xref: /netbsd/sys/netinet6/ip6_input.c (revision bf9ec67e)
1 /*	$NetBSD: ip6_input.c,v 1.54 2002/05/28 10:11:51 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 1.54 2002/05/28 10:11:51 itojun Exp $");
70 
71 #include "opt_inet.h"
72 #include "opt_ipsec.h"
73 #include "opt_pfil_hooks.h"
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 #include <sys/proc.h>
88 #include <sys/sysctl.h>
89 
90 #include <net/if.h>
91 #include <net/if_types.h>
92 #include <net/if_dl.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #ifdef PFIL_HOOKS
96 #include <net/pfil.h>
97 #endif
98 
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/in6_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/in6_ifattach.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_prefix.h>
113 
114 #ifdef IPSEC
115 #include <netinet6/ipsec.h>
116 #endif
117 
118 #include <netinet6/ip6protosw.h>
119 
120 /* we need it for NLOOP. */
121 #include "loop.h"
122 #include "faith.h"
123 #include "gif.h"
124 #include "bpfilter.h"
125 
126 #if NGIF > 0
127 #include <netinet6/in6_gif.h>
128 #endif
129 
130 #include <net/net_osdep.h>
131 
132 extern struct domain inet6domain;
133 
134 u_char ip6_protox[IPPROTO_MAX];
135 static int ip6qmaxlen = IFQ_MAXLEN;
136 struct in6_ifaddr *in6_ifaddr;
137 struct ifqueue ip6intrq;
138 
139 extern struct ifnet loif[NLOOP];
140 int ip6_forward_srcrt;			/* XXX */
141 int ip6_sourcecheck;			/* XXX */
142 int ip6_sourcecheck_interval;		/* XXX */
143 
144 #ifdef PFIL_HOOKS
145 struct pfil_head inet6_pfil_hook;
146 #endif
147 
148 struct ip6stat ip6stat;
149 
150 static void ip6_init2 __P((void *));
151 
152 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
153 
154 /*
155  * IP6 initialization: fill in IP6 protocol switch table.
156  * All protocols not implemented in kernel go to raw IP6 protocol handler.
157  */
158 void
159 ip6_init()
160 {
161 	struct ip6protosw *pr;
162 	int i;
163 
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 	 * to route local address of p2p link to loopback,
198 	 * assign loopback address first.
199 	 */
200 	in6_ifattach(&loif[0], NULL);
201 
202 	/* nd6_timer_init */
203 	callout_init(&nd6_timer_ch);
204 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
205 	/* router renumbering prefix list maintenance */
206 	callout_init(&in6_rr_timer_ch);
207 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
208 }
209 
210 /*
211  * IP6 input interrupt handling. Just pass the packet to ip6_input.
212  */
213 void
214 ip6intr()
215 {
216 	int s;
217 	struct mbuf *m;
218 
219 	for (;;) {
220 		s = splnet();
221 		IF_DEQUEUE(&ip6intrq, m);
222 		splx(s);
223 		if (m == 0)
224 			return;
225 		ip6_input(m);
226 	}
227 }
228 
229 extern struct	route_in6 ip6_forward_rt;
230 
231 void
232 ip6_input(m)
233 	struct mbuf *m;
234 {
235 	struct ip6_hdr *ip6;
236 	int off = sizeof(struct ip6_hdr), nest;
237 	u_int32_t plen;
238 	u_int32_t rtalert = ~0;
239 	int nxt, ours = 0;
240 	struct ifnet *deliverifp = NULL;
241 
242 #ifdef IPSEC
243 	/*
244 	 * should the inner packet be considered authentic?
245 	 * see comment in ah4_input().
246 	 */
247 	m->m_flags &= ~M_AUTHIPHDR;
248 	m->m_flags &= ~M_AUTHIPDGM;
249 #endif
250 
251 	/*
252 	 * mbuf statistics
253 	 */
254 	if (m->m_flags & M_EXT) {
255 		if (m->m_next)
256 			ip6stat.ip6s_mext2m++;
257 		else
258 			ip6stat.ip6s_mext1++;
259 	} else {
260 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
261 		if (m->m_next) {
262 			if (m->m_flags & M_LOOP) {
263 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
264 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
265 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
266 			else
267 				ip6stat.ip6s_m2m[0]++;
268 		} else
269 			ip6stat.ip6s_m1++;
270 #undef M2MMAX
271 	}
272 
273 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
274 	ip6stat.ip6s_total++;
275 
276 #ifndef PULLDOWN_TEST
277 	/* XXX is the line really necessary? */
278 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
279 #endif
280 
281 	if (m->m_len < sizeof(struct ip6_hdr)) {
282 		struct ifnet *inifp;
283 		inifp = m->m_pkthdr.rcvif;
284 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
285 			ip6stat.ip6s_toosmall++;
286 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
287 			return;
288 		}
289 	}
290 
291 	ip6 = mtod(m, struct ip6_hdr *);
292 
293 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
294 		ip6stat.ip6s_badvers++;
295 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
296 		goto bad;
297 	}
298 
299 #ifdef PFIL_HOOKS
300 	/*
301 	 * Run through list of hooks for input packets.  If there are any
302 	 * filters which require that additional packets in the flow are
303 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
304 	 * Note that filters must _never_ set this flag, as another filter
305 	 * in the list may have previously cleared it.
306 	 */
307 	/*
308 	 * let ipfilter look at packet on the wire,
309 	 * not the decapsulated packet.
