xref: /freebsd/sys/netinet6/ip6_input.c (revision 6af83ee0d2941d18880b6aaa2b4facd1d30c6106)
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_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/proc.h>
74 #include <sys/domain.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 
83 #include <net/if.h>
84 #include <net/if_types.h>
85 #include <net/if_dl.h>
86 #include <net/route.h>
87 #include <net/netisr.h>
88 #include <net/pfil.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_systm.h>
92 #ifdef INET
93 #include <netinet/ip.h>
94 #include <netinet/ip_icmp.h>
95 #endif /* INET */
96 #include <netinet/ip6.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/scope6_var.h>
102 #include <netinet6/in6_ifattach.h>
103 #include <netinet6/nd6.h>
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #ifdef INET6
108 #include <netinet6/ipsec6.h>
109 #endif
110 #endif
111 
112 #ifdef FAST_IPSEC
113 #include <netipsec/ipsec.h>
114 #include <netipsec/ipsec6.h>
115 #define	IPSEC
116 #endif /* FAST_IPSEC */
117 
118 #include <netinet6/ip6_fw.h>
119 
120 #include <netinet6/ip6protosw.h>
121 
122 #include <net/net_osdep.h>
123 
124 extern struct domain inet6domain;
125 
126 u_char ip6_protox[IPPROTO_MAX];
127 static struct ifqueue ip6intrq;
128 static int ip6qmaxlen = IFQ_MAXLEN;
129 struct in6_ifaddr *in6_ifaddr;
130 
131 extern struct callout in6_tmpaddrtimer_ch;
132 
133 int ip6_forward_srcrt;			/* XXX */
134 int ip6_sourcecheck;			/* XXX */
135 int ip6_sourcecheck_interval;		/* XXX */
136 
137 int ip6_ours_check_algorithm;
138 
139 struct pfil_head inet6_pfil_hook;
140 
141 /* firewall hooks */
142 ip6_fw_chk_t *ip6_fw_chk_ptr;
143 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
144 int ip6_fw_enable = 1;
145 
146 struct ip6stat ip6stat;
147 
148 static void ip6_init2 __P((void *));
149 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
150 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
151 #ifdef PULLDOWN_TEST
152 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
153 #endif
154 
155 /*
156  * IP6 initialization: fill in IP6 protocol switch table.
157  * All protocols not implemented in kernel go to raw IP6 protocol handler.
158  */
159 void
160 ip6_init()
161 {
162 	struct ip6protosw *pr;
163 	int i;
164 
165 #ifdef DIAGNOSTIC
166 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
167 		panic("sizeof(protosw) != sizeof(ip6protosw)");
168 #endif
169 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
170 	if (pr == 0)
171 		panic("ip6_init");
172 
173 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
174 	for (i = 0; i < IPPROTO_MAX; i++)
175 		ip6_protox[i] = pr - inet6sw;
176 	/*
177 	 * Cycle through IP protocols and put them into the appropriate place
178 	 * in ip6_protox[].
179 	 */
180 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
181 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
182 		if (pr->pr_domain->dom_family == PF_INET6 &&
183 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
184 			/* Be careful to only index valid IP protocols. */
185 			if (pr->pr_protocol <= IPPROTO_MAX)
186 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
187 		}
188 
189 	/* Initialize packet filter hooks. */
190 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
191 	inet6_pfil_hook.ph_af = AF_INET6;
192 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
193 		printf("%s: WARNING: unable to register pfil hook, "
194 			"error %d\n", __func__, i);
195 
196 	ip6intrq.ifq_maxlen = ip6qmaxlen;
197 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
198 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
199 	scope6_init();
200 	addrsel_policy_init();
201 	nd6_init();
202 	frag6_init();
203 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
204 }
205 
206 static void
207 ip6_init2(dummy)
208 	void *dummy;
209 {
210 
211 	/* nd6_timer_init */
212 	callout_init(&nd6_timer_ch, 0);
213 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
214 
215 	/* timer for regeneranation of temporary addresses randomize ID */
216 	callout_init(&in6_tmpaddrtimer_ch, 0);
217 	callout_reset(&in6_tmpaddrtimer_ch,
218 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
219 		       ip6_temp_regen_advance) * hz,
220 		      in6_tmpaddrtimer, NULL);
221 }
222 
223 /* cheat */
224 /* This must be after route_init(), which is now SI_ORDER_THIRD */
225 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
226 
227 extern struct	route_in6 ip6_forward_rt;
228 
229 void
230 ip6_input(m)
231 	struct mbuf *m;
232 {
233 	struct ip6_hdr *ip6;
234 	int off = sizeof(struct ip6_hdr), nest;
235 	u_int32_t plen;
236 	u_int32_t rtalert = ~0;
237 	int nxt, ours = 0;
238 	struct ifnet *deliverifp = NULL;
239 	struct sockaddr_in6 sa6;
240 	u_int32_t srczone, dstzone;
241 	struct in6_addr odst;
242 	int srcrt = 0;
243 
244 	GIANT_REQUIRED;			/* XXX for now */
245 #ifdef IPSEC
246 	/*
247 	 * should the inner packet be considered authentic?
248 	 * see comment in ah4_input().
249 	 */
250 	if (m) {
251 		m->m_flags &= ~M_AUTHIPHDR;
252 		m->m_flags &= ~M_AUTHIPDGM;
253 	}
254 #endif
255 
256 	/*
257 	 * make sure we don't have onion peering information into m_tag.
258 	 */
259 	ip6_delaux(m);
260 
261 	/*
262 	 * mbuf statistics
263 	 */
264 	if (m->m_flags & M_EXT) {
265 		if (m->m_next)
266 			ip6stat.ip6s_mext2m++;
267 		else
268 			ip6stat.ip6s_mext1++;
269 	} else {
270 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
271 		if (m->m_next) {
272 			if (m->m_flags & M_LOOP) {
273 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
274 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
275 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
276 			else
277 				ip6stat.ip6s_m2m[0]++;
278 		} else
279 			ip6stat.ip6s_m1++;
280 #undef M2MMAX
281 	}
282 
283 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
284 	ip6stat.ip6s_total++;
285 
286 #ifndef PULLDOWN_TEST
287 	/*
288 	 * L2 bridge code and some other code can return mbuf chain
289 	 * that does not conform to KAME requirement.  too bad.
