xref: /dragonfly/sys/netinet6/ip6_input.c (revision 9dbf638f)
1 /*	$FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.15 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/ip6_input.c,v 1.9 2003/09/16 01:58:00 hsu Exp $	*/
3 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  */
68 
69 #include "opt_ip6fw.h"
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.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 #include <sys/proc.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 
95 #include <netinet/in.h>
96 #include <netinet/in_systm.h>
97 #ifdef INET
98 #include <netinet/ip.h>
99 #include <netinet/ip_icmp.h>
100 #endif /* INET */
101 #include <netinet/ip6.h>
102 #include <netinet6/in6_var.h>
103 #include <netinet6/ip6_var.h>
104 #include <netinet/in_pcb.h>
105 #include <netinet/icmp6.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_prefix.h>
109 
110 #ifdef IPSEC
111 #include <netinet6/ipsec.h>
112 #ifdef INET6
113 #include <netinet6/ipsec6.h>
114 #endif
115 #endif
116 
117 #ifdef FAST_IPSEC
118 #include "ipsec.h"
119 #include "ipsec6.h"
120 #define	IPSEC
121 #endif /* FAST_IPSEC */
122 
123 #include <net/ip6fw/ip6_fw.h>
124 
125 #include <netinet6/ip6protosw.h>
126 
127 /* we need it for NLOOP. */
128 #include "use_loop.h"
129 
130 #include <net/net_osdep.h>
131 
132 extern int (*fr_checkp) (struct ip *, int, struct ifnet *, int, struct mbuf **);
133 
134 extern struct domain inet6domain;
135 extern struct ip6protosw inet6sw[];
136 
137 u_char ip6_protox[IPPROTO_MAX];
138 static struct ifqueue ip6intrq;
139 static int ip6qmaxlen = IFQ_MAXLEN;
140 struct in6_ifaddr *in6_ifaddr;
141 
142 extern struct callout in6_tmpaddrtimer_ch;
143 
144 int ip6_forward_srcrt;			/* XXX */
145 int ip6_sourcecheck;			/* XXX */
146 int ip6_sourcecheck_interval;		/* XXX */
147 const int int6intrq_present = 1;
148 
149 int ip6_ours_check_algorithm;
150 
151 /* firewall hooks */
152 ip6_fw_chk_t *ip6_fw_chk_ptr;
153 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
154 int ip6_fw_enable = 1;
155 
156 struct ip6stat ip6stat;
157 
158 static void ip6_init2 (void *);
159 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
160 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
161 #ifdef PULLDOWN_TEST
162 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
163 #endif
164 
165 /*
166  * IP6 initialization: fill in IP6 protocol switch table.
167  * All protocols not implemented in kernel go to raw IP6 protocol handler.
168  */
169 void
170 ip6_init()
171 {
172 	struct ip6protosw *pr;
173 	int i;
174 	struct timeval tv;
175 
176 #ifdef DIAGNOSTIC
177 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
178 		panic("sizeof(protosw) != sizeof(ip6protosw)");
179 #endif
180 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
181 	if (pr == 0)
182 		panic("ip6_init");
183 	for (i = 0; i < IPPROTO_MAX; i++)
184 		ip6_protox[i] = pr - inet6sw;
185 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
186 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
187 		if (pr->pr_domain->dom_family == PF_INET6 &&
188 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
189 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
190 	ip6intrq.ifq_maxlen = ip6qmaxlen;
191 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq);
192 	nd6_init();
193 	frag6_init();
194 	/*
195 	 * in many cases, random() here does NOT return random number
196 	 * as initialization during bootstrap time occur in fixed order.
197 	 */
198 	microtime(&tv);
199 	ip6_flow_seq = random() ^ tv.tv_usec;
200 	microtime(&tv);
201 	ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
202 }
203 
204 static void
205 ip6_init2(dummy)
206 	void *dummy;
207 {
208 
209 	/*
210 	 * to route local address of p2p link to loopback,
211 	 * assign loopback address first.
212 	 */
213 	in6_ifattach(&loif[0], NULL);
214 
215 	/* nd6_timer_init */
216 	callout_init(&nd6_timer_ch);
217 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
218 
219 	/* router renumbering prefix list maintenance */
220 	callout_init(&in6_rr_timer_ch);
221 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
222 
223 	/* timer for regeneranation of temporary addresses randomize ID */
224 	callout_reset(&in6_tmpaddrtimer_ch,
225 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
226 		       ip6_temp_regen_advance) * hz,
227 		      in6_tmpaddrtimer, NULL);
228 }
229 
230 /* cheat */
231 /* This must be after route_init(), which is now SI_ORDER_THIRD */
232 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
233 
234 extern struct	route_in6 ip6_forward_rt;
235 
236 void
237 ip6_input(m)
238 	struct mbuf *m;
239 {
240 	struct ip6_hdr *ip6;
241 	int off = sizeof(struct ip6_hdr), nest;
242 	u_int32_t plen;
243 	u_int32_t rtalert = ~0;
244 	int nxt, ours = 0;
245 	struct ifnet *deliverifp = NULL;
246 
247 #ifdef IPSEC
248 	/*
249 	 * should the inner packet be considered authentic?
250 	 * see comment in ah4_input().
251 	 */
252 	if (m) {
253 		m->m_flags &= ~M_AUTHIPHDR;
254 		m->m_flags &= ~M_AUTHIPDGM;
255 	}
256 #endif
257 
258 	/*
259 	 * make sure we don't have onion peering information into m_aux.
260 	 */
261 	ip6_delaux(m);
262 
263 	/*
264 	 * mbuf statistics
265 	 */
266 	if (m->m_flags & M_EXT) {
267 		if (m->m_next)
268 			ip6stat.ip6s_mext2m++;
269 		else
270 			ip6stat.ip6s_mext1++;
271 	} else {
272 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
273 		if (m->m_next) {
274 			if (m->m_flags & M_LOOP) {
275 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
276 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
277 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
278 			else
279 				ip6stat.ip6s_m2m[0]++;
280 		} else
281 			ip6stat.ip6s_m1++;
282 #undef M2MMAX
283 	}
284 
285 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
286 	ip6stat.ip6s_total++;
287 
288 #ifndef PULLDOWN_TEST
289 	/*
290 	 * L2 bridge code and some other code can return mbuf chain
291 	 * that does not conform to KAME requirement.  too bad.