310 	 */
311 #ifdef IPSEC
312 	if (!ipsec_getnhist(m))
313 #else
314 	if (1)
315 #endif
316 	{
317 		if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif,
318 				   PFIL_IN) != 0)
319 			return;
320 		if (m == NULL)
321 			return;
322 		ip6 = mtod(m, struct ip6_hdr *);
323 	}
324 #endif /* PFIL_HOOKS */
325 
326 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
327 
328 #ifdef ALTQ
329 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
330 		/* packet is dropped by traffic conditioner */
331 		return;
332 	}
333 #endif
334 
335 	/*
336 	 * Check against address spoofing/corruption.
337 	 */
338 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
339 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
340 		ip6stat.ip6s_badscope++;
341 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
342 		goto bad;
343 	}
344 	/*
345 	 * The following check is not documented in specs.  A malicious
346 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
347 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
348 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
349 	 *
350 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
351 	 * support IPv4-less kernel compilation, we cannot support SIIT
352 	 * environment at all.  So, it makes more sense for us to reject any
353 	 * malicious packets for non-SIIT environment, than try to do a
354 	 * partial support for SIIT environment.
355 	 */
356 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
357 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
358 		ip6stat.ip6s_badscope++;
359 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
360 		goto bad;
361 	}
362 #if 0
363 	/*
364 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
365 	 *
366 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
367 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
368 	 * is revised to forbid relaying case.
369 	 */
370 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
371 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
372 		ip6stat.ip6s_badscope++;
373 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
374 		goto bad;
375 	}
376 #endif
377 
378 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
379 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
380 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
381 			ours = 1;
382 			deliverifp = m->m_pkthdr.rcvif;
383 			goto hbhcheck;
384 		} else {
385 			ip6stat.ip6s_badscope++;
386 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
387 			goto bad;
388 		}
389 	}
390 
391 	/* drop packets if interface ID portion is already filled */
392 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
393 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
394 		    ip6->ip6_src.s6_addr16[1]) {
395 			ip6stat.ip6s_badscope++;
396 			goto bad;
397 		}
398 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
399 		    ip6->ip6_dst.s6_addr16[1]) {
400 			ip6stat.ip6s_badscope++;
401 			goto bad;
402 		}
403 	}
404 
405 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
406 		ip6->ip6_src.s6_addr16[1]
407 			= htons(m->m_pkthdr.rcvif->if_index);
408 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
409 		ip6->ip6_dst.s6_addr16[1]
410 			= htons(m->m_pkthdr.rcvif->if_index);
411 
412 	/*
413 	 * We use rt->rt_ifp to determine if the address is ours or not.
414 	 * If rt_ifp is lo0, the address is ours.
415 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
416 	 * so any address under fe80::%lo0/64 will be mistakenly considered
417 	 * local.  The special case is supplied to handle the case properly
418 	 * by actually looking at interface addresses
419 	 * (using in6ifa_ifpwithaddr).
420 	 */
421 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
422 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
423 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
424 			icmp6_error(m, ICMP6_DST_UNREACH,
425 			    ICMP6_DST_UNREACH_ADDR, 0);
426 			/* m is already freed */
427 			return;
428 		}
429 
430 		ours = 1;
431 		deliverifp = m->m_pkthdr.rcvif;
432 		goto hbhcheck;
433 	}
434 
435 	/*
436 	 * Multicast check
437 	 */
438 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
439 	  	struct	in6_multi *in6m = 0;
440 
441 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
442 		/*
443 		 * See if we belong to the destination multicast group on the
444 		 * arrival interface.
445 		 */
446 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
447 		if (in6m)
448 			ours = 1;
449 		else if (!ip6_mrouter) {
450 			ip6stat.ip6s_notmember++;
451 			ip6stat.ip6s_cantforward++;
452 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
453 			goto bad;
454 		}
455 		deliverifp = m->m_pkthdr.rcvif;
456 		goto hbhcheck;
457 	}
458 
459 	/*
460 	 *  Unicast check
461 	 */
462 	if (ip6_forward_rt.ro_rt != NULL &&
463 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
464 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
465 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
466 		ip6stat.ip6s_forward_cachehit++;
467 	else {
468 		struct sockaddr_in6 *dst6;
469 
470 		if (ip6_forward_rt.ro_rt) {
471 			/* route is down or destination is different */
472 			ip6stat.ip6s_forward_cachemiss++;
473 			RTFREE(ip6_forward_rt.ro_rt);
474 			ip6_forward_rt.ro_rt = 0;
475 		}
476 
477 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
478 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
479 		dst6->sin6_len = sizeof(struct sockaddr_in6);
480 		dst6->sin6_family = AF_INET6;
481 		dst6->sin6_addr = ip6->ip6_dst;
482 
483 		rtalloc((struct route *)&ip6_forward_rt);
484 	}
485 
486 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
487 
488 	/*
489 	 * Accept the packet if the forwarding interface to the destination
490 	 * according to the routing table is the loopback interface,
491 	 * unless the associated route has a gateway.
492 	 * Note that this approach causes to accept a packet if there is a
493 	 * route to the loopback interface for the destination of the packet.
494 	 * But we think it's even useful in some situations, e.g. when using
495 	 * a special daemon which wants to intercept the packet.
