xref: /freebsd/sys/netinet6/ip6_input.c (revision d2b2128a286a00ee53d79cb88b4e59bf42525cf9)
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 /*
159  * IP6 initialization: fill in IP6 protocol switch table.
160  * All protocols not implemented in kernel go to raw IP6 protocol handler.
161  */
162 void
163 ip6_init(void)
164 {
165 	INIT_VNET_INET6(curvnet);
166 	struct ip6protosw *pr;
167 	int i;
168 
169 	V_ip6qmaxlen = IFQ_MAXLEN;
170 	V_in6_maxmtu = 0;
171 #ifdef IP6_AUTO_LINKLOCAL
172 	V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL;
173 #else
174 	V_ip6_auto_linklocal = 1;	/* enable by default */
175 #endif
176 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
177 	    &V_ip6_auto_linklocal);
178 
179 #ifndef IPV6FORWARDING
180 #ifdef GATEWAY6
181 #define IPV6FORWARDING	1	/* forward IP6 packets not for us */
182 #else
183 #define IPV6FORWARDING	0	/* don't forward IP6 packets not for us */
184 #endif /* GATEWAY6 */
185 #endif /* !IPV6FORWARDING */
186 
187 #ifndef IPV6_SENDREDIRECTS
188 #define IPV6_SENDREDIRECTS	1
189 #endif
190 
191 	V_ip6_forwarding = IPV6FORWARDING; /* act as router? */
192 	V_ip6_sendredirects = IPV6_SENDREDIRECTS;
193 	V_ip6_defhlim = IPV6_DEFHLIM;
194 	V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS;
195 	V_ip6_accept_rtadv = 0;	 /* "IPV6FORWARDING ? 0 : 1" is dangerous */
196 	V_ip6_log_interval = 5;
197 	V_ip6_hdrnestlimit = 15; /* How many header options will we process? */
198 	V_ip6_dad_count = 1;	 /* DupAddrDetectionTransmits */
199 	V_ip6_auto_flowlabel = 1;
200 	V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */
201 	V_ip6_rr_prune = 5;	 /* router renumbering prefix
202                                   * walk list every 5 sec. */
203 	V_ip6_mcast_pmtu = 0;	 /* enable pMTU discovery for multicast? */
204 	V_ip6_v6only = 1;
205 	V_ip6_keepfaith = 0;
206 	V_ip6_log_time = (time_t)0L;
207 #ifdef IPSTEALTH
208 	V_ip6stealth = 0;
209 #endif
210 	V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */
211 
212 	V_pmtu_expire = 60*10;
213 	V_pmtu_probe = 60*2;
214 
215 	/* raw IP6 parameters */
216 	/*
217 	 * Nominal space allocated to a raw ip socket.
218 	 */
219 #define RIPV6SNDQ	8192
220 #define RIPV6RCVQ	8192
221 	V_rip6_sendspace = RIPV6SNDQ;
222 	V_rip6_recvspace = RIPV6RCVQ;
223 
224 	/* ICMPV6 parameters */
225 	V_icmp6_rediraccept = 1;	/* accept and process redirects */
226 	V_icmp6_redirtimeout = 10 * 60;	/* 10 minutes */
227 	V_icmp6errppslim = 100;		/* 100pps */
228 	/* control how to respond to NI queries */
229 	V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK);
230 
231 	/* UDP on IP6 parameters */
232 	V_udp6_sendspace = 9216;	/* really max datagram size */
233 	V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
234 					/* 40 1K datagrams */
235 	V_dad_init = 0;
236 
237 #ifdef DIAGNOSTIC
238 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
239 		panic("sizeof(protosw) != sizeof(ip6protosw)");
240 #endif
241 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
242 	if (pr == 0)
243 		panic("ip6_init");
244 
245 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
246 	for (i = 0; i < IPPROTO_MAX; i++)
247 		ip6_protox[i] = pr - inet6sw;
248 	/*
249 	 * Cycle through IP protocols and put them into the appropriate place
250 	 * in ip6_protox[].
251 	 */
252 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
253 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
254 		if (pr->pr_domain->dom_family == PF_INET6 &&
255 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
256 			/* Be careful to only index valid IP protocols. */
257 			if (pr->pr_protocol < IPPROTO_MAX)
258 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
259 		}
260 
261 	/* Initialize packet filter hooks. */
262 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
263 	inet6_pfil_hook.ph_af = AF_INET6;
264 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
265 		printf("%s: WARNING: unable to register pfil hook, "
266 			"error %d\n", __func__, i);
267 
268 	ip6intrq.ifq_maxlen = V_ip6qmaxlen;
269 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
270 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
271 	scope6_init();
272 	addrsel_policy_init();
273 	nd6_init();
274 	frag6_init();
275 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
276 }
277 
278 static void
279 ip6_init2(void *dummy)
280 {
281 	INIT_VNET_INET6(curvnet);
282 
283 	/* nd6_timer_init */
284 	callout_init(&V_nd6_timer_ch, 0);
285 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, NULL);
286 
287 	/* timer for regeneranation of temporary addresses randomize ID */
288 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
289 	callout_reset(&V_in6_tmpaddrtimer_ch,
290 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
291 		       V_ip6_temp_regen_advance) * hz,
292 		      in6_tmpaddrtimer, NULL);
293 }
294 
295 /* cheat */
296 /* This must be after route_init(), which is now SI_ORDER_THIRD */
297 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
298 
299 void
300 ip6_input(struct mbuf *m)
301 {
302 	INIT_VNET_NET(curvnet);
303 	INIT_VNET_INET6(curvnet);
304 	struct ip6_hdr *ip6;
305 	int off = sizeof(struct ip6_hdr), nest;
306 	u_int32_t plen;
307 	u_int32_t rtalert = ~0;
308 	int nxt, ours = 0;
309 	struct ifnet *deliverifp = NULL, *ifp = NULL;
310 	struct in6_addr odst;
311 	struct route_in6 rin6;
312 	int srcrt = 0;
313 	struct llentry *lle = NULL;
314 	struct sockaddr_in6 dst6, *dst;
315 
316 	bzero(&rin6, sizeof(struct route_in6));
317 #ifdef IPSEC
318 	/*
319 	 * should the inner packet be considered authentic?
