xref: /freebsd/sys/netinet6/ip6_input.c (revision 5521ff5a4d1929056e7ffc982fac3341ca54df7c)
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 
125 #include <net/net_osdep.h>
126 
127 extern struct domain inet6domain;
128 extern struct ip6protosw inet6sw[];
129 
130 u_char ip6_protox[IPPROTO_MAX];
131 static int ip6qmaxlen = IFQ_MAXLEN;
132 struct in6_ifaddr *in6_ifaddr;
133 
134 extern struct callout in6_tmpaddrtimer_ch;
135 
136 int ip6_forward_srcrt;			/* XXX */
137 int ip6_sourcecheck;			/* XXX */
138 int ip6_sourcecheck_interval;		/* XXX */
139 const int int6intrq_present = 1;
140 
141 int ip6_ours_check_algorithm;
142 
143 
144 /* firewall hooks */
145 ip6_fw_chk_t *ip6_fw_chk_ptr;
146 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
147 int ip6_fw_enable = 1;
148 
149 struct ip6stat ip6stat;
150 
151 static void ip6_init2 __P((void *));
152 static struct mbuf *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
153 
154 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
155 #ifdef PULLDOWN_TEST
156 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
157 #endif
158 
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
165 ip6_init()
166 {
167 	struct ip6protosw *pr;
168 	int i;
169 	struct timeval tv;
170 
171 #ifdef DIAGNOSTIC
172 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
173 		panic("sizeof(protosw) != sizeof(ip6protosw)");
174 #endif
175 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
176 	if (pr == 0)
177 		panic("ip6_init");
178 	for (i = 0; i < IPPROTO_MAX; i++)
179 		ip6_protox[i] = pr - inet6sw;
180 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
181 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
182 		if (pr->pr_domain->dom_family == PF_INET6 &&
183 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
184 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
185 	ip6intrq.ifq_maxlen = ip6qmaxlen;
186 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", MTX_DEF);
187 	register_netisr(NETISR_IPV6, ip6intr);
188 	nd6_init();
189 	frag6_init();
190 	/*
191 	 * in many cases, random() here does NOT return random number
192 	 * as initialization during bootstrap time occur in fixed order.
193 	 */
194 	microtime(&tv);
195 	ip6_flow_seq = random() ^ tv.tv_usec;
196 	microtime(&tv);
197 	ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
198 }
199 
200 static void
201 ip6_init2(dummy)
202 	void *dummy;
203 {
204 
205 	/*
206 	 * to route local address of p2p link to loopback,
207 	 * assign loopback address first.
208 	 */
209 	in6_ifattach(&loif[0], NULL);
210 
211 	/* nd6_timer_init */
212 	callout_init(&nd6_timer_ch, 0);
213 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
214 
215 	/* router renumbering prefix list maintenance */
216 	callout_init(&in6_rr_timer_ch, 0);
217 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
218 
219 	/* timer for regeneranation of temporary addresses randomize ID */
220 	callout_reset(&in6_tmpaddrtimer_ch,
221 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
222 		       ip6_temp_regen_advance) * hz,
223 		      in6_tmpaddrtimer, NULL);
224 }
225 
226 /* cheat */
227 /* This must be after route_init(), which is now SI_ORDER_THIRD */
228 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
229 
230 /*
231  * IP6 input interrupt handling. Just pass the packet to ip6_input.
232  */
233 void
234 ip6intr()
235 {
236 	int s;
237 	struct mbuf *m;
238 
239 	for (;;) {
240 		s = splimp();
241 		IF_DEQUEUE(&ip6intrq, m);
242 		splx(s);
243 		if (m == 0)
244 			return;
245 		ip6_input(m);
246 	}
247 }
248 
249 extern struct	route_in6 ip6_forward_rt;
250 
251 void
252 ip6_input(m)
253 	struct mbuf *m;
254 {
255 	struct ip6_hdr *ip6;
256 	int off = sizeof(struct ip6_hdr), nest;
257 	u_int32_t plen;
258 	u_int32_t rtalert = ~0;
259 	int nxt, ours = 0;
260 	struct ifnet *deliverifp = NULL;
261 #ifdef  PFIL_HOOKS
262 	struct packet_filter_hook *pfh;
263 	struct mbuf *m0;
264 	int rv;
265 #endif  /* PFIL_HOOKS */
266 
267 #ifdef IPSEC
268 	/*
269 	 * should the inner packet be considered authentic?
270 	 * see comment in ah4_input().
271 	 */
272 	if (m) {
273 		m->m_flags &= ~M_AUTHIPHDR;
274 		m->m_flags &= ~M_AUTHIPDGM;
275 	}
276 #endif
277 
278 	/*
279 	 * make sure we don't have onion peering information into m_aux.
280 	 */
281 	ip6_delaux(m);
282 
283 	/*
284 	 * mbuf statistics by kazu
285 	 */
286 	if (m->m_flags & M_EXT) {
287 		if (m->m_next)
288 			ip6stat.ip6s_mext2m++;
289 		else
290 			ip6stat.ip6s_mext1++;
291 	} else {
292 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
293 		if (m->m_next) {
294 			if (m->m_flags & M_LOOP) {
295 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/*XXX*/
296 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
297 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
298 			else
299 				ip6stat.ip6s_m2m[0]++;
300 		} else
301 			ip6stat.ip6s_m1++;
302 #undef M2MMAX
303 	}
304 
305 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
306 	ip6stat.ip6s_total++;
307 
308 #ifndef PULLDOWN_TEST
309 	/*
310 	 * L2 bridge code and some other code can return mbuf chain
311 	 * that does not conform to KAME requirement.  too bad.
