xref: /freebsd/sys/netinet6/ip6_input.c (revision 262e143bd46171a6415a5b28af260a5efa2a3db8)
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 in6_addr odst;
240 	int srcrt = 0;
241 
242 	GIANT_REQUIRED;			/* XXX for now */
243 #ifdef IPSEC
244 	/*
245 	 * should the inner packet be considered authentic?
246 	 * see comment in ah4_input().
247 	 */
248 	if (m) {
249 		m->m_flags &= ~M_AUTHIPHDR;
250 		m->m_flags &= ~M_AUTHIPDGM;
251 	}
252 #endif
253 
254 	/*
255 	 * make sure we don't have onion peering information into m_tag.
256 	 */
257 	ip6_delaux(m);
258 
259 	/*
260 	 * mbuf statistics
261 	 */
262 	if (m->m_flags & M_EXT) {
263 		if (m->m_next)
264 			ip6stat.ip6s_mext2m++;
265 		else
266 			ip6stat.ip6s_mext1++;
267 	} else {
268 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
269 		if (m->m_next) {
270 			if (m->m_flags & M_LOOP) {
271 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
272 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
273 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
274 			else
275 				ip6stat.ip6s_m2m[0]++;
276 		} else
277 			ip6stat.ip6s_m1++;
278 #undef M2MMAX
279 	}
280 
281 	/* drop the packet if IPv6 operation is disabled on the IF */
282 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
283 		m_freem(m);
284 		return;
285 	}
286 
287 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
288 	ip6stat.ip6s_total++;
289 
290 #ifndef PULLDOWN_TEST
291 	/*
292 	 * L2 bridge code and some other code can return mbuf chain
293 	 * that does not conform to KAME requirement.  too bad.
294 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
295 	 */
296 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
297 		struct mbuf *n;
298 
299 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
300 		if (n)
301 			M_MOVE_PKTHDR(n, m);
302 		if (n && n->m_pkthdr.len > MHLEN) {
303 			MCLGET(n, M_DONTWAIT);
304 			if ((n->m_flags & M_EXT) == 0) {
305 				m_freem(n);
306 				n = NULL;
307 			}
308 		}
309 		if (n == NULL) {
310 			m_freem(m);
311 			return;	/* ENOBUFS */
312 		}
313 
314 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
315 		n->m_len = n->m_pkthdr.len;
316 		m_freem(m);
317 		m = n;
318 	}
319 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
320 #endif
321 
322 	if (m->m_len < sizeof(struct ip6_hdr)) {
323 		struct ifnet *inifp;
324 		inifp = m->m_pkthdr.rcvif;
325 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
326 			ip6stat.ip6s_toosmall++;
327 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
328 			return;
329 		}
330 	}
331 
332 	ip6 = mtod(m, struct ip6_hdr *);
333 
334 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
335 		ip6stat.ip6s_badvers++;
336 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
337 		goto bad;
338 	}
339 
340 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
341 
342 	/*
343 	 * Check against address spoofing/corruption.
344 	 */
345 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
346 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
347 		/*
348 		 * XXX: "badscope" is not very suitable for a multicast source.
349 		 */
350 		ip6stat.ip6s_badscope++;
351 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
352 		goto bad;
353 	}
354 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
355 	    !(m->m_flags & M_LOOP)) {
356 		/*
357 		 * In this case, the packet should come from the loopback
358 		 * interface.  However, we cannot just check the if_flags,
359 		 * because ip6_mloopback() passes the "actual" interface
360 		 * as the outgoing/incoming interface.
361 		 */
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 
367 #ifdef ALTQ
368 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
369 		/* packet is dropped by traffic conditioner */
370 		return;
371 	}
372 #endif
373 	/*
374 	 * The following check is not documented in specs.  A malicious
375 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
376 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
377 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
378 	 *
379 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
380 	 * support IPv4-less kernel compilation, we cannot support SIIT
381 	 * environment at all.  So, it makes more sense for us to reject any
382 	 * malicious packets for non-SIIT environment, than try to do a
383 	 * partial support for SIIT environment.
384 	 */
385 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
386 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
387 		ip6stat.ip6s_badscope++;
388 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
389 		goto bad;
390 	}
391 #if 0
392 	/*
393 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
394 	 *
395 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
396 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
397 	 * is revised to forbid relaying case.
398 	 */
399 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
400 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
401 		ip6stat.ip6s_badscope++;
402 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
403 		goto bad;
404 	}
405 #endif
406 
407 	/*
408 	 * Disambiguate address scope zones (if there is ambiguity).
409 	 * We first make sure that the original source or destination address
410 	 * is not in our internal form for scoped addresses.  Such addresses
411 	 * are not necessarily invalid spec-wise, but we cannot accept them due
412 	 * to the usage conflict.
413 	 * in6_setscope() then also checks and rejects the cases where src or
414 	 * dst are the loopback address and the receiving interface
415 	 * is not loopback.
416 	 */
417 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
418 		ip6stat.ip6s_badscope++; /* XXX */
419 		goto bad;
420 	}
421 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
422 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
423 		ip6stat.ip6s_badscope++;
424 		goto bad;
425 	}
426 
427 	/*
428 	 * Run through list of hooks for input packets.
