xref: /freebsd/sys/netinet6/ip6_input.c (revision acd3428b7d3e94cef0e1881c868cb4b131d4ff41)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*-
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/proc.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/kernel.h>
80 #include <sys/syslog.h>
81 
82 #include <net/if.h>
83 #include <net/if_types.h>
84 #include <net/if_dl.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 #include <net/pfil.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_systm.h>
91 #ifdef INET
92 #include <netinet/ip.h>
93 #include <netinet/ip_icmp.h>
94 #endif /* INET */
95 #include <netinet/ip6.h>
96 #include <netinet6/in6_var.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/scope6_var.h>
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/nd6.h>
103 
104 #ifdef IPSEC
105 #include <netinet6/ipsec.h>
106 #ifdef INET6
107 #include <netinet6/ipsec6.h>
108 #endif
109 #endif
110 
111 #ifdef FAST_IPSEC
112 #include <netipsec/ipsec.h>
113 #include <netipsec/ipsec6.h>
114 #define	IPSEC
115 #endif /* FAST_IPSEC */
116 
117 #include <netinet6/ip6protosw.h>
118 
119 extern struct domain inet6domain;
120 
121 u_char ip6_protox[IPPROTO_MAX];
122 static struct ifqueue ip6intrq;
123 static int ip6qmaxlen = IFQ_MAXLEN;
124 struct in6_ifaddr *in6_ifaddr;
125 
126 extern struct callout in6_tmpaddrtimer_ch;
127 
128 int ip6_forward_srcrt;			/* XXX */
129 int ip6_sourcecheck;			/* XXX */
130 int ip6_sourcecheck_interval;		/* XXX */
131 
132 int ip6_ours_check_algorithm;
133 
134 struct pfil_head inet6_pfil_hook;
135 
136 struct ip6stat ip6stat;
137 
138 static void ip6_init2 __P((void *));
139 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
140 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
141 #ifdef PULLDOWN_TEST
142 static struct mbuf *ip6_pullexthdr __P((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()
151 {
152 	struct ip6protosw *pr;
153 	int i;
154 
155 #ifdef DIAGNOSTIC
156 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
157 		panic("sizeof(protosw) != sizeof(ip6protosw)");
158 #endif
159 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
160 	if (pr == 0)
161 		panic("ip6_init");
162 
163 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
164 	for (i = 0; i < IPPROTO_MAX; i++)
165 		ip6_protox[i] = pr - inet6sw;
166 	/*
167 	 * Cycle through IP protocols and put them into the appropriate place
168 	 * in ip6_protox[].
169 	 */
170 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
171 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
172 		if (pr->pr_domain->dom_family == PF_INET6 &&
173 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
174 			/* Be careful to only index valid IP protocols. */
175 			if (pr->pr_protocol < IPPROTO_MAX)
176 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
177 		}
178 
179 	/* Initialize packet filter hooks. */
180 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
181 	inet6_pfil_hook.ph_af = AF_INET6;
182 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
183 		printf("%s: WARNING: unable to register pfil hook, "
184 			"error %d\n", __func__, i);
185 
186 	ip6intrq.ifq_maxlen = ip6qmaxlen;
187 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
188 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
189 	scope6_init();
190 	addrsel_policy_init();
191 	nd6_init();
192 	frag6_init();
193 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
194 }
195 
196 static void
197 ip6_init2(dummy)
198 	void *dummy;
199 {
200 
201 	/* nd6_timer_init */
202 	callout_init(&nd6_timer_ch, 0);
203 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
204 
205 	/* timer for regeneranation of temporary addresses randomize ID */
206 	callout_init(&in6_tmpaddrtimer_ch, 0);
207 	callout_reset(&in6_tmpaddrtimer_ch,
208 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
209 		       ip6_temp_regen_advance) * hz,
210 		      in6_tmpaddrtimer, NULL);
211 }
212 
213 /* cheat */
214 /* This must be after route_init(), which is now SI_ORDER_THIRD */
215 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
216 
217 extern struct	route_in6 ip6_forward_rt;
218 
219 void
220 ip6_input(m)
221 	struct mbuf *m;
222 {
223 	struct ip6_hdr *ip6;
224 	int off = sizeof(struct ip6_hdr), nest;
225 	u_int32_t plen;
226 	u_int32_t rtalert = ~0;
227 	int nxt, ours = 0;
228 	struct ifnet *deliverifp = NULL;
229 	struct in6_addr odst;
230 	int srcrt = 0;
231 
232 	GIANT_REQUIRED;			/* XXX for now */
233 #ifdef IPSEC
234 	/*
235 	 * should the inner packet be considered authentic?
236 	 * see comment in ah4_input().
237 	 */
238 	if (m) {
239 		m->m_flags &= ~M_AUTHIPHDR;
240 		m->m_flags &= ~M_AUTHIPDGM;
241 	}
242 #endif
243 
244 	/*
245 	 * make sure we don't have onion peering information into m_tag.
246 	 */
247 	ip6_delaux(m);
248 
249 	/*
250 	 * mbuf statistics
251 	 */
252 	if (m->m_flags & M_EXT) {
253 		if (m->m_next)
254 			ip6stat.ip6s_mext2m++;
255 		else
256 			ip6stat.ip6s_mext1++;
257 	} else {
258 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
259 		if (m->m_next) {
260 			if (m->m_flags & M_LOOP) {
261 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
262 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
263 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
264 			else
265 				ip6stat.ip6s_m2m[0]++;
266 		} else
267 			ip6stat.ip6s_m1++;
268 #undef M2MMAX
269 	}
270 
271 	/* drop the packet if IPv6 operation is disabled on the IF */
272 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
273 		m_freem(m);
274 		return;
275 	}
276 
277 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
278 	ip6stat.ip6s_total++;
279 
280 #ifndef PULLDOWN_TEST
281 	/*
282 	 * L2 bridge code and some other code can return mbuf chain
283 	 * that does not conform to KAME requirement.  too bad.