320 	 * see comment in ah4_input().
321 	 * NB: m cannot be NULL when passed to the input routine
322 	 */
323 
324 	m->m_flags &= ~M_AUTHIPHDR;
325 	m->m_flags &= ~M_AUTHIPDGM;
326 
327 #endif /* IPSEC */
328 
329 	/*
330 	 * make sure we don't have onion peering information into m_tag.
331 	 */
332 	ip6_delaux(m);
333 
334 	/*
335 	 * mbuf statistics
336 	 */
337 	if (m->m_flags & M_EXT) {
338 		if (m->m_next)
339 			V_ip6stat.ip6s_mext2m++;
340 		else
341 			V_ip6stat.ip6s_mext1++;
342 	} else {
343 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
344 		if (m->m_next) {
345 			if (m->m_flags & M_LOOP) {
346 				V_ip6stat.ip6s_m2m[V_loif[0].if_index]++; /* XXX */
347 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
348 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
349 			else
350 				V_ip6stat.ip6s_m2m[0]++;
351 		} else
352 			V_ip6stat.ip6s_m1++;
353 #undef M2MMAX
354 	}
355 
356 	/* drop the packet if IPv6 operation is disabled on the IF */
357 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
358 		m_freem(m);
359 		return;
360 	}
361 
362 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
363 	V_ip6stat.ip6s_total++;
364 
365 #ifndef PULLDOWN_TEST
366 	/*
367 	 * L2 bridge code and some other code can return mbuf chain
368 	 * that does not conform to KAME requirement.  too bad.
369 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
370 	 */
371 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
372 		struct mbuf *n;
373 
374 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
375 		if (n)
376 			M_MOVE_PKTHDR(n, m);
377 		if (n && n->m_pkthdr.len > MHLEN) {
378 			MCLGET(n, M_DONTWAIT);
379 			if ((n->m_flags & M_EXT) == 0) {
380 				m_freem(n);
381 				n = NULL;
382 			}
383 		}
384 		if (n == NULL) {
385 			m_freem(m);
386 			return;	/* ENOBUFS */
387 		}
388 
389 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
390 		n->m_len = n->m_pkthdr.len;
391 		m_freem(m);
392 		m = n;
393 	}
394 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
395 #endif
396 
397 	if (m->m_len < sizeof(struct ip6_hdr)) {
398 		struct ifnet *inifp;
399 		inifp = m->m_pkthdr.rcvif;
400 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
401 			V_ip6stat.ip6s_toosmall++;
402 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
403 			return;
404 		}
405 	}
406 
407 	ip6 = mtod(m, struct ip6_hdr *);
408 
409 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
410 		V_ip6stat.ip6s_badvers++;
411 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
412 		goto bad;
413 	}
414 
415 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
416 
417 	/*
418 	 * Check against address spoofing/corruption.
419 	 */
420 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
421 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
422 		/*
423 		 * XXX: "badscope" is not very suitable for a multicast source.
424 		 */
425 		V_ip6stat.ip6s_badscope++;
426 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
427 		goto bad;
428 	}
429 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
430 	    !(m->m_flags & M_LOOP)) {
431 		/*
432 		 * In this case, the packet should come from the loopback
433 		 * interface.  However, we cannot just check the if_flags,
434 		 * because ip6_mloopback() passes the "actual" interface
435 		 * as the outgoing/incoming interface.
436 		 */
437 		V_ip6stat.ip6s_badscope++;
438 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
439 		goto bad;
440 	}
441 
442 #ifdef ALTQ
443 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
444 		/* packet is dropped by traffic conditioner */
445 		return;
446 	}
447 #endif
448 	/*
449 	 * The following check is not documented in specs.  A malicious
450 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
451 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
452 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
453 	 *
454 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
455 	 * support IPv4-less kernel compilation, we cannot support SIIT
456 	 * environment at all.  So, it makes more sense for us to reject any
457 	 * malicious packets for non-SIIT environment, than try to do a
458 	 * partial support for SIIT environment.
459 	 */
460 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
461 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
462 		V_ip6stat.ip6s_badscope++;
463 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
464 		goto bad;
465 	}
466 #if 0
467 	/*
468 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
469 	 *
470 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
471 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
472 	 * is revised to forbid relaying case.
473 	 */
474 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
475 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
476 		V_ip6stat.ip6s_badscope++;
477 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
478 		goto bad;
479 	}
480 #endif
481 
482 	/*
483 	 * Run through list of hooks for input packets.
484 	 *
485 	 * NB: Beware of the destination address changing
486 	 *     (e.g. by NAT rewriting).  When this happens,
487 	 *     tell ip6_forward to do the right thing.
488 	 */
489 	odst = ip6->ip6_dst;
490 
491 	/* Jump over all PFIL processing if hooks are not active. */
492 	if (!PFIL_HOOKED(&inet6_pfil_hook))
493 		goto passin;
494 
495 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
496 		return;
497 	if (m == NULL)			/* consumed by filter */
498 		return;
499 	ip6 = mtod(m, struct ip6_hdr *);
500 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
501 
502 passin:
503 	/*
504 	 * Disambiguate address scope zones (if there is ambiguity).
505 	 * We first make sure that the original source or destination address
506 	 * is not in our internal form for scoped addresses.  Such addresses
507 	 * are not necessarily invalid spec-wise, but we cannot accept them due
508 	 * to the usage conflict.
509 	 * in6_setscope() then also checks and rejects the cases where src or
510 	 * dst are the loopback address and the receiving interface
511 	 * is not loopback.
512 	 */
513 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
514 		V_ip6stat.ip6s_badscope++; /* XXX */
515 		goto bad;
516 	}
517 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
518 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
519 		V_ip6stat.ip6s_badscope++;
520 		goto bad;
521 	}
522 
523 	/*
524 	 * Multicast check
525 	 */
526 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
527 		struct in6_multi *in6m = 0;
528 
529 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
530 		/*
531 		 * See if we belong to the destination multicast group on the
532 		 * arrival interface.
