xref: /freebsd/sys/netinet6/ip6_input.c (revision 94942af266ac119ede0ca836f9aa5a5ac0582938)
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 			char ip6bufs[INET6_ADDRSTRLEN];
552 			char ip6bufd[INET6_ADDRSTRLEN];
553 			/* address is not ready, so discard the packet. */
554 			nd6log((LOG_INFO,
555 			    "ip6_input: packet to an unready address %s->%s\n",
556 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
557 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
558 
559 			goto bad;
560 		}
561 	}
562 
563 	/*
564 	 * FAITH (Firewall Aided Internet Translator)
565 	 */
566 	if (ip6_keepfaith) {
567 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
568 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
569 			/* XXX do we need more sanity checks? */
570 			ours = 1;
571 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
572 			goto hbhcheck;
573 		}
574 	}
575 
576 	/*
577 	 * Now there is no reason to process the packet if it's not our own
578 	 * and we're not a router.
579 	 */
580 	if (!ip6_forwarding) {
581 		ip6stat.ip6s_cantforward++;
582 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
583 		goto bad;
584 	}
585 
586   hbhcheck:
587 	/*
588 	 * record address information into m_tag, if we don't have one yet.
589 	 * note that we are unable to record it, if the address is not listed
590 	 * as our interface address (e.g. multicast addresses, addresses
591 	 * within FAITH prefixes and such).
592 	 */
593 	if (deliverifp && !ip6_getdstifaddr(m)) {
594 		struct in6_ifaddr *ia6;
595 
596 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
597 		if (ia6) {
598 			if (!ip6_setdstifaddr(m, ia6)) {
599 				/*
600 				 * XXX maybe we should drop the packet here,
601 				 * as we could not provide enough information
602 				 * to the upper layers.
603 				 */
604 			}
605 		}
606 	}
607 
608 	/*
609 	 * Process Hop-by-Hop options header if it's contained.
610 	 * m may be modified in ip6_hopopts_input().
611 	 * If a JumboPayload option is included, plen will also be modified.
612 	 */
613 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
614 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
615 		struct ip6_hbh *hbh;
616 
617 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
618 #if 0	/*touches NULL pointer*/
619 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
620 #endif
621 			return;	/* m have already been freed */
622 		}
623 
624 		/* adjust pointer */
625 		ip6 = mtod(m, struct ip6_hdr *);
626 
627 		/*
628 		 * if the payload length field is 0 and the next header field
629 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
630 		 * option MUST be included.
631 		 */
632 		if (ip6->ip6_plen == 0 && plen == 0) {
633 			/*
634 			 * Note that if a valid jumbo payload option is
635 			 * contained, ip6_hopopts_input() must set a valid
636 			 * (non-zero) payload length to the variable plen.
637 			 */
638 			ip6stat.ip6s_badoptions++;
639 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
640 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
641 			icmp6_error(m, ICMP6_PARAM_PROB,
642 				    ICMP6_PARAMPROB_HEADER,
643 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
644 			return;
645 		}
646 #ifndef PULLDOWN_TEST
647 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
648 		hbh = (struct ip6_hbh *)(ip6 + 1);
649 #else
650 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
651 			sizeof(struct ip6_hbh));
652 		if (hbh == NULL) {
653 			ip6stat.ip6s_tooshort++;
654 			return;
655 		}
656 #endif
657 		nxt = hbh->ip6h_nxt;
658 
659 		/*
660 		 * If we are acting as a router and the packet contains a
661 		 * router alert option, see if we know the option value.
662 		 * Currently, we only support the option value for MLD, in which
663 		 * case we should pass the packet to the multicast routing
664 		 * daemon.
665 		 */
666 		if (rtalert != ~0 && ip6_forwarding) {
667 			switch (rtalert) {
668 			case IP6OPT_RTALERT_MLD:
669 				ours = 1;
670 				break;
671 			default:
672 				/*
673 				 * RFC2711 requires unrecognized values must be
674 				 * silently ignored.
