xref: /freebsd/sys/netinet6/ip6_input.c (revision 1b6c76a2fe091c74f08427e6c870851025a9cf67)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.194 2001/05/27 13:28:35 itojun 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/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_types.h>
89 #include <net/if_dl.h>
90 #include <net/route.h>
91 #include <net/netisr.h>
92 #include <net/intrq.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 #include <netinet6/ip6_fw.h>
120 
121 #include <netinet6/ip6protosw.h>
122 
123 #include "faith.h"
124 #include "gif.h"
125 
126 #include <net/net_osdep.h>
127 
128 extern struct domain inet6domain;
129 extern struct ip6protosw inet6sw[];
130 
131 u_char ip6_protox[IPPROTO_MAX];
132 static int ip6qmaxlen = IFQ_MAXLEN;
133 struct in6_ifaddr *in6_ifaddr;
134 
135 extern struct callout in6_tmpaddrtimer_ch;
136 
137 int ip6_forward_srcrt;			/* XXX */
138 int ip6_sourcecheck;			/* XXX */
139 int ip6_sourcecheck_interval;		/* XXX */
140 const int int6intrq_present = 1;
141 
142 int ip6_ours_check_algorithm;
143 
144 
145 /* firewall hooks */
146 ip6_fw_chk_t *ip6_fw_chk_ptr;
147 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
148 int ip6_fw_enable = 1;
149 
150 struct ip6stat ip6stat;
151 
152 static void ip6_init2 __P((void *));
153 static struct mbuf *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
154 
155 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
156 #ifdef PULLDOWN_TEST
157 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
158 #endif
159 
160 
161 /*
162  * IP6 initialization: fill in IP6 protocol switch table.
163  * All protocols not implemented in kernel go to raw IP6 protocol handler.
164  */
165 void
166 ip6_init()
167 {
168 	struct ip6protosw *pr;
169 	int i;
170 	struct timeval tv;
171 
172 #ifdef DIAGNOSTIC
173 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
174 		panic("sizeof(protosw) != sizeof(ip6protosw)");
175 #endif
176 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
177 	if (pr == 0)
178 		panic("ip6_init");
179 	for (i = 0; i < IPPROTO_MAX; i++)
180 		ip6_protox[i] = pr - inet6sw;
181 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
182 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
183 		if (pr->pr_domain->dom_family == PF_INET6 &&
184 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
185 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
186 	ip6intrq.ifq_maxlen = ip6qmaxlen;
187 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", MTX_DEF);
188 	register_netisr(NETISR_IPV6, ip6intr);
189 	nd6_init();
190 	frag6_init();
191 	/*
192 	 * in many cases, random() here does NOT return random number
193 	 * as initialization during bootstrap time occur in fixed order.
194 	 */
195 	microtime(&tv);
196 	ip6_flow_seq = random() ^ tv.tv_usec;
197 	microtime(&tv);
198 	ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
199 }
200 
201 static void
202 ip6_init2(dummy)
203 	void *dummy;
204 {
205 
206 	/*
207 	 * to route local address of p2p link to loopback,
208 	 * assign loopback address first.
209 	 */
210 	in6_ifattach(&loif[0], NULL);
211 
212 	/* nd6_timer_init */
213 	callout_init(&nd6_timer_ch, 0);
214 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
215 
216 	/* router renumbering prefix list maintenance */
217 	callout_init(&in6_rr_timer_ch, 0);
218 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
219 
220 	/* timer for regeneranation of temporary addresses randomize ID */
221 	callout_reset(&in6_tmpaddrtimer_ch,
222 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
223 		       ip6_temp_regen_advance) * hz,
224 		      in6_tmpaddrtimer, NULL);
225 }
226 
227 /* cheat */
228 /* This must be after route_init(), which is now SI_ORDER_THIRD */
229 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
230 
231 /*
232  * IP6 input interrupt handling. Just pass the packet to ip6_input.
233  */
234 void
235 ip6intr()
236 {
237 	int s;
238 	struct mbuf *m;
239 
240 	for (;;) {
241 		s = splimp();
242 		IF_DEQUEUE(&ip6intrq, m);
243 		splx(s);
244 		if (m == 0)
245 			return;
246 		ip6_input(m);
247 	}
248 }
249 
250 extern struct	route_in6 ip6_forward_rt;
251 
252 void
253 ip6_input(m)
254 	struct mbuf *m;
255 {
256 	struct ip6_hdr *ip6;
257 	int off = sizeof(struct ip6_hdr), nest;
258 	u_int32_t plen;
259 	u_int32_t rtalert = ~0;
260 	int nxt, ours = 0;
261 	struct ifnet *deliverifp = NULL;
262 #ifdef  PFIL_HOOKS
263 	struct packet_filter_hook *pfh;
264 	struct mbuf *m0;
265 	int rv;
266 #endif  /* PFIL_HOOKS */
267 
268 #ifdef IPSEC
269 	/*
270 	 * should the inner packet be considered authentic?
271 	 * see comment in ah4_input().
272 	 */
273 	if (m) {
274 		m->m_flags &= ~M_AUTHIPHDR;
275 		m->m_flags &= ~M_AUTHIPDGM;
276 	}
277 #endif
278 
279 	/*
280 	 * make sure we don't have onion peering information into m_aux.
281 	 */
282 	ip6_delaux(m);
283 
284 	/*
285 	 * mbuf statistics by kazu
286 	 */
287 	if (m->m_flags & M_EXT) {
288 		if (m->m_next)
289 			ip6stat.ip6s_mext2m++;
290 		else
291 			ip6stat.ip6s_mext1++;
292 	} else {
293 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
294 		if (m->m_next) {
295 			if (m->m_flags & M_LOOP) {
296 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/*XXX*/
297 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
298 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
299 			else
300 				ip6stat.ip6s_m2m[0]++;
301 		} else
302 			ip6stat.ip6s_m1++;
303 #undef M2MMAX
304 	}
305 
306 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
307 	ip6stat.ip6s_total++;
308 
309 #ifndef PULLDOWN_TEST
310 	/*
311 	 * L2 bridge code and some other code can return mbuf chain
312 	 * that does not conform to KAME requirement.  too bad.