292 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
293 	 */
294 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
295 		struct mbuf *n;
296 
297 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
298 		if (n)
299 			M_MOVE_PKTHDR(n, m);
300 		if (n && n->m_pkthdr.len > MHLEN) {
301 			MCLGET(n, M_DONTWAIT);
302 			if ((n->m_flags & M_EXT) == 0) {
303 				m_freem(n);
304 				n = NULL;
305 			}
306 		}
307 		if (n == NULL) {
308 			m_freem(m);
309 			return;	/*ENOBUFS*/
310 		}
311 
312 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
313 		n->m_len = n->m_pkthdr.len;
314 		m_freem(m);
315 		m = n;
316 	}
317 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
318 #endif
319 
320 	if (m->m_len < sizeof(struct ip6_hdr)) {
321 		struct ifnet *inifp;
322 		inifp = m->m_pkthdr.rcvif;
323 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
324 			ip6stat.ip6s_toosmall++;
325 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
326 			return;
327 		}
328 	}
329 
330 	ip6 = mtod(m, struct ip6_hdr *);
331 
332 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
333 		ip6stat.ip6s_badvers++;
334 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
335 		goto bad;
336 	}
337 
338 	/*
339 	 * Check if we want to allow this packet to be processed.
340 	 * Consider it to be bad if not.
341 	 */
342 	if (fr_checkp) {
343 		struct	mbuf	*m1 = m;
344 
345 		if ((*fr_checkp)((struct ip *)ip6, sizeof(*ip6),
346 				 m->m_pkthdr.rcvif, 0, &m1) != 0)
347 			return;
348 		m = m1;
349 		if (m == NULL)
350 			return;
351 		ip6 = mtod(m, struct ip6_hdr *);
352 	}
353 
354 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
355 
356 	/*
357 	 * Check with the firewall...
358 	 */
359 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
360 		u_short port = 0;
361 		/* If ipfw says divert, we have to just drop packet */
362 		/* use port as a dummy argument */
363 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
364 			m_freem(m);
365 			m = NULL;
366 		}
367 		if (!m)
368 			return;
369 	}
370 
371 	/*
372 	 * Check against address spoofing/corruption.
373 	 */
374 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
375 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
376 		/*
377 		 * XXX: "badscope" is not very suitable for a multicast source.
378 		 */
379 		ip6stat.ip6s_badscope++;
380 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
381 		goto bad;
382 	}
383 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
384 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
385 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
386 		ip6stat.ip6s_badscope++;
387 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
388 		goto bad;
389 	}
390 
391 	/*
392 	 * The following check is not documented in specs.  A malicious
393 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
394 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
395 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
396 	 *
397 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
398 	 * support IPv4-less kernel compilation, we cannot support SIIT
399 	 * environment at all.  So, it makes more sense for us to reject any
400 	 * malicious packets for non-SIIT environment, than try to do a
401 	 * partical support for SIIT environment.
402 	 */
403 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
404 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
405 		ip6stat.ip6s_badscope++;
406 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
407 		goto bad;
408 	}
409 #if 0
410 	/*
411 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
412 	 *
413 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
414 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
415 	 * is revised to forbid relaying case.
416 	 */
417 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
418 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
419 		ip6stat.ip6s_badscope++;
420 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
421 		goto bad;
422 	}
423 #endif
424 
425 	/* drop packets if interface ID portion is already filled */
426 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
427 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
428 		    ip6->ip6_src.s6_addr16[1]) {
429 			ip6stat.ip6s_badscope++;
430 			goto bad;
431 		}
432 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
433 		    ip6->ip6_dst.s6_addr16[1]) {
434 			ip6stat.ip6s_badscope++;
435 			goto bad;
436 		}
437 	}
438 
439 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
440 		ip6->ip6_src.s6_addr16[1]
441 			= htons(m->m_pkthdr.rcvif->if_index);
442 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
443 		ip6->ip6_dst.s6_addr16[1]
444 			= htons(m->m_pkthdr.rcvif->if_index);
445 
446 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
447 	/*
448 	 * We use rt->rt_ifp to determine if the address is ours or not.
449 	 * If rt_ifp is lo0, the address is ours.
450 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
451 	 * so any address under fe80::%lo0/64 will be mistakenly considered
452 	 * local.  The special case is supplied to handle the case properly
453 	 * by actually looking at interface addresses
454 	 * (using in6ifa_ifpwithaddr).
455 	 */
456 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
457 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
458 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
459 			icmp6_error(m, ICMP6_DST_UNREACH,
460 			    ICMP6_DST_UNREACH_ADDR, 0);
461 			/* m is already freed */
462 			return;
463 		}
464 
465 		ours = 1;
466 		deliverifp = m->m_pkthdr.rcvif;
467 		goto hbhcheck;
468 	}
469 #endif
470 
471 	/*
472 	 * Multicast check
473 	 */
474 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
475 	  	struct	in6_multi *in6m = 0;
476 
477 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
478 		/*
479 		 * See if we belong to the destination multicast group on the
480 		 * arrival interface.
481 		 */
482 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
483 		if (in6m)
484 			ours = 1;
485 		else if (!ip6_mrouter) {
486 			ip6stat.ip6s_notmember++;
487 			ip6stat.ip6s_cantforward++;
488 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
489 			goto bad;
490 		}
491 		deliverifp = m->m_pkthdr.rcvif;
492 		goto hbhcheck;
493 	}
494 
495 	/*
496 	 *  Unicast check
497 	 */
498 	switch (ip6_ours_check_algorithm) {
499 	default:
500 		/*
501 		 * XXX: I intentionally broke our indentation rule here,
502 		 *      since this switch-case is just for measurement and
503 		 *      therefore should soon be removed.