675 				 */
676 				break;
677 			}
678 		}
679 	} else
680 		nxt = ip6->ip6_nxt;
681 
682 	/*
683 	 * Check that the amount of data in the buffers
684 	 * is as at least much as the IPv6 header would have us expect.
685 	 * Trim mbufs if longer than we expect.
686 	 * Drop packet if shorter than we expect.
687 	 */
688 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
689 		ip6stat.ip6s_tooshort++;
690 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
691 		goto bad;
692 	}
693 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
694 		if (m->m_len == m->m_pkthdr.len) {
695 			m->m_len = sizeof(struct ip6_hdr) + plen;
696 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
697 		} else
698 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
699 	}
700 
701 	/*
702 	 * Forward if desirable.
703 	 */
704 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
705 		/*
706 		 * If we are acting as a multicast router, all
707 		 * incoming multicast packets are passed to the
708 		 * kernel-level multicast forwarding function.
709 		 * The packet is returned (relatively) intact; if
710 		 * ip6_mforward() returns a non-zero value, the packet
711 		 * must be discarded, else it may be accepted below.
712 		 */
713 		if (ip6_mrouter && ip6_mforward &&
714 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
715 			ip6stat.ip6s_cantforward++;
716 			m_freem(m);
717 			return;
718 		}
719 		if (!ours) {
720 			m_freem(m);
721 			return;
722 		}
723 	} else if (!ours) {
724 		ip6_forward(m, srcrt);
725 		return;
726 	}
727 
728 	ip6 = mtod(m, struct ip6_hdr *);
729 
730 	/*
731 	 * Malicious party may be able to use IPv4 mapped addr to confuse
732 	 * tcp/udp stack and bypass security checks (act as if it was from
733 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
734 	 *
735 	 * For SIIT end node behavior, you may want to disable the check.
736 	 * However, you will  become vulnerable to attacks using IPv4 mapped
737 	 * source.
738 	 */
739 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
740 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
741 		ip6stat.ip6s_badscope++;
742 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
743 		goto bad;
744 	}
745 
746 	/*
747 	 * Tell launch routine the next header
748 	 */
749 	ip6stat.ip6s_delivered++;
750 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
751 	nest = 0;
752 
753 	while (nxt != IPPROTO_DONE) {
754 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
755 			ip6stat.ip6s_toomanyhdr++;
756 			goto bad;
757 		}
758 
759 		/*
760 		 * protection against faulty packet - there should be
761 		 * more sanity checks in header chain processing.
762 		 */
763 		if (m->m_pkthdr.len < off) {
764 			ip6stat.ip6s_tooshort++;
765 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
766 			goto bad;
767 		}
768 
769 #ifdef IPSEC
770 		/*
771 		 * enforce IPsec policy checking if we are seeing last header.
772 		 * note that we do not visit this with protocols with pcb layer
773 		 * code - like udp/tcp/raw ip.
774 		 */
775 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
776 		    ipsec6_in_reject(m, NULL)) {
777 			ipsec6stat.in_polvio++;
778 			goto bad;
779 		}
780 #endif
781 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
782 	}
783 	return;
784  bad:
785 	m_freem(m);
786 }
787 
788 /*
789  * set/grab in6_ifaddr correspond to IPv6 destination address.
790  * XXX backward compatibility wrapper
791  */
792 static struct ip6aux *
793 ip6_setdstifaddr(m, ia6)
794 	struct mbuf *m;
795 	struct in6_ifaddr *ia6;
796 {
797 	struct ip6aux *ip6a;
798 
799 	ip6a = ip6_addaux(m);
800 	if (ip6a)
801 		ip6a->ip6a_dstia6 = ia6;
802 	return ip6a;	/* NULL if failed to set */
803 }
804 
805 struct in6_ifaddr *
806 ip6_getdstifaddr(m)
807 	struct mbuf *m;
808 {
809 	struct ip6aux *ip6a;
810 
811 	ip6a = ip6_findaux(m);
812 	if (ip6a)
813 		return ip6a->ip6a_dstia6;
814 	else
815 		return NULL;
816 }
817 
818 /*
819  * Hop-by-Hop options header processing. If a valid jumbo payload option is
820  * included, the real payload length will be stored in plenp.
