xref: /netbsd-src/sys/netinet6/ip6_input.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: ip6_input.c,v 1.63 2003/05/14 14:41:33 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 1.63 2003/05/14 14:41:33 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 
113 #ifdef IPSEC
114 #include <netinet6/ipsec.h>
115 #endif
116 
117 #include <netinet6/ip6protosw.h>
118 
119 /* we need it for NLOOP. */
120 #include "loop.h"
121 #include "faith.h"
122 #include "gif.h"
123 #include "bpfilter.h"
124 
125 #if NGIF > 0
126 #include <netinet6/in6_gif.h>
127 #endif
128 
129 #include <net/net_osdep.h>
130 
131 extern struct domain inet6domain;
132 
133 u_char ip6_protox[IPPROTO_MAX];
134 static int ip6qmaxlen = IFQ_MAXLEN;
135 struct in6_ifaddr *in6_ifaddr;
136 struct ifqueue ip6intrq;
137 
138 extern struct ifnet loif[NLOOP];
139 int ip6_forward_srcrt;			/* XXX */
140 int ip6_sourcecheck;			/* XXX */
141 int ip6_sourcecheck_interval;		/* XXX */
142 
143 #ifdef PFIL_HOOKS
144 struct pfil_head inet6_pfil_hook;
145 #endif
146 
147 struct ip6stat ip6stat;
148 
149 static void ip6_init2 __P((void *));
150 
151 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
152 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, 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 
164 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
165 	if (pr == 0)
166 		panic("ip6_init");
167 	for (i = 0; i < IPPROTO_MAX; i++)
168 		ip6_protox[i] = pr - inet6sw;
169 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
170 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
171 		if (pr->pr_domain->dom_family == PF_INET6 &&
172 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
173 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
174 	ip6intrq.ifq_maxlen = ip6qmaxlen;
175 	nd6_init();
176 	frag6_init();
177 	ip6_flow_seq = arc4random();
178 
179 	ip6_init2((void *)0);
180 
181 #ifdef PFIL_HOOKS
182 	/* Register our Packet Filter hook. */
183 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
184 	inet6_pfil_hook.ph_af   = AF_INET6;
185 	i = pfil_head_register(&inet6_pfil_hook);
186 	if (i != 0)
187 		printf("ip6_init: WARNING: unable to register pfil hook, "
188 		    "error %d\n", i);
189 #endif /* PFIL_HOOKS */
190 }
191 
192 static void
193 ip6_init2(dummy)
194 	void *dummy;
195 {
196 
197 	/* nd6_timer_init */
198 	callout_init(&nd6_timer_ch);
199 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
200 }
201 
202 /*
203  * IP6 input interrupt handling. Just pass the packet to ip6_input.
204  */
205 void
206 ip6intr()
207 {
208 	int s;
209 	struct mbuf *m;
210 
211 	for (;;) {
212 		s = splnet();
213 		IF_DEQUEUE(&ip6intrq, m);
214 		splx(s);
215 		if (m == 0)
216 			return;
217 		ip6_input(m);
218 	}
219 }
220 
221 extern struct	route_in6 ip6_forward_rt;
222 
223 void
224 ip6_input(m)
225 	struct mbuf *m;
226 {
227 	struct ip6_hdr *ip6;
228 	int off = sizeof(struct ip6_hdr), nest;
229 	u_int32_t plen;
230 	u_int32_t rtalert = ~0;
231 	int nxt, ours = 0;
232 	struct ifnet *deliverifp = NULL;
233 
234 #ifdef IPSEC
235 	/*
236 	 * should the inner packet be considered authentic?
237 	 * see comment in ah4_input().
238 	 */
239 	m->m_flags &= ~M_AUTHIPHDR;
240 	m->m_flags &= ~M_AUTHIPDGM;
241 #endif
242 
243 	/*
244 	 * mbuf statistics
245 	 */
246 	if (m->m_flags & M_EXT) {
247 		if (m->m_next)
248 			ip6stat.ip6s_mext2m++;
249 		else
250 			ip6stat.ip6s_mext1++;
251 	} else {
252 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
253 		if (m->m_next) {
254 			if (m->m_flags & M_LOOP) {
255 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
256 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
257 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
258 			else
259 				ip6stat.ip6s_m2m[0]++;
260 		} else
261 			ip6stat.ip6s_m1++;
262 #undef M2MMAX
263 	}
264 
265 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
266 	ip6stat.ip6s_total++;
267 
268 	/*
269 	 * If the IPv6 header is not aligned, slurp it up into a new
270 	 * mbuf with space for link headers, in the event we forward
271 	 * it.  OTherwise, if it is aligned, make sure the entire base
272 	 * IPv6 header is in the first mbuf of the chain.
273 	 */
274 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
275 		struct ifnet *inifp = m->m_pkthdr.rcvif;
276 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
277 				  (max_linkhdr + 3) & ~3)) == NULL) {
278 			/* XXXJRT new stat, please */
279 			ip6stat.ip6s_toosmall++;
280 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
281 			return;
282 		}
283 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
284 		struct ifnet *inifp = m->m_pkthdr.rcvif;
285 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
286 			ip6stat.ip6s_toosmall++;
287 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
288 			return;
289 		}
290 	}
291 
292 	ip6 = mtod(m, struct ip6_hdr *);
293 
294 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
295 		ip6stat.ip6s_badvers++;
296 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
297 		goto bad;
298 	}
299 
300 #ifdef PFIL_HOOKS
301 	/*
302 	 * Run through list of hooks for input packets.  If there are any
303 	 * filters which require that additional packets in the flow are
304 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
305 	 * Note that filters must _never_ set this flag, as another filter
306 	 * in the list may have previously cleared it.
