);
1035 	if (ip6a)
1036 		ip6a->ip6a_dstia6 = ia6;
1037 	return ip6a;	/* NULL if failed to set */
1038 }
1039 
1040 struct in6_ifaddr *
1041 ip6_getdstifaddr(struct mbuf *m)
1042 {
1043 	struct ip6aux *ip6a;
1044 	struct in6_ifaddr *ia;
1045 
1046 	ip6a = ip6_findaux(m);
1047 	if (ip6a) {
1048 		ia = ip6a->ip6a_dstia6;
1049 		ifa_ref(&ia->ia_ifa);
1050 		return ia;
1051 	} else
1052 		return NULL;
1053 }
1054 
1055 /*
1056  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1057  * included, the real payload length will be stored in plenp.
1058  *
1059  * rtalertp - XXX: should be stored more smart way
1060  */
1061 static int
1062 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1063     struct mbuf **mp, int *offp)
1064 {
1065 	struct mbuf *m = *mp;
1066 	int off = *offp, hbhlen;
1067 	struct ip6_hbh *hbh;
1068 	u_int8_t *opt;
1069 
1070 	/* validation of the length of the header */
1071 #ifndef PULLDOWN_TEST
1072 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1073 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1074 	hbhlen = (hbh->ip6h_len + 1) << 3;
1075 
1076 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1077 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1078 #else
1079 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1080 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1081 	if (hbh == NULL) {
1082 		IP6STAT_INC(ip6s_tooshort);
1083 		return -1;
1084 	}
1085 	hbhlen = (hbh->ip6h_len + 1) << 3;
1086 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1087 		hbhlen);
1088 	if (hbh == NULL) {
1089 		IP6STAT_INC(ip6s_tooshort);
1090 		return -1;
1091 	}
1092 #endif
1093 	off += hbhlen;
1094 	hbhlen -= sizeof(struct ip6_hbh);
1095 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1096 
1097 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1098 				hbhlen, rtalertp, plenp) < 0)
1099 		return (-1);
1100 
1101 	*offp = off;
1102 	*mp = m;
1103 	return (0);
1104 }
1105 
1106 /*
1107  * Search header for all Hop-by-hop options and process each option.
1108  * This function is separate from ip6_hopopts_input() in order to
1109  * handle a case where the sending node itself process its hop-by-hop
1110  * options header. In such a case, the function is called from ip6_output().
1111  *
1112  * The function assumes that hbh header is located right after the IPv6 header
1113  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1114  * opthead + hbhlen is located in contiguous memory region.
1115  */
1116 int
1117 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1118     u_int32_t *rtalertp, u_int32_t *plenp)
1119 {
1120 	struct ip6_hdr *ip6;
1121 	int optlen = 0;
1122 	u_int8_t *opt = opthead;
1123 	u_int16_t rtalert_val;
1124 	u_int32_t jumboplen;
1125 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1126 
1127 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1128 		switch (*opt) {
1129 		case IP6OPT_PAD1:
1130 			optlen = 1;
1131 			break;
1132 		case IP6OPT_PADN:
1133 			if (hbhlen < IP6OPT_MINLEN) {
1134 				IP6STAT_INC(ip6s_toosmall);
1135 				goto bad;
1136 			}
1137 			optlen = *(opt + 1) + 2;
1138 			break;
1139 		case IP6OPT_ROUTER_ALERT:
1140 			/* XXX may need check for alignment */
1141 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1142 				IP6STAT_INC(ip6s_toosmall);
1143 				goto bad;
1144 			}
1145 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1146 				/* XXX stat */
1147 				icmp6_error(m, ICMP6_PARAM_PROB,
1148 				    ICMP6_PARAMPROB_HEADER,
1149 				    erroff + opt + 1 - opthead);
1150 				return (-1);
1151 			}
1152 			optlen = IP6OPT_RTALERT_LEN;
1153 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1154 			*rtalertp = ntohs(rtalert_val);
1155 			break;
1156 		case IP6OPT_JUMBO:
1157 			/* XXX may need check for alignment */
1158 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1159 				IP6STAT_INC(ip6s_toosmall);
1160 				goto bad;
1161 			}
1162 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1163 				/* XXX stat */
1164 				icmp6_error(m, ICMP6_PARAM_PROB,
1165 				    ICMP6_PARAMPROB_HEADER,
1166 				    erroff + opt + 1 - opthead);
1167 				return (-1);
1168 			}
1169 			optlen = IP6OPT_JUMBO_LEN;
1170 
1171 			/*
1172 			 * IPv6 packets that have non 0 payload length
1173 			 * must not contain a jumbo payload option.
