xref: /freebsd/sys/netinet6/ip6_input.c (revision 9c151e1fbf1a71a4912afa9693a39d55a00db4eb)
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 
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include "opt_ipsec.h"
66 #include "opt_route.h"
67 #include "opt_rss.h"
68 #include "opt_sctp.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/hhook.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/proc.h>
76 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/sdt.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/rmlock.h>
86 #include <sys/syslog.h>
87 #include <sys/sysctl.h>
88 #include <sys/eventhandler.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_types.h>
93 #include <net/if_private.h>
94 #include <net/if_dl.h>
95 #include <net/route.h>
96 #include <net/netisr.h>
97 #include <net/rss_config.h>
98 #include <net/pfil.h>
99 #include <net/vnet.h>
100 
101 #include <netinet/in.h>
102 #include <netinet/in_kdtrace.h>
103 #include <netinet/ip_var.h>
104 #include <netinet/in_systm.h>
105 #include <net/if_llatbl.h>
106 #ifdef INET
107 #include <netinet/ip.h>
108 #include <netinet/ip_icmp.h>
109 #endif /* INET */
110 #include <netinet/ip6.h>
111 #include <netinet6/in6_var.h>
112 #include <netinet6/ip6_var.h>
113 #include <netinet/ip_encap.h>
114 #include <netinet/in_pcb.h>
115 #include <netinet/icmp6.h>
116 #include <netinet6/scope6_var.h>
117 #include <netinet6/in6_ifattach.h>
118 #include <netinet6/mld6_var.h>
119 #include <netinet6/nd6.h>
120 #include <netinet6/in6_rss.h>
121 #ifdef SCTP
122 #include <netinet/sctp_pcb.h>
123 #include <netinet6/sctp6_var.h>
124 #endif
125 
126 #include <netipsec/ipsec_support.h>
127 
128 ip6proto_input_t	*ip6_protox[IPPROTO_MAX] = {
129 			    [0 ... IPPROTO_MAX - 1] = rip6_input };
130 ip6proto_ctlinput_t	*ip6_ctlprotox[IPPROTO_MAX] = {
131 			    [0 ... IPPROTO_MAX - 1] = rip6_ctlinput };
132 
133 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
134 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
135 VNET_DEFINE(u_long, in6_ifaddrhmask);
136 
137 static struct netisr_handler ip6_nh = {
138 	.nh_name = "ip6",
139 	.nh_handler = ip6_input,
140 	.nh_proto = NETISR_IPV6,
141 #ifdef RSS
142 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
143 	.nh_policy = NETISR_POLICY_CPU,
144 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
145 #else
146 	.nh_policy = NETISR_POLICY_FLOW,
147 #endif
148 };
149 
150 static int
151 sysctl_netinet6_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
152 {
153 	int error, qlimit;
154 
155 	netisr_getqlimit(&ip6_nh, &qlimit);
156 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
157 	if (error || !req->newptr)
158 		return (error);
159 	if (qlimit < 1)
160 		return (EINVAL);
161 	return (netisr_setqlimit(&ip6_nh, qlimit));
162 }
163 SYSCTL_DECL(_net_inet6_ip6);
164 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRQMAXLEN, intr_queue_maxlen,
165     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
166     0, 0, sysctl_netinet6_intr_queue_maxlen, "I",
167     "Maximum size of the IPv6 input queue");
168 
169 VNET_DEFINE_STATIC(bool, ip6_sav) = true;
170 #define	V_ip6_sav	VNET(ip6_sav)
171 SYSCTL_BOOL(_net_inet6_ip6, OID_AUTO, source_address_validation,
172     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_sav), true,
173     "Drop incoming packets with source address that is a local address");
174 
175 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, temp_max_desync_factor,
176     CTLFLAG_RD | CTLFLAG_VNET,
177     &VNET_NAME(ip6_temp_max_desync_factor), 0,
178     "RFC 8981 max desync factor");
179 
180 #ifdef RSS
181 static struct netisr_handler ip6_direct_nh = {
182 	.nh_name = "ip6_direct",
183 	.nh_handler = ip6_direct_input,
184 	.nh_proto = NETISR_IPV6_DIRECT,
185 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
186 	.nh_policy = NETISR_POLICY_CPU,
187 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
188 };
189 
190 static int
191 sysctl_netinet6_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
192 {
193 	int error, qlimit;
194 
195 	netisr_getqlimit(&ip6_direct_nh, &qlimit);
196 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
197 	if (error || !req->newptr)
198 		return (error);
199 	if (qlimit < 1)
200 		return (EINVAL);
201 	return (netisr_setqlimit(&ip6_direct_nh, qlimit));
202 }
203 SYSCTL_PROC(_net_inet6_ip6, IPV6CTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
204     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
205     0, 0, sysctl_netinet6_intr_direct_queue_maxlen, "I",
206     "Maximum size of the IPv6 direct input queue");
207 
208 #endif
209 
210 VNET_DEFINE(pfil_head_t, inet6_pfil_head);
211 VNET_DEFINE(pfil_head_t, inet6_local_pfil_head);
212 
213 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
214 VNET_PCPUSTAT_SYSINIT(ip6stat);
215 #ifdef VIMAGE
216 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
217 #endif /* VIMAGE */
218 
219 struct rmlock in6_ifaddr_lock;
220 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
221 
222 static int ip6_hopopts_input(u_int32_t *, struct mbuf **, int *);
223 
224 /*
225  * IP6 initialization: fill in IP6 protocol switch table.
226  * All protocols not implemented in kernel go to raw IP6 protocol handler.
227  */
228 static void
229 ip6_vnet_init(void *arg __unused)
230 {
231 	struct pfil_head_args args;
232 
233 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
234 	    &V_ip6_auto_linklocal);
235 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
236 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
237 	TUNABLE_BOOL_FETCH("net.inet6.ip6.use_stableaddr", &V_ip6_use_stableaddr);
238 
239 	CK_STAILQ_INIT(&V_in6_ifaddrhead);
240 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
241 	    &V_in6_ifaddrhmask);
242 
243 	/* Initialize packet filter hooks. */
244 	args.pa_version = PFIL_VERSION;
245 	args.pa_flags = PFIL_IN | PFIL_OUT;
246 	args.pa_type = PFIL_TYPE_IP6;
247 	args.pa_headname = PFIL_INET6_NAME;
248 	V_inet6_pfil_head = pfil_head_register(&args);
249 
250 	args.pa_flags = PFIL_OUT;
251 	args.pa_headname = PFIL_INET6_LOCAL_NAME;
252 	V_inet6_local_pfil_head = pfil_head_register(&args);
253 
254 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
255 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
256 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
257 		printf("%s: WARNING: unable to register input helper hook\n",
258 		    __func__);
259 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
260 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
261 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
262 		printf("%s: WARNING: unable to register output helper hook\n",
263 		    __func__);
264 
265 	scope6_init();
266 	addrsel_policy_init();
267 	nd6_init();
268 	frag6_init();
269 
270 	V_ip6_temp_max_desync_factor = TEMP_MAX_DESYNC_FACTOR_BASE +
271 	    (V_ip6_temp_preferred_lifetime >> 2) +
272 	    (V_ip6_temp_preferred_lifetime >> 3);
273 	V_ip6_desync_factor = arc4random() % V_ip6_temp_max_desync_factor;
274 
275 	/* Skip global initialization stuff for non-default instances. */
276 #ifdef VIMAGE
277 	netisr_register_vnet(&ip6_nh);
278 #ifdef RSS
279 	netisr_register_vnet(&ip6_direct_nh);
280 #endif
281 #endif
282 }
283 VNET_SYSINIT(ip6_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
284     ip6_vnet_init, NULL);
285 
286 static void
287 ip6_init(void *arg __unused)
288 {
289 	struct ifnet *ifp;
290 
291 	/*
292 	 * Register statically those protocols that are unlikely to ever go
293 	 * dynamic.