821  */
822 static int
823 ip6_hopopts_input(plenp, rtalertp, mp, offp)
824 	u_int32_t *plenp;
825 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
826 	struct mbuf **mp;
827 	int *offp;
828 {
829 	struct mbuf *m = *mp;
830 	int off = *offp, hbhlen;
831 	struct ip6_hbh *hbh;
832 	u_int8_t *opt;
833 
834 	/* validation of the length of the header */
835 #ifndef PULLDOWN_TEST
836 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
837 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
838 	hbhlen = (hbh->ip6h_len + 1) << 3;
839 
840 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
841 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
842 #else
843 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
844 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
845 	if (hbh == NULL) {
846 		ip6stat.ip6s_tooshort++;
847 		return -1;
848 	}
849 	hbhlen = (hbh->ip6h_len + 1) << 3;
850 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
851 		hbhlen);
852 	if (hbh == NULL) {
853 		ip6stat.ip6s_tooshort++;
854 		return -1;
855 	}
856 #endif
857 	off += hbhlen;
858 	hbhlen -= sizeof(struct ip6_hbh);
859 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
860 
861 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
862 				hbhlen, rtalertp, plenp) < 0)
863 		return (-1);
864 
865 	*offp = off;
866 	*mp = m;
867 	return (0);
868 }
869 
870 /*
871  * Search header for all Hop-by-hop options and process each option.
872  * This function is separate from ip6_hopopts_input() in order to
873  * handle a case where the sending node itself process its hop-by-hop
874  * options header. In such a case, the function is called from ip6_output().
875  *
876  * The function assumes that hbh header is located right after the IPv6 header
877  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
878  * opthead + hbhlen is located in continuous memory region.
879  */
880 int
881 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
882 	struct mbuf *m;
883 	u_int8_t *opthead;
884 	int hbhlen;
885 	u_int32_t *rtalertp;
886 	u_int32_t *plenp;
887 {
888 	struct ip6_hdr *ip6;
889 	int optlen = 0;
890 	u_int8_t *opt = opthead;
891 	u_int16_t rtalert_val;
892 	u_int32_t jumboplen;
893 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
894 
895 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
896 		switch (*opt) {
897 		case IP6OPT_PAD1:
898 			optlen = 1;
899 			break;
900 		case IP6OPT_PADN:
901 			if (hbhlen < IP6OPT_MINLEN) {
902 				ip6stat.ip6s_toosmall++;
903 				goto bad;
904 			}
905 			optlen = *(opt + 1) + 2;
906 			break;
907 		case IP6OPT_ROUTER_ALERT:
908 			/* XXX may need check for alignment */
909 			if (hbhlen < IP6OPT_RTALERT_LEN) {
910 				ip6stat.ip6s_toosmall++;
911 				goto bad;
912 			}
913 			if (*(opt + 1) != IP6OPT_RTALERT_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_RTALERT_LEN;
921 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
922 			*rtalertp = ntohs(rtalert_val);
923 			break;
924 		case IP6OPT_JUMBO:
925 			/* XXX may need check for alignment */
926 			if (hbhlen < IP6OPT_JUMBO_LEN) {
927 				ip6stat.ip6s_toosmall++;
928 				goto bad;
929 			}
930 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
931 				/* XXX stat */
932 				icmp6_error(m, ICMP6_PARAM_PROB,
933 				    ICMP6_PARAMPROB_HEADER,
934 				    erroff + opt + 1 - opthead);
935 				return (-1);
936 			}
937 			optlen = IP6OPT_JUMBO_LEN;
938 
939 			/*
940 			 * IPv6 packets that have non 0 payload length
941 			 * must not contain a jumbo payload option.
942 			 */
943 			ip6 = mtod(m, struct ip6_hdr *);
944 			if (ip6->ip6_plen) {
945 				ip6stat.ip6s_badoptions++;
946 				icmp6_error(m, ICMP6_PARAM_PROB,
947 				    ICMP6_PARAMPROB_HEADER,
948 				    erroff + opt - opthead);
949 				return (-1);
950 			}
951 
952 			/*
953 			 * We may see jumbolen in unaligned location, so
954 			 * we'd need to perform bcopy().
