xref: /netbsd-src/sys/netinet6/ip6_input.c (revision 1ca5c1b28139779176bd5c13ad7c5f25c0bcd5f8)
1 /*	$NetBSD: ip6_input.c,v 1.51 2001/12/22 01:40:03 itojun Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.188 2001/03/29 05:34:31 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 <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: ip6_input.c,v 1.51 2001/12/22 01:40:03 itojun Exp $");
70 
71 #include "opt_inet.h"
72 #include "opt_ipsec.h"
73 #include "opt_pfil_hooks.h"
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 #include <sys/proc.h>
88 #include <sys/sysctl.h>
89 
90 #include <net/if.h>
91 #include <net/if_types.h>
92 #include <net/if_dl.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #ifdef PFIL_HOOKS
96 #include <net/pfil.h>
97 #endif
98 
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/in6_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/in6_ifattach.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_prefix.h>
113 
114 #ifdef IPSEC
115 #include <netinet6/ipsec.h>
116 #endif
117 
118 #include <netinet6/ip6protosw.h>
119 
120 /* we need it for NLOOP. */
121 #include "loop.h"
122 #include "faith.h"
123 #include "gif.h"
124 #include "bpfilter.h"
125 
126 #if NGIF > 0
127 #include <netinet6/in6_gif.h>
128 #endif
129 
130 #include <net/net_osdep.h>
131 
132 extern struct domain inet6domain;
133 
134 u_char ip6_protox[IPPROTO_MAX];
135 static int ip6qmaxlen = IFQ_MAXLEN;
136 struct in6_ifaddr *in6_ifaddr;
137 struct ifqueue ip6intrq;
138 
139 extern struct ifnet loif[NLOOP];
140 int ip6_forward_srcrt;			/* XXX */
141 int ip6_sourcecheck;			/* XXX */
142 int ip6_sourcecheck_interval;		/* XXX */
143 
144 #ifdef PFIL_HOOKS
145 struct pfil_head inet6_pfil_hook;
146 #endif
147 
148 struct ip6stat ip6stat;
149 
150 static void ip6_init2 __P((void *));
151 
152 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
153 
154 /*
155  * IP6 initialization: fill in IP6 protocol switch table.
156  * All protocols not implemented in kernel go to raw IP6 protocol handler.
157  */
158 void
159 ip6_init()
160 {
161 	struct ip6protosw *pr;
162 	int i;
163 	struct timeval tv;
164 
165 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
166 	if (pr == 0)
167 		panic("ip6_init");
168 	for (i = 0; i < IPPROTO_MAX; i++)
169 		ip6_protox[i] = pr - inet6sw;
170 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
171 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
172 		if (pr->pr_domain->dom_family == PF_INET6 &&
173 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
174 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
175 	ip6intrq.ifq_maxlen = ip6qmaxlen;
176 	nd6_init();
177 	frag6_init();
178 	/*
179 	 * in many cases, random() here does NOT return random number
180 	 * as initialization during bootstrap time occur in fixed order.
181 	 */
182 	microtime(&tv);
183 	ip6_flow_seq = random() ^ tv.tv_usec;
184 
185 	ip6_init2((void *)0);
186 
187 #ifdef PFIL_HOOKS
188 	/* Register our Packet Filter hook. */
189 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
190 	inet6_pfil_hook.ph_af   = AF_INET6;
191 	i = pfil_head_register(&inet6_pfil_hook);
192 	if (i != 0)
193 		printf("ip6_init: WARNING: unable to register pfil hook, "
194 		    "error %d\n", i);
195 #endif /* PFIL_HOOKS */
196 }
197 
198 static void
199 ip6_init2(dummy)
200 	void *dummy;
201 {
202 	/*
203 	 * to route local address of p2p link to loopback,
204 	 * assign loopback address first.
205 	 */
206 	in6_ifattach(&loif[0], NULL);
207 
208 	/* nd6_timer_init */
209 	callout_init(&nd6_timer_ch);
210 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
211 	/* router renumbering prefix list maintenance */
212 	callout_init(&in6_rr_timer_ch);
213 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
214 }
215 
216 /*
217  * IP6 input interrupt handling. Just pass the packet to ip6_input.
218  */
219 void
220 ip6intr()
221 {
222 	int s;
223 	struct mbuf *m;
224 
225 	for (;;) {
226 		s = splnet();
227 		IF_DEQUEUE(&ip6intrq, m);
228 		splx(s);
229 		if (m == 0)
230 			return;
231 		ip6_input(m);
232 	}
233 }
234 
235 extern struct	route_in6 ip6_forward_rt;
236 
237 void
238 ip6_input(m)
239 	struct mbuf *m;
240 {
241 	struct ip6_hdr *ip6;
242 	int off = sizeof(struct ip6_hdr), nest;
243 	u_int32_t plen;
244 	u_int32_t rtalert = ~0;
245 	int nxt, ours = 0;
246 	struct ifnet *deliverifp = NULL;
247 
248 #ifdef IPSEC
249 	/*
250 	 * should the inner packet be considered authentic?
251 	 * see comment in ah4_input().