294 	 */
295 	IP6PROTO_REGISTER(IPPROTO_ICMPV6, icmp6_input, rip6_ctlinput);
296 	IP6PROTO_REGISTER(IPPROTO_DSTOPTS, dest6_input, NULL);
297 	IP6PROTO_REGISTER(IPPROTO_ROUTING, route6_input, NULL);
298 	IP6PROTO_REGISTER(IPPROTO_FRAGMENT, frag6_input, NULL);
299 	IP6PROTO_REGISTER(IPPROTO_IPV4, encap6_input, NULL);
300 	IP6PROTO_REGISTER(IPPROTO_IPV6, encap6_input, NULL);
301 	IP6PROTO_REGISTER(IPPROTO_ETHERIP, encap6_input, NULL);
302 	IP6PROTO_REGISTER(IPPROTO_GRE, encap6_input, NULL);
303 	IP6PROTO_REGISTER(IPPROTO_PIM, encap6_input, NULL);
304 #ifdef SCTP	/* XXX: has a loadable & static version */
305 	IP6PROTO_REGISTER(IPPROTO_SCTP, sctp6_input, sctp6_ctlinput);
306 #endif
307 
308 	EVENTHANDLER_REGISTER(vm_lowmem, frag6_drain, NULL, LOWMEM_PRI_DEFAULT);
309 	EVENTHANDLER_REGISTER(mbuf_lowmem, frag6_drain, NULL,
310 	    LOWMEM_PRI_DEFAULT);
311 
312 	netisr_register(&ip6_nh);
313 #ifdef RSS
314 	netisr_register(&ip6_direct_nh);
315 #endif
316 	/*
317          * XXXGL: we use SYSINIT() here, but go over V_ifnet.  See comment
318 	 * in sys/netinet/ip_input.c:ip_init().
319          */
320         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
321                 in6_ifarrival(NULL, ifp);
322 }
323 SYSINIT(ip6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init, NULL);
324 
325 int
326 ip6proto_register(uint8_t proto, ip6proto_input_t input,
327     ip6proto_ctlinput_t ctl)
328 {
329 
330 	MPASS(proto > 0);
331 
332 	if (ip6_protox[proto] == rip6_input) {
333 		ip6_protox[proto] = input;
334 		ip6_ctlprotox[proto] = ctl;
335 		return (0);
336 	} else
337 		return (EEXIST);
338 }
339 
340 int
341 ip6proto_unregister(uint8_t proto)
342 {
343 
344 	MPASS(proto > 0);
345 
346 	if (ip6_protox[proto] != rip6_input) {
347 		ip6_protox[proto] = rip6_input;
348 		ip6_ctlprotox[proto] = rip6_ctlinput;
349 		return (0);
350 	} else
351 		return (ENOENT);
352 }
353 
354 #ifdef VIMAGE
355 static void
356 ip6_destroy(void *unused __unused)
357 {
358 	struct ifaddr *ifa, *nifa;
359 	struct ifnet *ifp;
360 	int error;
361 
362 #ifdef RSS
363 	netisr_unregister_vnet(&ip6_direct_nh);
364 #endif
365 	netisr_unregister_vnet(&ip6_nh);
366 
367 	pfil_head_unregister(V_inet6_pfil_head);
368 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
369 	if (error != 0) {
370 		printf("%s: WARNING: unable to deregister input helper hook "
371 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
372 		    "error %d returned\n", __func__, error);
373 	}
374 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
375 	if (error != 0) {
376 		printf("%s: WARNING: unable to deregister output helper hook "
377 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
378 		    "error %d returned\n", __func__, error);
379 	}
380 
381 	/* Cleanup addresses. */
382 	IFNET_RLOCK();
383 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
384 		/* Cannot lock here - lock recursion. */
385 		/* IF_ADDR_LOCK(ifp); */
386 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
387 			if (ifa->ifa_addr->sa_family != AF_INET6)
388 				continue;
389 			in6_purgeaddr(ifa);
390 		}
391 		/* IF_ADDR_UNLOCK(ifp); */
392 		in6_ifdetach_destroy(ifp);
393 	}
394 	IFNET_RUNLOCK();
395 
396 	/* Make sure any routes are gone as well. */
397 	rib_flush_routes_family(AF_INET6);
398 
399 	frag6_destroy();
400 	nd6_destroy();
401 	in6_ifattach_destroy();
402 
403 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
404 }
405 
406 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
407 #endif
408 
409 static int
410 ip6_input_hbh(struct mbuf **mp, uint32_t *rtalert, int *off,
411     int *nxt, int *ours)
412 {
413 	struct mbuf *m;
414 	struct ip6_hdr *ip6;
415 	struct ip6_hbh *hbh;
416 
417 	if (ip6_hopopts_input(rtalert, mp, off)) {
418 #if 0	/*touches NULL pointer*/
419 		in6_ifstat_inc((*mp)->m_pkthdr.rcvif, ifs6_in_discard);
420 #endif
421 		goto out;	/* m have already been freed */
422 	}
423 
424 	/* adjust pointer */
425 	m = *mp;
426 	ip6 = mtod(m, struct ip6_hdr *);
427 
428 	/*
429 	 * If the payload length field is 0 and the next header field indicates
430 	 * Hop-by-Hop Options header, then a Jumbo Payload option MUST be
431 	 * included. We no not support Jumbo Payloads so report an error.
432 	 */
433 	if (ip6->ip6_plen == 0) {
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 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
443 	hbh = (struct ip6_hbh *)(ip6 + 1);
444 	*nxt = hbh->ip6h_nxt;
445 
446 	/*
447 	 * If we are acting as a router and the packet contains a
448 	 * router alert option, see if we know the option value.
449 	 * Currently, we only support the option value for MLD, in which
450 	 * case we should pass the packet to the multicast routing
451 	 * daemon.
452 	 */
453 	if (*rtalert != ~0) {
454 		switch (*rtalert) {
455 		case IP6OPT_RTALERT_MLD:
456 			if (V_ip6_forwarding)
457 				*ours = 1;
458 			break;
459 		default:
460 			/*
461 			 * RFC2711 requires unrecognized values must be
462 			 * silently ignored.
