xref: /freebsd/sys/netinet6/ip6_input.c (revision 6ef6ba9950260f42b47499d17874d00ca9290955)
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_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_kdtrace.h"
71 #include "opt_route.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/sdt.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_types.h>
91 #include <net/if_dl.h>
92 #include <net/route.h>
93 #include <net/netisr.h>
94 #include <net/pfil.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_kdtrace.h>
99 #include <netinet/ip_var.h>
100 #include <netinet/in_systm.h>
101 #include <net/if_llatbl.h>
102 #ifdef INET
103 #include <netinet/ip.h>
104 #include <netinet/ip_icmp.h>
105 #endif /* INET */
106 #include <netinet/ip6.h>
107 #include <netinet6/in6_var.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet/in_pcb.h>
110 #include <netinet/icmp6.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/in6_ifattach.h>
113 #include <netinet6/nd6.h>
114 
115 #ifdef IPSEC
116 #include <netipsec/ipsec.h>
117 #include <netinet6/ip6_ipsec.h>
118 #include <netipsec/ipsec6.h>
119 #endif /* IPSEC */
120 
121 #include <netinet6/ip6protosw.h>
122 
123 #ifdef FLOWTABLE
124 #include <net/flowtable.h>
125 VNET_DECLARE(int, ip6_output_flowtable_size);
126 #define	V_ip6_output_flowtable_size	VNET(ip6_output_flowtable_size)
127 #endif
128 
129 extern struct domain inet6domain;
130 
131 u_char ip6_protox[IPPROTO_MAX];
132 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
133 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
134 VNET_DEFINE(u_long, in6_ifaddrhmask);
135 
136 static struct netisr_handler ip6_nh = {
137 	.nh_name = "ip6",
138 	.nh_handler = ip6_input,
139 	.nh_proto = NETISR_IPV6,
140 	.nh_policy = NETISR_POLICY_FLOW,
141 };
142 
143 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
144 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
145 
146 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
147 
148 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
149 VNET_PCPUSTAT_SYSINIT(ip6stat);
150 #ifdef VIMAGE
151 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
152 #endif /* VIMAGE */
153 
154 struct rwlock in6_ifaddr_lock;
155 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
156 
157 static void ip6_init2(void *);
158 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
159 static struct ip6aux *ip6_addaux(struct mbuf *);
160 static struct ip6aux *ip6_findaux(struct mbuf *m);
161 static void ip6_delaux (struct mbuf *);
162 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
163 #ifdef PULLDOWN_TEST
164 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
165 #endif
166 
167 /*
168  * IP6 initialization: fill in IP6 protocol switch table.
169  * All protocols not implemented in kernel go to raw IP6 protocol handler.
170  */
171 void
172 ip6_init(void)
173 {
174 	struct ip6protosw *pr;
175 	int i;
176 
177 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
178 	    &V_ip6_auto_linklocal);
179 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
180 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
181 
182 	TAILQ_INIT(&V_in6_ifaddrhead);
183 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
184 	    &V_in6_ifaddrhmask);
185 
186 	/* Initialize packet filter hooks. */
187 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
188 	V_inet6_pfil_hook.ph_af = AF_INET6;
189 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
190 		printf("%s: WARNING: unable to register pfil hook, "
191 			"error %d\n", __func__, i);
192 
193 	scope6_init();
194 	addrsel_policy_init();
195 	nd6_init();
196 	frag6_init();
197 
198 #ifdef FLOWTABLE
199 	if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size",
200 		&V_ip6_output_flowtable_size)) {
201 		if (V_ip6_output_flowtable_size < 256)
202 			V_ip6_output_flowtable_size = 256;
203 		if (!powerof2(V_ip6_output_flowtable_size)) {
204 			printf("flowtable must be power of 2 size\n");
205 			V_ip6_output_flowtable_size = 2048;
206 		}
207 	} else {
208 		/*
209 		 * round up to the next power of 2
210 		 */
211 		V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
212 	}
213 	V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU);
214 #endif
215 
216 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
217 
218 	/* Skip global initialization stuff for non-default instances. */
219 	if (!IS_DEFAULT_VNET(curvnet))
220 		return;
221 
222 #ifdef DIAGNOSTIC
223 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
224 		panic("sizeof(protosw) != sizeof(ip6protosw)");
225 #endif
226 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
227 	if (pr == NULL)
228 		panic("ip6_init");
229 
230 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
231 	for (i = 0; i < IPPROTO_MAX; i++)
232 		ip6_protox[i] = pr - inet6sw;
233 	/*
234 	 * Cycle through IP protocols and put them into the appropriate place
235 	 * in ip6_protox[].
236 	 */
237 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
238 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
239 		if (pr->pr_domain->dom_family == PF_INET6 &&
240 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
241 			/* Be careful to only index valid IP protocols. */
242 			if (pr->pr_protocol < IPPROTO_MAX)
243 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
244 		}
245 
246 	netisr_register(&ip6_nh);
247 }
248 
249 /*
250  * The protocol to be inserted into ip6_protox[] must be already registered
251  * in inet6sw[], either statically or through pf_proto_register().
252  */
253 int
254 ip6proto_register(short ip6proto)
255 {
256 	struct ip6protosw *pr;
257 
258 	/* Sanity checks. */
259 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
260 		return (EPROTONOSUPPORT);
261 
262 	/*
263 	 * The protocol slot must not be occupied by another protocol
264 	 * already.  An index pointing to IPPROTO_RAW is unused.
265 	 */
266 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
267 	if (pr == NULL)
268 		return (EPFNOSUPPORT);
269 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
270 		return (EEXIST);
271 
272 	/*
273 	 * Find the protocol position in inet6sw[] and set the index.
