xref: /freebsd/sys/netinet6/ip6_input.c (revision c0020399a650364d0134f79f3fa319f84064372d)
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 #include "opt_route.h"
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/proc.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/vimage.h>
85 
86 #include <net/if.h>
87 #include <net/if_types.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/netisr.h>
91 #include <net/pfil.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_systm.h>
96 #include <net/if_llatbl.h>
97 #ifdef INET
98 #include <netinet/ip.h>
99 #include <netinet/ip_icmp.h>
100 #include <netinet/vinet.h>
101 #endif /* INET */
102 #include <netinet/ip6.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/vinet6.h>
111 
112 #ifdef IPSEC
113 #include <netipsec/ipsec.h>
114 #include <netinet6/ip6_ipsec.h>
115 #include <netipsec/ipsec6.h>
116 #endif /* IPSEC */
117 
118 #include <netinet6/ip6protosw.h>
119 
120 extern struct domain inet6domain;
121 
122 u_char ip6_protox[IPPROTO_MAX];
123 static struct ifqueue ip6intrq;
124 
125 #ifndef VIMAGE
126 #ifndef VIMAGE_GLOBALS
127 struct vnet_inet6 vnet_inet6_0;
128 #endif
129 #endif
130 
131 #ifdef VIMAGE_GLOBALS
132 static int ip6qmaxlen;
133 struct in6_ifaddr *in6_ifaddr;
134 struct ip6stat ip6stat;
135 
136 extern struct callout in6_tmpaddrtimer_ch;
137 
138 extern int dad_init;
139 extern int pmtu_expire;
140 extern int pmtu_probe;
141 extern u_long rip6_sendspace;
142 extern u_long rip6_recvspace;
143 extern int icmp6errppslim;
144 extern int icmp6_nodeinfo;
145 extern int udp6_sendspace;
146 extern int udp6_recvspace;
147 #endif
148 
149 struct pfil_head inet6_pfil_hook;
150 
151 static void ip6_init2(void *);
152 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
153 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
154 #ifdef PULLDOWN_TEST
155 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
156 #endif
157 
158 #ifndef VIMAGE_GLOBALS
159 static void vnet_inet6_register(void);
160 
161 static const vnet_modinfo_t vnet_inet6_modinfo = {
162 	.vmi_id		= VNET_MOD_INET6,
163 	.vmi_name	= "inet6",
164 	.vmi_dependson	= VNET_MOD_INET	/* XXX revisit - TCP/UDP needs this? */
165 };
166 
167 static void
168 vnet_inet6_register(void)
169 {
170 
171 	vnet_mod_register(&vnet_inet6_modinfo);
172 }
173 
174 SYSINIT(inet6, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet6_register, 0);
175 #endif
176 
177 /*
178  * IP6 initialization: fill in IP6 protocol switch table.
179  * All protocols not implemented in kernel go to raw IP6 protocol handler.
180  */
181 void
182 ip6_init(void)
183 {
184 	INIT_VNET_INET6(curvnet);
185 	struct ip6protosw *pr;
186 	int i;
187 
188 	V_ip6qmaxlen = IFQ_MAXLEN;
189 	V_in6_maxmtu = 0;
190 #ifdef IP6_AUTO_LINKLOCAL
191 	V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL;
192 #else
193 	V_ip6_auto_linklocal = 1;	/* enable by default */
194 #endif
195 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
196 	    &V_ip6_auto_linklocal);
197 
198 #ifndef IPV6FORWARDING
199 #ifdef GATEWAY6
200 #define IPV6FORWARDING	1	/* forward IP6 packets not for us */
201 #else
202 #define IPV6FORWARDING	0	/* don't forward IP6 packets not for us */
203 #endif /* GATEWAY6 */
204 #endif /* !IPV6FORWARDING */
205 
206 #ifndef IPV6_SENDREDIRECTS
207 #define IPV6_SENDREDIRECTS	1
208 #endif
209 
210 	V_ip6_forwarding = IPV6FORWARDING; /* act as router? */
211 	V_ip6_sendredirects = IPV6_SENDREDIRECTS;
212 	V_ip6_defhlim = IPV6_DEFHLIM;
213 	V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS;
214 	V_ip6_accept_rtadv = 0;	 /* "IPV6FORWARDING ? 0 : 1" is dangerous */
215 	V_ip6_log_interval = 5;
216 	V_ip6_hdrnestlimit = 15; /* How many header options will we process? */
217 	V_ip6_dad_count = 1;	 /* DupAddrDetectionTransmits */
218 	V_ip6_auto_flowlabel = 1;
219 	V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */
220 	V_ip6_rr_prune = 5;	 /* router renumbering prefix
221                                   * walk list every 5 sec. */
222 	V_ip6_mcast_pmtu = 0;	 /* enable pMTU discovery for multicast? */
223 	V_ip6_v6only = 1;
224 	V_ip6_keepfaith = 0;
225 	V_ip6_log_time = (time_t)0L;
226 #ifdef IPSTEALTH
227 	V_ip6stealth = 0;
228 #endif
229 	V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */
230 
231 	V_pmtu_expire = 60*10;
232 	V_pmtu_probe = 60*2;
233 
234 	/* raw IP6 parameters */
235 	/*
236 	 * Nominal space allocated to a raw ip socket.
237 	 */
238 #define RIPV6SNDQ	8192
239 #define RIPV6RCVQ	8192
240 	V_rip6_sendspace = RIPV6SNDQ;
241 	V_rip6_recvspace = RIPV6RCVQ;
242 
243 	/* ICMPV6 parameters */
244 	V_icmp6_rediraccept = 1;	/* accept and process redirects */
245 	V_icmp6_redirtimeout = 10 * 60;	/* 10 minutes */
246 	V_icmp6errppslim = 100;		/* 100pps */
247 	/* control how to respond to NI queries */
248 	V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK);
249 
250 	/* UDP on IP6 parameters */
251 	V_udp6_sendspace = 9216;	/* really max datagram size */
252 	V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
253 					/* 40 1K datagrams */
254 	V_dad_init = 0;
255 
256 	scope6_init();
257 	addrsel_policy_init();
258 	nd6_init();
259 	frag6_init();
260 
261 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
262 
263 	/* Skip global initialization stuff for non-default instances. */
264 	if (!IS_DEFAULT_VNET(curvnet))
265 		return;
266 
267 #ifdef DIAGNOSTIC
268 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
269 		panic("sizeof(protosw) != sizeof(ip6protosw)");
270 #endif
271 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
272 	if (pr == 0)
273 		panic("ip6_init");
274 
275 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
276 	for (i = 0; i < IPPROTO_MAX; i++)
277 		ip6_protox[i] = pr - inet6sw;
278 	/*
279 	 * Cycle through IP protocols and put them into the appropriate place
280 	 * in ip6_protox[].