429 	 *
430 	 * NB: Beware of the destination address changing
431 	 *     (e.g. by NAT rewriting).  When this happens,
432 	 *     tell ip6_forward to do the right thing.
433 	 */
434 	odst = ip6->ip6_dst;
435 
436 	/* Jump over all PFIL processing if hooks are not active. */
437 	if (inet6_pfil_hook.ph_busy_count == -1)
438 		goto passin;
439 
440 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
441 		return;
442 	if (m == NULL)			/* consumed by filter */
443 		return;
444 	ip6 = mtod(m, struct ip6_hdr *);
445 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
446 
447 passin:
448 	/*
449 	 * Check with the firewall...
450 	 */
451 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
452 		u_short port = 0;
453 		/* If ipfw says divert, we have to just drop packet */
454 		/* use port as a dummy argument */
455 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
456 			m_freem(m);
457 			m = NULL;
458 		}
459 		if (!m)
460 			return;
461 	}
462 
463 	/*
464 	 * Multicast check
465 	 */
466 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
467 	  	struct in6_multi *in6m = 0;
468 
469 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
470 		/*
471 		 * See if we belong to the destination multicast group on the
472 		 * arrival interface.
473 		 */
474 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
475 		if (in6m)
476 			ours = 1;
477 		else if (!ip6_mrouter) {
478 			ip6stat.ip6s_notmember++;
479 			ip6stat.ip6s_cantforward++;
480 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
481 			goto bad;
482 		}
483 		deliverifp = m->m_pkthdr.rcvif;
484 		goto hbhcheck;
485 	}
486 
487 	/*
488 	 *  Unicast check
489 	 */
490 	if (ip6_forward_rt.ro_rt != NULL &&
491 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
492 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
493 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
494 		ip6stat.ip6s_forward_cachehit++;
495 	else {
496 		struct sockaddr_in6 *dst6;
497 
498 		if (ip6_forward_rt.ro_rt) {
499 			/* route is down or destination is different */
500 			ip6stat.ip6s_forward_cachemiss++;
501 			RTFREE(ip6_forward_rt.ro_rt);
502 			ip6_forward_rt.ro_rt = 0;
503 		}
504 
505 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
506 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
507 		dst6->sin6_len = sizeof(struct sockaddr_in6);
508 		dst6->sin6_family = AF_INET6;
509 		dst6->sin6_addr = ip6->ip6_dst;
510 
511 		rtalloc((struct route *)&ip6_forward_rt);
512 	}
513 
514 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
515 
516 	/*
517 	 * Accept the packet if the forwarding interface to the destination
518 	 * according to the routing table is the loopback interface,
519 	 * unless the associated route has a gateway.
520 	 * Note that this approach causes to accept a packet if there is a
521 	 * route to the loopback interface for the destination of the packet.
522 	 * But we think it's even useful in some situations, e.g. when using
523 	 * a special daemon which wants to intercept the packet.
524 	 *
525 	 * XXX: some OSes automatically make a cloned route for the destination
526 	 * of an outgoing packet.  If the outgoing interface of the packet
527 	 * is a loopback one, the kernel would consider the packet to be
528 	 * accepted, even if we have no such address assinged on the interface.
529 	 * We check the cloned flag of the route entry to reject such cases,
530 	 * assuming that route entries for our own addresses are not made by
531 	 * cloning (it should be true because in6_addloop explicitly installs
532 	 * the host route).  However, we might have to do an explicit check
533 	 * while it would be less efficient.  Or, should we rather install a
534 	 * reject route for such a case?
535 	 */
536 	if (ip6_forward_rt.ro_rt &&
537 	    (ip6_forward_rt.ro_rt->rt_flags &
538 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
539 #ifdef RTF_WASCLONED
540 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
541 #endif
542 #ifdef RTF_CLONED
543 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
544 #endif
545 #if 0
546 	    /*
547 	     * The check below is redundant since the comparison of
548 	     * the destination and the key of the rtentry has
549 	     * already done through looking up the routing table.
550 	     */
551 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
552 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
553 #endif
554 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
555 		struct in6_ifaddr *ia6 =
556 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
557 
558 		/*
559 		 * record address information into m_tag.
560 		 */
561 		(void)ip6_setdstifaddr(m, ia6);
562 
563 		/*
564 		 * packets to a tentative, duplicated, or somehow invalid
565 		 * address must not be accepted.
566 		 */
567 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
568 			/* this address is ready */
569 			ours = 1;
570 			deliverifp = ia6->ia_ifp;	/* correct? */
571 			/* Count the packet in the ip address stats */
572 			ia6->ia_ifa.if_ipackets++;
573 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
574 			goto hbhcheck;
575 		} else {
576 			/* address is not ready, so discard the packet. */
577 			nd6log((LOG_INFO,
578 			    "ip6_input: packet to an unready address %s->%s\n",
579 			    ip6_sprintf(&ip6->ip6_src),
580 			    ip6_sprintf(&ip6->ip6_dst)));
581 
582 			goto bad;
583 		}
584 	}
585 
586 	/*
587 	 * FAITH (Firewall Aided Internet Translator)
588 	 */
589 	if (ip6_keepfaith) {
590 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
591 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
592 			/* XXX do we need more sanity checks? */
593 			ours = 1;
594 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
595 			goto hbhcheck;
596 		}
597 	}
598 
599 	/*
600 	 * Now there is no reason to process the packet if it's not our own
601 	 * and we're not a router.
