1 /*	$OpenBSD: ip6_input.c,v 1.61 2005/03/06 16:27:01 dhartmei 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. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "pf.h"
65 
66 #include <sys/param.h>
67 #include <sys/systm.h>
68 #include <sys/malloc.h>
69 #include <sys/mbuf.h>
70 #include <sys/domain.h>
71 #include <sys/protosw.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/errno.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 #include <sys/proc.h>
79 
80 #include <net/if.h>
81 #include <net/if_types.h>
82 #include <net/if_dl.h>
83 #include <net/route.h>
84 #include <net/netisr.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 
89 #ifdef INET
90 #include <netinet/ip.h>
91 #include <netinet/ip_icmp.h>
92 #endif /*INET*/
93 
94 #include <netinet/in_pcb.h>
95 #include <netinet6/in6_var.h>
96 #include <netinet/ip6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/in6_ifattach.h>
100 #include <netinet6/nd6.h>
101 
102 #include <netinet6/ip6protosw.h>
103 
104 #include "faith.h"
105 #include "gif.h"
106 #include "bpfilter.h"
107 
108 #if NPF > 0
109 #include <net/pfvar.h>
110 #endif
111 
112 extern struct domain inet6domain;
113 extern struct ip6protosw inet6sw[];
114 
115 u_char ip6_protox[IPPROTO_MAX];
116 static int ip6qmaxlen = IFQ_MAXLEN;
117 struct in6_ifaddr *in6_ifaddr;
118 struct ifqueue ip6intrq;
119 
120 int ip6_forward_srcrt;			/* XXX */
121 int ip6_sourcecheck;			/* XXX */
122 int ip6_sourcecheck_interval;		/* XXX */
123 
124 struct ip6stat ip6stat;
125 
126 static void ip6_init2(void *);
127 
128 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
129 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
130 
131 /*
132  * IP6 initialization: fill in IP6 protocol switch table.
133  * All protocols not implemented in kernel go to raw IP6 protocol handler.
134  */
135 void
ip6_init()136 ip6_init()
137 {
138 	struct ip6protosw *pr;
139 	int i;
140 
141 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
142 	if (pr == 0)
143 		panic("ip6_init");
144 	for (i = 0; i < IPPROTO_MAX; i++)
145 		ip6_protox[i] = pr - inet6sw;
146 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
147 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
148 		if (pr->pr_domain->dom_family == PF_INET6 &&
149 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
150 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
151 	ip6intrq.ifq_maxlen = ip6qmaxlen;
152 	nd6_init();
153 	frag6_init();
154 	ip6_init2((void *)0);
155 }
156 
157 static void
ip6_init2(dummy)158 ip6_init2(dummy)
159 	void *dummy;
160 {
161 
162 	/* nd6_timer_init */
163 	bzero(&nd6_timer_ch, sizeof(nd6_timer_ch));
164 	timeout_set(&nd6_timer_ch, nd6_timer, NULL);
165 	timeout_add(&nd6_timer_ch, hz);
166 }
167 
168 /*
169  * IP6 input interrupt handling. Just pass the packet to ip6_input.
170  */
171 void
ip6intr()172 ip6intr()
173 {
174 	int s;
175 	struct mbuf *m;
176 
177 	for (;;) {
178 		s = splimp();
179 		IF_DEQUEUE(&ip6intrq, m);
180 		splx(s);
181 		if (m == 0)
182 			return;
183 		ip6_input(m);
184 	}
185 }
186 
187 extern struct	route_in6 ip6_forward_rt;
188 
189 void
ip6_input(m)190 ip6_input(m)
191 	struct mbuf *m;
192 {
193 	struct ip6_hdr *ip6;
194 	int off = sizeof(struct ip6_hdr), nest;
195 	u_int32_t plen;
196 	u_int32_t rtalert = ~0;
197 	int nxt, ours = 0;
198 	struct ifnet *deliverifp = NULL;
199 #if NPF > 0
200 	struct in6_addr odst;
201 #endif
202 	int srcrt = 0;
203 
204 	/*
205 	 * mbuf statistics by kazu
206 	 */
207 	if (m->m_flags & M_EXT) {
208 		if (m->m_next)
209 			ip6stat.ip6s_mext2m++;
210 		else
211 			ip6stat.ip6s_mext1++;
212 	} else {
213 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
214 		if (m->m_next) {
215 			if (m->m_flags & M_LOOP) {
216 				ip6stat.ip6s_m2m[lo0ifp->if_index]++;	/*XXX*/
217 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
218 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
219 			else
220 				ip6stat.ip6s_m2m[0]++;
221 		} else
222 			ip6stat.ip6s_m1++;
223 #undef M2MMAX
224 	}
225 
226 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
227 	ip6stat.ip6s_total++;
228 
229 	if (m->m_len < sizeof(struct ip6_hdr)) {
230 		struct ifnet *inifp;
231 		inifp = m->m_pkthdr.rcvif;
232 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
233 			ip6stat.ip6s_toosmall++;
234 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
235 			return;
236 		}
237 	}
238 
239 	ip6 = mtod(m, struct ip6_hdr *);
240 
241 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
242 		ip6stat.ip6s_badvers++;
243 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
244 		goto bad;
245 	}
246 
247 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
248 
249 	/*
250 	 * Check against address spoofing/corruption.
