xref: /openbsd-src/sys/netinet6/ip6_input.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*	$OpenBSD: ip6_input.c,v 1.32 2001/06/27 05:50:07 kjc 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/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/domain.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/errno.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/proc.h>
81 
82 #include <net/if.h>
83 #include <net/if_types.h>
84 #include <net/if_dl.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 
91 #ifdef INET
92 #include <netinet/ip.h>
93 #include <netinet/ip_icmp.h>
94 #endif /*INET*/
95 
96 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet6/ip6_var.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/nd6.h>
103 #include <netinet6/in6_prefix.h>
104 
105 #include <netinet6/ip6protosw.h>
106 
107 #include "faith.h"
108 #include "gif.h"
109 #include "bpfilter.h"
110 
111 extern struct domain inet6domain;
112 extern struct ip6protosw inet6sw[];
113 
114 u_char ip6_protox[IPPROTO_MAX];
115 static int ip6qmaxlen = IFQ_MAXLEN;
116 struct in6_ifaddr *in6_ifaddr;
117 struct ifqueue ip6intrq;
118 
119 int ip6_forward_srcrt;			/* XXX */
120 int ip6_sourcecheck;			/* XXX */
121 int ip6_sourcecheck_interval;		/* XXX */
122 
123 struct ip6stat ip6stat;
124 
125 static void ip6_init2 __P((void *));
126 
127 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
128 #ifdef PULLDOWN_TEST
129 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
130 #endif
131 
132 /*
133  * IP6 initialization: fill in IP6 protocol switch table.
134  * All protocols not implemented in kernel go to raw IP6 protocol handler.
135  */
136 void
137 ip6_init()
138 {
139 	struct ip6protosw *pr;
140 	int i;
141 
142 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
143 	if (pr == 0)
144 		panic("ip6_init");
145 	for (i = 0; i < IPPROTO_MAX; i++)
146 		ip6_protox[i] = pr - inet6sw;
147 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
148 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
149 		if (pr->pr_domain->dom_family == PF_INET6 &&
150 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
151 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
152 	ip6intrq.ifq_maxlen = ip6qmaxlen;
153 	nd6_init();
154 	frag6_init();
155 	ip6_flow_seq = arc4random();
156 	ip6_init2((void *)0);
157 }
158 
159 static void
160 ip6_init2(dummy)
161 	void *dummy;
162 {
163 
164 #if 1
165 	/*
166 	 * to route local address of p2p link to loopback,
167 	 * assign loopback address first.
168 	 */
169 	in6_ifattach(lo0ifp, NULL);
170 #else
171 	/* you MUST bring lo0 up manually, in rc script. */
172 #endif
173 
174 	/* nd6_timer_init */
175 	bzero(&nd6_timer_ch, sizeof(nd6_timer_ch));
176 	timeout_set(&nd6_timer_ch, nd6_timer, NULL);
177 	timeout_add(&nd6_timer_ch, hz);
178 
179 	/* router renumbering prefix list maintenance */
180 	bzero(&in6_rr_timer_ch, sizeof(in6_rr_timer_ch));
181 	timeout_set(&in6_rr_timer_ch, in6_rr_timer, (caddr_t)0);
182 	timeout_add(&in6_rr_timer_ch, hz);
183 }
184 
185 /*
186  * IP6 input interrupt handling. Just pass the packet to ip6_input.
187  */
188 void
189 ip6intr()
190 {
191 	int s;
192 	struct mbuf *m;
193 
194 	for (;;) {
195 		s = splimp();
196 		IF_DEQUEUE(&ip6intrq, m);
197 		splx(s);
198 		if (m == 0)
199 			return;
200 		ip6_input(m);
201 	}
202 }
203 
204 extern struct	route_in6 ip6_forward_rt;
205 
206 void
207 ip6_input(m)
208 	struct mbuf *m;
209 {
210 	struct ip6_hdr *ip6;
211 	int off = sizeof(struct ip6_hdr), nest;
212 	u_int32_t plen;
213 	u_int32_t rtalert = ~0;
214 	int nxt, ours = 0;
215 	struct ifnet *deliverifp = NULL;
216 
217 #if 0 /* IPSEC */
218 	/*
219 	 * should the inner packet be considered authentic?
220 	 * see comment in ah4_input().
