xref: /freebsd/sys/netinet6/ip6_input.c (revision 6d732c66bca5da4d261577aad2c8ea84519b0bea)
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 #ifdef FLOWTABLE
123 #include <net/flowtable.h>
124 VNET_DECLARE(int, ip6_output_flowtable_size);
125 #define	V_ip6_output_flowtable_size	VNET(ip6_output_flowtable_size)
126 #endif
127 
128 extern struct domain inet6domain;
129 
130 u_char ip6_protox[IPPROTO_MAX];
131 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
132 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
133 VNET_DEFINE(u_long, in6_ifaddrhmask);
134 
135 static struct netisr_handler ip6_nh = {
136 	.nh_name = "ip6",
137 	.nh_handler = ip6_input,
138 	.nh_proto = NETISR_IPV6,
139 	.nh_policy = NETISR_POLICY_FLOW,
140 };
141 
142 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
143 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
144 
145 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
146 
147 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
148 VNET_PCPUSTAT_SYSINIT(ip6stat);
149 #ifdef VIMAGE
150 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
151 #endif /* VIMAGE */
152 
153 struct rwlock in6_ifaddr_lock;
154 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
155 
156 static void ip6_init2(void *);
157 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
158 static struct ip6aux *ip6_addaux(struct mbuf *);
159 static struct ip6aux *ip6_findaux(struct mbuf *m);
160 static void ip6_delaux (struct mbuf *);
161 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
162 #ifdef PULLDOWN_TEST
163 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
164 #endif
165 
166 /*
167  * IP6 initialization: fill in IP6 protocol switch table.
168  * All protocols not implemented in kernel go to raw IP6 protocol handler.
169  */
170 void
171 ip6_init(void)
172 {
173 	struct ip6protosw *pr;
174 	int i;
175 
176 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
177 	    &V_ip6_auto_linklocal);
178 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
179 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
180 
181 	TAILQ_INIT(&V_in6_ifaddrhead);
182 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
183 	    &V_in6_ifaddrhmask);
184 
185 	/* Initialize packet filter hooks. */
186 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
187 	V_inet6_pfil_hook.ph_af = AF_INET6;
188 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
189 		printf("%s: WARNING: unable to register pfil hook, "
190 			"error %d\n", __func__, i);
191 
192 	scope6_init();
193 	addrsel_policy_init();
194 	nd6_init();
195 	frag6_init();
196 
197 #ifdef FLOWTABLE
198 	if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size",
199 		&V_ip6_output_flowtable_size)) {
200 		if (V_ip6_output_flowtable_size < 256)
201 			V_ip6_output_flowtable_size = 256;
202 		if (!powerof2(V_ip6_output_flowtable_size)) {
203 			printf("flowtable must be power of 2 size\n");
204 			V_ip6_output_flowtable_size = 2048;
205 		}
206 	} else {
207 		/*
208 		 * round up to the next power of 2
209 		 */
210 		V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
211 	}
212 	V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU);
213 #endif
214 
215 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
216 
217 	/* Skip global initialization stuff for non-default instances. */
218 	if (!IS_DEFAULT_VNET(curvnet))
219 		return;
220 
221 #ifdef DIAGNOSTIC
222 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
223 		panic("sizeof(protosw) != sizeof(ip6protosw)");
224 #endif
225 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
226 	if (pr == NULL)
227 		panic("ip6_init");
228 
229 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
230 	for (i = 0; i < IPPROTO_MAX; i++)
231 		ip6_protox[i] = pr - inet6sw;
232 	/*
233 	 * Cycle through IP protocols and put them into the appropriate place
234 	 * in ip6_protox[].
235 	 */
236 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
237 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
238 		if (pr->pr_domain->dom_family == PF_INET6 &&
239 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
240 			/* Be careful to only index valid IP protocols. */
241 			if (pr->pr_protocol < IPPROTO_MAX)
242 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
243 		}
244 
245 	netisr_register(&ip6_nh);
246 }
247 
248 /*
249  * The protocol to be inserted into ip6_protox[] must be already registered
250  * in inet6sw[], either statically or through pf_proto_register().
251  */
252 int
253 ip6proto_register(short ip6proto)
254 {
255 	struct ip6protosw *pr;
256 
257 	/* Sanity checks. */
258 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
259 		return (EPROTONOSUPPORT);
260 
261 	/*
262 	 * The protocol slot must not be occupied by another protocol
263 	 * already.  An index pointing to IPPROTO_RAW is unused.
264 	 */
265 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
266 	if (pr == NULL)
267 		return (EPFNOSUPPORT);
268 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
269 		return (EEXIST);
270 
271 	/*
272 	 * Find the protocol position in inet6sw[] and set the index.
273 	 */
274 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
275 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
276 		if (pr->pr_domain->dom_family == PF_INET6 &&
277 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
278 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
279 			return (0);
280 		}
281 	}
282 	return (EPROTONOSUPPORT);
283 }
284 
285 int
286 ip6proto_unregister(short ip6proto)
287 {
288 	struct ip6protosw *pr;
289 
290 	/* Sanity checks. */
291 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
292 		return (EPROTONOSUPPORT);
293 
294 	/* Check if the protocol was indeed registered. */
295 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
296 	if (pr == NULL)
297 		return (EPFNOSUPPORT);
298 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
299 		return (ENOENT);
300 
301 	/* Reset the protocol slot to IPPROTO_RAW. */
302 	ip6_protox[ip6proto] = pr - inet6sw;
303 	return (0);
304 }
305 
306 #ifdef VIMAGE
307 void
308 ip6_destroy()
309 {
310 	int i;
311 
312 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
313 		printf("%s: WARNING: unable to unregister pfil hook, "
314 		    "error %d\n", __func__, i);
315 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
316 	nd6_destroy();
317 	callout_drain(&V_in6_tmpaddrtimer_ch);
318 }
319 #endif
320 
321 static int
322 ip6_init2_vnet(const void *unused __unused)
323 {
324 
325 	/* nd6_timer_init */
326 	callout_init(&V_nd6_timer_ch, 0);
327 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
328 
329 	/* timer for regeneranation of temporary addresses randomize ID */
330 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
331 	callout_reset(&V_in6_tmpaddrtimer_ch,
332 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
333 		       V_ip6_temp_regen_advance) * hz,
334 		      in6_tmpaddrtimer, curvnet);
335 
336 	return (0);
337 }
338 
339 static void
340 ip6_init2(void *dummy)
341 {
342 
343 	ip6_init2_vnet(NULL);
344 }
345 
346 /* cheat */
347 /* This must be after route_init(), which is now SI_ORDER_THIRD */
348 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
349 
350 static int
351 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
352     int *nxt, int *ours)
353 {
354 	struct ip6_hdr *ip6;
355 	struct ip6_hbh *hbh;
356 
357 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
358 #if 0	/*touches NULL pointer*/
359 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
360 #endif
361 		goto out;	/* m have already been freed */
362 	}
363 
364 	/* adjust pointer */
365 	ip6 = mtod(m, struct ip6_hdr *);
366 
367 	/*
368 	 * if the payload length field is 0 and the next header field
369 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
370 	 * option MUST be included.