504 		 */
505 	if (ip6_forward_rt.ro_rt != NULL &&
506 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
507 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
508 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
509 		ip6stat.ip6s_forward_cachehit++;
510 	else {
511 		struct sockaddr_in6 *dst6;
512 
513 		if (ip6_forward_rt.ro_rt) {
514 			/* route is down or destination is different */
515 			ip6stat.ip6s_forward_cachemiss++;
516 			RTFREE(ip6_forward_rt.ro_rt);
517 			ip6_forward_rt.ro_rt = 0;
518 		}
519 
520 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
521 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
522 		dst6->sin6_len = sizeof(struct sockaddr_in6);
523 		dst6->sin6_family = AF_INET6;
524 		dst6->sin6_addr = ip6->ip6_dst;
525 #ifdef SCOPEDROUTING
526 		ip6_forward_rt.ro_dst.sin6_scope_id =
527 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
528 #endif
529 
530 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
531 	}
532 
533 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
534 
535 	/*
536 	 * Accept the packet if the forwarding interface to the destination
537 	 * according to the routing table is the loopback interface,
538 	 * unless the associated route has a gateway.
539 	 * Note that this approach causes to accept a packet if there is a
540 	 * route to the loopback interface for the destination of the packet.
541 	 * But we think it's even useful in some situations, e.g. when using
542 	 * a special daemon which wants to intercept the packet.
543 	 *
544 	 * XXX: some OSes automatically make a cloned route for the destination
545 	 * of an outgoing packet.  If the outgoing interface of the packet
546 	 * is a loopback one, the kernel would consider the packet to be
547 	 * accepted, even if we have no such address assinged on the interface.
548 	 * We check the cloned flag of the route entry to reject such cases,
549 	 * assuming that route entries for our own addresses are not made by
550 	 * cloning (it should be true because in6_addloop explicitly installs
551 	 * the host route).  However, we might have to do an explicit check
552 	 * while it would be less efficient.  Or, should we rather install a
553 	 * reject route for such a case?
554 	 */
555 	if (ip6_forward_rt.ro_rt &&
556 	    (ip6_forward_rt.ro_rt->rt_flags &
557 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
558 #ifdef RTF_WASCLONED
559 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
560 #endif
561 #ifdef RTF_CLONED
562 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
563 #endif
564 #if 0
565 	    /*
566 	     * The check below is redundant since the comparison of
567 	     * the destination and the key of the rtentry has
568 	     * already done through looking up the routing table.
569 	     */
570 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
571 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
572 #endif
573 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
574 		struct in6_ifaddr *ia6 =
575 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
576 
577 		/*
578 		 * record address information into m_aux.
579 		 */
580 		(void)ip6_setdstifaddr(m, ia6);
581 
582 		/*
583 		 * packets to a tentative, duplicated, or somehow invalid
584 		 * address must not be accepted.
585 		 */
586 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
587 			/* this address is ready */
588 			ours = 1;
589 			deliverifp = ia6->ia_ifp;	/* correct? */
590 			/* Count the packet in the ip address stats */
591 			ia6->ia_ifa.if_ipackets++;
592 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
593 			goto hbhcheck;
594 		} else {
595 			/* address is not ready, so discard the packet. */
596 			nd6log((LOG_INFO,
597 			    "ip6_input: packet to an unready address %s->%s\n",
598 			    ip6_sprintf(&ip6->ip6_src),
599 			    ip6_sprintf(&ip6->ip6_dst)));
600 
601 			goto bad;
602 		}
603 	}
604 	} /* XXX indentation (see above) */
605 
606 	/*
607 	 * FAITH(Firewall Aided Internet Translator)
608 	 */
609 	if (ip6_keepfaith) {
610 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
611 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
612 			/* XXX do we need more sanity checks? */
613 			ours = 1;
614 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
615 			goto hbhcheck;
616 		}
617 	}
618 
619 	/*
620 	 * Now there is no reason to process the packet if it's not our own
621 	 * and we're not a router.
622 	 */
623 	if (!ip6_forwarding) {
624 		ip6stat.ip6s_cantforward++;
625 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
626 		goto bad;
627 	}
628 
629   hbhcheck:
630 	/*
631 	 * record address information into m_aux, if we don't have one yet.
632 	 * note that we are unable to record it, if the address is not listed
633 	 * as our interface address (e.g. multicast addresses, addresses
634 	 * within FAITH prefixes and such).
635 	 */
636 	if (deliverifp && !ip6_getdstifaddr(m)) {
637 		struct in6_ifaddr *ia6;
638 
639 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
640 		if (ia6) {
641 			if (!ip6_setdstifaddr(m, ia6)) {
642 				/*
643 				 * XXX maybe we should drop the packet here,
644 				 * as we could not provide enough information
645 				 * to the upper layers.
646 				 */
647 			}
648 		}
649 	}
650 
651 	/*
652 	 * Process Hop-by-Hop options header if it's contained.
653 	 * m may be modified in ip6_hopopts_input().
654 	 * If a JumboPayload option is included, plen will also be modified.
655 	 */
656 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
657 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
658 		struct ip6_hbh *hbh;
659 
660 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
661 #if 0	/*touches NULL pointer*/
662 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
663 #endif
664 			return;	/* m have already been freed */
665 		}
666 
667 		/* adjust pointer */
668 		ip6 = mtod(m, struct ip6_hdr *);
669 
670 		/*
671 		 * if the payload length field is 0 and the next header field
672 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
673 		 * option MUST be included.
674 		 */
675 		if (ip6->ip6_plen == 0 && plen == 0) {
676 			/*
677 			 * Note that if a valid jumbo payload option is
678 			 * contained, ip6_hoptops_input() must set a valid
679 			 * (non-zero) payload length to the variable plen.