307 	 */
308 	/*
309 	 * let ipfilter look at packet on the wire,
310 	 * not the decapsulated packet.
311 	 */
312 #ifdef IPSEC
313 	if (!ipsec_getnhist(m))
314 #else
315 	if (1)
316 #endif
317 	{
318 		if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif,
319 				   PFIL_IN) != 0)
320 			return;
321 		if (m == NULL)
322 			return;
323 		ip6 = mtod(m, struct ip6_hdr *);
324 	}
325 #endif /* PFIL_HOOKS */
326 
327 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
328 
329 #ifdef ALTQ
330 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
331 		/* packet is dropped by traffic conditioner */
332 		return;
333 	}
334 #endif
335 
336 	/*
337 	 * Check against address spoofing/corruption.
338 	 */
339 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
340 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
341 		/*
342 		 * XXX: "badscope" is not very suitable for a multicast source.
343 		 */
344 		ip6stat.ip6s_badscope++;
345 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
346 		goto bad;
347 	}
348 	/*
349 	 * The following check is not documented in specs.  A malicious
350 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
351 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
352 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
353 	 *
354 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
355 	 * support IPv4-less kernel compilation, we cannot support SIIT
356 	 * environment at all.  So, it makes more sense for us to reject any
357 	 * malicious packets for non-SIIT environment, than try to do a
358 	 * partial support for SIIT environment.
359 	 */
360 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
361 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
364 		goto bad;
365 	}
366 #if 0
367 	/*
368 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
369 	 *
370 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
371 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
372 	 * is revised to forbid relaying case.
373 	 */
374 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
375 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
376 		ip6stat.ip6s_badscope++;
377 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
378 		goto bad;
379 	}
380 #endif
381 
382 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
383 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
384 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
385 			ours = 1;
386 			deliverifp = m->m_pkthdr.rcvif;
387 			goto hbhcheck;
388 		} else {
389 			ip6stat.ip6s_badscope++;
390 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
391 			goto bad;
392 		}
393 	}
394 
395 	/* drop packets if interface ID portion is already filled */
396 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
397 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
398 		    ip6->ip6_src.s6_addr16[1]) {
399 			ip6stat.ip6s_badscope++;
400 			goto bad;
401 		}
402 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
403 		    ip6->ip6_dst.s6_addr16[1]) {
404 			ip6stat.ip6s_badscope++;
405 			goto bad;
406 		}
407 	}
408 
409 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
410 		ip6->ip6_src.s6_addr16[1]
411 			= htons(m->m_pkthdr.rcvif->if_index);
412 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
413 		ip6->ip6_dst.s6_addr16[1]
414 			= htons(m->m_pkthdr.rcvif->if_index);
415 
416 	/*
417 	 * We use rt->rt_ifp to determine if the address is ours or not.
418 	 * If rt_ifp is lo0, the address is ours.
419 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
420 	 * so any address under fe80::%lo0/64 will be mistakenly considered
421 	 * local.  The special case is supplied to handle the case properly
422 	 * by actually looking at interface addresses
423 	 * (using in6ifa_ifpwithaddr).
424 	 */
425 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
426 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
427 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
428 			icmp6_error(m, ICMP6_DST_UNREACH,
429 			    ICMP6_DST_UNREACH_ADDR, 0);
430 			/* m is already freed */
431 			return;
432 		}
433 
434 		ours = 1;
435 		deliverifp = m->m_pkthdr.rcvif;
436 		goto hbhcheck;
437 	}
438 
439 	/*
440 	 * Multicast check
441 	 */
442 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
443 	  	struct	in6_multi *in6m = 0;
444 
445 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
446 		/*
447 		 * See if we belong to the destination multicast group on the
448 		 * arrival interface.
449 		 */
450 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
451 		if (in6m)
452 			ours = 1;
453 		else if (!ip6_mrouter) {
454 			ip6stat.ip6s_notmember++;
455 			ip6stat.ip6s_cantforward++;
456 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
457 			goto bad;
458 		}
459 		deliverifp = m->m_pkthdr.rcvif;
460 		goto hbhcheck;
461 	}
462 
463 	/*
464 	 *  Unicast check
465 	 */
466 	if (ip6_forward_rt.ro_rt != NULL &&
467 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
468 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
469 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
470 		ip6stat.ip6s_forward_cachehit++;
471 	else {
472 		struct sockaddr_in6 *dst6;
473 
474 		if (ip6_forward_rt.ro_rt) {
475 			/* route is down or destination is different */
476 			ip6stat.ip6s_forward_cachemiss++;
477 			RTFREE(ip6_forward_rt.ro_rt);
478 			ip6_forward_rt.ro_rt = 0;
479 		}
480 
481 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
482 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
483 		dst6->sin6_len = sizeof(struct sockaddr_in6);
484 		dst6->sin6_family = AF_INET6;
485 		dst6->sin6_addr = ip6->ip6_dst;
486 
487 		rtalloc((struct route *)&ip6_forward_rt);
488 	}
489 
490 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
491 
492 	/*
493 	 * Accept the packet if the forwarding interface to the destination
494 	 * according to the routing table is the loopback interface,
495 	 * unless the associated route has a gateway.
496 	 * Note that this approach causes to accept a packet if there is a
497 	 * route to the loopback interface for the destination of the packet.
498 	 * But we think it's even useful in some situations, e.g. when using
499 	 * a special daemon which wants to intercept the packet.
500 	 */
501 	if (ip6_forward_rt.ro_rt &&
502 	    (ip6_forward_rt.ro_rt->rt_flags &
503 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
504 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
505 #if 0
506 	    /*
507 	     * The check below is redundant since the comparison of
508 	     * the destination and the key of the rtentry has
509 	     * already done through looking up the routing table.
