xref: /freebsd/sys/netinet6/ip6_input.c (revision 74bf4e164ba5851606a27d4feff27717452583e5)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/proc.h>
74 #include <sys/domain.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 
83 #include <net/if.h>
84 #include <net/if_types.h>
85 #include <net/if_dl.h>
86 #include <net/route.h>
87 #include <net/netisr.h>
88 #include <net/pfil.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_systm.h>
92 #ifdef INET
93 #include <netinet/ip.h>
94 #include <netinet/ip_icmp.h>
95 #endif /* INET */
96 #include <netinet/ip6.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/scope6_var.h>
102 #include <netinet6/in6_ifattach.h>
103 #include <netinet6/nd6.h>
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #ifdef INET6
108 #include <netinet6/ipsec6.h>
109 #endif
110 #endif
111 
112 #ifdef FAST_IPSEC
113 #include <netipsec/ipsec.h>
114 #include <netipsec/ipsec6.h>
115 #define	IPSEC
116 #endif /* FAST_IPSEC */
117 
118 #include <netinet6/ip6_fw.h>
119 
120 #include <netinet6/ip6protosw.h>
121 
122 #include <net/net_osdep.h>
123 
124 extern struct domain inet6domain;
125 
126 u_char ip6_protox[IPPROTO_MAX];
127 static struct ifqueue ip6intrq;
128 static int ip6qmaxlen = IFQ_MAXLEN;
129 struct in6_ifaddr *in6_ifaddr;
130 
131 extern struct callout in6_tmpaddrtimer_ch;
132 
133 int ip6_forward_srcrt;			/* XXX */
134 int ip6_sourcecheck;			/* XXX */
135 int ip6_sourcecheck_interval;		/* XXX */
136 
137 int ip6_ours_check_algorithm;
138 
139 struct pfil_head inet6_pfil_hook;
140 
141 /* firewall hooks */
142 ip6_fw_chk_t *ip6_fw_chk_ptr;
143 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
144 int ip6_fw_enable = 1;
145 
146 struct ip6stat ip6stat;
147 
148 static void ip6_init2 __P((void *));
149 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
150 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
151 #ifdef PULLDOWN_TEST
152 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
153 #endif
154 
155 /*
156  * IP6 initialization: fill in IP6 protocol switch table.
157  * All protocols not implemented in kernel go to raw IP6 protocol handler.
158  */
159 void
160 ip6_init()
161 {
162 	struct ip6protosw *pr;
163 	int i;
164 
165 #ifdef DIAGNOSTIC
166 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
167 		panic("sizeof(protosw) != sizeof(ip6protosw)");
168 #endif
169 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
170 	if (pr == 0)
171 		panic("ip6_init");
172 	for (i = 0; i < IPPROTO_MAX; i++)
173 		ip6_protox[i] = pr - inet6sw;
174 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
175 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
176 		if (pr->pr_domain->dom_family == PF_INET6 &&
177 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
178 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
179 
180 	/* Initialize packet filter hooks. */
181 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
182 	inet6_pfil_hook.ph_af = AF_INET6;
183 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
184 		printf("%s: WARNING: unable to register pfil hook, "
185 			"error %d\n", __func__, i);
186 
187 	ip6intrq.ifq_maxlen = ip6qmaxlen;
188 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
189 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
190 	scope6_init();
191 	addrsel_policy_init();
192 	nd6_init();
193 	frag6_init();
194 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
195 }
196 
197 static void
198 ip6_init2(dummy)
199 	void *dummy;
200 {
201 
202 	/* nd6_timer_init */
203 	callout_init(&nd6_timer_ch, 0);
204 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
205 
206 	/* timer for regeneranation of temporary addresses randomize ID */
207 	callout_init(&in6_tmpaddrtimer_ch, 0);
208 	callout_reset(&in6_tmpaddrtimer_ch,
209 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
210 		       ip6_temp_regen_advance) * hz,
211 		      in6_tmpaddrtimer, NULL);
212 }
213 
214 /* cheat */
215 /* This must be after route_init(), which is now SI_ORDER_THIRD */
216 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
217 
218 extern struct	route_in6 ip6_forward_rt;
219 
220 void
221 ip6_input(m)
222 	struct mbuf *m;
223 {
224 	struct ip6_hdr *ip6;
225 	int off = sizeof(struct ip6_hdr), nest;
226 	u_int32_t plen;
227 	u_int32_t rtalert = ~0;
228 	int nxt, ours = 0;
229 	struct ifnet *deliverifp = NULL;
230 	struct sockaddr_in6 sa6;
231 	u_int32_t srczone, dstzone;
232 	struct in6_addr odst;
233 	int srcrt = 0;
234 
235 	GIANT_REQUIRED;			/* XXX for now */
236 #ifdef IPSEC
237 	/*
238 	 * should the inner packet be considered authentic?
239 	 * see comment in ah4_input().
240 	 */
241 	if (m) {
242 		m->m_flags &= ~M_AUTHIPHDR;
243 		m->m_flags &= ~M_AUTHIPDGM;
244 	}
245 #endif
246 
247 	/*
248 	 * make sure we don't have onion peering information into m_tag.
249 	 */
250 	ip6_delaux(m);
251 
252 	/*
253 	 * mbuf statistics
254 	 */
255 	if (m->m_flags & M_EXT) {
256 		if (m->m_next)
257 			ip6stat.ip6s_mext2m++;
258 		else
259 			ip6stat.ip6s_mext1++;
260 	} else {
261 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
262 		if (m->m_next) {
263 			if (m->m_flags & M_LOOP) {
264 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
265 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
266 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
267 			else
268 				ip6stat.ip6s_m2m[0]++;
269 		} else
270 			ip6stat.ip6s_m1++;
271 #undef M2MMAX
272 	}
273 
274 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
275 	ip6stat.ip6s_total++;
276 
277 #ifndef PULLDOWN_TEST
278 	/*
279 	 * L2 bridge code and some other code can return mbuf chain
280 	 * that does not conform to KAME requirement.  too bad.