221 	 */
222 	if (m) {
223 		m->m_flags &= ~M_AUTHIPHDR;
224 		m->m_flags &= ~M_AUTHIPDGM;
225 	}
226 #endif
227 
228 	/*
229 	 * mbuf statistics by kazu
230 	 */
231 	if (m->m_flags & M_EXT) {
232 		if (m->m_next)
233 			ip6stat.ip6s_mext2m++;
234 		else
235 			ip6stat.ip6s_mext1++;
236 	} else {
237 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
238 		if (m->m_next) {
239 			if (m->m_flags & M_LOOP) {
240 				ip6stat.ip6s_m2m[lo0ifp->if_index]++;	/*XXX*/
241 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
242 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
243 			else
244 				ip6stat.ip6s_m2m[0]++;
245 		} else
246 			ip6stat.ip6s_m1++;
247 #undef M2MMAX
248 	}
249 
250 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
251 	ip6stat.ip6s_total++;
252 
253 #ifndef PULLDOWN_TEST
254 	/* XXX is the line really necessary? */
255 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
256 #endif
257 
258 	if (m->m_len < sizeof(struct ip6_hdr)) {
259 		struct ifnet *inifp;
260 		inifp = m->m_pkthdr.rcvif;
261 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
262 			ip6stat.ip6s_toosmall++;
263 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
264 			return;
265 		}
266 	}
267 
268 	ip6 = mtod(m, struct ip6_hdr *);
269 
270 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
271 		ip6stat.ip6s_badvers++;
272 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
273 		goto bad;
274 	}
275 
276 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
277 
278 #ifdef ALTQ
279 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
280 		/* packet is dropped by traffic conditioner */
281 		return;
282 	}
283 #endif
284 
285 	/*
286 	 * Scope check
287 	 */
288 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
289 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
290 		ip6stat.ip6s_badscope++;
291 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
292 		goto bad;
293 	}
294 	/*
295 	 * The following check is not documented in the spec.  Malicious party
296 	 * may be able to use IPv4 mapped addr to confuse tcp/udp stack and
297 	 * bypass security checks (act as if it was from 127.0.0.1 by using
298 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
299 	 *
300 	 * This check chokes if we are in SIIT cloud.  As none of BSDs support
301 	 * IPv4-less kernel compilation, we cannot support SIIT environment
302 	 * at all.  So, it makes more sense for us to reject any malicious
303 	 * packets for non-SIIT environment, than try to do a partical support
304 	 * for SIIT environment.
305 	 */
306 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
307 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
308 		ip6stat.ip6s_badscope++;
309 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
310 		goto bad;
311 	}
312 #if 0
313 	/*
314 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
315 	 *
316 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
317 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
318 	 * is revised to forbid relaying case.
319 	 */
320 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
321 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
322 		ip6stat.ip6s_badscope++;
323 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
324 		goto bad;
325 	}
326 #endif
327 
328 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
329 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
330 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
331 			ours = 1;
332 			deliverifp = m->m_pkthdr.rcvif;
333 			goto hbhcheck;
334 		} else {
335 			ip6stat.ip6s_badscope++;
336 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
337 			goto bad;
338 		}
339 	}
340 
341 	/* drop packets if interface ID portion is already filled */
342 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
343 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
344 		    ip6->ip6_src.s6_addr16[1]) {
345 			ip6stat.ip6s_badscope++;
346 			goto bad;
347 		}
348 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
349 		    ip6->ip6_dst.s6_addr16[1]) {
350 			ip6stat.ip6s_badscope++;
351 			goto bad;
352 		}
353 	}
354 
355 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
356 		ip6->ip6_src.s6_addr16[1]
357 			= htons(m->m_pkthdr.rcvif->if_index);
358 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
359 		ip6->ip6_dst.s6_addr16[1]
360 			= htons(m->m_pkthdr.rcvif->if_index);
361 
362 	/*
363 	 * We use rt->rt_ifp to determine if the address is ours or not.
364 	 * If rt_ifp is lo0, the address is ours.
365 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
366 	 * so any address under fe80::%lo0/64 will be mistakenly considered
367 	 * local.  The special case is supplied to handle the case properly
368 	 * by actually looking at interface addresses
369 	 * (using in6ifa_ifpwithaddr).
370 	 */
371 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
372 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
373 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
374 			icmp6_error(m, ICMP6_DST_UNREACH,
375 			    ICMP6_DST_UNREACH_ADDR, 0);
376 			/* m is already freed */
377 			return;
378 		}
379 
380 		ours = 1;
381 		deliverifp = m->m_pkthdr.rcvif;
382 		goto hbhcheck;
383 	}
384 
385 	/*
386 	 * Multicast check
387 	 */
388 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
389 	  	struct	in6_multi *in6m = 0;
390 
391 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
392 		/*
393 		 * See if we belong to the destination multicast group on the
394 		 * arrival interface.
395 		 */
396 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
397 		if (in6m)
398 			ours = 1;
399 		else if (!ip6_mrouter) {
400 			ip6stat.ip6s_notmember++;
401 			ip6stat.ip6s_cantforward++;
402 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
403 			goto bad;
404 		}
405 		deliverifp = m->m_pkthdr.rcvif;
406 		goto hbhcheck;
407 	}
408 
409 	/*
410 	 *  Unicast check
411 	 */
412 	if (ip6_forward_rt.ro_rt != NULL &&
413 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
414 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
415 			       &ip6_forward_rt.ro_dst.sin6_addr))
416 		ip6stat.ip6s_forward_cachehit++;
417 	else {
418 		if (ip6_forward_rt.ro_rt) {
419 			/* route is down or destination is different */
420 			ip6stat.ip6s_forward_cachemiss++;
421 			RTFREE(ip6_forward_rt.ro_rt);
422 			ip6_forward_rt.ro_rt = 0;
423 		}
424 
425 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
426 		ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
427 		ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
428 		ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
429 
430 		rtalloc((struct route *)&ip6_forward_rt);
431 	}
432 
433 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
434 
435 	/*
436 	 * Accept the packet if the forwarding interface to the destination
437 	 * according to the routing table is the loopback interface,
438 	 * unless the associated route has a gateway.