312 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
313 	 */
314 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
315 		struct mbuf *n;
316 
317 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
318 		if (n)
319 			M_COPY_PKTHDR(n, m);
320 		if (n && m->m_pkthdr.len > MHLEN) {
321 			MCLGET(n, M_DONTWAIT);
322 			if ((n->m_flags & M_EXT) == 0) {
323 				m_freem(n);
324 				n = NULL;
325 			}
326 		}
327 		if (!n) {
328 			m_freem(m);
329 			return;	/*ENOBUFS*/
330 		}
331 
332 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
333 		n->m_len = m->m_pkthdr.len;
334 		m_freem(m);
335 		m = n;
336 	}
337 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
338 #endif
339 
340 	if (m->m_len < sizeof(struct ip6_hdr)) {
341 		struct ifnet *inifp;
342 		inifp = m->m_pkthdr.rcvif;
343 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
344 			ip6stat.ip6s_toosmall++;
345 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
346 			return;
347 		}
348 	}
349 
350 	ip6 = mtod(m, struct ip6_hdr *);
351 
352 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
353 		ip6stat.ip6s_badvers++;
354 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
355 		goto bad;
356 	}
357 
358 #ifdef PFIL_HOOKS
359 	/*
360 	 * Run through list of hooks for input packets.  If there are any
361 	 * filters which require that additional packets in the flow are
362 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
363 	 * Note that filters must _never_ set this flag, as another filter
364 	 * in the list may have previously cleared it.
365 	 */
366 	m0 = m;
367 	pfh = pfil_hook_get(PFIL_IN, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
368 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
369 		if (pfh->pfil_func) {
370 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
371 					    m->m_pkthdr.rcvif, 0, &m0);
372 			if (rv)
373 				return;
374 			m = m0;
375 			if (m == NULL)
376 				return;
377 			ip6 = mtod(m, struct ip6_hdr *);
378 		}
379 #endif /* PFIL_HOOKS */
380 
381 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
382 
383 	/*
384 	 * Check with the firewall...
385 	 */
386 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
387 		u_short port = 0;
388 		/* If ipfw says divert, we have to just drop packet */
389 		/* use port as a dummy argument */
390 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
391 			m_freem(m);
392 			m = NULL;
393 		}
394 		if (!m)
395 			return;
396 	}
397 
398 	/*
399 	 * Check against address spoofing/corruption.
400 	 */
401 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
402 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
403 		/*
404 		 * XXX: "badscope" is not very suitable for a multicast source.
405 		 */
406 		ip6stat.ip6s_badscope++;
407 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
408 		goto bad;
409 	}
410 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
411 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
412 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
413 		ip6stat.ip6s_badscope++;
414 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
415 		goto bad;
416 	}
417 	/*
418 	 * The following check is not documented in specs.  A malicious
419 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
420 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
421 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
422 	 *
423 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
424 	 * support IPv4-less kernel compilation, we cannot support SIIT
425 	 * environment at all.  So, it makes more sense for us to reject any
426 	 * malicious packets for non-SIIT environment, than try to do a
427 	 * partical support for SIIT environment.
428 	 */
429 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
430 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
431 		ip6stat.ip6s_badscope++;
432 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
433 		goto bad;
434 	}
435 #if 0
436 	/*
437 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
438 	 *
439 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
440 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
441 	 * is revised to forbid relaying case.
442 	 */
443 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
444 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
445 		ip6stat.ip6s_badscope++;
446 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
447 		goto bad;
448 	}
449 #endif
450 
451 	/* drop packets if interface ID portion is already filled */
452 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
453 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
454 		    ip6->ip6_src.s6_addr16[1]) {
455 			ip6stat.ip6s_badscope++;
456 			goto bad;
457 		}
458 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
459 		    ip6->ip6_dst.s6_addr16[1]) {
460 			ip6stat.ip6s_badscope++;
461 			goto bad;
462 		}
463 	}
464 
465 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
466 		ip6->ip6_src.s6_addr16[1]
467 			= htons(m->m_pkthdr.rcvif->if_index);
468 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
469 		ip6->ip6_dst.s6_addr16[1]
470 			= htons(m->m_pkthdr.rcvif->if_index);
471 
472 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
473 	/*
474 	 * We use rt->rt_ifp to determine if the address is ours or not.
475 	 * If rt_ifp is lo0, the address is ours.
476 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
477 	 * so any address under fe80::%lo0/64 will be mistakenly considered
478 	 * local.  The special case is supplied to handle the case properly
479 	 * by actually looking at interface addresses
480 	 * (using in6ifa_ifpwithaddr).
481 	 */
482 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
483 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
484 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
485 			icmp6_error(m, ICMP6_DST_UNREACH,
486 			    ICMP6_DST_UNREACH_ADDR, 0);
487 			/* m is already freed */
488 			return;
489 		}
490 
491 		ours = 1;
492 		deliverifp = m->m_pkthdr.rcvif;
493 		goto hbhcheck;
494 	}
495 #endif
496 
497 	/*
498 	 * Multicast check
499 	 */
500 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
501 	  	struct	in6_multi *in6m = 0;
502 
503 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
504 		/*
505 		 * See if we belong to the destination multicast group on the
506 		 * arrival interface.
507 		 */
508 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
509 		if (in6m)
510 			ours = 1;
511 		else if (!ip6_mrouter) {
512 			ip6stat.ip6s_notmember++;
513 			ip6stat.ip6s_cantforward++;
514 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
515 			goto bad;
516 		}
517 		deliverifp = m->m_pkthdr.rcvif;
518 		goto hbhcheck;
519 	}
520 
521 	/*
522 	 *  Unicast check
523 	 */
524 	switch (ip6_ours_check_algorithm) {
525 	default:
526 		/*
527 		 * XXX: I intentionally broke our indentation rule here,
528 		 *      since this switch-case is just for measurement and
529 		 *      therefore should soon be removed.
