xref: /freebsd/sys/netinet6/ip6_input.c (revision 830940567b49bb0c08dfaed40418999e76616909)
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 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;	/* enable 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;	 /* "IPV6FORWARDING ? 0 : 1" is dangerous */
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 	/* Skip global initialization stuff for non-default instances. */
251 	if (!IS_DEFAULT_VNET(curvnet))
252 		return;
253 
254 #ifdef DIAGNOSTIC
255 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
256 		panic("sizeof(protosw) != sizeof(ip6protosw)");
257 #endif
258 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
259 	if (pr == NULL)
260 		panic("ip6_init");
261 
262 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
263 	for (i = 0; i < IPPROTO_MAX; i++)
264 		ip6_protox[i] = pr - inet6sw;
265 	/*
266 	 * Cycle through IP protocols and put them into the appropriate place
267 	 * in ip6_protox[].
268 	 */
269 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
270 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
271 		if (pr->pr_domain->dom_family == PF_INET6 &&
272 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
273 			/* Be careful to only index valid IP protocols. */
274 			if (pr->pr_protocol < IPPROTO_MAX)
275 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
276 		}
277 
278 	/* Initialize packet filter hooks. */
279 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
280 	inet6_pfil_hook.ph_af = AF_INET6;
281 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
282 		printf("%s: WARNING: unable to register pfil hook, "
283 			"error %d\n", __func__, i);
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(&inet6_pfil_hook))
519 		goto passin;
520 
521 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
522 		return;
523 	if (m == NULL)			/* consumed by filter */
524 		return;
525 	ip6 = mtod(m, struct ip6_hdr *);
526 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
527 
528 passin:
529 	/*
530 	 * Disambiguate address scope zones (if there is ambiguity).
531 	 * We first make sure that the original source or destination address
532 	 * is not in our internal form for scoped addresses.  Such addresses
533 	 * are not necessarily invalid spec-wise, but we cannot accept them due
534 	 * to the usage conflict.
535 	 * in6_setscope() then also checks and rejects the cases where src or
536 	 * dst are the loopback address and the receiving interface
537 	 * is not loopback.
538 	 */
539 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
540 		V_ip6stat.ip6s_badscope++; /* XXX */
541 		goto bad;
542 	}
543 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
544 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
545 		V_ip6stat.ip6s_badscope++;
546 		goto bad;
547 	}
548 
549 	/*
550 	 * Multicast check. Assume packet is for us to avoid
551 	 * prematurely taking locks.
552 	 */
553 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
554 		ours = 1;
555 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
556 		deliverifp = m->m_pkthdr.rcvif;
557 		goto hbhcheck;
558 	}
559 
560 	/*
561 	 *  Unicast check
562 	 */
563 
564 	bzero(&dst6, sizeof(dst6));
565 	dst6.sin6_family = AF_INET6;
566 	dst6.sin6_len = sizeof(struct sockaddr_in6);
567 	dst6.sin6_addr = ip6->ip6_dst;
568 	ifp = m->m_pkthdr.rcvif;
569 	IF_AFDATA_LOCK(ifp);
570 	lle = lla_lookup(LLTABLE6(ifp), 0,
571 	     (struct sockaddr *)&dst6);
572 	IF_AFDATA_UNLOCK(ifp);
573 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
574 		ours = 1;
575 		deliverifp = ifp;
576 		LLE_RUNLOCK(lle);
577 		goto hbhcheck;
578 	}
579 	if (lle != NULL)
580 		LLE_RUNLOCK(lle);
581 
582 	dst = &rin6.ro_dst;
583 	dst->sin6_len = sizeof(struct sockaddr_in6);
584 	dst->sin6_family = AF_INET6;
585 	dst->sin6_addr = ip6->ip6_dst;
586 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
587 	if (rin6.ro_rt)
588 		RT_UNLOCK(rin6.ro_rt);
589 
590 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
591 
592 	/*
593 	 * Accept the packet if the forwarding interface to the destination
594 	 * according to the routing table is the loopback interface,
595 	 * unless the associated route has a gateway.
596 	 * Note that this approach causes to accept a packet if there is a
597 	 * route to the loopback interface for the destination of the packet.
598 	 * But we think it's even useful in some situations, e.g. when using
599 	 * a special daemon which wants to intercept the packet.
600 	 *
601 	 * XXX: some OSes automatically make a cloned route for the destination
602 	 * of an outgoing packet.  If the outgoing interface of the packet
603 	 * is a loopback one, the kernel would consider the packet to be
604 	 * accepted, even if we have no such address assinged on the interface.
605 	 * We check the cloned flag of the route entry to reject such cases,
606 	 * assuming that route entries for our own addresses are not made by
607 	 * cloning (it should be true because in6_addloop explicitly installs
608 	 * the host route).  However, we might have to do an explicit check
609 	 * while it would be less efficient.  Or, should we rather install a
610 	 * reject route for such a case?
