xref: /freebsd/sys/netinet6/ip6_input.c (revision 2357939bc239bd5334a169b62313806178dd8f30)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei 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  * 4. 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 "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 #include "opt_pfil_hooks.h"
69 #include "opt_random_ip_id.h"
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/proc.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 
85 #include <net/if.h>
86 #include <net/if_types.h>
87 #include <net/if_dl.h>
88 #include <net/route.h>
89 #include <net/netisr.h>
90 #ifdef PFIL_HOOKS
91 #include <net/pfil.h>
92 #endif
93 
94 #include <netinet/in.h>
95 #include <netinet/in_systm.h>
96 #ifdef INET
97 #include <netinet/ip.h>
98 #include <netinet/ip_icmp.h>
99 #endif /* INET */
100 #include <netinet/ip6.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/icmp6.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_prefix.h>
109 
110 #ifdef IPSEC
111 #include <netinet6/ipsec.h>
112 #ifdef INET6
113 #include <netinet6/ipsec6.h>
114 #endif
115 #endif
116 
117 #ifdef FAST_IPSEC
118 #include <netipsec/ipsec.h>
119 #include <netipsec/ipsec6.h>
120 #define	IPSEC
121 #endif /* FAST_IPSEC */
122 
123 #include <netinet6/ip6_fw.h>
124 
125 #include <netinet6/ip6protosw.h>
126 
127 #include <net/net_osdep.h>
128 
129 extern struct domain inet6domain;
130 
131 u_char ip6_protox[IPPROTO_MAX];
132 static struct ifqueue ip6intrq;
133 static int ip6qmaxlen = IFQ_MAXLEN;
134 struct in6_ifaddr *in6_ifaddr;
135 
136 extern struct callout in6_tmpaddrtimer_ch;
137 
138 int ip6_forward_srcrt;			/* XXX */
139 int ip6_sourcecheck;			/* XXX */
140 int ip6_sourcecheck_interval;		/* XXX */
141 
142 int ip6_ours_check_algorithm;
143 
144 #ifdef PFIL_HOOKS
145 struct pfil_head inet6_pfil_hook;
146 #endif
147 
148 /* firewall hooks */
149 ip6_fw_chk_t *ip6_fw_chk_ptr;
150 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
151 int ip6_fw_enable = 1;
152 
153 struct ip6stat ip6stat;
154 
155 static void ip6_init2 __P((void *));
156 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
157 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
158 #ifdef PULLDOWN_TEST
159 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
160 #endif
161 
162 /*
163  * IP6 initialization: fill in IP6 protocol switch table.
164  * All protocols not implemented in kernel go to raw IP6 protocol handler.
165  */
166 void
167 ip6_init()
168 {
169 	struct ip6protosw *pr;
170 	int i;
171 
172 #ifdef DIAGNOSTIC
173 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
174 		panic("sizeof(protosw) != sizeof(ip6protosw)");
175 #endif
176 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
177 	if (pr == 0)
178 		panic("ip6_init");
179 	for (i = 0; i < IPPROTO_MAX; i++)
180 		ip6_protox[i] = pr - inet6sw;
181 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
182 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
183 		if (pr->pr_domain->dom_family == PF_INET6 &&
184 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
185 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
186 #ifdef PFIL_HOOKS
187 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
188 	inet6_pfil_hook.ph_af = AF_INET6;
189 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
190 		printf("%s: WARNING: unable to register pfil hook, "
191 			"error %d\n", __func__, i);
192 #endif /* PFIL_HOOKS */
193 	ip6intrq.ifq_maxlen = ip6qmaxlen;
194 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
195 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
196 	scope6_init();
197 	addrsel_policy_init();
198 	nd6_init();
199 	frag6_init();
200 #ifndef RANDOM_IP_ID
201 	ip6_flow_seq = arc4random();
202 #endif
203 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
204 }
205 
206 static void
207 ip6_init2(dummy)
208 	void *dummy;
209 {
210 
211 	/* nd6_timer_init */
212 	callout_init(&nd6_timer_ch, 0);
213 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
214 
215 	/* router renumbering prefix list maintenance */
216 	callout_init(&in6_rr_timer_ch, 0);
217 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
218 
219 	/* timer for regeneranation of temporary addresses randomize ID */
220 	callout_init(&in6_tmpaddrtimer_ch, 0);
221 	callout_reset(&in6_tmpaddrtimer_ch,
222 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
223 		       ip6_temp_regen_advance) * hz,
224 		      in6_tmpaddrtimer, NULL);
225 }
226 
227 /* cheat */
228 /* This must be after route_init(), which is now SI_ORDER_THIRD */
229 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
230 
231 extern struct	route_in6 ip6_forward_rt;
232 
233 void
234 ip6_input(m)
235 	struct mbuf *m;
236 {
237 	struct ip6_hdr *ip6;
238 	int off = sizeof(struct ip6_hdr), nest;
239 	u_int32_t plen;
240 	u_int32_t rtalert = ~0;
241 	int nxt, ours = 0;
242 	struct ifnet *deliverifp = NULL;
243 	struct sockaddr_in6 sa6;
244 	u_int32_t srczone, dstzone;
245 #ifdef PFIL_HOOKS
246 	struct in6_addr odst;
247 #endif
248 	int srcrt = 0;
249 
250 	GIANT_REQUIRED;			/* XXX for now */
251 #ifdef IPSEC
252 	/*
253 	 * should the inner packet be considered authentic?
254 	 * see comment in ah4_input().
255 	 */
256 	if (m) {
257 		m->m_flags &= ~M_AUTHIPHDR;
258 		m->m_flags &= ~M_AUTHIPDGM;
259 	}
260 #endif
261 
262 	/*
263 	 * make sure we don't have onion peering information into m_tag.