252 	 */
253 	m->m_flags &= ~M_AUTHIPHDR;
254 	m->m_flags &= ~M_AUTHIPDGM;
255 #endif
256 
257 	/*
258 	 * mbuf statistics
259 	 */
260 	if (m->m_flags & M_EXT) {
261 		if (m->m_next)
262 			ip6stat.ip6s_mext2m++;
263 		else
264 			ip6stat.ip6s_mext1++;
265 	} else {
266 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
267 		if (m->m_next) {
268 			if (m->m_flags & M_LOOP) {
269 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
270 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
271 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
272 			else
273 				ip6stat.ip6s_m2m[0]++;
274 		} else
275 			ip6stat.ip6s_m1++;
276 #undef M2MMAX
277 	}
278 
279 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
280 	ip6stat.ip6s_total++;
281 
282 #ifndef PULLDOWN_TEST
283 	/* XXX is the line really necessary? */
284 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
285 #endif
286 
287 	if (m->m_len < sizeof(struct ip6_hdr)) {
288 		struct ifnet *inifp;
289 		inifp = m->m_pkthdr.rcvif;
290 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
291 			ip6stat.ip6s_toosmall++;
292 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
293 			return;
294 		}
295 	}
296 
297 	ip6 = mtod(m, struct ip6_hdr *);
298 
299 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
300 		ip6stat.ip6s_badvers++;
301 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
302 		goto bad;
303 	}
304 
305 #ifdef PFIL_HOOKS
306 	/*
307 	 * Run through list of hooks for input packets.  If there are any
308 	 * filters which require that additional packets in the flow are
309 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
310 	 * Note that filters must _never_ set this flag, as another filter
311 	 * in the list may have previously cleared it.
312 	 */
313 	/*
314 	 * let ipfilter look at packet on the wire,
315 	 * not the decapsulated packet.
316 	 */
317 #ifdef IPSEC
318 	if (!ipsec_getnhist(m))
319 #else
320 	if (1)
321 #endif
322 	{
323 		if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif,
324 				   PFIL_IN) != 0)
325 			return;
326 		if (m == NULL)
327 			return;
328 		ip6 = mtod(m, struct ip6_hdr *);
329 	}
330 #endif /* PFIL_HOOKS */
331 
332 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
333 
334 #ifdef ALTQ
335 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
336 		/* packet is dropped by traffic conditioner */
337 		return;
338 	}
339 #endif
340 
341 	/*
342 	 * Check against address spoofing/corruption.
343 	 */
344 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
345 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
346 		ip6stat.ip6s_badscope++;
347 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
348 		goto bad;
349 	}
350 	/*
351 	 * The following check is not documented in specs.  A malicious
352 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
353 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
354 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
355 	 *
356 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
357 	 * support IPv4-less kernel compilation, we cannot support SIIT
358 	 * environment at all.  So, it makes more sense for us to reject any
359 	 * malicious packets for non-SIIT environment, than try to do a
360 	 * partical support for SIIT environment.
361 	 */
362 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
363 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
364 		ip6stat.ip6s_badscope++;
365 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
366 		goto bad;
367 	}
368 #if 0
369 	/*
370 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
371 	 *
372 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
373 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
374 	 * is revised to forbid relaying case.
375 	 */
376 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
377 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
378 		ip6stat.ip6s_badscope++;
379 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
380 		goto bad;
381 	}
382 #endif
383 
384 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
385 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
386 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
387 			ours = 1;
388 			deliverifp = m->m_pkthdr.rcvif;
389 			goto hbhcheck;
390 		} else {
391 			ip6stat.ip6s_badscope++;
392 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
393 			goto bad;
394 		}
395 	}
396 
397 	/* drop packets if interface ID portion is already filled */
398 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
399 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
400 		    ip6->ip6_src.s6_addr16[1]) {
401 			ip6stat.ip6s_badscope++;
402 			goto bad;
403 		}
404 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
405 		    ip6->ip6_dst.s6_addr16[1]) {
406 			ip6stat.ip6s_badscope++;
407 			goto bad;
408 		}
409 	}
410 
411 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
412 		ip6->ip6_src.s6_addr16[1]
413 			= htons(m->m_pkthdr.rcvif->if_index);
414 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
415 		ip6->ip6_dst.s6_addr16[1]
416 			= htons(m->m_pkthdr.rcvif->if_index);
417 
418 	/*
419 	 * We use rt->rt_ifp to determine if the address is ours or not.
420 	 * If rt_ifp is lo0, the address is ours.
421 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
422 	 * so any address under fe80::%lo0/64 will be mistakenly considered
423 	 * local.  The special case is supplied to handle the case properly
424 	 * by actually looking at interface addresses
425 	 * (using in6ifa_ifpwithaddr).
426 	 */
427 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
428 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
429 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
430 			icmp6_error(m, ICMP6_DST_UNREACH,
431 			    ICMP6_DST_UNREACH_ADDR, 0);
432 			/* m is already freed */
433 			return;
434 		}
435 
436 		ours = 1;
437 		deliverifp = m->m_pkthdr.rcvif;
438 		goto hbhcheck;
439 	}
440 
441 	/*
442 	 * Multicast check
443 	 */
444 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
445 	  	struct	in6_multi *in6m = 0;
446 
447 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
448 		/*
449 		 * See if we belong to the destination multicast group on the
450 		 * arrival interface.
451 		 */
452 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
453 		if (in6m)
454 			ours = 1;
455 		else if (!ip6_mrouter) {
456 			ip6stat.ip6s_notmember++;
457 			ip6stat.ip6s_cantforward++;
458 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
459 			goto bad;
460 		}
461 		deliverifp = m->m_pkthdr.rcvif;
462 		goto hbhcheck;
463 	}
464 
465 	/*
466 	 *  Unicast check
467 	 */
468 	if (ip6_forward_rt.ro_rt != NULL &&
469 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
470 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
471 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
472 		ip6stat.ip6s_forward_cachehit++;
473 	else {
474 		struct sockaddr_in6 *dst6;
475 
476 		if (ip6_forward_rt.ro_rt) {
477 			/* route is down or destination is different */
478 			ip6stat.ip6s_forward_cachemiss++;
479 			RTFREE(ip6_forward_rt.ro_rt);
480 			ip6_forward_rt.ro_rt = 0;
481 		}
482 
483 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
484 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
485 		dst6->sin6_len = sizeof(struct sockaddr_in6);
486 		dst6->sin6_family = AF_INET6;
487 		dst6->sin6_addr = ip6->ip6_dst;
488 
489 		rtalloc((struct route *)&ip6_forward_rt);
490 	}
491 
492 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
493 
494 	/*
495 	 * Accept the packet if the forwarding interface to the destination
496 	 * according to the routing table is the loopback interface,
497 	 * unless the associated route has a gateway.
