xref: /freebsd/sys/netinet6/ip6_input.c (revision 729362425c09cf6b362366aabc6fb547eee8035a)
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  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_ip6fw.h"
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 #include "opt_pfil_hooks.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/proc.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #ifdef PFIL_HOOKS
94 #include <net/pfil.h>
95 #endif
96 
97 #include <netinet/in.h>
98 #include <netinet/in_systm.h>
99 #ifdef INET
100 #include <netinet/ip.h>
101 #include <netinet/ip_icmp.h>
102 #endif /* INET */
103 #include <netinet/ip6.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/icmp6.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/in6_prefix.h>
111 
112 #ifdef IPSEC
113 #include <netinet6/ipsec.h>
114 #ifdef INET6
115 #include <netinet6/ipsec6.h>
116 #endif
117 #endif
118 
119 #ifdef FAST_IPSEC
120 #include <netipsec/ipsec.h>
121 #include <netipsec/ipsec6.h>
122 #define	IPSEC
123 #endif /* FAST_IPSEC */
124 
125 #include <netinet6/ip6_fw.h>
126 
127 #include <netinet6/ip6protosw.h>
128 
129 #include <net/net_osdep.h>
130 
131 extern struct domain inet6domain;
132 
133 u_char ip6_protox[IPPROTO_MAX];
134 static struct ifqueue ip6intrq;
135 static int ip6qmaxlen = IFQ_MAXLEN;
136 struct in6_ifaddr *in6_ifaddr;
137 
138 extern struct callout in6_tmpaddrtimer_ch;
139 
140 int ip6_forward_srcrt;			/* XXX */
141 int ip6_sourcecheck;			/* XXX */
142 int ip6_sourcecheck_interval;		/* XXX */
143 
144 int ip6_ours_check_algorithm;
145 
146 
147 /* firewall hooks */
148 ip6_fw_chk_t *ip6_fw_chk_ptr;
149 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
150 int ip6_fw_enable = 1;
151 
152 struct ip6stat ip6stat;
153 
154 static void ip6_init2 __P((void *));
155 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
156 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
157 #ifdef PULLDOWN_TEST
158 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
159 #endif
160 
161 
162 /*
163  * IP6 initialization: fill in IP6 protocol switch table.
164  * All protocols not implemented in kernel go to raw IP6 protocol handler.
165  */
166 void
167 ip6_init()
168 {
169 	struct ip6protosw *pr;
170 	int i;
171 	struct timeval tv;
172 
173 #ifdef DIAGNOSTIC
174 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
175 		panic("sizeof(protosw) != sizeof(ip6protosw)");
176 #endif
177 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
178 	if (pr == 0)
179 		panic("ip6_init");
180 	for (i = 0; i < IPPROTO_MAX; i++)
181 		ip6_protox[i] = pr - inet6sw;
182 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
183 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
184 		if (pr->pr_domain->dom_family == PF_INET6 &&
185 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
186 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
187 	ip6intrq.ifq_maxlen = ip6qmaxlen;
188 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
189 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq);
190 	nd6_init();
191 	frag6_init();
192 	/*
193 	 * in many cases, random() here does NOT return random number
194 	 * as initialization during bootstrap time occur in fixed order.
195 	 */
196 	microtime(&tv);
197 	ip6_flow_seq = random() ^ tv.tv_usec;
198 	microtime(&tv);
199 	ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
200 }
201 
202 static void
203 ip6_init2(dummy)
204 	void *dummy;
205 {
206 
207 	/*
208 	 * to route local address of p2p link to loopback,
209 	 * assign loopback address first.
210 	 */
211 	in6_ifattach(&loif[0], NULL);
212 
213 	/* nd6_timer_init */
214 	callout_init(&nd6_timer_ch, 0);
215 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
216 
217 	/* router renumbering prefix list maintenance */
218 	callout_init(&in6_rr_timer_ch, 0);
219 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
220 
221 	/* timer for regeneranation of temporary addresses randomize ID */
222 	callout_reset(&in6_tmpaddrtimer_ch,
223 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
224 		       ip6_temp_regen_advance) * hz,
225 		      in6_tmpaddrtimer, NULL);
226 }
227 
228 /* cheat */
229 /* This must be after route_init(), which is now SI_ORDER_THIRD */
230 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
231 
232 extern struct	route_in6 ip6_forward_rt;
233 
234 void
235 ip6_input(m)
236 	struct mbuf *m;
237 {
238 	struct ip6_hdr *ip6;
239 	int off = sizeof(struct ip6_hdr), nest;
240 	u_int32_t plen;
241 	u_int32_t rtalert = ~0;
242 	int nxt, ours = 0;
243 	struct ifnet *deliverifp = NULL;
244 #ifdef  PFIL_HOOKS
245 	struct packet_filter_hook *pfh;
246 	struct mbuf *m0;
247 	int rv;
248 #endif  /* PFIL_HOOKS */
249 
250 #ifdef IPSEC
251 	/*
252 	 * should the inner packet be considered authentic?
253 	 * see comment in ah4_input().
254 	 */
255 	if (m) {
256 		m->m_flags &= ~M_AUTHIPHDR;
257 		m->m_flags &= ~M_AUTHIPDGM;
258 	}
259 #endif
260 
261 	/*
262 	 * make sure we don't have onion peering information into m_aux.
263 	 */
264 	ip6_delaux(m);
265 
266 	/*
267 	 * mbuf statistics
268 	 */
269 	if (m->m_flags & M_EXT) {
270 		if (m->m_next)
271 			ip6stat.ip6s_mext2m++;
272 		else
273 			ip6stat.ip6s_mext1++;
274 	} else {
275 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
276 		if (m->m_next) {
277 			if (m->m_flags & M_LOOP) {
278 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
279 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
280 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
281 			else
282 				ip6stat.ip6s_m2m[0]++;
283 		} else
284 			ip6stat.ip6s_m1++;
285 #undef M2MMAX
286 	}
287 
288 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
289 	ip6stat.ip6s_total++;
290 
291 #ifndef PULLDOWN_TEST
292 	/*
293 	 * L2 bridge code and some other code can return mbuf chain
294 	 * that does not conform to KAME requirement.  too bad.