264 	 */
265 	ip6_delaux(m);
266 
267 	/*
268 	 * mbuf statistics
269 	 */
270 	if (m->m_flags & M_EXT) {
271 		if (m->m_next)
272 			ip6stat.ip6s_mext2m++;
273 		else
274 			ip6stat.ip6s_mext1++;
275 	} else {
276 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
277 		if (m->m_next) {
278 			if (m->m_flags & M_LOOP) {
279 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
280 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
281 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
282 			else
283 				ip6stat.ip6s_m2m[0]++;
284 		} else
285 			ip6stat.ip6s_m1++;
286 #undef M2MMAX
287 	}
288 
289 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
290 	ip6stat.ip6s_total++;
291 
292 #ifndef PULLDOWN_TEST
293 	/*
294 	 * L2 bridge code and some other code can return mbuf chain
295 	 * that does not conform to KAME requirement.  too bad.
296 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
297 	 */
298 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
299 		struct mbuf *n;
300 
301 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
302 		if (n)
303 			M_MOVE_PKTHDR(n, m);
304 		if (n && n->m_pkthdr.len > MHLEN) {
305 			MCLGET(n, M_DONTWAIT);
306 			if ((n->m_flags & M_EXT) == 0) {
307 				m_freem(n);
308 				n = NULL;
309 			}
310 		}
311 		if (n == NULL) {
312 			m_freem(m);
313 			return;	/* ENOBUFS */
314 		}
315 
316 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
317 		n->m_len = n->m_pkthdr.len;
318 		m_freem(m);
319 		m = n;
320 	}
321 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
322 #endif
323 
324 	if (m->m_len < sizeof(struct ip6_hdr)) {
325 		struct ifnet *inifp;
326 		inifp = m->m_pkthdr.rcvif;
327 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
328 			ip6stat.ip6s_toosmall++;
329 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
330 			return;
331 		}
332 	}
333 
334 	ip6 = mtod(m, struct ip6_hdr *);
335 
336 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
337 		ip6stat.ip6s_badvers++;
338 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
339 		goto bad;
340 	}
341 
342 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
343 
344 	/*
345 	 * Check against address spoofing/corruption.
346 	 */
347 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
348 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
349 		/*
350 		 * XXX: "badscope" is not very suitable for a multicast source.
351 		 */
352 		ip6stat.ip6s_badscope++;
353 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
354 		goto bad;
355 	}
356 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
357 	    !(m->m_flags & M_LOOP)) {
358 		/*
359 		 * In this case, the packet should come from the loopback
360 		 * interface.  However, we cannot just check the if_flags,
361 		 * because ip6_mloopback() passes the "actual" interface
362 		 * as the outgoing/incoming interface.
363 		 */
364 		ip6stat.ip6s_badscope++;
365 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
366 		goto bad;
367 	}
368 
369 	/*
370 	 * The following check is not documented in specs.  A malicious
371 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
372 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
373 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
374 	 *
375 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
376 	 * support IPv4-less kernel compilation, we cannot support SIIT
377 	 * environment at all.  So, it makes more sense for us to reject any
378 	 * malicious packets for non-SIIT environment, than try to do a
379 	 * partial support for SIIT environment.
380 	 */
381 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
382 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
383 		ip6stat.ip6s_badscope++;
384 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
385 		goto bad;
386 	}
387 #if 0
388 	/*
389 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
390 	 *
391 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
392 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
393 	 * is revised to forbid relaying case.
394 	 */
395 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
396 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
397 		ip6stat.ip6s_badscope++;
398 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
399 		goto bad;
400 	}
401 #endif
402 
403 	/*
404 	 * Drop packets if the link ID portion is already filled.
405 	 * XXX: this is technically not a good behavior.  But, we internally
406 	 * use the field to disambiguate link-local addresses, so we cannot
407 	 * be generous against those a bit strange addresses.
408 	 */
409 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
410 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
411 		    ip6->ip6_src.s6_addr16[1]) {
412 			ip6stat.ip6s_badscope++;
413 			goto bad;
414 		}
415 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
416 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
417 		    ip6->ip6_dst.s6_addr16[1]) {
418 			ip6stat.ip6s_badscope++;
419 			goto bad;
420 		}
421 	}
422 
423 #ifdef PFIL_HOOKS
424 	/*
425 	 * Run through list of hooks for input packets.
426 	 *
427 	 * NB: Beware of the destination address changing
428 	 *     (e.g. by NAT rewriting).  When this happens,
429 	 *     tell ip6_forward to do the right thing.
430 	 */
431 	odst = ip6->ip6_dst;
432 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
433 		return;
434 	if (m == NULL)			/* consumed by filter */
435 		return;
436 	ip6 = mtod(m, struct ip6_hdr *);
437 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
438 #endif /* PFIL_HOOKS */
439 
440 	/*
441 	 * Check with the firewall...
442 	 */
443 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
444 		u_short port = 0;
445 		/* If ipfw says divert, we have to just drop packet */
446 		/* use port as a dummy argument */
447 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
448 			m_freem(m);
449 			m = NULL;
450 		}
451 		if (!m)
452 			return;
453 	}
454 
455 	/*
456 	 * construct source and destination address structures with
457 	 * disambiguating their scope zones (if there is ambiguity).
458 	 * XXX: sin6_family and sin6_len will NOT be referred to, but we fill
459 	 * in these fields just in case.
460 	 */
461 	if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_src, &srczone) ||
462 	    in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_dst, &dstzone)) {
463 		/*
464 		 * Note that these generic checks cover cases that src or
465 		 * dst are the loopback address and the receiving interface
466 		 * is not loopback.
467 		 */
468 		ip6stat.ip6s_badscope++;
469 		goto bad;
470 	}
471 
472 	bzero(&sa6, sizeof(sa6));
473 	sa6.sin6_family = AF_INET6;
474 	sa6.sin6_len = sizeof(struct sockaddr_in6);
475 
476 	sa6.sin6_addr = ip6->ip6_src;
477 	sa6.sin6_scope_id = srczone;
478 	if (in6_embedscope(&ip6->ip6_src, &sa6, NULL, NULL)) {
479 		/* XXX: should not happen */
480 		ip6stat.ip6s_badscope++;
481 		goto bad;
482 	}
483 
484 	sa6.sin6_addr = ip6->ip6_dst;
485 	sa6.sin6_scope_id = dstzone;
486 	if (in6_embedscope(&ip6->ip6_dst, &sa6, NULL, NULL)) {
487 		/* XXX: should not happen */
488 		ip6stat.ip6s_badscope++;
489 		goto bad;
490 	}
491 
492 	/* XXX: ff01::%ifN awareness is not merged, yet. */
493 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_src))
494 		ip6->ip6_src.s6_addr16[1] = 0;
495 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst))
496 		ip6->ip6_dst.s6_addr16[1] = 0;
497 
498 	/*
499 	 * Multicast check
500 	 */
501 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
502 	  	struct in6_multi *in6m = 0;
503 
504 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
505 		/*
506 		 * See if we belong to the destination multicast group on the
507 		 * arrival interface.
