xref: /netbsd-src/sys/netinet6/ip6_input.c (revision 08c81a9c2dc8c7300e893321eb65c0925d60871c)
1 /*	$NetBSD: ip6_input.c,v 1.58 2002/09/11 02:41:25 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 1.58 2002/09/11 02:41:25 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 
153 /*
154  * IP6 initialization: fill in IP6 protocol switch table.
155  * All protocols not implemented in kernel go to raw IP6 protocol handler.
156  */
157 void
158 ip6_init()
159 {
160 	struct ip6protosw *pr;
161 	int i;
162 
163 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
164 	if (pr == 0)
165 		panic("ip6_init");
166 	for (i = 0; i < IPPROTO_MAX; i++)
167 		ip6_protox[i] = pr - inet6sw;
168 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
169 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
170 		if (pr->pr_domain->dom_family == PF_INET6 &&
171 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
172 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
173 	ip6intrq.ifq_maxlen = ip6qmaxlen;
174 	nd6_init();
175 	frag6_init();
176 	ip6_flow_seq = arc4random();
177 
178 	ip6_init2((void *)0);
179 
180 #ifdef PFIL_HOOKS
181 	/* Register our Packet Filter hook. */
182 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
183 	inet6_pfil_hook.ph_af   = AF_INET6;
184 	i = pfil_head_register(&inet6_pfil_hook);
185 	if (i != 0)
186 		printf("ip6_init: WARNING: unable to register pfil hook, "
187 		    "error %d\n", i);
188 #endif /* PFIL_HOOKS */
189 }
190 
191 static void
192 ip6_init2(dummy)
193 	void *dummy;
194 {
195 
196 	/* nd6_timer_init */
197 	callout_init(&nd6_timer_ch);
198 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
199 }
200 
201 /*
202  * IP6 input interrupt handling. Just pass the packet to ip6_input.
203  */
204 void
205 ip6intr()
206 {
207 	int s;
208 	struct mbuf *m;
209 
210 	for (;;) {
211 		s = splnet();
212 		IF_DEQUEUE(&ip6intrq, m);
213 		splx(s);
214 		if (m == 0)
215 			return;
216 		ip6_input(m);
217 	}
218 }
219 
220 extern struct	route_in6 ip6_forward_rt;
221 
222 void
223 ip6_input(m)
224 	struct mbuf *m;
225 {
226 	struct ip6_hdr *ip6;
227 	int off = sizeof(struct ip6_hdr), nest;
228 	u_int32_t plen;
229 	u_int32_t rtalert = ~0;
230 	int nxt, ours = 0;
231 	struct ifnet *deliverifp = NULL;
232 
233 #ifdef IPSEC
234 	/*
235 	 * should the inner packet be considered authentic?
236 	 * see comment in ah4_input().
237 	 */
238 	m->m_flags &= ~M_AUTHIPHDR;
239 	m->m_flags &= ~M_AUTHIPDGM;
240 #endif
241 
242 	/*
243 	 * mbuf statistics
244 	 */
245 	if (m->m_flags & M_EXT) {
246 		if (m->m_next)
247 			ip6stat.ip6s_mext2m++;
248 		else
249 			ip6stat.ip6s_mext1++;
250 	} else {
251 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
252 		if (m->m_next) {
253 			if (m->m_flags & M_LOOP) {
254 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
255 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
256 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
257 			else
258 				ip6stat.ip6s_m2m[0]++;
259 		} else
260 			ip6stat.ip6s_m1++;
261 #undef M2MMAX
262 	}
263 
264 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
265 	ip6stat.ip6s_total++;
266 
267 #ifndef PULLDOWN_TEST
268 	/* XXX is the line really necessary? */
269 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
270 #endif
271 
272 	/*
273 	 * If the IPv6 header is not aligned, slurp it up into a new
274 	 * mbuf with space for link headers, in the event we forward
275 	 * it.  OTherwise, if it is aligned, make sure the entire base
276 	 * IPv6 header is in the first mbuf of the chain.
277 	 */
278 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
279 		struct ifnet *inifp = m->m_pkthdr.rcvif;
280 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
281 				  (max_linkhdr + 3) & ~3)) == NULL) {
282 			/* XXXJRT new stat, please */
283 			ip6stat.ip6s_toosmall++;
284 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
285 			return;
286 		}
287 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
288 		struct ifnet *inifp = m->m_pkthdr.rcvif;
289 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
290 			ip6stat.ip6s_toosmall++;
291 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
292 			return;
293 		}
294 	}
295 
296 	ip6 = mtod(m, struct ip6_hdr *);
297 
298 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
299 		ip6stat.ip6s_badvers++;
300 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
301 		goto bad;
302 	}
303 
304 #ifdef PFIL_HOOKS
305 	/*
306 	 * Run through list of hooks for input packets.  If there are any
307 	 * filters which require that additional packets in the flow are
308 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
309 	 * Note that filters must _never_ set this flag, as another filter
310 	 * in the list may have previously cleared it.
311 	 */
312 	/*
313 	 * let ipfilter look at packet on the wire,
314 	 * not the decapsulated packet.
315 	 */
316 #ifdef IPSEC
317 	if (!ipsec_getnhist(m))
318 #else
319 	if (1)
320 #endif
321 	{
322 		if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif,
323 				   PFIL_IN) != 0)
324 			return;
325 		if (m == NULL)
326 			return;
327 		ip6 = mtod(m, struct ip6_hdr *);
328 	}
329 #endif /* PFIL_HOOKS */
330 
331 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
332 
333 #ifdef ALTQ
334 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
335 		/* packet is dropped by traffic conditioner */
336 		return;
337 	}
338 #endif
339 
340 	/*
341 	 * Check against address spoofing/corruption.