510 	     */
511 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
512 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
513 #endif
514 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
515 		struct in6_ifaddr *ia6 =
516 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
517 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
518 			m->m_flags |= M_ANYCAST6;
519 		/*
520 		 * packets to a tentative, duplicated, or somehow invalid
521 		 * address must not be accepted.
522 		 */
523 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
524 			/* this address is ready */
525 			ours = 1;
526 			deliverifp = ia6->ia_ifp;	/* correct? */
527 			goto hbhcheck;
528 		} else {
529 			/* address is not ready, so discard the packet. */
530 			nd6log((LOG_INFO,
531 			    "ip6_input: packet to an unready address %s->%s\n",
532 			    ip6_sprintf(&ip6->ip6_src),
533 			    ip6_sprintf(&ip6->ip6_dst)));
534 
535 			goto bad;
536 		}
537 	}
538 
539 	/*
540 	 * FAITH (Firewall Aided Internet Translator)
541 	 */
542 #if defined(NFAITH) && 0 < NFAITH
543 	if (ip6_keepfaith) {
544 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
545 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
546 			/* XXX do we need more sanity checks? */
547 			ours = 1;
548 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
549 			goto hbhcheck;
550 		}
551 	}
552 #endif
553 
554 #if 0
555     {
556 	/*
557 	 * Last resort: check in6_ifaddr for incoming interface.
558 	 * The code is here until I update the "goto ours hack" code above
559 	 * working right.
560 	 */
561 	struct ifaddr *ifa;
562 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
563 	     ifa;
564 	     ifa = ifa->ifa_list.tqe_next) {
565 		if (ifa->ifa_addr == NULL)
566 			continue;	/* just for safety */
567 		if (ifa->ifa_addr->sa_family != AF_INET6)
568 			continue;
569 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
570 			ours = 1;
571 			deliverifp = ifa->ifa_ifp;
572 			goto hbhcheck;
573 		}
574 	}
575     }
576 #endif
577 
578 	/*
579 	 * Now there is no reason to process the packet if it's not our own
580 	 * and we're not a router.
581 	 */
582 	if (!ip6_forwarding) {
583 		ip6stat.ip6s_cantforward++;
584 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
585 		goto bad;
586 	}
587 
588   hbhcheck:
589 	/*
590 	 * Process Hop-by-Hop options header if it's contained.
591 	 * m may be modified in ip6_hopopts_input().
592 	 * If a JumboPayload option is included, plen will also be modified.
593 	 */
594 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
595 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
596 		struct ip6_hbh *hbh;
597 
598 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
599 #if 0	/*touches NULL pointer*/
600 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
601 #endif
602 			return;	/* m have already been freed */
603 		}
604 
605 		/* adjust pointer */
606 		ip6 = mtod(m, struct ip6_hdr *);
607 
608 		/*
609 		 * if the payload length field is 0 and the next header field
610 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
611 		 * option MUST be included.
612 		 */
613 		if (ip6->ip6_plen == 0 && plen == 0) {
614 			/*
615 			 * Note that if a valid jumbo payload option is
616 			 * contained, ip6_hoptops_input() must set a valid
617 			 * (non-zero) payload length to the variable plen.
618 			 */
619 			ip6stat.ip6s_badoptions++;
620 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
621 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
622 			icmp6_error(m, ICMP6_PARAM_PROB,
623 				    ICMP6_PARAMPROB_HEADER,
624 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
625 			return;
626 		}
627 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
628 			sizeof(struct ip6_hbh));
629 		if (hbh == NULL) {
630 			ip6stat.ip6s_tooshort++;
631 			return;
632 		}
633 		KASSERT(IP6_HDR_ALIGNED_P(hbh));
634 		nxt = hbh->ip6h_nxt;
635 
636 		/*
637 		 * accept the packet if a router alert option is included
638 		 * and we act as an IPv6 router.
639 		 */
640 		if (rtalert != ~0 && ip6_forwarding)
641 			ours = 1;
642 	} else
643 		nxt = ip6->ip6_nxt;
644 
645 	/*
646 	 * Check that the amount of data in the buffers
647 	 * is as at least much as the IPv6 header would have us expect.
648 	 * Trim mbufs if longer than we expect.
649 	 * Drop packet if shorter than we expect.
650 	 */
651 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
652 		ip6stat.ip6s_tooshort++;
653 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
654 		goto bad;
655 	}
656 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
657 		if (m->m_len == m->m_pkthdr.len) {
658 			m->m_len = sizeof(struct ip6_hdr) + plen;
659 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
660 		} else
661 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
662 	}
663 
664 	/*
665 	 * Forward if desirable.
666 	 */
667 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
668 		/*
669 		 * If we are acting as a multicast router, all
670 		 * incoming multicast packets are passed to the
671 		 * kernel-level multicast forwarding function.
672 		 * The packet is returned (relatively) intact; if
673 		 * ip6_mforward() returns a non-zero value, the packet
674 		 * must be discarded, else it may be accepted below.
675 		 */
676 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
677 			ip6stat.ip6s_cantforward++;
678 			m_freem(m);
679 			return;
680 		}
681 		if (!ours) {
682 			m_freem(m);
683 			return;
684 		}
685 	} else if (!ours) {
686 		ip6_forward(m, 0);
687 		return;
688 	}
689 
690 	ip6 = mtod(m, struct ip6_hdr *);
691 
692 	/*
693 	 * Malicious party may be able to use IPv4 mapped addr to confuse
694 	 * tcp/udp stack and bypass security checks (act as if it was from
695 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
696 	 *
697 	 * For SIIT end node behavior, you may want to disable the check.
698 	 * However, you will  become vulnerable to attacks using IPv4 mapped
699 	 * source.