290 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
291 	 */
292 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
293 		struct mbuf *n;
294 
295 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
296 		if (n)
297 			M_MOVE_PKTHDR(n, m);
298 		if (n && n->m_pkthdr.len > MHLEN) {
299 			MCLGET(n, M_DONTWAIT);
300 			if ((n->m_flags & M_EXT) == 0) {
301 				m_freem(n);
302 				n = NULL;
303 			}
304 		}
305 		if (n == NULL) {
306 			m_freem(m);
307 			return;	/* ENOBUFS */
308 		}
309 
310 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
311 		n->m_len = n->m_pkthdr.len;
312 		m_freem(m);
313 		m = n;
314 	}
315 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
316 #endif
317 
318 	if (m->m_len < sizeof(struct ip6_hdr)) {
319 		struct ifnet *inifp;
320 		inifp = m->m_pkthdr.rcvif;
321 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
322 			ip6stat.ip6s_toosmall++;
323 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
324 			return;
325 		}
326 	}
327 
328 	ip6 = mtod(m, struct ip6_hdr *);
329 
330 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
331 		ip6stat.ip6s_badvers++;
332 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
333 		goto bad;
334 	}
335 
336 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
337 
338 	/*
339 	 * Check against address spoofing/corruption.
340 	 */
341 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
342 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
343 		/*
344 		 * XXX: "badscope" is not very suitable for a multicast source.
345 		 */
346 		ip6stat.ip6s_badscope++;
347 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
348 		goto bad;
349 	}
350 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
351 	    !(m->m_flags & M_LOOP)) {
352 		/*
353 		 * In this case, the packet should come from the loopback
354 		 * interface.  However, we cannot just check the if_flags,
355 		 * because ip6_mloopback() passes the "actual" interface
356 		 * as the outgoing/incoming interface.
357 		 */
358 		ip6stat.ip6s_badscope++;
359 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
360 		goto bad;
361 	}
362 
363 #ifdef ALTQ
364 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
365 		/* packet is dropped by traffic conditioner */
366 		return;
367 	}
368 #endif
369 	/*
370 	 * The following check is not documented in specs.  A malicious
371 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
372 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
373 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
374 	 *
375 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
376 	 * support IPv4-less kernel compilation, we cannot support SIIT
377 	 * environment at all.  So, it makes more sense for us to reject any
378 	 * malicious packets for non-SIIT environment, than try to do a
379 	 * partial support for SIIT environment.
380 	 */
381 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
382 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
383 		ip6stat.ip6s_badscope++;
384 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
385 		goto bad;
386 	}
387 #if 0
388 	/*
389 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
390 	 *
391 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
392 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
393 	 * is revised to forbid relaying case.
394 	 */
395 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
396 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
397 		ip6stat.ip6s_badscope++;
398 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
399 		goto bad;
400 	}
401 #endif
402 
403 	/*
404 	 * Drop packets if the link ID portion is already filled.
405 	 * XXX: this is technically not a good behavior.  But, we internally
406 	 * use the field to disambiguate link-local addresses, so we cannot
407 	 * be generous against those a bit strange addresses.
408 	 */
409 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
410 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
411 		    ip6->ip6_src.s6_addr16[1]) {
412 			ip6stat.ip6s_badscope++;
413 			goto bad;
414 		}
415 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
416 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
417 		    ip6->ip6_dst.s6_addr16[1]) {
418 			ip6stat.ip6s_badscope++;
419 			goto bad;
420 		}
421 	}
422 
423 	/*
424 	 * Run through list of hooks for input packets.
425 	 *
426 	 * NB: Beware of the destination address changing
427 	 *     (e.g. by NAT rewriting).  When this happens,
428 	 *     tell ip6_forward to do the right thing.
429 	 */
430 	odst = ip6->ip6_dst;
431 
432 	/* Jump over all PFIL processing if hooks are not active. */
433 	if (inet6_pfil_hook.ph_busy_count == -1)
434 		goto passin;
435 
436 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
437 		return;
438 	if (m == NULL)			/* consumed by filter */
439 		return;
440 	ip6 = mtod(m, struct ip6_hdr *);
441 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
442 
443 passin:
444 	/*
445 	 * Check with the firewall...
446 	 */
447 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
448 		u_short port = 0;
449 		/* If ipfw says divert, we have to just drop packet */
450 		/* use port as a dummy argument */
451 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
452 			m_freem(m);
453 			m = NULL;
454 		}
455 		if (!m)
456 			return;
457 	}
458 
459 	/*
460 	 * construct source and destination address structures with
461 	 * disambiguating their scope zones (if there is ambiguity).
462 	 * XXX: sin6_family and sin6_len will NOT be referred to, but we fill
463 	 * in these fields just in case.
464 	 */
465 	if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_src, &srczone) ||
466 	    in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_dst, &dstzone)) {
467 		/*
468 		 * Note that these generic checks cover cases that src or
469 		 * dst are the loopback address and the receiving interface
470 		 * is not loopback.
471 		 */
472 		ip6stat.ip6s_badscope++;
473 		goto bad;
474 	}
475 
476 	bzero(&sa6, sizeof(sa6));
477 	sa6.sin6_family = AF_INET6;
478 	sa6.sin6_len = sizeof(struct sockaddr_in6);
479 
480 	sa6.sin6_addr = ip6->ip6_src;
481 	sa6.sin6_scope_id = srczone;
482 	if (in6_embedscope(&ip6->ip6_src, &sa6, NULL, NULL)) {
483 		/* XXX: should not happen */
484 		ip6stat.ip6s_badscope++;
485 		goto bad;
486 	}
487 
488 	sa6.sin6_addr = ip6->ip6_dst;
489 	sa6.sin6_scope_id = dstzone;
490 	if (in6_embedscope(&ip6->ip6_dst, &sa6, NULL, NULL)) {
491 		/* XXX: should not happen */
492 		ip6stat.ip6s_badscope++;
493 		goto bad;
494 	}
495 
496 	/* XXX: ff01::%ifN awareness is not merged, yet. */
497 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_src))
498 		ip6->ip6_src.s6_addr16[1] = 0;
499 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst))
500 		ip6->ip6_dst.s6_addr16[1] = 0;
501 
502 	/*
503 	 * Multicast check
504 	 */
505 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
506 	  	struct in6_multi *in6m = 0;
507 
508 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
509 		/*
510 		 * See if we belong to the destination multicast group on the
511 		 * arrival interface.