463 			 */
464 			break;
465 		}
466 	}
467 
468 	return (0);
469 
470 out:
471 	return (1);
472 }
473 
474 #ifdef RSS
475 /*
476  * IPv6 direct input routine.
477  *
478  * This is called when reinjecting completed fragments where
479  * all of the previous checking and book-keeping has been done.
480  */
481 void
482 ip6_direct_input(struct mbuf *m)
483 {
484 	int off, nxt;
485 	int nest;
486 	struct m_tag *mtag;
487 	struct ip6_direct_ctx *ip6dc;
488 
489 	mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
490 	KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
491 
492 	ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
493 	nxt = ip6dc->ip6dc_nxt;
494 	off = ip6dc->ip6dc_off;
495 
496 	nest = 0;
497 
498 	m_tag_delete(m, mtag);
499 
500 	while (nxt != IPPROTO_DONE) {
501 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
502 			IP6STAT_INC(ip6s_toomanyhdr);
503 			goto bad;
504 		}
505 
506 		/*
507 		 * protection against faulty packet - there should be
508 		 * more sanity checks in header chain processing.
509 		 */
510 		if (m->m_pkthdr.len < off) {
511 			IP6STAT_INC(ip6s_tooshort);
512 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
513 			goto bad;
514 		}
515 
516 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
517 		if (IPSEC_ENABLED(ipv6)) {
518 			if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
519 				return;
520 		}
521 #endif /* IPSEC */
522 
523 		nxt = ip6_protox[nxt](&m, &off, nxt);
524 	}
525 	return;
526 bad:
527 	m_freem(m);
528 }
529 #endif
530 
531 void
532 ip6_input(struct mbuf *m)
533 {
534 	struct in6_addr odst;
535 	struct ip6_hdr *ip6;
536 	struct in6_ifaddr *ia;
537 	struct ifnet *rcvif;
538 	u_int32_t plen;
539 	u_int32_t rtalert = ~0;
540 	int off = sizeof(struct ip6_hdr), nest;
541 	int nxt, ours = 0;
542 	int srcrt = 0;
543 
544 	/*
545 	 * Drop the packet if IPv6 operation is disabled on the interface.
546 	 */
547 	rcvif = m->m_pkthdr.rcvif;
548 	if ((rcvif->if_inet6->nd_flags & ND6_IFF_IFDISABLED))
549 		goto bad;
550 
551 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
552 	/*
553 	 * should the inner packet be considered authentic?
554 	 * see comment in ah4_input().
555 	 * NB: m cannot be NULL when passed to the input routine
556 	 */
557 
558 	m->m_flags &= ~M_AUTHIPHDR;
559 	m->m_flags &= ~M_AUTHIPDGM;
560 
561 #endif /* IPSEC */
562 
563 	if (m->m_flags & M_FASTFWD_OURS) {
564 		/*
565 		 * Firewall changed destination to local.
566 		 */
567 		ip6 = mtod(m, struct ip6_hdr *);
568 		goto passin;
569 	}
570 
571 	/*
572 	 * mbuf statistics
573 	 */
574 	if (m->m_flags & M_EXT) {
575 		if (m->m_next)
576 			IP6STAT_INC(ip6s_mext2m);
577 		else
578 			IP6STAT_INC(ip6s_mext1);
579 	} else {
580 		if (m->m_next) {
581 			struct ifnet *ifp = (m->m_flags & M_LOOP) ? V_loif : rcvif;
582 			int ifindex = ifp->if_index;
583 			if (ifindex >= IP6S_M2MMAX)
584 				ifindex = 0;
585 			IP6STAT_INC2(ip6s_m2m, ifindex);
586 		} else
587 			IP6STAT_INC(ip6s_m1);
588 	}
589 
590 	in6_ifstat_inc(rcvif, ifs6_in_receive);
591 	IP6STAT_INC(ip6s_total);
592 
593 	/*
594 	 * L2 bridge code and some other code can return mbuf chain
595 	 * that does not conform to KAME requirement.  too bad.
596 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
597 	 */
598 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
599 		struct mbuf *n;
600 
601 		if (m->m_pkthdr.len > MHLEN)
602 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
603 		else
604 			n = m_gethdr(M_NOWAIT, MT_DATA);
605 		if (n == NULL)
606 			goto bad;
607 
608 		m_move_pkthdr(n, m);
609 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
610 		n->m_len = n->m_pkthdr.len;
611 		m_freem(m);
612 		m = n;
613 	}
614 	if (m->m_len < sizeof(struct ip6_hdr)) {
615 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
616 			IP6STAT_INC(ip6s_toosmall);
617 			in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
618 			goto bad;
619 		}
620 	}
621 
622 	ip6 = mtod(m, struct ip6_hdr *);
623 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
624 		IP6STAT_INC(ip6s_badvers);
625 		in6_ifstat_inc(rcvif, ifs6_in_hdrerr);
626 		goto bad;
627 	}
628 
629 	IP6STAT_INC2(ip6s_nxthist, ip6->ip6_nxt);
630 	IP_PROBE(receive, NULL, NULL, ip6, rcvif, NULL, ip6);
631 
632 	/*
633 	 * Check against address spoofing/corruption.  The unspecified address
634 	 * is checked further below.
635 	 */
636 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
637 		/*
638 		 * XXX: "badscope" is not very suitable for a multicast source.
639 		 */
640 		IP6STAT_INC(ip6s_badscope);
641 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
642 		goto bad;
643 	}
644 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
645 	    !(m->m_flags & M_LOOP)) {
646 		/*
647 		 * In this case, the packet should come from the loopback
648 		 * interface.  However, we cannot just check the if_flags,
649 		 * because ip6_mloopback() passes the "actual" interface
650 		 * as the outgoing/incoming interface.
651 		 */
652 		IP6STAT_INC(ip6s_badscope);
653 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
654 		goto bad;
655 	}
656 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
657 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
658 		/*
659 		 * RFC4291 2.7:
660 		 * Nodes must not originate a packet to a multicast address
661 		 * whose scop field contains the reserved value 0; if such
662 		 * a packet is received, it must be silently dropped.
663 		 */
664 		IP6STAT_INC(ip6s_badscope);
665 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
666 		goto bad;
667 	}
668 	/*
669 	 * The following check is not documented in specs.  A malicious
670 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
671 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
672 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
673 	 *
674 	 * We have supported IPv6-only kernels for a few years and this issue
675 	 * has not come up.  The world seems to move mostly towards not using
676 	 * v4mapped on the wire, so it makes sense for us to keep rejecting
677 	 * any such packets.
678 	 */
679 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
680 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
681 		IP6STAT_INC(ip6s_badscope);
682 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
683 		goto bad;
684 	}
685 #if 0
686 	/*
687 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
688 	 *
689 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
690 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
691 	 * is revised to forbid relaying case.
