xref: /openbsd/sys/netinet6/ip6_input.c (revision 3d8817e4)
1 /*	$OpenBSD: ip6_input.c,v 1.99 2011/04/03 13:56:05 stsp 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 #include "carp.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/domain.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/errno.h>
76 #include <sys/time.h>
77 #include <sys/kernel.h>
78 #include <sys/syslog.h>
79 #include <sys/proc.h>
80 
81 #include <net/if.h>
82 #include <net/if_types.h>
83 #include <net/if_dl.h>
84 #include <net/route.h>
85 #include <net/netisr.h>
86 
87 #include <netinet/in.h>
88 #include <netinet/in_systm.h>
89 
90 #ifdef INET
91 #include <netinet/ip.h>
92 #endif
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 #ifdef MROUTING
109 #include <netinet6/ip6_mroute.h>
110 #endif
111 
112 #if NPF > 0
113 #include <net/pfvar.h>
114 #endif
115 
116 #if NCARP > 0
117 #include <netinet/in_var.h>
118 #include <netinet/ip_carp.h>
119 #endif
120 
121 extern struct domain inet6domain;
122 extern struct ip6protosw inet6sw[];
123 
124 u_char ip6_protox[IPPROTO_MAX];
125 static int ip6qmaxlen = IFQ_MAXLEN;
126 struct in6_ifaddr *in6_ifaddr;
127 struct ifqueue ip6intrq;
128 
129 struct ip6stat ip6stat;
130 
131 void ip6_init2(void *);
132 int ip6_check_rh0hdr(struct mbuf *);
133 
134 int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
135 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
136 
137 /*
138  * IP6 initialization: fill in IP6 protocol switch table.
139  * All protocols not implemented in kernel go to raw IP6 protocol handler.
140  */
141 void
142 ip6_init(void)
143 {
144 	struct ip6protosw *pr;
145 	int i;
146 
147 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
148 	if (pr == 0)
149 		panic("ip6_init");
150 	for (i = 0; i < IPPROTO_MAX; i++)
151 		ip6_protox[i] = pr - inet6sw;
152 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
153 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
154 		if (pr->pr_domain->dom_family == PF_INET6 &&
155 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
156 		    pr->pr_protocol < IPPROTO_MAX)
157 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
158 	ip6intrq.ifq_maxlen = ip6qmaxlen;
159 	ip6_randomid_init();
160 	nd6_init();
161 	frag6_init();
162 	ip6_init2((void *)0);
163 }
164 
165 void
166 ip6_init2(void *dummy)
167 {
168 
169 	/* nd6_timer_init */
170 	bzero(&nd6_timer_ch, sizeof(nd6_timer_ch));
171 	timeout_set(&nd6_timer_ch, nd6_timer, NULL);
172 	timeout_add_sec(&nd6_timer_ch, 1);
173 }
174 
175 /*
176  * IP6 input interrupt handling. Just pass the packet to ip6_input.
177  */
178 void
179 ip6intr(void)
180 {
181 	int s;
182 	struct mbuf *m;
183 
184 	for (;;) {
185 		s = splnet();
186 		IF_DEQUEUE(&ip6intrq, m);
187 		splx(s);
188 		if (m == NULL)
189 			return;
190 		ip6_input(m);
191 	}
192 }
193 
194 extern struct	route_in6 ip6_forward_rt;
195 
196 void
197 ip6_input(struct mbuf *m)
198 {
199 	struct ip6_hdr *ip6;
200 	int off = sizeof(struct ip6_hdr), nest;
201 	u_int32_t plen;
202 	u_int32_t rtalert = ~0;
203 	int nxt, ours = 0;
204 	struct ifnet *deliverifp = NULL;
205 #if NPF > 0
206 	struct in6_addr odst;
207 #endif
208 	int srcrt = 0, isanycast = 0;
209 	u_int rtableid = 0;
210 
211 	/*
212 	 * mbuf statistics by kazu
213 	 */
214 	if (m->m_flags & M_EXT) {
215 		if (m->m_next)
216 			ip6stat.ip6s_mext2m++;
217 		else
218 			ip6stat.ip6s_mext1++;
219 	} else {
220 		if (m->m_next) {
221 			if (m->m_flags & M_LOOP) {
222 				ip6stat.ip6s_m2m[lo0ifp->if_index]++;	/*XXX*/
223 			} else if (m->m_pkthdr.rcvif->if_index < nitems(ip6stat.ip6s_m2m))
224 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
225 			else
226 				ip6stat.ip6s_m2m[0]++;
227 		} else
228 			ip6stat.ip6s_m1++;
229 	}
230 
231 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
232 	ip6stat.ip6s_total++;
233 
234 	if (m->m_len < sizeof(struct ip6_hdr)) {
235 		struct ifnet *inifp;
236 		inifp = m->m_pkthdr.rcvif;
237 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
238 			ip6stat.ip6s_toosmall++;
239 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
240 			return;
241 		}
242 	}
243 
244 	ip6 = mtod(m, struct ip6_hdr *);
245 
246 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
247 		ip6stat.ip6s_badvers++;
248 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
249 		goto bad;
250 	}
251 
252 #if NCARP > 0
253 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
254 	    ip6->ip6_nxt != IPPROTO_ICMPV6 &&
255 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
256 	    ip6->ip6_dst.s6_addr32))
257 		goto bad;
258 #endif
259 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
260 
261 	/*
262 	 * Check against address spoofing/corruption.