498 	 * Note that this approach causes to accept a packet if there is a
499 	 * route to the loopback interface for the destination of the packet.
500 	 * But we think it's even useful in some situations, e.g. when using
501 	 * a special daemon which wants to intercept the packet.
502 	 */
503 	if (ip6_forward_rt.ro_rt &&
504 	    (ip6_forward_rt.ro_rt->rt_flags &
505 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
506 #if 0
507 	    /*
508 	     * The check below is redundant since the comparison of
509 	     * the destination and the key of the rtentry has
510 	     * already done through looking up the routing table.
511 	     */
512 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
513 			       &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
514 #endif
515 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
516 		struct in6_ifaddr *ia6 =
517 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
518 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
519 			m->m_flags |= M_ANYCAST6;
520 		/*
521 		 * packets to a tentative, duplicated, or somehow invalid
522 		 * address must not be accepted.
523 		 */
524 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
525 			/* this address is ready */
526 			ours = 1;
527 			deliverifp = ia6->ia_ifp;	/* correct? */
528 			goto hbhcheck;
529 		} else {
530 			/* address is not ready, so discard the packet. */
531 			nd6log((LOG_INFO,
532 			    "ip6_input: packet to an unready address %s->%s\n",
533 			    ip6_sprintf(&ip6->ip6_src),
534 			    ip6_sprintf(&ip6->ip6_dst)));
535 
536 			goto bad;
537 		}
538 	}
539 
540 	/*
541 	 * FAITH(Firewall Aided Internet Translator)
542 	 */
543 #if defined(NFAITH) && 0 < NFAITH
544 	if (ip6_keepfaith) {
545 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
546 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
547 			/* XXX do we need more sanity checks? */
548 			ours = 1;
549 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
550 			goto hbhcheck;
551 		}
552 	}
553 #endif
554 
555 #if 0
556     {
557 	/*
558 	 * Last resort: check in6_ifaddr for incoming interface.
559 	 * The code is here until I update the "goto ours hack" code above
560 	 * working right.
561 	 */
562 	struct ifaddr *ifa;
563 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
564 	     ifa;
565 	     ifa = ifa->ifa_list.tqe_next) {
566 		if (ifa->ifa_addr == NULL)
567 			continue;	/* just for safety */
568 		if (ifa->ifa_addr->sa_family != AF_INET6)
569 			continue;
570 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
571 			ours = 1;
572 			deliverifp = ifa->ifa_ifp;
573 			goto hbhcheck;
574 		}
575 	}
576     }
577 #endif
578 
579 	/*
580 	 * Now there is no reason to process the packet if it's not our own
581 	 * and we're not a router.
582 	 */
583 	if (!ip6_forwarding) {
584 		ip6stat.ip6s_cantforward++;
585 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
586 		goto bad;
587 	}
588 
589   hbhcheck:
590 	/*
591 	 * Process Hop-by-Hop options header if it's contained.
592 	 * m may be modified in ip6_hopopts_input().
593 	 * If a JumboPayload option is included, plen will also be modified.
594 	 */
595 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
596 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
597 		struct ip6_hbh *hbh;
598 
599 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
600 #if 0	/*touches NULL pointer*/
601 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
602 #endif
603 			return;	/* m have already been freed */
604 		}
605 
606 		/* adjust pointer */
607 		ip6 = mtod(m, struct ip6_hdr *);
608 
609 		/*
610 		 * if the payload length field is 0 and the next header field
611 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
612 		 * option MUST be included.
613 		 */
614 		if (ip6->ip6_plen == 0 && plen == 0) {
615 			/*
616 			 * Note that if a valid jumbo payload option is
617 			 * contained, ip6_hoptops_input() must set a valid
618 			 * (non-zero) payload length to the variable plen.
619 			 */
620 			ip6stat.ip6s_badoptions++;
621 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
622 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
623 			icmp6_error(m, ICMP6_PARAM_PROB,
624 				    ICMP6_PARAMPROB_HEADER,
625 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
626 			return;
627 		}
628 #ifndef PULLDOWN_TEST
629 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
630 		hbh = (struct ip6_hbh *)(ip6 + 1);
631 #else
632 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
633 			sizeof(struct ip6_hbh));
634 		if (hbh == NULL) {
635 			ip6stat.ip6s_tooshort++;
636 			return;
637 		}
638 #endif
639 		nxt = hbh->ip6h_nxt;
640 
641 		/*
642 		 * accept the packet if a router alert option is included
643 		 * and we act as an IPv6 router.
644 		 */
645 		if (rtalert != ~0 && ip6_forwarding)
646 			ours = 1;
647 	} else
648 		nxt = ip6->ip6_nxt;
649 
650 	/*
651 	 * Check that the amount of data in the buffers
652 	 * is as at least much as the IPv6 header would have us expect.
653 	 * Trim mbufs if longer than we expect.
654 	 * Drop packet if shorter than we expect.