602 	 */
603 	if (!ip6_forwarding) {
604 		ip6stat.ip6s_cantforward++;
605 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
606 		goto bad;
607 	}
608 
609   hbhcheck:
610 	/*
611 	 * record address information into m_tag, if we don't have one yet.
612 	 * note that we are unable to record it, if the address is not listed
613 	 * as our interface address (e.g. multicast addresses, addresses
614 	 * within FAITH prefixes and such).
615 	 */
616 	if (deliverifp && !ip6_getdstifaddr(m)) {
617 		struct in6_ifaddr *ia6;
618 
619 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
620 		if (ia6) {
621 			if (!ip6_setdstifaddr(m, ia6)) {
622 				/*
623 				 * XXX maybe we should drop the packet here,
624 				 * as we could not provide enough information
625 				 * to the upper layers.
626 				 */
627 			}
628 		}
629 	}
630 
631 	/*
632 	 * Process Hop-by-Hop options header if it's contained.
633 	 * m may be modified in ip6_hopopts_input().
634 	 * If a JumboPayload option is included, plen will also be modified.
635 	 */
636 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
637 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
638 		struct ip6_hbh *hbh;
639 
640 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
641 #if 0	/*touches NULL pointer*/
642 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
643 #endif
644 			return;	/* m have already been freed */
645 		}
646 
647 		/* adjust pointer */
648 		ip6 = mtod(m, struct ip6_hdr *);
649 
650 		/*
651 		 * if the payload length field is 0 and the next header field
652 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
653 		 * option MUST be included.
654 		 */
655 		if (ip6->ip6_plen == 0 && plen == 0) {
656 			/*
657 			 * Note that if a valid jumbo payload option is
658 			 * contained, ip6_hopopts_input() must set a valid
659 			 * (non-zero) payload length to the variable plen.
660 			 */
661 			ip6stat.ip6s_badoptions++;
662 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
663 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
664 			icmp6_error(m, ICMP6_PARAM_PROB,
665 				    ICMP6_PARAMPROB_HEADER,
666 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
667 			return;
668 		}
669 #ifndef PULLDOWN_TEST
670 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
671 		hbh = (struct ip6_hbh *)(ip6 + 1);
672 #else
673 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
674 			sizeof(struct ip6_hbh));
675 		if (hbh == NULL) {
676 			ip6stat.ip6s_tooshort++;
677 			return;
678 		}
679 #endif
680 		nxt = hbh->ip6h_nxt;
681 
682 		/*
683 		 * accept the packet if a router alert option is included
684 		 * and we act as an IPv6 router.
685 		 */
686 		if (rtalert != ~0 && ip6_forwarding)
687 			ours = 1;
688 	} else
689 		nxt = ip6->ip6_nxt;
690 
691 	/*
692 	 * Check that the amount of data in the buffers
693 	 * is as at least much as the IPv6 header would have us expect.
694 	 * Trim mbufs if longer than we expect.
695 	 * Drop packet if shorter than we expect.
696 	 */
697 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
698 		ip6stat.ip6s_tooshort++;
699 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
700 		goto bad;
701 	}
702 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
703 		if (m->m_len == m->m_pkthdr.len) {
704 			m->m_len = sizeof(struct ip6_hdr) + plen;
705 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
706 		} else
707 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
708 	}
709 
710 	/*
711 	 * Forward if desirable.
712 	 */
713 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
714 		/*
715 		 * If we are acting as a multicast router, all
716 		 * incoming multicast packets are passed to the
717 		 * kernel-level multicast forwarding function.
718 		 * The packet is returned (relatively) intact; if
719 		 * ip6_mforward() returns a non-zero value, the packet
720 		 * must be discarded, else it may be accepted below.
721 		 */
722 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
723 			ip6stat.ip6s_cantforward++;
724 			m_freem(m);
725 			return;
726 		}
727 		if (!ours) {
728 			m_freem(m);
729 			return;
730 		}
731 	} else if (!ours) {
732 		ip6_forward(m, srcrt);
733 		return;
734 	}
735 
736 	ip6 = mtod(m, struct ip6_hdr *);
737 
738 	/*
739 	 * Malicious party may be able to use IPv4 mapped addr to confuse
740 	 * tcp/udp stack and bypass security checks (act as if it was from
741 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
742 	 *
743 	 * For SIIT end node behavior, you may want to disable the check.
744 	 * However, you will  become vulnerable to attacks using IPv4 mapped
745 	 * source.
746 	 */
747 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
748 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
749 		ip6stat.ip6s_badscope++;
750 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
751 		goto bad;
752 	}
753 
754 	/*
755 	 * Tell launch routine the next header
756 	 */
757 	ip6stat.ip6s_delivered++;
758 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
759 	nest = 0;
760 
761 	while (nxt != IPPROTO_DONE) {
762 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
763 			ip6stat.ip6s_toomanyhdr++;
764 			goto bad;
765 		}
766 
767 		/*
768 		 * protection against faulty packet - there should be
769 		 * more sanity checks in header chain processing.
