xref: /freebsd/sys/netinet6/ip6_input.c (revision bd81e07d2761cf1c13063eb49a5c0cb4a6951318)
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_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/sdt.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/lock.h>
86 #include <sys/rmlock.h>
87 #include <sys/syslog.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_types.h>
92 #include <net/if_dl.h>
93 #include <net/route.h>
94 #include <net/netisr.h>
95 #include <net/pfil.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_kdtrace.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/in_systm.h>
102 #include <net/if_llatbl.h>
103 #ifdef INET
104 #include <netinet/ip.h>
105 #include <netinet/ip_icmp.h>
106 #endif /* INET */
107 #include <netinet/ip6.h>
108 #include <netinet6/in6_var.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet/icmp6.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/in6_ifattach.h>
114 #include <netinet6/nd6.h>
115 
116 #ifdef IPSEC
117 #include <netipsec/ipsec.h>
118 #include <netinet6/ip6_ipsec.h>
119 #include <netipsec/ipsec6.h>
120 #endif /* IPSEC */
121 
122 #include <netinet6/ip6protosw.h>
123 
124 extern struct domain inet6domain;
125 
126 u_char ip6_protox[IPPROTO_MAX];
127 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
128 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
129 VNET_DEFINE(u_long, in6_ifaddrhmask);
130 
131 static struct netisr_handler ip6_nh = {
132 	.nh_name = "ip6",
133 	.nh_handler = ip6_input,
134 	.nh_proto = NETISR_IPV6,
135 	.nh_policy = NETISR_POLICY_FLOW,
136 };
137 
138 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
139 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
140 
141 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
142 
143 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
144 VNET_PCPUSTAT_SYSINIT(ip6stat);
145 #ifdef VIMAGE
146 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
147 #endif /* VIMAGE */
148 
149 struct rmlock in6_ifaddr_lock;
150 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
151 
152 static void ip6_init2(void *);
153 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
154 #ifdef PULLDOWN_TEST
155 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
156 #endif
157 
158 /*
159  * IP6 initialization: fill in IP6 protocol switch table.
160  * All protocols not implemented in kernel go to raw IP6 protocol handler.
161  */
162 void
163 ip6_init(void)
164 {
165 	struct protosw *pr;
166 	int i;
167 
168 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
169 	    &V_ip6_auto_linklocal);
170 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
171 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
172 
173 	TAILQ_INIT(&V_in6_ifaddrhead);
174 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
175 	    &V_in6_ifaddrhmask);
176 
177 	/* Initialize packet filter hooks. */
178 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
179 	V_inet6_pfil_hook.ph_af = AF_INET6;
180 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
181 		printf("%s: WARNING: unable to register pfil hook, "
182 			"error %d\n", __func__, i);
183 
184 	scope6_init();
185 	addrsel_policy_init();
186 	nd6_init();
187 	frag6_init();
188 
189 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
190 
191 	/* Skip global initialization stuff for non-default instances. */
192 	if (!IS_DEFAULT_VNET(curvnet))
193 		return;
194 
195 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
196 	if (pr == NULL)
197 		panic("ip6_init");
198 
199 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
200 	for (i = 0; i < IPPROTO_MAX; i++)
201 		ip6_protox[i] = pr - inet6sw;
202 	/*
203 	 * Cycle through IP protocols and put them into the appropriate place
204 	 * in ip6_protox[].
205 	 */
206 	for (pr = inet6domain.dom_protosw;
207 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
208 		if (pr->pr_domain->dom_family == PF_INET6 &&
209 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
210 			/* Be careful to only index valid IP protocols. */
211 			if (pr->pr_protocol < IPPROTO_MAX)
212 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
213 		}
214 
215 	netisr_register(&ip6_nh);
216 }
217 
218 /*
219  * The protocol to be inserted into ip6_protox[] must be already registered
220  * in inet6sw[], either statically or through pf_proto_register().
221  */
222 int
223 ip6proto_register(short ip6proto)
224 {
225 	struct protosw *pr;
226 
227 	/* Sanity checks. */
228 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
229 		return (EPROTONOSUPPORT);
230 
231 	/*
232 	 * The protocol slot must not be occupied by another protocol
233 	 * already.  An index pointing to IPPROTO_RAW is unused.
234 	 */
235 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
236 	if (pr == NULL)
237 		return (EPFNOSUPPORT);
238 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
239 		return (EEXIST);
240 
241 	/*
242 	 * Find the protocol position in inet6sw[] and set the index.
