xref: /freebsd/sys/netinet6/ip6_input.c (revision 9608d7e2cd58c1a7fff6562810f2ce519e6ec50a)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_ip6fw.h"
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 #include "opt_pfil_hooks.h"
73 #include "opt_random_ip_id.h"
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/proc.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/errno.h>
85 #include <sys/time.h>
86 #include <sys/kernel.h>
87 #include <sys/syslog.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 #ifdef PFIL_HOOKS
95 #include <net/pfil.h>
96 #endif
97 
98 #include <netinet/in.h>
99 #include <netinet/in_systm.h>
100 #ifdef INET
101 #include <netinet/ip.h>
102 #include <netinet/ip_icmp.h>
103 #endif /* INET */
104 #include <netinet/ip6.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet/in_pcb.h>
108 #include <netinet/icmp6.h>
109 #include <netinet6/scope6_var.h>
110 #include <netinet6/in6_ifattach.h>
111 #include <netinet6/nd6.h>
112 #include <netinet6/in6_prefix.h>
113 
114 #ifdef IPSEC
115 #include <netinet6/ipsec.h>
116 #ifdef INET6
117 #include <netinet6/ipsec6.h>
118 #endif
119 #endif
120 
121 #ifdef FAST_IPSEC
122 #include <netipsec/ipsec.h>
123 #include <netipsec/ipsec6.h>
124 #define	IPSEC
125 #endif /* FAST_IPSEC */
126 
127 #include <netinet6/ip6_fw.h>
128 
129 #include <netinet6/ip6protosw.h>
130 
131 #include <net/net_osdep.h>
132 
133 extern struct domain inet6domain;
134 
135 u_char ip6_protox[IPPROTO_MAX];
136 static struct ifqueue ip6intrq;
137 static int ip6qmaxlen = IFQ_MAXLEN;
138 struct in6_ifaddr *in6_ifaddr;
139 
140 extern struct callout in6_tmpaddrtimer_ch;
141 
142 int ip6_forward_srcrt;			/* XXX */
143 int ip6_sourcecheck;			/* XXX */
144 int ip6_sourcecheck_interval;		/* XXX */
145 
146 int ip6_ours_check_algorithm;
147 
148 #ifdef PFIL_HOOKS
149 struct pfil_head inet6_pfil_hook;
150 #endif
151 
152 /* firewall hooks */
153 ip6_fw_chk_t *ip6_fw_chk_ptr;
154 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
155 int ip6_fw_enable = 1;
156 
157 struct ip6stat ip6stat;
158 
159 static void ip6_init2 __P((void *));
160 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
161 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
162 #ifdef PULLDOWN_TEST
163 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
164 #endif
165 
166 /*
167  * IP6 initialization: fill in IP6 protocol switch table.
168  * All protocols not implemented in kernel go to raw IP6 protocol handler.
169  */
170 void
171 ip6_init()
172 {
173 	struct ip6protosw *pr;
174 	int i;
175 
176 #ifdef DIAGNOSTIC
177 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
178 		panic("sizeof(protosw) != sizeof(ip6protosw)");
179 #endif
180 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
181 	if (pr == 0)
182 		panic("ip6_init");
183 	for (i = 0; i < IPPROTO_MAX; i++)
184 		ip6_protox[i] = pr - inet6sw;
185 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
186 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
187 		if (pr->pr_domain->dom_family == PF_INET6 &&
188 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
189 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
190 #ifdef PFIL_HOOKS
191 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
192 	inet6_pfil_hook.ph_af = AF_INET6;
193 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
194 		printf("%s: WARNING: unable to register pfil hook, "
195 			"error %d\n", __func__, i);
196 #endif /* PFIL_HOOKS */
197 	ip6intrq.ifq_maxlen = ip6qmaxlen;
198 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
199 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
200 	scope6_init();
201 	addrsel_policy_init();
202 	nd6_init();
203 	frag6_init();
204 #ifndef RANDOM_IP_ID
205 	ip6_flow_seq = arc4random();
206 #endif
207 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
208 }
209 
210 static void
211 ip6_init2(dummy)
212 	void *dummy;
213 {
214 
215 	/* nd6_timer_init */
216 	callout_init(&nd6_timer_ch, 0);
217 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
218 
219 	/* router renumbering prefix list maintenance */
220 	callout_init(&in6_rr_timer_ch, 0);
221 	callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
222 
223 	/* timer for regeneranation of temporary addresses randomize ID */
224 	callout_init(&in6_tmpaddrtimer_ch, 0);
225 	callout_reset(&in6_tmpaddrtimer_ch,
226 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
227 		       ip6_temp_regen_advance) * hz,
228 		      in6_tmpaddrtimer, NULL);
229 }
230 
231 /* cheat */
232 /* This must be after route_init(), which is now SI_ORDER_THIRD */
233 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
234 
235 extern struct	route_in6 ip6_forward_rt;
236 
237 void
238 ip6_input(m)
239 	struct mbuf *m;
240 {
241 	struct ip6_hdr *ip6;
242 	int off = sizeof(struct ip6_hdr), nest;
243 	u_int32_t plen;
244 	u_int32_t rtalert = ~0;
245 	int nxt, ours = 0;
246 	struct ifnet *deliverifp = NULL;
247 #ifdef PFIL_HOOKS
248 	struct in6_addr odst;
249 #endif
250 	int srcrt = 0;
251 
252 	GIANT_REQUIRED;			/* XXX for now */
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_tag.
