xref: /freebsd/sys/netinet6/ip6_input.c (revision 6356dba0b403daa023dec24559ab1f8e602e4f14)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/proc.h>
75 #include <sys/domain.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/errno.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/vimage.h>
84 
85 #include <net/if.h>
86 #include <net/if_types.h>
87 #include <net/if_dl.h>
88 #include <net/route.h>
89 #include <net/netisr.h>
90 #include <net/pfil.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #ifdef INET
95 #include <netinet/ip.h>
96 #include <netinet/ip_icmp.h>
97 #endif /* INET */
98 #include <netinet/ip6.h>
99 #include <netinet6/in6_var.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/icmp6.h>
103 #include <netinet6/scope6_var.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/nd6.h>
106 
107 #ifdef IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netinet6/ip6_ipsec.h>
110 #include <netipsec/ipsec6.h>
111 #endif /* IPSEC */
112 
113 #include <netinet6/ip6protosw.h>
114 
115 extern struct domain inet6domain;
116 
117 u_char ip6_protox[IPPROTO_MAX];
118 static struct ifqueue ip6intrq;
119 static int ip6qmaxlen = IFQ_MAXLEN;
120 struct in6_ifaddr *in6_ifaddr;
121 
122 extern struct callout in6_tmpaddrtimer_ch;
123 
124 int ip6_forward_srcrt;			/* XXX */
125 int ip6_sourcecheck;			/* XXX */
126 int ip6_sourcecheck_interval;		/* XXX */
127 
128 int ip6_ours_check_algorithm;
129 
130 struct pfil_head inet6_pfil_hook;
131 
132 struct ip6stat ip6stat;
133 
134 static void ip6_init2(void *);
135 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
136 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
137 #ifdef PULLDOWN_TEST
138 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
139 #endif
140 
141 /*
142  * IP6 initialization: fill in IP6 protocol switch table.
143  * All protocols not implemented in kernel go to raw IP6 protocol handler.
144  */
145 void
146 ip6_init(void)
147 {
148 	struct ip6protosw *pr;
149 	int i;
150 
151 #ifdef DIAGNOSTIC
152 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
153 		panic("sizeof(protosw) != sizeof(ip6protosw)");
154 #endif
155 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
156 	if (pr == 0)
157 		panic("ip6_init");
158 
159 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
160 	for (i = 0; i < IPPROTO_MAX; i++)
161 		ip6_protox[i] = pr - inet6sw;
162 	/*
163 	 * Cycle through IP protocols and put them into the appropriate place
164 	 * in ip6_protox[].
165 	 */
166 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
167 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
168 		if (pr->pr_domain->dom_family == PF_INET6 &&
169 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
170 			/* Be careful to only index valid IP protocols. */
171 			if (pr->pr_protocol < IPPROTO_MAX)
172 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
173 		}
174 
175 	/* Initialize packet filter hooks. */
176 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
177 	inet6_pfil_hook.ph_af = AF_INET6;
178 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
179 		printf("%s: WARNING: unable to register pfil hook, "
180 			"error %d\n", __func__, i);
181 
182 	ip6intrq.ifq_maxlen = V_ip6qmaxlen;
183 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
184 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
185 	scope6_init();
186 	addrsel_policy_init();
187 	nd6_init();
188 	frag6_init();
189 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
190 }
191 
192 static void
193 ip6_init2(void *dummy)
194 {
195 
196 	/* nd6_timer_init */
197 	callout_init(&V_nd6_timer_ch, 0);
198 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, NULL);
199 
200 	/* timer for regeneranation of temporary addresses randomize ID */
201 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
202 	callout_reset(&V_in6_tmpaddrtimer_ch,
203 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
204 		       V_ip6_temp_regen_advance) * hz,
205 		      in6_tmpaddrtimer, NULL);
206 }
207 
208 /* cheat */
209 /* This must be after route_init(), which is now SI_ORDER_THIRD */
210 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
211 
212 extern struct	route_in6 ip6_forward_rt;
213 
214 void
215 ip6_input(struct mbuf *m)
216 {
217 	struct ip6_hdr *ip6;
218 	int off = sizeof(struct ip6_hdr), nest;
219 	u_int32_t plen;
220 	u_int32_t rtalert = ~0;
221 	int nxt, ours = 0;
222 	struct ifnet *deliverifp = NULL;
223 	struct in6_addr odst;
224 	int srcrt = 0;
225 
226 #ifdef IPSEC
227 	/*
228 	 * should the inner packet be considered authentic?
229 	 * see comment in ah4_input().
230 	 * NB: m cannot be NULL when passed to the input routine
231 	 */
232 
233 	m->m_flags &= ~M_AUTHIPHDR;
234 	m->m_flags &= ~M_AUTHIPDGM;
235 
236 #endif /* IPSEC */
237 
238 	/*
239 	 * make sure we don't have onion peering information into m_tag.
240 	 */
241 	ip6_delaux(m);
242 
243 	/*
244 	 * mbuf statistics
245 	 */
246 	if (m->m_flags & M_EXT) {
247 		if (m->m_next)
248 			V_ip6stat.ip6s_mext2m++;
249 		else
250 			V_ip6stat.ip6s_mext1++;
251 	} else {
252 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
253 		if (m->m_next) {
254 			if (m->m_flags & M_LOOP) {
255 				V_ip6stat.ip6s_m2m[V_loif[0].if_index]++; /* XXX */
256 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
257 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
258 			else
259 				V_ip6stat.ip6s_m2m[0]++;
260 		} else
261 			V_ip6stat.ip6s_m1++;
262 #undef M2MMAX
263 	}
264 
265 	/* drop the packet if IPv6 operation is disabled on the IF */
266 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
267 		m_freem(m);
268 		return;
269 	}
270 
271 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
272 	V_ip6stat.ip6s_total++;
273 
274 #ifndef PULLDOWN_TEST
275 	/*
276 	 * L2 bridge code and some other code can return mbuf chain
277 	 * that does not conform to KAME requirement.  too bad.