243 	 */
244 	for (pr = inet6domain.dom_protosw;
245 	    pr < inet6domain.dom_protoswNPROTOSW; pr++) {
246 		if (pr->pr_domain->dom_family == PF_INET6 &&
247 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
248 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
249 			return (0);
250 		}
251 	}
252 	return (EPROTONOSUPPORT);
253 }
254 
255 int
256 ip6proto_unregister(short ip6proto)
257 {
258 	struct protosw *pr;
259 
260 	/* Sanity checks. */
261 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
262 		return (EPROTONOSUPPORT);
263 
264 	/* Check if the protocol was indeed registered. */
265 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
266 	if (pr == NULL)
267 		return (EPFNOSUPPORT);
268 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
269 		return (ENOENT);
270 
271 	/* Reset the protocol slot to IPPROTO_RAW. */
272 	ip6_protox[ip6proto] = pr - inet6sw;
273 	return (0);
274 }
275 
276 #ifdef VIMAGE
277 void
278 ip6_destroy()
279 {
280 	int i;
281 
282 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
283 		printf("%s: WARNING: unable to unregister pfil hook, "
284 		    "error %d\n", __func__, i);
285 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
286 	nd6_destroy();
287 	callout_drain(&V_in6_tmpaddrtimer_ch);
288 }
289 #endif
290 
291 static int
292 ip6_init2_vnet(const void *unused __unused)
293 {
294 
295 	/* nd6_timer_init */
296 	callout_init(&V_nd6_timer_ch, 0);
297 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
298 
299 	/* timer for regeneranation of temporary addresses randomize ID */
300 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
301 	callout_reset(&V_in6_tmpaddrtimer_ch,
302 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
303 		       V_ip6_temp_regen_advance) * hz,
304 		      in6_tmpaddrtimer, curvnet);
305 
306 	return (0);
307 }
308 
309 static void
310 ip6_init2(void *dummy)
311 {
312 
313 	ip6_init2_vnet(NULL);
314 }
315 
316 /* cheat */
317 /* This must be after route_init(), which is now SI_ORDER_THIRD */
318 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
319 
320 static int
321 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
322     int *nxt, int *ours)
323 {
324 	struct ip6_hdr *ip6;
325 	struct ip6_hbh *hbh;
326 
327 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
328 #if 0	/*touches NULL pointer*/
329 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
330 #endif
331 		goto out;	/* m have already been freed */
332 	}
333 
334 	/* adjust pointer */
335 	ip6 = mtod(m, struct ip6_hdr *);
336 
337 	/*
338 	 * if the payload length field is 0 and the next header field
339 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
340 	 * option MUST be included.
341 	 */
342 	if (ip6->ip6_plen == 0 && *plen == 0) {
343 		/*
344 		 * Note that if a valid jumbo payload option is
345 		 * contained, ip6_hopopts_input() must set a valid
346 		 * (non-zero) payload length to the variable plen.
347 		 */
348 		IP6STAT_INC(ip6s_badoptions);
349 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
350 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
351 		icmp6_error(m, ICMP6_PARAM_PROB,
352 			    ICMP6_PARAMPROB_HEADER,
353 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
354 		goto out;
355 	}
356 #ifndef PULLDOWN_TEST
357 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
358 	hbh = (struct ip6_hbh *)(ip6 + 1);
359 #else
360 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
361 		sizeof(struct ip6_hbh));
362 	if (hbh == NULL) {
363 		IP6STAT_INC(ip6s_tooshort);
364 		goto out;
365 	}
366 #endif
367 	*nxt = hbh->ip6h_nxt;
368 
369 	/*
370 	 * If we are acting as a router and the packet contains a
371 	 * router alert option, see if we know the option value.
372 	 * Currently, we only support the option value for MLD, in which
373 	 * case we should pass the packet to the multicast routing
374 	 * daemon.
375 	 */
376 	if (*rtalert != ~0) {
377 		switch (*rtalert) {
378 		case IP6OPT_RTALERT_MLD:
379 			if (V_ip6_forwarding)
380 				*ours = 1;
381 			break;
382 		default:
383 			/*
384 			 * RFC2711 requires unrecognized values must be
385 			 * silently ignored.
386 			 */
387 			break;
388 		}
389 	}
390 
391 	return (0);
392 
393 out:
394 	return (1);
395 }
396 
397 void
398 ip6_input(struct mbuf *m)
399 {
400 	struct in6_addr odst;
401 	struct ip6_hdr *ip6;
402 	struct in6_ifaddr *ia;
403 	u_int32_t plen;
404 	u_int32_t rtalert = ~0;
405 	int off = sizeof(struct ip6_hdr), nest;
406 	int nxt, ours = 0;
407 	int srcrt = 0;
408 
409 #ifdef IPSEC
410 	/*
411 	 * should the inner packet be considered authentic?
412 	 * see comment in ah4_input().
413 	 * NB: m cannot be NULL when passed to the input routine
414 	 */
415 
416 	m->m_flags &= ~M_AUTHIPHDR;
417 	m->m_flags &= ~M_AUTHIPDGM;
418 
419 #endif /* IPSEC */
420 
421 	if (m->m_flags & M_FASTFWD_OURS) {
422 		/*
423 		 * Firewall changed destination to local.
424 		 */
425 		m->m_flags &= ~M_FASTFWD_OURS;
426 		ours = 1;
427 		ip6 = mtod(m, struct ip6_hdr *);
428 		goto hbhcheck;
429 	}
430 
431 	/*
432 	 * mbuf statistics
433 	 */
434 	if (m->m_flags & M_EXT) {
435 		if (m->m_next)
436 			IP6STAT_INC(ip6s_mext2m);
437 		else
438 			IP6STAT_INC(ip6s_mext1);
439 	} else {
440 		if (m->m_next) {
441 			if (m->m_flags & M_LOOP) {
442 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
443 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
444 				IP6STAT_INC(
445 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
446 			else
447 				IP6STAT_INC(ip6s_m2m[0]);
448 		} else
449 			IP6STAT_INC(ip6s_m1);
450 	}
451 
452 	/* drop the packet if IPv6 operation is disabled on the IF */
453 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED))
454 		goto bad;
455 
456 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
457 	IP6STAT_INC(ip6s_total);
458 
459 #ifndef PULLDOWN_TEST
460 	/*
461 	 * L2 bridge code and some other code can return mbuf chain
462 	 * that does not conform to KAME requirement.  too bad.