533 		 */
534 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
535 		if (in6m)
536 			ours = 1;
537 		else if (!ip6_mrouter) {
538 			V_ip6stat.ip6s_notmember++;
539 			V_ip6stat.ip6s_cantforward++;
540 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
541 			goto bad;
542 		}
543 		deliverifp = m->m_pkthdr.rcvif;
544 		goto hbhcheck;
545 	}
546 
547 	/*
548 	 *  Unicast check
549 	 */
550 
551 	bzero(&dst6, sizeof(dst6));
552 	dst6.sin6_family = AF_INET6;
553 	dst6.sin6_len = sizeof(struct sockaddr_in6);
554 	dst6.sin6_addr = ip6->ip6_dst;
555 	ifp = m->m_pkthdr.rcvif;
556 	IF_AFDATA_LOCK(ifp);
557 	lle = lla_lookup(LLTABLE6(ifp), 0,
558 	     (struct sockaddr *)&dst6);
559 	IF_AFDATA_UNLOCK(ifp);
560 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
561 		ours = 1;
562 		deliverifp = ifp;
563 		LLE_RUNLOCK(lle);
564 		goto hbhcheck;
565 	}
566 	if (lle != NULL)
567 		LLE_RUNLOCK(lle);
568 
569 	dst = &rin6.ro_dst;
570 	dst->sin6_len = sizeof(struct sockaddr_in6);
571 	dst->sin6_family = AF_INET6;
572 	dst->sin6_addr = ip6->ip6_dst;
573 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
574 	if (rin6.ro_rt)
575 		RT_UNLOCK(rin6.ro_rt);
576 
577 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
578 
579 	/*
580 	 * Accept the packet if the forwarding interface to the destination
581 	 * according to the routing table is the loopback interface,
582 	 * unless the associated route has a gateway.
583 	 * Note that this approach causes to accept a packet if there is a
584 	 * route to the loopback interface for the destination of the packet.
585 	 * But we think it's even useful in some situations, e.g. when using
586 	 * a special daemon which wants to intercept the packet.
587 	 *
588 	 * XXX: some OSes automatically make a cloned route for the destination
589 	 * of an outgoing packet.  If the outgoing interface of the packet
590 	 * is a loopback one, the kernel would consider the packet to be
591 	 * accepted, even if we have no such address assinged on the interface.
592 	 * We check the cloned flag of the route entry to reject such cases,
593 	 * assuming that route entries for our own addresses are not made by
594 	 * cloning (it should be true because in6_addloop explicitly installs
595 	 * the host route).  However, we might have to do an explicit check
596 	 * while it would be less efficient.  Or, should we rather install a
597 	 * reject route for such a case?
598 	 */
599 	if (rin6.ro_rt &&
600 	    (rin6.ro_rt->rt_flags &
601 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
602 #ifdef RTF_WASCLONED
603 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
604 #endif
605 #ifdef RTF_CLONED
606 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
607 #endif
608 #if 0
609 	    /*
610 	     * The check below is redundant since the comparison of
611 	     * the destination and the key of the rtentry has
612 	     * already done through looking up the routing table.
613 	     */
614 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
615 	    &rt6_key(rin6.ro_rt)->sin6_addr)
616 #endif
617 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
618 		struct in6_ifaddr *ia6 =
619 			(struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
620 
621 		/*
622 		 * record address information into m_tag.
623 		 */
624 		(void)ip6_setdstifaddr(m, ia6);
625 
626 		/*
627 		 * packets to a tentative, duplicated, or somehow invalid
628 		 * address must not be accepted.
629 		 */
630 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
631 			/* this address is ready */
632 			ours = 1;
633 			deliverifp = ia6->ia_ifp;	/* correct? */
634 			/* Count the packet in the ip address stats */
635 			ia6->ia_ifa.if_ipackets++;
636 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
637 			goto hbhcheck;
638 		} else {
639 			char ip6bufs[INET6_ADDRSTRLEN];
640 			char ip6bufd[INET6_ADDRSTRLEN];
641 			/* address is not ready, so discard the packet. */
642 			nd6log((LOG_INFO,
643 			    "ip6_input: packet to an unready address %s->%s\n",
644 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
645 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
646 
647 			goto bad;
648 		}
649 	}
650 
651 	/*
652 	 * FAITH (Firewall Aided Internet Translator)
653 	 */
654 	if (V_ip6_keepfaith) {
655 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
656 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
657 			/* XXX do we need more sanity checks? */
658 			ours = 1;
659 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
660 			goto hbhcheck;
661 		}
662 	}
663 
664 	/*
665 	 * Now there is no reason to process the packet if it's not our own
666 	 * and we're not a router.
667 	 */
668 	if (!V_ip6_forwarding) {
669 		V_ip6stat.ip6s_cantforward++;
670 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
671 		goto bad;
672 	}
673 
674   hbhcheck:
675 	/*
676 	 * record address information into m_tag, if we don't have one yet.
677 	 * note that we are unable to record it, if the address is not listed
678 	 * as our interface address (e.g. multicast addresses, addresses
679 	 * within FAITH prefixes and such).
680 	 */
681 	if (deliverifp && !ip6_getdstifaddr(m)) {
682 		struct in6_ifaddr *ia6;
683 
684 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
685 		if (ia6) {
686 			if (!ip6_setdstifaddr(m, ia6)) {
687 				/*
688 				 * XXX maybe we should drop the packet here,
689 				 * as we could not provide enough information
690 				 * to the upper layers.
691 				 */
692 			}
693 		}
694 	}
695 
696 	/*
697 	 * Process Hop-by-Hop options header if it's contained.
698 	 * m may be modified in ip6_hopopts_input().
699 	 * If a JumboPayload option is included, plen will also be modified.
