xref: /netbsd-src/sys/netinet6/ip6_input.c (revision 3cec974c61d7fac0a37c0377723a33214a458c8b)
1 /*	$NetBSD: ip6_input.c,v 1.37 2001/03/01 16:31:41 itojun Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.183 2001/03/01 15:15:23 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_inet.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/proc.h>
85 
86 #include <net/if.h>
87 #include <net/if_types.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/netisr.h>
91 #ifdef PFIL_HOOKS
92 #include <net/pfil.h>
93 #endif
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 <netinet6/in6_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 #endif
113 
114 #include <netinet6/ip6protosw.h>
115 
116 /* we need it for NLOOP. */
117 #include "loop.h"
118 #include "faith.h"
119 
120 #include "gif.h"
121 #include "bpfilter.h"
122 
123 #include <net/net_osdep.h>
124 
125 extern struct domain inet6domain;
126 
127 u_char ip6_protox[IPPROTO_MAX];
128 static int ip6qmaxlen = IFQ_MAXLEN;
129 struct in6_ifaddr *in6_ifaddr;
130 struct ifqueue ip6intrq;
131 
132 extern struct ifnet loif[NLOOP];
133 int ip6_forward_srcrt;			/* XXX */
134 int ip6_sourcecheck;			/* XXX */
135 int ip6_sourcecheck_interval;		/* XXX */
136 
137 #ifdef PFIL_HOOKS
138 struct pfil_head inet6_pfil_hook;
139 #endif
140 
141 struct ip6stat ip6stat;
142 
143 static void ip6_init2 __P((void *));
144 
145 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
146 
147 /*
148  * IP6 initialization: fill in IP6 protocol switch table.
149  * All protocols not implemented in kernel go to raw IP6 protocol handler.
150  */
151 void
152 ip6_init()
153 {
154 	struct ip6protosw *pr;
155 	int i;
156 	struct timeval tv;
157 
158 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
159 	if (pr == 0)
160 		panic("ip6_init");
161 	for (i = 0; i < IPPROTO_MAX; i++)
162 		ip6_protox[i] = pr - inet6sw;
163 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
164 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
165 		if (pr->pr_domain->dom_family == PF_INET6 &&
166 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
167 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
168 	ip6intrq.ifq_maxlen = ip6qmaxlen;
169 	nd6_init();
170 	frag6_init();
171 	/*
172 	 * in many cases, random() here does NOT return random number
173 	 * as initialization during bootstrap time occur in fixed order.
174 	 */
175 	microtime(&tv);
176 	ip6_flow_seq = random() ^ tv.tv_usec;
177 
178 	ip6_init2((void *)0);
179 
180 #ifdef PFIL_HOOKS
181 	/* Register our Packet Filter hook. */
182 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
183 	inet6_pfil_hook.ph_af   = AF_INET6;
184 	i = pfil_head_register(&inet6_pfil_hook);
185 	if (i != 0)
186 		printf("ip6_init: WARNING: unable to register pfil hook, "
187 		    "error %d\n", i);
188 #endif /* PFIL_HOOKS */
189 }
190 
191 static void
192 ip6_init2(dummy)
193 	void *dummy;
194 {
195 	/*
196 	 * to route local address of p2p link to loopback,
197 	 * assign loopback address first.
198 	 */
199 	in6_ifattach(&loif[0], NULL);
200 
201 	/* nd6_timer_init */
202 	callout_init(&nd6_timer_ch);
203 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
204 	/* router renumbering prefix list maintenance */
205 	callout_init(&in6_rr_timer_ch);
206 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
207 }
208 
209 /*
210  * IP6 input interrupt handling. Just pass the packet to ip6_input.
211  */
212 void
213 ip6intr()
214 {
215 	int s;
216 	struct mbuf *m;
217 
218 	for (;;) {
219 		s = splimp();
220 		IF_DEQUEUE(&ip6intrq, m);
221 		splx(s);
222 		if (m == 0)
223 			return;
224 		ip6_input(m);
225 	}
226 }
227 
228 extern struct	route_in6 ip6_forward_rt;
229 
230 void
231 ip6_input(m)
232 	struct mbuf *m;
233 {
234 	struct ip6_hdr *ip6;
235 	int off = sizeof(struct ip6_hdr), nest;
236 	u_int32_t plen;
237 	u_int32_t rtalert = ~0;
238 	int nxt, ours = 0;
239 	struct ifnet *deliverifp = NULL;
240 
241 #ifdef IPSEC
242 	/*
243 	 * should the inner packet be considered authentic?
244 	 * see comment in ah4_input().
245 	 */
246 	if (m) {
247 		m->m_flags &= ~M_AUTHIPHDR;
248 		m->m_flags &= ~M_AUTHIPDGM;
249 	}
250 #endif
251 
252 	/*
253 	 * mbuf statistics by kazu
254 	 */
255 	if (m->m_flags & M_EXT) {
256 		if (m->m_next)
257 			ip6stat.ip6s_mext2m++;
258 		else
259 			ip6stat.ip6s_mext1++;
260 	} else {
261 		if (m->m_next) {
262 			if (m->m_flags & M_LOOP) {
263 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/*XXX*/
264 			} else if (m->m_pkthdr.rcvif->if_index <= 31)
265 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
266 			else
267 				ip6stat.ip6s_m2m[0]++;
268 		} else
269 			ip6stat.ip6s_m1++;
270 	}
271 
272 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
273 	ip6stat.ip6s_total++;
274 
275 #ifndef PULLDOWN_TEST
276 	/* XXX is the line really necessary? */
277 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
278 #endif
279 
280 	if (m->m_len < sizeof(struct ip6_hdr)) {
281 		struct ifnet *inifp;
282 		inifp = m->m_pkthdr.rcvif;
283 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
284 			ip6stat.ip6s_toosmall++;
285 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
286 			return;
287 		}
288 	}
289 
290 	ip6 = mtod(m, struct ip6_hdr *);
291 
292 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
293 		ip6stat.ip6s_badvers++;
294 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
295 		goto bad;
296 	}
297 
298 #ifdef PFIL_HOOKS
299 	/*
300 	 * Run through list of hooks for input packets.  If there are any
301 	 * filters which require that additional packets in the flow are
302 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
303 	 * Note that filters must _never_ set this flag, as another filter
304 	 * in the list may have previously cleared it.
