xref: /freebsd/sys/netinet6/ip6_input.c (revision 190cef3d)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/hhook.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/proc.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/sdt.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/errno.h>
86 #include <sys/time.h>
87 #include <sys/kernel.h>
88 #include <sys/lock.h>
89 #include <sys/rmlock.h>
90 #include <sys/syslog.h>
91 #include <sys/sysctl.h>
92 
93 #include <net/if.h>
94 #include <net/if_var.h>
95 #include <net/if_types.h>
96 #include <net/if_dl.h>
97 #include <net/route.h>
98 #include <net/netisr.h>
99 #include <net/rss_config.h>
100 #include <net/pfil.h>
101 #include <net/vnet.h>
102 
103 #include <netinet/in.h>
104 #include <netinet/in_kdtrace.h>
105 #include <netinet/ip_var.h>
106 #include <netinet/in_systm.h>
107 #include <net/if_llatbl.h>
108 #ifdef INET
109 #include <netinet/ip.h>
110 #include <netinet/ip_icmp.h>
111 #endif /* INET */
112 #include <netinet/ip6.h>
113 #include <netinet6/in6_var.h>
114 #include <netinet6/ip6_var.h>
115 #include <netinet/in_pcb.h>
116 #include <netinet/icmp6.h>
117 #include <netinet6/scope6_var.h>
118 #include <netinet6/in6_ifattach.h>
119 #include <netinet6/mld6_var.h>
120 #include <netinet6/nd6.h>
121 #include <netinet6/in6_rss.h>
122 
123 #include <netipsec/ipsec_support.h>
124 
125 #include <netinet6/ip6protosw.h>
126 
127 extern struct domain inet6domain;
128 
129 u_char ip6_protox[IPPROTO_MAX];
130 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
131 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
132 VNET_DEFINE(u_long, in6_ifaddrhmask);
133 
134 static struct netisr_handler ip6_nh = {
135 	.nh_name = "ip6",
136 	.nh_handler = ip6_input,
137 	.nh_proto = NETISR_IPV6,
138 #ifdef RSS
139 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
140 	.nh_policy = NETISR_POLICY_CPU,
141 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
142 #else
143 	.nh_policy = NETISR_POLICY_FLOW,
144 #endif
145 };
146 
147 static int
148 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
149 {
150 	int error, qlimit;
151 
152 	netisr_getqlimit(&ip6_nh, &qlimit);
153 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
154 	if (error || !req->newptr)
155 		return (error);
156 	if (qlimit < 1)
157 		return (EINVAL);
158 	return (netisr_setqlimit(&ip6_nh, qlimit));
159 }
160 SYSCTL_DECL(_net_inet6_ip6);
161 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
162     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
163     "Maximum size of the IPv6 input queue");
164 
165 #ifdef RSS
166 static struct netisr_handler ip6_direct_nh = {
167 	.nh_name = "ip6_direct",
168 	.nh_handler = ip6_direct_input,
169 	.nh_proto = NETISR_IPV6_DIRECT,
170 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
171 	.nh_policy = NETISR_POLICY_CPU,
172 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
173 };
174 
175 static int
176 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
177 {
178 	int error, qlimit;
179 
180 	netisr_getqlimit(&ip6_direct_nh, &qlimit);
181 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
182 	if (error || !req->newptr)
183 		return (error);
184 	if (qlimit < 1)
185 		return (EINVAL);
186 	return (netisr_setqlimit(&ip6_direct_nh, qlimit));
187 }
188 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
189     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet6_intr_direct_queue_maxlen,
190     "I", "Maximum size of the IPv6 direct input queue");
191 
192 #endif
193 
194 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
195 
196 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
197 VNET_PCPUSTAT_SYSINIT(ip6stat);
198 #ifdef VIMAGE
199 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
200 #endif /* VIMAGE */
201 
202 struct rmlock in6_ifaddr_lock;
203 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
204 
205 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
206 #ifdef PULLDOWN_TEST
207 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
208 #endif
209 
210 /*
211  * IP6 initialization: fill in IP6 protocol switch table.
212  * All protocols not implemented in kernel go to raw IP6 protocol handler.
213  */
214 void
215 ip6_init(void)
216 {
217 	struct protosw *pr;
218 	int i;
219 
220 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
221 	    &V_ip6_auto_linklocal);
222 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
223 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
224 
225 	CK_STAILQ_INIT(&V_in6_ifaddrhead);
226 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
227 	    &V_in6_ifaddrhmask);
228 
229 	/* Initialize packet filter hooks. */
230 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
231 	V_inet6_pfil_hook.ph_af = AF_INET6;
232 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
233 		printf("%s: WARNING: unable to register pfil hook, "
234 			"error %d\n", __func__, i);
235 
236 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
237 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
238 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
239 		printf("%s: WARNING: unable to register input helper hook\n",
240 		    __func__);
241 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
242 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
243 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
244 		printf("%s: WARNING: unable to register output helper hook\n",
245 		    __func__);
246 
247 	scope6_init();
248 	addrsel_policy_init();
249 	nd6_init();
250 	frag6_init();
251 
252 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
253 
254 	/* Skip global initialization stuff for non-default instances. */
255 #ifdef VIMAGE
256 	if (!IS_DEFAULT_VNET(curvnet)) {
257 		netisr_register_vnet(&ip6_nh);
258 #ifdef RSS
259 		netisr_register_vnet(&ip6_direct_nh);
260 #endif
261 		return;
262 	}
263 #endif
264 
265 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
266 	if (pr == NULL)
267 		panic("ip6_init");
268 
269 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
270 	for (i = 0; i < IPPROTO_MAX; i++)
271 		ip6_protox[i] = pr - inet6sw;
272 	/*
273 	 * Cycle through IP protocols and put them into the appropriate place
274 	 * in ip6_protox[].
275 	 */
276 	for (pr = inet6domain.dom_protosw;
277 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
278 		if (pr->pr_domain->dom_family == PF_INET6 &&
279 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
280 			/* Be careful to only index valid IP protocols. */
281 			if (pr->pr_protocol < IPPROTO_MAX)
282 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
283 		}
284 
285 	netisr_register(&ip6_nh);
286 #ifdef RSS
287 	netisr_register(&ip6_direct_nh);
288 #endif
289 }
290 
291 /*
292  * The protocol to be inserted into ip6_protox[] must be already registered
293  * in inet6sw[], either statically or through pf_proto_register().
294  */
295 int
296 ip6proto_register(short ip6proto)
297 {
298 	struct protosw *pr;
299 
300 	/* Sanity checks. */
301 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
302 		return (EPROTONOSUPPORT);
303 
304 	/*
305 	 * The protocol slot must not be occupied by another protocol
306 	 * already.  An index pointing to IPPROTO_RAW is unused.
307 	 */
308 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
309 	if (pr == NULL)
310 		return (EPFNOSUPPORT);
311 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
312 		return (EEXIST);
313 
314 	/*
315 	 * Find the protocol position in inet6sw[] and set the index.