281 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
282 	 */
283 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
284 		struct mbuf *n;
285 
286 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
287 		if (n)
288 			M_MOVE_PKTHDR(n, m);
289 		if (n && n->m_pkthdr.len > MHLEN) {
290 			MCLGET(n, M_DONTWAIT);
291 			if ((n->m_flags & M_EXT) == 0) {
292 				m_freem(n);
293 				n = NULL;
294 			}
295 		}
296 		if (n == NULL) {
297 			m_freem(m);
298 			return;	/* ENOBUFS */
299 		}
300 
301 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
302 		n->m_len = n->m_pkthdr.len;
303 		m_freem(m);
304 		m = n;
305 	}
306 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
307 #endif
308 
309 	if (m->m_len < sizeof(struct ip6_hdr)) {
310 		struct ifnet *inifp;
311 		inifp = m->m_pkthdr.rcvif;
312 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
313 			ip6stat.ip6s_toosmall++;
314 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
315 			return;
316 		}
317 	}
318 
319 	ip6 = mtod(m, struct ip6_hdr *);
320 
321 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
322 		ip6stat.ip6s_badvers++;
323 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
324 		goto bad;
325 	}
326 
327 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
328 
329 	/*
330 	 * Check against address spoofing/corruption.
331 	 */
332 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
333 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
334 		/*
335 		 * XXX: "badscope" is not very suitable for a multicast source.
336 		 */
337 		ip6stat.ip6s_badscope++;
338 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
339 		goto bad;
340 	}
341 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
342 	    !(m->m_flags & M_LOOP)) {
343 		/*
344 		 * In this case, the packet should come from the loopback
345 		 * interface.  However, we cannot just check the if_flags,
346 		 * because ip6_mloopback() passes the "actual" interface
347 		 * as the outgoing/incoming interface.
348 		 */
349 		ip6stat.ip6s_badscope++;
350 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
351 		goto bad;
352 	}
353 
354 #ifdef ALTQ
355 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
356 		/* packet is dropped by traffic conditioner */
357 		return;
358 	}
359 #endif
360 	/*
361 	 * The following check is not documented in specs.  A malicious
362 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
363 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
364 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
365 	 *
366 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
367 	 * support IPv4-less kernel compilation, we cannot support SIIT
368 	 * environment at all.  So, it makes more sense for us to reject any
369 	 * malicious packets for non-SIIT environment, than try to do a
370 	 * partial support for SIIT environment.
371 	 */
372 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
373 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
374 		ip6stat.ip6s_badscope++;
375 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
376 		goto bad;
377 	}
378 #if 0
379 	/*
380 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
381 	 *
382 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
383 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
384 	 * is revised to forbid relaying case.
385 	 */
386 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
387 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
388 		ip6stat.ip6s_badscope++;
389 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
390 		goto bad;
391 	}
392 #endif
393 
394 	/*
395 	 * Drop packets if the link ID portion is already filled.
396 	 * XXX: this is technically not a good behavior.  But, we internally
397 	 * use the field to disambiguate link-local addresses, so we cannot
398 	 * be generous against those a bit strange addresses.
399 	 */
400 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
401 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
402 		    ip6->ip6_src.s6_addr16[1]) {
403 			ip6stat.ip6s_badscope++;
404 			goto bad;
405 		}
406 		if ((IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
407 		     IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) &&
408 		    ip6->ip6_dst.s6_addr16[1]) {
409 			ip6stat.ip6s_badscope++;
410 			goto bad;
411 		}
412 	}
413 
414 	/*
415 	 * Run through list of hooks for input packets.
416 	 *
417 	 * NB: Beware of the destination address changing
418 	 *     (e.g. by NAT rewriting).  When this happens,
419 	 *     tell ip6_forward to do the right thing.
420 	 */
421 	odst = ip6->ip6_dst;
422 
423 	/* Jump over all PFIL processing if hooks are not active. */
424 	if (inet6_pfil_hook.ph_busy_count == -1)
425 		goto passin;
426 
427 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
428 		return;
429 	if (m == NULL)			/* consumed by filter */
430 		return;
431 	ip6 = mtod(m, struct ip6_hdr *);
432 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
433 
434 passin:
435 	/*
436 	 * Check with the firewall...
437 	 */
438 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
439 		u_short port = 0;
440 		/* If ipfw says divert, we have to just drop packet */
441 		/* use port as a dummy argument */
442 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
443 			m_freem(m);
444 			m = NULL;
445 		}
446 		if (!m)
447 			return;
448 	}
449 
450 	/*
451 	 * construct source and destination address structures with
452 	 * disambiguating their scope zones (if there is ambiguity).
453 	 * XXX: sin6_family and sin6_len will NOT be referred to, but we fill
454 	 * in these fields just in case.
455 	 */
456 	if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_src, &srczone) ||
457 	    in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_dst, &dstzone)) {
458 		/*
459 		 * Note that these generic checks cover cases that src or
460 		 * dst are the loopback address and the receiving interface
461 		 * is not loopback.
462 		 */
463 		ip6stat.ip6s_badscope++;
464 		goto bad;
465 	}
466 
467 	bzero(&sa6, sizeof(sa6));
468 	sa6.sin6_family = AF_INET6;
469 	sa6.sin6_len = sizeof(struct sockaddr_in6);
470 
471 	sa6.sin6_addr = ip6->ip6_src;
472 	sa6.sin6_scope_id = srczone;
473 	if (in6_embedscope(&ip6->ip6_src, &sa6, NULL, NULL)) {
474 		/* XXX: should not happen */
475 		ip6stat.ip6s_badscope++;
476 		goto bad;
477 	}
478 
479 	sa6.sin6_addr = ip6->ip6_dst;
480 	sa6.sin6_scope_id = dstzone;
481 	if (in6_embedscope(&ip6->ip6_dst, &sa6, NULL, NULL)) {
482 		/* XXX: should not happen */
483 		ip6stat.ip6s_badscope++;
484 		goto bad;
485 	}
486 
487 	/* XXX: ff01::%ifN awareness is not merged, yet. */
488 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_src))
489 		ip6->ip6_src.s6_addr16[1] = 0;
490 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst))
491 		ip6->ip6_dst.s6_addr16[1] = 0;
492 
493 	/*
494 	 * Multicast check
495 	 */
496 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
497 	  	struct in6_multi *in6m = 0;
498 
499 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
500 		/*
501 		 * See if we belong to the destination multicast group on the
502 		 * arrival interface.