512 		 */
513 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
514 		if (in6m)
515 			ours = 1;
516 		else if (!ip6_mrouter) {
517 			ip6stat.ip6s_notmember++;
518 			ip6stat.ip6s_cantforward++;
519 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
520 			goto bad;
521 		}
522 		deliverifp = m->m_pkthdr.rcvif;
523 		goto hbhcheck;
524 	}
525 
526 	/*
527 	 *  Unicast check
528 	 */
529 	if (ip6_forward_rt.ro_rt != NULL &&
530 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
531 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
532 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
533 		ip6stat.ip6s_forward_cachehit++;
534 	else {
535 		struct sockaddr_in6 *dst6;
536 
537 		if (ip6_forward_rt.ro_rt) {
538 			/* route is down or destination is different */
539 			ip6stat.ip6s_forward_cachemiss++;
540 			RTFREE(ip6_forward_rt.ro_rt);
541 			ip6_forward_rt.ro_rt = 0;
542 		}
543 
544 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
545 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
546 		dst6->sin6_len = sizeof(struct sockaddr_in6);
547 		dst6->sin6_family = AF_INET6;
548 		dst6->sin6_addr = ip6->ip6_dst;
549 
550 		rtalloc((struct route *)&ip6_forward_rt);
551 	}
552 
553 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
554 
555 	/*
556 	 * Accept the packet if the forwarding interface to the destination
557 	 * according to the routing table is the loopback interface,
558 	 * unless the associated route has a gateway.
559 	 * Note that this approach causes to accept a packet if there is a
560 	 * route to the loopback interface for the destination of the packet.
561 	 * But we think it's even useful in some situations, e.g. when using
562 	 * a special daemon which wants to intercept the packet.
563 	 *
564 	 * XXX: some OSes automatically make a cloned route for the destination
565 	 * of an outgoing packet.  If the outgoing interface of the packet
566 	 * is a loopback one, the kernel would consider the packet to be
567 	 * accepted, even if we have no such address assinged on the interface.
568 	 * We check the cloned flag of the route entry to reject such cases,
569 	 * assuming that route entries for our own addresses are not made by
570 	 * cloning (it should be true because in6_addloop explicitly installs
571 	 * the host route).  However, we might have to do an explicit check
572 	 * while it would be less efficient.  Or, should we rather install a
573 	 * reject route for such a case?
574 	 */
575 	if (ip6_forward_rt.ro_rt &&
576 	    (ip6_forward_rt.ro_rt->rt_flags &
577 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
578 #ifdef RTF_WASCLONED
579 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
580 #endif
581 #ifdef RTF_CLONED
582 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
583 #endif
584 #if 0
585 	    /*
586 	     * The check below is redundant since the comparison of
587 	     * the destination and the key of the rtentry has
588 	     * already done through looking up the routing table.
589 	     */
590 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
591 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
592 #endif
593 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
594 		struct in6_ifaddr *ia6 =
595 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
596 
597 		/*
598 		 * record address information into m_tag.
599 		 */
600 		(void)ip6_setdstifaddr(m, ia6);
601 
602 		/*
603 		 * packets to a tentative, duplicated, or somehow invalid
604 		 * address must not be accepted.
605 		 */
606 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
607 			/* this address is ready */
608 			ours = 1;
609 			deliverifp = ia6->ia_ifp;	/* correct? */
610 			/* Count the packet in the ip address stats */
611 			ia6->ia_ifa.if_ipackets++;
612 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
613 			goto hbhcheck;
614 		} else {
615 			/* address is not ready, so discard the packet. */
616 			nd6log((LOG_INFO,
617 			    "ip6_input: packet to an unready address %s->%s\n",
618 			    ip6_sprintf(&ip6->ip6_src),
619 			    ip6_sprintf(&ip6->ip6_dst)));
620 
621 			goto bad;
622 		}
623 	}
624 
625 	/*
626 	 * FAITH (Firewall Aided Internet Translator)
627 	 */
628 	if (ip6_keepfaith) {
629 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
630 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
631 			/* XXX do we need more sanity checks? */
632 			ours = 1;
633 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
634 			goto hbhcheck;
635 		}
636 	}
637 
638 	/*
639 	 * Now there is no reason to process the packet if it's not our own
640 	 * and we're not a router.
641 	 */
642 	if (!ip6_forwarding) {
643 		ip6stat.ip6s_cantforward++;
644 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
645 		goto bad;
646 	}
647 
648   hbhcheck:
649 	/*
650 	 * record address information into m_tag, if we don't have one yet.
651 	 * note that we are unable to record it, if the address is not listed
652 	 * as our interface address (e.g. multicast addresses, addresses
653 	 * within FAITH prefixes and such).
654 	 */
655 	if (deliverifp && !ip6_getdstifaddr(m)) {
656 		struct in6_ifaddr *ia6;
657 
658 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
659 		if (ia6) {
660 			if (!ip6_setdstifaddr(m, ia6)) {
661 				/*
662 				 * XXX maybe we should drop the packet here,
663 				 * as we could not provide enough information
664 				 * to the upper layers.
665 				 */
666 			}
667 		}
668 	}
669 
670 	/*
671 	 * Process Hop-by-Hop options header if it's contained.
672 	 * m may be modified in ip6_hopopts_input().