263 	 */
264 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
265 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
266 		/*
267 		 * XXX: "badscope" is not very suitable for a multicast source.
268 		 */
269 		ip6stat.ip6s_badscope++;
270 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
271 		goto bad;
272 	}
273 
274 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
275 	    !(m->m_flags & M_LOOP)) {
276 		/*
277 		 * In this case, the packet should come from the loopback
278 		 * interface.  However, we cannot just check the if_flags,
279 		 * because ip6_mloopback() passes the "actual" interface
280 		 * as the outgoing/incoming interface.
281 		 */
282 		ip6stat.ip6s_badscope++;
283 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
284 		goto bad;
285 	}
286 
287 	/*
288 	 * The following check is not documented in specs.  A malicious
289 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
290 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
291 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
292 	 *
293 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
294 	 * support IPv4-less kernel compilation, we cannot support SIIT
295 	 * environment at all.  So, it makes more sense for us to reject any
296 	 * malicious packets for non-SIIT environment, than try to do a
297 	 * partial support for SIIT environment.
298 	 */
299 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
300 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
301 		ip6stat.ip6s_badscope++;
302 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
303 		goto bad;
304 	}
305 #if 0
306 	/*
307 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
308 	 *
309 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
310 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
311 	 * is revised to forbid relaying case.
312 	 */
313 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
314 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
315 		ip6stat.ip6s_badscope++;
316 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
317 		goto bad;
318 	}
319 #endif
320 
321 	if (ip6_check_rh0hdr(m)) {
322 		ip6stat.ip6s_badoptions++;
323 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
324 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
325 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, 0);
326 		/* m is already freed */
327 		return;
328 	}
329 
330 #if NPF > 0
331         /*
332          * Packet filter
333          */
334 	odst = ip6->ip6_dst;
335 	if (pf_test6(PF_IN, m->m_pkthdr.rcvif, &m, NULL) != PF_PASS)
336 		goto bad;
337 	if (m == NULL)
338 		return;
339 
340 	ip6 = mtod(m, struct ip6_hdr *);
341 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
342 #endif
343 
344 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
345 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
346 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
347 			ours = 1;
348 			deliverifp = m->m_pkthdr.rcvif;
349 			goto hbhcheck;
350 		} else {
351 			ip6stat.ip6s_badscope++;
352 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
353 			goto bad;
354 		}
355 	}
356 
357 	/* drop packets if interface ID portion is already filled */
358 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
359 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src) &&
360 		    ip6->ip6_src.s6_addr16[1]) {
361 			ip6stat.ip6s_badscope++;
362 			goto bad;
363 		}
364 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) &&
365 		    ip6->ip6_dst.s6_addr16[1]) {
366 			ip6stat.ip6s_badscope++;
367 			goto bad;
368 		}
369 	}
370 
371 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
372 		ip6->ip6_src.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
373 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
374 		ip6->ip6_dst.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
375 
376 	/*
377 	 * We use rt->rt_ifp to determine if the address is ours or not.
378 	 * If rt_ifp is lo0, the address is ours.
379 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
380 	 * so any address under fe80::%lo0/64 will be mistakenly considered
381 	 * local.  The special case is supplied to handle the case properly
382 	 * by actually looking at interface addresses
383 	 * (using in6ifa_ifpwithaddr).
384 	 */
385 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
386 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
387 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
388 			icmp6_error(m, ICMP6_DST_UNREACH,
389 			    ICMP6_DST_UNREACH_ADDR, 0);
390 			/* m is already freed */
391 			return;
392 		}
393 
394 		ours = 1;
395 		deliverifp = m->m_pkthdr.rcvif;
396 		goto hbhcheck;
397 	}
398 
399 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
400 		ours = 1;
401 		deliverifp = m->m_pkthdr.rcvif;
402 		goto hbhcheck;
403 	}
404 
405 	/*
406 	 * Multicast check
407 	 */
408 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
409 	  	struct	in6_multi *in6m = 0;
410 
411 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
412 		/*
413 		 * See if we belong to the destination multicast group on the
414 		 * arrival interface.
415 		 */
416 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
417 		if (in6m)
418 			ours = 1;
419 #ifdef MROUTING
420 		else if (!ip6_mforwarding || !ip6_mrouter)
421 #else
422 		else
423 #endif
424 		{
425 			ip6stat.ip6s_notmember++;
426 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
427 				ip6stat.ip6s_cantforward++;
428 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
429 			goto bad;
430 		}
431 		deliverifp = m->m_pkthdr.rcvif;
432 		goto hbhcheck;
433 	}
434 
435 #if NPF > 0
436 	rtableid = m->m_pkthdr.rdomain;
437 #endif
438 
439 	/*
440 	 *  Unicast check
441 	 */
442 	if (ip6_forward_rt.ro_rt != NULL &&
443 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
444 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
445 			       &ip6_forward_rt.ro_dst.sin6_addr) &&
446 	    rtableid == ip6_forward_rt.ro_tableid)
447 		ip6stat.ip6s_forward_cachehit++;
448 	else {
449 		if (ip6_forward_rt.ro_rt) {
450 			/* route is down or destination is different */
451 			ip6stat.ip6s_forward_cachemiss++;
452 			RTFREE(ip6_forward_rt.ro_rt);
453 			ip6_forward_rt.ro_rt = 0;
454 		}
455 
456 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
457 		ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
458 		ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
459 		ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
460 		ip6_forward_rt.ro_tableid = rtableid;
461 
462 		rtalloc_mpath((struct route *)&ip6_forward_rt,
463 		    &ip6->ip6_src.s6_addr32[0]);
464 	}
465 
466 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
467 
468 	/*
469 	 * Accept the packet if the forwarding interface to the destination
470 	 * according to the routing table is the loopback interface,
471 	 * unless the associated route has a gateway.
