xref: /dflybsd-src/sys/netinet6/ip6_input.c (revision 88abd8b5763f2e5d4b4db5c5dc1b5bb4c489698b)
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 protosw 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(netmsg_t 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 protosw *pr;
178 	int i;
179 	struct timeval tv;
180 
181 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
182 	if (pr == 0)
183 		panic("ip6_init");
184 	for (i = 0; i < IPPROTO_MAX; i++)
185 		ip6_protox[i] = pr - inet6sw;
186 	for (pr = inet6domain.dom_protosw;
187 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
188 		if (pr->pr_domain->dom_family == PF_INET6 &&
189 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
190 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
191 
192 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
193 	inet6_pfil_hook.ph_af = AF_INET6;
194 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
195 		kprintf("%s: WARNING: unable to register pfil hook, "
196 			"error %d\n", __func__, i);
197 	}
198 
199 	netisr_register(NETISR_IPV6, ip6_input, NULL); /* XXX cpufn */
200 	scope6_init();
201 	addrsel_policy_init();
202 	nd6_init();
203 	frag6_init();
204 	/*
205 	 * in many cases, random() here does NOT return random number
206 	 * as initialization during bootstrap time occur in fixed order.
207 	 */
208 	microtime(&tv);
209 	ip6_flow_seq = krandom() ^ tv.tv_usec;
210 	microtime(&tv);
211 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
212 }
213 
214 static void
215 ip6_init2(void *dummy)
216 {
217 	/* nd6_timer_init */
218 	callout_init(&nd6_timer_ch);
219 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
220 
221 	/* router renumbering prefix list maintenance */
222 	callout_init(&in6_rr_timer_ch);
223 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
224 
225 	/* timer for regeneranation of temporary addresses randomize ID */
226 	callout_init(&in6_tmpaddrtimer_ch);
227 	callout_reset(&in6_tmpaddrtimer_ch,
228 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
229 		       ip6_temp_regen_advance) * hz,
230 		      in6_tmpaddrtimer, NULL);
231 }
232 
233 /* cheat */
234 /* This must be after route_init(), which is now SI_ORDER_THIRD */
235 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
236 
237 extern struct	route_in6 ip6_forward_rt;
238 
239 static
240 void
241 ip6_input(netmsg_t msg)
242 {
243 	struct mbuf *m = msg->packet.nm_packet;
244 	struct ip6_hdr *ip6;
245 	int off = sizeof(struct ip6_hdr), nest;
246 	u_int32_t plen;
247 	u_int32_t rtalert = ~0;
248 	int nxt, ours = 0, rh_present = 0;
249 	struct ifnet *deliverifp = NULL;
250 	struct in6_addr odst;
251 	int srcrt = 0;
252 
253 #ifdef IPSEC
254 	/*
255 	 * should the inner packet be considered authentic?
256 	 * see comment in ah4_input().
257 	 */
258 	if (m) {
259 		m->m_flags &= ~M_AUTHIPHDR;
260 		m->m_flags &= ~M_AUTHIPDGM;
261 	}
262 #endif
263 
264 	/*
265 	 * make sure we don't have onion peering information into m_aux.
266 	 */
267 	ip6_delaux(m);
268 
269 	/*
270 	 * mbuf statistics
271 	 */
272 	if (m->m_flags & M_EXT) {
273 		if (m->m_next)
274 			ip6stat.ip6s_mext2m++;
275 		else
276 			ip6stat.ip6s_mext1++;
277 	} else {
278 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
279 		if (m->m_next) {
280 			if (m->m_flags & M_LOOP) {
281 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
282 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
283 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
284 			else
285 				ip6stat.ip6s_m2m[0]++;
286 		} else
287 			ip6stat.ip6s_m1++;
288 #undef M2MMAX
289 	}
290 
291 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
292 	ip6stat.ip6s_total++;
293 
294 #ifndef PULLDOWN_TEST
295 	/*
296 	 * L2 bridge code and some other code can return mbuf chain
297 	 * that does not conform to KAME requirement.  too bad.
298 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
299 	 */
300 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
301 		struct mbuf *n;
302 
303 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
304 		if (n == NULL)
305 			goto bad;
306 		M_MOVE_PKTHDR(n, m);
307 
308 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
309 		n->m_len = n->m_pkthdr.len;
310 		m_freem(m);
311 		m = n;
312 	}
313 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
314 #endif
315 
316 	if (m->m_len < sizeof(struct ip6_hdr)) {
317 		struct ifnet *inifp;
318 		inifp = m->m_pkthdr.rcvif;
319 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
320 			ip6stat.ip6s_toosmall++;
321 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
322 			goto bad2;
323 		}
324 	}
325 
326 	ip6 = mtod(m, struct ip6_hdr *);
327 
328 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
329 		ip6stat.ip6s_badvers++;
330 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
331 		goto bad;
332 	}
333 
334 	/*
335 	 * Run through list of hooks for input packets.
336 	 *
337 	 * NB: Beware of the destination address changing
338 	 *     (e.g. by NAT rewriting). When this happens,
339 	 *     tell ip6_forward to do the right thing.
340 	 */
341 	if (pfil_has_hooks(&inet6_pfil_hook)) {
342 		odst = ip6->ip6_dst;
343 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
344 		    m->m_pkthdr.rcvif, PFIL_IN)) {
345 			goto bad2;
346 		}
347 		if (m == NULL)			/* consumed by filter */
348 			goto bad2;
349 		ip6 = mtod(m, struct ip6_hdr *);
350 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
351 	}
352 
353 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
354 
355 #ifdef ALTQ
356 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
357 		/* packet is dropped by traffic conditioner */
358 		return;
359 	}
360 #endif
361 
362 	/*
363 	 * Check with the firewall...
364 	 */
365 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
366 		u_short port = 0;
367 		/* If ipfw says divert, we have to just drop packet */
368 		/* use port as a dummy argument */
369 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
370 			m_freem(m);
371 			m = NULL;
372 		}
373 		if (!m)
374 			goto bad2;
375 	}
376 
377 	/*
378 	 * Check against address spoofing/corruption.
