xref: /freebsd/sys/netinet6/ip6_input.c (revision 46c1105fbb6fbff6d6ccd0a18571342eb992d637)
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 	if (!IS_DEFAULT_VNET(curvnet))
221 		return;
222 
223 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
224 	if (pr == NULL)
225 		panic("ip6_init");
226 
227 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
228 	for (i = 0; i < IPPROTO_MAX; i++)
229 		ip6_protox[i] = pr - inet6sw;
230 	/*
231 	 * Cycle through IP protocols and put them into the appropriate place
232 	 * in ip6_protox[].
233 	 */
234 	for (pr = inet6domain.dom_protosw;
235 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
236 		if (pr->pr_domain->dom_family == PF_INET6 &&
237 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
238 			/* Be careful to only index valid IP protocols. */
239 			if (pr->pr_protocol < IPPROTO_MAX)
240 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
241 		}
242 
243 	netisr_register(&ip6_nh);
244 #ifdef RSS
245 	netisr_register(&ip6_direct_nh);
246 #endif
247 }
248 
249 /*
250  * The protocol to be inserted into ip6_protox[] must be already registered
251  * in inet6sw[], either statically or through pf_proto_register().
252  */
253 int
254 ip6proto_register(short ip6proto)
255 {
256 	struct protosw *pr;
257 
258 	/* Sanity checks. */
259 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
260 		return (EPROTONOSUPPORT);
261 
262 	/*
263 	 * The protocol slot must not be occupied by another protocol
264 	 * already.  An index pointing to IPPROTO_RAW is unused.
265 	 */
266 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
267 	if (pr == NULL)
268 		return (EPFNOSUPPORT);
269 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
270 		return (EEXIST);
271 
272 	/*
273 	 * Find the protocol position in inet6sw[] and set the index.
274 	 */
275 	for (pr = inet6domain.dom_protosw;
276 	    pr < inet6domain.dom_protoswNPROTOSW; pr++) {
277 		if (pr->pr_domain->dom_family == PF_INET6 &&
278 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
279 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
280 			return (0);
281 		}
282 	}
283 	return (EPROTONOSUPPORT);
284 }
285 
286 int
287 ip6proto_unregister(short ip6proto)
288 {
289 	struct protosw *pr;
290 
291 	/* Sanity checks. */
292 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
293 		return (EPROTONOSUPPORT);
294 
295 	/* Check if the protocol was indeed registered. */
296 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
297 	if (pr == NULL)
298 		return (EPFNOSUPPORT);
299 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
300 		return (ENOENT);
301 
302 	/* Reset the protocol slot to IPPROTO_RAW. */
303 	ip6_protox[ip6proto] = pr - inet6sw;
304 	return (0);
305 }
306 
307 #ifdef VIMAGE
308 void
309 ip6_destroy()
310 {
311 	int error;
312 
313 	if ((error = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
314 		printf("%s: WARNING: unable to unregister pfil hook, "
315 		    "error %d\n", __func__, error);
316 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET6]);
317 	if (error != 0) {
318 		printf("%s: WARNING: unable to deregister input helper hook "
319 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET6: "
320 		    "error %d returned\n", __func__, error);
321 	}
322 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET6]);
323 	if (error != 0) {
324 		printf("%s: WARNING: unable to deregister output helper hook "
325 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET6: "
326 		    "error %d returned\n", __func__, error);
327 	}
328 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
329 	nd6_destroy();
330 	in6_ifattach_destroy();
331 }
332 #endif
333 
334 static int
335 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
336     int *nxt, int *ours)
337 {
338 	struct ip6_hdr *ip6;
339 	struct ip6_hbh *hbh;
340 
341 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
342 #if 0	/*touches NULL pointer*/
343 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
344 #endif
345 		goto out;	/* m have already been freed */
346 	}
347 
348 	/* adjust pointer */
349 	ip6 = mtod(m, struct ip6_hdr *);
350 
351 	/*
352 	 * if the payload length field is 0 and the next header field
353 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
354 	 * option MUST be included.
355 	 */
356 	if (ip6->ip6_plen == 0 && *plen == 0) {
357 		/*
358 		 * Note that if a valid jumbo payload option is
359 		 * contained, ip6_hopopts_input() must set a valid
360 		 * (non-zero) payload length to the variable plen.