655 	 */
656 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
657 		ip6stat.ip6s_tooshort++;
658 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
659 		goto bad;
660 	}
661 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
662 		if (m->m_len == m->m_pkthdr.len) {
663 			m->m_len = sizeof(struct ip6_hdr) + plen;
664 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
665 		} else
666 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
667 	}
668 
669 	/*
670 	 * Forward if desirable.
671 	 */
672 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
673 		/*
674 		 * If we are acting as a multicast router, all
675 		 * incoming multicast packets are passed to the
676 		 * kernel-level multicast forwarding function.
677 		 * The packet is returned (relatively) intact; if
678 		 * ip6_mforward() returns a non-zero value, the packet
679 		 * must be discarded, else it may be accepted below.
680 		 */
681 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
682 			ip6stat.ip6s_cantforward++;
683 			m_freem(m);
684 			return;
685 		}
686 		if (!ours) {
687 			m_freem(m);
688 			return;
689 		}
690 	} else if (!ours) {
691 		ip6_forward(m, 0);
692 		return;
693 	}
694 
695 	ip6 = mtod(m, struct ip6_hdr *);
696 
697 	/*
698 	 * Malicious party may be able to use IPv4 mapped addr to confuse
699 	 * tcp/udp stack and bypass security checks (act as if it was from
700 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
701 	 *
702 	 * For SIIT end node behavior, you may want to disable the check.
703 	 * However, you will  become vulnerable to attacks using IPv4 mapped
704 	 * source.
705 	 */
706 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
707 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
708 		ip6stat.ip6s_badscope++;
709 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
710 		goto bad;
711 	}
712 
713 	/*
714 	 * Tell launch routine the next header
715 	 */
716 #ifdef IFA_STATS
717 	if (deliverifp != NULL) {
718 		struct in6_ifaddr *ia6;
719 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
720 		if (ia6)
721 			ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
722 	}
723 #endif
724 	ip6stat.ip6s_delivered++;
725 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
726 	nest = 0;
727 
728 	while (nxt != IPPROTO_DONE) {
729 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
730 			ip6stat.ip6s_toomanyhdr++;
731 			goto bad;
732 		}
733 
734 		/*
735 		 * protection against faulty packet - there should be
736 		 * more sanity checks in header chain processing.
737 		 */
738 		if (m->m_pkthdr.len < off) {
739 			ip6stat.ip6s_tooshort++;
740 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
741 			goto bad;
742 		}
743 
744 #ifdef IPSEC
745 		/*
746 		 * enforce IPsec policy checking if we are seeing last header.
747 		 * note that we do not visit this with protocols with pcb layer
748 		 * code - like udp/tcp/raw ip.
749 		 */
750 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
751 		    ipsec6_in_reject(m, NULL)) {
752 			ipsec6stat.in_polvio++;
753 			goto bad;
754 		}
755 #endif
756 
757 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
758 	}
759 	return;
760  bad:
761 	m_freem(m);
762 }
763 
764 /*
765  * Hop-by-Hop options header processing. If a valid jumbo payload option is
766  * included, the real payload length will be stored in plenp.
767  */
768 static int
769 ip6_hopopts_input(plenp, rtalertp, mp, offp)
770 	u_int32_t *plenp;
771 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
772 	struct mbuf **mp;
773 	int *offp;
774 {
775 	struct mbuf *m = *mp;
776 	int off = *offp, hbhlen;
777 	struct ip6_hbh *hbh;
778 	u_int8_t *opt;
779 
780 	/* validation of the length of the header */
781 #ifndef PULLDOWN_TEST
782 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
783 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
784 	hbhlen = (hbh->ip6h_len + 1) << 3;
785 
786 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
787 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
788 #else
789 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
790 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
791 	if (hbh == NULL) {
792 		ip6stat.ip6s_tooshort++;
793 		return -1;
794 	}
795 	hbhlen = (hbh->ip6h_len + 1) << 3;
796 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
797 		hbhlen);
798 	if (hbh == NULL) {
799 		ip6stat.ip6s_tooshort++;
800 		return -1;
801 	}
802 #endif
803 	off += hbhlen;
804 	hbhlen -= sizeof(struct ip6_hbh);
805 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
806 
807 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
808 				hbhlen, rtalertp, plenp) < 0)
809 		return(-1);
810 
811 	*offp = off;
812 	*mp = m;
813 	return(0);
814 }
815 
816 /*
817  * Search header for all Hop-by-hop options and process each option.
818  * This function is separate from ip6_hopopts_input() in order to
819  * handle a case where the sending node itself process its hop-by-hop
820  * options header. In such a case, the function is called from ip6_output().
821  */
822 int
823 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
824 	struct mbuf *m;
825 	u_int8_t *opthead;
826 	int hbhlen;
827 	u_int32_t *rtalertp;
828 	u_int32_t *plenp;
829 {
830 	struct ip6_hdr *ip6;
831 	int optlen = 0;
832 	u_int8_t *opt = opthead;
833 	u_int16_t rtalert_val;
834 	u_int32_t jumboplen;
835 
836 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
837 		switch (*opt) {
838 		case IP6OPT_PAD1:
839 			optlen = 1;
840 			break;
841 		case IP6OPT_PADN:
842 			if (hbhlen < IP6OPT_MINLEN) {
843 				ip6stat.ip6s_toosmall++;
844 				goto bad;
845 			}
846 			optlen = *(opt + 1) + 2;
847 			break;
848 		case IP6OPT_RTALERT:
849 			/* XXX may need check for alignment */
850 			if (hbhlen < IP6OPT_RTALERT_LEN) {
851 				ip6stat.ip6s_toosmall++;
852 				goto bad;
853 			}
854 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
855 				 /* XXX: should we discard the packet? */
856 				log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
857 				    *(opt + 1));
858 			}
859 			optlen = IP6OPT_RTALERT_LEN;
860 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
861 			*rtalertp = ntohs(rtalert_val);
862 			break;
863 		case IP6OPT_JUMBO:
864 			/* XXX may need check for alignment */
865 			if (hbhlen < IP6OPT_JUMBO_LEN) {
866 				ip6stat.ip6s_toosmall++;
867 				goto bad;
868 			}
869 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
870 				/* XXX: should we discard the packet? */
871 				log(LOG_ERR, "length of jumbopayload opt "
872 				    "is inconsistent(%d)\n",
873 				    *(opt + 1));
874 			}
875 			optlen = IP6OPT_JUMBO_LEN;
876 
877 			/*
878 			 * IPv6 packets that have non 0 payload length
879 			 * must not contain a jumbo payload option.