611 	 */
612 	if (rin6.ro_rt &&
613 	    (rin6.ro_rt->rt_flags &
614 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
615 #ifdef RTF_WASCLONED
616 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
617 #endif
618 #ifdef RTF_CLONED
619 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
620 #endif
621 #if 0
622 	    /*
623 	     * The check below is redundant since the comparison of
624 	     * the destination and the key of the rtentry has
625 	     * already done through looking up the routing table.
626 	     */
627 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
628 	    &rt6_key(rin6.ro_rt)->sin6_addr)
629 #endif
630 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
631 		struct in6_ifaddr *ia6 =
632 			(struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
633 
634 		/*
635 		 * record address information into m_tag.
636 		 */
637 		(void)ip6_setdstifaddr(m, ia6);
638 
639 		/*
640 		 * packets to a tentative, duplicated, or somehow invalid
641 		 * address must not be accepted.
642 		 */
643 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
644 			/* this address is ready */
645 			ours = 1;
646 			deliverifp = ia6->ia_ifp;	/* correct? */
647 			/* Count the packet in the ip address stats */
648 			ia6->ia_ifa.if_ipackets++;
649 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
650 			goto hbhcheck;
651 		} else {
652 			char ip6bufs[INET6_ADDRSTRLEN];
653 			char ip6bufd[INET6_ADDRSTRLEN];
654 			/* address is not ready, so discard the packet. */
655 			nd6log((LOG_INFO,
656 			    "ip6_input: packet to an unready address %s->%s\n",
657 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
658 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
659 
660 			goto bad;
661 		}
662 	}
663 
664 	/*
665 	 * FAITH (Firewall Aided Internet Translator)
666 	 */
667 	if (V_ip6_keepfaith) {
668 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
669 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
670 			/* XXX do we need more sanity checks? */
671 			ours = 1;
672 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
673 			goto hbhcheck;
674 		}
675 	}
676 
677 	/*
678 	 * Now there is no reason to process the packet if it's not our own
679 	 * and we're not a router.
680 	 */
681 	if (!V_ip6_forwarding) {
682 		V_ip6stat.ip6s_cantforward++;
683 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
684 		goto bad;
685 	}
686 
687   hbhcheck:
688 	/*
689 	 * record address information into m_tag, if we don't have one yet.
690 	 * note that we are unable to record it, if the address is not listed
691 	 * as our interface address (e.g. multicast addresses, addresses
692 	 * within FAITH prefixes and such).
693 	 */
694 	if (deliverifp && !ip6_getdstifaddr(m)) {
695 		struct in6_ifaddr *ia6;
696 
697 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
698 		if (ia6) {
699 			if (!ip6_setdstifaddr(m, ia6)) {
700 				/*
701 				 * XXX maybe we should drop the packet here,
702 				 * as we could not provide enough information
703 				 * to the upper layers.
704 				 */
705 			}
706 			ifa_free(&ia6->ia_ifa);
707 		}
708 	}
709 
710 	/*
711 	 * Process Hop-by-Hop options header if it's contained.
712 	 * m may be modified in ip6_hopopts_input().
713 	 * If a JumboPayload option is included, plen will also be modified.
714 	 */
715 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
716 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
717 		struct ip6_hbh *hbh;
718 
719 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
720 #if 0	/*touches NULL pointer*/
721 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
722 #endif
723 			goto out;	/* m have already been freed */
724 		}
725 
726 		/* adjust pointer */
727 		ip6 = mtod(m, struct ip6_hdr *);
728 
729 		/*
730 		 * if the payload length field is 0 and the next header field
731 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
732 		 * option MUST be included.
733 		 */
734 		if (ip6->ip6_plen == 0 && plen == 0) {
735 			/*
736 			 * Note that if a valid jumbo payload option is
737 			 * contained, ip6_hopopts_input() must set a valid
738 			 * (non-zero) payload length to the variable plen.
739 			 */
740 			V_ip6stat.ip6s_badoptions++;
741 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
742 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
743 			icmp6_error(m, ICMP6_PARAM_PROB,
744 				    ICMP6_PARAMPROB_HEADER,
745 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
746 			goto out;
747 		}
748 #ifndef PULLDOWN_TEST
749 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
750 		hbh = (struct ip6_hbh *)(ip6 + 1);
751 #else
752 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
753 			sizeof(struct ip6_hbh));
754 		if (hbh == NULL) {
755 			V_ip6stat.ip6s_tooshort++;
756 			goto out;
757 		}
758 #endif
759 		nxt = hbh->ip6h_nxt;
760 
761 		/*
762 		 * If we are acting as a router and the packet contains a
763 		 * router alert option, see if we know the option value.
764 		 * Currently, we only support the option value for MLD, in which
765 		 * case we should pass the packet to the multicast routing
766 		 * daemon.