342 	 */
343 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
344 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
345 		/*
346 		 * XXX: "badscope" is not very suitable for a multicast source.
347 		 */
348 		ip6stat.ip6s_badscope++;
349 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
350 		goto bad;
351 	}
352 	/*
353 	 * The following check is not documented in specs.  A malicious
354 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
355 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
356 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
357 	 *
358 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
359 	 * support IPv4-less kernel compilation, we cannot support SIIT
360 	 * environment at all.  So, it makes more sense for us to reject any
361 	 * malicious packets for non-SIIT environment, than try to do a
362 	 * partial support for SIIT environment.
363 	 */
364 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
365 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
366 		ip6stat.ip6s_badscope++;
367 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
368 		goto bad;
369 	}
370 #if 0
371 	/*
372 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
373 	 *
374 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
375 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
376 	 * is revised to forbid relaying case.
377 	 */
378 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
379 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
380 		ip6stat.ip6s_badscope++;
381 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
382 		goto bad;
383 	}
384 #endif
385 
386 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
387 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
388 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
389 			ours = 1;
390 			deliverifp = m->m_pkthdr.rcvif;
391 			goto hbhcheck;
392 		} else {
393 			ip6stat.ip6s_badscope++;
394 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
395 			goto bad;
396 		}
397 	}
398 
399 	/* drop packets if interface ID portion is already filled */
400 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
401 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
402 		    ip6->ip6_src.s6_addr16[1]) {
403 			ip6stat.ip6s_badscope++;
404 			goto bad;
405 		}
406 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
407 		    ip6->ip6_dst.s6_addr16[1]) {
408 			ip6stat.ip6s_badscope++;
409 			goto bad;
410 		}
411 	}
412 
413 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
414 		ip6->ip6_src.s6_addr16[1]
415 			= htons(m->m_pkthdr.rcvif->if_index);
416 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
417 		ip6->ip6_dst.s6_addr16[1]
418 			= htons(m->m_pkthdr.rcvif->if_index);
419 
420 	/*
421 	 * We use rt->rt_ifp to determine if the address is ours or not.
422 	 * If rt_ifp is lo0, the address is ours.
423 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
424 	 * so any address under fe80::%lo0/64 will be mistakenly considered
425 	 * local.  The special case is supplied to handle the case properly
426 	 * by actually looking at interface addresses
427 	 * (using in6ifa_ifpwithaddr).
428 	 */
429 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
430 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
431 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
432 			icmp6_error(m, ICMP6_DST_UNREACH,
433 			    ICMP6_DST_UNREACH_ADDR, 0);
434 			/* m is already freed */
435 			return;
436 		}
437 
438 		ours = 1;
439 		deliverifp = m->m_pkthdr.rcvif;
440 		goto hbhcheck;
441 	}
442 
443 	/*
444 	 * Multicast check
445 	 */
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
447 	  	struct	in6_multi *in6m = 0;
448 
449 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
450 		/*
451 		 * See if we belong to the destination multicast group on the
452 		 * arrival interface.
453 		 */
454 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
455 		if (in6m)
456 			ours = 1;
457 		else if (!ip6_mrouter) {
458 			ip6stat.ip6s_notmember++;
459 			ip6stat.ip6s_cantforward++;
460 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
461 			goto bad;
462 		}
463 		deliverifp = m->m_pkthdr.rcvif;
464 		goto hbhcheck;
465 	}
466 
467 	/*
468 	 *  Unicast check
469 	 */
470 	if (ip6_forward_rt.ro_rt != NULL &&
471 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
472 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
473 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
474 		ip6stat.ip6s_forward_cachehit++;
475 	else {
476 		struct sockaddr_in6 *dst6;
477 
478 		if (ip6_forward_rt.ro_rt) {
479 			/* route is down or destination is different */
480 			ip6stat.ip6s_forward_cachemiss++;
481 			RTFREE(ip6_forward_rt.ro_rt);
482 			ip6_forward_rt.ro_rt = 0;
483 		}
484 
485 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
486 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
487 		dst6->sin6_len = sizeof(struct sockaddr_in6);
488 		dst6->sin6_family = AF_INET6;
489 		dst6->sin6_addr = ip6->ip6_dst;
490 
491 		rtalloc((struct route *)&ip6_forward_rt);
492 	}
493 
494 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
495 
496 	/*
497 	 * Accept the packet if the forwarding interface to the destination
498 	 * according to the routing table is the loopback interface,
499 	 * unless the associated route has a gateway.
500 	 * Note that this approach causes to accept a packet if there is a
501 	 * route to the loopback interface for the destination of the packet.
502 	 * But we think it's even useful in some situations, e.g. when using
503 	 * a special daemon which wants to intercept the packet.
504 	 */
505 	if (ip6_forward_rt.ro_rt &&
506 	    (ip6_forward_rt.ro_rt->rt_flags &
507 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
508 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
509 #if 0
510 	    /*
511 	     * The check below is redundant since the comparison of
512 	     * the destination and the key of the rtentry has
513 	     * already done through looking up the routing table.
