xref: /freebsd/sys/netinet6/ip6_input.c (revision c17d43407fe04133a94055b0dbc7ea8965654a9f)
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/proc.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #include <net/intrq.h>
94 #ifdef PFIL_HOOKS
95 #include <net/pfil.h>
96 #endif
97 
98 #include <netinet/in.h>
99 #include <netinet/in_systm.h>
100 #ifdef INET
101 #include <netinet/ip.h>
102 #include <netinet/ip_icmp.h>
103 #endif /*INET*/
104 #include <netinet/ip6.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet/in_pcb.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/in6_ifattach.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/in6_prefix.h>
112 
113 #ifdef IPSEC
114 #include <netinet6/ipsec.h>
115 #ifdef INET6
116 #include <netinet6/ipsec6.h>
117 #endif
118 #endif
119 
120 #include <netinet6/ip6_fw.h>
121 
122 #include <netinet6/ip6protosw.h>
123 
124 #include <net/net_osdep.h>
125 
126 extern struct domain inet6domain;
127 
128 u_char ip6_protox[IPPROTO_MAX];
129 static int ip6qmaxlen = IFQ_MAXLEN;
130 struct in6_ifaddr *in6_ifaddr;
131 
132 extern struct callout in6_tmpaddrtimer_ch;
133 
134 int ip6_forward_srcrt;			/* XXX */
135 int ip6_sourcecheck;			/* XXX */
136 int ip6_sourcecheck_interval;		/* XXX */
137 
138 int ip6_ours_check_algorithm;
139 
140 
141 /* firewall hooks */
142 ip6_fw_chk_t *ip6_fw_chk_ptr;
143 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
144 int ip6_fw_enable = 1;
145 
146 struct ip6stat ip6stat;
147 
148 static void ip6_init2 __P((void *));
149 static struct mbuf *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
150 
151 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
152 #ifdef PULLDOWN_TEST
153 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
154 #endif
155 
156 
157 /*
158  * IP6 initialization: fill in IP6 protocol switch table.
159  * All protocols not implemented in kernel go to raw IP6 protocol handler.
160  */
161 void
162 ip6_init()
163 {
164 	struct ip6protosw *pr;
165 	int i;
166 	struct timeval tv;
167 
168 #ifdef DIAGNOSTIC
169 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
170 		panic("sizeof(protosw) != sizeof(ip6protosw)");
171 #endif
172 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
173 	if (pr == 0)
174 		panic("ip6_init");
175 	for (i = 0; i < IPPROTO_MAX; i++)
176 		ip6_protox[i] = pr - inet6sw;
177 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
178 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
179 		if (pr->pr_domain->dom_family == PF_INET6 &&
180 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
181 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
182 	ip6intrq.ifq_maxlen = ip6qmaxlen;
183 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", MTX_DEF);
184 	ip6intrq_present = 1;
185 	register_netisr(NETISR_IPV6, ip6intr);
186 	nd6_init();
187 	frag6_init();
188 	/*
189 	 * in many cases, random() here does NOT return random number
190 	 * as initialization during bootstrap time occur in fixed order.
191 	 */
192 	microtime(&tv);
193 	ip6_flow_seq = random() ^ tv.tv_usec;
194 	microtime(&tv);
195 	ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
196 }
197 
198 static void
199 ip6_init2(dummy)
200 	void *dummy;
201 {
202 
203 	/*
204 	 * to route local address of p2p link to loopback,
205 	 * assign loopback address first.
206 	 */
207 	in6_ifattach(&loif[0], NULL);
208 
209 	/* nd6_timer_init */
210 	callout_init(&nd6_timer_ch, 0);
211 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
212 
213 	/* router renumbering prefix list maintenance */
214 	callout_init(&in6_rr_timer_ch, 0);
215 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
216 
217 	/* timer for regeneranation of temporary addresses randomize ID */
218 	callout_reset(&in6_tmpaddrtimer_ch,
219 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
220 		       ip6_temp_regen_advance) * hz,
221 		      in6_tmpaddrtimer, NULL);
222 }
223 
224 /* cheat */
225 /* This must be after route_init(), which is now SI_ORDER_THIRD */
226 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
227 
228 /*
229  * IP6 input interrupt handling. Just pass the packet to ip6_input.
230  */
231 void
232 ip6intr()
233 {
234 	int s;
235 	struct mbuf *m;
236 
237 	for (;;) {
238 		s = splimp();
239 		IF_DEQUEUE(&ip6intrq, m);
240 		splx(s);
241 		if (m == 0)
242 			return;
243 		ip6_input(m);
244 	}
245 }
246 
247 extern struct	route_in6 ip6_forward_rt;
248 
249 void
250 ip6_input(m)
251 	struct mbuf *m;
252 {
253 	struct ip6_hdr *ip6;
254 	int off = sizeof(struct ip6_hdr), nest;
255 	u_int32_t plen;
256 	u_int32_t rtalert = ~0;
257 	int nxt, ours = 0;
258 	struct ifnet *deliverifp = NULL;
259 #ifdef  PFIL_HOOKS
260 	struct packet_filter_hook *pfh;
261 	struct mbuf *m0;
262 	int rv;
263 #endif  /* PFIL_HOOKS */
264 
265 #ifdef IPSEC
266 	/*
267 	 * should the inner packet be considered authentic?
268 	 * see comment in ah4_input().
269 	 */
270 	if (m) {
271 		m->m_flags &= ~M_AUTHIPHDR;
272 		m->m_flags &= ~M_AUTHIPDGM;
273 	}
274 #endif
275 
276 	/*
277 	 * make sure we don't have onion peering information into m_aux.
278 	 */
279 	ip6_delaux(m);
280 
281 	/*
282 	 * mbuf statistics by kazu
283 	 */
284 	if (m->m_flags & M_EXT) {
285 		if (m->m_next)
286 			ip6stat.ip6s_mext2m++;
287 		else
288 			ip6stat.ip6s_mext1++;
289 	} else {
290 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
291 		if (m->m_next) {
292 			if (m->m_flags & M_LOOP) {
293 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/*XXX*/
294 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
295 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
296 			else
297 				ip6stat.ip6s_m2m[0]++;
298 		} else
299 			ip6stat.ip6s_m1++;
300 #undef M2MMAX
301 	}
302 
303 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
304 	ip6stat.ip6s_total++;
305 
306 #ifndef PULLDOWN_TEST
307 	/*
308 	 * L2 bridge code and some other code can return mbuf chain
309 	 * that does not conform to KAME requirement.  too bad.