880 			 */
881 			ip6 = mtod(m, struct ip6_hdr *);
882 			if (ip6->ip6_plen) {
883 				ip6stat.ip6s_badoptions++;
884 				icmp6_error(m, ICMP6_PARAM_PROB,
885 					    ICMP6_PARAMPROB_HEADER,
886 					    sizeof(struct ip6_hdr) +
887 					    sizeof(struct ip6_hbh) +
888 					    opt - opthead);
889 				return(-1);
890 			}
891 
892 			/*
893 			 * We may see jumbolen in unaligned location, so
894 			 * we'd need to perform bcopy().
895 			 */
896 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
897 			jumboplen = (u_int32_t)htonl(jumboplen);
898 
899 #if 1
900 			/*
901 			 * if there are multiple jumbo payload options,
902 			 * *plenp will be non-zero and the packet will be
903 			 * rejected.
904 			 * the behavior may need some debate in ipngwg -
905 			 * multiple options does not make sense, however,
906 			 * there's no explicit mention in specification.
907 			 */
908 			if (*plenp != 0) {
909 				ip6stat.ip6s_badoptions++;
910 				icmp6_error(m, ICMP6_PARAM_PROB,
911 					    ICMP6_PARAMPROB_HEADER,
912 					    sizeof(struct ip6_hdr) +
913 					    sizeof(struct ip6_hbh) +
914 					    opt + 2 - opthead);
915 				return(-1);
916 			}
917 #endif
918 
919 			/*
920 			 * jumbo payload length must be larger than 65535.
921 			 */
922 			if (jumboplen <= IPV6_MAXPACKET) {
923 				ip6stat.ip6s_badoptions++;
924 				icmp6_error(m, ICMP6_PARAM_PROB,
925 					    ICMP6_PARAMPROB_HEADER,
926 					    sizeof(struct ip6_hdr) +
927 					    sizeof(struct ip6_hbh) +
928 					    opt + 2 - opthead);
929 				return(-1);
930 			}
931 			*plenp = jumboplen;
932 
933 			break;
934 		default:		/* unknown option */
935 			if (hbhlen < IP6OPT_MINLEN) {
936 				ip6stat.ip6s_toosmall++;
937 				goto bad;
938 			}
939 			if ((optlen = ip6_unknown_opt(opt, m,
940 						      sizeof(struct ip6_hdr) +
941 						      sizeof(struct ip6_hbh) +
942 						      opt - opthead)) == -1)
943 				return(-1);
944 			optlen += 2;
945 			break;
946 		}
947 	}
948 
949 	return(0);
950 
951   bad:
952 	m_freem(m);
953 	return(-1);
954 }
955 
956 /*
957  * Unknown option processing.
958  * The third argument `off' is the offset from the IPv6 header to the option,
959  * which is necessary if the IPv6 header the and option header and IPv6 header
960  * is not continuous in order to return an ICMPv6 error.
961  */
962 int
963 ip6_unknown_opt(optp, m, off)
964 	u_int8_t *optp;
965 	struct mbuf *m;
966 	int off;
967 {
968 	struct ip6_hdr *ip6;
969 
970 	switch (IP6OPT_TYPE(*optp)) {
971 	case IP6OPT_TYPE_SKIP: /* ignore the option */
972 		return((int)*(optp + 1));
973 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
974 		m_freem(m);
975 		return(-1);
976 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
977 		ip6stat.ip6s_badoptions++;
978 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
979 		return(-1);
980 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
981 		ip6stat.ip6s_badoptions++;
982 		ip6 = mtod(m, struct ip6_hdr *);
983 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
984 		    (m->m_flags & (M_BCAST|M_MCAST)))
985 			m_freem(m);
986 		else
987 			icmp6_error(m, ICMP6_PARAM_PROB,
988 				    ICMP6_PARAMPROB_OPTION, off);
989 		return(-1);
990 	}
991 
992 	m_freem(m);		/* XXX: NOTREACHED */
993 	return(-1);
994 }
995 
996 /*
997  * Create the "control" list for this pcb.
998  *
999  * The routine will be called from upper layer handlers like tcp6_input().
1000  * Thus the routine assumes that the caller (tcp6_input) have already
1001  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1002  * very first mbuf on the mbuf chain.
1003  * We may want to add some infinite loop prevention or sanity checks for safety.