313 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
314 	 */
315 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
316 		struct mbuf *n;
317 
318 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
319 		if (n)
320 			M_COPY_PKTHDR(n, m);
321 		if (n && m->m_pkthdr.len > MHLEN) {
322 			MCLGET(n, M_DONTWAIT);
323 			if ((n->m_flags & M_EXT) == 0) {
324 				m_freem(n);
325 				n = NULL;
326 			}
327 		}
328 		if (!n) {
329 			m_freem(m);
330 			return;	/*ENOBUFS*/
331 		}
332 
333 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
334 		n->m_len = m->m_pkthdr.len;
335 		m_freem(m);
336 		m = n;
337 	}
338 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
339 #endif
340 
341 	if (m->m_len < sizeof(struct ip6_hdr)) {
342 		struct ifnet *inifp;
343 		inifp = m->m_pkthdr.rcvif;
344 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
345 			ip6stat.ip6s_toosmall++;
346 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
347 			return;
348 		}
349 	}
350 
351 	ip6 = mtod(m, struct ip6_hdr *);
352 
353 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
354 		ip6stat.ip6s_badvers++;
355 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
356 		goto bad;
357 	}
358 
359 #ifdef PFIL_HOOKS
360 	/*
361 	 * Run through list of hooks for input packets.  If there are any
362 	 * filters which require that additional packets in the flow are
363 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
364 	 * Note that filters must _never_ set this flag, as another filter
365 	 * in the list may have previously cleared it.
366 	 */
367 	m0 = m;
368 	pfh = pfil_hook_get(PFIL_IN, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
369 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
370 		if (pfh->pfil_func) {
371 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
372 					    m->m_pkthdr.rcvif, 0, &m0);
373 			if (rv)
374 				return;
375 			m = m0;
376 			if (m == NULL)
377 				return;
378 			ip6 = mtod(m, struct ip6_hdr *);
379 		}
380 #endif /* PFIL_HOOKS */
381 
382 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
383 
384 	/*
385 	 * Check with the firewall...
386 	 */
387 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
388 		u_short port = 0;
389 		/* If ipfw says divert, we have to just drop packet */
390 		/* use port as a dummy argument */
391 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
392 			m_freem(m);
393 			m = NULL;
394 		}
395 		if (!m)
396 			return;
397 	}
398 
399 	/*
400 	 * Check against address spoofing/corruption.
401 	 */
402 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
403 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
404 		/*
405 		 * XXX: "badscope" is not very suitable for a multicast source.
406 		 */
407 		ip6stat.ip6s_badscope++;
408 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
409 		goto bad;
410 	}
411 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
412 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
413 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
414 		ip6stat.ip6s_badscope++;
415 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
416 		goto bad;
417 	}
418 	/*
419 	 * The following check is not documented in specs.  A malicious
420 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
421 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
422 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
423 	 *
424 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
425 	 * support IPv4-less kernel compilation, we cannot support SIIT
426 	 * environment at all.  So, it makes more sense for us to reject any
427 	 * malicious packets for non-SIIT environment, than try to do a
428 	 * partical support for SIIT environment.
429 	 */
430 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
431 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
432 		ip6stat.ip6s_badscope++;
433 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
434 		goto bad;
435 	}
436 #if 0
437 	/*
438 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
439 	 *
440 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
441 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
442 	 * is revised to forbid relaying case.
443 	 */
444 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
445 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
446 		ip6stat.ip6s_badscope++;
447 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
448 		goto bad;
449 	}
450 #endif
451 
452 	/* drop packets if interface ID portion is already filled */
453 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
454 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
455 		    ip6->ip6_src.s6_addr16[1]) {
456 			ip6stat.ip6s_badscope++;
457 			goto bad;
458 		}
459 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
460 		    ip6->ip6_dst.s6_addr16[1]) {
461 			ip6stat.ip6s_badscope++;
462 			goto bad;
463 		}
464 	}
465 
466 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
467 		ip6->ip6_src.s6_addr16[1]
468 			= htons(m->m_pkthdr.rcvif->if_index);
469 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
470 		ip6->ip6_dst.s6_addr16[1]
471 			= htons(m->m_pkthdr.rcvif->if_index);
472 
473 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
474 	/*
475 	 * We use rt->rt_ifp to determine if the address is ours or not.
476 	 * If rt_ifp is lo0, the address is ours.
477 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
478 	 * so any address under fe80::%lo0/64 will be mistakenly considered
479 	 * local.  The special case is supplied to handle the case properly
480 	 * by actually looking at interface addresses
481 	 * (using in6ifa_ifpwithaddr).
482 	 */
483 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
484 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
485 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
486 			icmp6_error(m, ICMP6_DST_UNREACH,
487 			    ICMP6_DST_UNREACH_ADDR, 0);
488 			/* m is already freed */
489 			return;
490 		}
491 
492 		ours = 1;
493 		deliverifp = m->m_pkthdr.rcvif;
494 		goto hbhcheck;
495 	}
496 #endif
497 
498 	/*
499 	 * Multicast check
500 	 */
501 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
502 	  	struct	in6_multi *in6m = 0;
503 
504 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
505 		/*
506 		 * See if we belong to the destination multicast group on the
507 		 * arrival interface.
508 		 */
509 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
510 		if (in6m)
511 			ours = 1;
512 		else if (!ip6_mrouter) {
513 			ip6stat.ip6s_notmember++;
514 			ip6stat.ip6s_cantforward++;
515 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
516 			goto bad;
517 		}
518 		deliverifp = m->m_pkthdr.rcvif;
519 		goto hbhcheck;
520 	}
521 
522 	/*
523 	 *  Unicast check
524 	 */
525 	switch (ip6_ours_check_algorithm) {
526 	default:
527 		/*
528 		 * XXX: I intentionally broke our indentation rule here,
529 		 *      since this switch-case is just for measurement and
530 		 *      therefore should soon be removed.
