xref: /freebsd/sys/netinet6/ip6_input.c (revision 4ed925457ab06e83238a5db33e89ccc94b99a713)
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 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/if_dl.h>
87 #include <net/route.h>
88 #include <net/netisr.h>
89 #include <net/pfil.h>
90 #include <net/vnet.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #include <net/if_llatbl.h>
95 #ifdef INET
96 #include <netinet/ip.h>
97 #include <netinet/ip_icmp.h>
98 #endif /* INET */
99 #include <netinet/ip6.h>
100 #include <netinet6/in6_var.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet/in_pcb.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/nd6.h>
107 
108 #ifdef IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netinet6/ip6_ipsec.h>
111 #include <netipsec/ipsec6.h>
112 #endif /* IPSEC */
113 
114 #include <netinet6/ip6protosw.h>
115 
116 extern struct domain inet6domain;
117 
118 u_char ip6_protox[IPPROTO_MAX];
119 
120 static struct netisr_handler ip6_nh = {
121 	.nh_name = "ip6",
122 	.nh_handler = ip6_input,
123 	.nh_proto = NETISR_IPV6,
124 	.nh_policy = NETISR_POLICY_FLOW,
125 };
126 
127 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
128 VNET_DEFINE(struct ip6stat, ip6stat);
129 
130 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
131 VNET_DECLARE(int, dad_init);
132 VNET_DECLARE(int, pmtu_expire);
133 VNET_DECLARE(int, pmtu_probe);
134 VNET_DECLARE(u_long, rip6_sendspace);
135 VNET_DECLARE(u_long, rip6_recvspace);
136 VNET_DECLARE(int, icmp6errppslim);
137 VNET_DECLARE(int, icmp6_nodeinfo);
138 VNET_DECLARE(int, udp6_sendspace);
139 VNET_DECLARE(int, udp6_recvspace);
140 
141 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
142 #define	V_dad_init			VNET(dad_init)
143 #define	V_pmtu_expire			VNET(pmtu_expire)
144 #define	V_pmtu_probe			VNET(pmtu_probe)
145 #define	V_rip6_sendspace		VNET(rip6_sendspace)
146 #define	V_rip6_recvspace		VNET(rip6_recvspace)
147 #define	V_icmp6errppslim		VNET(icmp6errppslim)
148 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
149 #define	V_udp6_sendspace		VNET(udp6_sendspace)
150 #define	V_udp6_recvspace		VNET(udp6_recvspace)
151 
152 struct rwlock in6_ifaddr_lock;
153 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
154 
155 VNET_DEFINE (struct pfil_head, inet6_pfil_hook);
156 
157 static void ip6_init2(void *);
158 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
159 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
160 #ifdef PULLDOWN_TEST
161 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
162 #endif
163 
164 /*
165  * IP6 initialization: fill in IP6 protocol switch table.
166  * All protocols not implemented in kernel go to raw IP6 protocol handler.
167  */
168 void
169 ip6_init(void)
170 {
171 	struct ip6protosw *pr;
172 	int i;
173 
174 	V_in6_maxmtu = 0;
175 #ifdef IP6_AUTO_LINKLOCAL
176 	V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL;
177 #else
178 	V_ip6_auto_linklocal = 1;	/* enabled by default */
179 #endif
180 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
181 	    &V_ip6_auto_linklocal);
182 
183 #ifndef IPV6FORWARDING
184 #ifdef GATEWAY6
185 #define IPV6FORWARDING	1	/* forward IP6 packets not for us */
186 #else
187 #define IPV6FORWARDING	0	/* don't forward IP6 packets not for us */
188 #endif /* GATEWAY6 */
189 #endif /* !IPV6FORWARDING */
190 
191 #ifndef IPV6_SENDREDIRECTS
192 #define IPV6_SENDREDIRECTS	1
193 #endif
194 
195 	V_ip6_forwarding = IPV6FORWARDING; /* act as router? */
196 	V_ip6_sendredirects = IPV6_SENDREDIRECTS;
197 	V_ip6_defhlim = IPV6_DEFHLIM;
198 	V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS;
199 	V_ip6_accept_rtadv = 0;
200 	V_ip6_log_interval = 5;
201 	V_ip6_hdrnestlimit = 15; /* How many header options will we process? */
202 	V_ip6_dad_count = 1;	 /* DupAddrDetectionTransmits */
203 	V_ip6_auto_flowlabel = 1;
204 	V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */
205 	V_ip6_rr_prune = 5;	 /* router renumbering prefix
206                                   * walk list every 5 sec. */
207 	V_ip6_mcast_pmtu = 0;	 /* enable pMTU discovery for multicast? */
208 	V_ip6_v6only = 1;
209 	V_ip6_keepfaith = 0;
210 	V_ip6_log_time = (time_t)0L;
211 #ifdef IPSTEALTH
212 	V_ip6stealth = 0;
213 #endif
214 	V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */
215 
216 	V_pmtu_expire = 60*10;
217 	V_pmtu_probe = 60*2;
218 
219 	/* raw IP6 parameters */
220 	/*
221 	 * Nominal space allocated to a raw ip socket.
222 	 */
223 #define RIPV6SNDQ	8192
224 #define RIPV6RCVQ	8192
225 	V_rip6_sendspace = RIPV6SNDQ;
226 	V_rip6_recvspace = RIPV6RCVQ;
227 
228 	/* ICMPV6 parameters */
229 	V_icmp6_rediraccept = 1;	/* accept and process redirects */
230 	V_icmp6_redirtimeout = 10 * 60;	/* 10 minutes */
231 	V_icmp6errppslim = 100;		/* 100pps */
232 	/* control how to respond to NI queries */
233 	V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK);
234 
235 	/* UDP on IP6 parameters */
236 	V_udp6_sendspace = 9216;	/* really max datagram size */
237 	V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
238 					/* 40 1K datagrams */
239 	V_dad_init = 0;
240 
241 	TAILQ_INIT(&V_in6_ifaddrhead);
242 
243 	scope6_init();
244 	addrsel_policy_init();
245 	nd6_init();
246 	frag6_init();
247 
248 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
249 
250 	/* Initialize packet filter hooks. */
251 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
252 	V_inet6_pfil_hook.ph_af = AF_INET6;
253 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
254 		printf("%s: WARNING: unable to register pfil hook, "
255 			"error %d\n", __func__, i);
256 
257 	/* Skip global initialization stuff for non-default instances. */
258 	if (!IS_DEFAULT_VNET(curvnet))
259 		return;
260 
261 #ifdef DIAGNOSTIC
262 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
263 		panic("sizeof(protosw) != sizeof(ip6protosw)");
264 #endif
265 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
266 	if (pr == NULL)
267 		panic("ip6_init");
268 
269 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
270 	for (i = 0; i < IPPROTO_MAX; i++)
271 		ip6_protox[i] = pr - inet6sw;
272 	/*
273 	 * Cycle through IP protocols and put them into the appropriate place
274 	 * in ip6_protox[].