278 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
279 	 */
280 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
281 		struct mbuf *n;
282 
283 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
284 		if (n)
285 			M_MOVE_PKTHDR(n, m);
286 		if (n && n->m_pkthdr.len > MHLEN) {
287 			MCLGET(n, M_DONTWAIT);
288 			if ((n->m_flags & M_EXT) == 0) {
289 				m_freem(n);
290 				n = NULL;
291 			}
292 		}
293 		if (n == NULL) {
294 			m_freem(m);
295 			return;	/* ENOBUFS */
296 		}
297 
298 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
299 		n->m_len = n->m_pkthdr.len;
300 		m_freem(m);
301 		m = n;
302 	}
303 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
304 #endif
305 
306 	if (m->m_len < sizeof(struct ip6_hdr)) {
307 		struct ifnet *inifp;
308 		inifp = m->m_pkthdr.rcvif;
309 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
310 			V_ip6stat.ip6s_toosmall++;
311 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
312 			return;
313 		}
314 	}
315 
316 	ip6 = mtod(m, struct ip6_hdr *);
317 
318 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
319 		V_ip6stat.ip6s_badvers++;
320 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
321 		goto bad;
322 	}
323 
324 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
325 
326 	/*
327 	 * Check against address spoofing/corruption.
328 	 */
329 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
330 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
331 		/*
332 		 * XXX: "badscope" is not very suitable for a multicast source.
333 		 */
334 		V_ip6stat.ip6s_badscope++;
335 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
336 		goto bad;
337 	}
338 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
339 	    !(m->m_flags & M_LOOP)) {
340 		/*
341 		 * In this case, the packet should come from the loopback
342 		 * interface.  However, we cannot just check the if_flags,
343 		 * because ip6_mloopback() passes the "actual" interface
344 		 * as the outgoing/incoming interface.
345 		 */
346 		V_ip6stat.ip6s_badscope++;
347 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
348 		goto bad;
349 	}
350 
351 #ifdef ALTQ
352 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
353 		/* packet is dropped by traffic conditioner */
354 		return;
355 	}
356 #endif
357 	/*
358 	 * The following check is not documented in specs.  A malicious
359 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
360 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
361 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
362 	 *
363 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
364 	 * support IPv4-less kernel compilation, we cannot support SIIT
365 	 * environment at all.  So, it makes more sense for us to reject any
366 	 * malicious packets for non-SIIT environment, than try to do a
367 	 * partial support for SIIT environment.
368 	 */
369 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
370 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
371 		V_ip6stat.ip6s_badscope++;
372 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
373 		goto bad;
374 	}
375 #if 0
376 	/*
377 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
378 	 *
379 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
380 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
381 	 * is revised to forbid relaying case.
382 	 */
383 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
384 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
385 		V_ip6stat.ip6s_badscope++;
386 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
387 		goto bad;
388 	}
389 #endif
390 
391 	/*
392 	 * Run through list of hooks for input packets.
393 	 *
394 	 * NB: Beware of the destination address changing
395 	 *     (e.g. by NAT rewriting).  When this happens,
396 	 *     tell ip6_forward to do the right thing.
397 	 */
398 	odst = ip6->ip6_dst;
399 
400 	/* Jump over all PFIL processing if hooks are not active. */
401 	if (!PFIL_HOOKED(&inet6_pfil_hook))
402 		goto passin;
403 
404 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
405 		return;
406 	if (m == NULL)			/* consumed by filter */
407 		return;
408 	ip6 = mtod(m, struct ip6_hdr *);
409 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
410 
411 passin:
412 	/*
413 	 * Disambiguate address scope zones (if there is ambiguity).
414 	 * We first make sure that the original source or destination address
415 	 * is not in our internal form for scoped addresses.  Such addresses
416 	 * are not necessarily invalid spec-wise, but we cannot accept them due
417 	 * to the usage conflict.
418 	 * in6_setscope() then also checks and rejects the cases where src or
419 	 * dst are the loopback address and the receiving interface
420 	 * is not loopback.
421 	 */
422 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
423 		V_ip6stat.ip6s_badscope++; /* XXX */
424 		goto bad;
425 	}
426 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
427 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
428 		V_ip6stat.ip6s_badscope++;
429 		goto bad;
430 	}
431 
432 	/*
433 	 * Multicast check
434 	 */
435 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
436 		struct in6_multi *in6m = 0;
437 
438 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
439 		/*
440 		 * See if we belong to the destination multicast group on the
441 		 * arrival interface.
442 		 */
443 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
444 		if (in6m)
445 			ours = 1;
446 		else if (!ip6_mrouter) {
447 			V_ip6stat.ip6s_notmember++;
448 			V_ip6stat.ip6s_cantforward++;
449 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
450 			goto bad;
451 		}
452 		deliverifp = m->m_pkthdr.rcvif;
453 		goto hbhcheck;
454 	}
455 
456 	/*
457 	 *  Unicast check
458 	 */
459 	if (V_ip6_forward_rt.ro_rt != NULL &&
460 	    (V_ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
461 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
462 	    &((struct sockaddr_in6 *)(&V_ip6_forward_rt.ro_dst))->sin6_addr))
463 		V_ip6stat.ip6s_forward_cachehit++;
464 	else {
465 		struct sockaddr_in6 *dst6;
466 
467 		if (V_ip6_forward_rt.ro_rt) {
468 			/* route is down or destination is different */
469 			V_ip6stat.ip6s_forward_cachemiss++;
470 			RTFREE(V_ip6_forward_rt.ro_rt);
471 			V_ip6_forward_rt.ro_rt = 0;
472 		}
473 
474 		bzero(&V_ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
475 		dst6 = (struct sockaddr_in6 *)&V_ip6_forward_rt.ro_dst;
476 		dst6->sin6_len = sizeof(struct sockaddr_in6);
477 		dst6->sin6_family = AF_INET6;
478 		dst6->sin6_addr = ip6->ip6_dst;
479 
480 		rtalloc((struct route *)&V_ip6_forward_rt);
481 	}
482 
483 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
484 
485 	/*
486 	 * Accept the packet if the forwarding interface to the destination
487 	 * according to the routing table is the loopback interface,
488 	 * unless the associated route has a gateway.