692 	 */
693 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
694 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
695 		IP6STAT_INC(ip6s_badscope);
696 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
697 		goto bad;
698 	}
699 #endif
700 	/*
701 	 * Try to forward the packet, but if we fail continue.
702 	 * ip6_tryforward() does not generate redirects, so fall
703 	 * through to normal processing if redirects are required.
704 	 * ip6_tryforward() does inbound and outbound packet firewall
705 	 * processing. If firewall has decided that destination becomes
706 	 * our local address, it sets M_FASTFWD_OURS flag. In this
707 	 * case skip another inbound firewall processing and update
708 	 * ip6 pointer.
709 	 */
710 	if (V_ip6_forwarding != 0 && V_ip6_sendredirects == 0
711 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
712 	    && (!IPSEC_ENABLED(ipv6) ||
713 	    IPSEC_CAPS(ipv6, m, IPSEC_CAP_OPERABLE) == 0)
714 #endif
715 	    ) {
716 		if ((m = ip6_tryforward(m)) == NULL)
717 			return;
718 		if (m->m_flags & M_FASTFWD_OURS) {
719 			ip6 = mtod(m, struct ip6_hdr *);
720 			goto passin;
721 		}
722 	}
723 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
724 	/*
725 	 * Bypass packet filtering for packets previously handled by IPsec.
726 	 */
727 	if (IPSEC_ENABLED(ipv6) &&
728 	    IPSEC_CAPS(ipv6, m, IPSEC_CAP_BYPASS_FILTER) != 0)
729 			goto passin;
730 #endif
731 	/*
732 	 * Run through list of hooks for input packets.
733 	 *
734 	 * NB: Beware of the destination address changing
735 	 *     (e.g. by NAT rewriting).  When this happens,
736 	 *     tell ip6_forward to do the right thing.
737 	 */
738 
739 	/* Jump over all PFIL processing if hooks are not active. */
740 	if (!PFIL_HOOKED_IN(V_inet6_pfil_head))
741 		goto passin;
742 
743 	odst = ip6->ip6_dst;
744 	if (pfil_mbuf_in(V_inet6_pfil_head, &m, m->m_pkthdr.rcvif,
745 	    NULL) != PFIL_PASS)
746 		return;
747 	ip6 = mtod(m, struct ip6_hdr *);
748 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
749 	if ((m->m_flags & (M_IP6_NEXTHOP | M_FASTFWD_OURS)) == M_IP6_NEXTHOP &&
750 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
751 		/*
752 		 * Directly ship the packet on.  This allows forwarding
753 		 * packets originally destined to us to some other directly
754 		 * connected host.
755 		 */
756 		ip6_forward(m, 1);
757 		return;
758 	}
759 
760 passin:
761 	/*
762 	 * The check is deferred to here to give firewalls a chance to block
763 	 * (and log) such packets.  ip6_tryforward() will not process such
764 	 * packets.
765 	 */
766 	if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst))) {
767 		IP6STAT_INC(ip6s_badscope);
768 		in6_ifstat_inc(rcvif, ifs6_in_addrerr);
769 		goto bad;
770 	}
771 
772 	plen = (uint32_t)ntohs(ip6->ip6_plen);
773 
774 	/*
775 	 * We don't support Jumbograms, reject packets with plen == 0 as early
776 	 * as we can.
777 	 */
778 	if (plen == 0)
779 		goto bad;
780 
781 	/*
782 	 * Disambiguate address scope zones (if there is ambiguity).
783 	 * We first make sure that the original source or destination address
784 	 * is not in our internal form for scoped addresses.  Such addresses
785 	 * are not necessarily invalid spec-wise, but we cannot accept them due
786 	 * to the usage conflict.
787 	 * in6_setscope() then also checks and rejects the cases where src or
788 	 * dst are the loopback address and the receiving interface
789 	 * is not loopback.
790 	 */
791 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
792 		IP6STAT_INC(ip6s_badscope); /* XXX */
793 		goto bad;
794 	}
795 	if (in6_setscope(&ip6->ip6_src, rcvif, NULL) ||
796 	    in6_setscope(&ip6->ip6_dst, rcvif, NULL)) {
797 		IP6STAT_INC(ip6s_badscope);
798 		goto bad;
799 	}
800 	if (m->m_flags & M_FASTFWD_OURS) {
801 		m->m_flags &= ~M_FASTFWD_OURS;
802 		ours = 1;
803 		goto hbhcheck;
804 	}
805 	/*
806 	 * Multicast check. Assume packet is for us to avoid
807 	 * prematurely taking locks.
808 	 */
809 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
810 		ours = 1;
811 		in6_ifstat_inc(rcvif, ifs6_in_mcast);
812 		goto hbhcheck;
813 	}
814 	/*
815 	 * Unicast check
816 	 * XXX: For now we keep link-local IPv6 addresses with embedded
817 	 *      scope zone id, therefore we use zero zoneid here.
818 	 */
819 	ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
820 	if (ia != NULL) {
821 		if (ia->ia6_flags & IN6_IFF_NOTREADY) {
822 			char ip6bufs[INET6_ADDRSTRLEN];
823 			char ip6bufd[INET6_ADDRSTRLEN];
824 			/* address is not ready, so discard the packet. */
825 			nd6log((LOG_INFO,
826 			    "ip6_input: packet to an unready address %s->%s\n",
827 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
828 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
829 			goto bad;
830 		}
831 		if (V_ip6_sav && !(m->m_flags & M_LOOP) &&
832 		    __predict_false(in6_localip_fib(&ip6->ip6_src,
833 			    rcvif->if_fib))) {
834 			IP6STAT_INC(ip6s_badscope); /* XXX */
835 			goto bad;
836 		}
837 		/* Count the packet in the ip address stats */
838 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
839 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
840 		ours = 1;
841 		goto hbhcheck;
842 	}
843 
844 	/*
845 	 * Now there is no reason to process the packet if it's not our own
846 	 * and we're not a router.
847 	 */
848 	if (!V_ip6_forwarding) {
849 		IP6STAT_INC(ip6s_cantforward);
850 		goto bad;
851 	}
852 
853   hbhcheck:
854 	/*
855 	 * Process Hop-by-Hop options header if it's contained.
856 	 * m may be modified in ip6_hopopts_input().
857 	 */
858 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
859 		if (ip6_input_hbh(&m, &rtalert, &off, &nxt, &ours) != 0)
860 			return;
861 	} else
862 		nxt = ip6->ip6_nxt;
863 
864 	/*
865 	 * Use mbuf flags to propagate Router Alert option to
866 	 * ICMPv6 layer, as hop-by-hop options have been stripped.
867 	 */
868 	if (rtalert != ~0)
869 		m->m_flags |= M_RTALERT_MLD;
870 
871 	/*
872 	 * Check that the amount of data in the buffers
873 	 * is as at least much as the IPv6 header would have us expect.
874 	 * Trim mbufs if longer than we expect.
875 	 * Drop packet if shorter than we expect.
