xref: /freebsd/sys/netinet6/ip6_input.c (revision dd48af360fdbbb9552f9fc6de7abe50d68ad5331)
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_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/proc.h>
75 #include <sys/domain.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/errno.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_dl.h>
87 #include <net/route.h>
88 #include <net/netisr.h>
89 #include <net/pfil.h>
90 #include <net/vnet.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #include <net/if_llatbl.h>
95 #ifdef INET
96 #include <netinet/ip.h>
97 #include <netinet/ip_icmp.h>
98 #endif /* INET */
99 #include <netinet/ip6.h>
100 #include <netinet6/in6_var.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet/in_pcb.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/nd6.h>
107 
108 #ifdef IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netinet6/ip6_ipsec.h>
111 #include <netipsec/ipsec6.h>
112 #endif /* IPSEC */
113 
114 #include <netinet6/ip6protosw.h>
115 
116 extern struct domain inet6domain;
117 
118 u_char ip6_protox[IPPROTO_MAX];
119 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
120 
121 static struct netisr_handler ip6_nh = {
122 	.nh_name = "ip6",
123 	.nh_handler = ip6_input,
124 	.nh_proto = NETISR_IPV6,
125 	.nh_policy = NETISR_POLICY_FLOW,
126 };
127 
128 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
129 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
130 
131 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
132 
133 VNET_DEFINE(struct ip6stat, ip6stat);
134 
135 struct rwlock in6_ifaddr_lock;
136 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
137 
138 static void ip6_init2(void *);
139 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
140 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
141 #ifdef PULLDOWN_TEST
142 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
143 #endif
144 
145 /*
146  * IP6 initialization: fill in IP6 protocol switch table.
147  * All protocols not implemented in kernel go to raw IP6 protocol handler.
148  */
149 void
150 ip6_init(void)
151 {
152 	struct ip6protosw *pr;
153 	int i;
154 
155 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
156 	    &V_ip6_auto_linklocal);
157 
158 	TAILQ_INIT(&V_in6_ifaddrhead);
159 
160 	/* Initialize packet filter hooks. */
161 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
162 	V_inet6_pfil_hook.ph_af = AF_INET6;
163 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
164 		printf("%s: WARNING: unable to register pfil hook, "
165 			"error %d\n", __func__, i);
166 
167 	scope6_init();
168 	addrsel_policy_init();
169 	nd6_init();
170 	frag6_init();
171 
172 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
173 
174 	/* Skip global initialization stuff for non-default instances. */
175 	if (!IS_DEFAULT_VNET(curvnet))
176 		return;
177 
178 #ifdef DIAGNOSTIC
179 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
180 		panic("sizeof(protosw) != sizeof(ip6protosw)");
181 #endif
182 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
183 	if (pr == NULL)
184 		panic("ip6_init");
185 
186 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
187 	for (i = 0; i < IPPROTO_MAX; i++)
188 		ip6_protox[i] = pr - inet6sw;
189 	/*
190 	 * Cycle through IP protocols and put them into the appropriate place
191 	 * in ip6_protox[].
192 	 */
193 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
194 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
195 		if (pr->pr_domain->dom_family == PF_INET6 &&
196 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
197 			/* Be careful to only index valid IP protocols. */
198 			if (pr->pr_protocol < IPPROTO_MAX)
199 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
200 		}
201 
202 	netisr_register(&ip6_nh);
203 }
204 
205 #ifdef VIMAGE
206 void
207 ip6_destroy()
208 {
209 
210 	nd6_destroy();
211 	callout_drain(&V_in6_tmpaddrtimer_ch);
212 }
213 #endif
214 
215 static int
216 ip6_init2_vnet(const void *unused __unused)
217 {
218 
219 	/* nd6_timer_init */
220 	callout_init(&V_nd6_timer_ch, 0);
221 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
222 
223 	/* timer for regeneranation of temporary addresses randomize ID */
224 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
225 	callout_reset(&V_in6_tmpaddrtimer_ch,
226 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
227 		       V_ip6_temp_regen_advance) * hz,
228 		      in6_tmpaddrtimer, curvnet);
229 
230 	return (0);
231 }
232 
233 static void
234 ip6_init2(void *dummy)
235 {
236 
237 	ip6_init2_vnet(NULL);
238 }
239 
240 /* cheat */
241 /* This must be after route_init(), which is now SI_ORDER_THIRD */
242 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
243 
244 void
245 ip6_input(struct mbuf *m)
246 {
247 	struct ip6_hdr *ip6;
248 	int off = sizeof(struct ip6_hdr), nest;
249 	u_int32_t plen;
250 	u_int32_t rtalert = ~0;
251 	int nxt, ours = 0;
252 	struct ifnet *deliverifp = NULL, *ifp = NULL;
253 	struct in6_addr odst;
254 	struct route_in6 rin6;
255 	int srcrt = 0;
256 	struct llentry *lle = NULL;
257 	struct sockaddr_in6 dst6, *dst;
258 
259 	bzero(&rin6, sizeof(struct route_in6));
260 #ifdef IPSEC
261 	/*
262 	 * should the inner packet be considered authentic?
263 	 * see comment in ah4_input().
264 	 * NB: m cannot be NULL when passed to the input routine
265 	 */
266 
267 	m->m_flags &= ~M_AUTHIPHDR;
268 	m->m_flags &= ~M_AUTHIPDGM;
269 
270 #endif /* IPSEC */
271 
272 	/*
273 	 * make sure we don't have onion peering information into m_tag.