514 	     */
515 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
516 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
517 #endif
518 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
519 		struct in6_ifaddr *ia6 =
520 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
521 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
522 			m->m_flags |= M_ANYCAST6;
523 		/*
524 		 * packets to a tentative, duplicated, or somehow invalid
525 		 * address must not be accepted.
526 		 */
527 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
528 			/* this address is ready */
529 			ours = 1;
530 			deliverifp = ia6->ia_ifp;	/* correct? */
531 			goto hbhcheck;
532 		} else {
533 			/* address is not ready, so discard the packet. */
534 			nd6log((LOG_INFO,
535 			    "ip6_input: packet to an unready address %s->%s\n",
536 			    ip6_sprintf(&ip6->ip6_src),
537 			    ip6_sprintf(&ip6->ip6_dst)));
538 
539 			goto bad;
540 		}
541 	}
542 
543 	/*
544 	 * FAITH (Firewall Aided Internet Translator)
545 	 */
546 #if defined(NFAITH) && 0 < NFAITH
547 	if (ip6_keepfaith) {
548 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
549 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
550 			/* XXX do we need more sanity checks? */
551 			ours = 1;
552 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
553 			goto hbhcheck;
554 		}
555 	}
556 #endif
557 
558 #if 0
559     {
560 	/*
561 	 * Last resort: check in6_ifaddr for incoming interface.
562 	 * The code is here until I update the "goto ours hack" code above
563 	 * working right.
564 	 */
565 	struct ifaddr *ifa;
566 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
567 	     ifa;
568 	     ifa = ifa->ifa_list.tqe_next) {
569 		if (ifa->ifa_addr == NULL)
570 			continue;	/* just for safety */
571 		if (ifa->ifa_addr->sa_family != AF_INET6)
572 			continue;
573 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
574 			ours = 1;
575 			deliverifp = ifa->ifa_ifp;
576 			goto hbhcheck;
577 		}
578 	}
579     }
580 #endif
581 
582 	/*
583 	 * Now there is no reason to process the packet if it's not our own
584 	 * and we're not a router.
585 	 */
586 	if (!ip6_forwarding) {
587 		ip6stat.ip6s_cantforward++;
588 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
589 		goto bad;
590 	}
591 
592   hbhcheck:
593 	/*
594 	 * Process Hop-by-Hop options header if it's contained.
595 	 * m may be modified in ip6_hopopts_input().
596 	 * If a JumboPayload option is included, plen will also be modified.
597 	 */
598 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
599 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
600 		struct ip6_hbh *hbh;
601 
602 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
603 #if 0	/*touches NULL pointer*/
604 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
605 #endif
606 			return;	/* m have already been freed */
607 		}
608 
609 		/* adjust pointer */
610 		ip6 = mtod(m, struct ip6_hdr *);
611 
612 		/*
613 		 * if the payload length field is 0 and the next header field
614 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
615 		 * option MUST be included.
616 		 */
617 		if (ip6->ip6_plen == 0 && plen == 0) {
618 			/*
619 			 * Note that if a valid jumbo payload option is
620 			 * contained, ip6_hoptops_input() must set a valid
621 			 * (non-zero) payload length to the variable plen.
622 			 */
623 			ip6stat.ip6s_badoptions++;
624 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
625 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
626 			icmp6_error(m, ICMP6_PARAM_PROB,
627 				    ICMP6_PARAMPROB_HEADER,
628 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
629 			return;
630 		}
631 #ifndef PULLDOWN_TEST
632 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
633 		hbh = (struct ip6_hbh *)(ip6 + 1);
634 #else
635 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
636 			sizeof(struct ip6_hbh));
637 		if (hbh == NULL) {
638 			ip6stat.ip6s_tooshort++;
639 			return;
640 		}
641 #endif
642 		KASSERT(IP6_HDR_ALIGNED_P(hbh));
643 		nxt = hbh->ip6h_nxt;
644 
645 		/*
646 		 * accept the packet if a router alert option is included
647 		 * and we act as an IPv6 router.
648 		 */
649 		if (rtalert != ~0 && ip6_forwarding)
650 			ours = 1;
651 	} else
652 		nxt = ip6->ip6_nxt;
653 
654 	/*
655 	 * Check that the amount of data in the buffers
656 	 * is as at least much as the IPv6 header would have us expect.
657 	 * Trim mbufs if longer than we expect.
658 	 * Drop packet if shorter than we expect.
659 	 */
660 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
661 		ip6stat.ip6s_tooshort++;
662 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
663 		goto bad;
664 	}
665 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
666 		if (m->m_len == m->m_pkthdr.len) {
667 			m->m_len = sizeof(struct ip6_hdr) + plen;
668 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
669 		} else
670 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
671 	}
672 
673 	/*
674 	 * Forward if desirable.
675 	 */
676 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
677 		/*
678 		 * If we are acting as a multicast router, all
679 		 * incoming multicast packets are passed to the
680 		 * kernel-level multicast forwarding function.
681 		 * The packet is returned (relatively) intact; if
682 		 * ip6_mforward() returns a non-zero value, the packet
683 		 * must be discarded, else it may be accepted below.
684 		 */
685 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
686 			ip6stat.ip6s_cantforward++;
687 			m_freem(m);
688 			return;
689 		}
690 		if (!ours) {
691 			m_freem(m);
692 			return;
693 		}
694 	} else if (!ours) {
695 		ip6_forward(m, 0);
696 		return;
697 	}
698 
699 	ip6 = mtod(m, struct ip6_hdr *);
700 
701 	/*
702 	 * Malicious party may be able to use IPv4 mapped addr to confuse
703 	 * tcp/udp stack and bypass security checks (act as if it was from
704 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
705 	 *
706 	 * For SIIT end node behavior, you may want to disable the check.