489 	 * Note that this approach causes to accept a packet if there is a
490 	 * route to the loopback interface for the destination of the packet.
491 	 * But we think it's even useful in some situations, e.g. when using
492 	 * a special daemon which wants to intercept the packet.
493 	 *
494 	 * XXX: some OSes automatically make a cloned route for the destination
495 	 * of an outgoing packet.  If the outgoing interface of the packet
496 	 * is a loopback one, the kernel would consider the packet to be
497 	 * accepted, even if we have no such address assinged on the interface.
498 	 * We check the cloned flag of the route entry to reject such cases,
499 	 * assuming that route entries for our own addresses are not made by
500 	 * cloning (it should be true because in6_addloop explicitly installs
501 	 * the host route).  However, we might have to do an explicit check
502 	 * while it would be less efficient.  Or, should we rather install a
503 	 * reject route for such a case?
504 	 */
505 	if (V_ip6_forward_rt.ro_rt &&
506 	    (V_ip6_forward_rt.ro_rt->rt_flags &
507 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
508 #ifdef RTF_WASCLONED
509 	    !(V_ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
510 #endif
511 #ifdef RTF_CLONED
512 	    !(V_ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
513 #endif
514 #if 0
515 	    /*
516 	     * The check below is redundant since the comparison of
517 	     * the destination and the key of the rtentry has
518 	     * already done through looking up the routing table.
519 	     */
520 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
521 	    &rt6_key(V_ip6_forward_rt.ro_rt)->sin6_addr)
522 #endif
523 	    V_ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
524 		struct in6_ifaddr *ia6 =
525 			(struct in6_ifaddr *)V_ip6_forward_rt.ro_rt->rt_ifa;
526 
527 		/*
528 		 * record address information into m_tag.
529 		 */
530 		(void)ip6_setdstifaddr(m, ia6);
531 
532 		/*
533 		 * packets to a tentative, duplicated, or somehow invalid
534 		 * address must not be accepted.
535 		 */
536 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
537 			/* this address is ready */
538 			ours = 1;
539 			deliverifp = ia6->ia_ifp;	/* correct? */
540 			/* Count the packet in the ip address stats */
541 			ia6->ia_ifa.if_ipackets++;
542 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
543 			goto hbhcheck;
544 		} else {
545 			char ip6bufs[INET6_ADDRSTRLEN];
546 			char ip6bufd[INET6_ADDRSTRLEN];
547 			/* address is not ready, so discard the packet. */
548 			nd6log((LOG_INFO,
549 			    "ip6_input: packet to an unready address %s->%s\n",
550 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
551 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
552 
553 			goto bad;
554 		}
555 	}
556 
557 	/*
558 	 * FAITH (Firewall Aided Internet Translator)
559 	 */
560 	if (V_ip6_keepfaith) {
561 		if (V_ip6_forward_rt.ro_rt && V_ip6_forward_rt.ro_rt->rt_ifp
562 		 && V_ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
563 			/* XXX do we need more sanity checks? */
564 			ours = 1;
565 			deliverifp = V_ip6_forward_rt.ro_rt->rt_ifp; /* faith */
566 			goto hbhcheck;
567 		}
568 	}
569 
570 	/*
571 	 * Now there is no reason to process the packet if it's not our own
572 	 * and we're not a router.
573 	 */
574 	if (!V_ip6_forwarding) {
575 		V_ip6stat.ip6s_cantforward++;
576 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
577 		goto bad;
578 	}
579 
580   hbhcheck:
581 	/*
582 	 * record address information into m_tag, if we don't have one yet.
583 	 * note that we are unable to record it, if the address is not listed
584 	 * as our interface address (e.g. multicast addresses, addresses
585 	 * within FAITH prefixes and such).
586 	 */
587 	if (deliverifp && !ip6_getdstifaddr(m)) {
588 		struct in6_ifaddr *ia6;
589 
590 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
591 		if (ia6) {
592 			if (!ip6_setdstifaddr(m, ia6)) {
593 				/*
594 				 * XXX maybe we should drop the packet here,
595 				 * as we could not provide enough information
596 				 * to the upper layers.
597 				 */
598 			}
599 		}
600 	}
601 
602 	/*
603 	 * Process Hop-by-Hop options header if it's contained.
604 	 * m may be modified in ip6_hopopts_input().
605 	 * If a JumboPayload option is included, plen will also be modified.
606 	 */
607 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
608 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
609 		struct ip6_hbh *hbh;
610 
611 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
612 #if 0	/*touches NULL pointer*/
613 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
614 #endif
615 			return;	/* m have already been freed */
616 		}
617 
618 		/* adjust pointer */
619 		ip6 = mtod(m, struct ip6_hdr *);
620 
621 		/*
622 		 * if the payload length field is 0 and the next header field
623 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
624 		 * option MUST be included.
625 		 */
626 		if (ip6->ip6_plen == 0 && plen == 0) {
627 			/*
628 			 * Note that if a valid jumbo payload option is
629 			 * contained, ip6_hopopts_input() must set a valid
630 			 * (non-zero) payload length to the variable plen.