310 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
311 	 */
312 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
313 		struct mbuf *n;
314 
315 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
316 		if (n)
317 			M_COPY_PKTHDR(n, m);
318 		if (n && m->m_pkthdr.len > MHLEN) {
319 			MCLGET(n, M_DONTWAIT);
320 			if ((n->m_flags & M_EXT) == 0) {
321 				m_freem(n);
322 				n = NULL;
323 			}
324 		}
325 		if (!n) {
326 			m_freem(m);
327 			return;	/*ENOBUFS*/
328 		}
329 
330 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
331 		n->m_len = m->m_pkthdr.len;
332 		m_freem(m);
333 		m = n;
334 	}
335 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
336 #endif
337 
338 	if (m->m_len < sizeof(struct ip6_hdr)) {
339 		struct ifnet *inifp;
340 		inifp = m->m_pkthdr.rcvif;
341 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
342 			ip6stat.ip6s_toosmall++;
343 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
344 			return;
345 		}
346 	}
347 
348 	ip6 = mtod(m, struct ip6_hdr *);
349 
350 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
351 		ip6stat.ip6s_badvers++;
352 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
353 		goto bad;
354 	}
355 
356 #ifdef PFIL_HOOKS
357 	/*
358 	 * Run through list of hooks for input packets.  If there are any
359 	 * filters which require that additional packets in the flow are
360 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
361 	 * Note that filters must _never_ set this flag, as another filter
362 	 * in the list may have previously cleared it.
363 	 */
364 	m0 = m;
365 	pfh = pfil_hook_get(PFIL_IN, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
366 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
367 		if (pfh->pfil_func) {
368 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
369 					    m->m_pkthdr.rcvif, 0, &m0);
370 			if (rv)
371 				return;
372 			m = m0;
373 			if (m == NULL)
374 				return;
375 			ip6 = mtod(m, struct ip6_hdr *);
376 		}
377 #endif /* PFIL_HOOKS */
378 
379 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
380 
381 	/*
382 	 * Check with the firewall...
383 	 */
384 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
385 		u_short port = 0;
386 		/* If ipfw says divert, we have to just drop packet */
387 		/* use port as a dummy argument */
388 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
389 			m_freem(m);
390 			m = NULL;
391 		}
392 		if (!m)
393 			return;
394 	}
395 
396 	/*
397 	 * Check against address spoofing/corruption.
398 	 */
399 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
400 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
401 		/*
402 		 * XXX: "badscope" is not very suitable for a multicast source.
403 		 */
404 		ip6stat.ip6s_badscope++;
405 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
406 		goto bad;
407 	}
408 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
409 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
410 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
411 		ip6stat.ip6s_badscope++;
412 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
413 		goto bad;
414 	}
415 	/*
416 	 * The following check is not documented in specs.  A malicious
417 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
418 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
419 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
420 	 *
421 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
422 	 * support IPv4-less kernel compilation, we cannot support SIIT
423 	 * environment at all.  So, it makes more sense for us to reject any
424 	 * malicious packets for non-SIIT environment, than try to do a
425 	 * partical support for SIIT environment.
426 	 */
427 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
428 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
429 		ip6stat.ip6s_badscope++;
430 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
431 		goto bad;
432 	}
433 #if 0
434 	/*
435 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
436 	 *
437 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
438 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
439 	 * is revised to forbid relaying case.
440 	 */
441 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
442 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
443 		ip6stat.ip6s_badscope++;
444 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
445 		goto bad;
446 	}
447 #endif
448 
449 	/* drop packets if interface ID portion is already filled */
450 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
451 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
452 		    ip6->ip6_src.s6_addr16[1]) {
453 			ip6stat.ip6s_badscope++;
454 			goto bad;
455 		}
456 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
457 		    ip6->ip6_dst.s6_addr16[1]) {
458 			ip6stat.ip6s_badscope++;
459 			goto bad;
460 		}
461 	}
462 
463 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
464 		ip6->ip6_src.s6_addr16[1]
465 			= htons(m->m_pkthdr.rcvif->if_index);
466 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
467 		ip6->ip6_dst.s6_addr16[1]
468 			= htons(m->m_pkthdr.rcvif->if_index);
469 
470 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
471 	/*
472 	 * We use rt->rt_ifp to determine if the address is ours or not.
473 	 * If rt_ifp is lo0, the address is ours.
474 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
475 	 * so any address under fe80::%lo0/64 will be mistakenly considered
476 	 * local.  The special case is supplied to handle the case properly
477 	 * by actually looking at interface addresses
478 	 * (using in6ifa_ifpwithaddr).
479 	 */
480 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
481 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
482 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
483 			icmp6_error(m, ICMP6_DST_UNREACH,
484 			    ICMP6_DST_UNREACH_ADDR, 0);
485 			/* m is already freed */
486 			return;
487 		}
488 
489 		ours = 1;
490 		deliverifp = m->m_pkthdr.rcvif;
491 		goto hbhcheck;
492 	}
493 #endif
494 
495 	/*
496 	 * Multicast check
497 	 */
498 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
499 	  	struct	in6_multi *in6m = 0;
500 
501 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
502 		/*
503 		 * See if we belong to the destination multicast group on the
504 		 * arrival interface.
505 		 */
506 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
507 		if (in6m)
508 			ours = 1;
509 		else if (!ip6_mrouter) {
510 			ip6stat.ip6s_notmember++;
511 			ip6stat.ip6s_cantforward++;
512 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
513 			goto bad;
514 		}
515 		deliverifp = m->m_pkthdr.rcvif;
516 		goto hbhcheck;
517 	}
518 
519 	/*
520 	 *  Unicast check
521 	 */
522 	switch (ip6_ours_check_algorithm) {
523 	default:
524 		/*
525 		 * XXX: I intentionally broke our indentation rule here,
526 		 *      since this switch-case is just for measurement and
527 		 *      therefore should soon be removed.