876 	 */
877 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
878 		IP6STAT_INC(ip6s_tooshort);
879 		in6_ifstat_inc(rcvif, ifs6_in_truncated);
880 		goto bad;
881 	}
882 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
883 		if (m->m_len == m->m_pkthdr.len) {
884 			m->m_len = sizeof(struct ip6_hdr) + plen;
885 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
886 		} else
887 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
888 	}
889 
890 	/*
891 	 * Forward if desirable.
892 	 */
893 	if (V_ip6_mrouter &&
894 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
895 		/*
896 		 * If we are acting as a multicast router, all
897 		 * incoming multicast packets are passed to the
898 		 * kernel-level multicast forwarding function.
899 		 * The packet is returned (relatively) intact; if
900 		 * ip6_mforward() returns a non-zero value, the packet
901 		 * must be discarded, else it may be accepted below.
902 		 *
903 		 * XXX TODO: Check hlim and multicast scope here to avoid
904 		 * unnecessarily calling into ip6_mforward().
905 		 */
906 		if (ip6_mforward && ip6_mforward(ip6, rcvif, m)) {
907 			IP6STAT_INC(ip6s_cantforward);
908 			goto bad;
909 		}
910 	} else if (!ours) {
911 		ip6_forward(m, srcrt);
912 		return;
913 	}
914 
915 	/*
916 	 * We are going to ship the packet to the local protocol stack. Call the
917 	 * filter again for this 'output' action, allowing redirect-like rules
918 	 * to adjust the source address.
919 	 */
920 	if (PFIL_HOOKED_OUT(V_inet6_local_pfil_head)) {
921 		if (pfil_mbuf_out(V_inet6_local_pfil_head, &m, V_loif, NULL) !=
922 		    PFIL_PASS)
923 			return;
924 		ip6 = mtod(m, struct ip6_hdr *);
925 	}
926 
927 	/*
928 	 * Tell launch routine the next header
929 	 */
930 	IP6STAT_INC(ip6s_delivered);
931 	in6_ifstat_inc(rcvif, ifs6_in_deliver);
932 	nest = 0;
933 
934 	while (nxt != IPPROTO_DONE) {
935 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
936 			IP6STAT_INC(ip6s_toomanyhdr);
937 			goto bad;
938 		}
939 
940 		/*
941 		 * protection against faulty packet - there should be
942 		 * more sanity checks in header chain processing.
943 		 */
944 		if (m->m_pkthdr.len < off) {
945 			IP6STAT_INC(ip6s_tooshort);
946 			in6_ifstat_inc(rcvif, ifs6_in_truncated);
947 			goto bad;
948 		}
949 
950 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
951 		if (IPSEC_ENABLED(ipv6)) {
952 			if (IPSEC_INPUT(ipv6, m, off, nxt) != 0)
953 				return;
954 		}
955 #endif /* IPSEC */
956 
957 		nxt = ip6_protox[nxt](&m, &off, nxt);
958 	}
959 	return;
960 bad:
961 	in6_ifstat_inc(rcvif, ifs6_in_discard);
962 	if (m != NULL)
963 		m_freem(m);
964 }
965 
966 /*
967  * Hop-by-Hop options header processing. If a valid jumbo payload option is
968  * included report an error.
969  *
970  * rtalertp - XXX: should be stored more smart way
971  */
972 static int
973 ip6_hopopts_input(u_int32_t *rtalertp, struct mbuf **mp, int *offp)
974 {
975 	struct mbuf *m = *mp;
976 	int off = *offp, hbhlen;
977 	struct ip6_hbh *hbh;
978 
979 	/* validation of the length of the header */
980 	if (m->m_len < off + sizeof(*hbh)) {
981 		m = m_pullup(m, off + sizeof(*hbh));
982 		if (m == NULL) {
983 			IP6STAT_INC(ip6s_exthdrtoolong);
984 			*mp = NULL;
985 			return (-1);
986 		}
987 	}
988 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
989 	hbhlen = (hbh->ip6h_len + 1) << 3;
990 
991 	if (m->m_len < off + hbhlen) {
992 		m = m_pullup(m, off + hbhlen);
993 		if (m == NULL) {
994 			IP6STAT_INC(ip6s_exthdrtoolong);
995 			*mp = NULL;
996 			return (-1);
997 		}
998 	}
999 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1000 	off += hbhlen;
1001 	hbhlen -= sizeof(struct ip6_hbh);
1002 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1003 				hbhlen, rtalertp) < 0) {
1004 		*mp = NULL;
1005 		return (-1);
1006 	}
1007 
1008 	*offp = off;
1009 	*mp = m;
1010 	return (0);
1011 }
1012 
1013 /*
1014  * Search header for all Hop-by-hop options and process each option.
1015  * This function is separate from ip6_hopopts_input() in order to
1016  * handle a case where the sending node itself process its hop-by-hop
1017  * options header. In such a case, the function is called from ip6_output().
1018  *
1019  * The function assumes that hbh header is located right after the IPv6 header
1020  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1021  * opthead + hbhlen is located in contiguous memory region.
1022  */
1023 int
1024 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1025     u_int32_t *rtalertp)
1026 {
1027 	int optlen = 0;
1028 	u_int8_t *opt = opthead;
1029 	u_int16_t rtalert_val;
1030 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1031 
1032 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1033 		switch (*opt) {
1034 		case IP6OPT_PAD1:
1035 			optlen = 1;
1036 			break;
1037 		case IP6OPT_PADN:
1038 			if (hbhlen < IP6OPT_MINLEN) {
1039 				IP6STAT_INC(ip6s_toosmall);
1040 				goto bad;
1041 			}
1042 			optlen = *(opt + 1) + 2;
1043 			break;
1044 		case IP6OPT_ROUTER_ALERT:
1045 			/* XXX may need check for alignment */
1046 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1047 				IP6STAT_INC(ip6s_toosmall);
1048 				goto bad;
1049 			}
1050 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1051 				/* XXX stat */
1052 				icmp6_error(m, ICMP6_PARAM_PROB,
1053 				    ICMP6_PARAMPROB_HEADER,
1054 				    erroff + opt + 1 - opthead);
1055 				return (-1);
1056 			}
1057 			optlen = IP6OPT_RTALERT_LEN;
1058 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1059 			*rtalertp = ntohs(rtalert_val);
1060 			break;
1061 		case IP6OPT_JUMBO:
1062 			/* We do not support the Jumbo Payload option. */
1063 			goto bad;
1064 		default:		/* unknown option */
1065 			if (hbhlen < IP6OPT_MINLEN) {
1066 				IP6STAT_INC(ip6s_toosmall);
1067 				goto bad;
1068 			}
1069 			optlen = ip6_unknown_opt(opt, m,
1070 			    erroff + opt - opthead);
1071 			if (optlen == -1)
1072 				return (-1);
1073 			optlen += 2;
1074 			break;
1075 		}
1076 	}
1077 
1078 	return (0);
1079 
1080   bad:
1081 	m_freem(m);
1082 	return (-1);
1083 }
1084 
1085 /*
1086  * Unknown option processing.