955 			 */
956 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
957 			jumboplen = (u_int32_t)htonl(jumboplen);
958 
959 #if 1
960 			/*
961 			 * if there are multiple jumbo payload options,
962 			 * *plenp will be non-zero and the packet will be
963 			 * rejected.
964 			 * the behavior may need some debate in ipngwg -
965 			 * multiple options does not make sense, however,
966 			 * there's no explicit mention in specification.
967 			 */
968 			if (*plenp != 0) {
969 				ip6stat.ip6s_badoptions++;
970 				icmp6_error(m, ICMP6_PARAM_PROB,
971 				    ICMP6_PARAMPROB_HEADER,
972 				    erroff + opt + 2 - opthead);
973 				return (-1);
974 			}
975 #endif
976 
977 			/*
978 			 * jumbo payload length must be larger than 65535.
979 			 */
980 			if (jumboplen <= IPV6_MAXPACKET) {
981 				ip6stat.ip6s_badoptions++;
982 				icmp6_error(m, ICMP6_PARAM_PROB,
983 				    ICMP6_PARAMPROB_HEADER,
984 				    erroff + opt + 2 - opthead);
985 				return (-1);
986 			}
987 			*plenp = jumboplen;
988 
989 			break;
990 		default:		/* unknown option */
991 			if (hbhlen < IP6OPT_MINLEN) {
992 				ip6stat.ip6s_toosmall++;
993 				goto bad;
994 			}
995 			optlen = ip6_unknown_opt(opt, m,
996 			    erroff + opt - opthead);
997 			if (optlen == -1)
998 				return (-1);
999 			optlen += 2;
1000 			break;
1001 		}
1002 	}
1003 
1004 	return (0);
1005 
1006   bad:
1007 	m_freem(m);
1008 	return (-1);
1009 }
1010 
1011 /*
1012  * Unknown option processing.
1013  * The third argument `off' is the offset from the IPv6 header to the option,
1014  * which is necessary if the IPv6 header the and option header and IPv6 header
1015  * is not continuous in order to return an ICMPv6 error.
1016  */
1017 int
1018 ip6_unknown_opt(optp, m, off)
1019 	u_int8_t *optp;
1020 	struct mbuf *m;
1021 	int off;
1022 {
1023 	struct ip6_hdr *ip6;
1024 
1025 	switch (IP6OPT_TYPE(*optp)) {
1026 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1027 		return ((int)*(optp + 1));
1028 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1029 		m_freem(m);
1030 		return (-1);
1031 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1032 		ip6stat.ip6s_badoptions++;
1033 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1034 		return (-1);
1035 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1036 		ip6stat.ip6s_badoptions++;
1037 		ip6 = mtod(m, struct ip6_hdr *);
1038 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1039 		    (m->m_flags & (M_BCAST|M_MCAST)))
1040 			m_freem(m);
1041 		else
1042 			icmp6_error(m, ICMP6_PARAM_PROB,
1043 				    ICMP6_PARAMPROB_OPTION, off);
1044 		return (-1);
1045 	}
1046 
1047 	m_freem(m);		/* XXX: NOTREACHED */
1048 	return (-1);
1049 }
1050 
1051 /*
1052  * Create the "control" list for this pcb.
1053  * The function will not modify mbuf chain at all.
1054  *
1055  * with KAME mbuf chain restriction:
1056  * The routine will be called from upper layer handlers like tcp6_input().
1057  * Thus the routine assumes that the caller (tcp6_input) have already
1058  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1059  * very first mbuf on the mbuf chain.