305 	 */
306 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif,
307 			   PFIL_IN) != 0)
308 		return;
309 	if (m == NULL)
310 		return;
311 	ip6 = mtod(m, struct ip6_hdr *);
312 #endif /* PFIL_HOOKS */
313 
314 
315 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
316 
317 #ifdef ALTQ
318 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
319 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
320 		/* packet is dropped by traffic conditioner */
321 		return;
322 	}
323 #endif
324 
325 	/*
326 	 * Scope check
327 	 */
328 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
329 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
330 		ip6stat.ip6s_badscope++;
331 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
332 		goto bad;
333 	}
334 	/*
335 	 * The following check is not documented in the spec.  Malicious party
336 	 * may be able to use IPv4 mapped addr to confuse tcp/udp stack and
337 	 * bypass security checks (act as if it was from 127.0.0.1 by using
338 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
339 	 *
340 	 * This check chokes if we are in SIIT cloud.  As none of BSDs support
341 	 * IPv4-less kernel compilation, we cannot support SIIT environment
342 	 * at all.  So, it makes more sense for us to reject any malicious
343 	 * packets for non-SIIT environment, than try to do a partical support
344 	 * for SIIT environment.
345 	 */
346 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
347 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
348 		ip6stat.ip6s_badscope++;
349 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
350 		goto bad;
351 	}
352 #if 0
353 	/*
354 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
355 	 *
356 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
357 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
358 	 * is revised to forbid relaying case.
359 	 */
360 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
361 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 #endif
367 
368 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
369 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
370 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
371 			ours = 1;
372 			deliverifp = m->m_pkthdr.rcvif;
373 			goto hbhcheck;
374 		} else {
375 			ip6stat.ip6s_badscope++;
376 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
377 			goto bad;
378 		}
379 	}
380 
381 #ifndef FAKE_LOOPBACK_IF
382 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0)
383 #else
384 	if (1)
385 #endif
386 	{
387 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
388 			ip6->ip6_src.s6_addr16[1]
389 				= htons(m->m_pkthdr.rcvif->if_index);
390 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
391 			ip6->ip6_dst.s6_addr16[1]
392 				= htons(m->m_pkthdr.rcvif->if_index);
393 	}
394 
395 	/*
396 	 * XXX we need this since we do not have "goto ours" hack route
397 	 * for some of our ifaddrs on loopback interface.
398 	 * we should correct it by changing in6_ifattach to install
399 	 * "goto ours" hack route.
400 	 */
401 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0) {
402 		if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
403 			ours = 1;
404 			deliverifp = m->m_pkthdr.rcvif;
405 			goto hbhcheck;
406 		}
407 	}
408 
409 	/*
410 	 * Multicast check
411 	 */
412 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
413 	  	struct	in6_multi *in6m = 0;
414 
415 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
416 		/*
417 		 * See if we belong to the destination multicast group on the
418 		 * arrival interface.
419 		 */
420 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
421 		if (in6m)
422 			ours = 1;
423 		else if (!ip6_mrouter) {
424 			ip6stat.ip6s_notmember++;
425 			ip6stat.ip6s_cantforward++;
426 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
427 			goto bad;
428 		}
429 		deliverifp = m->m_pkthdr.rcvif;
430 		goto hbhcheck;
431 	}
432 
433 	/*
434 	 *  Unicast check
435 	 */
436 	if (ip6_forward_rt.ro_rt != NULL &&
437 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
438 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
439 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
440 		ip6stat.ip6s_forward_cachehit++;
441 	else {
442 		struct sockaddr_in6 *dst6;
443 
444 		if (ip6_forward_rt.ro_rt) {
445 			/* route is down or destination is different */
446 			ip6stat.ip6s_forward_cachemiss++;
447 			RTFREE(ip6_forward_rt.ro_rt);
448 			ip6_forward_rt.ro_rt = 0;
449 		}
450 
451 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
452 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
453 		dst6->sin6_len = sizeof(struct sockaddr_in6);
454 		dst6->sin6_family = AF_INET6;
455 		dst6->sin6_addr = ip6->ip6_dst;
456 
457 		rtalloc((struct route *)&ip6_forward_rt);
458 	}
459 
460 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
461 
462 	/*
463 	 * Accept the packet if the forwarding interface to the destination
464 	 * according to the routing table is the loopback interface,
465 	 * unless the associated route has a gateway.
466 	 * Note that this approach causes to accept a packet if there is a
467 	 * route to the loopback interface for the destination of the packet.
468 	 * But we think it's even useful in some situations, e.g. when using
469 	 * a special daemon which wants to intercept the packet.