379 	 */
380 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
381 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
382 		/*
383 		 * XXX: "badscope" is not very suitable for a multicast source.
384 		 */
385 		ip6stat.ip6s_badscope++;
386 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
387 		goto bad;
388 	}
389 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
390 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
391 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
392 		ip6stat.ip6s_badscope++;
393 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
394 		goto bad;
395 	}
396 
397 	/*
398 	 * The following check is not documented in specs.  A malicious
399 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
400 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
401 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
402 	 *
403 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
404 	 * support IPv4-less kernel compilation, we cannot support SIIT
405 	 * environment at all.  So, it makes more sense for us to reject any
406 	 * malicious packets for non-SIIT environment, than try to do a
407 	 * partical support for SIIT environment.
408 	 */
409 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
410 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
411 		ip6stat.ip6s_badscope++;
412 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
413 		goto bad;
414 	}
415 #if 0
416 	/*
417 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
418 	 *
419 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
420 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
421 	 * is revised to forbid relaying case.
422 	 */
423 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
424 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
425 		ip6stat.ip6s_badscope++;
426 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
427 		goto bad;
428 	}
429 #endif
430 
431 	/* drop packets if interface ID portion is already filled */
432 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
433 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
434 		    ip6->ip6_src.s6_addr16[1]) {
435 			ip6stat.ip6s_badscope++;
436 			goto bad;
437 		}
438 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
439 		    ip6->ip6_dst.s6_addr16[1]) {
440 			ip6stat.ip6s_badscope++;
441 			goto bad;
442 		}
443 	}
444 
445 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
446 		ip6->ip6_src.s6_addr16[1]
447 			= htons(m->m_pkthdr.rcvif->if_index);
448 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
449 		ip6->ip6_dst.s6_addr16[1]
450 			= htons(m->m_pkthdr.rcvif->if_index);
451 
452 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
453 	/*
454 	 * We use rt->rt_ifp to determine if the address is ours or not.
455 	 * If rt_ifp is lo0, the address is ours.
456 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
457 	 * so any address under fe80::%lo0/64 will be mistakenly considered
458 	 * local.  The special case is supplied to handle the case properly
459 	 * by actually looking at interface addresses
460 	 * (using in6ifa_ifpwithaddr).
461 	 */
462 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
463 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
464 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
465 			icmp6_error(m, ICMP6_DST_UNREACH,
466 			    ICMP6_DST_UNREACH_ADDR, 0);
467 			/* m is already freed */
468 			goto bad2;
469 		}
470 
471 		ours = 1;
472 		deliverifp = m->m_pkthdr.rcvif;
473 		goto hbhcheck;
474 	}
475 #endif
476 
477 	/*
478 	 * Multicast check
479 	 */
480 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
481 	  	struct	in6_multi *in6m = 0;
482 
483 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
484 		/*
485 		 * See if we belong to the destination multicast group on the
486 		 * arrival interface.
487 		 */
488 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
489 		if (in6m)
490 			ours = 1;
491 		else if (!ip6_mrouter) {
492 			ip6stat.ip6s_notmember++;
493 			ip6stat.ip6s_cantforward++;
494 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
495 			goto bad;
496 		}
497 		deliverifp = m->m_pkthdr.rcvif;
498 		goto hbhcheck;
499 	}
500 
501 	/*
502 	 *  Unicast check
503 	 */
504 	switch (ip6_ours_check_algorithm) {
505 	default:
506 		/*
507 		 * XXX: I intentionally broke our indentation rule here,
508 		 *      since this switch-case is just for measurement and
509 		 *      therefore should soon be removed.
510 		 */
511 	if (ip6_forward_rt.ro_rt != NULL &&
512 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
513 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
514 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
515 		ip6stat.ip6s_forward_cachehit++;
516 	else {
517 		struct sockaddr_in6 *dst6;
518 
519 		if (ip6_forward_rt.ro_rt) {
520 			/* route is down or destination is different */
521 			ip6stat.ip6s_forward_cachemiss++;
522 			RTFREE(ip6_forward_rt.ro_rt);
523 			ip6_forward_rt.ro_rt = 0;
524 		}
525 
526 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
527 		dst6 = &ip6_forward_rt.ro_dst;
528 		dst6->sin6_len = sizeof(struct sockaddr_in6);
529 		dst6->sin6_family = AF_INET6;
530 		dst6->sin6_addr = ip6->ip6_dst;
531 
532 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
533 	}
534 
535 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
536 
537 	/*
538 	 * Accept the packet if the forwarding interface to the destination
539 	 * according to the routing table is the loopback interface,
540 	 * unless the associated route has a gateway.
541 	 * Note that this approach causes to accept a packet if there is a
542 	 * route to the loopback interface for the destination of the packet.
543 	 * But we think it's even useful in some situations, e.g. when using
544 	 * a special daemon which wants to intercept the packet.
545 	 *
546 	 * XXX: some OSes automatically make a cloned route for the destination
547 	 * of an outgoing packet.  If the outgoing interface of the packet
548 	 * is a loopback one, the kernel would consider the packet to be
549 	 * accepted, even if we have no such address assinged on the interface.
550 	 * We check the cloned flag of the route entry to reject such cases,
551 	 * assuming that route entries for our own addresses are not made by
552 	 * cloning (it should be true because in6_addloop explicitly installs
553 	 * the host route).  However, we might have to do an explicit check
554 	 * while it would be less efficient.  Or, should we rather install a
555 	 * reject route for such a case?
556 	 */
557 	if (ip6_forward_rt.ro_rt &&
558 	    (ip6_forward_rt.ro_rt->rt_flags &
559 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
560 #ifdef RTF_WASCLONED
561 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
562 #endif
563 #ifdef RTF_CLONED
564 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
565 #endif
566 #if 0
567 	    /*
568 	     * The check below is redundant since the comparison of
569 	     * the destination and the key of the rtentry has
570 	     * already done through looking up the routing table.