274 	 */
275 	ip6_delaux(m);
276 
277 	/*
278 	 * mbuf statistics
279 	 */
280 	if (m->m_flags & M_EXT) {
281 		if (m->m_next)
282 			V_ip6stat.ip6s_mext2m++;
283 		else
284 			V_ip6stat.ip6s_mext1++;
285 	} else {
286 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
287 		if (m->m_next) {
288 			if (m->m_flags & M_LOOP) {
289 				V_ip6stat.ip6s_m2m[V_loif->if_index]++;
290 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
291 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
292 			else
293 				V_ip6stat.ip6s_m2m[0]++;
294 		} else
295 			V_ip6stat.ip6s_m1++;
296 #undef M2MMAX
297 	}
298 
299 	/* drop the packet if IPv6 operation is disabled on the IF */
300 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
301 		m_freem(m);
302 		return;
303 	}
304 
305 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
306 	V_ip6stat.ip6s_total++;
307 
308 #ifndef PULLDOWN_TEST
309 	/*
310 	 * L2 bridge code and some other code can return mbuf chain
311 	 * that does not conform to KAME requirement.  too bad.
312 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
313 	 */
314 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
315 		struct mbuf *n;
316 
317 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
318 		if (n)
319 			M_MOVE_PKTHDR(n, m);
320 		if (n && n->m_pkthdr.len > MHLEN) {
321 			MCLGET(n, M_DONTWAIT);
322 			if ((n->m_flags & M_EXT) == 0) {
323 				m_freem(n);
324 				n = NULL;
325 			}
326 		}
327 		if (n == NULL) {
328 			m_freem(m);
329 			return;	/* ENOBUFS */
330 		}
331 
332 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
333 		n->m_len = n->m_pkthdr.len;
334 		m_freem(m);
335 		m = n;
336 	}
337 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
338 #endif
339 
340 	if (m->m_len < sizeof(struct ip6_hdr)) {
341 		struct ifnet *inifp;
342 		inifp = m->m_pkthdr.rcvif;
343 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
344 			V_ip6stat.ip6s_toosmall++;
345 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
346 			return;
347 		}
348 	}
349 
350 	ip6 = mtod(m, struct ip6_hdr *);
351 
352 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
353 		V_ip6stat.ip6s_badvers++;
354 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
355 		goto bad;
356 	}
357 
358 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
359 
360 	/*
361 	 * Check against address spoofing/corruption.
362 	 */
363 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
364 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
365 		/*
366 		 * XXX: "badscope" is not very suitable for a multicast source.
367 		 */
368 		V_ip6stat.ip6s_badscope++;
369 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
370 		goto bad;
371 	}
372 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
373 	    !(m->m_flags & M_LOOP)) {
374 		/*
375 		 * In this case, the packet should come from the loopback
376 		 * interface.  However, we cannot just check the if_flags,
377 		 * because ip6_mloopback() passes the "actual" interface
378 		 * as the outgoing/incoming interface.
379 		 */
380 		V_ip6stat.ip6s_badscope++;
381 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
382 		goto bad;
383 	}
384 
385 #ifdef ALTQ
386 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
387 		/* packet is dropped by traffic conditioner */
388 		return;
389 	}
390 #endif
391 	/*
392 	 * The following check is not documented in specs.  A malicious
393 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
394 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
395 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
396 	 *
397 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
398 	 * support IPv4-less kernel compilation, we cannot support SIIT
399 	 * environment at all.  So, it makes more sense for us to reject any
400 	 * malicious packets for non-SIIT environment, than try to do a
401 	 * partial support for SIIT environment.
402 	 */
403 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
404 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
405 		V_ip6stat.ip6s_badscope++;
406 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
407 		goto bad;
408 	}
409 #if 0
410 	/*
411 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
412 	 *
413 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
414 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
415 	 * is revised to forbid relaying case.
416 	 */
417 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
418 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
419 		V_ip6stat.ip6s_badscope++;
420 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
421 		goto bad;
422 	}
423 #endif
424 
425 	/*
426 	 * Run through list of hooks for input packets.
427 	 *
428 	 * NB: Beware of the destination address changing
429 	 *     (e.g. by NAT rewriting).  When this happens,
430 	 *     tell ip6_forward to do the right thing.
431 	 */
432 	odst = ip6->ip6_dst;
433 
434 	/* Jump over all PFIL processing if hooks are not active. */
435 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
436 		goto passin;
437 
438 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
439 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
440 		return;
441 	if (m == NULL)			/* consumed by filter */
442 		return;
443 	ip6 = mtod(m, struct ip6_hdr *);
444 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
445 
446 passin:
447 	/*
448 	 * Disambiguate address scope zones (if there is ambiguity).
449 	 * We first make sure that the original source or destination address
450 	 * is not in our internal form for scoped addresses.  Such addresses
451 	 * are not necessarily invalid spec-wise, but we cannot accept them due
452 	 * to the usage conflict.
453 	 * in6_setscope() then also checks and rejects the cases where src or
454 	 * dst are the loopback address and the receiving interface
455 	 * is not loopback.
456 	 */
457 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
458 		V_ip6stat.ip6s_badscope++; /* XXX */
459 		goto bad;
460 	}
461 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
462 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
463 		V_ip6stat.ip6s_badscope++;
464 		goto bad;
465 	}
466 
467 	/*
468 	 * Multicast check. Assume packet is for us to avoid
469 	 * prematurely taking locks.
470 	 */
471 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
472 		ours = 1;
473 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
474 		deliverifp = m->m_pkthdr.rcvif;
475 		goto hbhcheck;
476 	}
477 
478 	/*
479 	 *  Unicast check
480 	 */
481 
482 	bzero(&dst6, sizeof(dst6));
483 	dst6.sin6_family = AF_INET6;
484 	dst6.sin6_len = sizeof(struct sockaddr_in6);
485 	dst6.sin6_addr = ip6->ip6_dst;
486 	ifp = m->m_pkthdr.rcvif;
487 	IF_AFDATA_LOCK(ifp);
488 	lle = lla_lookup(LLTABLE6(ifp), 0,
489 	     (struct sockaddr *)&dst6);
490 	IF_AFDATA_UNLOCK(ifp);
491 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
492 		ours = 1;
493 		deliverifp = ifp;
494 		LLE_RUNLOCK(lle);
495 		goto hbhcheck;
496 	}
497 	if (lle != NULL)
498 		LLE_RUNLOCK(lle);
499 
500 	dst = &rin6.ro_dst;
501 	dst->sin6_len = sizeof(struct sockaddr_in6);
502 	dst->sin6_family = AF_INET6;
503 	dst->sin6_addr = ip6->ip6_dst;
504 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
505 	if (rin6.ro_rt)
506 		RT_UNLOCK(rin6.ro_rt);
507 
508 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
509 
510 	/*
511 	 * Accept the packet if the forwarding interface to the destination
512 	 * according to the routing table is the loopback interface,
513 	 * unless the associated route has a gateway.