496 	 */
497 	if (ip6_forward_rt.ro_rt &&
498 	    (ip6_forward_rt.ro_rt->rt_flags &
499 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
500 #if 0
501 	    /*
502 	     * The check below is redundant since the comparison of
503 	     * the destination and the key of the rtentry has
504 	     * already done through looking up the routing table.
505 	     */
506 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
507 			       &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
508 #endif
509 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
510 		struct in6_ifaddr *ia6 =
511 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
512 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
513 			m->m_flags |= M_ANYCAST6;
514 		/*
515 		 * packets to a tentative, duplicated, or somehow invalid
516 		 * address must not be accepted.
517 		 */
518 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
519 			/* this address is ready */
520 			ours = 1;
521 			deliverifp = ia6->ia_ifp;	/* correct? */
522 			goto hbhcheck;
523 		} else {
524 			/* address is not ready, so discard the packet. */
525 			nd6log((LOG_INFO,
526 			    "ip6_input: packet to an unready address %s->%s\n",
527 			    ip6_sprintf(&ip6->ip6_src),
528 			    ip6_sprintf(&ip6->ip6_dst)));
529 
530 			goto bad;
531 		}
532 	}
533 
534 	/*
535 	 * FAITH(Firewall Aided Internet Translator)
536 	 */
537 #if defined(NFAITH) && 0 < NFAITH
538 	if (ip6_keepfaith) {
539 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
540 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
541 			/* XXX do we need more sanity checks? */
542 			ours = 1;
543 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
544 			goto hbhcheck;
545 		}
546 	}
547 #endif
548 
549 #if 0
550     {
551 	/*
552 	 * Last resort: check in6_ifaddr for incoming interface.
553 	 * The code is here until I update the "goto ours hack" code above
554 	 * working right.
555 	 */
556 	struct ifaddr *ifa;
557 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
558 	     ifa;
559 	     ifa = ifa->ifa_list.tqe_next) {
560 		if (ifa->ifa_addr == NULL)
561 			continue;	/* just for safety */
562 		if (ifa->ifa_addr->sa_family != AF_INET6)
563 			continue;
564 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
565 			ours = 1;
566 			deliverifp = ifa->ifa_ifp;
567 			goto hbhcheck;
568 		}
569 	}
570     }
571 #endif
572 
573 	/*
574 	 * Now there is no reason to process the packet if it's not our own
575 	 * and we're not a router.
576 	 */
577 	if (!ip6_forwarding) {
578 		ip6stat.ip6s_cantforward++;
579 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
580 		goto bad;
581 	}
582 
583   hbhcheck:
584 	/*
585 	 * Process Hop-by-Hop options header if it's contained.
586 	 * m may be modified in ip6_hopopts_input().
587 	 * If a JumboPayload option is included, plen will also be modified.
588 	 */
589 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
590 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
591 		struct ip6_hbh *hbh;
592 
593 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
594 #if 0	/*touches NULL pointer*/
595 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
596 #endif
597 			return;	/* m have already been freed */
598 		}
599 
600 		/* adjust pointer */
601 		ip6 = mtod(m, struct ip6_hdr *);
602 
603 		/*
604 		 * if the payload length field is 0 and the next header field
605 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
606 		 * option MUST be included.
607 		 */
608 		if (ip6->ip6_plen == 0 && plen == 0) {
609 			/*
610 			 * Note that if a valid jumbo payload option is
611 			 * contained, ip6_hoptops_input() must set a valid
612 			 * (non-zero) payload length to the variable plen.
613 			 */
614 			ip6stat.ip6s_badoptions++;
615 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
616 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
617 			icmp6_error(m, ICMP6_PARAM_PROB,
618 				    ICMP6_PARAMPROB_HEADER,
619 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
620 			return;
621 		}
622 #ifndef PULLDOWN_TEST
623 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
624 		hbh = (struct ip6_hbh *)(ip6 + 1);
625 #else
626 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
627 			sizeof(struct ip6_hbh));
628 		if (hbh == NULL) {
629 			ip6stat.ip6s_tooshort++;
630 			return;
631 		}
632 #endif
633 		nxt = hbh->ip6h_nxt;
634 
635 		/*
636 		 * accept the packet if a router alert option is included
637 		 * and we act as an IPv6 router.
638 		 */
639 		if (rtalert != ~0 && ip6_forwarding)
640 			ours = 1;
641 	} else
642 		nxt = ip6->ip6_nxt;
643 
644 	/*
645 	 * Check that the amount of data in the buffers
646 	 * is as at least much as the IPv6 header would have us expect.
647 	 * Trim mbufs if longer than we expect.
648 	 * Drop packet if shorter than we expect.
649 	 */
650 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
651 		ip6stat.ip6s_tooshort++;
652 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
653 		goto bad;
654 	}
655 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
656 		if (m->m_len == m->m_pkthdr.len) {
657 			m->m_len = sizeof(struct ip6_hdr) + plen;
658 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
659 		} else
660 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
661 	}
662 
663 	/*
664 	 * Forward if desirable.
665 	 */
666 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
667 		/*
668 		 * If we are acting as a multicast router, all
669 		 * incoming multicast packets are passed to the
670 		 * kernel-level multicast forwarding function.
671 		 * The packet is returned (relatively) intact; if
672 		 * ip6_mforward() returns a non-zero value, the packet
673 		 * must be discarded, else it may be accepted below.
