xref: /dflybsd-src/sys/netinet6/ip6_input.c (revision 895c1f850848cf50a3243e134345f0d4ff33ac59)
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 <sys/thread2.h>
90 #include <sys/msgport2.h>
91 
92 #include <net/if.h>
93 #include <net/if_types.h>
94 #include <net/if_dl.h>
95 #include <net/route.h>
96 #include <net/netisr.h>
97 #include <net/pfil.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/in6_prefix.h>
114 
115 #ifdef IPSEC
116 #include <netinet6/ipsec.h>
117 #ifdef INET6
118 #include <netinet6/ipsec6.h>
119 #endif
120 #endif
121 
122 #ifdef FAST_IPSEC
123 #include <netproto/ipsec/ipsec.h>
124 #include <netproto/ipsec/ipsec6.h>
125 #define	IPSEC
126 #endif /* FAST_IPSEC */
127 
128 #include <net/ip6fw/ip6_fw.h>
129 
130 #include <netinet6/ip6protosw.h>
131 
132 /* we need it for NLOOP. */
133 #include "use_loop.h"
134 
135 #include <net/net_osdep.h>
136 
137 extern struct domain inet6domain;
138 extern struct ip6protosw inet6sw[];
139 
140 u_char ip6_protox[IPPROTO_MAX];
141 struct in6_ifaddr *in6_ifaddr;
142 
143 extern struct callout in6_tmpaddrtimer_ch;
144 
145 int ip6_forward_srcrt;			/* XXX */
146 int ip6_sourcecheck;			/* XXX */
147 int ip6_sourcecheck_interval;		/* XXX */
148 
149 int ip6_ours_check_algorithm;
150 
151 struct pfil_head inet6_pfil_hook;
152 
153 /* firewall hooks */
154 ip6_fw_chk_t *ip6_fw_chk_ptr;
155 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
156 int ip6_fw_enable = 1;
157 
158 struct ip6stat ip6stat;
159 
160 static void ip6_init2 (void *);
161 static struct ip6aux *ip6_setdstifaddr (struct mbuf *, struct in6_ifaddr *);
162 static int ip6_hopopts_input (u_int32_t *, u_int32_t *, struct mbuf **, int *);
163 static void ip6_input(struct netmsg *msg);
164 #ifdef PULLDOWN_TEST
165 static struct mbuf *ip6_pullexthdr (struct mbuf *, size_t, int);
166 #endif
167 
168 /*
169  * IP6 initialization: fill in IP6 protocol switch table.
170  * All protocols not implemented in kernel go to raw IP6 protocol handler.
171  */
172 void
173 ip6_init(void)
174 {
175 	struct ip6protosw *pr;
176 	int i;
177 	struct timeval tv;
178 
179 #ifdef DIAGNOSTIC
180 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
181 		panic("sizeof(protosw) != sizeof(ip6protosw)");
182 #endif
183 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
184 	if (pr == 0)
185 		panic("ip6_init");
186 	for (i = 0; i < IPPROTO_MAX; i++)
187 		ip6_protox[i] = pr - inet6sw;
188 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
189 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
190 		if (pr->pr_domain->dom_family == PF_INET6 &&
191 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
192 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
193 
194 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
195 	inet6_pfil_hook.ph_af = AF_INET6;
196 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0) {
197 		kprintf("%s: WARNING: unable to register pfil hook, "
198 			"error %d\n", __func__, i);
199 	}
200 
201 	netisr_register(NETISR_IPV6, cpu0_portfn, ip6_input,
202 			NETISR_FLAG_NOTMPSAFE);
203 	scope6_init();
204 	nd6_init();
205 	frag6_init();
206 	/*
207 	 * in many cases, random() here does NOT return random number
208 	 * as initialization during bootstrap time occur in fixed order.
209 	 */
210 	microtime(&tv);
211 	ip6_flow_seq = krandom() ^ tv.tv_usec;
212 	microtime(&tv);
213 	ip6_desync_factor = (krandom() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
214 }
215 
216 static void
217 ip6_init2(void *dummy)
218 {
219 	/* nd6_timer_init */
220 	callout_init(&nd6_timer_ch);
221 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
222 
223 	/* router renumbering prefix list maintenance */
224 	callout_init(&in6_rr_timer_ch);
225 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
226 
227 	/* timer for regeneranation of temporary addresses randomize ID */
228 	callout_init(&in6_tmpaddrtimer_ch);
229 	callout_reset(&in6_tmpaddrtimer_ch,
230 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
231 		       ip6_temp_regen_advance) * hz,
232 		      in6_tmpaddrtimer, NULL);
233 }
234 
235 /* cheat */
236 /* This must be after route_init(), which is now SI_ORDER_THIRD */
237 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
238 
239 extern struct	route_in6 ip6_forward_rt;
240 
241 static
242 void
243 ip6_input(struct netmsg *msg)
244 {
245 	struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
246 	struct ip6_hdr *ip6;
247 	int off = sizeof(struct ip6_hdr), nest;
248 	u_int32_t plen;
249 	u_int32_t rtalert = ~0;
250 	int nxt, ours = 0, rh_present = 0;
251 	struct ifnet *deliverifp = NULL;
252 	struct in6_addr odst;
253 	int srcrt = 0;
254 
255 #ifdef IPSEC
256 	/*
257 	 * should the inner packet be considered authentic?
258 	 * see comment in ah4_input().
259 	 */
260 	if (m) {
261 		m->m_flags &= ~M_AUTHIPHDR;
262 		m->m_flags &= ~M_AUTHIPDGM;
263 	}
264 #endif
265 
266 	/*
267 	 * make sure we don't have onion peering information into m_aux.