528 		 */
529 	if (ip6_forward_rt.ro_rt != NULL &&
530 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
531 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
532 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
533 		ip6stat.ip6s_forward_cachehit++;
534 	else {
535 		struct sockaddr_in6 *dst6;
536 
537 		if (ip6_forward_rt.ro_rt) {
538 			/* route is down or destination is different */
539 			ip6stat.ip6s_forward_cachemiss++;
540 			RTFREE(ip6_forward_rt.ro_rt);
541 			ip6_forward_rt.ro_rt = 0;
542 		}
543 
544 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
545 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
546 		dst6->sin6_len = sizeof(struct sockaddr_in6);
547 		dst6->sin6_family = AF_INET6;
548 		dst6->sin6_addr = ip6->ip6_dst;
549 #ifdef SCOPEDROUTING
550 		ip6_forward_rt.ro_dst.sin6_scope_id =
551 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
552 #endif
553 
554 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
555 	}
556 
557 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
558 
559 	/*
560 	 * Accept the packet if the forwarding interface to the destination
561 	 * according to the routing table is the loopback interface,
562 	 * unless the associated route has a gateway.
563 	 * Note that this approach causes to accept a packet if there is a
564 	 * route to the loopback interface for the destination of the packet.
565 	 * But we think it's even useful in some situations, e.g. when using
566 	 * a special daemon which wants to intercept the packet.
567 	 *
568 	 * XXX: some OSes automatically make a cloned route for the destination
569 	 * of an outgoing packet.  If the outgoing interface of the packet
570 	 * is a loopback one, the kernel would consider the packet to be
571 	 * accepted, even if we have no such address assinged on the interface.
572 	 * We check the cloned flag of the route entry to reject such cases,
573 	 * assuming that route entries for our own addresses are not made by
574 	 * cloning (it should be true because in6_addloop explicitly installs
575 	 * the host route).  However, we might have to do an explicit check
576 	 * while it would be less efficient.  Or, should we rather install a
577 	 * reject route for such a case?
578 	 */
579 	if (ip6_forward_rt.ro_rt &&
580 	    (ip6_forward_rt.ro_rt->rt_flags &
581 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
582 #ifdef RTF_WASCLONED
583 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
584 #endif
585 #ifdef RTF_CLONED
586 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
587 #endif
588 #if 0
589 	    /*
590 	     * The check below is redundant since the comparison of
591 	     * the destination and the key of the rtentry has
592 	     * already done through looking up the routing table.
593 	     */
594 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
595 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
596 #endif
597 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
598 		struct in6_ifaddr *ia6 =
599 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
600 
601 		/*
602 		 * record address information into m_aux.
603 		 */
604 		(void)ip6_setdstifaddr(m, ia6);
605 
606 		/*
607 		 * packets to a tentative, duplicated, or somehow invalid
608 		 * address must not be accepted.
609 		 */
610 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
611 			/* this address is ready */
612 			ours = 1;
613 			deliverifp = ia6->ia_ifp;	/* correct? */
614 			/* Count the packet in the ip address stats */
615 			ia6->ia_ifa.if_ipackets++;
616 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
617 			goto hbhcheck;
618 		} else {
619 			/* address is not ready, so discard the packet. */
620 			nd6log((LOG_INFO,
621 			    "ip6_input: packet to an unready address %s->%s\n",
622 			    ip6_sprintf(&ip6->ip6_src),
623 			    ip6_sprintf(&ip6->ip6_dst)));
624 
625 			goto bad;
626 		}
627 	}
628 	} /* XXX indentation (see above) */
629 
630 	/*
631 	 * FAITH(Firewall Aided Internet Translator)
632 	 */
633 	if (ip6_keepfaith) {
634 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
635 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
636 			/* XXX do we need more sanity checks? */
637 			ours = 1;
638 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
639 			goto hbhcheck;
640 		}
641 	}
642 
643 	/*
644 	 * Now there is no reason to process the packet if it's not our own
645 	 * and we're not a router.
646 	 */
647 	if (!ip6_forwarding) {
648 		ip6stat.ip6s_cantforward++;
649 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
650 		goto bad;
651 	}
652 
653   hbhcheck:
654 	/*
655 	 * record address information into m_aux, if we don't have one yet.
656 	 * note that we are unable to record it, if the address is not listed
657 	 * as our interface address (e.g. multicast addresses, addresses
658 	 * within FAITH prefixes and such).
659 	 */
660 	if (deliverifp && !ip6_getdstifaddr(m)) {
661 		struct in6_ifaddr *ia6;
662 
663 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
664 		if (ia6) {
665 			if (!ip6_setdstifaddr(m, ia6)) {
666 				/*
667 				 * XXX maybe we should drop the packet here,
668 				 * as we could not provide enough information
669 				 * to the upper layers.
670 				 */
671 			}
672 		}
673 	}
674 
675 	/*
676 	 * Process Hop-by-Hop options header if it's contained.
677 	 * m may be modified in ip6_hopopts_input().
678 	 * If a JumboPayload option is included, plen will also be modified.
679 	 */
680 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
681 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
682 		struct ip6_hbh *hbh;
683 
684 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
685 #if 0	/*touches NULL pointer*/
686 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
687 #endif
688 			return;	/* m have already been freed */
689 		}
690 
691 		/* adjust pointer */
692 		ip6 = mtod(m, struct ip6_hdr *);
693 
694 		/*
695 		 * if the payload length field is 0 and the next header field
696 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
697 		 * option MUST be included.
698 		 */
699 		if (ip6->ip6_plen == 0 && plen == 0) {
700 			/*
701 			 * Note that if a valid jumbo payload option is
702 			 * contained, ip6_hoptops_input() must set a valid
703 			 * (non-zero) payload length to the variable plen.