680 			 */
681 			ip6stat.ip6s_badoptions++;
682 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
683 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
684 			icmp6_error(m, ICMP6_PARAM_PROB,
685 				    ICMP6_PARAMPROB_HEADER,
686 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
687 			return;
688 		}
689 #ifndef PULLDOWN_TEST
690 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
691 		hbh = (struct ip6_hbh *)(ip6 + 1);
692 #else
693 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
694 			sizeof(struct ip6_hbh));
695 		if (hbh == NULL) {
696 			ip6stat.ip6s_tooshort++;
697 			return;
698 		}
699 #endif
700 		nxt = hbh->ip6h_nxt;
701 
702 		/*
703 		 * accept the packet if a router alert option is included
704 		 * and we act as an IPv6 router.
705 		 */
706 		if (rtalert != ~0 && ip6_forwarding)
707 			ours = 1;
708 	} else
709 		nxt = ip6->ip6_nxt;
710 
711 	/*
712 	 * Check that the amount of data in the buffers
713 	 * is as at least much as the IPv6 header would have us expect.
714 	 * Trim mbufs if longer than we expect.
715 	 * Drop packet if shorter than we expect.
716 	 */
717 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
718 		ip6stat.ip6s_tooshort++;
719 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
720 		goto bad;
721 	}
722 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
723 		if (m->m_len == m->m_pkthdr.len) {
724 			m->m_len = sizeof(struct ip6_hdr) + plen;
725 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
726 		} else
727 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
728 	}
729 
730 	/*
731 	 * Forward if desirable.
732 	 */
733 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
734 		/*
735 		 * If we are acting as a multicast router, all
736 		 * incoming multicast packets are passed to the
737 		 * kernel-level multicast forwarding function.
738 		 * The packet is returned (relatively) intact; if
739 		 * ip6_mforward() returns a non-zero value, the packet
740 		 * must be discarded, else it may be accepted below.
741 		 */
742 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
743 			ip6stat.ip6s_cantforward++;
744 			m_freem(m);
745 			return;
746 		}
747 		if (!ours) {
748 			m_freem(m);
749 			return;
750 		}
751 	} else if (!ours) {
752 		ip6_forward(m, 0);
753 		return;
754 	}
755 
756 	ip6 = mtod(m, struct ip6_hdr *);
757 
758 	/*
759 	 * Malicious party may be able to use IPv4 mapped addr to confuse
760 	 * tcp/udp stack and bypass security checks (act as if it was from
761 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
762 	 *
763 	 * For SIIT end node behavior, you may want to disable the check.
764 	 * However, you will  become vulnerable to attacks using IPv4 mapped
765 	 * source.
766 	 */
767 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
768 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
769 		ip6stat.ip6s_badscope++;
770 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
771 		goto bad;
772 	}
773 
774 	/*
775 	 * Tell launch routine the next header
776 	 */
777 	ip6stat.ip6s_delivered++;
778 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
779 	nest = 0;
780 
781 	while (nxt != IPPROTO_DONE) {
782 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
783 			ip6stat.ip6s_toomanyhdr++;
784 			goto bad;
785 		}
786 
787 		/*
788 		 * protection against faulty packet - there should be
789 		 * more sanity checks in header chain processing.
790 		 */
791 		if (m->m_pkthdr.len < off) {
792 			ip6stat.ip6s_tooshort++;
793 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
794 			goto bad;
795 		}
796 
797 #if 0
798 		/*
799 		 * do we need to do it for every header?  yeah, other
800 		 * functions can play with it (like re-allocate and copy).
801 		 */
802 		mhist = ip6_addaux(m);
803 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
804 			hist = mtod(mhist, caddr_t) + mhist->m_len;
805 			bcopy(&nxt, hist, sizeof(nxt));
806 			mhist->m_len += sizeof(nxt);
807 		} else {
808 			ip6stat.ip6s_toomanyhdr++;
809 			goto bad;
810 		}
811 #endif
812 
813 #ifdef IPSEC
814 		/*
815 		 * enforce IPsec policy checking if we are seeing last header.
816 		 * note that we do not visit this with protocols with pcb layer
817 		 * code - like udp/tcp/raw ip.
818 		 */
819 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
820 		    ipsec6_in_reject(m, NULL)) {
821 			ipsec6stat.in_polvio++;
822 			goto bad;
823 		}
824 #endif
825 
826 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
827 	}
828 	return;
829  bad:
830 	m_freem(m);
831 }
832 
833 /*
834  * set/grab in6_ifaddr correspond to IPv6 destination address.
835  * XXX backward compatibility wrapper
836  */
837 static struct ip6aux *
838 ip6_setdstifaddr(m, ia6)
839 	struct mbuf *m;
840 	struct in6_ifaddr *ia6;
841 {
842 	struct ip6aux *n;
843 
844 	n = ip6_addaux(m);
845 	if (n)
846 		n->ip6a_dstia6 = ia6;
847 	return n;	/* NULL if failed to set */
848 }
849 
850 struct in6_ifaddr *
851 ip6_getdstifaddr(m)
852 	struct mbuf *m;
853 {
854 	struct ip6aux *n;
855 
856 	n = ip6_findaux(m);
857 	if (n)
858 		return n->ip6a_dstia6;
859 	else
860 		return NULL;
861 }
862 
863 /*
864  * Hop-by-Hop options header processing. If a valid jumbo payload option is
865  * included, the real payload length will be stored in plenp.
