xref: /dflybsd-src/sys/netinet6/ip6_input.c (revision c3c96e4421a1087a390825eac6c01c9ed9182387)
1 /*	$FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.15 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/ip6_input.c,v 1.38 2008/09/24 14:26:39 sephe Exp $	*/
3 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  */
68 
69 #include "opt_ip6fw.h"
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/syslog.h>
86 #include <sys/proc.h>
87 #include <sys/priv.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/if_dl.h>
92 #include <net/route.h>
93 #include <net/netisr.h>
94 #include <net/pfil.h>
95 
96 #include <sys/thread2.h>
97 #include <sys/msgport2.h>
98 #include <net/netmsg2.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_systm.h>
102 #ifdef INET
103 #include <netinet/ip.h>
104 #include <netinet/ip_icmp.h>
105 #endif /* INET */
106 #include <netinet/ip6.h>
107 #include <netinet6/in6_var.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/icmp6.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/in6_ifattach.h>
113 #include <netinet6/nd6.h>
114 #include <netinet6/in6_prefix.h>
115 
116 #ifdef IPSEC
117 #include <netinet6/ipsec.h>
118 #ifdef INET6
119 #include <netinet6/ipsec6.h>
120 #endif
121 #endif
122 
123 #ifdef FAST_IPSEC
124 #include <netproto/ipsec/ipsec.h>
125 #include <netproto/ipsec/ipsec6.h>
126 #define	IPSEC
127 #endif /* FAST_IPSEC */
128 
129 #include <net/ip6fw/ip6_fw.h>
130 
131 #include <netinet6/ip6protosw.h>
132 
133 /* we need it for NLOOP. */
134 #include "use_loop.h"
135 
136 #include <net/net_osdep.h>
137 
138 extern struct domain inet6domain;
139 extern struct ip6protosw inet6sw[];
140 
141 u_char ip6_protox[IPPROTO_MAX];
142 struct in6_ifaddr *in6_ifaddr;
143 
144 extern struct callout in6_tmpaddrtimer_ch;
145 
146 int ip6_forward_srcrt;			/* XXX */
147 int ip6_sourcecheck;			/* XXX */
148 int ip6_sourcecheck_interval;		/* XXX */
149 
150 int ip6_ours_check_algorithm;
151 
152 struct pfil_head inet6_pfil_hook;
153 
154 /* firewall hooks */
155 ip6_fw_chk_t *ip6_fw_chk_ptr;
156 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
157 int ip6_fw_enable = 1;
158 
159 struct ip6stat ip6stat;
160 
161 static void ip6_init2 (void *);
162 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
163 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
164 static void ip6_input(struct netmsg *msg);
165 #ifdef PULLDOWN_TEST
166 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
167 #endif
168 static void transport6_processing_handler(netmsg_t netmsg);
169 
170 /*
171  * IP6 initialization: fill in IP6 protocol switch table.
172  * All protocols not implemented in kernel go to raw IP6 protocol handler.
173  */
174 void
175 ip6_init(void)
176 {
177 	struct ip6protosw *pr;
178 	int i;
179 	struct timeval tv;
180 
181 #ifdef DIAGNOSTIC
182 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
183 		panic("sizeof(protosw) != sizeof(ip6protosw)");
184 #endif
185 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
186 	if (pr == 0)
187 		panic("ip6_init");
188 	for (i = 0; i < IPPROTO_MAX; i++)
189 		ip6_protox[i] = pr - inet6sw;
190 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
191 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
192 		if (pr->pr_domain->dom_family == PF_INET6 &&
193 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
194 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
195 
196 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
197 	inet6_pfil_hook.ph_af = AF_INET6;
198 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
199 		kprintf("%s: WARNING: unable to register pfil hook, "
200 			"error %d\n", __func__, i);
201 	}
202 
203 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX cpufn */
204 	scope6_init();
205 	addrsel_policy_init();
206 	nd6_init();
207 	frag6_init();
208 	/*
209 	 * in many cases, random() here does NOT return random number
210 	 * as initialization during bootstrap time occur in fixed order.
211 	 */
212 	microtime(&tv);
213 	ip6_flow_seq = krandom() ^ tv.tv_usec;
214 	microtime(&tv);
215 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
216 }
217 
218 static void
219 ip6_init2(void *dummy)
220 {
221 	/* nd6_timer_init */
222 	callout_init(&nd6_timer_ch);
223 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
224 
225 	/* router renumbering prefix list maintenance */
226 	callout_init(&in6_rr_timer_ch);
227 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
228 
229 	/* timer for regeneranation of temporary addresses randomize ID */
230 	callout_init(&in6_tmpaddrtimer_ch);
231 	callout_reset(&in6_tmpaddrtimer_ch,
232 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
233 		       ip6_temp_regen_advance) * hz,
234 		      in6_tmpaddrtimer, NULL);
235 }
236 
237 /* cheat */
238 /* This must be after route_init(), which is now SI_ORDER_THIRD */
239 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
240 
241 extern struct	route_in6 ip6_forward_rt;
242 
243 static
244 void
245 ip6_input(struct netmsg *msg)
246 {
247 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
248 	struct ip6_hdr *ip6;
249 	int off = sizeof(struct ip6_hdr), nest;
250 	u_int32_t plen;
251 	u_int32_t rtalert = ~0;
252 	int nxt, ours = 0, rh_present = 0;
253 	struct ifnet *deliverifp = NULL;
254 	struct in6_addr odst;
255 	int srcrt = 0;
256 
257 #ifdef IPSEC
258 	/*
259 	 * should the inner packet be considered authentic?
260 	 * see comment in ah4_input().
261 	 */
262 	if (m) {
263 		m->m_flags &= ~M_AUTHIPHDR;
264 		m->m_flags &= ~M_AUTHIPDGM;
265 	}
266 #endif
267 
268 	/*
269 	 * make sure we don't have onion peering information into m_aux.