1060  */
1061 void
1062 ip6_savecontrol(in6p, m, mp)
1063 	struct inpcb *in6p;
1064 	struct mbuf *m, **mp;
1065 {
1066 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1067 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1068 
1069 #ifdef SO_TIMESTAMP
1070 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1071 		struct timeval tv;
1072 
1073 		microtime(&tv);
1074 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1075 		    SCM_TIMESTAMP, SOL_SOCKET);
1076 		if (*mp)
1077 			mp = &(*mp)->m_next;
1078 	}
1079 #endif
1080 
1081 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1082 		return;
1083 
1084 	/* RFC 2292 sec. 5 */
1085 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1086 		struct in6_pktinfo pi6;
1087 
1088 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1089 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1090 		pi6.ipi6_ifindex =
1091 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1092 
1093 		*mp = sbcreatecontrol((caddr_t) &pi6,
1094 		    sizeof(struct in6_pktinfo),
1095 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1096 		if (*mp)
1097 			mp = &(*mp)->m_next;
1098 	}
1099 
1100 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1101 		int hlim = ip6->ip6_hlim & 0xff;
1102 
1103 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1104 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1105 		if (*mp)
1106 			mp = &(*mp)->m_next;
1107 	}
1108 
1109 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1110 		u_int32_t flowinfo;
1111 		int tclass;
1112 
1113 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1114 		flowinfo >>= 20;
1115 
1116 		tclass = flowinfo & 0xff;
1117 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1118 		    IPV6_TCLASS, IPPROTO_IPV6);
1119 		if (*mp)
1120 			mp = &(*mp)->m_next;
1121 	}
1122 
1123 	/*
1124 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1125 	 * privilege for the option (see ip6_ctloutput), but it might be too
1126 	 * strict, since there might be some hop-by-hop options which can be
1127 	 * returned to normal user.
1128 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1129 	 */
1130 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1131 		/*
1132 		 * Check if a hop-by-hop options header is contatined in the
1133 		 * received packet, and if so, store the options as ancillary
1134 		 * data. Note that a hop-by-hop options header must be
1135 		 * just after the IPv6 header, which is assured through the
1136 		 * IPv6 input processing.
1137 		 */
1138 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1139 			struct ip6_hbh *hbh;
1140 			int hbhlen = 0;
1141 #ifdef PULLDOWN_TEST
1142 			struct mbuf *ext;
1143 #endif
1144 
1145 #ifndef PULLDOWN_TEST
1146 			hbh = (struct ip6_hbh *)(ip6 + 1);
1147 			hbhlen = (hbh->ip6h_len + 1) << 3;
1148 #else
1149 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1150 			    ip6->ip6_nxt);
1151 			if (ext == NULL) {
1152 				ip6stat.ip6s_tooshort++;
1153 				return;
1154 			}
1155 			hbh = mtod(ext, struct ip6_hbh *);
1156 			hbhlen = (hbh->ip6h_len + 1) << 3;
1157 			if (hbhlen != ext->m_len) {
1158 				m_freem(ext);
1159 				ip6stat.ip6s_tooshort++;
1160 				return;
1161 			}
1162 #endif
1163 
1164 			/*
1165 			 * XXX: We copy the whole header even if a
1166 			 * jumbo payload option is included, the option which
1167 			 * is to be removed before returning according to
1168 			 * RFC2292.
1169 			 * Note: this constraint is removed in RFC3542
1170 			 */
1171 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1172 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1173 			    IPPROTO_IPV6);
1174 			if (*mp)
1175 				mp = &(*mp)->m_next;
1176 #ifdef PULLDOWN_TEST
1177 			m_freem(ext);
1178 #endif
1179 		}
1180 	}
1181 
1182 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1183 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1184 
1185 		/*
1186 		 * Search for destination options headers or routing
1187 		 * header(s) through the header chain, and stores each
1188 		 * header as ancillary data.
1189 		 * Note that the order of the headers remains in
1190 		 * the chain of ancillary data.
1191 		 */
1192 		while (1) {	/* is explicit loop prevention necessary? */
1193 			struct ip6_ext *ip6e = NULL;
1194 			int elen;
1195 #ifdef PULLDOWN_TEST
1196 			struct mbuf *ext = NULL;
1197 #endif
1198 
1199 			/*
1200 			 * if it is not an extension header, don't try to
1201 			 * pull it from the chain.