503 		 */
504 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
505 		if (in6m)
506 			ours = 1;
507 		else if (!ip6_mrouter) {
508 			ip6stat.ip6s_notmember++;
509 			ip6stat.ip6s_cantforward++;
510 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
511 			goto bad;
512 		}
513 		deliverifp = m->m_pkthdr.rcvif;
514 		goto hbhcheck;
515 	}
516 
517 	/*
518 	 *  Unicast check
519 	 */
520 	if (ip6_forward_rt.ro_rt != NULL &&
521 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
522 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
523 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
524 		ip6stat.ip6s_forward_cachehit++;
525 	else {
526 		struct sockaddr_in6 *dst6;
527 
528 		if (ip6_forward_rt.ro_rt) {
529 			/* route is down or destination is different */
530 			ip6stat.ip6s_forward_cachemiss++;
531 			RTFREE(ip6_forward_rt.ro_rt);
532 			ip6_forward_rt.ro_rt = 0;
533 		}
534 
535 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
536 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
537 		dst6->sin6_len = sizeof(struct sockaddr_in6);
538 		dst6->sin6_family = AF_INET6;
539 		dst6->sin6_addr = ip6->ip6_dst;
540 
541 		rtalloc((struct route *)&ip6_forward_rt);
542 	}
543 
544 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
545 
546 	/*
547 	 * Accept the packet if the forwarding interface to the destination
548 	 * according to the routing table is the loopback interface,
549 	 * unless the associated route has a gateway.
550 	 * Note that this approach causes to accept a packet if there is a
551 	 * route to the loopback interface for the destination of the packet.
552 	 * But we think it's even useful in some situations, e.g. when using
553 	 * a special daemon which wants to intercept the packet.
554 	 *
555 	 * XXX: some OSes automatically make a cloned route for the destination
556 	 * of an outgoing packet.  If the outgoing interface of the packet
557 	 * is a loopback one, the kernel would consider the packet to be
558 	 * accepted, even if we have no such address assinged on the interface.
559 	 * We check the cloned flag of the route entry to reject such cases,
560 	 * assuming that route entries for our own addresses are not made by
561 	 * cloning (it should be true because in6_addloop explicitly installs
562 	 * the host route).  However, we might have to do an explicit check
563 	 * while it would be less efficient.  Or, should we rather install a
564 	 * reject route for such a case?
565 	 */
566 	if (ip6_forward_rt.ro_rt &&
567 	    (ip6_forward_rt.ro_rt->rt_flags &
568 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
569 #ifdef RTF_WASCLONED
570 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
571 #endif
572 #ifdef RTF_CLONED
573 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
574 #endif
575 #if 0
576 	    /*
577 	     * The check below is redundant since the comparison of
578 	     * the destination and the key of the rtentry has
579 	     * already done through looking up the routing table.
580 	     */
581 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
582 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
583 #endif
584 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
585 		struct in6_ifaddr *ia6 =
586 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
587 
588 		/*
589 		 * record address information into m_tag.
590 		 */
591 		(void)ip6_setdstifaddr(m, ia6);
592 
593 		/*
594 		 * packets to a tentative, duplicated, or somehow invalid
595 		 * address must not be accepted.
596 		 */
597 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
598 			/* this address is ready */
599 			ours = 1;
600 			deliverifp = ia6->ia_ifp;	/* correct? */
601 			/* Count the packet in the ip address stats */
602 			ia6->ia_ifa.if_ipackets++;
603 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
604 			goto hbhcheck;
605 		} else {
606 			/* address is not ready, so discard the packet. */
607 			nd6log((LOG_INFO,
608 			    "ip6_input: packet to an unready address %s->%s\n",
609 			    ip6_sprintf(&ip6->ip6_src),
610 			    ip6_sprintf(&ip6->ip6_dst)));
611 
612 			goto bad;
613 		}
614 	}
615 
616 	/*
617 	 * FAITH (Firewall Aided Internet Translator)
618 	 */
619 	if (ip6_keepfaith) {
620 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
621 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
622 			/* XXX do we need more sanity checks? */
623 			ours = 1;
624 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
625 			goto hbhcheck;
626 		}
627 	}
628 
629 	/*
630 	 * Now there is no reason to process the packet if it's not our own
631 	 * and we're not a router.
632 	 */
633 	if (!ip6_forwarding) {
634 		ip6stat.ip6s_cantforward++;
635 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
636 		goto bad;
637 	}
638 
639   hbhcheck:
640 	/*
641 	 * record address information into m_tag, if we don't have one yet.
642 	 * note that we are unable to record it, if the address is not listed
643 	 * as our interface address (e.g. multicast addresses, addresses
644 	 * within FAITH prefixes and such).
645 	 */
646 	if (deliverifp && !ip6_getdstifaddr(m)) {
647 		struct in6_ifaddr *ia6;
648 
649 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
650 		if (ia6) {
651 			if (!ip6_setdstifaddr(m, ia6)) {
652 				/*
653 				 * XXX maybe we should drop the packet here,
654 				 * as we could not provide enough information
655 				 * to the upper layers.
656 				 */
657 			}
658 		}
659 	}
660 
661 	/*
662 	 * Process Hop-by-Hop options header if it's contained.
663 	 * m may be modified in ip6_hopopts_input().
664 	 * If a JumboPayload option is included, plen will also be modified.