1174 			 */
1175 			ip6 = mtod(m, struct ip6_hdr *);
1176 			if (ip6->ip6_plen) {
1177 				IP6STAT_INC(ip6s_badoptions);
1178 				icmp6_error(m, ICMP6_PARAM_PROB,
1179 				    ICMP6_PARAMPROB_HEADER,
1180 				    erroff + opt - opthead);
1181 				return (-1);
1182 			}
1183 
1184 			/*
1185 			 * We may see jumbolen in unaligned location, so
1186 			 * we'd need to perform bcopy().
1187 			 */
1188 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1189 			jumboplen = (u_int32_t)htonl(jumboplen);
1190 
1191 #if 1
1192 			/*
1193 			 * if there are multiple jumbo payload options,
1194 			 * *plenp will be non-zero and the packet will be
1195 			 * rejected.
1196 			 * the behavior may need some debate in ipngwg -
1197 			 * multiple options does not make sense, however,
1198 			 * there's no explicit mention in specification.
1199 			 */
1200 			if (*plenp != 0) {
1201 				IP6STAT_INC(ip6s_badoptions);
1202 				icmp6_error(m, ICMP6_PARAM_PROB,
1203 				    ICMP6_PARAMPROB_HEADER,
1204 				    erroff + opt + 2 - opthead);
1205 				return (-1);
1206 			}
1207 #endif
1208 
1209 			/*
1210 			 * jumbo payload length must be larger than 65535.
1211 			 */
1212 			if (jumboplen <= IPV6_MAXPACKET) {
1213 				IP6STAT_INC(ip6s_badoptions);
1214 				icmp6_error(m, ICMP6_PARAM_PROB,
1215 				    ICMP6_PARAMPROB_HEADER,
1216 				    erroff + opt + 2 - opthead);
1217 				return (-1);
1218 			}
1219 			*plenp = jumboplen;
1220 
1221 			break;
1222 		default:		/* unknown option */
1223 			if (hbhlen < IP6OPT_MINLEN) {
1224 				IP6STAT_INC(ip6s_toosmall);
1225 				goto bad;
1226 			}
1227 			optlen = ip6_unknown_opt(opt, m,
1228 			    erroff + opt - opthead);
1229 			if (optlen == -1)
1230 				return (-1);
1231 			optlen += 2;
1232 			break;
1233 		}
1234 	}
1235 
1236 	return (0);
1237 
1238   bad:
1239 	m_freem(m);
1240 	return (-1);
1241 }
1242 
1243 /*
1244  * Unknown option processing.
1245  * The third argument `off' is the offset from the IPv6 header to the option,
1246  * which is necessary if the IPv6 header the and option header and IPv6 header
1247  * is not contiguous in order to return an ICMPv6 error.
1248  */
1249 int
1250 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1251 {
1252 	struct ip6_hdr *ip6;
1253 
1254 	switch (IP6OPT_TYPE(*optp)) {
1255 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1256 		return ((int)*(optp + 1));
1257 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1258 		m_freem(m);
1259 		return (-1);
1260 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1261 		IP6STAT_INC(ip6s_badoptions);
1262 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1263 		return (-1);
1264 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1265 		IP6STAT_INC(ip6s_badoptions);
1266 		ip6 = mtod(m, struct ip6_hdr *);
1267 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1268 		    (m->m_flags & (M_BCAST|M_MCAST)))
1269 			m_freem(m);
1270 		else
1271 			icmp6_error(m, ICMP6_PARAM_PROB,
1272 				    ICMP6_PARAMPROB_OPTION, off);
1273 		return (-1);
1274 	}
1275 
1276 	m_freem(m);		/* XXX: NOTREACHED */
1277 	return (-1);
1278 }
1279 
1280 /*
1281  * Create the "control" list for this pcb.
1282  * These functions will not modify mbuf chain at all.