530 		 */
531 	if (ip6_forward_rt.ro_rt != NULL &&
532 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
533 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
534 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
535 		ip6stat.ip6s_forward_cachehit++;
536 	else {
537 		struct sockaddr_in6 *dst6;
538 
539 		if (ip6_forward_rt.ro_rt) {
540 			/* route is down or destination is different */
541 			ip6stat.ip6s_forward_cachemiss++;
542 			RTFREE(ip6_forward_rt.ro_rt);
543 			ip6_forward_rt.ro_rt = 0;
544 		}
545 
546 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
547 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
548 		dst6->sin6_len = sizeof(struct sockaddr_in6);
549 		dst6->sin6_family = AF_INET6;
550 		dst6->sin6_addr = ip6->ip6_dst;
551 #ifdef SCOPEDROUTING
552 		ip6_forward_rt.ro_dst.sin6_scope_id =
553 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
554 #endif
555 
556 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
557 	}
558 
559 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
560 
561 	/*
562 	 * Accept the packet if the forwarding interface to the destination
563 	 * according to the routing table is the loopback interface,
564 	 * unless the associated route has a gateway.
565 	 * Note that this approach causes to accept a packet if there is a
566 	 * route to the loopback interface for the destination of the packet.
567 	 * But we think it's even useful in some situations, e.g. when using
568 	 * a special daemon which wants to intercept the packet.
569 	 *
570 	 * XXX: some OSes automatically make a cloned route for the destination
571 	 * of an outgoing packet.  If the outgoing interface of the packet
572 	 * is a loopback one, the kernel would consider the packet to be
573 	 * accepted, even if we have no such address assinged on the interface.
574 	 * We check the cloned flag of the route entry to reject such cases,
575 	 * assuming that route entries for our own addresses are not made by
576 	 * cloning (it should be true because in6_addloop explicitly installs
577 	 * the host route).  However, we might have to do an explicit check
578 	 * while it would be less efficient.  Or, should we rather install a
579 	 * reject route for such a case?
580 	 */
581 	if (ip6_forward_rt.ro_rt &&
582 	    (ip6_forward_rt.ro_rt->rt_flags &
583 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
584 #ifdef RTF_WASCLONED
585 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
586 #endif
587 #ifdef RTF_CLONED
588 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
589 #endif
590 #if 0
591 	    /*
592 	     * The check below is redundant since the comparison of
593 	     * the destination and the key of the rtentry has
594 	     * already done through looking up the routing table.
595 	     */
596 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
597 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
598 #endif
599 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
600 		struct in6_ifaddr *ia6 =
601 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
602 
603 		/*
604 		 * record address information into m_aux.
605 		 */
606 		(void)ip6_setdstifaddr(m, ia6);
607 
608 		/*
609 		 * packets to a tentative, duplicated, or somehow invalid
610 		 * address must not be accepted.
611 		 */
612 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
613 			/* this address is ready */
614 			ours = 1;
615 			deliverifp = ia6->ia_ifp;	/* correct? */
616 			/* Count the packet in the ip address stats */
617 			ia6->ia_ifa.if_ipackets++;
618 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
619 			goto hbhcheck;
620 		} else {
621 			/* address is not ready, so discard the packet. */
622 			nd6log((LOG_INFO,
623 			    "ip6_input: packet to an unready address %s->%s\n",
624 			    ip6_sprintf(&ip6->ip6_src),
625 			    ip6_sprintf(&ip6->ip6_dst)));
626 
627 			goto bad;
628 		}
629 	}
630 	} /* XXX indentation (see above) */
631 
632 	/*
633 	 * FAITH(Firewall Aided Internet Translator)
634 	 */
635 #if defined(NFAITH) && 0 < NFAITH
636 	if (ip6_keepfaith) {
637 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
638 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
639 			/* XXX do we need more sanity checks? */
640 			ours = 1;
641 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
642 			goto hbhcheck;
643 		}
644 	}
645 #endif
646 
647 	/*
648 	 * Now there is no reason to process the packet if it's not our own
649 	 * and we're not a router.
650 	 */
651 	if (!ip6_forwarding) {
652 		ip6stat.ip6s_cantforward++;
653 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
654 		goto bad;
655 	}
656 
657   hbhcheck:
658 	/*
659 	 * record address information into m_aux, if we don't have one yet.
660 	 * note that we are unable to record it, if the address is not listed
661 	 * as our interface address (e.g. multicast addresses, addresses
662 	 * within FAITH prefixes and such).
663 	 */
664 	if (deliverifp && !ip6_getdstifaddr(m)) {
665 		struct in6_ifaddr *ia6;
666 
667 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
668 		if (ia6) {
669 			if (!ip6_setdstifaddr(m, ia6)) {
670 				/*
671 				 * XXX maybe we should drop the packet here,
672 				 * as we could not provide enough information
673 				 * to the upper layers.
674 				 */
675 			}
676 		}
677 	}
678 
679 	/*
680 	 * Process Hop-by-Hop options header if it's contained.
681 	 * m may be modified in ip6_hopopts_input().
682 	 * If a JumboPayload option is included, plen will also be modified.
683 	 */
684 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
685 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
686 		struct ip6_hbh *hbh;
687 
688 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
689 #if 0	/*touches NULL pointer*/
690 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
691 #endif
692 			return;	/* m have already been freed */
693 		}
694 
695 		/* adjust pointer */
696 		ip6 = mtod(m, struct ip6_hdr *);
697 
698 		/*
699 		 * if the payload length field is 0 and the next header field
700 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
701 		 * option MUST be included.
702 		 */
703 		if (ip6->ip6_plen == 0 && plen == 0) {
704 			/*
705 			 * Note that if a valid jumbo payload option is
706 			 * contained, ip6_hoptops_input() must set a valid
707 			 * (non-zero) payload length to the variable plen.
708 			 */
709 			ip6stat.ip6s_badoptions++;
710 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
711 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
712 			icmp6_error(m, ICMP6_PARAM_PROB,
713 				    ICMP6_PARAMPROB_HEADER,
714 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
715 			return;
716 		}
717 #ifndef PULLDOWN_TEST
718 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
719 		hbh = (struct ip6_hbh *)(ip6 + 1);
720 #else
721 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
722 			sizeof(struct ip6_hbh));
723 		if (hbh == NULL) {
724 			ip6stat.ip6s_tooshort++;
725 			return;
726 		}
727 #endif
728 		nxt = hbh->ip6h_nxt;
729 
730 		/*
731 		 * accept the packet if a router alert option is included
732 		 * and we act as an IPv6 router.