316 	 */
317 	for (pr = inet6domain.dom_protosw;
318 	    pr < inet6domain.dom_protoswNPROTOSW; pr++) {
319 		if (pr->pr_domain->dom_family == PF_INET6 &&
320 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
321 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
322 			return (0);
323 		}
324 	}
325 	return (EPROTONOSUPPORT);
326 }
327 
328 int
329 ip6proto_unregister(short ip6proto)
330 {
331 	struct protosw *pr;
332 
333 	/* Sanity checks. */
334 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
335 		return (EPROTONOSUPPORT);
336 
337 	/* Check if the protocol was indeed registered. */
338 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
339 	if (pr == NULL)
340 		return (EPFNOSUPPORT);
341 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
342 		return (ENOENT);
343 
344 	/* Reset the protocol slot to IPPROTO_RAW. */
345 	ip6_protox[ip6proto] = pr - inet6sw;
346 	return (0);
347 }
348 
349 #ifdef VIMAGE
350 static void
351 ip6_destroy(void *unused __unused)
352 {
353 	struct ifaddr *ifa, *nifa;
354 	struct ifnet *ifp;
355 	int error;
356 
357 #ifdef RSS
358 	netisr_unregister_vnet(&ip6_direct_nh);
359 #endif
360 	netisr_unregister_vnet(&ip6_nh);
361 
362 	if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
363 		printf("%s: WARNING: unable to unregister pfil hook, "
364 		    "error %d\n", __func__, error);
365 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
366 	if (error != 0) {
367 		printf("%s: WARNING: unable to deregister input helper hook "
368 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
369 		    "error %d returned\n", __func__, error);
370 	}
371 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
372 	if (error != 0) {
373 		printf("%s: WARNING: unable to deregister output helper hook "
374 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
375 		    "error %d returned\n", __func__, error);
376 	}
377 
378 	/* Cleanup addresses. */
379 	IFNET_RLOCK();
380 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
381 		/* Cannot lock here - lock recursion. */
382 		/* IF_ADDR_LOCK(ifp); */
383 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
384 
385 			if (ifa->ifa_addr->sa_family != AF_INET6)
386 				continue;
387 			in6_purgeaddr(ifa);
388 		}
389 		/* IF_ADDR_UNLOCK(ifp); */
390 		in6_ifdetach_destroy(ifp);
391 		mld_domifdetach(ifp);
392 		/* Make sure any routes are gone as well. */
393 		rt_flushifroutes_af(ifp, AF_INET6);
394 	}
395 	IFNET_RUNLOCK();
396 
397 	nd6_destroy();
398 	in6_ifattach_destroy();
399 
400 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
401 }
402 
403 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
404 #endif
405 
406 static int
407 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
408     int *nxt, int *ours)
409 {
410 	struct ip6_hdr *ip6;
411 	struct ip6_hbh *hbh;
412 
413 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
414 #if 0	/*touches NULL pointer*/
415 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
416 #endif
417 		goto out;	/* m have already been freed */
418 	}
419 
420 	/* adjust pointer */
421 	ip6 = mtod(m, struct ip6_hdr *);
422 
423 	/*
424 	 * if the payload length field is 0 and the next header field
425 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
426 	 * option MUST be included.
427 	 */
428 	if (ip6->ip6_plen == 0 && *plen == 0) {
429 		/*
430 		 * Note that if a valid jumbo payload option is
431 		 * contained, ip6_hopopts_input() must set a valid
432 		 * (non-zero) payload length to the variable plen.
433 		 */
434 		IP6STAT_INC(ip6s_badoptions);
435 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
436 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
437 		icmp6_error(m, ICMP6_PARAM_PROB,
438 			    ICMP6_PARAMPROB_HEADER,
439 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
440 		goto out;
441 	}
442 #ifndef PULLDOWN_TEST
443 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
444 	hbh = (struct ip6_hbh *)(ip6 + 1);
445 #else
446 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
447 		sizeof(struct ip6_hbh));
448 	if (hbh == NULL) {
449 		IP6STAT_INC(ip6s_tooshort);
450 		goto out;
451 	}
452 #endif
453 	*nxt = hbh->ip6h_nxt;
454 
455 	/*
456 	 * If we are acting as a router and the packet contains a
457 	 * router alert option, see if we know the option value.
458 	 * Currently, we only support the option value for MLD, in which
459 	 * case we should pass the packet to the multicast routing
460 	 * daemon.
461 	 */
462 	if (*rtalert != ~0) {
463 		switch (*rtalert) {
464 		case IP6OPT_RTALERT_MLD:
465 			if (V_ip6_forwarding)
466 				*ours = 1;
467 			break;
468 		default:
469 			/*
470 			 * RFC2711 requires unrecognized values must be
471 			 * silently ignored.
472 			 */
473 			break;
474 		}
475 	}
476 
477 	return (0);
478 
479 out:
480 	return (1);
481 }
482 
483 #ifdef RSS
484 /*
485  * IPv6 direct input routine.
486  *
487  * This is called when reinjecting completed fragments where
488  * all of the previous checking and book-keeping has been done.
489  */
490 void
491 ip6_direct_input(struct mbuf *m)
492 {
493 	int off, nxt;
494 	int nest;
495 	struct m_tag *mtag;
496 	struct ip6_direct_ctx *ip6dc;
497 
498 	mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
499 	KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
500 
501 	ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
502 	nxt = ip6dc->ip6dc_nxt;
503 	off = ip6dc->ip6dc_off;
504 
505 	nest = 0;
506 
507 	m_tag_delete(m, mtag);
508 
509 	while (nxt != IPPROTO_DONE) {
510 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
511 			IP6STAT_INC(ip6s_toomanyhdr);
512 			goto bad;
513 		}
514 
515 		/*
516 		 * protection against faulty packet - there should be
517 		 * more sanity checks in header chain processing.
518 		 */
519 		if (m->m_pkthdr.len < off) {
520 			IP6STAT_INC(ip6s_tooshort);
521 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
522 			goto bad;
523 		}
524 
525 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
526 		if (IPSEC_ENABLED(ipv6)) {
527 			if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
528 				return;
529 		}
530 #endif /* IPSEC */
531 
532 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
533 	}
534 	return;
535 bad:
536 	m_freem(m);
537 }
538 #endif
539 
540 void
541 ip6_input(struct mbuf *m)
542 {
543 	struct in6_addr odst;
544 	struct ip6_hdr *ip6;
545 	struct in6_ifaddr *ia;
546 	struct ifnet *rcvif;
547 	u_int32_t plen;
548 	u_int32_t rtalert = ~0;
549 	int off = sizeof(struct ip6_hdr), nest;
550 	int nxt, ours = 0;
551 	int srcrt = 0;
552 
553 	/*
554 	 * Drop the packet if IPv6 operation is disabled on the interface.
