xref: /freebsd/sys/netinet6/ip6_input.c (revision 94bb24b38701354b6a34607982483227823de864)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/hhook.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/proc.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/sdt.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/errno.h>
85 #include <sys/time.h>
86 #include <sys/kernel.h>
87 #include <sys/lock.h>
88 #include <sys/rmlock.h>
89 #include <sys/syslog.h>
90 
91 #include <net/if.h>
92 #include <net/if_var.h>
93 #include <net/if_types.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/in_pcb.h>
114 #include <netinet/icmp6.h>
115 #include <netinet6/scope6_var.h>
116 #include <netinet6/in6_ifattach.h>
117 #include <netinet6/nd6.h>
118 #include <netinet6/in6_rss.h>
119 
120 #ifdef IPSEC
121 #include <netipsec/ipsec.h>
122 #include <netinet6/ip6_ipsec.h>
123 #include <netipsec/ipsec6.h>
124 #endif /* IPSEC */
125 
126 #include <netinet6/ip6protosw.h>
127 
128 extern struct domain inet6domain;
129 
130 u_char ip6_protox[IPPROTO_MAX];
131 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
132 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
133 VNET_DEFINE(u_long, in6_ifaddrhmask);
134 
135 static struct netisr_handler ip6_nh = {
136 	.nh_name = "ip6",
137 	.nh_handler = ip6_input,
138 	.nh_proto = NETISR_IPV6,
139 #ifdef RSS
140 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
141 	.nh_policy = NETISR_POLICY_CPU,
142 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
143 #else
144 	.nh_policy = NETISR_POLICY_FLOW,
145 #endif
146 };
147 
148 #ifdef RSS
149 static struct netisr_handler ip6_direct_nh = {
150 	.nh_name = "ip6_direct",
151 	.nh_handler = ip6_direct_input,
152 	.nh_proto = NETISR_IPV6_DIRECT,
153 	.nh_m2cpuid = rss_soft_m2cpuid_v6,
154 	.nh_policy = NETISR_POLICY_CPU,
155 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
156 };
157 #endif
158 
159 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
160 
161 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
162 VNET_PCPUSTAT_SYSINIT(ip6stat);
163 #ifdef VIMAGE
164 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
165 #endif /* VIMAGE */
166 
167 struct rmlock in6_ifaddr_lock;
168 RM_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
169 
170 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
171 #ifdef PULLDOWN_TEST
172 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
173 #endif
174 
175 /*
176  * IP6 initialization: fill in IP6 protocol switch table.
177  * All protocols not implemented in kernel go to raw IP6 protocol handler.
178  */
179 void
180 ip6_init(void)
181 {
182 	struct protosw *pr;
183 	int i;
184 
185 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
186 	    &V_ip6_auto_linklocal);
187 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
188 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
189 
190 	TAILQ_INIT(&V_in6_ifaddrhead);
191 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
192 	    &V_in6_ifaddrhmask);
193 
194 	/* Initialize packet filter hooks. */
195 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
196 	V_inet6_pfil_hook.ph_af = AF_INET6;
197 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
198 		printf("%s: WARNING: unable to register pfil hook, "
199 			"error %d\n", __func__, i);
200 
201 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET6,
202 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET6],
203 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
204 		printf("%s: WARNING: unable to register input helper hook\n",
205 		    __func__);
206 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET6,
207 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET6],
208 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
209 		printf("%s: WARNING: unable to register output helper hook\n",
210 		    __func__);
211 
212 	scope6_init();
213 	addrsel_policy_init();
214 	nd6_init();
215 	frag6_init();
216 
217 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
218 
219 	/* Skip global initialization stuff for non-default instances. */
220 #ifdef VIMAGE
221 	if (!IS_DEFAULT_VNET(curvnet)) {
222 		netisr_register_vnet(&ip6_nh);
223 #ifdef RSS
224 		netisr_register_vnet(&ip6_direct_nh);
225 #endif
226 		return;
227 	}
228 #endif
229 
230 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
231 	if (pr == NULL)
232 		panic("ip6_init");
233 
234 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
235 	for (i = 0; i < IPPROTO_MAX; i++)
236 		ip6_protox[i] = pr - inet6sw;
237 	/*
238 	 * Cycle through IP protocols and put them into the appropriate place
239 	 * in ip6_protox[].
240 	 */
241 	for (pr = inet6domain.dom_protosw;
242 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
243 		if (pr->pr_domain->dom_family == PF_INET6 &&
244 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
245 			/* Be careful to only index valid IP protocols. */
246 			if (pr->pr_protocol < IPPROTO_MAX)
247 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
248 		}
249 
250 	netisr_register(&ip6_nh);
251 #ifdef RSS
252 	netisr_register(&ip6_direct_nh);
253 #endif
254 }
255 
256 /*
257  * The protocol to be inserted into ip6_protox[] must be already registered
258  * in inet6sw[], either statically or through pf_proto_register().
259  */
260 int
261 ip6proto_register(short ip6proto)
262 {
263 	struct protosw *pr;
264 
265 	/* Sanity checks. */
266 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
267 		return (EPROTONOSUPPORT);
268 
269 	/*
270 	 * The protocol slot must not be occupied by another protocol
271 	 * already.  An index pointing to IPPROTO_RAW is unused.
272 	 */
273 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
274 	if (pr == NULL)
275 		return (EPFNOSUPPORT);
276 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
277 		return (EEXIST);
278 
279 	/*
280 	 * Find the protocol position in inet6sw[] and set the index.