472 	 * Note that this approach causes to accept a packet if there is a
473 	 * route to the loopback interface for the destination of the packet.
474 	 * But we think it's even useful in some situations, e.g. when using
475 	 * a special daemon which wants to intercept the packet.
476 	 */
477 	if (ip6_forward_rt.ro_rt &&
478 	    (ip6_forward_rt.ro_rt->rt_flags &
479 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
480 #if 0
481 	    /*
482 	     * The check below is redundant since the comparison of
483 	     * the destination and the key of the rtentry has
484 	     * already done through looking up the routing table.
485 	     */
486 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
487 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
488 #endif
489 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
490 		struct in6_ifaddr *ia6 =
491 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
492 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
493 			isanycast = 1;
494 		/*
495 		 * packets to a tentative, duplicated, or somehow invalid
496 		 * address must not be accepted.
497 		 */
498 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
499 			/* this address is ready */
500 			ours = 1;
501 			deliverifp = ia6->ia_ifp;	/* correct? */
502 			goto hbhcheck;
503 		} else {
504 			/* address is not ready, so discard the packet. */
505 			nd6log((LOG_INFO,
506 			    "ip6_input: packet to an unready address %s->%s\n",
507 			    ip6_sprintf(&ip6->ip6_src),
508 			    ip6_sprintf(&ip6->ip6_dst)));
509 
510 			goto bad;
511 		}
512 	}
513 
514 	/*
515 	 * FAITH (Firewall Aided Internet Translator)
516 	 */
517 #if defined(NFAITH) && 0 < NFAITH
518 	if (ip6_keepfaith) {
519 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
520 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
521 			/* XXX do we need more sanity checks? */
522 			ours = 1;
523 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
524 			goto hbhcheck;
525 		}
526 	}
527 #endif
528 
529 #if 0
530     {
531 	/*
532 	 * Last resort: check in6_ifaddr for incoming interface.
533 	 * The code is here until I update the "goto ours hack" code above
534 	 * working right.
535 	 */
536 	struct ifaddr *ifa;
537 	TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
538 		if (ifa->ifa_addr == NULL)
539 			continue;	/* just for safety */
540 		if (ifa->ifa_addr->sa_family != AF_INET6)
541 			continue;
542 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
543 			ours = 1;
544 			deliverifp = ifa->ifa_ifp;
545 			goto hbhcheck;
546 		}
547 	}
548     }
549 #endif
550 
551 #if NCARP > 0
552 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
553 	    ip6->ip6_nxt == IPPROTO_ICMPV6 &&
554 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
555 	    ip6->ip6_dst.s6_addr32))
556 		goto bad;
557 #endif
558 	/*
559 	 * Now there is no reason to process the packet if it's not our own
560 	 * and we're not a router.
561 	 */
562 	if (!ip6_forwarding) {
563 		ip6stat.ip6s_cantforward++;
564 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
565 		goto bad;
566 	}
567 
568   hbhcheck:
569 	/*
570 	 * Process Hop-by-Hop options header if it's contained.
571 	 * m may be modified in ip6_hopopts_input().
572 	 * If a JumboPayload option is included, plen will also be modified.
573 	 */
574 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
575 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
576 		struct ip6_hbh *hbh;
577 
578 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
579 #if 0	/*touches NULL pointer*/
580 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
581 #endif
582 			return;	/* m have already been freed */
583 		}
584 
585 		/* adjust pointer */
586 		ip6 = mtod(m, struct ip6_hdr *);
587 
588 		/*
589 		 * if the payload length field is 0 and the next header field
590 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
591 		 * option MUST be included.
592 		 */
593 		if (ip6->ip6_plen == 0 && plen == 0) {
594 			/*
595 			 * Note that if a valid jumbo payload option is
596 			 * contained, ip6_hoptops_input() must set a valid
597 			 * (non-zero) payload length to the variable plen.
598 			 */
599 			ip6stat.ip6s_badoptions++;
600 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
601 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
602 			icmp6_error(m, ICMP6_PARAM_PROB,
603 				    ICMP6_PARAMPROB_HEADER,
604 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
605 			return;
606 		}
607 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
608 			sizeof(struct ip6_hbh));
609 		if (hbh == NULL) {
610 			ip6stat.ip6s_tooshort++;
611 			return;
612 		}
613 		nxt = hbh->ip6h_nxt;
614 
615 		/*
616 		 * accept the packet if a router alert option is included
617 		 * and we act as an IPv6 router.
618 		 */
619 		if (rtalert != ~0 && ip6_forwarding)
620 			ours = 1;
621 	} else
622 		nxt = ip6->ip6_nxt;
623 
624 	/*
625 	 * Check that the amount of data in the buffers
626 	 * is as at least much as the IPv6 header would have us expect.
627 	 * Trim mbufs if longer than we expect.