733 		 */
734 		if (rtalert != ~0 && ip6_forwarding)
735 			ours = 1;
736 	} else
737 		nxt = ip6->ip6_nxt;
738 
739 	/*
740 	 * Check that the amount of data in the buffers
741 	 * is as at least much as the IPv6 header would have us expect.
742 	 * Trim mbufs if longer than we expect.
743 	 * Drop packet if shorter than we expect.
744 	 */
745 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
746 		ip6stat.ip6s_tooshort++;
747 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
748 		goto bad;
749 	}
750 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
751 		if (m->m_len == m->m_pkthdr.len) {
752 			m->m_len = sizeof(struct ip6_hdr) + plen;
753 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
754 		} else
755 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
756 	}
757 
758 	/*
759 	 * Forward if desirable.
760 	 */
761 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
762 		/*
763 		 * If we are acting as a multicast router, all
764 		 * incoming multicast packets are passed to the
765 		 * kernel-level multicast forwarding function.
766 		 * The packet is returned (relatively) intact; if
767 		 * ip6_mforward() returns a non-zero value, the packet
768 		 * must be discarded, else it may be accepted below.
769 		 */
770 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
771 			ip6stat.ip6s_cantforward++;
772 			m_freem(m);
773 			return;
774 		}
775 		if (!ours) {
776 			m_freem(m);
777 			return;
778 		}
779 	} else if (!ours) {
780 		ip6_forward(m, 0);
781 		return;
782 	}
783 
784 	ip6 = mtod(m, struct ip6_hdr *);
785 
786 	/*
787 	 * Malicious party may be able to use IPv4 mapped addr to confuse
788 	 * tcp/udp stack and bypass security checks (act as if it was from
789 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
790 	 *
791 	 * For SIIT end node behavior, you may want to disable the check.
792 	 * However, you will  become vulnerable to attacks using IPv4 mapped
793 	 * source.
794 	 */
795 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
796 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
797 		ip6stat.ip6s_badscope++;
798 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
799 		goto bad;
800 	}
801 
802 	/*
803 	 * Tell launch routine the next header
804 	 */
805 	ip6stat.ip6s_delivered++;
806 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
807 	nest = 0;
808 
809 	while (nxt != IPPROTO_DONE) {
810 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
811 			ip6stat.ip6s_toomanyhdr++;
812 			goto bad;
813 		}
814 
815 		/*
816 		 * protection against faulty packet - there should be
817 		 * more sanity checks in header chain processing.
818 		 */
819 		if (m->m_pkthdr.len < off) {
820 			ip6stat.ip6s_tooshort++;
821 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
822 			goto bad;
823 		}
824 
825 #if 0
826 		/*
827 		 * do we need to do it for every header?  yeah, other
828 		 * functions can play with it (like re-allocate and copy).
829 		 */
830 		mhist = ip6_addaux(m);
831 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
832 			hist = mtod(mhist, caddr_t) + mhist->m_len;
833 			bcopy(&nxt, hist, sizeof(nxt));
834 			mhist->m_len += sizeof(nxt);
835 		} else {
836 			ip6stat.ip6s_toomanyhdr++;
837 			goto bad;
838 		}
839 #endif
840 
841 #ifdef IPSEC
842 		/*
843 		 * enforce IPsec policy checking if we are seeing last header.
844 		 * note that we do not visit this with protocols with pcb layer
845 		 * code - like udp/tcp/raw ip.
846 		 */
847 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
848 		    ipsec6_in_reject(m, NULL)) {
849 			ipsec6stat.in_polvio++;
850 			goto bad;
851 		}
852 #endif
853 
854 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
855 	}
856 	return;
857  bad:
858 	m_freem(m);
859 }
860 
861 /*
862  * set/grab in6_ifaddr correspond to IPv6 destination address.
863  * XXX backward compatibility wrapper
864  */
865 static struct mbuf *
866 ip6_setdstifaddr(m, ia6)
867 	struct mbuf *m;
868 	struct in6_ifaddr *ia6;
869 {
870 	struct mbuf *n;
871 
872 	n = ip6_addaux(m);
873 	if (n)
874 		mtod(n, struct ip6aux *)->ip6a_dstia6 = ia6;
875 	return n;	/* NULL if failed to set */
876 }
877 
878 struct in6_ifaddr *
879 ip6_getdstifaddr(m)
880 	struct mbuf *m;
881 {
882 	struct mbuf *n;
883 
884 	n = ip6_findaux(m);
885 	if (n)
886 		return mtod(n, struct ip6aux *)->ip6a_dstia6;
887 	else
888 		return NULL;
889 }
890 
891 /*
892  * Hop-by-Hop options header processing. If a valid jumbo payload option is
893  * included, the real payload length will be stored in plenp.
894  */
895 static int
896 ip6_hopopts_input(plenp, rtalertp, mp, offp)
897 	u_int32_t *plenp;
898 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
899 	struct mbuf **mp;
900 	int *offp;
901 {
902 	struct mbuf *m = *mp;
903 	int off = *offp, hbhlen;
904 	struct ip6_hbh *hbh;
905 	u_int8_t *opt;
906 
907 	/* validation of the length of the header */
908 #ifndef PULLDOWN_TEST
909 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
910 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
911 	hbhlen = (hbh->ip6h_len + 1) << 3;
912 
913 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
914 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
915 #else
916 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
917 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
918 	if (hbh == NULL) {
919 		ip6stat.ip6s_tooshort++;
920 		return -1;
921 	}
922 	hbhlen = (hbh->ip6h_len + 1) << 3;
923 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
924 		hbhlen);
925 	if (hbh == NULL) {
926 		ip6stat.ip6s_tooshort++;
927 		return -1;
928 	}
929 #endif
930 	off += hbhlen;
931 	hbhlen -= sizeof(struct ip6_hbh);
932 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
933 
934 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
935 				hbhlen, rtalertp, plenp) < 0)
936 		return(-1);
937 
938 	*offp = off;
939 	*mp = m;
940 	return(0);
941 }
942 
943 /*
944  * Search header for all Hop-by-hop options and process each option.