281 	 */
282 	for (pr = inet6domain.dom_protosw;
283 	    pr < inet6domain.dom_protoswNPROTOSW; pr++) {
284 		if (pr->pr_domain->dom_family == PF_INET6 &&
285 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
286 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
287 			return (0);
288 		}
289 	}
290 	return (EPROTONOSUPPORT);
291 }
292 
293 int
294 ip6proto_unregister(short ip6proto)
295 {
296 	struct protosw *pr;
297 
298 	/* Sanity checks. */
299 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
300 		return (EPROTONOSUPPORT);
301 
302 	/* Check if the protocol was indeed registered. */
303 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
304 	if (pr == NULL)
305 		return (EPFNOSUPPORT);
306 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
307 		return (ENOENT);
308 
309 	/* Reset the protocol slot to IPPROTO_RAW. */
310 	ip6_protox[ip6proto] = pr - inet6sw;
311 	return (0);
312 }
313 
314 #ifdef VIMAGE
315 static void
316 ip6_destroy(void *unused __unused)
317 {
318 	int error;
319 
320 #ifdef RSS
321 	netisr_unregister_vnet(&ip6_direct_nh);
322 #endif
323 	netisr_unregister_vnet(&ip6_nh);
324 
325 	if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
326 		printf("%s: WARNING: unable to unregister pfil hook, "
327 		    "error %d\n", __func__, error);
328 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
329 	if (error != 0) {
330 		printf("%s: WARNING: unable to deregister input helper hook "
331 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
332 		    "error %d returned\n", __func__, error);
333 	}
334 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
335 	if (error != 0) {
336 		printf("%s: WARNING: unable to deregister output helper hook "
337 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
338 		    "error %d returned\n", __func__, error);
339 	}
340 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
341 	nd6_destroy();
342 	in6_ifattach_destroy();
343 }
344 
345 VNET_SYSUNINIT(inet6, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_destroy, NULL);
346 #endif
347 
348 static int
349 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
350     int *nxt, int *ours)
351 {
352 	struct ip6_hdr *ip6;
353 	struct ip6_hbh *hbh;
354 
355 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
356 #if 0	/*touches NULL pointer*/
357 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
358 #endif
359 		goto out;	/* m have already been freed */
360 	}
361 
362 	/* adjust pointer */
363 	ip6 = mtod(m, struct ip6_hdr *);
364 
365 	/*
366 	 * if the payload length field is 0 and the next header field
367 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
368 	 * option MUST be included.
369 	 */
370 	if (ip6->ip6_plen == 0 && *plen == 0) {
371 		/*
372 		 * Note that if a valid jumbo payload option is
373 		 * contained, ip6_hopopts_input() must set a valid
374 		 * (non-zero) payload length to the variable plen.
375 		 */
376 		IP6STAT_INC(ip6s_badoptions);
377 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
378 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
379 		icmp6_error(m, ICMP6_PARAM_PROB,
380 			    ICMP6_PARAMPROB_HEADER,
381 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
382 		goto out;
383 	}
384 #ifndef PULLDOWN_TEST
385 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
386 	hbh = (struct ip6_hbh *)(ip6 + 1);
387 #else
388 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
389 		sizeof(struct ip6_hbh));
390 	if (hbh == NULL) {
391 		IP6STAT_INC(ip6s_tooshort);
392 		goto out;
393 	}
394 #endif
395 	*nxt = hbh->ip6h_nxt;
396 
397 	/*
398 	 * If we are acting as a router and the packet contains a
399 	 * router alert option, see if we know the option value.
400 	 * Currently, we only support the option value for MLD, in which
401 	 * case we should pass the packet to the multicast routing
402 	 * daemon.
403 	 */
404 	if (*rtalert != ~0) {
405 		switch (*rtalert) {
406 		case IP6OPT_RTALERT_MLD:
407 			if (V_ip6_forwarding)
408 				*ours = 1;
409 			break;
410 		default:
411 			/*
412 			 * RFC2711 requires unrecognized values must be
413 			 * silently ignored.
414 			 */
415 			break;
416 		}
417 	}
418 
419 	return (0);
420 
421 out:
422 	return (1);
423 }
424 
425 #ifdef RSS
426 /*
427  * IPv6 direct input routine.
428  *
429  * This is called when reinjecting completed fragments where
430  * all of the previous checking and book-keeping has been done.
431  */
432 void
433 ip6_direct_input(struct mbuf *m)
434 {
435 	int off, nxt;
436 	int nest;
437 	struct m_tag *mtag;
438 	struct ip6_direct_ctx *ip6dc;
439 
440 	mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
441 	KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
442 
443 	ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
444 	nxt = ip6dc->ip6dc_nxt;
445 	off = ip6dc->ip6dc_off;
446 
447 	nest = 0;
448 
449 	m_tag_delete(m, mtag);
450 
451 	while (nxt != IPPROTO_DONE) {
452 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
453 			IP6STAT_INC(ip6s_toomanyhdr);
454 			goto bad;
455 		}
456 
457 		/*
458 		 * protection against faulty packet - there should be
459 		 * more sanity checks in header chain processing.
460 		 */
461 		if (m->m_pkthdr.len < off) {
462 			IP6STAT_INC(ip6s_tooshort);
463 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
464 			goto bad;
465 		}
466 
467 #ifdef IPSEC
468 		/*
469 		 * enforce IPsec policy checking if we are seeing last header.
470 		 * note that we do not visit this with protocols with pcb layer
471 		 * code - like udp/tcp/raw ip.
472 		 */
473 		if (ip6_ipsec_input(m, nxt))
474 			goto bad;
475 #endif /* IPSEC */
476 
477 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
478 	}
479 	return;
480 bad:
481 	m_freem(m);
482 }
483 #endif
484 
485 void
486 ip6_input(struct mbuf *m)
487 {
488 	struct in6_addr odst;
489 	struct ip6_hdr *ip6;
490 	struct in6_ifaddr *ia;
491 	u_int32_t plen;
492 	u_int32_t rtalert = ~0;
493 	int off = sizeof(struct ip6_hdr), nest;
494 	int nxt, ours = 0;
495 	int srcrt = 0;
496 
497 #ifdef IPSEC
498 	/*
499 	 * should the inner packet be considered authentic?
500 	 * see comment in ah4_input().
501 	 * NB: m cannot be NULL when passed to the input routine
502 	 */
503 
504 	m->m_flags &= ~M_AUTHIPHDR;
505 	m->m_flags &= ~M_AUTHIPDGM;
506 
507 #endif /* IPSEC */
508 
509 	if (m->m_flags & M_FASTFWD_OURS) {
510 		/*
511 		 * Firewall changed destination to local.