700 	 */
701 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
702 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
703 		ip6stat.ip6s_badscope++;
704 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
705 		goto bad;
706 	}
707 
708 	/*
709 	 * Tell launch routine the next header
710 	 */
711 #ifdef IFA_STATS
712 	if (deliverifp != NULL) {
713 		struct in6_ifaddr *ia6;
714 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
715 		if (ia6)
716 			ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
717 	}
718 #endif
719 	ip6stat.ip6s_delivered++;
720 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
721 	nest = 0;
722 
723 	while (nxt != IPPROTO_DONE) {
724 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
725 			ip6stat.ip6s_toomanyhdr++;
726 			goto bad;
727 		}
728 
729 		/*
730 		 * protection against faulty packet - there should be
731 		 * more sanity checks in header chain processing.
732 		 */
733 		if (m->m_pkthdr.len < off) {
734 			ip6stat.ip6s_tooshort++;
735 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
736 			goto bad;
737 		}
738 
739 #ifdef IPSEC
740 		/*
741 		 * enforce IPsec policy checking if we are seeing last header.
742 		 * note that we do not visit this with protocols with pcb layer
743 		 * code - like udp/tcp/raw ip.
744 		 */
745 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
746 		    ipsec6_in_reject(m, NULL)) {
747 			ipsec6stat.in_polvio++;
748 			goto bad;
749 		}
750 #endif
751 
752 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
753 	}
754 	return;
755  bad:
756 	m_freem(m);
757 }
758 
759 /*
760  * Hop-by-Hop options header processing. If a valid jumbo payload option is
761  * included, the real payload length will be stored in plenp.
762  */
763 static int
764 ip6_hopopts_input(plenp, rtalertp, mp, offp)
765 	u_int32_t *plenp;
766 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
767 	struct mbuf **mp;
768 	int *offp;
769 {
770 	struct mbuf *m = *mp;
771 	int off = *offp, hbhlen;
772 	struct ip6_hbh *hbh;
773 	u_int8_t *opt;
774 
775 	/* validation of the length of the header */
776 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
777 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
778 	if (hbh == NULL) {
779 		ip6stat.ip6s_tooshort++;
780 		return -1;
781 	}
782 	hbhlen = (hbh->ip6h_len + 1) << 3;
783 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
784 		hbhlen);
785 	if (hbh == NULL) {
786 		ip6stat.ip6s_tooshort++;
787 		return -1;
788 	}
789 	KASSERT(IP6_HDR_ALIGNED_P(hbh));
790 	off += hbhlen;
791 	hbhlen -= sizeof(struct ip6_hbh);
792 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
793 
794 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
795 				hbhlen, rtalertp, plenp) < 0)
796 		return (-1);
797 
798 	*offp = off;
799 	*mp = m;
800 	return (0);
801 }
802 
803 /*
804  * Search header for all Hop-by-hop options and process each option.
805  * This function is separate from ip6_hopopts_input() in order to
806  * handle a case where the sending node itself process its hop-by-hop
807  * options header. In such a case, the function is called from ip6_output().
808  *
809  * The function assumes that hbh header is located right after the IPv6 header
810  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
811  * opthead + hbhlen is located in continuous memory region.
812  */
813 int
814 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
815 	struct mbuf *m;
816 	u_int8_t *opthead;
817 	int hbhlen;
818 	u_int32_t *rtalertp;
819 	u_int32_t *plenp;
820 {
821 	struct ip6_hdr *ip6;
822 	int optlen = 0;
823 	u_int8_t *opt = opthead;
824 	u_int16_t rtalert_val;
825 	u_int32_t jumboplen;
826 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
827 
828 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
829 		switch (*opt) {
830 		case IP6OPT_PAD1:
831 			optlen = 1;
832 			break;
833 		case IP6OPT_PADN:
834 			if (hbhlen < IP6OPT_MINLEN) {
835 				ip6stat.ip6s_toosmall++;
836 				goto bad;
837 			}
838 			optlen = *(opt + 1) + 2;
839 			break;
840 		case IP6OPT_RTALERT:
841 			/* XXX may need check for alignment */
842 			if (hbhlen < IP6OPT_RTALERT_LEN) {
843 				ip6stat.ip6s_toosmall++;
844 				goto bad;
845 			}
846 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
847 				/* XXX stat */
848 				icmp6_error(m, ICMP6_PARAM_PROB,
849 				    ICMP6_PARAMPROB_HEADER,
850 				    erroff + opt + 1 - opthead);
851 				return (-1);
852 			}
853 			optlen = IP6OPT_RTALERT_LEN;
854 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
855 			*rtalertp = ntohs(rtalert_val);
856 			break;
857 		case IP6OPT_JUMBO:
858 			/* XXX may need check for alignment */
859 			if (hbhlen < IP6OPT_JUMBO_LEN) {
860 				ip6stat.ip6s_toosmall++;
861 				goto bad;
862 			}
863 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
864 				/* XXX stat */
865 				icmp6_error(m, ICMP6_PARAM_PROB,
866 				    ICMP6_PARAMPROB_HEADER,
867 				    erroff + opt + 1 - opthead);
868 				return (-1);
869 			}
870 			optlen = IP6OPT_JUMBO_LEN;
871 
872 			/*
873 			 * IPv6 packets that have non 0 payload length
874 			 * must not contain a jumbo payload option.
875 			 */
876 			ip6 = mtod(m, struct ip6_hdr *);
877 			if (ip6->ip6_plen) {
878 				ip6stat.ip6s_badoptions++;
879 				icmp6_error(m, ICMP6_PARAM_PROB,
880 				    ICMP6_PARAMPROB_HEADER,
881 				    erroff + opt - opthead);
882 				return (-1);
883 			}
884 
885 			/*
886 			 * We may see jumbolen in unaligned location, so
887 			 * we'd need to perform bcopy().