673 	 * If a JumboPayload option is included, plen will also be modified.
674 	 */
675 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
676 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
677 		struct ip6_hbh *hbh;
678 
679 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
680 #if 0	/*touches NULL pointer*/
681 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
682 #endif
683 			return;	/* m have already been freed */
684 		}
685 
686 		/* adjust pointer */
687 		ip6 = mtod(m, struct ip6_hdr *);
688 
689 		/*
690 		 * if the payload length field is 0 and the next header field
691 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
692 		 * option MUST be included.
693 		 */
694 		if (ip6->ip6_plen == 0 && plen == 0) {
695 			/*
696 			 * Note that if a valid jumbo payload option is
697 			 * contained, ip6_hopopts_input() must set a valid
698 			 * (non-zero) payload length to the variable plen.
699 			 */
700 			ip6stat.ip6s_badoptions++;
701 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
702 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
703 			icmp6_error(m, ICMP6_PARAM_PROB,
704 				    ICMP6_PARAMPROB_HEADER,
705 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
706 			return;
707 		}
708 #ifndef PULLDOWN_TEST
709 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
710 		hbh = (struct ip6_hbh *)(ip6 + 1);
711 #else
712 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
713 			sizeof(struct ip6_hbh));
714 		if (hbh == NULL) {
715 			ip6stat.ip6s_tooshort++;
716 			return;
717 		}
718 #endif
719 		nxt = hbh->ip6h_nxt;
720 
721 		/*
722 		 * accept the packet if a router alert option is included
723 		 * and we act as an IPv6 router.
724 		 */
725 		if (rtalert != ~0 && ip6_forwarding)
726 			ours = 1;
727 	} else
728 		nxt = ip6->ip6_nxt;
729 
730 	/*
731 	 * Check that the amount of data in the buffers
732 	 * is as at least much as the IPv6 header would have us expect.
733 	 * Trim mbufs if longer than we expect.
734 	 * Drop packet if shorter than we expect.
735 	 */
736 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
737 		ip6stat.ip6s_tooshort++;
738 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
739 		goto bad;
740 	}
741 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
742 		if (m->m_len == m->m_pkthdr.len) {
743 			m->m_len = sizeof(struct ip6_hdr) + plen;
744 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
745 		} else
746 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
747 	}
748 
749 	/*
750 	 * Forward if desirable.
751 	 */
752 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
753 		/*
754 		 * If we are acting as a multicast router, all
755 		 * incoming multicast packets are passed to the
756 		 * kernel-level multicast forwarding function.
757 		 * The packet is returned (relatively) intact; if
758 		 * ip6_mforward() returns a non-zero value, the packet
759 		 * must be discarded, else it may be accepted below.
760 		 */
761 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
762 			ip6stat.ip6s_cantforward++;
763 			m_freem(m);
764 			return;
765 		}
766 		if (!ours) {
767 			m_freem(m);
768 			return;
769 		}
770 	} else if (!ours) {
771 		ip6_forward(m, srcrt);
772 		return;
773 	}
774 
775 	ip6 = mtod(m, struct ip6_hdr *);
776 
777 	/*
778 	 * Malicious party may be able to use IPv4 mapped addr to confuse
779 	 * tcp/udp stack and bypass security checks (act as if it was from
780 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
781 	 *
782 	 * For SIIT end node behavior, you may want to disable the check.
783 	 * However, you will  become vulnerable to attacks using IPv4 mapped
784 	 * source.
785 	 */
786 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
787 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
788 		ip6stat.ip6s_badscope++;
789 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
790 		goto bad;
791 	}
792 
793 	/*
794 	 * Tell launch routine the next header
795 	 */
796 	ip6stat.ip6s_delivered++;
797 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
798 	nest = 0;
799 
800 	while (nxt != IPPROTO_DONE) {
801 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
802 			ip6stat.ip6s_toomanyhdr++;
803 			goto bad;
804 		}
805 
806 		/*
807 		 * protection against faulty packet - there should be
808 		 * more sanity checks in header chain processing.
809 		 */
810 		if (m->m_pkthdr.len < off) {
811 			ip6stat.ip6s_tooshort++;
812 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
813 			goto bad;
814 		}
815 
816 #ifdef IPSEC
817 		/*
818 		 * enforce IPsec policy checking if we are seeing last header.
819 		 * note that we do not visit this with protocols with pcb layer
820 		 * code - like udp/tcp/raw ip.
821 		 */
822 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
823 		    ipsec6_in_reject(m, NULL)) {
824 			ipsec6stat.in_polvio++;
825 			goto bad;
826 		}
827 #endif
828 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
829 	}
830 	return;
831  bad:
832 	m_freem(m);
833 }
834 
835 /*
836  * set/grab in6_ifaddr correspond to IPv6 destination address.
837  * XXX backward compatibility wrapper
838  */
839 static struct ip6aux *
840 ip6_setdstifaddr(m, ia6)
841 	struct mbuf *m;
842 	struct in6_ifaddr *ia6;
843 {
844 	struct ip6aux *ip6a;
845 
846 	ip6a = ip6_addaux(m);
847 	if (ip6a)
848 		ip6a->ip6a_dstia6 = ia6;
849 	return ip6a;	/* NULL if failed to set */
850 }
851 
852 struct in6_ifaddr *
853 ip6_getdstifaddr(m)
854 	struct mbuf *m;
855 {
856 	struct ip6aux *ip6a;
857 
858 	ip6a = ip6_findaux(m);
859 	if (ip6a)
860 		return ip6a->ip6a_dstia6;
861 	else
862 		return NULL;
863 }
864 
865 /*
866  * Hop-by-Hop options header processing. If a valid jumbo payload option is
867  * included, the real payload length will be stored in plenp.