700 	 */
701 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
702 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
703 		struct ip6_hbh *hbh;
704 
705 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
706 #if 0	/*touches NULL pointer*/
707 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
708 #endif
709 			goto out;	/* m have already been freed */
710 		}
711 
712 		/* adjust pointer */
713 		ip6 = mtod(m, struct ip6_hdr *);
714 
715 		/*
716 		 * if the payload length field is 0 and the next header field
717 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
718 		 * option MUST be included.
719 		 */
720 		if (ip6->ip6_plen == 0 && plen == 0) {
721 			/*
722 			 * Note that if a valid jumbo payload option is
723 			 * contained, ip6_hopopts_input() must set a valid
724 			 * (non-zero) payload length to the variable plen.
725 			 */
726 			V_ip6stat.ip6s_badoptions++;
727 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
728 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
729 			icmp6_error(m, ICMP6_PARAM_PROB,
730 				    ICMP6_PARAMPROB_HEADER,
731 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
732 			goto out;
733 		}
734 #ifndef PULLDOWN_TEST
735 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
736 		hbh = (struct ip6_hbh *)(ip6 + 1);
737 #else
738 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
739 			sizeof(struct ip6_hbh));
740 		if (hbh == NULL) {
741 			V_ip6stat.ip6s_tooshort++;
742 			goto out;
743 		}
744 #endif
745 		nxt = hbh->ip6h_nxt;
746 
747 		/*
748 		 * If we are acting as a router and the packet contains a
749 		 * router alert option, see if we know the option value.
750 		 * Currently, we only support the option value for MLD, in which
751 		 * case we should pass the packet to the multicast routing
752 		 * daemon.
753 		 */
754 		if (rtalert != ~0 && V_ip6_forwarding) {
755 			switch (rtalert) {
756 			case IP6OPT_RTALERT_MLD:
757 				ours = 1;
758 				break;
759 			default:
760 				/*
761 				 * RFC2711 requires unrecognized values must be
762 				 * silently ignored.
763 				 */
764 				break;
765 			}
766 		}
767 	} else
768 		nxt = ip6->ip6_nxt;
769 
770 	/*
771 	 * Check that the amount of data in the buffers
772 	 * is as at least much as the IPv6 header would have us expect.
773 	 * Trim mbufs if longer than we expect.
774 	 * Drop packet if shorter than we expect.
775 	 */
776 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
777 		V_ip6stat.ip6s_tooshort++;
778 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
779 		goto bad;
780 	}
781 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
782 		if (m->m_len == m->m_pkthdr.len) {
783 			m->m_len = sizeof(struct ip6_hdr) + plen;
784 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
785 		} else
786 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
787 	}
788 
789 	/*
790 	 * Forward if desirable.
791 	 */
792 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
793 		/*
794 		 * If we are acting as a multicast router, all
795 		 * incoming multicast packets are passed to the
796 		 * kernel-level multicast forwarding function.
797 		 * The packet is returned (relatively) intact; if
798 		 * ip6_mforward() returns a non-zero value, the packet
799 		 * must be discarded, else it may be accepted below.
800 		 */
801 		if (ip6_mrouter && ip6_mforward &&
802 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
803 			V_ip6stat.ip6s_cantforward++;
804 			goto bad;
805 		}
806 		if (!ours)
807 			goto bad;
808 	} else if (!ours) {
809 		ip6_forward(m, srcrt);
810 		goto out;
811 	}
812 
813 	ip6 = mtod(m, struct ip6_hdr *);
814 
815 	/*
816 	 * Malicious party may be able to use IPv4 mapped addr to confuse
817 	 * tcp/udp stack and bypass security checks (act as if it was from
818 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
819 	 *
820 	 * For SIIT end node behavior, you may want to disable the check.
821 	 * However, you will  become vulnerable to attacks using IPv4 mapped
822 	 * source.
823 	 */
824 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
825 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
826 		V_ip6stat.ip6s_badscope++;
827 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
828 		goto bad;
829 	}
830 
831 	/*
832 	 * Tell launch routine the next header
833 	 */
834 	V_ip6stat.ip6s_delivered++;
835 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
836 	nest = 0;
837 
838 	while (nxt != IPPROTO_DONE) {
839 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
840 			V_ip6stat.ip6s_toomanyhdr++;
841 			goto bad;
842 		}
843 
844 		/*
845 		 * protection against faulty packet - there should be
846 		 * more sanity checks in header chain processing.
847 		 */
848 		if (m->m_pkthdr.len < off) {
849 			V_ip6stat.ip6s_tooshort++;
850 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
851 			goto bad;
852 		}
853 
854 #ifdef IPSEC
855 		/*
856 		 * enforce IPsec policy checking if we are seeing last header.
857 		 * note that we do not visit this with protocols with pcb layer
858 		 * code - like udp/tcp/raw ip.
859 		 */
860 		if (ip6_ipsec_input(m, nxt))
861 			goto bad;
862 #endif /* IPSEC */
863 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
864 	}
865 	goto out;
866 bad:
867 	m_freem(m);
868 out:
869 	if (rin6.ro_rt)
870 		RTFREE(rin6.ro_rt);
871 }
872 
873 /*
874  * set/grab in6_ifaddr correspond to IPv6 destination address.
875  * XXX backward compatibility wrapper
876  */
877 static struct ip6aux *
878 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
879 {
880 	struct ip6aux *ip6a;
881 
882 	ip6a = ip6_addaux(m);
883 	if (ip6a)
884 		ip6a->ip6a_dstia6 = ia6;
885 	return ip6a;	/* NULL if failed to set */
886 }
887 
888 struct in6_ifaddr *
889 ip6_getdstifaddr(struct mbuf *m)
890 {
891 	struct ip6aux *ip6a;
892 
893 	ip6a = ip6_findaux(m);
894 	if (ip6a)
895 		return ip6a->ip6a_dstia6;
896 	else
897 		return NULL;
898 }
899 
900 /*
901  * Hop-by-Hop options header processing. If a valid jumbo payload option is
902  * included, the real payload length will be stored in plenp.