274 	 */
275 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
276 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
277 		if (pr->pr_domain->dom_family == PF_INET6 &&
278 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
279 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
280 			return (0);
281 		}
282 	}
283 	return (EPROTONOSUPPORT);
284 }
285 
286 int
287 ip6proto_unregister(short ip6proto)
288 {
289 	struct ip6protosw *pr;
290 
291 	/* Sanity checks. */
292 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
293 		return (EPROTONOSUPPORT);
294 
295 	/* Check if the protocol was indeed registered. */
296 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
297 	if (pr == NULL)
298 		return (EPFNOSUPPORT);
299 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
300 		return (ENOENT);
301 
302 	/* Reset the protocol slot to IPPROTO_RAW. */
303 	ip6_protox[ip6proto] = pr - inet6sw;
304 	return (0);
305 }
306 
307 #ifdef VIMAGE
308 void
309 ip6_destroy()
310 {
311 	int i;
312 
313 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
314 		printf("%s: WARNING: unable to unregister pfil hook, "
315 		    "error %d\n", __func__, i);
316 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
317 	nd6_destroy();
318 	callout_drain(&V_in6_tmpaddrtimer_ch);
319 }
320 #endif
321 
322 static int
323 ip6_init2_vnet(const void *unused __unused)
324 {
325 
326 	/* nd6_timer_init */
327 	callout_init(&V_nd6_timer_ch, 0);
328 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
329 
330 	/* timer for regeneranation of temporary addresses randomize ID */
331 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
332 	callout_reset(&V_in6_tmpaddrtimer_ch,
333 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
334 		       V_ip6_temp_regen_advance) * hz,
335 		      in6_tmpaddrtimer, curvnet);
336 
337 	return (0);
338 }
339 
340 static void
341 ip6_init2(void *dummy)
342 {
343 
344 	ip6_init2_vnet(NULL);
345 }
346 
347 /* cheat */
348 /* This must be after route_init(), which is now SI_ORDER_THIRD */
349 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
350 
351 static int
352 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
353     int *nxt, int *ours)
354 {
355 	struct ip6_hdr *ip6;
356 	struct ip6_hbh *hbh;
357 
358 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
359 #if 0	/*touches NULL pointer*/
360 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
361 #endif
362 		goto out;	/* m have already been freed */
363 	}
364 
365 	/* adjust pointer */
366 	ip6 = mtod(m, struct ip6_hdr *);
367 
368 	/*
369 	 * if the payload length field is 0 and the next header field
370 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
371 	 * option MUST be included.
372 	 */
373 	if (ip6->ip6_plen == 0 && *plen == 0) {
374 		/*
375 		 * Note that if a valid jumbo payload option is
376 		 * contained, ip6_hopopts_input() must set a valid
377 		 * (non-zero) payload length to the variable plen.
378 		 */
379 		IP6STAT_INC(ip6s_badoptions);
380 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
381 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
382 		icmp6_error(m, ICMP6_PARAM_PROB,
383 			    ICMP6_PARAMPROB_HEADER,
384 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
385 		goto out;
386 	}
387 #ifndef PULLDOWN_TEST
388 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
389 	hbh = (struct ip6_hbh *)(ip6 + 1);
390 #else
391 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
392 		sizeof(struct ip6_hbh));
393 	if (hbh == NULL) {
394 		IP6STAT_INC(ip6s_tooshort);
395 		goto out;
396 	}
397 #endif
398 	*nxt = hbh->ip6h_nxt;
399 
400 	/*
401 	 * If we are acting as a router and the packet contains a
402 	 * router alert option, see if we know the option value.
403 	 * Currently, we only support the option value for MLD, in which
404 	 * case we should pass the packet to the multicast routing
405 	 * daemon.
406 	 */
407 	if (*rtalert != ~0) {
408 		switch (*rtalert) {
409 		case IP6OPT_RTALERT_MLD:
410 			if (V_ip6_forwarding)
411 				*ours = 1;
412 			break;
413 		default:
414 			/*
415 			 * RFC2711 requires unrecognized values must be
416 			 * silently ignored.
417 			 */
418 			break;
419 		}
420 	}
421 
422 	return (0);
423 
424 out:
425 	return (1);
426 }
427 
428 void
429 ip6_input(struct mbuf *m)
430 {
431 	struct ip6_hdr *ip6;
432 	int off = sizeof(struct ip6_hdr), nest;
433 	u_int32_t plen;
434 	u_int32_t rtalert = ~0;
435 	int nxt, ours = 0;
436 	struct ifnet *deliverifp = NULL, *ifp = NULL;
437 	struct in6_addr odst;
438 	struct route_in6 rin6;
439 	int srcrt = 0;
440 	struct llentry *lle = NULL;
441 	struct sockaddr_in6 dst6, *dst;
442 
443 	bzero(&rin6, sizeof(struct route_in6));
444 #ifdef IPSEC
445 	/*
446 	 * should the inner packet be considered authentic?
447 	 * see comment in ah4_input().
448 	 * NB: m cannot be NULL when passed to the input routine
449 	 */
450 
451 	m->m_flags &= ~M_AUTHIPHDR;
452 	m->m_flags &= ~M_AUTHIPDGM;
453 
454 #endif /* IPSEC */
455 
456 	/*
457 	 * make sure we don't have onion peering information into m_tag.
458 	 */
459 	ip6_delaux(m);
460 
461 	if (m->m_flags & M_FASTFWD_OURS) {
462 		/*
463 		 * Firewall changed destination to local.
464 		 */
465 		m->m_flags &= ~M_FASTFWD_OURS;
466 		ours = 1;
467 		deliverifp = m->m_pkthdr.rcvif;
468 		ip6 = mtod(m, struct ip6_hdr *);
469 		goto hbhcheck;
470 	}
471 
472 	/*
473 	 * mbuf statistics
474 	 */
475 	if (m->m_flags & M_EXT) {
476 		if (m->m_next)
477 			IP6STAT_INC(ip6s_mext2m);
478 		else
479 			IP6STAT_INC(ip6s_mext1);
480 	} else {
481 		if (m->m_next) {
482 			if (m->m_flags & M_LOOP) {
483 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
484 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
485 				IP6STAT_INC(
486 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
487 			else
488 				IP6STAT_INC(ip6s_m2m[0]);
489 		} else
490 			IP6STAT_INC(ip6s_m1);
491 	}
492 
493 	/* drop the packet if IPv6 operation is disabled on the IF */
494 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
495 		m_freem(m);
496 		return;
497 	}
498 
499 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
500 	IP6STAT_INC(ip6s_total);
501 
502 #ifndef PULLDOWN_TEST
503 	/*
504 	 * L2 bridge code and some other code can return mbuf chain
505 	 * that does not conform to KAME requirement.  too bad.
506 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
507 	 */
508 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
509 		struct mbuf *n;
510 
511 		if (m->m_pkthdr.len > MHLEN)
512 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
513 		else
514 			n = m_gethdr(M_NOWAIT, MT_DATA);
515 		if (n == NULL) {
516 			m_freem(m);
517 			return;	/* ENOBUFS */
518 		}
519 
520 		m_move_pkthdr(n, m);
521 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
522 		n->m_len = n->m_pkthdr.len;
523 		m_freem(m);
524 		m = n;
525 	}
526 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
527 #endif
528 
529 	if (m->m_len < sizeof(struct ip6_hdr)) {
530 		struct ifnet *inifp;
531 		inifp = m->m_pkthdr.rcvif;
532 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
533 			IP6STAT_INC(ip6s_toosmall);
534 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
535 			return;
536 		}
537 	}
538 
539 	ip6 = mtod(m, struct ip6_hdr *);
540 
541 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
542 		IP6STAT_INC(ip6s_badvers);
543 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
544 		goto bad;
545 	}
546 
547 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
548 
549 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
550 
551 	/*
552 	 * Check against address spoofing/corruption.
553 	 */
554 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
555 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
556 		/*
557 		 * XXX: "badscope" is not very suitable for a multicast source.