770 		 */
771 		if (m->m_pkthdr.len < off) {
772 			ip6stat.ip6s_tooshort++;
773 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
774 			goto bad;
775 		}
776 
777 #ifdef IPSEC
778 		/*
779 		 * enforce IPsec policy checking if we are seeing last header.
780 		 * note that we do not visit this with protocols with pcb layer
781 		 * code - like udp/tcp/raw ip.
782 		 */
783 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
784 		    ipsec6_in_reject(m, NULL)) {
785 			ipsec6stat.in_polvio++;
786 			goto bad;
787 		}
788 #endif
789 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
790 	}
791 	return;
792  bad:
793 	m_freem(m);
794 }
795 
796 /*
797  * set/grab in6_ifaddr correspond to IPv6 destination address.
798  * XXX backward compatibility wrapper
799  */
800 static struct ip6aux *
801 ip6_setdstifaddr(m, ia6)
802 	struct mbuf *m;
803 	struct in6_ifaddr *ia6;
804 {
805 	struct ip6aux *ip6a;
806 
807 	ip6a = ip6_addaux(m);
808 	if (ip6a)
809 		ip6a->ip6a_dstia6 = ia6;
810 	return ip6a;	/* NULL if failed to set */
811 }
812 
813 struct in6_ifaddr *
814 ip6_getdstifaddr(m)
815 	struct mbuf *m;
816 {
817 	struct ip6aux *ip6a;
818 
819 	ip6a = ip6_findaux(m);
820 	if (ip6a)
821 		return ip6a->ip6a_dstia6;
822 	else
823 		return NULL;
824 }
825 
826 /*
827  * Hop-by-Hop options header processing. If a valid jumbo payload option is
828  * included, the real payload length will be stored in plenp.
829  */
830 static int
831 ip6_hopopts_input(plenp, rtalertp, mp, offp)
832 	u_int32_t *plenp;
833 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
834 	struct mbuf **mp;
835 	int *offp;
836 {
837 	struct mbuf *m = *mp;
838 	int off = *offp, hbhlen;
839 	struct ip6_hbh *hbh;
840 	u_int8_t *opt;
841 
842 	/* validation of the length of the header */
843 #ifndef PULLDOWN_TEST
844 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
845 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
846 	hbhlen = (hbh->ip6h_len + 1) << 3;
847 
848 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
849 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
850 #else
851 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
852 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
853 	if (hbh == NULL) {
854 		ip6stat.ip6s_tooshort++;
855 		return -1;
856 	}
857 	hbhlen = (hbh->ip6h_len + 1) << 3;
858 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
859 		hbhlen);
860 	if (hbh == NULL) {
861 		ip6stat.ip6s_tooshort++;
862 		return -1;
863 	}
864 #endif
865 	off += hbhlen;
866 	hbhlen -= sizeof(struct ip6_hbh);
867 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
868 
869 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
870 				hbhlen, rtalertp, plenp) < 0)
871 		return (-1);
872 
873 	*offp = off;
874 	*mp = m;
875 	return (0);
876 }
877 
878 /*
879  * Search header for all Hop-by-hop options and process each option.
880  * This function is separate from ip6_hopopts_input() in order to
881  * handle a case where the sending node itself process its hop-by-hop
882  * options header. In such a case, the function is called from ip6_output().
883  *
884  * The function assumes that hbh header is located right after the IPv6 header
885  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
886  * opthead + hbhlen is located in continuous memory region.
887  */
888 int
889 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
890 	struct mbuf *m;
891 	u_int8_t *opthead;
892 	int hbhlen;
893 	u_int32_t *rtalertp;
894 	u_int32_t *plenp;
895 {
896 	struct ip6_hdr *ip6;
897 	int optlen = 0;
898 	u_int8_t *opt = opthead;
899 	u_int16_t rtalert_val;
900 	u_int32_t jumboplen;
901 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
902 
903 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
904 		switch (*opt) {
905 		case IP6OPT_PAD1:
906 			optlen = 1;
907 			break;
908 		case IP6OPT_PADN:
909 			if (hbhlen < IP6OPT_MINLEN) {
910 				ip6stat.ip6s_toosmall++;
911 				goto bad;
912 			}
913 			optlen = *(opt + 1) + 2;
914 			break;
915 		case IP6OPT_ROUTER_ALERT:
916 			/* XXX may need check for alignment */
917 			if (hbhlen < IP6OPT_RTALERT_LEN) {
918 				ip6stat.ip6s_toosmall++;
919 				goto bad;
920 			}
921 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
922 				/* XXX stat */
923 				icmp6_error(m, ICMP6_PARAM_PROB,
924 				    ICMP6_PARAMPROB_HEADER,
925 				    erroff + opt + 1 - opthead);
926 				return (-1);
927 			}
928 			optlen = IP6OPT_RTALERT_LEN;
929 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
930 			*rtalertp = ntohs(rtalert_val);
931 			break;
932 		case IP6OPT_JUMBO:
933 			/* XXX may need check for alignment */
934 			if (hbhlen < IP6OPT_JUMBO_LEN) {
935 				ip6stat.ip6s_toosmall++;
936 				goto bad;
937 			}
938 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
939 				/* XXX stat */
940 				icmp6_error(m, ICMP6_PARAM_PROB,
941 				    ICMP6_PARAMPROB_HEADER,
942 				    erroff + opt + 1 - opthead);
943 				return (-1);
944 			}
945 			optlen = IP6OPT_JUMBO_LEN;
946 
947 			/*
948 			 * IPv6 packets that have non 0 payload length
949 			 * must not contain a jumbo payload option.