281 	 */
282 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
283 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
284 		if (pr->pr_domain->dom_family == PF_INET6 &&
285 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
286 			/* Be careful to only index valid IP protocols. */
287 			if (pr->pr_protocol < IPPROTO_MAX)
288 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
289 		}
290 
291 	/* Initialize packet filter hooks. */
292 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
293 	inet6_pfil_hook.ph_af = AF_INET6;
294 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
295 		printf("%s: WARNING: unable to register pfil hook, "
296 			"error %d\n", __func__, i);
297 
298 	ip6intrq.ifq_maxlen = V_ip6qmaxlen; /* XXX */
299 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
300 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
301 }
302 
303 static int
304 ip6_init2_vnet(const void *unused __unused)
305 {
306 	INIT_VNET_INET6(curvnet);
307 
308 	/* nd6_timer_init */
309 	callout_init(&V_nd6_timer_ch, 0);
310 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, NULL);
311 
312 	/* timer for regeneranation of temporary addresses randomize ID */
313 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
314 	callout_reset(&V_in6_tmpaddrtimer_ch,
315 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
316 		       V_ip6_temp_regen_advance) * hz,
317 		      in6_tmpaddrtimer, NULL);
318 
319 	return (0);
320 }
321 
322 static void
323 ip6_init2(void *dummy)
324 {
325 
326 	ip6_init2_vnet(NULL);
327 }
328 
329 /* cheat */
330 /* This must be after route_init(), which is now SI_ORDER_THIRD */
331 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
332 
333 void
334 ip6_input(struct mbuf *m)
335 {
336 	INIT_VNET_NET(curvnet);
337 	INIT_VNET_INET6(curvnet);
338 	struct ip6_hdr *ip6;
339 	int off = sizeof(struct ip6_hdr), nest;
340 	u_int32_t plen;
341 	u_int32_t rtalert = ~0;
342 	int nxt, ours = 0;
343 	struct ifnet *deliverifp = NULL, *ifp = NULL;
344 	struct in6_addr odst;
345 	struct route_in6 rin6;
346 	int srcrt = 0;
347 	struct llentry *lle = NULL;
348 	struct sockaddr_in6 dst6, *dst;
349 
350 	bzero(&rin6, sizeof(struct route_in6));
351 #ifdef IPSEC
352 	/*
353 	 * should the inner packet be considered authentic?
354 	 * see comment in ah4_input().
355 	 * NB: m cannot be NULL when passed to the input routine
356 	 */
357 
358 	m->m_flags &= ~M_AUTHIPHDR;
359 	m->m_flags &= ~M_AUTHIPDGM;
360 
361 #endif /* IPSEC */
362 
363 	/*
364 	 * make sure we don't have onion peering information into m_tag.
365 	 */
366 	ip6_delaux(m);
367 
368 	/*
369 	 * mbuf statistics
370 	 */
371 	if (m->m_flags & M_EXT) {
372 		if (m->m_next)
373 			V_ip6stat.ip6s_mext2m++;
374 		else
375 			V_ip6stat.ip6s_mext1++;
376 	} else {
377 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
378 		if (m->m_next) {
379 			if (m->m_flags & M_LOOP) {
380 				V_ip6stat.ip6s_m2m[V_loif[0].if_index]++; /* XXX */
381 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
382 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
383 			else
384 				V_ip6stat.ip6s_m2m[0]++;
385 		} else
386 			V_ip6stat.ip6s_m1++;
387 #undef M2MMAX
388 	}
389 
390 	/* drop the packet if IPv6 operation is disabled on the IF */
391 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
392 		m_freem(m);
393 		return;
394 	}
395 
396 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
397 	V_ip6stat.ip6s_total++;
398 
399 #ifndef PULLDOWN_TEST
400 	/*
401 	 * L2 bridge code and some other code can return mbuf chain
402 	 * that does not conform to KAME requirement.  too bad.
403 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
404 	 */
405 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
406 		struct mbuf *n;
407 
408 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
409 		if (n)
410 			M_MOVE_PKTHDR(n, m);
411 		if (n && n->m_pkthdr.len > MHLEN) {
412 			MCLGET(n, M_DONTWAIT);
413 			if ((n->m_flags & M_EXT) == 0) {
414 				m_freem(n);
415 				n = NULL;
416 			}
417 		}
418 		if (n == NULL) {
419 			m_freem(m);
420 			return;	/* ENOBUFS */
421 		}
422 
423 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
424 		n->m_len = n->m_pkthdr.len;
425 		m_freem(m);
426 		m = n;
427 	}
428 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
429 #endif
430 
431 	if (m->m_len < sizeof(struct ip6_hdr)) {
432 		struct ifnet *inifp;
433 		inifp = m->m_pkthdr.rcvif;
434 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
435 			V_ip6stat.ip6s_toosmall++;
436 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
437 			return;
438 		}
439 	}
440 
441 	ip6 = mtod(m, struct ip6_hdr *);
442 
443 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
444 		V_ip6stat.ip6s_badvers++;
445 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
446 		goto bad;
447 	}
448 
449 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
450 
451 	/*
452 	 * Check against address spoofing/corruption.
453 	 */
454 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
455 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
456 		/*
457 		 * XXX: "badscope" is not very suitable for a multicast source.
458 		 */
459 		V_ip6stat.ip6s_badscope++;
460 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
461 		goto bad;
462 	}
463 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
464 	    !(m->m_flags & M_LOOP)) {
465 		/*
466 		 * In this case, the packet should come from the loopback
467 		 * interface.  However, we cannot just check the if_flags,
468 		 * because ip6_mloopback() passes the "actual" interface
469 		 * as the outgoing/incoming interface.