439 	 * Note that this approach causes to accept a packet if there is a
440 	 * route to the loopback interface for the destination of the packet.
441 	 * But we think it's even useful in some situations, e.g. when using
442 	 * a special daemon which wants to intercept the packet.
443 	 */
444 	if (ip6_forward_rt.ro_rt &&
445 	    (ip6_forward_rt.ro_rt->rt_flags &
446 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
447 #if 0
448 	    /*
449 	     * The check below is redundant since the comparison of
450 	     * the destination and the key of the rtentry has
451 	     * already done through looking up the routing table.
452 	     */
453 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
454 			       &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
455 #endif
456 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
457 		struct in6_ifaddr *ia6 =
458 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
459 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
460 			m->m_flags |= M_ANYCAST6;
461 		/*
462 		 * packets to a tentative, duplicated, or somehow invalid
463 		 * address must not be accepted.
464 		 */
465 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
466 			/* this address is ready */
467 			ours = 1;
468 			deliverifp = ia6->ia_ifp;	/* correct? */
469 			goto hbhcheck;
470 		} else {
471 			/* address is not ready, so discard the packet. */
472 			nd6log((LOG_INFO,
473 			    "ip6_input: packet to an unready address %s->%s\n",
474 			    ip6_sprintf(&ip6->ip6_src),
475 			    ip6_sprintf(&ip6->ip6_dst)));
476 
477 			goto bad;
478 		}
479 	}
480 
481 	/*
482 	 * FAITH(Firewall Aided Internet Translator)
483 	 */
484 #if defined(NFAITH) && 0 < NFAITH
485 	if (ip6_keepfaith) {
486 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
487 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
488 			/* XXX do we need more sanity checks? */
489 			ours = 1;
490 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
491 			goto hbhcheck;
492 		}
493 	}
494 #endif
495 
496 #if 0
497     {
498 	/*
499 	 * Last resort: check in6_ifaddr for incoming interface.
500 	 * The code is here until I update the "goto ours hack" code above
501 	 * working right.
502 	 */
503 	struct ifaddr *ifa;
504 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
505 	     ifa;
506 	     ifa = ifa->ifa_list.tqe_next) {
507 		if (ifa->ifa_addr == NULL)
508 			continue;	/* just for safety */
509 		if (ifa->ifa_addr->sa_family != AF_INET6)
510 			continue;
511 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
512 			ours = 1;
513 			deliverifp = ifa->ifa_ifp;
514 			goto hbhcheck;
515 		}
516 	}
517     }
518 #endif
519 
520 	/*
521 	 * Now there is no reason to process the packet if it's not our own
522 	 * and we're not a router.
523 	 */
524 	if (!ip6_forwarding) {
525 		ip6stat.ip6s_cantforward++;
526 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
527 		goto bad;
528 	}
529 
530   hbhcheck:
531 	/*
532 	 * Process Hop-by-Hop options header if it's contained.
533 	 * m may be modified in ip6_hopopts_input().
534 	 * If a JumboPayload option is included, plen will also be modified.
535 	 */
536 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
537 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
538 		struct ip6_hbh *hbh;
539 
540 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
541 #if 0	/*touches NULL pointer*/
542 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
543 #endif
544 			return;	/* m have already been freed */
545 		}
546 
547 		/* adjust pointer */
548 		ip6 = mtod(m, struct ip6_hdr *);
549 
550 		/*
551 		 * if the payload length field is 0 and the next header field
552 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
553 		 * option MUST be included.
554 		 */
555 		if (ip6->ip6_plen == 0 && plen == 0) {
556 			/*
557 			 * Note that if a valid jumbo payload option is
558 			 * contained, ip6_hoptops_input() must set a valid
559 			 * (non-zero) payload length to the variable plen.
560 			 */
561 			ip6stat.ip6s_badoptions++;
562 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
563 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
564 			icmp6_error(m, ICMP6_PARAM_PROB,
565 				    ICMP6_PARAMPROB_HEADER,
566 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
567 			return;
568 		}
569 #ifndef PULLDOWN_TEST
570 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
571 		hbh = (struct ip6_hbh *)(ip6 + 1);
572 #else
573 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
574 			sizeof(struct ip6_hbh));
575 		if (hbh == NULL) {
576 			ip6stat.ip6s_tooshort++;
577 			return;
578 		}
579 #endif
580 		nxt = hbh->ip6h_nxt;
581 
582 		/*
583 		 * accept the packet if a router alert option is included
584 		 * and we act as an IPv6 router.