555 	 */
556 	rcvif = m->m_pkthdr.rcvif;
557 	if ((ND_IFINFO(rcvif)->flags & ND6_IFF_IFDISABLED))
558 		goto bad;
559 
560 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
561 	/*
562 	 * should the inner packet be considered authentic?
563 	 * see comment in ah4_input().
564 	 * NB: m cannot be NULL when passed to the input routine
565 	 */
566 
567 	m->m_flags &= ~M_AUTHIPHDR;
568 	m->m_flags &= ~M_AUTHIPDGM;
569 
570 #endif /* IPSEC */
571 
572 	if (m->m_flags & M_FASTFWD_OURS) {
573 		/*
574 		 * Firewall changed destination to local.
575 		 */
576 		ip6 = mtod(m, struct ip6_hdr *);
577 		goto passin;
578 	}
579 
580 	/*
581 	 * mbuf statistics
582 	 */
583 	if (m->m_flags & M_EXT) {
584 		if (m->m_next)
585 			IP6STAT_INC(ip6s_mext2m);
586 		else
587 			IP6STAT_INC(ip6s_mext1);
588 	} else {
589 		if (m->m_next) {
590 			if (m->m_flags & M_LOOP) {
591 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
592 			} else if (rcvif->if_index < IP6S_M2MMAX)
593 				IP6STAT_INC(ip6s_m2m[rcvif->if_index]);
594 			else
595 				IP6STAT_INC(ip6s_m2m[0]);
596 		} else
597 			IP6STAT_INC(ip6s_m1);
598 	}
599 
600 	in6_ifstat_inc(rcvif, ifs6_in_receive);
601 	IP6STAT_INC(ip6s_total);
602 
603 #ifndef PULLDOWN_TEST
604 	/*
605 	 * L2 bridge code and some other code can return mbuf chain
606 	 * that does not conform to KAME requirement.  too bad.
607 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
608 	 */
609 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
610 		struct mbuf *n;
611 
612 		if (m->m_pkthdr.len > MHLEN)
613 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
614 		else
615 			n = m_gethdr(M_NOWAIT, MT_DATA);
616 		if (n == NULL)
617 			goto bad;
618 
619 		m_move_pkthdr(n, m);
620 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
621 		n->m_len = n->m_pkthdr.len;
622 		m_freem(m);
623 		m = n;
624 	}
625 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
626 #endif
627 
628 	if (m->m_len < sizeof(struct ip6_hdr)) {
629 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
630 			IP6STAT_INC(ip6s_toosmall);
631 			in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
632 			goto bad;
633 		}
634 	}
635 
636 	ip6 = mtod(m, struct ip6_hdr *);
637 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
638 		IP6STAT_INC(ip6s_badvers);
639 		in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
640 		goto bad;
641 	}
642 
643 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
644 	IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
645 
646 	/*
647 	 * Check against address spoofing/corruption.
648 	 */
649 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
650 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
651 		/*
652 		 * XXX: "badscope" is not very suitable for a multicast source.
653 		 */
654 		IP6STAT_INC(ip6s_badscope);
655 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
656 		goto bad;
657 	}
658 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
659 	    !(m->m_flags & M_LOOP)) {
660 		/*
661 		 * In this case, the packet should come from the loopback
662 		 * interface.  However, we cannot just check the if_flags,
663 		 * because ip6_mloopback() passes the "actual" interface
664 		 * as the outgoing/incoming interface.
665 		 */
666 		IP6STAT_INC(ip6s_badscope);
667 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
668 		goto bad;
669 	}
670 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
671 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
672 		/*
673 		 * RFC4291 2.7:
674 		 * Nodes must not originate a packet to a multicast address
675 		 * whose scop field contains the reserved value 0; if such
676 		 * a packet is received, it must be silently dropped.
677 		 */
678 		IP6STAT_INC(ip6s_badscope);
679 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
680 		goto bad;
681 	}
682 #ifdef ALTQ
683 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
684 		/* packet is dropped by traffic conditioner */
685 		return;
686 	}
687 #endif
688 	/*
689 	 * The following check is not documented in specs.  A malicious
690 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
691 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
692 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
693 	 *
694 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
695 	 * support IPv4-less kernel compilation, we cannot support SIIT
696 	 * environment at all.  So, it makes more sense for us to reject any
697 	 * malicious packets for non-SIIT environment, than try to do a
698 	 * partial support for SIIT environment.
699 	 */
700 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
701 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
702 		IP6STAT_INC(ip6s_badscope);
703 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
704 		goto bad;
705 	}
706 #if 0
707 	/*
708 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
709 	 *
710 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
711 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
712 	 * is revised to forbid relaying case.
713 	 */
714 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
715 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
716 		IP6STAT_INC(ip6s_badscope);
717 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
718 		goto bad;
719 	}
720 #endif
721 	/*
722 	 * Try to forward the packet, but if we fail continue.
723 	 * ip6_tryforward() does not generate redirects, so fall
724 	 * through to normal processing if redirects are required.
725 	 * ip6_tryforward() does inbound and outbound packet firewall
726 	 * processing. If firewall has decided that destination becomes
727 	 * our local address, it sets M_FASTFWD_OURS flag. In this
728 	 * case skip another inbound firewall processing and update
729 	 * ip6 pointer.
730 	 */
731 	if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
732 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
733 	    && (!IPSEC_ENABLED(ipv6) ||
734 	    IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
735 #endif
736 	    ) {
737 		if ((m = ip6_tryforward(m)) == NULL)
738 			return;
739 		if (m->m_flags & M_FASTFWD_OURS) {
740 			ip6 = mtod(m, struct ip6_hdr *);
741 			goto passin;
742 		}
743 	}
744 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
745 	/*
746 	 * Bypass packet filtering for packets previously handled by IPsec.
747 	 */
748 	if (IPSEC_ENABLED(ipv6) &&
749 	    IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
750 			goto passin;
751 #endif
752 	/*
753 	 * Run through list of hooks for input packets.
754 	 *
755 	 * NB: Beware of the destination address changing
756 	 *     (e.g. by NAT rewriting).  When this happens,
757 	 *     tell ip6_forward to do the right thing.
758 	 */
759 
760 	/* Jump over all PFIL processing if hooks are not active. */
761 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
762 		goto passin;
763 
764 	odst = ip6->ip6_dst;
765 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
766 	    m->m_pkthdr.rcvif, PFIL_IN, 0, NULL))
767 		return;
768 	if (m == NULL)			/* consumed by filter */
769 		return;
770 	ip6 = mtod(m, struct ip6_hdr *);
771 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
772 	if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
773 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
774 		/*
775 		 * Directly ship the packet on.  This allows forwarding
776 		 * packets originally destined to us to some other directly
777 		 * connected host.
778 		 */
779 		ip6_forward(m, 1);
780 		return;
781 	}
782 
783 passin:
784 	/*
785 	 * Disambiguate address scope zones (if there is ambiguity).