1202 			 */
1203 			switch (nxt) {
1204 			case IPPROTO_DSTOPTS:
1205 			case IPPROTO_ROUTING:
1206 			case IPPROTO_HOPOPTS:
1207 			case IPPROTO_AH: /* is it possible? */
1208 				break;
1209 			default:
1210 				goto loopend;
1211 			}
1212 
1213 #ifndef PULLDOWN_TEST
1214 			if (off + sizeof(*ip6e) > m->m_len)
1215 				goto loopend;
1216 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1217 			if (nxt == IPPROTO_AH)
1218 				elen = (ip6e->ip6e_len + 2) << 2;
1219 			else
1220 				elen = (ip6e->ip6e_len + 1) << 3;
1221 			if (off + elen > m->m_len)
1222 				goto loopend;
1223 #else
1224 			ext = ip6_pullexthdr(m, off, nxt);
1225 			if (ext == NULL) {
1226 				ip6stat.ip6s_tooshort++;
1227 				return;
1228 			}
1229 			ip6e = mtod(ext, struct ip6_ext *);
1230 			if (nxt == IPPROTO_AH)
1231 				elen = (ip6e->ip6e_len + 2) << 2;
1232 			else
1233 				elen = (ip6e->ip6e_len + 1) << 3;
1234 			if (elen != ext->m_len) {
1235 				m_freem(ext);
1236 				ip6stat.ip6s_tooshort++;
1237 				return;
1238 			}
1239 #endif
1240 
1241 			switch (nxt) {
1242 			case IPPROTO_DSTOPTS:
1243 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1244 					break;
1245 
1246 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1247 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1248 				    IPPROTO_IPV6);
1249 				if (*mp)
1250 					mp = &(*mp)->m_next;
1251 				break;
1252 			case IPPROTO_ROUTING:
1253 				if (!in6p->in6p_flags & IN6P_RTHDR)
1254 					break;
1255 
1256 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1257 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1258 				    IPPROTO_IPV6);
1259 				if (*mp)
1260 					mp = &(*mp)->m_next;
1261 				break;
1262 			case IPPROTO_HOPOPTS:
1263 			case IPPROTO_AH: /* is it possible? */
1264 				break;
1265 
1266 			default:
1267 				/*
1268 			 	 * other cases have been filtered in the above.
1269 				 * none will visit this case.  here we supply
1270 				 * the code just in case (nxt overwritten or
1271 				 * other cases).
1272 				 */
1273 #ifdef PULLDOWN_TEST
1274 				m_freem(ext);
1275 #endif
1276 				goto loopend;
1277 
1278 			}
1279 
1280 			/* proceed with the next header. */
1281 			off += elen;
1282 			nxt = ip6e->ip6e_nxt;
1283 			ip6e = NULL;
1284 #ifdef PULLDOWN_TEST
1285 			m_freem(ext);
1286 			ext = NULL;
1287 #endif
1288 		}
1289 	  loopend:
1290 		;
1291 	}
1292 
1293 #undef IS2292
1294 }
1295 
1296 void
1297 ip6_notify_pmtu(in6p, dst, mtu)
1298 	struct inpcb *in6p;
1299 	struct sockaddr_in6 *dst;
1300 	u_int32_t *mtu;
1301 {
1302 	struct socket *so;
1303 	struct mbuf *m_mtu;
1304 	struct ip6_mtuinfo mtuctl;
1305 
1306 	so =  in6p->inp_socket;
1307 
1308 	if (mtu == NULL)
1309 		return;
1310 
1311 #ifdef DIAGNOSTIC
1312 	if (so == NULL)		/* I believe this is impossible */
1313 		panic("ip6_notify_pmtu: socket is NULL");
1314 #endif
1315 
1316 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1317 	mtuctl.ip6m_mtu = *mtu;
1318 	mtuctl.ip6m_addr = *dst;
1319 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1320 		return;
1321 
1322 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1323 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1324 		return;
1325 
1326 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1327 	    == 0) {
1328 		m_freem(m_mtu);
1329 		/* XXX: should count statistics */
1330 	} else
1331 		sorwakeup(so);
1332 
1333 	return;
1334 }
1335 
1336 #ifdef PULLDOWN_TEST
1337 /*
1338  * pull single extension header from mbuf chain.  returns single mbuf that
1339  * contains the result, or NULL on error.