531 		 */
532 	if (ip6_forward_rt.ro_rt != NULL &&
533 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
534 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
535 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
536 		ip6stat.ip6s_forward_cachehit++;
537 	else {
538 		struct sockaddr_in6 *dst6;
539 
540 		if (ip6_forward_rt.ro_rt) {
541 			/* route is down or destination is different */
542 			ip6stat.ip6s_forward_cachemiss++;
543 			RTFREE(ip6_forward_rt.ro_rt);
544 			ip6_forward_rt.ro_rt = 0;
545 		}
546 
547 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
548 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
549 		dst6->sin6_len = sizeof(struct sockaddr_in6);
550 		dst6->sin6_family = AF_INET6;
551 		dst6->sin6_addr = ip6->ip6_dst;
552 #ifdef SCOPEDROUTING
553 		ip6_forward_rt.ro_dst.sin6_scope_id =
554 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
555 #endif
556 
557 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
558 	}
559 
560 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
561 
562 	/*
563 	 * Accept the packet if the forwarding interface to the destination
564 	 * according to the routing table is the loopback interface,
565 	 * unless the associated route has a gateway.
566 	 * Note that this approach causes to accept a packet if there is a
567 	 * route to the loopback interface for the destination of the packet.
568 	 * But we think it's even useful in some situations, e.g. when using
569 	 * a special daemon which wants to intercept the packet.
570 	 *
571 	 * XXX: some OSes automatically make a cloned route for the destination
572 	 * of an outgoing packet.  If the outgoing interface of the packet
573 	 * is a loopback one, the kernel would consider the packet to be
574 	 * accepted, even if we have no such address assinged on the interface.
575 	 * We check the cloned flag of the route entry to reject such cases,
576 	 * assuming that route entries for our own addresses are not made by
577 	 * cloning (it should be true because in6_addloop explicitly installs
578 	 * the host route).  However, we might have to do an explicit check
579 	 * while it would be less efficient.  Or, should we rather install a
580 	 * reject route for such a case?
581 	 */
582 	if (ip6_forward_rt.ro_rt &&
583 	    (ip6_forward_rt.ro_rt->rt_flags &
584 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
585 #ifdef RTF_WASCLONED
586 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
587 #endif
588 #ifdef RTF_CLONED
589 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
590 #endif
591 #if 0
592 	    /*
593 	     * The check below is redundant since the comparison of
594 	     * the destination and the key of the rtentry has
595 	     * already done through looking up the routing table.
596 	     */
597 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
598 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
599 #endif
600 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
601 		struct in6_ifaddr *ia6 =
602 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
603 
604 		/*
605 		 * record address information into m_aux.
606 		 */
607 		(void)ip6_setdstifaddr(m, ia6);
608 
609 		/*
610 		 * packets to a tentative, duplicated, or somehow invalid
611 		 * address must not be accepted.
612 		 */
613 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
614 			/* this address is ready */
615 			ours = 1;
616 			deliverifp = ia6->ia_ifp;	/* correct? */
617 			/* Count the packet in the ip address stats */
618 			ia6->ia_ifa.if_ipackets++;
619 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
620 			goto hbhcheck;
621 		} else {
622 			/* address is not ready, so discard the packet. */
623 			nd6log((LOG_INFO,
624 			    "ip6_input: packet to an unready address %s->%s\n",
625 			    ip6_sprintf(&ip6->ip6_src),
626 			    ip6_sprintf(&ip6->ip6_dst)));
627 
628 			goto bad;
629 		}
630 	}
631 	} /* XXX indentation (see above) */
632 
633 	/*
634 	 * FAITH(Firewall Aided Internet Translator)
635 	 */
636 #if defined(NFAITH) && 0 < NFAITH
637 	if (ip6_keepfaith) {
638 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
639 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
640 			/* XXX do we need more sanity checks? */
641 			ours = 1;
642 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
643 			goto hbhcheck;
644 		}
645 	}
646 #endif
647 
648 	/*
649 	 * Now there is no reason to process the packet if it's not our own
650 	 * and we're not a router.
651 	 */
652 	if (!ip6_forwarding) {
653 		ip6stat.ip6s_cantforward++;
654 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
655 		goto bad;
656 	}
657 
658   hbhcheck:
659 	/*
660 	 * record address information into m_aux, if we don't have one yet.
661 	 * note that we are unable to record it, if the address is not listed
662 	 * as our interface address (e.g. multicast addresses, addresses
663 	 * within FAITH prefixes and such).
664 	 */
665 	if (deliverifp && !ip6_getdstifaddr(m)) {
666 		struct in6_ifaddr *ia6;
667 
668 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
669 		if (ia6) {
670 			if (!ip6_setdstifaddr(m, ia6)) {
671 				/*
672 				 * XXX maybe we should drop the packet here,
673 				 * as we could not provide enough information
674 				 * to the upper layers.
675 				 */
676 			}
677 		}
678 	}
679 
680 	/*
681 	 * Process Hop-by-Hop options header if it's contained.
682 	 * m may be modified in ip6_hopopts_input().
683 	 * If a JumboPayload option is included, plen will also be modified.
684 	 */
685 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
686 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
687 		struct ip6_hbh *hbh;
688 
689 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
690 #if 0	/*touches NULL pointer*/
691 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
692 #endif
693 			return;	/* m have already been freed */
694 		}
695 
696 		/* adjust pointer */
697 		ip6 = mtod(m, struct ip6_hdr *);
698 
699 		/*
700 		 * if the payload length field is 0 and the next header field
701 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
702 		 * option MUST be included.
703 		 */
704 		if (ip6->ip6_plen == 0 && plen == 0) {
705 			/*
706 			 * Note that if a valid jumbo payload option is
707 			 * contained, ip6_hoptops_input() must set a valid
708 			 * (non-zero) payload length to the variable plen.