371 	 */
372 	if (ip6->ip6_plen == 0 && *plen == 0) {
373 		/*
374 		 * Note that if a valid jumbo payload option is
375 		 * contained, ip6_hopopts_input() must set a valid
376 		 * (non-zero) payload length to the variable plen.
377 		 */
378 		IP6STAT_INC(ip6s_badoptions);
379 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
380 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
381 		icmp6_error(m, ICMP6_PARAM_PROB,
382 			    ICMP6_PARAMPROB_HEADER,
383 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
384 		goto out;
385 	}
386 #ifndef PULLDOWN_TEST
387 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
388 	hbh = (struct ip6_hbh *)(ip6 + 1);
389 #else
390 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
391 		sizeof(struct ip6_hbh));
392 	if (hbh == NULL) {
393 		IP6STAT_INC(ip6s_tooshort);
394 		goto out;
395 	}
396 #endif
397 	*nxt = hbh->ip6h_nxt;
398 
399 	/*
400 	 * If we are acting as a router and the packet contains a
401 	 * router alert option, see if we know the option value.
402 	 * Currently, we only support the option value for MLD, in which
403 	 * case we should pass the packet to the multicast routing
404 	 * daemon.
405 	 */
406 	if (*rtalert != ~0) {
407 		switch (*rtalert) {
408 		case IP6OPT_RTALERT_MLD:
409 			if (V_ip6_forwarding)
410 				*ours = 1;
411 			break;
412 		default:
413 			/*
414 			 * RFC2711 requires unrecognized values must be
415 			 * silently ignored.
416 			 */
417 			break;
418 		}
419 	}
420 
421 	return (0);
422 
423 out:
424 	return (1);
425 }
426 
427 void
428 ip6_input(struct mbuf *m)
429 {
430 	struct ip6_hdr *ip6;
431 	int off = sizeof(struct ip6_hdr), nest;
432 	u_int32_t plen;
433 	u_int32_t rtalert = ~0;
434 	int nxt, ours = 0;
435 	struct ifnet *deliverifp = NULL, *ifp = NULL;
436 	struct in6_addr odst;
437 	struct route_in6 rin6;
438 	int srcrt = 0;
439 	struct llentry *lle = NULL;
440 	struct sockaddr_in6 dst6, *dst;
441 
442 	bzero(&rin6, sizeof(struct route_in6));
443 #ifdef IPSEC
444 	/*
445 	 * should the inner packet be considered authentic?
446 	 * see comment in ah4_input().
447 	 * NB: m cannot be NULL when passed to the input routine
448 	 */
449 
450 	m->m_flags &= ~M_AUTHIPHDR;
451 	m->m_flags &= ~M_AUTHIPDGM;
452 
453 #endif /* IPSEC */
454 
455 	/*
456 	 * make sure we don't have onion peering information into m_tag.
457 	 */
458 	ip6_delaux(m);
459 
460 	if (m->m_flags & M_FASTFWD_OURS) {
461 		/*
462 		 * Firewall changed destination to local.
463 		 */
464 		m->m_flags &= ~M_FASTFWD_OURS;
465 		ours = 1;
466 		deliverifp = m->m_pkthdr.rcvif;
467 		ip6 = mtod(m, struct ip6_hdr *);
468 		goto hbhcheck;
469 	}
470 
471 	/*
472 	 * mbuf statistics
473 	 */
474 	if (m->m_flags & M_EXT) {
475 		if (m->m_next)
476 			IP6STAT_INC(ip6s_mext2m);
477 		else
478 			IP6STAT_INC(ip6s_mext1);
479 	} else {
480 		if (m->m_next) {
481 			if (m->m_flags & M_LOOP) {
482 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
483 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
484 				IP6STAT_INC(
485 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
486 			else
487 				IP6STAT_INC(ip6s_m2m[0]);
488 		} else
489 			IP6STAT_INC(ip6s_m1);
490 	}
491 
492 	/* drop the packet if IPv6 operation is disabled on the IF */
493 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
494 		m_freem(m);
495 		return;
496 	}
497 
498 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
499 	IP6STAT_INC(ip6s_total);
500 
501 #ifndef PULLDOWN_TEST
502 	/*
503 	 * L2 bridge code and some other code can return mbuf chain
504 	 * that does not conform to KAME requirement.  too bad.
505 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
506 	 */
507 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
508 		struct mbuf *n;
509 
510 		if (m->m_pkthdr.len > MHLEN)
511 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
512 		else
513 			n = m_gethdr(M_NOWAIT, MT_DATA);
514 		if (n == NULL) {
515 			m_freem(m);
516 			return;	/* ENOBUFS */
517 		}
518 
519 		m_move_pkthdr(n, m);
520 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
521 		n->m_len = n->m_pkthdr.len;
522 		m_freem(m);
523 		m = n;
524 	}
525 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
526 #endif
527 
528 	if (m->m_len < sizeof(struct ip6_hdr)) {
529 		struct ifnet *inifp;
530 		inifp = m->m_pkthdr.rcvif;
531 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
532 			IP6STAT_INC(ip6s_toosmall);
533 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
534 			return;
535 		}
536 	}
537 
538 	ip6 = mtod(m, struct ip6_hdr *);
539 
540 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
541 		IP6STAT_INC(ip6s_badvers);
542 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
543 		goto bad;
544 	}
545 
546 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
547 
548 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
549 
550 	/*
551 	 * Check against address spoofing/corruption.
552 	 */
553 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
554 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
555 		/*
556 		 * XXX: "badscope" is not very suitable for a multicast source.
557 		 */
558 		IP6STAT_INC(ip6s_badscope);
559 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
560 		goto bad;
561 	}
562 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
563 	    !(m->m_flags & M_LOOP)) {
564 		/*
565 		 * In this case, the packet should come from the loopback
566 		 * interface.  However, we cannot just check the if_flags,
567 		 * because ip6_mloopback() passes the "actual" interface
568 		 * as the outgoing/incoming interface.