470 		 */
471 		V_ip6stat.ip6s_badscope++;
472 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
473 		goto bad;
474 	}
475 
476 #ifdef ALTQ
477 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
478 		/* packet is dropped by traffic conditioner */
479 		return;
480 	}
481 #endif
482 	/*
483 	 * The following check is not documented in specs.  A malicious
484 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
485 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
486 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
487 	 *
488 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
489 	 * support IPv4-less kernel compilation, we cannot support SIIT
490 	 * environment at all.  So, it makes more sense for us to reject any
491 	 * malicious packets for non-SIIT environment, than try to do a
492 	 * partial support for SIIT environment.
493 	 */
494 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
495 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
496 		V_ip6stat.ip6s_badscope++;
497 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
498 		goto bad;
499 	}
500 #if 0
501 	/*
502 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
503 	 *
504 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
505 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
506 	 * is revised to forbid relaying case.
507 	 */
508 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
509 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
510 		V_ip6stat.ip6s_badscope++;
511 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
512 		goto bad;
513 	}
514 #endif
515 
516 	/*
517 	 * Run through list of hooks for input packets.
518 	 *
519 	 * NB: Beware of the destination address changing
520 	 *     (e.g. by NAT rewriting).  When this happens,
521 	 *     tell ip6_forward to do the right thing.
522 	 */
523 	odst = ip6->ip6_dst;
524 
525 	/* Jump over all PFIL processing if hooks are not active. */
526 	if (!PFIL_HOOKED(&inet6_pfil_hook))
527 		goto passin;
528 
529 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
530 		return;
531 	if (m == NULL)			/* consumed by filter */
532 		return;
533 	ip6 = mtod(m, struct ip6_hdr *);
534 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
535 
536 passin:
537 	/*
538 	 * Disambiguate address scope zones (if there is ambiguity).
539 	 * We first make sure that the original source or destination address
540 	 * is not in our internal form for scoped addresses.  Such addresses
541 	 * are not necessarily invalid spec-wise, but we cannot accept them due
542 	 * to the usage conflict.
543 	 * in6_setscope() then also checks and rejects the cases where src or
544 	 * dst are the loopback address and the receiving interface
545 	 * is not loopback.
546 	 */
547 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
548 		V_ip6stat.ip6s_badscope++; /* XXX */
549 		goto bad;
550 	}
551 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
552 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
553 		V_ip6stat.ip6s_badscope++;
554 		goto bad;
555 	}
556 
557 	/*
558 	 * Multicast check
559 	 */
560 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
561 		struct in6_multi *in6m = 0;
562 
563 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
564 		/*
565 		 * See if we belong to the destination multicast group on the
566 		 * arrival interface.
567 		 */
568 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
569 		if (in6m)
570 			ours = 1;
571 		else if (!ip6_mrouter) {
572 			V_ip6stat.ip6s_notmember++;
573 			V_ip6stat.ip6s_cantforward++;
574 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
575 			goto bad;
576 		}
577 		deliverifp = m->m_pkthdr.rcvif;
578 		goto hbhcheck;
579 	}
580 
581 	/*
582 	 *  Unicast check
583 	 */
584 
585 	bzero(&dst6, sizeof(dst6));
586 	dst6.sin6_family = AF_INET6;
587 	dst6.sin6_len = sizeof(struct sockaddr_in6);
588 	dst6.sin6_addr = ip6->ip6_dst;
589 	ifp = m->m_pkthdr.rcvif;
590 	IF_AFDATA_LOCK(ifp);
591 	lle = lla_lookup(LLTABLE6(ifp), 0,
592 	     (struct sockaddr *)&dst6);
593 	IF_AFDATA_UNLOCK(ifp);
594 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
595 		ours = 1;
596 		deliverifp = ifp;
597 		LLE_RUNLOCK(lle);
598 		goto hbhcheck;
599 	}
600 	if (lle != NULL)
601 		LLE_RUNLOCK(lle);
602 
603 	dst = &rin6.ro_dst;
604 	dst->sin6_len = sizeof(struct sockaddr_in6);
605 	dst->sin6_family = AF_INET6;
606 	dst->sin6_addr = ip6->ip6_dst;
607 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
608 	if (rin6.ro_rt)
609 		RT_UNLOCK(rin6.ro_rt);
610 
611 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
612 
613 	/*
614 	 * Accept the packet if the forwarding interface to the destination
615 	 * according to the routing table is the loopback interface,
616 	 * unless the associated route has a gateway.
617 	 * Note that this approach causes to accept a packet if there is a
618 	 * route to the loopback interface for the destination of the packet.
619 	 * But we think it's even useful in some situations, e.g. when using
620 	 * a special daemon which wants to intercept the packet.
621 	 *
622 	 * XXX: some OSes automatically make a cloned route for the destination
623 	 * of an outgoing packet.  If the outgoing interface of the packet
624 	 * is a loopback one, the kernel would consider the packet to be
625 	 * accepted, even if we have no such address assinged on the interface.
626 	 * We check the cloned flag of the route entry to reject such cases,
627 	 * assuming that route entries for our own addresses are not made by
628 	 * cloning (it should be true because in6_addloop explicitly installs
629 	 * the host route).  However, we might have to do an explicit check
630 	 * while it would be less efficient.  Or, should we rather install a
631 	 * reject route for such a case?
632 	 */
633 	if (rin6.ro_rt &&
634 	    (rin6.ro_rt->rt_flags &
635 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
636 #ifdef RTF_WASCLONED
637 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
638 #endif
639 #ifdef RTF_CLONED
640 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
641 #endif
642 #if 0
643 	    /*
644 	     * The check below is redundant since the comparison of
645 	     * the destination and the key of the rtentry has
646 	     * already done through looking up the routing table.
647 	     */
648 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
649 	    &rt6_key(rin6.ro_rt)->sin6_addr)
650 #endif
651 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
652 		struct in6_ifaddr *ia6 =
653 			(struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
654 
655 		/*
656 		 * record address information into m_tag.
657 		 */
658 		(void)ip6_setdstifaddr(m, ia6);
659 
660 		/*
661 		 * packets to a tentative, duplicated, or somehow invalid
662 		 * address must not be accepted.