945  * This function is separate from ip6_hopopts_input() in order to
946  * handle a case where the sending node itself process its hop-by-hop
947  * options header. In such a case, the function is called from ip6_output().
948  *
949  * The function assumes that hbh header is located right after the IPv6 header
950  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
951  * opthead + hbhlen is located in continuous memory region.
952  */
953 int
954 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
955 	struct mbuf *m;
956 	u_int8_t *opthead;
957 	int hbhlen;
958 	u_int32_t *rtalertp;
959 	u_int32_t *plenp;
960 {
961 	struct ip6_hdr *ip6;
962 	int optlen = 0;
963 	u_int8_t *opt = opthead;
964 	u_int16_t rtalert_val;
965 	u_int32_t jumboplen;
966 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
967 
968 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
969 		switch (*opt) {
970 		case IP6OPT_PAD1:
971 			optlen = 1;
972 			break;
973 		case IP6OPT_PADN:
974 			if (hbhlen < IP6OPT_MINLEN) {
975 				ip6stat.ip6s_toosmall++;
976 				goto bad;
977 			}
978 			optlen = *(opt + 1) + 2;
979 			break;
980 		case IP6OPT_RTALERT:
981 			/* XXX may need check for alignment */
982 			if (hbhlen < IP6OPT_RTALERT_LEN) {
983 				ip6stat.ip6s_toosmall++;
984 				goto bad;
985 			}
986 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
987 				/* XXX stat */
988 				icmp6_error(m, ICMP6_PARAM_PROB,
989 					    ICMP6_PARAMPROB_HEADER,
990 					    erroff + opt + 1 - opthead);
991 				return(-1);
992 			}
993 			optlen = IP6OPT_RTALERT_LEN;
994 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
995 			*rtalertp = ntohs(rtalert_val);
996 			break;
997 		case IP6OPT_JUMBO:
998 			/* XXX may need check for alignment */
999 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1000 				ip6stat.ip6s_toosmall++;
1001 				goto bad;
1002 			}
1003 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1004 				/* XXX stat */
1005 				icmp6_error(m, ICMP6_PARAM_PROB,
1006 					    ICMP6_PARAMPROB_HEADER,
1007 					    erroff + opt + 1 - opthead);
1008 				return(-1);
1009 			}
1010 			optlen = IP6OPT_JUMBO_LEN;
1011 
1012 			/*
1013 			 * IPv6 packets that have non 0 payload length
1014 			 * must not contain a jumbo payload option.
1015 			 */
1016 			ip6 = mtod(m, struct ip6_hdr *);
1017 			if (ip6->ip6_plen) {
1018 				ip6stat.ip6s_badoptions++;
1019 				icmp6_error(m, ICMP6_PARAM_PROB,
1020 					    ICMP6_PARAMPROB_HEADER,
1021 					    erroff + opt - opthead);
1022 				return(-1);
1023 			}
1024 
1025 			/*
1026 			 * We may see jumbolen in unaligned location, so
1027 			 * we'd need to perform bcopy().
1028 			 */
1029 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1030 			jumboplen = (u_int32_t)htonl(jumboplen);
1031 
1032 #if 1
1033 			/*
1034 			 * if there are multiple jumbo payload options,
1035 			 * *plenp will be non-zero and the packet will be
1036 			 * rejected.
1037 			 * the behavior may need some debate in ipngwg -
1038 			 * multiple options does not make sense, however,
1039 			 * there's no explicit mention in specification.
1040 			 */
1041 			if (*plenp != 0) {
1042 				ip6stat.ip6s_badoptions++;
1043 				icmp6_error(m, ICMP6_PARAM_PROB,
1044 					    ICMP6_PARAMPROB_HEADER,
1045 					    erroff + opt + 2 - opthead);
1046 				return(-1);
1047 			}
1048 #endif
1049 
1050 			/*
1051 			 * jumbo payload length must be larger than 65535.
1052 			 */
1053 			if (jumboplen <= IPV6_MAXPACKET) {
1054 				ip6stat.ip6s_badoptions++;
1055 				icmp6_error(m, ICMP6_PARAM_PROB,
1056 					    ICMP6_PARAMPROB_HEADER,
1057 					    erroff + opt + 2 - opthead);
1058 				return(-1);
1059 			}
1060 			*plenp = jumboplen;
1061 
1062 			break;
1063 		default:		/* unknown option */
1064 			if (hbhlen < IP6OPT_MINLEN) {
1065 				ip6stat.ip6s_toosmall++;
1066 				goto bad;
1067 			}
1068 			optlen = ip6_unknown_opt(opt, m,
1069 			    erroff + opt - opthead);
1070 			if (optlen == -1)
1071 				return(-1);
1072 			optlen += 2;
1073 			break;
1074 		}
1075 	}
1076 
1077 	return(0);
1078 
1079   bad:
1080 	m_freem(m);
1081 	return(-1);
1082 }
1083 
1084 /*
1085  * Unknown option processing.
1086  * The third argument `off' is the offset from the IPv6 header to the option,
1087  * which is necessary if the IPv6 header the and option header and IPv6 header
1088  * is not continuous in order to return an ICMPv6 error.