569 		 */
570 		IP6STAT_INC(ip6s_badscope);
571 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
572 		goto bad;
573 	}
574 
575 #ifdef ALTQ
576 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
577 		/* packet is dropped by traffic conditioner */
578 		return;
579 	}
580 #endif
581 	/*
582 	 * The following check is not documented in specs.  A malicious
583 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
584 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
585 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
586 	 *
587 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
588 	 * support IPv4-less kernel compilation, we cannot support SIIT
589 	 * environment at all.  So, it makes more sense for us to reject any
590 	 * malicious packets for non-SIIT environment, than try to do a
591 	 * partial support for SIIT environment.
592 	 */
593 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
594 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
595 		IP6STAT_INC(ip6s_badscope);
596 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
597 		goto bad;
598 	}
599 #if 0
600 	/*
601 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
602 	 *
603 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
604 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
605 	 * is revised to forbid relaying case.
606 	 */
607 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
608 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
609 		IP6STAT_INC(ip6s_badscope);
610 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
611 		goto bad;
612 	}
613 #endif
614 #ifdef IPSEC
615 	/*
616 	 * Bypass packet filtering for packets previously handled by IPsec.
617 	 */
618 	if (ip6_ipsec_filtertunnel(m))
619 		goto passin;
620 #endif /* IPSEC */
621 
622 	/*
623 	 * Run through list of hooks for input packets.
624 	 *
625 	 * NB: Beware of the destination address changing
626 	 *     (e.g. by NAT rewriting).  When this happens,
627 	 *     tell ip6_forward to do the right thing.
628 	 */
629 	odst = ip6->ip6_dst;
630 
631 	/* Jump over all PFIL processing if hooks are not active. */
632 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
633 		goto passin;
634 
635 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
636 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
637 		return;
638 	if (m == NULL)			/* consumed by filter */
639 		return;
640 	ip6 = mtod(m, struct ip6_hdr *);
641 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
642 
643 	if (m->m_flags & M_FASTFWD_OURS) {
644 		m->m_flags &= ~M_FASTFWD_OURS;
645 		ours = 1;
646 		deliverifp = m->m_pkthdr.rcvif;
647 		goto hbhcheck;
648 	}
649 	if ((m->m_flags & M_IP6_NEXTHOP) &&
650 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
651 		/*
652 		 * Directly ship the packet on.  This allows forwarding
653 		 * packets originally destined to us to some other directly
654 		 * connected host.
655 		 */
656 		ip6_forward(m, 1);
657 		goto out;
658 	}
659 
660 passin:
661 	/*
662 	 * Disambiguate address scope zones (if there is ambiguity).
663 	 * We first make sure that the original source or destination address
664 	 * is not in our internal form for scoped addresses.  Such addresses
665 	 * are not necessarily invalid spec-wise, but we cannot accept them due
666 	 * to the usage conflict.
667 	 * in6_setscope() then also checks and rejects the cases where src or
668 	 * dst are the loopback address and the receiving interface
669 	 * is not loopback.
670 	 */
671 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
672 		IP6STAT_INC(ip6s_badscope); /* XXX */
673 		goto bad;
674 	}
675 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
676 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
677 		IP6STAT_INC(ip6s_badscope);
678 		goto bad;
679 	}
680 
681 	/*
682 	 * Multicast check. Assume packet is for us to avoid
683 	 * prematurely taking locks.
684 	 */
685 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
686 		ours = 1;
687 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
688 		deliverifp = m->m_pkthdr.rcvif;
689 		goto hbhcheck;
690 	}
691 
692 	/*
693 	 *  Unicast check
694 	 */
695 
696 	bzero(&dst6, sizeof(dst6));
697 	dst6.sin6_family = AF_INET6;
698 	dst6.sin6_len = sizeof(struct sockaddr_in6);
699 	dst6.sin6_addr = ip6->ip6_dst;
700 	ifp = m->m_pkthdr.rcvif;
701 	IF_AFDATA_RLOCK(ifp);
702 	lle = lla_lookup(LLTABLE6(ifp), 0,
703 	     (struct sockaddr *)&dst6);
704 	IF_AFDATA_RUNLOCK(ifp);
705 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
706 		struct ifaddr *ifa;
707 		struct in6_ifaddr *ia6;
708 		int bad;
709 
710 		bad = 1;
711 #define	sa_equal(a1, a2)						\
712 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
713 		IF_ADDR_RLOCK(ifp);
714 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
715 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
716 				continue;
717 			if (sa_equal(&dst6, ifa->ifa_addr))
718 				break;
719 		}
720 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
721 		    __func__, lle));
722 #undef sa_equal
723 
724 		ia6 = (struct in6_ifaddr *)ifa;
725 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
726 			/* Count the packet in the ip address stats */
727 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
728 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
729 			    m->m_pkthdr.len);
730 
731 			/*
732 			 * record address information into m_tag.
733 			 */
734 			(void)ip6_setdstifaddr(m, ia6);
735 
736 			bad = 0;
737 		} else {
738 			char ip6bufs[INET6_ADDRSTRLEN];
739 			char ip6bufd[INET6_ADDRSTRLEN];
740 			/* address is not ready, so discard the packet. */
741 			nd6log((LOG_INFO,
742 			    "ip6_input: packet to an unready address %s->%s\n",
743 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
744 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
745 		}
746 		IF_ADDR_RUNLOCK(ifp);
747 		LLE_RUNLOCK(lle);
748 		if (bad)
749 			goto bad;
750 		else {
751 			ours = 1;
752 			deliverifp = ifp;
753 			goto hbhcheck;
754 		}
755 	}
756 	if (lle != NULL)
757 		LLE_RUNLOCK(lle);
758 
759 	dst = &rin6.ro_dst;
760 	dst->sin6_len = sizeof(struct sockaddr_in6);
761 	dst->sin6_family = AF_INET6;
762 	dst->sin6_addr = ip6->ip6_dst;
763 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
764 	if (rin6.ro_rt)
765 		RT_UNLOCK(rin6.ro_rt);
766 
767 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
768 
769 	/*
770 	 * Accept the packet if the forwarding interface to the destination
771 	 * according to the routing table is the loopback interface,
772 	 * unless the associated route has a gateway.
773 	 * Note that this approach causes to accept a packet if there is a
774 	 * route to the loopback interface for the destination of the packet.
775 	 * But we think it's even useful in some situations, e.g. when using
776 	 * a special daemon which wants to intercept the packet.
777 	 *
778 	 * XXX: some OSes automatically make a cloned route for the destination
779 	 * of an outgoing packet.  If the outgoing interface of the packet
780 	 * is a loopback one, the kernel would consider the packet to be
781 	 * accepted, even if we have no such address assinged on the interface.
782 	 * We check the cloned flag of the route entry to reject such cases,
783 	 * assuming that route entries for our own addresses are not made by
784 	 * cloning (it should be true because in6_addloop explicitly installs
785 	 * the host route).  However, we might have to do an explicit check
786 	 * while it would be less efficient.  Or, should we rather install a
787 	 * reject route for such a case?