270 	 */
271 	ip6_delaux(m);
272 
273 	/*
274 	 * mbuf statistics
275 	 */
276 	if (m->m_flags & M_EXT) {
277 		if (m->m_next)
278 			ip6stat.ip6s_mext2m++;
279 		else
280 			ip6stat.ip6s_mext1++;
281 	} else {
282 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
283 		if (m->m_next) {
284 			if (m->m_flags & M_LOOP) {
285 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
286 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
287 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
288 			else
289 				ip6stat.ip6s_m2m[0]++;
290 		} else
291 			ip6stat.ip6s_m1++;
292 #undef M2MMAX
293 	}
294 
295 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
296 	ip6stat.ip6s_total++;
297 
298 #ifndef PULLDOWN_TEST
299 	/*
300 	 * L2 bridge code and some other code can return mbuf chain
301 	 * that does not conform to KAME requirement.  too bad.
302 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
303 	 */
304 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
305 		struct mbuf *n;
306 
307 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
308 		if (n == NULL)
309 			goto bad;
310 		M_MOVE_PKTHDR(n, m);
311 
312 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
313 		n->m_len = n->m_pkthdr.len;
314 		m_freem(m);
315 		m = n;
316 	}
317 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
318 #endif
319 
320 	if (m->m_len < sizeof(struct ip6_hdr)) {
321 		struct ifnet *inifp;
322 		inifp = m->m_pkthdr.rcvif;
323 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
324 			ip6stat.ip6s_toosmall++;
325 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
326 			goto bad2;
327 		}
328 	}
329 
330 	ip6 = mtod(m, struct ip6_hdr *);
331 
332 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
333 		ip6stat.ip6s_badvers++;
334 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
335 		goto bad;
336 	}
337 
338 	/*
339 	 * Run through list of hooks for input packets.
340 	 *
341 	 * NB: Beware of the destination address changing
342 	 *     (e.g. by NAT rewriting). When this happens,
343 	 *     tell ip6_forward to do the right thing.
344 	 */
345 	if (pfil_has_hooks(&inet6_pfil_hook)) {
346 		odst = ip6->ip6_dst;
347 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
348 		    m->m_pkthdr.rcvif, PFIL_IN)) {
349 			goto bad2;
350 		}
351 		if (m == NULL)			/* consumed by filter */
352 			goto bad2;
353 		ip6 = mtod(m, struct ip6_hdr *);
354 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
355 	}
356 
357 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
358 
359 #ifdef ALTQ
360 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
361 		/* packet is dropped by traffic conditioner */
362 		return;
363 	}
364 #endif
365 
366 	/*
367 	 * Check with the firewall...
368 	 */
369 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
370 		u_short port = 0;
371 		/* If ipfw says divert, we have to just drop packet */
372 		/* use port as a dummy argument */
373 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
374 			m_freem(m);
375 			m = NULL;
376 		}
377 		if (!m)
378 			goto bad2;
379 	}
380 
381 	/*
382 	 * Check against address spoofing/corruption.
383 	 */
384 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
385 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
386 		/*
387 		 * XXX: "badscope" is not very suitable for a multicast source.
388 		 */
389 		ip6stat.ip6s_badscope++;
390 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
391 		goto bad;
392 	}
393 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
394 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
395 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
396 		ip6stat.ip6s_badscope++;
397 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
398 		goto bad;
399 	}
400 
401 	/*
402 	 * The following check is not documented in specs.  A malicious
403 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
404 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
405 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
406 	 *
407 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
408 	 * support IPv4-less kernel compilation, we cannot support SIIT
409 	 * environment at all.  So, it makes more sense for us to reject any
410 	 * malicious packets for non-SIIT environment, than try to do a
411 	 * partical support for SIIT environment.
412 	 */
413 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
414 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
415 		ip6stat.ip6s_badscope++;
416 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
417 		goto bad;
418 	}
419 #if 0
420 	/*
421 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
422 	 *
423 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
424 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
425 	 * is revised to forbid relaying case.
426 	 */
427 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
428 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
429 		ip6stat.ip6s_badscope++;
430 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
431 		goto bad;
432 	}
433 #endif
434 
435 	/* drop packets if interface ID portion is already filled */
436 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
437 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
438 		    ip6->ip6_src.s6_addr16[1]) {
439 			ip6stat.ip6s_badscope++;
440 			goto bad;
441 		}
442 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
443 		    ip6->ip6_dst.s6_addr16[1]) {
444 			ip6stat.ip6s_badscope++;
445 			goto bad;
446 		}
447 	}
448 
449 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
450 		ip6->ip6_src.s6_addr16[1]
451 			= htons(m->m_pkthdr.rcvif->if_index);
452 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
453 		ip6->ip6_dst.s6_addr16[1]
454 			= htons(m->m_pkthdr.rcvif->if_index);
455 
456 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
457 	/*
458 	 * We use rt->rt_ifp to determine if the address is ours or not.
459 	 * If rt_ifp is lo0, the address is ours.
460 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
461 	 * so any address under fe80::%lo0/64 will be mistakenly considered
462 	 * local.  The special case is supplied to handle the case properly
463 	 * by actually looking at interface addresses
464 	 * (using in6ifa_ifpwithaddr).
465 	 */
466 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
467 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
468 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
469 			icmp6_error(m, ICMP6_DST_UNREACH,
470 			    ICMP6_DST_UNREACH_ADDR, 0);
471 			/* m is already freed */
472 			goto bad2;
473 		}
474 
475 		ours = 1;
476 		deliverifp = m->m_pkthdr.rcvif;
477 		goto hbhcheck;
478 	}
479 #endif
480 
481 	/*
482 	 * Multicast check
483 	 */
484 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
485 	  	struct	in6_multi *in6m = 0;
486 
487 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
488 		/*
489 		 * See if we belong to the destination multicast group on the
490 		 * arrival interface.
491 		 */
492 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
493 		if (in6m)
494 			ours = 1;
495 		else if (!ip6_mrouter) {
496 			ip6stat.ip6s_notmember++;
497 			ip6stat.ip6s_cantforward++;
498 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
499 			goto bad;
500 		}
501 		deliverifp = m->m_pkthdr.rcvif;
502 		goto hbhcheck;
503 	}
504 
505 	/*
506 	 *  Unicast check
507 	 */
508 	switch (ip6_ours_check_algorithm) {
509 	default:
510 		/*
511 		 * XXX: I intentionally broke our indentation rule here,
512 		 *      since this switch-case is just for measurement and
513 		 *      therefore should soon be removed.