868  */
869 static int
870 ip6_hopopts_input(plenp, rtalertp, mp, offp)
871 	u_int32_t *plenp;
872 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
873 	struct mbuf **mp;
874 	int *offp;
875 {
876 	struct mbuf *m = *mp;
877 	int off = *offp, hbhlen;
878 	struct ip6_hbh *hbh;
879 	u_int8_t *opt;
880 
881 	/* validation of the length of the header */
882 #ifndef PULLDOWN_TEST
883 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
884 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
885 	hbhlen = (hbh->ip6h_len + 1) << 3;
886 
887 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
888 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
889 #else
890 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
891 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
892 	if (hbh == NULL) {
893 		ip6stat.ip6s_tooshort++;
894 		return -1;
895 	}
896 	hbhlen = (hbh->ip6h_len + 1) << 3;
897 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
898 		hbhlen);
899 	if (hbh == NULL) {
900 		ip6stat.ip6s_tooshort++;
901 		return -1;
902 	}
903 #endif
904 	off += hbhlen;
905 	hbhlen -= sizeof(struct ip6_hbh);
906 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
907 
908 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
909 				hbhlen, rtalertp, plenp) < 0)
910 		return (-1);
911 
912 	*offp = off;
913 	*mp = m;
914 	return (0);
915 }
916 
917 /*
918  * Search header for all Hop-by-hop options and process each option.
919  * This function is separate from ip6_hopopts_input() in order to
920  * handle a case where the sending node itself process its hop-by-hop
921  * options header. In such a case, the function is called from ip6_output().
922  *
923  * The function assumes that hbh header is located right after the IPv6 header
924  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
925  * opthead + hbhlen is located in continuous memory region.
926  */
927 int
928 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
929 	struct mbuf *m;
930 	u_int8_t *opthead;
931 	int hbhlen;
932 	u_int32_t *rtalertp;
933 	u_int32_t *plenp;
934 {
935 	struct ip6_hdr *ip6;
936 	int optlen = 0;
937 	u_int8_t *opt = opthead;
938 	u_int16_t rtalert_val;
939 	u_int32_t jumboplen;
940 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
941 
942 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
943 		switch (*opt) {
944 		case IP6OPT_PAD1:
945 			optlen = 1;
946 			break;
947 		case IP6OPT_PADN:
948 			if (hbhlen < IP6OPT_MINLEN) {
949 				ip6stat.ip6s_toosmall++;
950 				goto bad;
951 			}
952 			optlen = *(opt + 1) + 2;
953 			break;
954 		case IP6OPT_ROUTER_ALERT:
955 			/* XXX may need check for alignment */
956 			if (hbhlen < IP6OPT_RTALERT_LEN) {
957 				ip6stat.ip6s_toosmall++;
958 				goto bad;
959 			}
960 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
961 				/* XXX stat */
962 				icmp6_error(m, ICMP6_PARAM_PROB,
963 				    ICMP6_PARAMPROB_HEADER,
964 				    erroff + opt + 1 - opthead);
965 				return (-1);
966 			}
967 			optlen = IP6OPT_RTALERT_LEN;
968 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
969 			*rtalertp = ntohs(rtalert_val);
970 			break;
971 		case IP6OPT_JUMBO:
972 			/* XXX may need check for alignment */
973 			if (hbhlen < IP6OPT_JUMBO_LEN) {
974 				ip6stat.ip6s_toosmall++;
975 				goto bad;
976 			}
977 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
978 				/* XXX stat */
979 				icmp6_error(m, ICMP6_PARAM_PROB,
980 				    ICMP6_PARAMPROB_HEADER,
981 				    erroff + opt + 1 - opthead);
982 				return (-1);
983 			}
984 			optlen = IP6OPT_JUMBO_LEN;
985 
986 			/*
987 			 * IPv6 packets that have non 0 payload length
988 			 * must not contain a jumbo payload option.
989 			 */
990 			ip6 = mtod(m, struct ip6_hdr *);
991 			if (ip6->ip6_plen) {
992 				ip6stat.ip6s_badoptions++;
993 				icmp6_error(m, ICMP6_PARAM_PROB,
994 				    ICMP6_PARAMPROB_HEADER,
995 				    erroff + opt - opthead);
996 				return (-1);
997 			}
998 
999 			/*
1000 			 * We may see jumbolen in unaligned location, so
1001 			 * we'd need to perform bcopy().
1002 			 */
1003 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1004 			jumboplen = (u_int32_t)htonl(jumboplen);
1005 
1006 #if 1
1007 			/*
1008 			 * if there are multiple jumbo payload options,
1009 			 * *plenp will be non-zero and the packet will be
1010 			 * rejected.
1011 			 * the behavior may need some debate in ipngwg -
1012 			 * multiple options does not make sense, however,
1013 			 * there's no explicit mention in specification.
1014 			 */
1015 			if (*plenp != 0) {
1016 				ip6stat.ip6s_badoptions++;
1017 				icmp6_error(m, ICMP6_PARAM_PROB,
1018 				    ICMP6_PARAMPROB_HEADER,
1019 				    erroff + opt + 2 - opthead);
1020 				return (-1);
1021 			}
1022 #endif
1023 
1024 			/*
1025 			 * jumbo payload length must be larger than 65535.
1026 			 */
1027 			if (jumboplen <= IPV6_MAXPACKET) {
1028 				ip6stat.ip6s_badoptions++;
1029 				icmp6_error(m, ICMP6_PARAM_PROB,
1030 				    ICMP6_PARAMPROB_HEADER,
1031 				    erroff + opt + 2 - opthead);
1032 				return (-1);
1033 			}
1034 			*plenp = jumboplen;
1035 
1036 			break;
1037 		default:		/* unknown option */
1038 			if (hbhlen < IP6OPT_MINLEN) {
1039 				ip6stat.ip6s_toosmall++;
1040 				goto bad;
1041 			}
1042 			optlen = ip6_unknown_opt(opt, m,
1043 			    erroff + opt - opthead);
1044 			if (optlen == -1)
1045 				return (-1);
1046 			optlen += 2;
1047 			break;
1048 		}
1049 	}
1050 
1051 	return (0);
1052 
1053   bad:
1054 	m_freem(m);
1055 	return (-1);
1056 }
1057 
1058 /*
1059  * Unknown option processing.