585 		 */
586 		if (rtalert != ~0 && ip6_forwarding)
587 			ours = 1;
588 	} else
589 		nxt = ip6->ip6_nxt;
590 
591 	/*
592 	 * Check that the amount of data in the buffers
593 	 * is as at least much as the IPv6 header would have us expect.
594 	 * Trim mbufs if longer than we expect.
595 	 * Drop packet if shorter than we expect.
596 	 */
597 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
598 		ip6stat.ip6s_tooshort++;
599 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
600 		goto bad;
601 	}
602 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
603 		if (m->m_len == m->m_pkthdr.len) {
604 			m->m_len = sizeof(struct ip6_hdr) + plen;
605 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
606 		} else
607 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
608 	}
609 
610 	/*
611 	 * Forward if desirable.
612 	 */
613 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
614 		/*
615 		 * If we are acting as a multicast router, all
616 		 * incoming multicast packets are passed to the
617 		 * kernel-level multicast forwarding function.
618 		 * The packet is returned (relatively) intact; if
619 		 * ip6_mforward() returns a non-zero value, the packet
620 		 * must be discarded, else it may be accepted below.
621 		 */
622 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
623 			ip6stat.ip6s_cantforward++;
624 			m_freem(m);
625 			return;
626 		}
627 		if (!ours) {
628 			m_freem(m);
629 			return;
630 		}
631 	} else if (!ours) {
632 		ip6_forward(m, 0);
633 		return;
634 	}
635 
636 	ip6 = mtod(m, struct ip6_hdr *);
637 
638 	/*
639 	 * Malicious party may be able to use IPv4 mapped addr to confuse
640 	 * tcp/udp stack and bypass security checks (act as if it was from
641 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
642 	 *
643 	 * For SIIT end node behavior, you may want to disable the check.
644 	 * However, you will  become vulnerable to attacks using IPv4 mapped
645 	 * source.
646 	 */
647 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
648 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
649 		ip6stat.ip6s_badscope++;
650 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
651 		goto bad;
652 	}
653 
654 	/*
655 	 * Tell launch routine the next header
656 	 */
657 	ip6stat.ip6s_delivered++;
658 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
659 	nest = 0;
660 
661 	while (nxt != IPPROTO_DONE) {
662 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
663 			ip6stat.ip6s_toomanyhdr++;
664 			goto bad;
665 		}
666 
667 		/*
668 		 * protection against faulty packet - there should be
669 		 * more sanity checks in header chain processing.
670 		 */
671 		if (m->m_pkthdr.len < off) {
672 			ip6stat.ip6s_tooshort++;
673 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
674 			goto bad;
675 		}
676 
677 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
678 	}
679 	return;
680  bad:
681 	m_freem(m);
682 }
683 
684 /*
685  * Hop-by-Hop options header processing. If a valid jumbo payload option is
686  * included, the real payload length will be stored in plenp.
687  */
688 static int
689 ip6_hopopts_input(plenp, rtalertp, mp, offp)
690 	u_int32_t *plenp;
691 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
692 	struct mbuf **mp;
693 	int *offp;
694 {
695 	struct mbuf *m = *mp;
696 	int off = *offp, hbhlen;
697 	struct ip6_hbh *hbh;
698 	u_int8_t *opt;
699 
700 	/* validation of the length of the header */
701 #ifndef PULLDOWN_TEST
702 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
703 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
704 	hbhlen = (hbh->ip6h_len + 1) << 3;
705 
706 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
707 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
708 #else
709 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
710 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
711 	if (hbh == NULL) {
712 		ip6stat.ip6s_tooshort++;
713 		return -1;
714 	}
715 	hbhlen = (hbh->ip6h_len + 1) << 3;
716 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
717 		hbhlen);
718 	if (hbh == NULL) {
719 		ip6stat.ip6s_tooshort++;
720 		return -1;
721 	}
722 #endif
723 	off += hbhlen;
724 	hbhlen -= sizeof(struct ip6_hbh);
725 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
726 
727 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
728 				hbhlen, rtalertp, plenp) < 0)
729 		return(-1);
730 
731 	*offp = off;
732 	*mp = m;
733 	return(0);
734 }
735 
736 /*
737  * Search header for all Hop-by-hop options and process each option.
738  * This function is separate from ip6_hopopts_input() in order to
739  * handle a case where the sending node itself process its hop-by-hop
740  * options header. In such a case, the function is called from ip6_output().