275 	 */
276 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
277 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
278 		if (pr->pr_domain->dom_family == PF_INET6 &&
279 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
280 			/* Be careful to only index valid IP protocols. */
281 			if (pr->pr_protocol < IPPROTO_MAX)
282 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
283 		}
284 
285 	netisr_register(&ip6_nh);
286 }
287 
288 #ifdef VIMAGE
289 void
290 ip6_destroy()
291 {
292 
293 	nd6_destroy();
294 	callout_drain(&V_in6_tmpaddrtimer_ch);
295 }
296 #endif
297 
298 static int
299 ip6_init2_vnet(const void *unused __unused)
300 {
301 
302 	/* nd6_timer_init */
303 	callout_init(&V_nd6_timer_ch, 0);
304 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
305 
306 	/* timer for regeneranation of temporary addresses randomize ID */
307 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
308 	callout_reset(&V_in6_tmpaddrtimer_ch,
309 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
310 		       V_ip6_temp_regen_advance) * hz,
311 		      in6_tmpaddrtimer, curvnet);
312 
313 	return (0);
314 }
315 
316 static void
317 ip6_init2(void *dummy)
318 {
319 
320 	ip6_init2_vnet(NULL);
321 }
322 
323 /* cheat */
324 /* This must be after route_init(), which is now SI_ORDER_THIRD */
325 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
326 
327 void
328 ip6_input(struct mbuf *m)
329 {
330 	struct ip6_hdr *ip6;
331 	int off = sizeof(struct ip6_hdr), nest;
332 	u_int32_t plen;
333 	u_int32_t rtalert = ~0;
334 	int nxt, ours = 0;
335 	struct ifnet *deliverifp = NULL, *ifp = NULL;
336 	struct in6_addr odst;
337 	struct route_in6 rin6;
338 	int srcrt = 0;
339 	struct llentry *lle = NULL;
340 	struct sockaddr_in6 dst6, *dst;
341 
342 	bzero(&rin6, sizeof(struct route_in6));
343 #ifdef IPSEC
344 	/*
345 	 * should the inner packet be considered authentic?
346 	 * see comment in ah4_input().
347 	 * NB: m cannot be NULL when passed to the input routine
348 	 */
349 
350 	m->m_flags &= ~M_AUTHIPHDR;
351 	m->m_flags &= ~M_AUTHIPDGM;
352 
353 #endif /* IPSEC */
354 
355 	/*
356 	 * make sure we don't have onion peering information into m_tag.
357 	 */
358 	ip6_delaux(m);
359 
360 	/*
361 	 * mbuf statistics
362 	 */
363 	if (m->m_flags & M_EXT) {
364 		if (m->m_next)
365 			V_ip6stat.ip6s_mext2m++;
366 		else
367 			V_ip6stat.ip6s_mext1++;
368 	} else {
369 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
370 		if (m->m_next) {
371 			if (m->m_flags & M_LOOP) {
372 				V_ip6stat.ip6s_m2m[V_loif->if_index]++;
373 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
374 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
375 			else
376 				V_ip6stat.ip6s_m2m[0]++;
377 		} else
378 			V_ip6stat.ip6s_m1++;
379 #undef M2MMAX
380 	}
381 
382 	/* drop the packet if IPv6 operation is disabled on the IF */
383 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
384 		m_freem(m);
385 		return;
386 	}
387 
388 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
389 	V_ip6stat.ip6s_total++;
390 
391 #ifndef PULLDOWN_TEST
392 	/*
393 	 * L2 bridge code and some other code can return mbuf chain
394 	 * that does not conform to KAME requirement.  too bad.
395 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
396 	 */
397 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
398 		struct mbuf *n;
399 
400 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
401 		if (n)
402 			M_MOVE_PKTHDR(n, m);
403 		if (n && n->m_pkthdr.len > MHLEN) {
404 			MCLGET(n, M_DONTWAIT);
405 			if ((n->m_flags & M_EXT) == 0) {
406 				m_freem(n);
407 				n = NULL;
408 			}
409 		}
410 		if (n == NULL) {
411 			m_freem(m);
412 			return;	/* ENOBUFS */
413 		}
414 
415 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
416 		n->m_len = n->m_pkthdr.len;
417 		m_freem(m);
418 		m = n;
419 	}
420 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
421 #endif
422 
423 	if (m->m_len < sizeof(struct ip6_hdr)) {
424 		struct ifnet *inifp;
425 		inifp = m->m_pkthdr.rcvif;
426 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
427 			V_ip6stat.ip6s_toosmall++;
428 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
429 			return;
430 		}
431 	}
432 
433 	ip6 = mtod(m, struct ip6_hdr *);
434 
435 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
436 		V_ip6stat.ip6s_badvers++;
437 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
438 		goto bad;
439 	}
440 
441 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
442 
443 	/*
444 	 * Check against address spoofing/corruption.
445 	 */
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
447 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
448 		/*
449 		 * XXX: "badscope" is not very suitable for a multicast source.
450 		 */
451 		V_ip6stat.ip6s_badscope++;
452 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
453 		goto bad;
454 	}
455 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
456 	    !(m->m_flags & M_LOOP)) {
457 		/*
458 		 * In this case, the packet should come from the loopback
459 		 * interface.  However, we cannot just check the if_flags,
460 		 * because ip6_mloopback() passes the "actual" interface
461 		 * as the outgoing/incoming interface.
462 		 */
463 		V_ip6stat.ip6s_badscope++;
464 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
465 		goto bad;
466 	}
467 
468 #ifdef ALTQ
469 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
470 		/* packet is dropped by traffic conditioner */
471 		return;
472 	}
473 #endif
474 	/*
475 	 * The following check is not documented in specs.  A malicious
476 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
477 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
478 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
479 	 *
480 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
481 	 * support IPv4-less kernel compilation, we cannot support SIIT
482 	 * environment at all.  So, it makes more sense for us to reject any
483 	 * malicious packets for non-SIIT environment, than try to do a
484 	 * partial support for SIIT environment.