284 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
285 	 */
286 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
287 		struct mbuf *n;
288 
289 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
290 		if (n)
291 			M_MOVE_PKTHDR(n, m);
292 		if (n && n->m_pkthdr.len > MHLEN) {
293 			MCLGET(n, M_DONTWAIT);
294 			if ((n->m_flags & M_EXT) == 0) {
295 				m_freem(n);
296 				n = NULL;
297 			}
298 		}
299 		if (n == NULL) {
300 			m_freem(m);
301 			return;	/* ENOBUFS */
302 		}
303 
304 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
305 		n->m_len = n->m_pkthdr.len;
306 		m_freem(m);
307 		m = n;
308 	}
309 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
310 #endif
311 
312 	if (m->m_len < sizeof(struct ip6_hdr)) {
313 		struct ifnet *inifp;
314 		inifp = m->m_pkthdr.rcvif;
315 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
316 			ip6stat.ip6s_toosmall++;
317 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
318 			return;
319 		}
320 	}
321 
322 	ip6 = mtod(m, struct ip6_hdr *);
323 
324 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
325 		ip6stat.ip6s_badvers++;
326 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
327 		goto bad;
328 	}
329 
330 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
331 
332 	/*
333 	 * Check against address spoofing/corruption.
334 	 */
335 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
336 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
337 		/*
338 		 * XXX: "badscope" is not very suitable for a multicast source.
339 		 */
340 		ip6stat.ip6s_badscope++;
341 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
342 		goto bad;
343 	}
344 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
345 	    !(m->m_flags & M_LOOP)) {
346 		/*
347 		 * In this case, the packet should come from the loopback
348 		 * interface.  However, we cannot just check the if_flags,
349 		 * because ip6_mloopback() passes the "actual" interface
350 		 * as the outgoing/incoming interface.
351 		 */
352 		ip6stat.ip6s_badscope++;
353 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
354 		goto bad;
355 	}
356 
357 #ifdef ALTQ
358 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
359 		/* packet is dropped by traffic conditioner */
360 		return;
361 	}
362 #endif
363 	/*
364 	 * The following check is not documented in specs.  A malicious
365 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
366 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
367 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
368 	 *
369 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
370 	 * support IPv4-less kernel compilation, we cannot support SIIT
371 	 * environment at all.  So, it makes more sense for us to reject any
372 	 * malicious packets for non-SIIT environment, than try to do a
373 	 * partial support for SIIT environment.
374 	 */
375 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
376 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
377 		ip6stat.ip6s_badscope++;
378 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
379 		goto bad;
380 	}
381 #if 0
382 	/*
383 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
384 	 *
385 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
386 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
387 	 * is revised to forbid relaying case.
388 	 */
389 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
390 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
391 		ip6stat.ip6s_badscope++;
392 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
393 		goto bad;
394 	}
395 #endif
396 
397 	/*
398 	 * Run through list of hooks for input packets.
399 	 *
400 	 * NB: Beware of the destination address changing
401 	 *     (e.g. by NAT rewriting).  When this happens,
402 	 *     tell ip6_forward to do the right thing.
403 	 */
404 	odst = ip6->ip6_dst;
405 
406 	/* Jump over all PFIL processing if hooks are not active. */
407 	if (!PFIL_HOOKED(&inet6_pfil_hook))
408 		goto passin;
409 
410 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
411 		return;
412 	if (m == NULL)			/* consumed by filter */
413 		return;
414 	ip6 = mtod(m, struct ip6_hdr *);
415 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
416 
417 passin:
418 	/*
419 	 * Disambiguate address scope zones (if there is ambiguity).
420 	 * We first make sure that the original source or destination address
421 	 * is not in our internal form for scoped addresses.  Such addresses
422 	 * are not necessarily invalid spec-wise, but we cannot accept them due
423 	 * to the usage conflict.
424 	 * in6_setscope() then also checks and rejects the cases where src or
425 	 * dst are the loopback address and the receiving interface
426 	 * is not loopback.
427 	 */
428 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
429 		ip6stat.ip6s_badscope++; /* XXX */
430 		goto bad;
431 	}
432 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
433 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
434 		ip6stat.ip6s_badscope++;
435 		goto bad;
436 	}
437 
438 	/*
439 	 * Multicast check
440 	 */
441 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
442 	  	struct in6_multi *in6m = 0;
443 
444 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
445 		/*
446 		 * See if we belong to the destination multicast group on the
447 		 * arrival interface.
448 		 */
449 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
450 		if (in6m)
451 			ours = 1;
452 		else if (!ip6_mrouter) {
453 			ip6stat.ip6s_notmember++;
454 			ip6stat.ip6s_cantforward++;
455 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
456 			goto bad;
457 		}
458 		deliverifp = m->m_pkthdr.rcvif;
459 		goto hbhcheck;
460 	}
461 
462 	/*
463 	 *  Unicast check
464 	 */
465 	if (ip6_forward_rt.ro_rt != NULL &&
466 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
467 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
468 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
469 		ip6stat.ip6s_forward_cachehit++;
470 	else {
471 		struct sockaddr_in6 *dst6;
472 
473 		if (ip6_forward_rt.ro_rt) {
474 			/* route is down or destination is different */
475 			ip6stat.ip6s_forward_cachemiss++;
476 			RTFREE(ip6_forward_rt.ro_rt);
477 			ip6_forward_rt.ro_rt = 0;
478 		}
479 
480 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
481 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
482 		dst6->sin6_len = sizeof(struct sockaddr_in6);
483 		dst6->sin6_family = AF_INET6;
484 		dst6->sin6_addr = ip6->ip6_dst;
485 
486 		rtalloc((struct route *)&ip6_forward_rt);
487 	}
488 
489 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
490 
491 	/*
492 	 * Accept the packet if the forwarding interface to the destination
493 	 * according to the routing table is the loopback interface,
494 	 * unless the associated route has a gateway.