1089  */
1090 int
1091 ip6_unknown_opt(optp, m, off)
1092 	u_int8_t *optp;
1093 	struct mbuf *m;
1094 	int off;
1095 {
1096 	struct ip6_hdr *ip6;
1097 
1098 	switch (IP6OPT_TYPE(*optp)) {
1099 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1100 		return((int)*(optp + 1));
1101 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1102 		m_freem(m);
1103 		return(-1);
1104 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1105 		ip6stat.ip6s_badoptions++;
1106 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1107 		return(-1);
1108 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1109 		ip6stat.ip6s_badoptions++;
1110 		ip6 = mtod(m, struct ip6_hdr *);
1111 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1112 		    (m->m_flags & (M_BCAST|M_MCAST)))
1113 			m_freem(m);
1114 		else
1115 			icmp6_error(m, ICMP6_PARAM_PROB,
1116 				    ICMP6_PARAMPROB_OPTION, off);
1117 		return(-1);
1118 	}
1119 
1120 	m_freem(m);		/* XXX: NOTREACHED */
1121 	return(-1);
1122 }
1123 
1124 /*
1125  * Create the "control" list for this pcb.
1126  * The function will not modify mbuf chain at all.
1127  *
1128  * with KAME mbuf chain restriction:
1129  * The routine will be called from upper layer handlers like tcp6_input().
1130  * Thus the routine assumes that the caller (tcp6_input) have already
1131  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1132  * very first mbuf on the mbuf chain.
1133  */
1134 void
1135 ip6_savecontrol(in6p, mp, ip6, m)
1136 	struct inpcb *in6p;
1137 	struct mbuf **mp;
1138 	struct ip6_hdr *ip6;
1139 	struct mbuf *m;
1140 {
1141 	struct proc *p = curproc;	/* XXX */
1142 	int privileged = 0;
1143 	int rthdr_exist = 0;
1144 
1145 
1146 	if (p && !suser(p))
1147  		privileged++;
1148 
1149 #ifdef SO_TIMESTAMP
1150 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1151 		struct timeval tv;
1152 
1153 		microtime(&tv);
1154 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1155 				      SCM_TIMESTAMP, SOL_SOCKET);
1156 		if (*mp) {
1157 			mp = &(*mp)->m_next;
1158 		}
1159 	}
1160 #endif
1161 
1162 	/* RFC 2292 sec. 5 */
1163 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1164 		struct in6_pktinfo pi6;
1165 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1166 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1167 			pi6.ipi6_addr.s6_addr16[1] = 0;
1168 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1169 					? m->m_pkthdr.rcvif->if_index
1170 					: 0;
1171 		*mp = sbcreatecontrol((caddr_t) &pi6,
1172 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1173 			IPPROTO_IPV6);
1174 		if (*mp)
1175 			mp = &(*mp)->m_next;
1176 	}
1177 
1178 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1179 		int hlim = ip6->ip6_hlim & 0xff;
1180 		*mp = sbcreatecontrol((caddr_t) &hlim,
1181 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1182 		if (*mp)
1183 			mp = &(*mp)->m_next;
1184 	}
1185 
1186 	/*
1187 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1188 	 * for the option, but it might be too strict, since there might
1189 	 * be some hop-by-hop options which can be returned to normal user.
1190 	 * See RFC 2292 section 6.
1191 	 */
1192 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1193 		/*
1194 		 * Check if a hop-by-hop options header is contatined in the
1195 		 * received packet, and if so, store the options as ancillary
1196 		 * data. Note that a hop-by-hop options header must be
1197 		 * just after the IPv6 header, which fact is assured through
1198 		 * the IPv6 input processing.
1199 		 */
1200 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1201 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1202 			struct ip6_hbh *hbh;
1203 			int hbhlen = 0;
1204 #ifdef PULLDOWN_TEST
1205 			struct mbuf *ext;
1206 #endif
1207 
1208 #ifndef PULLDOWN_TEST
1209 			hbh = (struct ip6_hbh *)(ip6 + 1);
1210 			hbhlen = (hbh->ip6h_len + 1) << 3;
1211 #else
1212 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1213 			    ip6->ip6_nxt);
1214 			if (ext == NULL) {
1215 				ip6stat.ip6s_tooshort++;
1216 				return;
1217 			}
1218 			hbh = mtod(ext, struct ip6_hbh *);
1219 			hbhlen = (hbh->ip6h_len + 1) << 3;
1220 			if (hbhlen != ext->m_len) {
1221 				m_freem(ext);
1222 				ip6stat.ip6s_tooshort++;
1223 				return;
1224 			}
1225 #endif
1226 
1227 			/*
1228 			 * XXX: We copy whole the header even if a jumbo
1229 			 * payload option is included, which option is to
1230 			 * be removed before returning in the RFC 2292.
1231 			 * Note: this constraint is removed in 2292bis.
1232 			 */
1233 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1234 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1235 			if (*mp)
1236 				mp = &(*mp)->m_next;
1237 #ifdef PULLDOWN_TEST
1238 			m_freem(ext);
1239 #endif
1240 		}
1241 	}
1242 
1243 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1244 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1245 		int proto, off, nxt;
1246 
1247 		/*
1248 		 * go through the header chain to see if a routing header is
1249 		 * contained in the packet. We need this information to store
1250 		 * destination options headers (if any) properly.
1251 		 * XXX: performance issue. We should record this info when
1252 		 * processing extension headers in incoming routine.
1253 		 * (todo) use m_aux?
1254 		 */
1255 		proto = IPPROTO_IPV6;
1256 		off = 0;
1257 		nxt = -1;
1258 		while (1) {
1259 			int newoff;
1260 
1261 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1262 			if (newoff < 0)
1263 				break;
1264 			if (newoff < off) /* invalid, check for safety */
1265 				break;
1266 			if ((proto = nxt) == IPPROTO_ROUTING) {
1267 				rthdr_exist = 1;
1268 				break;
1269 			}
1270 			off = newoff;
1271 		}
1272 	}
1273 
1274 	if ((in6p->in6p_flags &
1275 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1276 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1277 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1278 
1279 		/*
1280 		 * Search for destination options headers or routing
1281 		 * header(s) through the header chain, and stores each
1282 		 * header as ancillary data.
1283 		 * Note that the order of the headers remains in
1284 		 * the chain of ancillary data.