674 		 */
675 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
676 			ip6stat.ip6s_cantforward++;
677 			m_freem(m);
678 			return;
679 		}
680 		if (!ours) {
681 			m_freem(m);
682 			return;
683 		}
684 	} else if (!ours) {
685 		ip6_forward(m, 0);
686 		return;
687 	}
688 
689 	ip6 = mtod(m, struct ip6_hdr *);
690 
691 	/*
692 	 * Malicious party may be able to use IPv4 mapped addr to confuse
693 	 * tcp/udp stack and bypass security checks (act as if it was from
694 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
695 	 *
696 	 * For SIIT end node behavior, you may want to disable the check.
697 	 * However, you will  become vulnerable to attacks using IPv4 mapped
698 	 * source.
699 	 */
700 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
701 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
702 		ip6stat.ip6s_badscope++;
703 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
704 		goto bad;
705 	}
706 
707 	/*
708 	 * Tell launch routine the next header
709 	 */
710 #ifdef IFA_STATS
711 	if (deliverifp != NULL) {
712 		struct in6_ifaddr *ia6;
713 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
714 		if (ia6)
715 			ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
716 	}
717 #endif
718 	ip6stat.ip6s_delivered++;
719 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
720 	nest = 0;
721 
722 	while (nxt != IPPROTO_DONE) {
723 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
724 			ip6stat.ip6s_toomanyhdr++;
725 			goto bad;
726 		}
727 
728 		/*
729 		 * protection against faulty packet - there should be
730 		 * more sanity checks in header chain processing.
731 		 */
732 		if (m->m_pkthdr.len < off) {
733 			ip6stat.ip6s_tooshort++;
734 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
735 			goto bad;
736 		}
737 
738 #ifdef IPSEC
739 		/*
740 		 * enforce IPsec policy checking if we are seeing last header.
741 		 * note that we do not visit this with protocols with pcb layer
742 		 * code - like udp/tcp/raw ip.
743 		 */
744 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
745 		    ipsec6_in_reject(m, NULL)) {
746 			ipsec6stat.in_polvio++;
747 			goto bad;
748 		}
749 #endif
750 
751 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
752 	}
753 	return;
754  bad:
755 	m_freem(m);
756 }
757 
758 /*
759  * Hop-by-Hop options header processing. If a valid jumbo payload option is
760  * included, the real payload length will be stored in plenp.
761  */
762 static int
763 ip6_hopopts_input(plenp, rtalertp, mp, offp)
764 	u_int32_t *plenp;
765 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
766 	struct mbuf **mp;
767 	int *offp;
768 {
769 	struct mbuf *m = *mp;
770 	int off = *offp, hbhlen;
771 	struct ip6_hbh *hbh;
772 	u_int8_t *opt;
773 
774 	/* validation of the length of the header */
775 #ifndef PULLDOWN_TEST
776 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
777 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
778 	hbhlen = (hbh->ip6h_len + 1) << 3;
779 
780 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
781 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
782 #else
783 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
784 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
785 	if (hbh == NULL) {
786 		ip6stat.ip6s_tooshort++;
787 		return -1;
788 	}
789 	hbhlen = (hbh->ip6h_len + 1) << 3;
790 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
791 		hbhlen);
792 	if (hbh == NULL) {
793 		ip6stat.ip6s_tooshort++;
794 		return -1;
795 	}
796 #endif
797 	off += hbhlen;
798 	hbhlen -= sizeof(struct ip6_hbh);
799 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
800 
801 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
802 				hbhlen, rtalertp, plenp) < 0)
803 		return(-1);
804 
805 	*offp = off;
806 	*mp = m;
807 	return(0);
808 }
809 
810 /*
811  * Search header for all Hop-by-hop options and process each option.
812  * This function is separate from ip6_hopopts_input() in order to
813  * handle a case where the sending node itself process its hop-by-hop
814  * options header. In such a case, the function is called from ip6_output().
815  */
816 int
817 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
818 	struct mbuf *m;
819 	u_int8_t *opthead;
820 	int hbhlen;
821 	u_int32_t *rtalertp;
822 	u_int32_t *plenp;
823 {
824 	struct ip6_hdr *ip6;
825 	int optlen = 0;
826 	u_int8_t *opt = opthead;
827 	u_int16_t rtalert_val;
828 	u_int32_t jumboplen;
829 
830 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
831 		switch (*opt) {
832 		case IP6OPT_PAD1:
833 			optlen = 1;
834 			break;
835 		case IP6OPT_PADN:
836 			if (hbhlen < IP6OPT_MINLEN) {
837 				ip6stat.ip6s_toosmall++;
838 				goto bad;
839 			}
840 			optlen = *(opt + 1) + 2;
841 			break;
842 		case IP6OPT_RTALERT:
843 			/* XXX may need check for alignment */
844 			if (hbhlen < IP6OPT_RTALERT_LEN) {
845 				ip6stat.ip6s_toosmall++;
846 				goto bad;
847 			}
848 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
849 				 /* XXX: should we discard the packet? */
850 				log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
851 				    *(opt + 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: should we discard the packet? */
865 				log(LOG_ERR, "length of jumbopayload opt "
866 				    "is inconsistent(%d)\n",
867 				    *(opt + 1));
868 			}
869 			optlen = IP6OPT_JUMBO_LEN;
870 
871 			/*
872 			 * IPv6 packets that have non 0 payload length
873 			 * must not contain a jumbo payload option.