268 	 */
269 	ip6_delaux(m);
270 
271 	/*
272 	 * mbuf statistics
273 	 */
274 	if (m->m_flags & M_EXT) {
275 		if (m->m_next)
276 			ip6stat.ip6s_mext2m++;
277 		else
278 			ip6stat.ip6s_mext1++;
279 	} else {
280 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
281 		if (m->m_next) {
282 			if (m->m_flags & M_LOOP) {
283 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/* XXX */
284 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
285 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
286 			else
287 				ip6stat.ip6s_m2m[0]++;
288 		} else
289 			ip6stat.ip6s_m1++;
290 #undef M2MMAX
291 	}
292 
293 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
294 	ip6stat.ip6s_total++;
295 
296 #ifndef PULLDOWN_TEST
297 	/*
298 	 * L2 bridge code and some other code can return mbuf chain
299 	 * that does not conform to KAME requirement.  too bad.
300 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
301 	 */
302 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
303 		struct mbuf *n;
304 
305 		n = m_getb(m->m_pkthdr.len, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
306 		if (n == NULL)
307 			goto bad;
308 		M_MOVE_PKTHDR(n, m);
309 
310 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
311 		n->m_len = n->m_pkthdr.len;
312 		m_freem(m);
313 		m = n;
314 	}
315 	IP6_EXTHDR_CHECK_VOIDRET(m, 0, sizeof(struct ip6_hdr));
316 #endif
317 
318 	if (m->m_len < sizeof(struct ip6_hdr)) {
319 		struct ifnet *inifp;
320 		inifp = m->m_pkthdr.rcvif;
321 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
322 			ip6stat.ip6s_toosmall++;
323 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
324 			goto bad2;
325 		}
326 	}
327 
328 	ip6 = mtod(m, struct ip6_hdr *);
329 
330 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
331 		ip6stat.ip6s_badvers++;
332 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
333 		goto bad;
334 	}
335 
336 	/*
337 	 * Run through list of hooks for input packets.
338 	 *
339 	 * NB: Beware of the destination address changing
340 	 *     (e.g. by NAT rewriting). When this happens,
341 	 *     tell ip6_forward to do the right thing.
342 	 */
343 	if (pfil_has_hooks(&inet6_pfil_hook)) {
344 		odst = ip6->ip6_dst;
345 		if (pfil_run_hooks(&inet6_pfil_hook, &m,
346 		    m->m_pkthdr.rcvif, PFIL_IN)) {
347 			goto bad2;
348 		}
349 		if (m == NULL)			/* consumed by filter */
350 			goto bad2;
351 		ip6 = mtod(m, struct ip6_hdr *);
352 		srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
353 	}
354 
355 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
356 
357 #ifdef ALTQ
358 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
359 		/* packet is dropped by traffic conditioner */
360 		return;
361 	}
362 #endif
363 
364 	/*
365 	 * Check with the firewall...
366 	 */
367 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
368 		u_short port = 0;
369 		/* If ipfw says divert, we have to just drop packet */
370 		/* use port as a dummy argument */
371 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
372 			m_freem(m);
373 			m = NULL;
374 		}
375 		if (!m)
376 			goto bad2;
377 	}
378 
379 	/*
380 	 * Check against address spoofing/corruption.
381 	 */
382 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
383 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
384 		/*
385 		 * XXX: "badscope" is not very suitable for a multicast source.
386 		 */
387 		ip6stat.ip6s_badscope++;
388 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
389 		goto bad;
390 	}
391 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
392 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
393 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
394 		ip6stat.ip6s_badscope++;
395 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
396 		goto bad;
397 	}
398 
399 	/*
400 	 * The following check is not documented in specs.  A malicious
401 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
402 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
403 	 * IPv6 src ::ffff:127.0.0.1).	Be cautious.
404 	 *
405 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
406 	 * support IPv4-less kernel compilation, we cannot support SIIT
407 	 * environment at all.  So, it makes more sense for us to reject any
408 	 * malicious packets for non-SIIT environment, than try to do a
409 	 * partical support for SIIT environment.
410 	 */
411 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
412 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
413 		ip6stat.ip6s_badscope++;
414 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
415 		goto bad;
416 	}
417 #if 0
418 	/*
419 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
420 	 *
421 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
422 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
423 	 * is revised to forbid relaying case.
424 	 */
425 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
426 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
427 		ip6stat.ip6s_badscope++;
428 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
429 		goto bad;
430 	}
431 #endif
432 
433 	/* drop packets if interface ID portion is already filled */
434 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
435 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
436 		    ip6->ip6_src.s6_addr16[1]) {
437 			ip6stat.ip6s_badscope++;
438 			goto bad;
439 		}
440 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
441 		    ip6->ip6_dst.s6_addr16[1]) {
442 			ip6stat.ip6s_badscope++;
443 			goto bad;
444 		}
445 	}
446 
447 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
448 		ip6->ip6_src.s6_addr16[1]
449 			= htons(m->m_pkthdr.rcvif->if_index);
450 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
451 		ip6->ip6_dst.s6_addr16[1]
452 			= htons(m->m_pkthdr.rcvif->if_index);
453 
454 #if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
455 	/*
456 	 * We use rt->rt_ifp to determine if the address is ours or not.
457 	 * If rt_ifp is lo0, the address is ours.
458 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
459 	 * so any address under fe80::%lo0/64 will be mistakenly considered
460 	 * local.  The special case is supplied to handle the case properly
461 	 * by actually looking at interface addresses
462 	 * (using in6ifa_ifpwithaddr).
463 	 */
464 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
465 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
466 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
467 			icmp6_error(m, ICMP6_DST_UNREACH,
468 			    ICMP6_DST_UNREACH_ADDR, 0);
469 			/* m is already freed */
470 			goto bad2;
471 		}
472 
473 		ours = 1;
474 		deliverifp = m->m_pkthdr.rcvif;
475 		goto hbhcheck;
476 	}
477 #endif
478 
479 	/*
480 	 * Multicast check
481 	 */
482 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
483 	  	struct	in6_multi *in6m = 0;
484 
485 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
486 		/*
487 		 * See if we belong to the destination multicast group on the
488 		 * arrival interface.