508 		 */
509 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
510 		if (in6m)
511 			ours = 1;
512 		else if (!ip6_mrouter) {
513 			ip6stat.ip6s_notmember++;
514 			ip6stat.ip6s_cantforward++;
515 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
516 			goto bad;
517 		}
518 		deliverifp = m->m_pkthdr.rcvif;
519 		goto hbhcheck;
520 	}
521 
522 	/*
523 	 *  Unicast check
524 	 */
525 	if (ip6_forward_rt.ro_rt != NULL &&
526 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
527 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
528 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
529 		ip6stat.ip6s_forward_cachehit++;
530 	else {
531 		struct sockaddr_in6 *dst6;
532 
533 		if (ip6_forward_rt.ro_rt) {
534 			/* route is down or destination is different */
535 			ip6stat.ip6s_forward_cachemiss++;
536 			RTFREE(ip6_forward_rt.ro_rt);
537 			ip6_forward_rt.ro_rt = 0;
538 		}
539 
540 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
541 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
542 		dst6->sin6_len = sizeof(struct sockaddr_in6);
543 		dst6->sin6_family = AF_INET6;
544 		dst6->sin6_addr = ip6->ip6_dst;
545 
546 		rtalloc((struct route *)&ip6_forward_rt);
547 	}
548 
549 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
550 
551 	/*
552 	 * Accept the packet if the forwarding interface to the destination
553 	 * according to the routing table is the loopback interface,
554 	 * unless the associated route has a gateway.
555 	 * Note that this approach causes to accept a packet if there is a
556 	 * route to the loopback interface for the destination of the packet.
557 	 * But we think it's even useful in some situations, e.g. when using
558 	 * a special daemon which wants to intercept the packet.
559 	 *
560 	 * XXX: some OSes automatically make a cloned route for the destination
561 	 * of an outgoing packet.  If the outgoing interface of the packet
562 	 * is a loopback one, the kernel would consider the packet to be
563 	 * accepted, even if we have no such address assinged on the interface.
564 	 * We check the cloned flag of the route entry to reject such cases,
565 	 * assuming that route entries for our own addresses are not made by
566 	 * cloning (it should be true because in6_addloop explicitly installs
567 	 * the host route).  However, we might have to do an explicit check
568 	 * while it would be less efficient.  Or, should we rather install a
569 	 * reject route for such a case?
570 	 */
571 	if (ip6_forward_rt.ro_rt &&
572 	    (ip6_forward_rt.ro_rt->rt_flags &
573 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
574 #ifdef RTF_WASCLONED
575 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
576 #endif
577 #ifdef RTF_CLONED
578 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
579 #endif
580 #if 0
581 	    /*
582 	     * The check below is redundant since the comparison of
583 	     * the destination and the key of the rtentry has
584 	     * already done through looking up the routing table.
585 	     */
586 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
587 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
588 #endif
589 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
590 		struct in6_ifaddr *ia6 =
591 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
592 
593 		/*
594 		 * record address information into m_tag.
595 		 */
596 		(void)ip6_setdstifaddr(m, ia6);
597 
598 		/*
599 		 * packets to a tentative, duplicated, or somehow invalid
600 		 * address must not be accepted.
601 		 */
602 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
603 			/* this address is ready */
604 			ours = 1;
605 			deliverifp = ia6->ia_ifp;	/* correct? */
606 			/* Count the packet in the ip address stats */
607 			ia6->ia_ifa.if_ipackets++;
608 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
609 			goto hbhcheck;
610 		} else {
611 			/* address is not ready, so discard the packet. */
612 			nd6log((LOG_INFO,
613 			    "ip6_input: packet to an unready address %s->%s\n",
614 			    ip6_sprintf(&ip6->ip6_src),
615 			    ip6_sprintf(&ip6->ip6_dst)));
616 
617 			goto bad;
618 		}
619 	}
620 
621 	/*
622 	 * FAITH (Firewall Aided Internet Translator)
623 	 */
624 	if (ip6_keepfaith) {
625 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
626 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
627 			/* XXX do we need more sanity checks? */
628 			ours = 1;
629 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
630 			goto hbhcheck;
631 		}
632 	}
633 
634 	/*
635 	 * Now there is no reason to process the packet if it's not our own
636 	 * and we're not a router.
637 	 */
638 	if (!ip6_forwarding) {
639 		ip6stat.ip6s_cantforward++;
640 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
641 		goto bad;
642 	}
643 
644   hbhcheck:
645 	/*
646 	 * record address information into m_tag, if we don't have one yet.
647 	 * note that we are unable to record it, if the address is not listed
648 	 * as our interface address (e.g. multicast addresses, addresses
649 	 * within FAITH prefixes and such).
650 	 */
651 	if (deliverifp && !ip6_getdstifaddr(m)) {
652 		struct in6_ifaddr *ia6;
653 
654 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
655 		if (ia6) {
656 			if (!ip6_setdstifaddr(m, ia6)) {
657 				/*
658 				 * XXX maybe we should drop the packet here,
659 				 * as we could not provide enough information
660 				 * to the upper layers.
661 				 */
662 			}
663 		}
664 	}
665 
666 	/*
667 	 * Process Hop-by-Hop options header if it's contained.
668 	 * m may be modified in ip6_hopopts_input().
669 	 * If a JumboPayload option is included, plen will also be modified.