950 			 */
951 			ip6 = mtod(m, struct ip6_hdr *);
952 			if (ip6->ip6_plen) {
953 				ip6stat.ip6s_badoptions++;
954 				icmp6_error(m, ICMP6_PARAM_PROB,
955 				    ICMP6_PARAMPROB_HEADER,
956 				    erroff + opt - opthead);
957 				return (-1);
958 			}
959 
960 			/*
961 			 * We may see jumbolen in unaligned location, so
962 			 * we'd need to perform bcopy().
963 			 */
964 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
965 			jumboplen = (u_int32_t)htonl(jumboplen);
966 
967 #if 1
968 			/*
969 			 * if there are multiple jumbo payload options,
970 			 * *plenp will be non-zero and the packet will be
971 			 * rejected.
972 			 * the behavior may need some debate in ipngwg -
973 			 * multiple options does not make sense, however,
974 			 * there's no explicit mention in specification.
975 			 */
976 			if (*plenp != 0) {
977 				ip6stat.ip6s_badoptions++;
978 				icmp6_error(m, ICMP6_PARAM_PROB,
979 				    ICMP6_PARAMPROB_HEADER,
980 				    erroff + opt + 2 - opthead);
981 				return (-1);
982 			}
983 #endif
984 
985 			/*
986 			 * jumbo payload length must be larger than 65535.
987 			 */
988 			if (jumboplen <= IPV6_MAXPACKET) {
989 				ip6stat.ip6s_badoptions++;
990 				icmp6_error(m, ICMP6_PARAM_PROB,
991 				    ICMP6_PARAMPROB_HEADER,
992 				    erroff + opt + 2 - opthead);
993 				return (-1);
994 			}
995 			*plenp = jumboplen;
996 
997 			break;
998 		default:		/* unknown option */
999 			if (hbhlen < IP6OPT_MINLEN) {
1000 				ip6stat.ip6s_toosmall++;
1001 				goto bad;
1002 			}
1003 			optlen = ip6_unknown_opt(opt, m,
1004 			    erroff + opt - opthead);
1005 			if (optlen == -1)
1006 				return (-1);
1007 			optlen += 2;
1008 			break;
1009 		}
1010 	}
1011 
1012 	return (0);
1013 
1014   bad:
1015 	m_freem(m);
1016 	return (-1);
1017 }
1018 
1019 /*
1020  * Unknown option processing.
1021  * The third argument `off' is the offset from the IPv6 header to the option,
1022  * which is necessary if the IPv6 header the and option header and IPv6 header
1023  * is not continuous in order to return an ICMPv6 error.
1024  */
1025 int
1026 ip6_unknown_opt(optp, m, off)
1027 	u_int8_t *optp;
1028 	struct mbuf *m;
1029 	int off;
1030 {
1031 	struct ip6_hdr *ip6;
1032 
1033 	switch (IP6OPT_TYPE(*optp)) {
1034 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1035 		return ((int)*(optp + 1));
1036 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1037 		m_freem(m);
1038 		return (-1);
1039 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1040 		ip6stat.ip6s_badoptions++;
1041 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1042 		return (-1);
1043 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1044 		ip6stat.ip6s_badoptions++;
1045 		ip6 = mtod(m, struct ip6_hdr *);
1046 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1047 		    (m->m_flags & (M_BCAST|M_MCAST)))
1048 			m_freem(m);
1049 		else
1050 			icmp6_error(m, ICMP6_PARAM_PROB,
1051 				    ICMP6_PARAMPROB_OPTION, off);
1052 		return (-1);
1053 	}
1054 
1055 	m_freem(m);		/* XXX: NOTREACHED */
1056 	return (-1);
1057 }
1058 
1059 /*
1060  * Create the "control" list for this pcb.
1061  * The function will not modify mbuf chain at all.
1062  *
1063  * with KAME mbuf chain restriction:
1064  * The routine will be called from upper layer handlers like tcp6_input().
1065  * Thus the routine assumes that the caller (tcp6_input) have already
1066  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1067  * very first mbuf on the mbuf chain.