295 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
296 	 */
297 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
298 		struct mbuf *n;
299 
300 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
301 		if (n)
302 			M_MOVE_PKTHDR(n, m);
303 		if (n && n->m_pkthdr.len > MHLEN) {
304 			MCLGET(n, M_DONTWAIT);
305 			if ((n->m_flags & M_EXT) == 0) {
306 				m_freem(n);
307 				n = NULL;
308 			}
309 		}
310 		if (n == NULL) {
311 			m_freem(m);
312 			return;	/*ENOBUFS*/
313 		}
314 
315 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
316 		n->m_len = n->m_pkthdr.len;
317 		m_freem(m);
318 		m = n;
319 	}
320 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
321 #endif
322 
323 	if (m->m_len < sizeof(struct ip6_hdr)) {
324 		struct ifnet *inifp;
325 		inifp = m->m_pkthdr.rcvif;
326 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
327 			ip6stat.ip6s_toosmall++;
328 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
329 			return;
330 		}
331 	}
332 
333 	ip6 = mtod(m, struct ip6_hdr *);
334 
335 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
336 		ip6stat.ip6s_badvers++;
337 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
338 		goto bad;
339 	}
340 
341 #ifdef PFIL_HOOKS
342 	/*
343 	 * Run through list of hooks for input packets.  If there are any
344 	 * filters which require that additional packets in the flow are
345 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
346 	 * Note that filters must _never_ set this flag, as another filter
347 	 * in the list may have previously cleared it.
348 	 */
349 	m0 = m;
350 	pfh = pfil_hook_get(PFIL_IN, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
351 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
352 		if (pfh->pfil_func) {
353 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
354 					    m->m_pkthdr.rcvif, 0, &m0);
355 			if (rv)
356 				return;
357 			m = m0;
358 			if (m == NULL)
359 				return;
360 			ip6 = mtod(m, struct ip6_hdr *);
361 		}
362 #endif /* PFIL_HOOKS */
363 
364 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
365 
366 	/*
367 	 * Check with the firewall...
368 	 */
369 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
370 		u_short port = 0;
371 		/* If ipfw says divert, we have to just drop packet */
372 		/* use port as a dummy argument */
373 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
374 			m_freem(m);
375 			m = NULL;
376 		}
377 		if (!m)
378 			return;
379 	}
380 
381 	/*
382 	 * Check against address spoofing/corruption.
383 	 */
384 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
385 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
386 		/*
387 		 * XXX: "badscope" is not very suitable for a multicast source.
388 		 */
389 		ip6stat.ip6s_badscope++;
390 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
391 		goto bad;
392 	}
393 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
394 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
395 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
396 		ip6stat.ip6s_badscope++;
397 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
398 		goto bad;
399 	}
400 
401 	/*
402 	 * The following check is not documented in specs.  A malicious
403 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
404 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
405 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
406 	 *
407 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
408 	 * support IPv4-less kernel compilation, we cannot support SIIT
409 	 * environment at all.  So, it makes more sense for us to reject any
410 	 * malicious packets for non-SIIT environment, than try to do a
411 	 * partical support for SIIT environment.
412 	 */
413 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
414 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
415 		ip6stat.ip6s_badscope++;
416 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
417 		goto bad;
418 	}
419 #if 0
420 	/*
421 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
422 	 *
423 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
424 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
425 	 * is revised to forbid relaying case.
426 	 */
427 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
428 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
429 		ip6stat.ip6s_badscope++;
430 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
431 		goto bad;
432 	}
433 #endif
434 
435 	/* drop packets if interface ID portion is already filled */
436 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
437 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
438 		    ip6->ip6_src.s6_addr16[1]) {
439 			ip6stat.ip6s_badscope++;
440 			goto bad;
441 		}
442 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
443 		    ip6->ip6_dst.s6_addr16[1]) {
444 			ip6stat.ip6s_badscope++;
445 			goto bad;
446 		}
447 	}
448 
449 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
450 		ip6->ip6_src.s6_addr16[1]
451 			= htons(m->m_pkthdr.rcvif->if_index);
452 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
453 		ip6->ip6_dst.s6_addr16[1]
454 			= htons(m->m_pkthdr.rcvif->if_index);
455 
456 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
457 	/*
458 	 * We use rt->rt_ifp to determine if the address is ours or not.
459 	 * If rt_ifp is lo0, the address is ours.
460 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
461 	 * so any address under fe80::%lo0/64 will be mistakenly considered
462 	 * local.  The special case is supplied to handle the case properly
463 	 * by actually looking at interface addresses
464 	 * (using in6ifa_ifpwithaddr).
465 	 */
466 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
467 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
468 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
469 			icmp6_error(m, ICMP6_DST_UNREACH,
470 			    ICMP6_DST_UNREACH_ADDR, 0);
471 			/* m is already freed */
472 			return;
473 		}
474 
475 		ours = 1;
476 		deliverifp = m->m_pkthdr.rcvif;
477 		goto hbhcheck;
478 	}
479 #endif
480 
481 	/*
482 	 * Multicast check
483 	 */
484 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
485 	  	struct	in6_multi *in6m = 0;
486 
487 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
488 		/*
489 		 * See if we belong to the destination multicast group on the
490 		 * arrival interface.
491 		 */
492 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
493 		if (in6m)
494 			ours = 1;
495 		else if (!ip6_mrouter) {
496 			ip6stat.ip6s_notmember++;
497 			ip6stat.ip6s_cantforward++;
498 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
499 			goto bad;
500 		}
501 		deliverifp = m->m_pkthdr.rcvif;
502 		goto hbhcheck;
503 	}
504 
505 	/*
506 	 *  Unicast check
507 	 */
508 	switch (ip6_ours_check_algorithm) {
509 	default:
510 		/*
511 		 * XXX: I intentionally broke our indentation rule here,
512 		 *      since this switch-case is just for measurement and
513 		 *      therefore should soon be removed.