665 	 */
666 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
667 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
668 		struct ip6_hbh *hbh;
669 
670 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
671 #if 0	/*touches NULL pointer*/
672 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
673 #endif
674 			return;	/* m have already been freed */
675 		}
676 
677 		/* adjust pointer */
678 		ip6 = mtod(m, struct ip6_hdr *);
679 
680 		/*
681 		 * if the payload length field is 0 and the next header field
682 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
683 		 * option MUST be included.
684 		 */
685 		if (ip6->ip6_plen == 0 && plen == 0) {
686 			/*
687 			 * Note that if a valid jumbo payload option is
688 			 * contained, ip6_hopopts_input() must set a valid
689 			 * (non-zero) payload length to the variable plen.
690 			 */
691 			ip6stat.ip6s_badoptions++;
692 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
693 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
694 			icmp6_error(m, ICMP6_PARAM_PROB,
695 				    ICMP6_PARAMPROB_HEADER,
696 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
697 			return;
698 		}
699 #ifndef PULLDOWN_TEST
700 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
701 		hbh = (struct ip6_hbh *)(ip6 + 1);
702 #else
703 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
704 			sizeof(struct ip6_hbh));
705 		if (hbh == NULL) {
706 			ip6stat.ip6s_tooshort++;
707 			return;
708 		}
709 #endif
710 		nxt = hbh->ip6h_nxt;
711 
712 		/*
713 		 * accept the packet if a router alert option is included
714 		 * and we act as an IPv6 router.
715 		 */
716 		if (rtalert != ~0 && ip6_forwarding)
717 			ours = 1;
718 	} else
719 		nxt = ip6->ip6_nxt;
720 
721 	/*
722 	 * Check that the amount of data in the buffers
723 	 * is as at least much as the IPv6 header would have us expect.
724 	 * Trim mbufs if longer than we expect.
725 	 * Drop packet if shorter than we expect.
726 	 */
727 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
728 		ip6stat.ip6s_tooshort++;
729 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
730 		goto bad;
731 	}
732 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
733 		if (m->m_len == m->m_pkthdr.len) {
734 			m->m_len = sizeof(struct ip6_hdr) + plen;
735 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
736 		} else
737 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
738 	}
739 
740 	/*
741 	 * Forward if desirable.
742 	 */
743 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
744 		/*
745 		 * If we are acting as a multicast router, all
746 		 * incoming multicast packets are passed to the
747 		 * kernel-level multicast forwarding function.
748 		 * The packet is returned (relatively) intact; if
749 		 * ip6_mforward() returns a non-zero value, the packet
750 		 * must be discarded, else it may be accepted below.
751 		 */
752 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
753 			ip6stat.ip6s_cantforward++;
754 			m_freem(m);
755 			return;
756 		}
757 		if (!ours) {
758 			m_freem(m);
759 			return;
760 		}
761 	} else if (!ours) {
762 		ip6_forward(m, srcrt);
763 		return;
764 	}
765 
766 	ip6 = mtod(m, struct ip6_hdr *);
767 
768 	/*
769 	 * Malicious party may be able to use IPv4 mapped addr to confuse
770 	 * tcp/udp stack and bypass security checks (act as if it was from
771 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
772 	 *
773 	 * For SIIT end node behavior, you may want to disable the check.
774 	 * However, you will  become vulnerable to attacks using IPv4 mapped
775 	 * source.
776 	 */
777 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
778 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
779 		ip6stat.ip6s_badscope++;
780 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
781 		goto bad;
782 	}
783 
784 	/*
785 	 * Tell launch routine the next header
786 	 */
787 	ip6stat.ip6s_delivered++;
788 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
789 	nest = 0;
790 
791 	while (nxt != IPPROTO_DONE) {
792 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
793 			ip6stat.ip6s_toomanyhdr++;
794 			goto bad;
795 		}
796 
797 		/*
798 		 * protection against faulty packet - there should be
799 		 * more sanity checks in header chain processing.
800 		 */
801 		if (m->m_pkthdr.len < off) {
802 			ip6stat.ip6s_tooshort++;
803 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
804 			goto bad;
805 		}
806 
807 #ifdef IPSEC
808 		/*
809 		 * enforce IPsec policy checking if we are seeing last header.
810 		 * note that we do not visit this with protocols with pcb layer
811 		 * code - like udp/tcp/raw ip.
812 		 */
813 		if ((inet6sw[ip6_protox[nxt]].pr_flags & PR_LASTHDR) != 0 &&
814 		    ipsec6_in_reject(m, NULL)) {
815 			ipsec6stat.in_polvio++;
816 			goto bad;
817 		}
818 #endif
819 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
820 	}
821 	return;
822  bad:
823 	m_freem(m);
824 }
825 
826 /*
827  * set/grab in6_ifaddr correspond to IPv6 destination address.
828  * XXX backward compatibility wrapper
829  */
830 static struct ip6aux *
831 ip6_setdstifaddr(m, ia6)
832 	struct mbuf *m;
833 	struct in6_ifaddr *ia6;
834 {
835 	struct ip6aux *ip6a;
836 
837 	ip6a = ip6_addaux(m);
838 	if (ip6a)
839 		ip6a->ip6a_dstia6 = ia6;
840 	return ip6a;	/* NULL if failed to set */
841 }
842 
843 struct in6_ifaddr *
844 ip6_getdstifaddr(m)
845 	struct mbuf *m;
846 {
847 	struct ip6aux *ip6a;
848 
849 	ip6a = ip6_findaux(m);
850 	if (ip6a)
851 		return ip6a->ip6a_dstia6;
852 	else
853 		return NULL;
854 }
855 
856 /*
857  * Hop-by-Hop options header processing. If a valid jumbo payload option is
858  * included, the real payload length will be stored in plenp.