558 		 */
559 		IP6STAT_INC(ip6s_badscope);
560 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
561 		goto bad;
562 	}
563 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
564 	    !(m->m_flags & M_LOOP)) {
565 		/*
566 		 * In this case, the packet should come from the loopback
567 		 * interface.  However, we cannot just check the if_flags,
568 		 * because ip6_mloopback() passes the "actual" interface
569 		 * as the outgoing/incoming interface.
570 		 */
571 		IP6STAT_INC(ip6s_badscope);
572 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
573 		goto bad;
574 	}
575 
576 #ifdef ALTQ
577 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
578 		/* packet is dropped by traffic conditioner */
579 		return;
580 	}
581 #endif
582 	/*
583 	 * The following check is not documented in specs.  A malicious
584 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
585 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
586 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
587 	 *
588 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
589 	 * support IPv4-less kernel compilation, we cannot support SIIT
590 	 * environment at all.  So, it makes more sense for us to reject any
591 	 * malicious packets for non-SIIT environment, than try to do a
592 	 * partial support for SIIT environment.
593 	 */
594 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
595 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
596 		IP6STAT_INC(ip6s_badscope);
597 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
598 		goto bad;
599 	}
600 #if 0
601 	/*
602 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
603 	 *
604 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
605 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
606 	 * is revised to forbid relaying case.
607 	 */
608 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
609 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
610 		IP6STAT_INC(ip6s_badscope);
611 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
612 		goto bad;
613 	}
614 #endif
615 #ifdef IPSEC
616 	/*
617 	 * Bypass packet filtering for packets previously handled by IPsec.
618 	 */
619 	if (ip6_ipsec_filtertunnel(m))
620 		goto passin;
621 #endif /* IPSEC */
622 
623 	/*
624 	 * Run through list of hooks for input packets.
625 	 *
626 	 * NB: Beware of the destination address changing
627 	 *     (e.g. by NAT rewriting).  When this happens,
628 	 *     tell ip6_forward to do the right thing.
629 	 */
630 	odst = ip6->ip6_dst;
631 
632 	/* Jump over all PFIL processing if hooks are not active. */
633 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
634 		goto passin;
635 
636 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
637 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
638 		return;
639 	if (m == NULL)			/* consumed by filter */
640 		return;
641 	ip6 = mtod(m, struct ip6_hdr *);
642 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
643 
644 	if (m->m_flags & M_FASTFWD_OURS) {
645 		m->m_flags &= ~M_FASTFWD_OURS;
646 		ours = 1;
647 		deliverifp = m->m_pkthdr.rcvif;
648 		goto hbhcheck;
649 	}
650 	if ((m->m_flags & M_IP6_NEXTHOP) &&
651 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
652 		/*
653 		 * Directly ship the packet on.  This allows forwarding
654 		 * packets originally destined to us to some other directly
655 		 * connected host.
656 		 */
657 		ip6_forward(m, 1);
658 		goto out;
659 	}
660 
661 passin:
662 	/*
663 	 * Disambiguate address scope zones (if there is ambiguity).
664 	 * We first make sure that the original source or destination address
665 	 * is not in our internal form for scoped addresses.  Such addresses
666 	 * are not necessarily invalid spec-wise, but we cannot accept them due
667 	 * to the usage conflict.
668 	 * in6_setscope() then also checks and rejects the cases where src or
669 	 * dst are the loopback address and the receiving interface
670 	 * is not loopback.
671 	 */
672 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
673 		IP6STAT_INC(ip6s_badscope); /* XXX */
674 		goto bad;
675 	}
676 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
677 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
678 		IP6STAT_INC(ip6s_badscope);
679 		goto bad;
680 	}
681 
682 	/*
683 	 * Multicast check. Assume packet is for us to avoid
684 	 * prematurely taking locks.
685 	 */
686 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
687 		ours = 1;
688 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
689 		deliverifp = m->m_pkthdr.rcvif;
690 		goto hbhcheck;
691 	}
692 
693 	/*
694 	 *  Unicast check
695 	 */
696 
697 	bzero(&dst6, sizeof(dst6));
698 	dst6.sin6_family = AF_INET6;
699 	dst6.sin6_len = sizeof(struct sockaddr_in6);
700 	dst6.sin6_addr = ip6->ip6_dst;
701 	ifp = m->m_pkthdr.rcvif;
702 	IF_AFDATA_RLOCK(ifp);
703 	lle = lla_lookup(LLTABLE6(ifp), 0,
704 	     (struct sockaddr *)&dst6);
705 	IF_AFDATA_RUNLOCK(ifp);
706 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
707 		struct ifaddr *ifa;
708 		struct in6_ifaddr *ia6;
709 		int bad;
710 
711 		bad = 1;
712 #define	sa_equal(a1, a2)						\
713 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
714 		IF_ADDR_RLOCK(ifp);
715 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
716 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
717 				continue;
718 			if (sa_equal(&dst6, ifa->ifa_addr))
719 				break;
720 		}
721 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
722 		    __func__, lle));
723 #undef sa_equal
724 
725 		ia6 = (struct in6_ifaddr *)ifa;
726 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
727 			/* Count the packet in the ip address stats */
728 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
729 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
730 			    m->m_pkthdr.len);
731 
732 			/*
733 			 * record address information into m_tag.
734 			 */
735 			(void)ip6_setdstifaddr(m, ia6);
736 
737 			bad = 0;
738 		} else {
739 			char ip6bufs[INET6_ADDRSTRLEN];
740 			char ip6bufd[INET6_ADDRSTRLEN];
741 			/* address is not ready, so discard the packet. */
742 			nd6log((LOG_INFO,
743 			    "ip6_input: packet to an unready address %s->%s\n",
744 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
745 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
746 		}
747 		IF_ADDR_RUNLOCK(ifp);
748 		LLE_RUNLOCK(lle);
749 		if (bad)
750 			goto bad;
751 		else {
752 			ours = 1;
753 			deliverifp = ifp;
754 			goto hbhcheck;
755 		}
756 	}
757 	if (lle != NULL)
758 		LLE_RUNLOCK(lle);
759 
760 	dst = &rin6.ro_dst;
761 	dst->sin6_len = sizeof(struct sockaddr_in6);
762 	dst->sin6_family = AF_INET6;
763 	dst->sin6_addr = ip6->ip6_dst;
764 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
765 	if (rin6.ro_rt)
766 		RT_UNLOCK(rin6.ro_rt);
767 
768 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
769 
770 	/*
771 	 * Accept the packet if the forwarding interface to the destination
772 	 * according to the routing table is the loopback interface,
773 	 * unless the associated route has a gateway.
774 	 * Note that this approach causes to accept a packet if there is a
775 	 * route to the loopback interface for the destination of the packet.