512 		 */
513 		m->m_flags &= ~M_FASTFWD_OURS;
514 		ours = 1;
515 		ip6 = mtod(m, struct ip6_hdr *);
516 		goto hbhcheck;
517 	}
518 
519 	/*
520 	 * mbuf statistics
521 	 */
522 	if (m->m_flags & M_EXT) {
523 		if (m->m_next)
524 			IP6STAT_INC(ip6s_mext2m);
525 		else
526 			IP6STAT_INC(ip6s_mext1);
527 	} else {
528 		if (m->m_next) {
529 			if (m->m_flags & M_LOOP) {
530 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
531 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
532 				IP6STAT_INC(
533 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
534 			else
535 				IP6STAT_INC(ip6s_m2m[0]);
536 		} else
537 			IP6STAT_INC(ip6s_m1);
538 	}
539 
540 	/* drop the packet if IPv6 operation is disabled on the IF */
541 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED))
542 		goto bad;
543 
544 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
545 	IP6STAT_INC(ip6s_total);
546 
547 #ifndef PULLDOWN_TEST
548 	/*
549 	 * L2 bridge code and some other code can return mbuf chain
550 	 * that does not conform to KAME requirement.  too bad.
551 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
552 	 */
553 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
554 		struct mbuf *n;
555 
556 		if (m->m_pkthdr.len > MHLEN)
557 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
558 		else
559 			n = m_gethdr(M_NOWAIT, MT_DATA);
560 		if (n == NULL) {
561 			m_freem(m);
562 			return;	/* ENOBUFS */
563 		}
564 
565 		m_move_pkthdr(n, m);
566 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
567 		n->m_len = n->m_pkthdr.len;
568 		m_freem(m);
569 		m = n;
570 	}
571 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
572 #endif
573 
574 	if (m->m_len < sizeof(struct ip6_hdr)) {
575 		struct ifnet *inifp;
576 		inifp = m->m_pkthdr.rcvif;
577 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
578 			IP6STAT_INC(ip6s_toosmall);
579 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
580 			return;
581 		}
582 	}
583 
584 	ip6 = mtod(m, struct ip6_hdr *);
585 
586 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
587 		IP6STAT_INC(ip6s_badvers);
588 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
589 		goto bad;
590 	}
591 
592 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
593 
594 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
595 
596 	/*
597 	 * Check against address spoofing/corruption.
598 	 */
599 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
600 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
601 		/*
602 		 * XXX: "badscope" is not very suitable for a multicast source.
603 		 */
604 		IP6STAT_INC(ip6s_badscope);
605 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
606 		goto bad;
607 	}
608 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
609 	    !(m->m_flags & M_LOOP)) {
610 		/*
611 		 * In this case, the packet should come from the loopback
612 		 * interface.  However, we cannot just check the if_flags,
613 		 * because ip6_mloopback() passes the "actual" interface
614 		 * as the outgoing/incoming interface.
615 		 */
616 		IP6STAT_INC(ip6s_badscope);
617 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
618 		goto bad;
619 	}
620 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
621 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
622 		/*
623 		 * RFC4291 2.7:
624 		 * Nodes must not originate a packet to a multicast address
625 		 * whose scop field contains the reserved value 0; if such
626 		 * a packet is received, it must be silently dropped.
627 		 */
628 		IP6STAT_INC(ip6s_badscope);
629 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
630 		goto bad;
631 	}
632 #ifdef ALTQ
633 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
634 		/* packet is dropped by traffic conditioner */
635 		return;
636 	}
637 #endif
638 	/*
639 	 * The following check is not documented in specs.  A malicious
640 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
641 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
642 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
643 	 *
644 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
645 	 * support IPv4-less kernel compilation, we cannot support SIIT
646 	 * environment at all.  So, it makes more sense for us to reject any
647 	 * malicious packets for non-SIIT environment, than try to do a
648 	 * partial support for SIIT environment.
649 	 */
650 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
651 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
652 		IP6STAT_INC(ip6s_badscope);
653 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
654 		goto bad;
655 	}
656 #if 0
657 	/*
658 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
659 	 *
660 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
661 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
662 	 * is revised to forbid relaying case.
663 	 */
664 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
665 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
666 		IP6STAT_INC(ip6s_badscope);
667 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
668 		goto bad;
669 	}
670 #endif
671 #ifdef IPSEC
672 	/*
673 	 * Bypass packet filtering for packets previously handled by IPsec.
674 	 */
675 	if (ip6_ipsec_filtertunnel(m))
676 		goto passin;
677 #endif /* IPSEC */
678 
679 	/*
680 	 * Run through list of hooks for input packets.
681 	 *
682 	 * NB: Beware of the destination address changing
683 	 *     (e.g. by NAT rewriting).  When this happens,
684 	 *     tell ip6_forward to do the right thing.
685 	 */
686 	odst = ip6->ip6_dst;
687 
688 	/* Jump over all PFIL processing if hooks are not active. */
689 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
690 		goto passin;
691 
692 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
693 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
694 		return;
695 	if (m == NULL)			/* consumed by filter */
696 		return;
697 	ip6 = mtod(m, struct ip6_hdr *);
698 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
699 
700 	if (m->m_flags & M_FASTFWD_OURS) {
701 		m->m_flags &= ~M_FASTFWD_OURS;
702 		ours = 1;
703 		goto hbhcheck;
704 	}
705 	if ((m->m_flags & M_IP6_NEXTHOP) &&
706 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
707 		/*
708 		 * Directly ship the packet on.  This allows forwarding
709 		 * packets originally destined to us to some other directly
710 		 * connected host.
711 		 */
712 		ip6_forward(m, 1);
713 		return;
714 	}
715 
716 passin:
717 	/*
718 	 * Disambiguate address scope zones (if there is ambiguity).
719 	 * We first make sure that the original source or destination address
720 	 * is not in our internal form for scoped addresses.  Such addresses
721 	 * are not necessarily invalid spec-wise, but we cannot accept them due
722 	 * to the usage conflict.
723 	 * in6_setscope() then also checks and rejects the cases where src or
724 	 * dst are the loopback address and the receiving interface
725 	 * is not loopback.