495 	 * Note that this approach causes to accept a packet if there is a
496 	 * route to the loopback interface for the destination of the packet.
497 	 * But we think it's even useful in some situations, e.g. when using
498 	 * a special daemon which wants to intercept the packet.
499 	 *
500 	 * XXX: some OSes automatically make a cloned route for the destination
501 	 * of an outgoing packet.  If the outgoing interface of the packet
502 	 * is a loopback one, the kernel would consider the packet to be
503 	 * accepted, even if we have no such address assinged on the interface.
504 	 * We check the cloned flag of the route entry to reject such cases,
505 	 * assuming that route entries for our own addresses are not made by
506 	 * cloning (it should be true because in6_addloop explicitly installs
507 	 * the host route).  However, we might have to do an explicit check
508 	 * while it would be less efficient.  Or, should we rather install a
509 	 * reject route for such a case?
510 	 */
511 	if (ip6_forward_rt.ro_rt &&
512 	    (ip6_forward_rt.ro_rt->rt_flags &
513 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
514 #ifdef RTF_WASCLONED
515 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
516 #endif
517 #ifdef RTF_CLONED
518 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
519 #endif
520 #if 0
521 	    /*
522 	     * The check below is redundant since the comparison of
523 	     * the destination and the key of the rtentry has
524 	     * already done through looking up the routing table.
525 	     */
526 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
527 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
528 #endif
529 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
530 		struct in6_ifaddr *ia6 =
531 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
532 
533 		/*
534 		 * record address information into m_tag.
535 		 */
536 		(void)ip6_setdstifaddr(m, ia6);
537 
538 		/*
539 		 * packets to a tentative, duplicated, or somehow invalid
540 		 * address must not be accepted.
541 		 */
542 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
543 			/* this address is ready */
544 			ours = 1;
545 			deliverifp = ia6->ia_ifp;	/* correct? */
546 			/* Count the packet in the ip address stats */
547 			ia6->ia_ifa.if_ipackets++;
548 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
549 			goto hbhcheck;
550 		} else {
551 			/* address is not ready, so discard the packet. */
552 			nd6log((LOG_INFO,
553 			    "ip6_input: packet to an unready address %s->%s\n",
554 			    ip6_sprintf(&ip6->ip6_src),
555 			    ip6_sprintf(&ip6->ip6_dst)));
556 
557 			goto bad;
558 		}
559 	}
560 
561 	/*
562 	 * FAITH (Firewall Aided Internet Translator)
563 	 */
564 	if (ip6_keepfaith) {
565 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
566 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
567 			/* XXX do we need more sanity checks? */
568 			ours = 1;
569 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
570 			goto hbhcheck;
571 		}
572 	}
573 
574 	/*
575 	 * Now there is no reason to process the packet if it's not our own
576 	 * and we're not a router.
577 	 */
578 	if (!ip6_forwarding) {
579 		ip6stat.ip6s_cantforward++;
580 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
581 		goto bad;
582 	}
583 
584   hbhcheck:
585 	/*
586 	 * record address information into m_tag, if we don't have one yet.
587 	 * note that we are unable to record it, if the address is not listed
588 	 * as our interface address (e.g. multicast addresses, addresses
589 	 * within FAITH prefixes and such).
590 	 */
591 	if (deliverifp && !ip6_getdstifaddr(m)) {
592 		struct in6_ifaddr *ia6;
593 
594 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
595 		if (ia6) {
596 			if (!ip6_setdstifaddr(m, ia6)) {
597 				/*
598 				 * XXX maybe we should drop the packet here,
599 				 * as we could not provide enough information
600 				 * to the upper layers.
601 				 */
602 			}
603 		}
604 	}
605 
606 	/*
607 	 * Process Hop-by-Hop options header if it's contained.
608 	 * m may be modified in ip6_hopopts_input().
609 	 * If a JumboPayload option is included, plen will also be modified.
610 	 */
611 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
612 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
613 		struct ip6_hbh *hbh;
614 
615 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
616 #if 0	/*touches NULL pointer*/
617 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
618 #endif
619 			return;	/* m have already been freed */
620 		}
621 
622 		/* adjust pointer */
623 		ip6 = mtod(m, struct ip6_hdr *);
624 
625 		/*
626 		 * if the payload length field is 0 and the next header field
627 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
628 		 * option MUST be included.
629 		 */
630 		if (ip6->ip6_plen == 0 && plen == 0) {
631 			/*
632 			 * Note that if a valid jumbo payload option is
633 			 * contained, ip6_hopopts_input() must set a valid
634 			 * (non-zero) payload length to the variable plen.
635 			 */
636 			ip6stat.ip6s_badoptions++;
637 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
638 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
639 			icmp6_error(m, ICMP6_PARAM_PROB,
640 				    ICMP6_PARAMPROB_HEADER,
641 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
642 			return;
643 		}
644 #ifndef PULLDOWN_TEST
645 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
646 		hbh = (struct ip6_hbh *)(ip6 + 1);
647 #else
648 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
649 			sizeof(struct ip6_hbh));
650 		if (hbh == NULL) {
651 			ip6stat.ip6s_tooshort++;
652 			return;
653 		}
654 #endif
655 		nxt = hbh->ip6h_nxt;
656 
657 		/*
658 		 * accept the packet if a router alert option is included
659 		 * and we act as an IPv6 router.