361 		 */
362 		IP6STAT_INC(ip6s_badoptions);
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
364 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
365 		icmp6_error(m, ICMP6_PARAM_PROB,
366 			    ICMP6_PARAMPROB_HEADER,
367 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
368 		goto out;
369 	}
370 #ifndef PULLDOWN_TEST
371 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
372 	hbh = (struct ip6_hbh *)(ip6 + 1);
373 #else
374 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
375 		sizeof(struct ip6_hbh));
376 	if (hbh == NULL) {
377 		IP6STAT_INC(ip6s_tooshort);
378 		goto out;
379 	}
380 #endif
381 	*nxt = hbh->ip6h_nxt;
382 
383 	/*
384 	 * If we are acting as a router and the packet contains a
385 	 * router alert option, see if we know the option value.
386 	 * Currently, we only support the option value for MLD, in which
387 	 * case we should pass the packet to the multicast routing
388 	 * daemon.
389 	 */
390 	if (*rtalert != ~0) {
391 		switch (*rtalert) {
392 		case IP6OPT_RTALERT_MLD:
393 			if (V_ip6_forwarding)
394 				*ours = 1;
395 			break;
396 		default:
397 			/*
398 			 * RFC2711 requires unrecognized values must be
399 			 * silently ignored.
400 			 */
401 			break;
402 		}
403 	}
404 
405 	return (0);
406 
407 out:
408 	return (1);
409 }
410 
411 #ifdef RSS
412 /*
413  * IPv6 direct input routine.
414  *
415  * This is called when reinjecting completed fragments where
416  * all of the previous checking and book-keeping has been done.
417  */
418 void
419 ip6_direct_input(struct mbuf *m)
420 {
421 	int off, nxt;
422 	int nest;
423 	struct m_tag *mtag;
424 	struct ip6_direct_ctx *ip6dc;
425 
426 	mtag = m_tag_locate(m, MTAG_ABI_IPV6, IPV6_TAG_DIRECT, NULL);
427 	KASSERT(mtag != NULL, ("Reinjected packet w/o direct ctx tag!"));
428 
429 	ip6dc = (struct ip6_direct_ctx *)(mtag + 1);
430 	nxt = ip6dc->ip6dc_nxt;
431 	off = ip6dc->ip6dc_off;
432 
433 	nest = 0;
434 
435 	m_tag_delete(m, mtag);
436 
437 	while (nxt != IPPROTO_DONE) {
438 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
439 			IP6STAT_INC(ip6s_toomanyhdr);
440 			goto bad;
441 		}
442 
443 		/*
444 		 * protection against faulty packet - there should be
445 		 * more sanity checks in header chain processing.
446 		 */
447 		if (m->m_pkthdr.len < off) {
448 			IP6STAT_INC(ip6s_tooshort);
449 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
450 			goto bad;
451 		}
452 
453 #ifdef IPSEC
454 		/*
455 		 * enforce IPsec policy checking if we are seeing last header.
456 		 * note that we do not visit this with protocols with pcb layer
457 		 * code - like udp/tcp/raw ip.
458 		 */
459 		if (ip6_ipsec_input(m, nxt))
460 			goto bad;
461 #endif /* IPSEC */
462 
463 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
464 	}
465 	return;
466 bad:
467 	m_freem(m);
468 }
469 #endif
470 
471 void
472 ip6_input(struct mbuf *m)
473 {
474 	struct in6_addr odst;
475 	struct ip6_hdr *ip6;
476 	struct in6_ifaddr *ia;
477 	u_int32_t plen;
478 	u_int32_t rtalert = ~0;
479 	int off = sizeof(struct ip6_hdr), nest;
480 	int nxt, ours = 0;
481 	int srcrt = 0;
482 
483 #ifdef IPSEC
484 	/*
485 	 * should the inner packet be considered authentic?
486 	 * see comment in ah4_input().
487 	 * NB: m cannot be NULL when passed to the input routine
488 	 */
489 
490 	m->m_flags &= ~M_AUTHIPHDR;
491 	m->m_flags &= ~M_AUTHIPDGM;
492 
493 #endif /* IPSEC */
494 
495 	if (m->m_flags & M_FASTFWD_OURS) {
496 		/*
497 		 * Firewall changed destination to local.