726 	 */
727 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
728 		IP6STAT_INC(ip6s_badscope); /* XXX */
729 		goto bad;
730 	}
731 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
732 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
733 		IP6STAT_INC(ip6s_badscope);
734 		goto bad;
735 	}
736 	/*
737 	 * Multicast check. Assume packet is for us to avoid
738 	 * prematurely taking locks.
739 	 */
740 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
741 		ours = 1;
742 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
743 		goto hbhcheck;
744 	}
745 	/*
746 	 * Unicast check
747 	 * XXX: For now we keep link-local IPv6 addresses with embedded
748 	 *      scope zone id, therefore we use zero zoneid here.
749 	 */
750 	ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
751 	if (ia != NULL) {
752 		if (ia->ia6_flags & IN6_IFF_NOTREADY) {
753 			char ip6bufs[INET6_ADDRSTRLEN];
754 			char ip6bufd[INET6_ADDRSTRLEN];
755 			/* address is not ready, so discard the packet. */
756 			nd6log((LOG_INFO,
757 			    "ip6_input: packet to an unready address %s->%s\n",
758 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
759 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
760 			ifa_free(&ia->ia_ifa);
761 			goto bad;
762 		}
763 		/* Count the packet in the ip address stats */
764 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
765 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
766 		ifa_free(&ia->ia_ifa);
767 		ours = 1;
768 		goto hbhcheck;
769 	}
770 
771 	/*
772 	 * Now there is no reason to process the packet if it's not our own
773 	 * and we're not a router.
774 	 */
775 	if (!V_ip6_forwarding) {
776 		IP6STAT_INC(ip6s_cantforward);
777 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
778 		goto bad;
779 	}
780 
781   hbhcheck:
782 	/*
783 	 * Process Hop-by-Hop options header if it's contained.
784 	 * m may be modified in ip6_hopopts_input().
785 	 * If a JumboPayload option is included, plen will also be modified.
786 	 */
787 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
788 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
789 		if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0)
790 			return;
791 	} else
792 		nxt = ip6->ip6_nxt;
793 
794 	/*
795 	 * Use mbuf flags to propagate Router Alert option to
796 	 * ICMPv6 layer, as hop-by-hop options have been stripped.
797 	 */
798 	if (rtalert != ~0)
799 		m->m_flags |= M_RTALERT_MLD;
800 
801 	/*
802 	 * Check that the amount of data in the buffers
803 	 * is as at least much as the IPv6 header would have us expect.
804 	 * Trim mbufs if longer than we expect.
805 	 * Drop packet if shorter than we expect.
806 	 */
807 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
808 		IP6STAT_INC(ip6s_tooshort);
809 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
810 		goto bad;
811 	}
812 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
813 		if (m->m_len == m->m_pkthdr.len) {
814 			m->m_len = sizeof(struct ip6_hdr) + plen;
815 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
816 		} else
817 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
818 	}
819 
820 	/*
821 	 * Forward if desirable.
822 	 */
823 	if (V_ip6_mrouter &&
824 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
825 		/*
826 		 * If we are acting as a multicast router, all
827 		 * incoming multicast packets are passed to the
828 		 * kernel-level multicast forwarding function.
829 		 * The packet is returned (relatively) intact; if
830 		 * ip6_mforward() returns a non-zero value, the packet
831 		 * must be discarded, else it may be accepted below.
832 		 *
833 		 * XXX TODO: Check hlim and multicast scope here to avoid
834 		 * unnecessarily calling into ip6_mforward().
835 		 */
836 		if (ip6_mforward &&
837 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
838 			IP6STAT_INC(ip6s_cantforward);
839 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
840 			goto bad;
841 		}
842 	} else if (!ours) {
843 		ip6_forward(m, srcrt);
844 		return;
845 	}
846 
847 	ip6 = mtod(m, struct ip6_hdr *);
848 
849 	/*
850 	 * Malicious party may be able to use IPv4 mapped addr to confuse
851 	 * tcp/udp stack and bypass security checks (act as if it was from
852 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
853 	 *
854 	 * For SIIT end node behavior, you may want to disable the check.
855 	 * However, you will  become vulnerable to attacks using IPv4 mapped
856 	 * source.
857 	 */
858 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
859 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
860 		IP6STAT_INC(ip6s_badscope);
861 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
862 		goto bad;
863 	}
864 
865 	/*
866 	 * Tell launch routine the next header
867 	 */
868 	IP6STAT_INC(ip6s_delivered);
869 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver);
870 	nest = 0;
871 
872 	while (nxt != IPPROTO_DONE) {
873 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
874 			IP6STAT_INC(ip6s_toomanyhdr);
875 			goto bad;
876 		}
877 
878 		/*
879 		 * protection against faulty packet - there should be
880 		 * more sanity checks in header chain processing.
881 		 */
882 		if (m->m_pkthdr.len < off) {
883 			IP6STAT_INC(ip6s_tooshort);
884 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
885 			goto bad;
886 		}
887 
888 #ifdef IPSEC
889 		/*
890 		 * enforce IPsec policy checking if we are seeing last header.
891 		 * note that we do not visit this with protocols with pcb layer
892 		 * code - like udp/tcp/raw ip.
893 		 */
894 		if (ip6_ipsec_input(m, nxt))
895 			goto bad;
896 #endif /* IPSEC */
897 
898 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
899 	}
900 	return;
901 bad:
902 	m_freem(m);
903 }
904 
905 /*
906  * Hop-by-Hop options header processing. If a valid jumbo payload option is
907  * included, the real payload length will be stored in plenp.