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 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
304 		if (n)
305 			M_MOVE_PKTHDR(n, m);
306 		if (n && n->m_pkthdr.len > MHLEN) {
307 			MCLGET(n, M_DONTWAIT);
308 			if ((n->m_flags & M_EXT) == 0) {
309 				m_freem(n);
310 				n = NULL;
311 			}
312 		}
313 		if (n == NULL) {
314 			m_freem(m);
315 			return;	/* ENOBUFS */
316 		}
317 
318 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
319 		n->m_len = n->m_pkthdr.len;
320 		m_freem(m);
321 		m = n;
322 	}
323 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
324 #endif
325 
326 	if (m->m_len < sizeof(struct ip6_hdr)) {
327 		struct ifnet *inifp;
328 		inifp = m->m_pkthdr.rcvif;
329 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
330 			ip6stat.ip6s_toosmall++;
331 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
332 			return;
333 		}
334 	}
335 
336 	ip6 = mtod(m, struct ip6_hdr *);
337 
338 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
339 		ip6stat.ip6s_badvers++;
340 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
341 		goto bad;
342 	}
343 
344 #ifdef PFIL_HOOKS
345 	/*
346 	 * Run through list of hooks for input packets.
347 	 *
348 	 * NB: Beware of the destination address changing
349 	 *     (e.g. by NAT rewriting).  When this happens,
350 	 *     tell ip6_forward to do the right thing.
351 	 */
352 	odst = ip6->ip6_dst;
353 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
354 		return;
355 	if (m == NULL)			/* consumed by filter */
356 		return;
357 	ip6 = mtod(m, struct ip6_hdr *);
358 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
359 #endif /* PFIL_HOOKS */
360 
361 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
362 
363 	/*
364 	 * Check with the firewall...
365 	 */
366 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
367 		u_short port = 0;
368 		/* If ipfw says divert, we have to just drop packet */
369 		/* use port as a dummy argument */
370 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
371 			m_freem(m);
372 			m = NULL;
373 		}
374 		if (!m)
375 			return;
376 	}
377 
378 	/*
379 	 * Check against address spoofing/corruption.
380 	 */
381 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
382 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
383 		/*
384 		 * XXX: "badscope" is not very suitable for a multicast source.
385 		 */
386 		ip6stat.ip6s_badscope++;
387 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
388 		goto bad;
389 	}
390 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
391 	     IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
392 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
393 		ip6stat.ip6s_badscope++;
394 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
395 		goto bad;
396 	}
397 
398 	/*
399 	 * The following check is not documented in specs.  A malicious
400 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
401 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
402 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
403 	 *
404 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
405 	 * support IPv4-less kernel compilation, we cannot support SIIT
406 	 * environment at all.  So, it makes more sense for us to reject any
407 	 * malicious packets for non-SIIT environment, than try to do a
408 	 * partial support for SIIT environment.
409 	 */
410 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
411 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
412 		ip6stat.ip6s_badscope++;
413 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
414 		goto bad;
415 	}
416 #if 0
417 	/*
418 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
419 	 *
420 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
421 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
422 	 * is revised to forbid relaying case.
423 	 */
424 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
425 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
426 		ip6stat.ip6s_badscope++;
427 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
428 		goto bad;
429 	}
430 #endif
431 
432 	/* drop packets if interface ID portion is already filled */
433 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
434 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
435 		    ip6->ip6_src.s6_addr16[1]) {
436 			ip6stat.ip6s_badscope++;
437 			goto bad;
438 		}
439 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
440 		     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 	/*
455 	 * Multicast check
456 	 */
457 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
458 	  	struct in6_multi *in6m = 0;
459 
460 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
461 		/*
462 		 * See if we belong to the destination multicast group on the
463 		 * arrival interface.
464 		 */
465 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
466 		if (in6m)
467 			ours = 1;
468 		else if (!ip6_mrouter) {
469 			ip6stat.ip6s_notmember++;
470 			ip6stat.ip6s_cantforward++;
471 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
472 			goto bad;
473 		}
474 		deliverifp = m->m_pkthdr.rcvif;
475 		goto hbhcheck;
476 	}
477 
478 	/*
479 	 *  Unicast check
480 	 */
481 	if (ip6_forward_rt.ro_rt != NULL &&
482 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
483 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
484 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
485 		ip6stat.ip6s_forward_cachehit++;
486 	else {
487 		struct sockaddr_in6 *dst6;
488 
489 		if (ip6_forward_rt.ro_rt) {
490 			/* route is down or destination is different */
491 			ip6stat.ip6s_forward_cachemiss++;
492 			RTFREE(ip6_forward_rt.ro_rt);
493 			ip6_forward_rt.ro_rt = 0;
494 		}
495 
496 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
497 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
498 		dst6->sin6_len = sizeof(struct sockaddr_in6);
499 		dst6->sin6_family = AF_INET6;
500 		dst6->sin6_addr = ip6->ip6_dst;
501 
502 		rtalloc((struct route *)&ip6_forward_rt);
503 	}
504 
505 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
506 
507 	/*
508 	 * Accept the packet if the forwarding interface to the destination
509 	 * according to the routing table is the loopback interface,
510 	 * unless the associated route has a gateway.
511 	 * Note that this approach causes to accept a packet if there is a
512 	 * route to the loopback interface for the destination of the packet.
513 	 * But we think it's even useful in some situations, e.g. when using
514 	 * a special daemon which wants to intercept the packet.