660 		 */
661 		if (rtalert != ~0 && ip6_forwarding)
662 			ours = 1;
663 	} else
664 		nxt = ip6->ip6_nxt;
665 
666 	/*
667 	 * Check that the amount of data in the buffers
668 	 * is as at least much as the IPv6 header would have us expect.
669 	 * Trim mbufs if longer than we expect.
670 	 * Drop packet if shorter than we expect.
671 	 */
672 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
673 		ip6stat.ip6s_tooshort++;
674 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
675 		goto bad;
676 	}
677 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
678 		if (m->m_len == m->m_pkthdr.len) {
679 			m->m_len = sizeof(struct ip6_hdr) + plen;
680 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
681 		} else
682 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
683 	}
684 
685 	/*
686 	 * Forward if desirable.
687 	 */
688 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
689 		/*
690 		 * If we are acting as a multicast router, all
691 		 * incoming multicast packets are passed to the
692 		 * kernel-level multicast forwarding function.
693 		 * The packet is returned (relatively) intact; if
694 		 * ip6_mforward() returns a non-zero value, the packet
695 		 * must be discarded, else it may be accepted below.
696 		 */
697 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
698 			ip6stat.ip6s_cantforward++;
699 			m_freem(m);
700 			return;
701 		}
702 		if (!ours) {
703 			m_freem(m);
704 			return;
705 		}
706 	} else if (!ours) {
707 		ip6_forward(m, srcrt);
708 		return;
709 	}
710 
711 	ip6 = mtod(m, struct ip6_hdr *);
712 
713 	/*
714 	 * Malicious party may be able to use IPv4 mapped addr to confuse
715 	 * tcp/udp stack and bypass security checks (act as if it was from
716 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
717 	 *
718 	 * For SIIT end node behavior, you may want to disable the check.
719 	 * However, you will  become vulnerable to attacks using IPv4 mapped
720 	 * source.
721 	 */
722 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
723 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
724 		ip6stat.ip6s_badscope++;
725 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
726 		goto bad;
727 	}
728 
729 	/*
730 	 * Tell launch routine the next header
731 	 */
732 	ip6stat.ip6s_delivered++;
733 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
734 	nest = 0;
735 
736 	while (nxt != IPPROTO_DONE) {
737 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
738 			ip6stat.ip6s_toomanyhdr++;
739 			goto bad;
740 		}
741 
742 		/*
743 		 * protection against faulty packet - there should be
744 		 * more sanity checks in header chain processing.
745 		 */
746 		if (m->m_pkthdr.len < off) {
747 			ip6stat.ip6s_tooshort++;
748 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
749 			goto bad;
750 		}
751 
752 #ifdef IPSEC
753 		/*
754 		 * enforce IPsec policy checking if we are seeing last header.
755 		 * note that we do not visit this with protocols with pcb layer
756 		 * code - like udp/tcp/raw ip.
757 		 */
758 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
759 		    ipsec6_in_reject(m, NULL)) {
760 			ipsec6stat.in_polvio++;
761 			goto bad;
762 		}
763 #endif
764 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
765 	}
766 	return;
767  bad:
768 	m_freem(m);
769 }
770 
771 /*
772  * set/grab in6_ifaddr correspond to IPv6 destination address.
773  * XXX backward compatibility wrapper
774  */
775 static struct ip6aux *
776 ip6_setdstifaddr(m, ia6)
777 	struct mbuf *m;
778 	struct in6_ifaddr *ia6;
779 {
780 	struct ip6aux *ip6a;
781 
782 	ip6a = ip6_addaux(m);
783 	if (ip6a)
784 		ip6a->ip6a_dstia6 = ia6;
785 	return ip6a;	/* NULL if failed to set */
786 }
787 
788 struct in6_ifaddr *
789 ip6_getdstifaddr(m)
790 	struct mbuf *m;
791 {
792 	struct ip6aux *ip6a;
793 
794 	ip6a = ip6_findaux(m);
795 	if (ip6a)
796 		return ip6a->ip6a_dstia6;
797 	else
798 		return NULL;
799 }
800 
801 /*
802  * Hop-by-Hop options header processing. If a valid jumbo payload option is
803  * included, the real payload length will be stored in plenp.
804  */
805 static int
806 ip6_hopopts_input(plenp, rtalertp, mp, offp)
807 	u_int32_t *plenp;
808 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
809 	struct mbuf **mp;
810 	int *offp;
811 {
812 	struct mbuf *m = *mp;
813 	int off = *offp, hbhlen;
814 	struct ip6_hbh *hbh;
815 	u_int8_t *opt;
816 
817 	/* validation of the length of the header */
818 #ifndef PULLDOWN_TEST
819 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
820 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
821 	hbhlen = (hbh->ip6h_len + 1) << 3;
822 
823 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
824 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
825 #else
826 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
827 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
828 	if (hbh == NULL) {
829 		ip6stat.ip6s_tooshort++;
830 		return -1;
831 	}
832 	hbhlen = (hbh->ip6h_len + 1) << 3;
833 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
834 		hbhlen);
835 	if (hbh == NULL) {
836 		ip6stat.ip6s_tooshort++;
837 		return -1;
838 	}
839 #endif
840 	off += hbhlen;
841 	hbhlen -= sizeof(struct ip6_hbh);
842 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
843 
844 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
845 				hbhlen, rtalertp, plenp) < 0)
846 		return (-1);
847 
848 	*offp = off;
849 	*mp = m;
850 	return (0);
851 }
852 
853 /*
854  * Search header for all Hop-by-hop options and process each option.