866  */
867 static int
868 ip6_hopopts_input(plenp, rtalertp, mp, offp)
869 	u_int32_t *plenp;
870 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
871 	struct mbuf **mp;
872 	int *offp;
873 {
874 	struct mbuf *m = *mp;
875 	int off = *offp, hbhlen;
876 	struct ip6_hbh *hbh;
877 	u_int8_t *opt;
878 
879 	/* validation of the length of the header */
880 #ifndef PULLDOWN_TEST
881 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
882 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
883 	hbhlen = (hbh->ip6h_len + 1) << 3;
884 
885 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
886 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
887 #else
888 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
889 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
890 	if (hbh == NULL) {
891 		ip6stat.ip6s_tooshort++;
892 		return -1;
893 	}
894 	hbhlen = (hbh->ip6h_len + 1) << 3;
895 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
896 		hbhlen);
897 	if (hbh == NULL) {
898 		ip6stat.ip6s_tooshort++;
899 		return -1;
900 	}
901 #endif
902 	off += hbhlen;
903 	hbhlen -= sizeof(struct ip6_hbh);
904 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
905 
906 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
907 				hbhlen, rtalertp, plenp) < 0)
908 		return(-1);
909 
910 	*offp = off;
911 	*mp = m;
912 	return(0);
913 }
914 
915 /*
916  * Search header for all Hop-by-hop options and process each option.
917  * This function is separate from ip6_hopopts_input() in order to
918  * handle a case where the sending node itself process its hop-by-hop
919  * options header. In such a case, the function is called from ip6_output().
920  *
921  * The function assumes that hbh header is located right after the IPv6 header
922  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
923  * opthead + hbhlen is located in continuous memory region.
924  */
925 int
926 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
927 	struct mbuf *m;
928 	u_int8_t *opthead;
929 	int hbhlen;
930 	u_int32_t *rtalertp;
931 	u_int32_t *plenp;
932 {
933 	struct ip6_hdr *ip6;
934 	int optlen = 0;
935 	u_int8_t *opt = opthead;
936 	u_int16_t rtalert_val;
937 	u_int32_t jumboplen;
938 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
939 
940 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
941 		switch (*opt) {
942 		case IP6OPT_PAD1:
943 			optlen = 1;
944 			break;
945 		case IP6OPT_PADN:
946 			if (hbhlen < IP6OPT_MINLEN) {
947 				ip6stat.ip6s_toosmall++;
948 				goto bad;
949 			}
950 			optlen = *(opt + 1) + 2;
951 			break;
952 		case IP6OPT_RTALERT:
953 			/* XXX may need check for alignment */
954 			if (hbhlen < IP6OPT_RTALERT_LEN) {
955 				ip6stat.ip6s_toosmall++;
956 				goto bad;
957 			}
958 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
959 				/* XXX stat */
960 				icmp6_error(m, ICMP6_PARAM_PROB,
961 					    ICMP6_PARAMPROB_HEADER,
962 					    erroff + opt + 1 - opthead);
963 				return(-1);
964 			}
965 			optlen = IP6OPT_RTALERT_LEN;
966 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
967 			*rtalertp = ntohs(rtalert_val);
968 			break;
969 		case IP6OPT_JUMBO:
970 			/* XXX may need check for alignment */
971 			if (hbhlen < IP6OPT_JUMBO_LEN) {
972 				ip6stat.ip6s_toosmall++;
973 				goto bad;
974 			}
975 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
976 				/* XXX stat */
977 				icmp6_error(m, ICMP6_PARAM_PROB,
978 					    ICMP6_PARAMPROB_HEADER,
979 					    erroff + opt + 1 - opthead);
980 				return(-1);
981 			}
982 			optlen = IP6OPT_JUMBO_LEN;
983 
984 			/*
985 			 * IPv6 packets that have non 0 payload length
986 			 * must not contain a jumbo payload option.
987 			 */
988 			ip6 = mtod(m, struct ip6_hdr *);
989 			if (ip6->ip6_plen) {
990 				ip6stat.ip6s_badoptions++;
991 				icmp6_error(m, ICMP6_PARAM_PROB,
992 					    ICMP6_PARAMPROB_HEADER,
993 					    erroff + opt - opthead);
994 				return(-1);
995 			}
996 
997 			/*
998 			 * We may see jumbolen in unaligned location, so
999 			 * we'd need to perform bcopy().
1000 			 */
1001 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1002 			jumboplen = (u_int32_t)htonl(jumboplen);
1003 
1004 #if 1
1005 			/*
1006 			 * if there are multiple jumbo payload options,
1007 			 * *plenp will be non-zero and the packet will be
1008 			 * rejected.
1009 			 * the behavior may need some debate in ipngwg -
1010 			 * multiple options does not make sense, however,
1011 			 * there's no explicit mention in specification.
1012 			 */
1013 			if (*plenp != 0) {
1014 				ip6stat.ip6s_badoptions++;
1015 				icmp6_error(m, ICMP6_PARAM_PROB,
1016 					    ICMP6_PARAMPROB_HEADER,
1017 					    erroff + opt + 2 - opthead);
1018 				return(-1);
1019 			}
1020 #endif
1021 
1022 			/*
1023 			 * jumbo payload length must be larger than 65535.
1024 			 */
1025 			if (jumboplen <= IPV6_MAXPACKET) {
1026 				ip6stat.ip6s_badoptions++;
1027 				icmp6_error(m, ICMP6_PARAM_PROB,
1028 					    ICMP6_PARAMPROB_HEADER,
1029 					    erroff + opt + 2 - opthead);
1030 				return(-1);
1031 			}
1032 			*plenp = jumboplen;
1033 
1034 			break;
1035 		default:		/* unknown option */
1036 			if (hbhlen < IP6OPT_MINLEN) {
1037 				ip6stat.ip6s_toosmall++;
1038 				goto bad;
1039 			}
1040 			optlen = ip6_unknown_opt(opt, m,
1041 			    erroff + opt - opthead);
1042 			if (optlen == -1)
1043 				return(-1);
1044 			optlen += 2;
1045 			break;
1046 		}
1047 	}
1048 
1049 	return(0);
1050 
1051   bad:
1052 	m_freem(m);
1053 	return(-1);
1054 }
1055 
1056 /*
1057  * Unknown option processing.
1058  * The third argument `off' is the offset from the IPv6 header to the option,
1059  * which is necessary if the IPv6 header the and option header and IPv6 header
1060  * is not continuous in order to return an ICMPv6 error.