514 	 * Note that this approach causes to accept a packet if there is a
515 	 * route to the loopback interface for the destination of the packet.
516 	 * But we think it's even useful in some situations, e.g. when using
517 	 * a special daemon which wants to intercept the packet.
518 	 *
519 	 * XXX: some OSes automatically make a cloned route for the destination
520 	 * of an outgoing packet.  If the outgoing interface of the packet
521 	 * is a loopback one, the kernel would consider the packet to be
522 	 * accepted, even if we have no such address assinged on the interface.
523 	 * We check the cloned flag of the route entry to reject such cases,
524 	 * assuming that route entries for our own addresses are not made by
525 	 * cloning (it should be true because in6_addloop explicitly installs
526 	 * the host route).  However, we might have to do an explicit check
527 	 * while it would be less efficient.  Or, should we rather install a
528 	 * reject route for such a case?
529 	 */
530 	if (rin6.ro_rt &&
531 	    (rin6.ro_rt->rt_flags &
532 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
533 #ifdef RTF_WASCLONED
534 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
535 #endif
536 #ifdef RTF_CLONED
537 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
538 #endif
539 #if 0
540 	    /*
541 	     * The check below is redundant since the comparison of
542 	     * the destination and the key of the rtentry has
543 	     * already done through looking up the routing table.
544 	     */
545 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
546 	    &rt6_key(rin6.ro_rt)->sin6_addr)
547 #endif
548 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
549 		int free_ia6 = 0;
550 		struct in6_ifaddr *ia6;
551 
552 		/*
553 		 * found the loopback route to the interface address
554 		 */
555 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
556 			struct sockaddr_in6 dest6;
557 
558 			bzero(&dest6, sizeof(dest6));
559 			dest6.sin6_family = AF_INET6;
560 			dest6.sin6_len = sizeof(dest6);
561 			dest6.sin6_addr = ip6->ip6_dst;
562 			ia6 = (struct in6_ifaddr *)
563 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
564 			if (ia6 == NULL)
565 				goto bad;
566 			free_ia6 = 1;
567 		}
568 		else
569 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
570 
571 		/*
572 		 * record address information into m_tag.
573 		 */
574 		(void)ip6_setdstifaddr(m, ia6);
575 
576 		/*
577 		 * packets to a tentative, duplicated, or somehow invalid
578 		 * address must not be accepted.
579 		 */
580 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
581 			/* this address is ready */
582 			ours = 1;
583 			deliverifp = ia6->ia_ifp;	/* correct? */
584 			/* Count the packet in the ip address stats */
585 			ia6->ia_ifa.if_ipackets++;
586 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
587 			if (ia6 != NULL && free_ia6 != 0)
588 				ifa_free(&ia6->ia_ifa);
589 			goto hbhcheck;
590 		} else {
591 			char ip6bufs[INET6_ADDRSTRLEN];
592 			char ip6bufd[INET6_ADDRSTRLEN];
593 			/* address is not ready, so discard the packet. */
594 			nd6log((LOG_INFO,
595 			    "ip6_input: packet to an unready address %s->%s\n",
596 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
597 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
598 
599 			if (ia6 != NULL && free_ia6 != 0)
600 				ifa_free(&ia6->ia_ifa);
601 			goto bad;
602 		}
603 	}
604 
605 	/*
606 	 * FAITH (Firewall Aided Internet Translator)
607 	 */
608 	if (V_ip6_keepfaith) {
609 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
610 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
611 			/* XXX do we need more sanity checks? */
612 			ours = 1;
613 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
614 			goto hbhcheck;
615 		}
616 	}
617 
618 	/*
619 	 * Now there is no reason to process the packet if it's not our own
620 	 * and we're not a router.
621 	 */
622 	if (!V_ip6_forwarding) {
623 		V_ip6stat.ip6s_cantforward++;
624 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
625 		goto bad;
626 	}
627 
628   hbhcheck:
629 	/*
630 	 * record address information into m_tag, if we don't have one yet.
631 	 * note that we are unable to record it, if the address is not listed
632 	 * as our interface address (e.g. multicast addresses, addresses
633 	 * within FAITH prefixes and such).
634 	 */
635 	if (deliverifp && !ip6_getdstifaddr(m)) {
636 		struct in6_ifaddr *ia6;
637 
638 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
639 		if (ia6) {
640 			if (!ip6_setdstifaddr(m, ia6)) {
641 				/*
642 				 * XXX maybe we should drop the packet here,
643 				 * as we could not provide enough information
644 				 * to the upper layers.
645 				 */
646 			}
647 			ifa_free(&ia6->ia_ifa);
648 		}
649 	}
650 
651 	/*
652 	 * Process Hop-by-Hop options header if it's contained.
653 	 * m may be modified in ip6_hopopts_input().
654 	 * If a JumboPayload option is included, plen will also be modified.
655 	 */
656 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
657 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
658 		struct ip6_hbh *hbh;
659 
660 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
661 #if 0	/*touches NULL pointer*/
662 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
663 #endif
664 			goto out;	/* m have already been freed */
665 		}
666 
667 		/* adjust pointer */
668 		ip6 = mtod(m, struct ip6_hdr *);
669 
670 		/*
671 		 * if the payload length field is 0 and the next header field
672 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
673 		 * option MUST be included.
674 		 */
675 		if (ip6->ip6_plen == 0 && plen == 0) {
676 			/*
677 			 * Note that if a valid jumbo payload option is
678 			 * contained, ip6_hopopts_input() must set a valid
679 			 * (non-zero) payload length to the variable plen.