741  */
742 int
743 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
744 	struct mbuf *m;
745 	u_int8_t *opthead;
746 	int hbhlen;
747 	u_int32_t *rtalertp;
748 	u_int32_t *plenp;
749 {
750 	struct ip6_hdr *ip6;
751 	int optlen = 0;
752 	u_int8_t *opt = opthead;
753 	u_int16_t rtalert_val;
754 	u_int32_t jumboplen;
755 
756 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
757 		switch (*opt) {
758 		case IP6OPT_PAD1:
759 			optlen = 1;
760 			break;
761 		case IP6OPT_PADN:
762 			if (hbhlen < IP6OPT_MINLEN) {
763 				ip6stat.ip6s_toosmall++;
764 				goto bad;
765 			}
766 			optlen = *(opt + 1) + 2;
767 			break;
768 		case IP6OPT_RTALERT:
769 			/* XXX may need check for alignment */
770 			if (hbhlen < IP6OPT_RTALERT_LEN) {
771 				ip6stat.ip6s_toosmall++;
772 				goto bad;
773 			}
774 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2)
775 				/* XXX: should we discard the packet? */
776 				log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
777 				    *(opt + 1));
778 			optlen = IP6OPT_RTALERT_LEN;
779 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
780 			*rtalertp = ntohs(rtalert_val);
781 			break;
782 		case IP6OPT_JUMBO:
783 			/* XXX may need check for alignment */
784 			if (hbhlen < IP6OPT_JUMBO_LEN) {
785 				ip6stat.ip6s_toosmall++;
786 				goto bad;
787 			}
788 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2)
789 				/* XXX: should we discard the packet? */
790 				log(LOG_ERR, "length of jumbopayload opt "
791 				    "is inconsistent(%d)\n",
792 				    *(opt + 1));
793 			optlen = IP6OPT_JUMBO_LEN;
794 
795 			/*
796 			 * IPv6 packets that have non 0 payload length
797 			 * must not contain a jumbo payload option.
798 			 */
799 			ip6 = mtod(m, struct ip6_hdr *);
800 			if (ip6->ip6_plen) {
801 				ip6stat.ip6s_badoptions++;
802 				icmp6_error(m, ICMP6_PARAM_PROB,
803 					    ICMP6_PARAMPROB_HEADER,
804 					    sizeof(struct ip6_hdr) +
805 					    sizeof(struct ip6_hbh) +
806 					    opt - opthead);
807 				return(-1);
808 			}
809 
810 			/*
811 			 * We may see jumbolen in unaligned location, so
812 			 * we'd need to perform bcopy().
813 			 */
814 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
815 			jumboplen = (u_int32_t)htonl(jumboplen);
816 
817 #if 1
818 			/*
819 			 * if there are multiple jumbo payload options,
820 			 * *plenp will be non-zero and the packet will be
821 			 * rejected.
822 			 * the behavior may need some debate in ipngwg -
823 			 * multiple options does not make sense, however,
824 			 * there's no explicit mention in specification.
825 			 */
826 			if (*plenp != 0) {
827 				ip6stat.ip6s_badoptions++;
828 				icmp6_error(m, ICMP6_PARAM_PROB,
829 					    ICMP6_PARAMPROB_HEADER,
830 					    sizeof(struct ip6_hdr) +
831 					    sizeof(struct ip6_hbh) +
832 					    opt + 2 - opthead);
833 				return(-1);
834 			}
835 #endif
836 
837 			/*
838 			 * jumbo payload length must be larger than 65535.
839 			 */
840 			if (jumboplen <= IPV6_MAXPACKET) {
841 				ip6stat.ip6s_badoptions++;
842 				icmp6_error(m, ICMP6_PARAM_PROB,
843 					    ICMP6_PARAMPROB_HEADER,
844 					    sizeof(struct ip6_hdr) +
845 					    sizeof(struct ip6_hbh) +
846 					    opt + 2 - opthead);
847 				return(-1);
848 			}
849 			*plenp = jumboplen;
850 
851 			break;
852 		default:		/* unknown option */
853 			if (hbhlen < IP6OPT_MINLEN) {
854 				ip6stat.ip6s_toosmall++;
855 				goto bad;
856 			}
857 			if ((optlen = ip6_unknown_opt(opt, m,
858 						      sizeof(struct ip6_hdr) +
859 						      sizeof(struct ip6_hbh) +
860 						      opt - opthead)) == -1)
861 				return(-1);
862 			optlen += 2;
863 			break;
864 		}
865 	}
866 
867 	return(0);
868 
869   bad:
870 	m_freem(m);
871 	return(-1);
872 }
873 
874 /*
875  * Unknown option processing.
876  * The third argument `off' is the offset from the IPv6 header to the option,
877  * which is necessary if the IPv6 header the and option header and IPv6 header
878  * is not continuous in order to return an ICMPv6 error.
879  */
880 int
881 ip6_unknown_opt(optp, m, off)
882 	u_int8_t *optp;
883 	struct mbuf *m;
884 	int off;
885 {
886 	struct ip6_hdr *ip6;
887 
888 	switch (IP6OPT_TYPE(*optp)) {
889 	case IP6OPT_TYPE_SKIP: /* ignore the option */
890 		return((int)*(optp + 1));
891 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
892 		m_freem(m);
893 		return(-1);
894 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
895 		ip6stat.ip6s_badoptions++;
896 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
897 		return(-1);
898 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
899 		ip6stat.ip6s_badoptions++;
900 		ip6 = mtod(m, struct ip6_hdr *);
901 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
902 		    (m->m_flags & (M_BCAST|M_MCAST)))
903 			m_freem(m);
904 		else
905 			icmp6_error(m, ICMP6_PARAM_PROB,
906 				    ICMP6_PARAMPROB_OPTION, off);
907 		return(-1);
908 	}
909 
910 	m_freem(m);		/* XXX: NOTREACHED */
911 	return(-1);
912 }
913 
914 /*
915  * Create the "control" list for this pcb.