628 	 * Drop packet if shorter than we expect.
629 	 */
630 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
631 		ip6stat.ip6s_tooshort++;
632 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
633 		goto bad;
634 	}
635 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
636 		if (m->m_len == m->m_pkthdr.len) {
637 			m->m_len = sizeof(struct ip6_hdr) + plen;
638 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
639 		} else
640 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
641 	}
642 
643 	/*
644 	 * Forward if desirable.
645 	 */
646 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
647 		/*
648 		 * If we are acting as a multicast router, all
649 		 * incoming multicast packets are passed to the
650 		 * kernel-level multicast forwarding function.
651 		 * The packet is returned (relatively) intact; if
652 		 * ip6_mforward() returns a non-zero value, the packet
653 		 * must be discarded, else it may be accepted below.
654 		 */
655 #ifdef MROUTING
656 		if (ip6_mforwarding && ip6_mrouter &&
657 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
658 			ip6stat.ip6s_cantforward++;
659 			m_freem(m);
660 			return;
661 		}
662 #endif
663 		if (!ours) {
664 			m_freem(m);
665 			return;
666 		}
667 	} else if (!ours) {
668 		ip6_forward(m, srcrt);
669 		return;
670 	}
671 
672 	ip6 = mtod(m, struct ip6_hdr *);
673 
674 	/*
675 	 * Malicious party may be able to use IPv4 mapped addr to confuse
676 	 * tcp/udp stack and bypass security checks (act as if it was from
677 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
678 	 *
679 	 * For SIIT end node behavior, you may want to disable the check.
680 	 * However, you will  become vulnerable to attacks using IPv4 mapped
681 	 * source.
682 	 */
683 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
684 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
685 		ip6stat.ip6s_badscope++;
686 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
687 		goto bad;
688 	}
689 
690 	/*
691 	 * Tell launch routine the next header
692 	 */
693 	ip6stat.ip6s_delivered++;
694 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
695 	nest = 0;
696 
697 	while (nxt != IPPROTO_DONE) {
698 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
699 			ip6stat.ip6s_toomanyhdr++;
700 			goto bad;
701 		}
702 
703 		/*
704 		 * protection against faulty packet - there should be
705 		 * more sanity checks in header chain processing.
706 		 */
707 		if (m->m_pkthdr.len < off) {
708 			ip6stat.ip6s_tooshort++;
709 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
710 			goto bad;
711 		}
712 
713 		/* draft-itojun-ipv6-tcp-to-anycast */
714 		if (isanycast && nxt == IPPROTO_TCP) {
715 			if (m->m_len >= sizeof(struct ip6_hdr)) {
716 				ip6 = mtod(m, struct ip6_hdr *);
717 				icmp6_error(m, ICMP6_DST_UNREACH,
718 					ICMP6_DST_UNREACH_ADDR,
719 					(caddr_t)&ip6->ip6_dst - (caddr_t)ip6);
720 				break;
721 			} else
722 				goto bad;
723 		}
724 
725 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
726 	}
727 	return;
728  bad:
729 	m_freem(m);
730 }
731 
732 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test6() */
733 int
734 ip6_check_rh0hdr(struct mbuf *m)
735 {
736 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
737 	struct ip6_rthdr rthdr;
738 	struct ip6_ext opt6;
739 	u_int8_t proto = ip6->ip6_nxt;
740 	int done = 0, lim, off, rh_cnt = 0;
741 
742 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
743 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
744 	do {
745 		switch (proto) {
746 		case IPPROTO_ROUTING:
747 			if (rh_cnt++) {
748 				/* more then one rh header present */
749 				return (1);
750 			}
751 
752 			if (off + sizeof(rthdr) > lim) {
753 				/* packet to short to make sense */
754 				return (1);
755 			}
756 
757 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
758 
759 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0)
760 				return (1);
761 
762 			off += (rthdr.ip6r_len + 1) * 8;
763 			proto = rthdr.ip6r_nxt;
764 			break;
765 		case IPPROTO_AH:
766 		case IPPROTO_HOPOPTS:
767 		case IPPROTO_DSTOPTS:
768 			/* get next header and header length */
769 			if (off + sizeof(opt6) > lim) {
770 				/*
771 				 * Packet to short to make sense, we could
772 				 * reject the packet but as a router we
773 				 * should not do that so forward it.
774 				 */
775 				return (0);
776 			}
777 
778 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
779 
780 			if (proto == IPPROTO_AH)
781 				off += (opt6.ip6e_len + 2) * 4;
782 			else
783 				off += (opt6.ip6e_len + 1) * 8;
784 			proto = opt6.ip6e_nxt;
785 			break;
786 		case IPPROTO_FRAGMENT:
787 		default:
788 			/* end of header stack */
789 			done = 1;
790 			break;
791 		}
792 	} while (!done);
793 
794 	return (0);
795 }
796 
797 /*
798  * Hop-by-Hop options header processing. If a valid jumbo payload option is
799  * included, the real payload length will be stored in plenp.