680 			 */
681 			V_ip6stat.ip6s_badoptions++;
682 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
683 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
684 			icmp6_error(m, ICMP6_PARAM_PROB,
685 				    ICMP6_PARAMPROB_HEADER,
686 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
687 			goto out;
688 		}
689 #ifndef PULLDOWN_TEST
690 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
691 		hbh = (struct ip6_hbh *)(ip6 + 1);
692 #else
693 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
694 			sizeof(struct ip6_hbh));
695 		if (hbh == NULL) {
696 			V_ip6stat.ip6s_tooshort++;
697 			goto out;
698 		}
699 #endif
700 		nxt = hbh->ip6h_nxt;
701 
702 		/*
703 		 * If we are acting as a router and the packet contains a
704 		 * router alert option, see if we know the option value.
705 		 * Currently, we only support the option value for MLD, in which
706 		 * case we should pass the packet to the multicast routing
707 		 * daemon.
708 		 */
709 		if (rtalert != ~0) {
710 			switch (rtalert) {
711 			case IP6OPT_RTALERT_MLD:
712 				if (V_ip6_forwarding)
713 					ours = 1;
714 				break;
715 			default:
716 				/*
717 				 * RFC2711 requires unrecognized values must be
718 				 * silently ignored.
719 				 */
720 				break;
721 			}
722 		}
723 	} else
724 		nxt = ip6->ip6_nxt;
725 
726 	/*
727 	 * Check that the amount of data in the buffers
728 	 * is as at least much as the IPv6 header would have us expect.
729 	 * Trim mbufs if longer than we expect.
730 	 * Drop packet if shorter than we expect.
731 	 */
732 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
733 		V_ip6stat.ip6s_tooshort++;
734 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
735 		goto bad;
736 	}
737 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
738 		if (m->m_len == m->m_pkthdr.len) {
739 			m->m_len = sizeof(struct ip6_hdr) + plen;
740 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
741 		} else
742 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
743 	}
744 
745 	/*
746 	 * Forward if desirable.
747 	 */
748 	if (V_ip6_mrouter &&
749 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
750 		/*
751 		 * If we are acting as a multicast router, all
752 		 * incoming multicast packets are passed to the
753 		 * kernel-level multicast forwarding function.
754 		 * The packet is returned (relatively) intact; if
755 		 * ip6_mforward() returns a non-zero value, the packet
756 		 * must be discarded, else it may be accepted below.
757 		 *
758 		 * XXX TODO: Check hlim and multicast scope here to avoid
759 		 * unnecessarily calling into ip6_mforward().
760 		 */
761 		if (ip6_mforward &&
762 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
763 			IP6STAT_INC(ip6s_cantforward);
764 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
765 			goto bad;
766 		}
767 	} else if (!ours) {
768 		ip6_forward(m, srcrt);
769 		goto out;
770 	}
771 
772 	ip6 = mtod(m, struct ip6_hdr *);
773 
774 	/*
775 	 * Malicious party may be able to use IPv4 mapped addr to confuse
776 	 * tcp/udp stack and bypass security checks (act as if it was from
777 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
778 	 *
779 	 * For SIIT end node behavior, you may want to disable the check.
780 	 * However, you will  become vulnerable to attacks using IPv4 mapped
781 	 * source.
782 	 */
783 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
784 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
785 		V_ip6stat.ip6s_badscope++;
786 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
787 		goto bad;
788 	}
789 
790 	/*
791 	 * Tell launch routine the next header
792 	 */
793 	V_ip6stat.ip6s_delivered++;
794 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
795 	nest = 0;
796 
797 	while (nxt != IPPROTO_DONE) {
798 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
799 			V_ip6stat.ip6s_toomanyhdr++;
800 			goto bad;
801 		}
802 
803 		/*
804 		 * protection against faulty packet - there should be
805 		 * more sanity checks in header chain processing.
806 		 */
807 		if (m->m_pkthdr.len < off) {
808 			V_ip6stat.ip6s_tooshort++;
809 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
810 			goto bad;
811 		}
812 
813 #ifdef IPSEC
814 		/*
815 		 * enforce IPsec policy checking if we are seeing last header.
816 		 * note that we do not visit this with protocols with pcb layer
817 		 * code - like udp/tcp/raw ip.
818 		 */
819 		if (ip6_ipsec_input(m, nxt))
820 			goto bad;
821 #endif /* IPSEC */
822 
823 		/*
824 		 * Use mbuf flags to propagate Router Alert option to
825 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
826 		 */
827 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
828 			m->m_flags |= M_RTALERT_MLD;
829 
830 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
831 	}
832 	goto out;
833 bad:
834 	m_freem(m);
835 out:
836 	if (rin6.ro_rt)
837 		RTFREE(rin6.ro_rt);
838 }
839 
840 /*
841  * set/grab in6_ifaddr correspond to IPv6 destination address.
842  * XXX backward compatibility wrapper
843  *
844  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
845  * and then bump it when the tag is copied, and release it when the tag is
846  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
847  * we just bump the ia refcount when we receive it.  This should be fixed.
848  */
849 static struct ip6aux *
850 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
851 {
852 	struct ip6aux *ip6a;
853 
854 	ip6a = ip6_addaux(m);
855 	if (ip6a)
856 		ip6a->ip6a_dstia6 = ia6;
857 	return ip6a;	/* NULL if failed to set */
858 }
859 
860 struct in6_ifaddr *
861 ip6_getdstifaddr(struct mbuf *m)
862 {
863 	struct ip6aux *ip6a;
864 	struct in6_ifaddr *ia;
865 
866 	ip6a = ip6_findaux(m);
867 	if (ip6a) {
868 		ia = ip6a->ip6a_dstia6;
869 		ifa_ref(&ia->ia_ifa);
870 		return ia;
871 	} else
872 		return NULL;
873 }
874 
875 /*
876  * Hop-by-Hop options header processing. If a valid jumbo payload option is
877  * included, the real payload length will be stored in plenp.