707 	 * However, you will  become vulnerable to attacks using IPv4 mapped
708 	 * source.
709 	 */
710 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
711 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
712 		ip6stat.ip6s_badscope++;
713 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
714 		goto bad;
715 	}
716 
717 	/*
718 	 * Tell launch routine the next header
719 	 */
720 #ifdef IFA_STATS
721 	if (deliverifp != NULL) {
722 		struct in6_ifaddr *ia6;
723 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
724 		if (ia6)
725 			ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
726 	}
727 #endif
728 	ip6stat.ip6s_delivered++;
729 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
730 	nest = 0;
731 
732 	while (nxt != IPPROTO_DONE) {
733 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
734 			ip6stat.ip6s_toomanyhdr++;
735 			goto bad;
736 		}
737 
738 		/*
739 		 * protection against faulty packet - there should be
740 		 * more sanity checks in header chain processing.
741 		 */
742 		if (m->m_pkthdr.len < off) {
743 			ip6stat.ip6s_tooshort++;
744 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
745 			goto bad;
746 		}
747 
748 #ifdef IPSEC
749 		/*
750 		 * enforce IPsec policy checking if we are seeing last header.
751 		 * note that we do not visit this with protocols with pcb layer
752 		 * code - like udp/tcp/raw ip.
753 		 */
754 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
755 		    ipsec6_in_reject(m, NULL)) {
756 			ipsec6stat.in_polvio++;
757 			goto bad;
758 		}
759 #endif
760 
761 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
762 	}
763 	return;
764  bad:
765 	m_freem(m);
766 }
767 
768 /*
769  * Hop-by-Hop options header processing. If a valid jumbo payload option is
770  * included, the real payload length will be stored in plenp.
771  */
772 static int
773 ip6_hopopts_input(plenp, rtalertp, mp, offp)
774 	u_int32_t *plenp;
775 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
776 	struct mbuf **mp;
777 	int *offp;
778 {
779 	struct mbuf *m = *mp;
780 	int off = *offp, hbhlen;
781 	struct ip6_hbh *hbh;
782 	u_int8_t *opt;
783 
784 	/* validation of the length of the header */
785 #ifndef PULLDOWN_TEST
786 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
787 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
788 	hbhlen = (hbh->ip6h_len + 1) << 3;
789 
790 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
791 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
792 #else
793 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
794 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
795 	if (hbh == NULL) {
796 		ip6stat.ip6s_tooshort++;
797 		return -1;
798 	}
799 	hbhlen = (hbh->ip6h_len + 1) << 3;
800 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
801 		hbhlen);
802 	if (hbh == NULL) {
803 		ip6stat.ip6s_tooshort++;
804 		return -1;
805 	}
806 #endif
807 	KASSERT(IP6_HDR_ALIGNED_P(hbh));
808 	off += hbhlen;
809 	hbhlen -= sizeof(struct ip6_hbh);
810 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
811 
812 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
813 				hbhlen, rtalertp, plenp) < 0)
814 		return (-1);
815 
816 	*offp = off;
817 	*mp = m;
818 	return (0);
819 }
820 
821 /*
822  * Search header for all Hop-by-hop options and process each option.
823  * This function is separate from ip6_hopopts_input() in order to
824  * handle a case where the sending node itself process its hop-by-hop
825  * options header. In such a case, the function is called from ip6_output().
826  *
827  * The function assumes that hbh header is located right after the IPv6 header
828  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
829  * opthead + hbhlen is located in continuous memory region.
830  */
831 int
832 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
833 	struct mbuf *m;
834 	u_int8_t *opthead;
835 	int hbhlen;
836 	u_int32_t *rtalertp;
837 	u_int32_t *plenp;
838 {
839 	struct ip6_hdr *ip6;
840 	int optlen = 0;
841 	u_int8_t *opt = opthead;
842 	u_int16_t rtalert_val;
843 	u_int32_t jumboplen;
844 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
845 
846 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
847 		switch (*opt) {
848 		case IP6OPT_PAD1:
849 			optlen = 1;
850 			break;
851 		case IP6OPT_PADN:
852 			if (hbhlen < IP6OPT_MINLEN) {
853 				ip6stat.ip6s_toosmall++;
854 				goto bad;
855 			}
856 			optlen = *(opt + 1) + 2;
857 			break;
858 		case IP6OPT_RTALERT:
859 			/* XXX may need check for alignment */
860 			if (hbhlen < IP6OPT_RTALERT_LEN) {
861 				ip6stat.ip6s_toosmall++;
862 				goto bad;
863 			}
864 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
865 				/* XXX stat */
866 				icmp6_error(m, ICMP6_PARAM_PROB,
867 				    ICMP6_PARAMPROB_HEADER,
868 				    erroff + opt + 1 - opthead);
869 				return (-1);
870 			}
871 			optlen = IP6OPT_RTALERT_LEN;
872 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
873 			*rtalertp = ntohs(rtalert_val);
874 			break;
875 		case IP6OPT_JUMBO:
876 			/* XXX may need check for alignment */
877 			if (hbhlen < IP6OPT_JUMBO_LEN) {
878 				ip6stat.ip6s_toosmall++;
879 				goto bad;
880 			}
881 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
882 				/* XXX stat */
883 				icmp6_error(m, ICMP6_PARAM_PROB,
884 				    ICMP6_PARAMPROB_HEADER,
885 				    erroff + opt + 1 - opthead);
886 				return (-1);
887 			}
888 			optlen = IP6OPT_JUMBO_LEN;
889 
890 			/*
891 			 * IPv6 packets that have non 0 payload length
892 			 * must not contain a jumbo payload option.