776 	 * But we think it's even useful in some situations, e.g. when using
777 	 * a special daemon which wants to intercept the packet.
778 	 *
779 	 * XXX: some OSes automatically make a cloned route for the destination
780 	 * of an outgoing packet.  If the outgoing interface of the packet
781 	 * is a loopback one, the kernel would consider the packet to be
782 	 * accepted, even if we have no such address assinged on the interface.
783 	 * We check the cloned flag of the route entry to reject such cases,
784 	 * assuming that route entries for our own addresses are not made by
785 	 * cloning (it should be true because in6_addloop explicitly installs
786 	 * the host route).  However, we might have to do an explicit check
787 	 * while it would be less efficient.  Or, should we rather install a
788 	 * reject route for such a case?
789 	 */
790 	if (rin6.ro_rt &&
791 	    (rin6.ro_rt->rt_flags &
792 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
793 #ifdef RTF_WASCLONED
794 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
795 #endif
796 #ifdef RTF_CLONED
797 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
798 #endif
799 #if 0
800 	    /*
801 	     * The check below is redundant since the comparison of
802 	     * the destination and the key of the rtentry has
803 	     * already done through looking up the routing table.
804 	     */
805 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
806 	    &rt6_key(rin6.ro_rt)->sin6_addr)
807 #endif
808 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
809 		int free_ia6 = 0;
810 		struct in6_ifaddr *ia6;
811 
812 		/*
813 		 * found the loopback route to the interface address
814 		 */
815 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
816 			struct sockaddr_in6 dest6;
817 
818 			bzero(&dest6, sizeof(dest6));
819 			dest6.sin6_family = AF_INET6;
820 			dest6.sin6_len = sizeof(dest6);
821 			dest6.sin6_addr = ip6->ip6_dst;
822 			ia6 = (struct in6_ifaddr *)
823 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
824 			if (ia6 == NULL)
825 				goto bad;
826 			free_ia6 = 1;
827 		}
828 		else
829 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
830 
831 		/*
832 		 * record address information into m_tag.
833 		 */
834 		(void)ip6_setdstifaddr(m, ia6);
835 
836 		/*
837 		 * packets to a tentative, duplicated, or somehow invalid
838 		 * address must not be accepted.
839 		 */
840 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
841 			/* this address is ready */
842 			ours = 1;
843 			deliverifp = ia6->ia_ifp;	/* correct? */
844 			/* Count the packet in the ip address stats */
845 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
846 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
847 			    m->m_pkthdr.len);
848 			if (free_ia6)
849 				ifa_free(&ia6->ia_ifa);
850 			goto hbhcheck;
851 		} else {
852 			char ip6bufs[INET6_ADDRSTRLEN];
853 			char ip6bufd[INET6_ADDRSTRLEN];
854 			/* address is not ready, so discard the packet. */
855 			nd6log((LOG_INFO,
856 			    "ip6_input: packet to an unready address %s->%s\n",
857 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
858 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
859 
860 			if (free_ia6)
861 				ifa_free(&ia6->ia_ifa);
862 			goto bad;
863 		}
864 	}
865 
866 	/*
867 	 * FAITH (Firewall Aided Internet Translator)
868 	 */
869 	if (V_ip6_keepfaith) {
870 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
871 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
872 			/* XXX do we need more sanity checks? */
873 			ours = 1;
874 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
875 			goto hbhcheck;
876 		}
877 	}
878 
879 	/*
880 	 * Now there is no reason to process the packet if it's not our own
881 	 * and we're not a router.
882 	 */
883 	if (!V_ip6_forwarding) {
884 		IP6STAT_INC(ip6s_cantforward);
885 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
886 		goto bad;
887 	}
888 
889   hbhcheck:
890 	/*
891 	 * record address information into m_tag, if we don't have one yet.
892 	 * note that we are unable to record it, if the address is not listed
893 	 * as our interface address (e.g. multicast addresses, addresses
894 	 * within FAITH prefixes and such).
895 	 */
896 	if (deliverifp) {
897 		struct in6_ifaddr *ia6;
898 
899  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
900 			ifa_free(&ia6->ia_ifa);
901 		} else {
902 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
903 			if (ia6) {
904 				if (!ip6_setdstifaddr(m, ia6)) {
905 					/*
906 					 * XXX maybe we should drop the packet here,
907 					 * as we could not provide enough information
908 					 * to the upper layers.
909 					 */
910 				}
911 				ifa_free(&ia6->ia_ifa);
912 			}
913 		}
914 	}
915 
916 	/*
917 	 * Process Hop-by-Hop options header if it's contained.
918 	 * m may be modified in ip6_hopopts_input().
919 	 * If a JumboPayload option is included, plen will also be modified.
920 	 */
921 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
922 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
923 		int error;
924 
925 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
926 		if (error != 0)
927 			goto out;
928 	} else
929 		nxt = ip6->ip6_nxt;
930 
931 	/*
932 	 * Check that the amount of data in the buffers
933 	 * is as at least much as the IPv6 header would have us expect.
934 	 * Trim mbufs if longer than we expect.
935 	 * Drop packet if shorter than we expect.
936 	 */
937 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
938 		IP6STAT_INC(ip6s_tooshort);
939 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
940 		goto bad;
941 	}
942 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
943 		if (m->m_len == m->m_pkthdr.len) {
944 			m->m_len = sizeof(struct ip6_hdr) + plen;
945 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
946 		} else
947 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
948 	}
949 
950 	/*
951 	 * Forward if desirable.
952 	 */
953 	if (V_ip6_mrouter &&
954 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
955 		/*
956 		 * If we are acting as a multicast router, all
957 		 * incoming multicast packets are passed to the
958 		 * kernel-level multicast forwarding function.
959 		 * The packet is returned (relatively) intact; if
960 		 * ip6_mforward() returns a non-zero value, the packet
961 		 * must be discarded, else it may be accepted below.
962 		 *
963 		 * XXX TODO: Check hlim and multicast scope here to avoid
964 		 * unnecessarily calling into ip6_mforward().
965 		 */
966 		if (ip6_mforward &&
967 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
968 			IP6STAT_INC(ip6s_cantforward);
969 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
970 			goto bad;
971 		}
972 	} else if (!ours) {
973 		ip6_forward(m, srcrt);
974 		goto out;
975 	}
976 
977 	ip6 = mtod(m, struct ip6_hdr *);
978 
979 	/*
980 	 * Malicious party may be able to use IPv4 mapped addr to confuse
981 	 * tcp/udp stack and bypass security checks (act as if it was from
982 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
983 	 *
984 	 * For SIIT end node behavior, you may want to disable the check.
985 	 * However, you will  become vulnerable to attacks using IPv4 mapped
986 	 * source.