874 			 */
875 			ip6 = mtod(m, struct ip6_hdr *);
876 			if (ip6->ip6_plen) {
877 				ip6stat.ip6s_badoptions++;
878 				icmp6_error(m, ICMP6_PARAM_PROB,
879 					    ICMP6_PARAMPROB_HEADER,
880 					    sizeof(struct ip6_hdr) +
881 					    sizeof(struct ip6_hbh) +
882 					    opt - opthead);
883 				return(-1);
884 			}
885 
886 			/*
887 			 * We may see jumbolen in unaligned location, so
888 			 * we'd need to perform bcopy().
889 			 */
890 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
891 			jumboplen = (u_int32_t)htonl(jumboplen);
892 
893 #if 1
894 			/*
895 			 * if there are multiple jumbo payload options,
896 			 * *plenp will be non-zero and the packet will be
897 			 * rejected.
898 			 * the behavior may need some debate in ipngwg -
899 			 * multiple options does not make sense, however,
900 			 * there's no explicit mention in specification.
901 			 */
902 			if (*plenp != 0) {
903 				ip6stat.ip6s_badoptions++;
904 				icmp6_error(m, ICMP6_PARAM_PROB,
905 					    ICMP6_PARAMPROB_HEADER,
906 					    sizeof(struct ip6_hdr) +
907 					    sizeof(struct ip6_hbh) +
908 					    opt + 2 - opthead);
909 				return(-1);
910 			}
911 #endif
912 
913 			/*
914 			 * jumbo payload length must be larger than 65535.
915 			 */
916 			if (jumboplen <= IPV6_MAXPACKET) {
917 				ip6stat.ip6s_badoptions++;
918 				icmp6_error(m, ICMP6_PARAM_PROB,
919 					    ICMP6_PARAMPROB_HEADER,
920 					    sizeof(struct ip6_hdr) +
921 					    sizeof(struct ip6_hbh) +
922 					    opt + 2 - opthead);
923 				return(-1);
924 			}
925 			*plenp = jumboplen;
926 
927 			break;
928 		default:		/* unknown option */
929 			if (hbhlen < IP6OPT_MINLEN) {
930 				ip6stat.ip6s_toosmall++;
931 				goto bad;
932 			}
933 			if ((optlen = ip6_unknown_opt(opt, m,
934 						      sizeof(struct ip6_hdr) +
935 						      sizeof(struct ip6_hbh) +
936 						      opt - opthead)) == -1)
937 				return(-1);
938 			optlen += 2;
939 			break;
940 		}
941 	}
942 
943 	return(0);
944 
945   bad:
946 	m_freem(m);
947 	return(-1);
948 }
949 
950 /*
951  * Unknown option processing.
952  * The third argument `off' is the offset from the IPv6 header to the option,
953  * which is necessary if the IPv6 header the and option header and IPv6 header
954  * is not continuous in order to return an ICMPv6 error.
955  */
956 int
957 ip6_unknown_opt(optp, m, off)
958 	u_int8_t *optp;
959 	struct mbuf *m;
960 	int off;
961 {
962 	struct ip6_hdr *ip6;
963 
964 	switch (IP6OPT_TYPE(*optp)) {
965 	case IP6OPT_TYPE_SKIP: /* ignore the option */
966 		return((int)*(optp + 1));
967 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
968 		m_freem(m);
969 		return(-1);
970 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
971 		ip6stat.ip6s_badoptions++;
972 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
973 		return(-1);
974 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
975 		ip6stat.ip6s_badoptions++;
976 		ip6 = mtod(m, struct ip6_hdr *);
977 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
978 		    (m->m_flags & (M_BCAST|M_MCAST)))
979 			m_freem(m);
980 		else
981 			icmp6_error(m, ICMP6_PARAM_PROB,
982 				    ICMP6_PARAMPROB_OPTION, off);
983 		return(-1);
984 	}
985 
986 	m_freem(m);		/* XXX: NOTREACHED */
987 	return(-1);
988 }
989 
990 /*
991  * Create the "control" list for this pcb.
992  *
993  * The routine will be called from upper layer handlers like tcp6_input().
994  * Thus the routine assumes that the caller (tcp6_input) have already
995  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
996  * very first mbuf on the mbuf chain.
997  * We may want to add some infinite loop prevention or sanity checks for safety.
998  * (This applies only when you are using KAME mbuf chain restriction, i.e.