1087  * The third argument `off' is the offset from the IPv6 header to the option,
1088  * which is necessary if the IPv6 header the and option header and IPv6 header
1089  * is not contiguous in order to return an ICMPv6 error.
1090  */
1091 int
1092 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1093 {
1094 	struct ip6_hdr *ip6;
1095 
1096 	switch (IP6OPT_TYPE(*optp)) {
1097 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1098 		return ((int)*(optp + 1));
1099 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1100 		m_freem(m);
1101 		return (-1);
1102 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1103 		IP6STAT_INC(ip6s_badoptions);
1104 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1105 		return (-1);
1106 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1107 		IP6STAT_INC(ip6s_badoptions);
1108 		ip6 = mtod(m, struct ip6_hdr *);
1109 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1110 		    (m->m_flags & (M_BCAST|M_MCAST)))
1111 			m_freem(m);
1112 		else
1113 			icmp6_error(m, ICMP6_PARAM_PROB,
1114 				    ICMP6_PARAMPROB_OPTION, off);
1115 		return (-1);
1116 	}
1117 
1118 	m_freem(m);		/* XXX: NOTREACHED */
1119 	return (-1);
1120 }
1121 
1122 /*
1123  * Create the "control" list for this pcb.
1124  * These functions will not modify mbuf chain at all.
1125  *
1126  * The routine will be called from upper layer handlers like tcp6_input().
1127  * Thus the routine assumes that the caller (tcp6_input) have already
1128  * called m_pullup() and all the extension headers are located in the
1129  * very first mbuf on the mbuf chain.
1130  *
1131  * ip6_savecontrol_v4 will handle those options that are possible to be
1132  * set on a v4-mapped socket.
1133  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1134  * options and handle the v6-only ones itself.
1135  */
1136 struct mbuf **
1137 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1138     int *v4only)
1139 {
1140 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1141 
1142 #if defined(SO_TIMESTAMP) && defined(SO_BINTIME)
1143 	if ((inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) != 0) {
1144 		union {
1145 			struct timeval tv;
1146 			struct bintime bt;
1147 			struct timespec ts;
1148 		} t;
1149 		struct bintime boottimebin, bt1;
1150 		struct timespec ts1;
1151 		int ts_clock;
1152 		bool stamped;
1153 
1154 		ts_clock = inp->inp_socket->so_ts_clock;
1155 		stamped = false;
1156 
1157 		/*
1158 		 * Handle BINTIME first. We create the same output options
1159 		 * for both SO_BINTIME and the case where SO_TIMESTAMP is
1160 		 * set with the timestamp clock set to SO_TS_BINTIME.
1161 		 */
1162 		if ((inp->inp_socket->so_options & SO_BINTIME) != 0 ||
1163 		    ts_clock == SO_TS_BINTIME) {
1164 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1165 			    M_TSTMP)) {
1166 				mbuf_tstmp2timespec(m, &ts1);
1167 				timespec2bintime(&ts1, &t.bt);
1168 				getboottimebin(&boottimebin);
1169 				bintime_add(&t.bt, &boottimebin);
1170 			} else {
1171 				bintime(&t.bt);
1172 			}
1173 			*mp = sbcreatecontrol(&t.bt, sizeof(t.bt), SCM_BINTIME,
1174 			    SOL_SOCKET, M_NOWAIT);
1175 			if (*mp != NULL) {
1176 				mp = &(*mp)->m_next;
1177 				stamped = true;
1178 			}
1179 
1180 			/*
1181 			 * Suppress other timestamps if SO_TIMESTAMP is not
1182 			 * set.
1183 			 */
1184 			if ((inp->inp_socket->so_options & SO_TIMESTAMP) == 0)
1185 				ts_clock = SO_TS_BINTIME;
1186 		}
1187 
1188 		switch (ts_clock) {
1189 		case SO_TS_REALTIME_MICRO:
1190 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1191 			    M_TSTMP)) {
1192 				mbuf_tstmp2timespec(m, &ts1);
1193 				timespec2bintime(&ts1, &bt1);
1194 				getboottimebin(&boottimebin);
1195 				bintime_add(&bt1, &boottimebin);
1196 				bintime2timeval(&bt1, &t.tv);
1197 			} else {
1198 				microtime(&t.tv);
1199 			}
1200 			*mp = sbcreatecontrol(&t.tv, sizeof(t.tv),
1201 			    SCM_TIMESTAMP, SOL_SOCKET, M_NOWAIT);
1202 			if (*mp != NULL) {
1203 				mp = &(*mp)->m_next;
1204 				stamped = true;
1205 			}
1206 			break;
1207 
1208 		case SO_TS_BINTIME:
1209 			break;
1210 
1211 		case SO_TS_REALTIME:
1212 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1213 			    M_TSTMP)) {
1214 				mbuf_tstmp2timespec(m, &t.ts);
1215 				getboottimebin(&boottimebin);
1216 				bintime2timespec(&boottimebin, &ts1);
1217 				timespecadd(&t.ts, &ts1, &t.ts);
1218 			} else {
1219 				nanotime(&t.ts);
1220 			}
1221 			*mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1222 			    SCM_REALTIME, SOL_SOCKET, M_NOWAIT);
1223 			if (*mp != NULL) {
1224 				mp = &(*mp)->m_next;
1225 				stamped = true;
1226 			}
1227 			break;
1228 
1229 		case SO_TS_MONOTONIC:
1230 			if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1231 			    M_TSTMP))
1232 				mbuf_tstmp2timespec(m, &t.ts);
1233 			else
1234 				nanouptime(&t.ts);
1235 			*mp = sbcreatecontrol(&t.ts, sizeof(t.ts),
1236 			    SCM_MONOTONIC, SOL_SOCKET, M_NOWAIT);
1237 			if (*mp != NULL) {
1238 				mp = &(*mp)->m_next;
1239 				stamped = true;
1240 			}
1241 			break;
1242 
1243 		default:
1244 			panic("unknown (corrupted) so_ts_clock");
1245 		}
1246 		if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) ==
1247 		    (M_PKTHDR | M_TSTMP)) {
1248 			struct sock_timestamp_info sti;
1249 
1250 			bzero(&sti, sizeof(sti));
1251 			sti.st_info_flags = ST_INFO_HW;
1252 			if ((m->m_flags & M_TSTMP_HPREC) != 0)
1253 				sti.st_info_flags |= ST_INFO_HW_HPREC;
1254 			*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1255 			    SOL_SOCKET, M_NOWAIT);
1256 			if (*mp != NULL)
1257 				mp = &(*mp)->m_next;
1258 		}
1259 	}
1260 #endif
1261 
1262 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1263 	/* RFC 2292 sec. 5 */
1264 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1265 		struct in6_pktinfo pi6;
1266 
1267 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1268 #ifdef INET
1269 			struct ip *ip;
1270 
1271 			ip = mtod(m, struct ip *);
1272 			pi6.ipi6_addr.s6_addr32[0] = 0;