663 		 */
664 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
665 			/* this address is ready */
666 			ours = 1;
667 			deliverifp = ia6->ia_ifp;	/* correct? */
668 			/* Count the packet in the ip address stats */
669 			ia6->ia_ifa.if_ipackets++;
670 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
671 			goto hbhcheck;
672 		} else {
673 			char ip6bufs[INET6_ADDRSTRLEN];
674 			char ip6bufd[INET6_ADDRSTRLEN];
675 			/* address is not ready, so discard the packet. */
676 			nd6log((LOG_INFO,
677 			    "ip6_input: packet to an unready address %s->%s\n",
678 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
679 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
680 
681 			goto bad;
682 		}
683 	}
684 
685 	/*
686 	 * FAITH (Firewall Aided Internet Translator)
687 	 */
688 	if (V_ip6_keepfaith) {
689 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
690 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
691 			/* XXX do we need more sanity checks? */
692 			ours = 1;
693 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
694 			goto hbhcheck;
695 		}
696 	}
697 
698 	/*
699 	 * Now there is no reason to process the packet if it's not our own
700 	 * and we're not a router.
701 	 */
702 	if (!V_ip6_forwarding) {
703 		V_ip6stat.ip6s_cantforward++;
704 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
705 		goto bad;
706 	}
707 
708   hbhcheck:
709 	/*
710 	 * record address information into m_tag, if we don't have one yet.
711 	 * note that we are unable to record it, if the address is not listed
712 	 * as our interface address (e.g. multicast addresses, addresses
713 	 * within FAITH prefixes and such).
714 	 */
715 	if (deliverifp && !ip6_getdstifaddr(m)) {
716 		struct in6_ifaddr *ia6;
717 
718 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
719 		if (ia6) {
720 			if (!ip6_setdstifaddr(m, ia6)) {
721 				/*
722 				 * XXX maybe we should drop the packet here,
723 				 * as we could not provide enough information
724 				 * to the upper layers.
725 				 */
726 			}
727 		}
728 	}
729 
730 	/*
731 	 * Process Hop-by-Hop options header if it's contained.
732 	 * m may be modified in ip6_hopopts_input().
733 	 * If a JumboPayload option is included, plen will also be modified.
734 	 */
735 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
736 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
737 		struct ip6_hbh *hbh;
738 
739 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
740 #if 0	/*touches NULL pointer*/
741 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
742 #endif
743 			goto out;	/* m have already been freed */
744 		}
745 
746 		/* adjust pointer */
747 		ip6 = mtod(m, struct ip6_hdr *);
748 
749 		/*
750 		 * if the payload length field is 0 and the next header field
751 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
752 		 * option MUST be included.
753 		 */
754 		if (ip6->ip6_plen == 0 && plen == 0) {
755 			/*
756 			 * Note that if a valid jumbo payload option is
757 			 * contained, ip6_hopopts_input() must set a valid
758 			 * (non-zero) payload length to the variable plen.
759 			 */
760 			V_ip6stat.ip6s_badoptions++;
761 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
762 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
763 			icmp6_error(m, ICMP6_PARAM_PROB,
764 				    ICMP6_PARAMPROB_HEADER,
765 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
766 			goto out;
767 		}
768 #ifndef PULLDOWN_TEST
769 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
770 		hbh = (struct ip6_hbh *)(ip6 + 1);
771 #else
772 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
773 			sizeof(struct ip6_hbh));
774 		if (hbh == NULL) {
775 			V_ip6stat.ip6s_tooshort++;
776 			goto out;
777 		}
778 #endif
779 		nxt = hbh->ip6h_nxt;
780 
781 		/*
782 		 * If we are acting as a router and the packet contains a
783 		 * router alert option, see if we know the option value.
784 		 * Currently, we only support the option value for MLD, in which
785 		 * case we should pass the packet to the multicast routing
786 		 * daemon.
787 		 */
788 		if (rtalert != ~0 && V_ip6_forwarding) {
789 			switch (rtalert) {
790 			case IP6OPT_RTALERT_MLD:
791 				ours = 1;
792 				break;
793 			default:
794 				/*
795 				 * RFC2711 requires unrecognized values must be
796 				 * silently ignored.
797 				 */
798 				break;
799 			}
800 		}
801 	} else
802 		nxt = ip6->ip6_nxt;
803 
804 	/*
805 	 * Check that the amount of data in the buffers
806 	 * is as at least much as the IPv6 header would have us expect.
807 	 * Trim mbufs if longer than we expect.
808 	 * Drop packet if shorter than we expect.
809 	 */
810 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
811 		V_ip6stat.ip6s_tooshort++;
812 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
813 		goto bad;
814 	}
815 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
816 		if (m->m_len == m->m_pkthdr.len) {
817 			m->m_len = sizeof(struct ip6_hdr) + plen;
818 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
819 		} else
820 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
821 	}
822 
823 	/*
824 	 * Forward if desirable.
825 	 */
826 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
827 		/*
828 		 * If we are acting as a multicast router, all
829 		 * incoming multicast packets are passed to the
830 		 * kernel-level multicast forwarding function.
831 		 * The packet is returned (relatively) intact; if
832 		 * ip6_mforward() returns a non-zero value, the packet
833 		 * must be discarded, else it may be accepted below.
834 		 */
835 		if (ip6_mrouter && ip6_mforward &&
836 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
837 			V_ip6stat.ip6s_cantforward++;
838 			goto bad;
839 		}
840 		if (!ours)
841 			goto bad;
842 	} else if (!ours) {
843 		ip6_forward(m, srcrt);
844 		goto out;
845 	}
846 
847 	ip6 = mtod(m, struct ip6_hdr *);
848 
849 	/*
850 	 * Malicious party may be able to use IPv4 mapped addr to confuse
851 	 * tcp/udp stack and bypass security checks (act as if it was from
852 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
853 	 *
854 	 * For SIIT end node behavior, you may want to disable the check.
855 	 * However, you will  become vulnerable to attacks using IPv4 mapped
856 	 * source.
857 	 */
858 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
859 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
860 		V_ip6stat.ip6s_badscope++;
861 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
862 		goto bad;
863 	}
864 
865 	/*
866 	 * Tell launch routine the next header
867 	 */
868 	V_ip6stat.ip6s_delivered++;
869 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
870 	nest = 0;
871 
872 	while (nxt != IPPROTO_DONE) {
873 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
874 			V_ip6stat.ip6s_toomanyhdr++;
875 			goto bad;
876 		}
877 
878 		/*
879 		 * protection against faulty packet - there should be
880 		 * more sanity checks in header chain processing.