788 	 */
789 	if (rin6.ro_rt &&
790 	    (rin6.ro_rt->rt_flags &
791 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
792 #ifdef RTF_WASCLONED
793 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
794 #endif
795 #ifdef RTF_CLONED
796 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
797 #endif
798 #if 0
799 	    /*
800 	     * The check below is redundant since the comparison of
801 	     * the destination and the key of the rtentry has
802 	     * already done through looking up the routing table.
803 	     */
804 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
805 	    &rt6_key(rin6.ro_rt)->sin6_addr)
806 #endif
807 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
808 		int free_ia6 = 0;
809 		struct in6_ifaddr *ia6;
810 
811 		/*
812 		 * found the loopback route to the interface address
813 		 */
814 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
815 			struct sockaddr_in6 dest6;
816 
817 			bzero(&dest6, sizeof(dest6));
818 			dest6.sin6_family = AF_INET6;
819 			dest6.sin6_len = sizeof(dest6);
820 			dest6.sin6_addr = ip6->ip6_dst;
821 			ia6 = (struct in6_ifaddr *)
822 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
823 			if (ia6 == NULL)
824 				goto bad;
825 			free_ia6 = 1;
826 		}
827 		else
828 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
829 
830 		/*
831 		 * record address information into m_tag.
832 		 */
833 		(void)ip6_setdstifaddr(m, ia6);
834 
835 		/*
836 		 * packets to a tentative, duplicated, or somehow invalid
837 		 * address must not be accepted.
838 		 */
839 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
840 			/* this address is ready */
841 			ours = 1;
842 			deliverifp = ia6->ia_ifp;	/* correct? */
843 			/* Count the packet in the ip address stats */
844 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
845 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
846 			    m->m_pkthdr.len);
847 			if (free_ia6)
848 				ifa_free(&ia6->ia_ifa);
849 			goto hbhcheck;
850 		} else {
851 			char ip6bufs[INET6_ADDRSTRLEN];
852 			char ip6bufd[INET6_ADDRSTRLEN];
853 			/* address is not ready, so discard the packet. */
854 			nd6log((LOG_INFO,
855 			    "ip6_input: packet to an unready address %s->%s\n",
856 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
857 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
858 
859 			if (free_ia6)
860 				ifa_free(&ia6->ia_ifa);
861 			goto bad;
862 		}
863 	}
864 
865 	/*
866 	 * FAITH (Firewall Aided Internet Translator)
867 	 */
868 	if (V_ip6_keepfaith) {
869 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
870 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
871 			/* XXX do we need more sanity checks? */
872 			ours = 1;
873 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
874 			goto hbhcheck;
875 		}
876 	}
877 
878 	/*
879 	 * Now there is no reason to process the packet if it's not our own
880 	 * and we're not a router.
881 	 */
882 	if (!V_ip6_forwarding) {
883 		IP6STAT_INC(ip6s_cantforward);
884 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
885 		goto bad;
886 	}
887 
888   hbhcheck:
889 	/*
890 	 * record address information into m_tag, if we don't have one yet.
891 	 * note that we are unable to record it, if the address is not listed
892 	 * as our interface address (e.g. multicast addresses, addresses
893 	 * within FAITH prefixes and such).
894 	 */
895 	if (deliverifp) {
896 		struct in6_ifaddr *ia6;
897 
898  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
899 			ifa_free(&ia6->ia_ifa);
900 		} else {
901 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
902 			if (ia6) {
903 				if (!ip6_setdstifaddr(m, ia6)) {
904 					/*
905 					 * XXX maybe we should drop the packet here,
906 					 * as we could not provide enough information
907 					 * to the upper layers.
908 					 */
909 				}
910 				ifa_free(&ia6->ia_ifa);
911 			}
912 		}
913 	}
914 
915 	/*
916 	 * Process Hop-by-Hop options header if it's contained.
917 	 * m may be modified in ip6_hopopts_input().
918 	 * If a JumboPayload option is included, plen will also be modified.
919 	 */
920 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
921 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
922 		int error;
923 
924 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
925 		if (error != 0)
926 			goto out;
927 	} else
928 		nxt = ip6->ip6_nxt;
929 
930 	/*
931 	 * Check that the amount of data in the buffers
932 	 * is as at least much as the IPv6 header would have us expect.
933 	 * Trim mbufs if longer than we expect.
934 	 * Drop packet if shorter than we expect.
935 	 */
936 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
937 		IP6STAT_INC(ip6s_tooshort);
938 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
939 		goto bad;
940 	}
941 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
942 		if (m->m_len == m->m_pkthdr.len) {
943 			m->m_len = sizeof(struct ip6_hdr) + plen;
944 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
945 		} else
946 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
947 	}
948 
949 	/*
950 	 * Forward if desirable.
951 	 */
952 	if (V_ip6_mrouter &&
953 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
954 		/*
955 		 * If we are acting as a multicast router, all
956 		 * incoming multicast packets are passed to the
957 		 * kernel-level multicast forwarding function.
958 		 * The packet is returned (relatively) intact; if
959 		 * ip6_mforward() returns a non-zero value, the packet
960 		 * must be discarded, else it may be accepted below.
961 		 *
962 		 * XXX TODO: Check hlim and multicast scope here to avoid
963 		 * unnecessarily calling into ip6_mforward().
964 		 */
965 		if (ip6_mforward &&
966 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
967 			IP6STAT_INC(ip6s_cantforward);
968 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
969 			goto bad;
970 		}
971 	} else if (!ours) {
972 		ip6_forward(m, srcrt);
973 		goto out;
974 	}
975 
976 	ip6 = mtod(m, struct ip6_hdr *);
977 
978 	/*
979 	 * Malicious party may be able to use IPv4 mapped addr to confuse
980 	 * tcp/udp stack and bypass security checks (act as if it was from
981 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
982 	 *
983 	 * For SIIT end node behavior, you may want to disable the check.
984 	 * However, you will  become vulnerable to attacks using IPv4 mapped
985 	 * source.
986 	 */
987 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
988 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
989 		IP6STAT_INC(ip6s_badscope);
990 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
991 		goto bad;
992 	}
993 
994 	/*
995 	 * Tell launch routine the next header
996 	 */
997 	IP6STAT_INC(ip6s_delivered);
998 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
999 	nest = 0;
1000 
1001 	while (nxt != IPPROTO_DONE) {
1002 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
1003 			IP6STAT_INC(ip6s_toomanyhdr);
1004 			goto bad;
1005 		}
1006 
1007 		/*
1008 		 * protection against faulty packet - there should be
1009 		 * more sanity checks in header chain processing.