908  *
909  * rtalertp - XXX: should be stored more smart way
910  */
911 static int
912 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
913     struct mbuf **mp, int *offp)
914 {
915 	struct mbuf *m = *mp;
916 	int off = *offp, hbhlen;
917 	struct ip6_hbh *hbh;
918 
919 	/* validation of the length of the header */
920 #ifndef PULLDOWN_TEST
921 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
922 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
923 	hbhlen = (hbh->ip6h_len + 1) << 3;
924 
925 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
926 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
927 #else
928 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
929 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
930 	if (hbh == NULL) {
931 		IP6STAT_INC(ip6s_tooshort);
932 		return -1;
933 	}
934 	hbhlen = (hbh->ip6h_len + 1) << 3;
935 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
936 		hbhlen);
937 	if (hbh == NULL) {
938 		IP6STAT_INC(ip6s_tooshort);
939 		return -1;
940 	}
941 #endif
942 	off += hbhlen;
943 	hbhlen -= sizeof(struct ip6_hbh);
944 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
945 				hbhlen, rtalertp, plenp) < 0)
946 		return (-1);
947 
948 	*offp = off;
949 	*mp = m;
950 	return (0);
951 }
952 
953 /*
954  * Search header for all Hop-by-hop options and process each option.
955  * This function is separate from ip6_hopopts_input() in order to
956  * handle a case where the sending node itself process its hop-by-hop
957  * options header. In such a case, the function is called from ip6_output().
958  *
959  * The function assumes that hbh header is located right after the IPv6 header
960  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
961  * opthead + hbhlen is located in contiguous memory region.
962  */
963 int
964 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
965     u_int32_t *rtalertp, u_int32_t *plenp)
966 {
967 	struct ip6_hdr *ip6;
968 	int optlen = 0;
969 	u_int8_t *opt = opthead;
970 	u_int16_t rtalert_val;
971 	u_int32_t jumboplen;
972 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
973 
974 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
975 		switch (*opt) {
976 		case IP6OPT_PAD1:
977 			optlen = 1;
978 			break;
979 		case IP6OPT_PADN:
980 			if (hbhlen < IP6OPT_MINLEN) {
981 				IP6STAT_INC(ip6s_toosmall);
982 				goto bad;
983 			}
984 			optlen = *(opt + 1) + 2;
985 			break;
986 		case IP6OPT_ROUTER_ALERT:
987 			/* XXX may need check for alignment */
988 			if (hbhlen < IP6OPT_RTALERT_LEN) {
989 				IP6STAT_INC(ip6s_toosmall);
990 				goto bad;
991 			}
992 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
993 				/* XXX stat */
994 				icmp6_error(m, ICMP6_PARAM_PROB,
995 				    ICMP6_PARAMPROB_HEADER,
996 				    erroff + opt + 1 - opthead);
997 				return (-1);
998 			}
999 			optlen = IP6OPT_RTALERT_LEN;
1000 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1001 			*rtalertp = ntohs(rtalert_val);
1002 			break;
1003 		case IP6OPT_JUMBO:
1004 			/* XXX may need check for alignment */
1005 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1006 				IP6STAT_INC(ip6s_toosmall);
1007 				goto bad;
1008 			}
1009 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1010 				/* XXX stat */
1011 				icmp6_error(m, ICMP6_PARAM_PROB,
1012 				    ICMP6_PARAMPROB_HEADER,
1013 				    erroff + opt + 1 - opthead);
1014 				return (-1);
1015 			}
1016 			optlen = IP6OPT_JUMBO_LEN;
1017 
1018 			/*
1019 			 * IPv6 packets that have non 0 payload length
1020 			 * must not contain a jumbo payload option.
1021 			 */
1022 			ip6 = mtod(m, struct ip6_hdr *);
1023 			if (ip6->ip6_plen) {
1024 				IP6STAT_INC(ip6s_badoptions);
1025 				icmp6_error(m, ICMP6_PARAM_PROB,
1026 				    ICMP6_PARAMPROB_HEADER,
1027 				    erroff + opt - opthead);
1028 				return (-1);
1029 			}
1030 
1031 			/*
1032 			 * We may see jumbolen in unaligned location, so
1033 			 * we'd need to perform bcopy().
1034 			 */
1035 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1036 			jumboplen = (u_int32_t)htonl(jumboplen);
1037 
1038 #if 1
1039 			/*
1040 			 * if there are multiple jumbo payload options,
1041 			 * *plenp will be non-zero and the packet will be
1042 			 * rejected.
1043 			 * the behavior may need some debate in ipngwg -
1044 			 * multiple options does not make sense, however,
1045 			 * there's no explicit mention in specification.
1046 			 */
1047 			if (*plenp != 0) {
1048 				IP6STAT_INC(ip6s_badoptions);
1049 				icmp6_error(m, ICMP6_PARAM_PROB,
1050 				    ICMP6_PARAMPROB_HEADER,
1051 				    erroff + opt + 2 - opthead);
1052 				return (-1);
1053 			}
1054 #endif
1055 
1056 			/*
1057 			 * jumbo payload length must be larger than 65535.
1058 			 */
1059 			if (jumboplen <= IPV6_MAXPACKET) {
1060 				IP6STAT_INC(ip6s_badoptions);
1061 				icmp6_error(m, ICMP6_PARAM_PROB,
1062 				    ICMP6_PARAMPROB_HEADER,
1063 				    erroff + opt + 2 - opthead);
1064 				return (-1);
1065 			}
1066 			*plenp = jumboplen;
1067 
1068 			break;
1069 		default:		/* unknown option */
1070 			if (hbhlen < IP6OPT_MINLEN) {
1071 				IP6STAT_INC(ip6s_toosmall);
1072 				goto bad;
1073 			}
1074 			optlen = ip6_unknown_opt(opt, m,
1075 			    erroff + opt - opthead);
1076 			if (optlen == -1)
1077 				return (-1);
1078 			optlen += 2;
1079 			break;
1080 		}
1081 	}
1082 
1083 	return (0);
1084 
1085   bad:
1086 	m_freem(m);
1087 	return (-1);
1088 }
1089 
1090 /*
1091  * Unknown option processing.
1092  * The third argument `off' is the offset from the IPv6 header to the option,
1093  * which is necessary if the IPv6 header the and option header and IPv6 header
1094  * is not contiguous in order to return an ICMPv6 error.