470 	 */
471 	if (ip6_forward_rt.ro_rt &&
472 	    (ip6_forward_rt.ro_rt->rt_flags &
473 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
474 #if 0
475 	    /*
476 	     * The check below is redundant since the comparison of
477 	     * the destination and the key of the rtentry has
478 	     * already done through looking up the routing table.
479 	     */
480 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
481 			       &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
482 #endif
483 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
484 		struct in6_ifaddr *ia6 =
485 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
486 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
487 			m->m_flags |= M_ANYCAST6;
488 		/*
489 		 * packets to a tentative, duplicated, or somehow invalid
490 		 * address must not be accepted.
491 		 */
492 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
493 			/* this address is ready */
494 			ours = 1;
495 			deliverifp = ia6->ia_ifp;	/* correct? */
496 			goto hbhcheck;
497 		} else {
498 			/* address is not ready, so discard the packet. */
499 			nd6log((LOG_INFO,
500 			    "ip6_input: packet to an unready address %s->%s\n",
501 			    ip6_sprintf(&ip6->ip6_src),
502 			    ip6_sprintf(&ip6->ip6_dst)));
503 
504 			goto bad;
505 		}
506 	}
507 
508 	/*
509 	 * FAITH(Firewall Aided Internet Translator)
510 	 */
511 #if defined(NFAITH) && 0 < NFAITH
512 	if (ip6_keepfaith) {
513 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
514 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
515 			/* XXX do we need more sanity checks? */
516 			ours = 1;
517 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
518 			goto hbhcheck;
519 		}
520 	}
521 #endif
522 
523 #if 0
524     {
525 	/*
526 	 * Last resort: check in6_ifaddr for incoming interface.
527 	 * The code is here until I update the "goto ours hack" code above
528 	 * working right.
529 	 */
530 	struct ifaddr *ifa;
531 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
532 	     ifa;
533 	     ifa = ifa->ifa_list.tqe_next) {
534 		if (ifa->ifa_addr == NULL)
535 			continue;	/* just for safety */
536 		if (ifa->ifa_addr->sa_family != AF_INET6)
537 			continue;
538 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
539 			ours = 1;
540 			deliverifp = ifa->ifa_ifp;
541 			goto hbhcheck;
542 		}
543 	}
544     }
545 #endif
546 
547 	/*
548 	 * Now there is no reason to process the packet if it's not our own
549 	 * and we're not a router.
550 	 */
551 	if (!ip6_forwarding) {
552 		ip6stat.ip6s_cantforward++;
553 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
554 		goto bad;
555 	}
556 
557   hbhcheck:
558 	/*
559 	 * Process Hop-by-Hop options header if it's contained.
560 	 * m may be modified in ip6_hopopts_input().
561 	 * If a JumboPayload option is included, plen will also be modified.
562 	 */
563 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
564 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
565 		struct ip6_hbh *hbh;
566 
567 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
568 #if 0	/*touches NULL pointer*/
569 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
570 #endif
571 			return;	/* m have already been freed */
572 		}
573 
574 		/* adjust pointer */
575 		ip6 = mtod(m, struct ip6_hdr *);
576 
577 		/*
578 		 * if the payload length field is 0 and the next header field
579 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
580 		 * option MUST be included.
581 		 */
582 		if (ip6->ip6_plen == 0 && plen == 0) {
583 			/*
584 			 * Note that if a valid jumbo payload option is
585 			 * contained, ip6_hoptops_input() must set a valid
586 			 * (non-zero) payload length to the variable plen.
587 			 */
588 			ip6stat.ip6s_badoptions++;
589 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
590 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
591 			icmp6_error(m, ICMP6_PARAM_PROB,
592 				    ICMP6_PARAMPROB_HEADER,
593 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
594 			return;
595 		}
596 #ifndef PULLDOWN_TEST
597 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
598 		hbh = (struct ip6_hbh *)(ip6 + 1);
599 #else
600 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
601 			sizeof(struct ip6_hbh));
602 		if (hbh == NULL) {
603 			ip6stat.ip6s_tooshort++;
604 			return;
605 		}
606 #endif
607 		nxt = hbh->ip6h_nxt;
608 
609 		/*
610 		 * accept the packet if a router alert option is included
611 		 * and we act as an IPv6 router.
612 		 */
613 		if (rtalert != ~0 && ip6_forwarding)
614 			ours = 1;
615 	} else
616 		nxt = ip6->ip6_nxt;
617 
618 	/*
619 	 * Check that the amount of data in the buffers
620 	 * is as at least much as the IPv6 header would have us expect.
621 	 * Trim mbufs if longer than we expect.
622 	 * Drop packet if shorter than we expect.
623 	 */
624 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
625 		ip6stat.ip6s_tooshort++;
626 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
627 		goto bad;
628 	}
629 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
630 		if (m->m_len == m->m_pkthdr.len) {
631 			m->m_len = sizeof(struct ip6_hdr) + plen;
632 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
633 		} else
634 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
635 	}
636 
637 	/*
638 	 * Forward if desirable.
639 	 */
640 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
641 		/*
642 		 * If we are acting as a multicast router, all
643 		 * incoming multicast packets are passed to the
644 		 * kernel-level multicast forwarding function.
645 		 * The packet is returned (relatively) intact; if
646 		 * ip6_mforward() returns a non-zero value, the packet
647 		 * must be discarded, else it may be accepted below.