571 	     */
572 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
573 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
574 #endif
575 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
576 		struct in6_ifaddr *ia6 =
577 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
578 
579 		/*
580 		 * record address information into m_aux.
581 		 */
582 		ip6_setdstifaddr(m, ia6);
583 
584 		/*
585 		 * packets to a tentative, duplicated, or somehow invalid
586 		 * address must not be accepted.
587 		 */
588 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
589 			/* this address is ready */
590 			ours = 1;
591 			deliverifp = ia6->ia_ifp;	/* correct? */
592 			/* Count the packet in the ip address stats */
593 			ia6->ia_ifa.if_ipackets++;
594 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
595 			goto hbhcheck;
596 		} else {
597 			/* address is not ready, so discard the packet. */
598 			nd6log((LOG_INFO,
599 			    "ip6_input: packet to an unready address %s->%s\n",
600 			    ip6_sprintf(&ip6->ip6_src),
601 			    ip6_sprintf(&ip6->ip6_dst)));
602 
603 			goto bad;
604 		}
605 	}
606 	} /* XXX indentation (see above) */
607 
608 	/*
609 	 * FAITH(Firewall Aided Internet Translator)
610 	 */
611 	if (ip6_keepfaith) {
612 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
613 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
614 			/* XXX do we need more sanity checks? */
615 			ours = 1;
616 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
617 			goto hbhcheck;
618 		}
619 	}
620 
621 	/*
622 	 * Now there is no reason to process the packet if it's not our own
623 	 * and we're not a router.
624 	 */
625 	if (!ip6_forwarding) {
626 		ip6stat.ip6s_cantforward++;
627 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
628 		goto bad;
629 	}
630 
631 hbhcheck:
632 	/*
633 	 * record address information into m_aux, if we don't have one yet.
634 	 * note that we are unable to record it, if the address is not listed
635 	 * as our interface address (e.g. multicast addresses, addresses
636 	 * within FAITH prefixes and such).
637 	 */
638 	if (deliverifp && !ip6_getdstifaddr(m)) {
639 		struct in6_ifaddr *ia6;
640 
641 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
642 		if (ia6) {
643 			if (!ip6_setdstifaddr(m, ia6)) {
644 				/*
645 				 * XXX maybe we should drop the packet here,
646 				 * as we could not provide enough information
647 				 * to the upper layers.
648 				 */
649 			}
650 		}
651 	}
652 
653 	/*
654 	 * Process Hop-by-Hop options header if it's contained.
655 	 * m may be modified in ip6_hopopts_input().
656 	 * If a JumboPayload option is included, plen will also be modified.
657 	 */
658 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
659 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
660 		struct ip6_hbh *hbh;
661 
662 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
663 #if 0	/*touches NULL pointer*/
664 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
665 #endif
666 			goto bad2;	/* m have already been freed */
667 		}
668 
669 		/* adjust pointer */
670 		ip6 = mtod(m, struct ip6_hdr *);
671 
672 		/*
673 		 * if the payload length field is 0 and the next header field
674 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
675 		 * option MUST be included.
676 		 */
677 		if (ip6->ip6_plen == 0 && plen == 0) {
678 			/*
679 			 * Note that if a valid jumbo payload option is
680 			 * contained, ip6_hoptops_input() must set a valid
681 			 * (non-zero) payload length to the variable plen.
682 			 */
683 			ip6stat.ip6s_badoptions++;
684 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
685 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
686 			icmp6_error(m, ICMP6_PARAM_PROB,
687 				    ICMP6_PARAMPROB_HEADER,
688 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
689 			goto bad2;
690 		}
691 #ifndef PULLDOWN_TEST
692 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
693 		hbh = (struct ip6_hbh *)(ip6 + 1);
694 #else
695 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
696 			sizeof(struct ip6_hbh));
697 		if (hbh == NULL) {
698 			ip6stat.ip6s_tooshort++;
699 			goto bad2;
700 		}
701 #endif
702 		nxt = hbh->ip6h_nxt;
703 
704 		/*
705 		 * If we are acting as a router and the packet contains a
706 		 * router alert option, see if we know the option value.
707 		 * Currently, we only support the option value for MLD, in which
708 		 * case we should pass the packet to the multicast routing
709 		 * daemon.
710 		 */
711 		if (rtalert != ~0 && ip6_forwarding) {
712 			switch (rtalert) {
713 			case IP6OPT_RTALERT_MLD:
714 				ours = 1;
715 				break;
716 			default:
717 				/*
718 				 * RFC2711 requires unrecognized values must be
719 				 * silently ignored.
720 				 */
721 				break;
722 			}
723 		}
724 	} else
725 		nxt = ip6->ip6_nxt;
726 
727 	/*
728 	 * Check that the amount of data in the buffers
729 	 * is as at least much as the IPv6 header would have us expect.
730 	 * Trim mbufs if longer than we expect.
731 	 * Drop packet if shorter than we expect.
732 	 */
733 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
734 		ip6stat.ip6s_tooshort++;
735 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
736 		goto bad;
737 	}
738 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
739 		if (m->m_len == m->m_pkthdr.len) {
740 			m->m_len = sizeof(struct ip6_hdr) + plen;
741 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
742 		} else
743 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
744 	}
745 
746 	/*
747 	 * Forward if desirable.
748 	 */
749 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
750 		/*
751 		 * If we are acting as a multicast router, all
752 		 * incoming multicast packets are passed to the
753 		 * kernel-level multicast forwarding function.
754 		 * The packet is returned (relatively) intact; if
755 		 * ip6_mforward() returns a non-zero value, the packet
756 		 * must be discarded, else it may be accepted below.