670 	 */
671 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
672 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
673 		struct ip6_hbh *hbh;
674 
675 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
676 #if 0	/*touches NULL pointer*/
677 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
678 #endif
679 			return;	/* m have already been freed */
680 		}
681 
682 		/* adjust pointer */
683 		ip6 = mtod(m, struct ip6_hdr *);
684 
685 		/*
686 		 * if the payload length field is 0 and the next header field
687 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
688 		 * option MUST be included.
689 		 */
690 		if (ip6->ip6_plen == 0 && plen == 0) {
691 			/*
692 			 * Note that if a valid jumbo payload option is
693 			 * contained, ip6_hopopts_input() must set a valid
694 			 * (non-zero) payload length to the variable plen.
695 			 */
696 			ip6stat.ip6s_badoptions++;
697 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
698 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
699 			icmp6_error(m, ICMP6_PARAM_PROB,
700 				    ICMP6_PARAMPROB_HEADER,
701 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
702 			return;
703 		}
704 #ifndef PULLDOWN_TEST
705 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
706 		hbh = (struct ip6_hbh *)(ip6 + 1);
707 #else
708 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
709 			sizeof(struct ip6_hbh));
710 		if (hbh == NULL) {
711 			ip6stat.ip6s_tooshort++;
712 			return;
713 		}
714 #endif
715 		nxt = hbh->ip6h_nxt;
716 
717 		/*
718 		 * accept the packet if a router alert option is included
719 		 * and we act as an IPv6 router.
720 		 */
721 		if (rtalert != ~0 && ip6_forwarding)
722 			ours = 1;
723 	} else
724 		nxt = ip6->ip6_nxt;
725 
726 	/*
727 	 * Check that the amount of data in the buffers
728 	 * is as at least much as the IPv6 header would have us expect.
729 	 * Trim mbufs if longer than we expect.
730 	 * Drop packet if shorter than we expect.
731 	 */
732 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
733 		ip6stat.ip6s_tooshort++;
734 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
735 		goto bad;
736 	}
737 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
738 		if (m->m_len == m->m_pkthdr.len) {
739 			m->m_len = sizeof(struct ip6_hdr) + plen;
740 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
741 		} else
742 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
743 	}
744 
745 	/*
746 	 * Forward if desirable.
747 	 */
748 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
749 		/*
750 		 * If we are acting as a multicast router, all
751 		 * incoming multicast packets are passed to the
752 		 * kernel-level multicast forwarding function.
753 		 * The packet is returned (relatively) intact; if
754 		 * ip6_mforward() returns a non-zero value, the packet
755 		 * must be discarded, else it may be accepted below.
756 		 */
757 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
758 			ip6stat.ip6s_cantforward++;
759 			m_freem(m);
760 			return;
761 		}
762 		if (!ours) {
763 			m_freem(m);
764 			return;
765 		}
766 	} else if (!ours) {
767 		ip6_forward(m, srcrt);
768 		return;
769 	}
770 
771 	ip6 = mtod(m, struct ip6_hdr *);
772 
773 	/*
774 	 * Malicious party may be able to use IPv4 mapped addr to confuse
775 	 * tcp/udp stack and bypass security checks (act as if it was from
776 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
777 	 *
778 	 * For SIIT end node behavior, you may want to disable the check.
779 	 * However, you will  become vulnerable to attacks using IPv4 mapped
780 	 * source.
781 	 */
782 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
783 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
784 		ip6stat.ip6s_badscope++;
785 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
786 		goto bad;
787 	}
788 
789 	/*
790 	 * Tell launch routine the next header
791 	 */
792 	ip6stat.ip6s_delivered++;
793 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
794 	nest = 0;
795 
796 	while (nxt != IPPROTO_DONE) {
797 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
798 			ip6stat.ip6s_toomanyhdr++;
799 			goto bad;
800 		}
801 
802 		/*
803 		 * protection against faulty packet - there should be
804 		 * more sanity checks in header chain processing.
805 		 */
806 		if (m->m_pkthdr.len < off) {
807 			ip6stat.ip6s_tooshort++;
808 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
809 			goto bad;
810 		}
811 
812 #ifdef IPSEC
813 		/*
814 		 * enforce IPsec policy checking if we are seeing last header.
815 		 * note that we do not visit this with protocols with pcb layer
816 		 * code - like udp/tcp/raw ip.
817 		 */
818 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
819 		    ipsec6_in_reject(m, NULL)) {
820 			ipsec6stat.in_polvio++;
821 			goto bad;
822 		}
823 #endif
824 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
825 	}
826 	return;
827  bad:
828 	m_freem(m);
829 }
830 
831 /*
832  * set/grab in6_ifaddr correspond to IPv6 destination address.
833  * XXX backward compatibility wrapper
834  */
835 static struct ip6aux *
836 ip6_setdstifaddr(m, ia6)
837 	struct mbuf *m;
838 	struct in6_ifaddr *ia6;
839 {
840 	struct ip6aux *ip6a;
841 
842 	ip6a = ip6_addaux(m);
843 	if (ip6a)
844 		ip6a->ip6a_dstia6 = ia6;
845 	return ip6a;	/* NULL if failed to set */
846 }
847 
848 struct in6_ifaddr *
849 ip6_getdstifaddr(m)
850 	struct mbuf *m;
851 {
852 	struct ip6aux *ip6a;
853 
854 	ip6a = ip6_findaux(m);
855 	if (ip6a)
856 		return ip6a->ip6a_dstia6;
857 	else
858 		return NULL;
859 }
860 
861 /*
862  * Hop-by-Hop options header processing. If a valid jumbo payload option is
863  * included, the real payload length will be stored in plenp.