463 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
464 	 */
465 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
466 		struct mbuf *n;
467 
468 		if (m->m_pkthdr.len > MHLEN)
469 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
470 		else
471 			n = m_gethdr(M_NOWAIT, MT_DATA);
472 		if (n == NULL) {
473 			m_freem(m);
474 			return;	/* ENOBUFS */
475 		}
476 
477 		m_move_pkthdr(n, m);
478 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
479 		n->m_len = n->m_pkthdr.len;
480 		m_freem(m);
481 		m = n;
482 	}
483 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
484 #endif
485 
486 	if (m->m_len < sizeof(struct ip6_hdr)) {
487 		struct ifnet *inifp;
488 		inifp = m->m_pkthdr.rcvif;
489 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
490 			IP6STAT_INC(ip6s_toosmall);
491 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
492 			return;
493 		}
494 	}
495 
496 	ip6 = mtod(m, struct ip6_hdr *);
497 
498 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
499 		IP6STAT_INC(ip6s_badvers);
500 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
501 		goto bad;
502 	}
503 
504 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
505 
506 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
507 
508 	/*
509 	 * Check against address spoofing/corruption.
510 	 */
511 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
512 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
513 		/*
514 		 * XXX: "badscope" is not very suitable for a multicast source.
515 		 */
516 		IP6STAT_INC(ip6s_badscope);
517 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
518 		goto bad;
519 	}
520 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
521 	    !(m->m_flags & M_LOOP)) {
522 		/*
523 		 * In this case, the packet should come from the loopback
524 		 * interface.  However, we cannot just check the if_flags,
525 		 * because ip6_mloopback() passes the "actual" interface
526 		 * as the outgoing/incoming interface.
527 		 */
528 		IP6STAT_INC(ip6s_badscope);
529 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
530 		goto bad;
531 	}
532 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
533 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
534 		/*
535 		 * RFC4291 2.7:
536 		 * Nodes must not originate a packet to a multicast address
537 		 * whose scop field contains the reserved value 0; if such
538 		 * a packet is received, it must be silently dropped.
539 		 */
540 		IP6STAT_INC(ip6s_badscope);
541 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
542 		goto bad;
543 	}
544 #ifdef ALTQ
545 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
546 		/* packet is dropped by traffic conditioner */
547 		return;
548 	}
549 #endif
550 	/*
551 	 * The following check is not documented in specs.  A malicious
552 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
553 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
554 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
555 	 *
556 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
557 	 * support IPv4-less kernel compilation, we cannot support SIIT
558 	 * environment at all.  So, it makes more sense for us to reject any
559 	 * malicious packets for non-SIIT environment, than try to do a
560 	 * partial support for SIIT environment.
561 	 */
562 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
563 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
564 		IP6STAT_INC(ip6s_badscope);
565 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
566 		goto bad;
567 	}
568 #if 0
569 	/*
570 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
571 	 *
572 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
573 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
574 	 * is revised to forbid relaying case.
575 	 */
576 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
577 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
578 		IP6STAT_INC(ip6s_badscope);
579 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
580 		goto bad;
581 	}
582 #endif
583 #ifdef IPSEC
584 	/*
585 	 * Bypass packet filtering for packets previously handled by IPsec.
586 	 */
587 	if (ip6_ipsec_filtertunnel(m))
588 		goto passin;
589 #endif /* IPSEC */
590 
591 	/*
592 	 * Run through list of hooks for input packets.
593 	 *
594 	 * NB: Beware of the destination address changing
595 	 *     (e.g. by NAT rewriting).  When this happens,
596 	 *     tell ip6_forward to do the right thing.
597 	 */
598 	odst = ip6->ip6_dst;
599 
600 	/* Jump over all PFIL processing if hooks are not active. */
601 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
602 		goto passin;
603 
604 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
605 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
606 		return;
607 	if (m == NULL)			/* consumed by filter */
608 		return;
609 	ip6 = mtod(m, struct ip6_hdr *);
610 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
611 
612 	if (m->m_flags & M_FASTFWD_OURS) {
613 		m->m_flags &= ~M_FASTFWD_OURS;
614 		ours = 1;
615 		goto hbhcheck;
616 	}
617 	if ((m->m_flags & M_IP6_NEXTHOP) &&
618 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
619 		/*
620 		 * Directly ship the packet on.  This allows forwarding
621 		 * packets originally destined to us to some other directly
622 		 * connected host.
623 		 */
624 		ip6_forward(m, 1);
625 		return;
626 	}
627 
628 passin:
629 	/*
630 	 * Disambiguate address scope zones (if there is ambiguity).
631 	 * We first make sure that the original source or destination address
632 	 * is not in our internal form for scoped addresses.  Such addresses
633 	 * are not necessarily invalid spec-wise, but we cannot accept them due
634 	 * to the usage conflict.
635 	 * in6_setscope() then also checks and rejects the cases where src or
636 	 * dst are the loopback address and the receiving interface
637 	 * is not loopback.
638 	 */
639 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
640 		IP6STAT_INC(ip6s_badscope); /* XXX */
641 		goto bad;
642 	}
643 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
644 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
645 		IP6STAT_INC(ip6s_badscope);
646 		goto bad;
647 	}
648 	/*
649 	 * Multicast check. Assume packet is for us to avoid
650 	 * prematurely taking locks.
651 	 */
652 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
653 		ours = 1;
654 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
655 		goto hbhcheck;
656 	}
657 	/*
658 	 * Unicast check
659 	 * XXX: For now we keep link-local IPv6 addresses with embedded
660 	 *      scope zone id, therefore we use zero zoneid here.