485 	 */
486 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
487 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
488 		V_ip6stat.ip6s_badscope++;
489 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
490 		goto bad;
491 	}
492 #if 0
493 	/*
494 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
495 	 *
496 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
497 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
498 	 * is revised to forbid relaying case.
499 	 */
500 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
501 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
502 		V_ip6stat.ip6s_badscope++;
503 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
504 		goto bad;
505 	}
506 #endif
507 
508 	/*
509 	 * Run through list of hooks for input packets.
510 	 *
511 	 * NB: Beware of the destination address changing
512 	 *     (e.g. by NAT rewriting).  When this happens,
513 	 *     tell ip6_forward to do the right thing.
514 	 */
515 	odst = ip6->ip6_dst;
516 
517 	/* Jump over all PFIL processing if hooks are not active. */
518 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
519 		goto passin;
520 
521 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
522 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
523 		return;
524 	if (m == NULL)			/* consumed by filter */
525 		return;
526 	ip6 = mtod(m, struct ip6_hdr *);
527 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
528 
529 passin:
530 	/*
531 	 * Disambiguate address scope zones (if there is ambiguity).
532 	 * We first make sure that the original source or destination address
533 	 * is not in our internal form for scoped addresses.  Such addresses
534 	 * are not necessarily invalid spec-wise, but we cannot accept them due
535 	 * to the usage conflict.
536 	 * in6_setscope() then also checks and rejects the cases where src or
537 	 * dst are the loopback address and the receiving interface
538 	 * is not loopback.
539 	 */
540 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
541 		V_ip6stat.ip6s_badscope++; /* XXX */
542 		goto bad;
543 	}
544 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
545 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
546 		V_ip6stat.ip6s_badscope++;
547 		goto bad;
548 	}
549 
550 	/*
551 	 * Multicast check. Assume packet is for us to avoid
552 	 * prematurely taking locks.
553 	 */
554 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
555 		ours = 1;
556 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
557 		deliverifp = m->m_pkthdr.rcvif;
558 		goto hbhcheck;
559 	}
560 
561 	/*
562 	 *  Unicast check
563 	 */
564 
565 	bzero(&dst6, sizeof(dst6));
566 	dst6.sin6_family = AF_INET6;
567 	dst6.sin6_len = sizeof(struct sockaddr_in6);
568 	dst6.sin6_addr = ip6->ip6_dst;
569 	ifp = m->m_pkthdr.rcvif;
570 	IF_AFDATA_LOCK(ifp);
571 	lle = lla_lookup(LLTABLE6(ifp), 0,
572 	     (struct sockaddr *)&dst6);
573 	IF_AFDATA_UNLOCK(ifp);
574 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
575 		ours = 1;
576 		deliverifp = ifp;
577 		LLE_RUNLOCK(lle);
578 		goto hbhcheck;
579 	}
580 	if (lle != NULL)
581 		LLE_RUNLOCK(lle);
582 
583 	dst = &rin6.ro_dst;
584 	dst->sin6_len = sizeof(struct sockaddr_in6);
585 	dst->sin6_family = AF_INET6;
586 	dst->sin6_addr = ip6->ip6_dst;
587 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
588 	if (rin6.ro_rt)
589 		RT_UNLOCK(rin6.ro_rt);
590 
591 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
592 
593 	/*
594 	 * Accept the packet if the forwarding interface to the destination
595 	 * according to the routing table is the loopback interface,
596 	 * unless the associated route has a gateway.
597 	 * Note that this approach causes to accept a packet if there is a
598 	 * route to the loopback interface for the destination of the packet.
599 	 * But we think it's even useful in some situations, e.g. when using
600 	 * a special daemon which wants to intercept the packet.
601 	 *
602 	 * XXX: some OSes automatically make a cloned route for the destination
603 	 * of an outgoing packet.  If the outgoing interface of the packet
604 	 * is a loopback one, the kernel would consider the packet to be
605 	 * accepted, even if we have no such address assinged on the interface.
606 	 * We check the cloned flag of the route entry to reject such cases,
607 	 * assuming that route entries for our own addresses are not made by
608 	 * cloning (it should be true because in6_addloop explicitly installs
609 	 * the host route).  However, we might have to do an explicit check
610 	 * while it would be less efficient.  Or, should we rather install a
611 	 * reject route for such a case?
612 	 */
613 	if (rin6.ro_rt &&
614 	    (rin6.ro_rt->rt_flags &
615 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
616 #ifdef RTF_WASCLONED
617 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
618 #endif
619 #ifdef RTF_CLONED
620 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
621 #endif
622 #if 0
623 	    /*
624 	     * The check below is redundant since the comparison of
625 	     * the destination and the key of the rtentry has
626 	     * already done through looking up the routing table.
627 	     */
628 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
629 	    &rt6_key(rin6.ro_rt)->sin6_addr)
630 #endif
631 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
632 		int free_ia6 = 0;
633 		struct in6_ifaddr *ia6;
634 
635 		/*
636 		 * found the loopback route to the interface address
637 		 */
638 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
639 			struct sockaddr_in6 dest6;
640 
641 			bzero(&dest6, sizeof(dest6));
642 			dest6.sin6_family = AF_INET6;
643 			dest6.sin6_len = sizeof(dest6);
644 			dest6.sin6_addr = ip6->ip6_dst;
645 			ia6 = (struct in6_ifaddr *)
646 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
647 			if (ia6 == NULL)
648 				goto bad;
649 			free_ia6 = 1;
650 		}
651 		else
652 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
653 
654 		/*
655 		 * record address information into m_tag.
656 		 */
657 		(void)ip6_setdstifaddr(m, ia6);
658 
659 		/*
660 		 * packets to a tentative, duplicated, or somehow invalid
661 		 * address must not be accepted.
662 		 */
663 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
664 			/* this address is ready */
665 			ours = 1;
666 			deliverifp = ia6->ia_ifp;	/* correct? */
667 			/* Count the packet in the ip address stats */
668 			ia6->ia_ifa.if_ipackets++;
669 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
670 			if (ia6 != NULL && free_ia6 != 0)
671 				ifa_free(&ia6->ia_ifa);
672 			goto hbhcheck;
673 		} else {
674 			char ip6bufs[INET6_ADDRSTRLEN];
675 			char ip6bufd[INET6_ADDRSTRLEN];
676 			/* address is not ready, so discard the packet. */
677 			nd6log((LOG_INFO,
678 			    "ip6_input: packet to an unready address %s->%s\n",
679 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
680 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
681 
682 			if (ia6 != NULL && free_ia6 != 0)
683 				ifa_free(&ia6->ia_ifa);
684 			goto bad;
685 		}
686 	}
687 
688 	/*
689 	 * FAITH (Firewall Aided Internet Translator)
690 	 */
691 	if (V_ip6_keepfaith) {
692 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
693 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
694 			/* XXX do we need more sanity checks? */
695 			ours = 1;
696 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
697 			goto hbhcheck;
698 		}
699 	}
700 
701 	/*
702 	 * Now there is no reason to process the packet if it's not our own
703 	 * and we're not a router.