916  *
917  * The routine will be called from upper layer handlers like tcp6_input().
918  * Thus the routine assumes that the caller (tcp6_input) have already
919  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
920  * very first mbuf on the mbuf chain.
921  * We may want to add some infinite loop prevention or sanity checks for safety.
922  * (This applies only when you are using KAME mbuf chain restriction, i.e.
923  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
924  */
925 void
926 ip6_savecontrol(in6p, mp, ip6, m)
927 	struct inpcb *in6p;
928 	struct mbuf **mp;
929 	struct ip6_hdr *ip6;
930 	struct mbuf *m;
931 {
932 # define in6p_flags	inp_flags
933 	int privileged = 0;
934 
935 	if ((in6p->inp_socket->so_state & SS_PRIV) != 0)
936 		privileged++;
937 
938 #ifdef SO_TIMESTAMP
939 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
940 		struct timeval tv;
941 
942 		microtime(&tv);
943 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
944 			SCM_TIMESTAMP, SOL_SOCKET);
945 		if (*mp)
946 			mp = &(*mp)->m_next;
947 	}
948 #endif
949 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
950 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
951 			sizeof(struct in6_addr), IPV6_RECVDSTADDR,
952 			IPPROTO_IPV6);
953 		if (*mp)
954 			mp = &(*mp)->m_next;
955 	}
956 
957 #ifdef noyet
958 	/* options were tossed above */
959 	if (in6p->in6p_flags & IN6P_RECVOPTS)
960 		/* broken */
961 	/* ip6_srcroute doesn't do what we want here, need to fix */
962 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
963 		/* broken */
964 #endif
965 
966 	/* RFC 2292 sec. 5 */
967 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
968 		struct in6_pktinfo pi6;
969 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
970 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
971 			pi6.ipi6_addr.s6_addr16[1] = 0;
972 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
973 					? m->m_pkthdr.rcvif->if_index
974 					: 0;
975 		*mp = sbcreatecontrol((caddr_t) &pi6,
976 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
977 			IPPROTO_IPV6);
978 		if (*mp)
979 			mp = &(*mp)->m_next;
980 	}
981 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
982 		int hlim = ip6->ip6_hlim & 0xff;
983 		*mp = sbcreatecontrol((caddr_t) &hlim,
984 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
985 		if (*mp)
986 			mp = &(*mp)->m_next;
987 	}
988 	/* IN6P_NEXTHOP - for outgoing packet only */
989 
990 	/*
991 	 * IPV6_HOPOPTS socket option. We require super-user privilege
992 	 * for the option, but it might be too strict, since there might
993 	 * be some hop-by-hop options which can be returned to normal user.
994 	 * See RFC 2292 section 6.
995 	 */
996 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0 && privileged) {
997 		/*
998 		 * Check if a hop-by-hop options header is contatined in the
999 		 * received packet, and if so, store the options as ancillary
1000 		 * data. Note that a hop-by-hop options header must be
1001 		 * just after the IPv6 header, which fact is assured through
1002 		 * the IPv6 input processing.
1003 		 */
1004 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1005 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1006 			struct ip6_hbh *hbh;
1007 			int hbhlen;
1008 #ifdef PULLDOWN_TEST
1009 			struct mbuf *ext;
1010 #endif
1011 
1012 #ifndef PULLDOWN_TEST
1013 			hbh = (struct ip6_hbh *)(ip6 + 1);
1014 			hbhlen = (hbh->ip6h_len + 1) << 3;
1015 #else
1016 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1017 			    ip6->ip6_nxt);
1018 			if (ext == NULL) {
1019 				ip6stat.ip6s_tooshort++;
1020 				return;
1021 			}
1022 			hbh = mtod(ext, struct ip6_hbh *);
1023 			hbhlen = (hbh->ip6h_len + 1) << 3;
1024 			if (hbhlen != ext->m_len) {
1025 				m_freem(ext);
1026 				ip6stat.ip6s_tooshort++;
1027 				return;
1028 			}
1029 #endif
1030 
1031 			/*
1032 			 * XXX: We copy whole the header even if a jumbo
1033 			 * payload option is included, which option is to
1034 			 * be removed before returning in the RFC 2292.
1035 			 * But it's too painful operation...
1036 			 */
1037 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1038 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1039 			if (*mp)
1040 				mp = &(*mp)->m_next;
1041 #ifdef PULLDOWN_TEST
1042 			m_freem(ext);
1043 #endif
1044 		}
1045 	}
1046 
1047 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1048 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1049 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1050 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1051 
1052 		/*
1053 		 * Search for destination options headers or routing
1054 		 * header(s) through the header chain, and stores each
1055 		 * header as ancillary data.
1056 		 * Note that the order of the headers remains in
1057 		 * the chain of ancillary data.