631 			 */
632 			V_ip6stat.ip6s_badoptions++;
633 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
634 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
635 			icmp6_error(m, ICMP6_PARAM_PROB,
636 				    ICMP6_PARAMPROB_HEADER,
637 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
638 			return;
639 		}
640 #ifndef PULLDOWN_TEST
641 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
642 		hbh = (struct ip6_hbh *)(ip6 + 1);
643 #else
644 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
645 			sizeof(struct ip6_hbh));
646 		if (hbh == NULL) {
647 			V_ip6stat.ip6s_tooshort++;
648 			return;
649 		}
650 #endif
651 		nxt = hbh->ip6h_nxt;
652 
653 		/*
654 		 * If we are acting as a router and the packet contains a
655 		 * router alert option, see if we know the option value.
656 		 * Currently, we only support the option value for MLD, in which
657 		 * case we should pass the packet to the multicast routing
658 		 * daemon.
659 		 */
660 		if (rtalert != ~0 && V_ip6_forwarding) {
661 			switch (rtalert) {
662 			case IP6OPT_RTALERT_MLD:
663 				ours = 1;
664 				break;
665 			default:
666 				/*
667 				 * RFC2711 requires unrecognized values must be
668 				 * silently ignored.
669 				 */
670 				break;
671 			}
672 		}
673 	} else
674 		nxt = ip6->ip6_nxt;
675 
676 	/*
677 	 * Check that the amount of data in the buffers
678 	 * is as at least much as the IPv6 header would have us expect.
679 	 * Trim mbufs if longer than we expect.
680 	 * Drop packet if shorter than we expect.
681 	 */
682 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
683 		V_ip6stat.ip6s_tooshort++;
684 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
685 		goto bad;
686 	}
687 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
688 		if (m->m_len == m->m_pkthdr.len) {
689 			m->m_len = sizeof(struct ip6_hdr) + plen;
690 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
691 		} else
692 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
693 	}
694 
695 	/*
696 	 * Forward if desirable.
697 	 */
698 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
699 		/*
700 		 * If we are acting as a multicast router, all
701 		 * incoming multicast packets are passed to the
702 		 * kernel-level multicast forwarding function.
703 		 * The packet is returned (relatively) intact; if
704 		 * ip6_mforward() returns a non-zero value, the packet
705 		 * must be discarded, else it may be accepted below.
706 		 */
707 		if (ip6_mrouter && ip6_mforward &&
708 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
709 			V_ip6stat.ip6s_cantforward++;
710 			m_freem(m);
711 			return;
712 		}
713 		if (!ours) {
714 			m_freem(m);
715 			return;
716 		}
717 	} else if (!ours) {
718 		ip6_forward(m, srcrt);
719 		return;
720 	}
721 
722 	ip6 = mtod(m, struct ip6_hdr *);
723 
724 	/*
725 	 * Malicious party may be able to use IPv4 mapped addr to confuse
726 	 * tcp/udp stack and bypass security checks (act as if it was from
727 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
728 	 *
729 	 * For SIIT end node behavior, you may want to disable the check.
730 	 * However, you will  become vulnerable to attacks using IPv4 mapped
731 	 * source.
732 	 */
733 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
734 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
735 		V_ip6stat.ip6s_badscope++;
736 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
737 		goto bad;
738 	}
739 
740 	/*
741 	 * Tell launch routine the next header
742 	 */
743 	V_ip6stat.ip6s_delivered++;
744 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
745 	nest = 0;
746 
747 	while (nxt != IPPROTO_DONE) {
748 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
749 			V_ip6stat.ip6s_toomanyhdr++;
750 			goto bad;
751 		}
752 
753 		/*
754 		 * protection against faulty packet - there should be
755 		 * more sanity checks in header chain processing.
756 		 */
757 		if (m->m_pkthdr.len < off) {
758 			V_ip6stat.ip6s_tooshort++;
759 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
760 			goto bad;
761 		}
762 
763 #ifdef IPSEC
764 		/*
765 		 * enforce IPsec policy checking if we are seeing last header.
766 		 * note that we do not visit this with protocols with pcb layer
767 		 * code - like udp/tcp/raw ip.
768 		 */
769 		if (ip6_ipsec_input(m, nxt))
770 			goto bad;
771 #endif /* IPSEC */
772 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
773 	}
774 	return;
775  bad:
776 	m_freem(m);
777 }
778 
779 /*
780  * set/grab in6_ifaddr correspond to IPv6 destination address.
781  * XXX backward compatibility wrapper
782  */
783 static struct ip6aux *
784 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
785 {
786 	struct ip6aux *ip6a;
787 
788 	ip6a = ip6_addaux(m);
789 	if (ip6a)
790 		ip6a->ip6a_dstia6 = ia6;
791 	return ip6a;	/* NULL if failed to set */
792 }
793 
794 struct in6_ifaddr *
795 ip6_getdstifaddr(struct mbuf *m)
796 {
797 	struct ip6aux *ip6a;
798 
799 	ip6a = ip6_findaux(m);
800 	if (ip6a)
801 		return ip6a->ip6a_dstia6;
802 	else
803 		return NULL;
804 }
805 
806 /*
807  * Hop-by-Hop options header processing. If a valid jumbo payload option is
808  * included, the real payload length will be stored in plenp.