878  *
879  * rtalertp - XXX: should be stored more smart way
880  */
881 static int
882 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
883     struct mbuf **mp, int *offp)
884 {
885 	struct mbuf *m = *mp;
886 	int off = *offp, hbhlen;
887 	struct ip6_hbh *hbh;
888 	u_int8_t *opt;
889 
890 	/* validation of the length of the header */
891 #ifndef PULLDOWN_TEST
892 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
893 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
894 	hbhlen = (hbh->ip6h_len + 1) << 3;
895 
896 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
897 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
898 #else
899 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
900 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
901 	if (hbh == NULL) {
902 		V_ip6stat.ip6s_tooshort++;
903 		return -1;
904 	}
905 	hbhlen = (hbh->ip6h_len + 1) << 3;
906 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
907 		hbhlen);
908 	if (hbh == NULL) {
909 		V_ip6stat.ip6s_tooshort++;
910 		return -1;
911 	}
912 #endif
913 	off += hbhlen;
914 	hbhlen -= sizeof(struct ip6_hbh);
915 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
916 
917 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
918 				hbhlen, rtalertp, plenp) < 0)
919 		return (-1);
920 
921 	*offp = off;
922 	*mp = m;
923 	return (0);
924 }
925 
926 /*
927  * Search header for all Hop-by-hop options and process each option.
928  * This function is separate from ip6_hopopts_input() in order to
929  * handle a case where the sending node itself process its hop-by-hop
930  * options header. In such a case, the function is called from ip6_output().
931  *
932  * The function assumes that hbh header is located right after the IPv6 header
933  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
934  * opthead + hbhlen is located in continuous memory region.
935  */
936 int
937 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
938     u_int32_t *rtalertp, u_int32_t *plenp)
939 {
940 	struct ip6_hdr *ip6;
941 	int optlen = 0;
942 	u_int8_t *opt = opthead;
943 	u_int16_t rtalert_val;
944 	u_int32_t jumboplen;
945 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
946 
947 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
948 		switch (*opt) {
949 		case IP6OPT_PAD1:
950 			optlen = 1;
951 			break;
952 		case IP6OPT_PADN:
953 			if (hbhlen < IP6OPT_MINLEN) {
954 				V_ip6stat.ip6s_toosmall++;
955 				goto bad;
956 			}
957 			optlen = *(opt + 1) + 2;
958 			break;
959 		case IP6OPT_ROUTER_ALERT:
960 			/* XXX may need check for alignment */
961 			if (hbhlen < IP6OPT_RTALERT_LEN) {
962 				V_ip6stat.ip6s_toosmall++;
963 				goto bad;
964 			}
965 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
966 				/* XXX stat */
967 				icmp6_error(m, ICMP6_PARAM_PROB,
968 				    ICMP6_PARAMPROB_HEADER,
969 				    erroff + opt + 1 - opthead);
970 				return (-1);
971 			}
972 			optlen = IP6OPT_RTALERT_LEN;
973 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
974 			*rtalertp = ntohs(rtalert_val);
975 			break;
976 		case IP6OPT_JUMBO:
977 			/* XXX may need check for alignment */
978 			if (hbhlen < IP6OPT_JUMBO_LEN) {
979 				V_ip6stat.ip6s_toosmall++;
980 				goto bad;
981 			}
982 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
983 				/* XXX stat */
984 				icmp6_error(m, ICMP6_PARAM_PROB,
985 				    ICMP6_PARAMPROB_HEADER,
986 				    erroff + opt + 1 - opthead);
987 				return (-1);
988 			}
989 			optlen = IP6OPT_JUMBO_LEN;
990 
991 			/*
992 			 * IPv6 packets that have non 0 payload length
993 			 * must not contain a jumbo payload option.
994 			 */
995 			ip6 = mtod(m, struct ip6_hdr *);
996 			if (ip6->ip6_plen) {
997 				V_ip6stat.ip6s_badoptions++;
998 				icmp6_error(m, ICMP6_PARAM_PROB,
999 				    ICMP6_PARAMPROB_HEADER,
1000 				    erroff + opt - opthead);
1001 				return (-1);
1002 			}
1003 
1004 			/*
1005 			 * We may see jumbolen in unaligned location, so
1006 			 * we'd need to perform bcopy().
1007 			 */
1008 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1009 			jumboplen = (u_int32_t)htonl(jumboplen);
1010 
1011 #if 1
1012 			/*
1013 			 * if there are multiple jumbo payload options,
1014 			 * *plenp will be non-zero and the packet will be
1015 			 * rejected.
1016 			 * the behavior may need some debate in ipngwg -
1017 			 * multiple options does not make sense, however,
1018 			 * there's no explicit mention in specification.
1019 			 */
1020 			if (*plenp != 0) {
1021 				V_ip6stat.ip6s_badoptions++;
1022 				icmp6_error(m, ICMP6_PARAM_PROB,
1023 				    ICMP6_PARAMPROB_HEADER,
1024 				    erroff + opt + 2 - opthead);
1025 				return (-1);
1026 			}
1027 #endif
1028 
1029 			/*
1030 			 * jumbo payload length must be larger than 65535.
1031 			 */
1032 			if (jumboplen <= IPV6_MAXPACKET) {
1033 				V_ip6stat.ip6s_badoptions++;
1034 				icmp6_error(m, ICMP6_PARAM_PROB,
1035 				    ICMP6_PARAMPROB_HEADER,
1036 				    erroff + opt + 2 - opthead);
1037 				return (-1);
1038 			}
1039 			*plenp = jumboplen;
1040 
1041 			break;
1042 		default:		/* unknown option */
1043 			if (hbhlen < IP6OPT_MINLEN) {
1044 				V_ip6stat.ip6s_toosmall++;
1045 				goto bad;
1046 			}
1047 			optlen = ip6_unknown_opt(opt, m,
1048 			    erroff + opt - opthead);
1049 			if (optlen == -1)
1050 				return (-1);
1051 			optlen += 2;
1052 			break;
1053 		}
1054 	}
1055 
1056 	return (0);
1057 
1058   bad:
1059 	m_freem(m);
1060 	return (-1);
1061 }
1062 
1063 /*
1064  * Unknown option processing.