800  *
801  * rtalertp - XXX: should be stored in a more smart way
802  */
803 int
804 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
805     int *offp)
806 {
807 	struct mbuf *m = *mp;
808 	int off = *offp, hbhlen;
809 	struct ip6_hbh *hbh;
810 
811 	/* validation of the length of the header */
812 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
813 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
814 	if (hbh == NULL) {
815 		ip6stat.ip6s_tooshort++;
816 		return -1;
817 	}
818 	hbhlen = (hbh->ip6h_len + 1) << 3;
819 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
820 		hbhlen);
821 	if (hbh == NULL) {
822 		ip6stat.ip6s_tooshort++;
823 		return -1;
824 	}
825 	off += hbhlen;
826 	hbhlen -= sizeof(struct ip6_hbh);
827 
828 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
829 				hbhlen, rtalertp, plenp) < 0)
830 		return (-1);
831 
832 	*offp = off;
833 	*mp = m;
834 	return (0);
835 }
836 
837 /*
838  * Search header for all Hop-by-hop options and process each option.
839  * This function is separate from ip6_hopopts_input() in order to
840  * handle a case where the sending node itself process its hop-by-hop
841  * options header. In such a case, the function is called from ip6_output().
842  *
843  * The function assumes that hbh header is located right after the IPv6 header
844  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
845  * opthead + hbhlen is located in continuous memory region.
846  */
847 int
848 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
849     u_int32_t *rtalertp, u_int32_t *plenp)
850 {
851 	struct ip6_hdr *ip6;
852 	int optlen = 0;
853 	u_int8_t *opt = opthead;
854 	u_int16_t rtalert_val;
855 	u_int32_t jumboplen;
856 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
857 
858 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
859 		switch (*opt) {
860 		case IP6OPT_PAD1:
861 			optlen = 1;
862 			break;
863 		case IP6OPT_PADN:
864 			if (hbhlen < IP6OPT_MINLEN) {
865 				ip6stat.ip6s_toosmall++;
866 				goto bad;
867 			}
868 			optlen = *(opt + 1) + 2;
869 			break;
870 		case IP6OPT_ROUTER_ALERT:
871 			/* XXX may need check for alignment */
872 			if (hbhlen < IP6OPT_RTALERT_LEN) {
873 				ip6stat.ip6s_toosmall++;
874 				goto bad;
875 			}
876 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
877 				/* XXX stat */
878 				icmp6_error(m, ICMP6_PARAM_PROB,
879 				    ICMP6_PARAMPROB_HEADER,
880 				    erroff + opt + 1 - opthead);
881 				return (-1);
882 			}
883 			optlen = IP6OPT_RTALERT_LEN;
884 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
885 			*rtalertp = ntohs(rtalert_val);
886 			break;
887 		case IP6OPT_JUMBO:
888 			/* XXX may need check for alignment */
889 			if (hbhlen < IP6OPT_JUMBO_LEN) {
890 				ip6stat.ip6s_toosmall++;
891 				goto bad;
892 			}
893 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
894 				/* XXX stat */
895 				icmp6_error(m, ICMP6_PARAM_PROB,
896 				    ICMP6_PARAMPROB_HEADER,
897 				    erroff + opt + 1 - opthead);
898 				return (-1);
899 			}
900 			optlen = IP6OPT_JUMBO_LEN;
901 
902 			/*
903 			 * IPv6 packets that have non 0 payload length
904 			 * must not contain a jumbo payload option.
905 			 */
906 			ip6 = mtod(m, struct ip6_hdr *);
907 			if (ip6->ip6_plen) {
908 				ip6stat.ip6s_badoptions++;
909 				icmp6_error(m, ICMP6_PARAM_PROB,
910 				    ICMP6_PARAMPROB_HEADER,
911 				    erroff + opt - opthead);
912 				return (-1);
913 			}
914 
915 			/*
916 			 * We may see jumbolen in unaligned location, so
917 			 * we'd need to perform bcopy().
918 			 */
919 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
920 			jumboplen = (u_int32_t)htonl(jumboplen);
921 
922 #if 1
923 			/*
924 			 * if there are multiple jumbo payload options,
925 			 * *plenp will be non-zero and the packet will be
926 			 * rejected.
927 			 * the behavior may need some debate in ipngwg -
928 			 * multiple options does not make sense, however,
929 			 * there's no explicit mention in specification.
930 			 */
931 			if (*plenp != 0) {
932 				ip6stat.ip6s_badoptions++;
933 				icmp6_error(m, ICMP6_PARAM_PROB,
934 				    ICMP6_PARAMPROB_HEADER,
935 				    erroff + opt + 2 - opthead);
936 				return (-1);
937 			}
938 #endif
939 
940 			/*
941 			 * jumbo payload length must be larger than 65535.
942 			 */
943 			if (jumboplen <= IPV6_MAXPACKET) {
944 				ip6stat.ip6s_badoptions++;
945 				icmp6_error(m, ICMP6_PARAM_PROB,
946 				    ICMP6_PARAMPROB_HEADER,
947 				    erroff + opt + 2 - opthead);
948 				return (-1);
949 			}
950 			*plenp = jumboplen;
951 
952 			break;
953 		default:		/* unknown option */
954 			if (hbhlen < IP6OPT_MINLEN) {
955 				ip6stat.ip6s_toosmall++;
956 				goto bad;
957 			}
958 			optlen = ip6_unknown_opt(opt, m,
959 			    erroff + opt - opthead);
960 			if (optlen == -1)
961 				return (-1);
962 			optlen += 2;
963 			break;
964 		}
965 	}
966 
967 	return (0);
968 
969   bad:
970 	m_freem(m);
971 	return (-1);
972 }
973 
974 /*
975  * Unknown option processing.