855  * This function is separate from ip6_hopopts_input() in order to
856  * handle a case where the sending node itself process its hop-by-hop
857  * options header. In such a case, the function is called from ip6_output().
858  *
859  * The function assumes that hbh header is located right after the IPv6 header
860  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
861  * opthead + hbhlen is located in continuous memory region.
862  */
863 int
864 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
865 	struct mbuf *m;
866 	u_int8_t *opthead;
867 	int hbhlen;
868 	u_int32_t *rtalertp;
869 	u_int32_t *plenp;
870 {
871 	struct ip6_hdr *ip6;
872 	int optlen = 0;
873 	u_int8_t *opt = opthead;
874 	u_int16_t rtalert_val;
875 	u_int32_t jumboplen;
876 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
877 
878 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
879 		switch (*opt) {
880 		case IP6OPT_PAD1:
881 			optlen = 1;
882 			break;
883 		case IP6OPT_PADN:
884 			if (hbhlen < IP6OPT_MINLEN) {
885 				ip6stat.ip6s_toosmall++;
886 				goto bad;
887 			}
888 			optlen = *(opt + 1) + 2;
889 			break;
890 		case IP6OPT_ROUTER_ALERT:
891 			/* XXX may need check for alignment */
892 			if (hbhlen < IP6OPT_RTALERT_LEN) {
893 				ip6stat.ip6s_toosmall++;
894 				goto bad;
895 			}
896 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
897 				/* XXX stat */
898 				icmp6_error(m, ICMP6_PARAM_PROB,
899 				    ICMP6_PARAMPROB_HEADER,
900 				    erroff + opt + 1 - opthead);
901 				return (-1);
902 			}
903 			optlen = IP6OPT_RTALERT_LEN;
904 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
905 			*rtalertp = ntohs(rtalert_val);
906 			break;
907 		case IP6OPT_JUMBO:
908 			/* XXX may need check for alignment */
909 			if (hbhlen < IP6OPT_JUMBO_LEN) {
910 				ip6stat.ip6s_toosmall++;
911 				goto bad;
912 			}
913 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
914 				/* XXX stat */
915 				icmp6_error(m, ICMP6_PARAM_PROB,
916 				    ICMP6_PARAMPROB_HEADER,
917 				    erroff + opt + 1 - opthead);
918 				return (-1);
919 			}
920 			optlen = IP6OPT_JUMBO_LEN;
921 
922 			/*
923 			 * IPv6 packets that have non 0 payload length
924 			 * must not contain a jumbo payload option.
925 			 */
926 			ip6 = mtod(m, struct ip6_hdr *);
927 			if (ip6->ip6_plen) {
928 				ip6stat.ip6s_badoptions++;
929 				icmp6_error(m, ICMP6_PARAM_PROB,
930 				    ICMP6_PARAMPROB_HEADER,
931 				    erroff + opt - opthead);
932 				return (-1);
933 			}
934 
935 			/*
936 			 * We may see jumbolen in unaligned location, so
937 			 * we'd need to perform bcopy().
938 			 */
939 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
940 			jumboplen = (u_int32_t)htonl(jumboplen);
941 
942 #if 1
943 			/*
944 			 * if there are multiple jumbo payload options,
945 			 * *plenp will be non-zero and the packet will be
946 			 * rejected.
947 			 * the behavior may need some debate in ipngwg -
948 			 * multiple options does not make sense, however,
949 			 * there's no explicit mention in specification.
950 			 */
951 			if (*plenp != 0) {
952 				ip6stat.ip6s_badoptions++;
953 				icmp6_error(m, ICMP6_PARAM_PROB,
954 				    ICMP6_PARAMPROB_HEADER,
955 				    erroff + opt + 2 - opthead);
956 				return (-1);
957 			}
958 #endif
959 
960 			/*
961 			 * jumbo payload length must be larger than 65535.
962 			 */
963 			if (jumboplen <= IPV6_MAXPACKET) {
964 				ip6stat.ip6s_badoptions++;
965 				icmp6_error(m, ICMP6_PARAM_PROB,
966 				    ICMP6_PARAMPROB_HEADER,
967 				    erroff + opt + 2 - opthead);
968 				return (-1);
969 			}
970 			*plenp = jumboplen;
971 
972 			break;
973 		default:		/* unknown option */
974 			if (hbhlen < IP6OPT_MINLEN) {
975 				ip6stat.ip6s_toosmall++;
976 				goto bad;
977 			}
978 			optlen = ip6_unknown_opt(opt, m,
979 			    erroff + opt - opthead);
980 			if (optlen == -1)
981 				return (-1);
982 			optlen += 2;
983 			break;
984 		}
985 	}
986 
987 	return (0);
988 
989   bad:
990 	m_freem(m);
991 	return (-1);
992 }
993 
994 /*
995  * Unknown option processing.
996  * The third argument `off' is the offset from the IPv6 header to the option,
997  * which is necessary if the IPv6 header the and option header and IPv6 header
998  * is not continuous in order to return an ICMPv6 error.
999  */
1000 int
1001 ip6_unknown_opt(optp, m, off)
1002 	u_int8_t *optp;
1003 	struct mbuf *m;
1004 	int off;
1005 {
1006 	struct ip6_hdr *ip6;
1007 
1008 	switch (IP6OPT_TYPE(*optp)) {
1009 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1010 		return ((int)*(optp + 1));
1011 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1012 		m_freem(m);
1013 		return (-1);
1014 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1015 		ip6stat.ip6s_badoptions++;
1016 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1017 		return (-1);
1018 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1019 		ip6stat.ip6s_badoptions++;
1020 		ip6 = mtod(m, struct ip6_hdr *);
1021 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1022 		    (m->m_flags & (M_BCAST|M_MCAST)))
1023 			m_freem(m);
1024 		else
1025 			icmp6_error(m, ICMP6_PARAM_PROB,
1026 				    ICMP6_PARAMPROB_OPTION, off);
1027 		return (-1);
1028 	}
1029 
1030 	m_freem(m);		/* XXX: NOTREACHED */
1031 	return (-1);
1032 }
1033 
1034 /*
1035  * Create the "control" list for this pcb.