489 		 */
490 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
491 		if (in6m)
492 			ours = 1;
493 		else if (!ip6_mrouter) {
494 			ip6stat.ip6s_notmember++;
495 			ip6stat.ip6s_cantforward++;
496 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
497 			goto bad;
498 		}
499 		deliverifp = m->m_pkthdr.rcvif;
500 		goto hbhcheck;
501 	}
502 
503 	/*
504 	 *  Unicast check
505 	 */
506 	switch (ip6_ours_check_algorithm) {
507 	default:
508 		/*
509 		 * XXX: I intentionally broke our indentation rule here,
510 		 *      since this switch-case is just for measurement and
511 		 *      therefore should soon be removed.
512 		 */
513 	if (ip6_forward_rt.ro_rt != NULL &&
514 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
515 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
516 			       &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
517 		ip6stat.ip6s_forward_cachehit++;
518 	else {
519 		struct sockaddr_in6 *dst6;
520 
521 		if (ip6_forward_rt.ro_rt) {
522 			/* route is down or destination is different */
523 			ip6stat.ip6s_forward_cachemiss++;
524 			RTFREE(ip6_forward_rt.ro_rt);
525 			ip6_forward_rt.ro_rt = 0;
526 		}
527 
528 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
529 		dst6 = &ip6_forward_rt.ro_dst;
530 		dst6->sin6_len = sizeof(struct sockaddr_in6);
531 		dst6->sin6_family = AF_INET6;
532 		dst6->sin6_addr = ip6->ip6_dst;
533 
534 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
535 	}
536 
537 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
538 
539 	/*
540 	 * Accept the packet if the forwarding interface to the destination
541 	 * according to the routing table is the loopback interface,
542 	 * unless the associated route has a gateway.
543 	 * Note that this approach causes to accept a packet if there is a
544 	 * route to the loopback interface for the destination of the packet.
545 	 * But we think it's even useful in some situations, e.g. when using
546 	 * a special daemon which wants to intercept the packet.
547 	 *
548 	 * XXX: some OSes automatically make a cloned route for the destination
549 	 * of an outgoing packet.  If the outgoing interface of the packet
550 	 * is a loopback one, the kernel would consider the packet to be
551 	 * accepted, even if we have no such address assinged on the interface.
552 	 * We check the cloned flag of the route entry to reject such cases,
553 	 * assuming that route entries for our own addresses are not made by
554 	 * cloning (it should be true because in6_addloop explicitly installs
555 	 * the host route).  However, we might have to do an explicit check
556 	 * while it would be less efficient.  Or, should we rather install a
557 	 * reject route for such a case?
558 	 */
559 	if (ip6_forward_rt.ro_rt &&
560 	    (ip6_forward_rt.ro_rt->rt_flags &
561 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
562 #ifdef RTF_WASCLONED
563 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
564 #endif
565 #ifdef RTF_CLONED
566 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
567 #endif
568 #if 0
569 	    /*
570 	     * The check below is redundant since the comparison of
571 	     * the destination and the key of the rtentry has
572 	     * already done through looking up the routing table.
573 	     */
574 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
575 				&rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
576 #endif
577 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
578 		struct in6_ifaddr *ia6 =
579 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
580 
581 		/*
582 		 * record address information into m_aux.
583 		 */
584 		ip6_setdstifaddr(m, ia6);
585 
586 		/*
587 		 * packets to a tentative, duplicated, or somehow invalid
588 		 * address must not be accepted.
589 		 */
590 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
591 			/* this address is ready */
592 			ours = 1;
593 			deliverifp = ia6->ia_ifp;	/* correct? */
594 			/* Count the packet in the ip address stats */
595 			ia6->ia_ifa.if_ipackets++;
596 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
597 			goto hbhcheck;
598 		} else {
599 			/* address is not ready, so discard the packet. */
600 			nd6log((LOG_INFO,
601 			    "ip6_input: packet to an unready address %s->%s\n",
602 			    ip6_sprintf(&ip6->ip6_src),
603 			    ip6_sprintf(&ip6->ip6_dst)));
604 
605 			goto bad;
606 		}
607 	}
608 	} /* XXX indentation (see above) */
609 
610 	/*
611 	 * FAITH(Firewall Aided Internet Translator)
612 	 */
613 	if (ip6_keepfaith) {
614 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
615 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
616 			/* XXX do we need more sanity checks? */
617 			ours = 1;
618 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
619 			goto hbhcheck;
620 		}
621 	}
622 
623 	/*
624 	 * Now there is no reason to process the packet if it's not our own
625 	 * and we're not a router.
626 	 */
627 	if (!ip6_forwarding) {
628 		ip6stat.ip6s_cantforward++;
629 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
630 		goto bad;
631 	}
632 
633 hbhcheck:
634 	/*
635 	 * record address information into m_aux, if we don't have one yet.
636 	 * note that we are unable to record it, if the address is not listed
637 	 * as our interface address (e.g. multicast addresses, addresses
638 	 * within FAITH prefixes and such).
639 	 */
640 	if (deliverifp && !ip6_getdstifaddr(m)) {
641 		struct in6_ifaddr *ia6;
642 
643 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
644 		if (ia6) {
645 			if (!ip6_setdstifaddr(m, ia6)) {
646 				/*
647 				 * XXX maybe we should drop the packet here,
648 				 * as we could not provide enough information
649 				 * to the upper layers.
650 				 */
651 			}
652 		}
653 	}
654 
655 	/*
656 	 * Process Hop-by-Hop options header if it's contained.
657 	 * m may be modified in ip6_hopopts_input().
658 	 * If a JumboPayload option is included, plen will also be modified.
659 	 */
660 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
661 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
662 		struct ip6_hbh *hbh;
663 
664 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
665 #if 0	/*touches NULL pointer*/
666 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
667 #endif
668 			goto bad2;	/* m have already been freed */
669 		}
670 
671 		/* adjust pointer */
672 		ip6 = mtod(m, struct ip6_hdr *);
673 
674 		/*
675 		 * if the payload length field is 0 and the next header field
676 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
677 		 * option MUST be included.