859  */
860 static int
861 ip6_hopopts_input(plenp, rtalertp, mp, offp)
862 	u_int32_t *plenp;
863 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
864 	struct mbuf **mp;
865 	int *offp;
866 {
867 	struct mbuf *m = *mp;
868 	int off = *offp, hbhlen;
869 	struct ip6_hbh *hbh;
870 	u_int8_t *opt;
871 
872 	/* validation of the length of the header */
873 #ifndef PULLDOWN_TEST
874 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
875 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
876 	hbhlen = (hbh->ip6h_len + 1) << 3;
877 
878 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
879 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
880 #else
881 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
882 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
883 	if (hbh == NULL) {
884 		ip6stat.ip6s_tooshort++;
885 		return -1;
886 	}
887 	hbhlen = (hbh->ip6h_len + 1) << 3;
888 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
889 		hbhlen);
890 	if (hbh == NULL) {
891 		ip6stat.ip6s_tooshort++;
892 		return -1;
893 	}
894 #endif
895 	off += hbhlen;
896 	hbhlen -= sizeof(struct ip6_hbh);
897 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
898 
899 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
900 				hbhlen, rtalertp, plenp) < 0)
901 		return (-1);
902 
903 	*offp = off;
904 	*mp = m;
905 	return (0);
906 }
907 
908 /*
909  * Search header for all Hop-by-hop options and process each option.
910  * This function is separate from ip6_hopopts_input() in order to
911  * handle a case where the sending node itself process its hop-by-hop
912  * options header. In such a case, the function is called from ip6_output().
913  *
914  * The function assumes that hbh header is located right after the IPv6 header
915  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
916  * opthead + hbhlen is located in continuous memory region.
917  */
918 int
919 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
920 	struct mbuf *m;
921 	u_int8_t *opthead;
922 	int hbhlen;
923 	u_int32_t *rtalertp;
924 	u_int32_t *plenp;
925 {
926 	struct ip6_hdr *ip6;
927 	int optlen = 0;
928 	u_int8_t *opt = opthead;
929 	u_int16_t rtalert_val;
930 	u_int32_t jumboplen;
931 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
932 
933 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
934 		switch (*opt) {
935 		case IP6OPT_PAD1:
936 			optlen = 1;
937 			break;
938 		case IP6OPT_PADN:
939 			if (hbhlen < IP6OPT_MINLEN) {
940 				ip6stat.ip6s_toosmall++;
941 				goto bad;
942 			}
943 			optlen = *(opt + 1) + 2;
944 			break;
945 		case IP6OPT_ROUTER_ALERT:
946 			/* XXX may need check for alignment */
947 			if (hbhlen < IP6OPT_RTALERT_LEN) {
948 				ip6stat.ip6s_toosmall++;
949 				goto bad;
950 			}
951 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
952 				/* XXX stat */
953 				icmp6_error(m, ICMP6_PARAM_PROB,
954 				    ICMP6_PARAMPROB_HEADER,
955 				    erroff + opt + 1 - opthead);
956 				return (-1);
957 			}
958 			optlen = IP6OPT_RTALERT_LEN;
959 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
960 			*rtalertp = ntohs(rtalert_val);
961 			break;
962 		case IP6OPT_JUMBO:
963 			/* XXX may need check for alignment */
964 			if (hbhlen < IP6OPT_JUMBO_LEN) {
965 				ip6stat.ip6s_toosmall++;
966 				goto bad;
967 			}
968 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
969 				/* XXX stat */
970 				icmp6_error(m, ICMP6_PARAM_PROB,
971 				    ICMP6_PARAMPROB_HEADER,
972 				    erroff + opt + 1 - opthead);
973 				return (-1);
974 			}
975 			optlen = IP6OPT_JUMBO_LEN;
976 
977 			/*
978 			 * IPv6 packets that have non 0 payload length
979 			 * must not contain a jumbo payload option.
980 			 */
981 			ip6 = mtod(m, struct ip6_hdr *);
982 			if (ip6->ip6_plen) {
983 				ip6stat.ip6s_badoptions++;
984 				icmp6_error(m, ICMP6_PARAM_PROB,
985 				    ICMP6_PARAMPROB_HEADER,
986 				    erroff + opt - opthead);
987 				return (-1);
988 			}
989 
990 			/*
991 			 * We may see jumbolen in unaligned location, so
992 			 * we'd need to perform bcopy().
993 			 */
994 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
995 			jumboplen = (u_int32_t)htonl(jumboplen);
996 
997 #if 1
998 			/*
999 			 * if there are multiple jumbo payload options,
1000 			 * *plenp will be non-zero and the packet will be
1001 			 * rejected.
1002 			 * the behavior may need some debate in ipngwg -
1003 			 * multiple options does not make sense, however,
1004 			 * there's no explicit mention in specification.
1005 			 */
1006 			if (*plenp != 0) {
1007 				ip6stat.ip6s_badoptions++;
1008 				icmp6_error(m, ICMP6_PARAM_PROB,
1009 				    ICMP6_PARAMPROB_HEADER,
1010 				    erroff + opt + 2 - opthead);
1011 				return (-1);
1012 			}
1013 #endif
1014 
1015 			/*
1016 			 * jumbo payload length must be larger than 65535.
1017 			 */
1018 			if (jumboplen <= IPV6_MAXPACKET) {
1019 				ip6stat.ip6s_badoptions++;
1020 				icmp6_error(m, ICMP6_PARAM_PROB,
1021 				    ICMP6_PARAMPROB_HEADER,
1022 				    erroff + opt + 2 - opthead);
1023 				return (-1);
1024 			}
1025 			*plenp = jumboplen;
1026 
1027 			break;
1028 		default:		/* unknown option */
1029 			if (hbhlen < IP6OPT_MINLEN) {
1030 				ip6stat.ip6s_toosmall++;
1031 				goto bad;
1032 			}
1033 			optlen = ip6_unknown_opt(opt, m,
1034 			    erroff + opt - opthead);
1035 			if (optlen == -1)
1036 				return (-1);
1037 			optlen += 2;
1038 			break;
1039 		}
1040 	}
1041 
1042 	return (0);
1043 
1044   bad:
1045 	m_freem(m);
1046 	return (-1);
1047 }
1048 
1049 /*
1050  * Unknown option processing.