864  */
865 static int
866 ip6_hopopts_input(plenp, rtalertp, mp, offp)
867 	u_int32_t *plenp;
868 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
869 	struct mbuf **mp;
870 	int *offp;
871 {
872 	struct mbuf *m = *mp;
873 	int off = *offp, hbhlen;
874 	struct ip6_hbh *hbh;
875 	u_int8_t *opt;
876 
877 	/* validation of the length of the header */
878 #ifndef PULLDOWN_TEST
879 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
880 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
881 	hbhlen = (hbh->ip6h_len + 1) << 3;
882 
883 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
884 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
885 #else
886 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
887 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
888 	if (hbh == NULL) {
889 		ip6stat.ip6s_tooshort++;
890 		return -1;
891 	}
892 	hbhlen = (hbh->ip6h_len + 1) << 3;
893 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
894 		hbhlen);
895 	if (hbh == NULL) {
896 		ip6stat.ip6s_tooshort++;
897 		return -1;
898 	}
899 #endif
900 	off += hbhlen;
901 	hbhlen -= sizeof(struct ip6_hbh);
902 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
903 
904 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
905 				hbhlen, rtalertp, plenp) < 0)
906 		return (-1);
907 
908 	*offp = off;
909 	*mp = m;
910 	return (0);
911 }
912 
913 /*
914  * Search header for all Hop-by-hop options and process each option.
915  * This function is separate from ip6_hopopts_input() in order to
916  * handle a case where the sending node itself process its hop-by-hop
917  * options header. In such a case, the function is called from ip6_output().
918  *
919  * The function assumes that hbh header is located right after the IPv6 header
920  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
921  * opthead + hbhlen is located in continuous memory region.
922  */
923 int
924 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
925 	struct mbuf *m;
926 	u_int8_t *opthead;
927 	int hbhlen;
928 	u_int32_t *rtalertp;
929 	u_int32_t *plenp;
930 {
931 	struct ip6_hdr *ip6;
932 	int optlen = 0;
933 	u_int8_t *opt = opthead;
934 	u_int16_t rtalert_val;
935 	u_int32_t jumboplen;
936 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
937 
938 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
939 		switch (*opt) {
940 		case IP6OPT_PAD1:
941 			optlen = 1;
942 			break;
943 		case IP6OPT_PADN:
944 			if (hbhlen < IP6OPT_MINLEN) {
945 				ip6stat.ip6s_toosmall++;
946 				goto bad;
947 			}
948 			optlen = *(opt + 1) + 2;
949 			break;
950 		case IP6OPT_ROUTER_ALERT:
951 			/* XXX may need check for alignment */
952 			if (hbhlen < IP6OPT_RTALERT_LEN) {
953 				ip6stat.ip6s_toosmall++;
954 				goto bad;
955 			}
956 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
957 				/* XXX stat */
958 				icmp6_error(m, ICMP6_PARAM_PROB,
959 				    ICMP6_PARAMPROB_HEADER,
960 				    erroff + opt + 1 - opthead);
961 				return (-1);
962 			}
963 			optlen = IP6OPT_RTALERT_LEN;
964 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
965 			*rtalertp = ntohs(rtalert_val);
966 			break;
967 		case IP6OPT_JUMBO:
968 			/* XXX may need check for alignment */
969 			if (hbhlen < IP6OPT_JUMBO_LEN) {
970 				ip6stat.ip6s_toosmall++;
971 				goto bad;
972 			}
973 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
974 				/* XXX stat */
975 				icmp6_error(m, ICMP6_PARAM_PROB,
976 				    ICMP6_PARAMPROB_HEADER,
977 				    erroff + opt + 1 - opthead);
978 				return (-1);
979 			}
980 			optlen = IP6OPT_JUMBO_LEN;
981 
982 			/*
983 			 * IPv6 packets that have non 0 payload length
984 			 * must not contain a jumbo payload option.
985 			 */
986 			ip6 = mtod(m, struct ip6_hdr *);
987 			if (ip6->ip6_plen) {
988 				ip6stat.ip6s_badoptions++;
989 				icmp6_error(m, ICMP6_PARAM_PROB,
990 				    ICMP6_PARAMPROB_HEADER,
991 				    erroff + opt - opthead);
992 				return (-1);
993 			}
994 
995 			/*
996 			 * We may see jumbolen in unaligned location, so
997 			 * we'd need to perform bcopy().
998 			 */
999 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1000 			jumboplen = (u_int32_t)htonl(jumboplen);
1001 
1002 #if 1
1003 			/*
1004 			 * if there are multiple jumbo payload options,
1005 			 * *plenp will be non-zero and the packet will be
1006 			 * rejected.
1007 			 * the behavior may need some debate in ipngwg -
1008 			 * multiple options does not make sense, however,
1009 			 * there's no explicit mention in specification.
1010 			 */
1011 			if (*plenp != 0) {
1012 				ip6stat.ip6s_badoptions++;
1013 				icmp6_error(m, ICMP6_PARAM_PROB,
1014 				    ICMP6_PARAMPROB_HEADER,
1015 				    erroff + opt + 2 - opthead);
1016 				return (-1);
1017 			}
1018 #endif
1019 
1020 			/*
1021 			 * jumbo payload length must be larger than 65535.
1022 			 */
1023 			if (jumboplen <= IPV6_MAXPACKET) {
1024 				ip6stat.ip6s_badoptions++;
1025 				icmp6_error(m, ICMP6_PARAM_PROB,
1026 				    ICMP6_PARAMPROB_HEADER,
1027 				    erroff + opt + 2 - opthead);
1028 				return (-1);
1029 			}
1030 			*plenp = jumboplen;
1031 
1032 			break;
1033 		default:		/* unknown option */
1034 			if (hbhlen < IP6OPT_MINLEN) {
1035 				ip6stat.ip6s_toosmall++;
1036 				goto bad;
1037 			}
1038 			optlen = ip6_unknown_opt(opt, m,
1039 			    erroff + opt - opthead);
1040 			if (optlen == -1)
1041 				return (-1);
1042 			optlen += 2;
1043 			break;
1044 		}
1045 	}
1046 
1047 	return (0);
1048 
1049   bad:
1050 	m_freem(m);
1051 	return (-1);
1052 }
1053 
1054 /*
1055  * Unknown option processing.
1056  * The third argument `off' is the offset from the IPv6 header to the option,
1057  * which is necessary if the IPv6 header the and option header and IPv6 header
1058  * is not continuous in order to return an ICMPv6 error.