498 		 */
499 		m->m_flags &= ~M_FASTFWD_OURS;
500 		ours = 1;
501 		ip6 = mtod(m, struct ip6_hdr *);
502 		goto hbhcheck;
503 	}
504 
505 	/*
506 	 * mbuf statistics
507 	 */
508 	if (m->m_flags & M_EXT) {
509 		if (m->m_next)
510 			IP6STAT_INC(ip6s_mext2m);
511 		else
512 			IP6STAT_INC(ip6s_mext1);
513 	} else {
514 		if (m->m_next) {
515 			if (m->m_flags & M_LOOP) {
516 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
517 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
518 				IP6STAT_INC(
519 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
520 			else
521 				IP6STAT_INC(ip6s_m2m[0]);
522 		} else
523 			IP6STAT_INC(ip6s_m1);
524 	}
525 
526 	/* drop the packet if IPv6 operation is disabled on the IF */
527 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED))
528 		goto bad;
529 
530 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
531 	IP6STAT_INC(ip6s_total);
532 
533 #ifndef PULLDOWN_TEST
534 	/*
535 	 * L2 bridge code and some other code can return mbuf chain
536 	 * that does not conform to KAME requirement.  too bad.
537 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
538 	 */
539 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
540 		struct mbuf *n;
541 
542 		if (m->m_pkthdr.len > MHLEN)
543 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
544 		else
545 			n = m_gethdr(M_NOWAIT, MT_DATA);
546 		if (n == NULL) {
547 			m_freem(m);
548 			return;	/* ENOBUFS */
549 		}
550 
551 		m_move_pkthdr(n, m);
552 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
553 		n->m_len = n->m_pkthdr.len;
554 		m_freem(m);
555 		m = n;
556 	}
557 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
558 #endif
559 
560 	if (m->m_len < sizeof(struct ip6_hdr)) {
561 		struct ifnet *inifp;
562 		inifp = m->m_pkthdr.rcvif;
563 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
564 			IP6STAT_INC(ip6s_toosmall);
565 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
566 			return;
567 		}
568 	}
569 
570 	ip6 = mtod(m, struct ip6_hdr *);
571 
572 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
573 		IP6STAT_INC(ip6s_badvers);
574 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
575 		goto bad;
576 	}
577 
578 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
579 
580 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
581 
582 	/*
583 	 * Check against address spoofing/corruption.
584 	 */
585 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
586 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
587 		/*
588 		 * XXX: "badscope" is not very suitable for a multicast source.
589 		 */
590 		IP6STAT_INC(ip6s_badscope);
591 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
592 		goto bad;
593 	}
594 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
595 	    !(m->m_flags & M_LOOP)) {
596 		/*
597 		 * In this case, the packet should come from the loopback
598 		 * interface.  However, we cannot just check the if_flags,
599 		 * because ip6_mloopback() passes the "actual" interface
600 		 * as the outgoing/incoming interface.
601 		 */
602 		IP6STAT_INC(ip6s_badscope);
603 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
604 		goto bad;
605 	}
606 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
607 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
608 		/*
609 		 * RFC4291 2.7:
610 		 * Nodes must not originate a packet to a multicast address
611 		 * whose scop field contains the reserved value 0; if such
612 		 * a packet is received, it must be silently dropped.
613 		 */
614 		IP6STAT_INC(ip6s_badscope);
615 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
616 		goto bad;
617 	}
618 #ifdef ALTQ
619 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
620 		/* packet is dropped by traffic conditioner */
621 		return;
622 	}
623 #endif
624 	/*
625 	 * The following check is not documented in specs.  A malicious
626 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
627 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
628 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
629 	 *
630 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
631 	 * support IPv4-less kernel compilation, we cannot support SIIT
632 	 * environment at all.  So, it makes more sense for us to reject any
633 	 * malicious packets for non-SIIT environment, than try to do a
634 	 * partial support for SIIT environment.
635 	 */
636 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
637 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
638 		IP6STAT_INC(ip6s_badscope);
639 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
640 		goto bad;
641 	}
642 #if 0
643 	/*
644 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
645 	 *
646 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
647 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
648 	 * is revised to forbid relaying case.
649 	 */
650 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
651 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
652 		IP6STAT_INC(ip6s_badscope);
653 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
654 		goto bad;
655 	}
656 #endif
657 #ifdef IPSEC
658 	/*
659 	 * Bypass packet filtering for packets previously handled by IPsec.
660 	 */
661 	if (ip6_ipsec_filtertunnel(m))
662 		goto passin;
663 #endif /* IPSEC */
664 
665 	/*
666 	 * Run through list of hooks for input packets.
667 	 *
668 	 * NB: Beware of the destination address changing
669 	 *     (e.g. by NAT rewriting).  When this happens,
670 	 *     tell ip6_forward to do the right thing.
