xref: /freebsd/sys/netinet6/ip6_input.c (revision f4f8f02054f3abb6ceb84aefcdecc78d5c8b462f)
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_size	= sizeof(struct vnet_inet6),
165 	.vmi_dependson	= VNET_MOD_INET	/* XXX revisit - TCP/UDP needs this? */
166 };
167 
168 static void
169 vnet_inet6_register(void)
170 {
171 
172 	vnet_mod_register(&vnet_inet6_modinfo);
173 }
174 
175 SYSINIT(inet6, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet6_register, 0);
176 #endif
177 
178 /*
179  * IP6 initialization: fill in IP6 protocol switch table.
180  * All protocols not implemented in kernel go to raw IP6 protocol handler.
181  */
182 void
183 ip6_init(void)
184 {
185 	INIT_VNET_INET6(curvnet);
186 	struct ip6protosw *pr;
187 	int i;
188 
189 	V_ip6qmaxlen = IFQ_MAXLEN;
190 	V_in6_maxmtu = 0;
191 #ifdef IP6_AUTO_LINKLOCAL
192 	V_ip6_auto_linklocal = IP6_AUTO_LINKLOCAL;
193 #else
194 	V_ip6_auto_linklocal = 1;	/* enable by default */
195 #endif
196 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
197 	    &V_ip6_auto_linklocal);
198 
199 #ifndef IPV6FORWARDING
200 #ifdef GATEWAY6
201 #define IPV6FORWARDING	1	/* forward IP6 packets not for us */
202 #else
203 #define IPV6FORWARDING	0	/* don't forward IP6 packets not for us */
204 #endif /* GATEWAY6 */
205 #endif /* !IPV6FORWARDING */
206 
207 #ifndef IPV6_SENDREDIRECTS
208 #define IPV6_SENDREDIRECTS	1
209 #endif
210 
211 	V_ip6_forwarding = IPV6FORWARDING; /* act as router? */
212 	V_ip6_sendredirects = IPV6_SENDREDIRECTS;
213 	V_ip6_defhlim = IPV6_DEFHLIM;
214 	V_ip6_defmcasthlim = IPV6_DEFAULT_MULTICAST_HOPS;
215 	V_ip6_accept_rtadv = 0;	 /* "IPV6FORWARDING ? 0 : 1" is dangerous */
216 	V_ip6_log_interval = 5;
217 	V_ip6_hdrnestlimit = 15; /* How many header options will we process? */
218 	V_ip6_dad_count = 1;	 /* DupAddrDetectionTransmits */
219 	V_ip6_auto_flowlabel = 1;
220 	V_ip6_use_deprecated = 1;/* allow deprecated addr (RFC2462 5.5.4) */
221 	V_ip6_rr_prune = 5;	 /* router renumbering prefix
222                                   * walk list every 5 sec. */
223 	V_ip6_mcast_pmtu = 0;	 /* enable pMTU discovery for multicast? */
224 	V_ip6_v6only = 1;
225 	V_ip6_keepfaith = 0;
226 	V_ip6_log_time = (time_t)0L;
227 #ifdef IPSTEALTH
228 	V_ip6stealth = 0;
229 #endif
230 	V_nd6_onlink_ns_rfc4861 = 0; /* allow 'on-link' nd6 NS (RFC 4861) */
231 
232 	V_pmtu_expire = 60*10;
233 	V_pmtu_probe = 60*2;
234 
235 	/* raw IP6 parameters */
236 	/*
237 	 * Nominal space allocated to a raw ip socket.
238 	 */
239 #define RIPV6SNDQ	8192
240 #define RIPV6RCVQ	8192
241 	V_rip6_sendspace = RIPV6SNDQ;
242 	V_rip6_recvspace = RIPV6RCVQ;
243 
244 	/* ICMPV6 parameters */
245 	V_icmp6_rediraccept = 1;	/* accept and process redirects */
246 	V_icmp6_redirtimeout = 10 * 60;	/* 10 minutes */
247 	V_icmp6errppslim = 100;		/* 100pps */
248 	/* control how to respond to NI queries */
249 	V_icmp6_nodeinfo = (ICMP6_NODEINFO_FQDNOK|ICMP6_NODEINFO_NODEADDROK);
250 
251 	/* UDP on IP6 parameters */
252 	V_udp6_sendspace = 9216;	/* really max datagram size */
253 	V_udp6_recvspace = 40 * (1024 + sizeof(struct sockaddr_in6));
254 					/* 40 1K datagrams */
255 	V_dad_init = 0;
256 
257 	scope6_init();
258 	addrsel_policy_init();
259 	nd6_init();
260 	frag6_init();
261 
262 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
263 
264 	/* Skip global initialization stuff for non-default instances. */
265 	if (!IS_DEFAULT_VNET(curvnet))
266 		return;
267 
268 #ifdef DIAGNOSTIC
269 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
270 		panic("sizeof(protosw) != sizeof(ip6protosw)");
271 #endif
272 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
273 	if (pr == NULL)
274 		panic("ip6_init");
275 
276 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
277 	for (i = 0; i < IPPROTO_MAX; i++)
278 		ip6_protox[i] = pr - inet6sw;
279 	/*
280 	 * Cycle through IP protocols and put them into the appropriate place
281 	 * in ip6_protox[].