1283  *
1284  * With KAME mbuf chain restriction:
1285  * The routine will be called from upper layer handlers like tcp6_input().
1286  * Thus the routine assumes that the caller (tcp6_input) have already
1287  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1288  * very first mbuf on the mbuf chain.
1289  *
1290  * ip6_savecontrol_v4 will handle those options that are possible to be
1291  * set on a v4-mapped socket.
1292  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1293  * options and handle the v6-only ones itself.
1294  */
1295 struct mbuf **
1296 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1297     int *v4only)
1298 {
1299 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1300 
1301 #ifdef SO_TIMESTAMP
1302 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1303 		struct timeval tv;
1304 
1305 		microtime(&tv);
1306 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1307 		    SCM_TIMESTAMP, SOL_SOCKET);
1308 		if (*mp)
1309 			mp = &(*mp)->m_next;
1310 	}
1311 #endif
1312 
1313 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1314 	/* RFC 2292 sec. 5 */
1315 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1316 		struct in6_pktinfo pi6;
1317 
1318 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1319 #ifdef INET
1320 			struct ip *ip;
1321 
1322 			ip = mtod(m, struct ip *);
1323 			pi6.ipi6_addr.s6_addr32[0] = 0;
1324 			pi6.ipi6_addr.s6_addr32[1] = 0;
1325 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1326 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1327 #else
1328 			/* We won't hit this code */
1329 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1330 #endif
1331 		} else {
1332 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1333 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1334 		}
1335 		pi6.ipi6_ifindex =
1336 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1337 
1338 		*mp = sbcreatecontrol((caddr_t) &pi6,
1339 		    sizeof(struct in6_pktinfo),
1340 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1341 		if (*mp)
1342 			mp = &(*mp)->m_next;
1343 	}
1344 
1345 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1346 		int hlim;
1347 
1348 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1349 #ifdef INET
1350 			struct ip *ip;
1351 
1352 			ip = mtod(m, struct ip *);
1353 			hlim = ip->ip_ttl;
1354 #else
1355 			/* We won't hit this code */
1356 			hlim = 0;
1357 #endif
1358 		} else {
1359 			hlim = ip6->ip6_hlim & 0xff;
1360 		}
1361 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1362 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1363 		    IPPROTO_IPV6);
1364 		if (*mp)
1365 			mp = &(*mp)->m_next;
1366 	}
1367 
1368 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1369 		int tclass;
1370 
1371 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1372 #ifdef INET
1373 			struct ip *ip;
1374 
1375 			ip = mtod(m, struct ip *);
1376 			tclass = ip->ip_tos;
1377 #else
1378 			/* We won't hit this code */
1379 			tclass = 0;
1380 #endif
1381 		} else {
1382 			u_int32_t flowinfo;
1383 
1384 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1385 			flowinfo >>= 20;
1386 			tclass = flowinfo & 0xff;
1387 		}
1388 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1389 		    IPV6_TCLASS, IPPROTO_IPV6);
1390 		if (*mp)
1391 			mp = &(*mp)->m_next;
1392 	}
1393 
1394 	if (v4only != NULL) {
1395 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1396 			*v4only = 1;
1397 		} else {
1398 			*v4only = 0;
1399 		}
1400 	}
1401 
1402 	return (mp);
1403 }
1404 
1405 void
1406 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1407 {
1408 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1409 	int v4only = 0;
1410 
1411 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1412 	if (v4only)
1413 		return;
1414 
1415 	/*
1416 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1417 	 * privilege for the option (see ip6_ctloutput), but it might be too
1418 	 * strict, since there might be some hop-by-hop options which can be
1419 	 * returned to normal user.
1420 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1421 	 */
1422 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1423 		/*
1424 		 * Check if a hop-by-hop options header is contatined in the
1425 		 * received packet, and if so, store the options as ancillary
1426 		 * data. Note that a hop-by-hop options header must be
1427 		 * just after the IPv6 header, which is assured through the
1428 		 * IPv6 input processing.