1010 		 */
1011 		if (m->m_pkthdr.len < off) {
1012 			IP6STAT_INC(ip6s_tooshort);
1013 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
1014 			goto bad;
1015 		}
1016 
1017 #ifdef IPSEC
1018 		/*
1019 		 * enforce IPsec policy checking if we are seeing last header.
1020 		 * note that we do not visit this with protocols with pcb layer
1021 		 * code - like udp/tcp/raw ip.
1022 		 */
1023 		if (ip6_ipsec_input(m, nxt))
1024 			goto bad;
1025 #endif /* IPSEC */
1026 
1027 		/*
1028 		 * Use mbuf flags to propagate Router Alert option to
1029 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1030 		 */
1031 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1032 			m->m_flags |= M_RTALERT_MLD;
1033 
1034 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1035 	}
1036 	goto out;
1037 bad:
1038 	m_freem(m);
1039 out:
1040 	if (rin6.ro_rt)
1041 		RTFREE(rin6.ro_rt);
1042 }
1043 
1044 /*
1045  * set/grab in6_ifaddr correspond to IPv6 destination address.
1046  * XXX backward compatibility wrapper
1047  *
1048  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1049  * and then bump it when the tag is copied, and release it when the tag is
1050  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1051  * we just bump the ia refcount when we receive it.  This should be fixed.
1052  */
1053 static struct ip6aux *
1054 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1055 {
1056 	struct ip6aux *ip6a;
1057 
1058 	ip6a = ip6_addaux(m);
1059 	if (ip6a)
1060 		ip6a->ip6a_dstia6 = ia6;
1061 	return ip6a;	/* NULL if failed to set */
1062 }
1063 
1064 struct in6_ifaddr *
1065 ip6_getdstifaddr(struct mbuf *m)
1066 {
1067 	struct ip6aux *ip6a;
1068 	struct in6_ifaddr *ia;
1069 
1070 	ip6a = ip6_findaux(m);
1071 	if (ip6a) {
1072 		ia = ip6a->ip6a_dstia6;
1073 		ifa_ref(&ia->ia_ifa);
1074 		return ia;
1075 	} else
1076 		return NULL;
1077 }
1078 
1079 /*
1080  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1081  * included, the real payload length will be stored in plenp.
1082  *
1083  * rtalertp - XXX: should be stored more smart way
1084  */
1085 static int
1086 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1087     struct mbuf **mp, int *offp)
1088 {
1089 	struct mbuf *m = *mp;
1090 	int off = *offp, hbhlen;
1091 	struct ip6_hbh *hbh;
1092 	u_int8_t *opt;
1093 
1094 	/* validation of the length of the header */
1095 #ifndef PULLDOWN_TEST
1096 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1097 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1098 	hbhlen = (hbh->ip6h_len + 1) << 3;
1099 
1100 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1101 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1102 #else
1103 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1104 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1105 	if (hbh == NULL) {
1106 		IP6STAT_INC(ip6s_tooshort);
1107 		return -1;
1108 	}
1109 	hbhlen = (hbh->ip6h_len + 1) << 3;
1110 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1111 		hbhlen);
1112 	if (hbh == NULL) {
1113 		IP6STAT_INC(ip6s_tooshort);
1114 		return -1;
1115 	}
1116 #endif
1117 	off += hbhlen;
1118 	hbhlen -= sizeof(struct ip6_hbh);
1119 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1120 
1121 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1122 				hbhlen, rtalertp, plenp) < 0)
1123 		return (-1);
1124 
1125 	*offp = off;
1126 	*mp = m;
1127 	return (0);
1128 }
1129 
1130 /*
1131  * Search header for all Hop-by-hop options and process each option.
1132  * This function is separate from ip6_hopopts_input() in order to
1133  * handle a case where the sending node itself process its hop-by-hop
1134  * options header. In such a case, the function is called from ip6_output().
1135  *
1136  * The function assumes that hbh header is located right after the IPv6 header
1137  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1138  * opthead + hbhlen is located in contiguous memory region.
1139  */
1140 int
1141 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1142     u_int32_t *rtalertp, u_int32_t *plenp)
1143 {
1144 	struct ip6_hdr *ip6;
1145 	int optlen = 0;
1146 	u_int8_t *opt = opthead;
1147 	u_int16_t rtalert_val;
1148 	u_int32_t jumboplen;
1149 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1150 
1151 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1152 		switch (*opt) {
1153 		case IP6OPT_PAD1:
1154 			optlen = 1;
1155 			break;
1156 		case IP6OPT_PADN:
1157 			if (hbhlen < IP6OPT_MINLEN) {
1158 				IP6STAT_INC(ip6s_toosmall);
1159 				goto bad;
1160 			}
1161 			optlen = *(opt + 1) + 2;
1162 			break;
1163 		case IP6OPT_ROUTER_ALERT:
1164 			/* XXX may need check for alignment */
1165 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1166 				IP6STAT_INC(ip6s_toosmall);
1167 				goto bad;
1168 			}
1169 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1170 				/* XXX stat */
1171 				icmp6_error(m, ICMP6_PARAM_PROB,
1172 				    ICMP6_PARAMPROB_HEADER,
1173 				    erroff + opt + 1 - opthead);
1174 				return (-1);
1175 			}
1176 			optlen = IP6OPT_RTALERT_LEN;
1177 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1178 			*rtalertp = ntohs(rtalert_val);
1179 			break;
1180 		case IP6OPT_JUMBO:
1181 			/* XXX may need check for alignment */
1182 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1183 				IP6STAT_INC(ip6s_toosmall);
1184 				goto bad;
1185 			}
1186 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1187 				/* XXX stat */
1188 				icmp6_error(m, ICMP6_PARAM_PROB,
1189 				    ICMP6_PARAMPROB_HEADER,
1190 				    erroff + opt + 1 - opthead);
1191 				return (-1);
1192 			}
1193 			optlen = IP6OPT_JUMBO_LEN;
1194 
1195 			/*
1196 			 * IPv6 packets that have non 0 payload length
1197 			 * must not contain a jumbo payload option.
1198 			 */
1199 			ip6 = mtod(m, struct ip6_hdr *);
1200 			if (ip6->ip6_plen) {
1201 				IP6STAT_INC(ip6s_badoptions);
1202 				icmp6_error(m, ICMP6_PARAM_PROB,
1203 				    ICMP6_PARAMPROB_HEADER,
1204 				    erroff + opt - opthead);
1205 				return (-1);
1206 			}
1207 
1208 			/*
1209 			 * We may see jumbolen in unaligned location, so
1210 			 * we'd need to perform bcopy().
1211 			 */
1212 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1213 			jumboplen = (u_int32_t)htonl(jumboplen);
1214 
1215 #if 1
1216 			/*
1217 			 * if there are multiple jumbo payload options,
1218 			 * *plenp will be non-zero and the packet will be
1219 			 * rejected.