1060  * The third argument `off' is the offset from the IPv6 header to the option,
1061  * which is necessary if the IPv6 header the and option header and IPv6 header
1062  * is not continuous in order to return an ICMPv6 error.
1063  */
1064 int
1065 ip6_unknown_opt(optp, m, off)
1066 	u_int8_t *optp;
1067 	struct mbuf *m;
1068 	int off;
1069 {
1070 	struct ip6_hdr *ip6;
1071 
1072 	switch (IP6OPT_TYPE(*optp)) {
1073 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1074 		return ((int)*(optp + 1));
1075 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1076 		m_freem(m);
1077 		return (-1);
1078 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1079 		ip6stat.ip6s_badoptions++;
1080 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1081 		return (-1);
1082 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1083 		ip6stat.ip6s_badoptions++;
1084 		ip6 = mtod(m, struct ip6_hdr *);
1085 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1086 		    (m->m_flags & (M_BCAST|M_MCAST)))
1087 			m_freem(m);
1088 		else
1089 			icmp6_error(m, ICMP6_PARAM_PROB,
1090 				    ICMP6_PARAMPROB_OPTION, off);
1091 		return (-1);
1092 	}
1093 
1094 	m_freem(m);		/* XXX: NOTREACHED */
1095 	return (-1);
1096 }
1097 
1098 /*
1099  * Create the "control" list for this pcb.
1100  * The function will not modify mbuf chain at all.
1101  *
1102  * with KAME mbuf chain restriction:
1103  * The routine will be called from upper layer handlers like tcp6_input().
1104  * Thus the routine assumes that the caller (tcp6_input) have already
1105  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1106  * very first mbuf on the mbuf chain.
1107  */
1108 void
1109 ip6_savecontrol(in6p, m, mp)
1110 	struct inpcb *in6p;
1111 	struct mbuf *m, **mp;
1112 {
1113 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1114 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1115 
1116 #ifdef SO_TIMESTAMP
1117 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1118 		struct timeval tv;
1119 
1120 		microtime(&tv);
1121 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1122 		    SCM_TIMESTAMP, SOL_SOCKET);
1123 		if (*mp)
1124 			mp = &(*mp)->m_next;
1125 	}
1126 #endif
1127 
1128 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1129 		return;
1130 
1131 	/* RFC 2292 sec. 5 */
1132 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1133 		struct in6_pktinfo pi6;
1134 
1135 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1136 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1137 		pi6.ipi6_ifindex =
1138 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1139 
1140 		*mp = sbcreatecontrol((caddr_t) &pi6,
1141 		    sizeof(struct in6_pktinfo),
1142 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1143 		if (*mp)
1144 			mp = &(*mp)->m_next;
1145 	}
1146 
1147 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1148 		int hlim = ip6->ip6_hlim & 0xff;
1149 
1150 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1151 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1152 		if (*mp)
1153 			mp = &(*mp)->m_next;
1154 	}
1155 
1156 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1157 		u_int32_t flowinfo;
1158 		int tclass;
1159 
1160 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1161 		flowinfo >>= 20;
1162 
1163 		tclass = flowinfo & 0xff;
1164 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1165 		    IPV6_TCLASS, IPPROTO_IPV6);
1166 		if (*mp)
1167 			mp = &(*mp)->m_next;
1168 	}
1169 
1170 	/*
1171 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1172 	 * privilege for the option (see ip6_ctloutput), but it might be too
1173 	 * strict, since there might be some hop-by-hop options which can be
1174 	 * returned to normal user.
1175 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1176 	 */
1177 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1178 		/*
1179 		 * Check if a hop-by-hop options header is contatined in the
1180 		 * received packet, and if so, store the options as ancillary
1181 		 * data. Note that a hop-by-hop options header must be
1182 		 * just after the IPv6 header, which is assured through the
1183 		 * IPv6 input processing.
1184 		 */
1185 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1186 			struct ip6_hbh *hbh;
1187 			int hbhlen = 0;
1188 #ifdef PULLDOWN_TEST
1189 			struct mbuf *ext;
1190 #endif
1191 
1192 #ifndef PULLDOWN_TEST
1193 			hbh = (struct ip6_hbh *)(ip6 + 1);
1194 			hbhlen = (hbh->ip6h_len + 1) << 3;
1195 #else
1196 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1197 			    ip6->ip6_nxt);
1198 			if (ext == NULL) {
1199 				ip6stat.ip6s_tooshort++;
1200 				return;
1201 			}
1202 			hbh = mtod(ext, struct ip6_hbh *);
1203 			hbhlen = (hbh->ip6h_len + 1) << 3;
1204 			if (hbhlen != ext->m_len) {
1205 				m_freem(ext);
1206 				ip6stat.ip6s_tooshort++;
1207 				return;
1208 			}
1209 #endif
1210 
1211 			/*
1212 			 * XXX: We copy the whole header even if a
1213 			 * jumbo payload option is included, the option which
1214 			 * is to be removed before returning according to
1215 			 * RFC2292.
1216 			 * Note: this constraint is removed in 2292bis.
1217 			 */
1218 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1219 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1220 			    IPPROTO_IPV6);
1221 			if (*mp)
1222 				mp = &(*mp)->m_next;
1223 #ifdef PULLDOWN_TEST
1224 			m_freem(ext);
1225 #endif
1226 		}
1227 	}
1228 
1229 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1230 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1231 
1232 		/*
1233 		 * Search for destination options headers or routing
1234 		 * header(s) through the header chain, and stores each
1235 		 * header as ancillary data.
1236 		 * Note that the order of the headers remains in
1237 		 * the chain of ancillary data.
1238 		 */
1239 		while (1) {	/* is explicit loop prevention necessary? */
1240 			struct ip6_ext *ip6e = NULL;
1241 			int elen;
1242 #ifdef PULLDOWN_TEST
1243 			struct mbuf *ext = NULL;
1244 #endif
1245 
1246 			/*
1247 			 * if it is not an extension header, don't try to
1248 			 * pull it from the chain.