1065  * The third argument `off' is the offset from the IPv6 header to the option,
1066  * which is necessary if the IPv6 header the and option header and IPv6 header
1067  * is not continuous in order to return an ICMPv6 error.
1068  */
1069 int
1070 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1071 {
1072 	struct ip6_hdr *ip6;
1073 
1074 	switch (IP6OPT_TYPE(*optp)) {
1075 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1076 		return ((int)*(optp + 1));
1077 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1078 		m_freem(m);
1079 		return (-1);
1080 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1081 		V_ip6stat.ip6s_badoptions++;
1082 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1083 		return (-1);
1084 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1085 		V_ip6stat.ip6s_badoptions++;
1086 		ip6 = mtod(m, struct ip6_hdr *);
1087 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1088 		    (m->m_flags & (M_BCAST|M_MCAST)))
1089 			m_freem(m);
1090 		else
1091 			icmp6_error(m, ICMP6_PARAM_PROB,
1092 				    ICMP6_PARAMPROB_OPTION, off);
1093 		return (-1);
1094 	}
1095 
1096 	m_freem(m);		/* XXX: NOTREACHED */
1097 	return (-1);
1098 }
1099 
1100 /*
1101  * Create the "control" list for this pcb.
1102  * These functions will not modify mbuf chain at all.
1103  *
1104  * With KAME mbuf chain restriction:
1105  * The routine will be called from upper layer handlers like tcp6_input().
1106  * Thus the routine assumes that the caller (tcp6_input) have already
1107  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1108  * very first mbuf on the mbuf chain.
1109  *
1110  * ip6_savecontrol_v4 will handle those options that are possible to be
1111  * set on a v4-mapped socket.
1112  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1113  * options and handle the v6-only ones itself.
1114  */
1115 struct mbuf **
1116 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1117     int *v4only)
1118 {
1119 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1120 
1121 #ifdef SO_TIMESTAMP
1122 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1123 		struct timeval tv;
1124 
1125 		microtime(&tv);
1126 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1127 		    SCM_TIMESTAMP, SOL_SOCKET);
1128 		if (*mp)
1129 			mp = &(*mp)->m_next;
1130 	}
1131 #endif
1132 
1133 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1134 		if (v4only != NULL)
1135 			*v4only = 1;
1136 		return (mp);
1137 	}
1138 
1139 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1140 	/* RFC 2292 sec. 5 */
1141 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1142 		struct in6_pktinfo pi6;
1143 
1144 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1145 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1146 		pi6.ipi6_ifindex =
1147 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1148 
1149 		*mp = sbcreatecontrol((caddr_t) &pi6,
1150 		    sizeof(struct in6_pktinfo),
1151 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1152 		if (*mp)
1153 			mp = &(*mp)->m_next;
1154 	}
1155 
1156 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1157 		int hlim = ip6->ip6_hlim & 0xff;
1158 
1159 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1160 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1161 		    IPPROTO_IPV6);
1162 		if (*mp)
1163 			mp = &(*mp)->m_next;
1164 	}
1165 
1166 	if (v4only != NULL)
1167 		*v4only = 0;
1168 	return (mp);
1169 }
1170 
1171 void
1172 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1173 {
1174 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1175 	int v4only = 0;
1176 
1177 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1178 	if (v4only)
1179 		return;
1180 
1181 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1182 		u_int32_t flowinfo;
1183 		int tclass;
1184 
1185 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1186 		flowinfo >>= 20;
1187 
1188 		tclass = flowinfo & 0xff;
1189 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1190 		    IPV6_TCLASS, IPPROTO_IPV6);
1191 		if (*mp)
1192 			mp = &(*mp)->m_next;
1193 	}
1194 
1195 	/*
1196 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1197 	 * privilege for the option (see ip6_ctloutput), but it might be too
1198 	 * strict, since there might be some hop-by-hop options which can be
1199 	 * returned to normal user.
1200 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1201 	 */
1202 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1203 		/*
1204 		 * Check if a hop-by-hop options header is contatined in the
1205 		 * received packet, and if so, store the options as ancillary
1206 		 * data. Note that a hop-by-hop options header must be
1207 		 * just after the IPv6 header, which is assured through the
1208 		 * IPv6 input processing.
1209 		 */
1210 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1211 			struct ip6_hbh *hbh;
1212 			int hbhlen = 0;
1213 #ifdef PULLDOWN_TEST
1214 			struct mbuf *ext;
1215 #endif
1216 
1217 #ifndef PULLDOWN_TEST
1218 			hbh = (struct ip6_hbh *)(ip6 + 1);
1219 			hbhlen = (hbh->ip6h_len + 1) << 3;
1220 #else
1221 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1222 			    ip6->ip6_nxt);
1223 			if (ext == NULL) {
1224 				V_ip6stat.ip6s_tooshort++;
1225 				return;
1226 			}
1227 			hbh = mtod(ext, struct ip6_hbh *);
1228 			hbhlen = (hbh->ip6h_len + 1) << 3;
1229 			if (hbhlen != ext->m_len) {
1230 				m_freem(ext);
1231 				V_ip6stat.ip6s_tooshort++;
1232 				return;
1233 			}
1234 #endif
1235 
1236 			/*
1237 			 * XXX: We copy the whole header even if a
1238 			 * jumbo payload option is included, the option which
1239 			 * is to be removed before returning according to
1240 			 * RFC2292.
1241 			 * Note: this constraint is removed in RFC3542
1242 			 */
1243 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1244 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1245 			    IPPROTO_IPV6);
1246 			if (*mp)
1247 				mp = &(*mp)->m_next;
1248 #ifdef PULLDOWN_TEST
1249 			m_freem(ext);
1250 #endif
1251 		}
1252 	}
1253 
1254 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1255 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1256 
1257 		/*
1258 		 * Search for destination options headers or routing
1259 		 * header(s) through the header chain, and stores each
1260 		 * header as ancillary data.