709 			 */
710 			ip6stat.ip6s_badoptions++;
711 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
712 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
713 			icmp6_error(m, ICMP6_PARAM_PROB,
714 				    ICMP6_PARAMPROB_HEADER,
715 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
716 			return;
717 		}
718 #ifndef PULLDOWN_TEST
719 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
720 		hbh = (struct ip6_hbh *)(ip6 + 1);
721 #else
722 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
723 			sizeof(struct ip6_hbh));
724 		if (hbh == NULL) {
725 			ip6stat.ip6s_tooshort++;
726 			return;
727 		}
728 #endif
729 		nxt = hbh->ip6h_nxt;
730 
731 		/*
732 		 * accept the packet if a router alert option is included
733 		 * and we act as an IPv6 router.
734 		 */
735 		if (rtalert != ~0 && ip6_forwarding)
736 			ours = 1;
737 	} else
738 		nxt = ip6->ip6_nxt;
739 
740 	/*
741 	 * Check that the amount of data in the buffers
742 	 * is as at least much as the IPv6 header would have us expect.
743 	 * Trim mbufs if longer than we expect.
744 	 * Drop packet if shorter than we expect.
745 	 */
746 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
747 		ip6stat.ip6s_tooshort++;
748 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
749 		goto bad;
750 	}
751 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
752 		if (m->m_len == m->m_pkthdr.len) {
753 			m->m_len = sizeof(struct ip6_hdr) + plen;
754 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
755 		} else
756 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
757 	}
758 
759 	/*
760 	 * Forward if desirable.
761 	 */
762 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
763 		/*
764 		 * If we are acting as a multicast router, all
765 		 * incoming multicast packets are passed to the
766 		 * kernel-level multicast forwarding function.
767 		 * The packet is returned (relatively) intact; if
768 		 * ip6_mforward() returns a non-zero value, the packet
769 		 * must be discarded, else it may be accepted below.
770 		 */
771 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
772 			ip6stat.ip6s_cantforward++;
773 			m_freem(m);
774 			return;
775 		}
776 		if (!ours) {
777 			m_freem(m);
778 			return;
779 		}
780 	} else if (!ours) {
781 		ip6_forward(m, 0);
782 		return;
783 	}
784 
785 	ip6 = mtod(m, struct ip6_hdr *);
786 
787 	/*
788 	 * Malicious party may be able to use IPv4 mapped addr to confuse
789 	 * tcp/udp stack and bypass security checks (act as if it was from
790 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
791 	 *
792 	 * For SIIT end node behavior, you may want to disable the check.
793 	 * However, you will  become vulnerable to attacks using IPv4 mapped
794 	 * source.
795 	 */
796 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
797 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
798 		ip6stat.ip6s_badscope++;
799 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
800 		goto bad;
801 	}
802 
803 	/*
804 	 * Tell launch routine the next header
805 	 */
806 	ip6stat.ip6s_delivered++;
807 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
808 	nest = 0;
809 
810 	while (nxt != IPPROTO_DONE) {
811 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
812 			ip6stat.ip6s_toomanyhdr++;
813 			goto bad;
814 		}
815 
816 		/*
817 		 * protection against faulty packet - there should be
818 		 * more sanity checks in header chain processing.
819 		 */
820 		if (m->m_pkthdr.len < off) {
821 			ip6stat.ip6s_tooshort++;
822 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
823 			goto bad;
824 		}
825 
826 #if 0
827 		/*
828 		 * do we need to do it for every header?  yeah, other
829 		 * functions can play with it (like re-allocate and copy).
830 		 */
831 		mhist = ip6_addaux(m);
832 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
833 			hist = mtod(mhist, caddr_t) + mhist->m_len;
834 			bcopy(&nxt, hist, sizeof(nxt));
835 			mhist->m_len += sizeof(nxt);
836 		} else {
837 			ip6stat.ip6s_toomanyhdr++;
838 			goto bad;
839 		}
840 #endif
841 
842 #ifdef IPSEC
843 		/*
844 		 * enforce IPsec policy checking if we are seeing last header.
845 		 * note that we do not visit this with protocols with pcb layer
846 		 * code - like udp/tcp/raw ip.
847 		 */
848 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
849 		    ipsec6_in_reject(m, NULL)) {
850 			ipsec6stat.in_polvio++;
851 			goto bad;
852 		}
853 #endif
854 
855 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
856 	}
857 	return;
858  bad:
859 	m_freem(m);
860 }
861 
862 /*
863  * set/grab in6_ifaddr correspond to IPv6 destination address.
864  * XXX backward compatibility wrapper
865  */
866 static struct mbuf *
867 ip6_setdstifaddr(m, ia6)
868 	struct mbuf *m;
869 	struct in6_ifaddr *ia6;
870 {
871 	struct mbuf *n;
872 
873 	n = ip6_addaux(m);
874 	if (n)
875 		mtod(n, struct ip6aux *)->ip6a_dstia6 = ia6;
876 	return n;	/* NULL if failed to set */
877 }
878 
879 struct in6_ifaddr *
880 ip6_getdstifaddr(m)
881 	struct mbuf *m;
882 {
883 	struct mbuf *n;
884 
885 	n = ip6_findaux(m);
886 	if (n)
887 		return mtod(n, struct ip6aux *)->ip6a_dstia6;
888 	else
889 		return NULL;
890 }
891 
892 /*
893  * Hop-by-Hop options header processing. If a valid jumbo payload option is
894  * included, the real payload length will be stored in plenp.
895  */
896 static int
897 ip6_hopopts_input(plenp, rtalertp, mp, offp)
898 	u_int32_t *plenp;
899 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
900 	struct mbuf **mp;
901 	int *offp;
902 {
903 	struct mbuf *m = *mp;
904 	int off = *offp, hbhlen;
905 	struct ip6_hbh *hbh;
906 	u_int8_t *opt;
907 
908 	/* validation of the length of the header */
909 #ifndef PULLDOWN_TEST
910 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
911 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
912 	hbhlen = (hbh->ip6h_len + 1) << 3;
913 
914 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
915 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
916 #else
917 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
918 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
919 	if (hbh == NULL) {
920 		ip6stat.ip6s_tooshort++;
921 		return -1;
922 	}
923 	hbhlen = (hbh->ip6h_len + 1) << 3;
924 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
925 		hbhlen);
926 	if (hbh == NULL) {
927 		ip6stat.ip6s_tooshort++;
928 		return -1;
929 	}
930 #endif
931 	off += hbhlen;
932 	hbhlen -= sizeof(struct ip6_hbh);
933 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
934 
935 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
936 				hbhlen, rtalertp, plenp) < 0)
937 		return(-1);
938 
939 	*offp = off;
940 	*mp = m;
941 	return(0);
942 }
943 
944 /*
945  * Search header for all Hop-by-hop options and process each option.