514 		 */
515 	if (ip6_forward_rt.ro_rt != NULL &&
516 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
517 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
518 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
519 		ip6stat.ip6s_forward_cachehit++;
520 	else {
521 		struct sockaddr_in6 *dst6;
522 
523 		if (ip6_forward_rt.ro_rt) {
524 			/* route is down or destination is different */
525 			ip6stat.ip6s_forward_cachemiss++;
526 			RTFREE(ip6_forward_rt.ro_rt);
527 			ip6_forward_rt.ro_rt = 0;
528 		}
529 
530 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
531 		dst6 = &ip6_forward_rt.ro_dst;
532 		dst6->sin6_len = sizeof(struct sockaddr_in6);
533 		dst6->sin6_family = AF_INET6;
534 		dst6->sin6_addr = ip6->ip6_dst;
535 
536 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
537 	}
538 
539 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
540 
541 	/*
542 	 * Accept the packet if the forwarding interface to the destination
543 	 * according to the routing table is the loopback interface,
544 	 * unless the associated route has a gateway.
545 	 * Note that this approach causes to accept a packet if there is a
546 	 * route to the loopback interface for the destination of the packet.
547 	 * But we think it's even useful in some situations, e.g. when using
548 	 * a special daemon which wants to intercept the packet.
549 	 *
550 	 * XXX: some OSes automatically make a cloned route for the destination
551 	 * of an outgoing packet.  If the outgoing interface of the packet
552 	 * is a loopback one, the kernel would consider the packet to be
553 	 * accepted, even if we have no such address assinged on the interface.
554 	 * We check the cloned flag of the route entry to reject such cases,
555 	 * assuming that route entries for our own addresses are not made by
556 	 * cloning (it should be true because in6_addloop explicitly installs
557 	 * the host route).  However, we might have to do an explicit check
558 	 * while it would be less efficient.  Or, should we rather install a
559 	 * reject route for such a case?
560 	 */
561 	if (ip6_forward_rt.ro_rt &&
562 	    (ip6_forward_rt.ro_rt->rt_flags &
563 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
564 #ifdef RTF_WASCLONED
565 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
566 #endif
567 #ifdef RTF_CLONED
568 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
569 #endif
570 #if 0
571 	    /*
572 	     * The check below is redundant since the comparison of
573 	     * the destination and the key of the rtentry has
574 	     * already done through looking up the routing table.
575 	     */
576 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
577 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
578 #endif
579 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
580 		struct in6_ifaddr *ia6 =
581 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
582 
583 		/*
584 		 * record address information into m_aux.
585 		 */
586 		ip6_setdstifaddr(m, ia6);
587 
588 		/*
589 		 * packets to a tentative, duplicated, or somehow invalid
590 		 * address must not be accepted.
591 		 */
592 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
593 			/* this address is ready */
594 			ours = 1;
595 			deliverifp = ia6->ia_ifp;	/* correct? */
596 			/* Count the packet in the ip address stats */
597 			ia6->ia_ifa.if_ipackets++;
598 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
599 			goto hbhcheck;
600 		} else {
601 			/* address is not ready, so discard the packet. */
602 			nd6log((LOG_INFO,
603 			    "ip6_input: packet to an unready address %s->%s\n",
604 			    ip6_sprintf(&ip6->ip6_src),
605 			    ip6_sprintf(&ip6->ip6_dst)));
606 
607 			goto bad;
608 		}
609 	}
610 	} /* XXX indentation (see above) */
611 
612 	/*
613 	 * FAITH(Firewall Aided Internet Translator)
614 	 */
615 	if (ip6_keepfaith) {
616 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
617 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
618 			/* XXX do we need more sanity checks? */
619 			ours = 1;
620 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
621 			goto hbhcheck;
622 		}
623 	}
624 
625 	/*
626 	 * Now there is no reason to process the packet if it's not our own
627 	 * and we're not a router.
628 	 */
629 	if (!ip6_forwarding) {
630 		ip6stat.ip6s_cantforward++;
631 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
632 		goto bad;
633 	}
634 
635 hbhcheck:
636 	/*
637 	 * record address information into m_aux, if we don't have one yet.
638 	 * note that we are unable to record it, if the address is not listed
639 	 * as our interface address (e.g. multicast addresses, addresses
640 	 * within FAITH prefixes and such).
641 	 */
642 	if (deliverifp && !ip6_getdstifaddr(m)) {
643 		struct in6_ifaddr *ia6;
644 
645 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
646 		if (ia6) {
647 			if (!ip6_setdstifaddr(m, ia6)) {
648 				/*
649 				 * XXX maybe we should drop the packet here,
650 				 * as we could not provide enough information
651 				 * to the upper layers.
652 				 */
653 			}
654 		}
655 	}
656 
657 	/*
658 	 * Process Hop-by-Hop options header if it's contained.
659 	 * m may be modified in ip6_hopopts_input().
660 	 * If a JumboPayload option is included, plen will also be modified.
661 	 */
662 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
663 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
664 		struct ip6_hbh *hbh;
665 
666 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
667 #if 0	/*touches NULL pointer*/
668 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
669 #endif
670 			goto bad2;	/* m have already been freed */
671 		}
672 
673 		/* adjust pointer */
674 		ip6 = mtod(m, struct ip6_hdr *);
675 
676 		/*
677 		 * if the payload length field is 0 and the next header field
678 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
679 		 * option MUST be included.
680 		 */
681 		if (ip6->ip6_plen == 0 && plen == 0) {
682 			/*
683 			 * Note that if a valid jumbo payload option is
684 			 * contained, ip6_hoptops_input() must set a valid
685 			 * (non-zero) payload length to the variable plen.
686 			 */
687 			ip6stat.ip6s_badoptions++;
688 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
689 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
690 			icmp6_error(m, ICMP6_PARAM_PROB,
691 				    ICMP6_PARAMPROB_HEADER,
692 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
693 			goto bad2;
694 		}
695 #ifndef PULLDOWN_TEST
696 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
697 		hbh = (struct ip6_hbh *)(ip6 + 1);
698 #else
699 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
700 			sizeof(struct ip6_hbh));
701 		if (hbh == NULL) {
702 			ip6stat.ip6s_tooshort++;
703 			goto bad2;
704 		}
705 #endif
706 		nxt = hbh->ip6h_nxt;
707 
708 		/*
709 		 * If we are acting as a router and the packet contains a
710 		 * router alert option, see if we know the option value.