282 	 */
283 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
284 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
285 		if (pr->pr_domain->dom_family == PF_INET6 &&
286 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
287 			/* Be careful to only index valid IP protocols. */
288 			if (pr->pr_protocol < IPPROTO_MAX)
289 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
290 		}
291 
292 	/* Initialize packet filter hooks. */
293 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
294 	inet6_pfil_hook.ph_af = AF_INET6;
295 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
296 		printf("%s: WARNING: unable to register pfil hook, "
297 			"error %d\n", __func__, i);
298 
299 	ip6intrq.ifq_maxlen = V_ip6qmaxlen; /* XXX */
300 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
301 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
302 }
303 
304 static int
305 ip6_init2_vnet(const void *unused __unused)
306 {
307 	INIT_VNET_INET6(curvnet);
308 
309 	/* nd6_timer_init */
310 	callout_init(&V_nd6_timer_ch, 0);
311 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
312 
313 	/* timer for regeneranation of temporary addresses randomize ID */
314 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
315 	callout_reset(&V_in6_tmpaddrtimer_ch,
316 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
317 		       V_ip6_temp_regen_advance) * hz,
318 		      in6_tmpaddrtimer, curvnet);
319 
320 	return (0);
321 }
322 
323 static void
324 ip6_init2(void *dummy)
325 {
326 
327 	ip6_init2_vnet(NULL);
328 }
329 
330 /* cheat */
331 /* This must be after route_init(), which is now SI_ORDER_THIRD */
332 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
333 
334 void
335 ip6_input(struct mbuf *m)
336 {
337 	INIT_VNET_NET(curvnet);
338 	INIT_VNET_INET6(curvnet);
339 	struct ip6_hdr *ip6;
340 	int off = sizeof(struct ip6_hdr), nest;
341 	u_int32_t plen;
342 	u_int32_t rtalert = ~0;
343 	int nxt, ours = 0;
344 	struct ifnet *deliverifp = NULL, *ifp = NULL;
345 	struct in6_addr odst;
346 	struct route_in6 rin6;
347 	int srcrt = 0;
348 	struct llentry *lle = NULL;
349 	struct sockaddr_in6 dst6, *dst;
350 
351 	bzero(&rin6, sizeof(struct route_in6));
352 #ifdef IPSEC
353 	/*
354 	 * should the inner packet be considered authentic?
355 	 * see comment in ah4_input().
356 	 * NB: m cannot be NULL when passed to the input routine
357 	 */
358 
359 	m->m_flags &= ~M_AUTHIPHDR;
360 	m->m_flags &= ~M_AUTHIPDGM;
361 
362 #endif /* IPSEC */
363 
364 	/*
365 	 * make sure we don't have onion peering information into m_tag.
366 	 */
367 	ip6_delaux(m);
368 
369 	/*
370 	 * mbuf statistics
371 	 */
372 	if (m->m_flags & M_EXT) {
373 		if (m->m_next)
374 			V_ip6stat.ip6s_mext2m++;
375 		else
376 			V_ip6stat.ip6s_mext1++;
377 	} else {
378 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
379 		if (m->m_next) {
380 			if (m->m_flags & M_LOOP) {
381 				V_ip6stat.ip6s_m2m[V_loif[0].if_index]++; /* XXX */
382 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
383 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
384 			else
385 				V_ip6stat.ip6s_m2m[0]++;
386 		} else
387 			V_ip6stat.ip6s_m1++;
388 #undef M2MMAX
389 	}
390 
391 	/* drop the packet if IPv6 operation is disabled on the IF */
392 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
393 		m_freem(m);
394 		return;
395 	}
396 
397 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
398 	V_ip6stat.ip6s_total++;
399 
400 #ifndef PULLDOWN_TEST
401 	/*
402 	 * L2 bridge code and some other code can return mbuf chain
403 	 * that does not conform to KAME requirement.  too bad.
404 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
405 	 */
406 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
407 		struct mbuf *n;
408 
409 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
410 		if (n)
411 			M_MOVE_PKTHDR(n, m);
412 		if (n && n->m_pkthdr.len > MHLEN) {
413 			MCLGET(n, M_DONTWAIT);
414 			if ((n->m_flags & M_EXT) == 0) {
415 				m_freem(n);
416 				n = NULL;
417 			}
418 		}
419 		if (n == NULL) {
420 			m_freem(m);
421 			return;	/* ENOBUFS */
422 		}
423 
424 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
425 		n->m_len = n->m_pkthdr.len;
426 		m_freem(m);
427 		m = n;
428 	}
429 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
430 #endif
431 
432 	if (m->m_len < sizeof(struct ip6_hdr)) {
433 		struct ifnet *inifp;
434 		inifp = m->m_pkthdr.rcvif;
435 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
436 			V_ip6stat.ip6s_toosmall++;
437 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
438 			return;
439 		}
440 	}
441 
442 	ip6 = mtod(m, struct ip6_hdr *);
443 
444 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
445 		V_ip6stat.ip6s_badvers++;
446 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
447 		goto bad;
448 	}
449 
450 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
451 
452 	/*
453 	 * Check against address spoofing/corruption.
454 	 */
455 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
456 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
457 		/*
458 		 * XXX: "badscope" is not very suitable for a multicast source.
459 		 */
460 		V_ip6stat.ip6s_badscope++;
461 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
462 		goto bad;
463 	}
464 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
465 	    !(m->m_flags & M_LOOP)) {
466 		/*
467 		 * In this case, the packet should come from the loopback
468 		 * interface.  However, we cannot just check the if_flags,
469 		 * because ip6_mloopback() passes the "actual" interface
470 		 * as the outgoing/incoming interface.
471 		 */
472 		V_ip6stat.ip6s_badscope++;
473 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
474 		goto bad;
475 	}
476 
477 #ifdef ALTQ
478 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
479 		/* packet is dropped by traffic conditioner */
480 		return;
481 	}
482 #endif
483 	/*
484 	 * The following check is not documented in specs.  A malicious
485 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
486 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
487 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
488 	 *
489 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
490 	 * support IPv4-less kernel compilation, we cannot support SIIT
491 	 * environment at all.  So, it makes more sense for us to reject any
492 	 * malicious packets for non-SIIT environment, than try to do a
493 	 * partial support for SIIT environment.
494 	 */
495 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
496 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
497 		V_ip6stat.ip6s_badscope++;
498 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
499 		goto bad;
500 	}
501 #if 0
502 	/*
503 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
504 	 *
505 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
506 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
507 	 * is revised to forbid relaying case.
508 	 */
509 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
510 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
511 		V_ip6stat.ip6s_badscope++;
512 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
513 		goto bad;
514 	}
515 #endif
516 
517 	/*
518 	 * Run through list of hooks for input packets.
519 	 *
520 	 * NB: Beware of the destination address changing
521 	 *     (e.g. by NAT rewriting).  When this happens,
522 	 *     tell ip6_forward to do the right thing.