999  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1000  */
1001 void
1002 ip6_savecontrol(in6p, mp, ip6, m)
1003 	struct in6pcb *in6p;
1004 	struct mbuf **mp;
1005 	struct ip6_hdr *ip6;
1006 	struct mbuf *m;
1007 {
1008 	struct proc *p = curproc;	/* XXX */
1009 	int privileged;
1010 
1011 	privileged = 0;
1012 	if (p && !suser(p->p_ucred, &p->p_acflag))
1013 		privileged++;
1014 
1015 #ifdef SO_TIMESTAMP
1016 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1017 		struct timeval tv;
1018 
1019 		microtime(&tv);
1020 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1021 			SCM_TIMESTAMP, SOL_SOCKET);
1022 		if (*mp)
1023 			mp = &(*mp)->m_next;
1024 	}
1025 #endif
1026 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
1027 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
1028 			sizeof(struct in6_addr), IPV6_RECVDSTADDR,
1029 			IPPROTO_IPV6);
1030 		if (*mp)
1031 			mp = &(*mp)->m_next;
1032 	}
1033 
1034 #ifdef noyet
1035 	/* options were tossed above */
1036 	if (in6p->in6p_flags & IN6P_RECVOPTS)
1037 		/* broken */
1038 	/* ip6_srcroute doesn't do what we want here, need to fix */
1039 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
1040 		/* broken */
1041 #endif
1042 
1043 	/* RFC 2292 sec. 5 */
1044 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1045 		struct in6_pktinfo pi6;
1046 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1047 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1048 			pi6.ipi6_addr.s6_addr16[1] = 0;
1049 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1050 					? m->m_pkthdr.rcvif->if_index
1051 					: 0;
1052 		*mp = sbcreatecontrol((caddr_t) &pi6,
1053 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
1054 			IPPROTO_IPV6);
1055 		if (*mp)
1056 			mp = &(*mp)->m_next;
1057 	}
1058 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1059 		int hlim = ip6->ip6_hlim & 0xff;
1060 		*mp = sbcreatecontrol((caddr_t) &hlim,
1061 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
1062 		if (*mp)
1063 			mp = &(*mp)->m_next;
1064 	}
1065 	/* IN6P_NEXTHOP - for outgoing packet only */
1066 
1067 	/*
1068 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1069 	 * for the option, but it might be too strict, since there might
1070 	 * be some hop-by-hop options which can be returned to normal user.
1071 	 * See RFC 2292 section 6.
1072 	 */
1073 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
1074 		/*
1075 		 * Check if a hop-by-hop options header is contatined in the
1076 		 * received packet, and if so, store the options as ancillary
1077 		 * data. Note that a hop-by-hop options header must be
1078 		 * just after the IPv6 header, which fact is assured through
1079 		 * the IPv6 input processing.
1080 		 */
1081 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1082 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1083 			struct ip6_hbh *hbh;
1084 			int hbhlen;
1085 
1086 #ifndef PULLDOWN_TEST
1087 			hbh = (struct ip6_hbh *)(ip6 + 1);
1088 			hbhlen = (hbh->ip6h_len + 1) << 3;
1089 #else
1090 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1091 				sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1092 			if (hbh == NULL) {
1093 				ip6stat.ip6s_tooshort++;
1094 				return;
1095 			}
1096 			hbhlen = (hbh->ip6h_len + 1) << 3;
1097 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1098 				sizeof(struct ip6_hdr), hbhlen);
1099 			if (hbh == NULL) {
1100 				ip6stat.ip6s_tooshort++;
1101 				return;
1102 			}
1103 #endif
1104 
1105 			/*
1106 			 * XXX: We copy whole the header even if a jumbo
1107 			 * payload option is included, which option is to
1108 			 * be removed before returning in the RFC 2292.
1109 			 * But it's too painful operation...
1110 			 */
1111 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1112 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1113 			if (*mp)
1114 				mp = &(*mp)->m_next;
1115 		}
1116 	}
1117 
1118 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1119 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1120 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1121 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);;
1122 
1123 		/*
1124 		 * Search for destination options headers or routing
1125 		 * header(s) through the header chain, and stores each
1126 		 * header as ancillary data.
1127 		 * Note that the order of the headers remains in
1128 		 * the chain of ancillary data.
1129 		 */
1130 		while (1) {	/* is explicit loop prevention necessary? */
1131 			struct ip6_ext *ip6e;
1132 			int elen;
1133 
1134 #ifndef PULLDOWN_TEST
1135 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1136 			if (nxt == IPPROTO_AH)
1137 				elen = (ip6e->ip6e_len + 2) << 2;
1138 			else
1139 				elen = (ip6e->ip6e_len + 1) << 3;
1140 #else
1141 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off,
1142 				sizeof(struct ip6_ext));
1143 			if (ip6e == NULL) {
1144 				ip6stat.ip6s_tooshort++;
1145 				return;
1146 			}
1147 			if (nxt == IPPROTO_AH)
1148 				elen = (ip6e->ip6e_len + 2) << 2;
1149 			else
1150 				elen = (ip6e->ip6e_len + 1) << 3;
1151 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off, elen);
1152 			if (ip6e == NULL) {
1153 				ip6stat.ip6s_tooshort++;
1154 				return;
1155 			}
1156 #endif
1157 
1158 			switch (nxt) {
1159 		        case IPPROTO_DSTOPTS:
1160 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1161 					break;
1162 
1163 				/*
1164 				 * We also require super-user privilege for
1165 				 * the option.
1166 				 * See the comments on IN6_HOPOPTS.
1167 				 */
1168 				if (!privileged)
1169 					break;
1170 
1171 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1172 						      IPV6_DSTOPTS,
1173 						      IPPROTO_IPV6);
1174 				if (*mp)
1175 					mp = &(*mp)->m_next;
1176 				break;
1177 
1178 			case IPPROTO_ROUTING:
1179 				if (!in6p->in6p_flags & IN6P_RTHDR)
1180 					break;
1181 
1182 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1183 						      IPV6_RTHDR,
1184 						      IPPROTO_IPV6);
1185 				if (*mp)
1186 					mp = &(*mp)->m_next;
1187 				break;
1188 
1189 			case IPPROTO_UDP:
1190 			case IPPROTO_TCP:
1191 			case IPPROTO_ICMPV6:
1192 			default:
1193 				/*
1194 				 * stop search if we encounter an upper
1195 				 * layer protocol headers.