1051  * The third argument `off' is the offset from the IPv6 header to the option,
1052  * which is necessary if the IPv6 header the and option header and IPv6 header
1053  * is not continuous in order to return an ICMPv6 error.
1054  */
1055 int
1056 ip6_unknown_opt(optp, m, off)
1057 	u_int8_t *optp;
1058 	struct mbuf *m;
1059 	int off;
1060 {
1061 	struct ip6_hdr *ip6;
1062 
1063 	switch (IP6OPT_TYPE(*optp)) {
1064 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1065 		return ((int)*(optp + 1));
1066 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1067 		m_freem(m);
1068 		return (-1);
1069 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1070 		ip6stat.ip6s_badoptions++;
1071 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1072 		return (-1);
1073 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1074 		ip6stat.ip6s_badoptions++;
1075 		ip6 = mtod(m, struct ip6_hdr *);
1076 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1077 		    (m->m_flags & (M_BCAST|M_MCAST)))
1078 			m_freem(m);
1079 		else
1080 			icmp6_error(m, ICMP6_PARAM_PROB,
1081 				    ICMP6_PARAMPROB_OPTION, off);
1082 		return (-1);
1083 	}
1084 
1085 	m_freem(m);		/* XXX: NOTREACHED */
1086 	return (-1);
1087 }
1088 
1089 /*
1090  * Create the "control" list for this pcb.
1091  * The function will not modify mbuf chain at all.
1092  *
1093  * with KAME mbuf chain restriction:
1094  * The routine will be called from upper layer handlers like tcp6_input().
1095  * Thus the routine assumes that the caller (tcp6_input) have already
1096  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1097  * very first mbuf on the mbuf chain.
1098  */
1099 void
1100 ip6_savecontrol(in6p, m, mp)
1101 	struct inpcb *in6p;
1102 	struct mbuf *m, **mp;
1103 {
1104 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1105 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1106 
1107 #ifdef SO_TIMESTAMP
1108 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1109 		struct timeval tv;
1110 
1111 		microtime(&tv);
1112 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1113 		    SCM_TIMESTAMP, SOL_SOCKET);
1114 		if (*mp)
1115 			mp = &(*mp)->m_next;
1116 	}
1117 #endif
1118 
1119 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1120 		return;
1121 
1122 	/* RFC 2292 sec. 5 */
1123 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1124 		struct in6_pktinfo pi6;
1125 
1126 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1127 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1128 		pi6.ipi6_ifindex =
1129 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1130 
1131 		*mp = sbcreatecontrol((caddr_t) &pi6,
1132 		    sizeof(struct in6_pktinfo),
1133 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1134 		if (*mp)
1135 			mp = &(*mp)->m_next;
1136 	}
1137 
1138 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1139 		int hlim = ip6->ip6_hlim & 0xff;
1140 
1141 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1142 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1143 		if (*mp)
1144 			mp = &(*mp)->m_next;
1145 	}
1146 
1147 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1148 		u_int32_t flowinfo;
1149 		int tclass;
1150 
1151 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1152 		flowinfo >>= 20;
1153 
1154 		tclass = flowinfo & 0xff;
1155 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1156 		    IPV6_TCLASS, IPPROTO_IPV6);
1157 		if (*mp)
1158 			mp = &(*mp)->m_next;
1159 	}
1160 
1161 	/*
1162 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1163 	 * privilege for the option (see ip6_ctloutput), but it might be too
1164 	 * strict, since there might be some hop-by-hop options which can be
1165 	 * returned to normal user.
1166 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1167 	 */
1168 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1169 		/*
1170 		 * Check if a hop-by-hop options header is contatined in the
1171 		 * received packet, and if so, store the options as ancillary
1172 		 * data. Note that a hop-by-hop options header must be
1173 		 * just after the IPv6 header, which is assured through the
1174 		 * IPv6 input processing.
1175 		 */
1176 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1177 			struct ip6_hbh *hbh;
1178 			int hbhlen = 0;
1179 #ifdef PULLDOWN_TEST
1180 			struct mbuf *ext;
1181 #endif
1182 
1183 #ifndef PULLDOWN_TEST
1184 			hbh = (struct ip6_hbh *)(ip6 + 1);
1185 			hbhlen = (hbh->ip6h_len + 1) << 3;
1186 #else
1187 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1188 			    ip6->ip6_nxt);
1189 			if (ext == NULL) {
1190 				ip6stat.ip6s_tooshort++;
1191 				return;
1192 			}
1193 			hbh = mtod(ext, struct ip6_hbh *);
1194 			hbhlen = (hbh->ip6h_len + 1) << 3;
1195 			if (hbhlen != ext->m_len) {
1196 				m_freem(ext);
1197 				ip6stat.ip6s_tooshort++;
1198 				return;
1199 			}
1200 #endif
1201 
1202 			/*
1203 			 * XXX: We copy the whole header even if a
1204 			 * jumbo payload option is included, the option which
1205 			 * is to be removed before returning according to
1206 			 * RFC2292.
1207 			 * Note: this constraint is removed in 2292bis.
1208 			 */
1209 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1210 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1211 			    IPPROTO_IPV6);
1212 			if (*mp)
1213 				mp = &(*mp)->m_next;
1214 #ifdef PULLDOWN_TEST
1215 			m_freem(ext);
1216 #endif
1217 		}
1218 	}
1219 
1220 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1221 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1222 
1223 		/*
1224 		 * Search for destination options headers or routing
1225 		 * header(s) through the header chain, and stores each
1226 		 * header as ancillary data.
1227 		 * Note that the order of the headers remains in
1228 		 * the chain of ancillary data.
1229 		 */
1230 		while (1) {	/* is explicit loop prevention necessary? */
1231 			struct ip6_ext *ip6e = NULL;
1232 			int elen;
1233 #ifdef PULLDOWN_TEST
1234 			struct mbuf *ext = NULL;
1235 #endif
1236 
1237 			/*
1238 			 * if it is not an extension header, don't try to
1239 			 * pull it from the chain.