881 		 */
882 		if (m->m_pkthdr.len < off) {
883 			V_ip6stat.ip6s_tooshort++;
884 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
885 			goto bad;
886 		}
887 
888 #ifdef IPSEC
889 		/*
890 		 * enforce IPsec policy checking if we are seeing last header.
891 		 * note that we do not visit this with protocols with pcb layer
892 		 * code - like udp/tcp/raw ip.
893 		 */
894 		if (ip6_ipsec_input(m, nxt))
895 			goto bad;
896 #endif /* IPSEC */
897 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
898 	}
899 	goto out;
900 bad:
901 	m_freem(m);
902 out:
903 	if (rin6.ro_rt)
904 		RTFREE(rin6.ro_rt);
905 }
906 
907 /*
908  * set/grab in6_ifaddr correspond to IPv6 destination address.
909  * XXX backward compatibility wrapper
910  */
911 static struct ip6aux *
912 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
913 {
914 	struct ip6aux *ip6a;
915 
916 	ip6a = ip6_addaux(m);
917 	if (ip6a)
918 		ip6a->ip6a_dstia6 = ia6;
919 	return ip6a;	/* NULL if failed to set */
920 }
921 
922 struct in6_ifaddr *
923 ip6_getdstifaddr(struct mbuf *m)
924 {
925 	struct ip6aux *ip6a;
926 
927 	ip6a = ip6_findaux(m);
928 	if (ip6a)
929 		return ip6a->ip6a_dstia6;
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 	INIT_VNET_INET6(curvnet);
945 	struct mbuf *m = *mp;
946 	int off = *offp, hbhlen;
947 	struct ip6_hbh *hbh;
948 	u_int8_t *opt;
949 
950 	/* validation of the length of the header */
951 #ifndef PULLDOWN_TEST
952 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
953 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
954 	hbhlen = (hbh->ip6h_len + 1) << 3;
955 
956 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
957 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
958 #else
959 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
960 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
961 	if (hbh == NULL) {
962 		V_ip6stat.ip6s_tooshort++;
963 		return -1;
964 	}
965 	hbhlen = (hbh->ip6h_len + 1) << 3;
966 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
967 		hbhlen);
968 	if (hbh == NULL) {
969 		V_ip6stat.ip6s_tooshort++;
970 		return -1;
971 	}
972 #endif
973 	off += hbhlen;
974 	hbhlen -= sizeof(struct ip6_hbh);
975 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
976 
977 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
978 				hbhlen, rtalertp, plenp) < 0)
979 		return (-1);
980 
981 	*offp = off;
982 	*mp = m;
983 	return (0);
984 }
985 
986 /*
987  * Search header for all Hop-by-hop options and process each option.
988  * This function is separate from ip6_hopopts_input() in order to
989  * handle a case where the sending node itself process its hop-by-hop
990  * options header. In such a case, the function is called from ip6_output().
991  *
992  * The function assumes that hbh header is located right after the IPv6 header
993  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
994  * opthead + hbhlen is located in continuous memory region.
995  */
996 int
997 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
998     u_int32_t *rtalertp, u_int32_t *plenp)
999 {
1000 	INIT_VNET_INET6(curvnet);
1001 	struct ip6_hdr *ip6;
1002 	int optlen = 0;
1003 	u_int8_t *opt = opthead;
1004 	u_int16_t rtalert_val;
1005 	u_int32_t jumboplen;
1006 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1007 
1008 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1009 		switch (*opt) {
1010 		case IP6OPT_PAD1:
1011 			optlen = 1;
1012 			break;
1013 		case IP6OPT_PADN:
1014 			if (hbhlen < IP6OPT_MINLEN) {
1015 				V_ip6stat.ip6s_toosmall++;
1016 				goto bad;
1017 			}
1018 			optlen = *(opt + 1) + 2;
1019 			break;
1020 		case IP6OPT_ROUTER_ALERT:
1021 			/* XXX may need check for alignment */
1022 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1023 				V_ip6stat.ip6s_toosmall++;
1024 				goto bad;
1025 			}
1026 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1027 				/* XXX stat */
1028 				icmp6_error(m, ICMP6_PARAM_PROB,
1029 				    ICMP6_PARAMPROB_HEADER,
1030 				    erroff + opt + 1 - opthead);
1031 				return (-1);
1032 			}
1033 			optlen = IP6OPT_RTALERT_LEN;
1034 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1035 			*rtalertp = ntohs(rtalert_val);
1036 			break;
1037 		case IP6OPT_JUMBO:
1038 			/* XXX may need check for alignment */
1039 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1040 				V_ip6stat.ip6s_toosmall++;
1041 				goto bad;
1042 			}
1043 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1044 				/* XXX stat */
1045 				icmp6_error(m, ICMP6_PARAM_PROB,
1046 				    ICMP6_PARAMPROB_HEADER,
1047 				    erroff + opt + 1 - opthead);
1048 				return (-1);
1049 			}
1050 			optlen = IP6OPT_JUMBO_LEN;
1051 
1052 			/*
1053 			 * IPv6 packets that have non 0 payload length
1054 			 * must not contain a jumbo payload option.
1055 			 */
1056 			ip6 = mtod(m, struct ip6_hdr *);
1057 			if (ip6->ip6_plen) {
1058 				V_ip6stat.ip6s_badoptions++;
1059 				icmp6_error(m, ICMP6_PARAM_PROB,
1060 				    ICMP6_PARAMPROB_HEADER,
1061 				    erroff + opt - opthead);
1062 				return (-1);
1063 			}
1064 
1065 			/*
1066 			 * We may see jumbolen in unaligned location, so
1067 			 * we'd need to perform bcopy().
1068 			 */
1069 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1070 			jumboplen = (u_int32_t)htonl(jumboplen);
1071 
1072 #if 1
1073 			/*
1074 			 * if there are multiple jumbo payload options,
1075 			 * *plenp will be non-zero and the packet will be
1076 			 * rejected.
1077 			 * the behavior may need some debate in ipngwg -
1078 			 * multiple options does not make sense, however,
1079 			 * there's no explicit mention in specification.
1080 			 */
1081 			if (*plenp != 0) {
1082 				V_ip6stat.ip6s_badoptions++;
1083 				icmp6_error(m, ICMP6_PARAM_PROB,
1084 				    ICMP6_PARAMPROB_HEADER,
1085 				    erroff + opt + 2 - opthead);
1086 				return (-1);
1087 			}
1088 #endif
1089 
1090 			/*
1091 			 * jumbo payload length must be larger than 65535.