711 		 * Currently, we only support the option value for MLD, in which
712 		 * case we should pass the packet to the multicast routing
713 		 * daemon.
714 		 */
715 		if (rtalert != ~0 && ip6_forwarding) {
716 			switch (rtalert) {
717 			case IP6OPT_RTALERT_MLD:
718 				ours = 1;
719 				break;
720 			default:
721 				/*
722 				 * RFC2711 requires unrecognized values must be
723 				 * silently ignored.
724 				 */
725 				break;
726 			}
727 		}
728 	} else
729 		nxt = ip6->ip6_nxt;
730 
731 	/*
732 	 * Check that the amount of data in the buffers
733 	 * is as at least much as the IPv6 header would have us expect.
734 	 * Trim mbufs if longer than we expect.
735 	 * Drop packet if shorter than we expect.
736 	 */
737 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
738 		ip6stat.ip6s_tooshort++;
739 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
740 		goto bad;
741 	}
742 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
743 		if (m->m_len == m->m_pkthdr.len) {
744 			m->m_len = sizeof(struct ip6_hdr) + plen;
745 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
746 		} else
747 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
748 	}
749 
750 	/*
751 	 * Forward if desirable.
752 	 */
753 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
754 		/*
755 		 * If we are acting as a multicast router, all
756 		 * incoming multicast packets are passed to the
757 		 * kernel-level multicast forwarding function.
758 		 * The packet is returned (relatively) intact; if
759 		 * ip6_mforward() returns a non-zero value, the packet
760 		 * must be discarded, else it may be accepted below.
761 		 */
762 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
763 			ip6stat.ip6s_cantforward++;
764 			goto bad;
765 		}
766 		if (!ours)
767 			goto bad;
768 	} else if (!ours) {
769 		ip6_forward(m, srcrt);
770 		goto bad2;
771 	}
772 
773 	ip6 = mtod(m, struct ip6_hdr *);
774 
775 	/*
776 	 * Malicious party may be able to use IPv4 mapped addr to confuse
777 	 * tcp/udp stack and bypass security checks (act as if it was from
778 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
779 	 *
780 	 * For SIIT end node behavior, you may want to disable the check.
781 	 * However, you will  become vulnerable to attacks using IPv4 mapped
782 	 * source.
783 	 */
784 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
785 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
786 		ip6stat.ip6s_badscope++;
787 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
788 		goto bad;
789 	}
790 
791 	/*
792 	 * Tell launch routine the next header
793 	 */
794 	ip6stat.ip6s_delivered++;
795 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
796 	nest = 0;
797 
798 	rh_present = 0;
799 	while (nxt != IPPROTO_DONE) {
800 		struct ip6protosw *sw6;
801 
802 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
803 			ip6stat.ip6s_toomanyhdr++;
804 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
805 			goto bad;
806 		}
807 
808 		/*
809 		 * protection against faulty packet - there should be
810 		 * more sanity checks in header chain processing.
811 		 */
812 		if (m->m_pkthdr.len < off) {
813 			ip6stat.ip6s_tooshort++;
814 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
815 			goto bad;
816 		}
817 
818 #if 0
819 		/*
820 		 * do we need to do it for every header?  yeah, other
821 		 * functions can play with it (like re-allocate and copy).
822 		 */
823 		mhist = ip6_addaux(m);
824 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
825 			hist = mtod(mhist, caddr_t) + mhist->m_len;
826 			bcopy(&nxt, hist, sizeof(nxt));
827 			mhist->m_len += sizeof(nxt);
828 		} else {
829 			ip6stat.ip6s_toomanyhdr++;
830 			goto bad;
831 		}
832 #endif
833 
834 		if (nxt == IPPROTO_ROUTING) {
835 			if (rh_present++) {
836 				in6_ifstat_inc(m->m_pkthdr.rcvif,
837 				    ifs6_in_hdrerr);
838 				ip6stat.ip6s_badoptions++;
839 				goto bad;
840 			}
841 		}
842 
843 		sw6 = &inet6sw[ip6_protox[nxt]];
844 #ifdef IPSEC
845 		/*
846 		 * enforce IPsec policy checking if we are seeing last header.
847 		 * note that we do not visit this with protocols with pcb layer
848 		 * code - like udp/tcp/raw ip.
849 		 */
850 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
851 			ipsec6stat.in_polvio++;
852 			goto bad;
853 		}
854 #endif
855 		/*
856 		 * If this is a terminal header forward to the port, otherwise
857 		 * process synchronously for more headers.
858 		 */
859 		if (sw6->pr_flags & PR_LASTHDR) {
860 			struct netmsg_packet *pmsg;
861 			lwkt_port_t port;
862 
863 			port = sw6->pr_soport(NULL, NULL, &m);
864 			KKASSERT(port != NULL);
865 			pmsg = &m->m_hdr.mh_netmsg;
866 			netmsg_init(&pmsg->nm_netmsg, NULL,
867 				    &netisr_apanic_rport,
868 				    0, transport6_processing_handler);
869 			pmsg->nm_packet = m;
870 			pmsg->nm_nxt = nxt;
871 			pmsg->nm_netmsg.nm_lmsg.u.ms_result = off;
872 			lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
873 			/* done with m */
874 			nxt = IPPROTO_DONE;
875 		} else {
876 			nxt = sw6->pr_input(&m, &off, nxt);
877 		}
878 	}
879 	goto bad2;
880 bad:
881 	m_freem(m);
882 bad2:
883 	;
884 	/* msg was embedded in the mbuf, do not reply! */
885 }
886 
887 /*
888  * We have to call the pr_input() function from the correct protocol
889  * thread.  The sw6->pr_soport() request at the end of ip6_input()
890  * returns the port and we forward a netmsg to the port to execute
891  * this function.
892  */
893 static void
894 transport6_processing_handler(netmsg_t netmsg)
895 {
896 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
897 	struct ip6protosw *sw6;
898 	int hlen;
899 	int nxt;
900 
901 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
902 	hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
903 
904 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
905 	KKASSERT(nxt == IPPROTO_DONE);
906 	/* netmsg was embedded in the mbuf, do not reply! */
907 }
908 
909 /*
910  * set/grab in6_ifaddr correspond to IPv6 destination address.