523 	 */
524 	odst = ip6->ip6_dst;
525 
526 	/* Jump over all PFIL processing if hooks are not active. */
527 	if (!PFIL_HOOKED(&inet6_pfil_hook))
528 		goto passin;
529 
530 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
531 		return;
532 	if (m == NULL)			/* consumed by filter */
533 		return;
534 	ip6 = mtod(m, struct ip6_hdr *);
535 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
536 
537 passin:
538 	/*
539 	 * Disambiguate address scope zones (if there is ambiguity).
540 	 * We first make sure that the original source or destination address
541 	 * is not in our internal form for scoped addresses.  Such addresses
542 	 * are not necessarily invalid spec-wise, but we cannot accept them due
543 	 * to the usage conflict.
544 	 * in6_setscope() then also checks and rejects the cases where src or
545 	 * dst are the loopback address and the receiving interface
546 	 * is not loopback.
547 	 */
548 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
549 		V_ip6stat.ip6s_badscope++; /* XXX */
550 		goto bad;
551 	}
552 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
553 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
554 		V_ip6stat.ip6s_badscope++;
555 		goto bad;
556 	}
557 
558 	/*
559 	 * Multicast check. Assume packet is for us to avoid
560 	 * prematurely taking locks.
561 	 */
562 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
563 		ours = 1;
564 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
565 		deliverifp = m->m_pkthdr.rcvif;
566 		goto hbhcheck;
567 	}
568 
569 	/*
570 	 *  Unicast check
571 	 */
572 
573 	bzero(&dst6, sizeof(dst6));
574 	dst6.sin6_family = AF_INET6;
575 	dst6.sin6_len = sizeof(struct sockaddr_in6);
576 	dst6.sin6_addr = ip6->ip6_dst;
577 	ifp = m->m_pkthdr.rcvif;
578 	IF_AFDATA_LOCK(ifp);
579 	lle = lla_lookup(LLTABLE6(ifp), 0,
580 	     (struct sockaddr *)&dst6);
581 	IF_AFDATA_UNLOCK(ifp);
582 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
583 		ours = 1;
584 		deliverifp = ifp;
585 		LLE_RUNLOCK(lle);
586 		goto hbhcheck;
587 	}
588 	if (lle != NULL)
589 		LLE_RUNLOCK(lle);
590 
591 	dst = &rin6.ro_dst;
592 	dst->sin6_len = sizeof(struct sockaddr_in6);
593 	dst->sin6_family = AF_INET6;
594 	dst->sin6_addr = ip6->ip6_dst;
595 	rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0);
596 	if (rin6.ro_rt)
597 		RT_UNLOCK(rin6.ro_rt);
598 
599 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
600 
601 	/*
602 	 * Accept the packet if the forwarding interface to the destination
603 	 * according to the routing table is the loopback interface,
604 	 * unless the associated route has a gateway.
605 	 * Note that this approach causes to accept a packet if there is a
606 	 * route to the loopback interface for the destination of the packet.
607 	 * But we think it's even useful in some situations, e.g. when using
608 	 * a special daemon which wants to intercept the packet.
609 	 *
610 	 * XXX: some OSes automatically make a cloned route for the destination
611 	 * of an outgoing packet.  If the outgoing interface of the packet
612 	 * is a loopback one, the kernel would consider the packet to be
613 	 * accepted, even if we have no such address assinged on the interface.
614 	 * We check the cloned flag of the route entry to reject such cases,
615 	 * assuming that route entries for our own addresses are not made by
616 	 * cloning (it should be true because in6_addloop explicitly installs
617 	 * the host route).  However, we might have to do an explicit check
618 	 * while it would be less efficient.  Or, should we rather install a
619 	 * reject route for such a case?
620 	 */
621 	if (rin6.ro_rt &&
622 	    (rin6.ro_rt->rt_flags &
623 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
624 #ifdef RTF_WASCLONED
625 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
626 #endif
627 #ifdef RTF_CLONED
628 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
629 #endif
630 #if 0
631 	    /*
632 	     * The check below is redundant since the comparison of
633 	     * the destination and the key of the rtentry has
634 	     * already done through looking up the routing table.
635 	     */
636 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
637 	    &rt6_key(rin6.ro_rt)->sin6_addr)
638 #endif
639 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
640 		struct in6_ifaddr *ia6 =
641 			(struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
642 
643 		/*
644 		 * record address information into m_tag.
645 		 */
646 		(void)ip6_setdstifaddr(m, ia6);
647 
648 		/*
649 		 * packets to a tentative, duplicated, or somehow invalid
650 		 * address must not be accepted.
651 		 */
652 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
653 			/* this address is ready */
654 			ours = 1;
655 			deliverifp = ia6->ia_ifp;	/* correct? */
656 			/* Count the packet in the ip address stats */
657 			ia6->ia_ifa.if_ipackets++;
658 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
659 			goto hbhcheck;
660 		} else {
661 			char ip6bufs[INET6_ADDRSTRLEN];
662 			char ip6bufd[INET6_ADDRSTRLEN];
663 			/* address is not ready, so discard the packet. */
664 			nd6log((LOG_INFO,
665 			    "ip6_input: packet to an unready address %s->%s\n",
666 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
667 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
668 
669 			goto bad;
670 		}
671 	}
672 
673 	/*
674 	 * FAITH (Firewall Aided Internet Translator)
675 	 */
676 	if (V_ip6_keepfaith) {
677 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
678 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
679 			/* XXX do we need more sanity checks? */
680 			ours = 1;
681 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
682 			goto hbhcheck;
683 		}
684 	}
685 
686 	/*
687 	 * Now there is no reason to process the packet if it's not our own
688 	 * and we're not a router.
689 	 */
690 	if (!V_ip6_forwarding) {
691 		V_ip6stat.ip6s_cantforward++;
692 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
693 		goto bad;
694 	}
695 
696   hbhcheck:
697 	/*
698 	 * record address information into m_tag, if we don't have one yet.
699 	 * note that we are unable to record it, if the address is not listed
700 	 * as our interface address (e.g. multicast addresses, addresses
701 	 * within FAITH prefixes and such).