1429 		 */
1430 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1431 			struct ip6_hbh *hbh;
1432 			int hbhlen = 0;
1433 #ifdef PULLDOWN_TEST
1434 			struct mbuf *ext;
1435 #endif
1436 
1437 #ifndef PULLDOWN_TEST
1438 			hbh = (struct ip6_hbh *)(ip6 + 1);
1439 			hbhlen = (hbh->ip6h_len + 1) << 3;
1440 #else
1441 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1442 			    ip6->ip6_nxt);
1443 			if (ext == NULL) {
1444 				IP6STAT_INC(ip6s_tooshort);
1445 				return;
1446 			}
1447 			hbh = mtod(ext, struct ip6_hbh *);
1448 			hbhlen = (hbh->ip6h_len + 1) << 3;
1449 			if (hbhlen != ext->m_len) {
1450 				m_freem(ext);
1451 				IP6STAT_INC(ip6s_tooshort);
1452 				return;
1453 			}
1454 #endif
1455 
1456 			/*
1457 			 * XXX: We copy the whole header even if a
1458 			 * jumbo payload option is included, the option which
1459 			 * is to be removed before returning according to
1460 			 * RFC2292.
1461 			 * Note: this constraint is removed in RFC3542
1462 			 */
1463 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1464 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1465 			    IPPROTO_IPV6);
1466 			if (*mp)
1467 				mp = &(*mp)->m_next;
1468 #ifdef PULLDOWN_TEST
1469 			m_freem(ext);
1470 #endif
1471 		}
1472 	}
1473 
1474 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1475 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1476 
1477 		/*
1478 		 * Search for destination options headers or routing
1479 		 * header(s) through the header chain, and stores each
1480 		 * header as ancillary data.
1481 		 * Note that the order of the headers remains in
1482 		 * the chain of ancillary data.
1483 		 */
1484 		while (1) {	/* is explicit loop prevention necessary? */
1485 			struct ip6_ext *ip6e = NULL;
1486 			int elen;
1487 #ifdef PULLDOWN_TEST
1488 			struct mbuf *ext = NULL;
1489 #endif
1490 
1491 			/*
1492 			 * if it is not an extension header, don't try to
1493 			 * pull it from the chain.
1494 			 */
1495 			switch (nxt) {
1496 			case IPPROTO_DSTOPTS:
1497 			case IPPROTO_ROUTING:
1498 			case IPPROTO_HOPOPTS:
1499 			case IPPROTO_AH: /* is it possible? */
1500 				break;
1501 			default:
1502 				goto loopend;
1503 			}
1504 
1505 #ifndef PULLDOWN_TEST
1506 			if (off + sizeof(*ip6e) > m->m_len)
1507 				goto loopend;
1508 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1509 			if (nxt == IPPROTO_AH)
1510 				elen = (ip6e->ip6e_len + 2) << 2;
1511 			else
1512 				elen = (ip6e->ip6e_len + 1) << 3;
1513 			if (off + elen > m->m_len)
1514 				goto loopend;
1515 #else
1516 			ext = ip6_pullexthdr(m, off, nxt);
1517 			if (ext == NULL) {
1518 				IP6STAT_INC(ip6s_tooshort);
1519 				return;
1520 			}
1521 			ip6e = mtod(ext, struct ip6_ext *);
1522 			if (nxt == IPPROTO_AH)
1523 				elen = (ip6e->ip6e_len + 2) << 2;
1524 			else
1525 				elen = (ip6e->ip6e_len + 1) << 3;
1526 			if (elen != ext->m_len) {
1527 				m_freem(ext);
1528 				IP6STAT_INC(ip6s_tooshort);
1529 				return;
1530 			}
1531 #endif
1532 
1533 			switch (nxt) {
1534 			case IPPROTO_DSTOPTS:
1535 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1536 					break;
1537 
1538 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1539 				    IS2292(in6p,
1540 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1541 				    IPPROTO_IPV6);
1542 				if (*mp)
1543 					mp = &(*mp)->m_next;
1544 				break;
1545 			case IPPROTO_ROUTING:
1546 				if (!(in6p->inp_flags & IN6P_RTHDR))
1547 					break;
1548 
1549 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1550 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1551 				    IPPROTO_IPV6);
1552 				if (*mp)
1553 					mp = &(*mp)->m_next;
1554 				break;
1555 			case IPPROTO_HOPOPTS:
1556 			case IPPROTO_AH: /* is it possible? */
1557 				break;
1558 
1559 			default:
1560 				/*
1561 				 * other cases have been filtered in the above.