1196 				 */
1197 				goto loopend;
1198 
1199 			case IPPROTO_HOPOPTS:
1200 			case IPPROTO_AH: /* is it possible? */
1201 				break;
1202 			}
1203 
1204 			/* proceed with the next header. */
1205 			off += elen;
1206 			nxt = ip6e->ip6e_nxt;
1207 		}
1208 	  loopend:
1209 		;
1210 	}
1211 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1212 		/* to be done */
1213 	}
1214 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1215 		/* to be done */
1216 	}
1217 	/* IN6P_RTHDR - to be done */
1218 
1219 }
1220 
1221 /*
1222  * Get pointer to the previous header followed by the header
1223  * currently processed.
1224  * XXX: This function supposes that
1225  *	M includes all headers,
1226  *	the next header field and the header length field of each header
1227  *	are valid, and
1228  *	the sum of each header length equals to OFF.
1229  * Because of these assumptions, this function must be called very
1230  * carefully. Moreover, it will not be used in the near future when
1231  * we develop `neater' mechanism to process extension headers.
1232  */
1233 char *
1234 ip6_get_prevhdr(m, off)
1235 	struct mbuf *m;
1236 	int off;
1237 {
1238 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1239 
1240 	if (off == sizeof(struct ip6_hdr))
1241 		return(&ip6->ip6_nxt);
1242 	else {
1243 		int len, nxt;
1244 		struct ip6_ext *ip6e = NULL;
1245 
1246 		nxt = ip6->ip6_nxt;
1247 		len = sizeof(struct ip6_hdr);
1248 		while (len < off) {
1249 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1250 
1251 			switch (nxt) {
1252 			case IPPROTO_FRAGMENT:
1253 				len += sizeof(struct ip6_frag);
1254 				break;
1255 			case IPPROTO_AH:
1256 				len += (ip6e->ip6e_len + 2) << 2;
1257 				break;
1258 			default:
1259 				len += (ip6e->ip6e_len + 1) << 3;
1260 				break;
1261 			}
1262 			nxt = ip6e->ip6e_nxt;
1263 		}
1264 		if (ip6e)
1265 			return(&ip6e->ip6e_nxt);
1266 		else
1267 			return NULL;
1268 	}
1269 }
1270 
1271 /*
1272  * get next header offset.  m will be retained.
1273  */
1274 int
1275 ip6_nexthdr(m, off, proto, nxtp)
1276 	struct mbuf *m;
1277 	int off;
1278 	int proto;
1279 	int *nxtp;
1280 {
1281 	struct ip6_hdr ip6;
1282 	struct ip6_ext ip6e;
1283 	struct ip6_frag fh;
1284 
1285 	/* just in case */
1286 	if (m == NULL)
1287 		panic("ip6_nexthdr: m == NULL");
1288 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1289 		return -1;
1290 
1291 	switch (proto) {
1292 	case IPPROTO_IPV6:
1293 		if (m->m_pkthdr.len < off + sizeof(ip6))
1294 			return -1;
1295 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1296 		if (nxtp)
1297 			*nxtp = ip6.ip6_nxt;
1298 		off += sizeof(ip6);
1299 		return off;
1300 
1301 	case IPPROTO_FRAGMENT:
1302 		/*
1303 		 * terminate parsing if it is not the first fragment,
1304 		 * it does not make sense to parse through it.
1305 		 */
1306 		if (m->m_pkthdr.len < off + sizeof(fh))
1307 			return -1;
1308 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1309 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1310 			return -1;
1311 		if (nxtp)
1312 			*nxtp = fh.ip6f_nxt;
1313 		off += sizeof(struct ip6_frag);
1314 		return off;
1315 
1316 	case IPPROTO_AH:
1317 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1318 			return -1;
1319 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1320 		if (nxtp)
1321 			*nxtp = ip6e.ip6e_nxt;
1322 		off += (ip6e.ip6e_len + 2) << 2;
1323 		if (m->m_pkthdr.len < off)
1324 			return -1;
1325 		return off;
1326 
1327 	case IPPROTO_HOPOPTS:
1328 	case IPPROTO_ROUTING:
1329 	case IPPROTO_DSTOPTS:
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 + 1) << 3;
1336 		if (m->m_pkthdr.len < off)
1337 			return -1;
1338 		return off;
1339 
1340 	case IPPROTO_NONE:
1341 	case IPPROTO_ESP:
1342 	case IPPROTO_IPCOMP:
1343 		/* give up */
1344 		return -1;
1345 
1346 	default:
1347 		return -1;
1348 	}
1349 
1350 	return -1;
1351 }
1352 
1353 /*
1354  * get offset for the last header in the chain.  m will be kept untainted.