987 	 */
988 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
989 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
990 		IP6STAT_INC(ip6s_badscope);
991 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
992 		goto bad;
993 	}
994 
995 	/*
996 	 * Tell launch routine the next header
997 	 */
998 	IP6STAT_INC(ip6s_delivered);
999 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
1000 	nest = 0;
1001 
1002 	while (nxt != IPPROTO_DONE) {
1003 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
1004 			IP6STAT_INC(ip6s_toomanyhdr);
1005 			goto bad;
1006 		}
1007 
1008 		/*
1009 		 * protection against faulty packet - there should be
1010 		 * more sanity checks in header chain processing.
1011 		 */
1012 		if (m->m_pkthdr.len < off) {
1013 			IP6STAT_INC(ip6s_tooshort);
1014 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
1015 			goto bad;
1016 		}
1017 
1018 #ifdef IPSEC
1019 		/*
1020 		 * enforce IPsec policy checking if we are seeing last header.
1021 		 * note that we do not visit this with protocols with pcb layer
1022 		 * code - like udp/tcp/raw ip.
1023 		 */
1024 		if (ip6_ipsec_input(m, nxt))
1025 			goto bad;
1026 #endif /* IPSEC */
1027 
1028 		/*
1029 		 * Use mbuf flags to propagate Router Alert option to
1030 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1031 		 */
1032 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1033 			m->m_flags |= M_RTALERT_MLD;
1034 
1035 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1036 	}
1037 	goto out;
1038 bad:
1039 	m_freem(m);
1040 out:
1041 	if (rin6.ro_rt)
1042 		RTFREE(rin6.ro_rt);
1043 }
1044 
1045 /*
1046  * set/grab in6_ifaddr correspond to IPv6 destination address.
1047  * XXX backward compatibility wrapper
1048  *
1049  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1050  * and then bump it when the tag is copied, and release it when the tag is
1051  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1052  * we just bump the ia refcount when we receive it.  This should be fixed.
1053  */
1054 static struct ip6aux *
1055 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1056 {
1057 	struct ip6aux *ip6a;
1058 
1059 	ip6a = ip6_addaux(m);
1060 	if (ip6a)
1061 		ip6a->ip6a_dstia6 = ia6;
1062 	return ip6a;	/* NULL if failed to set */
1063 }
1064 
1065 struct in6_ifaddr *
1066 ip6_getdstifaddr(struct mbuf *m)
1067 {
1068 	struct ip6aux *ip6a;
1069 	struct in6_ifaddr *ia;
1070 
1071 	ip6a = ip6_findaux(m);
1072 	if (ip6a) {
1073 		ia = ip6a->ip6a_dstia6;
1074 		ifa_ref(&ia->ia_ifa);
1075 		return ia;
1076 	} else
1077 		return NULL;
1078 }
1079 
1080 /*
1081  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1082  * included, the real payload length will be stored in plenp.
1083  *
1084  * rtalertp - XXX: should be stored more smart way
1085  */
1086 static int
1087 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1088     struct mbuf **mp, int *offp)
1089 {
1090 	struct mbuf *m = *mp;
1091 	int off = *offp, hbhlen;
1092 	struct ip6_hbh *hbh;
1093 	u_int8_t *opt;
1094 
1095 	/* validation of the length of the header */
1096 #ifndef PULLDOWN_TEST
1097 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1098 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1099 	hbhlen = (hbh->ip6h_len + 1) << 3;
1100 
1101 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1102 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1103 #else
1104 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1105 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1106 	if (hbh == NULL) {
1107 		IP6STAT_INC(ip6s_tooshort);
1108 		return -1;
1109 	}
1110 	hbhlen = (hbh->ip6h_len + 1) << 3;
1111 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1112 		hbhlen);
1113 	if (hbh == NULL) {
1114 		IP6STAT_INC(ip6s_tooshort);
1115 		return -1;
1116 	}
1117 #endif
1118 	off += hbhlen;
1119 	hbhlen -= sizeof(struct ip6_hbh);
1120 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1121 
1122 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1123 				hbhlen, rtalertp, plenp) < 0)
1124 		return (-1);
1125 
1126 	*offp = off;
1127 	*mp = m;
1128 	return (0);
1129 }
1130 
1131 /*
1132  * Search header for all Hop-by-hop options and process each option.
1133  * This function is separate from ip6_hopopts_input() in order to
1134  * handle a case where the sending node itself process its hop-by-hop
1135  * options header. In such a case, the function is called from ip6_output().
1136  *
1137  * The function assumes that hbh header is located right after the IPv6 header
1138  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1139  * opthead + hbhlen is located in contiguous memory region.
1140  */
1141 int
1142 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1143     u_int32_t *rtalertp, u_int32_t *plenp)
1144 {
1145 	struct ip6_hdr *ip6;
1146 	int optlen = 0;
1147 	u_int8_t *opt = opthead;
1148 	u_int16_t rtalert_val;
1149 	u_int32_t jumboplen;
1150 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1151 
1152 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1153 		switch (*opt) {
1154 		case IP6OPT_PAD1:
1155 			optlen = 1;
1156 			break;
1157 		case IP6OPT_PADN:
1158 			if (hbhlen < IP6OPT_MINLEN) {
1159 				IP6STAT_INC(ip6s_toosmall);
1160 				goto bad;
1161 			}
1162 			optlen = *(opt + 1) + 2;
1163 			break;
1164 		case IP6OPT_ROUTER_ALERT:
1165 			/* XXX may need check for alignment */
1166 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1167 				IP6STAT_INC(ip6s_toosmall);
1168 				goto bad;
1169 			}
1170 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1171 				/* XXX stat */
1172 				icmp6_error(m, ICMP6_PARAM_PROB,
1173 				    ICMP6_PARAMPROB_HEADER,
1174 				    erroff + opt + 1 - opthead);
1175 				return (-1);
1176 			}
1177 			optlen = IP6OPT_RTALERT_LEN;
1178 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1179 			*rtalertp = ntohs(rtalert_val);
1180 			break;
1181 		case IP6OPT_JUMBO:
1182 			/* XXX may need check for alignment */
1183 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1184 				IP6STAT_INC(ip6s_toosmall);
1185 				goto bad;
1186 			}
1187 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1188 				/* XXX stat */
1189 				icmp6_error(m, ICMP6_PARAM_PROB,
1190 				    ICMP6_PARAMPROB_HEADER,
1191 				    erroff + opt + 1 - opthead);
1192 				return (-1);
1193 			}
1194 			optlen = IP6OPT_JUMBO_LEN;
1195 
1196 			/*
1197 			 * IPv6 packets that have non 0 payload length
1198 			 * must not contain a jumbo payload option.
1199 			 */
1200 			ip6 = mtod(m, struct ip6_hdr *);
1201 			if (ip6->ip6_plen) {
1202 				IP6STAT_INC(ip6s_badoptions);
1203 				icmp6_error(m, ICMP6_PARAM_PROB,
1204 				    ICMP6_PARAMPROB_HEADER,
1205 				    erroff + opt - opthead);
1206 				return (-1);
1207 			}
1208 
1209 			/*
1210 			 * We may see jumbolen in unaligned location, so
1211 			 * we'd need to perform bcopy().