702 	 */
703 	if (deliverifp && !ip6_getdstifaddr(m)) {
704 		struct in6_ifaddr *ia6;
705 
706 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
707 		if (ia6) {
708 			if (!ip6_setdstifaddr(m, ia6)) {
709 				/*
710 				 * XXX maybe we should drop the packet here,
711 				 * as we could not provide enough information
712 				 * to the upper layers.
713 				 */
714 			}
715 		}
716 	}
717 
718 	/*
719 	 * Process Hop-by-Hop options header if it's contained.
720 	 * m may be modified in ip6_hopopts_input().
721 	 * If a JumboPayload option is included, plen will also be modified.
722 	 */
723 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
724 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
725 		struct ip6_hbh *hbh;
726 
727 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
728 #if 0	/*touches NULL pointer*/
729 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
730 #endif
731 			goto out;	/* m have already been freed */
732 		}
733 
734 		/* adjust pointer */
735 		ip6 = mtod(m, struct ip6_hdr *);
736 
737 		/*
738 		 * if the payload length field is 0 and the next header field
739 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
740 		 * option MUST be included.
741 		 */
742 		if (ip6->ip6_plen == 0 && plen == 0) {
743 			/*
744 			 * Note that if a valid jumbo payload option is
745 			 * contained, ip6_hopopts_input() must set a valid
746 			 * (non-zero) payload length to the variable plen.
747 			 */
748 			V_ip6stat.ip6s_badoptions++;
749 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
750 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
751 			icmp6_error(m, ICMP6_PARAM_PROB,
752 				    ICMP6_PARAMPROB_HEADER,
753 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
754 			goto out;
755 		}
756 #ifndef PULLDOWN_TEST
757 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
758 		hbh = (struct ip6_hbh *)(ip6 + 1);
759 #else
760 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
761 			sizeof(struct ip6_hbh));
762 		if (hbh == NULL) {
763 			V_ip6stat.ip6s_tooshort++;
764 			goto out;
765 		}
766 #endif
767 		nxt = hbh->ip6h_nxt;
768 
769 		/*
770 		 * If we are acting as a router and the packet contains a
771 		 * router alert option, see if we know the option value.
772 		 * Currently, we only support the option value for MLD, in which
773 		 * case we should pass the packet to the multicast routing
774 		 * daemon.
775 		 */
776 		if (rtalert != ~0 && V_ip6_forwarding) {
777 			switch (rtalert) {
778 			case IP6OPT_RTALERT_MLD:
779 				ours = 1;
780 				break;
781 			default:
782 				/*
783 				 * RFC2711 requires unrecognized values must be
784 				 * silently ignored.
785 				 */
786 				break;
787 			}
788 		}
789 	} else
790 		nxt = ip6->ip6_nxt;
791 
792 	/*
793 	 * Check that the amount of data in the buffers
794 	 * is as at least much as the IPv6 header would have us expect.
795 	 * Trim mbufs if longer than we expect.
796 	 * Drop packet if shorter than we expect.
797 	 */
798 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
799 		V_ip6stat.ip6s_tooshort++;
800 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
801 		goto bad;
802 	}
803 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
804 		if (m->m_len == m->m_pkthdr.len) {
805 			m->m_len = sizeof(struct ip6_hdr) + plen;
806 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
807 		} else
808 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
809 	}
810 
811 	/*
812 	 * Forward if desirable.
813 	 */
814 	if (V_ip6_mrouter &&
815 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
816 		/*
817 		 * If we are acting as a multicast router, all
818 		 * incoming multicast packets are passed to the
819 		 * kernel-level multicast forwarding function.
820 		 * The packet is returned (relatively) intact; if
821 		 * ip6_mforward() returns a non-zero value, the packet
822 		 * must be discarded, else it may be accepted below.
823 		 */
824 		if (ip6_mforward &&
825 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
826 			IP6STAT_INC(ip6s_cantforward);
827 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
828 			goto bad;
829 		}
830 	} else if (!ours) {
831 		ip6_forward(m, srcrt);
832 		goto out;
833 	}
834 
835 	ip6 = mtod(m, struct ip6_hdr *);
836 
837 	/*
838 	 * Malicious party may be able to use IPv4 mapped addr to confuse
839 	 * tcp/udp stack and bypass security checks (act as if it was from
840 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
841 	 *
842 	 * For SIIT end node behavior, you may want to disable the check.
843 	 * However, you will  become vulnerable to attacks using IPv4 mapped
844 	 * source.
845 	 */
846 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
847 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
848 		V_ip6stat.ip6s_badscope++;
849 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
850 		goto bad;
851 	}
852 
853 	/*
854 	 * Tell launch routine the next header
855 	 */
856 	V_ip6stat.ip6s_delivered++;
857 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
858 	nest = 0;
859 
860 	while (nxt != IPPROTO_DONE) {
861 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
862 			V_ip6stat.ip6s_toomanyhdr++;
863 			goto bad;
864 		}
865 
866 		/*
867 		 * protection against faulty packet - there should be
868 		 * more sanity checks in header chain processing.
869 		 */
870 		if (m->m_pkthdr.len < off) {
871 			V_ip6stat.ip6s_tooshort++;
872 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
873 			goto bad;
874 		}
875 
876 #ifdef IPSEC
877 		/*
878 		 * enforce IPsec policy checking if we are seeing last header.
879 		 * note that we do not visit this with protocols with pcb layer
880 		 * code - like udp/tcp/raw ip.
881 		 */
882 		if (ip6_ipsec_input(m, nxt))
883 			goto bad;
884 #endif /* IPSEC */
885 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
886 	}
887 	goto out;
888 bad:
889 	m_freem(m);
890 out:
891 	if (rin6.ro_rt)
892 		RTFREE(rin6.ro_rt);
893 }
894 
895 /*
896  * set/grab in6_ifaddr correspond to IPv6 destination address.