1261 		 * Note that the order of the headers remains in
1262 		 * the chain of ancillary data.
1263 		 */
1264 		while (1) {	/* is explicit loop prevention necessary? */
1265 			struct ip6_ext *ip6e = NULL;
1266 			int elen;
1267 #ifdef PULLDOWN_TEST
1268 			struct mbuf *ext = NULL;
1269 #endif
1270 
1271 			/*
1272 			 * if it is not an extension header, don't try to
1273 			 * pull it from the chain.
1274 			 */
1275 			switch (nxt) {
1276 			case IPPROTO_DSTOPTS:
1277 			case IPPROTO_ROUTING:
1278 			case IPPROTO_HOPOPTS:
1279 			case IPPROTO_AH: /* is it possible? */
1280 				break;
1281 			default:
1282 				goto loopend;
1283 			}
1284 
1285 #ifndef PULLDOWN_TEST
1286 			if (off + sizeof(*ip6e) > m->m_len)
1287 				goto loopend;
1288 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1289 			if (nxt == IPPROTO_AH)
1290 				elen = (ip6e->ip6e_len + 2) << 2;
1291 			else
1292 				elen = (ip6e->ip6e_len + 1) << 3;
1293 			if (off + elen > m->m_len)
1294 				goto loopend;
1295 #else
1296 			ext = ip6_pullexthdr(m, off, nxt);
1297 			if (ext == NULL) {
1298 				V_ip6stat.ip6s_tooshort++;
1299 				return;
1300 			}
1301 			ip6e = mtod(ext, struct ip6_ext *);
1302 			if (nxt == IPPROTO_AH)
1303 				elen = (ip6e->ip6e_len + 2) << 2;
1304 			else
1305 				elen = (ip6e->ip6e_len + 1) << 3;
1306 			if (elen != ext->m_len) {
1307 				m_freem(ext);
1308 				V_ip6stat.ip6s_tooshort++;
1309 				return;
1310 			}
1311 #endif
1312 
1313 			switch (nxt) {
1314 			case IPPROTO_DSTOPTS:
1315 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1316 					break;
1317 
1318 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1319 				    IS2292(in6p,
1320 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1321 				    IPPROTO_IPV6);
1322 				if (*mp)
1323 					mp = &(*mp)->m_next;
1324 				break;
1325 			case IPPROTO_ROUTING:
1326 				if (!(in6p->inp_flags & IN6P_RTHDR))
1327 					break;
1328 
1329 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1330 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1331 				    IPPROTO_IPV6);
1332 				if (*mp)
1333 					mp = &(*mp)->m_next;
1334 				break;
1335 			case IPPROTO_HOPOPTS:
1336 			case IPPROTO_AH: /* is it possible? */
1337 				break;
1338 
1339 			default:
1340 				/*
1341 				 * other cases have been filtered in the above.
1342 				 * none will visit this case.  here we supply
1343 				 * the code just in case (nxt overwritten or
1344 				 * other cases).
1345 				 */
1346 #ifdef PULLDOWN_TEST
1347 				m_freem(ext);
1348 #endif
1349 				goto loopend;
1350 
1351 			}
1352 
1353 			/* proceed with the next header. */
1354 			off += elen;
1355 			nxt = ip6e->ip6e_nxt;
1356 			ip6e = NULL;
1357 #ifdef PULLDOWN_TEST
1358 			m_freem(ext);
1359 			ext = NULL;
1360 #endif
1361 		}
1362 	  loopend:
1363 		;
1364 	}
1365 }
1366 #undef IS2292
1367 
1368 void
1369 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1370 {
1371 	struct socket *so;
1372 	struct mbuf *m_mtu;
1373 	struct ip6_mtuinfo mtuctl;
1374 
1375 	so =  in6p->inp_socket;
1376 
1377 	if (mtu == NULL)
1378 		return;
1379 
1380 #ifdef DIAGNOSTIC
1381 	if (so == NULL)		/* I believe this is impossible */
1382 		panic("ip6_notify_pmtu: socket is NULL");
1383 #endif
1384 
1385 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1386 	mtuctl.ip6m_mtu = *mtu;
1387 	mtuctl.ip6m_addr = *dst;
1388 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1389 		return;
1390 
1391 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1392 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1393 		return;
1394 
1395 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1396 	    == 0) {
1397 		m_freem(m_mtu);
1398 		/* XXX: should count statistics */
1399 	} else
1400 		sorwakeup(so);
1401 
1402 	return;
1403 }
1404 
1405 #ifdef PULLDOWN_TEST
1406 /*
1407  * pull single extension header from mbuf chain.  returns single mbuf that
1408  * contains the result, or NULL on error.
1409  */
1410 static struct mbuf *
1411 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1412 {
1413 	struct ip6_ext ip6e;
1414 	size_t elen;
1415 	struct mbuf *n;
1416 
1417 #ifdef DIAGNOSTIC
1418 	switch (nxt) {
1419 	case IPPROTO_DSTOPTS:
1420 	case IPPROTO_ROUTING:
1421 	case IPPROTO_HOPOPTS:
1422 	case IPPROTO_AH: /* is it possible? */
1423 		break;
1424 	default:
1425 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1426 	}
1427 #endif
1428 
1429 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1430 	if (nxt == IPPROTO_AH)
1431 		elen = (ip6e.ip6e_len + 2) << 2;
1432 	else
1433 		elen = (ip6e.ip6e_len + 1) << 3;
1434 
1435 	MGET(n, M_DONTWAIT, MT_DATA);
1436 	if (n && elen >= MLEN) {
1437 		MCLGET(n, M_DONTWAIT);
1438 		if ((n->m_flags & M_EXT) == 0) {
1439 			m_free(n);
1440 			n = NULL;
1441 		}
1442 	}
1443 	if (!n)
1444 		return NULL;
1445 
1446 	n->m_len = 0;
1447 	if (elen >= M_TRAILINGSPACE(n)) {
1448 		m_free(n);
1449 		return NULL;
1450 	}
1451 
1452 	m_copydata(m, off, elen, mtod(n, caddr_t));
1453 	n->m_len = elen;
1454 	return n;
1455 }
1456 #endif
1457 
1458 /*
1459  * Get pointer to the previous header followed by the header
1460  * currently processed.