1061  */
1062 int
1063 ip6_unknown_opt(optp, m, off)
1064 	u_int8_t *optp;
1065 	struct mbuf *m;
1066 	int off;
1067 {
1068 	struct ip6_hdr *ip6;
1069 
1070 	switch (IP6OPT_TYPE(*optp)) {
1071 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1072 		return((int)*(optp + 1));
1073 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1074 		m_freem(m);
1075 		return(-1);
1076 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1077 		ip6stat.ip6s_badoptions++;
1078 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1079 		return(-1);
1080 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1081 		ip6stat.ip6s_badoptions++;
1082 		ip6 = mtod(m, struct ip6_hdr *);
1083 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1084 		    (m->m_flags & (M_BCAST|M_MCAST)))
1085 			m_freem(m);
1086 		else
1087 			icmp6_error(m, ICMP6_PARAM_PROB,
1088 				    ICMP6_PARAMPROB_OPTION, off);
1089 		return(-1);
1090 	}
1091 
1092 	m_freem(m);		/* XXX: NOTREACHED */
1093 	return(-1);
1094 }
1095 
1096 /*
1097  * Create the "control" list for this pcb.
1098  * The function will not modify mbuf chain at all.
1099  *
1100  * with KAME mbuf chain restriction:
1101  * The routine will be called from upper layer handlers like tcp6_input().
1102  * Thus the routine assumes that the caller (tcp6_input) have already
1103  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1104  * very first mbuf on the mbuf chain.
1105  */
1106 void
1107 ip6_savecontrol(in6p, mp, ip6, m)
1108 	struct inpcb *in6p;
1109 	struct mbuf **mp;
1110 	struct ip6_hdr *ip6;
1111 	struct mbuf *m;
1112 {
1113 	struct thread *td = curthread;	/* XXX */
1114 	int privileged = 0;
1115 	int rthdr_exist = 0;
1116 
1117 
1118 	if (suser(td) == 0)
1119  		privileged++;
1120 
1121 #ifdef SO_TIMESTAMP
1122 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1123 		struct timeval tv;
1124 
1125 		microtime(&tv);
1126 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1127 				      SCM_TIMESTAMP, SOL_SOCKET);
1128 		if (*mp) {
1129 			mp = &(*mp)->m_next;
1130 		}
1131 	}
1132 #endif
1133 
1134 	/* RFC 2292 sec. 5 */
1135 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1136 		struct in6_pktinfo pi6;
1137 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1138 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1139 			pi6.ipi6_addr.s6_addr16[1] = 0;
1140 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1141 					? m->m_pkthdr.rcvif->if_index
1142 					: 0;
1143 		*mp = sbcreatecontrol((caddr_t) &pi6,
1144 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1145 			IPPROTO_IPV6);
1146 		if (*mp)
1147 			mp = &(*mp)->m_next;
1148 	}
1149 
1150 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1151 		int hlim = ip6->ip6_hlim & 0xff;
1152 		*mp = sbcreatecontrol((caddr_t) &hlim,
1153 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1154 		if (*mp)
1155 			mp = &(*mp)->m_next;
1156 	}
1157 
1158 	/*
1159 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1160 	 * for the option, but it might be too strict, since there might
1161 	 * be some hop-by-hop options which can be returned to normal user.
1162 	 * See RFC 2292 section 6.
1163 	 */
1164 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1165 		/*
1166 		 * Check if a hop-by-hop options header is contatined in the
1167 		 * received packet, and if so, store the options as ancillary
1168 		 * data. Note that a hop-by-hop options header must be
1169 		 * just after the IPv6 header, which fact is assured through
1170 		 * the IPv6 input processing.
1171 		 */
1172 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1173 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1174 			struct ip6_hbh *hbh;
1175 			int hbhlen = 0;
1176 #ifdef PULLDOWN_TEST
1177 			struct mbuf *ext;
1178 #endif
1179 
1180 #ifndef PULLDOWN_TEST
1181 			hbh = (struct ip6_hbh *)(ip6 + 1);
1182 			hbhlen = (hbh->ip6h_len + 1) << 3;
1183 #else
1184 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1185 			    ip6->ip6_nxt);
1186 			if (ext == NULL) {
1187 				ip6stat.ip6s_tooshort++;
1188 				return;
1189 			}
1190 			hbh = mtod(ext, struct ip6_hbh *);
1191 			hbhlen = (hbh->ip6h_len + 1) << 3;
1192 			if (hbhlen != ext->m_len) {
1193 				m_freem(ext);
1194 				ip6stat.ip6s_tooshort++;
1195 				return;
1196 			}
1197 #endif
1198 
1199 			/*
1200 			 * XXX: We copy whole the header even if a jumbo
1201 			 * payload option is included, which option is to
1202 			 * be removed before returning in the RFC 2292.
1203 			 * Note: this constraint is removed in 2292bis.
1204 			 */
1205 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1206 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1207 			if (*mp)
1208 				mp = &(*mp)->m_next;
1209 #ifdef PULLDOWN_TEST
1210 			m_freem(ext);
1211 #endif
1212 		}
1213 	}
1214 
1215 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1216 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1217 		int proto, off, nxt;
1218 
1219 		/*
1220 		 * go through the header chain to see if a routing header is
1221 		 * contained in the packet. We need this information to store
1222 		 * destination options headers (if any) properly.
1223 		 * XXX: performance issue. We should record this info when
1224 		 * processing extension headers in incoming routine.
1225 		 * (todo) use m_aux?