514 		 */
515 	if (ip6_forward_rt.ro_rt != NULL &&
516 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
517 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
518 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
519 		ip6stat.ip6s_forward_cachehit++;
520 	else {
521 		struct sockaddr_in6 *dst6;
522 
523 		if (ip6_forward_rt.ro_rt) {
524 			/* route is down or destination is different */
525 			ip6stat.ip6s_forward_cachemiss++;
526 			RTFREE(ip6_forward_rt.ro_rt);
527 			ip6_forward_rt.ro_rt = 0;
528 		}
529 
530 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
531 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
532 		dst6->sin6_len = sizeof(struct sockaddr_in6);
533 		dst6->sin6_family = AF_INET6;
534 		dst6->sin6_addr = ip6->ip6_dst;
535 #ifdef SCOPEDROUTING
536 		ip6_forward_rt.ro_dst.sin6_scope_id =
537 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
538 #endif
539 
540 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
541 	}
542 
543 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
544 
545 	/*
546 	 * Accept the packet if the forwarding interface to the destination
547 	 * according to the routing table is the loopback interface,
548 	 * unless the associated route has a gateway.
549 	 * Note that this approach causes to accept a packet if there is a
550 	 * route to the loopback interface for the destination of the packet.
551 	 * But we think it's even useful in some situations, e.g. when using
552 	 * a special daemon which wants to intercept the packet.
553 	 *
554 	 * XXX: some OSes automatically make a cloned route for the destination
555 	 * of an outgoing packet.  If the outgoing interface of the packet
556 	 * is a loopback one, the kernel would consider the packet to be
557 	 * accepted, even if we have no such address assinged on the interface.
558 	 * We check the cloned flag of the route entry to reject such cases,
559 	 * assuming that route entries for our own addresses are not made by
560 	 * cloning (it should be true because in6_addloop explicitly installs
561 	 * the host route).  However, we might have to do an explicit check
562 	 * while it would be less efficient.  Or, should we rather install a
563 	 * reject route for such a case?
564 	 */
565 	if (ip6_forward_rt.ro_rt &&
566 	    (ip6_forward_rt.ro_rt->rt_flags &
567 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
568 #ifdef RTF_WASCLONED
569 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
570 #endif
571 #ifdef RTF_CLONED
572 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
573 #endif
574 #if 0
575 	    /*
576 	     * The check below is redundant since the comparison of
577 	     * the destination and the key of the rtentry has
578 	     * already done through looking up the routing table.
579 	     */
580 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
581 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
582 #endif
583 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
584 		struct in6_ifaddr *ia6 =
585 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
586 
587 		/*
588 		 * record address information into m_aux.
589 		 */
590 		(void)ip6_setdstifaddr(m, ia6);
591 
592 		/*
593 		 * packets to a tentative, duplicated, or somehow invalid
594 		 * address must not be accepted.
595 		 */
596 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
597 			/* this address is ready */
598 			ours = 1;
599 			deliverifp = ia6->ia_ifp;	/* correct? */
600 			/* Count the packet in the ip address stats */
601 			ia6->ia_ifa.if_ipackets++;
602 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
603 			goto hbhcheck;
604 		} else {
605 			/* address is not ready, so discard the packet. */
606 			nd6log((LOG_INFO,
607 			    "ip6_input: packet to an unready address %s->%s\n",
608 			    ip6_sprintf(&ip6->ip6_src),
609 			    ip6_sprintf(&ip6->ip6_dst)));
610 
611 			goto bad;
612 		}
613 	}
614 	} /* XXX indentation (see above) */
615 
616 	/*
617 	 * FAITH(Firewall Aided Internet Translator)
618 	 */
619 	if (ip6_keepfaith) {
620 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
621 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
622 			/* XXX do we need more sanity checks? */
623 			ours = 1;
624 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
625 			goto hbhcheck;
626 		}
627 	}
628 
629 	/*
630 	 * Now there is no reason to process the packet if it's not our own
631 	 * and we're not a router.
632 	 */
633 	if (!ip6_forwarding) {
634 		ip6stat.ip6s_cantforward++;
635 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
636 		goto bad;
637 	}
638 
639   hbhcheck:
640 	/*
641 	 * record address information into m_aux, if we don't have one yet.
642 	 * note that we are unable to record it, if the address is not listed
643 	 * as our interface address (e.g. multicast addresses, addresses
644 	 * within FAITH prefixes and such).
645 	 */
646 	if (deliverifp && !ip6_getdstifaddr(m)) {
647 		struct in6_ifaddr *ia6;
648 
649 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
650 		if (ia6) {
651 			if (!ip6_setdstifaddr(m, ia6)) {
652 				/*
653 				 * XXX maybe we should drop the packet here,
654 				 * as we could not provide enough information
655 				 * to the upper layers.
656 				 */
657 			}
658 		}
659 	}
660 
661 	/*
662 	 * Process Hop-by-Hop options header if it's contained.
663 	 * m may be modified in ip6_hopopts_input().
664 	 * If a JumboPayload option is included, plen will also be modified.
665 	 */
666 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
667 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
668 		struct ip6_hbh *hbh;
669 
670 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
671 #if 0	/*touches NULL pointer*/
672 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
673 #endif
674 			return;	/* m have already been freed */
675 		}
676 
677 		/* adjust pointer */
678 		ip6 = mtod(m, struct ip6_hdr *);
679 
680 		/*
681 		 * if the payload length field is 0 and the next header field
682 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
683 		 * option MUST be included.
684 		 */
685 		if (ip6->ip6_plen == 0 && plen == 0) {
686 			/*
687 			 * Note that if a valid jumbo payload option is
688 			 * contained, ip6_hoptops_input() must set a valid
689 			 * (non-zero) payload length to the variable plen.