1212 			 */
1213 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1214 			jumboplen = (u_int32_t)htonl(jumboplen);
1215 
1216 #if 1
1217 			/*
1218 			 * if there are multiple jumbo payload options,
1219 			 * *plenp will be non-zero and the packet will be
1220 			 * rejected.
1221 			 * the behavior may need some debate in ipngwg -
1222 			 * multiple options does not make sense, however,
1223 			 * there's no explicit mention in specification.
1224 			 */
1225 			if (*plenp != 0) {
1226 				IP6STAT_INC(ip6s_badoptions);
1227 				icmp6_error(m, ICMP6_PARAM_PROB,
1228 				    ICMP6_PARAMPROB_HEADER,
1229 				    erroff + opt + 2 - opthead);
1230 				return (-1);
1231 			}
1232 #endif
1233 
1234 			/*
1235 			 * jumbo payload length must be larger than 65535.
1236 			 */
1237 			if (jumboplen <= IPV6_MAXPACKET) {
1238 				IP6STAT_INC(ip6s_badoptions);
1239 				icmp6_error(m, ICMP6_PARAM_PROB,
1240 				    ICMP6_PARAMPROB_HEADER,
1241 				    erroff + opt + 2 - opthead);
1242 				return (-1);
1243 			}
1244 			*plenp = jumboplen;
1245 
1246 			break;
1247 		default:		/* unknown option */
1248 			if (hbhlen < IP6OPT_MINLEN) {
1249 				IP6STAT_INC(ip6s_toosmall);
1250 				goto bad;
1251 			}
1252 			optlen = ip6_unknown_opt(opt, m,
1253 			    erroff + opt - opthead);
1254 			if (optlen == -1)
1255 				return (-1);
1256 			optlen += 2;
1257 			break;
1258 		}
1259 	}
1260 
1261 	return (0);
1262 
1263   bad:
1264 	m_freem(m);
1265 	return (-1);
1266 }
1267 
1268 /*
1269  * Unknown option processing.
1270  * The third argument `off' is the offset from the IPv6 header to the option,
1271  * which is necessary if the IPv6 header the and option header and IPv6 header
1272  * is not contiguous in order to return an ICMPv6 error.
1273  */
1274 int
1275 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1276 {
1277 	struct ip6_hdr *ip6;
1278 
1279 	switch (IP6OPT_TYPE(*optp)) {
1280 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1281 		return ((int)*(optp + 1));
1282 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1283 		m_freem(m);
1284 		return (-1);
1285 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1286 		IP6STAT_INC(ip6s_badoptions);
1287 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1288 		return (-1);
1289 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1290 		IP6STAT_INC(ip6s_badoptions);
1291 		ip6 = mtod(m, struct ip6_hdr *);
1292 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1293 		    (m->m_flags & (M_BCAST|M_MCAST)))
1294 			m_freem(m);
1295 		else
1296 			icmp6_error(m, ICMP6_PARAM_PROB,
1297 				    ICMP6_PARAMPROB_OPTION, off);
1298 		return (-1);
1299 	}
1300 
1301 	m_freem(m);		/* XXX: NOTREACHED */
1302 	return (-1);
1303 }
1304 
1305 /*
1306  * Create the "control" list for this pcb.
1307  * These functions will not modify mbuf chain at all.
1308  *
1309  * With KAME mbuf chain restriction:
1310  * The routine will be called from upper layer handlers like tcp6_input().
1311  * Thus the routine assumes that the caller (tcp6_input) have already
1312  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1313  * very first mbuf on the mbuf chain.
1314  *
1315  * ip6_savecontrol_v4 will handle those options that are possible to be
1316  * set on a v4-mapped socket.
1317  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1318  * options and handle the v6-only ones itself.
1319  */
1320 struct mbuf **
1321 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1322     int *v4only)
1323 {
1324 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1325 
1326 #ifdef SO_TIMESTAMP
1327 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1328 		struct timeval tv;
1329 
1330 		microtime(&tv);
1331 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1332 		    SCM_TIMESTAMP, SOL_SOCKET);
1333 		if (*mp)
1334 			mp = &(*mp)->m_next;
1335 	}
1336 #endif
1337 
1338 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1339 	/* RFC 2292 sec. 5 */
1340 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1341 		struct in6_pktinfo pi6;
1342 
1343 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1344 #ifdef INET
1345 			struct ip *ip;
1346 
1347 			ip = mtod(m, struct ip *);
1348 			pi6.ipi6_addr.s6_addr32[0] = 0;
1349 			pi6.ipi6_addr.s6_addr32[1] = 0;
1350 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1351 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1352 #else
1353 			/* We won't hit this code */
1354 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1355 #endif
1356 		} else {
1357 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1358 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1359 		}
1360 		pi6.ipi6_ifindex =
1361 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1362 
1363 		*mp = sbcreatecontrol((caddr_t) &pi6,
1364 		    sizeof(struct in6_pktinfo),
1365 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1366 		if (*mp)
1367 			mp = &(*mp)->m_next;
1368 	}
1369 
1370 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1371 		int hlim;
1372 
1373 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1374 #ifdef INET
1375 			struct ip *ip;
1376 
1377 			ip = mtod(m, struct ip *);
1378 			hlim = ip->ip_ttl;
1379 #else
1380 			/* We won't hit this code */
1381 			hlim = 0;
1382 #endif
1383 		} else {
1384 			hlim = ip6->ip6_hlim & 0xff;
1385 		}
1386 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1387 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1388 		    IPPROTO_IPV6);
1389 		if (*mp)
1390 			mp = &(*mp)->m_next;
1391 	}
1392 
1393 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1394 		int tclass;
1395 
1396 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1397 #ifdef INET
1398 			struct ip *ip;
1399 
1400 			ip = mtod(m, struct ip *);
1401 			tclass = ip->ip_tos;
1402 #else
1403 			/* We won't hit this code */
1404 			tclass = 0;
1405 #endif
1406 		} else {
1407 			u_int32_t flowinfo;
1408 
1409 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1410 			flowinfo >>= 20;
1411 			tclass = flowinfo & 0xff;
1412 		}
1413 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1414 		    IPV6_TCLASS, IPPROTO_IPV6);
1415 		if (*mp)
1416 			mp = &(*mp)->m_next;
1417 	}
1418 
1419 	if (v4only != NULL) {
1420 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1421 			*v4only = 1;
1422 		} else {
1423 			*v4only = 0;
1424 		}
1425 	}
1426 
1427 	return (mp);
1428 }
1429 
1430 void
1431 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1432 {
1433 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1434 	int v4only = 0;
1435 
1436 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1437 	if (v4only)
1438 		return;
1439 
1440 	/*
1441 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1442 	 * privilege for the option (see ip6_ctloutput), but it might be too
1443 	 * strict, since there might be some hop-by-hop options which can be
1444 	 * returned to normal user.