1004  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1005  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1006  */
1007 void
1008 ip6_savecontrol(in6p, mp, ip6, m)
1009 	struct in6pcb *in6p;
1010 	struct mbuf **mp;
1011 	struct ip6_hdr *ip6;
1012 	struct mbuf *m;
1013 {
1014 	struct proc *p = curproc;	/* XXX */
1015 	int privileged;
1016 
1017 	privileged = 0;
1018 	if (p && !suser(p->p_ucred, &p->p_acflag))
1019 		privileged++;
1020 
1021 #ifdef SO_TIMESTAMP
1022 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1023 		struct timeval tv;
1024 
1025 		microtime(&tv);
1026 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1027 			SCM_TIMESTAMP, SOL_SOCKET);
1028 		if (*mp)
1029 			mp = &(*mp)->m_next;
1030 	}
1031 #endif
1032 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
1033 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
1034 			sizeof(struct in6_addr), IPV6_RECVDSTADDR,
1035 			IPPROTO_IPV6);
1036 		if (*mp)
1037 			mp = &(*mp)->m_next;
1038 	}
1039 
1040 #ifdef noyet
1041 	/* options were tossed above */
1042 	if (in6p->in6p_flags & IN6P_RECVOPTS)
1043 		/* broken */
1044 	/* ip6_srcroute doesn't do what we want here, need to fix */
1045 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
1046 		/* broken */
1047 #endif
1048 
1049 	/* RFC 2292 sec. 5 */
1050 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1051 		struct in6_pktinfo pi6;
1052 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1053 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1054 			pi6.ipi6_addr.s6_addr16[1] = 0;
1055 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1056 					? m->m_pkthdr.rcvif->if_index
1057 					: 0;
1058 		*mp = sbcreatecontrol((caddr_t) &pi6,
1059 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1060 			IPPROTO_IPV6);
1061 		if (*mp)
1062 			mp = &(*mp)->m_next;
1063 	}
1064 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1065 		int hlim = ip6->ip6_hlim & 0xff;
1066 		*mp = sbcreatecontrol((caddr_t) &hlim,
1067 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1068 		if (*mp)
1069 			mp = &(*mp)->m_next;
1070 	}
1071 	/* IN6P_NEXTHOP - for outgoing packet only */
1072 
1073 	/*
1074 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1075 	 * for the option, but it might be too strict, since there might
1076 	 * be some hop-by-hop options which can be returned to normal user.
1077 	 * See RFC 2292 section 6.
1078 	 */
1079 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1080 		/*
1081 		 * Check if a hop-by-hop options header is contatined in the
1082 		 * received packet, and if so, store the options as ancillary
1083 		 * data. Note that a hop-by-hop options header must be
1084 		 * just after the IPv6 header, which fact is assured through
1085 		 * the IPv6 input processing.
1086 		 */
1087 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1088 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1089 			struct ip6_hbh *hbh;
1090 			int hbhlen;
1091 
1092 #ifndef PULLDOWN_TEST
1093 			hbh = (struct ip6_hbh *)(ip6 + 1);
1094 			hbhlen = (hbh->ip6h_len + 1) << 3;
1095 #else
1096 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1097 				sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1098 			if (hbh == NULL) {
1099 				ip6stat.ip6s_tooshort++;
1100 				return;
1101 			}
1102 			hbhlen = (hbh->ip6h_len + 1) << 3;
1103 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1104 				sizeof(struct ip6_hdr), hbhlen);
1105 			if (hbh == NULL) {
1106 				ip6stat.ip6s_tooshort++;
1107 				return;
1108 			}
1109 #endif
1110 
1111 			/*
1112 			 * XXX: We copy whole the header even if a jumbo
1113 			 * payload option is included, which option is to
1114 			 * be removed before returning in the RFC 2292.
1115 			 * But it's too painful operation...
1116 			 */
1117 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1118 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1119 			if (*mp)
1120 				mp = &(*mp)->m_next;
1121 		}
1122 	}
1123 
1124 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1125 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1126 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1127 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);;
1128 
1129 		/*
1130 		 * Search for destination options headers or routing
1131 		 * header(s) through the header chain, and stores each
1132 		 * header as ancillary data.
1133 		 * Note that the order of the headers remains in
1134 		 * the chain of ancillary data.
1135 		 */
1136 		while (1) {	/* is explicit loop prevention necessary? */
1137 			struct ip6_ext *ip6e;
1138 			int elen;
1139 
1140 #ifndef PULLDOWN_TEST
1141 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1142 			if (nxt == IPPROTO_AH)
1143 				elen = (ip6e->ip6e_len + 2) << 2;
1144 			else
1145 				elen = (ip6e->ip6e_len + 1) << 3;
1146 #else
1147 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off,
1148 				sizeof(struct ip6_ext));
1149 			if (ip6e == NULL) {
1150 				ip6stat.ip6s_tooshort++;
1151 				return;
1152 			}
1153 			if (nxt == IPPROTO_AH)
1154 				elen = (ip6e->ip6e_len + 2) << 2;
1155 			else
1156 				elen = (ip6e->ip6e_len + 1) << 3;
1157 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off, elen);
1158 			if (ip6e == NULL) {
1159 				ip6stat.ip6s_tooshort++;
1160 				return;
1161 			}
1162 #endif
1163 
1164 			switch (nxt) {
1165 		        case IPPROTO_DSTOPTS:
1166 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1167 					break;
1168 
1169 				/*
1170 				 * We also require super-user privilege for
1171 				 * the option.
1172 				 * See the comments on IN6_HOPOPTS.
1173 				 */
1174 				if (!privileged)
1175 					break;
1176 
1177 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1178 						      IPV6_DSTOPTS,
1179 						      IPPROTO_IPV6);
1180 				if (*mp)
1181 					mp = &(*mp)->m_next;
1182 				break;
1183 
1184 			case IPPROTO_ROUTING:
1185 				if (!in6p->in6p_flags & IN6P_RTHDR)
1186 					break;
1187 
1188 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1189 						      IPV6_RTHDR,
1190 						      IPPROTO_IPV6);
1191 				if (*mp)
1192 					mp = &(*mp)->m_next;
1193 				break;
1194 
1195 			case IPPROTO_UDP:
1196 			case IPPROTO_TCP:
1197 			case IPPROTO_ICMPV6:
1198 			default:
1199 				/*
1200 				 * stop search if we encounter an upper
1201 				 * layer protocol headers.