515 	 *
516 	 * XXX: some OSes automatically make a cloned route for the destination
517 	 * of an outgoing packet.  If the outgoing interface of the packet
518 	 * is a loopback one, the kernel would consider the packet to be
519 	 * accepted, even if we have no such address assinged on the interface.
520 	 * We check the cloned flag of the route entry to reject such cases,
521 	 * assuming that route entries for our own addresses are not made by
522 	 * cloning (it should be true because in6_addloop explicitly installs
523 	 * the host route).  However, we might have to do an explicit check
524 	 * while it would be less efficient.  Or, should we rather install a
525 	 * reject route for such a case?
526 	 */
527 	if (ip6_forward_rt.ro_rt &&
528 	    (ip6_forward_rt.ro_rt->rt_flags &
529 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
530 #ifdef RTF_WASCLONED
531 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
532 #endif
533 #ifdef RTF_CLONED
534 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
535 #endif
536 #if 0
537 	    /*
538 	     * The check below is redundant since the comparison of
539 	     * the destination and the key of the rtentry has
540 	     * already done through looking up the routing table.
541 	     */
542 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
543 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
544 #endif
545 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
546 		struct in6_ifaddr *ia6 =
547 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
548 
549 		/*
550 		 * record address information into m_tag.
551 		 */
552 		(void)ip6_setdstifaddr(m, ia6);
553 
554 		/*
555 		 * packets to a tentative, duplicated, or somehow invalid
556 		 * address must not be accepted.
557 		 */
558 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
559 			/* this address is ready */
560 			ours = 1;
561 			deliverifp = ia6->ia_ifp;	/* correct? */
562 			/* Count the packet in the ip address stats */
563 			ia6->ia_ifa.if_ipackets++;
564 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
565 			goto hbhcheck;
566 		} else {
567 			/* address is not ready, so discard the packet. */
568 			nd6log((LOG_INFO,
569 			    "ip6_input: packet to an unready address %s->%s\n",
570 			    ip6_sprintf(&ip6->ip6_src),
571 			    ip6_sprintf(&ip6->ip6_dst)));
572 
573 			goto bad;
574 		}
575 	}
576 
577 	/*
578 	 * FAITH (Firewall Aided Internet Translator)
579 	 */
580 	if (ip6_keepfaith) {
581 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
582 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
583 			/* XXX do we need more sanity checks? */
584 			ours = 1;
585 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
586 			goto hbhcheck;
587 		}
588 	}
589 
590 	/*
591 	 * Now there is no reason to process the packet if it's not our own
592 	 * and we're not a router.
593 	 */
594 	if (!ip6_forwarding) {
595 		ip6stat.ip6s_cantforward++;
596 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
597 		goto bad;
598 	}
599 
600   hbhcheck:
601 	/*
602 	 * record address information into m_tag, if we don't have one yet.
603 	 * note that we are unable to record it, if the address is not listed
604 	 * as our interface address (e.g. multicast addresses, addresses
605 	 * within FAITH prefixes and such).
606 	 */
607 	if (deliverifp && !ip6_getdstifaddr(m)) {
608 		struct in6_ifaddr *ia6;
609 
610 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
611 		if (ia6) {
612 			if (!ip6_setdstifaddr(m, ia6)) {
613 				/*
614 				 * XXX maybe we should drop the packet here,
615 				 * as we could not provide enough information
616 				 * to the upper layers.
617 				 */
618 			}
619 		}
620 	}
621 
622 	/*
623 	 * Process Hop-by-Hop options header if it's contained.
624 	 * m may be modified in ip6_hopopts_input().
625 	 * If a JumboPayload option is included, plen will also be modified.
626 	 */
627 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
628 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
629 		struct ip6_hbh *hbh;
630 
631 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
632 #if 0	/*touches NULL pointer*/
633 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
634 #endif
635 			return;	/* m have already been freed */
636 		}
637 
638 		/* adjust pointer */
639 		ip6 = mtod(m, struct ip6_hdr *);
640 
641 		/*
642 		 * if the payload length field is 0 and the next header field
643 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
644 		 * option MUST be included.
645 		 */
646 		if (ip6->ip6_plen == 0 && plen == 0) {
647 			/*
648 			 * Note that if a valid jumbo payload option is
649 			 * contained, ip6_hopopts_input() must set a valid
650 			 * (non-zero) payload length to the variable plen.
651 			 */
652 			ip6stat.ip6s_badoptions++;
653 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
654 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
655 			icmp6_error(m, ICMP6_PARAM_PROB,
656 				    ICMP6_PARAMPROB_HEADER,
657 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
658 			return;
659 		}
660 #ifndef PULLDOWN_TEST
661 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
662 		hbh = (struct ip6_hbh *)(ip6 + 1);
663 #else
664 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
665 			sizeof(struct ip6_hbh));
666 		if (hbh == NULL) {
667 			ip6stat.ip6s_tooshort++;
668 			return;
669 		}
670 #endif
671 		nxt = hbh->ip6h_nxt;
672 
673 		/*
674 		 * accept the packet if a router alert option is included
675 		 * and we act as an IPv6 router.
676 		 */
677 		if (rtalert != ~0 && ip6_forwarding)
678 			ours = 1;
679 	} else
680 		nxt = ip6->ip6_nxt;
681 
682 	/*
683 	 * Check that the amount of data in the buffers
684 	 * is as at least much as the IPv6 header would have us expect.
685 	 * Trim mbufs if longer than we expect.
686 	 * Drop packet if shorter than we expect.