809  *
810  * rtalertp - XXX: should be stored more smart way
811  */
812 static int
813 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
814     struct mbuf **mp, int *offp)
815 {
816 	struct mbuf *m = *mp;
817 	int off = *offp, hbhlen;
818 	struct ip6_hbh *hbh;
819 	u_int8_t *opt;
820 
821 	/* validation of the length of the header */
822 #ifndef PULLDOWN_TEST
823 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
824 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
825 	hbhlen = (hbh->ip6h_len + 1) << 3;
826 
827 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
828 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
829 #else
830 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
831 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
832 	if (hbh == NULL) {
833 		V_ip6stat.ip6s_tooshort++;
834 		return -1;
835 	}
836 	hbhlen = (hbh->ip6h_len + 1) << 3;
837 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
838 		hbhlen);
839 	if (hbh == NULL) {
840 		V_ip6stat.ip6s_tooshort++;
841 		return -1;
842 	}
843 #endif
844 	off += hbhlen;
845 	hbhlen -= sizeof(struct ip6_hbh);
846 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
847 
848 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
849 				hbhlen, rtalertp, plenp) < 0)
850 		return (-1);
851 
852 	*offp = off;
853 	*mp = m;
854 	return (0);
855 }
856 
857 /*
858  * Search header for all Hop-by-hop options and process each option.
859  * This function is separate from ip6_hopopts_input() in order to
860  * handle a case where the sending node itself process its hop-by-hop
861  * options header. In such a case, the function is called from ip6_output().
862  *
863  * The function assumes that hbh header is located right after the IPv6 header
864  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
865  * opthead + hbhlen is located in continuous memory region.
866  */
867 int
868 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
869     u_int32_t *rtalertp, u_int32_t *plenp)
870 {
871 	struct ip6_hdr *ip6;
872 	int optlen = 0;
873 	u_int8_t *opt = opthead;
874 	u_int16_t rtalert_val;
875 	u_int32_t jumboplen;
876 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
877 
878 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
879 		switch (*opt) {
880 		case IP6OPT_PAD1:
881 			optlen = 1;
882 			break;
883 		case IP6OPT_PADN:
884 			if (hbhlen < IP6OPT_MINLEN) {
885 				V_ip6stat.ip6s_toosmall++;
886 				goto bad;
887 			}
888 			optlen = *(opt + 1) + 2;
889 			break;
890 		case IP6OPT_ROUTER_ALERT:
891 			/* XXX may need check for alignment */
892 			if (hbhlen < IP6OPT_RTALERT_LEN) {
893 				V_ip6stat.ip6s_toosmall++;
894 				goto bad;
895 			}
896 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
897 				/* XXX stat */
898 				icmp6_error(m, ICMP6_PARAM_PROB,
899 				    ICMP6_PARAMPROB_HEADER,
900 				    erroff + opt + 1 - opthead);
901 				return (-1);
902 			}
903 			optlen = IP6OPT_RTALERT_LEN;
904 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
905 			*rtalertp = ntohs(rtalert_val);
906 			break;
907 		case IP6OPT_JUMBO:
908 			/* XXX may need check for alignment */
909 			if (hbhlen < IP6OPT_JUMBO_LEN) {
910 				V_ip6stat.ip6s_toosmall++;
911 				goto bad;
912 			}
913 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
914 				/* XXX stat */
915 				icmp6_error(m, ICMP6_PARAM_PROB,
916 				    ICMP6_PARAMPROB_HEADER,
917 				    erroff + opt + 1 - opthead);
918 				return (-1);
919 			}
920 			optlen = IP6OPT_JUMBO_LEN;
921 
922 			/*
923 			 * IPv6 packets that have non 0 payload length
924 			 * must not contain a jumbo payload option.
925 			 */
926 			ip6 = mtod(m, struct ip6_hdr *);
927 			if (ip6->ip6_plen) {
928 				V_ip6stat.ip6s_badoptions++;
929 				icmp6_error(m, ICMP6_PARAM_PROB,
930 				    ICMP6_PARAMPROB_HEADER,
931 				    erroff + opt - opthead);
932 				return (-1);
933 			}
934 
935 			/*
936 			 * We may see jumbolen in unaligned location, so
937 			 * we'd need to perform bcopy().
938 			 */
939 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
940 			jumboplen = (u_int32_t)htonl(jumboplen);
941 
942 #if 1
943 			/*
944 			 * if there are multiple jumbo payload options,
945 			 * *plenp will be non-zero and the packet will be
946 			 * rejected.
947 			 * the behavior may need some debate in ipngwg -
948 			 * multiple options does not make sense, however,
949 			 * there's no explicit mention in specification.
950 			 */
951 			if (*plenp != 0) {
952 				V_ip6stat.ip6s_badoptions++;
953 				icmp6_error(m, ICMP6_PARAM_PROB,
954 				    ICMP6_PARAMPROB_HEADER,
955 				    erroff + opt + 2 - opthead);
956 				return (-1);
957 			}
958 #endif
959 
960 			/*
961 			 * jumbo payload length must be larger than 65535.
962 			 */
963 			if (jumboplen <= IPV6_MAXPACKET) {
964 				V_ip6stat.ip6s_badoptions++;
965 				icmp6_error(m, ICMP6_PARAM_PROB,
966 				    ICMP6_PARAMPROB_HEADER,
967 				    erroff + opt + 2 - opthead);
968 				return (-1);
969 			}
970 			*plenp = jumboplen;
971 
972 			break;
973 		default:		/* unknown option */
974 			if (hbhlen < IP6OPT_MINLEN) {
975 				V_ip6stat.ip6s_toosmall++;
976 				goto bad;
977 			}
978 			optlen = ip6_unknown_opt(opt, m,
979 			    erroff + opt - opthead);
980 			if (optlen == -1)
981 				return (-1);
982 			optlen += 2;
983 			break;
984 		}
985 	}
986 
987 	return (0);
988 
989   bad:
990 	m_freem(m);
991 	return (-1);
992 }
993 
994 /*
995  * Unknown option processing.
996  * The third argument `off' is the offset from the IPv6 header to the option,
997  * which is necessary if the IPv6 header the and option header and IPv6 header
998  * is not continuous in order to return an ICMPv6 error.