1240 			 */
1241 			switch (nxt) {
1242 			case IPPROTO_DSTOPTS:
1243 			case IPPROTO_ROUTING:
1244 			case IPPROTO_HOPOPTS:
1245 			case IPPROTO_AH: /* is it possible? */
1246 				break;
1247 			default:
1248 				goto loopend;
1249 			}
1250 
1251 #ifndef PULLDOWN_TEST
1252 			if (off + sizeof(*ip6e) > m->m_len)
1253 				goto loopend;
1254 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1255 			if (nxt == IPPROTO_AH)
1256 				elen = (ip6e->ip6e_len + 2) << 2;
1257 			else
1258 				elen = (ip6e->ip6e_len + 1) << 3;
1259 			if (off + elen > m->m_len)
1260 				goto loopend;
1261 #else
1262 			ext = ip6_pullexthdr(m, off, nxt);
1263 			if (ext == NULL) {
1264 				ip6stat.ip6s_tooshort++;
1265 				return;
1266 			}
1267 			ip6e = mtod(ext, struct ip6_ext *);
1268 			if (nxt == IPPROTO_AH)
1269 				elen = (ip6e->ip6e_len + 2) << 2;
1270 			else
1271 				elen = (ip6e->ip6e_len + 1) << 3;
1272 			if (elen != ext->m_len) {
1273 				m_freem(ext);
1274 				ip6stat.ip6s_tooshort++;
1275 				return;
1276 			}
1277 #endif
1278 
1279 			switch (nxt) {
1280 			case IPPROTO_DSTOPTS:
1281 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1282 					break;
1283 
1284 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1285 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1286 				    IPPROTO_IPV6);
1287 				if (*mp)
1288 					mp = &(*mp)->m_next;
1289 				break;
1290 			case IPPROTO_ROUTING:
1291 				if (!in6p->in6p_flags & IN6P_RTHDR)
1292 					break;
1293 
1294 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1295 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1296 				    IPPROTO_IPV6);
1297 				if (*mp)
1298 					mp = &(*mp)->m_next;
1299 				break;
1300 			case IPPROTO_HOPOPTS:
1301 			case IPPROTO_AH: /* is it possible? */
1302 				break;
1303 
1304 			default:
1305 				/*
1306 			 	 * other cases have been filtered in the above.
1307 				 * none will visit this case.  here we supply
1308 				 * the code just in case (nxt overwritten or
1309 				 * other cases).
1310 				 */
1311 #ifdef PULLDOWN_TEST
1312 				m_freem(ext);
1313 #endif
1314 				goto loopend;
1315 
1316 			}
1317 
1318 			/* proceed with the next header. */
1319 			off += elen;
1320 			nxt = ip6e->ip6e_nxt;
1321 			ip6e = NULL;
1322 #ifdef PULLDOWN_TEST
1323 			m_freem(ext);
1324 			ext = NULL;
1325 #endif
1326 		}
1327 	  loopend:
1328 		;
1329 	}
1330 
1331 #undef IS2292
1332 }
1333 
1334 void
1335 ip6_notify_pmtu(in6p, dst, mtu)
1336 	struct inpcb *in6p;
1337 	struct sockaddr_in6 *dst;
1338 	u_int32_t *mtu;
1339 {
1340 	struct socket *so;
1341 	struct mbuf *m_mtu;
1342 	struct ip6_mtuinfo mtuctl;
1343 
1344 	so =  in6p->inp_socket;
1345 
1346 	if (mtu == NULL)
1347 		return;
1348 
1349 #ifdef DIAGNOSTIC
1350 	if (so == NULL)		/* I believe this is impossible */
1351 		panic("ip6_notify_pmtu: socket is NULL");
1352 #endif
1353 
1354 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1355 	mtuctl.ip6m_mtu = *mtu;
1356 	mtuctl.ip6m_addr = *dst;
1357 	in6_recoverscope(&mtuctl.ip6m_addr, &mtuctl.ip6m_addr.sin6_addr, NULL);
1358 
1359 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1360 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1361 		return;
1362 
1363 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1364 	    == 0) {
1365 		m_freem(m_mtu);
1366 		/* XXX: should count statistics */
1367 	} else
1368 		sorwakeup(so);
1369 
1370 	return;
1371 }
1372 
1373 #ifdef PULLDOWN_TEST
1374 /*
1375  * pull single extension header from mbuf chain.  returns single mbuf that
1376  * contains the result, or NULL on error.
1377  */
1378 static struct mbuf *
1379 ip6_pullexthdr(m, off, nxt)
1380 	struct mbuf *m;
1381 	size_t off;
1382 	int nxt;
1383 {
1384 	struct ip6_ext ip6e;
1385 	size_t elen;
1386 	struct mbuf *n;
1387 
1388 #ifdef DIAGNOSTIC
1389 	switch (nxt) {
1390 	case IPPROTO_DSTOPTS:
1391 	case IPPROTO_ROUTING:
1392 	case IPPROTO_HOPOPTS:
1393 	case IPPROTO_AH: /* is it possible? */
1394 		break;
1395 	default:
1396 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1397 	}
1398 #endif
1399 
1400 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1401 	if (nxt == IPPROTO_AH)
1402 		elen = (ip6e.ip6e_len + 2) << 2;
1403 	else
1404 		elen = (ip6e.ip6e_len + 1) << 3;
1405 
1406 	MGET(n, M_DONTWAIT, MT_DATA);
1407 	if (n && elen >= MLEN) {
1408 		MCLGET(n, M_DONTWAIT);
1409 		if ((n->m_flags & M_EXT) == 0) {
1410 			m_free(n);
1411 			n = NULL;
1412 		}
1413 	}
1414 	if (!n)
1415 		return NULL;
1416 
1417 	n->m_len = 0;
1418 	if (elen >= M_TRAILINGSPACE(n)) {
1419 		m_free(n);
1420 		return NULL;
1421 	}
1422 
1423 	m_copydata(m, off, elen, mtod(n, caddr_t));
1424 	n->m_len = elen;
1425 	return n;
1426 }
1427 #endif
1428 
1429 /*
1430  * Get pointer to the previous header followed by the header
1431  * currently processed.