687 	 */
688 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
689 		ip6stat.ip6s_tooshort++;
690 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
691 		goto bad;
692 	}
693 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
694 		if (m->m_len == m->m_pkthdr.len) {
695 			m->m_len = sizeof(struct ip6_hdr) + plen;
696 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
697 		} else
698 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
699 	}
700 
701 	/*
702 	 * Forward if desirable.
703 	 */
704 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
705 		/*
706 		 * If we are acting as a multicast router, all
707 		 * incoming multicast packets are passed to the
708 		 * kernel-level multicast forwarding function.
709 		 * The packet is returned (relatively) intact; if
710 		 * ip6_mforward() returns a non-zero value, the packet
711 		 * must be discarded, else it may be accepted below.
712 		 */
713 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
714 			ip6stat.ip6s_cantforward++;
715 			m_freem(m);
716 			return;
717 		}
718 		if (!ours) {
719 			m_freem(m);
720 			return;
721 		}
722 	} else if (!ours) {
723 		ip6_forward(m, srcrt);
724 		return;
725 	}
726 
727 	ip6 = mtod(m, struct ip6_hdr *);
728 
729 	/*
730 	 * Malicious party may be able to use IPv4 mapped addr to confuse
731 	 * tcp/udp stack and bypass security checks (act as if it was from
732 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
733 	 *
734 	 * For SIIT end node behavior, you may want to disable the check.
735 	 * However, you will  become vulnerable to attacks using IPv4 mapped
736 	 * source.
737 	 */
738 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
739 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
740 		ip6stat.ip6s_badscope++;
741 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
742 		goto bad;
743 	}
744 
745 	/*
746 	 * Tell launch routine the next header
747 	 */
748 	ip6stat.ip6s_delivered++;
749 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
750 	nest = 0;
751 
752 	while (nxt != IPPROTO_DONE) {
753 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
754 			ip6stat.ip6s_toomanyhdr++;
755 			goto bad;
756 		}
757 
758 		/*
759 		 * protection against faulty packet - there should be
760 		 * more sanity checks in header chain processing.
761 		 */
762 		if (m->m_pkthdr.len < off) {
763 			ip6stat.ip6s_tooshort++;
764 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
765 			goto bad;
766 		}
767 
768 #ifdef IPSEC
769 		/*
770 		 * enforce IPsec policy checking if we are seeing last header.
771 		 * note that we do not visit this with protocols with pcb layer
772 		 * code - like udp/tcp/raw ip.
773 		 */
774 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
775 		    ipsec6_in_reject(m, NULL)) {
776 			ipsec6stat.in_polvio++;
777 			goto bad;
778 		}
779 #endif
780 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
781 	}
782 	return;
783  bad:
784 	m_freem(m);
785 }
786 
787 /*
788  * set/grab in6_ifaddr correspond to IPv6 destination address.
789  * XXX backward compatibility wrapper
790  */
791 static struct ip6aux *
792 ip6_setdstifaddr(m, ia6)
793 	struct mbuf *m;
794 	struct in6_ifaddr *ia6;
795 {
796 	struct ip6aux *ip6a;
797 
798 	ip6a = ip6_addaux(m);
799 	if (ip6a)
800 		ip6a->ip6a_dstia6 = ia6;
801 	return ip6a;	/* NULL if failed to set */
802 }
803 
804 struct in6_ifaddr *
805 ip6_getdstifaddr(m)
806 	struct mbuf *m;
807 {
808 	struct ip6aux *ip6a;
809 
810 	ip6a = ip6_findaux(m);
811 	if (ip6a)
812 		return ip6a->ip6a_dstia6;
813 	else
814 		return NULL;
815 }
816 
817 /*
818  * Hop-by-Hop options header processing. If a valid jumbo payload option is
819  * included, the real payload length will be stored in plenp.
820  */
821 static int
822 ip6_hopopts_input(plenp, rtalertp, mp, offp)
823 	u_int32_t *plenp;
824 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
825 	struct mbuf **mp;
826 	int *offp;
827 {
828 	struct mbuf *m = *mp;
829 	int off = *offp, hbhlen;
830 	struct ip6_hbh *hbh;
831 	u_int8_t *opt;
832 
833 	/* validation of the length of the header */
834 #ifndef PULLDOWN_TEST
835 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
836 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
837 	hbhlen = (hbh->ip6h_len + 1) << 3;
838 
839 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
840 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
841 #else
842 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
843 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
844 	if (hbh == NULL) {
845 		ip6stat.ip6s_tooshort++;
846 		return -1;
847 	}
848 	hbhlen = (hbh->ip6h_len + 1) << 3;
849 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
850 		hbhlen);
851 	if (hbh == NULL) {
852 		ip6stat.ip6s_tooshort++;
853 		return -1;
854 	}
855 #endif
856 	off += hbhlen;
857 	hbhlen -= sizeof(struct ip6_hbh);
858 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
859 
860 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
861 				hbhlen, rtalertp, plenp) < 0)
862 		return (-1);
863 
864 	*offp = off;
865 	*mp = m;
866 	return (0);
867 }
868 
869 /*
870  * Search header for all Hop-by-hop options and process each option.
871  * This function is separate from ip6_hopopts_input() in order to
872  * handle a case where the sending node itself process its hop-by-hop
873  * options header. In such a case, the function is called from ip6_output().
874  *
875  * The function assumes that hbh header is located right after the IPv6 header
876  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
877  * opthead + hbhlen is located in continuous memory region.