999  */
1000 int
1001 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1002 {
1003 	struct ip6_hdr *ip6;
1004 
1005 	switch (IP6OPT_TYPE(*optp)) {
1006 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1007 		return ((int)*(optp + 1));
1008 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1009 		m_freem(m);
1010 		return (-1);
1011 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1012 		V_ip6stat.ip6s_badoptions++;
1013 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1014 		return (-1);
1015 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1016 		V_ip6stat.ip6s_badoptions++;
1017 		ip6 = mtod(m, struct ip6_hdr *);
1018 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1019 		    (m->m_flags & (M_BCAST|M_MCAST)))
1020 			m_freem(m);
1021 		else
1022 			icmp6_error(m, ICMP6_PARAM_PROB,
1023 				    ICMP6_PARAMPROB_OPTION, off);
1024 		return (-1);
1025 	}
1026 
1027 	m_freem(m);		/* XXX: NOTREACHED */
1028 	return (-1);
1029 }
1030 
1031 /*
1032  * Create the "control" list for this pcb.
1033  * These functions will not modify mbuf chain at all.
1034  *
1035  * With KAME mbuf chain restriction:
1036  * The routine will be called from upper layer handlers like tcp6_input().
1037  * Thus the routine assumes that the caller (tcp6_input) have already
1038  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1039  * very first mbuf on the mbuf chain.
1040  *
1041  * ip6_savecontrol_v4 will handle those options that are possible to be
1042  * set on a v4-mapped socket.
1043  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1044  * options and handle the v6-only ones itself.
1045  */
1046 struct mbuf **
1047 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1048     int *v4only)
1049 {
1050 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1051 
1052 #ifdef SO_TIMESTAMP
1053 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1054 		struct timeval tv;
1055 
1056 		microtime(&tv);
1057 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1058 		    SCM_TIMESTAMP, SOL_SOCKET);
1059 		if (*mp)
1060 			mp = &(*mp)->m_next;
1061 	}
1062 #endif
1063 
1064 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1065 		if (v4only != NULL)
1066 			*v4only = 1;
1067 		return (mp);
1068 	}
1069 
1070 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1071 	/* RFC 2292 sec. 5 */
1072 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1073 		struct in6_pktinfo pi6;
1074 
1075 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1076 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1077 		pi6.ipi6_ifindex =
1078 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1079 
1080 		*mp = sbcreatecontrol((caddr_t) &pi6,
1081 		    sizeof(struct in6_pktinfo),
1082 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1083 		if (*mp)
1084 			mp = &(*mp)->m_next;
1085 	}
1086 
1087 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1088 		int hlim = ip6->ip6_hlim & 0xff;
1089 
1090 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1091 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1092 		    IPPROTO_IPV6);
1093 		if (*mp)
1094 			mp = &(*mp)->m_next;
1095 	}
1096 
1097 	if (v4only != NULL)
1098 		*v4only = 0;
1099 	return (mp);
1100 }
1101 
1102 void
1103 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1104 {
1105 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1106 	int v4only = 0;
1107 
1108 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1109 	if (v4only)
1110 		return;
1111 
1112 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1113 		u_int32_t flowinfo;
1114 		int tclass;
1115 
1116 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1117 		flowinfo >>= 20;
1118 
1119 		tclass = flowinfo & 0xff;
1120 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1121 		    IPV6_TCLASS, IPPROTO_IPV6);
1122 		if (*mp)
1123 			mp = &(*mp)->m_next;
1124 	}
1125 
1126 	/*
1127 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1128 	 * privilege for the option (see ip6_ctloutput), but it might be too
1129 	 * strict, since there might be some hop-by-hop options which can be
1130 	 * returned to normal user.
1131 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1132 	 */
1133 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1134 		/*
1135 		 * Check if a hop-by-hop options header is contatined in the
1136 		 * received packet, and if so, store the options as ancillary
1137 		 * data. Note that a hop-by-hop options header must be
1138 		 * just after the IPv6 header, which is assured through the
1139 		 * IPv6 input processing.
1140 		 */
1141 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1142 			struct ip6_hbh *hbh;
1143 			int hbhlen = 0;
1144 #ifdef PULLDOWN_TEST
1145 			struct mbuf *ext;
1146 #endif
1147 
1148 #ifndef PULLDOWN_TEST
1149 			hbh = (struct ip6_hbh *)(ip6 + 1);
1150 			hbhlen = (hbh->ip6h_len + 1) << 3;
1151 #else
1152 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1153 			    ip6->ip6_nxt);
1154 			if (ext == NULL) {
1155 				V_ip6stat.ip6s_tooshort++;
1156 				return;
1157 			}
1158 			hbh = mtod(ext, struct ip6_hbh *);
1159 			hbhlen = (hbh->ip6h_len + 1) << 3;
1160 			if (hbhlen != ext->m_len) {
1161 				m_freem(ext);
1162 				V_ip6stat.ip6s_tooshort++;
1163 				return;
1164 			}
1165 #endif
1166 
1167 			/*
1168 			 * XXX: We copy the whole header even if a
1169 			 * jumbo payload option is included, the option which
1170 			 * is to be removed before returning according to
1171 			 * RFC2292.
1172 			 * Note: this constraint is removed in RFC3542
1173 			 */
1174 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1175 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1176 			    IPPROTO_IPV6);
1177 			if (*mp)
1178 				mp = &(*mp)->m_next;
1179 #ifdef PULLDOWN_TEST
1180 			m_freem(ext);
1181 #endif
1182 		}
1183 	}
1184 
1185 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1186 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1187 
1188 		/*
1189 		 * Search for destination options headers or routing
1190 		 * header(s) through the header chain, and stores each
1191 		 * header as ancillary data.
1192 		 * Note that the order of the headers remains in
1193 		 * the chain of ancillary data.