911  * XXX backward compatibility wrapper
912  */
913 static struct ip6aux *
914 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
915 {
916 	struct ip6aux *n;
917 
918 	n = ip6_addaux(m);
919 	if (n)
920 		n->ip6a_dstia6 = ia6;
921 	return n;	/* NULL if failed to set */
922 }
923 
924 struct in6_ifaddr *
925 ip6_getdstifaddr(struct mbuf *m)
926 {
927 	struct ip6aux *n;
928 
929 	n = ip6_findaux(m);
930 	if (n)
931 		return n->ip6a_dstia6;
932 	else
933 		return NULL;
934 }
935 
936 /*
937  * Hop-by-Hop options header processing. If a valid jumbo payload option is
938  * included, the real payload length will be stored in plenp.
939  */
940 static int
941 ip6_hopopts_input(u_int32_t *plenp,
942 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
943 		  struct mbuf **mp,
944 		  int *offp)
945 {
946 	struct mbuf *m = *mp;
947 	int off = *offp, hbhlen;
948 	struct ip6_hbh *hbh;
949 	u_int8_t *opt;
950 
951 	/* validation of the length of the header */
952 #ifndef PULLDOWN_TEST
953 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
954 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
955 	hbhlen = (hbh->ip6h_len + 1) << 3;
956 
957 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
958 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
959 #else
960 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
961 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
962 	if (hbh == NULL) {
963 		ip6stat.ip6s_tooshort++;
964 		return -1;
965 	}
966 	hbhlen = (hbh->ip6h_len + 1) << 3;
967 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
968 		hbhlen);
969 	if (hbh == NULL) {
970 		ip6stat.ip6s_tooshort++;
971 		return -1;
972 	}
973 #endif
974 	off += hbhlen;
975 	hbhlen -= sizeof(struct ip6_hbh);
976 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
977 
978 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
979 				hbhlen, rtalertp, plenp) < 0)
980 		return (-1);
981 
982 	*offp = off;
983 	*mp = m;
984 	return (0);
985 }
986 
987 /*
988  * Search header for all Hop-by-hop options and process each option.
989  * This function is separate from ip6_hopopts_input() in order to
990  * handle a case where the sending node itself process its hop-by-hop
991  * options header. In such a case, the function is called from ip6_output().
992  *
993  * The function assumes that hbh header is located right after the IPv6 header
994  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
995  * opthead + hbhlen is located in continuous memory region.
996  */
997 int
998 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
999 		    u_int32_t *rtalertp, u_int32_t *plenp)
1000 {
1001 	struct ip6_hdr *ip6;
1002 	int optlen = 0;
1003 	u_int8_t *opt = opthead;
1004 	u_int16_t rtalert_val;
1005 	u_int32_t jumboplen;
1006 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1007 
1008 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1009 		switch (*opt) {
1010 		case IP6OPT_PAD1:
1011 			optlen = 1;
1012 			break;
1013 		case IP6OPT_PADN:
1014 			if (hbhlen < IP6OPT_MINLEN) {
1015 				ip6stat.ip6s_toosmall++;
1016 				goto bad;
1017 			}
1018 			optlen = *(opt + 1) + 2;
1019 			break;
1020 		case IP6OPT_RTALERT:
1021 			/* XXX may need check for alignment */
1022 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1023 				ip6stat.ip6s_toosmall++;
1024 				goto bad;
1025 			}
1026 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1027 				/* XXX stat */
1028 				icmp6_error(m, ICMP6_PARAM_PROB,
1029 					    ICMP6_PARAMPROB_HEADER,
1030 					    erroff + opt + 1 - opthead);
1031 				return (-1);
1032 			}
1033 			optlen = IP6OPT_RTALERT_LEN;
1034 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1035 			*rtalertp = ntohs(rtalert_val);
1036 			break;
1037 		case IP6OPT_JUMBO:
1038 			/* XXX may need check for alignment */
1039 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1040 				ip6stat.ip6s_toosmall++;
1041 				goto bad;
1042 			}
1043 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1044 				/* XXX stat */
1045 				icmp6_error(m, ICMP6_PARAM_PROB,
1046 					    ICMP6_PARAMPROB_HEADER,
1047 					    erroff + opt + 1 - opthead);
1048 				return (-1);
1049 			}
1050 			optlen = IP6OPT_JUMBO_LEN;
1051 
1052 			/*
1053 			 * IPv6 packets that have non 0 payload length
1054 			 * must not contain a jumbo payload option.
1055 			 */
1056 			ip6 = mtod(m, struct ip6_hdr *);
1057 			if (ip6->ip6_plen) {
1058 				ip6stat.ip6s_badoptions++;
1059 				icmp6_error(m, ICMP6_PARAM_PROB,
1060 					    ICMP6_PARAMPROB_HEADER,
1061 					    erroff + opt - opthead);
1062 				return (-1);
1063 			}
1064 
1065 			/*
1066 			 * We may see jumbolen in unaligned location, so
1067 			 * we'd need to perform bcopy().
1068 			 */
1069 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1070 			jumboplen = (u_int32_t)htonl(jumboplen);
1071 
1072 #if 1
1073 			/*
1074 			 * if there are multiple jumbo payload options,
1075 			 * *plenp will be non-zero and the packet will be
1076 			 * rejected.
1077 			 * the behavior may need some debate in ipngwg -
1078 			 * multiple options does not make sense, however,
1079 			 * there's no explicit mention in specification.
1080 			 */
1081 			if (*plenp != 0) {
1082 				ip6stat.ip6s_badoptions++;
1083 				icmp6_error(m, ICMP6_PARAM_PROB,
1084 					    ICMP6_PARAMPROB_HEADER,
1085 					    erroff + opt + 2 - opthead);
1086 				return (-1);
1087 			}
1088 #endif
1089 
1090 			/*
1091 			 * jumbo payload length must be larger than 65535.