690 			 */
691 			ip6stat.ip6s_badoptions++;
692 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
693 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
694 			icmp6_error(m, ICMP6_PARAM_PROB,
695 				    ICMP6_PARAMPROB_HEADER,
696 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
697 			return;
698 		}
699 #ifndef PULLDOWN_TEST
700 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
701 		hbh = (struct ip6_hbh *)(ip6 + 1);
702 #else
703 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
704 			sizeof(struct ip6_hbh));
705 		if (hbh == NULL) {
706 			ip6stat.ip6s_tooshort++;
707 			return;
708 		}
709 #endif
710 		nxt = hbh->ip6h_nxt;
711 
712 		/*
713 		 * accept the packet if a router alert option is included
714 		 * and we act as an IPv6 router.
715 		 */
716 		if (rtalert != ~0 && ip6_forwarding)
717 			ours = 1;
718 	} else
719 		nxt = ip6->ip6_nxt;
720 
721 	/*
722 	 * Check that the amount of data in the buffers
723 	 * is as at least much as the IPv6 header would have us expect.
724 	 * Trim mbufs if longer than we expect.
725 	 * Drop packet if shorter than we expect.
726 	 */
727 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
728 		ip6stat.ip6s_tooshort++;
729 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
730 		goto bad;
731 	}
732 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
733 		if (m->m_len == m->m_pkthdr.len) {
734 			m->m_len = sizeof(struct ip6_hdr) + plen;
735 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
736 		} else
737 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
738 	}
739 
740 	/*
741 	 * Forward if desirable.
742 	 */
743 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
744 		/*
745 		 * If we are acting as a multicast router, all
746 		 * incoming multicast packets are passed to the
747 		 * kernel-level multicast forwarding function.
748 		 * The packet is returned (relatively) intact; if
749 		 * ip6_mforward() returns a non-zero value, the packet
750 		 * must be discarded, else it may be accepted below.
751 		 */
752 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
753 			ip6stat.ip6s_cantforward++;
754 			m_freem(m);
755 			return;
756 		}
757 		if (!ours) {
758 			m_freem(m);
759 			return;
760 		}
761 	} else if (!ours) {
762 		ip6_forward(m, 0);
763 		return;
764 	}
765 
766 	ip6 = mtod(m, struct ip6_hdr *);
767 
768 	/*
769 	 * Malicious party may be able to use IPv4 mapped addr to confuse
770 	 * tcp/udp stack and bypass security checks (act as if it was from
771 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
772 	 *
773 	 * For SIIT end node behavior, you may want to disable the check.
774 	 * However, you will  become vulnerable to attacks using IPv4 mapped
775 	 * source.
776 	 */
777 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
778 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
779 		ip6stat.ip6s_badscope++;
780 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
781 		goto bad;
782 	}
783 
784 	/*
785 	 * Tell launch routine the next header
786 	 */
787 	ip6stat.ip6s_delivered++;
788 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
789 	nest = 0;
790 
791 	while (nxt != IPPROTO_DONE) {
792 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
793 			ip6stat.ip6s_toomanyhdr++;
794 			goto bad;
795 		}
796 
797 		/*
798 		 * protection against faulty packet - there should be
799 		 * more sanity checks in header chain processing.
800 		 */
801 		if (m->m_pkthdr.len < off) {
802 			ip6stat.ip6s_tooshort++;
803 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
804 			goto bad;
805 		}
806 
807 #if 0
808 		/*
809 		 * do we need to do it for every header?  yeah, other
810 		 * functions can play with it (like re-allocate and copy).
811 		 */
812 		mhist = ip6_addaux(m);
813 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
814 			hist = mtod(mhist, caddr_t) + mhist->m_len;
815 			bcopy(&nxt, hist, sizeof(nxt));
816 			mhist->m_len += sizeof(nxt);
817 		} else {
818 			ip6stat.ip6s_toomanyhdr++;
819 			goto bad;
820 		}
821 #endif
822 
823 #ifdef IPSEC
824 		/*
825 		 * enforce IPsec policy checking if we are seeing last header.
826 		 * note that we do not visit this with protocols with pcb layer
827 		 * code - like udp/tcp/raw ip.
828 		 */
829 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
830 		    ipsec6_in_reject(m, NULL)) {
831 			ipsec6stat.in_polvio++;
832 			goto bad;
833 		}
834 #endif
835 
836 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
837 	}
838 	return;
839  bad:
840 	m_freem(m);
841 }
842 
843 /*
844  * set/grab in6_ifaddr correspond to IPv6 destination address.
845  * XXX backward compatibility wrapper
846  */
847 static struct ip6aux *
848 ip6_setdstifaddr(m, ia6)
849 	struct mbuf *m;
850 	struct in6_ifaddr *ia6;
851 {
852 	struct ip6aux *n;
853 
854 	n = ip6_addaux(m);
855 	if (n)
856 		n->ip6a_dstia6 = ia6;
857 	return n;	/* NULL if failed to set */
858 }
859 
860 struct in6_ifaddr *
861 ip6_getdstifaddr(m)
862 	struct mbuf *m;
863 {
864 	struct ip6aux *n;
865 
866 	n = ip6_findaux(m);
867 	if (n)
868 		return n->ip6a_dstia6;
869 	else
870 		return NULL;
871 }
872 
873 /*
874  * Hop-by-Hop options header processing. If a valid jumbo payload option is
875  * included, the real payload length will be stored in plenp.