1059  */
1060 int
1061 ip6_unknown_opt(optp, m, off)
1062 	u_int8_t *optp;
1063 	struct mbuf *m;
1064 	int off;
1065 {
1066 	struct ip6_hdr *ip6;
1067 
1068 	switch (IP6OPT_TYPE(*optp)) {
1069 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1070 		return ((int)*(optp + 1));
1071 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1072 		m_freem(m);
1073 		return (-1);
1074 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1075 		ip6stat.ip6s_badoptions++;
1076 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1077 		return (-1);
1078 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1079 		ip6stat.ip6s_badoptions++;
1080 		ip6 = mtod(m, struct ip6_hdr *);
1081 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1082 		    (m->m_flags & (M_BCAST|M_MCAST)))
1083 			m_freem(m);
1084 		else
1085 			icmp6_error(m, ICMP6_PARAM_PROB,
1086 				    ICMP6_PARAMPROB_OPTION, off);
1087 		return (-1);
1088 	}
1089 
1090 	m_freem(m);		/* XXX: NOTREACHED */
1091 	return (-1);
1092 }
1093 
1094 /*
1095  * Create the "control" list for this pcb.
1096  * The function will not modify mbuf chain at all.
1097  *
1098  * with KAME mbuf chain restriction:
1099  * The routine will be called from upper layer handlers like tcp6_input().
1100  * Thus the routine assumes that the caller (tcp6_input) have already
1101  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1102  * very first mbuf on the mbuf chain.
1103  */
1104 void
1105 ip6_savecontrol(in6p, m, mp)
1106 	struct inpcb *in6p;
1107 	struct mbuf *m, **mp;
1108 {
1109 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1110 	struct thread *td = curthread;	/* XXX */
1111 	int privileged = 0;
1112 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1113 
1114 	if (td && !suser(td))
1115 		privileged++;
1116 
1117 #ifdef SO_TIMESTAMP
1118 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1119 		struct timeval tv;
1120 
1121 		microtime(&tv);
1122 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1123 		    SCM_TIMESTAMP, SOL_SOCKET);
1124 		if (*mp)
1125 			mp = &(*mp)->m_next;
1126 	}
1127 #endif
1128 
1129 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1130 		return;
1131 
1132 	/* RFC 2292 sec. 5 */
1133 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1134 		struct in6_pktinfo pi6;
1135 
1136 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1137 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1138 		pi6.ipi6_ifindex =
1139 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1140 
1141 		*mp = sbcreatecontrol((caddr_t) &pi6,
1142 		    sizeof(struct in6_pktinfo),
1143 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1144 		if (*mp)
1145 			mp = &(*mp)->m_next;
1146 	}
1147 
1148 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1149 		int hlim = ip6->ip6_hlim & 0xff;
1150 
1151 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1152 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1153 		if (*mp)
1154 			mp = &(*mp)->m_next;
1155 	}
1156 
1157 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1158 		u_int32_t flowinfo;
1159 		int tclass;
1160 
1161 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1162 		flowinfo >>= 20;
1163 
1164 		tclass = flowinfo & 0xff;
1165 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1166 		    IPV6_TCLASS, IPPROTO_IPV6);
1167 		if (*mp)
1168 			mp = &(*mp)->m_next;
1169 	}
1170 
1171 	/*
1172 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1173 	 * for the option, but it might be too strict, since there might
1174 	 * be some hop-by-hop options which can be returned to normal user.
1175 	 * See RFC 2292 section 6.
1176 	 */
1177 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1178 #ifdef DIAGNOSTIC
1179 		if (!privileged)
1180 			panic("IN6P_HOPOPTS is set for unprivileged socket");
1181 #endif
1182 		/*
1183 		 * Check if a hop-by-hop options header is contatined in the
1184 		 * received packet, and if so, store the options as ancillary
1185 		 * data. Note that a hop-by-hop options header must be
1186 		 * just after the IPv6 header, which is assured through the
1187 		 * IPv6 input processing.
1188 		 */
1189 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1190 			struct ip6_hbh *hbh;
1191 			int hbhlen = 0;
1192 #ifdef PULLDOWN_TEST
1193 			struct mbuf *ext;
1194 #endif
1195 
1196 #ifndef PULLDOWN_TEST
1197 			hbh = (struct ip6_hbh *)(ip6 + 1);
1198 			hbhlen = (hbh->ip6h_len + 1) << 3;
1199 #else
1200 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1201 			    ip6->ip6_nxt);
1202 			if (ext == NULL) {
1203 				ip6stat.ip6s_tooshort++;
1204 				return;
1205 			}
1206 			hbh = mtod(ext, struct ip6_hbh *);
1207 			hbhlen = (hbh->ip6h_len + 1) << 3;
1208 			if (hbhlen != ext->m_len) {
1209 				m_freem(ext);
1210 				ip6stat.ip6s_tooshort++;
1211 				return;
1212 			}
1213 #endif
1214 
1215 			/*
1216 			 * XXX: We copy the whole header even if a
1217 			 * jumbo payload option is included, the option which
1218 			 * is to be removed before returning according to
1219 			 * RFC2292.
1220 			 * Note: this constraint is removed in 2292bis.
1221 			 */
1222 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1223 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1224 			    IPPROTO_IPV6);
1225 			if (*mp)
1226 				mp = &(*mp)->m_next;
1227 #ifdef PULLDOWN_TEST
1228 			m_freem(ext);
1229 #endif
1230 		}
1231 	}
1232 
1233 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1234 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1235 
1236 		/*
1237 		 * Search for destination options headers or routing
1238 		 * header(s) through the header chain, and stores each
1239 		 * header as ancillary data.
1240 		 * Note that the order of the headers remains in
1241 		 * the chain of ancillary data.
1242 		 */
1243 		while (1) {	/* is explicit loop prevention necessary? */
1244 			struct ip6_ext *ip6e = NULL;
1245 			int elen;
1246 #ifdef PULLDOWN_TEST
1247 			struct mbuf *ext = NULL;
1248 #endif
1249 
1250 			/*
1251 			 * if it is not an extension header, don't try to
1252 			 * pull it from the chain.