661 	 */
662 	ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
663 	if (ia != NULL) {
664 		if (ia->ia6_flags & IN6_IFF_NOTREADY) {
665 			char ip6bufs[INET6_ADDRSTRLEN];
666 			char ip6bufd[INET6_ADDRSTRLEN];
667 			/* address is not ready, so discard the packet. */
668 			nd6log((LOG_INFO,
669 			    "ip6_input: packet to an unready address %s->%s\n",
670 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
671 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
672 			ifa_free(&ia->ia_ifa);
673 			goto bad;
674 		}
675 		/* Count the packet in the ip address stats */
676 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
677 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
678 		ifa_free(&ia->ia_ifa);
679 		ours = 1;
680 		goto hbhcheck;
681 	}
682 
683 	/*
684 	 * Now there is no reason to process the packet if it's not our own
685 	 * and we're not a router.
686 	 */
687 	if (!V_ip6_forwarding) {
688 		IP6STAT_INC(ip6s_cantforward);
689 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
690 		goto bad;
691 	}
692 
693   hbhcheck:
694 	/*
695 	 * Process Hop-by-Hop options header if it's contained.
696 	 * m may be modified in ip6_hopopts_input().
697 	 * If a JumboPayload option is included, plen will also be modified.
698 	 */
699 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
700 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
701 		if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0)
702 			return;
703 	} else
704 		nxt = ip6->ip6_nxt;
705 
706 	/*
707 	 * Check that the amount of data in the buffers
708 	 * is as at least much as the IPv6 header would have us expect.
709 	 * Trim mbufs if longer than we expect.
710 	 * Drop packet if shorter than we expect.
711 	 */
712 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
713 		IP6STAT_INC(ip6s_tooshort);
714 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
715 		goto bad;
716 	}
717 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
718 		if (m->m_len == m->m_pkthdr.len) {
719 			m->m_len = sizeof(struct ip6_hdr) + plen;
720 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
721 		} else
722 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
723 	}
724 
725 	/*
726 	 * Forward if desirable.
727 	 */
728 	if (V_ip6_mrouter &&
729 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
730 		/*
731 		 * If we are acting as a multicast router, all
732 		 * incoming multicast packets are passed to the
733 		 * kernel-level multicast forwarding function.
734 		 * The packet is returned (relatively) intact; if
735 		 * ip6_mforward() returns a non-zero value, the packet
736 		 * must be discarded, else it may be accepted below.
737 		 *
738 		 * XXX TODO: Check hlim and multicast scope here to avoid
739 		 * unnecessarily calling into ip6_mforward().
740 		 */
741 		if (ip6_mforward &&
742 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
743 			IP6STAT_INC(ip6s_cantforward);
744 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
745 			goto bad;
746 		}
747 	} else if (!ours) {
748 		ip6_forward(m, srcrt);
749 		return;
750 	}
751 
752 	ip6 = mtod(m, struct ip6_hdr *);
753 
754 	/*
755 	 * Malicious party may be able to use IPv4 mapped addr to confuse
756 	 * tcp/udp stack and bypass security checks (act as if it was from
757 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
758 	 *
759 	 * For SIIT end node behavior, you may want to disable the check.
760 	 * However, you will  become vulnerable to attacks using IPv4 mapped
761 	 * source.
762 	 */
763 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
764 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
765 		IP6STAT_INC(ip6s_badscope);
766 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
767 		goto bad;
768 	}
769 
770 	/*
771 	 * Tell launch routine the next header
772 	 */
773 	IP6STAT_INC(ip6s_delivered);
774 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver);
775 	nest = 0;
776 
777 	while (nxt != IPPROTO_DONE) {
778 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
779 			IP6STAT_INC(ip6s_toomanyhdr);
780 			goto bad;
781 		}
782 
783 		/*
784 		 * protection against faulty packet - there should be
785 		 * more sanity checks in header chain processing.
786 		 */
787 		if (m->m_pkthdr.len < off) {
788 			IP6STAT_INC(ip6s_tooshort);
789 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
790 			goto bad;
791 		}
792 
793 #ifdef IPSEC
794 		/*
795 		 * enforce IPsec policy checking if we are seeing last header.
796 		 * note that we do not visit this with protocols with pcb layer
797 		 * code - like udp/tcp/raw ip.
798 		 */
799 		if (ip6_ipsec_input(m, nxt))
800 			goto bad;
801 #endif /* IPSEC */
802 
803 		/*
804 		 * Use mbuf flags to propagate Router Alert option to
805 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
806 		 */
807 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
808 			m->m_flags |= M_RTALERT_MLD;
809 
810 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
811 	}
812 	return;
813 bad:
814 	m_freem(m);
815 }
816 
817 /*
818  * Hop-by-Hop options header processing. If a valid jumbo payload option is
819  * included, the real payload length will be stored in plenp.
820  *
821  * rtalertp - XXX: should be stored more smart way
822  */
823 static int
824 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
825     struct mbuf **mp, int *offp)
826 {
827 	struct mbuf *m = *mp;
828 	int off = *offp, hbhlen;
829 	struct ip6_hbh *hbh;
830 
831 	/* validation of the length of the header */
832 #ifndef PULLDOWN_TEST
833 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
834 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
835 	hbhlen = (hbh->ip6h_len + 1) << 3;
836 
837 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
838 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
839 #else
840 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
841 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
842 	if (hbh == NULL) {
843 		IP6STAT_INC(ip6s_tooshort);
844 		return -1;
845 	}
846 	hbhlen = (hbh->ip6h_len + 1) << 3;
847 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
848 		hbhlen);
849 	if (hbh == NULL) {
850 		IP6STAT_INC(ip6s_tooshort);
851 		return -1;
852 	}
853 #endif
854 	off += hbhlen;
855 	hbhlen -= sizeof(struct ip6_hbh);
856 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
857 				hbhlen, rtalertp, plenp) < 0)
858 		return (-1);
859 
860 	*offp = off;
861 	*mp = m;
862 	return (0);
863 }
864 
865 /*
866  * Search header for all Hop-by-hop options and process each option.