876  */
877 static int
878 ip6_hopopts_input(plenp, rtalertp, mp, offp)
879 	u_int32_t *plenp;
880 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
881 	struct mbuf **mp;
882 	int *offp;
883 {
884 	struct mbuf *m = *mp;
885 	int off = *offp, hbhlen;
886 	struct ip6_hbh *hbh;
887 	u_int8_t *opt;
888 
889 	/* validation of the length of the header */
890 #ifndef PULLDOWN_TEST
891 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
892 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
893 	hbhlen = (hbh->ip6h_len + 1) << 3;
894 
895 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
896 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
897 #else
898 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
899 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
900 	if (hbh == NULL) {
901 		ip6stat.ip6s_tooshort++;
902 		return -1;
903 	}
904 	hbhlen = (hbh->ip6h_len + 1) << 3;
905 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
906 		hbhlen);
907 	if (hbh == NULL) {
908 		ip6stat.ip6s_tooshort++;
909 		return -1;
910 	}
911 #endif
912 	off += hbhlen;
913 	hbhlen -= sizeof(struct ip6_hbh);
914 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
915 
916 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
917 				hbhlen, rtalertp, plenp) < 0)
918 		return(-1);
919 
920 	*offp = off;
921 	*mp = m;
922 	return(0);
923 }
924 
925 /*
926  * Search header for all Hop-by-hop options and process each option.
927  * This function is separate from ip6_hopopts_input() in order to
928  * handle a case where the sending node itself process its hop-by-hop
929  * options header. In such a case, the function is called from ip6_output().
930  *
931  * The function assumes that hbh header is located right after the IPv6 header
932  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
933  * opthead + hbhlen is located in continuous memory region.
934  */
935 int
936 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
937 	struct mbuf *m;
938 	u_int8_t *opthead;
939 	int hbhlen;
940 	u_int32_t *rtalertp;
941 	u_int32_t *plenp;
942 {
943 	struct ip6_hdr *ip6;
944 	int optlen = 0;
945 	u_int8_t *opt = opthead;
946 	u_int16_t rtalert_val;
947 	u_int32_t jumboplen;
948 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
949 
950 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
951 		switch (*opt) {
952 		case IP6OPT_PAD1:
953 			optlen = 1;
954 			break;
955 		case IP6OPT_PADN:
956 			if (hbhlen < IP6OPT_MINLEN) {
957 				ip6stat.ip6s_toosmall++;
958 				goto bad;
959 			}
960 			optlen = *(opt + 1) + 2;
961 			break;
962 		case IP6OPT_RTALERT:
963 			/* XXX may need check for alignment */
964 			if (hbhlen < IP6OPT_RTALERT_LEN) {
965 				ip6stat.ip6s_toosmall++;
966 				goto bad;
967 			}
968 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
969 				/* XXX stat */
970 				icmp6_error(m, ICMP6_PARAM_PROB,
971 					    ICMP6_PARAMPROB_HEADER,
972 					    erroff + opt + 1 - opthead);
973 				return(-1);
974 			}
975 			optlen = IP6OPT_RTALERT_LEN;
976 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
977 			*rtalertp = ntohs(rtalert_val);
978 			break;
979 		case IP6OPT_JUMBO:
980 			/* XXX may need check for alignment */
981 			if (hbhlen < IP6OPT_JUMBO_LEN) {
982 				ip6stat.ip6s_toosmall++;
983 				goto bad;
984 			}
985 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
986 				/* XXX stat */
987 				icmp6_error(m, ICMP6_PARAM_PROB,
988 					    ICMP6_PARAMPROB_HEADER,
989 					    erroff + opt + 1 - opthead);
990 				return(-1);
991 			}
992 			optlen = IP6OPT_JUMBO_LEN;
993 
994 			/*
995 			 * IPv6 packets that have non 0 payload length
996 			 * must not contain a jumbo payload option.
997 			 */
998 			ip6 = mtod(m, struct ip6_hdr *);
999 			if (ip6->ip6_plen) {
1000 				ip6stat.ip6s_badoptions++;
1001 				icmp6_error(m, ICMP6_PARAM_PROB,
1002 					    ICMP6_PARAMPROB_HEADER,
1003 					    erroff + opt - opthead);
1004 				return(-1);
1005 			}
1006 
1007 			/*
1008 			 * We may see jumbolen in unaligned location, so
1009 			 * we'd need to perform bcopy().
1010 			 */
1011 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1012 			jumboplen = (u_int32_t)htonl(jumboplen);
1013 
1014 #if 1
1015 			/*
1016 			 * if there are multiple jumbo payload options,
1017 			 * *plenp will be non-zero and the packet will be
1018 			 * rejected.
1019 			 * the behavior may need some debate in ipngwg -
1020 			 * multiple options does not make sense, however,
1021 			 * there's no explicit mention in specification.
1022 			 */
1023 			if (*plenp != 0) {
1024 				ip6stat.ip6s_badoptions++;
1025 				icmp6_error(m, ICMP6_PARAM_PROB,
1026 					    ICMP6_PARAMPROB_HEADER,
1027 					    erroff + opt + 2 - opthead);
1028 				return(-1);
1029 			}
1030 #endif
1031 
1032 			/*
1033 			 * jumbo payload length must be larger than 65535.
1034 			 */
1035 			if (jumboplen <= IPV6_MAXPACKET) {
1036 				ip6stat.ip6s_badoptions++;
1037 				icmp6_error(m, ICMP6_PARAM_PROB,
1038 					    ICMP6_PARAMPROB_HEADER,
1039 					    erroff + opt + 2 - opthead);
1040 				return(-1);
1041 			}
1042 			*plenp = jumboplen;
1043 
1044 			break;
1045 		default:		/* unknown option */
1046 			if (hbhlen < IP6OPT_MINLEN) {
1047 				ip6stat.ip6s_toosmall++;
1048 				goto bad;
1049 			}
1050 			optlen = ip6_unknown_opt(opt, m,
1051 			    erroff + opt - opthead);
1052 			if (optlen == -1)
1053 				return(-1);
1054 			optlen += 2;
1055 			break;
1056 		}
1057 	}
1058 
1059 	return(0);
1060 
1061   bad:
1062 	m_freem(m);
1063 	return(-1);
1064 }
1065 
1066 /*
1067  * Unknown option processing.
1068  * The third argument `off' is the offset from the IPv6 header to the option,
1069  * which is necessary if the IPv6 header the and option header and IPv6 header
1070  * is not continuous in order to return an ICMPv6 error.