1220 			 * the behavior may need some debate in ipngwg -
1221 			 * multiple options does not make sense, however,
1222 			 * there's no explicit mention in specification.
1223 			 */
1224 			if (*plenp != 0) {
1225 				IP6STAT_INC(ip6s_badoptions);
1226 				icmp6_error(m, ICMP6_PARAM_PROB,
1227 				    ICMP6_PARAMPROB_HEADER,
1228 				    erroff + opt + 2 - opthead);
1229 				return (-1);
1230 			}
1231 #endif
1232 
1233 			/*
1234 			 * jumbo payload length must be larger than 65535.
1235 			 */
1236 			if (jumboplen <= IPV6_MAXPACKET) {
1237 				IP6STAT_INC(ip6s_badoptions);
1238 				icmp6_error(m, ICMP6_PARAM_PROB,
1239 				    ICMP6_PARAMPROB_HEADER,
1240 				    erroff + opt + 2 - opthead);
1241 				return (-1);
1242 			}
1243 			*plenp = jumboplen;
1244 
1245 			break;
1246 		default:		/* unknown option */
1247 			if (hbhlen < IP6OPT_MINLEN) {
1248 				IP6STAT_INC(ip6s_toosmall);
1249 				goto bad;
1250 			}
1251 			optlen = ip6_unknown_opt(opt, m,
1252 			    erroff + opt - opthead);
1253 			if (optlen == -1)
1254 				return (-1);
1255 			optlen += 2;
1256 			break;
1257 		}
1258 	}
1259 
1260 	return (0);
1261 
1262   bad:
1263 	m_freem(m);
1264 	return (-1);
1265 }
1266 
1267 /*
1268  * Unknown option processing.
1269  * The third argument `off' is the offset from the IPv6 header to the option,
1270  * which is necessary if the IPv6 header the and option header and IPv6 header
1271  * is not contiguous in order to return an ICMPv6 error.
1272  */
1273 int
1274 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1275 {
1276 	struct ip6_hdr *ip6;
1277 
1278 	switch (IP6OPT_TYPE(*optp)) {
1279 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1280 		return ((int)*(optp + 1));
1281 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1282 		m_freem(m);
1283 		return (-1);
1284 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1285 		IP6STAT_INC(ip6s_badoptions);
1286 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1287 		return (-1);
1288 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1289 		IP6STAT_INC(ip6s_badoptions);
1290 		ip6 = mtod(m, struct ip6_hdr *);
1291 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1292 		    (m->m_flags & (M_BCAST|M_MCAST)))
1293 			m_freem(m);
1294 		else
1295 			icmp6_error(m, ICMP6_PARAM_PROB,
1296 				    ICMP6_PARAMPROB_OPTION, off);
1297 		return (-1);
1298 	}
1299 
1300 	m_freem(m);		/* XXX: NOTREACHED */
1301 	return (-1);
1302 }
1303 
1304 /*
1305  * Create the "control" list for this pcb.
1306  * These functions will not modify mbuf chain at all.
1307  *
1308  * With KAME mbuf chain restriction:
1309  * The routine will be called from upper layer handlers like tcp6_input().
1310  * Thus the routine assumes that the caller (tcp6_input) have already
1311  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1312  * very first mbuf on the mbuf chain.
1313  *
1314  * ip6_savecontrol_v4 will handle those options that are possible to be
1315  * set on a v4-mapped socket.
1316  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1317  * options and handle the v6-only ones itself.
1318  */
1319 struct mbuf **
1320 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1321     int *v4only)
1322 {
1323 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1324 
1325 #ifdef SO_TIMESTAMP
1326 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1327 		struct timeval tv;
1328 
1329 		microtime(&tv);
1330 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1331 		    SCM_TIMESTAMP, SOL_SOCKET);
1332 		if (*mp)
1333 			mp = &(*mp)->m_next;
1334 	}
1335 #endif
1336 
1337 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1338 	/* RFC 2292 sec. 5 */
1339 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1340 		struct in6_pktinfo pi6;
1341 
1342 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1343 #ifdef INET
1344 			struct ip *ip;
1345 
1346 			ip = mtod(m, struct ip *);
1347 			pi6.ipi6_addr.s6_addr32[0] = 0;
1348 			pi6.ipi6_addr.s6_addr32[1] = 0;
1349 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1350 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1351 #else
1352 			/* We won't hit this code */
1353 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1354 #endif
1355 		} else {
1356 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1357 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1358 		}
1359 		pi6.ipi6_ifindex =
1360 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1361 
1362 		*mp = sbcreatecontrol((caddr_t) &pi6,
1363 		    sizeof(struct in6_pktinfo),
1364 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1365 		if (*mp)
1366 			mp = &(*mp)->m_next;
1367 	}
1368 
1369 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1370 		int hlim;
1371 
1372 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1373 #ifdef INET
1374 			struct ip *ip;
1375 
1376 			ip = mtod(m, struct ip *);
1377 			hlim = ip->ip_ttl;
1378 #else
1379 			/* We won't hit this code */
1380 			hlim = 0;
1381 #endif
1382 		} else {
1383 			hlim = ip6->ip6_hlim & 0xff;
1384 		}
1385 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1386 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1387 		    IPPROTO_IPV6);
1388 		if (*mp)
1389 			mp = &(*mp)->m_next;
1390 	}
1391 
1392 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1393 		int tclass;
1394 
1395 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1396 #ifdef INET
1397 			struct ip *ip;
1398 
1399 			ip = mtod(m, struct ip *);
1400 			tclass = ip->ip_tos;
1401 #else
1402 			/* We won't hit this code */
1403 			tclass = 0;
1404 #endif
1405 		} else {
1406 			u_int32_t flowinfo;
1407 
1408 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1409 			flowinfo >>= 20;
1410 			tclass = flowinfo & 0xff;
1411 		}
1412 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1413 		    IPV6_TCLASS, IPPROTO_IPV6);
1414 		if (*mp)
1415 			mp = &(*mp)->m_next;
1416 	}
1417 
1418 	if (v4only != NULL) {
1419 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1420 			*v4only = 1;
1421 		} else {
1422 			*v4only = 0;
1423 		}
1424 	}
1425 
1426 	return (mp);
1427 }
1428 
1429 void
1430 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1431 {
1432 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1433 	int v4only = 0;
1434 
1435 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1436 	if (v4only)
1437 		return;
1438 
1439 	/*
1440 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1441 	 * privilege for the option (see ip6_ctloutput), but it might be too
1442 	 * strict, since there might be some hop-by-hop options which can be
1443 	 * returned to normal user.