757 		 */
758 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
759 			ip6stat.ip6s_cantforward++;
760 			goto bad;
761 		}
762 		if (!ours)
763 			goto bad;
764 	} else if (!ours) {
765 		ip6_forward(m, srcrt);
766 		goto bad2;
767 	}
768 
769 	ip6 = mtod(m, struct ip6_hdr *);
770 
771 	/*
772 	 * Malicious party may be able to use IPv4 mapped addr to confuse
773 	 * tcp/udp stack and bypass security checks (act as if it was from
774 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
775 	 *
776 	 * For SIIT end node behavior, you may want to disable the check.
777 	 * However, you will  become vulnerable to attacks using IPv4 mapped
778 	 * source.
779 	 */
780 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
781 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
782 		ip6stat.ip6s_badscope++;
783 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
784 		goto bad;
785 	}
786 
787 	/*
788 	 * Tell launch routine the next header
789 	 */
790 	ip6stat.ip6s_delivered++;
791 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
792 	nest = 0;
793 
794 	rh_present = 0;
795 	while (nxt != IPPROTO_DONE) {
796 		struct protosw *sw6;
797 
798 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
799 			ip6stat.ip6s_toomanyhdr++;
800 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
801 			goto bad;
802 		}
803 
804 		/*
805 		 * protection against faulty packet - there should be
806 		 * more sanity checks in header chain processing.
807 		 */
808 		if (m->m_pkthdr.len < off) {
809 			ip6stat.ip6s_tooshort++;
810 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
811 			goto bad;
812 		}
813 
814 #if 0
815 		/*
816 		 * do we need to do it for every header?  yeah, other
817 		 * functions can play with it (like re-allocate and copy).
818 		 */
819 		mhist = ip6_addaux(m);
820 		if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
821 			hist = mtod(mhist, caddr_t) + mhist->m_len;
822 			bcopy(&nxt, hist, sizeof(nxt));
823 			mhist->m_len += sizeof(nxt);
824 		} else {
825 			ip6stat.ip6s_toomanyhdr++;
826 			goto bad;
827 		}
828 #endif
829 
830 		if (nxt == IPPROTO_ROUTING) {
831 			if (rh_present++) {
832 				in6_ifstat_inc(m->m_pkthdr.rcvif,
833 				    ifs6_in_hdrerr);
834 				ip6stat.ip6s_badoptions++;
835 				goto bad;
836 			}
837 		}
838 
839 		sw6 = &inet6sw[ip6_protox[nxt]];
840 #ifdef IPSEC
841 		/*
842 		 * enforce IPsec policy checking if we are seeing last header.
843 		 * note that we do not visit this with protocols with pcb layer
844 		 * code - like udp/tcp/raw ip.
845 		 */
846 		if ((sw6->pr_flags & PR_LASTHDR) && ipsec6_in_reject(m, NULL)) {
847 			ipsec6stat.in_polvio++;
848 			goto bad;
849 		}
850 #endif
851 		/*
852 		 * If this is a terminal header forward to the port, otherwise
853 		 * process synchronously for more headers.
854 		 */
855 		if (sw6->pr_flags & PR_LASTHDR) {
856 			struct netmsg_packet *pmsg;
857 			lwkt_port_t port;
858 
859 			port = cpu_portfn(0); /* XXX */
860 			KKASSERT(port != NULL);
861 			pmsg = &m->m_hdr.mh_netmsg;
862 			netmsg_init(&pmsg->base, NULL,
863 				    &netisr_apanic_rport,
864 				    0, transport6_processing_handler);
865 			pmsg->nm_packet = m;
866 			pmsg->nm_nxt = nxt;
867 			pmsg->base.lmsg.u.ms_result = off;
868 			lwkt_sendmsg(port, &pmsg->base.lmsg);
869 			/* done with m */
870 			nxt = IPPROTO_DONE;
871 		} else {
872 			nxt = sw6->pr_input(&m, &off, nxt);
873 		}
874 	}
875 	goto bad2;
876 bad:
877 	m_freem(m);
878 bad2:
879 	;
880 	/* msg was embedded in the mbuf, do not reply! */
881 }
882 
883 /*
884  * We have to call the pr_input() function from the correct protocol
885  * thread.  The sw6->pr_soport() request at the end of ip6_input()
886  * returns the port and we forward a netmsg to the port to execute
887  * this function.
888  */
889 static void
890 transport6_processing_handler(netmsg_t netmsg)
891 {
892 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
893 	struct protosw *sw6;
894 	int hlen;
895 	int nxt;
896 
897 	sw6 = &inet6sw[ip6_protox[pmsg->nm_nxt]];
898 	hlen = pmsg->base.lmsg.u.ms_result;
899 
900 	nxt = sw6->pr_input(&pmsg->nm_packet, &hlen, pmsg->nm_nxt);
901 	KKASSERT(nxt == IPPROTO_DONE);
902 	/* netmsg was embedded in the mbuf, do not reply! */
903 }
904 
905 /*
906  * set/grab in6_ifaddr correspond to IPv6 destination address.
907  * XXX backward compatibility wrapper
908  */
909 static struct ip6aux *
910 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
911 {
912 	struct ip6aux *n;
913 
914 	n = ip6_addaux(m);
915 	if (n)
916 		n->ip6a_dstia6 = ia6;
917 	return n;	/* NULL if failed to set */
918 }
919 
920 struct in6_ifaddr *
921 ip6_getdstifaddr(struct mbuf *m)
922 {
923 	struct ip6aux *n;
924 
925 	n = ip6_findaux(m);
926 	if (n)
927 		return n->ip6a_dstia6;
928 	else
929 		return NULL;
930 }
931 
932 /*
933  * Hop-by-Hop options header processing. If a valid jumbo payload option is
934  * included, the real payload length will be stored in plenp.