946  * This function is separate from ip6_hopopts_input() in order to
947  * handle a case where the sending node itself process its hop-by-hop
948  * options header. In such a case, the function is called from ip6_output().
949  *
950  * The function assumes that hbh header is located right after the IPv6 header
951  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
952  * opthead + hbhlen is located in continuous memory region.
953  */
954 int
955 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
956 	struct mbuf *m;
957 	u_int8_t *opthead;
958 	int hbhlen;
959 	u_int32_t *rtalertp;
960 	u_int32_t *plenp;
961 {
962 	struct ip6_hdr *ip6;
963 	int optlen = 0;
964 	u_int8_t *opt = opthead;
965 	u_int16_t rtalert_val;
966 	u_int32_t jumboplen;
967 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
968 
969 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
970 		switch (*opt) {
971 		case IP6OPT_PAD1:
972 			optlen = 1;
973 			break;
974 		case IP6OPT_PADN:
975 			if (hbhlen < IP6OPT_MINLEN) {
976 				ip6stat.ip6s_toosmall++;
977 				goto bad;
978 			}
979 			optlen = *(opt + 1) + 2;
980 			break;
981 		case IP6OPT_RTALERT:
982 			/* XXX may need check for alignment */
983 			if (hbhlen < IP6OPT_RTALERT_LEN) {
984 				ip6stat.ip6s_toosmall++;
985 				goto bad;
986 			}
987 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
988 				/* XXX stat */
989 				icmp6_error(m, ICMP6_PARAM_PROB,
990 					    ICMP6_PARAMPROB_HEADER,
991 					    erroff + opt + 1 - opthead);
992 				return(-1);
993 			}
994 			optlen = IP6OPT_RTALERT_LEN;
995 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
996 			*rtalertp = ntohs(rtalert_val);
997 			break;
998 		case IP6OPT_JUMBO:
999 			/* XXX may need check for alignment */
1000 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1001 				ip6stat.ip6s_toosmall++;
1002 				goto bad;
1003 			}
1004 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1005 				/* XXX stat */
1006 				icmp6_error(m, ICMP6_PARAM_PROB,
1007 					    ICMP6_PARAMPROB_HEADER,
1008 					    erroff + opt + 1 - opthead);
1009 				return(-1);
1010 			}
1011 			optlen = IP6OPT_JUMBO_LEN;
1012 
1013 			/*
1014 			 * IPv6 packets that have non 0 payload length
1015 			 * must not contain a jumbo payload option.
1016 			 */
1017 			ip6 = mtod(m, struct ip6_hdr *);
1018 			if (ip6->ip6_plen) {
1019 				ip6stat.ip6s_badoptions++;
1020 				icmp6_error(m, ICMP6_PARAM_PROB,
1021 					    ICMP6_PARAMPROB_HEADER,
1022 					    erroff + opt - opthead);
1023 				return(-1);
1024 			}
1025 
1026 			/*
1027 			 * We may see jumbolen in unaligned location, so
1028 			 * we'd need to perform bcopy().
1029 			 */
1030 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1031 			jumboplen = (u_int32_t)htonl(jumboplen);
1032 
1033 #if 1
1034 			/*
1035 			 * if there are multiple jumbo payload options,
1036 			 * *plenp will be non-zero and the packet will be
1037 			 * rejected.
1038 			 * the behavior may need some debate in ipngwg -
1039 			 * multiple options does not make sense, however,
1040 			 * there's no explicit mention in specification.
1041 			 */
1042 			if (*plenp != 0) {
1043 				ip6stat.ip6s_badoptions++;
1044 				icmp6_error(m, ICMP6_PARAM_PROB,
1045 					    ICMP6_PARAMPROB_HEADER,
1046 					    erroff + opt + 2 - opthead);
1047 				return(-1);
1048 			}
1049 #endif
1050 
1051 			/*
1052 			 * jumbo payload length must be larger than 65535.
1053 			 */
1054 			if (jumboplen <= IPV6_MAXPACKET) {
1055 				ip6stat.ip6s_badoptions++;
1056 				icmp6_error(m, ICMP6_PARAM_PROB,
1057 					    ICMP6_PARAMPROB_HEADER,
1058 					    erroff + opt + 2 - opthead);
1059 				return(-1);
1060 			}
1061 			*plenp = jumboplen;
1062 
1063 			break;
1064 		default:		/* unknown option */
1065 			if (hbhlen < IP6OPT_MINLEN) {
1066 				ip6stat.ip6s_toosmall++;
1067 				goto bad;
1068 			}
1069 			optlen = ip6_unknown_opt(opt, m,
1070 			    erroff + opt - opthead);
1071 			if (optlen == -1)
1072 				return(-1);
1073 			optlen += 2;
1074 			break;
1075 		}
1076 	}
1077 
1078 	return(0);
1079 
1080   bad:
1081 	m_freem(m);
1082 	return(-1);
1083 }
1084 
1085 /*
1086  * Unknown option processing.
1087  * The third argument `off' is the offset from the IPv6 header to the option,
1088  * which is necessary if the IPv6 header the and option header and IPv6 header
1089  * is not continuous in order to return an ICMPv6 error.