251 	 */
252 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
253 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
254 		/*
255 		 * XXX: "badscope" is not very suitable for a multicast source.
256 		 */
257 		ip6stat.ip6s_badscope++;
258 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
259 		goto bad;
260 	}
261 	/*
262 	 * The following check is not documented in specs.  A malicious
263 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
264 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
265 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
266 	 *
267 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
268 	 * support IPv4-less kernel compilation, we cannot support SIIT
269 	 * environment at all.  So, it makes more sense for us to reject any
270 	 * malicious packets for non-SIIT environment, than try to do a
271 	 * partial support for SIIT environment.
272 	 */
273 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
274 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
275 		ip6stat.ip6s_badscope++;
276 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
277 		goto bad;
278 	}
279 #if 0
280 	/*
281 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
282 	 *
283 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
284 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
285 	 * is revised to forbid relaying case.
286 	 */
287 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
288 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
289 		ip6stat.ip6s_badscope++;
290 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
291 		goto bad;
292 	}
293 #endif
294 
295 #if NPF > 0
296         /*
297          * Packet filter
298          */
299 	odst = ip6->ip6_dst;
300 	if (pf_test6(PF_IN, m->m_pkthdr.rcvif, &m) != PF_PASS)
301 		goto bad;
302 	if (m == NULL)
303 		return;
304 
305 	ip6 = mtod(m, struct ip6_hdr *);
306 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
307 #endif
308 
309 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
310 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
311 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
312 			ours = 1;
313 			deliverifp = m->m_pkthdr.rcvif;
314 			goto hbhcheck;
315 		} else {
316 			ip6stat.ip6s_badscope++;
317 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
318 			goto bad;
319 		}
320 	}
321 
322 	/* drop packets if interface ID portion is already filled */
323 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
324 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
325 		    ip6->ip6_src.s6_addr16[1]) {
326 			ip6stat.ip6s_badscope++;
327 			goto bad;
328 		}
329 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
330 		    ip6->ip6_dst.s6_addr16[1]) {
331 			ip6stat.ip6s_badscope++;
332 			goto bad;
333 		}
334 	}
335 
336 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
337 		ip6->ip6_src.s6_addr16[1]
338 			= htons(m->m_pkthdr.rcvif->if_index);
339 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
340 		ip6->ip6_dst.s6_addr16[1]
341 			= htons(m->m_pkthdr.rcvif->if_index);
342 
343 	/*
344 	 * We use rt->rt_ifp to determine if the address is ours or not.
345 	 * If rt_ifp is lo0, the address is ours.
346 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
347 	 * so any address under fe80::%lo0/64 will be mistakenly considered
348 	 * local.  The special case is supplied to handle the case properly
349 	 * by actually looking at interface addresses
350 	 * (using in6ifa_ifpwithaddr).
351 	 */
352 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
353 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
354 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
355 			icmp6_error(m, ICMP6_DST_UNREACH,
356 			    ICMP6_DST_UNREACH_ADDR, 0);
357 			/* m is already freed */
358 			return;
359 		}
360 
361 		ours = 1;
362 		deliverifp = m->m_pkthdr.rcvif;
363 		goto hbhcheck;
364 	}
365 
366 	/*
367 	 * Multicast check
368 	 */
369 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
370 	  	struct	in6_multi *in6m = 0;
371 
372 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
373 		/*
374 		 * See if we belong to the destination multicast group on the
375 		 * arrival interface.
376 		 */
377 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
378 		if (in6m)
379 			ours = 1;
380 		else if (!ip6_mrouter) {
381 			ip6stat.ip6s_notmember++;
382 			ip6stat.ip6s_cantforward++;
383 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
384 			goto bad;
385 		}
386 		deliverifp = m->m_pkthdr.rcvif;
387 		goto hbhcheck;
388 	}
389 
390 	/*
391 	 *  Unicast check
392 	 */
393 	if (ip6_forward_rt.ro_rt != NULL &&
394 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
395 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
396 			       &ip6_forward_rt.ro_dst.sin6_addr))
397 		ip6stat.ip6s_forward_cachehit++;
398 	else {
399 		if (ip6_forward_rt.ro_rt) {
400 			/* route is down or destination is different */
401 			ip6stat.ip6s_forward_cachemiss++;
402 			RTFREE(ip6_forward_rt.ro_rt);
403 			ip6_forward_rt.ro_rt = 0;
404 		}
405 
406 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
407 		ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
408 		ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
409 		ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
410 
411 		rtalloc((struct route *)&ip6_forward_rt);
412 	}
413 
414 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
415 
416 	/*
417 	 * Accept the packet if the forwarding interface to the destination
418 	 * according to the routing table is the loopback interface,
419 	 * unless the associated route has a gateway.