1461  * XXX: This function supposes that
1462  *	M includes all headers,
1463  *	the next header field and the header length field of each header
1464  *	are valid, and
1465  *	the sum of each header length equals to OFF.
1466  * Because of these assumptions, this function must be called very
1467  * carefully. Moreover, it will not be used in the near future when
1468  * we develop `neater' mechanism to process extension headers.
1469  */
1470 char *
1471 ip6_get_prevhdr(struct mbuf *m, int off)
1472 {
1473 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1474 
1475 	if (off == sizeof(struct ip6_hdr))
1476 		return (&ip6->ip6_nxt);
1477 	else {
1478 		int len, nxt;
1479 		struct ip6_ext *ip6e = NULL;
1480 
1481 		nxt = ip6->ip6_nxt;
1482 		len = sizeof(struct ip6_hdr);
1483 		while (len < off) {
1484 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1485 
1486 			switch (nxt) {
1487 			case IPPROTO_FRAGMENT:
1488 				len += sizeof(struct ip6_frag);
1489 				break;
1490 			case IPPROTO_AH:
1491 				len += (ip6e->ip6e_len + 2) << 2;
1492 				break;
1493 			default:
1494 				len += (ip6e->ip6e_len + 1) << 3;
1495 				break;
1496 			}
1497 			nxt = ip6e->ip6e_nxt;
1498 		}
1499 		if (ip6e)
1500 			return (&ip6e->ip6e_nxt);
1501 		else
1502 			return NULL;
1503 	}
1504 }
1505 
1506 /*
1507  * get next header offset.  m will be retained.
1508  */
1509 int
1510 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1511 {
1512 	struct ip6_hdr ip6;
1513 	struct ip6_ext ip6e;
1514 	struct ip6_frag fh;
1515 
1516 	/* just in case */
1517 	if (m == NULL)
1518 		panic("ip6_nexthdr: m == NULL");
1519 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1520 		return -1;
1521 
1522 	switch (proto) {
1523 	case IPPROTO_IPV6:
1524 		if (m->m_pkthdr.len < off + sizeof(ip6))
1525 			return -1;
1526 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1527 		if (nxtp)
1528 			*nxtp = ip6.ip6_nxt;
1529 		off += sizeof(ip6);
1530 		return off;
1531 
1532 	case IPPROTO_FRAGMENT:
1533 		/*
1534 		 * terminate parsing if it is not the first fragment,
1535 		 * it does not make sense to parse through it.
1536 		 */
1537 		if (m->m_pkthdr.len < off + sizeof(fh))
1538 			return -1;
1539 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1540 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1541 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1542 			return -1;
1543 		if (nxtp)
1544 			*nxtp = fh.ip6f_nxt;
1545 		off += sizeof(struct ip6_frag);
1546 		return off;
1547 
1548 	case IPPROTO_AH:
1549 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1550 			return -1;
1551 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1552 		if (nxtp)
1553 			*nxtp = ip6e.ip6e_nxt;
1554 		off += (ip6e.ip6e_len + 2) << 2;
1555 		return off;
1556 
1557 	case IPPROTO_HOPOPTS:
1558 	case IPPROTO_ROUTING:
1559 	case IPPROTO_DSTOPTS:
1560 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1561 			return -1;
1562 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1563 		if (nxtp)
1564 			*nxtp = ip6e.ip6e_nxt;
1565 		off += (ip6e.ip6e_len + 1) << 3;
1566 		return off;
1567 
1568 	case IPPROTO_NONE:
1569 	case IPPROTO_ESP:
1570 	case IPPROTO_IPCOMP:
1571 		/* give up */
1572 		return -1;
1573 
1574 	default:
1575 		return -1;
1576 	}
1577 
1578 	return -1;
1579 }
1580 
1581 /*
1582  * get offset for the last header in the chain.  m will be kept untainted.
1583  */
1584 int
1585 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1586 {
1587 	int newoff;
1588 	int nxt;
1589 
1590 	if (!nxtp) {
1591 		nxt = -1;
1592 		nxtp = &nxt;
1593 	}
1594 	while (1) {
1595 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1596 		if (newoff < 0)
1597 			return off;
1598 		else if (newoff < off)
1599 			return -1;	/* invalid */
1600 		else if (newoff == off)
1601 			return newoff;
1602 
1603 		off = newoff;
1604 		proto = *nxtp;
1605 	}
1606 }
1607 
1608 struct ip6aux *
1609 ip6_addaux(struct mbuf *m)
1610 {
1611 	struct m_tag *mtag;
1612 
1613 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1614 	if (!mtag) {
1615 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1616 		    M_NOWAIT);
1617 		if (mtag) {
1618 			m_tag_prepend(m, mtag);
1619 			bzero(mtag + 1, sizeof(struct ip6aux));
1620 		}
1621 	}
1622 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1623 }
1624 
1625 struct ip6aux *
1626 ip6_findaux(struct mbuf *m)
1627 {
1628 	struct m_tag *mtag;
1629 
1630 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1631 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1632 }
1633 
1634 void
1635 ip6_delaux(struct mbuf *m)
1636 {
1637 	struct m_tag *mtag;
1638 
1639 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1640 	if (mtag)
1641 		m_tag_delete(m, mtag);
1642 }
1643 
1644 /*
1645  * System control for IP6
1646  */
1647 
1648 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1649 	0,		0,		0,		0,
1650 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1651 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1652 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1653 	0,		0,		0,		0,
1654 	ENOPROTOOPT
1655 };
1656