1194 		 */
1195 		while (1) {	/* is explicit loop prevention necessary? */
1196 			struct ip6_ext *ip6e = NULL;
1197 			int elen;
1198 #ifdef PULLDOWN_TEST
1199 			struct mbuf *ext = NULL;
1200 #endif
1201 
1202 			/*
1203 			 * if it is not an extension header, don't try to
1204 			 * pull it from the chain.
1205 			 */
1206 			switch (nxt) {
1207 			case IPPROTO_DSTOPTS:
1208 			case IPPROTO_ROUTING:
1209 			case IPPROTO_HOPOPTS:
1210 			case IPPROTO_AH: /* is it possible? */
1211 				break;
1212 			default:
1213 				goto loopend;
1214 			}
1215 
1216 #ifndef PULLDOWN_TEST
1217 			if (off + sizeof(*ip6e) > m->m_len)
1218 				goto loopend;
1219 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1220 			if (nxt == IPPROTO_AH)
1221 				elen = (ip6e->ip6e_len + 2) << 2;
1222 			else
1223 				elen = (ip6e->ip6e_len + 1) << 3;
1224 			if (off + elen > m->m_len)
1225 				goto loopend;
1226 #else
1227 			ext = ip6_pullexthdr(m, off, nxt);
1228 			if (ext == NULL) {
1229 				V_ip6stat.ip6s_tooshort++;
1230 				return;
1231 			}
1232 			ip6e = mtod(ext, struct ip6_ext *);
1233 			if (nxt == IPPROTO_AH)
1234 				elen = (ip6e->ip6e_len + 2) << 2;
1235 			else
1236 				elen = (ip6e->ip6e_len + 1) << 3;
1237 			if (elen != ext->m_len) {
1238 				m_freem(ext);
1239 				V_ip6stat.ip6s_tooshort++;
1240 				return;
1241 			}
1242 #endif
1243 
1244 			switch (nxt) {
1245 			case IPPROTO_DSTOPTS:
1246 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1247 					break;
1248 
1249 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1250 				    IS2292(in6p,
1251 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1252 				    IPPROTO_IPV6);
1253 				if (*mp)
1254 					mp = &(*mp)->m_next;
1255 				break;
1256 			case IPPROTO_ROUTING:
1257 				if (!in6p->in6p_flags & IN6P_RTHDR)
1258 					break;
1259 
1260 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1261 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1262 				    IPPROTO_IPV6);
1263 				if (*mp)
1264 					mp = &(*mp)->m_next;
1265 				break;
1266 			case IPPROTO_HOPOPTS:
1267 			case IPPROTO_AH: /* is it possible? */
1268 				break;
1269 
1270 			default:
1271 				/*
1272 				 * other cases have been filtered in the above.
1273 				 * none will visit this case.  here we supply
1274 				 * the code just in case (nxt overwritten or
1275 				 * other cases).
1276 				 */
1277 #ifdef PULLDOWN_TEST
1278 				m_freem(ext);
1279 #endif
1280 				goto loopend;
1281 
1282 			}
1283 
1284 			/* proceed with the next header. */
1285 			off += elen;
1286 			nxt = ip6e->ip6e_nxt;
1287 			ip6e = NULL;
1288 #ifdef PULLDOWN_TEST
1289 			m_freem(ext);
1290 			ext = NULL;
1291 #endif
1292 		}
1293 	  loopend:
1294 		;
1295 	}
1296 }
1297 #undef IS2292
1298 
1299 void
1300 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1301 {
1302 	struct socket *so;
1303 	struct mbuf *m_mtu;
1304 	struct ip6_mtuinfo mtuctl;
1305 
1306 	so =  in6p->inp_socket;
1307 
1308 	if (mtu == NULL)
1309 		return;
1310 
1311 #ifdef DIAGNOSTIC
1312 	if (so == NULL)		/* I believe this is impossible */
1313 		panic("ip6_notify_pmtu: socket is NULL");
1314 #endif
1315 
1316 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1317 	mtuctl.ip6m_mtu = *mtu;
1318 	mtuctl.ip6m_addr = *dst;
1319 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1320 		return;
1321 
1322 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1323 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1324 		return;
1325 
1326 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1327 	    == 0) {
1328 		m_freem(m_mtu);
1329 		/* XXX: should count statistics */
1330 	} else
1331 		sorwakeup(so);
1332 
1333 	return;
1334 }
1335 
1336 #ifdef PULLDOWN_TEST
1337 /*
1338  * pull single extension header from mbuf chain.  returns single mbuf that
1339  * contains the result, or NULL on error.
1340  */
1341 static struct mbuf *
1342 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1343 {
1344 	struct ip6_ext ip6e;
1345 	size_t elen;
1346 	struct mbuf *n;
1347 
1348 #ifdef DIAGNOSTIC
1349 	switch (nxt) {
1350 	case IPPROTO_DSTOPTS:
1351 	case IPPROTO_ROUTING:
1352 	case IPPROTO_HOPOPTS:
1353 	case IPPROTO_AH: /* is it possible? */
1354 		break;
1355 	default:
1356 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1357 	}
1358 #endif
1359 
1360 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1361 	if (nxt == IPPROTO_AH)
1362 		elen = (ip6e.ip6e_len + 2) << 2;
1363 	else
1364 		elen = (ip6e.ip6e_len + 1) << 3;
1365 
1366 	MGET(n, M_DONTWAIT, MT_DATA);
1367 	if (n && elen >= MLEN) {
1368 		MCLGET(n, M_DONTWAIT);
1369 		if ((n->m_flags & M_EXT) == 0) {
1370 			m_free(n);
1371 			n = NULL;
1372 		}
1373 	}
1374 	if (!n)
1375 		return NULL;
1376 
1377 	n->m_len = 0;
1378 	if (elen >= M_TRAILINGSPACE(n)) {
1379 		m_free(n);
1380 		return NULL;
1381 	}
1382 
1383 	m_copydata(m, off, elen, mtod(n, caddr_t));
1384 	n->m_len = elen;
1385 	return n;
1386 }
1387 #endif
1388 
1389 /*
1390  * Get pointer to the previous header followed by the header
1391  * currently processed.