1090  */
1091 int
1092 ip6_unknown_opt(optp, m, off)
1093 	u_int8_t *optp;
1094 	struct mbuf *m;
1095 	int off;
1096 {
1097 	struct ip6_hdr *ip6;
1098 
1099 	switch (IP6OPT_TYPE(*optp)) {
1100 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1101 		return((int)*(optp + 1));
1102 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1103 		m_freem(m);
1104 		return(-1);
1105 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1106 		ip6stat.ip6s_badoptions++;
1107 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1108 		return(-1);
1109 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1110 		ip6stat.ip6s_badoptions++;
1111 		ip6 = mtod(m, struct ip6_hdr *);
1112 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1113 		    (m->m_flags & (M_BCAST|M_MCAST)))
1114 			m_freem(m);
1115 		else
1116 			icmp6_error(m, ICMP6_PARAM_PROB,
1117 				    ICMP6_PARAMPROB_OPTION, off);
1118 		return(-1);
1119 	}
1120 
1121 	m_freem(m);		/* XXX: NOTREACHED */
1122 	return(-1);
1123 }
1124 
1125 /*
1126  * Create the "control" list for this pcb.
1127  * The function will not modify mbuf chain at all.
1128  *
1129  * with KAME mbuf chain restriction:
1130  * The routine will be called from upper layer handlers like tcp6_input().
1131  * Thus the routine assumes that the caller (tcp6_input) have already
1132  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1133  * very first mbuf on the mbuf chain.
1134  */
1135 void
1136 ip6_savecontrol(in6p, mp, ip6, m)
1137 	struct inpcb *in6p;
1138 	struct mbuf **mp;
1139 	struct ip6_hdr *ip6;
1140 	struct mbuf *m;
1141 {
1142 	struct proc *p = curproc;	/* XXX */
1143 	int privileged = 0;
1144 	int rthdr_exist = 0;
1145 
1146 
1147 	if (p && !suser(p))
1148  		privileged++;
1149 
1150 #ifdef SO_TIMESTAMP
1151 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1152 		struct timeval tv;
1153 
1154 		microtime(&tv);
1155 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1156 				      SCM_TIMESTAMP, SOL_SOCKET);
1157 		if (*mp) {
1158 			mp = &(*mp)->m_next;
1159 		}
1160 	}
1161 #endif
1162 
1163 	/* RFC 2292 sec. 5 */
1164 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1165 		struct in6_pktinfo pi6;
1166 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1167 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1168 			pi6.ipi6_addr.s6_addr16[1] = 0;
1169 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1170 					? m->m_pkthdr.rcvif->if_index
1171 					: 0;
1172 		*mp = sbcreatecontrol((caddr_t) &pi6,
1173 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1174 			IPPROTO_IPV6);
1175 		if (*mp)
1176 			mp = &(*mp)->m_next;
1177 	}
1178 
1179 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1180 		int hlim = ip6->ip6_hlim & 0xff;
1181 		*mp = sbcreatecontrol((caddr_t) &hlim,
1182 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1183 		if (*mp)
1184 			mp = &(*mp)->m_next;
1185 	}
1186 
1187 	/*
1188 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1189 	 * for the option, but it might be too strict, since there might
1190 	 * be some hop-by-hop options which can be returned to normal user.
1191 	 * See RFC 2292 section 6.
1192 	 */
1193 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1194 		/*
1195 		 * Check if a hop-by-hop options header is contatined in the
1196 		 * received packet, and if so, store the options as ancillary
1197 		 * data. Note that a hop-by-hop options header must be
1198 		 * just after the IPv6 header, which fact is assured through
1199 		 * the IPv6 input processing.
1200 		 */
1201 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1202 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1203 			struct ip6_hbh *hbh;
1204 			int hbhlen = 0;
1205 #ifdef PULLDOWN_TEST
1206 			struct mbuf *ext;
1207 #endif
1208 
1209 #ifndef PULLDOWN_TEST
1210 			hbh = (struct ip6_hbh *)(ip6 + 1);
1211 			hbhlen = (hbh->ip6h_len + 1) << 3;
1212 #else
1213 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1214 			    ip6->ip6_nxt);
1215 			if (ext == NULL) {
1216 				ip6stat.ip6s_tooshort++;
1217 				return;
1218 			}
1219 			hbh = mtod(ext, struct ip6_hbh *);
1220 			hbhlen = (hbh->ip6h_len + 1) << 3;
1221 			if (hbhlen != ext->m_len) {
1222 				m_freem(ext);
1223 				ip6stat.ip6s_tooshort++;
1224 				return;
1225 			}
1226 #endif
1227 
1228 			/*
1229 			 * XXX: We copy whole the header even if a jumbo
1230 			 * payload option is included, which option is to
1231 			 * be removed before returning in the RFC 2292.
1232 			 * Note: this constraint is removed in 2292bis.
1233 			 */
1234 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1235 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1236 			if (*mp)
1237 				mp = &(*mp)->m_next;
1238 #ifdef PULLDOWN_TEST
1239 			m_freem(ext);
1240 #endif
1241 		}
1242 	}
1243 
1244 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1245 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1246 		int proto, off, nxt;
1247 
1248 		/*
1249 		 * go through the header chain to see if a routing header is
1250 		 * contained in the packet. We need this information to store
1251 		 * destination options headers (if any) properly.
1252 		 * XXX: performance issue. We should record this info when
1253 		 * processing extension headers in incoming routine.
1254 		 * (todo) use m_aux?
1255 		 */
1256 		proto = IPPROTO_IPV6;
1257 		off = 0;
1258 		nxt = -1;
1259 		while (1) {
1260 			int newoff;
1261 
1262 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1263 			if (newoff < 0)
1264 				break;
1265 			if (newoff < off) /* invalid, check for safety */
1266 				break;
1267 			if ((proto = nxt) == IPPROTO_ROUTING) {
1268 				rthdr_exist = 1;
1269 				break;
1270 			}
1271 			off = newoff;
1272 		}
1273 	}
1274 
1275 	if ((in6p->in6p_flags &
1276 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1277 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1278 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1279 
1280 		/*
1281 		 * Search for destination options headers or routing
1282 		 * header(s) through the header chain, and stores each
1283 		 * header as ancillary data.
1284 		 * Note that the order of the headers remains in
1285 		 * the chain of ancillary data.