1092 			 */
1093 			if (jumboplen <= IPV6_MAXPACKET) {
1094 				ip6stat.ip6s_badoptions++;
1095 				icmp6_error(m, ICMP6_PARAM_PROB,
1096 					    ICMP6_PARAMPROB_HEADER,
1097 					    erroff + opt + 2 - opthead);
1098 				return (-1);
1099 			}
1100 			*plenp = jumboplen;
1101 
1102 			break;
1103 		default:		/* unknown option */
1104 			if (hbhlen < IP6OPT_MINLEN) {
1105 				ip6stat.ip6s_toosmall++;
1106 				goto bad;
1107 			}
1108 			optlen = ip6_unknown_opt(opt, m,
1109 			    erroff + opt - opthead);
1110 			if (optlen == -1)
1111 				return (-1);
1112 			optlen += 2;
1113 			break;
1114 		}
1115 	}
1116 
1117 	return (0);
1118 
1119 bad:
1120 	m_freem(m);
1121 	return (-1);
1122 }
1123 
1124 /*
1125  * Unknown option processing.
1126  * The third argument `off' is the offset from the IPv6 header to the option,
1127  * which is necessary if the IPv6 header the and option header and IPv6 header
1128  * is not continuous in order to return an ICMPv6 error.
1129  */
1130 int
1131 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1132 {
1133 	struct ip6_hdr *ip6;
1134 
1135 	switch (IP6OPT_TYPE(*optp)) {
1136 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1137 		return ((int)*(optp + 1));
1138 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1139 		m_freem(m);
1140 		return (-1);
1141 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1142 		ip6stat.ip6s_badoptions++;
1143 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1144 		return (-1);
1145 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1146 		ip6stat.ip6s_badoptions++;
1147 		ip6 = mtod(m, struct ip6_hdr *);
1148 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1149 		    (m->m_flags & (M_BCAST|M_MCAST)))
1150 			m_freem(m);
1151 		else
1152 			icmp6_error(m, ICMP6_PARAM_PROB,
1153 				    ICMP6_PARAMPROB_OPTION, off);
1154 		return (-1);
1155 	}
1156 
1157 	m_freem(m);		/* XXX: NOTREACHED */
1158 	return (-1);
1159 }
1160 
1161 /*
1162  * Create the "control" list for this pcb.
1163  * The function will not modify mbuf chain at all.
1164  *
1165  * with KAME mbuf chain restriction:
1166  * The routine will be called from upper layer handlers like tcp6_input().
1167  * Thus the routine assumes that the caller (tcp6_input) have already
1168  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1169  * very first mbuf on the mbuf chain.
1170  */
1171 void
1172 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1173 		struct mbuf *m)
1174 {
1175 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1176 	struct thread *td = curthread;	/* XXX */
1177 	int privileged = 0;
1178 	int rthdr_exist = 0;
1179 
1180 
1181 	if (priv_check(td, PRIV_ROOT) == 0)
1182  		privileged++;
1183 
1184 #ifdef SO_TIMESTAMP
1185 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1186 		struct timeval tv;
1187 
1188 		microtime(&tv);
1189 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1190 				      SCM_TIMESTAMP, SOL_SOCKET);
1191 		if (*mp) {
1192 			mp = &(*mp)->m_next;
1193 		}
1194 	}
1195 #endif
1196 
1197 	/* RFC 2292 sec. 5 */
1198 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1199 		struct in6_pktinfo pi6;
1200 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1201 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1202 			pi6.ipi6_addr.s6_addr16[1] = 0;
1203 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1204 					? m->m_pkthdr.rcvif->if_index
1205 					: 0;
1206 		*mp = sbcreatecontrol((caddr_t) &pi6,
1207 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1208 			IPPROTO_IPV6);
1209 		if (*mp)
1210 			mp = &(*mp)->m_next;
1211 	}
1212 
1213 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1214 		int hlim = ip6->ip6_hlim & 0xff;
1215 		*mp = sbcreatecontrol((caddr_t) &hlim,
1216 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1217 		if (*mp)
1218 			mp = &(*mp)->m_next;
1219 	}
1220 
1221 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1222 		u_int32_t flowinfo;
1223 		int tclass;
1224 
1225 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1226 		flowinfo >>= 20;
1227 
1228 		tclass = flowinfo & 0xff;
1229 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1230 		    IPV6_TCLASS, IPPROTO_IPV6);
1231 		if (*mp)
1232 			mp = &(*mp)->m_next;
1233 	}
1234 
1235 	/*
1236 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1237 	 * for the option, but it might be too strict, since there might
1238 	 * be some hop-by-hop options which can be returned to normal user.
1239 	 * See RFC 2292 section 6.
1240 	 */
1241 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1242 		/*
1243 		 * Check if a hop-by-hop options header is contatined in the
1244 		 * received packet, and if so, store the options as ancillary
1245 		 * data. Note that a hop-by-hop options header must be
1246 		 * just after the IPv6 header, which fact is assured through
1247 		 * the IPv6 input processing.
1248 		 */
1249 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1250 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1251 			struct ip6_hbh *hbh;
1252 			int hbhlen = 0;
1253 #ifdef PULLDOWN_TEST
1254 			struct mbuf *ext;
1255 #endif
1256 
1257 #ifndef PULLDOWN_TEST
1258 			hbh = (struct ip6_hbh *)(ip6 + 1);
1259 			hbhlen = (hbh->ip6h_len + 1) << 3;
1260 #else
1261 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1262 			    ip6->ip6_nxt);
1263 			if (ext == NULL) {
1264 				ip6stat.ip6s_tooshort++;
1265 				return;
1266 			}
1267 			hbh = mtod(ext, struct ip6_hbh *);
1268 			hbhlen = (hbh->ip6h_len + 1) << 3;
1269 			if (hbhlen != ext->m_len) {
1270 				m_freem(ext);
1271 				ip6stat.ip6s_tooshort++;
1272 				return;
1273 			}
1274 #endif
1275 
1276 			/*
1277 			 * XXX: We copy whole the header even if a jumbo
1278 			 * payload option is included, which option is to
1279 			 * be removed before returning in the RFC 2292.
1280 			 * Note: this constraint is removed in 2292bis.
1281 			 */
1282 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1283 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1284 			if (*mp)
1285 				mp = &(*mp)->m_next;
1286 #ifdef PULLDOWN_TEST
1287 			m_freem(ext);
1288 #endif
1289 		}
1290 	}
1291 
1292 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1293 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1294 		int proto, off, nxt;
1295 
1296 		/*
1297 		 * go through the header chain to see if a routing header is
1298 		 * contained in the packet. We need this information to store
1299 		 * destination options headers (if any) properly.