1095  */
1096 int
1097 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1098 {
1099 	struct ip6_hdr *ip6;
1100 
1101 	switch (IP6OPT_TYPE(*optp)) {
1102 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1103 		return ((int)*(optp + 1));
1104 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1105 		m_freem(m);
1106 		return (-1);
1107 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1108 		IP6STAT_INC(ip6s_badoptions);
1109 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1110 		return (-1);
1111 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1112 		IP6STAT_INC(ip6s_badoptions);
1113 		ip6 = mtod(m, struct ip6_hdr *);
1114 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1115 		    (m->m_flags & (M_BCAST|M_MCAST)))
1116 			m_freem(m);
1117 		else
1118 			icmp6_error(m, ICMP6_PARAM_PROB,
1119 				    ICMP6_PARAMPROB_OPTION, off);
1120 		return (-1);
1121 	}
1122 
1123 	m_freem(m);		/* XXX: NOTREACHED */
1124 	return (-1);
1125 }
1126 
1127 /*
1128  * Create the "control" list for this pcb.
1129  * These functions will not modify mbuf chain at all.
1130  *
1131  * With KAME mbuf chain restriction:
1132  * The routine will be called from upper layer handlers like tcp6_input().
1133  * Thus the routine assumes that the caller (tcp6_input) have already
1134  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1135  * very first mbuf on the mbuf chain.
1136  *
1137  * ip6_savecontrol_v4 will handle those options that are possible to be
1138  * set on a v4-mapped socket.
1139  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1140  * options and handle the v6-only ones itself.
1141  */
1142 struct mbuf **
1143 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1144     int *v4only)
1145 {
1146 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1147 
1148 #ifdef SO_TIMESTAMP
1149 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1150 		struct timeval tv;
1151 
1152 		microtime(&tv);
1153 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1154 		    SCM_TIMESTAMP, SOL_SOCKET);
1155 		if (*mp)
1156 			mp = &(*mp)->m_next;
1157 	}
1158 #endif
1159 
1160 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1161 	/* RFC 2292 sec. 5 */
1162 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1163 		struct in6_pktinfo pi6;
1164 
1165 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1166 #ifdef INET
1167 			struct ip *ip;
1168 
1169 			ip = mtod(m, struct ip *);
1170 			pi6.ipi6_addr.s6_addr32[0] = 0;
1171 			pi6.ipi6_addr.s6_addr32[1] = 0;
1172 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1173 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1174 #else
1175 			/* We won't hit this code */
1176 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1177 #endif
1178 		} else {
1179 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1180 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1181 		}
1182 		pi6.ipi6_ifindex =
1183 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1184 
1185 		*mp = sbcreatecontrol((caddr_t) &pi6,
1186 		    sizeof(struct in6_pktinfo),
1187 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1188 		if (*mp)
1189 			mp = &(*mp)->m_next;
1190 	}
1191 
1192 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1193 		int hlim;
1194 
1195 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1196 #ifdef INET
1197 			struct ip *ip;
1198 
1199 			ip = mtod(m, struct ip *);
1200 			hlim = ip->ip_ttl;
1201 #else
1202 			/* We won't hit this code */
1203 			hlim = 0;
1204 #endif
1205 		} else {
1206 			hlim = ip6->ip6_hlim & 0xff;
1207 		}
1208 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1209 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1210 		    IPPROTO_IPV6);
1211 		if (*mp)
1212 			mp = &(*mp)->m_next;
1213 	}
1214 
1215 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1216 		int tclass;
1217 
1218 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1219 #ifdef INET
1220 			struct ip *ip;
1221 
1222 			ip = mtod(m, struct ip *);
1223 			tclass = ip->ip_tos;
1224 #else
1225 			/* We won't hit this code */
1226 			tclass = 0;
1227 #endif
1228 		} else {
1229 			u_int32_t flowinfo;
1230 
1231 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1232 			flowinfo >>= 20;
1233 			tclass = flowinfo & 0xff;
1234 		}
1235 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1236 		    IPV6_TCLASS, IPPROTO_IPV6);
1237 		if (*mp)
1238 			mp = &(*mp)->m_next;
1239 	}
1240 
1241 	if (v4only != NULL) {
1242 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1243 			*v4only = 1;
1244 		} else {
1245 			*v4only = 0;
1246 		}
1247 	}
1248 
1249 	return (mp);
1250 }
1251 
1252 void
1253 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1254 {
1255 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1256 	int v4only = 0;
1257 
1258 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1259 	if (v4only)
1260 		return;
1261 
1262 	/*
1263 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1264 	 * privilege for the option (see ip6_ctloutput), but it might be too
1265 	 * strict, since there might be some hop-by-hop options which can be
1266 	 * returned to normal user.
1267 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1268 	 */
1269 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1270 		/*
1271 		 * Check if a hop-by-hop options header is contatined in the
1272 		 * received packet, and if so, store the options as ancillary
1273 		 * data. Note that a hop-by-hop options header must be
1274 		 * just after the IPv6 header, which is assured through the
1275 		 * IPv6 input processing.
1276 		 */
1277 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1278 			struct ip6_hbh *hbh;
1279 			int hbhlen = 0;
1280 #ifdef PULLDOWN_TEST
1281 			struct mbuf *ext;
1282 #endif
1283 
1284 #ifndef PULLDOWN_TEST
1285 			hbh = (struct ip6_hbh *)(ip6 + 1);
1286 			hbhlen = (hbh->ip6h_len + 1) << 3;
1287 #else
1288 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1289 			    ip6->ip6_nxt);
1290 			if (ext == NULL) {
1291 				IP6STAT_INC(ip6s_tooshort);
1292 				return;
1293 			}
1294 			hbh = mtod(ext, struct ip6_hbh *);
1295 			hbhlen = (hbh->ip6h_len + 1) << 3;
1296 			if (hbhlen != ext->m_len) {
1297 				m_freem(ext);
1298 				IP6STAT_INC(ip6s_tooshort);
1299 				return;
1300 			}
1301 #endif
1302 
1303 			/*
1304 			 * XXX: We copy the whole header even if a
1305 			 * jumbo payload option is included, the option which
1306 			 * is to be removed before returning according to
1307 			 * RFC2292.