976  * The third argument `off' is the offset from the IPv6 header to the option,
977  * which allows returning an ICMPv6 error even if the IPv6 header and the
978  * option header are not continuous.
979  */
980 int
981 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
982 {
983 	struct ip6_hdr *ip6;
984 
985 	switch (IP6OPT_TYPE(*optp)) {
986 	case IP6OPT_TYPE_SKIP: /* ignore the option */
987 		return ((int)*(optp + 1));
988 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
989 		m_freem(m);
990 		return (-1);
991 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
992 		ip6stat.ip6s_badoptions++;
993 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
994 		return (-1);
995 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
996 		ip6stat.ip6s_badoptions++;
997 		ip6 = mtod(m, struct ip6_hdr *);
998 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
999 		    (m->m_flags & (M_BCAST|M_MCAST)))
1000 			m_freem(m);
1001 		else
1002 			icmp6_error(m, ICMP6_PARAM_PROB,
1003 				    ICMP6_PARAMPROB_OPTION, off);
1004 		return (-1);
1005 	}
1006 
1007 	m_freem(m);		/* XXX: NOTREACHED */
1008 	return (-1);
1009 }
1010 
1011 /*
1012  * Create the "control" list for this pcb.
1013  *
1014  * The routine will be called from upper layer handlers like tcp6_input().
1015  * Thus the routine assumes that the caller (tcp6_input) have already
1016  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1017  * very first mbuf on the mbuf chain.
1018  * We may want to add some infinite loop prevention or sanity checks for safety.
1019  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1020  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1021  */
1022 void
1023 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1024 {
1025 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1026 # define in6p_flags	inp_flags
1027 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1028 
1029 #ifdef SO_TIMESTAMP
1030 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
1031 		struct timeval tv;
1032 
1033 		microtime(&tv);
1034 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1035 		    SCM_TIMESTAMP, SOL_SOCKET);
1036 		if (*mp)
1037 			mp = &(*mp)->m_next;
1038 	}
1039 #endif
1040 
1041 	/* RFC 2292 sec. 5 */
1042 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1043 		struct in6_pktinfo pi6;
1044 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1045 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
1046 			pi6.ipi6_addr.s6_addr16[1] = 0;
1047 		pi6.ipi6_ifindex =
1048 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1049 		*mp = sbcreatecontrol((caddr_t) &pi6,
1050 		    sizeof(struct in6_pktinfo),
1051 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1052 		if (*mp)
1053 			mp = &(*mp)->m_next;
1054 	}
1055 
1056 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1057 		int hlim = ip6->ip6_hlim & 0xff;
1058 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1059 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1060 		if (*mp)
1061 			mp = &(*mp)->m_next;
1062 	}
1063 
1064 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1065 		u_int32_t flowinfo;
1066 		int tclass;
1067 
1068 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1069 		flowinfo >>= 20;
1070 
1071 		tclass = flowinfo & 0xff;
1072 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
1073 		    IPV6_TCLASS, IPPROTO_IPV6);
1074 		if (*mp)
1075 			mp = &(*mp)->m_next;
1076 	}
1077 
1078 	/*
1079 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1080 	 * privilege for the option (see ip6_ctloutput), but it might be too
1081 	 * strict, since there might be some hop-by-hop options which can be
1082 	 * returned to normal user.
1083 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1084 	 */
1085 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1086 		/*
1087 		 * Check if a hop-by-hop options header is contatined in the
1088 		 * received packet, and if so, store the options as ancillary
1089 		 * data. Note that a hop-by-hop options header must be
1090 		 * just after the IPv6 header, which is assured through the
1091 		 * IPv6 input processing.
1092 		 */
1093 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1094 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1095 			struct ip6_hbh *hbh;
1096 			int hbhlen = 0;
1097 			struct mbuf *ext;
1098 
1099 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1100 			    ip6->ip6_nxt);
1101 			if (ext == NULL) {
1102 				ip6stat.ip6s_tooshort++;
1103 				return;
1104 			}
1105 			hbh = mtod(ext, struct ip6_hbh *);
1106 			hbhlen = (hbh->ip6h_len + 1) << 3;
1107 			if (hbhlen != ext->m_len) {
1108 				m_freem(ext);
1109 				ip6stat.ip6s_tooshort++;
1110 				return;
1111 			}
1112 
1113 			/*
1114 			 * XXX: We copy the whole header even if a
1115 			 * jumbo payload option is included, the option which
1116 			 * is to be removed before returning according to
1117 			 * RFC2292.
1118 			 * Note: this constraint is removed in RFC3542.
1119 			 */
1120 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1121 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1122 			    IPPROTO_IPV6);
1123 			if (*mp)
1124 				mp = &(*mp)->m_next;
1125 			m_freem(ext);
1126 		}
1127 	}
1128 
1129 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1130 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1131 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1132 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1133 
1134 		/*
1135 		 * Search for destination options headers or routing
1136 		 * header(s) through the header chain, and stores each
1137 		 * header as ancillary data.
1138 		 * Note that the order of the headers remains in
1139 		 * the chain of ancillary data.
1140 		 */
1141 		while (1) {	/* is explicit loop prevention necessary? */
1142 			struct ip6_ext *ip6e = NULL;
1143 			int elen;
1144 			struct mbuf *ext = NULL;
1145 
1146 			/*
1147 			 * if it is not an extension header, don't try to
1148 			 * pull it from the chain.