1068  */
1069 void
1070 ip6_savecontrol(in6p, m, mp)
1071 	struct inpcb *in6p;
1072 	struct mbuf *m, **mp;
1073 {
1074 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1075 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1076 
1077 #ifdef SO_TIMESTAMP
1078 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1079 		struct timeval tv;
1080 
1081 		microtime(&tv);
1082 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1083 		    SCM_TIMESTAMP, SOL_SOCKET);
1084 		if (*mp)
1085 			mp = &(*mp)->m_next;
1086 	}
1087 #endif
1088 
1089 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1090 		return;
1091 
1092 	/* RFC 2292 sec. 5 */
1093 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1094 		struct in6_pktinfo pi6;
1095 
1096 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1097 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1098 		pi6.ipi6_ifindex =
1099 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1100 
1101 		*mp = sbcreatecontrol((caddr_t) &pi6,
1102 		    sizeof(struct in6_pktinfo),
1103 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1104 		if (*mp)
1105 			mp = &(*mp)->m_next;
1106 	}
1107 
1108 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1109 		int hlim = ip6->ip6_hlim & 0xff;
1110 
1111 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1112 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1113 		if (*mp)
1114 			mp = &(*mp)->m_next;
1115 	}
1116 
1117 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1118 		u_int32_t flowinfo;
1119 		int tclass;
1120 
1121 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1122 		flowinfo >>= 20;
1123 
1124 		tclass = flowinfo & 0xff;
1125 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1126 		    IPV6_TCLASS, IPPROTO_IPV6);
1127 		if (*mp)
1128 			mp = &(*mp)->m_next;
1129 	}
1130 
1131 	/*
1132 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1133 	 * privilege for the option (see ip6_ctloutput), but it might be too
1134 	 * strict, since there might be some hop-by-hop options which can be
1135 	 * returned to normal user.
1136 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1137 	 */
1138 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1139 		/*
1140 		 * Check if a hop-by-hop options header is contatined in the
1141 		 * received packet, and if so, store the options as ancillary
1142 		 * data. Note that a hop-by-hop options header must be
1143 		 * just after the IPv6 header, which is assured through the
1144 		 * IPv6 input processing.
1145 		 */
1146 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1147 			struct ip6_hbh *hbh;
1148 			int hbhlen = 0;
1149 #ifdef PULLDOWN_TEST
1150 			struct mbuf *ext;
1151 #endif
1152 
1153 #ifndef PULLDOWN_TEST
1154 			hbh = (struct ip6_hbh *)(ip6 + 1);
1155 			hbhlen = (hbh->ip6h_len + 1) << 3;
1156 #else
1157 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1158 			    ip6->ip6_nxt);
1159 			if (ext == NULL) {
1160 				ip6stat.ip6s_tooshort++;
1161 				return;
1162 			}
1163 			hbh = mtod(ext, struct ip6_hbh *);
1164 			hbhlen = (hbh->ip6h_len + 1) << 3;
1165 			if (hbhlen != ext->m_len) {
1166 				m_freem(ext);
1167 				ip6stat.ip6s_tooshort++;
1168 				return;
1169 			}
1170 #endif
1171 
1172 			/*
1173 			 * XXX: We copy the whole header even if a
1174 			 * jumbo payload option is included, the option which
1175 			 * is to be removed before returning according to
1176 			 * RFC2292.
1177 			 * Note: this constraint is removed in RFC3542
1178 			 */
1179 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1180 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1181 			    IPPROTO_IPV6);
1182 			if (*mp)
1183 				mp = &(*mp)->m_next;
1184 #ifdef PULLDOWN_TEST
1185 			m_freem(ext);
1186 #endif
1187 		}
1188 	}
1189 
1190 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1191 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1192 
1193 		/*
1194 		 * Search for destination options headers or routing
1195 		 * header(s) through the header chain, and stores each
1196 		 * header as ancillary data.
1197 		 * Note that the order of the headers remains in
1198 		 * the chain of ancillary data.
1199 		 */
1200 		while (1) {	/* is explicit loop prevention necessary? */
1201 			struct ip6_ext *ip6e = NULL;
1202 			int elen;
1203 #ifdef PULLDOWN_TEST
1204 			struct mbuf *ext = NULL;
1205 #endif
1206 
1207 			/*
1208 			 * if it is not an extension header, don't try to
1209 			 * pull it from the chain.
1210 			 */
1211 			switch (nxt) {
1212 			case IPPROTO_DSTOPTS:
1213 			case IPPROTO_ROUTING:
1214 			case IPPROTO_HOPOPTS:
1215 			case IPPROTO_AH: /* is it possible? */
1216 				break;
1217 			default:
1218 				goto loopend;
1219 			}
1220 
1221 #ifndef PULLDOWN_TEST
1222 			if (off + sizeof(*ip6e) > m->m_len)
1223 				goto loopend;
1224 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1225 			if (nxt == IPPROTO_AH)
1226 				elen = (ip6e->ip6e_len + 2) << 2;
1227 			else
1228 				elen = (ip6e->ip6e_len + 1) << 3;
1229 			if (off + elen > m->m_len)
1230 				goto loopend;
1231 #else
1232 			ext = ip6_pullexthdr(m, off, nxt);
1233 			if (ext == NULL) {
1234 				ip6stat.ip6s_tooshort++;
1235 				return;
1236 			}
1237 			ip6e = mtod(ext, struct ip6_ext *);
1238 			if (nxt == IPPROTO_AH)
1239 				elen = (ip6e->ip6e_len + 2) << 2;
1240 			else
1241 				elen = (ip6e->ip6e_len + 1) << 3;
1242 			if (elen != ext->m_len) {
1243 				m_freem(ext);
1244 				ip6stat.ip6s_tooshort++;
1245 				return;
1246 			}
1247 #endif
1248 
1249 			switch (nxt) {
1250 			case IPPROTO_DSTOPTS:
1251 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1252 					break;
1253 
1254 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1255 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1256 				    IPPROTO_IPV6);
1257 				if (*mp)
1258 					mp = &(*mp)->m_next;
1259 				break;
1260 			case IPPROTO_ROUTING:
1261 				if (!in6p->in6p_flags & IN6P_RTHDR)
1262 					break;
1263 
1264 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1265 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1266 				    IPPROTO_IPV6);
1267 				if (*mp)
1268 					mp = &(*mp)->m_next;
1269 				break;
1270 			case IPPROTO_HOPOPTS:
1271 			case IPPROTO_AH: /* is it possible? */
1272 				break;
1273 
1274 			default:
1275 				/*
1276 			 	 * other cases have been filtered in the above.