897  * XXX backward compatibility wrapper
898  */
899 static struct ip6aux *
900 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
901 {
902 	struct ip6aux *ip6a;
903 
904 	ip6a = ip6_addaux(m);
905 	if (ip6a)
906 		ip6a->ip6a_dstia6 = ia6;
907 	return ip6a;	/* NULL if failed to set */
908 }
909 
910 struct in6_ifaddr *
911 ip6_getdstifaddr(struct mbuf *m)
912 {
913 	struct ip6aux *ip6a;
914 
915 	ip6a = ip6_findaux(m);
916 	if (ip6a)
917 		return ip6a->ip6a_dstia6;
918 	else
919 		return NULL;
920 }
921 
922 /*
923  * Hop-by-Hop options header processing. If a valid jumbo payload option is
924  * included, the real payload length will be stored in plenp.
925  *
926  * rtalertp - XXX: should be stored more smart way
927  */
928 static int
929 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
930     struct mbuf **mp, int *offp)
931 {
932 	INIT_VNET_INET6(curvnet);
933 	struct mbuf *m = *mp;
934 	int off = *offp, hbhlen;
935 	struct ip6_hbh *hbh;
936 	u_int8_t *opt;
937 
938 	/* validation of the length of the header */
939 #ifndef PULLDOWN_TEST
940 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
941 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
942 	hbhlen = (hbh->ip6h_len + 1) << 3;
943 
944 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
945 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
946 #else
947 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
948 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
949 	if (hbh == NULL) {
950 		V_ip6stat.ip6s_tooshort++;
951 		return -1;
952 	}
953 	hbhlen = (hbh->ip6h_len + 1) << 3;
954 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
955 		hbhlen);
956 	if (hbh == NULL) {
957 		V_ip6stat.ip6s_tooshort++;
958 		return -1;
959 	}
960 #endif
961 	off += hbhlen;
962 	hbhlen -= sizeof(struct ip6_hbh);
963 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
964 
965 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
966 				hbhlen, rtalertp, plenp) < 0)
967 		return (-1);
968 
969 	*offp = off;
970 	*mp = m;
971 	return (0);
972 }
973 
974 /*
975  * Search header for all Hop-by-hop options and process each option.
976  * This function is separate from ip6_hopopts_input() in order to
977  * handle a case where the sending node itself process its hop-by-hop
978  * options header. In such a case, the function is called from ip6_output().
979  *
980  * The function assumes that hbh header is located right after the IPv6 header
981  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
982  * opthead + hbhlen is located in continuous memory region.
983  */
984 int
985 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
986     u_int32_t *rtalertp, u_int32_t *plenp)
987 {
988 	INIT_VNET_INET6(curvnet);
989 	struct ip6_hdr *ip6;
990 	int optlen = 0;
991 	u_int8_t *opt = opthead;
992 	u_int16_t rtalert_val;
993 	u_int32_t jumboplen;
994 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
995 
996 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
997 		switch (*opt) {
998 		case IP6OPT_PAD1:
999 			optlen = 1;
1000 			break;
1001 		case IP6OPT_PADN:
1002 			if (hbhlen < IP6OPT_MINLEN) {
1003 				V_ip6stat.ip6s_toosmall++;
1004 				goto bad;
1005 			}
1006 			optlen = *(opt + 1) + 2;
1007 			break;
1008 		case IP6OPT_ROUTER_ALERT:
1009 			/* XXX may need check for alignment */
1010 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1011 				V_ip6stat.ip6s_toosmall++;
1012 				goto bad;
1013 			}
1014 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1015 				/* XXX stat */
1016 				icmp6_error(m, ICMP6_PARAM_PROB,
1017 				    ICMP6_PARAMPROB_HEADER,
1018 				    erroff + opt + 1 - opthead);
1019 				return (-1);
1020 			}
1021 			optlen = IP6OPT_RTALERT_LEN;
1022 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1023 			*rtalertp = ntohs(rtalert_val);
1024 			break;
1025 		case IP6OPT_JUMBO:
1026 			/* XXX may need check for alignment */
1027 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1028 				V_ip6stat.ip6s_toosmall++;
1029 				goto bad;
1030 			}
1031 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1032 				/* XXX stat */
1033 				icmp6_error(m, ICMP6_PARAM_PROB,
1034 				    ICMP6_PARAMPROB_HEADER,
1035 				    erroff + opt + 1 - opthead);
1036 				return (-1);
1037 			}
1038 			optlen = IP6OPT_JUMBO_LEN;
1039 
1040 			/*
1041 			 * IPv6 packets that have non 0 payload length
1042 			 * must not contain a jumbo payload option.
1043 			 */
1044 			ip6 = mtod(m, struct ip6_hdr *);
1045 			if (ip6->ip6_plen) {
1046 				V_ip6stat.ip6s_badoptions++;
1047 				icmp6_error(m, ICMP6_PARAM_PROB,
1048 				    ICMP6_PARAMPROB_HEADER,
1049 				    erroff + opt - opthead);
1050 				return (-1);
1051 			}
1052 
1053 			/*
1054 			 * We may see jumbolen in unaligned location, so
1055 			 * we'd need to perform bcopy().
1056 			 */
1057 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1058 			jumboplen = (u_int32_t)htonl(jumboplen);
1059 
1060 #if 1
1061 			/*
1062 			 * if there are multiple jumbo payload options,
1063 			 * *plenp will be non-zero and the packet will be
1064 			 * rejected.
1065 			 * the behavior may need some debate in ipngwg -
1066 			 * multiple options does not make sense, however,
1067 			 * there's no explicit mention in specification.
1068 			 */
1069 			if (*plenp != 0) {
1070 				V_ip6stat.ip6s_badoptions++;
1071 				icmp6_error(m, ICMP6_PARAM_PROB,
1072 				    ICMP6_PARAMPROB_HEADER,
1073 				    erroff + opt + 2 - opthead);
1074 				return (-1);
1075 			}
1076 #endif
1077 
1078 			/*
1079 			 * jumbo payload length must be larger than 65535.