903  *
904  * rtalertp - XXX: should be stored more smart way
905  */
906 static int
907 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
908     struct mbuf **mp, int *offp)
909 {
910 	INIT_VNET_INET6(curvnet);
911 	struct mbuf *m = *mp;
912 	int off = *offp, hbhlen;
913 	struct ip6_hbh *hbh;
914 	u_int8_t *opt;
915 
916 	/* validation of the length of the header */
917 #ifndef PULLDOWN_TEST
918 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
919 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
920 	hbhlen = (hbh->ip6h_len + 1) << 3;
921 
922 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
923 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
924 #else
925 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
926 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
927 	if (hbh == NULL) {
928 		V_ip6stat.ip6s_tooshort++;
929 		return -1;
930 	}
931 	hbhlen = (hbh->ip6h_len + 1) << 3;
932 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
933 		hbhlen);
934 	if (hbh == NULL) {
935 		V_ip6stat.ip6s_tooshort++;
936 		return -1;
937 	}
938 #endif
939 	off += hbhlen;
940 	hbhlen -= sizeof(struct ip6_hbh);
941 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
942 
943 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
944 				hbhlen, rtalertp, plenp) < 0)
945 		return (-1);
946 
947 	*offp = off;
948 	*mp = m;
949 	return (0);
950 }
951 
952 /*
953  * Search header for all Hop-by-hop options and process each option.
954  * This function is separate from ip6_hopopts_input() in order to
955  * handle a case where the sending node itself process its hop-by-hop
956  * options header. In such a case, the function is called from ip6_output().
957  *
958  * The function assumes that hbh header is located right after the IPv6 header
959  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
960  * opthead + hbhlen is located in continuous memory region.
961  */
962 int
963 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
964     u_int32_t *rtalertp, u_int32_t *plenp)
965 {
966 	INIT_VNET_INET6(curvnet);
967 	struct ip6_hdr *ip6;
968 	int optlen = 0;
969 	u_int8_t *opt = opthead;
970 	u_int16_t rtalert_val;
971 	u_int32_t jumboplen;
972 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
973 
974 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
975 		switch (*opt) {
976 		case IP6OPT_PAD1:
977 			optlen = 1;
978 			break;
979 		case IP6OPT_PADN:
980 			if (hbhlen < IP6OPT_MINLEN) {
981 				V_ip6stat.ip6s_toosmall++;
982 				goto bad;
983 			}
984 			optlen = *(opt + 1) + 2;
985 			break;
986 		case IP6OPT_ROUTER_ALERT:
987 			/* XXX may need check for alignment */
988 			if (hbhlen < IP6OPT_RTALERT_LEN) {
989 				V_ip6stat.ip6s_toosmall++;
990 				goto bad;
991 			}
992 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
993 				/* XXX stat */
994 				icmp6_error(m, ICMP6_PARAM_PROB,
995 				    ICMP6_PARAMPROB_HEADER,
996 				    erroff + opt + 1 - opthead);
997 				return (-1);
998 			}
999 			optlen = IP6OPT_RTALERT_LEN;
1000 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1001 			*rtalertp = ntohs(rtalert_val);
1002 			break;
1003 		case IP6OPT_JUMBO:
1004 			/* XXX may need check for alignment */
1005 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1006 				V_ip6stat.ip6s_toosmall++;
1007 				goto bad;
1008 			}
1009 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1010 				/* XXX stat */
1011 				icmp6_error(m, ICMP6_PARAM_PROB,
1012 				    ICMP6_PARAMPROB_HEADER,
1013 				    erroff + opt + 1 - opthead);
1014 				return (-1);
1015 			}
1016 			optlen = IP6OPT_JUMBO_LEN;
1017 
1018 			/*
1019 			 * IPv6 packets that have non 0 payload length
1020 			 * must not contain a jumbo payload option.
1021 			 */
1022 			ip6 = mtod(m, struct ip6_hdr *);
1023 			if (ip6->ip6_plen) {
1024 				V_ip6stat.ip6s_badoptions++;
1025 				icmp6_error(m, ICMP6_PARAM_PROB,
1026 				    ICMP6_PARAMPROB_HEADER,
1027 				    erroff + opt - opthead);
1028 				return (-1);
1029 			}
1030 
1031 			/*
1032 			 * We may see jumbolen in unaligned location, so
1033 			 * we'd need to perform bcopy().
1034 			 */
1035 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1036 			jumboplen = (u_int32_t)htonl(jumboplen);
1037 
1038 #if 1
1039 			/*
1040 			 * if there are multiple jumbo payload options,
1041 			 * *plenp will be non-zero and the packet will be
1042 			 * rejected.
1043 			 * the behavior may need some debate in ipngwg -
1044 			 * multiple options does not make sense, however,
1045 			 * there's no explicit mention in specification.
1046 			 */
1047 			if (*plenp != 0) {
1048 				V_ip6stat.ip6s_badoptions++;
1049 				icmp6_error(m, ICMP6_PARAM_PROB,
1050 				    ICMP6_PARAMPROB_HEADER,
1051 				    erroff + opt + 2 - opthead);
1052 				return (-1);
1053 			}
1054 #endif
1055 
1056 			/*
1057 			 * jumbo payload length must be larger than 65535.
1058 			 */
1059 			if (jumboplen <= IPV6_MAXPACKET) {
1060 				V_ip6stat.ip6s_badoptions++;
1061 				icmp6_error(m, ICMP6_PARAM_PROB,
1062 				    ICMP6_PARAMPROB_HEADER,
1063 				    erroff + opt + 2 - opthead);
1064 				return (-1);
1065 			}
1066 			*plenp = jumboplen;
1067 
1068 			break;
1069 		default:		/* unknown option */
1070 			if (hbhlen < IP6OPT_MINLEN) {
1071 				V_ip6stat.ip6s_toosmall++;
1072 				goto bad;
1073 			}
1074 			optlen = ip6_unknown_opt(opt, m,
1075 			    erroff + opt - opthead);
1076 			if (optlen == -1)
1077 				return (-1);
1078 			optlen += 2;
1079 			break;
1080 		}
1081 	}
1082 
1083 	return (0);
1084 
1085   bad:
1086 	m_freem(m);
1087 	return (-1);
1088 }
1089 
1090 /*
1091  * Unknown option processing.