893 			 */
894 			ip6 = mtod(m, struct ip6_hdr *);
895 			if (ip6->ip6_plen) {
896 				ip6stat.ip6s_badoptions++;
897 				icmp6_error(m, ICMP6_PARAM_PROB,
898 				    ICMP6_PARAMPROB_HEADER,
899 				    erroff + opt - opthead);
900 				return (-1);
901 			}
902 
903 			/*
904 			 * We may see jumbolen in unaligned location, so
905 			 * we'd need to perform bcopy().
906 			 */
907 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
908 			jumboplen = (u_int32_t)htonl(jumboplen);
909 
910 #if 1
911 			/*
912 			 * if there are multiple jumbo payload options,
913 			 * *plenp will be non-zero and the packet will be
914 			 * rejected.
915 			 * the behavior may need some debate in ipngwg -
916 			 * multiple options does not make sense, however,
917 			 * there's no explicit mention in specification.
918 			 */
919 			if (*plenp != 0) {
920 				ip6stat.ip6s_badoptions++;
921 				icmp6_error(m, ICMP6_PARAM_PROB,
922 				    ICMP6_PARAMPROB_HEADER,
923 				    erroff + opt + 2 - opthead);
924 				return (-1);
925 			}
926 #endif
927 
928 			/*
929 			 * jumbo payload length must be larger than 65535.
930 			 */
931 			if (jumboplen <= IPV6_MAXPACKET) {
932 				ip6stat.ip6s_badoptions++;
933 				icmp6_error(m, ICMP6_PARAM_PROB,
934 				    ICMP6_PARAMPROB_HEADER,
935 				    erroff + opt + 2 - opthead);
936 				return (-1);
937 			}
938 			*plenp = jumboplen;
939 
940 			break;
941 		default:		/* unknown option */
942 			if (hbhlen < IP6OPT_MINLEN) {
943 				ip6stat.ip6s_toosmall++;
944 				goto bad;
945 			}
946 			optlen = ip6_unknown_opt(opt, m,
947 			    erroff + opt - opthead);
948 			if (optlen == -1)
949 				return (-1);
950 			optlen += 2;
951 			break;
952 		}
953 	}
954 
955 	return (0);
956 
957   bad:
958 	m_freem(m);
959 	return (-1);
960 }
961 
962 /*
963  * Unknown option processing.
964  * The third argument `off' is the offset from the IPv6 header to the option,
965  * which is necessary if the IPv6 header the and option header and IPv6 header
966  * is not continuous in order to return an ICMPv6 error.
967  */
968 int
969 ip6_unknown_opt(optp, m, off)
970 	u_int8_t *optp;
971 	struct mbuf *m;
972 	int off;
973 {
974 	struct ip6_hdr *ip6;
975 
976 	switch (IP6OPT_TYPE(*optp)) {
977 	case IP6OPT_TYPE_SKIP: /* ignore the option */
978 		return ((int)*(optp + 1));
979 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
980 		m_freem(m);
981 		return (-1);
982 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
983 		ip6stat.ip6s_badoptions++;
984 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
985 		return (-1);
986 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
987 		ip6stat.ip6s_badoptions++;
988 		ip6 = mtod(m, struct ip6_hdr *);
989 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
990 		    (m->m_flags & (M_BCAST|M_MCAST)))
991 			m_freem(m);
992 		else
993 			icmp6_error(m, ICMP6_PARAM_PROB,
994 				    ICMP6_PARAMPROB_OPTION, off);
995 		return (-1);
996 	}
997 
998 	m_freem(m);		/* XXX: NOTREACHED */
999 	return (-1);
1000 }
1001 
1002 /*
1003  * Create the "control" list for this pcb.
1004  *
1005  * The routine will be called from upper layer handlers like tcp6_input().
1006  * Thus the routine assumes that the caller (tcp6_input) have already
1007  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1008  * very first mbuf on the mbuf chain.
1009  * We may want to add some infinite loop prevention or sanity checks for safety.