1036  * The function will not modify mbuf chain at all.
1037  *
1038  * with KAME mbuf chain restriction:
1039  * The routine will be called from upper layer handlers like tcp6_input().
1040  * Thus the routine assumes that the caller (tcp6_input) have already
1041  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1042  * very first mbuf on the mbuf chain.
1043  */
1044 void
1045 ip6_savecontrol(in6p, m, mp)
1046 	struct inpcb *in6p;
1047 	struct mbuf *m, **mp;
1048 {
1049 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1050 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1051 
1052 #ifdef SO_TIMESTAMP
1053 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1054 		struct timeval tv;
1055 
1056 		microtime(&tv);
1057 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1058 		    SCM_TIMESTAMP, SOL_SOCKET);
1059 		if (*mp)
1060 			mp = &(*mp)->m_next;
1061 	}
1062 #endif
1063 
1064 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1065 		return;
1066 
1067 	/* RFC 2292 sec. 5 */
1068 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1069 		struct in6_pktinfo pi6;
1070 
1071 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1072 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1073 		pi6.ipi6_ifindex =
1074 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1075 
1076 		*mp = sbcreatecontrol((caddr_t) &pi6,
1077 		    sizeof(struct in6_pktinfo),
1078 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1079 		if (*mp)
1080 			mp = &(*mp)->m_next;
1081 	}
1082 
1083 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1084 		int hlim = ip6->ip6_hlim & 0xff;
1085 
1086 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1087 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1088 		if (*mp)
1089 			mp = &(*mp)->m_next;
1090 	}
1091 
1092 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1093 		u_int32_t flowinfo;
1094 		int tclass;
1095 
1096 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1097 		flowinfo >>= 20;
1098 
1099 		tclass = flowinfo & 0xff;
1100 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1101 		    IPV6_TCLASS, IPPROTO_IPV6);
1102 		if (*mp)
1103 			mp = &(*mp)->m_next;
1104 	}
1105 
1106 	/*
1107 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1108 	 * privilege for the option (see ip6_ctloutput), but it might be too
1109 	 * strict, since there might be some hop-by-hop options which can be
1110 	 * returned to normal user.
1111 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1112 	 */
1113 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1114 		/*
1115 		 * Check if a hop-by-hop options header is contatined in the
1116 		 * received packet, and if so, store the options as ancillary
1117 		 * data. Note that a hop-by-hop options header must be
1118 		 * just after the IPv6 header, which is assured through the
1119 		 * IPv6 input processing.
1120 		 */
1121 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1122 			struct ip6_hbh *hbh;
1123 			int hbhlen = 0;
1124 #ifdef PULLDOWN_TEST
1125 			struct mbuf *ext;
1126 #endif
1127 
1128 #ifndef PULLDOWN_TEST
1129 			hbh = (struct ip6_hbh *)(ip6 + 1);
1130 			hbhlen = (hbh->ip6h_len + 1) << 3;
1131 #else
1132 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1133 			    ip6->ip6_nxt);
1134 			if (ext == NULL) {
1135 				ip6stat.ip6s_tooshort++;
1136 				return;
1137 			}
1138 			hbh = mtod(ext, struct ip6_hbh *);
1139 			hbhlen = (hbh->ip6h_len + 1) << 3;
1140 			if (hbhlen != ext->m_len) {
1141 				m_freem(ext);
1142 				ip6stat.ip6s_tooshort++;
1143 				return;
1144 			}
1145 #endif
1146 
1147 			/*
1148 			 * XXX: We copy the whole header even if a
1149 			 * jumbo payload option is included, the option which
1150 			 * is to be removed before returning according to
1151 			 * RFC2292.
1152 			 * Note: this constraint is removed in RFC3542
1153 			 */
1154 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1155 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1156 			    IPPROTO_IPV6);
1157 			if (*mp)
1158 				mp = &(*mp)->m_next;
1159 #ifdef PULLDOWN_TEST
1160 			m_freem(ext);
1161 #endif
1162 		}
1163 	}
1164 
1165 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1166 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1167 
1168 		/*
1169 		 * Search for destination options headers or routing
1170 		 * header(s) through the header chain, and stores each
1171 		 * header as ancillary data.
1172 		 * Note that the order of the headers remains in
1173 		 * the chain of ancillary data.
1174 		 */
1175 		while (1) {	/* is explicit loop prevention necessary? */
1176 			struct ip6_ext *ip6e = NULL;
1177 			int elen;
1178 #ifdef PULLDOWN_TEST
1179 			struct mbuf *ext = NULL;
1180 #endif
1181 
1182 			/*
1183 			 * if it is not an extension header, don't try to
1184 			 * pull it from the chain.