1092  * The third argument `off' is the offset from the IPv6 header to the option,
1093  * which is necessary if the IPv6 header the and option header and IPv6 header
1094  * is not continuous in order to return an ICMPv6 error.
1095  */
1096 int
1097 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1098 {
1099 	INIT_VNET_INET6(curvnet);
1100 	struct ip6_hdr *ip6;
1101 
1102 	switch (IP6OPT_TYPE(*optp)) {
1103 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1104 		return ((int)*(optp + 1));
1105 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1106 		m_freem(m);
1107 		return (-1);
1108 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1109 		V_ip6stat.ip6s_badoptions++;
1110 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1111 		return (-1);
1112 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1113 		V_ip6stat.ip6s_badoptions++;
1114 		ip6 = mtod(m, struct ip6_hdr *);
1115 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1116 		    (m->m_flags & (M_BCAST|M_MCAST)))
1117 			m_freem(m);
1118 		else
1119 			icmp6_error(m, ICMP6_PARAM_PROB,
1120 				    ICMP6_PARAMPROB_OPTION, off);
1121 		return (-1);
1122 	}
1123 
1124 	m_freem(m);		/* XXX: NOTREACHED */
1125 	return (-1);
1126 }
1127 
1128 /*
1129  * Create the "control" list for this pcb.
1130  * These functions will not modify mbuf chain at all.
1131  *
1132  * With KAME mbuf chain restriction:
1133  * The routine will be called from upper layer handlers like tcp6_input().
1134  * Thus the routine assumes that the caller (tcp6_input) have already
1135  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1136  * very first mbuf on the mbuf chain.
1137  *
1138  * ip6_savecontrol_v4 will handle those options that are possible to be
1139  * set on a v4-mapped socket.
1140  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1141  * options and handle the v6-only ones itself.
1142  */
1143 struct mbuf **
1144 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1145     int *v4only)
1146 {
1147 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1148 
1149 #ifdef SO_TIMESTAMP
1150 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1151 		struct timeval tv;
1152 
1153 		microtime(&tv);
1154 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1155 		    SCM_TIMESTAMP, SOL_SOCKET);
1156 		if (*mp)
1157 			mp = &(*mp)->m_next;
1158 	}
1159 #endif
1160 
1161 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1162 		if (v4only != NULL)
1163 			*v4only = 1;
1164 		return (mp);
1165 	}
1166 
1167 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1168 	/* RFC 2292 sec. 5 */
1169 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1170 		struct in6_pktinfo pi6;
1171 
1172 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1173 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1174 		pi6.ipi6_ifindex =
1175 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1176 
1177 		*mp = sbcreatecontrol((caddr_t) &pi6,
1178 		    sizeof(struct in6_pktinfo),
1179 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1180 		if (*mp)
1181 			mp = &(*mp)->m_next;
1182 	}
1183 
1184 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1185 		int hlim = ip6->ip6_hlim & 0xff;
1186 
1187 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1188 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1189 		    IPPROTO_IPV6);
1190 		if (*mp)
1191 			mp = &(*mp)->m_next;
1192 	}
1193 
1194 	if (v4only != NULL)
1195 		*v4only = 0;
1196 	return (mp);
1197 }
1198 
1199 void
1200 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1201 {
1202 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1203 	int v4only = 0;
1204 
1205 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1206 	if (v4only)
1207 		return;
1208 
1209 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1210 		u_int32_t flowinfo;
1211 		int tclass;
1212 
1213 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1214 		flowinfo >>= 20;
1215 
1216 		tclass = flowinfo & 0xff;
1217 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1218 		    IPV6_TCLASS, IPPROTO_IPV6);
1219 		if (*mp)
1220 			mp = &(*mp)->m_next;
1221 	}
1222 
1223 	/*
1224 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1225 	 * privilege for the option (see ip6_ctloutput), but it might be too
1226 	 * strict, since there might be some hop-by-hop options which can be
1227 	 * returned to normal user.
1228 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1229 	 */
1230 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1231 		/*
1232 		 * Check if a hop-by-hop options header is contatined in the
1233 		 * received packet, and if so, store the options as ancillary
1234 		 * data. Note that a hop-by-hop options header must be
1235 		 * just after the IPv6 header, which is assured through the
1236 		 * IPv6 input processing.
1237 		 */
1238 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1239 			struct ip6_hbh *hbh;
1240 			int hbhlen = 0;
1241 #ifdef PULLDOWN_TEST
1242 			struct mbuf *ext;
1243 #endif
1244 
1245 #ifndef PULLDOWN_TEST
1246 			hbh = (struct ip6_hbh *)(ip6 + 1);
1247 			hbhlen = (hbh->ip6h_len + 1) << 3;
1248 #else
1249 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1250 			    ip6->ip6_nxt);
1251 			if (ext == NULL) {
1252 				V_ip6stat.ip6s_tooshort++;
1253 				return;
1254 			}
1255 			hbh = mtod(ext, struct ip6_hbh *);
1256 			hbhlen = (hbh->ip6h_len + 1) << 3;
1257 			if (hbhlen != ext->m_len) {
1258 				m_freem(ext);
1259 				V_ip6stat.ip6s_tooshort++;
1260 				return;
1261 			}
1262 #endif
1263 
1264 			/*
1265 			 * XXX: We copy the whole header even if a
1266 			 * jumbo payload option is included, the option which
1267 			 * is to be removed before returning according to
1268 			 * RFC2292.
1269 			 * Note: this constraint is removed in RFC3542
1270 			 */
1271 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1272 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1273 			    IPPROTO_IPV6);
1274 			if (*mp)
1275 				mp = &(*mp)->m_next;
1276 #ifdef PULLDOWN_TEST
1277 			m_freem(ext);
1278 #endif
1279 		}
1280 	}
1281 
1282 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1283 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1284 
1285 		/*
1286 		 * Search for destination options headers or routing
1287 		 * header(s) through the header chain, and stores each
1288 		 * header as ancillary data.