1080 			 */
1081 			if (jumboplen <= IPV6_MAXPACKET) {
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 			*plenp = jumboplen;
1089 
1090 			break;
1091 		default:		/* unknown option */
1092 			if (hbhlen < IP6OPT_MINLEN) {
1093 				V_ip6stat.ip6s_toosmall++;
1094 				goto bad;
1095 			}
1096 			optlen = ip6_unknown_opt(opt, m,
1097 			    erroff + opt - opthead);
1098 			if (optlen == -1)
1099 				return (-1);
1100 			optlen += 2;
1101 			break;
1102 		}
1103 	}
1104 
1105 	return (0);
1106 
1107   bad:
1108 	m_freem(m);
1109 	return (-1);
1110 }
1111 
1112 /*
1113  * Unknown option processing.
1114  * The third argument `off' is the offset from the IPv6 header to the option,
1115  * which is necessary if the IPv6 header the and option header and IPv6 header
1116  * is not continuous in order to return an ICMPv6 error.
1117  */
1118 int
1119 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1120 {
1121 	INIT_VNET_INET6(curvnet);
1122 	struct ip6_hdr *ip6;
1123 
1124 	switch (IP6OPT_TYPE(*optp)) {
1125 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1126 		return ((int)*(optp + 1));
1127 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1128 		m_freem(m);
1129 		return (-1);
1130 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1131 		V_ip6stat.ip6s_badoptions++;
1132 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1133 		return (-1);
1134 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1135 		V_ip6stat.ip6s_badoptions++;
1136 		ip6 = mtod(m, struct ip6_hdr *);
1137 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1138 		    (m->m_flags & (M_BCAST|M_MCAST)))
1139 			m_freem(m);
1140 		else
1141 			icmp6_error(m, ICMP6_PARAM_PROB,
1142 				    ICMP6_PARAMPROB_OPTION, off);
1143 		return (-1);
1144 	}
1145 
1146 	m_freem(m);		/* XXX: NOTREACHED */
1147 	return (-1);
1148 }
1149 
1150 /*
1151  * Create the "control" list for this pcb.
1152  * These functions will not modify mbuf chain at all.
1153  *
1154  * With KAME mbuf chain restriction:
1155  * The routine will be called from upper layer handlers like tcp6_input().
1156  * Thus the routine assumes that the caller (tcp6_input) have already
1157  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1158  * very first mbuf on the mbuf chain.
1159  *
1160  * ip6_savecontrol_v4 will handle those options that are possible to be
1161  * set on a v4-mapped socket.
1162  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1163  * options and handle the v6-only ones itself.
1164  */
1165 struct mbuf **
1166 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1167     int *v4only)
1168 {
1169 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1170 
1171 #ifdef SO_TIMESTAMP
1172 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1173 		struct timeval tv;
1174 
1175 		microtime(&tv);
1176 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1177 		    SCM_TIMESTAMP, SOL_SOCKET);
1178 		if (*mp)
1179 			mp = &(*mp)->m_next;
1180 	}
1181 #endif
1182 
1183 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1184 		if (v4only != NULL)
1185 			*v4only = 1;
1186 		return (mp);
1187 	}
1188 
1189 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1190 	/* RFC 2292 sec. 5 */
1191 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1192 		struct in6_pktinfo pi6;
1193 
1194 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1195 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1196 		pi6.ipi6_ifindex =
1197 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1198 
1199 		*mp = sbcreatecontrol((caddr_t) &pi6,
1200 		    sizeof(struct in6_pktinfo),
1201 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1202 		if (*mp)
1203 			mp = &(*mp)->m_next;
1204 	}
1205 
1206 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1207 		int hlim = ip6->ip6_hlim & 0xff;
1208 
1209 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1210 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1211 		    IPPROTO_IPV6);
1212 		if (*mp)
1213 			mp = &(*mp)->m_next;
1214 	}
1215 
1216 	if (v4only != NULL)
1217 		*v4only = 0;
1218 	return (mp);
1219 }
1220 
1221 void
1222 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1223 {
1224 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1225 	int v4only = 0;
1226 
1227 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1228 	if (v4only)
1229 		return;
1230 
1231 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1232 		u_int32_t flowinfo;
1233 		int tclass;
1234 
1235 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1236 		flowinfo >>= 20;
1237 
1238 		tclass = flowinfo & 0xff;
1239 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1240 		    IPV6_TCLASS, IPPROTO_IPV6);
1241 		if (*mp)
1242 			mp = &(*mp)->m_next;
1243 	}
1244 
1245 	/*
1246 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1247 	 * privilege for the option (see ip6_ctloutput), but it might be too
1248 	 * strict, since there might be some hop-by-hop options which can be
1249 	 * returned to normal user.
1250 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1251 	 */
1252 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1253 		/*
1254 		 * Check if a hop-by-hop options header is contatined in the
1255 		 * received packet, and if so, store the options as ancillary
1256 		 * data. Note that a hop-by-hop options header must be
1257 		 * just after the IPv6 header, which is assured through the
1258 		 * IPv6 input processing.
1259 		 */
1260 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1261 			struct ip6_hbh *hbh;
1262 			int hbhlen = 0;
1263 #ifdef PULLDOWN_TEST
1264 			struct mbuf *ext;
1265 #endif
1266 
1267 #ifndef PULLDOWN_TEST
1268 			hbh = (struct ip6_hbh *)(ip6 + 1);
1269 			hbhlen = (hbh->ip6h_len + 1) << 3;
1270 #else
1271 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1272 			    ip6->ip6_nxt);
1273 			if (ext == NULL) {
1274 				V_ip6stat.ip6s_tooshort++;
1275 				return;
1276 			}
1277 			hbh = mtod(ext, struct ip6_hbh *);
1278 			hbhlen = (hbh->ip6h_len + 1) << 3;
1279 			if (hbhlen != ext->m_len) {
1280 				m_freem(ext);
1281 				V_ip6stat.ip6s_tooshort++;
1282 				return;
1283 			}
1284 #endif
1285 
1286 			/*
1287 			 * XXX: We copy the whole header even if a
1288 			 * jumbo payload option is included, the option which
1289 			 * is to be removed before returning according to
1290 			 * RFC2292.