648 		 */
649 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
650 			ip6stat.ip6s_cantforward++;
651 			m_freem(m);
652 			return;
653 		}
654 		if (!ours) {
655 			m_freem(m);
656 			return;
657 		}
658 	} else if (!ours) {
659 		ip6_forward(m, 0);
660 		return;
661 	}
662 
663 	ip6 = mtod(m, struct ip6_hdr *);
664 
665 	/*
666 	 * Malicious party may be able to use IPv4 mapped addr to confuse
667 	 * tcp/udp stack and bypass security checks (act as if it was from
668 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
669 	 *
670 	 * For SIIT end node behavior, you may want to disable the check.
671 	 * However, you will  become vulnerable to attacks using IPv4 mapped
672 	 * source.
673 	 */
674 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
675 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
676 		ip6stat.ip6s_badscope++;
677 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
678 		goto bad;
679 	}
680 
681 	/*
682 	 * Tell launch routine the next header
683 	 */
684 #ifdef IFA_STATS
685 	if (deliverifp != NULL) {
686 		struct in6_ifaddr *ia6;
687 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
688 		if (ia6)
689 			ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
690 	}
691 #endif
692 	ip6stat.ip6s_delivered++;
693 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
694 	nest = 0;
695 	while (nxt != IPPROTO_DONE) {
696 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
697 			ip6stat.ip6s_toomanyhdr++;
698 			goto bad;
699 		}
700 
701 		/*
702 		 * protection against faulty packet - there should be
703 		 * more sanity checks in header chain processing.
704 		 */
705 		if (m->m_pkthdr.len < off) {
706 			ip6stat.ip6s_tooshort++;
707 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
708 			goto bad;
709 		}
710 
711 #ifdef IPSEC
712 		/*
713 		 * enforce IPsec policy checking if we are seeing last header.
714 		 * note that we do not visit this with protocols with pcb layer
715 		 * code - like udp/tcp/raw ip.
716 		 */
717 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
718 		    ipsec6_in_reject(m, NULL)) {
719 			ipsec6stat.in_polvio++;
720 			goto bad;
721 		}
722 #endif
723 
724 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
725 	}
726 	return;
727  bad:
728 	m_freem(m);
729 }
730 
731 /*
732  * Hop-by-Hop options header processing. If a valid jumbo payload option is
733  * included, the real payload length will be stored in plenp.
734  */
735 static int
736 ip6_hopopts_input(plenp, rtalertp, mp, offp)
737 	u_int32_t *plenp;
738 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
739 	struct mbuf **mp;
740 	int *offp;
741 {
742 	struct mbuf *m = *mp;
743 	int off = *offp, hbhlen;
744 	struct ip6_hbh *hbh;
745 	u_int8_t *opt;
746 
747 	/* validation of the length of the header */
748 #ifndef PULLDOWN_TEST
749 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
750 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
751 	hbhlen = (hbh->ip6h_len + 1) << 3;
752 
753 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
754 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
755 #else
756 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
757 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
758 	if (hbh == NULL) {
759 		ip6stat.ip6s_tooshort++;
760 		return -1;
761 	}
762 	hbhlen = (hbh->ip6h_len + 1) << 3;
763 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
764 		hbhlen);
765 	if (hbh == NULL) {
766 		ip6stat.ip6s_tooshort++;
767 		return -1;
768 	}
769 #endif
770 	off += hbhlen;
771 	hbhlen -= sizeof(struct ip6_hbh);
772 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
773 
774 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
775 				hbhlen, rtalertp, plenp) < 0)
776 		return(-1);
777 
778 	*offp = off;
779 	*mp = m;
780 	return(0);
781 }
782 
783 /*
784  * Search header for all Hop-by-hop options and process each option.
785  * This function is separate from ip6_hopopts_input() in order to
786  * handle a case where the sending node itself process its hop-by-hop
787  * options header. In such a case, the function is called from ip6_output().
788  */
789 int
790 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
791 	struct mbuf *m;
792 	u_int8_t *opthead;
793 	int hbhlen;
794 	u_int32_t *rtalertp;
795 	u_int32_t *plenp;
796 {
797 	struct ip6_hdr *ip6;
798 	int optlen = 0;
799 	u_int8_t *opt = opthead;
800 	u_int16_t rtalert_val;
801 	u_int32_t jumboplen;
802 
803 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
804 		switch (*opt) {
805 		case IP6OPT_PAD1:
806 			optlen = 1;
807 			break;
808 		case IP6OPT_PADN:
809 			if (hbhlen < IP6OPT_MINLEN) {
810 				ip6stat.ip6s_toosmall++;
811 				goto bad;
812 			}
813 			optlen = *(opt + 1) + 2;
814 			break;
815 		case IP6OPT_RTALERT:
816 			/* XXX may need check for alignment */
817 			if (hbhlen < IP6OPT_RTALERT_LEN) {
818 				ip6stat.ip6s_toosmall++;
819 				goto bad;
820 			}
821 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
822 				 /* XXX: should we discard the packet? */
823 				log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
824 				    *(opt + 1));
825 			}
826 			optlen = IP6OPT_RTALERT_LEN;
827 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
828 			*rtalertp = ntohs(rtalert_val);
829 			break;
830 		case IP6OPT_JUMBO:
831 			/* XXX may need check for alignment */
832 			if (hbhlen < IP6OPT_JUMBO_LEN) {
833 				ip6stat.ip6s_toosmall++;
834 				goto bad;
835 			}
836 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
837 				/* XXX: should we discard the packet? */
838 				log(LOG_ERR, "length of jumbopayload opt "
839 				    "is inconsistent(%d)\n",
840 				    *(opt + 1));
841 			}
842 			optlen = IP6OPT_JUMBO_LEN;
843 
844 			/*
845 			 * IPv6 packets that have non 0 payload length
846 			 * must not contain a jumbo payload option.