1226 		 */
1227 		proto = IPPROTO_IPV6;
1228 		off = 0;
1229 		nxt = -1;
1230 		while (1) {
1231 			int newoff;
1232 
1233 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1234 			if (newoff < 0)
1235 				break;
1236 			if (newoff < off) /* invalid, check for safety */
1237 				break;
1238 			if ((proto = nxt) == IPPROTO_ROUTING) {
1239 				rthdr_exist = 1;
1240 				break;
1241 			}
1242 			off = newoff;
1243 		}
1244 	}
1245 
1246 	if ((in6p->in6p_flags &
1247 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1248 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1249 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1250 
1251 		/*
1252 		 * Search for destination options headers or routing
1253 		 * header(s) through the header chain, and stores each
1254 		 * header as ancillary data.
1255 		 * Note that the order of the headers remains in
1256 		 * the chain of ancillary data.
1257 		 */
1258 		while (1) {	/* is explicit loop prevention necessary? */
1259 			struct ip6_ext *ip6e = NULL;
1260 			int elen;
1261 #ifdef PULLDOWN_TEST
1262 			struct mbuf *ext = NULL;
1263 #endif
1264 
1265 			/*
1266 			 * if it is not an extension header, don't try to
1267 			 * pull it from the chain.
1268 			 */
1269 			switch (nxt) {
1270 			case IPPROTO_DSTOPTS:
1271 			case IPPROTO_ROUTING:
1272 			case IPPROTO_HOPOPTS:
1273 			case IPPROTO_AH: /* is it possible? */
1274 				break;
1275 			default:
1276 				goto loopend;
1277 			}
1278 
1279 #ifndef PULLDOWN_TEST
1280 			if (off + sizeof(*ip6e) > m->m_len)
1281 				goto loopend;
1282 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1283 			if (nxt == IPPROTO_AH)
1284 				elen = (ip6e->ip6e_len + 2) << 2;
1285 			else
1286 				elen = (ip6e->ip6e_len + 1) << 3;
1287 			if (off + elen > m->m_len)
1288 				goto loopend;
1289 #else
1290 			ext = ip6_pullexthdr(m, off, nxt);
1291 			if (ext == NULL) {
1292 				ip6stat.ip6s_tooshort++;
1293 				return;
1294 			}
1295 			ip6e = mtod(ext, struct ip6_ext *);
1296 			if (nxt == IPPROTO_AH)
1297 				elen = (ip6e->ip6e_len + 2) << 2;
1298 			else
1299 				elen = (ip6e->ip6e_len + 1) << 3;
1300 			if (elen != ext->m_len) {
1301 				m_freem(ext);
1302 				ip6stat.ip6s_tooshort++;
1303 				return;
1304 			}
1305 #endif
1306 
1307 			switch (nxt) {
1308 			case IPPROTO_DSTOPTS:
1309 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1310 					break;
1311 
1312 				/*
1313 				 * We also require super-user privilege for
1314 				 * the option.
1315 				 * See the comments on IN6_HOPOPTS.
1316 				 */
1317 				if (!privileged)
1318 					break;
1319 
1320 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1321 						      IPV6_DSTOPTS,
1322 						      IPPROTO_IPV6);
1323 				if (*mp)
1324 					mp = &(*mp)->m_next;
1325 				break;
1326 			case IPPROTO_ROUTING:
1327 				if (!in6p->in6p_flags & IN6P_RTHDR)
1328 					break;
1329 
1330 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1331 						      IPV6_RTHDR,
1332 						      IPPROTO_IPV6);
1333 				if (*mp)
1334 					mp = &(*mp)->m_next;
1335 				break;
1336 			case IPPROTO_HOPOPTS:
1337 			case IPPROTO_AH: /* is it possible? */
1338 				break;
1339 
1340 			default:
1341 				/*
1342 			 	 * other cases have been filtered in the above.
1343 				 * none will visit this case.  here we supply
1344 				 * the code just in case (nxt overwritten or
1345 				 * other cases).
1346 				 */
1347 #ifdef PULLDOWN_TEST
1348 				m_freem(ext);
1349 #endif
1350 				goto loopend;
1351 
1352 			}
1353 
1354 			/* proceed with the next header. */
1355 			off += elen;
1356 			nxt = ip6e->ip6e_nxt;
1357 			ip6e = NULL;
1358 #ifdef PULLDOWN_TEST
1359 			m_freem(ext);
1360 			ext = NULL;
1361 #endif
1362 		}
1363 	  loopend:
1364 		;
1365 	}
1366 
1367 }
1368 
1369 #ifdef PULLDOWN_TEST
1370 /*
1371  * pull single extension header from mbuf chain.  returns single mbuf that
1372  * contains the result, or NULL on error.
1373  */
1374 static struct mbuf *
1375 ip6_pullexthdr(m, off, nxt)
1376 	struct mbuf *m;
1377 	size_t off;
1378 	int nxt;
1379 {
1380 	struct ip6_ext ip6e;
1381 	size_t elen;
1382 	struct mbuf *n;
1383 
1384 #ifdef DIAGNOSTIC
1385 	switch (nxt) {
1386 	case IPPROTO_DSTOPTS:
1387 	case IPPROTO_ROUTING:
1388 	case IPPROTO_HOPOPTS:
1389 	case IPPROTO_AH: /* is it possible? */
1390 		break;
1391 	default:
1392 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1393 	}
1394 #endif
1395 
1396 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1397 	if (nxt == IPPROTO_AH)
1398 		elen = (ip6e.ip6e_len + 2) << 2;
1399 	else
1400 		elen = (ip6e.ip6e_len + 1) << 3;
1401 
1402 	MGET(n, M_DONTWAIT, MT_DATA);
1403 	if (n && elen >= MLEN) {
1404 		MCLGET(n, M_DONTWAIT);
1405 		if ((n->m_flags & M_EXT) == 0) {
1406 			m_free(n);
1407 			n = NULL;
1408 		}
1409 	}
1410 	if (!n)
1411 		return NULL;
1412 
1413 	n->m_len = 0;
1414 	if (elen >= M_TRAILINGSPACE(n)) {
1415 		m_free(n);
1416 		return NULL;
1417 	}
1418 
1419 	m_copydata(m, off, elen, mtod(n, caddr_t));
1420 	n->m_len = elen;
1421 	return n;
1422 }
1423 #endif
1424 
1425 /*
1426  * Get pointer to the previous header followed by the header
1427  * currently processed.