1010  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1011  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1012  */
1013 void
1014 ip6_savecontrol(in6p, mp, ip6, m)
1015 	struct in6pcb *in6p;
1016 	struct mbuf **mp;
1017 	struct ip6_hdr *ip6;
1018 	struct mbuf *m;
1019 {
1020 	struct proc *p = curproc;	/* XXX */
1021 	int privileged;
1022 
1023 	privileged = 0;
1024 	if (p && !suser(p->p_ucred, &p->p_acflag))
1025 		privileged++;
1026 
1027 #ifdef SO_TIMESTAMP
1028 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1029 		struct timeval tv;
1030 
1031 		microtime(&tv);
1032 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1033 			SCM_TIMESTAMP, SOL_SOCKET);
1034 		if (*mp)
1035 			mp = &(*mp)->m_next;
1036 	}
1037 #endif
1038 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
1039 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
1040 			sizeof(struct in6_addr), IPV6_RECVDSTADDR,
1041 			IPPROTO_IPV6);
1042 		if (*mp)
1043 			mp = &(*mp)->m_next;
1044 	}
1045 
1046 #ifdef noyet
1047 	/* options were tossed above */
1048 	if (in6p->in6p_flags & IN6P_RECVOPTS)
1049 		/* broken */
1050 	/* ip6_srcroute doesn't do what we want here, need to fix */
1051 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
1052 		/* broken */
1053 #endif
1054 
1055 	/* RFC 2292 sec. 5 */
1056 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1057 		struct in6_pktinfo pi6;
1058 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1059 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1060 			pi6.ipi6_addr.s6_addr16[1] = 0;
1061 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1062 					? m->m_pkthdr.rcvif->if_index
1063 					: 0;
1064 		*mp = sbcreatecontrol((caddr_t) &pi6,
1065 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1066 			IPPROTO_IPV6);
1067 		if (*mp)
1068 			mp = &(*mp)->m_next;
1069 	}
1070 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1071 		int hlim = ip6->ip6_hlim & 0xff;
1072 		*mp = sbcreatecontrol((caddr_t) &hlim,
1073 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1074 		if (*mp)
1075 			mp = &(*mp)->m_next;
1076 	}
1077 	/* IN6P_NEXTHOP - for outgoing packet only */
1078 
1079 	/*
1080 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1081 	 * for the option, but it might be too strict, since there might
1082 	 * be some hop-by-hop options which can be returned to normal user.
1083 	 * See RFC 2292 section 6.
1084 	 */
1085 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1086 		/*
1087 		 * Check if a hop-by-hop options header is contatined in the
1088 		 * received packet, and if so, store the options as ancillary
1089 		 * data. Note that a hop-by-hop options header must be
1090 		 * just after the IPv6 header, which fact is assured through
1091 		 * the IPv6 input processing.
1092 		 */
1093 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1094 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1095 			struct ip6_hbh *hbh;
1096 			int hbhlen;
1097 
1098 #ifndef PULLDOWN_TEST
1099 			hbh = (struct ip6_hbh *)(ip6 + 1);
1100 			hbhlen = (hbh->ip6h_len + 1) << 3;
1101 #else
1102 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1103 				sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1104 			if (hbh == NULL) {
1105 				ip6stat.ip6s_tooshort++;
1106 				return;
1107 			}
1108 			hbhlen = (hbh->ip6h_len + 1) << 3;
1109 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1110 				sizeof(struct ip6_hdr), hbhlen);
1111 			if (hbh == NULL) {
1112 				ip6stat.ip6s_tooshort++;
1113 				return;
1114 			}
1115 #endif
1116 
1117 			/*
1118 			 * XXX: We copy whole the header even if a jumbo
1119 			 * payload option is included, which option is to
1120 			 * be removed before returning in the RFC 2292.
1121 			 * But it's too painful operation...
1122 			 */
1123 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1124 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1125 			if (*mp)
1126 				mp = &(*mp)->m_next;
1127 		}
1128 	}
1129 
1130 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1131 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1132 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1133 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);;
1134 
1135 		/*
1136 		 * Search for destination options headers or routing
1137 		 * header(s) through the header chain, and stores each
1138 		 * header as ancillary data.
1139 		 * Note that the order of the headers remains in
1140 		 * the chain of ancillary data.
1141 		 */
1142 		while (1) {	/* is explicit loop prevention necessary? */
1143 			struct ip6_ext *ip6e;
1144 			int elen;
1145 
1146 #ifndef PULLDOWN_TEST
1147 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1148 			if (nxt == IPPROTO_AH)
1149 				elen = (ip6e->ip6e_len + 2) << 2;
1150 			else
1151 				elen = (ip6e->ip6e_len + 1) << 3;
1152 #else
1153 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off,
1154 				sizeof(struct ip6_ext));
1155 			if (ip6e == NULL) {
1156 				ip6stat.ip6s_tooshort++;
1157 				return;
1158 			}
1159 			if (nxt == IPPROTO_AH)
1160 				elen = (ip6e->ip6e_len + 2) << 2;
1161 			else
1162 				elen = (ip6e->ip6e_len + 1) << 3;
1163 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off, elen);
1164 			if (ip6e == NULL) {
1165 				ip6stat.ip6s_tooshort++;
1166 				return;
1167 			}
1168 #endif
1169 			KASSERT(IP6_HDR_ALIGNED_P(ip6e));
1170 
1171 			switch (nxt) {
1172 		        case IPPROTO_DSTOPTS:
1173 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1174 					break;
1175 
1176 				/*
1177 				 * We also require super-user privilege for
1178 				 * the option.
1179 				 * See the comments on IN6_HOPOPTS.
1180 				 */
1181 				if (!privileged)
1182 					break;
1183 
1184 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1185 						      IPV6_DSTOPTS,
1186 						      IPPROTO_IPV6);
1187 				if (*mp)
1188 					mp = &(*mp)->m_next;
1189 				break;
1190 
1191 			case IPPROTO_ROUTING:
1192 				if (!in6p->in6p_flags & IN6P_RTHDR)
1193 					break;
1194 
1195 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1196 						      IPV6_RTHDR,
1197 						      IPPROTO_IPV6);
1198 				if (*mp)
1199 					mp = &(*mp)->m_next;
1200 				break;
1201 
1202 			case IPPROTO_UDP:
1203 			case IPPROTO_TCP:
1204 			case IPPROTO_ICMPV6:
1205 			default:
1206 				/*
1207 				 * stop search if we encounter an upper
1208 				 * layer protocol headers.