1392  * XXX: This function supposes that
1393  *	M includes all headers,
1394  *	the next header field and the header length field of each header
1395  *	are valid, and
1396  *	the sum of each header length equals to OFF.
1397  * Because of these assumptions, this function must be called very
1398  * carefully. Moreover, it will not be used in the near future when
1399  * we develop `neater' mechanism to process extension headers.
1400  */
1401 char *
1402 ip6_get_prevhdr(struct mbuf *m, int off)
1403 {
1404 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1405 
1406 	if (off == sizeof(struct ip6_hdr))
1407 		return (&ip6->ip6_nxt);
1408 	else {
1409 		int len, nxt;
1410 		struct ip6_ext *ip6e = NULL;
1411 
1412 		nxt = ip6->ip6_nxt;
1413 		len = sizeof(struct ip6_hdr);
1414 		while (len < off) {
1415 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1416 
1417 			switch (nxt) {
1418 			case IPPROTO_FRAGMENT:
1419 				len += sizeof(struct ip6_frag);
1420 				break;
1421 			case IPPROTO_AH:
1422 				len += (ip6e->ip6e_len + 2) << 2;
1423 				break;
1424 			default:
1425 				len += (ip6e->ip6e_len + 1) << 3;
1426 				break;
1427 			}
1428 			nxt = ip6e->ip6e_nxt;
1429 		}
1430 		if (ip6e)
1431 			return (&ip6e->ip6e_nxt);
1432 		else
1433 			return NULL;
1434 	}
1435 }
1436 
1437 /*
1438  * get next header offset.  m will be retained.
1439  */
1440 int
1441 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1442 {
1443 	struct ip6_hdr ip6;
1444 	struct ip6_ext ip6e;
1445 	struct ip6_frag fh;
1446 
1447 	/* just in case */
1448 	if (m == NULL)
1449 		panic("ip6_nexthdr: m == NULL");
1450 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1451 		return -1;
1452 
1453 	switch (proto) {
1454 	case IPPROTO_IPV6:
1455 		if (m->m_pkthdr.len < off + sizeof(ip6))
1456 			return -1;
1457 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1458 		if (nxtp)
1459 			*nxtp = ip6.ip6_nxt;
1460 		off += sizeof(ip6);
1461 		return off;
1462 
1463 	case IPPROTO_FRAGMENT:
1464 		/*
1465 		 * terminate parsing if it is not the first fragment,
1466 		 * it does not make sense to parse through it.
1467 		 */
1468 		if (m->m_pkthdr.len < off + sizeof(fh))
1469 			return -1;
1470 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1471 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1472 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1473 			return -1;
1474 		if (nxtp)
1475 			*nxtp = fh.ip6f_nxt;
1476 		off += sizeof(struct ip6_frag);
1477 		return off;
1478 
1479 	case IPPROTO_AH:
1480 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1481 			return -1;
1482 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1483 		if (nxtp)
1484 			*nxtp = ip6e.ip6e_nxt;
1485 		off += (ip6e.ip6e_len + 2) << 2;
1486 		return off;
1487 
1488 	case IPPROTO_HOPOPTS:
1489 	case IPPROTO_ROUTING:
1490 	case IPPROTO_DSTOPTS:
1491 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1492 			return -1;
1493 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1494 		if (nxtp)
1495 			*nxtp = ip6e.ip6e_nxt;
1496 		off += (ip6e.ip6e_len + 1) << 3;
1497 		return off;
1498 
1499 	case IPPROTO_NONE:
1500 	case IPPROTO_ESP:
1501 	case IPPROTO_IPCOMP:
1502 		/* give up */
1503 		return -1;
1504 
1505 	default:
1506 		return -1;
1507 	}
1508 
1509 	return -1;
1510 }
1511 
1512 /*
1513  * get offset for the last header in the chain.  m will be kept untainted.
1514  */
1515 int
1516 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1517 {
1518 	int newoff;
1519 	int nxt;
1520 
1521 	if (!nxtp) {
1522 		nxt = -1;
1523 		nxtp = &nxt;
1524 	}
1525 	while (1) {
1526 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1527 		if (newoff < 0)
1528 			return off;
1529 		else if (newoff < off)
1530 			return -1;	/* invalid */
1531 		else if (newoff == off)
1532 			return newoff;
1533 
1534 		off = newoff;
1535 		proto = *nxtp;
1536 	}
1537 }
1538 
1539 struct ip6aux *
1540 ip6_addaux(struct mbuf *m)
1541 {
1542 	struct m_tag *mtag;
1543 
1544 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1545 	if (!mtag) {
1546 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1547 		    M_NOWAIT);
1548 		if (mtag) {
1549 			m_tag_prepend(m, mtag);
1550 			bzero(mtag + 1, sizeof(struct ip6aux));
1551 		}
1552 	}
1553 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1554 }
1555 
1556 struct ip6aux *
1557 ip6_findaux(struct mbuf *m)
1558 {
1559 	struct m_tag *mtag;
1560 
1561 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1562 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1563 }
1564 
1565 void
1566 ip6_delaux(struct mbuf *m)
1567 {
1568 	struct m_tag *mtag;
1569 
1570 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1571 	if (mtag)
1572 		m_tag_delete(m, mtag);
1573 }
1574 
1575 /*
1576  * System control for IP6
1577  */
1578 
1579 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1580 	0,		0,		0,		0,
1581 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1582 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1583 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1584 	0,		0,		0,		0,
1585 	ENOPROTOOPT
1586 };
1587