1277 				 * none will visit this case.  here we supply
1278 				 * the code just in case (nxt overwritten or
1279 				 * other cases).
1280 				 */
1281 #ifdef PULLDOWN_TEST
1282 				m_freem(ext);
1283 #endif
1284 				goto loopend;
1285 
1286 			}
1287 
1288 			/* proceed with the next header. */
1289 			off += elen;
1290 			nxt = ip6e->ip6e_nxt;
1291 			ip6e = NULL;
1292 #ifdef PULLDOWN_TEST
1293 			m_freem(ext);
1294 			ext = NULL;
1295 #endif
1296 		}
1297 	  loopend:
1298 		;
1299 	}
1300 
1301 #undef IS2292
1302 }
1303 
1304 void
1305 ip6_notify_pmtu(in6p, dst, mtu)
1306 	struct inpcb *in6p;
1307 	struct sockaddr_in6 *dst;
1308 	u_int32_t *mtu;
1309 {
1310 	struct socket *so;
1311 	struct mbuf *m_mtu;
1312 	struct ip6_mtuinfo mtuctl;
1313 
1314 	so =  in6p->inp_socket;
1315 
1316 	if (mtu == NULL)
1317 		return;
1318 
1319 #ifdef DIAGNOSTIC
1320 	if (so == NULL)		/* I believe this is impossible */
1321 		panic("ip6_notify_pmtu: socket is NULL");
1322 #endif
1323 
1324 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1325 	mtuctl.ip6m_mtu = *mtu;
1326 	mtuctl.ip6m_addr = *dst;
1327 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1328 		return;
1329 
1330 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1331 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1332 		return;
1333 
1334 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1335 	    == 0) {
1336 		m_freem(m_mtu);
1337 		/* XXX: should count statistics */
1338 	} else
1339 		sorwakeup(so);
1340 
1341 	return;
1342 }
1343 
1344 #ifdef PULLDOWN_TEST
1345 /*
1346  * pull single extension header from mbuf chain.  returns single mbuf that
1347  * contains the result, or NULL on error.
1348  */
1349 static struct mbuf *
1350 ip6_pullexthdr(m, off, nxt)
1351 	struct mbuf *m;
1352 	size_t off;
1353 	int nxt;
1354 {
1355 	struct ip6_ext ip6e;
1356 	size_t elen;
1357 	struct mbuf *n;
1358 
1359 #ifdef DIAGNOSTIC
1360 	switch (nxt) {
1361 	case IPPROTO_DSTOPTS:
1362 	case IPPROTO_ROUTING:
1363 	case IPPROTO_HOPOPTS:
1364 	case IPPROTO_AH: /* is it possible? */
1365 		break;
1366 	default:
1367 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1368 	}
1369 #endif
1370 
1371 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1372 	if (nxt == IPPROTO_AH)
1373 		elen = (ip6e.ip6e_len + 2) << 2;
1374 	else
1375 		elen = (ip6e.ip6e_len + 1) << 3;
1376 
1377 	MGET(n, M_DONTWAIT, MT_DATA);
1378 	if (n && elen >= MLEN) {
1379 		MCLGET(n, M_DONTWAIT);
1380 		if ((n->m_flags & M_EXT) == 0) {
1381 			m_free(n);
1382 			n = NULL;
1383 		}
1384 	}
1385 	if (!n)
1386 		return NULL;
1387 
1388 	n->m_len = 0;
1389 	if (elen >= M_TRAILINGSPACE(n)) {
1390 		m_free(n);
1391 		return NULL;
1392 	}
1393 
1394 	m_copydata(m, off, elen, mtod(n, caddr_t));
1395 	n->m_len = elen;
1396 	return n;
1397 }
1398 #endif
1399 
1400 /*
1401  * Get pointer to the previous header followed by the header
1402  * currently processed.
1403  * XXX: This function supposes that
1404  *	M includes all headers,
1405  *	the next header field and the header length field of each header
1406  *	are valid, and
1407  *	the sum of each header length equals to OFF.
1408  * Because of these assumptions, this function must be called very
1409  * carefully. Moreover, it will not be used in the near future when
1410  * we develop `neater' mechanism to process extension headers.