1253 			 */
1254 			switch (nxt) {
1255 			case IPPROTO_DSTOPTS:
1256 			case IPPROTO_ROUTING:
1257 			case IPPROTO_HOPOPTS:
1258 			case IPPROTO_AH: /* is it possible? */
1259 				break;
1260 			default:
1261 				goto loopend;
1262 			}
1263 
1264 #ifndef PULLDOWN_TEST
1265 			if (off + sizeof(*ip6e) > m->m_len)
1266 				goto loopend;
1267 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1268 			if (nxt == IPPROTO_AH)
1269 				elen = (ip6e->ip6e_len + 2) << 2;
1270 			else
1271 				elen = (ip6e->ip6e_len + 1) << 3;
1272 			if (off + elen > m->m_len)
1273 				goto loopend;
1274 #else
1275 			ext = ip6_pullexthdr(m, off, nxt);
1276 			if (ext == NULL) {
1277 				ip6stat.ip6s_tooshort++;
1278 				return;
1279 			}
1280 			ip6e = mtod(ext, struct ip6_ext *);
1281 			if (nxt == IPPROTO_AH)
1282 				elen = (ip6e->ip6e_len + 2) << 2;
1283 			else
1284 				elen = (ip6e->ip6e_len + 1) << 3;
1285 			if (elen != ext->m_len) {
1286 				m_freem(ext);
1287 				ip6stat.ip6s_tooshort++;
1288 				return;
1289 			}
1290 #endif
1291 
1292 			switch (nxt) {
1293 			case IPPROTO_DSTOPTS:
1294 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1295 					break;
1296 
1297 				/*
1298 				 * We also require super-user privilege for
1299 				 * the option.  See comments on IN6_HOPOPTS.
1300 				 */
1301 				if (!privileged)
1302 					break;
1303 
1304 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1305 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1306 				    IPPROTO_IPV6);
1307 				if (*mp)
1308 					mp = &(*mp)->m_next;
1309 				break;
1310 			case IPPROTO_ROUTING:
1311 				if (!in6p->in6p_flags & IN6P_RTHDR)
1312 					break;
1313 
1314 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1315 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1316 				    IPPROTO_IPV6);
1317 				if (*mp)
1318 					mp = &(*mp)->m_next;
1319 				break;
1320 			case IPPROTO_HOPOPTS:
1321 			case IPPROTO_AH: /* is it possible? */
1322 				break;
1323 
1324 			default:
1325 				/*
1326 			 	 * other cases have been filtered in the above.
1327 				 * none will visit this case.  here we supply
1328 				 * the code just in case (nxt overwritten or
1329 				 * other cases).
1330 				 */
1331 #ifdef PULLDOWN_TEST
1332 				m_freem(ext);
1333 #endif
1334 				goto loopend;
1335 
1336 			}
1337 
1338 			/* proceed with the next header. */
1339 			off += elen;
1340 			nxt = ip6e->ip6e_nxt;
1341 			ip6e = NULL;
1342 #ifdef PULLDOWN_TEST
1343 			m_freem(ext);
1344 			ext = NULL;
1345 #endif
1346 		}
1347 	  loopend:
1348 		;
1349 	}
1350 
1351 #undef IS2292
1352 }
1353 
1354 void
1355 ip6_notify_pmtu(in6p, dst, mtu)
1356 	struct inpcb *in6p;
1357 	struct sockaddr_in6 *dst;
1358 	u_int32_t *mtu;
1359 {
1360 	struct socket *so;
1361 	struct mbuf *m_mtu;
1362 	struct ip6_mtuinfo mtuctl;
1363 
1364 	so =  in6p->inp_socket;
1365 
1366 	if (mtu == NULL)
1367 		return;
1368 
1369 #ifdef DIAGNOSTIC
1370 	if (so == NULL)		/* I believe this is impossible */
1371 		panic("ip6_notify_pmtu: socket is NULL");
1372 #endif
1373 
1374 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1375 	mtuctl.ip6m_mtu = *mtu;
1376 	mtuctl.ip6m_addr = *dst;
1377 	in6_recoverscope(&mtuctl.ip6m_addr, &mtuctl.ip6m_addr.sin6_addr, NULL);
1378 
1379 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1380 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1381 		return;
1382 
1383 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1384 	    == 0) {
1385 		m_freem(m_mtu);
1386 		/* XXX: should count statistics */
1387 	} else
1388 		sorwakeup(so);
1389 
1390 	return;
1391 }
1392 
1393 #ifdef PULLDOWN_TEST
1394 /*
1395  * pull single extension header from mbuf chain.  returns single mbuf that
1396  * contains the result, or NULL on error.
1397  */
1398 static struct mbuf *
1399 ip6_pullexthdr(m, off, nxt)
1400 	struct mbuf *m;
1401 	size_t off;
1402 	int nxt;
1403 {
1404 	struct ip6_ext ip6e;
1405 	size_t elen;
1406 	struct mbuf *n;
1407 
1408 #ifdef DIAGNOSTIC
1409 	switch (nxt) {
1410 	case IPPROTO_DSTOPTS:
1411 	case IPPROTO_ROUTING:
1412 	case IPPROTO_HOPOPTS:
1413 	case IPPROTO_AH: /* is it possible? */
1414 		break;
1415 	default:
1416 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1417 	}
1418 #endif
1419 
1420 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1421 	if (nxt == IPPROTO_AH)
1422 		elen = (ip6e.ip6e_len + 2) << 2;
1423 	else
1424 		elen = (ip6e.ip6e_len + 1) << 3;
1425 
1426 	MGET(n, M_DONTWAIT, MT_DATA);
1427 	if (n && elen >= MLEN) {
1428 		MCLGET(n, M_DONTWAIT);
1429 		if ((n->m_flags & M_EXT) == 0) {
1430 			m_free(n);
1431 			n = NULL;
1432 		}
1433 	}
1434 	if (!n)
1435 		return NULL;
1436 
1437 	n->m_len = 0;
1438 	if (elen >= M_TRAILINGSPACE(n)) {
1439 		m_free(n);
1440 		return NULL;
1441 	}
1442 
1443 	m_copydata(m, off, elen, mtod(n, caddr_t));
1444 	n->m_len = elen;
1445 	return n;
1446 }
1447 #endif
1448 
1449 /*
1450  * Get pointer to the previous header followed by the header
1451  * currently processed.