704 	 */
705 	if (!V_ip6_forwarding) {
706 		V_ip6stat.ip6s_cantforward++;
707 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
708 		goto bad;
709 	}
710 
711   hbhcheck:
712 	/*
713 	 * record address information into m_tag, if we don't have one yet.
714 	 * note that we are unable to record it, if the address is not listed
715 	 * as our interface address (e.g. multicast addresses, addresses
716 	 * within FAITH prefixes and such).
717 	 */
718 	if (deliverifp && !ip6_getdstifaddr(m)) {
719 		struct in6_ifaddr *ia6;
720 
721 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
722 		if (ia6) {
723 			if (!ip6_setdstifaddr(m, ia6)) {
724 				/*
725 				 * XXX maybe we should drop the packet here,
726 				 * as we could not provide enough information
727 				 * to the upper layers.
728 				 */
729 			}
730 			ifa_free(&ia6->ia_ifa);
731 		}
732 	}
733 
734 	/*
735 	 * Process Hop-by-Hop options header if it's contained.
736 	 * m may be modified in ip6_hopopts_input().
737 	 * If a JumboPayload option is included, plen will also be modified.
738 	 */
739 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
740 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
741 		struct ip6_hbh *hbh;
742 
743 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
744 #if 0	/*touches NULL pointer*/
745 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
746 #endif
747 			goto out;	/* m have already been freed */
748 		}
749 
750 		/* adjust pointer */
751 		ip6 = mtod(m, struct ip6_hdr *);
752 
753 		/*
754 		 * if the payload length field is 0 and the next header field
755 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
756 		 * option MUST be included.
757 		 */
758 		if (ip6->ip6_plen == 0 && plen == 0) {
759 			/*
760 			 * Note that if a valid jumbo payload option is
761 			 * contained, ip6_hopopts_input() must set a valid
762 			 * (non-zero) payload length to the variable plen.
763 			 */
764 			V_ip6stat.ip6s_badoptions++;
765 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
766 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
767 			icmp6_error(m, ICMP6_PARAM_PROB,
768 				    ICMP6_PARAMPROB_HEADER,
769 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
770 			goto out;
771 		}
772 #ifndef PULLDOWN_TEST
773 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
774 		hbh = (struct ip6_hbh *)(ip6 + 1);
775 #else
776 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
777 			sizeof(struct ip6_hbh));
778 		if (hbh == NULL) {
779 			V_ip6stat.ip6s_tooshort++;
780 			goto out;
781 		}
782 #endif
783 		nxt = hbh->ip6h_nxt;
784 
785 		/*
786 		 * If we are acting as a router and the packet contains a
787 		 * router alert option, see if we know the option value.
788 		 * Currently, we only support the option value for MLD, in which
789 		 * case we should pass the packet to the multicast routing
790 		 * daemon.
791 		 */
792 		if (rtalert != ~0) {
793 			switch (rtalert) {
794 			case IP6OPT_RTALERT_MLD:
795 				if (V_ip6_forwarding)
796 					ours = 1;
797 				break;
798 			default:
799 				/*
800 				 * RFC2711 requires unrecognized values must be
801 				 * silently ignored.
802 				 */
803 				break;
804 			}
805 		}
806 	} else
807 		nxt = ip6->ip6_nxt;
808 
809 	/*
810 	 * Check that the amount of data in the buffers
811 	 * is as at least much as the IPv6 header would have us expect.
812 	 * Trim mbufs if longer than we expect.
813 	 * Drop packet if shorter than we expect.
814 	 */
815 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
816 		V_ip6stat.ip6s_tooshort++;
817 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
818 		goto bad;
819 	}
820 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
821 		if (m->m_len == m->m_pkthdr.len) {
822 			m->m_len = sizeof(struct ip6_hdr) + plen;
823 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
824 		} else
825 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
826 	}
827 
828 	/*
829 	 * Forward if desirable.
830 	 */
831 	if (V_ip6_mrouter &&
832 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
833 		/*
834 		 * If we are acting as a multicast router, all
835 		 * incoming multicast packets are passed to the
836 		 * kernel-level multicast forwarding function.
837 		 * The packet is returned (relatively) intact; if
838 		 * ip6_mforward() returns a non-zero value, the packet
839 		 * must be discarded, else it may be accepted below.
840 		 *
841 		 * XXX TODO: Check hlim and multicast scope here to avoid
842 		 * unnecessarily calling into ip6_mforward().
843 		 */
844 		if (ip6_mforward &&
845 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
846 			IP6STAT_INC(ip6s_cantforward);
847 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
848 			goto bad;
849 		}
850 	} else if (!ours) {
851 		ip6_forward(m, srcrt);
852 		goto out;
853 	}
854 
855 	ip6 = mtod(m, struct ip6_hdr *);
856 
857 	/*
858 	 * Malicious party may be able to use IPv4 mapped addr to confuse
859 	 * tcp/udp stack and bypass security checks (act as if it was from
860 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
861 	 *
862 	 * For SIIT end node behavior, you may want to disable the check.
863 	 * However, you will  become vulnerable to attacks using IPv4 mapped
864 	 * source.
865 	 */
866 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
867 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
868 		V_ip6stat.ip6s_badscope++;
869 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
870 		goto bad;
871 	}
872 
873 	/*
874 	 * Tell launch routine the next header
875 	 */
876 	V_ip6stat.ip6s_delivered++;
877 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
878 	nest = 0;
879 
880 	while (nxt != IPPROTO_DONE) {
881 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
882 			V_ip6stat.ip6s_toomanyhdr++;
883 			goto bad;
884 		}
885 
886 		/*
887 		 * protection against faulty packet - there should be
888 		 * more sanity checks in header chain processing.