704 			 */
705 			ip6stat.ip6s_badoptions++;
706 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
707 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
708 			icmp6_error(m, ICMP6_PARAM_PROB,
709 				    ICMP6_PARAMPROB_HEADER,
710 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
711 			return;
712 		}
713 #ifndef PULLDOWN_TEST
714 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
715 		hbh = (struct ip6_hbh *)(ip6 + 1);
716 #else
717 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
718 			sizeof(struct ip6_hbh));
719 		if (hbh == NULL) {
720 			ip6stat.ip6s_tooshort++;
721 			return;
722 		}
723 #endif
724 		nxt = hbh->ip6h_nxt;
725 
726 		/*
727 		 * accept the packet if a router alert option is included
728 		 * and we act as an IPv6 router.
729 		 */
730 		if (rtalert != ~0 && ip6_forwarding)
731 			ours = 1;
732 	} else
733 		nxt = ip6->ip6_nxt;
734 
735 	/*
736 	 * Check that the amount of data in the buffers
737 	 * is as at least much as the IPv6 header would have us expect.
738 	 * Trim mbufs if longer than we expect.
739 	 * Drop packet if shorter than we expect.
740 	 */
741 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
742 		ip6stat.ip6s_tooshort++;
743 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
744 		goto bad;
745 	}
746 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
747 		if (m->m_len == m->m_pkthdr.len) {
748 			m->m_len = sizeof(struct ip6_hdr) + plen;
749 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
750 		} else
751 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
752 	}
753 
754 	/*
755 	 * Forward if desirable.
756 	 */
757 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
758 		/*
759 		 * If we are acting as a multicast router, all
760 		 * incoming multicast packets are passed to the
761 		 * kernel-level multicast forwarding function.
762 		 * The packet is returned (relatively) intact; if
763 		 * ip6_mforward() returns a non-zero value, the packet
764 		 * must be discarded, else it may be accepted below.
765 		 */
766 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
767 			ip6stat.ip6s_cantforward++;
768 			m_freem(m);
769 			return;
770 		}
771 		if (!ours) {
772 			m_freem(m);
773 			return;
774 		}
775 	} else if (!ours) {
776 		ip6_forward(m, 0);
777 		return;
778 	}
779 
780 	ip6 = mtod(m, struct ip6_hdr *);
781 
782 	/*
783 	 * Malicious party may be able to use IPv4 mapped addr to confuse
784 	 * tcp/udp stack and bypass security checks (act as if it was from
785 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
786 	 *
787 	 * For SIIT end node behavior, you may want to disable the check.
788 	 * However, you will  become vulnerable to attacks using IPv4 mapped
789 	 * source.
790 	 */
791 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
792 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
793 		ip6stat.ip6s_badscope++;
794 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
795 		goto bad;
796 	}
797 
798 	/*
799 	 * Tell launch routine the next header
800 	 */
801 	ip6stat.ip6s_delivered++;
802 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
803 	nest = 0;
804 
805 	while (nxt != IPPROTO_DONE) {
806 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
807 			ip6stat.ip6s_toomanyhdr++;
808 			goto bad;
809 		}
810 
811 		/*
812 		 * protection against faulty packet - there should be
813 		 * more sanity checks in header chain processing.
814 		 */
815 		if (m->m_pkthdr.len < off) {
816 			ip6stat.ip6s_tooshort++;
817 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
818 			goto bad;
819 		}
820 
821 #if 0
822 		/*
823 		 * do we need to do it for every header?  yeah, other
824 		 * functions can play with it (like re-allocate and copy).
825 		 */
826 		mhist = ip6_addaux(m);
827 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
828 			hist = mtod(mhist, caddr_t) + mhist->m_len;
829 			bcopy(&nxt, hist, sizeof(nxt));
830 			mhist->m_len += sizeof(nxt);
831 		} else {
832 			ip6stat.ip6s_toomanyhdr++;
833 			goto bad;
834 		}
835 #endif
836 
837 #ifdef IPSEC
838 		/*
839 		 * enforce IPsec policy checking if we are seeing last header.
840 		 * note that we do not visit this with protocols with pcb layer
841 		 * code - like udp/tcp/raw ip.
842 		 */
843 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
844 		    ipsec6_in_reject(m, NULL)) {
845 			ipsec6stat.in_polvio++;
846 			goto bad;
847 		}
848 #endif
849 
850 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
851 	}
852 	return;
853  bad:
854 	m_freem(m);
855 }
856 
857 /*
858  * set/grab in6_ifaddr correspond to IPv6 destination address.
859  * XXX backward compatibility wrapper
860  */
861 static struct mbuf *
862 ip6_setdstifaddr(m, ia6)
863 	struct mbuf *m;
864 	struct in6_ifaddr *ia6;
865 {
866 	struct mbuf *n;
867 
868 	n = ip6_addaux(m);
869 	if (n)
870 		mtod(n, struct ip6aux *)->ip6a_dstia6 = ia6;
871 	return n;	/* NULL if failed to set */
872 }
873 
874 struct in6_ifaddr *
875 ip6_getdstifaddr(m)
876 	struct mbuf *m;
877 {
878 	struct mbuf *n;
879 
880 	n = ip6_findaux(m);
881 	if (n)
882 		return mtod(n, struct ip6aux *)->ip6a_dstia6;
883 	else
884 		return NULL;
885 }
886 
887 /*
888  * Hop-by-Hop options header processing. If a valid jumbo payload option is
889  * included, the real payload length will be stored in plenp.
890  */
891 static int
892 ip6_hopopts_input(plenp, rtalertp, mp, offp)
893 	u_int32_t *plenp;
894 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
895 	struct mbuf **mp;
896 	int *offp;
897 {
898 	struct mbuf *m = *mp;
899 	int off = *offp, hbhlen;
900 	struct ip6_hbh *hbh;
901 	u_int8_t *opt;
902 
903 	/* validation of the length of the header */
904 #ifndef PULLDOWN_TEST
905 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
906 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
907 	hbhlen = (hbh->ip6h_len + 1) << 3;
908 
909 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
910 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
911 #else
912 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
913 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
914 	if (hbh == NULL) {
915 		ip6stat.ip6s_tooshort++;
916 		return -1;
917 	}
918 	hbhlen = (hbh->ip6h_len + 1) << 3;
919 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
920 		hbhlen);
921 	if (hbh == NULL) {
922 		ip6stat.ip6s_tooshort++;
923 		return -1;
924 	}
925 #endif
926 	off += hbhlen;
927 	hbhlen -= sizeof(struct ip6_hbh);
928 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
929 
930 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
931 				hbhlen, rtalertp, plenp) < 0)
932 		return(-1);
933 
934 	*offp = off;
935 	*mp = m;
936 	return(0);
937 }
938 
939 /*
940  * Search header for all Hop-by-hop options and process each option.