767 		 */
768 		if (rtalert != ~0) {
769 			switch (rtalert) {
770 			case IP6OPT_RTALERT_MLD:
771 				if (V_ip6_forwarding)
772 					ours = 1;
773 				break;
774 			default:
775 				/*
776 				 * RFC2711 requires unrecognized values must be
777 				 * silently ignored.
778 				 */
779 				break;
780 			}
781 		}
782 	} else
783 		nxt = ip6->ip6_nxt;
784 
785 	/*
786 	 * Check that the amount of data in the buffers
787 	 * is as at least much as the IPv6 header would have us expect.
788 	 * Trim mbufs if longer than we expect.
789 	 * Drop packet if shorter than we expect.
790 	 */
791 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
792 		V_ip6stat.ip6s_tooshort++;
793 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
794 		goto bad;
795 	}
796 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
797 		if (m->m_len == m->m_pkthdr.len) {
798 			m->m_len = sizeof(struct ip6_hdr) + plen;
799 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
800 		} else
801 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
802 	}
803 
804 	/*
805 	 * Forward if desirable.
806 	 */
807 	if (V_ip6_mrouter &&
808 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
809 		/*
810 		 * If we are acting as a multicast router, all
811 		 * incoming multicast packets are passed to the
812 		 * kernel-level multicast forwarding function.
813 		 * The packet is returned (relatively) intact; if
814 		 * ip6_mforward() returns a non-zero value, the packet
815 		 * must be discarded, else it may be accepted below.
816 		 *
817 		 * XXX TODO: Check hlim and multicast scope here to avoid
818 		 * unnecessarily calling into ip6_mforward().
819 		 */
820 		if (ip6_mforward &&
821 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
822 			IP6STAT_INC(ip6s_cantforward);
823 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
824 			goto bad;
825 		}
826 	} else if (!ours) {
827 		ip6_forward(m, srcrt);
828 		goto out;
829 	}
830 
831 	ip6 = mtod(m, struct ip6_hdr *);
832 
833 	/*
834 	 * Malicious party may be able to use IPv4 mapped addr to confuse
835 	 * tcp/udp stack and bypass security checks (act as if it was from
836 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
837 	 *
838 	 * For SIIT end node behavior, you may want to disable the check.
839 	 * However, you will  become vulnerable to attacks using IPv4 mapped
840 	 * source.
841 	 */
842 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
843 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
844 		V_ip6stat.ip6s_badscope++;
845 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
846 		goto bad;
847 	}
848 
849 	/*
850 	 * Tell launch routine the next header
851 	 */
852 	V_ip6stat.ip6s_delivered++;
853 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
854 	nest = 0;
855 
856 	while (nxt != IPPROTO_DONE) {
857 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
858 			V_ip6stat.ip6s_toomanyhdr++;
859 			goto bad;
860 		}
861 
862 		/*
863 		 * protection against faulty packet - there should be
864 		 * more sanity checks in header chain processing.
865 		 */
866 		if (m->m_pkthdr.len < off) {
867 			V_ip6stat.ip6s_tooshort++;
868 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
869 			goto bad;
870 		}
871 
872 #ifdef IPSEC
873 		/*
874 		 * enforce IPsec policy checking if we are seeing last header.
875 		 * note that we do not visit this with protocols with pcb layer
876 		 * code - like udp/tcp/raw ip.
877 		 */
878 		if (ip6_ipsec_input(m, nxt))
879 			goto bad;
880 #endif /* IPSEC */
881 
882 		/*
883 		 * Use mbuf flags to propagate Router Alert option to
884 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
885 		 */
886 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
887 			m->m_flags |= M_RTALERT_MLD;
888 
889 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
890 	}
891 	goto out;
892 bad:
893 	m_freem(m);
894 out:
895 	if (rin6.ro_rt)
896 		RTFREE(rin6.ro_rt);
897 }
898 
899 /*
900  * set/grab in6_ifaddr correspond to IPv6 destination address.
901  * XXX backward compatibility wrapper
902  *
903  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
904  * and then bump it when the tag is copied, and release it when the tag is
905  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
906  * we just bump the ia refcount when we receive it.  This should be fixed.
907  */
908 static struct ip6aux *
909 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
910 {
911 	struct ip6aux *ip6a;
912 
913 	ip6a = ip6_addaux(m);
914 	if (ip6a)
915 		ip6a->ip6a_dstia6 = ia6;
916 	return ip6a;	/* NULL if failed to set */
917 }
918 
919 struct in6_ifaddr *
920 ip6_getdstifaddr(struct mbuf *m)
921 {
922 	struct ip6aux *ip6a;
923 	struct in6_ifaddr *ia;
924 
925 	ip6a = ip6_findaux(m);
926 	if (ip6a) {
927 		ia = ip6a->ip6a_dstia6;
928 		ifa_ref(&ia->ia_ifa);
929 		return ia;
930 	} else
931 		return NULL;
932 }
933 
934 /*
935  * Hop-by-Hop options header processing. If a valid jumbo payload option is
936  * included, the real payload length will be stored in plenp.