1340  */
1341 static struct mbuf *
1342 ip6_pullexthdr(m, off, nxt)
1343 	struct mbuf *m;
1344 	size_t off;
1345 	int nxt;
1346 {
1347 	struct ip6_ext ip6e;
1348 	size_t elen;
1349 	struct mbuf *n;
1350 
1351 #ifdef DIAGNOSTIC
1352 	switch (nxt) {
1353 	case IPPROTO_DSTOPTS:
1354 	case IPPROTO_ROUTING:
1355 	case IPPROTO_HOPOPTS:
1356 	case IPPROTO_AH: /* is it possible? */
1357 		break;
1358 	default:
1359 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1360 	}
1361 #endif
1362 
1363 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1364 	if (nxt == IPPROTO_AH)
1365 		elen = (ip6e.ip6e_len + 2) << 2;
1366 	else
1367 		elen = (ip6e.ip6e_len + 1) << 3;
1368 
1369 	MGET(n, M_DONTWAIT, MT_DATA);
1370 	if (n && elen >= MLEN) {
1371 		MCLGET(n, M_DONTWAIT);
1372 		if ((n->m_flags & M_EXT) == 0) {
1373 			m_free(n);
1374 			n = NULL;
1375 		}
1376 	}
1377 	if (!n)
1378 		return NULL;
1379 
1380 	n->m_len = 0;
1381 	if (elen >= M_TRAILINGSPACE(n)) {
1382 		m_free(n);
1383 		return NULL;
1384 	}
1385 
1386 	m_copydata(m, off, elen, mtod(n, caddr_t));
1387 	n->m_len = elen;
1388 	return n;
1389 }
1390 #endif
1391 
1392 /*
1393  * Get pointer to the previous header followed by the header
1394  * currently processed.
1395  * XXX: This function supposes that
1396  *	M includes all headers,
1397  *	the next header field and the header length field of each header
1398  *	are valid, and
1399  *	the sum of each header length equals to OFF.
1400  * Because of these assumptions, this function must be called very
1401  * carefully. Moreover, it will not be used in the near future when
1402  * we develop `neater' mechanism to process extension headers.
1403  */
1404 char *
1405 ip6_get_prevhdr(m, off)
1406 	struct mbuf *m;
1407 	int off;
1408 {
1409 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1410 
1411 	if (off == sizeof(struct ip6_hdr))
1412 		return (&ip6->ip6_nxt);
1413 	else {
1414 		int len, nxt;
1415 		struct ip6_ext *ip6e = NULL;
1416 
1417 		nxt = ip6->ip6_nxt;
1418 		len = sizeof(struct ip6_hdr);
1419 		while (len < off) {
1420 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1421 
1422 			switch (nxt) {
1423 			case IPPROTO_FRAGMENT:
1424 				len += sizeof(struct ip6_frag);
1425 				break;
1426 			case IPPROTO_AH:
1427 				len += (ip6e->ip6e_len + 2) << 2;
1428 				break;
1429 			default:
1430 				len += (ip6e->ip6e_len + 1) << 3;
1431 				break;
1432 			}
1433 			nxt = ip6e->ip6e_nxt;
1434 		}
1435 		if (ip6e)
1436 			return (&ip6e->ip6e_nxt);
1437 		else
1438 			return NULL;
1439 	}
1440 }
1441 
1442 /*
1443  * get next header offset.  m will be retained.