678 		 */
679 		if (ip6->ip6_plen == 0 && plen == 0) {
680 			/*
681 			 * Note that if a valid jumbo payload option is
682 			 * contained, ip6_hoptops_input() must set a valid
683 			 * (non-zero) payload length to the variable plen.
684 			 */
685 			ip6stat.ip6s_badoptions++;
686 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
687 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
688 			icmp6_error(m, ICMP6_PARAM_PROB,
689 				    ICMP6_PARAMPROB_HEADER,
690 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
691 			goto bad2;
692 		}
693 #ifndef PULLDOWN_TEST
694 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
695 		hbh = (struct ip6_hbh *)(ip6 + 1);
696 #else
697 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
698 			sizeof(struct ip6_hbh));
699 		if (hbh == NULL) {
700 			ip6stat.ip6s_tooshort++;
701 			goto bad2;
702 		}
703 #endif
704 		nxt = hbh->ip6h_nxt;
705 
706 		/*
707 		 * If we are acting as a router and the packet contains a
708 		 * router alert option, see if we know the option value.
709 		 * Currently, we only support the option value for MLD, in which
710 		 * case we should pass the packet to the multicast routing
711 		 * daemon.
712 		 */
713 		if (rtalert != ~0 && ip6_forwarding) {
714 			switch (rtalert) {
715 			case IP6OPT_RTALERT_MLD:
716 				ours = 1;
717 				break;
718 			default:
719 				/*
720 				 * RFC2711 requires unrecognized values must be
721 				 * silently ignored.
722 				 */
723 				break;
724 			}
725 		}
726 	} else
727 		nxt = ip6->ip6_nxt;
728 
729 	/*
730 	 * Check that the amount of data in the buffers
731 	 * is as at least much as the IPv6 header would have us expect.
732 	 * Trim mbufs if longer than we expect.
733 	 * Drop packet if shorter than we expect.
734 	 */
735 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
736 		ip6stat.ip6s_tooshort++;
737 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
738 		goto bad;
739 	}
740 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
741 		if (m->m_len == m->m_pkthdr.len) {
742 			m->m_len = sizeof(struct ip6_hdr) + plen;
743 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
744 		} else
745 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
746 	}
747 
748 	/*
749 	 * Forward if desirable.
750 	 */
751 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
752 		/*
753 		 * If we are acting as a multicast router, all
754 		 * incoming multicast packets are passed to the
755 		 * kernel-level multicast forwarding function.
756 		 * The packet is returned (relatively) intact; if
757 		 * ip6_mforward() returns a non-zero value, the packet
758 		 * must be discarded, else it may be accepted below.
759 		 */
760 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
761 			ip6stat.ip6s_cantforward++;
762 			goto bad;
763 		}
764 		if (!ours)
765 			goto bad;
766 	} else if (!ours) {
767 		ip6_forward(m, srcrt);
768 		goto bad2;
769 	}
770 
771 	ip6 = mtod(m, struct ip6_hdr *);
772 
773 	/*
774 	 * Malicious party may be able to use IPv4 mapped addr to confuse
775 	 * tcp/udp stack and bypass security checks (act as if it was from
776 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
777 	 *
778 	 * For SIIT end node behavior, you may want to disable the check.
779 	 * However, you will  become vulnerable to attacks using IPv4 mapped
780 	 * source.
781 	 */
782 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
783 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
784 		ip6stat.ip6s_badscope++;
785 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
786 		goto bad;
787 	}
788 
789 	/*
790 	 * Tell launch routine the next header
791 	 */
792 	ip6stat.ip6s_delivered++;
793 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
794 	nest = 0;
795 
796 	rh_present = 0;
797 	while (nxt != IPPROTO_DONE) {
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 #ifdef IPSEC
840 		/*
841 		 * enforce IPsec policy checking if we are seeing last header.
842 		 * note that we do not visit this with protocols with pcb layer
843 		 * code - like udp/tcp/raw ip.
844 		 */
845 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) &&
846 		    ipsec6_in_reject(m, NULL)) {
847 			ipsec6stat.in_polvio++;
848 			goto bad;
849 		}
850 #endif
851 
852 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
853 	}
854 	goto bad2;
855 bad:
856 	m_freem(m);
857 bad2:
858 	;
859 	/* msg was embedded in the mbuf, do not reply! */
860 }
861 
862 /*
863  * set/grab in6_ifaddr correspond to IPv6 destination address.
864  * XXX backward compatibility wrapper
865  */
866 static struct ip6aux *
867 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
868 {
869 	struct ip6aux *n;
870 
871 	n = ip6_addaux(m);
872 	if (n)
873 		n->ip6a_dstia6 = ia6;
874 	return n;	/* NULL if failed to set */
875 }
876 
877 struct in6_ifaddr *
878 ip6_getdstifaddr(struct mbuf *m)
879 {
880 	struct ip6aux *n;
881 
882 	n = ip6_findaux(m);
883 	if (n)
884 		return n->ip6a_dstia6;
885 	else
886 		return NULL;
887 }
888 
889 /*
890  * Hop-by-Hop options header processing. If a valid jumbo payload option is
891  * included, the real payload length will be stored in plenp.
892  */
893 static int
894 ip6_hopopts_input(u_int32_t *plenp,
895 		  u_int32_t *rtalertp,/* XXX: should be stored more smart way */
896 		  struct mbuf **mp,
897 		  int *offp)
898 {
899 	struct mbuf *m = *mp;
900 	int off = *offp, hbhlen;
901 	struct ip6_hbh *hbh;
902 	u_int8_t *opt;
903 
904 	/* validation of the length of the header */
905 #ifndef PULLDOWN_TEST
906 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
907 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
908 	hbhlen = (hbh->ip6h_len + 1) << 3;
909 
910 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
911 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
912 #else
913 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
914 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
915 	if (hbh == NULL) {
916 		ip6stat.ip6s_tooshort++;
917 		return -1;
918 	}
919 	hbhlen = (hbh->ip6h_len + 1) << 3;
920 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
921 		hbhlen);
922 	if (hbh == NULL) {
923 		ip6stat.ip6s_tooshort++;
924 		return -1;
925 	}
926 #endif
927 	off += hbhlen;
928 	hbhlen -= sizeof(struct ip6_hbh);
929 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
930 
931 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
932 				hbhlen, rtalertp, plenp) < 0)
933 		return (-1);
934 
935 	*offp = off;
936 	*mp = m;
937 	return (0);
938 }
939 
940 /*
941  * Search header for all Hop-by-hop options and process each option.