1308 			 * Note: this constraint is removed in RFC3542
1309 			 */
1310 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1311 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1312 			    IPPROTO_IPV6);
1313 			if (*mp)
1314 				mp = &(*mp)->m_next;
1315 #ifdef PULLDOWN_TEST
1316 			m_freem(ext);
1317 #endif
1318 		}
1319 	}
1320 
1321 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1322 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1323 
1324 		/*
1325 		 * Search for destination options headers or routing
1326 		 * header(s) through the header chain, and stores each
1327 		 * header as ancillary data.
1328 		 * Note that the order of the headers remains in
1329 		 * the chain of ancillary data.
1330 		 */
1331 		while (1) {	/* is explicit loop prevention necessary? */
1332 			struct ip6_ext *ip6e = NULL;
1333 			int elen;
1334 #ifdef PULLDOWN_TEST
1335 			struct mbuf *ext = NULL;
1336 #endif
1337 
1338 			/*
1339 			 * if it is not an extension header, don't try to
1340 			 * pull it from the chain.
1341 			 */
1342 			switch (nxt) {
1343 			case IPPROTO_DSTOPTS:
1344 			case IPPROTO_ROUTING:
1345 			case IPPROTO_HOPOPTS:
1346 			case IPPROTO_AH: /* is it possible? */
1347 				break;
1348 			default:
1349 				goto loopend;
1350 			}
1351 
1352 #ifndef PULLDOWN_TEST
1353 			if (off + sizeof(*ip6e) > m->m_len)
1354 				goto loopend;
1355 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1356 			if (nxt == IPPROTO_AH)
1357 				elen = (ip6e->ip6e_len + 2) << 2;
1358 			else
1359 				elen = (ip6e->ip6e_len + 1) << 3;
1360 			if (off + elen > m->m_len)
1361 				goto loopend;
1362 #else
1363 			ext = ip6_pullexthdr(m, off, nxt);
1364 			if (ext == NULL) {
1365 				IP6STAT_INC(ip6s_tooshort);
1366 				return;
1367 			}
1368 			ip6e = mtod(ext, struct ip6_ext *);
1369 			if (nxt == IPPROTO_AH)
1370 				elen = (ip6e->ip6e_len + 2) << 2;
1371 			else
1372 				elen = (ip6e->ip6e_len + 1) << 3;
1373 			if (elen != ext->m_len) {
1374 				m_freem(ext);
1375 				IP6STAT_INC(ip6s_tooshort);
1376 				return;
1377 			}
1378 #endif
1379 
1380 			switch (nxt) {
1381 			case IPPROTO_DSTOPTS:
1382 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1383 					break;
1384 
1385 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1386 				    IS2292(in6p,
1387 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1388 				    IPPROTO_IPV6);
1389 				if (*mp)
1390 					mp = &(*mp)->m_next;
1391 				break;
1392 			case IPPROTO_ROUTING:
1393 				if (!(in6p->inp_flags & IN6P_RTHDR))
1394 					break;
1395 
1396 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1397 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1398 				    IPPROTO_IPV6);
1399 				if (*mp)
1400 					mp = &(*mp)->m_next;
1401 				break;
1402 			case IPPROTO_HOPOPTS:
1403 			case IPPROTO_AH: /* is it possible? */
1404 				break;
1405 
1406 			default:
1407 				/*
1408 				 * other cases have been filtered in the above.
1409 				 * none will visit this case.  here we supply
1410 				 * the code just in case (nxt overwritten or
1411 				 * other cases).
1412 				 */
1413 #ifdef PULLDOWN_TEST
1414 				m_freem(ext);
1415 #endif
1416 				goto loopend;
1417 
1418 			}
1419 
1420 			/* proceed with the next header. */
1421 			off += elen;
1422 			nxt = ip6e->ip6e_nxt;
1423 			ip6e = NULL;
1424 #ifdef PULLDOWN_TEST
1425 			m_freem(ext);
1426 			ext = NULL;
1427 #endif
1428 		}
1429 	  loopend:
1430 		;
1431 	}
1432 
1433 	if (in6p->inp_flags2 & INP_RECVFLOWID) {
1434 		uint32_t flowid, flow_type;
1435 
1436 		flowid = m->m_pkthdr.flowid;
1437 		flow_type = M_HASHTYPE_GET(m);
1438 
1439 		/*
1440 		 * XXX should handle the failure of one or the
1441 		 * other - don't populate both?
1442 		 */
1443 		*mp = sbcreatecontrol((caddr_t) &flowid,
1444 		    sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6);
1445 		if (*mp)
1446 			mp = &(*mp)->m_next;
1447 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1448 		    sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6);
1449 		if (*mp)
1450 			mp = &(*mp)->m_next;
1451 	}
1452 
1453 #ifdef	RSS
1454 	if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) {
1455 		uint32_t flowid, flow_type;
1456 		uint32_t rss_bucketid;
1457 
1458 		flowid = m->m_pkthdr.flowid;
1459 		flow_type = M_HASHTYPE_GET(m);
1460 
1461 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1462 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1463 			   sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6);
1464 			if (*mp)
1465 				mp = &(*mp)->m_next;
1466 		}
1467 	}
1468 #endif
1469 
1470 }
1471 #undef IS2292
1472 
1473 void
1474 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1475 {
1476 	struct socket *so;
1477 	struct mbuf *m_mtu;
1478 	struct ip6_mtuinfo mtuctl;
1479 
1480 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1481 	/*
1482 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1483 	 * application wanted to know the MTU value.
1484 	 * NOTE: we notify disconnected sockets, because some udp
1485 	 * applications keep sending sockets disconnected.
1486 	 * NOTE: our implementation doesn't notify connected sockets that has
1487 	 * foreign address that is different than given destination addresses
1488 	 * (this is permitted by RFC 3542).
1489 	 */
1490 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1491 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1492 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1493 		return;
1494 
1495 	mtuctl.ip6m_mtu = mtu;
1496 	mtuctl.ip6m_addr = *dst;
1497 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1498 		return;
1499 
1500 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1501 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1502 		return;
1503 
1504 	so =  inp->inp_socket;
1505 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1506 	    == 0) {
1507 		m_freem(m_mtu);
1508 		/* XXX: should count statistics */
1509 	} else
1510 		sorwakeup(so);
1511 }
1512 
1513 #ifdef PULLDOWN_TEST
1514 /*
1515  * pull single extension header from mbuf chain.  returns single mbuf that
1516  * contains the result, or NULL on error.