867  * This function is separate from ip6_hopopts_input() in order to
868  * handle a case where the sending node itself process its hop-by-hop
869  * options header. In such a case, the function is called from ip6_output().
870  *
871  * The function assumes that hbh header is located right after the IPv6 header
872  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
873  * opthead + hbhlen is located in contiguous memory region.
874  */
875 int
876 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
877     u_int32_t *rtalertp, u_int32_t *plenp)
878 {
879 	struct ip6_hdr *ip6;
880 	int optlen = 0;
881 	u_int8_t *opt = opthead;
882 	u_int16_t rtalert_val;
883 	u_int32_t jumboplen;
884 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
885 
886 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
887 		switch (*opt) {
888 		case IP6OPT_PAD1:
889 			optlen = 1;
890 			break;
891 		case IP6OPT_PADN:
892 			if (hbhlen < IP6OPT_MINLEN) {
893 				IP6STAT_INC(ip6s_toosmall);
894 				goto bad;
895 			}
896 			optlen = *(opt + 1) + 2;
897 			break;
898 		case IP6OPT_ROUTER_ALERT:
899 			/* XXX may need check for alignment */
900 			if (hbhlen < IP6OPT_RTALERT_LEN) {
901 				IP6STAT_INC(ip6s_toosmall);
902 				goto bad;
903 			}
904 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
905 				/* XXX stat */
906 				icmp6_error(m, ICMP6_PARAM_PROB,
907 				    ICMP6_PARAMPROB_HEADER,
908 				    erroff + opt + 1 - opthead);
909 				return (-1);
910 			}
911 			optlen = IP6OPT_RTALERT_LEN;
912 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
913 			*rtalertp = ntohs(rtalert_val);
914 			break;
915 		case IP6OPT_JUMBO:
916 			/* XXX may need check for alignment */
917 			if (hbhlen < IP6OPT_JUMBO_LEN) {
918 				IP6STAT_INC(ip6s_toosmall);
919 				goto bad;
920 			}
921 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
922 				/* XXX stat */
923 				icmp6_error(m, ICMP6_PARAM_PROB,
924 				    ICMP6_PARAMPROB_HEADER,
925 				    erroff + opt + 1 - opthead);
926 				return (-1);
927 			}
928 			optlen = IP6OPT_JUMBO_LEN;
929 
930 			/*
931 			 * IPv6 packets that have non 0 payload length
932 			 * must not contain a jumbo payload option.
933 			 */
934 			ip6 = mtod(m, struct ip6_hdr *);
935 			if (ip6->ip6_plen) {
936 				IP6STAT_INC(ip6s_badoptions);
937 				icmp6_error(m, ICMP6_PARAM_PROB,
938 				    ICMP6_PARAMPROB_HEADER,
939 				    erroff + opt - opthead);
940 				return (-1);
941 			}
942 
943 			/*
944 			 * We may see jumbolen in unaligned location, so
945 			 * we'd need to perform bcopy().
946 			 */
947 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
948 			jumboplen = (u_int32_t)htonl(jumboplen);
949 
950 #if 1
951 			/*
952 			 * if there are multiple jumbo payload options,
953 			 * *plenp will be non-zero and the packet will be
954 			 * rejected.
955 			 * the behavior may need some debate in ipngwg -
956 			 * multiple options does not make sense, however,
957 			 * there's no explicit mention in specification.
958 			 */
959 			if (*plenp != 0) {
960 				IP6STAT_INC(ip6s_badoptions);
961 				icmp6_error(m, ICMP6_PARAM_PROB,
962 				    ICMP6_PARAMPROB_HEADER,
963 				    erroff + opt + 2 - opthead);
964 				return (-1);
965 			}
966 #endif
967 
968 			/*
969 			 * jumbo payload length must be larger than 65535.
970 			 */
971 			if (jumboplen <= IPV6_MAXPACKET) {
972 				IP6STAT_INC(ip6s_badoptions);
973 				icmp6_error(m, ICMP6_PARAM_PROB,
974 				    ICMP6_PARAMPROB_HEADER,
975 				    erroff + opt + 2 - opthead);
976 				return (-1);
977 			}
978 			*plenp = jumboplen;
979 
980 			break;
981 		default:		/* unknown option */
982 			if (hbhlen < IP6OPT_MINLEN) {
983 				IP6STAT_INC(ip6s_toosmall);
984 				goto bad;
985 			}
986 			optlen = ip6_unknown_opt(opt, m,
987 			    erroff + opt - opthead);
988 			if (optlen == -1)
989 				return (-1);
990 			optlen += 2;
991 			break;
992 		}
993 	}
994 
995 	return (0);
996 
997   bad:
998 	m_freem(m);
999 	return (-1);
1000 }
1001 
1002 /*
1003  * Unknown option processing.
1004  * The third argument `off' is the offset from the IPv6 header to the option,
1005  * which is necessary if the IPv6 header the and option header and IPv6 header
1006  * is not contiguous in order to return an ICMPv6 error.
1007  */
1008 int
1009 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1010 {
1011 	struct ip6_hdr *ip6;
1012 
1013 	switch (IP6OPT_TYPE(*optp)) {
1014 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1015 		return ((int)*(optp + 1));
1016 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1017 		m_freem(m);
1018 		return (-1);
1019 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1020 		IP6STAT_INC(ip6s_badoptions);
1021 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1022 		return (-1);
1023 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1024 		IP6STAT_INC(ip6s_badoptions);
1025 		ip6 = mtod(m, struct ip6_hdr *);
1026 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1027 		    (m->m_flags & (M_BCAST|M_MCAST)))
1028 			m_freem(m);
1029 		else
1030 			icmp6_error(m, ICMP6_PARAM_PROB,
1031 				    ICMP6_PARAMPROB_OPTION, off);
1032 		return (-1);
1033 	}
1034 
1035 	m_freem(m);		/* XXX: NOTREACHED */
1036 	return (-1);
1037 }
1038 
1039 /*
1040  * Create the "control" list for this pcb.