420 	 * Note that this approach causes to accept a packet if there is a
421 	 * route to the loopback interface for the destination of the packet.
422 	 * But we think it's even useful in some situations, e.g. when using
423 	 * a special daemon which wants to intercept the packet.
424 	 */
425 	if (ip6_forward_rt.ro_rt &&
426 	    (ip6_forward_rt.ro_rt->rt_flags &
427 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
428 #if 0
429 	    /*
430 	     * The check below is redundant since the comparison of
431 	     * the destination and the key of the rtentry has
432 	     * already done through looking up the routing table.
433 	     */
434 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
435 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
436 #endif
437 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
438 		struct in6_ifaddr *ia6 =
439 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
440 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
441 			m->m_flags |= M_ANYCAST6;
442 		/*
443 		 * packets to a tentative, duplicated, or somehow invalid
444 		 * address must not be accepted.
445 		 */
446 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
447 			/* this address is ready */
448 			ours = 1;
449 			deliverifp = ia6->ia_ifp;	/* correct? */
450 			goto hbhcheck;
451 		} else {
452 			/* address is not ready, so discard the packet. */
453 			nd6log((LOG_INFO,
454 			    "ip6_input: packet to an unready address %s->%s\n",
455 			    ip6_sprintf(&ip6->ip6_src),
456 			    ip6_sprintf(&ip6->ip6_dst)));
457 
458 			goto bad;
459 		}
460 	}
461 
462 	/*
463 	 * FAITH (Firewall Aided Internet Translator)
464 	 */
465 #if defined(NFAITH) && 0 < NFAITH
466 	if (ip6_keepfaith) {
467 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
468 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
469 			/* XXX do we need more sanity checks? */
470 			ours = 1;
471 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
472 			goto hbhcheck;
473 		}
474 	}
475 #endif
476 
477 #if 0
478     {
479 	/*
480 	 * Last resort: check in6_ifaddr for incoming interface.
481 	 * The code is here until I update the "goto ours hack" code above
482 	 * working right.
483 	 */
484 	struct ifaddr *ifa;
485 	for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
486 	     ifa;
487 	     ifa = ifa->ifa_list.tqe_next) {
488 		if (ifa->ifa_addr == NULL)
489 			continue;	/* just for safety */
490 		if (ifa->ifa_addr->sa_family != AF_INET6)
491 			continue;
492 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
493 			ours = 1;
494 			deliverifp = ifa->ifa_ifp;
495 			goto hbhcheck;
496 		}
497 	}
498     }
499 #endif
500 
501 	/*
502 	 * Now there is no reason to process the packet if it's not our own
503 	 * and we're not a router.
504 	 */
505 	if (!ip6_forwarding) {
506 		ip6stat.ip6s_cantforward++;
507 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
508 		goto bad;
509 	}
510 
511   hbhcheck:
512 	/*
513 	 * Process Hop-by-Hop options header if it's contained.
514 	 * m may be modified in ip6_hopopts_input().
515 	 * If a JumboPayload option is included, plen will also be modified.
516 	 */
517 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
518 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
519 		struct ip6_hbh *hbh;
520 
521 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
522 #if 0	/*touches NULL pointer*/
523 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
524 #endif
525 			return;	/* m have already been freed */
526 		}
527 
528 		/* adjust pointer */
529 		ip6 = mtod(m, struct ip6_hdr *);
530 
531 		/*
532 		 * if the payload length field is 0 and the next header field
533 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
534 		 * option MUST be included.
535 		 */
536 		if (ip6->ip6_plen == 0 && plen == 0) {
537 			/*
538 			 * Note that if a valid jumbo payload option is
539 			 * contained, ip6_hoptops_input() must set a valid
540 			 * (non-zero) payload length to the variable plen.
541 			 */
542 			ip6stat.ip6s_badoptions++;
543 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
544 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
545 			icmp6_error(m, ICMP6_PARAM_PROB,
546 				    ICMP6_PARAMPROB_HEADER,
547 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
548 			return;
549 		}
550 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
551 			sizeof(struct ip6_hbh));
552 		if (hbh == NULL) {
553 			ip6stat.ip6s_tooshort++;
554 			return;
555 		}
556 		nxt = hbh->ip6h_nxt;
557 
558 		/*
559 		 * accept the packet if a router alert option is included
560 		 * and we act as an IPv6 router.
561 		 */
562 		if (rtalert != ~0 && ip6_forwarding)
563 			ours = 1;
564 	} else
565 		nxt = ip6->ip6_nxt;
566 
567 	/*
568 	 * Check that the amount of data in the buffers
569 	 * is as at least much as the IPv6 header would have us expect.
570 	 * Trim mbufs if longer than we expect.
571 	 * Drop packet if shorter than we expect.
572 	 */
573 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
574 		ip6stat.ip6s_tooshort++;
575 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
576 		goto bad;
577 	}
578 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
579 		if (m->m_len == m->m_pkthdr.len) {
580 			m->m_len = sizeof(struct ip6_hdr) + plen;
581 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
582 		} else
583 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
584 	}
585 
586 	/*
587 	 * Forward if desirable.