786 	 * We first make sure that the original source or destination address
787 	 * is not in our internal form for scoped addresses.  Such addresses
788 	 * are not necessarily invalid spec-wise, but we cannot accept them due
789 	 * to the usage conflict.
790 	 * in6_setscope() then also checks and rejects the cases where src or
791 	 * dst are the loopback address and the receiving interface
792 	 * is not loopback.
793 	 */
794 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
795 		IP6STAT_INC(ip6s_badscope); /* XXX */
796 		goto bad;
797 	}
798 	if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
799 	    in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
800 		IP6STAT_INC(ip6s_badscope);
801 		goto bad;
802 	}
803 	if (m->m_flags & M_FASTFWD_OURS) {
804 		m->m_flags &= ~M_FASTFWD_OURS;
805 		ours = 1;
806 		goto hbhcheck;
807 	}
808 	/*
809 	 * Multicast check. Assume packet is for us to avoid
810 	 * prematurely taking locks.
811 	 */
812 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
813 		ours = 1;
814 		in6_ifstat_inc(rcvif, ifs6_in_mcast);
815 		goto hbhcheck;
816 	}
817 	/*
818 	 * Unicast check
819 	 * XXX: For now we keep link-local IPv6 addresses with embedded
820 	 *      scope zone id, therefore we use zero zoneid here.
821 	 */
822 	ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
823 	if (ia != NULL) {
824 		if (ia->ia6_flags & IN6_IFF_NOTREADY) {
825 			char ip6bufs[INET6_ADDRSTRLEN];
826 			char ip6bufd[INET6_ADDRSTRLEN];
827 			/* address is not ready, so discard the packet. */
828 			nd6log((LOG_INFO,
829 			    "ip6_input: packet to an unready address %s->%s\n",
830 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
831 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
832 			ifa_free(&ia->ia_ifa);
833 			goto bad;
834 		}
835 		/* Count the packet in the ip address stats */
836 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
837 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
838 		ifa_free(&ia->ia_ifa);
839 		ours = 1;
840 		goto hbhcheck;
841 	}
842 
843 	/*
844 	 * Now there is no reason to process the packet if it's not our own
845 	 * and we're not a router.
846 	 */
847 	if (!V_ip6_forwarding) {
848 		IP6STAT_INC(ip6s_cantforward);
849 		goto bad;
850 	}
851 
852   hbhcheck:
853 	/*
854 	 * Process Hop-by-Hop options header if it's contained.
855 	 * m may be modified in ip6_hopopts_input().
856 	 * If a JumboPayload option is included, plen will also be modified.
857 	 */
858 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
859 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
860 		if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0)
861 			return;
862 	} else
863 		nxt = ip6->ip6_nxt;
864 
865 	/*
866 	 * Use mbuf flags to propagate Router Alert option to
867 	 * ICMPv6 layer, as hop-by-hop options have been stripped.
868 	 */
869 	if (rtalert != ~0)
870 		m->m_flags |= M_RTALERT_MLD;
871 
872 	/*
873 	 * Check that the amount of data in the buffers
874 	 * is as at least much as the IPv6 header would have us expect.
875 	 * Trim mbufs if longer than we expect.
876 	 * Drop packet if shorter than we expect.
877 	 */
878 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
879 		IP6STAT_INC(ip6s_tooshort);
880 		in6_ifstat_inc(rcvif, ifs6_in_truncated);
881 		goto bad;
882 	}
883 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
884 		if (m->m_len == m->m_pkthdr.len) {
885 			m->m_len = sizeof(struct ip6_hdr) + plen;
886 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
887 		} else
888 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
889 	}
890 
891 	/*
892 	 * Forward if desirable.
893 	 */
894 	if (V_ip6_mrouter &&
895 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
896 		/*
897 		 * If we are acting as a multicast router, all
898 		 * incoming multicast packets are passed to the
899 		 * kernel-level multicast forwarding function.
900 		 * The packet is returned (relatively) intact; if
901 		 * ip6_mforward() returns a non-zero value, the packet
902 		 * must be discarded, else it may be accepted below.
903 		 *
904 		 * XXX TODO: Check hlim and multicast scope here to avoid
905 		 * unnecessarily calling into ip6_mforward().
906 		 */
907 		if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
908 			IP6STAT_INC(ip6s_cantforward);
909 			goto bad;
910 		}
911 	} else if (!ours) {
912 		ip6_forward(m, srcrt);
913 		return;
914 	}
915 
916 	ip6 = mtod(m, struct ip6_hdr *);
917 
918 	/*
919 	 * Malicious party may be able to use IPv4 mapped addr to confuse
920 	 * tcp/udp stack and bypass security checks (act as if it was from
921 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
922 	 *
923 	 * For SIIT end node behavior, you may want to disable the check.
924 	 * However, you will  become vulnerable to attacks using IPv4 mapped
925 	 * source.
926 	 */
927 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
928 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
929 		IP6STAT_INC(ip6s_badscope);
930 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
931 		goto bad;
932 	}
933 
934 	/*
935 	 * Tell launch routine the next header
936 	 */
937 	IP6STAT_INC(ip6s_delivered);
938 	in6_ifstat_inc(rcvif, ifs6_in_deliver);
939 	nest = 0;
940 
941 	while (nxt != IPPROTO_DONE) {
942 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
943 			IP6STAT_INC(ip6s_toomanyhdr);
944 			goto bad;
945 		}
946 
947 		/*
948 		 * protection against faulty packet - there should be
949 		 * more sanity checks in header chain processing.
950 		 */
951 		if (m->m_pkthdr.len < off) {
952 			IP6STAT_INC(ip6s_tooshort);
953 			in6_ifstat_inc(rcvif, ifs6_in_truncated);
954 			goto bad;
955 		}
956 
957 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
958 		if (IPSEC_ENABLED(ipv6)) {
959 			if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
960 				return;
961 		}
962 #endif /* IPSEC */
963 
964 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
965 	}
966 	return;
967 bad:
968 	in6_ifstat_inc(rcvif, ifs6_in_discard);
969 	if (m != NULL)
970 		m_freem(m);
971 }
972 
973 /*
974  * Hop-by-Hop options header processing. If a valid jumbo payload option is
975  * included, the real payload length will be stored in plenp.
976  *
977  * rtalertp - XXX: should be stored more smart way
978  */
979 static int
980 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
981     struct mbuf **mp, int *offp)
982 {
983 	struct mbuf *m = *mp;
984 	int off = *offp, hbhlen;
985 	struct ip6_hbh *hbh;
986 
987 	/* validation of the length of the header */
988 #ifndef PULLDOWN_TEST
989 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
990 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
991 	hbhlen = (hbh->ip6h_len + 1) << 3;
992 
993 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
994 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
995 #else
996 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
997 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
998 	if (hbh == NULL) {
999 		IP6STAT_INC(ip6s_tooshort);
1000 		return -1;
1001 	}
1002 	hbhlen = (hbh->ip6h_len + 1) << 3;
1003 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1004 		hbhlen);
1005 	if (hbh == NULL) {
1006 		IP6STAT_INC(ip6s_tooshort);
1007 		return -1;
1008 	}
1009 #endif
1010 	off += hbhlen;
1011 	hbhlen -= sizeof(struct ip6_hbh);
1012 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1013 				hbhlen, rtalertp, plenp) < 0)
1014 		return (-1);
1015 
1016 	*offp = off;
1017 	*mp = m;
1018 	return (0);
1019 }
1020 
1021 /*
1022  * Search header for all Hop-by-hop options and process each option.