941  * This function is separate from ip6_hopopts_input() in order to
942  * handle a case where the sending node itself process its hop-by-hop
943  * options header. In such a case, the function is called from ip6_output().
944  *
945  * The function assumes that hbh header is located right after the IPv6 header
946  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
947  * opthead + hbhlen is located in continuous memory region.
948  */
949 int
950 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
951 	struct mbuf *m;
952 	u_int8_t *opthead;
953 	int hbhlen;
954 	u_int32_t *rtalertp;
955 	u_int32_t *plenp;
956 {
957 	struct ip6_hdr *ip6;
958 	int optlen = 0;
959 	u_int8_t *opt = opthead;
960 	u_int16_t rtalert_val;
961 	u_int32_t jumboplen;
962 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
963 
964 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
965 		switch (*opt) {
966 		case IP6OPT_PAD1:
967 			optlen = 1;
968 			break;
969 		case IP6OPT_PADN:
970 			if (hbhlen < IP6OPT_MINLEN) {
971 				ip6stat.ip6s_toosmall++;
972 				goto bad;
973 			}
974 			optlen = *(opt + 1) + 2;
975 			break;
976 		case IP6OPT_RTALERT:
977 			/* XXX may need check for alignment */
978 			if (hbhlen < IP6OPT_RTALERT_LEN) {
979 				ip6stat.ip6s_toosmall++;
980 				goto bad;
981 			}
982 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
983 				/* XXX stat */
984 				icmp6_error(m, ICMP6_PARAM_PROB,
985 					    ICMP6_PARAMPROB_HEADER,
986 					    erroff + opt + 1 - opthead);
987 				return(-1);
988 			}
989 			optlen = IP6OPT_RTALERT_LEN;
990 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
991 			*rtalertp = ntohs(rtalert_val);
992 			break;
993 		case IP6OPT_JUMBO:
994 			/* XXX may need check for alignment */
995 			if (hbhlen < IP6OPT_JUMBO_LEN) {
996 				ip6stat.ip6s_toosmall++;
997 				goto bad;
998 			}
999 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1000 				/* XXX stat */
1001 				icmp6_error(m, ICMP6_PARAM_PROB,
1002 					    ICMP6_PARAMPROB_HEADER,
1003 					    erroff + opt + 1 - opthead);
1004 				return(-1);
1005 			}
1006 			optlen = IP6OPT_JUMBO_LEN;
1007 
1008 			/*
1009 			 * IPv6 packets that have non 0 payload length
1010 			 * must not contain a jumbo payload option.
1011 			 */
1012 			ip6 = mtod(m, struct ip6_hdr *);
1013 			if (ip6->ip6_plen) {
1014 				ip6stat.ip6s_badoptions++;
1015 				icmp6_error(m, ICMP6_PARAM_PROB,
1016 					    ICMP6_PARAMPROB_HEADER,
1017 					    erroff + opt - opthead);
1018 				return(-1);
1019 			}
1020 
1021 			/*
1022 			 * We may see jumbolen in unaligned location, so
1023 			 * we'd need to perform bcopy().
1024 			 */
1025 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1026 			jumboplen = (u_int32_t)htonl(jumboplen);
1027 
1028 #if 1
1029 			/*
1030 			 * if there are multiple jumbo payload options,
1031 			 * *plenp will be non-zero and the packet will be
1032 			 * rejected.
1033 			 * the behavior may need some debate in ipngwg -
1034 			 * multiple options does not make sense, however,
1035 			 * there's no explicit mention in specification.
1036 			 */
1037 			if (*plenp != 0) {
1038 				ip6stat.ip6s_badoptions++;
1039 				icmp6_error(m, ICMP6_PARAM_PROB,
1040 					    ICMP6_PARAMPROB_HEADER,
1041 					    erroff + opt + 2 - opthead);
1042 				return(-1);
1043 			}
1044 #endif
1045 
1046 			/*
1047 			 * jumbo payload length must be larger than 65535.
1048 			 */
1049 			if (jumboplen <= IPV6_MAXPACKET) {
1050 				ip6stat.ip6s_badoptions++;
1051 				icmp6_error(m, ICMP6_PARAM_PROB,
1052 					    ICMP6_PARAMPROB_HEADER,
1053 					    erroff + opt + 2 - opthead);
1054 				return(-1);
1055 			}
1056 			*plenp = jumboplen;
1057 
1058 			break;
1059 		default:		/* unknown option */
1060 			if (hbhlen < IP6OPT_MINLEN) {
1061 				ip6stat.ip6s_toosmall++;
1062 				goto bad;
1063 			}
1064 			optlen = ip6_unknown_opt(opt, m,
1065 			    erroff + opt - opthead);
1066 			if (optlen == -1)
1067 				return(-1);
1068 			optlen += 2;
1069 			break;
1070 		}
1071 	}
1072 
1073 	return(0);
1074 
1075   bad:
1076 	m_freem(m);
1077 	return(-1);
1078 }
1079 
1080 /*
1081  * Unknown option processing.
1082  * The third argument `off' is the offset from the IPv6 header to the option,
1083  * which is necessary if the IPv6 header the and option header and IPv6 header
1084  * is not continuous in order to return an ICMPv6 error.