937  *
938  * rtalertp - XXX: should be stored more smart way
939  */
940 static int
941 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
942     struct mbuf **mp, int *offp)
943 {
944 	struct mbuf *m = *mp;
945 	int off = *offp, hbhlen;
946 	struct ip6_hbh *hbh;
947 	u_int8_t *opt;
948 
949 	/* validation of the length of the header */
950 #ifndef PULLDOWN_TEST
951 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
952 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
953 	hbhlen = (hbh->ip6h_len + 1) << 3;
954 
955 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
956 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
957 #else
958 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
959 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
960 	if (hbh == NULL) {
961 		V_ip6stat.ip6s_tooshort++;
962 		return -1;
963 	}
964 	hbhlen = (hbh->ip6h_len + 1) << 3;
965 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
966 		hbhlen);
967 	if (hbh == NULL) {
968 		V_ip6stat.ip6s_tooshort++;
969 		return -1;
970 	}
971 #endif
972 	off += hbhlen;
973 	hbhlen -= sizeof(struct ip6_hbh);
974 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
975 
976 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
977 				hbhlen, rtalertp, plenp) < 0)
978 		return (-1);
979 
980 	*offp = off;
981 	*mp = m;
982 	return (0);
983 }
984 
985 /*
986  * Search header for all Hop-by-hop options and process each option.
987  * This function is separate from ip6_hopopts_input() in order to
988  * handle a case where the sending node itself process its hop-by-hop
989  * options header. In such a case, the function is called from ip6_output().
990  *
991  * The function assumes that hbh header is located right after the IPv6 header
992  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
993  * opthead + hbhlen is located in continuous memory region.
994  */
995 int
996 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
997     u_int32_t *rtalertp, u_int32_t *plenp)
998 {
999 	struct ip6_hdr *ip6;
1000 	int optlen = 0;
1001 	u_int8_t *opt = opthead;
1002 	u_int16_t rtalert_val;
1003 	u_int32_t jumboplen;
1004 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1005 
1006 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1007 		switch (*opt) {
1008 		case IP6OPT_PAD1:
1009 			optlen = 1;
1010 			break;
1011 		case IP6OPT_PADN:
1012 			if (hbhlen < IP6OPT_MINLEN) {
1013 				V_ip6stat.ip6s_toosmall++;
1014 				goto bad;
1015 			}
1016 			optlen = *(opt + 1) + 2;
1017 			break;
1018 		case IP6OPT_ROUTER_ALERT:
1019 			/* XXX may need check for alignment */
1020 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1021 				V_ip6stat.ip6s_toosmall++;
1022 				goto bad;
1023 			}
1024 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1025 				/* XXX stat */
1026 				icmp6_error(m, ICMP6_PARAM_PROB,
1027 				    ICMP6_PARAMPROB_HEADER,
1028 				    erroff + opt + 1 - opthead);
1029 				return (-1);
1030 			}
1031 			optlen = IP6OPT_RTALERT_LEN;
1032 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1033 			*rtalertp = ntohs(rtalert_val);
1034 			break;
1035 		case IP6OPT_JUMBO:
1036 			/* XXX may need check for alignment */
1037 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1038 				V_ip6stat.ip6s_toosmall++;
1039 				goto bad;
1040 			}
1041 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1042 				/* XXX stat */
1043 				icmp6_error(m, ICMP6_PARAM_PROB,
1044 				    ICMP6_PARAMPROB_HEADER,
1045 				    erroff + opt + 1 - opthead);
1046 				return (-1);
1047 			}
1048 			optlen = IP6OPT_JUMBO_LEN;
1049 
1050 			/*
1051 			 * IPv6 packets that have non 0 payload length
1052 			 * must not contain a jumbo payload option.
1053 			 */
1054 			ip6 = mtod(m, struct ip6_hdr *);
1055 			if (ip6->ip6_plen) {
1056 				V_ip6stat.ip6s_badoptions++;
1057 				icmp6_error(m, ICMP6_PARAM_PROB,
1058 				    ICMP6_PARAMPROB_HEADER,
1059 				    erroff + opt - opthead);
1060 				return (-1);
1061 			}
1062 
1063 			/*
1064 			 * We may see jumbolen in unaligned location, so
1065 			 * we'd need to perform bcopy().
1066 			 */
1067 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1068 			jumboplen = (u_int32_t)htonl(jumboplen);
1069 
1070 #if 1
1071 			/*
1072 			 * if there are multiple jumbo payload options,
1073 			 * *plenp will be non-zero and the packet will be
1074 			 * rejected.