888 			 */
889 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
890 			jumboplen = (u_int32_t)htonl(jumboplen);
891 
892 #if 1
893 			/*
894 			 * if there are multiple jumbo payload options,
895 			 * *plenp will be non-zero and the packet will be
896 			 * rejected.
897 			 * the behavior may need some debate in ipngwg -
898 			 * multiple options does not make sense, however,
899 			 * there's no explicit mention in specification.
900 			 */
901 			if (*plenp != 0) {
902 				ip6stat.ip6s_badoptions++;
903 				icmp6_error(m, ICMP6_PARAM_PROB,
904 				    ICMP6_PARAMPROB_HEADER,
905 				    erroff + opt + 2 - opthead);
906 				return (-1);
907 			}
908 #endif
909 
910 			/*
911 			 * jumbo payload length must be larger than 65535.
912 			 */
913 			if (jumboplen <= IPV6_MAXPACKET) {
914 				ip6stat.ip6s_badoptions++;
915 				icmp6_error(m, ICMP6_PARAM_PROB,
916 				    ICMP6_PARAMPROB_HEADER,
917 				    erroff + opt + 2 - opthead);
918 				return (-1);
919 			}
920 			*plenp = jumboplen;
921 
922 			break;
923 		default:		/* unknown option */
924 			if (hbhlen < IP6OPT_MINLEN) {
925 				ip6stat.ip6s_toosmall++;
926 				goto bad;
927 			}
928 			optlen = ip6_unknown_opt(opt, m,
929 			    erroff + opt - opthead);
930 			if (optlen == -1)
931 				return (-1);
932 			optlen += 2;
933 			break;
934 		}
935 	}
936 
937 	return (0);
938 
939   bad:
940 	m_freem(m);
941 	return (-1);
942 }
943 
944 /*
945  * Unknown option processing.
946  * The third argument `off' is the offset from the IPv6 header to the option,
947  * which is necessary if the IPv6 header the and option header and IPv6 header
948  * is not continuous in order to return an ICMPv6 error.
949  */
950 int
951 ip6_unknown_opt(optp, m, off)
952 	u_int8_t *optp;
953 	struct mbuf *m;
954 	int off;
955 {
956 	struct ip6_hdr *ip6;
957 
958 	switch (IP6OPT_TYPE(*optp)) {
959 	case IP6OPT_TYPE_SKIP: /* ignore the option */
960 		return ((int)*(optp + 1));
961 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
962 		m_freem(m);
963 		return (-1);
964 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
965 		ip6stat.ip6s_badoptions++;
966 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
967 		return (-1);
968 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
969 		ip6stat.ip6s_badoptions++;
970 		ip6 = mtod(m, struct ip6_hdr *);
971 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
972 		    (m->m_flags & (M_BCAST|M_MCAST)))
973 			m_freem(m);
974 		else
975 			icmp6_error(m, ICMP6_PARAM_PROB,
976 				    ICMP6_PARAMPROB_OPTION, off);
977 		return (-1);
978 	}
979 
980 	m_freem(m);		/* XXX: NOTREACHED */
981 	return (-1);
982 }
983 
984 /*
985  * Create the "control" list for this pcb.
986  *
987  * The routine will be called from upper layer handlers like tcp6_input().
988  * Thus the routine assumes that the caller (tcp6_input) have already
989  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
990  * very first mbuf on the mbuf chain.
991  * We may want to add some infinite loop prevention or sanity checks for safety.
992  * (This applies only when you are using KAME mbuf chain restriction, i.e.
993  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
994  */
995 void
996 ip6_savecontrol(in6p, mp, ip6, m)
997 	struct in6pcb *in6p;
998 	struct mbuf **mp;
999 	struct ip6_hdr *ip6;
1000 	struct mbuf *m;
1001 {
1002 	struct proc *p = curproc;	/* XXX */
1003 	int privileged;
1004 
1005 	privileged = 0;
1006 	if (p && !suser(p->p_ucred, &p->p_acflag))
1007 		privileged++;
1008 
1009 #ifdef SO_TIMESTAMP
1010 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1011 		struct timeval tv;
1012 
1013 		microtime(&tv);
1014 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1015 		    SCM_TIMESTAMP, SOL_SOCKET);
1016 		if (*mp)
1017 			mp = &(*mp)->m_next;
1018 	}
1019 #endif
1020 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
1021 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
1022 		    sizeof(struct in6_addr), IPV6_RECVDSTADDR, IPPROTO_IPV6);
1023 		if (*mp)
1024 			mp = &(*mp)->m_next;
1025 	}
1026 
1027 #ifdef noyet
1028 	/* options were tossed above */
1029 	if (in6p->in6p_flags & IN6P_RECVOPTS)
1030 		/* broken */
1031 	/* ip6_srcroute doesn't do what we want here, need to fix */
1032 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
1033 		/* broken */
1034 #endif
1035 
1036 	/* RFC 2292 sec. 5 */
1037 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1038 		struct in6_pktinfo pi6;
1039 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1040 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1041 			pi6.ipi6_addr.s6_addr16[1] = 0;
1042 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1043 					? m->m_pkthdr.rcvif->if_index
1044 					: 0;
1045 		*mp = sbcreatecontrol((caddr_t) &pi6,
1046 		    sizeof(struct in6_pktinfo), IPV6_PKTINFO, IPPROTO_IPV6);
1047 		if (*mp)
1048 			mp = &(*mp)->m_next;
1049 	}
1050 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1051 		int hlim = ip6->ip6_hlim & 0xff;
1052 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1053 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
1054 		if (*mp)
1055 			mp = &(*mp)->m_next;
1056 	}
1057 	/* IN6P_NEXTHOP - for outgoing packet only */
1058 
1059 	/*
1060 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1061 	 * for the option, but it might be too strict, since there might
1062 	 * be some hop-by-hop options which can be returned to normal user.