1273 			pi6.ipi6_addr.s6_addr32[1] = 0;
1274 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1275 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1276 #else
1277 			/* We won't hit this code */
1278 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1279 #endif
1280 		} else {
1281 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1282 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1283 		}
1284 		pi6.ipi6_ifindex =
1285 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1286 
1287 		*mp = sbcreatecontrol(&pi6, sizeof(struct in6_pktinfo),
1288 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6,
1289 		    M_NOWAIT);
1290 		if (*mp)
1291 			mp = &(*mp)->m_next;
1292 	}
1293 
1294 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1295 		int hlim;
1296 
1297 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1298 #ifdef INET
1299 			struct ip *ip;
1300 
1301 			ip = mtod(m, struct ip *);
1302 			hlim = ip->ip_ttl;
1303 #else
1304 			/* We won't hit this code */
1305 			hlim = 0;
1306 #endif
1307 		} else {
1308 			hlim = ip6->ip6_hlim & 0xff;
1309 		}
1310 		*mp = sbcreatecontrol(&hlim, sizeof(int),
1311 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1312 		    IPPROTO_IPV6, M_NOWAIT);
1313 		if (*mp)
1314 			mp = &(*mp)->m_next;
1315 	}
1316 
1317 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1318 		int tclass;
1319 
1320 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1321 #ifdef INET
1322 			struct ip *ip;
1323 
1324 			ip = mtod(m, struct ip *);
1325 			tclass = ip->ip_tos;
1326 #else
1327 			/* We won't hit this code */
1328 			tclass = 0;
1329 #endif
1330 		} else {
1331 			u_int32_t flowinfo;
1332 
1333 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1334 			flowinfo >>= 20;
1335 			tclass = flowinfo & 0xff;
1336 		}
1337 		*mp = sbcreatecontrol(&tclass, sizeof(int), IPV6_TCLASS,
1338 		    IPPROTO_IPV6, M_NOWAIT);
1339 		if (*mp)
1340 			mp = &(*mp)->m_next;
1341 	}
1342 
1343 	if (v4only != NULL) {
1344 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1345 			*v4only = 1;
1346 		} else {
1347 			*v4only = 0;
1348 		}
1349 	}
1350 
1351 	return (mp);
1352 }
1353 
1354 void
1355 ip6_savecontrol(struct inpcb *inp, struct mbuf *m, struct mbuf **mp)
1356 {
1357 	struct ip6_hdr *ip6;
1358 	int v4only = 0;
1359 
1360 	mp = ip6_savecontrol_v4(inp, m, mp, &v4only);
1361 	if (v4only)
1362 		return;
1363 
1364 	ip6 = mtod(m, struct ip6_hdr *);
1365 	/*
1366 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1367 	 * privilege for the option (see ip6_ctloutput), but it might be too
1368 	 * strict, since there might be some hop-by-hop options which can be
1369 	 * returned to normal user.
1370 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1371 	 */
1372 	if ((inp->inp_flags & IN6P_HOPOPTS) != 0) {
1373 		/*
1374 		 * Check if a hop-by-hop options header is contatined in the
1375 		 * received packet, and if so, store the options as ancillary
1376 		 * data. Note that a hop-by-hop options header must be
1377 		 * just after the IPv6 header, which is assured through the
1378 		 * IPv6 input processing.
1379 		 */
1380 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1381 			struct ip6_hbh *hbh;
1382 			u_int hbhlen;
1383 
1384 			hbh = (struct ip6_hbh *)(ip6 + 1);
1385 			hbhlen = (hbh->ip6h_len + 1) << 3;
1386 
1387 			/*
1388 			 * XXX: We copy the whole header even if a
1389 			 * jumbo payload option is included, the option which
1390 			 * is to be removed before returning according to
1391 			 * RFC2292.
1392 			 * Note: this constraint is removed in RFC3542
1393 			 */
1394 			*mp = sbcreatecontrol(hbh, hbhlen,
1395 			    IS2292(inp, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1396 			    IPPROTO_IPV6, M_NOWAIT);
1397 			if (*mp)
1398 				mp = &(*mp)->m_next;
1399 		}
1400 	}
1401 
1402 	if ((inp->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1403 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1404 
1405 		/*
1406 		 * Search for destination options headers or routing
1407 		 * header(s) through the header chain, and stores each
1408 		 * header as ancillary data.
1409 		 * Note that the order of the headers remains in
1410 		 * the chain of ancillary data.
1411 		 */
1412 		while (1) {	/* is explicit loop prevention necessary? */
1413 			struct ip6_ext *ip6e = NULL;
1414 			u_int elen;
1415 
1416 			/*
1417 			 * if it is not an extension header, don't try to
1418 			 * pull it from the chain.
1419 			 */
1420 			switch (nxt) {
1421 			case IPPROTO_DSTOPTS:
1422 			case IPPROTO_ROUTING:
1423 			case IPPROTO_HOPOPTS:
1424 			case IPPROTO_AH: /* is it possible? */
1425 				break;
1426 			default:
1427 				goto loopend;
1428 			}
1429 
1430 			if (off + sizeof(*ip6e) > m->m_len)
1431 				goto loopend;
1432 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1433 			if (nxt == IPPROTO_AH)
1434 				elen = (ip6e->ip6e_len + 2) << 2;
1435 			else
1436 				elen = (ip6e->ip6e_len + 1) << 3;
1437 			if (off + elen > m->m_len)
1438 				goto loopend;
1439 
1440 			switch (nxt) {
1441 			case IPPROTO_DSTOPTS:
1442 				if (!(inp->inp_flags & IN6P_DSTOPTS))
1443 					break;
1444 
1445 				*mp = sbcreatecontrol(ip6e, elen,
1446 				    IS2292(inp, IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1447 				    IPPROTO_IPV6, M_NOWAIT);
1448 				if (*mp)
1449 					mp = &(*mp)->m_next;
1450 				break;
1451 			case IPPROTO_ROUTING:
1452 				if (!(inp->inp_flags & IN6P_RTHDR))
1453 					break;
1454 
1455 				*mp = sbcreatecontrol(ip6e, elen,
1456 				    IS2292(inp, IPV6_2292RTHDR, IPV6_RTHDR),
1457 				    IPPROTO_IPV6, M_NOWAIT);
1458 				if (*mp)
1459 					mp = &(*mp)->m_next;
1460 				break;
1461 			case IPPROTO_HOPOPTS:
1462 			case IPPROTO_AH: /* is it possible? */
1463 				break;
1464 
1465 			default:
1466 				/*
1467 				 * other cases have been filtered in the above.
1468 				 * none will visit this case.  here we supply
1469 				 * the code just in case (nxt overwritten or
1470 				 * other cases).