935  */
936 static int
937 ip6_hopopts_input(u_int32_t *plenp,
938 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
939 		  struct mbuf **mp,
940 		  int *offp)
941 {
942 	struct mbuf *m = *mp;
943 	int off = *offp, hbhlen;
944 	struct ip6_hbh *hbh;
945 	u_int8_t *opt;
946 
947 	/* validation of the length of the header */
948 #ifndef PULLDOWN_TEST
949 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
950 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
951 	hbhlen = (hbh->ip6h_len + 1) << 3;
952 
953 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
954 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
955 #else
956 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
957 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
958 	if (hbh == NULL) {
959 		ip6stat.ip6s_tooshort++;
960 		return -1;
961 	}
962 	hbhlen = (hbh->ip6h_len + 1) << 3;
963 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
964 		hbhlen);
965 	if (hbh == NULL) {
966 		ip6stat.ip6s_tooshort++;
967 		return -1;
968 	}
969 #endif
970 	off += hbhlen;
971 	hbhlen -= sizeof(struct ip6_hbh);
972 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
973 
974 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
975 				hbhlen, rtalertp, plenp) < 0)
976 		return (-1);
977 
978 	*offp = off;
979 	*mp = m;
980 	return (0);
981 }
982 
983 /*
984  * Search header for all Hop-by-hop options and process each option.
985  * This function is separate from ip6_hopopts_input() in order to
986  * handle a case where the sending node itself process its hop-by-hop
987  * options header. In such a case, the function is called from ip6_output().
988  *
989  * The function assumes that hbh header is located right after the IPv6 header
990  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
991  * opthead + hbhlen is located in continuous memory region.
992  */
993 int
994 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
995 		    u_int32_t *rtalertp, u_int32_t *plenp)
996 {
997 	struct ip6_hdr *ip6;
998 	int optlen = 0;
999 	u_int8_t *opt = opthead;
1000 	u_int16_t rtalert_val;
1001 	u_int32_t jumboplen;
1002 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1003 
1004 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1005 		switch (*opt) {
1006 		case IP6OPT_PAD1:
1007 			optlen = 1;
1008 			break;
1009 		case IP6OPT_PADN:
1010 			if (hbhlen < IP6OPT_MINLEN) {
1011 				ip6stat.ip6s_toosmall++;
1012 				goto bad;
1013 			}
1014 			optlen = *(opt + 1) + 2;
1015 			break;
1016 		case IP6OPT_RTALERT:
1017 			/* XXX may need check for alignment */
1018 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1019 				ip6stat.ip6s_toosmall++;
1020 				goto bad;
1021 			}
1022 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1023 				/* XXX stat */
1024 				icmp6_error(m, ICMP6_PARAM_PROB,
1025 					    ICMP6_PARAMPROB_HEADER,
1026 					    erroff + opt + 1 - opthead);
1027 				return (-1);
1028 			}
1029 			optlen = IP6OPT_RTALERT_LEN;
1030 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1031 			*rtalertp = ntohs(rtalert_val);
1032 			break;
1033 		case IP6OPT_JUMBO:
1034 			/* XXX may need check for alignment */
1035 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1036 				ip6stat.ip6s_toosmall++;
1037 				goto bad;
1038 			}
1039 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1040 				/* XXX stat */
1041 				icmp6_error(m, ICMP6_PARAM_PROB,
1042 					    ICMP6_PARAMPROB_HEADER,
1043 					    erroff + opt + 1 - opthead);
1044 				return (-1);
1045 			}
1046 			optlen = IP6OPT_JUMBO_LEN;
1047 
1048 			/*
1049 			 * IPv6 packets that have non 0 payload length
1050 			 * must not contain a jumbo payload option.
1051 			 */
1052 			ip6 = mtod(m, struct ip6_hdr *);
1053 			if (ip6->ip6_plen) {
1054 				ip6stat.ip6s_badoptions++;
1055 				icmp6_error(m, ICMP6_PARAM_PROB,
1056 					    ICMP6_PARAMPROB_HEADER,
1057 					    erroff + opt - opthead);
1058 				return (-1);
1059 			}
1060 
1061 			/*
1062 			 * We may see jumbolen in unaligned location, so
1063 			 * we'd need to perform bcopy().
1064 			 */
1065 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1066 			jumboplen = (u_int32_t)htonl(jumboplen);
1067 
1068 #if 1
1069 			/*
1070 			 * if there are multiple jumbo payload options,
1071 			 * *plenp will be non-zero and the packet will be
1072 			 * rejected.
1073 			 * the behavior may need some debate in ipngwg -
1074 			 * multiple options does not make sense, however,
1075 			 * there's no explicit mention in specification.
1076 			 */
1077 			if (*plenp != 0) {
1078 				ip6stat.ip6s_badoptions++;
1079 				icmp6_error(m, ICMP6_PARAM_PROB,
1080 					    ICMP6_PARAMPROB_HEADER,
1081 					    erroff + opt + 2 - opthead);
1082 				return (-1);
1083 			}
1084 #endif
1085 
1086 			/*
1087 			 * jumbo payload length must be larger than 65535.
1088 			 */
1089 			if (jumboplen <= IPV6_MAXPACKET) {
1090 				ip6stat.ip6s_badoptions++;
1091 				icmp6_error(m, ICMP6_PARAM_PROB,
1092 					    ICMP6_PARAMPROB_HEADER,
1093 					    erroff + opt + 2 - opthead);
1094 				return (-1);
1095 			}
1096 			*plenp = jumboplen;
1097 
1098 			break;
1099 		default:		/* unknown option */
1100 			if (hbhlen < IP6OPT_MINLEN) {
1101 				ip6stat.ip6s_toosmall++;
1102 				goto bad;
1103 			}
1104 			optlen = ip6_unknown_opt(opt, m,
1105 			    erroff + opt - opthead);
1106 			if (optlen == -1)
1107 				return (-1);
1108 			optlen += 2;
1109 			break;
1110 		}
1111 	}
1112 
1113 	return (0);
1114 
1115 bad:
1116 	m_freem(m);
1117 	return (-1);
1118 }
1119 
1120 /*
1121  * Unknown option processing.
1122  * The third argument `off' is the offset from the IPv6 header to the option,
1123  * which is necessary if the IPv6 header the and option header and IPv6 header
1124  * is not continuous in order to return an ICMPv6 error.