1063 	 * See RFC 2292 section 6.
1064 	 */
1065 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1066 		/*
1067 		 * Check if a hop-by-hop options header is contatined in the
1068 		 * received packet, and if so, store the options as ancillary
1069 		 * data. Note that a hop-by-hop options header must be
1070 		 * just after the IPv6 header, which fact is assured through
1071 		 * the IPv6 input processing.
1072 		 */
1073 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1074 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1075 			struct ip6_hbh *hbh;
1076 			int hbhlen;
1077 			struct mbuf *ext;
1078 
1079 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1080 			    ip6->ip6_nxt);
1081 			if (ext == NULL) {
1082 				ip6stat.ip6s_tooshort++;
1083 				return;
1084 			}
1085 			hbh = mtod(ext, struct ip6_hbh *);
1086 			hbhlen = (hbh->ip6h_len + 1) << 3;
1087 			if (hbhlen != ext->m_len) {
1088 				m_freem(ext);
1089 				ip6stat.ip6s_tooshort++;
1090 				return;
1091 			}
1092 
1093 			/*
1094 			 * XXX: We copy whole the header even if a jumbo
1095 			 * payload option is included, which option is to
1096 			 * be removed before returning in the RFC 2292.
1097 			 * But it's too painful operation...
1098 			 */
1099 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1100 			    IPV6_HOPOPTS, IPPROTO_IPV6);
1101 			if (*mp)
1102 				mp = &(*mp)->m_next;
1103 			m_freem(ext);
1104 		}
1105 	}
1106 
1107 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1108 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1109 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1110 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1111 
1112 		/*
1113 		 * Search for destination options headers or routing
1114 		 * header(s) through the header chain, and stores each
1115 		 * header as ancillary data.
1116 		 * Note that the order of the headers remains in
1117 		 * the chain of ancillary data.
1118 		 */
1119 		while (1) {	/* is explicit loop prevention necessary? */
1120 			struct ip6_ext *ip6e = NULL;
1121 			int elen;
1122 			struct mbuf *ext = NULL;
1123 
1124 			/*
1125 			 * if it is not an extension header, don't try to
1126 			 * pull it from the chain.
1127 			 */
1128 			switch (nxt) {
1129 			case IPPROTO_DSTOPTS:
1130 			case IPPROTO_ROUTING:
1131 			case IPPROTO_HOPOPTS:
1132 			case IPPROTO_AH: /* is it possible? */
1133 				break;
1134 			default:
1135 				goto loopend;
1136 			}
1137 
1138 			ext = ip6_pullexthdr(m, off, nxt);
1139 			if (ext == NULL) {
1140 				ip6stat.ip6s_tooshort++;
1141 				return;
1142 			}
1143 			ip6e = mtod(ext, struct ip6_ext *);
1144 			if (nxt == IPPROTO_AH)
1145 				elen = (ip6e->ip6e_len + 2) << 2;
1146 			else
1147 				elen = (ip6e->ip6e_len + 1) << 3;
1148 			if (elen != ext->m_len) {
1149 				m_freem(ext);
1150 				ip6stat.ip6s_tooshort++;
1151 				return;
1152 			}
1153 			KASSERT(IP6_HDR_ALIGNED_P(ip6e));
1154 
1155 			switch (nxt) {
1156 			case IPPROTO_DSTOPTS:
1157 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1158 					break;
1159 
1160 				/*
1161 				 * We also require super-user privilege for
1162 				 * the option.
1163 				 * See the comments on IN6_HOPOPTS.
1164 				 */
1165 				if (!privileged)
1166 					break;
1167 
1168 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1169 				    IPV6_DSTOPTS, IPPROTO_IPV6);
1170 				if (*mp)
1171 					mp = &(*mp)->m_next;
1172 				break;
1173 
1174 			case IPPROTO_ROUTING:
1175 				if (!in6p->in6p_flags & IN6P_RTHDR)
1176 					break;
1177 
1178 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1179 				    IPV6_RTHDR, IPPROTO_IPV6);
1180 				if (*mp)
1181 					mp = &(*mp)->m_next;
1182 				break;
1183 
1184 			case IPPROTO_HOPOPTS:
1185 			case IPPROTO_AH: /* is it possible? */
1186 				break;
1187 
1188 			default:
1189 				/*
1190 			 	 * other cases have been filtered in the above.
1191 				 * none will visit this case.  here we supply
1192 				 * the code just in case (nxt overwritten or
1193 				 * other cases).
1194 				 */
1195 				m_freem(ext);
1196 				goto loopend;
1197 
1198 			}
1199 
1200 			/* proceed with the next header. */
1201 			off += elen;
1202 			nxt = ip6e->ip6e_nxt;
1203 			ip6e = NULL;
1204 			m_freem(ext);
1205 			ext = NULL;
1206 		}
1207 	  loopend:
1208 	  	;
1209 	}
1210 }
1211 
1212 /*
1213  * pull single extension header from mbuf chain.  returns single mbuf that
1214  * contains the result, or NULL on error.