1071  */
1072 int
1073 ip6_unknown_opt(optp, m, off)
1074 	u_int8_t *optp;
1075 	struct mbuf *m;
1076 	int off;
1077 {
1078 	struct ip6_hdr *ip6;
1079 
1080 	switch (IP6OPT_TYPE(*optp)) {
1081 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1082 		return((int)*(optp + 1));
1083 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1084 		m_freem(m);
1085 		return(-1);
1086 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1087 		ip6stat.ip6s_badoptions++;
1088 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1089 		return(-1);
1090 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1091 		ip6stat.ip6s_badoptions++;
1092 		ip6 = mtod(m, struct ip6_hdr *);
1093 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1094 		    (m->m_flags & (M_BCAST|M_MCAST)))
1095 			m_freem(m);
1096 		else
1097 			icmp6_error(m, ICMP6_PARAM_PROB,
1098 				    ICMP6_PARAMPROB_OPTION, off);
1099 		return(-1);
1100 	}
1101 
1102 	m_freem(m);		/* XXX: NOTREACHED */
1103 	return(-1);
1104 }
1105 
1106 /*
1107  * Create the "control" list for this pcb.
1108  * The function will not modify mbuf chain at all.
1109  *
1110  * with KAME mbuf chain restriction:
1111  * The routine will be called from upper layer handlers like tcp6_input().
1112  * Thus the routine assumes that the caller (tcp6_input) have already
1113  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1114  * very first mbuf on the mbuf chain.
1115  */
1116 void
1117 ip6_savecontrol(in6p, mp, ip6, m)
1118 	struct inpcb *in6p;
1119 	struct mbuf **mp;
1120 	struct ip6_hdr *ip6;
1121 	struct mbuf *m;
1122 {
1123 #if __FreeBSD_version >= 500000
1124 	struct thread *td = curthread;	/* XXX */
1125 #else
1126 	struct proc *td = curproc;	/* XXX */
1127 #endif
1128 	int privileged = 0;
1129 	int rthdr_exist = 0;
1130 
1131 
1132 	if (td && !suser(td))
1133  		privileged++;
1134 
1135 #ifdef SO_TIMESTAMP
1136 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1137 		struct timeval tv;
1138 
1139 		microtime(&tv);
1140 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1141 				      SCM_TIMESTAMP, SOL_SOCKET);
1142 		if (*mp) {
1143 			mp = &(*mp)->m_next;
1144 		}
1145 	}
1146 #endif
1147 
1148 	/* RFC 2292 sec. 5 */
1149 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1150 		struct in6_pktinfo pi6;
1151 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1152 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1153 			pi6.ipi6_addr.s6_addr16[1] = 0;
1154 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1155 					? m->m_pkthdr.rcvif->if_index
1156 					: 0;
1157 		*mp = sbcreatecontrol((caddr_t) &pi6,
1158 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1159 			IPPROTO_IPV6);
1160 		if (*mp)
1161 			mp = &(*mp)->m_next;
1162 	}
1163 
1164 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1165 		int hlim = ip6->ip6_hlim & 0xff;
1166 		*mp = sbcreatecontrol((caddr_t) &hlim,
1167 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1168 		if (*mp)
1169 			mp = &(*mp)->m_next;
1170 	}
1171 
1172 	/*
1173 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1174 	 * for the option, but it might be too strict, since there might
1175 	 * be some hop-by-hop options which can be returned to normal user.
1176 	 * See RFC 2292 section 6.
1177 	 */
1178 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1179 		/*
1180 		 * Check if a hop-by-hop options header is contatined in the
1181 		 * received packet, and if so, store the options as ancillary
1182 		 * data. Note that a hop-by-hop options header must be
1183 		 * just after the IPv6 header, which fact is assured through
1184 		 * the IPv6 input processing.
1185 		 */
1186 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1187 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1188 			struct ip6_hbh *hbh;
1189 			int hbhlen = 0;
1190 #ifdef PULLDOWN_TEST
1191 			struct mbuf *ext;
1192 #endif
1193 
1194 #ifndef PULLDOWN_TEST
1195 			hbh = (struct ip6_hbh *)(ip6 + 1);
1196 			hbhlen = (hbh->ip6h_len + 1) << 3;
1197 #else
1198 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1199 			    ip6->ip6_nxt);
1200 			if (ext == NULL) {
1201 				ip6stat.ip6s_tooshort++;
1202 				return;
1203 			}
1204 			hbh = mtod(ext, struct ip6_hbh *);
1205 			hbhlen = (hbh->ip6h_len + 1) << 3;
1206 			if (hbhlen != ext->m_len) {
1207 				m_freem(ext);
1208 				ip6stat.ip6s_tooshort++;
1209 				return;
1210 			}
1211 #endif
1212 
1213 			/*
1214 			 * XXX: We copy whole the header even if a jumbo
1215 			 * payload option is included, which option is to
1216 			 * be removed before returning in the RFC 2292.
1217 			 * Note: this constraint is removed in 2292bis.
1218 			 */
1219 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1220 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1221 			if (*mp)
1222 				mp = &(*mp)->m_next;
1223 #ifdef PULLDOWN_TEST
1224 			m_freem(ext);
1225 #endif
1226 		}
1227 	}
1228 
1229 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1230 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1231 		int proto, off, nxt;
1232 
1233 		/*
1234 		 * go through the header chain to see if a routing header is
1235 		 * contained in the packet. We need this information to store
1236 		 * destination options headers (if any) properly.
1237 		 * XXX: performance issue. We should record this info when
1238 		 * processing extension headers in incoming routine.
1239 		 * (todo) use m_aux?