1471 				 */
1472 				goto loopend;
1473 			}
1474 
1475 			/* proceed with the next header. */
1476 			off += elen;
1477 			nxt = ip6e->ip6e_nxt;
1478 			ip6e = NULL;
1479 		}
1480 	  loopend:
1481 		;
1482 	}
1483 
1484 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1485 		uint32_t flowid, flow_type;
1486 
1487 		flowid = m->m_pkthdr.flowid;
1488 		flow_type = M_HASHTYPE_GET(m);
1489 
1490 		/*
1491 		 * XXX should handle the failure of one or the
1492 		 * other - don't populate both?
1493 		 */
1494 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IPV6_FLOWID,
1495 		    IPPROTO_IPV6, M_NOWAIT);
1496 		if (*mp)
1497 			mp = &(*mp)->m_next;
1498 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1499 		    IPV6_FLOWTYPE, IPPROTO_IPV6, M_NOWAIT);
1500 		if (*mp)
1501 			mp = &(*mp)->m_next;
1502 	}
1503 
1504 #ifdef	RSS
1505 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1506 		uint32_t flowid, flow_type;
1507 		uint32_t rss_bucketid;
1508 
1509 		flowid = m->m_pkthdr.flowid;
1510 		flow_type = M_HASHTYPE_GET(m);
1511 
1512 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1513 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1514 			    IPV6_RSSBUCKETID, IPPROTO_IPV6, M_NOWAIT);
1515 			if (*mp)
1516 				mp = &(*mp)->m_next;
1517 		}
1518 	}
1519 #endif
1520 
1521 }
1522 #undef IS2292
1523 
1524 void
1525 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1526 {
1527 	struct socket *so;
1528 	struct mbuf *m_mtu;
1529 	struct ip6_mtuinfo mtuctl;
1530 
1531 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1532 	/*
1533 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1534 	 * application wanted to know the MTU value.
1535 	 * NOTE: we notify disconnected sockets, because some udp
1536 	 * applications keep sending sockets disconnected.
1537 	 * NOTE: our implementation doesn't notify connected sockets that has
1538 	 * foreign address that is different than given destination addresses
1539 	 * (this is permitted by RFC 3542).
1540 	 */
1541 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1542 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1543 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1544 		return;
1545 
1546 	mtuctl.ip6m_mtu = mtu;
1547 	mtuctl.ip6m_addr = *dst;
1548 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1549 		return;
1550 
1551 	if ((m_mtu = sbcreatecontrol(&mtuctl, sizeof(mtuctl), IPV6_PATHMTU,
1552 	    IPPROTO_IPV6, M_NOWAIT)) == NULL)
1553 		return;
1554 
1555 	so =  inp->inp_socket;
1556 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1557 	    == 0) {
1558 		soroverflow(so);
1559 		m_freem(m_mtu);
1560 		/* XXX: should count statistics */
1561 	} else
1562 		sorwakeup(so);
1563 }
1564 
1565 /*
1566  * Get pointer to the previous header followed by the header
1567  * currently processed.
1568  */
1569 int
1570 ip6_get_prevhdr(const struct mbuf *m, int off)
1571 {
1572 	struct ip6_ext ip6e;
1573 	struct ip6_hdr *ip6;
1574 	int len, nlen, nxt;
1575 
1576 	if (off == sizeof(struct ip6_hdr))
1577 		return (offsetof(struct ip6_hdr, ip6_nxt));
1578 	if (off < sizeof(struct ip6_hdr))
1579 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1580 
1581 	ip6 = mtod(m, struct ip6_hdr *);
1582 	nxt = ip6->ip6_nxt;
1583 	len = sizeof(struct ip6_hdr);
1584 	nlen = 0;
1585 	while (len < off) {
1586 		m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1587 		switch (nxt) {
1588 		case IPPROTO_FRAGMENT:
1589 			nlen = sizeof(struct ip6_frag);
1590 			break;
1591 		case IPPROTO_AH:
1592 			nlen = (ip6e.ip6e_len + 2) << 2;
1593 			break;
1594 		default:
1595 			nlen = (ip6e.ip6e_len + 1) << 3;
1596 		}
1597 		len += nlen;
1598 		nxt = ip6e.ip6e_nxt;
1599 	}
1600 	return (len - nlen);
1601 }
1602 
1603 /*
1604  * get next header offset.  m will be retained.
1605  */
1606 int
1607 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1608 {
1609 	struct ip6_hdr ip6;
1610 	struct ip6_ext ip6e;
1611 	struct ip6_frag fh;
1612 
1613 	/* just in case */
1614 	if (m == NULL)
1615 		panic("ip6_nexthdr: m == NULL");
1616 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1617 		return -1;
1618 
1619 	switch (proto) {
1620 	case IPPROTO_IPV6:
1621 		if (m->m_pkthdr.len < off + sizeof(ip6))
1622 			return -1;
1623 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1624 		if (nxtp)
1625 			*nxtp = ip6.ip6_nxt;
1626 		off += sizeof(ip6);
1627 		return off;
1628 
1629 	case IPPROTO_FRAGMENT:
1630 		/*
1631 		 * terminate parsing if it is not the first fragment,
1632 		 * it does not make sense to parse through it.
1633 		 */
1634 		if (m->m_pkthdr.len < off + sizeof(fh))
1635 			return -1;
1636 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1637 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1638 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1639 			return -1;
1640 		if (nxtp)
1641 			*nxtp = fh.ip6f_nxt;
1642 		off += sizeof(struct ip6_frag);
1643 		return off;
1644 
1645 	case IPPROTO_AH:
1646 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1647 			return -1;
1648 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1649 		if (nxtp)
1650 			*nxtp = ip6e.ip6e_nxt;
1651 		off += (ip6e.ip6e_len + 2) << 2;
1652 		return off;
1653 
1654 	case IPPROTO_HOPOPTS:
1655 	case IPPROTO_ROUTING:
1656 	case IPPROTO_DSTOPTS:
1657 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1658 			return -1;
1659 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1660 		if (nxtp)
1661 			*nxtp = ip6e.ip6e_nxt;
1662 		off += (ip6e.ip6e_len + 1) << 3;
1663 		return off;
1664 
1665 	case IPPROTO_NONE:
1666 	case IPPROTO_ESP:
1667 	case IPPROTO_IPCOMP:
1668 		/* give up */
1669 		return -1;
1670 
1671 	default:
1672 		return -1;
1673 	}
1674 
1675 	/* NOTREACHED */
1676 }
1677 
1678 /*
1679  * get offset for the last header in the chain.  m will be kept untainted.
1680  */
1681 int
1682 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1683 {
1684 	int newoff;
1685 	int nxt;
1686 
1687 	if (!nxtp) {
1688 		nxt = -1;
1689 		nxtp = &nxt;
1690 	}
1691 	while (1) {
1692 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1693 		if (newoff < 0)
1694 			return off;
1695 		else if (newoff < off)
1696 			return -1;	/* invalid */
1697 		else if (newoff == off)
1698 			return newoff;
1699 
1700 		off = newoff;
1701 		proto = *nxtp;
1702 	}
1703 }
1704