847 			 */
848 			ip6 = mtod(m, struct ip6_hdr *);
849 			if (ip6->ip6_plen) {
850 				ip6stat.ip6s_badoptions++;
851 				icmp6_error(m, ICMP6_PARAM_PROB,
852 					    ICMP6_PARAMPROB_HEADER,
853 					    sizeof(struct ip6_hdr) +
854 					    sizeof(struct ip6_hbh) +
855 					    opt - opthead);
856 				return(-1);
857 			}
858 
859 			/*
860 			 * We may see jumbolen in unaligned location, so
861 			 * we'd need to perform bcopy().
862 			 */
863 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
864 			jumboplen = (u_int32_t)htonl(jumboplen);
865 
866 #if 1
867 			/*
868 			 * if there are multiple jumbo payload options,
869 			 * *plenp will be non-zero and the packet will be
870 			 * rejected.
871 			 * the behavior may need some debate in ipngwg -
872 			 * multiple options does not make sense, however,
873 			 * there's no explicit mention in specification.
874 			 */
875 			if (*plenp != 0) {
876 				ip6stat.ip6s_badoptions++;
877 				icmp6_error(m, ICMP6_PARAM_PROB,
878 					    ICMP6_PARAMPROB_HEADER,
879 					    sizeof(struct ip6_hdr) +
880 					    sizeof(struct ip6_hbh) +
881 					    opt + 2 - opthead);
882 				return(-1);
883 			}
884 #endif
885 
886 			/*
887 			 * jumbo payload length must be larger than 65535.
888 			 */
889 			if (jumboplen <= IPV6_MAXPACKET) {
890 				ip6stat.ip6s_badoptions++;
891 				icmp6_error(m, ICMP6_PARAM_PROB,
892 					    ICMP6_PARAMPROB_HEADER,
893 					    sizeof(struct ip6_hdr) +
894 					    sizeof(struct ip6_hbh) +
895 					    opt + 2 - opthead);
896 				return(-1);
897 			}
898 			*plenp = jumboplen;
899 
900 			break;
901 		default:		/* unknown option */
902 			if (hbhlen < IP6OPT_MINLEN) {
903 				ip6stat.ip6s_toosmall++;
904 				goto bad;
905 			}
906 			if ((optlen = ip6_unknown_opt(opt, m,
907 						      sizeof(struct ip6_hdr) +
908 						      sizeof(struct ip6_hbh) +
909 						      opt - opthead)) == -1)
910 				return(-1);
911 			optlen += 2;
912 			break;
913 		}
914 	}
915 
916 	return(0);
917 
918   bad:
919 	m_freem(m);
920 	return(-1);
921 }
922 
923 /*
924  * Unknown option processing.
925  * The third argument `off' is the offset from the IPv6 header to the option,
926  * which is necessary if the IPv6 header the and option header and IPv6 header
927  * is not continuous in order to return an ICMPv6 error.
928  */
929 int
930 ip6_unknown_opt(optp, m, off)
931 	u_int8_t *optp;
932 	struct mbuf *m;
933 	int off;
934 {
935 	struct ip6_hdr *ip6;
936 
937 	switch (IP6OPT_TYPE(*optp)) {
938 	case IP6OPT_TYPE_SKIP: /* ignore the option */
939 		return((int)*(optp + 1));
940 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
941 		m_freem(m);
942 		return(-1);
943 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
944 		ip6stat.ip6s_badoptions++;
945 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
946 		return(-1);
947 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
948 		ip6stat.ip6s_badoptions++;
949 		ip6 = mtod(m, struct ip6_hdr *);
950 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
951 		    (m->m_flags & (M_BCAST|M_MCAST)))
952 			m_freem(m);
953 		else
954 			icmp6_error(m, ICMP6_PARAM_PROB,
955 				    ICMP6_PARAMPROB_OPTION, off);
956 		return(-1);
957 	}
958 
959 	m_freem(m);		/* XXX: NOTREACHED */
960 	return(-1);
961 }
962 
963 /*
964  * Create the "control" list for this pcb.
965  *
966  * The routine will be called from upper layer handlers like tcp6_input().
967  * Thus the routine assumes that the caller (tcp6_input) have already
968  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
969  * very first mbuf on the mbuf chain.
970  * We may want to add some infinite loop prevention or sanity checks for safety.
971  * (This applies only when you are using KAME mbuf chain restriction, i.e.