1085  */
1086 int
1087 ip6_unknown_opt(optp, m, off)
1088 	u_int8_t *optp;
1089 	struct mbuf *m;
1090 	int off;
1091 {
1092 	struct ip6_hdr *ip6;
1093 
1094 	switch (IP6OPT_TYPE(*optp)) {
1095 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1096 		return((int)*(optp + 1));
1097 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1098 		m_freem(m);
1099 		return(-1);
1100 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1101 		ip6stat.ip6s_badoptions++;
1102 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1103 		return(-1);
1104 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1105 		ip6stat.ip6s_badoptions++;
1106 		ip6 = mtod(m, struct ip6_hdr *);
1107 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1108 		    (m->m_flags & (M_BCAST|M_MCAST)))
1109 			m_freem(m);
1110 		else
1111 			icmp6_error(m, ICMP6_PARAM_PROB,
1112 				    ICMP6_PARAMPROB_OPTION, off);
1113 		return(-1);
1114 	}
1115 
1116 	m_freem(m);		/* XXX: NOTREACHED */
1117 	return(-1);
1118 }
1119 
1120 /*
1121  * Create the "control" list for this pcb.
1122  * The function will not modify mbuf chain at all.
1123  *
1124  * with KAME mbuf chain restriction:
1125  * The routine will be called from upper layer handlers like tcp6_input().
1126  * Thus the routine assumes that the caller (tcp6_input) have already
1127  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1128  * very first mbuf on the mbuf chain.
1129  */
1130 void
1131 ip6_savecontrol(in6p, mp, ip6, m)
1132 	struct inpcb *in6p;
1133 	struct mbuf **mp;
1134 	struct ip6_hdr *ip6;
1135 	struct mbuf *m;
1136 {
1137 	struct proc *p = curproc;	/* XXX */
1138 	int privileged = 0;
1139 	int rthdr_exist = 0;
1140 
1141 
1142 	if (p && !suser(p))
1143  		privileged++;
1144 
1145 #ifdef SO_TIMESTAMP
1146 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1147 		struct timeval tv;
1148 
1149 		microtime(&tv);
1150 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1151 				      SCM_TIMESTAMP, SOL_SOCKET);
1152 		if (*mp) {
1153 			mp = &(*mp)->m_next;
1154 		}
1155 	}
1156 #endif
1157 
1158 	/* RFC 2292 sec. 5 */
1159 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1160 		struct in6_pktinfo pi6;
1161 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1162 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1163 			pi6.ipi6_addr.s6_addr16[1] = 0;
1164 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1165 					? m->m_pkthdr.rcvif->if_index
1166 					: 0;
1167 		*mp = sbcreatecontrol((caddr_t) &pi6,
1168 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1169 			IPPROTO_IPV6);
1170 		if (*mp)
1171 			mp = &(*mp)->m_next;
1172 	}
1173 
1174 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1175 		int hlim = ip6->ip6_hlim & 0xff;
1176 		*mp = sbcreatecontrol((caddr_t) &hlim,
1177 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1178 		if (*mp)
1179 			mp = &(*mp)->m_next;
1180 	}
1181 
1182 	/*
1183 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1184 	 * for the option, but it might be too strict, since there might
1185 	 * be some hop-by-hop options which can be returned to normal user.
1186 	 * See RFC 2292 section 6.
1187 	 */
1188 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1189 		/*
1190 		 * Check if a hop-by-hop options header is contatined in the
1191 		 * received packet, and if so, store the options as ancillary
1192 		 * data. Note that a hop-by-hop options header must be
1193 		 * just after the IPv6 header, which fact is assured through
1194 		 * the IPv6 input processing.
1195 		 */
1196 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1197 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1198 			struct ip6_hbh *hbh;
1199 			int hbhlen = 0;
1200 #ifdef PULLDOWN_TEST
1201 			struct mbuf *ext;
1202 #endif
1203 
1204 #ifndef PULLDOWN_TEST
1205 			hbh = (struct ip6_hbh *)(ip6 + 1);
1206 			hbhlen = (hbh->ip6h_len + 1) << 3;
1207 #else
1208 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1209 			    ip6->ip6_nxt);
1210 			if (ext == NULL) {
1211 				ip6stat.ip6s_tooshort++;
1212 				return;
1213 			}
1214 			hbh = mtod(ext, struct ip6_hbh *);
1215 			hbhlen = (hbh->ip6h_len + 1) << 3;
1216 			if (hbhlen != ext->m_len) {
1217 				m_freem(ext);
1218 				ip6stat.ip6s_tooshort++;
1219 				return;
1220 			}
1221 #endif
1222 
1223 			/*
1224 			 * XXX: We copy whole the header even if a jumbo
1225 			 * payload option is included, which option is to
1226 			 * be removed before returning in the RFC 2292.
1227 			 * Note: this constraint is removed in 2292bis.
1228 			 */
1229 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1230 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1231 			if (*mp)
1232 				mp = &(*mp)->m_next;
1233 #ifdef PULLDOWN_TEST
1234 			m_freem(ext);
1235 #endif
1236 		}
1237 	}
1238 
1239 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1240 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1241 		int proto, off, nxt;
1242 
1243 		/*
1244 		 * go through the header chain to see if a routing header is
1245 		 * contained in the packet. We need this information to store
1246 		 * destination options headers (if any) properly.
1247 		 * XXX: performance issue. We should record this info when
1248 		 * processing extension headers in incoming routine.
1249 		 * (todo) use m_aux?
1250 		 */
1251 		proto = IPPROTO_IPV6;
1252 		off = 0;
1253 		nxt = -1;
1254 		while (1) {
1255 			int newoff;
1256 
1257 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1258 			if (newoff < 0)
1259 				break;
1260 			if (newoff < off) /* invalid, check for safety */
1261 				break;
1262 			if ((proto = nxt) == IPPROTO_ROUTING) {
1263 				rthdr_exist = 1;
1264 				break;
1265 			}
1266 			off = newoff;
1267 		}
1268 	}
1269 
1270 	if ((in6p->in6p_flags &
1271 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1272 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1273 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1274 
1275 		/*
1276 		 * Search for destination options headers or routing
1277 		 * header(s) through the header chain, and stores each
1278 		 * header as ancillary data.
1279 		 * Note that the order of the headers remains in
1280 		 * the chain of ancillary data.