1444  */
1445 int
1446 ip6_nexthdr(m, off, proto, nxtp)
1447 	struct mbuf *m;
1448 	int off;
1449 	int proto;
1450 	int *nxtp;
1451 {
1452 	struct ip6_hdr ip6;
1453 	struct ip6_ext ip6e;
1454 	struct ip6_frag fh;
1455 
1456 	/* just in case */
1457 	if (m == NULL)
1458 		panic("ip6_nexthdr: m == NULL");
1459 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1460 		return -1;
1461 
1462 	switch (proto) {
1463 	case IPPROTO_IPV6:
1464 		if (m->m_pkthdr.len < off + sizeof(ip6))
1465 			return -1;
1466 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1467 		if (nxtp)
1468 			*nxtp = ip6.ip6_nxt;
1469 		off += sizeof(ip6);
1470 		return off;
1471 
1472 	case IPPROTO_FRAGMENT:
1473 		/*
1474 		 * terminate parsing if it is not the first fragment,
1475 		 * it does not make sense to parse through it.
1476 		 */
1477 		if (m->m_pkthdr.len < off + sizeof(fh))
1478 			return -1;
1479 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1480 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1481 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1482 			return -1;
1483 		if (nxtp)
1484 			*nxtp = fh.ip6f_nxt;
1485 		off += sizeof(struct ip6_frag);
1486 		return off;
1487 
1488 	case IPPROTO_AH:
1489 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1490 			return -1;
1491 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1492 		if (nxtp)
1493 			*nxtp = ip6e.ip6e_nxt;
1494 		off += (ip6e.ip6e_len + 2) << 2;
1495 		return off;
1496 
1497 	case IPPROTO_HOPOPTS:
1498 	case IPPROTO_ROUTING:
1499 	case IPPROTO_DSTOPTS:
1500 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1501 			return -1;
1502 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1503 		if (nxtp)
1504 			*nxtp = ip6e.ip6e_nxt;
1505 		off += (ip6e.ip6e_len + 1) << 3;
1506 		return off;
1507 
1508 	case IPPROTO_NONE:
1509 	case IPPROTO_ESP:
1510 	case IPPROTO_IPCOMP:
1511 		/* give up */
1512 		return -1;
1513 
1514 	default:
1515 		return -1;
1516 	}
1517 
1518 	return -1;
1519 }
1520 
1521 /*
1522  * get offset for the last header in the chain.  m will be kept untainted.
1523  */
1524 int
1525 ip6_lasthdr(m, off, proto, nxtp)
1526 	struct mbuf *m;
1527 	int off;
1528 	int proto;
1529 	int *nxtp;
1530 {
1531 	int newoff;
1532 	int nxt;
1533 
1534 	if (!nxtp) {
1535 		nxt = -1;
1536 		nxtp = &nxt;
1537 	}
1538 	while (1) {
1539 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1540 		if (newoff < 0)
1541 			return off;
1542 		else if (newoff < off)
1543 			return -1;	/* invalid */
1544 		else if (newoff == off)
1545 			return newoff;
1546 
1547 		off = newoff;
1548 		proto = *nxtp;
1549 	}
1550 }
1551 
1552 struct ip6aux *
1553 ip6_addaux(m)
1554 	struct mbuf *m;
1555 {
1556 	struct m_tag *mtag;
1557 
1558 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1559 	if (!mtag) {
1560 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1561 		    M_NOWAIT);
1562 		if (mtag) {
1563 			m_tag_prepend(m, mtag);
1564 			bzero(mtag + 1, sizeof(struct ip6aux));
1565 		}
1566 	}
1567 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1568 }
1569 
1570 struct ip6aux *
1571 ip6_findaux(m)
1572 	struct mbuf *m;
1573 {
1574 	struct m_tag *mtag;
1575 
1576 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1577 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1578 }
1579 
1580 void
1581 ip6_delaux(m)
1582 	struct mbuf *m;
1583 {
1584 	struct m_tag *mtag;
1585 
1586 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1587 	if (mtag)
1588 		m_tag_delete(m, mtag);
1589 }
1590 
1591 /*
1592  * System control for IP6
1593  */
1594 
1595 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1596 	0,		0,		0,		0,
1597 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1598 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1599 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1600 	0,		0,		0,		0,
1601 	ENOPROTOOPT
1602 };
1603