878  */
879 int
880 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
881 	struct mbuf *m;
882 	u_int8_t *opthead;
883 	int hbhlen;
884 	u_int32_t *rtalertp;
885 	u_int32_t *plenp;
886 {
887 	struct ip6_hdr *ip6;
888 	int optlen = 0;
889 	u_int8_t *opt = opthead;
890 	u_int16_t rtalert_val;
891 	u_int32_t jumboplen;
892 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
893 
894 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
895 		switch (*opt) {
896 		case IP6OPT_PAD1:
897 			optlen = 1;
898 			break;
899 		case IP6OPT_PADN:
900 			if (hbhlen < IP6OPT_MINLEN) {
901 				ip6stat.ip6s_toosmall++;
902 				goto bad;
903 			}
904 			optlen = *(opt + 1) + 2;
905 			break;
906 		case IP6OPT_ROUTER_ALERT:
907 			/* XXX may need check for alignment */
908 			if (hbhlen < IP6OPT_RTALERT_LEN) {
909 				ip6stat.ip6s_toosmall++;
910 				goto bad;
911 			}
912 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
913 				/* XXX stat */
914 				icmp6_error(m, ICMP6_PARAM_PROB,
915 				    ICMP6_PARAMPROB_HEADER,
916 				    erroff + opt + 1 - opthead);
917 				return (-1);
918 			}
919 			optlen = IP6OPT_RTALERT_LEN;
920 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
921 			*rtalertp = ntohs(rtalert_val);
922 			break;
923 		case IP6OPT_JUMBO:
924 			/* XXX may need check for alignment */
925 			if (hbhlen < IP6OPT_JUMBO_LEN) {
926 				ip6stat.ip6s_toosmall++;
927 				goto bad;
928 			}
929 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
930 				/* XXX stat */
931 				icmp6_error(m, ICMP6_PARAM_PROB,
932 				    ICMP6_PARAMPROB_HEADER,
933 				    erroff + opt + 1 - opthead);
934 				return (-1);
935 			}
936 			optlen = IP6OPT_JUMBO_LEN;
937 
938 			/*
939 			 * IPv6 packets that have non 0 payload length
940 			 * must not contain a jumbo payload option.
941 			 */
942 			ip6 = mtod(m, struct ip6_hdr *);
943 			if (ip6->ip6_plen) {
944 				ip6stat.ip6s_badoptions++;
945 				icmp6_error(m, ICMP6_PARAM_PROB,
946 				    ICMP6_PARAMPROB_HEADER,
947 				    erroff + opt - opthead);
948 				return (-1);
949 			}
950 
951 			/*
952 			 * We may see jumbolen in unaligned location, so
953 			 * we'd need to perform bcopy().
954 			 */
955 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
956 			jumboplen = (u_int32_t)htonl(jumboplen);
957 
958 #if 1
959 			/*
960 			 * if there are multiple jumbo payload options,
961 			 * *plenp will be non-zero and the packet will be
962 			 * rejected.
963 			 * the behavior may need some debate in ipngwg -
964 			 * multiple options does not make sense, however,
965 			 * there's no explicit mention in specification.
966 			 */
967 			if (*plenp != 0) {
968 				ip6stat.ip6s_badoptions++;
969 				icmp6_error(m, ICMP6_PARAM_PROB,
970 				    ICMP6_PARAMPROB_HEADER,
971 				    erroff + opt + 2 - opthead);
972 				return (-1);
973 			}
974 #endif
975 
976 			/*
977 			 * jumbo payload length must be larger than 65535.
978 			 */
979 			if (jumboplen <= IPV6_MAXPACKET) {
980 				ip6stat.ip6s_badoptions++;
981 				icmp6_error(m, ICMP6_PARAM_PROB,
982 				    ICMP6_PARAMPROB_HEADER,
983 				    erroff + opt + 2 - opthead);
984 				return (-1);
985 			}
986 			*plenp = jumboplen;
987 
988 			break;
989 		default:		/* unknown option */
990 			if (hbhlen < IP6OPT_MINLEN) {
991 				ip6stat.ip6s_toosmall++;
992 				goto bad;
993 			}
994 			optlen = ip6_unknown_opt(opt, m,
995 			    erroff + opt - opthead);
996 			if (optlen == -1)
997 				return (-1);
998 			optlen += 2;
999 			break;
1000 		}
1001 	}
1002 
1003 	return (0);
1004 
1005   bad:
1006 	m_freem(m);
1007 	return (-1);
1008 }
1009 
1010 /*
1011  * Unknown option processing.
1012  * The third argument `off' is the offset from the IPv6 header to the option,
1013  * which is necessary if the IPv6 header the and option header and IPv6 header
1014  * is not continuous in order to return an ICMPv6 error.
1015  */
1016 int
1017 ip6_unknown_opt(optp, m, off)
1018 	u_int8_t *optp;
1019 	struct mbuf *m;
1020 	int off;
1021 {
1022 	struct ip6_hdr *ip6;
1023 
1024 	switch (IP6OPT_TYPE(*optp)) {
1025 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1026 		return ((int)*(optp + 1));
1027 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1028 		m_freem(m);
1029 		return (-1);
1030 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1031 		ip6stat.ip6s_badoptions++;
1032 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1033 		return (-1);
1034 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1035 		ip6stat.ip6s_badoptions++;
1036 		ip6 = mtod(m, struct ip6_hdr *);
1037 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1038 		    (m->m_flags & (M_BCAST|M_MCAST)))
1039 			m_freem(m);
1040 		else
1041 			icmp6_error(m, ICMP6_PARAM_PROB,
1042 				    ICMP6_PARAMPROB_OPTION, off);
1043 		return (-1);
1044 	}
1045 
1046 	m_freem(m);		/* XXX: NOTREACHED */
1047 	return (-1);
1048 }
1049 
1050 /*
1051  * Create the "control" list for this pcb.