1452  * XXX: This function supposes that
1453  *	M includes all headers,
1454  *	the next header field and the header length field of each header
1455  *	are valid, and
1456  *	the sum of each header length equals to OFF.
1457  * Because of these assumptions, this function must be called very
1458  * carefully. Moreover, it will not be used in the near future when
1459  * we develop `neater' mechanism to process extension headers.
1460  */
1461 char *
1462 ip6_get_prevhdr(m, off)
1463 	struct mbuf *m;
1464 	int off;
1465 {
1466 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1467 
1468 	if (off == sizeof(struct ip6_hdr))
1469 		return (&ip6->ip6_nxt);
1470 	else {
1471 		int len, nxt;
1472 		struct ip6_ext *ip6e = NULL;
1473 
1474 		nxt = ip6->ip6_nxt;
1475 		len = sizeof(struct ip6_hdr);
1476 		while (len < off) {
1477 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1478 
1479 			switch (nxt) {
1480 			case IPPROTO_FRAGMENT:
1481 				len += sizeof(struct ip6_frag);
1482 				break;
1483 			case IPPROTO_AH:
1484 				len += (ip6e->ip6e_len + 2) << 2;
1485 				break;
1486 			default:
1487 				len += (ip6e->ip6e_len + 1) << 3;
1488 				break;
1489 			}
1490 			nxt = ip6e->ip6e_nxt;
1491 		}
1492 		if (ip6e)
1493 			return (&ip6e->ip6e_nxt);
1494 		else
1495 			return NULL;
1496 	}
1497 }
1498 
1499 /*
1500  * get next header offset.  m will be retained.
1501  */
1502 int
1503 ip6_nexthdr(m, off, proto, nxtp)
1504 	struct mbuf *m;
1505 	int off;
1506 	int proto;
1507 	int *nxtp;
1508 {
1509 	struct ip6_hdr ip6;
1510 	struct ip6_ext ip6e;
1511 	struct ip6_frag fh;
1512 
1513 	/* just in case */
1514 	if (m == NULL)
1515 		panic("ip6_nexthdr: m == NULL");
1516 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1517 		return -1;
1518 
1519 	switch (proto) {
1520 	case IPPROTO_IPV6:
1521 		if (m->m_pkthdr.len < off + sizeof(ip6))
1522 			return -1;
1523 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1524 		if (nxtp)
1525 			*nxtp = ip6.ip6_nxt;
1526 		off += sizeof(ip6);
1527 		return off;
1528 
1529 	case IPPROTO_FRAGMENT:
1530 		/*
1531 		 * terminate parsing if it is not the first fragment,
1532 		 * it does not make sense to parse through it.
1533 		 */
1534 		if (m->m_pkthdr.len < off + sizeof(fh))
1535 			return -1;
1536 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1537 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1538 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1539 			return -1;
1540 		if (nxtp)
1541 			*nxtp = fh.ip6f_nxt;
1542 		off += sizeof(struct ip6_frag);
1543 		return off;
1544 
1545 	case IPPROTO_AH:
1546 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1547 			return -1;
1548 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1549 		if (nxtp)
1550 			*nxtp = ip6e.ip6e_nxt;
1551 		off += (ip6e.ip6e_len + 2) << 2;
1552 		return off;
1553 
1554 	case IPPROTO_HOPOPTS:
1555 	case IPPROTO_ROUTING:
1556 	case IPPROTO_DSTOPTS:
1557 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1558 			return -1;
1559 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1560 		if (nxtp)
1561 			*nxtp = ip6e.ip6e_nxt;
1562 		off += (ip6e.ip6e_len + 1) << 3;
1563 		return off;
1564 
1565 	case IPPROTO_NONE:
1566 	case IPPROTO_ESP:
1567 	case IPPROTO_IPCOMP:
1568 		/* give up */
1569 		return -1;
1570 
1571 	default:
1572 		return -1;
1573 	}
1574 
1575 	return -1;
1576 }
1577 
1578 /*
1579  * get offset for the last header in the chain.  m will be kept untainted.
1580  */
1581 int
1582 ip6_lasthdr(m, off, proto, nxtp)
1583 	struct mbuf *m;
1584 	int off;
1585 	int proto;
1586 	int *nxtp;
1587 {
1588 	int newoff;
1589 	int nxt;
1590 
1591 	if (!nxtp) {
1592 		nxt = -1;
1593 		nxtp = &nxt;
1594 	}
1595 	while (1) {
1596 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1597 		if (newoff < 0)
1598 			return off;
1599 		else if (newoff < off)
1600 			return -1;	/* invalid */
1601 		else if (newoff == off)
1602 			return newoff;
1603 
1604 		off = newoff;
1605 		proto = *nxtp;
1606 	}
1607 }
1608 
1609 struct ip6aux *
1610 ip6_addaux(m)
1611 	struct mbuf *m;
1612 {
1613 	struct m_tag *mtag;
1614 
1615 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1616 	if (!mtag) {
1617 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1618 		    M_NOWAIT);
1619 		if (mtag) {
1620 			m_tag_prepend(m, mtag);
1621 			bzero(mtag + 1, sizeof(struct ip6aux));
1622 		}
1623 	}
1624 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1625 }
1626 
1627 struct ip6aux *
1628 ip6_findaux(m)
1629 	struct mbuf *m;
1630 {
1631 	struct m_tag *mtag;
1632 
1633 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1634 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1635 }
1636 
1637 void
1638 ip6_delaux(m)
1639 	struct mbuf *m;
1640 {
1641 	struct m_tag *mtag;
1642 
1643 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1644 	if (mtag)
1645 		m_tag_delete(m, mtag);
1646 }
1647 
1648 /*
1649  * System control for IP6
1650  */
1651 
1652 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1653 	0,		0,		0,		0,
1654 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1655 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1656 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1657 	0,		0,		0,		0,
1658 	ENOPROTOOPT
1659 };
1660