889 		 */
890 		if (m->m_pkthdr.len < off) {
891 			V_ip6stat.ip6s_tooshort++;
892 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
893 			goto bad;
894 		}
895 
896 #ifdef IPSEC
897 		/*
898 		 * enforce IPsec policy checking if we are seeing last header.
899 		 * note that we do not visit this with protocols with pcb layer
900 		 * code - like udp/tcp/raw ip.
901 		 */
902 		if (ip6_ipsec_input(m, nxt))
903 			goto bad;
904 #endif /* IPSEC */
905 
906 		/*
907 		 * Use mbuf flags to propagate Router Alert option to
908 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
909 		 */
910 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
911 			m->m_flags |= M_RTALERT_MLD;
912 
913 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
914 	}
915 	goto out;
916 bad:
917 	m_freem(m);
918 out:
919 	if (rin6.ro_rt)
920 		RTFREE(rin6.ro_rt);
921 }
922 
923 /*
924  * set/grab in6_ifaddr correspond to IPv6 destination address.
925  * XXX backward compatibility wrapper
926  *
927  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
928  * and then bump it when the tag is copied, and release it when the tag is
929  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
930  * we just bump the ia refcount when we receive it.  This should be fixed.
931  */
932 static struct ip6aux *
933 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
934 {
935 	struct ip6aux *ip6a;
936 
937 	ip6a = ip6_addaux(m);
938 	if (ip6a)
939 		ip6a->ip6a_dstia6 = ia6;
940 	return ip6a;	/* NULL if failed to set */
941 }
942 
943 struct in6_ifaddr *
944 ip6_getdstifaddr(struct mbuf *m)
945 {
946 	struct ip6aux *ip6a;
947 	struct in6_ifaddr *ia;
948 
949 	ip6a = ip6_findaux(m);
950 	if (ip6a) {
951 		ia = ip6a->ip6a_dstia6;
952 		ifa_ref(&ia->ia_ifa);
953 		return ia;
954 	} else
955 		return NULL;
956 }
957 
958 /*
959  * Hop-by-Hop options header processing. If a valid jumbo payload option is
960  * included, the real payload length will be stored in plenp.
961  *
962  * rtalertp - XXX: should be stored more smart way
963  */
964 static int
965 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
966     struct mbuf **mp, int *offp)
967 {
968 	struct mbuf *m = *mp;
969 	int off = *offp, hbhlen;
970 	struct ip6_hbh *hbh;
971 	u_int8_t *opt;
972 
973 	/* validation of the length of the header */
974 #ifndef PULLDOWN_TEST
975 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
976 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
977 	hbhlen = (hbh->ip6h_len + 1) << 3;
978 
979 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
980 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
981 #else
982 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
983 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
984 	if (hbh == NULL) {
985 		V_ip6stat.ip6s_tooshort++;
986 		return -1;
987 	}
988 	hbhlen = (hbh->ip6h_len + 1) << 3;
989 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
990 		hbhlen);
991 	if (hbh == NULL) {
992 		V_ip6stat.ip6s_tooshort++;
993 		return -1;
994 	}
995 #endif
996 	off += hbhlen;
997 	hbhlen -= sizeof(struct ip6_hbh);
998 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
999 
1000 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1001 				hbhlen, rtalertp, plenp) < 0)
1002 		return (-1);
1003 
1004 	*offp = off;
1005 	*mp = m;
1006 	return (0);
1007 }
1008 
1009 /*
1010  * Search header for all Hop-by-hop options and process each option.
1011  * This function is separate from ip6_hopopts_input() in order to
1012  * handle a case where the sending node itself process its hop-by-hop
1013  * options header. In such a case, the function is called from ip6_output().
1014  *
1015  * The function assumes that hbh header is located right after the IPv6 header
1016  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1017  * opthead + hbhlen is located in continuous memory region.
1018  */
1019 int
1020 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1021     u_int32_t *rtalertp, u_int32_t *plenp)
1022 {
1023 	struct ip6_hdr *ip6;
1024 	int optlen = 0;
1025 	u_int8_t *opt = opthead;
1026 	u_int16_t rtalert_val;
1027 	u_int32_t jumboplen;
1028 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1029 
1030 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1031 		switch (*opt) {
1032 		case IP6OPT_PAD1:
1033 			optlen = 1;
1034 			break;
1035 		case IP6OPT_PADN:
1036 			if (hbhlen < IP6OPT_MINLEN) {
1037 				V_ip6stat.ip6s_toosmall++;
1038 				goto bad;
1039 			}
1040 			optlen = *(opt + 1) + 2;
1041 			break;
1042 		case IP6OPT_ROUTER_ALERT:
1043 			/* XXX may need check for alignment */
1044 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1045 				V_ip6stat.ip6s_toosmall++;
1046 				goto bad;
1047 			}
1048 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1049 				/* XXX stat */
1050 				icmp6_error(m, ICMP6_PARAM_PROB,
1051 				    ICMP6_PARAMPROB_HEADER,
1052 				    erroff + opt + 1 - opthead);
1053 				return (-1);
1054 			}
1055 			optlen = IP6OPT_RTALERT_LEN;
1056 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1057 			*rtalertp = ntohs(rtalert_val);
1058 			break;
1059 		case IP6OPT_JUMBO:
1060 			/* XXX may need check for alignment */
1061 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1062 				V_ip6stat.ip6s_toosmall++;
1063 				goto bad;
1064 			}
1065 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1066 				/* XXX stat */
1067 				icmp6_error(m, ICMP6_PARAM_PROB,
1068 				    ICMP6_PARAMPROB_HEADER,
1069 				    erroff + opt + 1 - opthead);
1070 				return (-1);
1071 			}
1072 			optlen = IP6OPT_JUMBO_LEN;
1073 
1074 			/*
1075 			 * IPv6 packets that have non 0 payload length
1076 			 * must not contain a jumbo payload option.
1077 			 */
1078 			ip6 = mtod(m, struct ip6_hdr *);
1079 			if (ip6->ip6_plen) {
1080 				V_ip6stat.ip6s_badoptions++;
1081 				icmp6_error(m, ICMP6_PARAM_PROB,
1082 				    ICMP6_PARAMPROB_HEADER,
1083 				    erroff + opt - opthead);
1084 				return (-1);
1085 			}
1086 
1087 			/*
1088 			 * We may see jumbolen in unaligned location, so
1089 			 * we'd need to perform bcopy().
1090 			 */
1091 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1092 			jumboplen = (u_int32_t)htonl(jumboplen);
1093 
1094 #if 1
1095 			/*
1096 			 * if there are multiple jumbo payload options,
1097 			 * *plenp will be non-zero and the packet will be
1098 			 * rejected.