1075 			 * the behavior may need some debate in ipngwg -
1076 			 * multiple options does not make sense, however,
1077 			 * there's no explicit mention in specification.
1078 			 */
1079 			if (*plenp != 0) {
1080 				V_ip6stat.ip6s_badoptions++;
1081 				icmp6_error(m, ICMP6_PARAM_PROB,
1082 				    ICMP6_PARAMPROB_HEADER,
1083 				    erroff + opt + 2 - opthead);
1084 				return (-1);
1085 			}
1086 #endif
1087 
1088 			/*
1089 			 * jumbo payload length must be larger than 65535.
1090 			 */
1091 			if (jumboplen <= IPV6_MAXPACKET) {
1092 				V_ip6stat.ip6s_badoptions++;
1093 				icmp6_error(m, ICMP6_PARAM_PROB,
1094 				    ICMP6_PARAMPROB_HEADER,
1095 				    erroff + opt + 2 - opthead);
1096 				return (-1);
1097 			}
1098 			*plenp = jumboplen;
1099 
1100 			break;
1101 		default:		/* unknown option */
1102 			if (hbhlen < IP6OPT_MINLEN) {
1103 				V_ip6stat.ip6s_toosmall++;
1104 				goto bad;
1105 			}
1106 			optlen = ip6_unknown_opt(opt, m,
1107 			    erroff + opt - opthead);
1108 			if (optlen == -1)
1109 				return (-1);
1110 			optlen += 2;
1111 			break;
1112 		}
1113 	}
1114 
1115 	return (0);
1116 
1117   bad:
1118 	m_freem(m);
1119 	return (-1);
1120 }
1121 
1122 /*
1123  * Unknown option processing.
1124  * The third argument `off' is the offset from the IPv6 header to the option,
1125  * which is necessary if the IPv6 header the and option header and IPv6 header
1126  * is not continuous in order to return an ICMPv6 error.
1127  */
1128 int
1129 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1130 {
1131 	struct ip6_hdr *ip6;
1132 
1133 	switch (IP6OPT_TYPE(*optp)) {
1134 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1135 		return ((int)*(optp + 1));
1136 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1137 		m_freem(m);
1138 		return (-1);
1139 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1140 		V_ip6stat.ip6s_badoptions++;
1141 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1142 		return (-1);
1143 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1144 		V_ip6stat.ip6s_badoptions++;
1145 		ip6 = mtod(m, struct ip6_hdr *);
1146 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1147 		    (m->m_flags & (M_BCAST|M_MCAST)))
1148 			m_freem(m);
1149 		else
1150 			icmp6_error(m, ICMP6_PARAM_PROB,
1151 				    ICMP6_PARAMPROB_OPTION, off);
1152 		return (-1);
1153 	}
1154 
1155 	m_freem(m);		/* XXX: NOTREACHED */
1156 	return (-1);
1157 }
1158 
1159 /*
1160  * Create the "control" list for this pcb.
1161  * These functions will not modify mbuf chain at all.
1162  *
1163  * With KAME mbuf chain restriction:
1164  * The routine will be called from upper layer handlers like tcp6_input().
1165  * Thus the routine assumes that the caller (tcp6_input) have already
1166  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1167  * very first mbuf on the mbuf chain.
1168  *
1169  * ip6_savecontrol_v4 will handle those options that are possible to be
1170  * set on a v4-mapped socket.
1171  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1172  * options and handle the v6-only ones itself.
1173  */
1174 struct mbuf **
1175 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1176     int *v4only)
1177 {
1178 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1179 
1180 #ifdef SO_TIMESTAMP
1181 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1182 		struct timeval tv;
1183 
1184 		microtime(&tv);
1185 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1186 		    SCM_TIMESTAMP, SOL_SOCKET);
1187 		if (*mp)
1188 			mp = &(*mp)->m_next;
1189 	}
1190 #endif
1191 
1192 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1193 		if (v4only != NULL)
1194 			*v4only = 1;
1195 		return (mp);
1196 	}
1197 
1198 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1199 	/* RFC 2292 sec. 5 */
1200 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1201 		struct in6_pktinfo pi6;
1202 
1203 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1204 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1205 		pi6.ipi6_ifindex =
1206 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1207 
1208 		*mp = sbcreatecontrol((caddr_t) &pi6,
1209 		    sizeof(struct in6_pktinfo),
1210 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1211 		if (*mp)
1212 			mp = &(*mp)->m_next;
1213 	}
1214 
1215 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1216 		int hlim = ip6->ip6_hlim & 0xff;
1217 
1218 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1219 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1220 		    IPPROTO_IPV6);
1221 		if (*mp)
1222 			mp = &(*mp)->m_next;
1223 	}
1224 
1225 	if (v4only != NULL)
1226 		*v4only = 0;
1227 	return (mp);
1228 }
1229 
1230 void
1231 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1232 {
1233 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1234 	int v4only = 0;
1235 
1236 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1237 	if (v4only)
1238 		return;
1239 
1240 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1241 		u_int32_t flowinfo;
1242 		int tclass;
1243 
1244 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1245 		flowinfo >>= 20;
1246 
1247 		tclass = flowinfo & 0xff;
1248 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1249 		    IPV6_TCLASS, IPPROTO_IPV6);
1250 		if (*mp)
1251 			mp = &(*mp)->m_next;
1252 	}
1253 
1254 	/*
1255 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1256 	 * privilege for the option (see ip6_ctloutput), but it might be too
1257 	 * strict, since there might be some hop-by-hop options which can be
1258 	 * returned to normal user.