942  * This function is separate from ip6_hopopts_input() in order to
943  * handle a case where the sending node itself process its hop-by-hop
944  * options header. In such a case, the function is called from ip6_output().
945  *
946  * The function assumes that hbh header is located right after the IPv6 header
947  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
948  * opthead + hbhlen is located in continuous memory region.
949  */
950 int
951 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
952 		    u_int32_t *rtalertp, u_int32_t *plenp)
953 {
954 	struct ip6_hdr *ip6;
955 	int optlen = 0;
956 	u_int8_t *opt = opthead;
957 	u_int16_t rtalert_val;
958 	u_int32_t jumboplen;
959 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
960 
961 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
962 		switch (*opt) {
963 		case IP6OPT_PAD1:
964 			optlen = 1;
965 			break;
966 		case IP6OPT_PADN:
967 			if (hbhlen < IP6OPT_MINLEN) {
968 				ip6stat.ip6s_toosmall++;
969 				goto bad;
970 			}
971 			optlen = *(opt + 1) + 2;
972 			break;
973 		case IP6OPT_RTALERT:
974 			/* XXX may need check for alignment */
975 			if (hbhlen < IP6OPT_RTALERT_LEN) {
976 				ip6stat.ip6s_toosmall++;
977 				goto bad;
978 			}
979 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
980 				/* XXX stat */
981 				icmp6_error(m, ICMP6_PARAM_PROB,
982 					    ICMP6_PARAMPROB_HEADER,
983 					    erroff + opt + 1 - opthead);
984 				return (-1);
985 			}
986 			optlen = IP6OPT_RTALERT_LEN;
987 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
988 			*rtalertp = ntohs(rtalert_val);
989 			break;
990 		case IP6OPT_JUMBO:
991 			/* XXX may need check for alignment */
992 			if (hbhlen < IP6OPT_JUMBO_LEN) {
993 				ip6stat.ip6s_toosmall++;
994 				goto bad;
995 			}
996 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
997 				/* XXX stat */
998 				icmp6_error(m, ICMP6_PARAM_PROB,
999 					    ICMP6_PARAMPROB_HEADER,
1000 					    erroff + opt + 1 - opthead);
1001 				return (-1);
1002 			}
1003 			optlen = IP6OPT_JUMBO_LEN;
1004 
1005 			/*
1006 			 * IPv6 packets that have non 0 payload length
1007 			 * must not contain a jumbo payload option.
1008 			 */
1009 			ip6 = mtod(m, struct ip6_hdr *);
1010 			if (ip6->ip6_plen) {
1011 				ip6stat.ip6s_badoptions++;
1012 				icmp6_error(m, ICMP6_PARAM_PROB,
1013 					    ICMP6_PARAMPROB_HEADER,
1014 					    erroff + opt - opthead);
1015 				return (-1);
1016 			}
1017 
1018 			/*
1019 			 * We may see jumbolen in unaligned location, so
1020 			 * we'd need to perform bcopy().
1021 			 */
1022 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1023 			jumboplen = (u_int32_t)htonl(jumboplen);
1024 
1025 #if 1
1026 			/*
1027 			 * if there are multiple jumbo payload options,
1028 			 * *plenp will be non-zero and the packet will be
1029 			 * rejected.
1030 			 * the behavior may need some debate in ipngwg -
1031 			 * multiple options does not make sense, however,
1032 			 * there's no explicit mention in specification.
1033 			 */
1034 			if (*plenp != 0) {
1035 				ip6stat.ip6s_badoptions++;
1036 				icmp6_error(m, ICMP6_PARAM_PROB,
1037 					    ICMP6_PARAMPROB_HEADER,
1038 					    erroff + opt + 2 - opthead);
1039 				return (-1);
1040 			}
1041 #endif
1042 
1043 			/*
1044 			 * jumbo payload length must be larger than 65535.
1045 			 */
1046 			if (jumboplen <= IPV6_MAXPACKET) {
1047 				ip6stat.ip6s_badoptions++;
1048 				icmp6_error(m, ICMP6_PARAM_PROB,
1049 					    ICMP6_PARAMPROB_HEADER,
1050 					    erroff + opt + 2 - opthead);
1051 				return (-1);
1052 			}
1053 			*plenp = jumboplen;
1054 
1055 			break;
1056 		default:		/* unknown option */
1057 			if (hbhlen < IP6OPT_MINLEN) {
1058 				ip6stat.ip6s_toosmall++;
1059 				goto bad;
1060 			}
1061 			optlen = ip6_unknown_opt(opt, m,
1062 			    erroff + opt - opthead);
1063 			if (optlen == -1)
1064 				return (-1);
1065 			optlen += 2;
1066 			break;
1067 		}
1068 	}
1069 
1070 	return (0);
1071 
1072 bad:
1073 	m_freem(m);
1074 	return (-1);
1075 }
1076 
1077 /*
1078  * Unknown option processing.
1079  * The third argument `off' is the offset from the IPv6 header to the option,
1080  * which is necessary if the IPv6 header the and option header and IPv6 header
1081  * is not continuous in order to return an ICMPv6 error.
1082  */
1083 int
1084 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1085 {
1086 	struct ip6_hdr *ip6;
1087 
1088 	switch (IP6OPT_TYPE(*optp)) {
1089 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1090 		return ((int)*(optp + 1));
1091 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1092 		m_freem(m);
1093 		return (-1);
1094 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1095 		ip6stat.ip6s_badoptions++;
1096 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1097 		return (-1);
1098 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1099 		ip6stat.ip6s_badoptions++;
1100 		ip6 = mtod(m, struct ip6_hdr *);
1101 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1102 		    (m->m_flags & (M_BCAST|M_MCAST)))
1103 			m_freem(m);
1104 		else
1105 			icmp6_error(m, ICMP6_PARAM_PROB,
1106 				    ICMP6_PARAMPROB_OPTION, off);
1107 		return (-1);
1108 	}
1109 
1110 	m_freem(m);		/* XXX: NOTREACHED */
1111 	return (-1);
1112 }
1113 
1114 /*
1115  * Create the "control" list for this pcb.