1023  * This function is separate from ip6_hopopts_input() in order to
1024  * handle a case where the sending node itself process its hop-by-hop
1025  * options header. In such a case, the function is called from ip6_output().
1026  *
1027  * The function assumes that hbh header is located right after the IPv6 header
1028  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1029  * opthead + hbhlen is located in contiguous memory region.
1030  */
1031 int
1032 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1033     u_int32_t *rtalertp, u_int32_t *plenp)
1034 {
1035 	struct ip6_hdr *ip6;
1036 	int optlen = 0;
1037 	u_int8_t *opt = opthead;
1038 	u_int16_t rtalert_val;
1039 	u_int32_t jumboplen;
1040 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1041 
1042 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1043 		switch (*opt) {
1044 		case IP6OPT_PAD1:
1045 			optlen = 1;
1046 			break;
1047 		case IP6OPT_PADN:
1048 			if (hbhlen < IP6OPT_MINLEN) {
1049 				IP6STAT_INC(ip6s_toosmall);
1050 				goto bad;
1051 			}
1052 			optlen = *(opt + 1) + 2;
1053 			break;
1054 		case IP6OPT_ROUTER_ALERT:
1055 			/* XXX may need check for alignment */
1056 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1057 				IP6STAT_INC(ip6s_toosmall);
1058 				goto bad;
1059 			}
1060 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1061 				/* XXX stat */
1062 				icmp6_error(m, ICMP6_PARAM_PROB,
1063 				    ICMP6_PARAMPROB_HEADER,
1064 				    erroff + opt + 1 - opthead);
1065 				return (-1);
1066 			}
1067 			optlen = IP6OPT_RTALERT_LEN;
1068 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1069 			*rtalertp = ntohs(rtalert_val);
1070 			break;
1071 		case IP6OPT_JUMBO:
1072 			/* XXX may need check for alignment */
1073 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1074 				IP6STAT_INC(ip6s_toosmall);
1075 				goto bad;
1076 			}
1077 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1078 				/* XXX stat */
1079 				icmp6_error(m, ICMP6_PARAM_PROB,
1080 				    ICMP6_PARAMPROB_HEADER,
1081 				    erroff + opt + 1 - opthead);
1082 				return (-1);
1083 			}
1084 			optlen = IP6OPT_JUMBO_LEN;
1085 
1086 			/*
1087 			 * IPv6 packets that have non 0 payload length
1088 			 * must not contain a jumbo payload option.
1089 			 */
1090 			ip6 = mtod(m, struct ip6_hdr *);
1091 			if (ip6->ip6_plen) {
1092 				IP6STAT_INC(ip6s_badoptions);
1093 				icmp6_error(m, ICMP6_PARAM_PROB,
1094 				    ICMP6_PARAMPROB_HEADER,
1095 				    erroff + opt - opthead);
1096 				return (-1);
1097 			}
1098 
1099 			/*
1100 			 * We may see jumbolen in unaligned location, so
1101 			 * we'd need to perform bcopy().
1102 			 */
1103 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1104 			jumboplen = (u_int32_t)htonl(jumboplen);
1105 
1106 #if 1
1107 			/*
1108 			 * if there are multiple jumbo payload options,
1109 			 * *plenp will be non-zero and the packet will be
1110 			 * rejected.
1111 			 * the behavior may need some debate in ipngwg -
1112 			 * multiple options does not make sense, however,
1113 			 * there's no explicit mention in specification.
1114 			 */
1115 			if (*plenp != 0) {
1116 				IP6STAT_INC(ip6s_badoptions);
1117 				icmp6_error(m, ICMP6_PARAM_PROB,
1118 				    ICMP6_PARAMPROB_HEADER,
1119 				    erroff + opt + 2 - opthead);
1120 				return (-1);
1121 			}
1122 #endif
1123 
1124 			/*
1125 			 * jumbo payload length must be larger than 65535.
1126 			 */
1127 			if (jumboplen <= IPV6_MAXPACKET) {
1128 				IP6STAT_INC(ip6s_badoptions);
1129 				icmp6_error(m, ICMP6_PARAM_PROB,
1130 				    ICMP6_PARAMPROB_HEADER,
1131 				    erroff + opt + 2 - opthead);
1132 				return (-1);
1133 			}
1134 			*plenp = jumboplen;
1135 
1136 			break;
1137 		default:		/* unknown option */
1138 			if (hbhlen < IP6OPT_MINLEN) {
1139 				IP6STAT_INC(ip6s_toosmall);
1140 				goto bad;
1141 			}
1142 			optlen = ip6_unknown_opt(opt, m,
1143 			    erroff + opt - opthead);
1144 			if (optlen == -1)
1145 				return (-1);
1146 			optlen += 2;
1147 			break;
1148 		}
1149 	}
1150 
1151 	return (0);
1152 
1153   bad:
1154 	m_freem(m);
1155 	return (-1);
1156 }
1157 
1158 /*
1159  * Unknown option processing.
1160  * The third argument `off' is the offset from the IPv6 header to the option,
1161  * which is necessary if the IPv6 header the and option header and IPv6 header
1162  * is not contiguous in order to return an ICMPv6 error.
1163  */
1164 int
1165 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1166 {
1167 	struct ip6_hdr *ip6;
1168 
1169 	switch (IP6OPT_TYPE(*optp)) {
1170 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1171 		return ((int)*(optp + 1));
1172 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1173 		m_freem(m);
1174 		return (-1);
1175 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1176 		IP6STAT_INC(ip6s_badoptions);
1177 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1178 		return (-1);
1179 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1180 		IP6STAT_INC(ip6s_badoptions);
1181 		ip6 = mtod(m, struct ip6_hdr *);
1182 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1183 		    (m->m_flags & (M_BCAST|M_MCAST)))
1184 			m_freem(m);
1185 		else
1186 			icmp6_error(m, ICMP6_PARAM_PROB,
1187 				    ICMP6_PARAMPROB_OPTION, off);
1188 		return (-1);
1189 	}
1190 
1191 	m_freem(m);		/* XXX: NOTREACHED */
1192 	return (-1);
1193 }
1194 
1195 /*
1196  * Create the "control" list for this pcb.
1197  * These functions will not modify mbuf chain at all.
1198  *
1199  * With KAME mbuf chain restriction:
1200  * The routine will be called from upper layer handlers like tcp6_input().