1125  */
1126 int
1127 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1128 {
1129 	struct ip6_hdr *ip6;
1130 
1131 	switch (IP6OPT_TYPE(*optp)) {
1132 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1133 		return ((int)*(optp + 1));
1134 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1135 		m_freem(m);
1136 		return (-1);
1137 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1138 		ip6stat.ip6s_badoptions++;
1139 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1140 		return (-1);
1141 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1142 		ip6stat.ip6s_badoptions++;
1143 		ip6 = mtod(m, struct ip6_hdr *);
1144 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1145 		    (m->m_flags & (M_BCAST|M_MCAST)))
1146 			m_freem(m);
1147 		else
1148 			icmp6_error(m, ICMP6_PARAM_PROB,
1149 				    ICMP6_PARAMPROB_OPTION, off);
1150 		return (-1);
1151 	}
1152 
1153 	m_freem(m);		/* XXX: NOTREACHED */
1154 	return (-1);
1155 }
1156 
1157 /*
1158  * Create the "control" list for this pcb.
1159  * The function will not modify mbuf chain at all.
1160  *
1161  * with KAME mbuf chain restriction:
1162  * The routine will be called from upper layer handlers like tcp6_input().
1163  * Thus the routine assumes that the caller (tcp6_input) have already
1164  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1165  * very first mbuf on the mbuf chain.
1166  */
1167 void
1168 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1169 		struct mbuf *m)
1170 {
1171 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1172 	struct thread *td = curthread;	/* XXX */
1173 	int privileged = 0;
1174 	int rthdr_exist = 0;
1175 
1176 
1177 	if (priv_check(td, PRIV_ROOT) == 0)
1178  		privileged++;
1179 
1180 #ifdef SO_TIMESTAMP
1181 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1182 		struct timeval tv;
1183 
1184 		microtime(&tv);
1185 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1186 				      SCM_TIMESTAMP, SOL_SOCKET);
1187 		if (*mp) {
1188 			mp = &(*mp)->m_next;
1189 		}
1190 	}
1191 #endif
1192 
1193 	/* RFC 2292 sec. 5 */
1194 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1195 		struct in6_pktinfo pi6;
1196 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1197 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1198 			pi6.ipi6_addr.s6_addr16[1] = 0;
1199 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1200 					? m->m_pkthdr.rcvif->if_index
1201 					: 0;
1202 		*mp = sbcreatecontrol((caddr_t) &pi6,
1203 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1204 			IPPROTO_IPV6);
1205 		if (*mp)
1206 			mp = &(*mp)->m_next;
1207 	}
1208 
1209 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1210 		int hlim = ip6->ip6_hlim & 0xff;
1211 		*mp = sbcreatecontrol((caddr_t) &hlim,
1212 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1213 		if (*mp)
1214 			mp = &(*mp)->m_next;
1215 	}
1216 
1217 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1218 		u_int32_t flowinfo;
1219 		int tclass;
1220 
1221 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1222 		flowinfo >>= 20;
1223 
1224 		tclass = flowinfo & 0xff;
1225 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1226 		    IPV6_TCLASS, IPPROTO_IPV6);
1227 		if (*mp)
1228 			mp = &(*mp)->m_next;
1229 	}
1230 
1231 	/*
1232 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1233 	 * for the option, but it might be too strict, since there might
1234 	 * be some hop-by-hop options which can be returned to normal user.
1235 	 * See RFC 2292 section 6.
1236 	 */
1237 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1238 		/*
1239 		 * Check if a hop-by-hop options header is contatined in the
1240 		 * received packet, and if so, store the options as ancillary
1241 		 * data. Note that a hop-by-hop options header must be
1242 		 * just after the IPv6 header, which fact is assured through
1243 		 * the IPv6 input processing.
1244 		 */
1245 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1246 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1247 			struct ip6_hbh *hbh;
1248 			int hbhlen = 0;
1249 #ifdef PULLDOWN_TEST
1250 			struct mbuf *ext;
1251 #endif
1252 
1253 #ifndef PULLDOWN_TEST
1254 			hbh = (struct ip6_hbh *)(ip6 + 1);
1255 			hbhlen = (hbh->ip6h_len + 1) << 3;
1256 #else
1257 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1258 			    ip6->ip6_nxt);
1259 			if (ext == NULL) {
1260 				ip6stat.ip6s_tooshort++;
1261 				return;
1262 			}
1263 			hbh = mtod(ext, struct ip6_hbh *);
1264 			hbhlen = (hbh->ip6h_len + 1) << 3;
1265 			if (hbhlen != ext->m_len) {
1266 				m_freem(ext);
1267 				ip6stat.ip6s_tooshort++;
1268 				return;
1269 			}
1270 #endif
1271 
1272 			/*
1273 			 * XXX: We copy whole the header even if a jumbo
1274 			 * payload option is included, which option is to
1275 			 * be removed before returning in the RFC 2292.
1276 			 * Note: this constraint is removed in 2292bis.
1277 			 */
1278 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1279 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1280 			if (*mp)
1281 				mp = &(*mp)->m_next;
1282 #ifdef PULLDOWN_TEST
1283 			m_freem(ext);
1284 #endif
1285 		}
1286 	}
1287 
1288 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1289 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1290 		int proto, off, nxt;
1291 
1292 		/*
1293 		 * go through the header chain to see if a routing header is
1294 		 * contained in the packet. We need this information to store
1295 		 * destination options headers (if any) properly.
1296 		 * XXX: performance issue. We should record this info when
1297 		 * processing extension headers in incoming routine.
1298 		 * (todo) use m_aux?
1299 		 */
1300 		proto = IPPROTO_IPV6;
1301 		off = 0;
1302 		nxt = -1;
1303 		while (1) {
1304 			int newoff;
1305 
1306 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1307 			if (newoff < 0)
1308 				break;
1309 			if (newoff < off) /* invalid, check for safety */
1310 				break;
1311 			if ((proto = nxt) == IPPROTO_ROUTING) {
1312 				rthdr_exist = 1;
1313 				break;
1314 			}
1315 			off = newoff;
1316 		}
1317 	}
1318 
1319 	if ((in6p->in6p_flags &
1320 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1321 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1322 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1323 
1324 		/*
1325 		 * Search for destination options headers or routing
1326 		 * header(s) through the header chain, and stores each
1327 		 * header as ancillary data.