588 	 */
589 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
590 		/*
591 		 * If we are acting as a multicast router, all
592 		 * incoming multicast packets are passed to the
593 		 * kernel-level multicast forwarding function.
594 		 * The packet is returned (relatively) intact; if
595 		 * ip6_mforward() returns a non-zero value, the packet
596 		 * must be discarded, else it may be accepted below.
597 		 */
598 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
599 			ip6stat.ip6s_cantforward++;
600 			m_freem(m);
601 			return;
602 		}
603 		if (!ours) {
604 			m_freem(m);
605 			return;
606 		}
607 	} else if (!ours) {
608 		ip6_forward(m, srcrt);
609 		return;
610 	}
611 
612 	ip6 = mtod(m, struct ip6_hdr *);
613 
614 	/*
615 	 * Malicious party may be able to use IPv4 mapped addr to confuse
616 	 * tcp/udp stack and bypass security checks (act as if it was from
617 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
618 	 *
619 	 * For SIIT end node behavior, you may want to disable the check.
620 	 * However, you will  become vulnerable to attacks using IPv4 mapped
621 	 * source.
622 	 */
623 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
624 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
625 		ip6stat.ip6s_badscope++;
626 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
627 		goto bad;
628 	}
629 
630 	/*
631 	 * Tell launch routine the next header
632 	 */
633 	ip6stat.ip6s_delivered++;
634 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
635 	nest = 0;
636 
637 	while (nxt != IPPROTO_DONE) {
638 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
639 			ip6stat.ip6s_toomanyhdr++;
640 			goto bad;
641 		}
642 
643 		/*
644 		 * protection against faulty packet - there should be
645 		 * more sanity checks in header chain processing.
646 		 */
647 		if (m->m_pkthdr.len < off) {
648 			ip6stat.ip6s_tooshort++;
649 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
650 			goto bad;
651 		}
652 
653 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
654 	}
655 	return;
656  bad:
657 	m_freem(m);
658 }
659 
660 /*
661  * Hop-by-Hop options header processing. If a valid jumbo payload option is
662  * included, the real payload length will be stored in plenp.
663  */
664 static int
ip6_hopopts_input(plenp,rtalertp,mp,offp)665 ip6_hopopts_input(plenp, rtalertp, mp, offp)
666 	u_int32_t *plenp;
667 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
668 	struct mbuf **mp;
669 	int *offp;
670 {
671 	struct mbuf *m = *mp;
672 	int off = *offp, hbhlen;
673 	struct ip6_hbh *hbh;
674 	u_int8_t *opt;
675 
676 	/* validation of the length of the header */
677 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
678 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
679 	if (hbh == NULL) {
680 		ip6stat.ip6s_tooshort++;
681 		return -1;
682 	}
683 	hbhlen = (hbh->ip6h_len + 1) << 3;
684 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
685 		hbhlen);
686 	if (hbh == NULL) {
687 		ip6stat.ip6s_tooshort++;
688 		return -1;
689 	}
690 	off += hbhlen;
691 	hbhlen -= sizeof(struct ip6_hbh);
692 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
693 
694 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
695 				hbhlen, rtalertp, plenp) < 0)
696 		return (-1);
697 
698 	*offp = off;
699 	*mp = m;
700 	return (0);
701 }
702 
703 /*
704  * Search header for all Hop-by-hop options and process each option.
705  * This function is separate from ip6_hopopts_input() in order to
706  * handle a case where the sending node itself process its hop-by-hop
707  * options header. In such a case, the function is called from ip6_output().
708  *
709  * The function assumes that hbh header is located right after the IPv6 header
710  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
711  * opthead + hbhlen is located in continuous memory region.
712  */
713 int
ip6_process_hopopts(m,opthead,hbhlen,rtalertp,plenp)714 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
715 	struct mbuf *m;
716 	u_int8_t *opthead;
717 	int hbhlen;
718 	u_int32_t *rtalertp;
719 	u_int32_t *plenp;
720 {
721 	struct ip6_hdr *ip6;
722 	int optlen = 0;
723 	u_int8_t *opt = opthead;
724 	u_int16_t rtalert_val;
725 	u_int32_t jumboplen;
726 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
727 
728 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
729 		switch (*opt) {
730 		case IP6OPT_PAD1:
731 			optlen = 1;
732 			break;
733 		case IP6OPT_PADN:
734 			if (hbhlen < IP6OPT_MINLEN) {
735 				ip6stat.ip6s_toosmall++;
736 				goto bad;
737 			}
738 			optlen = *(opt + 1) + 2;
739 			break;
740 		case IP6OPT_RTALERT:
741 			/* XXX may need check for alignment */
742 			if (hbhlen < IP6OPT_RTALERT_LEN) {
743 				ip6stat.ip6s_toosmall++;
744 				goto bad;
745 			}
746 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
747 				/* XXX stat */
748 				icmp6_error(m, ICMP6_PARAM_PROB,
749 				    ICMP6_PARAMPROB_HEADER,
750 				    erroff + opt + 1 - opthead);
751 				return (-1);
752 			}
753 			optlen = IP6OPT_RTALERT_LEN;
754 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
755 			*rtalertp = ntohs(rtalert_val);
756 			break;
757 		case IP6OPT_JUMBO:
758 			/* XXX may need check for alignment */
759 			if (hbhlen < IP6OPT_JUMBO_LEN) {
760 				ip6stat.ip6s_toosmall++;
761 				goto bad;
762 			}
763 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
764 				/* XXX stat */
765 				icmp6_error(m, ICMP6_PARAM_PROB,
766 				    ICMP6_PARAMPROB_HEADER,
767 				    erroff + opt + 1 - opthead);
768 				return (-1);
769 			}
770 			optlen = IP6OPT_JUMBO_LEN;
771 
772 			/*
773 			 * IPv6 packets that have non 0 payload length
774 			 * must not contain a jumbo payload option.