1285 		 */
1286 		while (1) {	/* is explicit loop prevention necessary? */
1287 			struct ip6_ext *ip6e = NULL;
1288 			int elen;
1289 #ifdef PULLDOWN_TEST
1290 			struct mbuf *ext = NULL;
1291 #endif
1292 
1293 			/*
1294 			 * if it is not an extension header, don't try to
1295 			 * pull it from the chain.
1296 			 */
1297 			switch (nxt) {
1298 			case IPPROTO_DSTOPTS:
1299 			case IPPROTO_ROUTING:
1300 			case IPPROTO_HOPOPTS:
1301 			case IPPROTO_AH: /* is it possible? */
1302 				break;
1303 			default:
1304 				goto loopend;
1305 			}
1306 
1307 #ifndef PULLDOWN_TEST
1308 			if (off + sizeof(*ip6e) > m->m_len)
1309 				goto loopend;
1310 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1311 			if (nxt == IPPROTO_AH)
1312 				elen = (ip6e->ip6e_len + 2) << 2;
1313 			else
1314 				elen = (ip6e->ip6e_len + 1) << 3;
1315 			if (off + elen > m->m_len)
1316 				goto loopend;
1317 #else
1318 			ext = ip6_pullexthdr(m, off, nxt);
1319 			if (ext == NULL) {
1320 				ip6stat.ip6s_tooshort++;
1321 				return;
1322 			}
1323 			ip6e = mtod(ext, struct ip6_ext *);
1324 			if (nxt == IPPROTO_AH)
1325 				elen = (ip6e->ip6e_len + 2) << 2;
1326 			else
1327 				elen = (ip6e->ip6e_len + 1) << 3;
1328 			if (elen != ext->m_len) {
1329 				m_freem(ext);
1330 				ip6stat.ip6s_tooshort++;
1331 				return;
1332 			}
1333 #endif
1334 
1335 			switch (nxt) {
1336 			case IPPROTO_DSTOPTS:
1337 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1338 					break;
1339 
1340 				/*
1341 				 * We also require super-user privilege for
1342 				 * the option.
1343 				 * See the comments on IN6_HOPOPTS.
1344 				 */
1345 				if (!privileged)
1346 					break;
1347 
1348 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1349 						      IPV6_DSTOPTS,
1350 						      IPPROTO_IPV6);
1351 				if (*mp)
1352 					mp = &(*mp)->m_next;
1353 				break;
1354 			case IPPROTO_ROUTING:
1355 				if (!in6p->in6p_flags & IN6P_RTHDR)
1356 					break;
1357 
1358 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1359 						      IPV6_RTHDR,
1360 						      IPPROTO_IPV6);
1361 				if (*mp)
1362 					mp = &(*mp)->m_next;
1363 				break;
1364 			case IPPROTO_HOPOPTS:
1365 			case IPPROTO_AH: /* is it possible? */
1366 				break;
1367 
1368 			default:
1369 				/*
1370 			 	 * other cases have been filtered in the above.
1371 				 * none will visit this case.  here we supply
1372 				 * the code just in case (nxt overwritten or
1373 				 * other cases).
1374 				 */
1375 #ifdef PULLDOWN_TEST
1376 				m_freem(ext);
1377 #endif
1378 				goto loopend;
1379 
1380 			}
1381 
1382 			/* proceed with the next header. */
1383 			off += elen;
1384 			nxt = ip6e->ip6e_nxt;
1385 			ip6e = NULL;
1386 #ifdef PULLDOWN_TEST
1387 			m_freem(ext);
1388 			ext = NULL;
1389 #endif
1390 		}
1391 	  loopend:
1392 		;
1393 	}
1394 
1395 }
1396 
1397 #ifdef PULLDOWN_TEST
1398 /*
1399  * pull single extension header from mbuf chain.  returns single mbuf that
1400  * contains the result, or NULL on error.
1401  */
1402 static struct mbuf *
1403 ip6_pullexthdr(m, off, nxt)
1404 	struct mbuf *m;
1405 	size_t off;
1406 	int nxt;
1407 {
1408 	struct ip6_ext ip6e;
1409 	size_t elen;
1410 	struct mbuf *n;
1411 
1412 #ifdef DIAGNOSTIC
1413 	switch (nxt) {
1414 	case IPPROTO_DSTOPTS:
1415 	case IPPROTO_ROUTING:
1416 	case IPPROTO_HOPOPTS:
1417 	case IPPROTO_AH: /* is it possible? */
1418 		break;
1419 	default:
1420 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1421 	}
1422 #endif
1423 
1424 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1425 	if (nxt == IPPROTO_AH)
1426 		elen = (ip6e.ip6e_len + 2) << 2;
1427 	else
1428 		elen = (ip6e.ip6e_len + 1) << 3;
1429 
1430 	MGET(n, M_DONTWAIT, MT_DATA);
1431 	if (n && elen >= MLEN) {
1432 		MCLGET(n, M_DONTWAIT);
1433 		if ((n->m_flags & M_EXT) == 0) {
1434 			m_free(n);
1435 			n = NULL;
1436 		}
1437 	}
1438 	if (!n)
1439 		return NULL;
1440 
1441 	n->m_len = 0;
1442 	if (elen >= M_TRAILINGSPACE(n)) {
1443 		m_free(n);
1444 		return NULL;
1445 	}
1446 
1447 	m_copydata(m, off, elen, mtod(n, caddr_t));
1448 	n->m_len = elen;
1449 	return n;
1450 }
1451 #endif
1452 
1453 /*
1454  * Get pointer to the previous header followed by the header
1455  * currently processed.
1456  * XXX: This function supposes that
1457  *	M includes all headers,
1458  *	the next header field and the header length field of each header
1459  *	are valid, and
1460  *	the sum of each header length equals to OFF.
1461  * Because of these assumptions, this function must be called very
1462  * carefully. Moreover, it will not be used in the near future when
1463  * we develop `neater' mechanism to process extension headers.