1099 			 * the behavior may need some debate in ipngwg -
1100 			 * multiple options does not make sense, however,
1101 			 * there's no explicit mention in specification.
1102 			 */
1103 			if (*plenp != 0) {
1104 				V_ip6stat.ip6s_badoptions++;
1105 				icmp6_error(m, ICMP6_PARAM_PROB,
1106 				    ICMP6_PARAMPROB_HEADER,
1107 				    erroff + opt + 2 - opthead);
1108 				return (-1);
1109 			}
1110 #endif
1111 
1112 			/*
1113 			 * jumbo payload length must be larger than 65535.
1114 			 */
1115 			if (jumboplen <= IPV6_MAXPACKET) {
1116 				V_ip6stat.ip6s_badoptions++;
1117 				icmp6_error(m, ICMP6_PARAM_PROB,
1118 				    ICMP6_PARAMPROB_HEADER,
1119 				    erroff + opt + 2 - opthead);
1120 				return (-1);
1121 			}
1122 			*plenp = jumboplen;
1123 
1124 			break;
1125 		default:		/* unknown option */
1126 			if (hbhlen < IP6OPT_MINLEN) {
1127 				V_ip6stat.ip6s_toosmall++;
1128 				goto bad;
1129 			}
1130 			optlen = ip6_unknown_opt(opt, m,
1131 			    erroff + opt - opthead);
1132 			if (optlen == -1)
1133 				return (-1);
1134 			optlen += 2;
1135 			break;
1136 		}
1137 	}
1138 
1139 	return (0);
1140 
1141   bad:
1142 	m_freem(m);
1143 	return (-1);
1144 }
1145 
1146 /*
1147  * Unknown option processing.
1148  * The third argument `off' is the offset from the IPv6 header to the option,
1149  * which is necessary if the IPv6 header the and option header and IPv6 header
1150  * is not continuous in order to return an ICMPv6 error.
1151  */
1152 int
1153 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1154 {
1155 	struct ip6_hdr *ip6;
1156 
1157 	switch (IP6OPT_TYPE(*optp)) {
1158 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1159 		return ((int)*(optp + 1));
1160 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1161 		m_freem(m);
1162 		return (-1);
1163 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1164 		V_ip6stat.ip6s_badoptions++;
1165 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1166 		return (-1);
1167 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1168 		V_ip6stat.ip6s_badoptions++;
1169 		ip6 = mtod(m, struct ip6_hdr *);
1170 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1171 		    (m->m_flags & (M_BCAST|M_MCAST)))
1172 			m_freem(m);
1173 		else
1174 			icmp6_error(m, ICMP6_PARAM_PROB,
1175 				    ICMP6_PARAMPROB_OPTION, off);
1176 		return (-1);
1177 	}
1178 
1179 	m_freem(m);		/* XXX: NOTREACHED */
1180 	return (-1);
1181 }
1182 
1183 /*
1184  * Create the "control" list for this pcb.
1185  * These functions will not modify mbuf chain at all.
1186  *
1187  * With KAME mbuf chain restriction:
1188  * The routine will be called from upper layer handlers like tcp6_input().
1189  * Thus the routine assumes that the caller (tcp6_input) have already
1190  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1191  * very first mbuf on the mbuf chain.
1192  *
1193  * ip6_savecontrol_v4 will handle those options that are possible to be
1194  * set on a v4-mapped socket.
1195  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1196  * options and handle the v6-only ones itself.
1197  */
1198 struct mbuf **
1199 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1200     int *v4only)
1201 {
1202 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1203 
1204 #ifdef SO_TIMESTAMP
1205 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1206 		struct timeval tv;
1207 
1208 		microtime(&tv);
1209 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1210 		    SCM_TIMESTAMP, SOL_SOCKET);
1211 		if (*mp)
1212 			mp = &(*mp)->m_next;
1213 	}
1214 #endif
1215 
1216 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1217 		if (v4only != NULL)
1218 			*v4only = 1;
1219 		return (mp);
1220 	}
1221 
1222 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1223 	/* RFC 2292 sec. 5 */
1224 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1225 		struct in6_pktinfo pi6;
1226 
1227 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1228 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1229 		pi6.ipi6_ifindex =
1230 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1231 
1232 		*mp = sbcreatecontrol((caddr_t) &pi6,
1233 		    sizeof(struct in6_pktinfo),
1234 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1235 		if (*mp)
1236 			mp = &(*mp)->m_next;
1237 	}
1238 
1239 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1240 		int hlim = ip6->ip6_hlim & 0xff;
1241 
1242 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1243 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1244 		    IPPROTO_IPV6);
1245 		if (*mp)
1246 			mp = &(*mp)->m_next;
1247 	}
1248 
1249 	if (v4only != NULL)
1250 		*v4only = 0;
1251 	return (mp);
1252 }
1253 
1254 void
1255 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1256 {
1257 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1258 	int v4only = 0;
1259 
1260 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1261 	if (v4only)
1262 		return;
1263 
1264 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1265 		u_int32_t flowinfo;
1266 		int tclass;
1267 
1268 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1269 		flowinfo >>= 20;
1270 
1271 		tclass = flowinfo & 0xff;
1272 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1273 		    IPV6_TCLASS, IPPROTO_IPV6);
1274 		if (*mp)
1275 			mp = &(*mp)->m_next;
1276 	}
1277 
1278 	/*
1279 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1280 	 * privilege for the option (see ip6_ctloutput), but it might be too
1281 	 * strict, since there might be some hop-by-hop options which can be
1282 	 * returned to normal user.
1283 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1284 	 */
1285 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1286 		/*
1287 		 * Check if a hop-by-hop options header is contatined in the
1288 		 * received packet, and if so, store the options as ancillary
1289 		 * data. Note that a hop-by-hop options header must be
1290 		 * just after the IPv6 header, which is assured through the
1291 		 * IPv6 input processing.