1249 			 */
1250 			switch (nxt) {
1251 			case IPPROTO_DSTOPTS:
1252 			case IPPROTO_ROUTING:
1253 			case IPPROTO_HOPOPTS:
1254 			case IPPROTO_AH: /* is it possible? */
1255 				break;
1256 			default:
1257 				goto loopend;
1258 			}
1259 
1260 #ifndef PULLDOWN_TEST
1261 			if (off + sizeof(*ip6e) > m->m_len)
1262 				goto loopend;
1263 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1264 			if (nxt == IPPROTO_AH)
1265 				elen = (ip6e->ip6e_len + 2) << 2;
1266 			else
1267 				elen = (ip6e->ip6e_len + 1) << 3;
1268 			if (off + elen > m->m_len)
1269 				goto loopend;
1270 #else
1271 			ext = ip6_pullexthdr(m, off, nxt);
1272 			if (ext == NULL) {
1273 				ip6stat.ip6s_tooshort++;
1274 				return;
1275 			}
1276 			ip6e = mtod(ext, struct ip6_ext *);
1277 			if (nxt == IPPROTO_AH)
1278 				elen = (ip6e->ip6e_len + 2) << 2;
1279 			else
1280 				elen = (ip6e->ip6e_len + 1) << 3;
1281 			if (elen != ext->m_len) {
1282 				m_freem(ext);
1283 				ip6stat.ip6s_tooshort++;
1284 				return;
1285 			}
1286 #endif
1287 
1288 			switch (nxt) {
1289 			case IPPROTO_DSTOPTS:
1290 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1291 					break;
1292 
1293 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1294 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1295 				    IPPROTO_IPV6);
1296 				if (*mp)
1297 					mp = &(*mp)->m_next;
1298 				break;
1299 			case IPPROTO_ROUTING:
1300 				if (!in6p->in6p_flags & IN6P_RTHDR)
1301 					break;
1302 
1303 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1304 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1305 				    IPPROTO_IPV6);
1306 				if (*mp)
1307 					mp = &(*mp)->m_next;
1308 				break;
1309 			case IPPROTO_HOPOPTS:
1310 			case IPPROTO_AH: /* is it possible? */
1311 				break;
1312 
1313 			default:
1314 				/*
1315 			 	 * other cases have been filtered in the above.
1316 				 * none will visit this case.  here we supply
1317 				 * the code just in case (nxt overwritten or
1318 				 * other cases).
1319 				 */
1320 #ifdef PULLDOWN_TEST
1321 				m_freem(ext);
1322 #endif
1323 				goto loopend;
1324 
1325 			}
1326 
1327 			/* proceed with the next header. */
1328 			off += elen;
1329 			nxt = ip6e->ip6e_nxt;
1330 			ip6e = NULL;
1331 #ifdef PULLDOWN_TEST
1332 			m_freem(ext);
1333 			ext = NULL;
1334 #endif
1335 		}
1336 	  loopend:
1337 		;
1338 	}
1339 
1340 #undef IS2292
1341 }
1342 
1343 void
1344 ip6_notify_pmtu(in6p, dst, mtu)
1345 	struct inpcb *in6p;
1346 	struct sockaddr_in6 *dst;
1347 	u_int32_t *mtu;
1348 {
1349 	struct socket *so;
1350 	struct mbuf *m_mtu;
1351 	struct ip6_mtuinfo mtuctl;
1352 
1353 	so =  in6p->inp_socket;
1354 
1355 	if (mtu == NULL)
1356 		return;
1357 
1358 #ifdef DIAGNOSTIC
1359 	if (so == NULL)		/* I believe this is impossible */
1360 		panic("ip6_notify_pmtu: socket is NULL");
1361 #endif
1362 
1363 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1364 	mtuctl.ip6m_mtu = *mtu;
1365 	mtuctl.ip6m_addr = *dst;
1366 	in6_recoverscope(&mtuctl.ip6m_addr, &mtuctl.ip6m_addr.sin6_addr, NULL);
1367 
1368 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1369 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1370 		return;
1371 
1372 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1373 	    == 0) {
1374 		m_freem(m_mtu);
1375 		/* XXX: should count statistics */
1376 	} else
1377 		sorwakeup(so);
1378 
1379 	return;
1380 }
1381 
1382 #ifdef PULLDOWN_TEST
1383 /*
1384  * pull single extension header from mbuf chain.  returns single mbuf that
1385  * contains the result, or NULL on error.
1386  */
1387 static struct mbuf *
1388 ip6_pullexthdr(m, off, nxt)
1389 	struct mbuf *m;
1390 	size_t off;
1391 	int nxt;
1392 {
1393 	struct ip6_ext ip6e;
1394 	size_t elen;
1395 	struct mbuf *n;
1396 
1397 #ifdef DIAGNOSTIC
1398 	switch (nxt) {
1399 	case IPPROTO_DSTOPTS:
1400 	case IPPROTO_ROUTING:
1401 	case IPPROTO_HOPOPTS:
1402 	case IPPROTO_AH: /* is it possible? */
1403 		break;
1404 	default:
1405 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1406 	}
1407 #endif
1408 
1409 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1410 	if (nxt == IPPROTO_AH)
1411 		elen = (ip6e.ip6e_len + 2) << 2;
1412 	else
1413 		elen = (ip6e.ip6e_len + 1) << 3;
1414 
1415 	MGET(n, M_DONTWAIT, MT_DATA);
1416 	if (n && elen >= MLEN) {
1417 		MCLGET(n, M_DONTWAIT);
1418 		if ((n->m_flags & M_EXT) == 0) {
1419 			m_free(n);
1420 			n = NULL;
1421 		}
1422 	}
1423 	if (!n)
1424 		return NULL;
1425 
1426 	n->m_len = 0;
1427 	if (elen >= M_TRAILINGSPACE(n)) {
1428 		m_free(n);
1429 		return NULL;
1430 	}
1431 
1432 	m_copydata(m, off, elen, mtod(n, caddr_t));
1433 	n->m_len = elen;
1434 	return n;
1435 }
1436 #endif
1437 
1438 /*
1439  * Get pointer to the previous header followed by the header
1440  * currently processed.