1041  * These functions will not modify mbuf chain at all.
1042  *
1043  * With KAME mbuf chain restriction:
1044  * The routine will be called from upper layer handlers like tcp6_input().
1045  * Thus the routine assumes that the caller (tcp6_input) have already
1046  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1047  * very first mbuf on the mbuf chain.
1048  *
1049  * ip6_savecontrol_v4 will handle those options that are possible to be
1050  * set on a v4-mapped socket.
1051  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1052  * options and handle the v6-only ones itself.
1053  */
1054 struct mbuf **
1055 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1056     int *v4only)
1057 {
1058 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1059 
1060 #ifdef SO_TIMESTAMP
1061 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1062 		struct timeval tv;
1063 
1064 		microtime(&tv);
1065 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1066 		    SCM_TIMESTAMP, SOL_SOCKET);
1067 		if (*mp)
1068 			mp = &(*mp)->m_next;
1069 	}
1070 #endif
1071 
1072 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1073 	/* RFC 2292 sec. 5 */
1074 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1075 		struct in6_pktinfo pi6;
1076 
1077 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1078 #ifdef INET
1079 			struct ip *ip;
1080 
1081 			ip = mtod(m, struct ip *);
1082 			pi6.ipi6_addr.s6_addr32[0] = 0;
1083 			pi6.ipi6_addr.s6_addr32[1] = 0;
1084 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1085 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1086 #else
1087 			/* We won't hit this code */
1088 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1089 #endif
1090 		} else {
1091 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1092 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1093 		}
1094 		pi6.ipi6_ifindex =
1095 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1096 
1097 		*mp = sbcreatecontrol((caddr_t) &pi6,
1098 		    sizeof(struct in6_pktinfo),
1099 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1100 		if (*mp)
1101 			mp = &(*mp)->m_next;
1102 	}
1103 
1104 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1105 		int hlim;
1106 
1107 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1108 #ifdef INET
1109 			struct ip *ip;
1110 
1111 			ip = mtod(m, struct ip *);
1112 			hlim = ip->ip_ttl;
1113 #else
1114 			/* We won't hit this code */
1115 			hlim = 0;
1116 #endif
1117 		} else {
1118 			hlim = ip6->ip6_hlim & 0xff;
1119 		}
1120 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1121 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1122 		    IPPROTO_IPV6);
1123 		if (*mp)
1124 			mp = &(*mp)->m_next;
1125 	}
1126 
1127 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1128 		int tclass;
1129 
1130 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1131 #ifdef INET
1132 			struct ip *ip;
1133 
1134 			ip = mtod(m, struct ip *);
1135 			tclass = ip->ip_tos;
1136 #else
1137 			/* We won't hit this code */
1138 			tclass = 0;
1139 #endif
1140 		} else {
1141 			u_int32_t flowinfo;
1142 
1143 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1144 			flowinfo >>= 20;
1145 			tclass = flowinfo & 0xff;
1146 		}
1147 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1148 		    IPV6_TCLASS, IPPROTO_IPV6);
1149 		if (*mp)
1150 			mp = &(*mp)->m_next;
1151 	}
1152 
1153 	if (v4only != NULL) {
1154 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1155 			*v4only = 1;
1156 		} else {
1157 			*v4only = 0;
1158 		}
1159 	}
1160 
1161 	return (mp);
1162 }
1163 
1164 void
1165 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1166 {
1167 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1168 	int v4only = 0;
1169 
1170 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1171 	if (v4only)
1172 		return;
1173 
1174 	/*
1175 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1176 	 * privilege for the option (see ip6_ctloutput), but it might be too
1177 	 * strict, since there might be some hop-by-hop options which can be
1178 	 * returned to normal user.
1179 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1180 	 */
1181 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1182 		/*
1183 		 * Check if a hop-by-hop options header is contatined in the
1184 		 * received packet, and if so, store the options as ancillary
1185 		 * data. Note that a hop-by-hop options header must be
1186 		 * just after the IPv6 header, which is assured through the
1187 		 * IPv6 input processing.
1188 		 */
1189 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1190 			struct ip6_hbh *hbh;
1191 			int hbhlen = 0;
1192 #ifdef PULLDOWN_TEST
1193 			struct mbuf *ext;
1194 #endif
1195 
1196 #ifndef PULLDOWN_TEST
1197 			hbh = (struct ip6_hbh *)(ip6 + 1);
1198 			hbhlen = (hbh->ip6h_len + 1) << 3;
1199 #else
1200 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1201 			    ip6->ip6_nxt);
1202 			if (ext == NULL) {
1203 				IP6STAT_INC(ip6s_tooshort);
1204 				return;
1205 			}
1206 			hbh = mtod(ext, struct ip6_hbh *);
1207 			hbhlen = (hbh->ip6h_len + 1) << 3;
1208 			if (hbhlen != ext->m_len) {
1209 				m_freem(ext);
1210 				IP6STAT_INC(ip6s_tooshort);
1211 				return;
1212 			}
1213 #endif
1214 
1215 			/*
1216 			 * XXX: We copy the whole header even if a
1217 			 * jumbo payload option is included, the option which
1218 			 * is to be removed before returning according to
1219 			 * RFC2292.