1289 		 * Note that the order of the headers remains in
1290 		 * the chain of ancillary data.
1291 		 */
1292 		while (1) {	/* is explicit loop prevention necessary? */
1293 			struct ip6_ext *ip6e = NULL;
1294 			int elen;
1295 #ifdef PULLDOWN_TEST
1296 			struct mbuf *ext = NULL;
1297 #endif
1298 
1299 			/*
1300 			 * if it is not an extension header, don't try to
1301 			 * pull it from the chain.
1302 			 */
1303 			switch (nxt) {
1304 			case IPPROTO_DSTOPTS:
1305 			case IPPROTO_ROUTING:
1306 			case IPPROTO_HOPOPTS:
1307 			case IPPROTO_AH: /* is it possible? */
1308 				break;
1309 			default:
1310 				goto loopend;
1311 			}
1312 
1313 #ifndef PULLDOWN_TEST
1314 			if (off + sizeof(*ip6e) > m->m_len)
1315 				goto loopend;
1316 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1317 			if (nxt == IPPROTO_AH)
1318 				elen = (ip6e->ip6e_len + 2) << 2;
1319 			else
1320 				elen = (ip6e->ip6e_len + 1) << 3;
1321 			if (off + elen > m->m_len)
1322 				goto loopend;
1323 #else
1324 			ext = ip6_pullexthdr(m, off, nxt);
1325 			if (ext == NULL) {
1326 				V_ip6stat.ip6s_tooshort++;
1327 				return;
1328 			}
1329 			ip6e = mtod(ext, struct ip6_ext *);
1330 			if (nxt == IPPROTO_AH)
1331 				elen = (ip6e->ip6e_len + 2) << 2;
1332 			else
1333 				elen = (ip6e->ip6e_len + 1) << 3;
1334 			if (elen != ext->m_len) {
1335 				m_freem(ext);
1336 				V_ip6stat.ip6s_tooshort++;
1337 				return;
1338 			}
1339 #endif
1340 
1341 			switch (nxt) {
1342 			case IPPROTO_DSTOPTS:
1343 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1344 					break;
1345 
1346 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1347 				    IS2292(in6p,
1348 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1349 				    IPPROTO_IPV6);
1350 				if (*mp)
1351 					mp = &(*mp)->m_next;
1352 				break;
1353 			case IPPROTO_ROUTING:
1354 				if (!in6p->inp_flags & IN6P_RTHDR)
1355 					break;
1356 
1357 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1358 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1359 				    IPPROTO_IPV6);
1360 				if (*mp)
1361 					mp = &(*mp)->m_next;
1362 				break;
1363 			case IPPROTO_HOPOPTS:
1364 			case IPPROTO_AH: /* is it possible? */
1365 				break;
1366 
1367 			default:
1368 				/*
1369 				 * other cases have been filtered in the above.
1370 				 * none will visit this case.  here we supply
1371 				 * the code just in case (nxt overwritten or
1372 				 * other cases).
1373 				 */
1374 #ifdef PULLDOWN_TEST
1375 				m_freem(ext);
1376 #endif
1377 				goto loopend;
1378 
1379 			}
1380 
1381 			/* proceed with the next header. */
1382 			off += elen;
1383 			nxt = ip6e->ip6e_nxt;
1384 			ip6e = NULL;
1385 #ifdef PULLDOWN_TEST
1386 			m_freem(ext);
1387 			ext = NULL;
1388 #endif
1389 		}
1390 	  loopend:
1391 		;
1392 	}
1393 }
1394 #undef IS2292
1395 
1396 void
1397 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1398 {
1399 	struct socket *so;
1400 	struct mbuf *m_mtu;
1401 	struct ip6_mtuinfo mtuctl;
1402 
1403 	so =  in6p->inp_socket;
1404 
1405 	if (mtu == NULL)
1406 		return;
1407 
1408 #ifdef DIAGNOSTIC
1409 	if (so == NULL)		/* I believe this is impossible */
1410 		panic("ip6_notify_pmtu: socket is NULL");
1411 #endif
1412 
1413 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1414 	mtuctl.ip6m_mtu = *mtu;
1415 	mtuctl.ip6m_addr = *dst;
1416 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1417 		return;
1418 
1419 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1420 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1421 		return;
1422 
1423 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1424 	    == 0) {
1425 		m_freem(m_mtu);
1426 		/* XXX: should count statistics */
1427 	} else
1428 		sorwakeup(so);
1429 
1430 	return;
1431 }
1432 
1433 #ifdef PULLDOWN_TEST
1434 /*
1435  * pull single extension header from mbuf chain.  returns single mbuf that
1436  * contains the result, or NULL on error.
1437  */
1438 static struct mbuf *
1439 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1440 {
1441 	struct ip6_ext ip6e;
1442 	size_t elen;
1443 	struct mbuf *n;
1444 
1445 #ifdef DIAGNOSTIC
1446 	switch (nxt) {
1447 	case IPPROTO_DSTOPTS:
1448 	case IPPROTO_ROUTING:
1449 	case IPPROTO_HOPOPTS:
1450 	case IPPROTO_AH: /* is it possible? */
1451 		break;
1452 	default:
1453 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1454 	}
1455 #endif
1456 
1457 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1458 	if (nxt == IPPROTO_AH)
1459 		elen = (ip6e.ip6e_len + 2) << 2;
1460 	else
1461 		elen = (ip6e.ip6e_len + 1) << 3;
1462 
1463 	MGET(n, M_DONTWAIT, MT_DATA);
1464 	if (n && elen >= MLEN) {
1465 		MCLGET(n, M_DONTWAIT);
1466 		if ((n->m_flags & M_EXT) == 0) {
1467 			m_free(n);
1468 			n = NULL;
1469 		}
1470 	}
1471 	if (!n)
1472 		return NULL;
1473 
1474 	n->m_len = 0;
1475 	if (elen >= M_TRAILINGSPACE(n)) {
1476 		m_free(n);
1477 		return NULL;
1478 	}
1479 
1480 	m_copydata(m, off, elen, mtod(n, caddr_t));
1481 	n->m_len = elen;
1482 	return n;
1483 }
1484 #endif
1485 
1486 /*
1487  * Get pointer to the previous header followed by the header
1488  * currently processed.