1428  * XXX: This function supposes that
1429  *	M includes all headers,
1430  *	the next header field and the header length field of each header
1431  *	are valid, and
1432  *	the sum of each header length equals to OFF.
1433  * Because of these assumptions, this function must be called very
1434  * carefully. Moreover, it will not be used in the near future when
1435  * we develop `neater' mechanism to process extension headers.
1436  */
1437 char *
1438 ip6_get_prevhdr(m, off)
1439 	struct mbuf *m;
1440 	int off;
1441 {
1442 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1443 
1444 	if (off == sizeof(struct ip6_hdr))
1445 		return(&ip6->ip6_nxt);
1446 	else {
1447 		int len, nxt;
1448 		struct ip6_ext *ip6e = NULL;
1449 
1450 		nxt = ip6->ip6_nxt;
1451 		len = sizeof(struct ip6_hdr);
1452 		while (len < off) {
1453 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1454 
1455 			switch (nxt) {
1456 			case IPPROTO_FRAGMENT:
1457 				len += sizeof(struct ip6_frag);
1458 				break;
1459 			case IPPROTO_AH:
1460 				len += (ip6e->ip6e_len + 2) << 2;
1461 				break;
1462 			default:
1463 				len += (ip6e->ip6e_len + 1) << 3;
1464 				break;
1465 			}
1466 			nxt = ip6e->ip6e_nxt;
1467 		}
1468 		if (ip6e)
1469 			return(&ip6e->ip6e_nxt);
1470 		else
1471 			return NULL;
1472 	}
1473 }
1474 
1475 /*
1476  * get next header offset.  m will be retained.
1477  */
1478 int
1479 ip6_nexthdr(m, off, proto, nxtp)
1480 	struct mbuf *m;
1481 	int off;
1482 	int proto;
1483 	int *nxtp;
1484 {
1485 	struct ip6_hdr ip6;
1486 	struct ip6_ext ip6e;
1487 	struct ip6_frag fh;
1488 
1489 	/* just in case */
1490 	if (m == NULL)
1491 		panic("ip6_nexthdr: m == NULL");
1492 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1493 		return -1;
1494 
1495 	switch (proto) {
1496 	case IPPROTO_IPV6:
1497 		if (m->m_pkthdr.len < off + sizeof(ip6))
1498 			return -1;
1499 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1500 		if (nxtp)
1501 			*nxtp = ip6.ip6_nxt;
1502 		off += sizeof(ip6);
1503 		return off;
1504 
1505 	case IPPROTO_FRAGMENT:
1506 		/*
1507 		 * terminate parsing if it is not the first fragment,
1508 		 * it does not make sense to parse through it.
1509 		 */
1510 		if (m->m_pkthdr.len < off + sizeof(fh))
1511 			return -1;
1512 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1513 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1514 			return -1;
1515 		if (nxtp)
1516 			*nxtp = fh.ip6f_nxt;
1517 		off += sizeof(struct ip6_frag);
1518 		return off;
1519 
1520 	case IPPROTO_AH:
1521 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1522 			return -1;
1523 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1524 		if (nxtp)
1525 			*nxtp = ip6e.ip6e_nxt;
1526 		off += (ip6e.ip6e_len + 2) << 2;
1527 		return off;
1528 
1529 	case IPPROTO_HOPOPTS:
1530 	case IPPROTO_ROUTING:
1531 	case IPPROTO_DSTOPTS:
1532 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1533 			return -1;
1534 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1535 		if (nxtp)
1536 			*nxtp = ip6e.ip6e_nxt;
1537 		off += (ip6e.ip6e_len + 1) << 3;
1538 		return off;
1539 
1540 	case IPPROTO_NONE:
1541 	case IPPROTO_ESP:
1542 	case IPPROTO_IPCOMP:
1543 		/* give up */
1544 		return -1;
1545 
1546 	default:
1547 		return -1;
1548 	}
1549 
1550 	return -1;
1551 }
1552 
1553 /*
1554  * get offset for the last header in the chain.  m will be kept untainted.
1555  */
1556 int
1557 ip6_lasthdr(m, off, proto, nxtp)
1558 	struct mbuf *m;
1559 	int off;
1560 	int proto;
1561 	int *nxtp;
1562 {
1563 	int newoff;
1564 	int nxt;
1565 
1566 	if (!nxtp) {
1567 		nxt = -1;
1568 		nxtp = &nxt;
1569 	}
1570 	while (1) {
1571 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1572 		if (newoff < 0)
1573 			return off;
1574 		else if (newoff < off)
1575 			return -1;	/* invalid */
1576 		else if (newoff == off)
1577 			return newoff;
1578 
1579 		off = newoff;
1580 		proto = *nxtp;
1581 	}
1582 }
1583 
1584 struct ip6aux *
1585 ip6_addaux(m)
1586 	struct mbuf *m;
1587 {
1588 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1589 	if (!tag) {
1590 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1591 				sizeof (struct ip6aux),
1592 				M_DONTWAIT);
1593 		if (tag)
1594 			m_tag_prepend(m, tag);
1595 	}
1596 	if (tag)
1597 		bzero(tag+1, sizeof (struct ip6aux));
1598 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1599 }
1600 
1601 struct ip6aux *
1602 ip6_findaux(m)
1603 	struct mbuf *m;
1604 {
1605 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1606 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1607 }
1608 
1609 void
1610 ip6_delaux(m)
1611 	struct mbuf *m;
1612 {
1613 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1614 	if (tag)
1615 		m_tag_delete(m, tag);
1616 }
1617 
1618 /*
1619  * System control for IP6
1620  */
1621 
1622 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1623 	0,		0,		0,		0,
1624 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1625 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1626 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1627 	0,		0,		0,		0,
1628 	ENOPROTOOPT
1629 };
1630