1058 		 */
1059 		while (1) {	/* is explicit loop prevention necessary? */
1060 			struct ip6_ext *ip6e = NULL;
1061 			int elen;
1062 #ifdef PULLDOWN_TEST
1063 			struct mbuf *ext = NULL;
1064 #endif
1065 
1066 			/*
1067 			 * if it is not an extension header, don't try to
1068 			 * pull it from the chain.
1069 			 */
1070 			switch (nxt) {
1071 			case IPPROTO_DSTOPTS:
1072 			case IPPROTO_ROUTING:
1073 			case IPPROTO_HOPOPTS:
1074 			case IPPROTO_AH: /* is it possible? */
1075 				break;
1076 			default:
1077 				goto loopend;
1078 			}
1079 
1080 #ifndef PULLDOWN_TEST
1081 			if (off + sizeof(*ip6e) > m->m_len)
1082 				goto loopend;
1083 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1084 			if (nxt == IPPROTO_AH)
1085 				elen = (ip6e->ip6e_len + 2) << 2;
1086 			else
1087 				elen = (ip6e->ip6e_len + 1) << 3;
1088 			if (off + elen > m->m_len)
1089 				goto loopend;
1090 #else
1091 			ext = ip6_pullexthdr(m, off, nxt);
1092 			if (ext == NULL) {
1093 				ip6stat.ip6s_tooshort++;
1094 				return;
1095 			}
1096 			ip6e = mtod(ext, struct ip6_ext *);
1097 			if (nxt == IPPROTO_AH)
1098 				elen = (ip6e->ip6e_len + 2) << 2;
1099 			else
1100 				elen = (ip6e->ip6e_len + 1) << 3;
1101 			if (elen != ext->m_len) {
1102 				m_freem(ext);
1103 				ip6stat.ip6s_tooshort++;
1104 				return;
1105 			}
1106 #endif
1107 
1108 			switch (nxt) {
1109 			case IPPROTO_DSTOPTS:
1110 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1111 					break;
1112 
1113 				/*
1114 				 * We also require super-user privilege for
1115 				 * the option.
1116 				 * See the comments on IN6_HOPOPTS.
1117 				 */
1118 				if (!privileged)
1119 					break;
1120 
1121 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1122 						      IPV6_DSTOPTS,
1123 						      IPPROTO_IPV6);
1124 				if (*mp)
1125 					mp = &(*mp)->m_next;
1126 				break;
1127 
1128 			case IPPROTO_ROUTING:
1129 				if (!in6p->in6p_flags & IN6P_RTHDR)
1130 					break;
1131 
1132 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1133 						      IPV6_RTHDR,
1134 						      IPPROTO_IPV6);
1135 				if (*mp)
1136 					mp = &(*mp)->m_next;
1137 				break;
1138 
1139 			case IPPROTO_HOPOPTS:
1140 			case IPPROTO_AH: /* is it possible? */
1141 				break;
1142 
1143 			default:
1144 				/*
1145 			 	 * other cases have been filtered in the above.
1146 				 * none will visit this case.  here we supply
1147 				 * the code just in case (nxt overwritten or
1148 				 * other cases).
1149 				 */
1150 #ifdef PULLDOWN_TEST
1151 				m_freem(ext);
1152 #endif
1153 				goto loopend;
1154 
1155 			}
1156 
1157 			/* proceed with the next header. */
1158 			off += elen;
1159 			nxt = ip6e->ip6e_nxt;
1160 			ip6e = NULL;
1161 #ifdef PULLDOWN_TEST
1162 			m_freem(ext);
1163 			ext = NULL;
1164 #endif
1165 		}
1166 	  loopend:
1167 	}
1168 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1169 		/* to be done */
1170 	}
1171 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1172 		/* to be done */
1173 	}
1174 	/* IN6P_RTHDR - to be done */
1175 # undef in6p_flags
1176 }
1177 
1178 #ifdef PULLDOWN_TEST
1179 /*
1180  * pull single extension header from mbuf chain.  returns single mbuf that
1181  * contains the result, or NULL on error.