1240 		 */
1241 		proto = IPPROTO_IPV6;
1242 		off = 0;
1243 		nxt = -1;
1244 		while (1) {
1245 			int newoff;
1246 
1247 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1248 			if (newoff < 0)
1249 				break;
1250 			if (newoff < off) /* invalid, check for safety */
1251 				break;
1252 			if ((proto = nxt) == IPPROTO_ROUTING) {
1253 				rthdr_exist = 1;
1254 				break;
1255 			}
1256 			off = newoff;
1257 		}
1258 	}
1259 
1260 	if ((in6p->in6p_flags &
1261 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1262 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1263 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1264 
1265 		/*
1266 		 * Search for destination options headers or routing
1267 		 * header(s) through the header chain, and stores each
1268 		 * header as ancillary data.
1269 		 * Note that the order of the headers remains in
1270 		 * the chain of ancillary data.
1271 		 */
1272 		while (1) {	/* is explicit loop prevention necessary? */
1273 			struct ip6_ext *ip6e = NULL;
1274 			int elen;
1275 #ifdef PULLDOWN_TEST
1276 			struct mbuf *ext = NULL;
1277 #endif
1278 
1279 			/*
1280 			 * if it is not an extension header, don't try to
1281 			 * pull it from the chain.
1282 			 */
1283 			switch (nxt) {
1284 			case IPPROTO_DSTOPTS:
1285 			case IPPROTO_ROUTING:
1286 			case IPPROTO_HOPOPTS:
1287 			case IPPROTO_AH: /* is it possible? */
1288 				break;
1289 			default:
1290 				goto loopend;
1291 			}
1292 
1293 #ifndef PULLDOWN_TEST
1294 			if (off + sizeof(*ip6e) > m->m_len)
1295 				goto loopend;
1296 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1297 			if (nxt == IPPROTO_AH)
1298 				elen = (ip6e->ip6e_len + 2) << 2;
1299 			else
1300 				elen = (ip6e->ip6e_len + 1) << 3;
1301 			if (off + elen > m->m_len)
1302 				goto loopend;
1303 #else
1304 			ext = ip6_pullexthdr(m, off, nxt);
1305 			if (ext == NULL) {
1306 				ip6stat.ip6s_tooshort++;
1307 				return;
1308 			}
1309 			ip6e = mtod(ext, struct ip6_ext *);
1310 			if (nxt == IPPROTO_AH)
1311 				elen = (ip6e->ip6e_len + 2) << 2;
1312 			else
1313 				elen = (ip6e->ip6e_len + 1) << 3;
1314 			if (elen != ext->m_len) {
1315 				m_freem(ext);
1316 				ip6stat.ip6s_tooshort++;
1317 				return;
1318 			}
1319 #endif
1320 
1321 			switch (nxt) {
1322 			case IPPROTO_DSTOPTS:
1323 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1324 					break;
1325 
1326 				/*
1327 				 * We also require super-user privilege for
1328 				 * the option.
1329 				 * See the comments on IN6_HOPOPTS.
1330 				 */
1331 				if (!privileged)
1332 					break;
1333 
1334 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1335 						      IPV6_DSTOPTS,
1336 						      IPPROTO_IPV6);
1337 				if (*mp)
1338 					mp = &(*mp)->m_next;
1339 				break;
1340 			case IPPROTO_ROUTING:
1341 				if (!in6p->in6p_flags & IN6P_RTHDR)
1342 					break;
1343 
1344 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1345 						      IPV6_RTHDR,
1346 						      IPPROTO_IPV6);
1347 				if (*mp)
1348 					mp = &(*mp)->m_next;
1349 				break;
1350 			case IPPROTO_HOPOPTS:
1351 			case IPPROTO_AH: /* is it possible? */
1352 				break;
1353 
1354 			default:
1355 				/*
1356 			 	 * other cases have been filtered in the above.
1357 				 * none will visit this case.  here we supply
1358 				 * the code just in case (nxt overwritten or
1359 				 * other cases).
1360 				 */
1361 #ifdef PULLDOWN_TEST
1362 				m_freem(ext);
1363 #endif
1364 				goto loopend;
1365 
1366 			}
1367 
1368 			/* proceed with the next header. */
1369 			off += elen;
1370 			nxt = ip6e->ip6e_nxt;
1371 			ip6e = NULL;
1372 #ifdef PULLDOWN_TEST
1373 			m_freem(ext);
1374 			ext = NULL;
1375 #endif
1376 		}
1377 	  loopend:
1378 		;
1379 	}
1380 
1381 }
1382 
1383 #ifdef PULLDOWN_TEST
1384 /*
1385  * pull single extension header from mbuf chain.  returns single mbuf that
1386  * contains the result, or NULL on error.
1387  */
1388 static struct mbuf *
1389 ip6_pullexthdr(m, off, nxt)
1390 	struct mbuf *m;
1391 	size_t off;
1392 	int nxt;
1393 {
1394 	struct ip6_ext ip6e;
1395 	size_t elen;
1396 	struct mbuf *n;
1397 
1398 #ifdef DIAGNOSTIC
1399 	switch (nxt) {
1400 	case IPPROTO_DSTOPTS:
1401 	case IPPROTO_ROUTING:
1402 	case IPPROTO_HOPOPTS:
1403 	case IPPROTO_AH: /* is it possible? */
1404 		break;
1405 	default:
1406 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1407 	}
1408 #endif
1409 
1410 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1411 	if (nxt == IPPROTO_AH)
1412 		elen = (ip6e.ip6e_len + 2) << 2;
1413 	else
1414 		elen = (ip6e.ip6e_len + 1) << 3;
1415 
1416 	MGET(n, M_DONTWAIT, MT_DATA);
1417 	if (n && elen >= MLEN) {
1418 		MCLGET(n, M_DONTWAIT);
1419 		if ((n->m_flags & M_EXT) == 0) {
1420 			m_free(n);
1421 			n = NULL;
1422 		}
1423 	}
1424 	if (!n)
1425 		return NULL;
1426 
1427 	n->m_len = 0;
1428 	if (elen >= M_TRAILINGSPACE(n)) {
1429 		m_free(n);
1430 		return NULL;
1431 	}
1432 
1433 	m_copydata(m, off, elen, mtod(n, caddr_t));
1434 	n->m_len = elen;
1435 	return n;
1436 }
1437 #endif
1438 
1439 /*
1440  * Get pointer to the previous header followed by the header
1441  * currently processed.