1464  */
1465 char *
1466 ip6_get_prevhdr(m, off)
1467 	struct mbuf *m;
1468 	int off;
1469 {
1470 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1471 
1472 	if (off == sizeof(struct ip6_hdr))
1473 		return(&ip6->ip6_nxt);
1474 	else {
1475 		int len, nxt;
1476 		struct ip6_ext *ip6e = NULL;
1477 
1478 		nxt = ip6->ip6_nxt;
1479 		len = sizeof(struct ip6_hdr);
1480 		while (len < off) {
1481 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1482 
1483 			switch (nxt) {
1484 			case IPPROTO_FRAGMENT:
1485 				len += sizeof(struct ip6_frag);
1486 				break;
1487 			case IPPROTO_AH:
1488 				len += (ip6e->ip6e_len + 2) << 2;
1489 				break;
1490 			default:
1491 				len += (ip6e->ip6e_len + 1) << 3;
1492 				break;
1493 			}
1494 			nxt = ip6e->ip6e_nxt;
1495 		}
1496 		if (ip6e)
1497 			return(&ip6e->ip6e_nxt);
1498 		else
1499 			return NULL;
1500 	}
1501 }
1502 
1503 /*
1504  * get next header offset.  m will be retained.
1505  */
1506 int
1507 ip6_nexthdr(m, off, proto, nxtp)
1508 	struct mbuf *m;
1509 	int off;
1510 	int proto;
1511 	int *nxtp;
1512 {
1513 	struct ip6_hdr ip6;
1514 	struct ip6_ext ip6e;
1515 	struct ip6_frag fh;
1516 
1517 	/* just in case */
1518 	if (m == NULL)
1519 		panic("ip6_nexthdr: m == NULL");
1520 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1521 		return -1;
1522 
1523 	switch (proto) {
1524 	case IPPROTO_IPV6:
1525 		if (m->m_pkthdr.len < off + sizeof(ip6))
1526 			return -1;
1527 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1528 		if (nxtp)
1529 			*nxtp = ip6.ip6_nxt;
1530 		off += sizeof(ip6);
1531 		return off;
1532 
1533 	case IPPROTO_FRAGMENT:
1534 		/*
1535 		 * terminate parsing if it is not the first fragment,
1536 		 * it does not make sense to parse through it.
1537 		 */
1538 		if (m->m_pkthdr.len < off + sizeof(fh))
1539 			return -1;
1540 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1541 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1542 			return -1;
1543 		if (nxtp)
1544 			*nxtp = fh.ip6f_nxt;
1545 		off += sizeof(struct ip6_frag);
1546 		return off;
1547 
1548 	case IPPROTO_AH:
1549 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1550 			return -1;
1551 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1552 		if (nxtp)
1553 			*nxtp = ip6e.ip6e_nxt;
1554 		off += (ip6e.ip6e_len + 2) << 2;
1555 		return off;
1556 
1557 	case IPPROTO_HOPOPTS:
1558 	case IPPROTO_ROUTING:
1559 	case IPPROTO_DSTOPTS:
1560 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1561 			return -1;
1562 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1563 		if (nxtp)
1564 			*nxtp = ip6e.ip6e_nxt;
1565 		off += (ip6e.ip6e_len + 1) << 3;
1566 		return off;
1567 
1568 	case IPPROTO_NONE:
1569 	case IPPROTO_ESP:
1570 	case IPPROTO_IPCOMP:
1571 		/* give up */
1572 		return -1;
1573 
1574 	default:
1575 		return -1;
1576 	}
1577 
1578 	return -1;
1579 }
1580 
1581 /*
1582  * get offset for the last header in the chain.  m will be kept untainted.
1583  */
1584 int
1585 ip6_lasthdr(m, off, proto, nxtp)
1586 	struct mbuf *m;
1587 	int off;
1588 	int proto;
1589 	int *nxtp;
1590 {
1591 	int newoff;
1592 	int nxt;
1593 
1594 	if (!nxtp) {
1595 		nxt = -1;
1596 		nxtp = &nxt;
1597 	}
1598 	while (1) {
1599 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1600 		if (newoff < 0)
1601 			return off;
1602 		else if (newoff < off)
1603 			return -1;	/* invalid */
1604 		else if (newoff == off)
1605 			return newoff;
1606 
1607 		off = newoff;
1608 		proto = *nxtp;
1609 	}
1610 }
1611 
1612 struct mbuf *
1613 ip6_addaux(m)
1614 	struct mbuf *m;
1615 {
1616 	struct mbuf *n;
1617 
1618 #ifdef DIAGNOSTIC
1619 	if (sizeof(struct ip6aux) > MHLEN)
1620 		panic("assumption failed on sizeof(ip6aux)");
1621 #endif
1622 	n = m_aux_find(m, AF_INET6, -1);
1623 	if (n) {
1624 		if (n->m_len < sizeof(struct ip6aux)) {
1625 			printf("conflicting use of ip6aux");
1626 			return NULL;
1627 		}
1628 	} else {
1629 		n = m_aux_add(m, AF_INET6, -1);
1630 		n->m_len = sizeof(struct ip6aux);
1631 		bzero(mtod(n, caddr_t), n->m_len);
1632 	}
1633 	return n;
1634 }
1635 
1636 struct mbuf *
1637 ip6_findaux(m)
1638 	struct mbuf *m;
1639 {
1640 	struct mbuf *n;
1641 
1642 	n = m_aux_find(m, AF_INET6, -1);
1643 	if (n && n->m_len < sizeof(struct ip6aux)) {
1644 		printf("conflicting use of ip6aux");
1645 		n = NULL;
1646 	}
1647 	return n;
1648 }
1649 
1650 void
1651 ip6_delaux(m)
1652 	struct mbuf *m;
1653 {
1654 	struct mbuf *n;
1655 
1656 	n = m_aux_find(m, AF_INET6, -1);
1657 	if (n)
1658 		m_aux_delete(m, n);
1659 }
1660 
1661 /*
1662  * System control for IP6
1663  */
1664 
1665 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1666 	0,		0,		0,		0,
1667 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1668 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1669 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1670 	0,		0,		0,		0,
1671 	ENOPROTOOPT
1672 };
1673