1292 		 */
1293 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1294 			struct ip6_hbh *hbh;
1295 			int hbhlen = 0;
1296 #ifdef PULLDOWN_TEST
1297 			struct mbuf *ext;
1298 #endif
1299 
1300 #ifndef PULLDOWN_TEST
1301 			hbh = (struct ip6_hbh *)(ip6 + 1);
1302 			hbhlen = (hbh->ip6h_len + 1) << 3;
1303 #else
1304 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1305 			    ip6->ip6_nxt);
1306 			if (ext == NULL) {
1307 				V_ip6stat.ip6s_tooshort++;
1308 				return;
1309 			}
1310 			hbh = mtod(ext, struct ip6_hbh *);
1311 			hbhlen = (hbh->ip6h_len + 1) << 3;
1312 			if (hbhlen != ext->m_len) {
1313 				m_freem(ext);
1314 				V_ip6stat.ip6s_tooshort++;
1315 				return;
1316 			}
1317 #endif
1318 
1319 			/*
1320 			 * XXX: We copy the whole header even if a
1321 			 * jumbo payload option is included, the option which
1322 			 * is to be removed before returning according to
1323 			 * RFC2292.
1324 			 * Note: this constraint is removed in RFC3542
1325 			 */
1326 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1327 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1328 			    IPPROTO_IPV6);
1329 			if (*mp)
1330 				mp = &(*mp)->m_next;
1331 #ifdef PULLDOWN_TEST
1332 			m_freem(ext);
1333 #endif
1334 		}
1335 	}
1336 
1337 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1338 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1339 
1340 		/*
1341 		 * Search for destination options headers or routing
1342 		 * header(s) through the header chain, and stores each
1343 		 * header as ancillary data.
1344 		 * Note that the order of the headers remains in
1345 		 * the chain of ancillary data.
1346 		 */
1347 		while (1) {	/* is explicit loop prevention necessary? */
1348 			struct ip6_ext *ip6e = NULL;
1349 			int elen;
1350 #ifdef PULLDOWN_TEST
1351 			struct mbuf *ext = NULL;
1352 #endif
1353 
1354 			/*
1355 			 * if it is not an extension header, don't try to
1356 			 * pull it from the chain.
1357 			 */
1358 			switch (nxt) {
1359 			case IPPROTO_DSTOPTS:
1360 			case IPPROTO_ROUTING:
1361 			case IPPROTO_HOPOPTS:
1362 			case IPPROTO_AH: /* is it possible? */
1363 				break;
1364 			default:
1365 				goto loopend;
1366 			}
1367 
1368 #ifndef PULLDOWN_TEST
1369 			if (off + sizeof(*ip6e) > m->m_len)
1370 				goto loopend;
1371 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1372 			if (nxt == IPPROTO_AH)
1373 				elen = (ip6e->ip6e_len + 2) << 2;
1374 			else
1375 				elen = (ip6e->ip6e_len + 1) << 3;
1376 			if (off + elen > m->m_len)
1377 				goto loopend;
1378 #else
1379 			ext = ip6_pullexthdr(m, off, nxt);
1380 			if (ext == NULL) {
1381 				V_ip6stat.ip6s_tooshort++;
1382 				return;
1383 			}
1384 			ip6e = mtod(ext, struct ip6_ext *);
1385 			if (nxt == IPPROTO_AH)
1386 				elen = (ip6e->ip6e_len + 2) << 2;
1387 			else
1388 				elen = (ip6e->ip6e_len + 1) << 3;
1389 			if (elen != ext->m_len) {
1390 				m_freem(ext);
1391 				V_ip6stat.ip6s_tooshort++;
1392 				return;
1393 			}
1394 #endif
1395 
1396 			switch (nxt) {
1397 			case IPPROTO_DSTOPTS:
1398 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1399 					break;
1400 
1401 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1402 				    IS2292(in6p,
1403 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1404 				    IPPROTO_IPV6);
1405 				if (*mp)
1406 					mp = &(*mp)->m_next;
1407 				break;
1408 			case IPPROTO_ROUTING:
1409 				if (!(in6p->inp_flags & IN6P_RTHDR))
1410 					break;
1411 
1412 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1413 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1414 				    IPPROTO_IPV6);
1415 				if (*mp)
1416 					mp = &(*mp)->m_next;
1417 				break;
1418 			case IPPROTO_HOPOPTS:
1419 			case IPPROTO_AH: /* is it possible? */
1420 				break;
1421 
1422 			default:
1423 				/*
1424 				 * other cases have been filtered in the above.
1425 				 * none will visit this case.  here we supply
1426 				 * the code just in case (nxt overwritten or
1427 				 * other cases).
1428 				 */
1429 #ifdef PULLDOWN_TEST
1430 				m_freem(ext);
1431 #endif
1432 				goto loopend;
1433 
1434 			}
1435 
1436 			/* proceed with the next header. */
1437 			off += elen;
1438 			nxt = ip6e->ip6e_nxt;
1439 			ip6e = NULL;
1440 #ifdef PULLDOWN_TEST
1441 			m_freem(ext);
1442 			ext = NULL;
1443 #endif
1444 		}
1445 	  loopend:
1446 		;
1447 	}
1448 }
1449 #undef IS2292
1450 
1451 void
1452 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1453 {
1454 	struct socket *so;
1455 	struct mbuf *m_mtu;
1456 	struct ip6_mtuinfo mtuctl;
1457 
1458 	so =  in6p->inp_socket;
1459 
1460 	if (mtu == NULL)
1461 		return;
1462 
1463 #ifdef DIAGNOSTIC
1464 	if (so == NULL)		/* I believe this is impossible */
1465 		panic("ip6_notify_pmtu: socket is NULL");
1466 #endif
1467 
1468 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1469 	mtuctl.ip6m_mtu = *mtu;
1470 	mtuctl.ip6m_addr = *dst;
1471 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1472 		return;
1473 
1474 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1475 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1476 		return;
1477 
1478 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1479 	    == 0) {
1480 		m_freem(m_mtu);
1481 		/* XXX: should count statistics */
1482 	} else
1483 		sorwakeup(so);
1484 
1485 	return;
1486 }
1487 
1488 #ifdef PULLDOWN_TEST
1489 /*
1490  * pull single extension header from mbuf chain.  returns single mbuf that
1491  * contains the result, or NULL on error.