775 			 */
776 			ip6 = mtod(m, struct ip6_hdr *);
777 			if (ip6->ip6_plen) {
778 				ip6stat.ip6s_badoptions++;
779 				icmp6_error(m, ICMP6_PARAM_PROB,
780 				    ICMP6_PARAMPROB_HEADER,
781 				    erroff + opt - opthead);
782 				return (-1);
783 			}
784 
785 			/*
786 			 * We may see jumbolen in unaligned location, so
787 			 * we'd need to perform bcopy().
788 			 */
789 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
790 			jumboplen = (u_int32_t)htonl(jumboplen);
791 
792 #if 1
793 			/*
794 			 * if there are multiple jumbo payload options,
795 			 * *plenp will be non-zero and the packet will be
796 			 * rejected.
797 			 * the behavior may need some debate in ipngwg -
798 			 * multiple options does not make sense, however,
799 			 * there's no explicit mention in specification.
800 			 */
801 			if (*plenp != 0) {
802 				ip6stat.ip6s_badoptions++;
803 				icmp6_error(m, ICMP6_PARAM_PROB,
804 				    ICMP6_PARAMPROB_HEADER,
805 				    erroff + opt + 2 - opthead);
806 				return (-1);
807 			}
808 #endif
809 
810 			/*
811 			 * jumbo payload length must be larger than 65535.
812 			 */
813 			if (jumboplen <= IPV6_MAXPACKET) {
814 				ip6stat.ip6s_badoptions++;
815 				icmp6_error(m, ICMP6_PARAM_PROB,
816 				    ICMP6_PARAMPROB_HEADER,
817 				    erroff + opt + 2 - opthead);
818 				return (-1);
819 			}
820 			*plenp = jumboplen;
821 
822 			break;
823 		default:		/* unknown option */
824 			if (hbhlen < IP6OPT_MINLEN) {
825 				ip6stat.ip6s_toosmall++;
826 				goto bad;
827 			}
828 			optlen = ip6_unknown_opt(opt, m,
829 			    erroff + opt - opthead);
830 			if (optlen == -1)
831 				return (-1);
832 			optlen += 2;
833 			break;
834 		}
835 	}
836 
837 	return (0);
838 
839   bad:
840 	m_freem(m);
841 	return (-1);
842 }
843 
844 /*
845  * Unknown option processing.
846  * The third argument `off' is the offset from the IPv6 header to the option,
847  * which is necessary if the IPv6 header the and option header and IPv6 header
848  * is not continuous in order to return an ICMPv6 error.
849  */
850 int
ip6_unknown_opt(optp,m,off)851 ip6_unknown_opt(optp, m, off)
852 	u_int8_t *optp;
853 	struct mbuf *m;
854 	int off;
855 {
856 	struct ip6_hdr *ip6;
857 
858 	switch (IP6OPT_TYPE(*optp)) {
859 	case IP6OPT_TYPE_SKIP: /* ignore the option */
860 		return ((int)*(optp + 1));
861 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
862 		m_freem(m);
863 		return (-1);
864 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
865 		ip6stat.ip6s_badoptions++;
866 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
867 		return (-1);
868 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
869 		ip6stat.ip6s_badoptions++;
870 		ip6 = mtod(m, struct ip6_hdr *);
871 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
872 		    (m->m_flags & (M_BCAST|M_MCAST)))
873 			m_freem(m);
874 		else
875 			icmp6_error(m, ICMP6_PARAM_PROB,
876 				    ICMP6_PARAMPROB_OPTION, off);
877 		return (-1);
878 	}
879 
880 	m_freem(m);		/* XXX: NOTREACHED */
881 	return (-1);
882 }
883 
884 /*
885  * Create the "control" list for this pcb.
886  *
887  * The routine will be called from upper layer handlers like tcp6_input().
888  * Thus the routine assumes that the caller (tcp6_input) have already
889  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
890  * very first mbuf on the mbuf chain.
891  * We may want to add some infinite loop prevention or sanity checks for safety.
892  * (This applies only when you are using KAME mbuf chain restriction, i.e.