1489  * XXX: This function supposes that
1490  *	M includes all headers,
1491  *	the next header field and the header length field of each header
1492  *	are valid, and
1493  *	the sum of each header length equals to OFF.
1494  * Because of these assumptions, this function must be called very
1495  * carefully. Moreover, it will not be used in the near future when
1496  * we develop `neater' mechanism to process extension headers.
1497  */
1498 char *
1499 ip6_get_prevhdr(struct mbuf *m, int off)
1500 {
1501 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1502 
1503 	if (off == sizeof(struct ip6_hdr))
1504 		return (&ip6->ip6_nxt);
1505 	else {
1506 		int len, nxt;
1507 		struct ip6_ext *ip6e = NULL;
1508 
1509 		nxt = ip6->ip6_nxt;
1510 		len = sizeof(struct ip6_hdr);
1511 		while (len < off) {
1512 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1513 
1514 			switch (nxt) {
1515 			case IPPROTO_FRAGMENT:
1516 				len += sizeof(struct ip6_frag);
1517 				break;
1518 			case IPPROTO_AH:
1519 				len += (ip6e->ip6e_len + 2) << 2;
1520 				break;
1521 			default:
1522 				len += (ip6e->ip6e_len + 1) << 3;
1523 				break;
1524 			}
1525 			nxt = ip6e->ip6e_nxt;
1526 		}
1527 		if (ip6e)
1528 			return (&ip6e->ip6e_nxt);
1529 		else
1530 			return NULL;
1531 	}
1532 }
1533 
1534 /*
1535  * get next header offset.  m will be retained.
1536  */
1537 int
1538 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1539 {
1540 	struct ip6_hdr ip6;
1541 	struct ip6_ext ip6e;
1542 	struct ip6_frag fh;
1543 
1544 	/* just in case */
1545 	if (m == NULL)
1546 		panic("ip6_nexthdr: m == NULL");
1547 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1548 		return -1;
1549 
1550 	switch (proto) {
1551 	case IPPROTO_IPV6:
1552 		if (m->m_pkthdr.len < off + sizeof(ip6))
1553 			return -1;
1554 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1555 		if (nxtp)
1556 			*nxtp = ip6.ip6_nxt;
1557 		off += sizeof(ip6);
1558 		return off;
1559 
1560 	case IPPROTO_FRAGMENT:
1561 		/*
1562 		 * terminate parsing if it is not the first fragment,
1563 		 * it does not make sense to parse through it.
1564 		 */
1565 		if (m->m_pkthdr.len < off + sizeof(fh))
1566 			return -1;
1567 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1568 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1569 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1570 			return -1;
1571 		if (nxtp)
1572 			*nxtp = fh.ip6f_nxt;
1573 		off += sizeof(struct ip6_frag);
1574 		return off;
1575 
1576 	case IPPROTO_AH:
1577 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1578 			return -1;
1579 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1580 		if (nxtp)
1581 			*nxtp = ip6e.ip6e_nxt;
1582 		off += (ip6e.ip6e_len + 2) << 2;
1583 		return off;
1584 
1585 	case IPPROTO_HOPOPTS:
1586 	case IPPROTO_ROUTING:
1587 	case IPPROTO_DSTOPTS:
1588 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1589 			return -1;
1590 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1591 		if (nxtp)
1592 			*nxtp = ip6e.ip6e_nxt;
1593 		off += (ip6e.ip6e_len + 1) << 3;
1594 		return off;
1595 
1596 	case IPPROTO_NONE:
1597 	case IPPROTO_ESP:
1598 	case IPPROTO_IPCOMP:
1599 		/* give up */
1600 		return -1;
1601 
1602 	default:
1603 		return -1;
1604 	}
1605 
1606 	return -1;
1607 }
1608 
1609 /*
1610  * get offset for the last header in the chain.  m will be kept untainted.
1611  */
1612 int
1613 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1614 {
1615 	int newoff;
1616 	int nxt;
1617 
1618 	if (!nxtp) {
1619 		nxt = -1;
1620 		nxtp = &nxt;
1621 	}
1622 	while (1) {
1623 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1624 		if (newoff < 0)
1625 			return off;
1626 		else if (newoff < off)
1627 			return -1;	/* invalid */
1628 		else if (newoff == off)
1629 			return newoff;
1630 
1631 		off = newoff;
1632 		proto = *nxtp;
1633 	}
1634 }
1635 
1636 struct ip6aux *
1637 ip6_addaux(struct mbuf *m)
1638 {
1639 	struct m_tag *mtag;
1640 
1641 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1642 	if (!mtag) {
1643 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1644 		    M_NOWAIT);
1645 		if (mtag) {
1646 			m_tag_prepend(m, mtag);
1647 			bzero(mtag + 1, sizeof(struct ip6aux));
1648 		}
1649 	}
1650 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1651 }
1652 
1653 struct ip6aux *
1654 ip6_findaux(struct mbuf *m)
1655 {
1656 	struct m_tag *mtag;
1657 
1658 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1659 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1660 }
1661 
1662 void
1663 ip6_delaux(struct mbuf *m)
1664 {
1665 	struct m_tag *mtag;
1666 
1667 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1668 	if (mtag)
1669 		m_tag_delete(m, mtag);
1670 }
1671 
1672 /*
1673  * System control for IP6
1674  */
1675 
1676 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1677 	0,		0,		0,		0,
1678 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1679 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1680 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1681 	0,		0,		0,		0,
1682 	ENOPROTOOPT
1683 };
1684