1185 			 */
1186 			switch (nxt) {
1187 			case IPPROTO_DSTOPTS:
1188 			case IPPROTO_ROUTING:
1189 			case IPPROTO_HOPOPTS:
1190 			case IPPROTO_AH: /* is it possible? */
1191 				break;
1192 			default:
1193 				goto loopend;
1194 			}
1195 
1196 #ifndef PULLDOWN_TEST
1197 			if (off + sizeof(*ip6e) > m->m_len)
1198 				goto loopend;
1199 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1200 			if (nxt == IPPROTO_AH)
1201 				elen = (ip6e->ip6e_len + 2) << 2;
1202 			else
1203 				elen = (ip6e->ip6e_len + 1) << 3;
1204 			if (off + elen > m->m_len)
1205 				goto loopend;
1206 #else
1207 			ext = ip6_pullexthdr(m, off, nxt);
1208 			if (ext == NULL) {
1209 				ip6stat.ip6s_tooshort++;
1210 				return;
1211 			}
1212 			ip6e = mtod(ext, struct ip6_ext *);
1213 			if (nxt == IPPROTO_AH)
1214 				elen = (ip6e->ip6e_len + 2) << 2;
1215 			else
1216 				elen = (ip6e->ip6e_len + 1) << 3;
1217 			if (elen != ext->m_len) {
1218 				m_freem(ext);
1219 				ip6stat.ip6s_tooshort++;
1220 				return;
1221 			}
1222 #endif
1223 
1224 			switch (nxt) {
1225 			case IPPROTO_DSTOPTS:
1226 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1227 					break;
1228 
1229 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1230 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1231 				    IPPROTO_IPV6);
1232 				if (*mp)
1233 					mp = &(*mp)->m_next;
1234 				break;
1235 			case IPPROTO_ROUTING:
1236 				if (!in6p->in6p_flags & IN6P_RTHDR)
1237 					break;
1238 
1239 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1240 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1241 				    IPPROTO_IPV6);
1242 				if (*mp)
1243 					mp = &(*mp)->m_next;
1244 				break;
1245 			case IPPROTO_HOPOPTS:
1246 			case IPPROTO_AH: /* is it possible? */
1247 				break;
1248 
1249 			default:
1250 				/*
1251 			 	 * other cases have been filtered in the above.
1252 				 * none will visit this case.  here we supply
1253 				 * the code just in case (nxt overwritten or
1254 				 * other cases).
1255 				 */
1256 #ifdef PULLDOWN_TEST
1257 				m_freem(ext);
1258 #endif
1259 				goto loopend;
1260 
1261 			}
1262 
1263 			/* proceed with the next header. */
1264 			off += elen;
1265 			nxt = ip6e->ip6e_nxt;
1266 			ip6e = NULL;
1267 #ifdef PULLDOWN_TEST
1268 			m_freem(ext);
1269 			ext = NULL;
1270 #endif
1271 		}
1272 	  loopend:
1273 		;
1274 	}
1275 
1276 #undef IS2292
1277 }
1278 
1279 void
1280 ip6_notify_pmtu(in6p, dst, mtu)
1281 	struct inpcb *in6p;
1282 	struct sockaddr_in6 *dst;
1283 	u_int32_t *mtu;
1284 {
1285 	struct socket *so;
1286 	struct mbuf *m_mtu;
1287 	struct ip6_mtuinfo mtuctl;
1288 
1289 	so =  in6p->inp_socket;
1290 
1291 	if (mtu == NULL)
1292 		return;
1293 
1294 #ifdef DIAGNOSTIC
1295 	if (so == NULL)		/* I believe this is impossible */
1296 		panic("ip6_notify_pmtu: socket is NULL");
1297 #endif
1298 
1299 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1300 	mtuctl.ip6m_mtu = *mtu;
1301 	mtuctl.ip6m_addr = *dst;
1302 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1303 		return;
1304 
1305 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1306 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1307 		return;
1308 
1309 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1310 	    == 0) {
1311 		m_freem(m_mtu);
1312 		/* XXX: should count statistics */
1313 	} else
1314 		sorwakeup(so);
1315 
1316 	return;
1317 }
1318 
1319 #ifdef PULLDOWN_TEST
1320 /*
1321  * pull single extension header from mbuf chain.  returns single mbuf that
1322  * contains the result, or NULL on error.
1323  */
1324 static struct mbuf *
1325 ip6_pullexthdr(m, off, nxt)
1326 	struct mbuf *m;
1327 	size_t off;
1328 	int nxt;
1329 {
1330 	struct ip6_ext ip6e;
1331 	size_t elen;
1332 	struct mbuf *n;
1333 
1334 #ifdef DIAGNOSTIC
1335 	switch (nxt) {
1336 	case IPPROTO_DSTOPTS:
1337 	case IPPROTO_ROUTING:
1338 	case IPPROTO_HOPOPTS:
1339 	case IPPROTO_AH: /* is it possible? */
1340 		break;
1341 	default:
1342 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1343 	}
1344 #endif
1345 
1346 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1347 	if (nxt == IPPROTO_AH)
1348 		elen = (ip6e.ip6e_len + 2) << 2;
1349 	else
1350 		elen = (ip6e.ip6e_len + 1) << 3;
1351 
1352 	MGET(n, M_DONTWAIT, MT_DATA);
1353 	if (n && elen >= MLEN) {
1354 		MCLGET(n, M_DONTWAIT);
1355 		if ((n->m_flags & M_EXT) == 0) {
1356 			m_free(n);
1357 			n = NULL;
1358 		}
1359 	}
1360 	if (!n)
1361 		return NULL;
1362 
1363 	n->m_len = 0;
1364 	if (elen >= M_TRAILINGSPACE(n)) {
1365 		m_free(n);
1366 		return NULL;
1367 	}
1368 
1369 	m_copydata(m, off, elen, mtod(n, caddr_t));
1370 	n->m_len = elen;
1371 	return n;
1372 }
1373 #endif
1374 
1375 /*
1376  * Get pointer to the previous header followed by the header
1377  * currently processed.
1378  * XXX: This function supposes that
1379  *	M includes all headers,
1380  *	the next header field and the header length field of each header
1381  *	are valid, and
1382  *	the sum of each header length equals to OFF.