1562 				 * none will visit this case.  here we supply
1563 				 * the code just in case (nxt overwritten or
1564 				 * other cases).
1565 				 */
1566 #ifdef PULLDOWN_TEST
1567 				m_freem(ext);
1568 #endif
1569 				goto loopend;
1570 
1571 			}
1572 
1573 			/* proceed with the next header. */
1574 			off += elen;
1575 			nxt = ip6e->ip6e_nxt;
1576 			ip6e = NULL;
1577 #ifdef PULLDOWN_TEST
1578 			m_freem(ext);
1579 			ext = NULL;
1580 #endif
1581 		}
1582 	  loopend:
1583 		;
1584 	}
1585 }
1586 #undef IS2292
1587 
1588 void
1589 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1590 {
1591 	struct socket *so;
1592 	struct mbuf *m_mtu;
1593 	struct ip6_mtuinfo mtuctl;
1594 
1595 	so =  in6p->inp_socket;
1596 
1597 	if (mtu == NULL)
1598 		return;
1599 
1600 #ifdef DIAGNOSTIC
1601 	if (so == NULL)		/* I believe this is impossible */
1602 		panic("ip6_notify_pmtu: socket is NULL");
1603 #endif
1604 
1605 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1606 	mtuctl.ip6m_mtu = *mtu;
1607 	mtuctl.ip6m_addr = *dst;
1608 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1609 		return;
1610 
1611 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1612 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1613 		return;
1614 
1615 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1616 	    == 0) {
1617 		m_freem(m_mtu);
1618 		/* XXX: should count statistics */
1619 	} else
1620 		sorwakeup(so);
1621 
1622 	return;
1623 }
1624 
1625 #ifdef PULLDOWN_TEST
1626 /*
1627  * pull single extension header from mbuf chain.  returns single mbuf that
1628  * contains the result, or NULL on error.
1629  */
1630 static struct mbuf *
1631 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1632 {
1633 	struct ip6_ext ip6e;
1634 	size_t elen;
1635 	struct mbuf *n;
1636 
1637 #ifdef DIAGNOSTIC
1638 	switch (nxt) {
1639 	case IPPROTO_DSTOPTS:
1640 	case IPPROTO_ROUTING:
1641 	case IPPROTO_HOPOPTS:
1642 	case IPPROTO_AH: /* is it possible? */
1643 		break;
1644 	default:
1645 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1646 	}
1647 #endif
1648 
1649 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1650 	if (nxt == IPPROTO_AH)
1651 		elen = (ip6e.ip6e_len + 2) << 2;
1652 	else
1653 		elen = (ip6e.ip6e_len + 1) << 3;
1654 
1655 	if (elen > MLEN)
1656 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1657 	else
1658 		n = m_get(M_NOWAIT, MT_DATA);
1659 	if (n == NULL)
1660 		return NULL;
1661 
1662 	m_copydata(m, off, elen, mtod(n, caddr_t));
1663 	n->m_len = elen;
1664 	return n;
1665 }
1666 #endif
1667 
1668 /*
1669  * Get pointer to the previous header followed by the header
1670  * currently processed.
1671  * XXX: This function supposes that
1672  *	M includes all headers,
1673  *	the next header field and the header length field of each header
1674  *	are valid, and
1675  *	the sum of each header length equals to OFF.
1676  * Because of these assumptions, this function must be called very
1677  * carefully. Moreover, it will not be used in the near future when
1678  * we develop `neater' mechanism to process extension headers.
1679  */
1680 char *
1681 ip6_get_prevhdr(struct mbuf *m, int off)
1682 {
1683 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1684 
1685 	if (off == sizeof(struct ip6_hdr))
1686 		return (&ip6->ip6_nxt);
1687 	else {
1688 		int len, nxt;
1689 		struct ip6_ext *ip6e = NULL;
1690 
1691 		nxt = ip6->ip6_nxt;
1692 		len = sizeof(struct ip6_hdr);
1693 		while (len < off) {
1694 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1695 
1696 			switch (nxt) {
1697 			case IPPROTO_FRAGMENT:
1698 				len += sizeof(struct ip6_frag);
1699 				break;
1700 			case IPPROTO_AH:
1701 				len += (ip6e->ip6e_len + 2) << 2;
1702 				break;
1703 			default:
1704 				len += (ip6e->ip6e_len + 1) << 3;
1705 				break;
1706 			}
1707 			nxt = ip6e->ip6e_nxt;
1708 		}
1709 		if (ip6e)
1710 			return (&ip6e->ip6e_nxt);
1711 		else
1712 			return NULL;
1713 	}
1714 }
1715 
1716 /*
1717  * get next header offset.  m will be retained.