1209 				 */
1210 				goto loopend;
1211 
1212 			case IPPROTO_HOPOPTS:
1213 			case IPPROTO_AH: /* is it possible? */
1214 				break;
1215 			}
1216 
1217 			/* proceed with the next header. */
1218 			off += elen;
1219 			nxt = ip6e->ip6e_nxt;
1220 		}
1221 	  loopend:
1222 		;
1223 	}
1224 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1225 		/* to be done */
1226 	}
1227 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1228 		/* to be done */
1229 	}
1230 	/* IN6P_RTHDR - to be done */
1231 
1232 }
1233 
1234 /*
1235  * Get pointer to the previous header followed by the header
1236  * currently processed.
1237  * XXX: This function supposes that
1238  *	M includes all headers,
1239  *	the next header field and the header length field of each header
1240  *	are valid, and
1241  *	the sum of each header length equals to OFF.
1242  * Because of these assumptions, this function must be called very
1243  * carefully. Moreover, it will not be used in the near future when
1244  * we develop `neater' mechanism to process extension headers.
1245  */
1246 u_int8_t *
1247 ip6_get_prevhdr(m, off)
1248 	struct mbuf *m;
1249 	int off;
1250 {
1251 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1252 
1253 	if (off == sizeof(struct ip6_hdr))
1254 		return (&ip6->ip6_nxt);
1255 	else {
1256 		int len, nxt;
1257 		struct ip6_ext *ip6e = NULL;
1258 
1259 		nxt = ip6->ip6_nxt;
1260 		len = sizeof(struct ip6_hdr);
1261 		while (len < off) {
1262 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1263 
1264 			switch (nxt) {
1265 			case IPPROTO_FRAGMENT:
1266 				len += sizeof(struct ip6_frag);
1267 				break;
1268 			case IPPROTO_AH:
1269 				len += (ip6e->ip6e_len + 2) << 2;
1270 				break;
1271 			default:
1272 				len += (ip6e->ip6e_len + 1) << 3;
1273 				break;
1274 			}
1275 			nxt = ip6e->ip6e_nxt;
1276 		}
1277 		if (ip6e)
1278 			return (&ip6e->ip6e_nxt);
1279 		else
1280 			return NULL;
1281 	}
1282 }
1283 
1284 /*
1285  * get next header offset.  m will be retained.
1286  */
1287 int
1288 ip6_nexthdr(m, off, proto, nxtp)
1289 	struct mbuf *m;
1290 	int off;
1291 	int proto;
1292 	int *nxtp;
1293 {
1294 	struct ip6_hdr ip6;
1295 	struct ip6_ext ip6e;
1296 	struct ip6_frag fh;
1297 
1298 	/* just in case */
1299 	if (m == NULL)
1300 		panic("ip6_nexthdr: m == NULL");
1301 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1302 		return -1;
1303 
1304 	switch (proto) {
1305 	case IPPROTO_IPV6:
1306 		if (m->m_pkthdr.len < off + sizeof(ip6))
1307 			return -1;
1308 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1309 		if (nxtp)
1310 			*nxtp = ip6.ip6_nxt;
1311 		off += sizeof(ip6);
1312 		return off;
1313 
1314 	case IPPROTO_FRAGMENT:
1315 		/*
1316 		 * terminate parsing if it is not the first fragment,
1317 		 * it does not make sense to parse through it.
1318 		 */
1319 		if (m->m_pkthdr.len < off + sizeof(fh))
1320 			return -1;
1321 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1322 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1323 			return -1;
1324 		if (nxtp)
1325 			*nxtp = fh.ip6f_nxt;
1326 		off += sizeof(struct ip6_frag);
1327 		return off;
1328 
1329 	case IPPROTO_AH:
1330 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1331 			return -1;
1332 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1333 		if (nxtp)
1334 			*nxtp = ip6e.ip6e_nxt;
1335 		off += (ip6e.ip6e_len + 2) << 2;
1336 		if (m->m_pkthdr.len < off)
1337 			return -1;
1338 		return off;
1339 
1340 	case IPPROTO_HOPOPTS:
1341 	case IPPROTO_ROUTING:
1342 	case IPPROTO_DSTOPTS:
1343 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1344 			return -1;
1345 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1346 		if (nxtp)
1347 			*nxtp = ip6e.ip6e_nxt;
1348 		off += (ip6e.ip6e_len + 1) << 3;
1349 		if (m->m_pkthdr.len < off)
1350 			return -1;
1351 		return off;
1352 
1353 	case IPPROTO_NONE:
1354 	case IPPROTO_ESP:
1355 	case IPPROTO_IPCOMP:
1356 		/* give up */
1357 		return -1;
1358 
1359 	default:
1360 		return -1;
1361 	}
1362 
1363 	return -1;
1364 }
1365 
1366 /*
1367  * get offset for the last header in the chain.  m will be kept untainted.