1441  * XXX: This function supposes that
1442  *	M includes all headers,
1443  *	the next header field and the header length field of each header
1444  *	are valid, and
1445  *	the sum of each header length equals to OFF.
1446  * Because of these assumptions, this function must be called very
1447  * carefully. Moreover, it will not be used in the near future when
1448  * we develop `neater' mechanism to process extension headers.
1449  */
1450 char *
1451 ip6_get_prevhdr(m, off)
1452 	struct mbuf *m;
1453 	int off;
1454 {
1455 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1456 
1457 	if (off == sizeof(struct ip6_hdr))
1458 		return (&ip6->ip6_nxt);
1459 	else {
1460 		int len, nxt;
1461 		struct ip6_ext *ip6e = NULL;
1462 
1463 		nxt = ip6->ip6_nxt;
1464 		len = sizeof(struct ip6_hdr);
1465 		while (len < off) {
1466 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1467 
1468 			switch (nxt) {
1469 			case IPPROTO_FRAGMENT:
1470 				len += sizeof(struct ip6_frag);
1471 				break;
1472 			case IPPROTO_AH:
1473 				len += (ip6e->ip6e_len + 2) << 2;
1474 				break;
1475 			default:
1476 				len += (ip6e->ip6e_len + 1) << 3;
1477 				break;
1478 			}
1479 			nxt = ip6e->ip6e_nxt;
1480 		}
1481 		if (ip6e)
1482 			return (&ip6e->ip6e_nxt);
1483 		else
1484 			return NULL;
1485 	}
1486 }
1487 
1488 /*
1489  * get next header offset.  m will be retained.
1490  */
1491 int
1492 ip6_nexthdr(m, off, proto, nxtp)
1493 	struct mbuf *m;
1494 	int off;
1495 	int proto;
1496 	int *nxtp;
1497 {
1498 	struct ip6_hdr ip6;
1499 	struct ip6_ext ip6e;
1500 	struct ip6_frag fh;
1501 
1502 	/* just in case */
1503 	if (m == NULL)
1504 		panic("ip6_nexthdr: m == NULL");
1505 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1506 		return -1;
1507 
1508 	switch (proto) {
1509 	case IPPROTO_IPV6:
1510 		if (m->m_pkthdr.len < off + sizeof(ip6))
1511 			return -1;
1512 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1513 		if (nxtp)
1514 			*nxtp = ip6.ip6_nxt;
1515 		off += sizeof(ip6);
1516 		return off;
1517 
1518 	case IPPROTO_FRAGMENT:
1519 		/*
1520 		 * terminate parsing if it is not the first fragment,
1521 		 * it does not make sense to parse through it.
1522 		 */
1523 		if (m->m_pkthdr.len < off + sizeof(fh))
1524 			return -1;
1525 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1526 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1527 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1528 			return -1;
1529 		if (nxtp)
1530 			*nxtp = fh.ip6f_nxt;
1531 		off += sizeof(struct ip6_frag);
1532 		return off;
1533 
1534 	case IPPROTO_AH:
1535 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1536 			return -1;
1537 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1538 		if (nxtp)
1539 			*nxtp = ip6e.ip6e_nxt;
1540 		off += (ip6e.ip6e_len + 2) << 2;
1541 		return off;
1542 
1543 	case IPPROTO_HOPOPTS:
1544 	case IPPROTO_ROUTING:
1545 	case IPPROTO_DSTOPTS:
1546 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1547 			return -1;
1548 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1549 		if (nxtp)
1550 			*nxtp = ip6e.ip6e_nxt;
1551 		off += (ip6e.ip6e_len + 1) << 3;
1552 		return off;
1553 
1554 	case IPPROTO_NONE:
1555 	case IPPROTO_ESP:
1556 	case IPPROTO_IPCOMP:
1557 		/* give up */
1558 		return -1;
1559 
1560 	default:
1561 		return -1;
1562 	}
1563 
1564 	return -1;
1565 }
1566 
1567 /*
1568  * get offset for the last header in the chain.  m will be kept untainted.
1569  */
1570 int
1571 ip6_lasthdr(m, off, proto, nxtp)
1572 	struct mbuf *m;
1573 	int off;
1574 	int proto;
1575 	int *nxtp;
1576 {
1577 	int newoff;
1578 	int nxt;
1579 
1580 	if (!nxtp) {
1581 		nxt = -1;
1582 		nxtp = &nxt;
1583 	}
1584 	while (1) {
1585 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1586 		if (newoff < 0)
1587 			return off;
1588 		else if (newoff < off)
1589 			return -1;	/* invalid */
1590 		else if (newoff == off)
1591 			return newoff;
1592 
1593 		off = newoff;
1594 		proto = *nxtp;
1595 	}
1596 }
1597 
1598 struct ip6aux *
1599 ip6_addaux(m)
1600 	struct mbuf *m;
1601 {
1602 	struct m_tag *mtag;
1603 
1604 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1605 	if (!mtag) {
1606 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1607 		    M_NOWAIT);
1608 		if (mtag) {
1609 			m_tag_prepend(m, mtag);
1610 			bzero(mtag + 1, sizeof(struct ip6aux));
1611 		}
1612 	}
1613 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1614 }
1615 
1616 struct ip6aux *
1617 ip6_findaux(m)
1618 	struct mbuf *m;
1619 {
1620 	struct m_tag *mtag;
1621 
1622 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1623 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1624 }
1625 
1626 void
1627 ip6_delaux(m)
1628 	struct mbuf *m;
1629 {
1630 	struct m_tag *mtag;
1631 
1632 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1633 	if (mtag)
1634 		m_tag_delete(m, mtag);
1635 }
1636 
1637 /*
1638  * System control for IP6
1639  */
1640 
1641 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1642 	0,		0,		0,		0,
1643 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1644 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1645 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1646 	0,		0,		0,		0,
1647 	ENOPROTOOPT
1648 };
1649