1328 		 * Note that the order of the headers remains in
1329 		 * the chain of ancillary data.
1330 		 */
1331 		while (1) {	/* is explicit loop prevention necessary? */
1332 			struct ip6_ext *ip6e = NULL;
1333 			int elen;
1334 #ifdef PULLDOWN_TEST
1335 			struct mbuf *ext = NULL;
1336 #endif
1337 
1338 			/*
1339 			 * if it is not an extension header, don't try to
1340 			 * pull it from the chain.
1341 			 */
1342 			switch (nxt) {
1343 			case IPPROTO_DSTOPTS:
1344 			case IPPROTO_ROUTING:
1345 			case IPPROTO_HOPOPTS:
1346 			case IPPROTO_AH: /* is it possible? */
1347 				break;
1348 			default:
1349 				goto loopend;
1350 			}
1351 
1352 #ifndef PULLDOWN_TEST
1353 			if (off + sizeof(*ip6e) > m->m_len)
1354 				goto loopend;
1355 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1356 			if (nxt == IPPROTO_AH)
1357 				elen = (ip6e->ip6e_len + 2) << 2;
1358 			else
1359 				elen = (ip6e->ip6e_len + 1) << 3;
1360 			if (off + elen > m->m_len)
1361 				goto loopend;
1362 #else
1363 			ext = ip6_pullexthdr(m, off, nxt);
1364 			if (ext == NULL) {
1365 				ip6stat.ip6s_tooshort++;
1366 				return;
1367 			}
1368 			ip6e = mtod(ext, struct ip6_ext *);
1369 			if (nxt == IPPROTO_AH)
1370 				elen = (ip6e->ip6e_len + 2) << 2;
1371 			else
1372 				elen = (ip6e->ip6e_len + 1) << 3;
1373 			if (elen != ext->m_len) {
1374 				m_freem(ext);
1375 				ip6stat.ip6s_tooshort++;
1376 				return;
1377 			}
1378 #endif
1379 
1380 			switch (nxt) {
1381 			case IPPROTO_DSTOPTS:
1382 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1383 					break;
1384 
1385 				/*
1386 				 * We also require super-user privilege for
1387 				 * the option.
1388 				 * See the comments on IN6_HOPOPTS.
1389 				 */
1390 				if (!privileged)
1391 					break;
1392 
1393 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1394 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1395 						      IPPROTO_IPV6);
1396 				if (*mp)
1397 					mp = &(*mp)->m_next;
1398 				break;
1399 			case IPPROTO_ROUTING:
1400 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1401 					break;
1402 
1403 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1404 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1405 						      IPPROTO_IPV6);
1406 				if (*mp)
1407 					mp = &(*mp)->m_next;
1408 				break;
1409 			case IPPROTO_HOPOPTS:
1410 			case IPPROTO_AH: /* is it possible? */
1411 				break;
1412 
1413 			default:
1414 				/*
1415 			 	 * other cases have been filtered in the above.
1416 				 * none will visit this case.  here we supply
1417 				 * the code just in case (nxt overwritten or
1418 				 * other cases).
1419 				 */
1420 #ifdef PULLDOWN_TEST
1421 				m_freem(ext);
1422 #endif
1423 				goto loopend;
1424 
1425 			}
1426 
1427 			/* proceed with the next header. */
1428 			off += elen;
1429 			nxt = ip6e->ip6e_nxt;
1430 			ip6e = NULL;
1431 #ifdef PULLDOWN_TEST
1432 			m_freem(ext);
1433 			ext = NULL;
1434 #endif
1435 		}
1436 	  loopend:
1437 		;
1438 	}
1439 #undef IS2292
1440 }
1441 
1442 void
1443 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1444 {
1445 	struct socket *so;
1446 	struct mbuf *m_mtu;
1447 	struct ip6_mtuinfo mtuctl;
1448 
1449 	so =  in6p->inp_socket;
1450 
1451 	if (mtu == NULL)
1452 		return;
1453 
1454 #ifdef DIAGNOSTIC
1455 	if (so == NULL)		/* I believe this is impossible */
1456 		panic("ip6_notify_pmtu: socket is NULL");
1457 #endif
1458 
1459 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1460 	mtuctl.ip6m_mtu = *mtu;
1461 	mtuctl.ip6m_addr = *dst;
1462 
1463 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1464 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1465 		return;
1466 
1467 	lwkt_gettoken(&so->so_rcv.ssb_token);
1468 	if (sbappendaddr(&so->so_rcv.sb, (struct sockaddr *)dst, NULL, m_mtu)
1469 	    == 0) {
1470 		m_freem(m_mtu);
1471 		/* XXX: should count statistics */
1472 	} else {
1473 		sorwakeup(so);
1474 	}
1475 	lwkt_reltoken(&so->so_rcv.ssb_token);
1476 }
1477 
1478 #ifdef PULLDOWN_TEST
1479 /*
1480  * pull single extension header from mbuf chain.  returns single mbuf that
1481  * contains the result, or NULL on error.
1482  */
1483 static struct mbuf *
1484 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1485 {
1486 	struct ip6_ext ip6e;
1487 	size_t elen;
1488 	struct mbuf *n;
1489 
1490 #ifdef DIAGNOSTIC
1491 	switch (nxt) {
1492 	case IPPROTO_DSTOPTS:
1493 	case IPPROTO_ROUTING:
1494 	case IPPROTO_HOPOPTS:
1495 	case IPPROTO_AH: /* is it possible? */
1496 		break;
1497 	default:
1498 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1499 	}
1500 #endif
1501 
1502 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1503 	if (nxt == IPPROTO_AH)
1504 		elen = (ip6e.ip6e_len + 2) << 2;
1505 	else
1506 		elen = (ip6e.ip6e_len + 1) << 3;
1507 
1508 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1509 	if (n == NULL)
1510 		return NULL;
1511 	n->m_len = 0;
1512 
1513 	if (elen >= M_TRAILINGSPACE(n)) {
1514 		m_free(n);
1515 		return NULL;
1516 	}
1517 
1518 	m_copydata(m, off, elen, mtod(n, caddr_t));
1519 	n->m_len = elen;
1520 	return n;
1521 }
1522 #endif
1523 
1524 /*
1525  * Get pointer to the previous header followed by the header
1526  * currently processed.