1445 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1446 	 */
1447 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1448 		/*
1449 		 * Check if a hop-by-hop options header is contatined in the
1450 		 * received packet, and if so, store the options as ancillary
1451 		 * data. Note that a hop-by-hop options header must be
1452 		 * just after the IPv6 header, which is assured through the
1453 		 * IPv6 input processing.
1454 		 */
1455 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1456 			struct ip6_hbh *hbh;
1457 			int hbhlen = 0;
1458 #ifdef PULLDOWN_TEST
1459 			struct mbuf *ext;
1460 #endif
1461 
1462 #ifndef PULLDOWN_TEST
1463 			hbh = (struct ip6_hbh *)(ip6 + 1);
1464 			hbhlen = (hbh->ip6h_len + 1) << 3;
1465 #else
1466 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1467 			    ip6->ip6_nxt);
1468 			if (ext == NULL) {
1469 				IP6STAT_INC(ip6s_tooshort);
1470 				return;
1471 			}
1472 			hbh = mtod(ext, struct ip6_hbh *);
1473 			hbhlen = (hbh->ip6h_len + 1) << 3;
1474 			if (hbhlen != ext->m_len) {
1475 				m_freem(ext);
1476 				IP6STAT_INC(ip6s_tooshort);
1477 				return;
1478 			}
1479 #endif
1480 
1481 			/*
1482 			 * XXX: We copy the whole header even if a
1483 			 * jumbo payload option is included, the option which
1484 			 * is to be removed before returning according to
1485 			 * RFC2292.
1486 			 * Note: this constraint is removed in RFC3542
1487 			 */
1488 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1489 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1490 			    IPPROTO_IPV6);
1491 			if (*mp)
1492 				mp = &(*mp)->m_next;
1493 #ifdef PULLDOWN_TEST
1494 			m_freem(ext);
1495 #endif
1496 		}
1497 	}
1498 
1499 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1500 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1501 
1502 		/*
1503 		 * Search for destination options headers or routing
1504 		 * header(s) through the header chain, and stores each
1505 		 * header as ancillary data.
1506 		 * Note that the order of the headers remains in
1507 		 * the chain of ancillary data.
1508 		 */
1509 		while (1) {	/* is explicit loop prevention necessary? */
1510 			struct ip6_ext *ip6e = NULL;
1511 			int elen;
1512 #ifdef PULLDOWN_TEST
1513 			struct mbuf *ext = NULL;
1514 #endif
1515 
1516 			/*
1517 			 * if it is not an extension header, don't try to
1518 			 * pull it from the chain.
1519 			 */
1520 			switch (nxt) {
1521 			case IPPROTO_DSTOPTS:
1522 			case IPPROTO_ROUTING:
1523 			case IPPROTO_HOPOPTS:
1524 			case IPPROTO_AH: /* is it possible? */
1525 				break;
1526 			default:
1527 				goto loopend;
1528 			}
1529 
1530 #ifndef PULLDOWN_TEST
1531 			if (off + sizeof(*ip6e) > m->m_len)
1532 				goto loopend;
1533 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1534 			if (nxt == IPPROTO_AH)
1535 				elen = (ip6e->ip6e_len + 2) << 2;
1536 			else
1537 				elen = (ip6e->ip6e_len + 1) << 3;
1538 			if (off + elen > m->m_len)
1539 				goto loopend;
1540 #else
1541 			ext = ip6_pullexthdr(m, off, nxt);
1542 			if (ext == NULL) {
1543 				IP6STAT_INC(ip6s_tooshort);
1544 				return;
1545 			}
1546 			ip6e = mtod(ext, struct ip6_ext *);
1547 			if (nxt == IPPROTO_AH)
1548 				elen = (ip6e->ip6e_len + 2) << 2;
1549 			else
1550 				elen = (ip6e->ip6e_len + 1) << 3;
1551 			if (elen != ext->m_len) {
1552 				m_freem(ext);
1553 				IP6STAT_INC(ip6s_tooshort);
1554 				return;
1555 			}
1556 #endif
1557 
1558 			switch (nxt) {
1559 			case IPPROTO_DSTOPTS:
1560 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1561 					break;
1562 
1563 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1564 				    IS2292(in6p,
1565 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1566 				    IPPROTO_IPV6);
1567 				if (*mp)
1568 					mp = &(*mp)->m_next;
1569 				break;
1570 			case IPPROTO_ROUTING:
1571 				if (!(in6p->inp_flags & IN6P_RTHDR))
1572 					break;
1573 
1574 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1575 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1576 				    IPPROTO_IPV6);
1577 				if (*mp)
1578 					mp = &(*mp)->m_next;
1579 				break;
1580 			case IPPROTO_HOPOPTS:
1581 			case IPPROTO_AH: /* is it possible? */
1582 				break;
1583 
1584 			default:
1585 				/*
1586 				 * other cases have been filtered in the above.
1587 				 * none will visit this case.  here we supply
1588 				 * the code just in case (nxt overwritten or
1589 				 * other cases).
1590 				 */
1591 #ifdef PULLDOWN_TEST
1592 				m_freem(ext);
1593 #endif
1594 				goto loopend;
1595 
1596 			}
1597 
1598 			/* proceed with the next header. */
1599 			off += elen;
1600 			nxt = ip6e->ip6e_nxt;
1601 			ip6e = NULL;
1602 #ifdef PULLDOWN_TEST
1603 			m_freem(ext);
1604 			ext = NULL;
1605 #endif
1606 		}
1607 	  loopend:
1608 		;
1609 	}
1610 }
1611 #undef IS2292
1612 
1613 void
1614 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1615 {
1616 	struct socket *so;
1617 	struct mbuf *m_mtu;
1618 	struct ip6_mtuinfo mtuctl;
1619 
1620 	so =  in6p->inp_socket;
1621 
1622 	if (mtu == NULL)
1623 		return;
1624 
1625 #ifdef DIAGNOSTIC
1626 	if (so == NULL)		/* I believe this is impossible */
1627 		panic("ip6_notify_pmtu: socket is NULL");
1628 #endif
1629 
1630 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1631 	mtuctl.ip6m_mtu = *mtu;
1632 	mtuctl.ip6m_addr = *dst;
1633 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1634 		return;
1635 
1636 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1637 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1638 		return;
1639 
1640 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1641 	    == 0) {
1642 		m_freem(m_mtu);
1643 		/* XXX: should count statistics */
1644 	} else
1645 		sorwakeup(so);
1646 
1647 	return;
1648 }
1649 
1650 #ifdef PULLDOWN_TEST
1651 /*
1652  * pull single extension header from mbuf chain.  returns single mbuf that
1653  * contains the result, or NULL on error.