1052  * The function will not modify mbuf chain at all.
1053  *
1054  * with KAME mbuf chain restriction:
1055  * The routine will be called from upper layer handlers like tcp6_input().
1056  * Thus the routine assumes that the caller (tcp6_input) have already
1057  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1058  * very first mbuf on the mbuf chain.
1059  */
1060 void
1061 ip6_savecontrol(in6p, m, mp)
1062 	struct inpcb *in6p;
1063 	struct mbuf *m, **mp;
1064 {
1065 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1066 	struct thread *td = curthread;	/* XXX */
1067 	int privileged = 0;
1068 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1069 
1070 	if (td && !suser(td))
1071 		privileged++;
1072 
1073 #ifdef SO_TIMESTAMP
1074 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1075 		struct timeval tv;
1076 
1077 		microtime(&tv);
1078 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1079 		    SCM_TIMESTAMP, SOL_SOCKET);
1080 		if (*mp)
1081 			mp = &(*mp)->m_next;
1082 	}
1083 #endif
1084 
1085 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1086 		return;
1087 
1088 	/* RFC 2292 sec. 5 */
1089 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1090 		struct in6_pktinfo pi6;
1091 
1092 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1093 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1094 		pi6.ipi6_ifindex =
1095 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1096 
1097 		*mp = sbcreatecontrol((caddr_t) &pi6,
1098 		    sizeof(struct in6_pktinfo),
1099 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1100 		if (*mp)
1101 			mp = &(*mp)->m_next;
1102 	}
1103 
1104 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1105 		int hlim = ip6->ip6_hlim & 0xff;
1106 
1107 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1108 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1109 		if (*mp)
1110 			mp = &(*mp)->m_next;
1111 	}
1112 
1113 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1114 		u_int32_t flowinfo;
1115 		int tclass;
1116 
1117 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1118 		flowinfo >>= 20;
1119 
1120 		tclass = flowinfo & 0xff;
1121 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1122 		    IPV6_TCLASS, IPPROTO_IPV6);
1123 		if (*mp)
1124 			mp = &(*mp)->m_next;
1125 	}
1126 
1127 	/*
1128 	 * IPV6_HOPOPTS socket option. We require super-user privilege
1129 	 * for the option, but it might be too strict, since there might
1130 	 * be some hop-by-hop options which can be returned to normal user.
1131 	 * See RFC 2292 section 6.
1132 	 */
1133 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1134 #ifdef DIAGNOSTIC
1135 		if (!privileged)
1136 			panic("IN6P_HOPOPTS is set for unprivileged socket");
1137 #endif
1138 		/*
1139 		 * Check if a hop-by-hop options header is contatined in the
1140 		 * received packet, and if so, store the options as ancillary
1141 		 * data. Note that a hop-by-hop options header must be
1142 		 * just after the IPv6 header, which is assured through the
1143 		 * IPv6 input processing.
1144 		 */
1145 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1146 			struct ip6_hbh *hbh;
1147 			int hbhlen = 0;
1148 #ifdef PULLDOWN_TEST
1149 			struct mbuf *ext;
1150 #endif
1151 
1152 #ifndef PULLDOWN_TEST
1153 			hbh = (struct ip6_hbh *)(ip6 + 1);
1154 			hbhlen = (hbh->ip6h_len + 1) << 3;
1155 #else
1156 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1157 			    ip6->ip6_nxt);
1158 			if (ext == NULL) {
1159 				ip6stat.ip6s_tooshort++;
1160 				return;
1161 			}
1162 			hbh = mtod(ext, struct ip6_hbh *);
1163 			hbhlen = (hbh->ip6h_len + 1) << 3;
1164 			if (hbhlen != ext->m_len) {
1165 				m_freem(ext);
1166 				ip6stat.ip6s_tooshort++;
1167 				return;
1168 			}
1169 #endif
1170 
1171 			/*
1172 			 * XXX: We copy the whole header even if a
1173 			 * jumbo payload option is included, the option which
1174 			 * is to be removed before returning according to
1175 			 * RFC2292.
1176 			 * Note: this constraint is removed in 2292bis.
1177 			 */
1178 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1179 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1180 			    IPPROTO_IPV6);
1181 			if (*mp)
1182 				mp = &(*mp)->m_next;
1183 #ifdef PULLDOWN_TEST
1184 			m_freem(ext);
1185 #endif
1186 		}
1187 	}
1188 
1189 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1190 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1191 
1192 		/*
1193 		 * Search for destination options headers or routing
1194 		 * header(s) through the header chain, and stores each
1195 		 * header as ancillary data.
1196 		 * Note that the order of the headers remains in
1197 		 * the chain of ancillary data.
1198 		 */
1199 		while (1) {	/* is explicit loop prevention necessary? */
1200 			struct ip6_ext *ip6e = NULL;
1201 			int elen;
1202 #ifdef PULLDOWN_TEST
1203 			struct mbuf *ext = NULL;
1204 #endif
1205 
1206 			/*
1207 			 * if it is not an extension header, don't try to
1208 			 * pull it from the chain.