1201  * Thus the routine assumes that the caller (tcp6_input) have already
1202  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1203  * very first mbuf on the mbuf chain.
1204  *
1205  * ip6_savecontrol_v4 will handle those options that are possible to be
1206  * set on a v4-mapped socket.
1207  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1208  * options and handle the v6-only ones itself.
1209  */
1210 struct mbuf **
1211 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1212     int *v4only)
1213 {
1214 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1215 
1216 #ifdef SO_TIMESTAMP
1217 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1218 		union {
1219 			struct timeval tv;
1220 			struct bintime bt;
1221 			struct timespec ts;
1222 		} t;
1223 		struct bintime boottimebin, bt1;
1224 		struct timespec ts1;
1225 		bool stamped;
1226 
1227 		stamped = false;
1228 		switch (inp->inp_socket->so_ts_clock) {
1229 		case SO_TS_REALTIME_MICRO:
1230 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1231 			    M_TSTMP)) {
1232 				mbuf_tstmp2timespec(m, &ts1);
1233 				timespec2bintime(&ts1, &bt1);
1234 				getboottimebin(&boottimebin);
1235 				bintime_add(&bt1, &boottimebin);
1236 				bintime2timeval(&bt1, &t.tv);
1237 			} else {
1238 				microtime(&t.tv);
1239 			}
1240 			*mp = sbcreatecontrol((caddr_t) &t.tv, sizeof(t.tv),
1241 			    SCM_TIMESTAMP, SOL_SOCKET);
1242 			if (*mp != NULL) {
1243 				mp = &(*mp)->m_next;
1244 				stamped = true;
1245 			}
1246 			break;
1247 
1248 		case SO_TS_BINTIME:
1249 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1250 			    M_TSTMP)) {
1251 				mbuf_tstmp2timespec(m, &ts1);
1252 				timespec2bintime(&ts1, &t.bt);
1253 				getboottimebin(&boottimebin);
1254 				bintime_add(&t.bt, &boottimebin);
1255 			} else {
1256 				bintime(&t.bt);
1257 			}
1258 			*mp = sbcreatecontrol((caddr_t)&t.bt, sizeof(t.bt),
1259 			    SCM_BINTIME, SOL_SOCKET);
1260 			if (*mp != NULL) {
1261 				mp = &(*mp)->m_next;
1262 				stamped = true;
1263 			}
1264 			break;
1265 
1266 		case SO_TS_REALTIME:
1267 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1268 			    M_TSTMP)) {
1269 				mbuf_tstmp2timespec(m, &t.ts);
1270 				getboottimebin(&boottimebin);
1271 				bintime2timespec(&boottimebin, &ts1);
1272 				timespecadd(&t.ts, &ts1, &t.ts);
1273 			} else {
1274 				nanotime(&t.ts);
1275 			}
1276 			*mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1277 			    SCM_REALTIME, SOL_SOCKET);
1278 			if (*mp != NULL) {
1279 				mp = &(*mp)->m_next;
1280 				stamped = true;
1281 			}
1282 			break;
1283 
1284 		case SO_TS_MONOTONIC:
1285 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1286 			    M_TSTMP))
1287 				mbuf_tstmp2timespec(m, &t.ts);
1288 			else
1289 				nanouptime(&t.ts);
1290 			*mp = sbcreatecontrol((caddr_t)&t.ts, sizeof(t.ts),
1291 			    SCM_MONOTONIC, SOL_SOCKET);
1292 			if (*mp != NULL) {
1293 				mp = &(*mp)->m_next;
1294 				stamped = true;
1295 			}
1296 			break;
1297 
1298 		default:
1299 			panic("unknown (corrupted) so_ts_clock");
1300 		}
1301 		if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1302 		    (M_PKTHDR | M_TSTMP)) {
1303 			struct sock_timestamp_info sti;
1304 
1305 			bzero(&sti, sizeof(sti));
1306 			sti.st_info_flags = ST_INFO_HW;
1307 			if ((m->m_flags & M_TSTMP_HPREC) != 0)
1308 				sti.st_info_flags |= ST_INFO_HW_HPREC;
1309 			*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti),
1310 			    SCM_TIME_INFO, SOL_SOCKET);
1311 			if (*mp != NULL)
1312 				mp = &(*mp)->m_next;
1313 		}
1314 	}
1315 #endif
1316 
1317 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1318 	/* RFC 2292 sec. 5 */
1319 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1320 		struct in6_pktinfo pi6;
1321 
1322 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1323 #ifdef INET
1324 			struct ip *ip;
1325 
1326 			ip = mtod(m, struct ip *);
1327 			pi6.ipi6_addr.s6_addr32[0] = 0;
1328 			pi6.ipi6_addr.s6_addr32[1] = 0;
1329 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1330 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1331 #else
1332 			/* We won't hit this code */
1333 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1334 #endif
1335 		} else {
1336 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1337 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1338 		}
1339 		pi6.ipi6_ifindex =
1340 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1341 
1342 		*mp = sbcreatecontrol((caddr_t) &pi6,
1343 		    sizeof(struct in6_pktinfo),
1344 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1345 		if (*mp)
1346 			mp = &(*mp)->m_next;
1347 	}
1348 
1349 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1350 		int hlim;
1351 
1352 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1353 #ifdef INET
1354 			struct ip *ip;
1355 
1356 			ip = mtod(m, struct ip *);
1357 			hlim = ip->ip_ttl;
1358 #else
1359 			/* We won't hit this code */
1360 			hlim = 0;
1361 #endif
1362 		} else {
1363 			hlim = ip6->ip6_hlim & 0xff;
1364 		}
1365 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1366 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1367 		    IPPROTO_IPV6);
1368 		if (*mp)
1369 			mp = &(*mp)->m_next;
1370 	}
1371 
1372 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1373 		int tclass;
1374 
1375 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1376 #ifdef INET
1377 			struct ip *ip;
1378 
1379 			ip = mtod(m, struct ip *);
1380 			tclass = ip->ip_tos;
1381 #else
1382 			/* We won't hit this code */
1383 			tclass = 0;
1384 #endif
1385 		} else {
1386 			u_int32_t flowinfo;
1387 
1388 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1389 			flowinfo >>= 20;
1390 			tclass = flowinfo & 0xff;
1391 		}
1392 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1393 		    IPV6_TCLASS, IPPROTO_IPV6);
1394 		if (*mp)
1395 			mp = &(*mp)->m_next;
1396 	}
1397 
1398 	if (v4only != NULL) {
1399 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1400 			*v4only = 1;
1401 		} else {
1402 			*v4only = 0;
1403 		}
1404 	}
1405 
1406 	return (mp);
1407 }
1408 
1409 void
1410 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1411 {
1412 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1413 	int v4only = 0;
1414 
1415 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1416 	if (v4only)
1417 		return;
1418 
1419 	/*
1420 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1421 	 * privilege for the option (see ip6_ctloutput), but it might be too
1422 	 * strict, since there might be some hop-by-hop options which can be
1423 	 * returned to normal user.