1220 			 * Note: this constraint is removed in RFC3542
1221 			 */
1222 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1223 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1224 			    IPPROTO_IPV6);
1225 			if (*mp)
1226 				mp = &(*mp)->m_next;
1227 #ifdef PULLDOWN_TEST
1228 			m_freem(ext);
1229 #endif
1230 		}
1231 	}
1232 
1233 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1234 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1235 
1236 		/*
1237 		 * Search for destination options headers or routing
1238 		 * header(s) through the header chain, and stores each
1239 		 * header as ancillary data.
1240 		 * Note that the order of the headers remains in
1241 		 * the chain of ancillary data.
1242 		 */
1243 		while (1) {	/* is explicit loop prevention necessary? */
1244 			struct ip6_ext *ip6e = NULL;
1245 			int elen;
1246 #ifdef PULLDOWN_TEST
1247 			struct mbuf *ext = NULL;
1248 #endif
1249 
1250 			/*
1251 			 * if it is not an extension header, don't try to
1252 			 * pull it from the chain.
1253 			 */
1254 			switch (nxt) {
1255 			case IPPROTO_DSTOPTS:
1256 			case IPPROTO_ROUTING:
1257 			case IPPROTO_HOPOPTS:
1258 			case IPPROTO_AH: /* is it possible? */
1259 				break;
1260 			default:
1261 				goto loopend;
1262 			}
1263 
1264 #ifndef PULLDOWN_TEST
1265 			if (off + sizeof(*ip6e) > m->m_len)
1266 				goto loopend;
1267 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1268 			if (nxt == IPPROTO_AH)
1269 				elen = (ip6e->ip6e_len + 2) << 2;
1270 			else
1271 				elen = (ip6e->ip6e_len + 1) << 3;
1272 			if (off + elen > m->m_len)
1273 				goto loopend;
1274 #else
1275 			ext = ip6_pullexthdr(m, off, nxt);
1276 			if (ext == NULL) {
1277 				IP6STAT_INC(ip6s_tooshort);
1278 				return;
1279 			}
1280 			ip6e = mtod(ext, struct ip6_ext *);
1281 			if (nxt == IPPROTO_AH)
1282 				elen = (ip6e->ip6e_len + 2) << 2;
1283 			else
1284 				elen = (ip6e->ip6e_len + 1) << 3;
1285 			if (elen != ext->m_len) {
1286 				m_freem(ext);
1287 				IP6STAT_INC(ip6s_tooshort);
1288 				return;
1289 			}
1290 #endif
1291 
1292 			switch (nxt) {
1293 			case IPPROTO_DSTOPTS:
1294 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1295 					break;
1296 
1297 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1298 				    IS2292(in6p,
1299 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1300 				    IPPROTO_IPV6);
1301 				if (*mp)
1302 					mp = &(*mp)->m_next;
1303 				break;
1304 			case IPPROTO_ROUTING:
1305 				if (!(in6p->inp_flags & IN6P_RTHDR))
1306 					break;
1307 
1308 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1309 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1310 				    IPPROTO_IPV6);
1311 				if (*mp)
1312 					mp = &(*mp)->m_next;
1313 				break;
1314 			case IPPROTO_HOPOPTS:
1315 			case IPPROTO_AH: /* is it possible? */
1316 				break;
1317 
1318 			default:
1319 				/*
1320 				 * other cases have been filtered in the above.
1321 				 * none will visit this case.  here we supply
1322 				 * the code just in case (nxt overwritten or
1323 				 * other cases).
1324 				 */
1325 #ifdef PULLDOWN_TEST
1326 				m_freem(ext);
1327 #endif
1328 				goto loopend;
1329 
1330 			}
1331 
1332 			/* proceed with the next header. */
1333 			off += elen;
1334 			nxt = ip6e->ip6e_nxt;
1335 			ip6e = NULL;
1336 #ifdef PULLDOWN_TEST
1337 			m_freem(ext);
1338 			ext = NULL;
1339 #endif
1340 		}
1341 	  loopend:
1342 		;
1343 	}
1344 }
1345 #undef IS2292
1346 
1347 void
1348 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1349 {
1350 	struct socket *so;
1351 	struct mbuf *m_mtu;
1352 	struct ip6_mtuinfo mtuctl;
1353 
1354 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1355 	/*
1356 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1357 	 * application wanted to know the MTU value.
1358 	 * NOTE: we notify disconnected sockets, because some udp
1359 	 * applications keep sending sockets disconnected.
1360 	 * NOTE: our implementation doesn't notify connected sockets that has
1361 	 * foreign address that is different than given destination addresses
1362 	 * (this is permitted by RFC 3542).
1363 	 */
1364 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1365 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1366 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1367 		return;
1368 
1369 	mtuctl.ip6m_mtu = mtu;
1370 	mtuctl.ip6m_addr = *dst;
1371 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1372 		return;
1373 
1374 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1375 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1376 		return;
1377 
1378 	so =  inp->inp_socket;
1379 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1380 	    == 0) {
1381 		m_freem(m_mtu);
1382 		/* XXX: should count statistics */
1383 	} else
1384 		sorwakeup(so);
1385 }
1386 
1387 #ifdef PULLDOWN_TEST
1388 /*
1389  * pull single extension header from mbuf chain.  returns single mbuf that
1390  * contains the result, or NULL on error.