1116  * The function will not modify mbuf chain at all.
1117  *
1118  * with KAME mbuf chain restriction:
1119  * The routine will be called from upper layer handlers like tcp6_input().
1120  * Thus the routine assumes that the caller (tcp6_input) have already
1121  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1122  * very first mbuf on the mbuf chain.
1123  */
1124 void
1125 ip6_savecontrol(struct inpcb *in6p, struct mbuf **mp, struct ip6_hdr *ip6,
1126 		struct mbuf *m)
1127 {
1128 	#define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1129 	struct thread *td = curthread;	/* XXX */
1130 	int privileged = 0;
1131 	int rthdr_exist = 0;
1132 
1133 
1134 	if (priv_check(td, PRIV_ROOT) == 0)
1135  		privileged++;
1136 
1137 #ifdef SO_TIMESTAMP
1138 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
1139 		struct timeval tv;
1140 
1141 		microtime(&tv);
1142 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1143 				      SCM_TIMESTAMP, SOL_SOCKET);
1144 		if (*mp) {
1145 			mp = &(*mp)->m_next;
1146 		}
1147 	}
1148 #endif
1149 
1150 	/* RFC 2292 sec. 5 */
1151 	if (in6p->in6p_flags & IN6P_PKTINFO) {
1152 		struct in6_pktinfo pi6;
1153 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1154 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
1155 			pi6.ipi6_addr.s6_addr16[1] = 0;
1156 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
1157 					? m->m_pkthdr.rcvif->if_index
1158 					: 0;
1159 		*mp = sbcreatecontrol((caddr_t) &pi6,
1160 			sizeof(struct in6_pktinfo), IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO),
1161 			IPPROTO_IPV6);
1162 		if (*mp)
1163 			mp = &(*mp)->m_next;
1164 	}
1165 
1166 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
1167 		int hlim = ip6->ip6_hlim & 0xff;
1168 		*mp = sbcreatecontrol((caddr_t) &hlim,
1169 			sizeof(int), IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1170 		if (*mp)
1171 			mp = &(*mp)->m_next;
1172 	}
1173 
1174 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1175 		u_int32_t flowinfo;
1176 		int tclass;
1177 
1178 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1179 		flowinfo >>= 20;
1180 
1181 		tclass = flowinfo & 0xff;
1182 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1183 		    IPV6_TCLASS, IPPROTO_IPV6);
1184 		if (*mp)
1185 			mp = &(*mp)->m_next;
1186 	}
1187 
1188 	/*
1189 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1190 	 * for the option, but it might be too strict, since there might
1191 	 * be some hop-by-hop options which can be returned to normal user.
1192 	 * See RFC 2292 section 6.
1193 	 */
1194 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1195 		/*
1196 		 * Check if a hop-by-hop options header is contatined in the
1197 		 * received packet, and if so, store the options as ancillary
1198 		 * data. Note that a hop-by-hop options header must be
1199 		 * just after the IPv6 header, which fact is assured through
1200 		 * the IPv6 input processing.
1201 		 */
1202 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1203 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1204 			struct ip6_hbh *hbh;
1205 			int hbhlen = 0;
1206 #ifdef PULLDOWN_TEST
1207 			struct mbuf *ext;
1208 #endif
1209 
1210 #ifndef PULLDOWN_TEST
1211 			hbh = (struct ip6_hbh *)(ip6 + 1);
1212 			hbhlen = (hbh->ip6h_len + 1) << 3;
1213 #else
1214 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1215 			    ip6->ip6_nxt);
1216 			if (ext == NULL) {
1217 				ip6stat.ip6s_tooshort++;
1218 				return;
1219 			}
1220 			hbh = mtod(ext, struct ip6_hbh *);
1221 			hbhlen = (hbh->ip6h_len + 1) << 3;
1222 			if (hbhlen != ext->m_len) {
1223 				m_freem(ext);
1224 				ip6stat.ip6s_tooshort++;
1225 				return;
1226 			}
1227 #endif
1228 
1229 			/*
1230 			 * XXX: We copy whole the header even if a jumbo
1231 			 * payload option is included, which option is to
1232 			 * be removed before returning in the RFC 2292.
1233 			 * Note: this constraint is removed in 2292bis.
1234 			 */
1235 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1236 				 IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS), IPPROTO_IPV6);
1237 			if (*mp)
1238 				mp = &(*mp)->m_next;
1239 #ifdef PULLDOWN_TEST
1240 			m_freem(ext);
1241 #endif
1242 		}
1243 	}
1244 
1245 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1246 	if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1247 		int proto, off, nxt;
1248 
1249 		/*
1250 		 * go through the header chain to see if a routing header is
1251 		 * contained in the packet. We need this information to store
1252 		 * destination options headers (if any) properly.
1253 		 * XXX: performance issue. We should record this info when
1254 		 * processing extension headers in incoming routine.
1255 		 * (todo) use m_aux?
1256 		 */
1257 		proto = IPPROTO_IPV6;
1258 		off = 0;
1259 		nxt = -1;
1260 		while (1) {
1261 			int newoff;
1262 
1263 			newoff = ip6_nexthdr(m, off, proto, &nxt);
1264 			if (newoff < 0)
1265 				break;
1266 			if (newoff < off) /* invalid, check for safety */
1267 				break;
1268 			if ((proto = nxt) == IPPROTO_ROUTING) {
1269 				rthdr_exist = 1;
1270 				break;
1271 			}
1272 			off = newoff;
1273 		}
1274 	}
1275 
1276 	if ((in6p->in6p_flags &
1277 	     (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
1278 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1279 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1280 
1281 		/*
1282 		 * Search for destination options headers or routing
1283 		 * header(s) through the header chain, and stores each
1284 		 * header as ancillary data.