1654  */
1655 static struct mbuf *
1656 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1657 {
1658 	struct ip6_ext ip6e;
1659 	size_t elen;
1660 	struct mbuf *n;
1661 
1662 #ifdef DIAGNOSTIC
1663 	switch (nxt) {
1664 	case IPPROTO_DSTOPTS:
1665 	case IPPROTO_ROUTING:
1666 	case IPPROTO_HOPOPTS:
1667 	case IPPROTO_AH: /* is it possible? */
1668 		break;
1669 	default:
1670 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1671 	}
1672 #endif
1673 
1674 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1675 	if (nxt == IPPROTO_AH)
1676 		elen = (ip6e.ip6e_len + 2) << 2;
1677 	else
1678 		elen = (ip6e.ip6e_len + 1) << 3;
1679 
1680 	if (elen > MLEN)
1681 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1682 	else
1683 		n = m_get(M_NOWAIT, MT_DATA);
1684 	if (n == NULL)
1685 		return NULL;
1686 
1687 	m_copydata(m, off, elen, mtod(n, caddr_t));
1688 	n->m_len = elen;
1689 	return n;
1690 }
1691 #endif
1692 
1693 /*
1694  * Get pointer to the previous header followed by the header
1695  * currently processed.
1696  * XXX: This function supposes that
1697  *	M includes all headers,
1698  *	the next header field and the header length field of each header
1699  *	are valid, and
1700  *	the sum of each header length equals to OFF.
1701  * Because of these assumptions, this function must be called very
1702  * carefully. Moreover, it will not be used in the near future when
1703  * we develop `neater' mechanism to process extension headers.
1704  */
1705 char *
1706 ip6_get_prevhdr(struct mbuf *m, int off)
1707 {
1708 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1709 
1710 	if (off == sizeof(struct ip6_hdr))
1711 		return (&ip6->ip6_nxt);
1712 	else {
1713 		int len, nxt;
1714 		struct ip6_ext *ip6e = NULL;
1715 
1716 		nxt = ip6->ip6_nxt;
1717 		len = sizeof(struct ip6_hdr);
1718 		while (len < off) {
1719 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1720 
1721 			switch (nxt) {
1722 			case IPPROTO_FRAGMENT:
1723 				len += sizeof(struct ip6_frag);
1724 				break;
1725 			case IPPROTO_AH:
1726 				len += (ip6e->ip6e_len + 2) << 2;
1727 				break;
1728 			default:
1729 				len += (ip6e->ip6e_len + 1) << 3;
1730 				break;
1731 			}
1732 			nxt = ip6e->ip6e_nxt;
1733 		}
1734 		if (ip6e)
1735 			return (&ip6e->ip6e_nxt);
1736 		else
1737 			return NULL;
1738 	}
1739 }
1740 
1741 /*
1742  * get next header offset.  m will be retained.
1743  */
1744 int
1745 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1746 {
1747 	struct ip6_hdr ip6;
1748 	struct ip6_ext ip6e;
1749 	struct ip6_frag fh;
1750 
1751 	/* just in case */
1752 	if (m == NULL)
1753 		panic("ip6_nexthdr: m == NULL");
1754 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1755 		return -1;
1756 
1757 	switch (proto) {
1758 	case IPPROTO_IPV6:
1759 		if (m->m_pkthdr.len < off + sizeof(ip6))
1760 			return -1;
1761 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1762 		if (nxtp)
1763 			*nxtp = ip6.ip6_nxt;
1764 		off += sizeof(ip6);
1765 		return off;
1766 
1767 	case IPPROTO_FRAGMENT:
1768 		/*
1769 		 * terminate parsing if it is not the first fragment,
1770 		 * it does not make sense to parse through it.
1771 		 */
1772 		if (m->m_pkthdr.len < off + sizeof(fh))
1773 			return -1;
1774 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1775 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1776 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1777 			return -1;
1778 		if (nxtp)
1779 			*nxtp = fh.ip6f_nxt;
1780 		off += sizeof(struct ip6_frag);
1781 		return off;
1782 
1783 	case IPPROTO_AH:
1784 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1785 			return -1;
1786 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1787 		if (nxtp)
1788 			*nxtp = ip6e.ip6e_nxt;
1789 		off += (ip6e.ip6e_len + 2) << 2;
1790 		return off;
1791 
1792 	case IPPROTO_HOPOPTS:
1793 	case IPPROTO_ROUTING:
1794 	case IPPROTO_DSTOPTS:
1795 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1796 			return -1;
1797 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1798 		if (nxtp)
1799 			*nxtp = ip6e.ip6e_nxt;
1800 		off += (ip6e.ip6e_len + 1) << 3;
1801 		return off;
1802 
1803 	case IPPROTO_NONE:
1804 	case IPPROTO_ESP:
1805 	case IPPROTO_IPCOMP:
1806 		/* give up */
1807 		return -1;
1808 
1809 	default:
1810 		return -1;
1811 	}
1812 
1813 	return -1;
1814 }
1815 
1816 /*
1817  * get offset for the last header in the chain.  m will be kept untainted.
1818  */
1819 int
1820 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1821 {
1822 	int newoff;
1823 	int nxt;
1824 
1825 	if (!nxtp) {
1826 		nxt = -1;
1827 		nxtp = &nxt;
1828 	}
1829 	while (1) {
1830 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1831 		if (newoff < 0)
1832 			return off;
1833 		else if (newoff < off)
1834 			return -1;	/* invalid */
1835 		else if (newoff == off)
1836 			return newoff;
1837 
1838 		off = newoff;
1839 		proto = *nxtp;
1840 	}
1841 }
1842 
1843 static struct ip6aux *
1844 ip6_addaux(struct mbuf *m)
1845 {
1846 	struct m_tag *mtag;
1847 
1848 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1849 	if (!mtag) {
1850 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1851 		    M_NOWAIT);
1852 		if (mtag) {
1853 			m_tag_prepend(m, mtag);
1854 			bzero(mtag + 1, sizeof(struct ip6aux));
1855 		}
1856 	}
1857 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1858 }
1859 
1860 static struct ip6aux *
1861 ip6_findaux(struct mbuf *m)
1862 {
1863 	struct m_tag *mtag;
1864 
1865 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1866 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1867 }
1868 
1869 static void
1870 ip6_delaux(struct mbuf *m)
1871 {
1872 	struct m_tag *mtag;
1873 
1874 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1875 	if (mtag)
1876 		m_tag_delete(m, mtag);
1877 }
1878 
1879 /*
1880  * System control for IP6
1881  */
1882 
1883 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1884 	0,		0,		0,		0,
1885 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1886 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1887 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1888 	0,		0,		0,		0,
1889 	ENOPROTOOPT
1890 };
1891