1209 			 */
1210 			switch (nxt) {
1211 			case IPPROTO_DSTOPTS:
1212 			case IPPROTO_ROUTING:
1213 			case IPPROTO_HOPOPTS:
1214 			case IPPROTO_AH: /* is it possible? */
1215 				break;
1216 			default:
1217 				goto loopend;
1218 			}
1219 
1220 #ifndef PULLDOWN_TEST
1221 			if (off + sizeof(*ip6e) > m->m_len)
1222 				goto loopend;
1223 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1224 			if (nxt == IPPROTO_AH)
1225 				elen = (ip6e->ip6e_len + 2) << 2;
1226 			else
1227 				elen = (ip6e->ip6e_len + 1) << 3;
1228 			if (off + elen > m->m_len)
1229 				goto loopend;
1230 #else
1231 			ext = ip6_pullexthdr(m, off, nxt);
1232 			if (ext == NULL) {
1233 				ip6stat.ip6s_tooshort++;
1234 				return;
1235 			}
1236 			ip6e = mtod(ext, struct ip6_ext *);
1237 			if (nxt == IPPROTO_AH)
1238 				elen = (ip6e->ip6e_len + 2) << 2;
1239 			else
1240 				elen = (ip6e->ip6e_len + 1) << 3;
1241 			if (elen != ext->m_len) {
1242 				m_freem(ext);
1243 				ip6stat.ip6s_tooshort++;
1244 				return;
1245 			}
1246 #endif
1247 
1248 			switch (nxt) {
1249 			case IPPROTO_DSTOPTS:
1250 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1251 					break;
1252 
1253 				/*
1254 				 * We also require super-user privilege for
1255 				 * the option.  See comments on IN6_HOPOPTS.
1256 				 */
1257 				if (!privileged)
1258 					break;
1259 
1260 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1261 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1262 				    IPPROTO_IPV6);
1263 				if (*mp)
1264 					mp = &(*mp)->m_next;
1265 				break;
1266 			case IPPROTO_ROUTING:
1267 				if (!in6p->in6p_flags & IN6P_RTHDR)
1268 					break;
1269 
1270 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1271 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1272 				    IPPROTO_IPV6);
1273 				if (*mp)
1274 					mp = &(*mp)->m_next;
1275 				break;
1276 			case IPPROTO_HOPOPTS:
1277 			case IPPROTO_AH: /* is it possible? */
1278 				break;
1279 
1280 			default:
1281 				/*
1282 			 	 * other cases have been filtered in the above.
1283 				 * none will visit this case.  here we supply
1284 				 * the code just in case (nxt overwritten or
1285 				 * other cases).
1286 				 */
1287 #ifdef PULLDOWN_TEST
1288 				m_freem(ext);
1289 #endif
1290 				goto loopend;
1291 
1292 			}
1293 
1294 			/* proceed with the next header. */
1295 			off += elen;
1296 			nxt = ip6e->ip6e_nxt;
1297 			ip6e = NULL;
1298 #ifdef PULLDOWN_TEST
1299 			m_freem(ext);
1300 			ext = NULL;
1301 #endif
1302 		}
1303 	  loopend:
1304 		;
1305 	}
1306 
1307 #undef IS2292
1308 }
1309 
1310 #ifdef PULLDOWN_TEST
1311 /*
1312  * pull single extension header from mbuf chain.  returns single mbuf that
1313  * contains the result, or NULL on error.
1314  */
1315 static struct mbuf *
1316 ip6_pullexthdr(m, off, nxt)
1317 	struct mbuf *m;
1318 	size_t off;
1319 	int nxt;
1320 {
1321 	struct ip6_ext ip6e;
1322 	size_t elen;
1323 	struct mbuf *n;
1324 
1325 #ifdef DIAGNOSTIC
1326 	switch (nxt) {
1327 	case IPPROTO_DSTOPTS:
1328 	case IPPROTO_ROUTING:
1329 	case IPPROTO_HOPOPTS:
1330 	case IPPROTO_AH: /* is it possible? */
1331 		break;
1332 	default:
1333 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1334 	}
1335 #endif
1336 
1337 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1338 	if (nxt == IPPROTO_AH)
1339 		elen = (ip6e.ip6e_len + 2) << 2;
1340 	else
1341 		elen = (ip6e.ip6e_len + 1) << 3;
1342 
1343 	MGET(n, M_DONTWAIT, MT_DATA);
1344 	if (n && elen >= MLEN) {
1345 		MCLGET(n, M_DONTWAIT);
1346 		if ((n->m_flags & M_EXT) == 0) {
1347 			m_free(n);
1348 			n = NULL;
1349 		}
1350 	}
1351 	if (!n)
1352 		return NULL;
1353 
1354 	n->m_len = 0;
1355 	if (elen >= M_TRAILINGSPACE(n)) {
1356 		m_free(n);
1357 		return NULL;
1358 	}
1359 
1360 	m_copydata(m, off, elen, mtod(n, caddr_t));
1361 	n->m_len = elen;
1362 	return n;
1363 }
1364 #endif
1365 
1366 /*
1367  * Get pointer to the previous header followed by the header
1368  * currently processed.
1369  * XXX: This function supposes that
1370  *	M includes all headers,
1371  *	the next header field and the header length field of each header
1372  *	are valid, and
1373  *	the sum of each header length equals to OFF.
1374  * Because of these assumptions, this function must be called very
1375  * carefully. Moreover, it will not be used in the near future when
1376  * we develop `neater' mechanism to process extension headers.