972  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
973  */
974 void
975 ip6_savecontrol(in6p, mp, ip6, m)
976 	struct in6pcb *in6p;
977 	struct mbuf **mp;
978 	struct ip6_hdr *ip6;
979 	struct mbuf *m;
980 {
981 	struct proc *p = curproc;	/* XXX */
982 	int privileged;
983 
984 	privileged = 0;
985 	if (p && !suser(p->p_ucred, &p->p_acflag))
986 		privileged++;
987 
988 #ifdef SO_TIMESTAMP
989 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
990 		struct timeval tv;
991 
992 		microtime(&tv);
993 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
994 			SCM_TIMESTAMP, SOL_SOCKET);
995 		if (*mp)
996 			mp = &(*mp)->m_next;
997 	}
998 #endif
999 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
1000 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
1001 			sizeof(struct in6_addr), IPV6_RECVDSTADDR,
1002 			IPPROTO_IPV6);
1003 		if (*mp)
1004 			mp = &(*mp)->m_next;
1005 	}
1006 
1007 #ifdef noyet
1008 	/* options were tossed above */
1009 	if (in6p->in6p_flags & IN6P_RECVOPTS)
1010 		/* broken */
1011 	/* ip6_srcroute doesn't do what we want here, need to fix */
1012 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
1013 		/* broken */
1014 #endif
1015 
1016 	/* RFC 2292 sec. 5 */
1017 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1018 		struct in6_pktinfo pi6;
1019 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1020 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1021 			pi6.ipi6_addr.s6_addr16[1] = 0;
1022 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1023 					? m->m_pkthdr.rcvif->if_index
1024 					: 0;
1025 		*mp = sbcreatecontrol((caddr_t) &pi6,
1026 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1027 			IPPROTO_IPV6);
1028 		if (*mp)
1029 			mp = &(*mp)->m_next;
1030 	}
1031 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1032 		int hlim = ip6->ip6_hlim & 0xff;
1033 		*mp = sbcreatecontrol((caddr_t) &hlim,
1034 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1035 		if (*mp)
1036 			mp = &(*mp)->m_next;
1037 	}
1038 	/* IN6P_NEXTHOP - for outgoing packet only */
1039 
1040 	/*
1041 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1042 	 * for the option, but it might be too strict, since there might
1043 	 * be some hop-by-hop options which can be returned to normal user.
1044 	 * See RFC 2292 section 6.
1045 	 */
1046 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1047 		/*
1048 		 * Check if a hop-by-hop options header is contatined in the
1049 		 * received packet, and if so, store the options as ancillary
1050 		 * data. Note that a hop-by-hop options header must be
1051 		 * just after the IPv6 header, which fact is assured through
1052 		 * the IPv6 input processing.
1053 		 */
1054 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1055 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1056 			struct ip6_hbh *hbh;
1057 			int hbhlen;
1058 
1059 #ifndef PULLDOWN_TEST
1060 			hbh = (struct ip6_hbh *)(ip6 + 1);
1061 			hbhlen = (hbh->ip6h_len + 1) << 3;
1062 #else
1063 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1064 				sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1065 			if (hbh == NULL) {
1066 				ip6stat.ip6s_tooshort++;
1067 				return;
1068 			}
1069 			hbhlen = (hbh->ip6h_len + 1) << 3;
1070 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1071 				sizeof(struct ip6_hdr), hbhlen);
1072 			if (hbh == NULL) {
1073 				ip6stat.ip6s_tooshort++;
1074 				return;
1075 			}
1076 #endif
1077 
1078 			/*
1079 			 * XXX: We copy whole the header even if a jumbo
1080 			 * payload option is included, which option is to
1081 			 * be removed before returning in the RFC 2292.
1082 			 * But it's too painful operation...
1083 			 */
1084 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1085 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1086 			if (*mp)
1087 				mp = &(*mp)->m_next;
1088 		}
1089 	}
1090 
1091 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1092 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1093 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1094 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);;
1095 
1096 		/*
1097 		 * Search for destination options headers or routing
1098 		 * header(s) through the header chain, and stores each
1099 		 * header as ancillary data.
1100 		 * Note that the order of the headers remains in
1101 		 * the chain of ancillary data.
1102 		 */
1103 		while (1) {	/* is explicit loop prevention necessary? */
1104 			struct ip6_ext *ip6e;
1105 			int elen;
1106 
1107 #ifndef PULLDOWN_TEST
1108 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1109 			if (nxt == IPPROTO_AH)
1110 				elen = (ip6e->ip6e_len + 2) << 2;
1111 			else
1112 				elen = (ip6e->ip6e_len + 1) << 3;
1113 #else
1114 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off,
1115 				sizeof(struct ip6_ext));
1116 			if (ip6e == NULL) {
1117 				ip6stat.ip6s_tooshort++;
1118 				return;
1119 			}
1120 			if (nxt == IPPROTO_AH)
1121 				elen = (ip6e->ip6e_len + 2) << 2;
1122 			else
1123 				elen = (ip6e->ip6e_len + 1) << 3;
1124 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off, elen);
1125 			if (ip6e == NULL) {
1126 				ip6stat.ip6s_tooshort++;
1127 				return;
1128 			}
1129 #endif
1130 
1131 			switch (nxt) {
1132 		        case IPPROTO_DSTOPTS:
1133 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1134 					break;
1135 
1136 				/*
1137 				 * We also require super-user privilege for
1138 				 * the option.
1139 				 * See the comments on IN6_HOPOPTS.
1140 				 */
1141 				if (!privileged)
1142 					break;
1143 
1144 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1145 						      IPV6_DSTOPTS,
1146 						      IPPROTO_IPV6);
1147 				if (*mp)
1148 					mp = &(*mp)->m_next;
1149 				break;
1150 
1151 			case IPPROTO_ROUTING:
1152 				if (!in6p->in6p_flags & IN6P_RTHDR)
1153 					break;
1154 
1155 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1156 						      IPV6_RTHDR,
1157 						      IPPROTO_IPV6);
1158 				if (*mp)
1159 					mp = &(*mp)->m_next;
1160 				break;
1161 
1162 			case IPPROTO_UDP:
1163 			case IPPROTO_TCP:
1164 			case IPPROTO_ICMPV6:
1165 			default:
1166 				/*
1167 				 * stop search if we encounter an upper
1168 				 * layer protocol headers.