1149 			 */
1150 			switch (nxt) {
1151 			case IPPROTO_DSTOPTS:
1152 			case IPPROTO_ROUTING:
1153 			case IPPROTO_HOPOPTS:
1154 			case IPPROTO_AH: /* is it possible? */
1155 				break;
1156 			default:
1157 				goto loopend;
1158 			}
1159 
1160 			ext = ip6_pullexthdr(m, off, nxt);
1161 			if (ext == NULL) {
1162 				ip6stat.ip6s_tooshort++;
1163 				return;
1164 			}
1165 			ip6e = mtod(ext, struct ip6_ext *);
1166 			if (nxt == IPPROTO_AH)
1167 				elen = (ip6e->ip6e_len + 2) << 2;
1168 			else
1169 				elen = (ip6e->ip6e_len + 1) << 3;
1170 			if (elen != ext->m_len) {
1171 				m_freem(ext);
1172 				ip6stat.ip6s_tooshort++;
1173 				return;
1174 			}
1175 
1176 			switch (nxt) {
1177 			case IPPROTO_DSTOPTS:
1178 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1179 					break;
1180 
1181 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1182 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1183 				    IPPROTO_IPV6);
1184 				if (*mp)
1185 					mp = &(*mp)->m_next;
1186 				break;
1187 
1188 			case IPPROTO_ROUTING:
1189 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1190 					break;
1191 
1192 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1193 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1194 				    IPPROTO_IPV6);
1195 				if (*mp)
1196 					mp = &(*mp)->m_next;
1197 				break;
1198 
1199 			case IPPROTO_HOPOPTS:
1200 			case IPPROTO_AH: /* is it possible? */
1201 				break;
1202 
1203 			default:
1204 				/*
1205 				 * other cases have been filtered in the above.
1206 				 * none will visit this case.  here we supply
1207 				 * the code just in case (nxt overwritten or
1208 				 * other cases).
1209 				 */
1210 				m_freem(ext);
1211 				goto loopend;
1212 
1213 			}
1214 
1215 			/* proceed with the next header. */
1216 			off += elen;
1217 			nxt = ip6e->ip6e_nxt;
1218 			ip6e = NULL;
1219 			m_freem(ext);
1220 			ext = NULL;
1221 		}
1222 	  loopend:
1223 	  	;
1224 	}
1225 # undef in6p_flags
1226 }
1227 
1228 /*
1229  * pull single extension header from mbuf chain.  returns single mbuf that
1230  * contains the result, or NULL on error.
1231  */
1232 struct mbuf *
1233 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1234 {
1235 	struct ip6_ext ip6e;
1236 	size_t elen;
1237 	struct mbuf *n;
1238 
1239 #ifdef DIAGNOSTIC
1240 	switch (nxt) {
1241 	case IPPROTO_DSTOPTS:
1242 	case IPPROTO_ROUTING:
1243 	case IPPROTO_HOPOPTS:
1244 	case IPPROTO_AH: /* is it possible? */
1245 		break;
1246 	default:
1247 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1248 	}
1249 #endif
1250 
1251 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1252 	if (nxt == IPPROTO_AH)
1253 		elen = (ip6e.ip6e_len + 2) << 2;
1254 	else
1255 		elen = (ip6e.ip6e_len + 1) << 3;
1256 
1257 	MGET(n, M_DONTWAIT, MT_DATA);
1258 	if (n && elen >= MLEN) {
1259 		MCLGET(n, M_DONTWAIT);
1260 		if ((n->m_flags & M_EXT) == 0) {
1261 			m_free(n);
1262 			n = NULL;
1263 		}
1264 	}
1265 	if (!n)
1266 		return NULL;
1267 
1268 	n->m_len = 0;
1269 	if (elen >= M_TRAILINGSPACE(n)) {
1270 		m_free(n);
1271 		return NULL;
1272 	}
1273 
1274 	m_copydata(m, off, elen, mtod(n, caddr_t));
1275 	n->m_len = elen;
1276 	return n;
1277 }
1278 
1279 /*
1280  * Get pointer to the previous header followed by the header
1281  * currently processed.
1282  * XXX: This function supposes that
1283  *	M includes all headers,
1284  *	the next header field and the header length field of each header
1285  *	are valid, and
1286  *	the sum of each header length equals to OFF.
1287  * Because of these assumptions, this function must be called very
1288  * carefully. Moreover, it will not be used in the near future when
1289  * we develop `neater' mechanism to process extension headers.
1290  */
1291 u_int8_t *
1292 ip6_get_prevhdr(struct mbuf *m, int off)
1293 {
1294 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1295 
1296 	if (off == sizeof(struct ip6_hdr))
1297 		return (&ip6->ip6_nxt);
1298 	else {
1299 		int len, nxt;
1300 		struct ip6_ext *ip6e = NULL;
1301 
1302 		nxt = ip6->ip6_nxt;
1303 		len = sizeof(struct ip6_hdr);
1304 		while (len < off) {
1305 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1306 
1307 			switch (nxt) {
1308 			case IPPROTO_FRAGMENT:
1309 				len += sizeof(struct ip6_frag);
1310 				break;
1311 			case IPPROTO_AH:
1312 				len += (ip6e->ip6e_len + 2) << 2;
1313 				break;
1314 			default:
1315 				len += (ip6e->ip6e_len + 1) << 3;
1316 				break;
1317 			}
1318 			nxt = ip6e->ip6e_nxt;
1319 		}
1320 		if (ip6e)
1321 			return (&ip6e->ip6e_nxt);
1322 		else
1323 			return NULL;
1324 	}
1325 }
1326 
1327 /*
1328  * get next header offset.  m will be retained.