1411  */
1412 char *
1413 ip6_get_prevhdr(m, off)
1414 	struct mbuf *m;
1415 	int off;
1416 {
1417 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1418 
1419 	if (off == sizeof(struct ip6_hdr))
1420 		return (&ip6->ip6_nxt);
1421 	else {
1422 		int len, nxt;
1423 		struct ip6_ext *ip6e = NULL;
1424 
1425 		nxt = ip6->ip6_nxt;
1426 		len = sizeof(struct ip6_hdr);
1427 		while (len < off) {
1428 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1429 
1430 			switch (nxt) {
1431 			case IPPROTO_FRAGMENT:
1432 				len += sizeof(struct ip6_frag);
1433 				break;
1434 			case IPPROTO_AH:
1435 				len += (ip6e->ip6e_len + 2) << 2;
1436 				break;
1437 			default:
1438 				len += (ip6e->ip6e_len + 1) << 3;
1439 				break;
1440 			}
1441 			nxt = ip6e->ip6e_nxt;
1442 		}
1443 		if (ip6e)
1444 			return (&ip6e->ip6e_nxt);
1445 		else
1446 			return NULL;
1447 	}
1448 }
1449 
1450 /*
1451  * get next header offset.  m will be retained.
1452  */
1453 int
1454 ip6_nexthdr(m, off, proto, nxtp)
1455 	struct mbuf *m;
1456 	int off;
1457 	int proto;
1458 	int *nxtp;
1459 {
1460 	struct ip6_hdr ip6;
1461 	struct ip6_ext ip6e;
1462 	struct ip6_frag fh;
1463 
1464 	/* just in case */
1465 	if (m == NULL)
1466 		panic("ip6_nexthdr: m == NULL");
1467 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1468 		return -1;
1469 
1470 	switch (proto) {
1471 	case IPPROTO_IPV6:
1472 		if (m->m_pkthdr.len < off + sizeof(ip6))
1473 			return -1;
1474 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1475 		if (nxtp)
1476 			*nxtp = ip6.ip6_nxt;
1477 		off += sizeof(ip6);
1478 		return off;
1479 
1480 	case IPPROTO_FRAGMENT:
1481 		/*
1482 		 * terminate parsing if it is not the first fragment,
1483 		 * it does not make sense to parse through it.
1484 		 */
1485 		if (m->m_pkthdr.len < off + sizeof(fh))
1486 			return -1;
1487 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1488 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1489 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1490 			return -1;
1491 		if (nxtp)
1492 			*nxtp = fh.ip6f_nxt;
1493 		off += sizeof(struct ip6_frag);
1494 		return off;
1495 
1496 	case IPPROTO_AH:
1497 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1498 			return -1;
1499 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1500 		if (nxtp)
1501 			*nxtp = ip6e.ip6e_nxt;
1502 		off += (ip6e.ip6e_len + 2) << 2;
1503 		return off;
1504 
1505 	case IPPROTO_HOPOPTS:
1506 	case IPPROTO_ROUTING:
1507 	case IPPROTO_DSTOPTS:
1508 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1509 			return -1;
1510 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1511 		if (nxtp)
1512 			*nxtp = ip6e.ip6e_nxt;
1513 		off += (ip6e.ip6e_len + 1) << 3;
1514 		return off;
1515 
1516 	case IPPROTO_NONE:
1517 	case IPPROTO_ESP:
1518 	case IPPROTO_IPCOMP:
1519 		/* give up */
1520 		return -1;
1521 
1522 	default:
1523 		return -1;
1524 	}
1525 
1526 	return -1;
1527 }
1528 
1529 /*
1530  * get offset for the last header in the chain.  m will be kept untainted.
1531  */
1532 int
1533 ip6_lasthdr(m, off, proto, nxtp)
1534 	struct mbuf *m;
1535 	int off;
1536 	int proto;
1537 	int *nxtp;
1538 {
1539 	int newoff;
1540 	int nxt;
1541 
1542 	if (!nxtp) {
1543 		nxt = -1;
1544 		nxtp = &nxt;
1545 	}
1546 	while (1) {
1547 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1548 		if (newoff < 0)
1549 			return off;
1550 		else if (newoff < off)
1551 			return -1;	/* invalid */
1552 		else if (newoff == off)
1553 			return newoff;
1554 
1555 		off = newoff;
1556 		proto = *nxtp;
1557 	}
1558 }
1559 
1560 struct ip6aux *
1561 ip6_addaux(m)
1562 	struct mbuf *m;
1563 {
1564 	struct m_tag *mtag;
1565 
1566 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1567 	if (!mtag) {
1568 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1569 		    M_NOWAIT);
1570 		if (mtag) {
1571 			m_tag_prepend(m, mtag);
1572 			bzero(mtag + 1, sizeof(struct ip6aux));
1573 		}
1574 	}
1575 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1576 }
1577 
1578 struct ip6aux *
1579 ip6_findaux(m)
1580 	struct mbuf *m;
1581 {
1582 	struct m_tag *mtag;
1583 
1584 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1585 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1586 }
1587 
1588 void
1589 ip6_delaux(m)
1590 	struct mbuf *m;
1591 {
1592 	struct m_tag *mtag;
1593 
1594 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1595 	if (mtag)
1596 		m_tag_delete(m, mtag);
1597 }
1598 
1599 /*
1600  * System control for IP6
1601  */
1602 
1603 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1604 	0,		0,		0,		0,
1605 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1606 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1607 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1608 	0,		0,		0,		0,
1609 	ENOPROTOOPT
1610 };
1611