1286 		 */
1287 		while (1) {	/* is explicit loop prevention necessary? */
1288 			struct ip6_ext *ip6e = NULL;
1289 			int elen;
1290 #ifdef PULLDOWN_TEST
1291 			struct mbuf *ext = NULL;
1292 #endif
1293 
1294 			/*
1295 			 * if it is not an extension header, don't try to
1296 			 * pull it from the chain.
1297 			 */
1298 			switch (nxt) {
1299 			case IPPROTO_DSTOPTS:
1300 			case IPPROTO_ROUTING:
1301 			case IPPROTO_HOPOPTS:
1302 			case IPPROTO_AH: /* is it possible? */
1303 				break;
1304 			default:
1305 				goto loopend;
1306 			}
1307 
1308 #ifndef PULLDOWN_TEST
1309 			if (off + sizeof(*ip6e) > m->m_len)
1310 				goto loopend;
1311 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1312 			if (nxt == IPPROTO_AH)
1313 				elen = (ip6e->ip6e_len + 2) << 2;
1314 			else
1315 				elen = (ip6e->ip6e_len + 1) << 3;
1316 			if (off + elen > m->m_len)
1317 				goto loopend;
1318 #else
1319 			ext = ip6_pullexthdr(m, off, nxt);
1320 			if (ext == NULL) {
1321 				ip6stat.ip6s_tooshort++;
1322 				return;
1323 			}
1324 			ip6e = mtod(ext, struct ip6_ext *);
1325 			if (nxt == IPPROTO_AH)
1326 				elen = (ip6e->ip6e_len + 2) << 2;
1327 			else
1328 				elen = (ip6e->ip6e_len + 1) << 3;
1329 			if (elen != ext->m_len) {
1330 				m_freem(ext);
1331 				ip6stat.ip6s_tooshort++;
1332 				return;
1333 			}
1334 #endif
1335 
1336 			switch (nxt) {
1337 			case IPPROTO_DSTOPTS:
1338 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1339 					break;
1340 
1341 				/*
1342 				 * We also require super-user privilege for
1343 				 * the option.
1344 				 * See the comments on IN6_HOPOPTS.
1345 				 */
1346 				if (!privileged)
1347 					break;
1348 
1349 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1350 						      IPV6_DSTOPTS,
1351 						      IPPROTO_IPV6);
1352 				if (*mp)
1353 					mp = &(*mp)->m_next;
1354 				break;
1355 			case IPPROTO_ROUTING:
1356 				if (!in6p->in6p_flags & IN6P_RTHDR)
1357 					break;
1358 
1359 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1360 						      IPV6_RTHDR,
1361 						      IPPROTO_IPV6);
1362 				if (*mp)
1363 					mp = &(*mp)->m_next;
1364 				break;
1365 			case IPPROTO_HOPOPTS:
1366 			case IPPROTO_AH: /* is it possible? */
1367 				break;
1368 
1369 			default:
1370 				/*
1371 			 	 * other cases have been filtered in the above.
1372 				 * none will visit this case.  here we supply
1373 				 * the code just in case (nxt overwritten or
1374 				 * other cases).
1375 				 */
1376 #ifdef PULLDOWN_TEST
1377 				m_freem(ext);
1378 #endif
1379 				goto loopend;
1380 
1381 			}
1382 
1383 			/* proceed with the next header. */
1384 			off += elen;
1385 			nxt = ip6e->ip6e_nxt;
1386 			ip6e = NULL;
1387 #ifdef PULLDOWN_TEST
1388 			m_freem(ext);
1389 			ext = NULL;
1390 #endif
1391 		}
1392 	  loopend:
1393 		;
1394 	}
1395 
1396 }
1397 
1398 #ifdef PULLDOWN_TEST
1399 /*
1400  * pull single extension header from mbuf chain.  returns single mbuf that
1401  * contains the result, or NULL on error.
1402  */
1403 static struct mbuf *
1404 ip6_pullexthdr(m, off, nxt)
1405 	struct mbuf *m;
1406 	size_t off;
1407 	int nxt;
1408 {
1409 	struct ip6_ext ip6e;
1410 	size_t elen;
1411 	struct mbuf *n;
1412 
1413 #ifdef DIAGNOSTIC
1414 	switch (nxt) {
1415 	case IPPROTO_DSTOPTS:
1416 	case IPPROTO_ROUTING:
1417 	case IPPROTO_HOPOPTS:
1418 	case IPPROTO_AH: /* is it possible? */
1419 		break;
1420 	default:
1421 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1422 	}
1423 #endif
1424 
1425 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1426 	if (nxt == IPPROTO_AH)
1427 		elen = (ip6e.ip6e_len + 2) << 2;
1428 	else
1429 		elen = (ip6e.ip6e_len + 1) << 3;
1430 
1431 	MGET(n, M_DONTWAIT, MT_DATA);
1432 	if (n && elen >= MLEN) {
1433 		MCLGET(n, M_DONTWAIT);
1434 		if ((n->m_flags & M_EXT) == 0) {
1435 			m_free(n);
1436 			n = NULL;
1437 		}
1438 	}
1439 	if (!n)
1440 		return NULL;
1441 
1442 	n->m_len = 0;
1443 	if (elen >= M_TRAILINGSPACE(n)) {
1444 		m_free(n);
1445 		return NULL;
1446 	}
1447 
1448 	m_copydata(m, off, elen, mtod(n, caddr_t));
1449 	n->m_len = elen;
1450 	return n;
1451 }
1452 #endif
1453 
1454 /*
1455  * Get pointer to the previous header followed by the header
1456  * currently processed.
1457  * XXX: This function supposes that
1458  *	M includes all headers,
1459  *	the next header field and the header length field of each header
1460  *	are valid, and
1461  *	the sum of each header length equals to OFF.
1462  * Because of these assumptions, this function must be called very
1463  * carefully. Moreover, it will not be used in the near future when
1464  * we develop `neater' mechanism to process extension headers.