1215  */
1216 static struct mbuf *
1217 ip6_pullexthdr(m, off, nxt)
1218 	struct mbuf *m;
1219 	size_t off;
1220 	int nxt;
1221 {
1222 	struct ip6_ext ip6e;
1223 	size_t elen;
1224 	struct mbuf *n;
1225 
1226 #ifdef DIAGNOSTIC
1227 	switch (nxt) {
1228 	case IPPROTO_DSTOPTS:
1229 	case IPPROTO_ROUTING:
1230 	case IPPROTO_HOPOPTS:
1231 	case IPPROTO_AH: /* is it possible? */
1232 		break;
1233 	default:
1234 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1235 	}
1236 #endif
1237 
1238 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1239 	if (nxt == IPPROTO_AH)
1240 		elen = (ip6e.ip6e_len + 2) << 2;
1241 	else
1242 		elen = (ip6e.ip6e_len + 1) << 3;
1243 
1244 	MGET(n, M_DONTWAIT, MT_DATA);
1245 	if (n && elen >= MLEN) {
1246 		MCLGET(n, M_DONTWAIT);
1247 		if ((n->m_flags & M_EXT) == 0) {
1248 			m_free(n);
1249 			n = NULL;
1250 		}
1251 	}
1252 	if (!n)
1253 		return NULL;
1254 
1255 	n->m_len = 0;
1256 	if (elen >= M_TRAILINGSPACE(n)) {
1257 		m_free(n);
1258 		return NULL;
1259 	}
1260 
1261 	m_copydata(m, off, elen, mtod(n, caddr_t));
1262 	n->m_len = elen;
1263 	return n;
1264 }
1265 
1266 /*
1267  * Get pointer to the previous header followed by the header
1268  * currently processed.
1269  * XXX: This function supposes that
1270  *	M includes all headers,
1271  *	the next header field and the header length field of each header
1272  *	are valid, and
1273  *	the sum of each header length equals to OFF.
1274  * Because of these assumptions, this function must be called very
1275  * carefully. Moreover, it will not be used in the near future when
1276  * we develop `neater' mechanism to process extension headers.
1277  */
1278 u_int8_t *
1279 ip6_get_prevhdr(m, off)
1280 	struct mbuf *m;
1281 	int off;
1282 {
1283 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1284 
1285 	if (off == sizeof(struct ip6_hdr))
1286 		return (&ip6->ip6_nxt);
1287 	else {
1288 		int len, nxt;
1289 		struct ip6_ext *ip6e = NULL;
1290 
1291 		nxt = ip6->ip6_nxt;
1292 		len = sizeof(struct ip6_hdr);
1293 		while (len < off) {
1294 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1295 
1296 			switch (nxt) {
1297 			case IPPROTO_FRAGMENT:
1298 				len += sizeof(struct ip6_frag);
1299 				break;
1300 			case IPPROTO_AH:
1301 				len += (ip6e->ip6e_len + 2) << 2;
1302 				break;
1303 			default:
1304 				len += (ip6e->ip6e_len + 1) << 3;
1305 				break;
1306 			}
1307 			nxt = ip6e->ip6e_nxt;
1308 		}
1309 		if (ip6e)
1310 			return (&ip6e->ip6e_nxt);
1311 		else
1312 			return NULL;
1313 	}
1314 }
1315 
1316 /*
1317  * get next header offset.  m will be retained.
1318  */
1319 int
1320 ip6_nexthdr(m, off, proto, nxtp)
1321 	struct mbuf *m;
1322 	int off;
1323 	int proto;
1324 	int *nxtp;
1325 {
1326 	struct ip6_hdr ip6;
1327 	struct ip6_ext ip6e;
1328 	struct ip6_frag fh;
1329 
1330 	/* just in case */
1331 	if (m == NULL)
1332 		panic("ip6_nexthdr: m == NULL");
1333 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1334 		return -1;
1335 
1336 	switch (proto) {
1337 	case IPPROTO_IPV6:
1338 		if (m->m_pkthdr.len < off + sizeof(ip6))
1339 			return -1;
1340 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1341 		if (nxtp)
1342 			*nxtp = ip6.ip6_nxt;
1343 		off += sizeof(ip6);
1344 		return off;
1345 
1346 	case IPPROTO_FRAGMENT:
1347 		/*
1348 		 * terminate parsing if it is not the first fragment,
1349 		 * it does not make sense to parse through it.
1350 		 */
1351 		if (m->m_pkthdr.len < off + sizeof(fh))
1352 			return -1;
1353 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1354 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1355 			return -1;
1356 		if (nxtp)
1357 			*nxtp = fh.ip6f_nxt;
1358 		off += sizeof(struct ip6_frag);
1359 		return off;
1360 
1361 	case IPPROTO_AH:
1362 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1363 			return -1;
1364 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1365 		if (nxtp)
1366 			*nxtp = ip6e.ip6e_nxt;
1367 		off += (ip6e.ip6e_len + 2) << 2;
1368 		if (m->m_pkthdr.len < off)
1369 			return -1;
1370 		return off;
1371 
1372 	case IPPROTO_HOPOPTS:
1373 	case IPPROTO_ROUTING:
1374 	case IPPROTO_DSTOPTS:
1375 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1376 			return -1;
1377 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1378 		if (nxtp)
1379 			*nxtp = ip6e.ip6e_nxt;
1380 		off += (ip6e.ip6e_len + 1) << 3;
1381 		if (m->m_pkthdr.len < off)
1382 			return -1;
1383 		return off;
1384 
1385 	case IPPROTO_NONE:
1386 	case IPPROTO_ESP:
1387 	case IPPROTO_IPCOMP:
1388 		/* give up */
1389 		return -1;
1390 
1391 	default:
1392 		return -1;
1393 	}
1394 }
1395 
1396 /*
1397  * get offset for the last header in the chain.  m will be kept untainted.