1169 				 */
1170 				goto loopend;
1171 
1172 			case IPPROTO_HOPOPTS:
1173 			case IPPROTO_AH: /* is it possible? */
1174 				break;
1175 			}
1176 
1177 			/* proceed with the next header. */
1178 			off += elen;
1179 			nxt = ip6e->ip6e_nxt;
1180 		}
1181 	  loopend:
1182 		;
1183 	}
1184 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1185 		/* to be done */
1186 	}
1187 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1188 		/* to be done */
1189 	}
1190 	/* IN6P_RTHDR - to be done */
1191 
1192 }
1193 
1194 /*
1195  * Get pointer to the previous header followed by the header
1196  * currently processed.
1197  * XXX: This function supposes that
1198  *	M includes all headers,
1199  *	the next header field and the header length field of each header
1200  *	are valid, and
1201  *	the sum of each header length equals to OFF.
1202  * Because of these assumptions, this function must be called very
1203  * carefully. Moreover, it will not be used in the near future when
1204  * we develop `neater' mechanism to process extension headers.
1205  */
1206 char *
1207 ip6_get_prevhdr(m, off)
1208 	struct mbuf *m;
1209 	int off;
1210 {
1211 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1212 
1213 	if (off == sizeof(struct ip6_hdr))
1214 		return(&ip6->ip6_nxt);
1215 	else {
1216 		int len, nxt;
1217 		struct ip6_ext *ip6e = NULL;
1218 
1219 		nxt = ip6->ip6_nxt;
1220 		len = sizeof(struct ip6_hdr);
1221 		while (len < off) {
1222 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1223 
1224 			switch (nxt) {
1225 			case IPPROTO_FRAGMENT:
1226 				len += sizeof(struct ip6_frag);
1227 				break;
1228 			case IPPROTO_AH:
1229 				len += (ip6e->ip6e_len + 2) << 2;
1230 				break;
1231 			default:
1232 				len += (ip6e->ip6e_len + 1) << 3;
1233 				break;
1234 			}
1235 			nxt = ip6e->ip6e_nxt;
1236 		}
1237 		if (ip6e)
1238 			return(&ip6e->ip6e_nxt);
1239 		else
1240 			return NULL;
1241 	}
1242 }
1243 
1244 /*
1245  * get next header offset.  m will be retained.
1246  */
1247 int
1248 ip6_nexthdr(m, off, proto, nxtp)
1249 	struct mbuf *m;
1250 	int off;
1251 	int proto;
1252 	int *nxtp;
1253 {
1254 	struct ip6_hdr ip6;
1255 	struct ip6_ext ip6e;
1256 	struct ip6_frag fh;
1257 
1258 	/* just in case */
1259 	if (m == NULL)
1260 		panic("ip6_nexthdr: m == NULL");
1261 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1262 		return -1;
1263 
1264 	switch (proto) {
1265 	case IPPROTO_IPV6:
1266 		if (m->m_pkthdr.len < off + sizeof(ip6))
1267 			return -1;
1268 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1269 		if (nxtp)
1270 			*nxtp = ip6.ip6_nxt;
1271 		off += sizeof(ip6);
1272 		return off;
1273 
1274 	case IPPROTO_FRAGMENT:
1275 		/*
1276 		 * terminate parsing if it is not the first fragment,
1277 		 * it does not make sense to parse through it.
1278 		 */
1279 		if (m->m_pkthdr.len < off + sizeof(fh))
1280 			return -1;
1281 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1282 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1283 			return -1;
1284 		if (nxtp)
1285 			*nxtp = fh.ip6f_nxt;
1286 		off += sizeof(struct ip6_frag);
1287 		return off;
1288 
1289 	case IPPROTO_AH:
1290 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1291 			return -1;
1292 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1293 		if (nxtp)
1294 			*nxtp = ip6e.ip6e_nxt;
1295 		off += (ip6e.ip6e_len + 2) << 2;
1296 		return off;
1297 
1298 	case IPPROTO_HOPOPTS:
1299 	case IPPROTO_ROUTING:
1300 	case IPPROTO_DSTOPTS:
1301 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1302 			return -1;
1303 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1304 		if (nxtp)
1305 			*nxtp = ip6e.ip6e_nxt;
1306 		off += (ip6e.ip6e_len + 1) << 3;
1307 		return off;
1308 
1309 	case IPPROTO_NONE:
1310 	case IPPROTO_ESP:
1311 	case IPPROTO_IPCOMP:
1312 		/* give up */
1313 		return -1;
1314 
1315 	default:
1316 		return -1;
1317 	}
1318 
1319 	return -1;
1320 }
1321 
1322 /*
1323  * get offset for the last header in the chain.  m will be kept untainted.