1492  */
1493 static struct mbuf *
1494 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1495 {
1496 	struct ip6_ext ip6e;
1497 	size_t elen;
1498 	struct mbuf *n;
1499 
1500 #ifdef DIAGNOSTIC
1501 	switch (nxt) {
1502 	case IPPROTO_DSTOPTS:
1503 	case IPPROTO_ROUTING:
1504 	case IPPROTO_HOPOPTS:
1505 	case IPPROTO_AH: /* is it possible? */
1506 		break;
1507 	default:
1508 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1509 	}
1510 #endif
1511 
1512 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1513 	if (nxt == IPPROTO_AH)
1514 		elen = (ip6e.ip6e_len + 2) << 2;
1515 	else
1516 		elen = (ip6e.ip6e_len + 1) << 3;
1517 
1518 	MGET(n, M_DONTWAIT, MT_DATA);
1519 	if (n && elen >= MLEN) {
1520 		MCLGET(n, M_DONTWAIT);
1521 		if ((n->m_flags & M_EXT) == 0) {
1522 			m_free(n);
1523 			n = NULL;
1524 		}
1525 	}
1526 	if (!n)
1527 		return NULL;
1528 
1529 	n->m_len = 0;
1530 	if (elen >= M_TRAILINGSPACE(n)) {
1531 		m_free(n);
1532 		return NULL;
1533 	}
1534 
1535 	m_copydata(m, off, elen, mtod(n, caddr_t));
1536 	n->m_len = elen;
1537 	return n;
1538 }
1539 #endif
1540 
1541 /*
1542  * Get pointer to the previous header followed by the header
1543  * currently processed.
1544  * XXX: This function supposes that
1545  *	M includes all headers,
1546  *	the next header field and the header length field of each header
1547  *	are valid, and
1548  *	the sum of each header length equals to OFF.
1549  * Because of these assumptions, this function must be called very
1550  * carefully. Moreover, it will not be used in the near future when
1551  * we develop `neater' mechanism to process extension headers.
1552  */
1553 char *
1554 ip6_get_prevhdr(struct mbuf *m, int off)
1555 {
1556 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1557 
1558 	if (off == sizeof(struct ip6_hdr))
1559 		return (&ip6->ip6_nxt);
1560 	else {
1561 		int len, nxt;
1562 		struct ip6_ext *ip6e = NULL;
1563 
1564 		nxt = ip6->ip6_nxt;
1565 		len = sizeof(struct ip6_hdr);
1566 		while (len < off) {
1567 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1568 
1569 			switch (nxt) {
1570 			case IPPROTO_FRAGMENT:
1571 				len += sizeof(struct ip6_frag);
1572 				break;
1573 			case IPPROTO_AH:
1574 				len += (ip6e->ip6e_len + 2) << 2;
1575 				break;
1576 			default:
1577 				len += (ip6e->ip6e_len + 1) << 3;
1578 				break;
1579 			}
1580 			nxt = ip6e->ip6e_nxt;
1581 		}
1582 		if (ip6e)
1583 			return (&ip6e->ip6e_nxt);
1584 		else
1585 			return NULL;
1586 	}
1587 }
1588 
1589 /*
1590  * get next header offset.  m will be retained.
1591  */
1592 int
1593 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1594 {
1595 	struct ip6_hdr ip6;
1596 	struct ip6_ext ip6e;
1597 	struct ip6_frag fh;
1598 
1599 	/* just in case */
1600 	if (m == NULL)
1601 		panic("ip6_nexthdr: m == NULL");
1602 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1603 		return -1;
1604 
1605 	switch (proto) {
1606 	case IPPROTO_IPV6:
1607 		if (m->m_pkthdr.len < off + sizeof(ip6))
1608 			return -1;
1609 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1610 		if (nxtp)
1611 			*nxtp = ip6.ip6_nxt;
1612 		off += sizeof(ip6);
1613 		return off;
1614 
1615 	case IPPROTO_FRAGMENT:
1616 		/*
1617 		 * terminate parsing if it is not the first fragment,
1618 		 * it does not make sense to parse through it.
1619 		 */
1620 		if (m->m_pkthdr.len < off + sizeof(fh))
1621 			return -1;
1622 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1623 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1624 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1625 			return -1;
1626 		if (nxtp)
1627 			*nxtp = fh.ip6f_nxt;
1628 		off += sizeof(struct ip6_frag);
1629 		return off;
1630 
1631 	case IPPROTO_AH:
1632 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1633 			return -1;
1634 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1635 		if (nxtp)
1636 			*nxtp = ip6e.ip6e_nxt;
1637 		off += (ip6e.ip6e_len + 2) << 2;
1638 		return off;
1639 
1640 	case IPPROTO_HOPOPTS:
1641 	case IPPROTO_ROUTING:
1642 	case IPPROTO_DSTOPTS:
1643 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1644 			return -1;
1645 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1646 		if (nxtp)
1647 			*nxtp = ip6e.ip6e_nxt;
1648 		off += (ip6e.ip6e_len + 1) << 3;
1649 		return off;
1650 
1651 	case IPPROTO_NONE:
1652 	case IPPROTO_ESP:
1653 	case IPPROTO_IPCOMP:
1654 		/* give up */
1655 		return -1;
1656 
1657 	default:
1658 		return -1;
1659 	}
1660 
1661 	return -1;
1662 }
1663 
1664 /*
1665  * get offset for the last header in the chain.  m will be kept untainted.
1666  */
1667 int
1668 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1669 {
1670 	int newoff;
1671 	int nxt;
1672 
1673 	if (!nxtp) {
1674 		nxt = -1;
1675 		nxtp = &nxt;
1676 	}
1677 	while (1) {
1678 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1679 		if (newoff < 0)
1680 			return off;
1681 		else if (newoff < off)
1682 			return -1;	/* invalid */
1683 		else if (newoff == off)
1684 			return newoff;
1685 
1686 		off = newoff;
1687 		proto = *nxtp;
1688 	}
1689 }
1690 
1691 struct ip6aux *
1692 ip6_addaux(struct mbuf *m)
1693 {
1694 	struct m_tag *mtag;
1695 
1696 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1697 	if (!mtag) {
1698 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1699 		    M_NOWAIT);
1700 		if (mtag) {
1701 			m_tag_prepend(m, mtag);
1702 			bzero(mtag + 1, sizeof(struct ip6aux));
1703 		}
1704 	}
1705 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1706 }
1707 
1708 struct ip6aux *
1709 ip6_findaux(struct mbuf *m)
1710 {
1711 	struct m_tag *mtag;
1712 
1713 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1714 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1715 }
1716 
1717 void
1718 ip6_delaux(struct mbuf *m)
1719 {
1720 	struct m_tag *mtag;
1721 
1722 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1723 	if (mtag)
1724 		m_tag_delete(m, mtag);
1725 }
1726 
1727 /*
1728  * System control for IP6
1729  */
1730 
1731 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1732 	0,		0,		0,		0,
1733 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1734 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1735 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1736 	0,		0,		0,		0,
1737 	ENOPROTOOPT
1738 };
1739