893  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
894  */
895 void
ip6_savecontrol(in6p,mp,ip6,m)896 ip6_savecontrol(in6p, mp, ip6, m)
897 	struct inpcb *in6p;
898 	struct mbuf **mp;
899 	struct ip6_hdr *ip6;
900 	struct mbuf *m;
901 {
902 # define in6p_flags	inp_flags
903 
904 #ifdef SO_TIMESTAMP
905 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
906 		struct timeval tv;
907 
908 		microtime(&tv);
909 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
910 		    SCM_TIMESTAMP, SOL_SOCKET);
911 		if (*mp)
912 			mp = &(*mp)->m_next;
913 	}
914 #endif
915 	if (in6p->in6p_flags & IN6P_RECVDSTADDR) {
916 		*mp = sbcreatecontrol((caddr_t) &ip6->ip6_dst,
917 		    sizeof(struct in6_addr), IPV6_RECVDSTADDR, IPPROTO_IPV6);
918 		if (*mp)
919 			mp = &(*mp)->m_next;
920 	}
921 
922 #ifdef noyet
923 	/* options were tossed above */
924 	if (in6p->in6p_flags & IN6P_RECVOPTS)
925 		/* broken */
926 	/* ip6_srcroute doesn't do what we want here, need to fix */
927 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
928 		/* broken */
929 #endif
930 
931 	/* RFC 2292 sec. 5 */
932 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
933 		struct in6_pktinfo pi6;
934 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
935 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
936 			pi6.ipi6_addr.s6_addr16[1] = 0;
937 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
938 					? m->m_pkthdr.rcvif->if_index
939 					: 0;
940 		*mp = sbcreatecontrol((caddr_t) &pi6,
941 		    sizeof(struct in6_pktinfo), IPV6_PKTINFO, IPPROTO_IPV6);
942 		if (*mp)
943 			mp = &(*mp)->m_next;
944 	}
945 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
946 		int hlim = ip6->ip6_hlim & 0xff;
947 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
948 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
949 		if (*mp)
950 			mp = &(*mp)->m_next;
951 	}
952 	/* IN6P_NEXTHOP - for outgoing packet only */
953 
954 	/*
955 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
956 	 * privilege for the option (see ip6_ctloutput), but it might be too
957 	 * strict, since there might be some hop-by-hop options which can be
958 	 * returned to normal user.
959 	 * See also RFC 2292 section 6.
960 	 */
961 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
962 		/*
963 		 * Check if a hop-by-hop options header is contatined in the
964 		 * received packet, and if so, store the options as ancillary
965 		 * data. Note that a hop-by-hop options header must be
966 		 * just after the IPv6 header, which fact is assured through
967 		 * the IPv6 input processing.
968 		 */
969 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
970 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
971 			struct ip6_hbh *hbh;
972 			int hbhlen;
973 			struct mbuf *ext;
974 
975 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
976 			    ip6->ip6_nxt);
977 			if (ext == NULL) {
978 				ip6stat.ip6s_tooshort++;
979 				return;
980 			}
981 			hbh = mtod(ext, struct ip6_hbh *);
982 			hbhlen = (hbh->ip6h_len + 1) << 3;
983 			if (hbhlen != ext->m_len) {
984 				m_freem(ext);
985 				ip6stat.ip6s_tooshort++;
986 				return;
987 			}
988 
989 			/*
990 			 * XXX: We copy whole the header even if a jumbo
991 			 * payload option is included, which option is to
992 			 * be removed before returning in the RFC 2292.
993 			 * But it's too painful operation...
994 			 */
995 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
996 			    IPV6_HOPOPTS, IPPROTO_IPV6);
997 			if (*mp)
998 				mp = &(*mp)->m_next;
999 			m_freem(ext);
1000 		}
1001 	}
1002 
1003 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1004 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1005 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1006 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1007 
1008 		/*
1009 		 * Search for destination options headers or routing
1010 		 * header(s) through the header chain, and stores each
1011 		 * header as ancillary data.
1012 		 * Note that the order of the headers remains in
1013 		 * the chain of ancillary data.
1014 		 */
1015 		while (1) {	/* is explicit loop prevention necessary? */
1016 			struct ip6_ext *ip6e = NULL;
1017 			int elen;
1018 			struct mbuf *ext = NULL;
1019 
1020 			/*
1021 			 * if it is not an extension header, don't try to
1022 			 * pull it from the chain.