1355  */
1356 int
1357 ip6_lasthdr(m, off, proto, nxtp)
1358 	struct mbuf *m;
1359 	int off;
1360 	int proto;
1361 	int *nxtp;
1362 {
1363 	int newoff;
1364 	int nxt;
1365 
1366 	if (!nxtp) {
1367 		nxt = -1;
1368 		nxtp = &nxt;
1369 	}
1370 	while (1) {
1371 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1372 		if (newoff < 0)
1373 			return off;
1374 		else if (newoff < off)
1375 			return -1;	/* invalid */
1376 		else if (newoff == off)
1377 			return newoff;
1378 
1379 		off = newoff;
1380 		proto = *nxtp;
1381 	}
1382 }
1383 
1384 /*
1385  * System control for IP6
1386  */
1387 
1388 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1389 	0,		0,		0,		0,
1390 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1391 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1392 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1393 	0,		0,		0,		0,
1394 	ENOPROTOOPT
1395 };
1396 
1397 int
1398 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1399 	int *name;
1400 	u_int namelen;
1401 	void *oldp;
1402 	size_t *oldlenp;
1403 	void *newp;
1404 	size_t newlen;
1405 {
1406 	int old, error;
1407 
1408 	/* All sysctl names at this level are terminal. */
1409 	if (namelen != 1)
1410 		return ENOTDIR;
1411 
1412 	switch (name[0]) {
1413 
1414 	case IPV6CTL_FORWARDING:
1415 		return sysctl_int(oldp, oldlenp, newp, newlen,
1416 				  &ip6_forwarding);
1417 	case IPV6CTL_SENDREDIRECTS:
1418 		return sysctl_int(oldp, oldlenp, newp, newlen,
1419 				&ip6_sendredirects);
1420 	case IPV6CTL_DEFHLIM:
1421 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1422 	case IPV6CTL_MAXFRAGPACKETS:
1423 		return sysctl_int(oldp, oldlenp, newp, newlen,
1424 				&ip6_maxfragpackets);
1425 	case IPV6CTL_ACCEPT_RTADV:
1426 		return sysctl_int(oldp, oldlenp, newp, newlen,
1427 				&ip6_accept_rtadv);
1428 	case IPV6CTL_KEEPFAITH:
1429 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1430 	case IPV6CTL_LOG_INTERVAL:
1431 		return sysctl_int(oldp, oldlenp, newp, newlen,
1432 				&ip6_log_interval);
1433 	case IPV6CTL_HDRNESTLIMIT:
1434 		return sysctl_int(oldp, oldlenp, newp, newlen,
1435 				&ip6_hdrnestlimit);
1436 	case IPV6CTL_DAD_COUNT:
1437 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1438 	case IPV6CTL_AUTO_FLOWLABEL:
1439 		return sysctl_int(oldp, oldlenp, newp, newlen,
1440 				&ip6_auto_flowlabel);
1441 	case IPV6CTL_DEFMCASTHLIM:
1442 		return sysctl_int(oldp, oldlenp, newp, newlen,
1443 				&ip6_defmcasthlim);
1444 #if NGIF > 0
1445 	case IPV6CTL_GIF_HLIM:
1446 		return sysctl_int(oldp, oldlenp, newp, newlen,
1447 				&ip6_gif_hlim);
1448 #endif
1449 	case IPV6CTL_KAME_VERSION:
1450 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1451 	case IPV6CTL_USE_DEPRECATED:
1452 		return sysctl_int(oldp, oldlenp, newp, newlen,
1453 				&ip6_use_deprecated);
1454 	case IPV6CTL_RR_PRUNE:
1455 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1456 	case IPV6CTL_V6ONLY:
1457 #ifdef INET6_BINDV6ONLY
1458 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1459 #else
1460 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_v6only);
1461 #endif
1462 	case IPV6CTL_ANONPORTMIN:
1463 		old = ip6_anonportmin;
1464 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1465 		    &ip6_anonportmin);
1466 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmin < 0 ||
1467 		    ip6_anonportmin > 65535
1468 #ifndef IPNOPRIVPORTS
1469 		    || ip6_anonportmin < IPV6PORT_RESERVED
1470 #endif
1471 		    ) {
1472 			ip6_anonportmin = old;
1473 			return (EINVAL);
1474 		}
1475 		return (error);
1476 	case IPV6CTL_ANONPORTMAX:
1477 		old = ip6_anonportmax;
1478 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1479 		    &ip6_anonportmax);
1480 		if (ip6_anonportmin >= ip6_anonportmax || ip6_anonportmax < 0 ||
1481 		    ip6_anonportmax > 65535
1482 #ifndef IPNOPRIVPORTS
1483 		    || ip6_anonportmax < IPV6PORT_RESERVED
1484 #endif
1485 		    ) {
1486 			ip6_anonportmax = old;
1487 			return (EINVAL);
1488 		}
1489 		return (error);
1490 #ifndef IPNOPRIVPORTS
1491 	case IPV6CTL_LOWPORTMIN:
1492 		old = ip6_lowportmin;
1493 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1494 		    &ip6_lowportmin);
1495 		if (ip6_lowportmin >= ip6_lowportmax ||
1496 		    ip6_lowportmin > IPV6PORT_RESERVEDMAX ||
1497 		    ip6_lowportmin < IPV6PORT_RESERVEDMIN) {
1498 			ip6_lowportmin = old;
1499 			return (EINVAL);
1500 		}
1501 		return (error);
1502 	case IPV6CTL_LOWPORTMAX:
1503 		old = ip6_lowportmax;
1504 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1505 		    &ip6_lowportmax);
1506 		if (ip6_lowportmin >= ip6_lowportmax ||
1507 		    ip6_lowportmax > IPV6PORT_RESERVEDMAX ||
1508 		    ip6_lowportmax < IPV6PORT_RESERVEDMIN) {
1509 			ip6_lowportmax = old;
1510 			return (EINVAL);
1511 		}
1512 		return (error);
1513 #endif
1514 	case IPV6CTL_MAXFRAGS:
1515 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_maxfrags);
1516 	default:
1517 		return EOPNOTSUPP;
1518 	}
1519 	/* NOTREACHED */
1520 }
1521