1324  */
1325 int
1326 ip6_lasthdr(m, off, proto, nxtp)
1327 	struct mbuf *m;
1328 	int off;
1329 	int proto;
1330 	int *nxtp;
1331 {
1332 	int newoff;
1333 	int nxt;
1334 
1335 	if (!nxtp) {
1336 		nxt = -1;
1337 		nxtp = &nxt;
1338 	}
1339 	while (1) {
1340 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1341 		if (newoff < 0)
1342 			return off;
1343 		else if (newoff < off)
1344 			return -1;	/* invalid */
1345 		else if (newoff == off)
1346 			return newoff;
1347 
1348 		off = newoff;
1349 		proto = *nxtp;
1350 	}
1351 }
1352 
1353 /*
1354  * System control for IP6
1355  */
1356 
1357 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1358 	0,		0,		0,		0,
1359 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1360 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1361 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1362 	0,		0,		0,		0,
1363 	ENOPROTOOPT
1364 };
1365 
1366 #include <uvm/uvm_extern.h>
1367 #include <sys/sysctl.h>
1368 
1369 int
1370 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1371 	int *name;
1372 	u_int namelen;
1373 	void *oldp;
1374 	size_t *oldlenp;
1375 	void *newp;
1376 	size_t newlen;
1377 {
1378 	int old, error;
1379 
1380 	/* All sysctl names at this level are terminal. */
1381 	if (namelen != 1)
1382 		return ENOTDIR;
1383 
1384 	switch (name[0]) {
1385 
1386 	case IPV6CTL_FORWARDING:
1387 		return sysctl_int(oldp, oldlenp, newp, newlen,
1388 				  &ip6_forwarding);
1389 	case IPV6CTL_SENDREDIRECTS:
1390 		return sysctl_int(oldp, oldlenp, newp, newlen,
1391 				&ip6_sendredirects);
1392 	case IPV6CTL_DEFHLIM:
1393 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1394 	case IPV6CTL_MAXFRAGPACKETS:
1395 		return sysctl_int(oldp, oldlenp, newp, newlen,
1396 				&ip6_maxfragpackets);
1397 	case IPV6CTL_ACCEPT_RTADV:
1398 		return sysctl_int(oldp, oldlenp, newp, newlen,
1399 				&ip6_accept_rtadv);
1400 	case IPV6CTL_KEEPFAITH:
1401 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1402 	case IPV6CTL_LOG_INTERVAL:
1403 		return sysctl_int(oldp, oldlenp, newp, newlen,
1404 				&ip6_log_interval);
1405 	case IPV6CTL_HDRNESTLIMIT:
1406 		return sysctl_int(oldp, oldlenp, newp, newlen,
1407 				&ip6_hdrnestlimit);
1408 	case IPV6CTL_DAD_COUNT:
1409 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1410 	case IPV6CTL_AUTO_FLOWLABEL:
1411 		return sysctl_int(oldp, oldlenp, newp, newlen,
1412 				&ip6_auto_flowlabel);
1413 	case IPV6CTL_DEFMCASTHLIM:
1414 		return sysctl_int(oldp, oldlenp, newp, newlen,
1415 				&ip6_defmcasthlim);
1416 	case IPV6CTL_GIF_HLIM:
1417 		return sysctl_int(oldp, oldlenp, newp, newlen,
1418 				&ip6_gif_hlim);
1419 	case IPV6CTL_KAME_VERSION:
1420 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1421 	case IPV6CTL_USE_DEPRECATED:
1422 		return sysctl_int(oldp, oldlenp, newp, newlen,
1423 				&ip6_use_deprecated);
1424 	case IPV6CTL_RR_PRUNE:
1425 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1426 #ifndef INET6_BINDV6ONLY
1427 	case IPV6CTL_BINDV6ONLY:
1428 		return sysctl_int(oldp, oldlenp, newp, newlen,
1429 				&ip6_bindv6only);
1430 #endif
1431 	case IPV6CTL_ANONPORTMIN:
1432 		old = ip6_anonportmin;
1433 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1434 		    &ip6_anonportmin);
1435 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmin < 0 ||
1436 		    ip6_anonportmin > 65535
1437 #ifndef IPNOPRIVPORTS
1438 		    || ip6_anonportmin < IPV6PORT_RESERVED
1439 #endif
1440 		    ) {
1441 			ip6_anonportmin = old;
1442 			return (EINVAL);
1443 		}
1444 		return (error);
1445 	case IPV6CTL_ANONPORTMAX:
1446 		old = ip6_anonportmax;
1447 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1448 		    &ip6_anonportmax);
1449 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmax < 0 ||
1450 		    ip6_anonportmax > 65535
1451 #ifndef IPNOPRIVPORTS
1452 		    || ip6_anonportmax < IPV6PORT_RESERVED
1453 #endif
1454 		    ) {
1455 			ip6_anonportmax = old;
1456 			return (EINVAL);
1457 		}
1458 		return (error);
1459 #ifndef IPNOPRIVPORTS
1460 	case IPV6CTL_LOWPORTMIN:
1461 		old = ip6_lowportmin;
1462 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1463 		    &ip6_lowportmin);
1464 		if (ip6_lowportmin >= ip6_lowportmax ||
1465 		    ip6_lowportmin > IPV6PORT_RESERVEDMAX ||
1466 		    ip6_lowportmin < IPV6PORT_RESERVEDMIN) {
1467 			ip6_lowportmin = old;
1468 			return (EINVAL);
1469 		}
1470 		return (error);
1471 	case IPV6CTL_LOWPORTMAX:
1472 		old = ip6_lowportmax;
1473 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1474 		    &ip6_lowportmax);
1475 		if (ip6_lowportmin >= ip6_lowportmax ||
1476 		    ip6_lowportmax > IPV6PORT_RESERVEDMAX ||
1477 		    ip6_lowportmax < IPV6PORT_RESERVEDMIN) {
1478 			ip6_lowportmax = old;
1479 			return (EINVAL);
1480 		}
1481 		return (error);
1482 #endif
1483 	default:
1484 		return EOPNOTSUPP;
1485 	}
1486 	/* NOTREACHED */
1487 }
1488