1023 			 */
1024 			switch (nxt) {
1025 			case IPPROTO_DSTOPTS:
1026 			case IPPROTO_ROUTING:
1027 			case IPPROTO_HOPOPTS:
1028 			case IPPROTO_AH: /* is it possible? */
1029 				break;
1030 			default:
1031 				goto loopend;
1032 			}
1033 
1034 			ext = ip6_pullexthdr(m, off, nxt);
1035 			if (ext == NULL) {
1036 				ip6stat.ip6s_tooshort++;
1037 				return;
1038 			}
1039 			ip6e = mtod(ext, struct ip6_ext *);
1040 			if (nxt == IPPROTO_AH)
1041 				elen = (ip6e->ip6e_len + 2) << 2;
1042 			else
1043 				elen = (ip6e->ip6e_len + 1) << 3;
1044 			if (elen != ext->m_len) {
1045 				m_freem(ext);
1046 				ip6stat.ip6s_tooshort++;
1047 				return;
1048 			}
1049 
1050 			switch (nxt) {
1051 			case IPPROTO_DSTOPTS:
1052 				if (!in6p->in6p_flags & IN6P_DSTOPTS)
1053 					break;
1054 
1055 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1056 				    IPV6_DSTOPTS, IPPROTO_IPV6);
1057 				if (*mp)
1058 					mp = &(*mp)->m_next;
1059 				break;
1060 
1061 			case IPPROTO_ROUTING:
1062 				if (!in6p->in6p_flags & IN6P_RTHDR)
1063 					break;
1064 
1065 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1066 				    IPV6_RTHDR, IPPROTO_IPV6);
1067 				if (*mp)
1068 					mp = &(*mp)->m_next;
1069 				break;
1070 
1071 			case IPPROTO_HOPOPTS:
1072 			case IPPROTO_AH: /* is it possible? */
1073 				break;
1074 
1075 			default:
1076 				/*
1077 			 	 * other cases have been filtered in the above.
1078 				 * none will visit this case.  here we supply
1079 				 * the code just in case (nxt overwritten or
1080 				 * other cases).
1081 				 */
1082 				m_freem(ext);
1083 				goto loopend;
1084 
1085 			}
1086 
1087 			/* proceed with the next header. */
1088 			off += elen;
1089 			nxt = ip6e->ip6e_nxt;
1090 			ip6e = NULL;
1091 			m_freem(ext);
1092 			ext = NULL;
1093 		}
1094 	  loopend:
1095 	  	;
1096 	}
1097 # undef in6p_flags
1098 }
1099 
1100 /*
1101  * pull single extension header from mbuf chain.  returns single mbuf that
1102  * contains the result, or NULL on error.
1103  */
1104 static struct mbuf *
ip6_pullexthdr(m,off,nxt)1105 ip6_pullexthdr(m, off, nxt)
1106 	struct mbuf *m;
1107 	size_t off;
1108 	int nxt;
1109 {
1110 	struct ip6_ext ip6e;
1111 	size_t elen;
1112 	struct mbuf *n;
1113 
1114 #ifdef DIAGNOSTIC
1115 	switch (nxt) {
1116 	case IPPROTO_DSTOPTS:
1117 	case IPPROTO_ROUTING:
1118 	case IPPROTO_HOPOPTS:
1119 	case IPPROTO_AH: /* is it possible? */
1120 		break;
1121 	default:
1122 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1123 	}
1124 #endif
1125 
1126 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1127 	if (nxt == IPPROTO_AH)
1128 		elen = (ip6e.ip6e_len + 2) << 2;
1129 	else
1130 		elen = (ip6e.ip6e_len + 1) << 3;
1131 
1132 	MGET(n, M_DONTWAIT, MT_DATA);
1133 	if (n && elen >= MLEN) {
1134 		MCLGET(n, M_DONTWAIT);
1135 		if ((n->m_flags & M_EXT) == 0) {
1136 			m_free(n);
1137 			n = NULL;
1138 		}
1139 	}
1140 	if (!n)
1141 		return NULL;
1142 
1143 	n->m_len = 0;
1144 	if (elen >= M_TRAILINGSPACE(n)) {
1145 		m_free(n);
1146 		return NULL;
1147 	}
1148 
1149 	m_copydata(m, off, elen, mtod(n, caddr_t));
1150 	n->m_len = elen;
1151 	return n;
1152 }
1153 
1154 /*
1155  * Get pointer to the previous header followed by the header
1156  * currently processed.
1157  * XXX: This function supposes that
1158  *	M includes all headers,
1159  *	the next header field and the header length field of each header
1160  *	are valid, and
1161  *	the sum of each header length equals to OFF.
1162  * Because of these assumptions, this function must be called very
1163  * carefully. Moreover, it will not be used in the near future when
1164  * we develop `neater' mechanism to process extension headers.
1165  */
1166 u_int8_t *
ip6_get_prevhdr(m,off)1167 ip6_get_prevhdr(m, off)
1168 	struct mbuf *m;
1169 	int off;
1170 {
1171 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1172 
1173 	if (off == sizeof(struct ip6_hdr))
1174 		return (&ip6->ip6_nxt);
1175 	else {
1176 		int len, nxt;
1177 		struct ip6_ext *ip6e = NULL;
1178 
1179 		nxt = ip6->ip6_nxt;
1180 		len = sizeof(struct ip6_hdr);
1181 		while (len < off) {
1182 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1183 
1184 			switch (nxt) {
1185 			case IPPROTO_FRAGMENT:
1186 				len += sizeof(struct ip6_frag);
1187 				break;
1188 			case IPPROTO_AH:
1189 				len += (ip6e->ip6e_len + 2) << 2;
1190 				break;
1191 			default:
1192 				len += (ip6e->ip6e_len + 1) << 3;
1193 				break;
1194 			}
1195 			nxt = ip6e->ip6e_nxt;
1196 		}
1197 		if (ip6e)
1198 			return (&ip6e->ip6e_nxt);
1199 		else
1200 			return NULL;
1201 	}
1202 }
1203 
1204 /*
1205  * get next header offset.  m will be retained.