1377  */
1378 char *
1379 ip6_get_prevhdr(m, off)
1380 	struct mbuf *m;
1381 	int off;
1382 {
1383 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1384 
1385 	if (off == sizeof(struct ip6_hdr))
1386 		return (&ip6->ip6_nxt);
1387 	else {
1388 		int len, nxt;
1389 		struct ip6_ext *ip6e = NULL;
1390 
1391 		nxt = ip6->ip6_nxt;
1392 		len = sizeof(struct ip6_hdr);
1393 		while (len < off) {
1394 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1395 
1396 			switch (nxt) {
1397 			case IPPROTO_FRAGMENT:
1398 				len += sizeof(struct ip6_frag);
1399 				break;
1400 			case IPPROTO_AH:
1401 				len += (ip6e->ip6e_len + 2) << 2;
1402 				break;
1403 			default:
1404 				len += (ip6e->ip6e_len + 1) << 3;
1405 				break;
1406 			}
1407 			nxt = ip6e->ip6e_nxt;
1408 		}
1409 		if (ip6e)
1410 			return (&ip6e->ip6e_nxt);
1411 		else
1412 			return NULL;
1413 	}
1414 }
1415 
1416 /*
1417  * get next header offset.  m will be retained.
1418  */
1419 int
1420 ip6_nexthdr(m, off, proto, nxtp)
1421 	struct mbuf *m;
1422 	int off;
1423 	int proto;
1424 	int *nxtp;
1425 {
1426 	struct ip6_hdr ip6;
1427 	struct ip6_ext ip6e;
1428 	struct ip6_frag fh;
1429 
1430 	/* just in case */
1431 	if (m == NULL)
1432 		panic("ip6_nexthdr: m == NULL");
1433 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1434 		return -1;
1435 
1436 	switch (proto) {
1437 	case IPPROTO_IPV6:
1438 		if (m->m_pkthdr.len < off + sizeof(ip6))
1439 			return -1;
1440 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1441 		if (nxtp)
1442 			*nxtp = ip6.ip6_nxt;
1443 		off += sizeof(ip6);
1444 		return off;
1445 
1446 	case IPPROTO_FRAGMENT:
1447 		/*
1448 		 * terminate parsing if it is not the first fragment,
1449 		 * it does not make sense to parse through it.
1450 		 */
1451 		if (m->m_pkthdr.len < off + sizeof(fh))
1452 			return -1;
1453 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1454 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1455 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1456 			return -1;
1457 		if (nxtp)
1458 			*nxtp = fh.ip6f_nxt;
1459 		off += sizeof(struct ip6_frag);
1460 		return off;
1461 
1462 	case IPPROTO_AH:
1463 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1464 			return -1;
1465 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1466 		if (nxtp)
1467 			*nxtp = ip6e.ip6e_nxt;
1468 		off += (ip6e.ip6e_len + 2) << 2;
1469 		return off;
1470 
1471 	case IPPROTO_HOPOPTS:
1472 	case IPPROTO_ROUTING:
1473 	case IPPROTO_DSTOPTS:
1474 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1475 			return -1;
1476 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1477 		if (nxtp)
1478 			*nxtp = ip6e.ip6e_nxt;
1479 		off += (ip6e.ip6e_len + 1) << 3;
1480 		return off;
1481 
1482 	case IPPROTO_NONE:
1483 	case IPPROTO_ESP:
1484 	case IPPROTO_IPCOMP:
1485 		/* give up */
1486 		return -1;
1487 
1488 	default:
1489 		return -1;
1490 	}
1491 
1492 	return -1;
1493 }
1494 
1495 /*
1496  * get offset for the last header in the chain.  m will be kept untainted.
1497  */
1498 int
1499 ip6_lasthdr(m, off, proto, nxtp)
1500 	struct mbuf *m;
1501 	int off;
1502 	int proto;
1503 	int *nxtp;
1504 {
1505 	int newoff;
1506 	int nxt;
1507 
1508 	if (!nxtp) {
1509 		nxt = -1;
1510 		nxtp = &nxt;
1511 	}
1512 	while (1) {
1513 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1514 		if (newoff < 0)
1515 			return off;
1516 		else if (newoff < off)
1517 			return -1;	/* invalid */
1518 		else if (newoff == off)
1519 			return newoff;
1520 
1521 		off = newoff;
1522 		proto = *nxtp;
1523 	}
1524 }
1525 
1526 struct ip6aux *
1527 ip6_addaux(m)
1528 	struct mbuf *m;
1529 {
1530 	struct m_tag *mtag;
1531 
1532 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1533 	if (!mtag) {
1534 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1535 		    M_NOWAIT);
1536 		if (mtag) {
1537 			m_tag_prepend(m, mtag);
1538 			bzero(mtag + 1, sizeof(struct ip6aux));
1539 		}
1540 	}
1541 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1542 }
1543 
1544 struct ip6aux *
1545 ip6_findaux(m)
1546 	struct mbuf *m;
1547 {
1548 	struct m_tag *mtag;
1549 
1550 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1551 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1552 }
1553 
1554 void
1555 ip6_delaux(m)
1556 	struct mbuf *m;
1557 {
1558 	struct m_tag *mtag;
1559 
1560 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1561 	if (mtag)
1562 		m_tag_delete(m, mtag);
1563 }
1564 
1565 /*
1566  * System control for IP6
1567  */
1568 
1569 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1570 	0,		0,		0,		0,
1571 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1572 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1573 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1574 	0,		0,		0,		0,
1575 	ENOPROTOOPT
1576 };
1577