1442  * XXX: This function supposes that
1443  *	M includes all headers,
1444  *	the next header field and the header length field of each header
1445  *	are valid, and
1446  *	the sum of each header length equals to OFF.
1447  * Because of these assumptions, this function must be called very
1448  * carefully. Moreover, it will not be used in the near future when
1449  * we develop `neater' mechanism to process extension headers.
1450  */
1451 char *
1452 ip6_get_prevhdr(m, off)
1453 	struct mbuf *m;
1454 	int off;
1455 {
1456 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1457 
1458 	if (off == sizeof(struct ip6_hdr))
1459 		return(&ip6->ip6_nxt);
1460 	else {
1461 		int len, nxt;
1462 		struct ip6_ext *ip6e = NULL;
1463 
1464 		nxt = ip6->ip6_nxt;
1465 		len = sizeof(struct ip6_hdr);
1466 		while (len < off) {
1467 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1468 
1469 			switch (nxt) {
1470 			case IPPROTO_FRAGMENT:
1471 				len += sizeof(struct ip6_frag);
1472 				break;
1473 			case IPPROTO_AH:
1474 				len += (ip6e->ip6e_len + 2) << 2;
1475 				break;
1476 			default:
1477 				len += (ip6e->ip6e_len + 1) << 3;
1478 				break;
1479 			}
1480 			nxt = ip6e->ip6e_nxt;
1481 		}
1482 		if (ip6e)
1483 			return(&ip6e->ip6e_nxt);
1484 		else
1485 			return NULL;
1486 	}
1487 }
1488 
1489 /*
1490  * get next header offset.  m will be retained.
1491  */
1492 int
1493 ip6_nexthdr(m, off, proto, nxtp)
1494 	struct mbuf *m;
1495 	int off;
1496 	int proto;
1497 	int *nxtp;
1498 {
1499 	struct ip6_hdr ip6;
1500 	struct ip6_ext ip6e;
1501 	struct ip6_frag fh;
1502 
1503 	/* just in case */
1504 	if (m == NULL)
1505 		panic("ip6_nexthdr: m == NULL");
1506 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1507 		return -1;
1508 
1509 	switch (proto) {
1510 	case IPPROTO_IPV6:
1511 		if (m->m_pkthdr.len < off + sizeof(ip6))
1512 			return -1;
1513 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1514 		if (nxtp)
1515 			*nxtp = ip6.ip6_nxt;
1516 		off += sizeof(ip6);
1517 		return off;
1518 
1519 	case IPPROTO_FRAGMENT:
1520 		/*
1521 		 * terminate parsing if it is not the first fragment,
1522 		 * it does not make sense to parse through it.
1523 		 */
1524 		if (m->m_pkthdr.len < off + sizeof(fh))
1525 			return -1;
1526 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1527 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1528 			return -1;
1529 		if (nxtp)
1530 			*nxtp = fh.ip6f_nxt;
1531 		off += sizeof(struct ip6_frag);
1532 		return off;
1533 
1534 	case IPPROTO_AH:
1535 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1536 			return -1;
1537 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1538 		if (nxtp)
1539 			*nxtp = ip6e.ip6e_nxt;
1540 		off += (ip6e.ip6e_len + 2) << 2;
1541 		return off;
1542 
1543 	case IPPROTO_HOPOPTS:
1544 	case IPPROTO_ROUTING:
1545 	case IPPROTO_DSTOPTS:
1546 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1547 			return -1;
1548 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1549 		if (nxtp)
1550 			*nxtp = ip6e.ip6e_nxt;
1551 		off += (ip6e.ip6e_len + 1) << 3;
1552 		return off;
1553 
1554 	case IPPROTO_NONE:
1555 	case IPPROTO_ESP:
1556 	case IPPROTO_IPCOMP:
1557 		/* give up */
1558 		return -1;
1559 
1560 	default:
1561 		return -1;
1562 	}
1563 
1564 	return -1;
1565 }
1566 
1567 /*
1568  * get offset for the last header in the chain.  m will be kept untainted.
1569  */
1570 int
1571 ip6_lasthdr(m, off, proto, nxtp)
1572 	struct mbuf *m;
1573 	int off;
1574 	int proto;
1575 	int *nxtp;
1576 {
1577 	int newoff;
1578 	int nxt;
1579 
1580 	if (!nxtp) {
1581 		nxt = -1;
1582 		nxtp = &nxt;
1583 	}
1584 	while (1) {
1585 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1586 		if (newoff < 0)
1587 			return off;
1588 		else if (newoff < off)
1589 			return -1;	/* invalid */
1590 		else if (newoff == off)
1591 			return newoff;
1592 
1593 		off = newoff;
1594 		proto = *nxtp;
1595 	}
1596 }
1597 
1598 struct ip6aux *
1599 ip6_addaux(m)
1600 	struct mbuf *m;
1601 {
1602 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1603 	if (!tag) {
1604 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1605 				sizeof (struct ip6aux),
1606 				M_NOWAIT);
1607 		if (tag)
1608 			m_tag_prepend(m, tag);
1609 	}
1610 	if (tag)
1611 		bzero(tag+1, sizeof (struct ip6aux));
1612 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1613 }
1614 
1615 struct ip6aux *
1616 ip6_findaux(m)
1617 	struct mbuf *m;
1618 {
1619 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1620 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1621 }
1622 
1623 void
1624 ip6_delaux(m)
1625 	struct mbuf *m;
1626 {
1627 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1628 	if (tag)
1629 		m_tag_delete(m, tag);
1630 }
1631 
1632 /*
1633  * System control for IP6
1634  */
1635 
1636 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1637 	0,		0,		0,		0,
1638 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1639 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1640 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1641 	0,		0,		0,		0,
1642 	ENOPROTOOPT
1643 };
1644