1444 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1445 	 */
1446 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1447 		/*
1448 		 * Check if a hop-by-hop options header is contatined in the
1449 		 * received packet, and if so, store the options as ancillary
1450 		 * data. Note that a hop-by-hop options header must be
1451 		 * just after the IPv6 header, which is assured through the
1452 		 * IPv6 input processing.
1453 		 */
1454 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1455 			struct ip6_hbh *hbh;
1456 			int hbhlen = 0;
1457 #ifdef PULLDOWN_TEST
1458 			struct mbuf *ext;
1459 #endif
1460 
1461 #ifndef PULLDOWN_TEST
1462 			hbh = (struct ip6_hbh *)(ip6 + 1);
1463 			hbhlen = (hbh->ip6h_len + 1) << 3;
1464 #else
1465 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1466 			    ip6->ip6_nxt);
1467 			if (ext == NULL) {
1468 				IP6STAT_INC(ip6s_tooshort);
1469 				return;
1470 			}
1471 			hbh = mtod(ext, struct ip6_hbh *);
1472 			hbhlen = (hbh->ip6h_len + 1) << 3;
1473 			if (hbhlen != ext->m_len) {
1474 				m_freem(ext);
1475 				IP6STAT_INC(ip6s_tooshort);
1476 				return;
1477 			}
1478 #endif
1479 
1480 			/*
1481 			 * XXX: We copy the whole header even if a
1482 			 * jumbo payload option is included, the option which
1483 			 * is to be removed before returning according to
1484 			 * RFC2292.
1485 			 * Note: this constraint is removed in RFC3542
1486 			 */
1487 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1488 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1489 			    IPPROTO_IPV6);
1490 			if (*mp)
1491 				mp = &(*mp)->m_next;
1492 #ifdef PULLDOWN_TEST
1493 			m_freem(ext);
1494 #endif
1495 		}
1496 	}
1497 
1498 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1499 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1500 
1501 		/*
1502 		 * Search for destination options headers or routing
1503 		 * header(s) through the header chain, and stores each
1504 		 * header as ancillary data.
1505 		 * Note that the order of the headers remains in
1506 		 * the chain of ancillary data.
1507 		 */
1508 		while (1) {	/* is explicit loop prevention necessary? */
1509 			struct ip6_ext *ip6e = NULL;
1510 			int elen;
1511 #ifdef PULLDOWN_TEST
1512 			struct mbuf *ext = NULL;
1513 #endif
1514 
1515 			/*
1516 			 * if it is not an extension header, don't try to
1517 			 * pull it from the chain.
1518 			 */
1519 			switch (nxt) {
1520 			case IPPROTO_DSTOPTS:
1521 			case IPPROTO_ROUTING:
1522 			case IPPROTO_HOPOPTS:
1523 			case IPPROTO_AH: /* is it possible? */
1524 				break;
1525 			default:
1526 				goto loopend;
1527 			}
1528 
1529 #ifndef PULLDOWN_TEST
1530 			if (off + sizeof(*ip6e) > m->m_len)
1531 				goto loopend;
1532 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1533 			if (nxt == IPPROTO_AH)
1534 				elen = (ip6e->ip6e_len + 2) << 2;
1535 			else
1536 				elen = (ip6e->ip6e_len + 1) << 3;
1537 			if (off + elen > m->m_len)
1538 				goto loopend;
1539 #else
1540 			ext = ip6_pullexthdr(m, off, nxt);
1541 			if (ext == NULL) {
1542 				IP6STAT_INC(ip6s_tooshort);
1543 				return;
1544 			}
1545 			ip6e = mtod(ext, struct ip6_ext *);
1546 			if (nxt == IPPROTO_AH)
1547 				elen = (ip6e->ip6e_len + 2) << 2;
1548 			else
1549 				elen = (ip6e->ip6e_len + 1) << 3;
1550 			if (elen != ext->m_len) {
1551 				m_freem(ext);
1552 				IP6STAT_INC(ip6s_tooshort);
1553 				return;
1554 			}
1555 #endif
1556 
1557 			switch (nxt) {
1558 			case IPPROTO_DSTOPTS:
1559 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1560 					break;
1561 
1562 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1563 				    IS2292(in6p,
1564 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1565 				    IPPROTO_IPV6);
1566 				if (*mp)
1567 					mp = &(*mp)->m_next;
1568 				break;
1569 			case IPPROTO_ROUTING:
1570 				if (!(in6p->inp_flags & IN6P_RTHDR))
1571 					break;
1572 
1573 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1574 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1575 				    IPPROTO_IPV6);
1576 				if (*mp)
1577 					mp = &(*mp)->m_next;
1578 				break;
1579 			case IPPROTO_HOPOPTS:
1580 			case IPPROTO_AH: /* is it possible? */
1581 				break;
1582 
1583 			default:
1584 				/*
1585 				 * other cases have been filtered in the above.
1586 				 * none will visit this case.  here we supply
1587 				 * the code just in case (nxt overwritten or
1588 				 * other cases).
1589 				 */
1590 #ifdef PULLDOWN_TEST
1591 				m_freem(ext);
1592 #endif
1593 				goto loopend;
1594 
1595 			}
1596 
1597 			/* proceed with the next header. */
1598 			off += elen;
1599 			nxt = ip6e->ip6e_nxt;
1600 			ip6e = NULL;
1601 #ifdef PULLDOWN_TEST
1602 			m_freem(ext);
1603 			ext = NULL;
1604 #endif
1605 		}
1606 	  loopend:
1607 		;
1608 	}
1609 }
1610 #undef IS2292
1611 
1612 void
1613 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1614 {
1615 	struct socket *so;
1616 	struct mbuf *m_mtu;
1617 	struct ip6_mtuinfo mtuctl;
1618 
1619 	so =  in6p->inp_socket;
1620 
1621 	if (mtu == NULL)
1622 		return;
1623 
1624 #ifdef DIAGNOSTIC
1625 	if (so == NULL)		/* I believe this is impossible */
1626 		panic("ip6_notify_pmtu: socket is NULL");
1627 #endif
1628 
1629 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1630 	mtuctl.ip6m_mtu = *mtu;
1631 	mtuctl.ip6m_addr = *dst;
1632 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1633 		return;
1634 
1635 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1636 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1637 		return;
1638 
1639 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1640 	    == 0) {
1641 		m_freem(m_mtu);
1642 		/* XXX: should count statistics */
1643 	} else
1644 		sorwakeup(so);
1645 
1646 	return;
1647 }
1648 
1649 #ifdef PULLDOWN_TEST
1650 /*
1651  * pull single extension header from mbuf chain.  returns single mbuf that
1652  * contains the result, or NULL on error.