1527  * XXX: This function supposes that
1528  *	M includes all headers,
1529  *	the next header field and the header length field of each header
1530  *	are valid, and
1531  *	the sum of each header length equals to OFF.
1532  * Because of these assumptions, this function must be called very
1533  * carefully. Moreover, it will not be used in the near future when
1534  * we develop `neater' mechanism to process extension headers.
1535  */
1536 char *
1537 ip6_get_prevhdr(struct mbuf *m, int off)
1538 {
1539 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1540 
1541 	if (off == sizeof(struct ip6_hdr))
1542 		return (&ip6->ip6_nxt);
1543 	else {
1544 		int len, nxt;
1545 		struct ip6_ext *ip6e = NULL;
1546 
1547 		nxt = ip6->ip6_nxt;
1548 		len = sizeof(struct ip6_hdr);
1549 		while (len < off) {
1550 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1551 
1552 			switch (nxt) {
1553 			case IPPROTO_FRAGMENT:
1554 				len += sizeof(struct ip6_frag);
1555 				break;
1556 			case IPPROTO_AH:
1557 				len += (ip6e->ip6e_len + 2) << 2;
1558 				break;
1559 			default:
1560 				len += (ip6e->ip6e_len + 1) << 3;
1561 				break;
1562 			}
1563 			nxt = ip6e->ip6e_nxt;
1564 		}
1565 		if (ip6e)
1566 			return (&ip6e->ip6e_nxt);
1567 		else
1568 			return NULL;
1569 	}
1570 }
1571 
1572 /*
1573  * get next header offset.  m will be retained.
1574  */
1575 int
1576 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1577 {
1578 	struct ip6_hdr ip6;
1579 	struct ip6_ext ip6e;
1580 	struct ip6_frag fh;
1581 
1582 	/* just in case */
1583 	if (m == NULL)
1584 		panic("ip6_nexthdr: m == NULL");
1585 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1586 		return -1;
1587 
1588 	switch (proto) {
1589 	case IPPROTO_IPV6:
1590 		if (m->m_pkthdr.len < off + sizeof(ip6))
1591 			return -1;
1592 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1593 		if (nxtp)
1594 			*nxtp = ip6.ip6_nxt;
1595 		off += sizeof(ip6);
1596 		return off;
1597 
1598 	case IPPROTO_FRAGMENT:
1599 		/*
1600 		 * terminate parsing if it is not the first fragment,
1601 		 * it does not make sense to parse through it.
1602 		 */
1603 		if (m->m_pkthdr.len < off + sizeof(fh))
1604 			return -1;
1605 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1606 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1607 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1608 			return -1;
1609 		if (nxtp)
1610 			*nxtp = fh.ip6f_nxt;
1611 		off += sizeof(struct ip6_frag);
1612 		return off;
1613 
1614 	case IPPROTO_AH:
1615 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1616 			return -1;
1617 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1618 		if (nxtp)
1619 			*nxtp = ip6e.ip6e_nxt;
1620 		off += (ip6e.ip6e_len + 2) << 2;
1621 		return off;
1622 
1623 	case IPPROTO_HOPOPTS:
1624 	case IPPROTO_ROUTING:
1625 	case IPPROTO_DSTOPTS:
1626 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1627 			return -1;
1628 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1629 		if (nxtp)
1630 			*nxtp = ip6e.ip6e_nxt;
1631 		off += (ip6e.ip6e_len + 1) << 3;
1632 		return off;
1633 
1634 	case IPPROTO_NONE:
1635 	case IPPROTO_ESP:
1636 	case IPPROTO_IPCOMP:
1637 		/* give up */
1638 		return -1;
1639 
1640 	default:
1641 		return -1;
1642 	}
1643 
1644 	return -1;
1645 }
1646 
1647 /*
1648  * get offset for the last header in the chain.  m will be kept untainted.
1649  */
1650 int
1651 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1652 {
1653 	int newoff;
1654 	int nxt;
1655 
1656 	if (!nxtp) {
1657 		nxt = -1;
1658 		nxtp = &nxt;
1659 	}
1660 	while (1) {
1661 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1662 		if (newoff < 0)
1663 			return off;
1664 		else if (newoff < off)
1665 			return -1;	/* invalid */
1666 		else if (newoff == off)
1667 			return newoff;
1668 
1669 		off = newoff;
1670 		proto = *nxtp;
1671 	}
1672 }
1673 
1674 struct ip6aux *
1675 ip6_addaux(struct mbuf *m)
1676 {
1677 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1678 	if (!tag) {
1679 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1680 				sizeof (struct ip6aux),
1681 				MB_DONTWAIT);
1682 		if (tag)
1683 			m_tag_prepend(m, tag);
1684 	}
1685 	if (tag)
1686 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1687 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1688 }
1689 
1690 struct ip6aux *
1691 ip6_findaux(struct mbuf *m)
1692 {
1693 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1694 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1695 }
1696 
1697 void
1698 ip6_delaux(struct mbuf *m)
1699 {
1700 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1701 	if (tag)
1702 		m_tag_delete(m, tag);
1703 }
1704 
1705 /*
1706  * System control for IP6
1707  */
1708 
1709 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1710 	0,		0,		0,		0,
1711 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1712 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1713 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1714 	0,		0,		0,		0,
1715 	ENOPROTOOPT
1716 };
1717