1465  */
1466 char *
1467 ip6_get_prevhdr(m, off)
1468 	struct mbuf *m;
1469 	int off;
1470 {
1471 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1472 
1473 	if (off == sizeof(struct ip6_hdr))
1474 		return(&ip6->ip6_nxt);
1475 	else {
1476 		int len, nxt;
1477 		struct ip6_ext *ip6e = NULL;
1478 
1479 		nxt = ip6->ip6_nxt;
1480 		len = sizeof(struct ip6_hdr);
1481 		while (len < off) {
1482 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1483 
1484 			switch (nxt) {
1485 			case IPPROTO_FRAGMENT:
1486 				len += sizeof(struct ip6_frag);
1487 				break;
1488 			case IPPROTO_AH:
1489 				len += (ip6e->ip6e_len + 2) << 2;
1490 				break;
1491 			default:
1492 				len += (ip6e->ip6e_len + 1) << 3;
1493 				break;
1494 			}
1495 			nxt = ip6e->ip6e_nxt;
1496 		}
1497 		if (ip6e)
1498 			return(&ip6e->ip6e_nxt);
1499 		else
1500 			return NULL;
1501 	}
1502 }
1503 
1504 /*
1505  * get next header offset.  m will be retained.
1506  */
1507 int
1508 ip6_nexthdr(m, off, proto, nxtp)
1509 	struct mbuf *m;
1510 	int off;
1511 	int proto;
1512 	int *nxtp;
1513 {
1514 	struct ip6_hdr ip6;
1515 	struct ip6_ext ip6e;
1516 	struct ip6_frag fh;
1517 
1518 	/* just in case */
1519 	if (m == NULL)
1520 		panic("ip6_nexthdr: m == NULL");
1521 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1522 		return -1;
1523 
1524 	switch (proto) {
1525 	case IPPROTO_IPV6:
1526 		if (m->m_pkthdr.len < off + sizeof(ip6))
1527 			return -1;
1528 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1529 		if (nxtp)
1530 			*nxtp = ip6.ip6_nxt;
1531 		off += sizeof(ip6);
1532 		return off;
1533 
1534 	case IPPROTO_FRAGMENT:
1535 		/*
1536 		 * terminate parsing if it is not the first fragment,
1537 		 * it does not make sense to parse through it.
1538 		 */
1539 		if (m->m_pkthdr.len < off + sizeof(fh))
1540 			return -1;
1541 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1542 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1543 			return -1;
1544 		if (nxtp)
1545 			*nxtp = fh.ip6f_nxt;
1546 		off += sizeof(struct ip6_frag);
1547 		return off;
1548 
1549 	case IPPROTO_AH:
1550 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1551 			return -1;
1552 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1553 		if (nxtp)
1554 			*nxtp = ip6e.ip6e_nxt;
1555 		off += (ip6e.ip6e_len + 2) << 2;
1556 		return off;
1557 
1558 	case IPPROTO_HOPOPTS:
1559 	case IPPROTO_ROUTING:
1560 	case IPPROTO_DSTOPTS:
1561 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1562 			return -1;
1563 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1564 		if (nxtp)
1565 			*nxtp = ip6e.ip6e_nxt;
1566 		off += (ip6e.ip6e_len + 1) << 3;
1567 		return off;
1568 
1569 	case IPPROTO_NONE:
1570 	case IPPROTO_ESP:
1571 	case IPPROTO_IPCOMP:
1572 		/* give up */
1573 		return -1;
1574 
1575 	default:
1576 		return -1;
1577 	}
1578 
1579 	return -1;
1580 }
1581 
1582 /*
1583  * get offset for the last header in the chain.  m will be kept untainted.
1584  */
1585 int
1586 ip6_lasthdr(m, off, proto, nxtp)
1587 	struct mbuf *m;
1588 	int off;
1589 	int proto;
1590 	int *nxtp;
1591 {
1592 	int newoff;
1593 	int nxt;
1594 
1595 	if (!nxtp) {
1596 		nxt = -1;
1597 		nxtp = &nxt;
1598 	}
1599 	while (1) {
1600 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1601 		if (newoff < 0)
1602 			return off;
1603 		else if (newoff < off)
1604 			return -1;	/* invalid */
1605 		else if (newoff == off)
1606 			return newoff;
1607 
1608 		off = newoff;
1609 		proto = *nxtp;
1610 	}
1611 }
1612 
1613 struct mbuf *
1614 ip6_addaux(m)
1615 	struct mbuf *m;
1616 {
1617 	struct mbuf *n;
1618 
1619 #ifdef DIAGNOSTIC
1620 	if (sizeof(struct ip6aux) > MHLEN)
1621 		panic("assumption failed on sizeof(ip6aux)");
1622 #endif
1623 	n = m_aux_find(m, AF_INET6, -1);
1624 	if (n) {
1625 		if (n->m_len < sizeof(struct ip6aux)) {
1626 			printf("conflicting use of ip6aux");
1627 			return NULL;
1628 		}
1629 	} else {
1630 		n = m_aux_add(m, AF_INET6, -1);
1631 		n->m_len = sizeof(struct ip6aux);
1632 		bzero(mtod(n, caddr_t), n->m_len);
1633 	}
1634 	return n;
1635 }
1636 
1637 struct mbuf *
1638 ip6_findaux(m)
1639 	struct mbuf *m;
1640 {
1641 	struct mbuf *n;
1642 
1643 	n = m_aux_find(m, AF_INET6, -1);
1644 	if (n && n->m_len < sizeof(struct ip6aux)) {
1645 		printf("conflicting use of ip6aux");
1646 		n = NULL;
1647 	}
1648 	return n;
1649 }
1650 
1651 void
1652 ip6_delaux(m)
1653 	struct mbuf *m;
1654 {
1655 	struct mbuf *n;
1656 
1657 	n = m_aux_find(m, AF_INET6, -1);
1658 	if (n)
1659 		m_aux_delete(m, n);
1660 }
1661 
1662 /*
1663  * System control for IP6
1664  */
1665 
1666 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1667 	0,		0,		0,		0,
1668 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1669 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1670 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1671 	0,		0,		0,		0,
1672 	ENOPROTOOPT
1673 };
1674