1291 			 * Note: this constraint is removed in RFC3542
1292 			 */
1293 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1294 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1295 			    IPPROTO_IPV6);
1296 			if (*mp)
1297 				mp = &(*mp)->m_next;
1298 #ifdef PULLDOWN_TEST
1299 			m_freem(ext);
1300 #endif
1301 		}
1302 	}
1303 
1304 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1305 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1306 
1307 		/*
1308 		 * Search for destination options headers or routing
1309 		 * header(s) through the header chain, and stores each
1310 		 * header as ancillary data.
1311 		 * Note that the order of the headers remains in
1312 		 * the chain of ancillary data.
1313 		 */
1314 		while (1) {	/* is explicit loop prevention necessary? */
1315 			struct ip6_ext *ip6e = NULL;
1316 			int elen;
1317 #ifdef PULLDOWN_TEST
1318 			struct mbuf *ext = NULL;
1319 #endif
1320 
1321 			/*
1322 			 * if it is not an extension header, don't try to
1323 			 * pull it from the chain.
1324 			 */
1325 			switch (nxt) {
1326 			case IPPROTO_DSTOPTS:
1327 			case IPPROTO_ROUTING:
1328 			case IPPROTO_HOPOPTS:
1329 			case IPPROTO_AH: /* is it possible? */
1330 				break;
1331 			default:
1332 				goto loopend;
1333 			}
1334 
1335 #ifndef PULLDOWN_TEST
1336 			if (off + sizeof(*ip6e) > m->m_len)
1337 				goto loopend;
1338 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1339 			if (nxt == IPPROTO_AH)
1340 				elen = (ip6e->ip6e_len + 2) << 2;
1341 			else
1342 				elen = (ip6e->ip6e_len + 1) << 3;
1343 			if (off + elen > m->m_len)
1344 				goto loopend;
1345 #else
1346 			ext = ip6_pullexthdr(m, off, nxt);
1347 			if (ext == NULL) {
1348 				V_ip6stat.ip6s_tooshort++;
1349 				return;
1350 			}
1351 			ip6e = mtod(ext, struct ip6_ext *);
1352 			if (nxt == IPPROTO_AH)
1353 				elen = (ip6e->ip6e_len + 2) << 2;
1354 			else
1355 				elen = (ip6e->ip6e_len + 1) << 3;
1356 			if (elen != ext->m_len) {
1357 				m_freem(ext);
1358 				V_ip6stat.ip6s_tooshort++;
1359 				return;
1360 			}
1361 #endif
1362 
1363 			switch (nxt) {
1364 			case IPPROTO_DSTOPTS:
1365 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1366 					break;
1367 
1368 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1369 				    IS2292(in6p,
1370 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1371 				    IPPROTO_IPV6);
1372 				if (*mp)
1373 					mp = &(*mp)->m_next;
1374 				break;
1375 			case IPPROTO_ROUTING:
1376 				if (!in6p->inp_flags & IN6P_RTHDR)
1377 					break;
1378 
1379 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1380 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1381 				    IPPROTO_IPV6);
1382 				if (*mp)
1383 					mp = &(*mp)->m_next;
1384 				break;
1385 			case IPPROTO_HOPOPTS:
1386 			case IPPROTO_AH: /* is it possible? */
1387 				break;
1388 
1389 			default:
1390 				/*
1391 				 * other cases have been filtered in the above.
1392 				 * none will visit this case.  here we supply
1393 				 * the code just in case (nxt overwritten or
1394 				 * other cases).
1395 				 */
1396 #ifdef PULLDOWN_TEST
1397 				m_freem(ext);
1398 #endif
1399 				goto loopend;
1400 
1401 			}
1402 
1403 			/* proceed with the next header. */
1404 			off += elen;
1405 			nxt = ip6e->ip6e_nxt;
1406 			ip6e = NULL;
1407 #ifdef PULLDOWN_TEST
1408 			m_freem(ext);
1409 			ext = NULL;
1410 #endif
1411 		}
1412 	  loopend:
1413 		;
1414 	}
1415 }
1416 #undef IS2292
1417 
1418 void
1419 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1420 {
1421 	struct socket *so;
1422 	struct mbuf *m_mtu;
1423 	struct ip6_mtuinfo mtuctl;
1424 
1425 	so =  in6p->inp_socket;
1426 
1427 	if (mtu == NULL)
1428 		return;
1429 
1430 #ifdef DIAGNOSTIC
1431 	if (so == NULL)		/* I believe this is impossible */
1432 		panic("ip6_notify_pmtu: socket is NULL");
1433 #endif
1434 
1435 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1436 	mtuctl.ip6m_mtu = *mtu;
1437 	mtuctl.ip6m_addr = *dst;
1438 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1439 		return;
1440 
1441 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1442 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1443 		return;
1444 
1445 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1446 	    == 0) {
1447 		m_freem(m_mtu);
1448 		/* XXX: should count statistics */
1449 	} else
1450 		sorwakeup(so);
1451 
1452 	return;
1453 }
1454 
1455 #ifdef PULLDOWN_TEST
1456 /*
1457  * pull single extension header from mbuf chain.  returns single mbuf that
1458  * contains the result, or NULL on error.
1459  */
1460 static struct mbuf *
1461 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1462 {
1463 	struct ip6_ext ip6e;
1464 	size_t elen;
1465 	struct mbuf *n;
1466 
1467 #ifdef DIAGNOSTIC
1468 	switch (nxt) {
1469 	case IPPROTO_DSTOPTS:
1470 	case IPPROTO_ROUTING:
1471 	case IPPROTO_HOPOPTS:
1472 	case IPPROTO_AH: /* is it possible? */
1473 		break;
1474 	default:
1475 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1476 	}
1477 #endif
1478 
1479 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1480 	if (nxt == IPPROTO_AH)
1481 		elen = (ip6e.ip6e_len + 2) << 2;
1482 	else
1483 		elen = (ip6e.ip6e_len + 1) << 3;
1484 
1485 	MGET(n, M_DONTWAIT, MT_DATA);
1486 	if (n && elen >= MLEN) {
1487 		MCLGET(n, M_DONTWAIT);
1488 		if ((n->m_flags & M_EXT) == 0) {
1489 			m_free(n);
1490 			n = NULL;
1491 		}
1492 	}
1493 	if (!n)
1494 		return NULL;
1495 
1496 	n->m_len = 0;
1497 	if (elen >= M_TRAILINGSPACE(n)) {
1498 		m_free(n);
1499 		return NULL;
1500 	}
1501 
1502 	m_copydata(m, off, elen, mtod(n, caddr_t));
1503 	n->m_len = elen;
1504 	return n;
1505 }
1506 #endif
1507 
1508 /*
1509  * Get pointer to the previous header followed by the header
1510  * currently processed.