1182  */
1183 static struct mbuf *
1184 ip6_pullexthdr(m, off, nxt)
1185 	struct mbuf *m;
1186 	size_t off;
1187 	int nxt;
1188 {
1189 	struct ip6_ext ip6e;
1190 	size_t elen;
1191 	struct mbuf *n;
1192 
1193 #ifdef DIAGNOSTIC
1194 	switch (nxt) {
1195 	case IPPROTO_DSTOPTS:
1196 	case IPPROTO_ROUTING:
1197 	case IPPROTO_HOPOPTS:
1198 	case IPPROTO_AH: /* is it possible? */
1199 		break;
1200 	default:
1201 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1202 	}
1203 #endif
1204 
1205 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1206 	if (nxt == IPPROTO_AH)
1207 		elen = (ip6e.ip6e_len + 2) << 2;
1208 	else
1209 		elen = (ip6e.ip6e_len + 1) << 3;
1210 
1211 	MGET(n, M_DONTWAIT, MT_DATA);
1212 	if (n && elen >= MLEN) {
1213 		MCLGET(n, M_DONTWAIT);
1214 		if ((n->m_flags & M_EXT) == 0) {
1215 			m_free(n);
1216 			n = NULL;
1217 		}
1218 	}
1219 	if (!n)
1220 		return NULL;
1221 
1222 	n->m_len = 0;
1223 	if (elen >= M_TRAILINGSPACE(n)) {
1224 		m_free(n);
1225 		return NULL;
1226 	}
1227 
1228 	m_copydata(m, off, elen, mtod(n, caddr_t));
1229 	n->m_len = elen;
1230 	return n;
1231 }
1232 #endif
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 char *
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 		return off;
1337 
1338 	case IPPROTO_HOPOPTS:
1339 	case IPPROTO_ROUTING:
1340 	case IPPROTO_DSTOPTS:
1341 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1342 			return -1;
1343 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1344 		if (nxtp)
1345 			*nxtp = ip6e.ip6e_nxt;
1346 		off += (ip6e.ip6e_len + 1) << 3;
1347 		return off;
1348 
1349 	case IPPROTO_NONE:
1350 	case IPPROTO_ESP:
1351 	case IPPROTO_IPCOMP:
1352 		/* give up */
1353 		return -1;
1354 
1355 	default:
1356 		return -1;
1357 	}
1358 
1359 	return -1;
1360 }
1361 
1362 /*
1363  * get offset for the last header in the chain.  m will be kept untainted.
1364  */
1365 int
1366 ip6_lasthdr(m, off, proto, nxtp)
1367 	struct mbuf *m;
1368 	int off;
1369 	int proto;
1370 	int *nxtp;
1371 {
1372 	int newoff;
1373 	int nxt;
1374 
1375 	if (!nxtp) {
1376 		nxt = -1;
1377 		nxtp = &nxt;
1378 	}
1379 	while (1) {
1380 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1381 		if (newoff < 0)
1382 			return off;
1383 		else if (newoff < off)
1384 			return -1;	/* invalid */
1385 		else if (newoff == off)
1386 			return newoff;
1387 
1388 		off = newoff;
1389 		proto = *nxtp;
1390 	}
1391 }
1392 
1393 /*
1394  * System control for IP6
1395  */
1396 
1397 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1398 	0,		0,		0,		0,
1399 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1400 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1401 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1402 	0,		0,		0,		0,
1403 	ENOPROTOOPT
1404 };
1405 
1406 #include <vm/vm.h>
1407 #include <sys/sysctl.h>
1408 
1409 int
1410 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1411 	int *name;
1412 	u_int namelen;
1413 	void *oldp;
1414 	size_t *oldlenp;
1415 	void *newp;
1416 	size_t newlen;
1417 {
1418 	/* All sysctl names at this level are terminal. */
1419 	if (namelen != 1)
1420 		return ENOTDIR;
1421 
1422 	switch (name[0]) {
1423 
1424 	case IPV6CTL_FORWARDING:
1425 		return sysctl_int(oldp, oldlenp, newp, newlen,
1426 				  &ip6_forwarding);
1427 	case IPV6CTL_SENDREDIRECTS:
1428 		return sysctl_int(oldp, oldlenp, newp, newlen,
1429 				&ip6_sendredirects);
1430 	case IPV6CTL_DEFHLIM:
1431 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
1432 	case IPV6CTL_MAXFRAGPACKETS:
1433 		return sysctl_int(oldp, oldlenp, newp, newlen,
1434 				&ip6_maxfragpackets);
1435 	case IPV6CTL_ACCEPT_RTADV:
1436 		return sysctl_int(oldp, oldlenp, newp, newlen,
1437 				&ip6_accept_rtadv);
1438 	case IPV6CTL_KEEPFAITH:
1439 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
1440 	case IPV6CTL_LOG_INTERVAL:
1441 		return sysctl_int(oldp, oldlenp, newp, newlen,
1442 				&ip6_log_interval);
1443 	case IPV6CTL_HDRNESTLIMIT:
1444 		return sysctl_int(oldp, oldlenp, newp, newlen,
1445 				&ip6_hdrnestlimit);
1446 	case IPV6CTL_DAD_COUNT:
1447 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
1448 	case IPV6CTL_AUTO_FLOWLABEL:
1449 		return sysctl_int(oldp, oldlenp, newp, newlen,
1450 				&ip6_auto_flowlabel);
1451 	case IPV6CTL_DEFMCASTHLIM:
1452 		return sysctl_int(oldp, oldlenp, newp, newlen,
1453 				&ip6_defmcasthlim);
1454 	case IPV6CTL_KAME_VERSION:
1455 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1456 	case IPV6CTL_USE_DEPRECATED:
1457 		return sysctl_int(oldp, oldlenp, newp, newlen,
1458 				&ip6_use_deprecated);
1459 	case IPV6CTL_RR_PRUNE:
1460 		return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
1461 	default:
1462 		return EOPNOTSUPP;
1463 	}
1464 	/* NOTREACHED */
1465 }
1466