1285 		 * Note that the order of the headers remains in
1286 		 * the chain of ancillary data.
1287 		 */
1288 		while (1) {	/* is explicit loop prevention necessary? */
1289 			struct ip6_ext *ip6e = NULL;
1290 			int elen;
1291 #ifdef PULLDOWN_TEST
1292 			struct mbuf *ext = NULL;
1293 #endif
1294 
1295 			/*
1296 			 * if it is not an extension header, don't try to
1297 			 * pull it from the chain.
1298 			 */
1299 			switch (nxt) {
1300 			case IPPROTO_DSTOPTS:
1301 			case IPPROTO_ROUTING:
1302 			case IPPROTO_HOPOPTS:
1303 			case IPPROTO_AH: /* is it possible? */
1304 				break;
1305 			default:
1306 				goto loopend;
1307 			}
1308 
1309 #ifndef PULLDOWN_TEST
1310 			if (off + sizeof(*ip6e) > m->m_len)
1311 				goto loopend;
1312 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1313 			if (nxt == IPPROTO_AH)
1314 				elen = (ip6e->ip6e_len + 2) << 2;
1315 			else
1316 				elen = (ip6e->ip6e_len + 1) << 3;
1317 			if (off + elen > m->m_len)
1318 				goto loopend;
1319 #else
1320 			ext = ip6_pullexthdr(m, off, nxt);
1321 			if (ext == NULL) {
1322 				ip6stat.ip6s_tooshort++;
1323 				return;
1324 			}
1325 			ip6e = mtod(ext, struct ip6_ext *);
1326 			if (nxt == IPPROTO_AH)
1327 				elen = (ip6e->ip6e_len + 2) << 2;
1328 			else
1329 				elen = (ip6e->ip6e_len + 1) << 3;
1330 			if (elen != ext->m_len) {
1331 				m_freem(ext);
1332 				ip6stat.ip6s_tooshort++;
1333 				return;
1334 			}
1335 #endif
1336 
1337 			switch (nxt) {
1338 			case IPPROTO_DSTOPTS:
1339 				if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
1340 					break;
1341 
1342 				/*
1343 				 * We also require super-user privilege for
1344 				 * the option.
1345 				 * See the comments on IN6_HOPOPTS.
1346 				 */
1347 				if (!privileged)
1348 					break;
1349 
1350 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1351 						      IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1352 						      IPPROTO_IPV6);
1353 				if (*mp)
1354 					mp = &(*mp)->m_next;
1355 				break;
1356 			case IPPROTO_ROUTING:
1357 				if (!in6p->in6p_flags & IN6P_RTHDR)
1358 					break;
1359 
1360 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1361 						      IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1362 						      IPPROTO_IPV6);
1363 				if (*mp)
1364 					mp = &(*mp)->m_next;
1365 				break;
1366 			case IPPROTO_HOPOPTS:
1367 			case IPPROTO_AH: /* is it possible? */
1368 				break;
1369 
1370 			default:
1371 				/*
1372 			 	 * other cases have been filtered in the above.
1373 				 * none will visit this case.  here we supply
1374 				 * the code just in case (nxt overwritten or
1375 				 * other cases).
1376 				 */
1377 #ifdef PULLDOWN_TEST
1378 				m_freem(ext);
1379 #endif
1380 				goto loopend;
1381 
1382 			}
1383 
1384 			/* proceed with the next header. */
1385 			off += elen;
1386 			nxt = ip6e->ip6e_nxt;
1387 			ip6e = NULL;
1388 #ifdef PULLDOWN_TEST
1389 			m_freem(ext);
1390 			ext = NULL;
1391 #endif
1392 		}
1393 	  loopend:
1394 		;
1395 	}
1396 #undef IS2292
1397 }
1398 
1399 void
1400 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1401 {
1402 	struct socket *so;
1403 	struct mbuf *m_mtu;
1404 	struct ip6_mtuinfo mtuctl;
1405 
1406 	so =  in6p->inp_socket;
1407 
1408 	if (mtu == NULL)
1409 		return;
1410 
1411 #ifdef DIAGNOSTIC
1412 	if (so == NULL)		/* I believe this is impossible */
1413 		panic("ip6_notify_pmtu: socket is NULL");
1414 #endif
1415 
1416 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1417 	mtuctl.ip6m_mtu = *mtu;
1418 	mtuctl.ip6m_addr = *dst;
1419 
1420 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1421 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1422 		return;
1423 
1424 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1425 	    == 0) {
1426 		m_freem(m_mtu);
1427 		/* XXX: should count statistics */
1428 	} else
1429 		sorwakeup(so);
1430 
1431 	return;
1432 }
1433 
1434 #ifdef PULLDOWN_TEST
1435 /*
1436  * pull single extension header from mbuf chain.  returns single mbuf that
1437  * contains the result, or NULL on error.
1438  */
1439 static struct mbuf *
1440 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1441 {
1442 	struct ip6_ext ip6e;
1443 	size_t elen;
1444 	struct mbuf *n;
1445 
1446 #ifdef DIAGNOSTIC
1447 	switch (nxt) {
1448 	case IPPROTO_DSTOPTS:
1449 	case IPPROTO_ROUTING:
1450 	case IPPROTO_HOPOPTS:
1451 	case IPPROTO_AH: /* is it possible? */
1452 		break;
1453 	default:
1454 		kprintf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1455 	}
1456 #endif
1457 
1458 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1459 	if (nxt == IPPROTO_AH)
1460 		elen = (ip6e.ip6e_len + 2) << 2;
1461 	else
1462 		elen = (ip6e.ip6e_len + 1) << 3;
1463 
1464 	n = m_getb(elen, MB_DONTWAIT, MT_DATA, 0);
1465 	if (n == NULL)
1466 		return NULL;
1467 	n->m_len = 0;
1468 
1469 	if (elen >= M_TRAILINGSPACE(n)) {
1470 		m_free(n);
1471 		return NULL;
1472 	}
1473 
1474 	m_copydata(m, off, elen, mtod(n, caddr_t));
1475 	n->m_len = elen;
1476 	return n;
1477 }
1478 #endif
1479 
1480 /*
1481  * Get pointer to the previous header followed by the header
1482  * currently processed.