1511  * XXX: This function supposes that
1512  *	M includes all headers,
1513  *	the next header field and the header length field of each header
1514  *	are valid, and
1515  *	the sum of each header length equals to OFF.
1516  * Because of these assumptions, this function must be called very
1517  * carefully. Moreover, it will not be used in the near future when
1518  * we develop `neater' mechanism to process extension headers.
1519  */
1520 char *
1521 ip6_get_prevhdr(struct mbuf *m, int off)
1522 {
1523 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1524 
1525 	if (off == sizeof(struct ip6_hdr))
1526 		return (&ip6->ip6_nxt);
1527 	else {
1528 		int len, nxt;
1529 		struct ip6_ext *ip6e = NULL;
1530 
1531 		nxt = ip6->ip6_nxt;
1532 		len = sizeof(struct ip6_hdr);
1533 		while (len < off) {
1534 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1535 
1536 			switch (nxt) {
1537 			case IPPROTO_FRAGMENT:
1538 				len += sizeof(struct ip6_frag);
1539 				break;
1540 			case IPPROTO_AH:
1541 				len += (ip6e->ip6e_len + 2) << 2;
1542 				break;
1543 			default:
1544 				len += (ip6e->ip6e_len + 1) << 3;
1545 				break;
1546 			}
1547 			nxt = ip6e->ip6e_nxt;
1548 		}
1549 		if (ip6e)
1550 			return (&ip6e->ip6e_nxt);
1551 		else
1552 			return NULL;
1553 	}
1554 }
1555 
1556 /*
1557  * get next header offset.  m will be retained.
1558  */
1559 int
1560 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1561 {
1562 	struct ip6_hdr ip6;
1563 	struct ip6_ext ip6e;
1564 	struct ip6_frag fh;
1565 
1566 	/* just in case */
1567 	if (m == NULL)
1568 		panic("ip6_nexthdr: m == NULL");
1569 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1570 		return -1;
1571 
1572 	switch (proto) {
1573 	case IPPROTO_IPV6:
1574 		if (m->m_pkthdr.len < off + sizeof(ip6))
1575 			return -1;
1576 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1577 		if (nxtp)
1578 			*nxtp = ip6.ip6_nxt;
1579 		off += sizeof(ip6);
1580 		return off;
1581 
1582 	case IPPROTO_FRAGMENT:
1583 		/*
1584 		 * terminate parsing if it is not the first fragment,
1585 		 * it does not make sense to parse through it.
1586 		 */
1587 		if (m->m_pkthdr.len < off + sizeof(fh))
1588 			return -1;
1589 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1590 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1591 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1592 			return -1;
1593 		if (nxtp)
1594 			*nxtp = fh.ip6f_nxt;
1595 		off += sizeof(struct ip6_frag);
1596 		return off;
1597 
1598 	case IPPROTO_AH:
1599 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1600 			return -1;
1601 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1602 		if (nxtp)
1603 			*nxtp = ip6e.ip6e_nxt;
1604 		off += (ip6e.ip6e_len + 2) << 2;
1605 		return off;
1606 
1607 	case IPPROTO_HOPOPTS:
1608 	case IPPROTO_ROUTING:
1609 	case IPPROTO_DSTOPTS:
1610 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1611 			return -1;
1612 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1613 		if (nxtp)
1614 			*nxtp = ip6e.ip6e_nxt;
1615 		off += (ip6e.ip6e_len + 1) << 3;
1616 		return off;
1617 
1618 	case IPPROTO_NONE:
1619 	case IPPROTO_ESP:
1620 	case IPPROTO_IPCOMP:
1621 		/* give up */
1622 		return -1;
1623 
1624 	default:
1625 		return -1;
1626 	}
1627 
1628 	return -1;
1629 }
1630 
1631 /*
1632  * get offset for the last header in the chain.  m will be kept untainted.
1633  */
1634 int
1635 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1636 {
1637 	int newoff;
1638 	int nxt;
1639 
1640 	if (!nxtp) {
1641 		nxt = -1;
1642 		nxtp = &nxt;
1643 	}
1644 	while (1) {
1645 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1646 		if (newoff < 0)
1647 			return off;
1648 		else if (newoff < off)
1649 			return -1;	/* invalid */
1650 		else if (newoff == off)
1651 			return newoff;
1652 
1653 		off = newoff;
1654 		proto = *nxtp;
1655 	}
1656 }
1657 
1658 struct ip6aux *
1659 ip6_addaux(struct mbuf *m)
1660 {
1661 	struct m_tag *mtag;
1662 
1663 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1664 	if (!mtag) {
1665 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1666 		    M_NOWAIT);
1667 		if (mtag) {
1668 			m_tag_prepend(m, mtag);
1669 			bzero(mtag + 1, sizeof(struct ip6aux));
1670 		}
1671 	}
1672 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1673 }
1674 
1675 struct ip6aux *
1676 ip6_findaux(struct mbuf *m)
1677 {
1678 	struct m_tag *mtag;
1679 
1680 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1681 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1682 }
1683 
1684 void
1685 ip6_delaux(struct mbuf *m)
1686 {
1687 	struct m_tag *mtag;
1688 
1689 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1690 	if (mtag)
1691 		m_tag_delete(m, mtag);
1692 }
1693 
1694 /*
1695  * System control for IP6
1696  */
1697 
1698 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1699 	0,		0,		0,		0,
1700 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1701 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1702 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1703 	0,		0,		0,		0,
1704 	ENOPROTOOPT
1705 };
1706