1202 				 */
1203 				goto loopend;
1204 
1205 			case IPPROTO_HOPOPTS:
1206 			case IPPROTO_AH: /* is it possible? */
1207 				break;
1208 			}
1209 
1210 			/* proceed with the next header. */
1211 			off += elen;
1212 			nxt = ip6e->ip6e_nxt;
1213 		}
1214 	  loopend:
1215 		;
1216 	}
1217 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1218 		/* to be done */
1219 	}
1220 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1221 		/* to be done */
1222 	}
1223 	/* IN6P_RTHDR - to be done */
1224 
1225 }
1226 
1227 /*
1228  * Get pointer to the previous header followed by the header
1229  * currently processed.
1230  * XXX: This function supposes that
1231  *	M includes all headers,
1232  *	the next header field and the header length field of each header
1233  *	are valid, and
1234  *	the sum of each header length equals to OFF.
1235  * Because of these assumptions, this function must be called very
1236  * carefully. Moreover, it will not be used in the near future when
1237  * we develop `neater' mechanism to process extension headers.
1238  */
1239 char *
1240 ip6_get_prevhdr(m, off)
1241 	struct mbuf *m;
1242 	int off;
1243 {
1244 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1245 
1246 	if (off == sizeof(struct ip6_hdr))
1247 		return(&ip6->ip6_nxt);
1248 	else {
1249 		int len, nxt;
1250 		struct ip6_ext *ip6e = NULL;
1251 
1252 		nxt = ip6->ip6_nxt;
1253 		len = sizeof(struct ip6_hdr);
1254 		while (len < off) {
1255 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1256 
1257 			switch (nxt) {
1258 			case IPPROTO_FRAGMENT:
1259 				len += sizeof(struct ip6_frag);
1260 				break;
1261 			case IPPROTO_AH:
1262 				len += (ip6e->ip6e_len + 2) << 2;
1263 				break;
1264 			default:
1265 				len += (ip6e->ip6e_len + 1) << 3;
1266 				break;
1267 			}
1268 			nxt = ip6e->ip6e_nxt;
1269 		}
1270 		if (ip6e)
1271 			return(&ip6e->ip6e_nxt);
1272 		else
1273 			return NULL;
1274 	}
1275 }
1276 
1277 /*
1278  * get next header offset.  m will be retained.
1279  */
1280 int
1281 ip6_nexthdr(m, off, proto, nxtp)
1282 	struct mbuf *m;
1283 	int off;
1284 	int proto;
1285 	int *nxtp;
1286 {
1287 	struct ip6_hdr ip6;
1288 	struct ip6_ext ip6e;
1289 	struct ip6_frag fh;
1290 
1291 	/* just in case */
1292 	if (m == NULL)
1293 		panic("ip6_nexthdr: m == NULL");
1294 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1295 		return -1;
1296 
1297 	switch (proto) {
1298 	case IPPROTO_IPV6:
1299 		if (m->m_pkthdr.len < off + sizeof(ip6))
1300 			return -1;
1301 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1302 		if (nxtp)
1303 			*nxtp = ip6.ip6_nxt;
1304 		off += sizeof(ip6);
1305 		return off;
1306 
1307 	case IPPROTO_FRAGMENT:
1308 		/*
1309 		 * terminate parsing if it is not the first fragment,
1310 		 * it does not make sense to parse through it.
1311 		 */
1312 		if (m->m_pkthdr.len < off + sizeof(fh))
1313 			return -1;
1314 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1315 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1316 			return -1;
1317 		if (nxtp)
1318 			*nxtp = fh.ip6f_nxt;
1319 		off += sizeof(struct ip6_frag);
1320 		return off;
1321 
1322 	case IPPROTO_AH:
1323 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1324 			return -1;
1325 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1326 		if (nxtp)
1327 			*nxtp = ip6e.ip6e_nxt;
1328 		off += (ip6e.ip6e_len + 2) << 2;
1329 		if (m->m_pkthdr.len < off)
1330 			return -1;
1331 		return off;
1332 
1333 	case IPPROTO_HOPOPTS:
1334 	case IPPROTO_ROUTING:
1335 	case IPPROTO_DSTOPTS:
1336 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1337 			return -1;
1338 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1339 		if (nxtp)
1340 			*nxtp = ip6e.ip6e_nxt;
1341 		off += (ip6e.ip6e_len + 1) << 3;
1342 		if (m->m_pkthdr.len < off)
1343 			return -1;
1344 		return off;
1345 
1346 	case IPPROTO_NONE:
1347 	case IPPROTO_ESP:
1348 	case IPPROTO_IPCOMP:
1349 		/* give up */
1350 		return -1;
1351 
1352 	default:
1353 		return -1;
1354 	}
1355 
1356 	return -1;
1357 }
1358 
1359 /*
1360  * get offset for the last header in the chain.  m will be kept untainted.