1300 		 * XXX: performance issue. We should record this info when
1301 		 * processing extension headers in incoming routine.
1302 		 * (todo) use m_aux?
1303 		 */
1304 		proto = IPPROTO_IPV6;
1305 		off = 0;
1306 		nxt = -1;
1307 		while (1) {
1308 			int newoff;
1309 
1310 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1311 			if (newoff < 0)
1312 				break;
1313 			if (newoff < off) /* invalid, check for safety */
1314 				break;
1315 			if ((proto = nxt) == IPPROTO_ROUTING) {
1316 				rthdr_exist = 1;
1317 				break;
1318 			}
1319 			off = newoff;
1320 		}
1321 	}
1322 
1323 	if ((in6p->in6p_flags &
1324 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1325 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1326 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1327 
1328 		/*
1329 		 * Search for destination options headers or routing
1330 		 * header(s) through the header chain, and stores each
1331 		 * header as ancillary data.
1332 		 * Note that the order of the headers remains in
1333 		 * the chain of ancillary data.
1334 		 */
1335 		while (1) {	/* is explicit loop prevention necessary? */
1336 			struct ip6_ext *ip6e = NULL;
1337 			int elen;
1338 #ifdef PULLDOWN_TEST
1339 			struct mbuf *ext = NULL;
1340 #endif
1341 
1342 			/*
1343 			 * if it is not an extension header, don't try to
1344 			 * pull it from the chain.
1345 			 */
1346 			switch (nxt) {
1347 			case IPPROTO_DSTOPTS:
1348 			case IPPROTO_ROUTING:
1349 			case IPPROTO_HOPOPTS:
1350 			case IPPROTO_AH: /* is it possible? */
1351 				break;
1352 			default:
1353 				goto loopend;
1354 			}
1355 
1356 #ifndef PULLDOWN_TEST
1357 			if (off + sizeof(*ip6e) > m->m_len)
1358 				goto loopend;
1359 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1360 			if (nxt == IPPROTO_AH)
1361 				elen = (ip6e->ip6e_len + 2) << 2;
1362 			else
1363 				elen = (ip6e->ip6e_len + 1) << 3;
1364 			if (off + elen > m->m_len)
1365 				goto loopend;
1366 #else
1367 			ext = ip6_pullexthdr(m, off, nxt);
1368 			if (ext == NULL) {
1369 				ip6stat.ip6s_tooshort++;
1370 				return;
1371 			}
1372 			ip6e = mtod(ext, struct ip6_ext *);
1373 			if (nxt == IPPROTO_AH)
1374 				elen = (ip6e->ip6e_len + 2) << 2;
1375 			else
1376 				elen = (ip6e->ip6e_len + 1) << 3;
1377 			if (elen != ext->m_len) {
1378 				m_freem(ext);
1379 				ip6stat.ip6s_tooshort++;
1380 				return;
1381 			}
1382 #endif
1383 
1384 			switch (nxt) {
1385 			case IPPROTO_DSTOPTS:
1386 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1387 					break;
1388 
1389 				/*
1390 				 * We also require super-user privilege for
1391 				 * the option.
1392 				 * See the comments on IN6_HOPOPTS.
1393 				 */
1394 				if (!privileged)
1395 					break;
1396 
1397 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1398 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1399 						      IPPROTO_IPV6);
1400 				if (*mp)
1401 					mp = &(*mp)->m_next;
1402 				break;
1403 			case IPPROTO_ROUTING:
1404 				if (!in6p->in6p_flags & IN6P_RTHDR)
1405 					break;
1406 
1407 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1408 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1409 						      IPPROTO_IPV6);
1410 				if (*mp)
1411 					mp = &(*mp)->m_next;
1412 				break;
1413 			case IPPROTO_HOPOPTS:
1414 			case IPPROTO_AH: /* is it possible? */
1415 				break;
1416 
1417 			default:
1418 				/*
1419 			 	 * other cases have been filtered in the above.
1420 				 * none will visit this case.  here we supply
1421 				 * the code just in case (nxt overwritten or
1422 				 * other cases).
1423 				 */
1424 #ifdef PULLDOWN_TEST
1425 				m_freem(ext);
1426 #endif
1427 				goto loopend;
1428 
1429 			}
1430 
1431 			/* proceed with the next header. */
1432 			off += elen;
1433 			nxt = ip6e->ip6e_nxt;
1434 			ip6e = NULL;
1435 #ifdef PULLDOWN_TEST
1436 			m_freem(ext);
1437 			ext = NULL;
1438 #endif
1439 		}
1440 	  loopend:
1441 		;
1442 	}
1443 #undef IS2292
1444 }
1445 
1446 void
1447 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1448 {
1449 	struct socket *so;
1450 	struct mbuf *m_mtu;
1451 	struct ip6_mtuinfo mtuctl;
1452 
1453 	so =  in6p->inp_socket;
1454 
1455 	if (mtu == NULL)
1456 		return;
1457 
1458 #ifdef DIAGNOSTIC
1459 	if (so == NULL)		/* I believe this is impossible */
1460 		panic("ip6_notify_pmtu: socket is NULL");
1461 #endif
1462 
1463 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1464 	mtuctl.ip6m_mtu = *mtu;
1465 	mtuctl.ip6m_addr = *dst;
1466 
1467 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1468 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1469 		return;
1470 
1471 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1472 	    == 0) {
1473 		m_freem(m_mtu);
1474 		/* XXX: should count statistics */
1475 	} else
1476 		sorwakeup(so);
1477 
1478 	return;
1479 }
1480 
1481 #ifdef PULLDOWN_TEST
1482 /*
1483  * pull single extension header from mbuf chain.  returns single mbuf that
1484  * contains the result, or NULL on error.