1383  * Because of these assumptions, this function must be called very
1384  * carefully. Moreover, it will not be used in the near future when
1385  * we develop `neater' mechanism to process extension headers.
1386  */
1387 char *
1388 ip6_get_prevhdr(m, off)
1389 	struct mbuf *m;
1390 	int off;
1391 {
1392 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1393 
1394 	if (off == sizeof(struct ip6_hdr))
1395 		return (&ip6->ip6_nxt);
1396 	else {
1397 		int len, nxt;
1398 		struct ip6_ext *ip6e = NULL;
1399 
1400 		nxt = ip6->ip6_nxt;
1401 		len = sizeof(struct ip6_hdr);
1402 		while (len < off) {
1403 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1404 
1405 			switch (nxt) {
1406 			case IPPROTO_FRAGMENT:
1407 				len += sizeof(struct ip6_frag);
1408 				break;
1409 			case IPPROTO_AH:
1410 				len += (ip6e->ip6e_len + 2) << 2;
1411 				break;
1412 			default:
1413 				len += (ip6e->ip6e_len + 1) << 3;
1414 				break;
1415 			}
1416 			nxt = ip6e->ip6e_nxt;
1417 		}
1418 		if (ip6e)
1419 			return (&ip6e->ip6e_nxt);
1420 		else
1421 			return NULL;
1422 	}
1423 }
1424 
1425 /*
1426  * get next header offset.  m will be retained.
1427  */
1428 int
1429 ip6_nexthdr(m, off, proto, nxtp)
1430 	struct mbuf *m;
1431 	int off;
1432 	int proto;
1433 	int *nxtp;
1434 {
1435 	struct ip6_hdr ip6;
1436 	struct ip6_ext ip6e;
1437 	struct ip6_frag fh;
1438 
1439 	/* just in case */
1440 	if (m == NULL)
1441 		panic("ip6_nexthdr: m == NULL");
1442 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1443 		return -1;
1444 
1445 	switch (proto) {
1446 	case IPPROTO_IPV6:
1447 		if (m->m_pkthdr.len < off + sizeof(ip6))
1448 			return -1;
1449 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1450 		if (nxtp)
1451 			*nxtp = ip6.ip6_nxt;
1452 		off += sizeof(ip6);
1453 		return off;
1454 
1455 	case IPPROTO_FRAGMENT:
1456 		/*
1457 		 * terminate parsing if it is not the first fragment,
1458 		 * it does not make sense to parse through it.
1459 		 */
1460 		if (m->m_pkthdr.len < off + sizeof(fh))
1461 			return -1;
1462 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1463 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1464 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1465 			return -1;
1466 		if (nxtp)
1467 			*nxtp = fh.ip6f_nxt;
1468 		off += sizeof(struct ip6_frag);
1469 		return off;
1470 
1471 	case IPPROTO_AH:
1472 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1473 			return -1;
1474 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1475 		if (nxtp)
1476 			*nxtp = ip6e.ip6e_nxt;
1477 		off += (ip6e.ip6e_len + 2) << 2;
1478 		return off;
1479 
1480 	case IPPROTO_HOPOPTS:
1481 	case IPPROTO_ROUTING:
1482 	case IPPROTO_DSTOPTS:
1483 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1484 			return -1;
1485 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1486 		if (nxtp)
1487 			*nxtp = ip6e.ip6e_nxt;
1488 		off += (ip6e.ip6e_len + 1) << 3;
1489 		return off;
1490 
1491 	case IPPROTO_NONE:
1492 	case IPPROTO_ESP:
1493 	case IPPROTO_IPCOMP:
1494 		/* give up */
1495 		return -1;
1496 
1497 	default:
1498 		return -1;
1499 	}
1500 
1501 	return -1;
1502 }
1503 
1504 /*
1505  * get offset for the last header in the chain.  m will be kept untainted.
1506  */
1507 int
1508 ip6_lasthdr(m, off, proto, nxtp)
1509 	struct mbuf *m;
1510 	int off;
1511 	int proto;
1512 	int *nxtp;
1513 {
1514 	int newoff;
1515 	int nxt;
1516 
1517 	if (!nxtp) {
1518 		nxt = -1;
1519 		nxtp = &nxt;
1520 	}
1521 	while (1) {
1522 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1523 		if (newoff < 0)
1524 			return off;
1525 		else if (newoff < off)
1526 			return -1;	/* invalid */
1527 		else if (newoff == off)
1528 			return newoff;
1529 
1530 		off = newoff;
1531 		proto = *nxtp;
1532 	}
1533 }
1534 
1535 struct ip6aux *
1536 ip6_addaux(m)
1537 	struct mbuf *m;
1538 {
1539 	struct m_tag *mtag;
1540 
1541 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1542 	if (!mtag) {
1543 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1544 		    M_NOWAIT);
1545 		if (mtag) {
1546 			m_tag_prepend(m, mtag);
1547 			bzero(mtag + 1, sizeof(struct ip6aux));
1548 		}
1549 	}
1550 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1551 }
1552 
1553 struct ip6aux *
1554 ip6_findaux(m)
1555 	struct mbuf *m;
1556 {
1557 	struct m_tag *mtag;
1558 
1559 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1560 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1561 }
1562 
1563 void
1564 ip6_delaux(m)
1565 	struct mbuf *m;
1566 {
1567 	struct m_tag *mtag;
1568 
1569 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1570 	if (mtag)
1571 		m_tag_delete(m, mtag);
1572 }
1573 
1574 /*
1575  * System control for IP6
1576  */
1577 
1578 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1579 	0,		0,		0,		0,
1580 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1581 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1582 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1583 	0,		0,		0,		0,
1584 	ENOPROTOOPT
1585 };
1586