1424 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1425 	 */
1426 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1427 		/*
1428 		 * Check if a hop-by-hop options header is contatined in the
1429 		 * received packet, and if so, store the options as ancillary
1430 		 * data. Note that a hop-by-hop options header must be
1431 		 * just after the IPv6 header, which is assured through the
1432 		 * IPv6 input processing.
1433 		 */
1434 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1435 			struct ip6_hbh *hbh;
1436 			int hbhlen = 0;
1437 #ifdef PULLDOWN_TEST
1438 			struct mbuf *ext;
1439 #endif
1440 
1441 #ifndef PULLDOWN_TEST
1442 			hbh = (struct ip6_hbh *)(ip6 + 1);
1443 			hbhlen = (hbh->ip6h_len + 1) << 3;
1444 #else
1445 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1446 			    ip6->ip6_nxt);
1447 			if (ext == NULL) {
1448 				IP6STAT_INC(ip6s_tooshort);
1449 				return;
1450 			}
1451 			hbh = mtod(ext, struct ip6_hbh *);
1452 			hbhlen = (hbh->ip6h_len + 1) << 3;
1453 			if (hbhlen != ext->m_len) {
1454 				m_freem(ext);
1455 				IP6STAT_INC(ip6s_tooshort);
1456 				return;
1457 			}
1458 #endif
1459 
1460 			/*
1461 			 * XXX: We copy the whole header even if a
1462 			 * jumbo payload option is included, the option which
1463 			 * is to be removed before returning according to
1464 			 * RFC2292.
1465 			 * Note: this constraint is removed in RFC3542
1466 			 */
1467 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1468 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1469 			    IPPROTO_IPV6);
1470 			if (*mp)
1471 				mp = &(*mp)->m_next;
1472 #ifdef PULLDOWN_TEST
1473 			m_freem(ext);
1474 #endif
1475 		}
1476 	}
1477 
1478 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1479 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1480 
1481 		/*
1482 		 * Search for destination options headers or routing
1483 		 * header(s) through the header chain, and stores each
1484 		 * header as ancillary data.
1485 		 * Note that the order of the headers remains in
1486 		 * the chain of ancillary data.
1487 		 */
1488 		while (1) {	/* is explicit loop prevention necessary? */
1489 			struct ip6_ext *ip6e = NULL;
1490 			int elen;
1491 #ifdef PULLDOWN_TEST
1492 			struct mbuf *ext = NULL;
1493 #endif
1494 
1495 			/*
1496 			 * if it is not an extension header, don't try to
1497 			 * pull it from the chain.
1498 			 */
1499 			switch (nxt) {
1500 			case IPPROTO_DSTOPTS:
1501 			case IPPROTO_ROUTING:
1502 			case IPPROTO_HOPOPTS:
1503 			case IPPROTO_AH: /* is it possible? */
1504 				break;
1505 			default:
1506 				goto loopend;
1507 			}
1508 
1509 #ifndef PULLDOWN_TEST
1510 			if (off + sizeof(*ip6e) > m->m_len)
1511 				goto loopend;
1512 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1513 			if (nxt == IPPROTO_AH)
1514 				elen = (ip6e->ip6e_len + 2) << 2;
1515 			else
1516 				elen = (ip6e->ip6e_len + 1) << 3;
1517 			if (off + elen > m->m_len)
1518 				goto loopend;
1519 #else
1520 			ext = ip6_pullexthdr(m, off, nxt);
1521 			if (ext == NULL) {
1522 				IP6STAT_INC(ip6s_tooshort);
1523 				return;
1524 			}
1525 			ip6e = mtod(ext, struct ip6_ext *);
1526 			if (nxt == IPPROTO_AH)
1527 				elen = (ip6e->ip6e_len + 2) << 2;
1528 			else
1529 				elen = (ip6e->ip6e_len + 1) << 3;
1530 			if (elen != ext->m_len) {
1531 				m_freem(ext);
1532 				IP6STAT_INC(ip6s_tooshort);
1533 				return;
1534 			}
1535 #endif
1536 
1537 			switch (nxt) {
1538 			case IPPROTO_DSTOPTS:
1539 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1540 					break;
1541 
1542 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1543 				    IS2292(in6p,
1544 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1545 				    IPPROTO_IPV6);
1546 				if (*mp)
1547 					mp = &(*mp)->m_next;
1548 				break;
1549 			case IPPROTO_ROUTING:
1550 				if (!(in6p->inp_flags & IN6P_RTHDR))
1551 					break;
1552 
1553 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1554 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1555 				    IPPROTO_IPV6);
1556 				if (*mp)
1557 					mp = &(*mp)->m_next;
1558 				break;
1559 			case IPPROTO_HOPOPTS:
1560 			case IPPROTO_AH: /* is it possible? */
1561 				break;
1562 
1563 			default:
1564 				/*
1565 				 * other cases have been filtered in the above.
1566 				 * none will visit this case.  here we supply
1567 				 * the code just in case (nxt overwritten or
1568 				 * other cases).
1569 				 */
1570 #ifdef PULLDOWN_TEST
1571 				m_freem(ext);
1572 #endif
1573 				goto loopend;
1574 
1575 			}
1576 
1577 			/* proceed with the next header. */
1578 			off += elen;
1579 			nxt = ip6e->ip6e_nxt;
1580 			ip6e = NULL;
1581 #ifdef PULLDOWN_TEST
1582 			m_freem(ext);
1583 			ext = NULL;
1584 #endif
1585 		}
1586 	  loopend:
1587 		;
1588 	}
1589 
1590 	if (in6p->inp_flags2 & INP_RECVFLOWID) {
1591 		uint32_t flowid, flow_type;
1592 
1593 		flowid = m->m_pkthdr.flowid;
1594 		flow_type = M_HASHTYPE_GET(m);
1595 
1596 		/*
1597 		 * XXX should handle the failure of one or the
1598 		 * other - don't populate both?
1599 		 */
1600 		*mp = sbcreatecontrol((caddr_t) &flowid,
1601 		    sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6);
1602 		if (*mp)
1603 			mp = &(*mp)->m_next;
1604 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1605 		    sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6);
1606 		if (*mp)
1607 			mp = &(*mp)->m_next;
1608 	}
1609 
1610 #ifdef	RSS
1611 	if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) {
1612 		uint32_t flowid, flow_type;
1613 		uint32_t rss_bucketid;
1614 
1615 		flowid = m->m_pkthdr.flowid;
1616 		flow_type = M_HASHTYPE_GET(m);
1617 
1618 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1619 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1620 			   sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6);
1621 			if (*mp)
1622 				mp = &(*mp)->m_next;
1623 		}
1624 	}
1625 #endif
1626 
1627 }
1628 #undef IS2292
1629 
1630 void
1631 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1632 {
1633 	struct socket *so;
1634 	struct mbuf *m_mtu;
1635 	struct ip6_mtuinfo mtuctl;
1636 
1637 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1638 	/*
1639 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1640 	 * application wanted to know the MTU value.