1485  */
1486 static struct mbuf *
1487 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1488 {
1489 	struct ip6_ext ip6e;
1490 	size_t elen;
1491 	struct mbuf *n;
1492 
1493 #ifdef DIAGNOSTIC
1494 	switch (nxt) {
1495 	case IPPROTO_DSTOPTS:
1496 	case IPPROTO_ROUTING:
1497 	case IPPROTO_HOPOPTS:
1498 	case IPPROTO_AH: /* is it possible? */
1499 		break;
1500 	default:
1501 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1502 	}
1503 #endif
1504 
1505 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1506 	if (nxt == IPPROTO_AH)
1507 		elen = (ip6e.ip6e_len + 2) << 2;
1508 	else
1509 		elen = (ip6e.ip6e_len + 1) << 3;
1510 
1511 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1512 	if (n == NULL)
1513 		return NULL;
1514 	n->m_len = 0;
1515 
1516 	if (elen >= M_TRAILINGSPACE(n)) {
1517 		m_free(n);
1518 		return NULL;
1519 	}
1520 
1521 	m_copydata(m, off, elen, mtod(n, caddr_t));
1522 	n->m_len = elen;
1523 	return n;
1524 }
1525 #endif
1526 
1527 /*
1528  * Get pointer to the previous header followed by the header
1529  * currently processed.
1530  * XXX: This function supposes that
1531  *	M includes all headers,
1532  *	the next header field and the header length field of each header
1533  *	are valid, and
1534  *	the sum of each header length equals to OFF.
1535  * Because of these assumptions, this function must be called very
1536  * carefully. Moreover, it will not be used in the near future when
1537  * we develop `neater' mechanism to process extension headers.
1538  */
1539 char *
1540 ip6_get_prevhdr(struct mbuf *m, int off)
1541 {
1542 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1543 
1544 	if (off == sizeof(struct ip6_hdr))
1545 		return (&ip6->ip6_nxt);
1546 	else {
1547 		int len, nxt;
1548 		struct ip6_ext *ip6e = NULL;
1549 
1550 		nxt = ip6->ip6_nxt;
1551 		len = sizeof(struct ip6_hdr);
1552 		while (len < off) {
1553 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1554 
1555 			switch (nxt) {
1556 			case IPPROTO_FRAGMENT:
1557 				len += sizeof(struct ip6_frag);
1558 				break;
1559 			case IPPROTO_AH:
1560 				len += (ip6e->ip6e_len + 2) << 2;
1561 				break;
1562 			default:
1563 				len += (ip6e->ip6e_len + 1) << 3;
1564 				break;
1565 			}
1566 			nxt = ip6e->ip6e_nxt;
1567 		}
1568 		if (ip6e)
1569 			return (&ip6e->ip6e_nxt);
1570 		else
1571 			return NULL;
1572 	}
1573 }
1574 
1575 /*
1576  * get next header offset.  m will be retained.
1577  */
1578 int
1579 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1580 {
1581 	struct ip6_hdr ip6;
1582 	struct ip6_ext ip6e;
1583 	struct ip6_frag fh;
1584 
1585 	/* just in case */
1586 	if (m == NULL)
1587 		panic("ip6_nexthdr: m == NULL");
1588 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1589 		return -1;
1590 
1591 	switch (proto) {
1592 	case IPPROTO_IPV6:
1593 		if (m->m_pkthdr.len < off + sizeof(ip6))
1594 			return -1;
1595 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1596 		if (nxtp)
1597 			*nxtp = ip6.ip6_nxt;
1598 		off += sizeof(ip6);
1599 		return off;
1600 
1601 	case IPPROTO_FRAGMENT:
1602 		/*
1603 		 * terminate parsing if it is not the first fragment,
1604 		 * it does not make sense to parse through it.
1605 		 */
1606 		if (m->m_pkthdr.len < off + sizeof(fh))
1607 			return -1;
1608 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1609 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1610 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1611 			return -1;
1612 		if (nxtp)
1613 			*nxtp = fh.ip6f_nxt;
1614 		off += sizeof(struct ip6_frag);
1615 		return off;
1616 
1617 	case IPPROTO_AH:
1618 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1619 			return -1;
1620 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1621 		if (nxtp)
1622 			*nxtp = ip6e.ip6e_nxt;
1623 		off += (ip6e.ip6e_len + 2) << 2;
1624 		return off;
1625 
1626 	case IPPROTO_HOPOPTS:
1627 	case IPPROTO_ROUTING:
1628 	case IPPROTO_DSTOPTS:
1629 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1630 			return -1;
1631 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1632 		if (nxtp)
1633 			*nxtp = ip6e.ip6e_nxt;
1634 		off += (ip6e.ip6e_len + 1) << 3;
1635 		return off;
1636 
1637 	case IPPROTO_NONE:
1638 	case IPPROTO_ESP:
1639 	case IPPROTO_IPCOMP:
1640 		/* give up */
1641 		return -1;
1642 
1643 	default:
1644 		return -1;
1645 	}
1646 
1647 	return -1;
1648 }
1649 
1650 /*
1651  * get offset for the last header in the chain.  m will be kept untainted.
1652  */
1653 int
1654 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1655 {
1656 	int newoff;
1657 	int nxt;
1658 
1659 	if (!nxtp) {
1660 		nxt = -1;
1661 		nxtp = &nxt;
1662 	}
1663 	while (1) {
1664 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1665 		if (newoff < 0)
1666 			return off;
1667 		else if (newoff < off)
1668 			return -1;	/* invalid */
1669 		else if (newoff == off)
1670 			return newoff;
1671 
1672 		off = newoff;
1673 		proto = *nxtp;
1674 	}
1675 }
1676 
1677 struct ip6aux *
1678 ip6_addaux(struct mbuf *m)
1679 {
1680 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1681 	if (!tag) {
1682 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1683 				sizeof (struct ip6aux),
1684 				MB_DONTWAIT);
1685 		if (tag)
1686 			m_tag_prepend(m, tag);
1687 	}
1688 	if (tag)
1689 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1690 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1691 }
1692 
1693 struct ip6aux *
1694 ip6_findaux(struct mbuf *m)
1695 {
1696 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1697 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1698 }
1699 
1700 void
1701 ip6_delaux(struct mbuf *m)
1702 {
1703 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1704 	if (tag)
1705 		m_tag_delete(m, tag);
1706 }
1707 
1708 /*
1709  * System control for IP6
1710  */
1711 
1712 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1713 	0,		0,		0,		0,
1714 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1715 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1716 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1717 	0,		0,		0,		0,
1718 	ENOPROTOOPT
1719 };
1720