1483  * XXX: This function supposes that
1484  *	M includes all headers,
1485  *	the next header field and the header length field of each header
1486  *	are valid, and
1487  *	the sum of each header length equals to OFF.
1488  * Because of these assumptions, this function must be called very
1489  * carefully. Moreover, it will not be used in the near future when
1490  * we develop `neater' mechanism to process extension headers.
1491  */
1492 char *
1493 ip6_get_prevhdr(struct mbuf *m, int off)
1494 {
1495 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1496 
1497 	if (off == sizeof(struct ip6_hdr))
1498 		return (&ip6->ip6_nxt);
1499 	else {
1500 		int len, nxt;
1501 		struct ip6_ext *ip6e = NULL;
1502 
1503 		nxt = ip6->ip6_nxt;
1504 		len = sizeof(struct ip6_hdr);
1505 		while (len < off) {
1506 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1507 
1508 			switch (nxt) {
1509 			case IPPROTO_FRAGMENT:
1510 				len += sizeof(struct ip6_frag);
1511 				break;
1512 			case IPPROTO_AH:
1513 				len += (ip6e->ip6e_len + 2) << 2;
1514 				break;
1515 			default:
1516 				len += (ip6e->ip6e_len + 1) << 3;
1517 				break;
1518 			}
1519 			nxt = ip6e->ip6e_nxt;
1520 		}
1521 		if (ip6e)
1522 			return (&ip6e->ip6e_nxt);
1523 		else
1524 			return NULL;
1525 	}
1526 }
1527 
1528 /*
1529  * get next header offset.  m will be retained.
1530  */
1531 int
1532 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1533 {
1534 	struct ip6_hdr ip6;
1535 	struct ip6_ext ip6e;
1536 	struct ip6_frag fh;
1537 
1538 	/* just in case */
1539 	if (m == NULL)
1540 		panic("ip6_nexthdr: m == NULL");
1541 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1542 		return -1;
1543 
1544 	switch (proto) {
1545 	case IPPROTO_IPV6:
1546 		if (m->m_pkthdr.len < off + sizeof(ip6))
1547 			return -1;
1548 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1549 		if (nxtp)
1550 			*nxtp = ip6.ip6_nxt;
1551 		off += sizeof(ip6);
1552 		return off;
1553 
1554 	case IPPROTO_FRAGMENT:
1555 		/*
1556 		 * terminate parsing if it is not the first fragment,
1557 		 * it does not make sense to parse through it.
1558 		 */
1559 		if (m->m_pkthdr.len < off + sizeof(fh))
1560 			return -1;
1561 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1562 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1563 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1564 			return -1;
1565 		if (nxtp)
1566 			*nxtp = fh.ip6f_nxt;
1567 		off += sizeof(struct ip6_frag);
1568 		return off;
1569 
1570 	case IPPROTO_AH:
1571 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1572 			return -1;
1573 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1574 		if (nxtp)
1575 			*nxtp = ip6e.ip6e_nxt;
1576 		off += (ip6e.ip6e_len + 2) << 2;
1577 		return off;
1578 
1579 	case IPPROTO_HOPOPTS:
1580 	case IPPROTO_ROUTING:
1581 	case IPPROTO_DSTOPTS:
1582 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1583 			return -1;
1584 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1585 		if (nxtp)
1586 			*nxtp = ip6e.ip6e_nxt;
1587 		off += (ip6e.ip6e_len + 1) << 3;
1588 		return off;
1589 
1590 	case IPPROTO_NONE:
1591 	case IPPROTO_ESP:
1592 	case IPPROTO_IPCOMP:
1593 		/* give up */
1594 		return -1;
1595 
1596 	default:
1597 		return -1;
1598 	}
1599 
1600 	return -1;
1601 }
1602 
1603 /*
1604  * get offset for the last header in the chain.  m will be kept untainted.
1605  */
1606 int
1607 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1608 {
1609 	int newoff;
1610 	int nxt;
1611 
1612 	if (!nxtp) {
1613 		nxt = -1;
1614 		nxtp = &nxt;
1615 	}
1616 	while (1) {
1617 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1618 		if (newoff < 0)
1619 			return off;
1620 		else if (newoff < off)
1621 			return -1;	/* invalid */
1622 		else if (newoff == off)
1623 			return newoff;
1624 
1625 		off = newoff;
1626 		proto = *nxtp;
1627 	}
1628 }
1629 
1630 struct ip6aux *
1631 ip6_addaux(struct mbuf *m)
1632 {
1633 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1634 	if (!tag) {
1635 		tag = m_tag_get(PACKET_TAG_IPV6_INPUT,
1636 				sizeof (struct ip6aux),
1637 				MB_DONTWAIT);
1638 		if (tag)
1639 			m_tag_prepend(m, tag);
1640 	}
1641 	if (tag)
1642 		bzero(m_tag_data(tag), sizeof (struct ip6aux));
1643 	return tag ? (struct ip6aux *)m_tag_data(tag) : NULL;
1644 }
1645 
1646 struct ip6aux *
1647 ip6_findaux(struct mbuf *m)
1648 {
1649 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1650 	return tag ? (struct ip6aux*)(tag+1) : NULL;
1651 }
1652 
1653 void
1654 ip6_delaux(struct mbuf *m)
1655 {
1656 	struct m_tag *tag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1657 	if (tag)
1658 		m_tag_delete(m, tag);
1659 }
1660 
1661 /*
1662  * System control for IP6
1663  */
1664 
1665 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1666 	0,		0,		0,		0,
1667 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1668 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1669 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1670 	0,		0,		0,		0,
1671 	ENOPROTOOPT
1672 };
1673