1641 	 * NOTE: we notify disconnected sockets, because some udp
1642 	 * applications keep sending sockets disconnected.
1643 	 * NOTE: our implementation doesn't notify connected sockets that has
1644 	 * foreign address that is different than given destination addresses
1645 	 * (this is permitted by RFC 3542).
1646 	 */
1647 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1648 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1649 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1650 		return;
1651 
1652 	mtuctl.ip6m_mtu = mtu;
1653 	mtuctl.ip6m_addr = *dst;
1654 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1655 		return;
1656 
1657 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1658 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1659 		return;
1660 
1661 	so =  inp->inp_socket;
1662 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1663 	    == 0) {
1664 		m_freem(m_mtu);
1665 		/* XXX: should count statistics */
1666 	} else
1667 		sorwakeup(so);
1668 }
1669 
1670 #ifdef PULLDOWN_TEST
1671 /*
1672  * pull single extension header from mbuf chain.  returns single mbuf that
1673  * contains the result, or NULL on error.
1674  */
1675 static struct mbuf *
1676 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1677 {
1678 	struct ip6_ext ip6e;
1679 	size_t elen;
1680 	struct mbuf *n;
1681 
1682 #ifdef DIAGNOSTIC
1683 	switch (nxt) {
1684 	case IPPROTO_DSTOPTS:
1685 	case IPPROTO_ROUTING:
1686 	case IPPROTO_HOPOPTS:
1687 	case IPPROTO_AH: /* is it possible? */
1688 		break;
1689 	default:
1690 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1691 	}
1692 #endif
1693 
1694 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1695 	if (nxt == IPPROTO_AH)
1696 		elen = (ip6e.ip6e_len + 2) << 2;
1697 	else
1698 		elen = (ip6e.ip6e_len + 1) << 3;
1699 
1700 	if (elen > MLEN)
1701 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1702 	else
1703 		n = m_get(M_NOWAIT, MT_DATA);
1704 	if (n == NULL)
1705 		return NULL;
1706 
1707 	m_copydata(m, off, elen, mtod(n, caddr_t));
1708 	n->m_len = elen;
1709 	return n;
1710 }
1711 #endif
1712 
1713 /*
1714  * Get pointer to the previous header followed by the header
1715  * currently processed.
1716  */
1717 int
1718 ip6_get_prevhdr(const struct mbuf *m, int off)
1719 {
1720 	struct ip6_ext ip6e;
1721 	struct ip6_hdr *ip6;
1722 	int len, nlen, nxt;
1723 
1724 	if (off == sizeof(struct ip6_hdr))
1725 		return (offsetof(struct ip6_hdr, ip6_nxt));
1726 	if (off < sizeof(struct ip6_hdr))
1727 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1728 
1729 	ip6 = mtod(m, struct ip6_hdr *);
1730 	nxt = ip6->ip6_nxt;
1731 	len = sizeof(struct ip6_hdr);
1732 	nlen = 0;
1733 	while (len < off) {
1734 		m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1735 		switch (nxt) {
1736 		case IPPROTO_FRAGMENT:
1737 			nlen = sizeof(struct ip6_frag);
1738 			break;
1739 		case IPPROTO_AH:
1740 			nlen = (ip6e.ip6e_len + 2) << 2;
1741 			break;
1742 		default:
1743 			nlen = (ip6e.ip6e_len + 1) << 3;
1744 		}
1745 		len += nlen;
1746 		nxt = ip6e.ip6e_nxt;
1747 	}
1748 	return (len - nlen);
1749 }
1750 
1751 /*
1752  * get next header offset.  m will be retained.
1753  */
1754 int
1755 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1756 {
1757 	struct ip6_hdr ip6;
1758 	struct ip6_ext ip6e;
1759 	struct ip6_frag fh;
1760 
1761 	/* just in case */
1762 	if (m == NULL)
1763 		panic("ip6_nexthdr: m == NULL");
1764 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1765 		return -1;
1766 
1767 	switch (proto) {
1768 	case IPPROTO_IPV6:
1769 		if (m->m_pkthdr.len < off + sizeof(ip6))
1770 			return -1;
1771 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1772 		if (nxtp)
1773 			*nxtp = ip6.ip6_nxt;
1774 		off += sizeof(ip6);
1775 		return off;
1776 
1777 	case IPPROTO_FRAGMENT:
1778 		/*
1779 		 * terminate parsing if it is not the first fragment,
1780 		 * it does not make sense to parse through it.
1781 		 */
1782 		if (m->m_pkthdr.len < off + sizeof(fh))
1783 			return -1;
1784 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1785 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1786 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1787 			return -1;
1788 		if (nxtp)
1789 			*nxtp = fh.ip6f_nxt;
1790 		off += sizeof(struct ip6_frag);
1791 		return off;
1792 
1793 	case IPPROTO_AH:
1794 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1795 			return -1;
1796 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1797 		if (nxtp)
1798 			*nxtp = ip6e.ip6e_nxt;
1799 		off += (ip6e.ip6e_len + 2) << 2;
1800 		return off;
1801 
1802 	case IPPROTO_HOPOPTS:
1803 	case IPPROTO_ROUTING:
1804 	case IPPROTO_DSTOPTS:
1805 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1806 			return -1;
1807 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1808 		if (nxtp)
1809 			*nxtp = ip6e.ip6e_nxt;
1810 		off += (ip6e.ip6e_len + 1) << 3;
1811 		return off;
1812 
1813 	case IPPROTO_NONE:
1814 	case IPPROTO_ESP:
1815 	case IPPROTO_IPCOMP:
1816 		/* give up */
1817 		return -1;
1818 
1819 	default:
1820 		return -1;
1821 	}
1822 
1823 	/* NOTREACHED */
1824 }
1825 
1826 /*
1827  * get offset for the last header in the chain.  m will be kept untainted.
1828  */
1829 int
1830 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1831 {
1832 	int newoff;
1833 	int nxt;
1834 
1835 	if (!nxtp) {
1836 		nxt = -1;
1837 		nxtp = &nxt;
1838 	}
1839 	while (1) {
1840 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1841 		if (newoff < 0)
1842 			return off;
1843 		else if (newoff < off)
1844 			return -1;	/* invalid */
1845 		else if (newoff == off)
1846 			return newoff;
1847 
1848 		off = newoff;
1849 		proto = *nxtp;
1850 	}
1851 }
1852 
1853 /*
1854  * System control for IP6
1855  */
1856 
1857 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1858 	0,		0,		0,		0,
1859 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1860 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1861 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1862 	0,		0,		EHOSTUNREACH,	0,
1863 	ENOPROTOOPT,	ECONNREFUSED
1864 };
1865