1281 		 */
1282 		while (1) {	/* is explicit loop prevention necessary? */
1283 			struct ip6_ext *ip6e = NULL;
1284 			int elen;
1285 #ifdef PULLDOWN_TEST
1286 			struct mbuf *ext = NULL;
1287 #endif
1288 
1289 			/*
1290 			 * if it is not an extension header, don't try to
1291 			 * pull it from the chain.
1292 			 */
1293 			switch (nxt) {
1294 			case IPPROTO_DSTOPTS:
1295 			case IPPROTO_ROUTING:
1296 			case IPPROTO_HOPOPTS:
1297 			case IPPROTO_AH: /* is it possible? */
1298 				break;
1299 			default:
1300 				goto loopend;
1301 			}
1302 
1303 #ifndef PULLDOWN_TEST
1304 			if (off + sizeof(*ip6e) > m->m_len)
1305 				goto loopend;
1306 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1307 			if (nxt == IPPROTO_AH)
1308 				elen = (ip6e->ip6e_len + 2) << 2;
1309 			else
1310 				elen = (ip6e->ip6e_len + 1) << 3;
1311 			if (off + elen > m->m_len)
1312 				goto loopend;
1313 #else
1314 			ext = ip6_pullexthdr(m, off, nxt);
1315 			if (ext == NULL) {
1316 				ip6stat.ip6s_tooshort++;
1317 				return;
1318 			}
1319 			ip6e = mtod(ext, struct ip6_ext *);
1320 			if (nxt == IPPROTO_AH)
1321 				elen = (ip6e->ip6e_len + 2) << 2;
1322 			else
1323 				elen = (ip6e->ip6e_len + 1) << 3;
1324 			if (elen != ext->m_len) {
1325 				m_freem(ext);
1326 				ip6stat.ip6s_tooshort++;
1327 				return;
1328 			}
1329 #endif
1330 
1331 			switch (nxt) {
1332 			case IPPROTO_DSTOPTS:
1333 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1334 					break;
1335 
1336 				/*
1337 				 * We also require super-user privilege for
1338 				 * the option.
1339 				 * See the comments on IN6_HOPOPTS.
1340 				 */
1341 				if (!privileged)
1342 					break;
1343 
1344 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1345 						      IPV6_DSTOPTS,
1346 						      IPPROTO_IPV6);
1347 				if (*mp)
1348 					mp = &(*mp)->m_next;
1349 				break;
1350 			case IPPROTO_ROUTING:
1351 				if (!in6p->in6p_flags & IN6P_RTHDR)
1352 					break;
1353 
1354 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1355 						      IPV6_RTHDR,
1356 						      IPPROTO_IPV6);
1357 				if (*mp)
1358 					mp = &(*mp)->m_next;
1359 				break;
1360 			case IPPROTO_HOPOPTS:
1361 			case IPPROTO_AH: /* is it possible? */
1362 				break;
1363 
1364 			default:
1365 				/*
1366 			 	 * other cases have been filtered in the above.
1367 				 * none will visit this case.  here we supply
1368 				 * the code just in case (nxt overwritten or
1369 				 * other cases).
1370 				 */
1371 #ifdef PULLDOWN_TEST
1372 				m_freem(ext);
1373 #endif
1374 				goto loopend;
1375 
1376 			}
1377 
1378 			/* proceed with the next header. */
1379 			off += elen;
1380 			nxt = ip6e->ip6e_nxt;
1381 			ip6e = NULL;
1382 #ifdef PULLDOWN_TEST
1383 			m_freem(ext);
1384 			ext = NULL;
1385 #endif
1386 		}
1387 	  loopend:
1388 		;
1389 	}
1390 
1391 }
1392 
1393 #ifdef PULLDOWN_TEST
1394 /*
1395  * pull single extension header from mbuf chain.  returns single mbuf that
1396  * contains the result, or NULL on error.
1397  */
1398 static struct mbuf *
1399 ip6_pullexthdr(m, off, nxt)
1400 	struct mbuf *m;
1401 	size_t off;
1402 	int nxt;
1403 {
1404 	struct ip6_ext ip6e;
1405 	size_t elen;
1406 	struct mbuf *n;
1407 
1408 #ifdef DIAGNOSTIC
1409 	switch (nxt) {
1410 	case IPPROTO_DSTOPTS:
1411 	case IPPROTO_ROUTING:
1412 	case IPPROTO_HOPOPTS:
1413 	case IPPROTO_AH: /* is it possible? */
1414 		break;
1415 	default:
1416 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1417 	}
1418 #endif
1419 
1420 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1421 	if (nxt == IPPROTO_AH)
1422 		elen = (ip6e.ip6e_len + 2) << 2;
1423 	else
1424 		elen = (ip6e.ip6e_len + 1) << 3;
1425 
1426 	MGET(n, M_DONTWAIT, MT_DATA);
1427 	if (n && elen >= MLEN) {
1428 		MCLGET(n, M_DONTWAIT);
1429 		if ((n->m_flags & M_EXT) == 0) {
1430 			m_free(n);
1431 			n = NULL;
1432 		}
1433 	}
1434 	if (!n)
1435 		return NULL;
1436 
1437 	n->m_len = 0;
1438 	if (elen >= M_TRAILINGSPACE(n)) {
1439 		m_free(n);
1440 		return NULL;
1441 	}
1442 
1443 	m_copydata(m, off, elen, mtod(n, caddr_t));
1444 	n->m_len = elen;
1445 	return n;
1446 }
1447 #endif
1448 
1449 /*
1450  * Get pointer to the previous header followed by the header
1451  * currently processed.
1452  * XXX: This function supposes that
1453  *	M includes all headers,
1454  *	the next header field and the header length field of each header
1455  *	are valid, and
1456  *	the sum of each header length equals to OFF.
1457  * Because of these assumptions, this function must be called very
1458  * carefully. Moreover, it will not be used in the near future when
1459  * we develop `neater' mechanism to process extension headers.