1092 			 */
1093 			if (jumboplen <= IPV6_MAXPACKET) {
1094 				V_ip6stat.ip6s_badoptions++;
1095 				icmp6_error(m, ICMP6_PARAM_PROB,
1096 				    ICMP6_PARAMPROB_HEADER,
1097 				    erroff + opt + 2 - opthead);
1098 				return (-1);
1099 			}
1100 			*plenp = jumboplen;
1101 
1102 			break;
1103 		default:		/* unknown option */
1104 			if (hbhlen < IP6OPT_MINLEN) {
1105 				V_ip6stat.ip6s_toosmall++;
1106 				goto bad;
1107 			}
1108 			optlen = ip6_unknown_opt(opt, m,
1109 			    erroff + opt - opthead);
1110 			if (optlen == -1)
1111 				return (-1);
1112 			optlen += 2;
1113 			break;
1114 		}
1115 	}
1116 
1117 	return (0);
1118 
1119   bad:
1120 	m_freem(m);
1121 	return (-1);
1122 }
1123 
1124 /*
1125  * Unknown option processing.
1126  * The third argument `off' is the offset from the IPv6 header to the option,
1127  * which is necessary if the IPv6 header the and option header and IPv6 header
1128  * is not continuous in order to return an ICMPv6 error.
1129  */
1130 int
1131 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1132 {
1133 	INIT_VNET_INET6(curvnet);
1134 	struct ip6_hdr *ip6;
1135 
1136 	switch (IP6OPT_TYPE(*optp)) {
1137 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1138 		return ((int)*(optp + 1));
1139 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1140 		m_freem(m);
1141 		return (-1);
1142 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1143 		V_ip6stat.ip6s_badoptions++;
1144 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1145 		return (-1);
1146 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1147 		V_ip6stat.ip6s_badoptions++;
1148 		ip6 = mtod(m, struct ip6_hdr *);
1149 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1150 		    (m->m_flags & (M_BCAST|M_MCAST)))
1151 			m_freem(m);
1152 		else
1153 			icmp6_error(m, ICMP6_PARAM_PROB,
1154 				    ICMP6_PARAMPROB_OPTION, off);
1155 		return (-1);
1156 	}
1157 
1158 	m_freem(m);		/* XXX: NOTREACHED */
1159 	return (-1);
1160 }
1161 
1162 /*
1163  * Create the "control" list for this pcb.
1164  * These functions will not modify mbuf chain at all.
1165  *
1166  * With KAME mbuf chain restriction:
1167  * The routine will be called from upper layer handlers like tcp6_input().
1168  * Thus the routine assumes that the caller (tcp6_input) have already
1169  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1170  * very first mbuf on the mbuf chain.
1171  *
1172  * ip6_savecontrol_v4 will handle those options that are possible to be
1173  * set on a v4-mapped socket.
1174  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1175  * options and handle the v6-only ones itself.
1176  */
1177 struct mbuf **
1178 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1179     int *v4only)
1180 {
1181 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1182 
1183 #ifdef SO_TIMESTAMP
1184 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1185 		struct timeval tv;
1186 
1187 		microtime(&tv);
1188 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1189 		    SCM_TIMESTAMP, SOL_SOCKET);
1190 		if (*mp)
1191 			mp = &(*mp)->m_next;
1192 	}
1193 #endif
1194 
1195 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1196 		if (v4only != NULL)
1197 			*v4only = 1;
1198 		return (mp);
1199 	}
1200 
1201 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1202 	/* RFC 2292 sec. 5 */
1203 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1204 		struct in6_pktinfo pi6;
1205 
1206 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1207 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1208 		pi6.ipi6_ifindex =
1209 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1210 
1211 		*mp = sbcreatecontrol((caddr_t) &pi6,
1212 		    sizeof(struct in6_pktinfo),
1213 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1214 		if (*mp)
1215 			mp = &(*mp)->m_next;
1216 	}
1217 
1218 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1219 		int hlim = ip6->ip6_hlim & 0xff;
1220 
1221 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1222 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1223 		    IPPROTO_IPV6);
1224 		if (*mp)
1225 			mp = &(*mp)->m_next;
1226 	}
1227 
1228 	if (v4only != NULL)
1229 		*v4only = 0;
1230 	return (mp);
1231 }
1232 
1233 void
1234 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1235 {
1236 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1237 	int v4only = 0;
1238 
1239 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1240 	if (v4only)
1241 		return;
1242 
1243 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1244 		u_int32_t flowinfo;
1245 		int tclass;
1246 
1247 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1248 		flowinfo >>= 20;
1249 
1250 		tclass = flowinfo & 0xff;
1251 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1252 		    IPV6_TCLASS, IPPROTO_IPV6);
1253 		if (*mp)
1254 			mp = &(*mp)->m_next;
1255 	}
1256 
1257 	/*
1258 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1259 	 * privilege for the option (see ip6_ctloutput), but it might be too
1260 	 * strict, since there might be some hop-by-hop options which can be
1261 	 * returned to normal user.
1262 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1263 	 */
1264 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1265 		/*
1266 		 * Check if a hop-by-hop options header is contatined in the
1267 		 * received packet, and if so, store the options as ancillary
1268 		 * data. Note that a hop-by-hop options header must be
1269 		 * just after the IPv6 header, which is assured through the
1270 		 * IPv6 input processing.
1271 		 */
1272 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1273 			struct ip6_hbh *hbh;
1274 			int hbhlen = 0;
1275 #ifdef PULLDOWN_TEST
1276 			struct mbuf *ext;
1277 #endif
1278 
1279 #ifndef PULLDOWN_TEST
1280 			hbh = (struct ip6_hbh *)(ip6 + 1);
1281 			hbhlen = (hbh->ip6h_len + 1) << 3;
1282 #else
1283 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1284 			    ip6->ip6_nxt);
1285 			if (ext == NULL) {
1286 				V_ip6stat.ip6s_tooshort++;
1287 				return;
1288 			}
1289 			hbh = mtod(ext, struct ip6_hbh *);
1290 			hbhlen = (hbh->ip6h_len + 1) << 3;
1291 			if (hbhlen != ext->m_len) {
1292 				m_freem(ext);
1293 				V_ip6stat.ip6s_tooshort++;
1294 				return;
1295 			}
1296 #endif
1297 
1298 			/*
1299 			 * XXX: We copy the whole header even if a
1300 			 * jumbo payload option is included, the option which
1301 			 * is to be removed before returning according to
1302 			 * RFC2292.