1398  */
1399 int
1400 ip6_lasthdr(m, off, proto, nxtp)
1401 	struct mbuf *m;
1402 	int off;
1403 	int proto;
1404 	int *nxtp;
1405 {
1406 	int newoff;
1407 	int nxt;
1408 
1409 	if (!nxtp) {
1410 		nxt = -1;
1411 		nxtp = &nxt;
1412 	}
1413 	while (1) {
1414 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1415 		if (newoff < 0)
1416 			return off;
1417 		else if (newoff < off)
1418 			return -1;	/* invalid */
1419 		else if (newoff == off)
1420 			return newoff;
1421 
1422 		off = newoff;
1423 		proto = *nxtp;
1424 	}
1425 }
1426 
1427 /*
1428  * System control for IP6
1429  */
1430 
1431 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1432 	0,		0,		0,		0,
1433 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1434 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1435 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1436 	0,		0,		0,		0,
1437 	ENOPROTOOPT
1438 };
1439 
1440 int
1441 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1442 	int *name;
1443 	u_int namelen;
1444 	void *oldp;
1445 	size_t *oldlenp;
1446 	void *newp;
1447 	size_t newlen;
1448 {
1449 	int old, error;
1450 
1451 	/* All sysctl names at this level are terminal. */
1452 	if (namelen != 1)
1453 		return ENOTDIR;
1454 
1455 	switch (name[0]) {
1456 
1457 	case IPV6CTL_FORWARDING:
1458 		return sysctl_int(oldp, oldlenp, newp, newlen,
1459 				  &ip6_forwarding);
1460 	case IPV6CTL_SENDREDIRECTS:
1461 		return sysctl_int(oldp, oldlenp, newp, newlen,
1462 				&ip6_sendredirects);
1463 	case IPV6CTL_DEFHLIM:
1464 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1465 	case IPV6CTL_MAXFRAGPACKETS:
1466 		return sysctl_int(oldp, oldlenp, newp, newlen,
1467 				&ip6_maxfragpackets);
1468 	case IPV6CTL_ACCEPT_RTADV:
1469 		return sysctl_int(oldp, oldlenp, newp, newlen,
1470 				&ip6_accept_rtadv);
1471 	case IPV6CTL_KEEPFAITH:
1472 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1473 	case IPV6CTL_LOG_INTERVAL:
1474 		return sysctl_int(oldp, oldlenp, newp, newlen,
1475 				&ip6_log_interval);
1476 	case IPV6CTL_HDRNESTLIMIT:
1477 		return sysctl_int(oldp, oldlenp, newp, newlen,
1478 				&ip6_hdrnestlimit);
1479 	case IPV6CTL_DAD_COUNT:
1480 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1481 	case IPV6CTL_AUTO_FLOWLABEL:
1482 		return sysctl_int(oldp, oldlenp, newp, newlen,
1483 				&ip6_auto_flowlabel);
1484 	case IPV6CTL_DEFMCASTHLIM:
1485 		return sysctl_int(oldp, oldlenp, newp, newlen,
1486 				&ip6_defmcasthlim);
1487 #if NGIF > 0
1488 	case IPV6CTL_GIF_HLIM:
1489 		return sysctl_int(oldp, oldlenp, newp, newlen,
1490 				&ip6_gif_hlim);
1491 #endif
1492 	case IPV6CTL_KAME_VERSION:
1493 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1494 	case IPV6CTL_USE_DEPRECATED:
1495 		return sysctl_int(oldp, oldlenp, newp, newlen,
1496 				&ip6_use_deprecated);
1497 	case IPV6CTL_RR_PRUNE:
1498 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1499 	case IPV6CTL_V6ONLY:
1500 #ifdef INET6_BINDV6ONLY
1501 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1502 #else
1503 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_v6only);
1504 #endif
1505 	case IPV6CTL_ANONPORTMIN:
1506 		old = ip6_anonportmin;
1507 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1508 		    &ip6_anonportmin);
1509 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmin < 0 ||
1510 		    ip6_anonportmin > 65535
1511 #ifndef IPNOPRIVPORTS
1512 		    || ip6_anonportmin < IPV6PORT_RESERVED
1513 #endif
1514 		    ) {
1515 			ip6_anonportmin = old;
1516 			return (EINVAL);
1517 		}
1518 		return (error);
1519 	case IPV6CTL_ANONPORTMAX:
1520 		old = ip6_anonportmax;
1521 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1522 		    &ip6_anonportmax);
1523 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmax < 0 ||
1524 		    ip6_anonportmax > 65535
1525 #ifndef IPNOPRIVPORTS
1526 		    || ip6_anonportmax < IPV6PORT_RESERVED
1527 #endif
1528 		    ) {
1529 			ip6_anonportmax = old;
1530 			return (EINVAL);
1531 		}
1532 		return (error);
1533 #ifndef IPNOPRIVPORTS
1534 	case IPV6CTL_LOWPORTMIN:
1535 		old = ip6_lowportmin;
1536 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1537 		    &ip6_lowportmin);
1538 		if (ip6_lowportmin >= ip6_lowportmax ||
1539 		    ip6_lowportmin > IPV6PORT_RESERVEDMAX ||
1540 		    ip6_lowportmin < IPV6PORT_RESERVEDMIN) {
1541 			ip6_lowportmin = old;
1542 			return (EINVAL);
1543 		}
1544 		return (error);
1545 	case IPV6CTL_LOWPORTMAX:
1546 		old = ip6_lowportmax;
1547 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1548 		    &ip6_lowportmax);
1549 		if (ip6_lowportmin >= ip6_lowportmax ||
1550 		    ip6_lowportmax > IPV6PORT_RESERVEDMAX ||
1551 		    ip6_lowportmax < IPV6PORT_RESERVEDMIN) {
1552 			ip6_lowportmax = old;
1553 			return (EINVAL);
1554 		}
1555 		return (error);
1556 #endif
1557 	case IPV6CTL_MAXFRAGS:
1558 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_maxfrags);
1559 	default:
1560 		return EOPNOTSUPP;
1561 	}
1562 	/* NOTREACHED */
1563 }
1564