1206  */
1207 int
ip6_nexthdr(m,off,proto,nxtp)1208 ip6_nexthdr(m, off, proto, nxtp)
1209 	struct mbuf *m;
1210 	int off;
1211 	int proto;
1212 	int *nxtp;
1213 {
1214 	struct ip6_hdr ip6;
1215 	struct ip6_ext ip6e;
1216 	struct ip6_frag fh;
1217 
1218 	/* just in case */
1219 	if (m == NULL)
1220 		panic("ip6_nexthdr: m == NULL");
1221 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1222 		return -1;
1223 
1224 	switch (proto) {
1225 	case IPPROTO_IPV6:
1226 		if (m->m_pkthdr.len < off + sizeof(ip6))
1227 			return -1;
1228 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1229 		if (nxtp)
1230 			*nxtp = ip6.ip6_nxt;
1231 		off += sizeof(ip6);
1232 		return off;
1233 
1234 	case IPPROTO_FRAGMENT:
1235 		/*
1236 		 * terminate parsing if it is not the first fragment,
1237 		 * it does not make sense to parse through it.
1238 		 */
1239 		if (m->m_pkthdr.len < off + sizeof(fh))
1240 			return -1;
1241 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1242 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1243 			return -1;
1244 		if (nxtp)
1245 			*nxtp = fh.ip6f_nxt;
1246 		off += sizeof(struct ip6_frag);
1247 		return off;
1248 
1249 	case IPPROTO_AH:
1250 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1251 			return -1;
1252 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1253 		if (nxtp)
1254 			*nxtp = ip6e.ip6e_nxt;
1255 		off += (ip6e.ip6e_len + 2) << 2;
1256 		if (m->m_pkthdr.len < off)
1257 			return -1;
1258 		return off;
1259 
1260 	case IPPROTO_HOPOPTS:
1261 	case IPPROTO_ROUTING:
1262 	case IPPROTO_DSTOPTS:
1263 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1264 			return -1;
1265 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1266 		if (nxtp)
1267 			*nxtp = ip6e.ip6e_nxt;
1268 		off += (ip6e.ip6e_len + 1) << 3;
1269 		if (m->m_pkthdr.len < off)
1270 			return -1;
1271 		return off;
1272 
1273 	case IPPROTO_NONE:
1274 	case IPPROTO_ESP:
1275 	case IPPROTO_IPCOMP:
1276 		/* give up */
1277 		return -1;
1278 
1279 	default:
1280 		return -1;
1281 	}
1282 
1283 	return -1;
1284 }
1285 
1286 /*
1287  * get offset for the last header in the chain.  m will be kept untainted.
1288  */
1289 int
ip6_lasthdr(m,off,proto,nxtp)1290 ip6_lasthdr(m, off, proto, nxtp)
1291 	struct mbuf *m;
1292 	int off;
1293 	int proto;
1294 	int *nxtp;
1295 {
1296 	int newoff;
1297 	int nxt;
1298 
1299 	if (!nxtp) {
1300 		nxt = -1;
1301 		nxtp = &nxt;
1302 	}
1303 	while (1) {
1304 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1305 		if (newoff < 0)
1306 			return off;
1307 		else if (newoff < off)
1308 			return -1;	/* invalid */
1309 		else if (newoff == off)
1310 			return newoff;
1311 
1312 		off = newoff;
1313 		proto = *nxtp;
1314 	}
1315 }
1316 
1317 /*
1318  * System control for IP6
1319  */
1320 
1321 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1322 	0,		0,		0,		0,
1323 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1324 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1325 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1326 	0,		0,		0,		0,
1327 	ENOPROTOOPT
1328 };
1329 
1330 #include <uvm/uvm_extern.h>
1331 #include <sys/sysctl.h>
1332 
1333 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1334 
1335 int
ip6_sysctl(name,namelen,oldp,oldlenp,newp,newlen)1336 ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1337 	int *name;
1338 	u_int namelen;
1339 	void *oldp;
1340 	size_t *oldlenp;
1341 	void *newp;
1342 	size_t newlen;
1343 {
1344 	/* All sysctl names at this level are terminal. */
1345 	if (namelen != 1)
1346 		return ENOTDIR;
1347 
1348 	switch (name[0]) {
1349 	case IPV6CTL_KAME_VERSION:
1350 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1351 	case IPV6CTL_V6ONLY:
1352 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1353 	default:
1354 		if (name[0] < IPV6CTL_MAXID)
1355 			return (sysctl_int_arr(ipv6ctl_vars, name, namelen,
1356 			    oldp, oldlenp, newp, newlen));
1357 		return (EOPNOTSUPP);
1358 	}
1359 	/* NOTREACHED */
1360 }
1361