1368  */
1369 int
1370 ip6_lasthdr(m, off, proto, nxtp)
1371 	struct mbuf *m;
1372 	int off;
1373 	int proto;
1374 	int *nxtp;
1375 {
1376 	int newoff;
1377 	int nxt;
1378 
1379 	if (!nxtp) {
1380 		nxt = -1;
1381 		nxtp = &nxt;
1382 	}
1383 	while (1) {
1384 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1385 		if (newoff < 0)
1386 			return off;
1387 		else if (newoff < off)
1388 			return -1;	/* invalid */
1389 		else if (newoff == off)
1390 			return newoff;
1391 
1392 		off = newoff;
1393 		proto = *nxtp;
1394 	}
1395 }
1396 
1397 /*
1398  * System control for IP6
1399  */
1400 
1401 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1402 	0,		0,		0,		0,
1403 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1404 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1405 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1406 	0,		0,		0,		0,
1407 	ENOPROTOOPT
1408 };
1409 
1410 int
1411 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1412 	int *name;
1413 	u_int namelen;
1414 	void *oldp;
1415 	size_t *oldlenp;
1416 	void *newp;
1417 	size_t newlen;
1418 {
1419 	int old, error;
1420 
1421 	/* All sysctl names at this level are terminal. */
1422 	if (namelen != 1)
1423 		return ENOTDIR;
1424 
1425 	switch (name[0]) {
1426 
1427 	case IPV6CTL_FORWARDING:
1428 		return sysctl_int(oldp, oldlenp, newp, newlen,
1429 				  &ip6_forwarding);
1430 	case IPV6CTL_SENDREDIRECTS:
1431 		return sysctl_int(oldp, oldlenp, newp, newlen,
1432 				&ip6_sendredirects);
1433 	case IPV6CTL_DEFHLIM:
1434 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1435 	case IPV6CTL_MAXFRAGPACKETS:
1436 		return sysctl_int(oldp, oldlenp, newp, newlen,
1437 				&ip6_maxfragpackets);
1438 	case IPV6CTL_ACCEPT_RTADV:
1439 		return sysctl_int(oldp, oldlenp, newp, newlen,
1440 				&ip6_accept_rtadv);
1441 	case IPV6CTL_KEEPFAITH:
1442 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1443 	case IPV6CTL_LOG_INTERVAL:
1444 		return sysctl_int(oldp, oldlenp, newp, newlen,
1445 				&ip6_log_interval);
1446 	case IPV6CTL_HDRNESTLIMIT:
1447 		return sysctl_int(oldp, oldlenp, newp, newlen,
1448 				&ip6_hdrnestlimit);
1449 	case IPV6CTL_DAD_COUNT:
1450 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1451 	case IPV6CTL_AUTO_FLOWLABEL:
1452 		return sysctl_int(oldp, oldlenp, newp, newlen,
1453 				&ip6_auto_flowlabel);
1454 	case IPV6CTL_DEFMCASTHLIM:
1455 		return sysctl_int(oldp, oldlenp, newp, newlen,
1456 				&ip6_defmcasthlim);
1457 #if NGIF > 0
1458 	case IPV6CTL_GIF_HLIM:
1459 		return sysctl_int(oldp, oldlenp, newp, newlen,
1460 				&ip6_gif_hlim);
1461 #endif
1462 	case IPV6CTL_KAME_VERSION:
1463 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1464 	case IPV6CTL_USE_DEPRECATED:
1465 		return sysctl_int(oldp, oldlenp, newp, newlen,
1466 				&ip6_use_deprecated);
1467 	case IPV6CTL_RR_PRUNE:
1468 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1469 	case IPV6CTL_V6ONLY:
1470 #ifdef INET6_BINDV6ONLY
1471 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1472 #else
1473 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_v6only);
1474 #endif
1475 	case IPV6CTL_ANONPORTMIN:
1476 		old = ip6_anonportmin;
1477 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1478 		    &ip6_anonportmin);
1479 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmin < 0 ||
1480 		    ip6_anonportmin > 65535
1481 #ifndef IPNOPRIVPORTS
1482 		    || ip6_anonportmin < IPV6PORT_RESERVED
1483 #endif
1484 		    ) {
1485 			ip6_anonportmin = old;
1486 			return (EINVAL);
1487 		}
1488 		return (error);
1489 	case IPV6CTL_ANONPORTMAX:
1490 		old = ip6_anonportmax;
1491 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1492 		    &ip6_anonportmax);
1493 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmax < 0 ||
1494 		    ip6_anonportmax > 65535
1495 #ifndef IPNOPRIVPORTS
1496 		    || ip6_anonportmax < IPV6PORT_RESERVED
1497 #endif
1498 		    ) {
1499 			ip6_anonportmax = old;
1500 			return (EINVAL);
1501 		}
1502 		return (error);
1503 #ifndef IPNOPRIVPORTS
1504 	case IPV6CTL_LOWPORTMIN:
1505 		old = ip6_lowportmin;
1506 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1507 		    &ip6_lowportmin);
1508 		if (ip6_lowportmin >= ip6_lowportmax ||
1509 		    ip6_lowportmin > IPV6PORT_RESERVEDMAX ||
1510 		    ip6_lowportmin < IPV6PORT_RESERVEDMIN) {
1511 			ip6_lowportmin = old;
1512 			return (EINVAL);
1513 		}
1514 		return (error);
1515 	case IPV6CTL_LOWPORTMAX:
1516 		old = ip6_lowportmax;
1517 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1518 		    &ip6_lowportmax);
1519 		if (ip6_lowportmin >= ip6_lowportmax ||
1520 		    ip6_lowportmax > IPV6PORT_RESERVEDMAX ||
1521 		    ip6_lowportmax < IPV6PORT_RESERVEDMIN) {
1522 			ip6_lowportmax = old;
1523 			return (EINVAL);
1524 		}
1525 		return (error);
1526 #endif
1527 	case IPV6CTL_MAXFRAGS:
1528 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_maxfrags);
1529 	default:
1530 		return EOPNOTSUPP;
1531 	}
1532 	/* NOTREACHED */
1533 }
1534