671 	 */
672 	odst = ip6->ip6_dst;
673 
674 	/* Jump over all PFIL processing if hooks are not active. */
675 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
676 		goto passin;
677 
678 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
679 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
680 		return;
681 	if (m == NULL)			/* consumed by filter */
682 		return;
683 	ip6 = mtod(m, struct ip6_hdr *);
684 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
685 
686 	if (m->m_flags & M_FASTFWD_OURS) {
687 		m->m_flags &= ~M_FASTFWD_OURS;
688 		ours = 1;
689 		goto hbhcheck;
690 	}
691 	if ((m->m_flags & M_IP6_NEXTHOP) &&
692 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
693 		/*
694 		 * Directly ship the packet on.  This allows forwarding
695 		 * packets originally destined to us to some other directly
696 		 * connected host.
697 		 */
698 		ip6_forward(m, 1);
699 		return;
700 	}
701 
702 passin:
703 	/*
704 	 * Disambiguate address scope zones (if there is ambiguity).
705 	 * We first make sure that the original source or destination address
706 	 * is not in our internal form for scoped addresses.  Such addresses
707 	 * are not necessarily invalid spec-wise, but we cannot accept them due
708 	 * to the usage conflict.
709 	 * in6_setscope() then also checks and rejects the cases where src or
710 	 * dst are the loopback address and the receiving interface
711 	 * is not loopback.
712 	 */
713 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
714 		IP6STAT_INC(ip6s_badscope); /* XXX */
715 		goto bad;
716 	}
717 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
718 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
719 		IP6STAT_INC(ip6s_badscope);
720 		goto bad;
721 	}
722 	/*
723 	 * Multicast check. Assume packet is for us to avoid
724 	 * prematurely taking locks.
725 	 */
726 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
727 		ours = 1;
728 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
729 		goto hbhcheck;
730 	}
731 	/*
732 	 * Unicast check
733 	 * XXX: For now we keep link-local IPv6 addresses with embedded
734 	 *      scope zone id, therefore we use zero zoneid here.
735 	 */
736 	ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
737 	if (ia != NULL) {
738 		if (ia->ia6_flags & IN6_IFF_NOTREADY) {
739 			char ip6bufs[INET6_ADDRSTRLEN];
740 			char ip6bufd[INET6_ADDRSTRLEN];
741 			/* address is not ready, so discard the packet. */
742 			nd6log((LOG_INFO,
743 			    "ip6_input: packet to an unready address %s->%s\n",
744 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
745 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
746 			ifa_free(&ia->ia_ifa);
747 			goto bad;
748 		}
749 		/* Count the packet in the ip address stats */
750 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
751 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
752 		ifa_free(&ia->ia_ifa);
753 		ours = 1;
754 		goto hbhcheck;
755 	}
756 
757 	/*
758 	 * Now there is no reason to process the packet if it's not our own
759 	 * and we're not a router.
760 	 */
761 	if (!V_ip6_forwarding) {
762 		IP6STAT_INC(ip6s_cantforward);
763 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
764 		goto bad;
765 	}
766 
767   hbhcheck:
768 	/*
769 	 * Process Hop-by-Hop options header if it's contained.
770 	 * m may be modified in ip6_hopopts_input().
771 	 * If a JumboPayload option is included, plen will also be modified.
772 	 */
773 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
774 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
775 		if (ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours) != 0)
776 			return;
777 	} else
778 		nxt = ip6->ip6_nxt;
779 
780 	/*
781 	 * Use mbuf flags to propagate Router Alert option to
782 	 * ICMPv6 layer, as hop-by-hop options have been stripped.
783 	 */
784 	if (rtalert != ~0)
785 		m->m_flags |= M_RTALERT_MLD;
786 
787 	/*
788 	 * Check that the amount of data in the buffers
789 	 * is as at least much as the IPv6 header would have us expect.
790 	 * Trim mbufs if longer than we expect.
791 	 * Drop packet if shorter than we expect.
792 	 */
793 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
794 		IP6STAT_INC(ip6s_tooshort);
795 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
796 		goto bad;
797 	}
798 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
799 		if (m->m_len == m->m_pkthdr.len) {
800 			m->m_len = sizeof(struct ip6_hdr) + plen;
801 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
802 		} else
803 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
804 	}
805 
806 	/*
807 	 * Forward if desirable.
808 	 */
809 	if (V_ip6_mrouter &&
810 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
811 		/*
812 		 * If we are acting as a multicast router, all
813 		 * incoming multicast packets are passed to the
814 		 * kernel-level multicast forwarding function.
815 		 * The packet is returned (relatively) intact; if
816 		 * ip6_mforward() returns a non-zero value, the packet
817 		 * must be discarded, else it may be accepted below.
818 		 *
819 		 * XXX TODO: Check hlim and multicast scope here to avoid
820 		 * unnecessarily calling into ip6_mforward().