1517  */
1518 static struct mbuf *
1519 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1520 {
1521 	struct ip6_ext ip6e;
1522 	size_t elen;
1523 	struct mbuf *n;
1524 
1525 #ifdef DIAGNOSTIC
1526 	switch (nxt) {
1527 	case IPPROTO_DSTOPTS:
1528 	case IPPROTO_ROUTING:
1529 	case IPPROTO_HOPOPTS:
1530 	case IPPROTO_AH: /* is it possible? */
1531 		break;
1532 	default:
1533 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1534 	}
1535 #endif
1536 
1537 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1538 	if (nxt == IPPROTO_AH)
1539 		elen = (ip6e.ip6e_len + 2) << 2;
1540 	else
1541 		elen = (ip6e.ip6e_len + 1) << 3;
1542 
1543 	if (elen > MLEN)
1544 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1545 	else
1546 		n = m_get(M_NOWAIT, MT_DATA);
1547 	if (n == NULL)
1548 		return NULL;
1549 
1550 	m_copydata(m, off, elen, mtod(n, caddr_t));
1551 	n->m_len = elen;
1552 	return n;
1553 }
1554 #endif
1555 
1556 /*
1557  * Get pointer to the previous header followed by the header
1558  * currently processed.
1559  * XXX: This function supposes that
1560  *	M includes all headers,
1561  *	the next header field and the header length field of each header
1562  *	are valid, and
1563  *	the sum of each header length equals to OFF.
1564  * Because of these assumptions, this function must be called very
1565  * carefully. Moreover, it will not be used in the near future when
1566  * we develop `neater' mechanism to process extension headers.
1567  */
1568 char *
1569 ip6_get_prevhdr(const struct mbuf *m, int off)
1570 {
1571 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1572 
1573 	if (off == sizeof(struct ip6_hdr))
1574 		return (&ip6->ip6_nxt);
1575 	else {
1576 		int len, nxt;
1577 		struct ip6_ext *ip6e = NULL;
1578 
1579 		nxt = ip6->ip6_nxt;
1580 		len = sizeof(struct ip6_hdr);
1581 		while (len < off) {
1582 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1583 
1584 			switch (nxt) {
1585 			case IPPROTO_FRAGMENT:
1586 				len += sizeof(struct ip6_frag);
1587 				break;
1588 			case IPPROTO_AH:
1589 				len += (ip6e->ip6e_len + 2) << 2;
1590 				break;
1591 			default:
1592 				len += (ip6e->ip6e_len + 1) << 3;
1593 				break;
1594 			}
1595 			nxt = ip6e->ip6e_nxt;
1596 		}
1597 		if (ip6e)
1598 			return (&ip6e->ip6e_nxt);
1599 		else
1600 			return NULL;
1601 	}
1602 }
1603 
1604 /*
1605  * get next header offset.  m will be retained.
1606  */
1607 int
1608 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1609 {
1610 	struct ip6_hdr ip6;
1611 	struct ip6_ext ip6e;
1612 	struct ip6_frag fh;
1613 
1614 	/* just in case */
1615 	if (m == NULL)
1616 		panic("ip6_nexthdr: m == NULL");
1617 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1618 		return -1;
1619 
1620 	switch (proto) {
1621 	case IPPROTO_IPV6:
1622 		if (m->m_pkthdr.len < off + sizeof(ip6))
1623 			return -1;
1624 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1625 		if (nxtp)
1626 			*nxtp = ip6.ip6_nxt;
1627 		off += sizeof(ip6);
1628 		return off;
1629 
1630 	case IPPROTO_FRAGMENT:
1631 		/*
1632 		 * terminate parsing if it is not the first fragment,
1633 		 * it does not make sense to parse through it.
1634 		 */
1635 		if (m->m_pkthdr.len < off + sizeof(fh))
1636 			return -1;
1637 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1638 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1639 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1640 			return -1;
1641 		if (nxtp)
1642 			*nxtp = fh.ip6f_nxt;
1643 		off += sizeof(struct ip6_frag);
1644 		return off;
1645 
1646 	case IPPROTO_AH:
1647 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1648 			return -1;
1649 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1650 		if (nxtp)
1651 			*nxtp = ip6e.ip6e_nxt;
1652 		off += (ip6e.ip6e_len + 2) << 2;
1653 		return off;
1654 
1655 	case IPPROTO_HOPOPTS:
1656 	case IPPROTO_ROUTING:
1657 	case IPPROTO_DSTOPTS:
1658 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1659 			return -1;
1660 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1661 		if (nxtp)
1662 			*nxtp = ip6e.ip6e_nxt;
1663 		off += (ip6e.ip6e_len + 1) << 3;
1664 		return off;
1665 
1666 	case IPPROTO_NONE:
1667 	case IPPROTO_ESP:
1668 	case IPPROTO_IPCOMP:
1669 		/* give up */
1670 		return -1;
1671 
1672 	default:
1673 		return -1;
1674 	}
1675 
1676 	/* NOTREACHED */
1677 }
1678 
1679 /*
1680  * get offset for the last header in the chain.  m will be kept untainted.
1681  */
1682 int
1683 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1684 {
1685 	int newoff;
1686 	int nxt;
1687 
1688 	if (!nxtp) {
1689 		nxt = -1;
1690 		nxtp = &nxt;
1691 	}
1692 	while (1) {
1693 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1694 		if (newoff < 0)
1695 			return off;
1696 		else if (newoff < off)
1697 			return -1;	/* invalid */
1698 		else if (newoff == off)
1699 			return newoff;
1700 
1701 		off = newoff;
1702 		proto = *nxtp;
1703 	}
1704 }
1705 
1706 /*
1707  * System control for IP6
1708  */
1709 
1710 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1711 	0,		0,		0,		0,
1712 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1713 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1714 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1715 	0,		0,		0,		0,
1716 	ENOPROTOOPT
1717 };
1718