1460  */
1461 char *
1462 ip6_get_prevhdr(m, off)
1463 	struct mbuf *m;
1464 	int off;
1465 {
1466 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1467 
1468 	if (off == sizeof(struct ip6_hdr))
1469 		return(&ip6->ip6_nxt);
1470 	else {
1471 		int len, nxt;
1472 		struct ip6_ext *ip6e = NULL;
1473 
1474 		nxt = ip6->ip6_nxt;
1475 		len = sizeof(struct ip6_hdr);
1476 		while (len < off) {
1477 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1478 
1479 			switch (nxt) {
1480 			case IPPROTO_FRAGMENT:
1481 				len += sizeof(struct ip6_frag);
1482 				break;
1483 			case IPPROTO_AH:
1484 				len += (ip6e->ip6e_len + 2) << 2;
1485 				break;
1486 			default:
1487 				len += (ip6e->ip6e_len + 1) << 3;
1488 				break;
1489 			}
1490 			nxt = ip6e->ip6e_nxt;
1491 		}
1492 		if (ip6e)
1493 			return(&ip6e->ip6e_nxt);
1494 		else
1495 			return NULL;
1496 	}
1497 }
1498 
1499 /*
1500  * get next header offset.  m will be retained.
1501  */
1502 int
1503 ip6_nexthdr(m, off, proto, nxtp)
1504 	struct mbuf *m;
1505 	int off;
1506 	int proto;
1507 	int *nxtp;
1508 {
1509 	struct ip6_hdr ip6;
1510 	struct ip6_ext ip6e;
1511 	struct ip6_frag fh;
1512 
1513 	/* just in case */
1514 	if (m == NULL)
1515 		panic("ip6_nexthdr: m == NULL");
1516 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1517 		return -1;
1518 
1519 	switch (proto) {
1520 	case IPPROTO_IPV6:
1521 		if (m->m_pkthdr.len < off + sizeof(ip6))
1522 			return -1;
1523 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1524 		if (nxtp)
1525 			*nxtp = ip6.ip6_nxt;
1526 		off += sizeof(ip6);
1527 		return off;
1528 
1529 	case IPPROTO_FRAGMENT:
1530 		/*
1531 		 * terminate parsing if it is not the first fragment,
1532 		 * it does not make sense to parse through it.
1533 		 */
1534 		if (m->m_pkthdr.len < off + sizeof(fh))
1535 			return -1;
1536 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1537 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1538 			return -1;
1539 		if (nxtp)
1540 			*nxtp = fh.ip6f_nxt;
1541 		off += sizeof(struct ip6_frag);
1542 		return off;
1543 
1544 	case IPPROTO_AH:
1545 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1546 			return -1;
1547 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1548 		if (nxtp)
1549 			*nxtp = ip6e.ip6e_nxt;
1550 		off += (ip6e.ip6e_len + 2) << 2;
1551 		return off;
1552 
1553 	case IPPROTO_HOPOPTS:
1554 	case IPPROTO_ROUTING:
1555 	case IPPROTO_DSTOPTS:
1556 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1557 			return -1;
1558 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1559 		if (nxtp)
1560 			*nxtp = ip6e.ip6e_nxt;
1561 		off += (ip6e.ip6e_len + 1) << 3;
1562 		return off;
1563 
1564 	case IPPROTO_NONE:
1565 	case IPPROTO_ESP:
1566 	case IPPROTO_IPCOMP:
1567 		/* give up */
1568 		return -1;
1569 
1570 	default:
1571 		return -1;
1572 	}
1573 
1574 	return -1;
1575 }
1576 
1577 /*
1578  * get offset for the last header in the chain.  m will be kept untainted.
1579  */
1580 int
1581 ip6_lasthdr(m, off, proto, nxtp)
1582 	struct mbuf *m;
1583 	int off;
1584 	int proto;
1585 	int *nxtp;
1586 {
1587 	int newoff;
1588 	int nxt;
1589 
1590 	if (!nxtp) {
1591 		nxt = -1;
1592 		nxtp = &nxt;
1593 	}
1594 	while (1) {
1595 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1596 		if (newoff < 0)
1597 			return off;
1598 		else if (newoff < off)
1599 			return -1;	/* invalid */
1600 		else if (newoff == off)
1601 			return newoff;
1602 
1603 		off = newoff;
1604 		proto = *nxtp;
1605 	}
1606 }
1607 
1608 struct mbuf *
1609 ip6_addaux(m)
1610 	struct mbuf *m;
1611 {
1612 	struct mbuf *n;
1613 
1614 #ifdef DIAGNOSTIC
1615 	if (sizeof(struct ip6aux) > MHLEN)
1616 		panic("assumption failed on sizeof(ip6aux)");
1617 #endif
1618 	n = m_aux_find(m, AF_INET6, -1);
1619 	if (n) {
1620 		if (n->m_len < sizeof(struct ip6aux)) {
1621 			printf("conflicting use of ip6aux");
1622 			return NULL;
1623 		}
1624 	} else {
1625 		n = m_aux_add(m, AF_INET6, -1);
1626 		n->m_len = sizeof(struct ip6aux);
1627 		bzero(mtod(n, caddr_t), n->m_len);
1628 	}
1629 	return n;
1630 }
1631 
1632 struct mbuf *
1633 ip6_findaux(m)
1634 	struct mbuf *m;
1635 {
1636 	struct mbuf *n;
1637 
1638 	n = m_aux_find(m, AF_INET6, -1);
1639 	if (n && n->m_len < sizeof(struct ip6aux)) {
1640 		printf("conflicting use of ip6aux");
1641 		n = NULL;
1642 	}
1643 	return n;
1644 }
1645 
1646 void
1647 ip6_delaux(m)
1648 	struct mbuf *m;
1649 {
1650 	struct mbuf *n;
1651 
1652 	n = m_aux_find(m, AF_INET6, -1);
1653 	if (n)
1654 		m_aux_delete(m, n);
1655 }
1656 
1657 /*
1658  * System control for IP6
1659  */
1660 
1661 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1662 	0,		0,		0,		0,
1663 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1664 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1665 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1666 	0,		0,		0,		0,
1667 	ENOPROTOOPT
1668 };
1669