1303 			 * Note: this constraint is removed in RFC3542
1304 			 */
1305 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1306 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1307 			    IPPROTO_IPV6);
1308 			if (*mp)
1309 				mp = &(*mp)->m_next;
1310 #ifdef PULLDOWN_TEST
1311 			m_freem(ext);
1312 #endif
1313 		}
1314 	}
1315 
1316 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1317 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1318 
1319 		/*
1320 		 * Search for destination options headers or routing
1321 		 * header(s) through the header chain, and stores each
1322 		 * header as ancillary data.
1323 		 * Note that the order of the headers remains in
1324 		 * the chain of ancillary data.
1325 		 */
1326 		while (1) {	/* is explicit loop prevention necessary? */
1327 			struct ip6_ext *ip6e = NULL;
1328 			int elen;
1329 #ifdef PULLDOWN_TEST
1330 			struct mbuf *ext = NULL;
1331 #endif
1332 
1333 			/*
1334 			 * if it is not an extension header, don't try to
1335 			 * pull it from the chain.
1336 			 */
1337 			switch (nxt) {
1338 			case IPPROTO_DSTOPTS:
1339 			case IPPROTO_ROUTING:
1340 			case IPPROTO_HOPOPTS:
1341 			case IPPROTO_AH: /* is it possible? */
1342 				break;
1343 			default:
1344 				goto loopend;
1345 			}
1346 
1347 #ifndef PULLDOWN_TEST
1348 			if (off + sizeof(*ip6e) > m->m_len)
1349 				goto loopend;
1350 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1351 			if (nxt == IPPROTO_AH)
1352 				elen = (ip6e->ip6e_len + 2) << 2;
1353 			else
1354 				elen = (ip6e->ip6e_len + 1) << 3;
1355 			if (off + elen > m->m_len)
1356 				goto loopend;
1357 #else
1358 			ext = ip6_pullexthdr(m, off, nxt);
1359 			if (ext == NULL) {
1360 				V_ip6stat.ip6s_tooshort++;
1361 				return;
1362 			}
1363 			ip6e = mtod(ext, struct ip6_ext *);
1364 			if (nxt == IPPROTO_AH)
1365 				elen = (ip6e->ip6e_len + 2) << 2;
1366 			else
1367 				elen = (ip6e->ip6e_len + 1) << 3;
1368 			if (elen != ext->m_len) {
1369 				m_freem(ext);
1370 				V_ip6stat.ip6s_tooshort++;
1371 				return;
1372 			}
1373 #endif
1374 
1375 			switch (nxt) {
1376 			case IPPROTO_DSTOPTS:
1377 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1378 					break;
1379 
1380 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1381 				    IS2292(in6p,
1382 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1383 				    IPPROTO_IPV6);
1384 				if (*mp)
1385 					mp = &(*mp)->m_next;
1386 				break;
1387 			case IPPROTO_ROUTING:
1388 				if (!in6p->inp_flags & IN6P_RTHDR)
1389 					break;
1390 
1391 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1392 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1393 				    IPPROTO_IPV6);
1394 				if (*mp)
1395 					mp = &(*mp)->m_next;
1396 				break;
1397 			case IPPROTO_HOPOPTS:
1398 			case IPPROTO_AH: /* is it possible? */
1399 				break;
1400 
1401 			default:
1402 				/*
1403 				 * other cases have been filtered in the above.
1404 				 * none will visit this case.  here we supply
1405 				 * the code just in case (nxt overwritten or
1406 				 * other cases).
1407 				 */
1408 #ifdef PULLDOWN_TEST
1409 				m_freem(ext);
1410 #endif
1411 				goto loopend;
1412 
1413 			}
1414 
1415 			/* proceed with the next header. */
1416 			off += elen;
1417 			nxt = ip6e->ip6e_nxt;
1418 			ip6e = NULL;
1419 #ifdef PULLDOWN_TEST
1420 			m_freem(ext);
1421 			ext = NULL;
1422 #endif
1423 		}
1424 	  loopend:
1425 		;
1426 	}
1427 }
1428 #undef IS2292
1429 
1430 void
1431 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1432 {
1433 	struct socket *so;
1434 	struct mbuf *m_mtu;
1435 	struct ip6_mtuinfo mtuctl;
1436 
1437 	so =  in6p->inp_socket;
1438 
1439 	if (mtu == NULL)
1440 		return;
1441 
1442 #ifdef DIAGNOSTIC
1443 	if (so == NULL)		/* I believe this is impossible */
1444 		panic("ip6_notify_pmtu: socket is NULL");
1445 #endif
1446 
1447 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1448 	mtuctl.ip6m_mtu = *mtu;
1449 	mtuctl.ip6m_addr = *dst;
1450 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1451 		return;
1452 
1453 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1454 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1455 		return;
1456 
1457 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1458 	    == 0) {
1459 		m_freem(m_mtu);
1460 		/* XXX: should count statistics */
1461 	} else
1462 		sorwakeup(so);
1463 
1464 	return;
1465 }
1466 
1467 #ifdef PULLDOWN_TEST
1468 /*
1469  * pull single extension header from mbuf chain.  returns single mbuf that
1470  * contains the result, or NULL on error.