1259 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1260 	 */
1261 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1262 		/*
1263 		 * Check if a hop-by-hop options header is contatined in the
1264 		 * received packet, and if so, store the options as ancillary
1265 		 * data. Note that a hop-by-hop options header must be
1266 		 * just after the IPv6 header, which is assured through the
1267 		 * IPv6 input processing.
1268 		 */
1269 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1270 			struct ip6_hbh *hbh;
1271 			int hbhlen = 0;
1272 #ifdef PULLDOWN_TEST
1273 			struct mbuf *ext;
1274 #endif
1275 
1276 #ifndef PULLDOWN_TEST
1277 			hbh = (struct ip6_hbh *)(ip6 + 1);
1278 			hbhlen = (hbh->ip6h_len + 1) << 3;
1279 #else
1280 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1281 			    ip6->ip6_nxt);
1282 			if (ext == NULL) {
1283 				V_ip6stat.ip6s_tooshort++;
1284 				return;
1285 			}
1286 			hbh = mtod(ext, struct ip6_hbh *);
1287 			hbhlen = (hbh->ip6h_len + 1) << 3;
1288 			if (hbhlen != ext->m_len) {
1289 				m_freem(ext);
1290 				V_ip6stat.ip6s_tooshort++;
1291 				return;
1292 			}
1293 #endif
1294 
1295 			/*
1296 			 * XXX: We copy the whole header even if a
1297 			 * jumbo payload option is included, the option which
1298 			 * is to be removed before returning according to
1299 			 * RFC2292.
1300 			 * Note: this constraint is removed in RFC3542
1301 			 */
1302 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1303 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1304 			    IPPROTO_IPV6);
1305 			if (*mp)
1306 				mp = &(*mp)->m_next;
1307 #ifdef PULLDOWN_TEST
1308 			m_freem(ext);
1309 #endif
1310 		}
1311 	}
1312 
1313 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1314 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1315 
1316 		/*
1317 		 * Search for destination options headers or routing
1318 		 * header(s) through the header chain, and stores each
1319 		 * header as ancillary data.
1320 		 * Note that the order of the headers remains in
1321 		 * the chain of ancillary data.
1322 		 */
1323 		while (1) {	/* is explicit loop prevention necessary? */
1324 			struct ip6_ext *ip6e = NULL;
1325 			int elen;
1326 #ifdef PULLDOWN_TEST
1327 			struct mbuf *ext = NULL;
1328 #endif
1329 
1330 			/*
1331 			 * if it is not an extension header, don't try to
1332 			 * pull it from the chain.
1333 			 */
1334 			switch (nxt) {
1335 			case IPPROTO_DSTOPTS:
1336 			case IPPROTO_ROUTING:
1337 			case IPPROTO_HOPOPTS:
1338 			case IPPROTO_AH: /* is it possible? */
1339 				break;
1340 			default:
1341 				goto loopend;
1342 			}
1343 
1344 #ifndef PULLDOWN_TEST
1345 			if (off + sizeof(*ip6e) > m->m_len)
1346 				goto loopend;
1347 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1348 			if (nxt == IPPROTO_AH)
1349 				elen = (ip6e->ip6e_len + 2) << 2;
1350 			else
1351 				elen = (ip6e->ip6e_len + 1) << 3;
1352 			if (off + elen > m->m_len)
1353 				goto loopend;
1354 #else
1355 			ext = ip6_pullexthdr(m, off, nxt);
1356 			if (ext == NULL) {
1357 				V_ip6stat.ip6s_tooshort++;
1358 				return;
1359 			}
1360 			ip6e = mtod(ext, struct ip6_ext *);
1361 			if (nxt == IPPROTO_AH)
1362 				elen = (ip6e->ip6e_len + 2) << 2;
1363 			else
1364 				elen = (ip6e->ip6e_len + 1) << 3;
1365 			if (elen != ext->m_len) {
1366 				m_freem(ext);
1367 				V_ip6stat.ip6s_tooshort++;
1368 				return;
1369 			}
1370 #endif
1371 
1372 			switch (nxt) {
1373 			case IPPROTO_DSTOPTS:
1374 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1375 					break;
1376 
1377 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1378 				    IS2292(in6p,
1379 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1380 				    IPPROTO_IPV6);
1381 				if (*mp)
1382 					mp = &(*mp)->m_next;
1383 				break;
1384 			case IPPROTO_ROUTING:
1385 				if (!in6p->inp_flags & IN6P_RTHDR)
1386 					break;
1387 
1388 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1389 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1390 				    IPPROTO_IPV6);
1391 				if (*mp)
1392 					mp = &(*mp)->m_next;
1393 				break;
1394 			case IPPROTO_HOPOPTS:
1395 			case IPPROTO_AH: /* is it possible? */
1396 				break;
1397 
1398 			default:
1399 				/*
1400 				 * other cases have been filtered in the above.