1391  */
1392 static struct mbuf *
1393 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1394 {
1395 	struct ip6_ext ip6e;
1396 	size_t elen;
1397 	struct mbuf *n;
1398 
1399 #ifdef DIAGNOSTIC
1400 	switch (nxt) {
1401 	case IPPROTO_DSTOPTS:
1402 	case IPPROTO_ROUTING:
1403 	case IPPROTO_HOPOPTS:
1404 	case IPPROTO_AH: /* is it possible? */
1405 		break;
1406 	default:
1407 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1408 	}
1409 #endif
1410 
1411 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1412 	if (nxt == IPPROTO_AH)
1413 		elen = (ip6e.ip6e_len + 2) << 2;
1414 	else
1415 		elen = (ip6e.ip6e_len + 1) << 3;
1416 
1417 	if (elen > MLEN)
1418 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1419 	else
1420 		n = m_get(M_NOWAIT, MT_DATA);
1421 	if (n == NULL)
1422 		return NULL;
1423 
1424 	m_copydata(m, off, elen, mtod(n, caddr_t));
1425 	n->m_len = elen;
1426 	return n;
1427 }
1428 #endif
1429 
1430 /*
1431  * Get pointer to the previous header followed by the header
1432  * currently processed.
1433  * XXX: This function supposes that
1434  *	M includes all headers,
1435  *	the next header field and the header length field of each header
1436  *	are valid, and
1437  *	the sum of each header length equals to OFF.
1438  * Because of these assumptions, this function must be called very
1439  * carefully. Moreover, it will not be used in the near future when
1440  * we develop `neater' mechanism to process extension headers.
1441  */
1442 char *
1443 ip6_get_prevhdr(struct mbuf *m, int off)
1444 {
1445 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1446 
1447 	if (off == sizeof(struct ip6_hdr))
1448 		return (&ip6->ip6_nxt);
1449 	else {
1450 		int len, nxt;
1451 		struct ip6_ext *ip6e = NULL;
1452 
1453 		nxt = ip6->ip6_nxt;
1454 		len = sizeof(struct ip6_hdr);
1455 		while (len < off) {
1456 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1457 
1458 			switch (nxt) {
1459 			case IPPROTO_FRAGMENT:
1460 				len += sizeof(struct ip6_frag);
1461 				break;
1462 			case IPPROTO_AH:
1463 				len += (ip6e->ip6e_len + 2) << 2;
1464 				break;
1465 			default:
1466 				len += (ip6e->ip6e_len + 1) << 3;
1467 				break;
1468 			}
1469 			nxt = ip6e->ip6e_nxt;
1470 		}
1471 		if (ip6e)
1472 			return (&ip6e->ip6e_nxt);
1473 		else
1474 			return NULL;
1475 	}
1476 }
1477 
1478 /*
1479  * get next header offset.  m will be retained.
1480  */
1481 int
1482 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1483 {
1484 	struct ip6_hdr ip6;
1485 	struct ip6_ext ip6e;
1486 	struct ip6_frag fh;
1487 
1488 	/* just in case */
1489 	if (m == NULL)
1490 		panic("ip6_nexthdr: m == NULL");
1491 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1492 		return -1;
1493 
1494 	switch (proto) {
1495 	case IPPROTO_IPV6:
1496 		if (m->m_pkthdr.len < off + sizeof(ip6))
1497 			return -1;
1498 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1499 		if (nxtp)
1500 			*nxtp = ip6.ip6_nxt;
1501 		off += sizeof(ip6);
1502 		return off;
1503 
1504 	case IPPROTO_FRAGMENT:
1505 		/*
1506 		 * terminate parsing if it is not the first fragment,
1507 		 * it does not make sense to parse through it.
1508 		 */
1509 		if (m->m_pkthdr.len < off + sizeof(fh))
1510 			return -1;
1511 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1512 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1513 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1514 			return -1;
1515 		if (nxtp)
1516 			*nxtp = fh.ip6f_nxt;
1517 		off += sizeof(struct ip6_frag);
1518 		return off;
1519 
1520 	case IPPROTO_AH:
1521 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1522 			return -1;
1523 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1524 		if (nxtp)
1525 			*nxtp = ip6e.ip6e_nxt;
1526 		off += (ip6e.ip6e_len + 2) << 2;
1527 		return off;
1528 
1529 	case IPPROTO_HOPOPTS:
1530 	case IPPROTO_ROUTING:
1531 	case IPPROTO_DSTOPTS:
1532 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1533 			return -1;
1534 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1535 		if (nxtp)
1536 			*nxtp = ip6e.ip6e_nxt;
1537 		off += (ip6e.ip6e_len + 1) << 3;
1538 		return off;
1539 
1540 	case IPPROTO_NONE:
1541 	case IPPROTO_ESP:
1542 	case IPPROTO_IPCOMP:
1543 		/* give up */
1544 		return -1;
1545 
1546 	default:
1547 		return -1;
1548 	}
1549 
1550 	return -1;
1551 }
1552 
1553 /*
1554  * get offset for the last header in the chain.  m will be kept untainted.
1555  */
1556 int
1557 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1558 {
1559 	int newoff;
1560 	int nxt;
1561 
1562 	if (!nxtp) {
1563 		nxt = -1;
1564 		nxtp = &nxt;
1565 	}
1566 	while (1) {
1567 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1568 		if (newoff < 0)
1569 			return off;
1570 		else if (newoff < off)
1571 			return -1;	/* invalid */
1572 		else if (newoff == off)
1573 			return newoff;
1574 
1575 		off = newoff;
1576 		proto = *nxtp;
1577 	}
1578 }
1579 
1580 /*
1581  * System control for IP6
1582  */
1583 
1584 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1585 	0,		0,		0,		0,
1586 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1587 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1588 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1589 	0,		0,		0,		0,
1590 	ENOPROTOOPT
1591 };
1592