1471  */
1472 static struct mbuf *
1473 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1474 {
1475 	struct ip6_ext ip6e;
1476 	size_t elen;
1477 	struct mbuf *n;
1478 
1479 #ifdef DIAGNOSTIC
1480 	switch (nxt) {
1481 	case IPPROTO_DSTOPTS:
1482 	case IPPROTO_ROUTING:
1483 	case IPPROTO_HOPOPTS:
1484 	case IPPROTO_AH: /* is it possible? */
1485 		break;
1486 	default:
1487 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1488 	}
1489 #endif
1490 
1491 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1492 	if (nxt == IPPROTO_AH)
1493 		elen = (ip6e.ip6e_len + 2) << 2;
1494 	else
1495 		elen = (ip6e.ip6e_len + 1) << 3;
1496 
1497 	MGET(n, M_DONTWAIT, MT_DATA);
1498 	if (n && elen >= MLEN) {
1499 		MCLGET(n, M_DONTWAIT);
1500 		if ((n->m_flags & M_EXT) == 0) {
1501 			m_free(n);
1502 			n = NULL;
1503 		}
1504 	}
1505 	if (!n)
1506 		return NULL;
1507 
1508 	n->m_len = 0;
1509 	if (elen >= M_TRAILINGSPACE(n)) {
1510 		m_free(n);
1511 		return NULL;
1512 	}
1513 
1514 	m_copydata(m, off, elen, mtod(n, caddr_t));
1515 	n->m_len = elen;
1516 	return n;
1517 }
1518 #endif
1519 
1520 /*
1521  * Get pointer to the previous header followed by the header
1522  * currently processed.
1523  * XXX: This function supposes that
1524  *	M includes all headers,
1525  *	the next header field and the header length field of each header
1526  *	are valid, and
1527  *	the sum of each header length equals to OFF.
1528  * Because of these assumptions, this function must be called very
1529  * carefully. Moreover, it will not be used in the near future when
1530  * we develop `neater' mechanism to process extension headers.
1531  */
1532 char *
1533 ip6_get_prevhdr(struct mbuf *m, int off)
1534 {
1535 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1536 
1537 	if (off == sizeof(struct ip6_hdr))
1538 		return (&ip6->ip6_nxt);
1539 	else {
1540 		int len, nxt;
1541 		struct ip6_ext *ip6e = NULL;
1542 
1543 		nxt = ip6->ip6_nxt;
1544 		len = sizeof(struct ip6_hdr);
1545 		while (len < off) {
1546 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1547 
1548 			switch (nxt) {
1549 			case IPPROTO_FRAGMENT:
1550 				len += sizeof(struct ip6_frag);
1551 				break;
1552 			case IPPROTO_AH:
1553 				len += (ip6e->ip6e_len + 2) << 2;
1554 				break;
1555 			default:
1556 				len += (ip6e->ip6e_len + 1) << 3;
1557 				break;
1558 			}
1559 			nxt = ip6e->ip6e_nxt;
1560 		}
1561 		if (ip6e)
1562 			return (&ip6e->ip6e_nxt);
1563 		else
1564 			return NULL;
1565 	}
1566 }
1567 
1568 /*
1569  * get next header offset.  m will be retained.
1570  */
1571 int
1572 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1573 {
1574 	struct ip6_hdr ip6;
1575 	struct ip6_ext ip6e;
1576 	struct ip6_frag fh;
1577 
1578 	/* just in case */
1579 	if (m == NULL)
1580 		panic("ip6_nexthdr: m == NULL");
1581 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1582 		return -1;
1583 
1584 	switch (proto) {
1585 	case IPPROTO_IPV6:
1586 		if (m->m_pkthdr.len < off + sizeof(ip6))
1587 			return -1;
1588 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1589 		if (nxtp)
1590 			*nxtp = ip6.ip6_nxt;
1591 		off += sizeof(ip6);
1592 		return off;
1593 
1594 	case IPPROTO_FRAGMENT:
1595 		/*
1596 		 * terminate parsing if it is not the first fragment,
1597 		 * it does not make sense to parse through it.
1598 		 */
1599 		if (m->m_pkthdr.len < off + sizeof(fh))
1600 			return -1;
1601 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1602 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1603 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1604 			return -1;
1605 		if (nxtp)
1606 			*nxtp = fh.ip6f_nxt;
1607 		off += sizeof(struct ip6_frag);
1608 		return off;
1609 
1610 	case IPPROTO_AH:
1611 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1612 			return -1;
1613 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1614 		if (nxtp)
1615 			*nxtp = ip6e.ip6e_nxt;
1616 		off += (ip6e.ip6e_len + 2) << 2;
1617 		return off;
1618 
1619 	case IPPROTO_HOPOPTS:
1620 	case IPPROTO_ROUTING:
1621 	case IPPROTO_DSTOPTS:
1622 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1623 			return -1;
1624 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1625 		if (nxtp)
1626 			*nxtp = ip6e.ip6e_nxt;
1627 		off += (ip6e.ip6e_len + 1) << 3;
1628 		return off;
1629 
1630 	case IPPROTO_NONE:
1631 	case IPPROTO_ESP:
1632 	case IPPROTO_IPCOMP:
1633 		/* give up */
1634 		return -1;
1635 
1636 	default:
1637 		return -1;
1638 	}
1639 
1640 	return -1;
1641 }
1642 
1643 /*
1644  * get offset for the last header in the chain.  m will be kept untainted.
1645  */
1646 int
1647 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1648 {
1649 	int newoff;
1650 	int nxt;
1651 
1652 	if (!nxtp) {
1653 		nxt = -1;
1654 		nxtp = &nxt;
1655 	}
1656 	while (1) {
1657 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1658 		if (newoff < 0)
1659 			return off;
1660 		else if (newoff < off)
1661 			return -1;	/* invalid */
1662 		else if (newoff == off)
1663 			return newoff;
1664 
1665 		off = newoff;
1666 		proto = *nxtp;
1667 	}
1668 }
1669 
1670 struct ip6aux *
1671 ip6_addaux(struct mbuf *m)
1672 {
1673 	struct m_tag *mtag;
1674 
1675 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1676 	if (!mtag) {
1677 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1678 		    M_NOWAIT);
1679 		if (mtag) {
1680 			m_tag_prepend(m, mtag);
1681 			bzero(mtag + 1, sizeof(struct ip6aux));
1682 		}
1683 	}
1684 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1685 }
1686 
1687 struct ip6aux *
1688 ip6_findaux(struct mbuf *m)
1689 {
1690 	struct m_tag *mtag;
1691 
1692 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1693 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1694 }
1695 
1696 void
1697 ip6_delaux(struct mbuf *m)
1698 {
1699 	struct m_tag *mtag;
1700 
1701 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1702 	if (mtag)
1703 		m_tag_delete(m, mtag);
1704 }
1705 
1706 /*
1707  * System control for IP6
1708  */
1709 
1710 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1711 	0,		0,		0,		0,
1712 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1713 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1714 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1715 	0,		0,		0,		0,
1716 	ENOPROTOOPT
1717 };
1718