1401 				 * none will visit this case.  here we supply
1402 				 * the code just in case (nxt overwritten or
1403 				 * other cases).
1404 				 */
1405 #ifdef PULLDOWN_TEST
1406 				m_freem(ext);
1407 #endif
1408 				goto loopend;
1409 
1410 			}
1411 
1412 			/* proceed with the next header. */
1413 			off += elen;
1414 			nxt = ip6e->ip6e_nxt;
1415 			ip6e = NULL;
1416 #ifdef PULLDOWN_TEST
1417 			m_freem(ext);
1418 			ext = NULL;
1419 #endif
1420 		}
1421 	  loopend:
1422 		;
1423 	}
1424 }
1425 #undef IS2292
1426 
1427 void
1428 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1429 {
1430 	struct socket *so;
1431 	struct mbuf *m_mtu;
1432 	struct ip6_mtuinfo mtuctl;
1433 
1434 	so =  in6p->inp_socket;
1435 
1436 	if (mtu == NULL)
1437 		return;
1438 
1439 #ifdef DIAGNOSTIC
1440 	if (so == NULL)		/* I believe this is impossible */
1441 		panic("ip6_notify_pmtu: socket is NULL");
1442 #endif
1443 
1444 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1445 	mtuctl.ip6m_mtu = *mtu;
1446 	mtuctl.ip6m_addr = *dst;
1447 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1448 		return;
1449 
1450 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1451 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1452 		return;
1453 
1454 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1455 	    == 0) {
1456 		m_freem(m_mtu);
1457 		/* XXX: should count statistics */
1458 	} else
1459 		sorwakeup(so);
1460 
1461 	return;
1462 }
1463 
1464 #ifdef PULLDOWN_TEST
1465 /*
1466  * pull single extension header from mbuf chain.  returns single mbuf that
1467  * contains the result, or NULL on error.
1468  */
1469 static struct mbuf *
1470 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1471 {
1472 	struct ip6_ext ip6e;
1473 	size_t elen;
1474 	struct mbuf *n;
1475 
1476 #ifdef DIAGNOSTIC
1477 	switch (nxt) {
1478 	case IPPROTO_DSTOPTS:
1479 	case IPPROTO_ROUTING:
1480 	case IPPROTO_HOPOPTS:
1481 	case IPPROTO_AH: /* is it possible? */
1482 		break;
1483 	default:
1484 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1485 	}
1486 #endif
1487 
1488 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1489 	if (nxt == IPPROTO_AH)
1490 		elen = (ip6e.ip6e_len + 2) << 2;
1491 	else
1492 		elen = (ip6e.ip6e_len + 1) << 3;
1493 
1494 	MGET(n, M_DONTWAIT, MT_DATA);
1495 	if (n && elen >= MLEN) {
1496 		MCLGET(n, M_DONTWAIT);
1497 		if ((n->m_flags & M_EXT) == 0) {
1498 			m_free(n);
1499 			n = NULL;
1500 		}
1501 	}
1502 	if (!n)
1503 		return NULL;
1504 
1505 	n->m_len = 0;
1506 	if (elen >= M_TRAILINGSPACE(n)) {
1507 		m_free(n);
1508 		return NULL;
1509 	}
1510 
1511 	m_copydata(m, off, elen, mtod(n, caddr_t));
1512 	n->m_len = elen;
1513 	return n;
1514 }
1515 #endif
1516 
1517 /*
1518  * Get pointer to the previous header followed by the header
1519  * currently processed.
1520  * XXX: This function supposes that
1521  *	M includes all headers,
1522  *	the next header field and the header length field of each header
1523  *	are valid, and
1524  *	the sum of each header length equals to OFF.
1525  * Because of these assumptions, this function must be called very
1526  * carefully. Moreover, it will not be used in the near future when
1527  * we develop `neater' mechanism to process extension headers.