1718  */
1719 int
1720 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1721 {
1722 	struct ip6_hdr ip6;
1723 	struct ip6_ext ip6e;
1724 	struct ip6_frag fh;
1725 
1726 	/* just in case */
1727 	if (m == NULL)
1728 		panic("ip6_nexthdr: m == NULL");
1729 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1730 		return -1;
1731 
1732 	switch (proto) {
1733 	case IPPROTO_IPV6:
1734 		if (m->m_pkthdr.len < off + sizeof(ip6))
1735 			return -1;
1736 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1737 		if (nxtp)
1738 			*nxtp = ip6.ip6_nxt;
1739 		off += sizeof(ip6);
1740 		return off;
1741 
1742 	case IPPROTO_FRAGMENT:
1743 		/*
1744 		 * terminate parsing if it is not the first fragment,
1745 		 * it does not make sense to parse through it.
1746 		 */
1747 		if (m->m_pkthdr.len < off + sizeof(fh))
1748 			return -1;
1749 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1750 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1751 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1752 			return -1;
1753 		if (nxtp)
1754 			*nxtp = fh.ip6f_nxt;
1755 		off += sizeof(struct ip6_frag);
1756 		return off;
1757 
1758 	case IPPROTO_AH:
1759 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1760 			return -1;
1761 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1762 		if (nxtp)
1763 			*nxtp = ip6e.ip6e_nxt;
1764 		off += (ip6e.ip6e_len + 2) << 2;
1765 		return off;
1766 
1767 	case IPPROTO_HOPOPTS:
1768 	case IPPROTO_ROUTING:
1769 	case IPPROTO_DSTOPTS:
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 + 1) << 3;
1776 		return off;
1777 
1778 	case IPPROTO_NONE:
1779 	case IPPROTO_ESP:
1780 	case IPPROTO_IPCOMP:
1781 		/* give up */
1782 		return -1;
1783 
1784 	default:
1785 		return -1;
1786 	}
1787 
1788 	return -1;
1789 }
1790 
1791 /*
1792  * get offset for the last header in the chain.  m will be kept untainted.
1793  */
1794 int
1795 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1796 {
1797 	int newoff;
1798 	int nxt;
1799 
1800 	if (!nxtp) {
1801 		nxt = -1;
1802 		nxtp = &nxt;
1803 	}
1804 	while (1) {
1805 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1806 		if (newoff < 0)
1807 			return off;
1808 		else if (newoff < off)
1809 			return -1;	/* invalid */
1810 		else if (newoff == off)
1811 			return newoff;
1812 
1813 		off = newoff;
1814 		proto = *nxtp;
1815 	}
1816 }
1817 
1818 static struct ip6aux *
1819 ip6_addaux(struct mbuf *m)
1820 {
1821 	struct m_tag *mtag;
1822 
1823 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1824 	if (!mtag) {
1825 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1826 		    M_NOWAIT);
1827 		if (mtag) {
1828 			m_tag_prepend(m, mtag);
1829 			bzero(mtag + 1, sizeof(struct ip6aux));
1830 		}
1831 	}
1832 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1833 }
1834 
1835 static struct ip6aux *
1836 ip6_findaux(struct mbuf *m)
1837 {
1838 	struct m_tag *mtag;
1839 
1840 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1841 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1842 }
1843 
1844 static void
1845 ip6_delaux(struct mbuf *m)
1846 {
1847 	struct m_tag *mtag;
1848 
1849 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1850 	if (mtag)
1851 		m_tag_delete(m, mtag);
1852 }
1853 
1854 /*
1855  * System control for IP6
1856  */
1857 
1858 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1859 	0,		0,		0,		0,
1860 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1861 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1862 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1863 	0,		0,		0,		0,
1864 	ENOPROTOOPT
1865 };
1866