1432  * XXX: This function supposes that
1433  *	M includes all headers,
1434  *	the next header field and the header length field of each header
1435  *	are valid, and
1436  *	the sum of each header length equals to OFF.
1437  * Because of these assumptions, this function must be called very
1438  * carefully. Moreover, it will not be used in the near future when
1439  * we develop `neater' mechanism to process extension headers.
1440  */
1441 char *
1442 ip6_get_prevhdr(m, off)
1443 	struct mbuf *m;
1444 	int off;
1445 {
1446 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1447 
1448 	if (off == sizeof(struct ip6_hdr))
1449 		return (&ip6->ip6_nxt);
1450 	else {
1451 		int len, nxt;
1452 		struct ip6_ext *ip6e = NULL;
1453 
1454 		nxt = ip6->ip6_nxt;
1455 		len = sizeof(struct ip6_hdr);
1456 		while (len < off) {
1457 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1458 
1459 			switch (nxt) {
1460 			case IPPROTO_FRAGMENT:
1461 				len += sizeof(struct ip6_frag);
1462 				break;
1463 			case IPPROTO_AH:
1464 				len += (ip6e->ip6e_len + 2) << 2;
1465 				break;
1466 			default:
1467 				len += (ip6e->ip6e_len + 1) << 3;
1468 				break;
1469 			}
1470 			nxt = ip6e->ip6e_nxt;
1471 		}
1472 		if (ip6e)
1473 			return (&ip6e->ip6e_nxt);
1474 		else
1475 			return NULL;
1476 	}
1477 }
1478 
1479 /*
1480  * get next header offset.  m will be retained.
1481  */
1482 int
1483 ip6_nexthdr(m, off, proto, nxtp)
1484 	struct mbuf *m;
1485 	int off;
1486 	int proto;
1487 	int *nxtp;
1488 {
1489 	struct ip6_hdr ip6;
1490 	struct ip6_ext ip6e;
1491 	struct ip6_frag fh;
1492 
1493 	/* just in case */
1494 	if (m == NULL)
1495 		panic("ip6_nexthdr: m == NULL");
1496 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1497 		return -1;
1498 
1499 	switch (proto) {
1500 	case IPPROTO_IPV6:
1501 		if (m->m_pkthdr.len < off + sizeof(ip6))
1502 			return -1;
1503 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1504 		if (nxtp)
1505 			*nxtp = ip6.ip6_nxt;
1506 		off += sizeof(ip6);
1507 		return off;
1508 
1509 	case IPPROTO_FRAGMENT:
1510 		/*
1511 		 * terminate parsing if it is not the first fragment,
1512 		 * it does not make sense to parse through it.
1513 		 */
1514 		if (m->m_pkthdr.len < off + sizeof(fh))
1515 			return -1;
1516 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1517 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1518 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1519 			return -1;
1520 		if (nxtp)
1521 			*nxtp = fh.ip6f_nxt;
1522 		off += sizeof(struct ip6_frag);
1523 		return off;
1524 
1525 	case IPPROTO_AH:
1526 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1527 			return -1;
1528 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1529 		if (nxtp)
1530 			*nxtp = ip6e.ip6e_nxt;
1531 		off += (ip6e.ip6e_len + 2) << 2;
1532 		return off;
1533 
1534 	case IPPROTO_HOPOPTS:
1535 	case IPPROTO_ROUTING:
1536 	case IPPROTO_DSTOPTS:
1537 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1538 			return -1;
1539 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1540 		if (nxtp)
1541 			*nxtp = ip6e.ip6e_nxt;
1542 		off += (ip6e.ip6e_len + 1) << 3;
1543 		return off;
1544 
1545 	case IPPROTO_NONE:
1546 	case IPPROTO_ESP:
1547 	case IPPROTO_IPCOMP:
1548 		/* give up */
1549 		return -1;
1550 
1551 	default:
1552 		return -1;
1553 	}
1554 
1555 	return -1;
1556 }
1557 
1558 /*
1559  * get offset for the last header in the chain.  m will be kept untainted.
1560  */
1561 int
1562 ip6_lasthdr(m, off, proto, nxtp)
1563 	struct mbuf *m;
1564 	int off;
1565 	int proto;
1566 	int *nxtp;
1567 {
1568 	int newoff;
1569 	int nxt;
1570 
1571 	if (!nxtp) {
1572 		nxt = -1;
1573 		nxtp = &nxt;
1574 	}
1575 	while (1) {
1576 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1577 		if (newoff < 0)
1578 			return off;
1579 		else if (newoff < off)
1580 			return -1;	/* invalid */
1581 		else if (newoff == off)
1582 			return newoff;
1583 
1584 		off = newoff;
1585 		proto = *nxtp;
1586 	}
1587 }
1588 
1589 struct ip6aux *
1590 ip6_addaux(m)
1591 	struct mbuf *m;
1592 {
1593 	struct m_tag *mtag;
1594 
1595 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1596 	if (!mtag) {
1597 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1598 		    M_NOWAIT);
1599 		if (mtag) {
1600 			m_tag_prepend(m, mtag);
1601 			bzero(mtag + 1, sizeof(struct ip6aux));
1602 		}
1603 	}
1604 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1605 }
1606 
1607 struct ip6aux *
1608 ip6_findaux(m)
1609 	struct mbuf *m;
1610 {
1611 	struct m_tag *mtag;
1612 
1613 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1614 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1615 }
1616 
1617 void
1618 ip6_delaux(m)
1619 	struct mbuf *m;
1620 {
1621 	struct m_tag *mtag;
1622 
1623 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1624 	if (mtag)
1625 		m_tag_delete(m, mtag);
1626 }
1627 
1628 /*
1629  * System control for IP6
1630  */
1631 
1632 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1633 	0,		0,		0,		0,
1634 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1635 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1636 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1637 	0,		0,		0,		0,
1638 	ENOPROTOOPT
1639 };
1640