1528  */
1529 char *
1530 ip6_get_prevhdr(struct mbuf *m, int off)
1531 {
1532 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1533 
1534 	if (off == sizeof(struct ip6_hdr))
1535 		return (&ip6->ip6_nxt);
1536 	else {
1537 		int len, nxt;
1538 		struct ip6_ext *ip6e = NULL;
1539 
1540 		nxt = ip6->ip6_nxt;
1541 		len = sizeof(struct ip6_hdr);
1542 		while (len < off) {
1543 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1544 
1545 			switch (nxt) {
1546 			case IPPROTO_FRAGMENT:
1547 				len += sizeof(struct ip6_frag);
1548 				break;
1549 			case IPPROTO_AH:
1550 				len += (ip6e->ip6e_len + 2) << 2;
1551 				break;
1552 			default:
1553 				len += (ip6e->ip6e_len + 1) << 3;
1554 				break;
1555 			}
1556 			nxt = ip6e->ip6e_nxt;
1557 		}
1558 		if (ip6e)
1559 			return (&ip6e->ip6e_nxt);
1560 		else
1561 			return NULL;
1562 	}
1563 }
1564 
1565 /*
1566  * get next header offset.  m will be retained.
1567  */
1568 int
1569 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1570 {
1571 	struct ip6_hdr ip6;
1572 	struct ip6_ext ip6e;
1573 	struct ip6_frag fh;
1574 
1575 	/* just in case */
1576 	if (m == NULL)
1577 		panic("ip6_nexthdr: m == NULL");
1578 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1579 		return -1;
1580 
1581 	switch (proto) {
1582 	case IPPROTO_IPV6:
1583 		if (m->m_pkthdr.len < off + sizeof(ip6))
1584 			return -1;
1585 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1586 		if (nxtp)
1587 			*nxtp = ip6.ip6_nxt;
1588 		off += sizeof(ip6);
1589 		return off;
1590 
1591 	case IPPROTO_FRAGMENT:
1592 		/*
1593 		 * terminate parsing if it is not the first fragment,
1594 		 * it does not make sense to parse through it.
1595 		 */
1596 		if (m->m_pkthdr.len < off + sizeof(fh))
1597 			return -1;
1598 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1599 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1600 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1601 			return -1;
1602 		if (nxtp)
1603 			*nxtp = fh.ip6f_nxt;
1604 		off += sizeof(struct ip6_frag);
1605 		return off;
1606 
1607 	case IPPROTO_AH:
1608 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1609 			return -1;
1610 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1611 		if (nxtp)
1612 			*nxtp = ip6e.ip6e_nxt;
1613 		off += (ip6e.ip6e_len + 2) << 2;
1614 		return off;
1615 
1616 	case IPPROTO_HOPOPTS:
1617 	case IPPROTO_ROUTING:
1618 	case IPPROTO_DSTOPTS:
1619 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1620 			return -1;
1621 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1622 		if (nxtp)
1623 			*nxtp = ip6e.ip6e_nxt;
1624 		off += (ip6e.ip6e_len + 1) << 3;
1625 		return off;
1626 
1627 	case IPPROTO_NONE:
1628 	case IPPROTO_ESP:
1629 	case IPPROTO_IPCOMP:
1630 		/* give up */
1631 		return -1;
1632 
1633 	default:
1634 		return -1;
1635 	}
1636 
1637 	return -1;
1638 }
1639 
1640 /*
1641  * get offset for the last header in the chain.  m will be kept untainted.
1642  */
1643 int
1644 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1645 {
1646 	int newoff;
1647 	int nxt;
1648 
1649 	if (!nxtp) {
1650 		nxt = -1;
1651 		nxtp = &nxt;
1652 	}
1653 	while (1) {
1654 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1655 		if (newoff < 0)
1656 			return off;
1657 		else if (newoff < off)
1658 			return -1;	/* invalid */
1659 		else if (newoff == off)
1660 			return newoff;
1661 
1662 		off = newoff;
1663 		proto = *nxtp;
1664 	}
1665 }
1666 
1667 struct ip6aux *
1668 ip6_addaux(struct mbuf *m)
1669 {
1670 	struct m_tag *mtag;
1671 
1672 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1673 	if (!mtag) {
1674 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1675 		    M_NOWAIT);
1676 		if (mtag) {
1677 			m_tag_prepend(m, mtag);
1678 			bzero(mtag + 1, sizeof(struct ip6aux));
1679 		}
1680 	}
1681 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1682 }
1683 
1684 struct ip6aux *
1685 ip6_findaux(struct mbuf *m)
1686 {
1687 	struct m_tag *mtag;
1688 
1689 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1690 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1691 }
1692 
1693 void
1694 ip6_delaux(struct mbuf *m)
1695 {
1696 	struct m_tag *mtag;
1697 
1698 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1699 	if (mtag)
1700 		m_tag_delete(m, mtag);
1701 }
1702 
1703 /*
1704  * System control for IP6
1705  */
1706 
1707 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1708 	0,		0,		0,		0,
1709 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1710 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1711 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1712 	0,		0,		0,		0,
1713 	ENOPROTOOPT
1714 };
1715