1329  */
1330 int
1331 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1332 {
1333 	struct ip6_hdr ip6;
1334 	struct ip6_ext ip6e;
1335 	struct ip6_frag fh;
1336 
1337 	/* just in case */
1338 	if (m == NULL)
1339 		panic("ip6_nexthdr: m == NULL");
1340 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1341 		return -1;
1342 
1343 	switch (proto) {
1344 	case IPPROTO_IPV6:
1345 		if (m->m_pkthdr.len < off + sizeof(ip6))
1346 			return -1;
1347 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1348 		if (nxtp)
1349 			*nxtp = ip6.ip6_nxt;
1350 		off += sizeof(ip6);
1351 		return off;
1352 
1353 	case IPPROTO_FRAGMENT:
1354 		/*
1355 		 * terminate parsing if it is not the first fragment,
1356 		 * it does not make sense to parse through it.
1357 		 */
1358 		if (m->m_pkthdr.len < off + sizeof(fh))
1359 			return -1;
1360 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1361 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1362 			return -1;
1363 		if (nxtp)
1364 			*nxtp = fh.ip6f_nxt;
1365 		off += sizeof(struct ip6_frag);
1366 		return off;
1367 
1368 	case IPPROTO_AH:
1369 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1370 			return -1;
1371 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1372 		if (nxtp)
1373 			*nxtp = ip6e.ip6e_nxt;
1374 		off += (ip6e.ip6e_len + 2) << 2;
1375 		if (m->m_pkthdr.len < off)
1376 			return -1;
1377 		return off;
1378 
1379 	case IPPROTO_HOPOPTS:
1380 	case IPPROTO_ROUTING:
1381 	case IPPROTO_DSTOPTS:
1382 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1383 			return -1;
1384 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1385 		if (nxtp)
1386 			*nxtp = ip6e.ip6e_nxt;
1387 		off += (ip6e.ip6e_len + 1) << 3;
1388 		if (m->m_pkthdr.len < off)
1389 			return -1;
1390 		return off;
1391 
1392 	case IPPROTO_NONE:
1393 	case IPPROTO_ESP:
1394 	case IPPROTO_IPCOMP:
1395 		/* give up */
1396 		return -1;
1397 
1398 	default:
1399 		return -1;
1400 	}
1401 
1402 	return -1;
1403 }
1404 
1405 /*
1406  * get offset for the last header in the chain.  m will be kept untainted.
1407  */
1408 int
1409 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1410 {
1411 	int newoff;
1412 	int nxt;
1413 
1414 	if (!nxtp) {
1415 		nxt = -1;
1416 		nxtp = &nxt;
1417 	}
1418 	while (1) {
1419 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1420 		if (newoff < 0)
1421 			return off;
1422 		else if (newoff < off)
1423 			return -1;	/* invalid */
1424 		else if (newoff == off)
1425 			return newoff;
1426 
1427 		off = newoff;
1428 		proto = *nxtp;
1429 	}
1430 }
1431 
1432 /*
1433  * System control for IP6
1434  */
1435 
1436 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1437 	0,		0,		0,		0,
1438 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1439 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1440 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1441 	0,		0,		0,		0,
1442 	ENOPROTOOPT
1443 };
1444 
1445 #include <uvm/uvm_extern.h>
1446 #include <sys/sysctl.h>
1447 
1448 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1449 
1450 int
1451 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1452     void *newp, size_t newlen)
1453 {
1454 #ifdef MROUTING
1455 	extern int ip6_mrtproto;
1456 	extern struct mrt6stat mrt6stat;
1457 #endif
1458 
1459 	/* All sysctl names at this level are terminal. */
1460 	if (namelen != 1)
1461 		return ENOTDIR;
1462 
1463 	switch (name[0]) {
1464 	case IPV6CTL_KAME_VERSION:
1465 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1466 	case IPV6CTL_V6ONLY:
1467 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1468 	case IPV6CTL_DAD_PENDING:
1469 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1470 	case IPV6CTL_STATS:
1471 		if (newp != NULL)
1472 			return (EPERM);
1473 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1474 		    &ip6stat, sizeof(ip6stat)));
1475 	case IPV6CTL_MRTSTATS:
1476 #ifdef MROUTING
1477 		if (newp != NULL)
1478 			return (EPERM);
1479 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1480 		    &mrt6stat, sizeof(mrt6stat)));
1481 #else
1482 		return (EOPNOTSUPP);
1483 #endif
1484 	case IPV6CTL_MRTPROTO:
1485 #ifdef MROUTING
1486 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1487 #else
1488 		return (EOPNOTSUPP);
1489 #endif
1490 	default:
1491 		if (name[0] < IPV6CTL_MAXID)
1492 			return (sysctl_int_arr(ipv6ctl_vars, name, namelen,
1493 			    oldp, oldlenp, newp, newlen));
1494 		return (EOPNOTSUPP);
1495 	}
1496 	/* NOTREACHED */
1497 }
1498