1361  */
1362 int
1363 ip6_lasthdr(m, off, proto, nxtp)
1364 	struct mbuf *m;
1365 	int off;
1366 	int proto;
1367 	int *nxtp;
1368 {
1369 	int newoff;
1370 	int nxt;
1371 
1372 	if (!nxtp) {
1373 		nxt = -1;
1374 		nxtp = &nxt;
1375 	}
1376 	while (1) {
1377 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1378 		if (newoff < 0)
1379 			return off;
1380 		else if (newoff < off)
1381 			return -1;	/* invalid */
1382 		else if (newoff == off)
1383 			return newoff;
1384 
1385 		off = newoff;
1386 		proto = *nxtp;
1387 	}
1388 }
1389 
1390 /*
1391  * System control for IP6
1392  */
1393 
1394 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1395 	0,		0,		0,		0,
1396 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1397 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1398 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1399 	0,		0,		0,		0,
1400 	ENOPROTOOPT
1401 };
1402 
1403 int
1404 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1405 	int *name;
1406 	u_int namelen;
1407 	void *oldp;
1408 	size_t *oldlenp;
1409 	void *newp;
1410 	size_t newlen;
1411 {
1412 	int old, error;
1413 
1414 	/* All sysctl names at this level are terminal. */
1415 	if (namelen != 1)
1416 		return ENOTDIR;
1417 
1418 	switch (name[0]) {
1419 
1420 	case IPV6CTL_FORWARDING:
1421 		return sysctl_int(oldp, oldlenp, newp, newlen,
1422 				  &ip6_forwarding);
1423 	case IPV6CTL_SENDREDIRECTS:
1424 		return sysctl_int(oldp, oldlenp, newp, newlen,
1425 				&ip6_sendredirects);
1426 	case IPV6CTL_DEFHLIM:
1427 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1428 	case IPV6CTL_MAXFRAGPACKETS:
1429 		return sysctl_int(oldp, oldlenp, newp, newlen,
1430 				&ip6_maxfragpackets);
1431 	case IPV6CTL_ACCEPT_RTADV:
1432 		return sysctl_int(oldp, oldlenp, newp, newlen,
1433 				&ip6_accept_rtadv);
1434 	case IPV6CTL_KEEPFAITH:
1435 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1436 	case IPV6CTL_LOG_INTERVAL:
1437 		return sysctl_int(oldp, oldlenp, newp, newlen,
1438 				&ip6_log_interval);
1439 	case IPV6CTL_HDRNESTLIMIT:
1440 		return sysctl_int(oldp, oldlenp, newp, newlen,
1441 				&ip6_hdrnestlimit);
1442 	case IPV6CTL_DAD_COUNT:
1443 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1444 	case IPV6CTL_AUTO_FLOWLABEL:
1445 		return sysctl_int(oldp, oldlenp, newp, newlen,
1446 				&ip6_auto_flowlabel);
1447 	case IPV6CTL_DEFMCASTHLIM:
1448 		return sysctl_int(oldp, oldlenp, newp, newlen,
1449 				&ip6_defmcasthlim);
1450 #if NGIF > 0
1451 	case IPV6CTL_GIF_HLIM:
1452 		return sysctl_int(oldp, oldlenp, newp, newlen,
1453 				&ip6_gif_hlim);
1454 #endif
1455 	case IPV6CTL_KAME_VERSION:
1456 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1457 	case IPV6CTL_USE_DEPRECATED:
1458 		return sysctl_int(oldp, oldlenp, newp, newlen,
1459 				&ip6_use_deprecated);
1460 	case IPV6CTL_RR_PRUNE:
1461 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1462 	case IPV6CTL_V6ONLY:
1463 #ifdef INET6_BINDV6ONLY
1464 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1465 #else
1466 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_v6only);
1467 #endif
1468 	case IPV6CTL_ANONPORTMIN:
1469 		old = ip6_anonportmin;
1470 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1471 		    &ip6_anonportmin);
1472 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmin < 0 ||
1473 		    ip6_anonportmin > 65535
1474 #ifndef IPNOPRIVPORTS
1475 		    || ip6_anonportmin < IPV6PORT_RESERVED
1476 #endif
1477 		    ) {
1478 			ip6_anonportmin = old;
1479 			return (EINVAL);
1480 		}
1481 		return (error);
1482 	case IPV6CTL_ANONPORTMAX:
1483 		old = ip6_anonportmax;
1484 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1485 		    &ip6_anonportmax);
1486 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmax < 0 ||
1487 		    ip6_anonportmax > 65535
1488 #ifndef IPNOPRIVPORTS
1489 		    || ip6_anonportmax < IPV6PORT_RESERVED
1490 #endif
1491 		    ) {
1492 			ip6_anonportmax = old;
1493 			return (EINVAL);
1494 		}
1495 		return (error);
1496 #ifndef IPNOPRIVPORTS
1497 	case IPV6CTL_LOWPORTMIN:
1498 		old = ip6_lowportmin;
1499 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1500 		    &ip6_lowportmin);
1501 		if (ip6_lowportmin >= ip6_lowportmax ||
1502 		    ip6_lowportmin > IPV6PORT_RESERVEDMAX ||
1503 		    ip6_lowportmin < IPV6PORT_RESERVEDMIN) {
1504 			ip6_lowportmin = old;
1505 			return (EINVAL);
1506 		}
1507 		return (error);
1508 	case IPV6CTL_LOWPORTMAX:
1509 		old = ip6_lowportmax;
1510 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1511 		    &ip6_lowportmax);
1512 		if (ip6_lowportmin >= ip6_lowportmax ||
1513 		    ip6_lowportmax > IPV6PORT_RESERVEDMAX ||
1514 		    ip6_lowportmax < IPV6PORT_RESERVEDMIN) {
1515 			ip6_lowportmax = old;
1516 			return (EINVAL);
1517 		}
1518 		return (error);
1519 #endif
1520 	default:
1521 		return EOPNOTSUPP;
1522 	}
1523 	/* NOTREACHED */
1524 }
1525