1653  */
1654 static struct mbuf *
1655 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1656 {
1657 	struct ip6_ext ip6e;
1658 	size_t elen;
1659 	struct mbuf *n;
1660 
1661 #ifdef DIAGNOSTIC
1662 	switch (nxt) {
1663 	case IPPROTO_DSTOPTS:
1664 	case IPPROTO_ROUTING:
1665 	case IPPROTO_HOPOPTS:
1666 	case IPPROTO_AH: /* is it possible? */
1667 		break;
1668 	default:
1669 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1670 	}
1671 #endif
1672 
1673 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1674 	if (nxt == IPPROTO_AH)
1675 		elen = (ip6e.ip6e_len + 2) << 2;
1676 	else
1677 		elen = (ip6e.ip6e_len + 1) << 3;
1678 
1679 	if (elen > MLEN)
1680 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1681 	else
1682 		n = m_get(M_NOWAIT, MT_DATA);
1683 	if (n == NULL)
1684 		return NULL;
1685 
1686 	m_copydata(m, off, elen, mtod(n, caddr_t));
1687 	n->m_len = elen;
1688 	return n;
1689 }
1690 #endif
1691 
1692 /*
1693  * Get pointer to the previous header followed by the header
1694  * currently processed.
1695  * XXX: This function supposes that
1696  *	M includes all headers,
1697  *	the next header field and the header length field of each header
1698  *	are valid, and
1699  *	the sum of each header length equals to OFF.
1700  * Because of these assumptions, this function must be called very
1701  * carefully. Moreover, it will not be used in the near future when
1702  * we develop `neater' mechanism to process extension headers.
1703  */
1704 char *
1705 ip6_get_prevhdr(struct mbuf *m, int off)
1706 {
1707 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1708 
1709 	if (off == sizeof(struct ip6_hdr))
1710 		return (&ip6->ip6_nxt);
1711 	else {
1712 		int len, nxt;
1713 		struct ip6_ext *ip6e = NULL;
1714 
1715 		nxt = ip6->ip6_nxt;
1716 		len = sizeof(struct ip6_hdr);
1717 		while (len < off) {
1718 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1719 
1720 			switch (nxt) {
1721 			case IPPROTO_FRAGMENT:
1722 				len += sizeof(struct ip6_frag);
1723 				break;
1724 			case IPPROTO_AH:
1725 				len += (ip6e->ip6e_len + 2) << 2;
1726 				break;
1727 			default:
1728 				len += (ip6e->ip6e_len + 1) << 3;
1729 				break;
1730 			}
1731 			nxt = ip6e->ip6e_nxt;
1732 		}
1733 		if (ip6e)
1734 			return (&ip6e->ip6e_nxt);
1735 		else
1736 			return NULL;
1737 	}
1738 }
1739 
1740 /*
1741  * get next header offset.  m will be retained.
1742  */
1743 int
1744 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1745 {
1746 	struct ip6_hdr ip6;
1747 	struct ip6_ext ip6e;
1748 	struct ip6_frag fh;
1749 
1750 	/* just in case */
1751 	if (m == NULL)
1752 		panic("ip6_nexthdr: m == NULL");
1753 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1754 		return -1;
1755 
1756 	switch (proto) {
1757 	case IPPROTO_IPV6:
1758 		if (m->m_pkthdr.len < off + sizeof(ip6))
1759 			return -1;
1760 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1761 		if (nxtp)
1762 			*nxtp = ip6.ip6_nxt;
1763 		off += sizeof(ip6);
1764 		return off;
1765 
1766 	case IPPROTO_FRAGMENT:
1767 		/*
1768 		 * terminate parsing if it is not the first fragment,
1769 		 * it does not make sense to parse through it.
1770 		 */
1771 		if (m->m_pkthdr.len < off + sizeof(fh))
1772 			return -1;
1773 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1774 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1775 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1776 			return -1;
1777 		if (nxtp)
1778 			*nxtp = fh.ip6f_nxt;
1779 		off += sizeof(struct ip6_frag);
1780 		return off;
1781 
1782 	case IPPROTO_AH:
1783 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1784 			return -1;
1785 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1786 		if (nxtp)
1787 			*nxtp = ip6e.ip6e_nxt;
1788 		off += (ip6e.ip6e_len + 2) << 2;
1789 		return off;
1790 
1791 	case IPPROTO_HOPOPTS:
1792 	case IPPROTO_ROUTING:
1793 	case IPPROTO_DSTOPTS:
1794 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1795 			return -1;
1796 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1797 		if (nxtp)
1798 			*nxtp = ip6e.ip6e_nxt;
1799 		off += (ip6e.ip6e_len + 1) << 3;
1800 		return off;
1801 
1802 	case IPPROTO_NONE:
1803 	case IPPROTO_ESP:
1804 	case IPPROTO_IPCOMP:
1805 		/* give up */
1806 		return -1;
1807 
1808 	default:
1809 		return -1;
1810 	}
1811 
1812 	return -1;
1813 }
1814 
1815 /*
1816  * get offset for the last header in the chain.  m will be kept untainted.
1817  */
1818 int
1819 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1820 {
1821 	int newoff;
1822 	int nxt;
1823 
1824 	if (!nxtp) {
1825 		nxt = -1;
1826 		nxtp = &nxt;
1827 	}
1828 	while (1) {
1829 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1830 		if (newoff < 0)
1831 			return off;
1832 		else if (newoff < off)
1833 			return -1;	/* invalid */
1834 		else if (newoff == off)
1835 			return newoff;
1836 
1837 		off = newoff;
1838 		proto = *nxtp;
1839 	}
1840 }
1841 
1842 static struct ip6aux *
1843 ip6_addaux(struct mbuf *m)
1844 {
1845 	struct m_tag *mtag;
1846 
1847 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1848 	if (!mtag) {
1849 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1850 		    M_NOWAIT);
1851 		if (mtag) {
1852 			m_tag_prepend(m, mtag);
1853 			bzero(mtag + 1, sizeof(struct ip6aux));
1854 		}
1855 	}
1856 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1857 }
1858 
1859 static struct ip6aux *
1860 ip6_findaux(struct mbuf *m)
1861 {
1862 	struct m_tag *mtag;
1863 
1864 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1865 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1866 }
1867 
1868 static void
1869 ip6_delaux(struct mbuf *m)
1870 {
1871 	struct m_tag *mtag;
1872 
1873 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1874 	if (mtag)
1875 		m_tag_delete(m, mtag);
1876 }
1877 
1878 /*
1879  * System control for IP6
1880  */
1881 
1882 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1883 	0,		0,		0,		0,
1884 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1885 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1886 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1887 	0,		0,		0,		0,
1888 	ENOPROTOOPT
1889 };
1890