821 		 */
822 		if (ip6_mforward &&
823 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
824 			IP6STAT_INC(ip6s_cantforward);
825 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
826 			goto bad;
827 		}
828 	} else if (!ours) {
829 		ip6_forward(m, srcrt);
830 		return;
831 	}
832 
833 	ip6 = mtod(m, struct ip6_hdr *);
834 
835 	/*
836 	 * Malicious party may be able to use IPv4 mapped addr to confuse
837 	 * tcp/udp stack and bypass security checks (act as if it was from
838 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
839 	 *
840 	 * For SIIT end node behavior, you may want to disable the check.
841 	 * However, you will  become vulnerable to attacks using IPv4 mapped
842 	 * source.
843 	 */
844 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
845 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
846 		IP6STAT_INC(ip6s_badscope);
847 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
848 		goto bad;
849 	}
850 
851 	/*
852 	 * Tell launch routine the next header
853 	 */
854 	IP6STAT_INC(ip6s_delivered);
855 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_deliver);
856 	nest = 0;
857 
858 	while (nxt != IPPROTO_DONE) {
859 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
860 			IP6STAT_INC(ip6s_toomanyhdr);
861 			goto bad;
862 		}
863 
864 		/*
865 		 * protection against faulty packet - there should be
866 		 * more sanity checks in header chain processing.
867 		 */
868 		if (m->m_pkthdr.len < off) {
869 			IP6STAT_INC(ip6s_tooshort);
870 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
871 			goto bad;
872 		}
873 
874 #ifdef IPSEC
875 		/*
876 		 * enforce IPsec policy checking if we are seeing last header.
877 		 * note that we do not visit this with protocols with pcb layer
878 		 * code - like udp/tcp/raw ip.
879 		 */
880 		if (ip6_ipsec_input(m, nxt))
881 			goto bad;
882 #endif /* IPSEC */
883 
884 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
885 	}
886 	return;
887 bad:
888 	m_freem(m);
889 }
890 
891 /*
892  * Hop-by-Hop options header processing. If a valid jumbo payload option is
893  * included, the real payload length will be stored in plenp.
894  *
895  * rtalertp - XXX: should be stored more smart way
896  */
897 static int
898 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
899     struct mbuf **mp, int *offp)
900 {
901 	struct mbuf *m = *mp;
902 	int off = *offp, hbhlen;
903 	struct ip6_hbh *hbh;
904 
905 	/* validation of the length of the header */
906 #ifndef PULLDOWN_TEST
907 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
908 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
909 	hbhlen = (hbh->ip6h_len + 1) << 3;
910 
911 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
912 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
913 #else
914 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
915 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
916 	if (hbh == NULL) {
917 		IP6STAT_INC(ip6s_tooshort);
918 		return -1;
919 	}
920 	hbhlen = (hbh->ip6h_len + 1) << 3;
921 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
922 		hbhlen);
923 	if (hbh == NULL) {
924 		IP6STAT_INC(ip6s_tooshort);
925 		return -1;
926 	}
927 #endif
928 	off += hbhlen;
929 	hbhlen -= sizeof(struct ip6_hbh);
930 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
931 				hbhlen, rtalertp, plenp) < 0)
932 		return (-1);
933 
934 	*offp = off;
935 	*mp = m;
936 	return (0);
937 }
938 
939 /*
940  * Search header for all Hop-by-hop options and process each option.
941  * This function is separate from ip6_hopopts_input() in order to
942  * handle a case where the sending node itself process its hop-by-hop
943  * options header. In such a case, the function is called from ip6_output().
944  *
945  * The function assumes that hbh header is located right after the IPv6 header
946  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
947  * opthead + hbhlen is located in contiguous memory region.
948  */
949 int
950 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
951     u_int32_t *rtalertp, u_int32_t *plenp)
952 {
953 	struct ip6_hdr *ip6;
954 	int optlen = 0;
955 	u_int8_t *opt = opthead;
956 	u_int16_t rtalert_val;
957 	u_int32_t jumboplen;
958 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
959 
960 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
961 		switch (*opt) {
962 		case IP6OPT_PAD1:
963 			optlen = 1;
964 			break;
965 		case IP6OPT_PADN:
966 			if (hbhlen < IP6OPT_MINLEN) {
967 				IP6STAT_INC(ip6s_toosmall);
968 				goto bad;
969 			}
970 			optlen = *(opt + 1) + 2;
971 			break;
972 		case IP6OPT_ROUTER_ALERT:
973 			/* XXX may need check for alignment */
974 			if (hbhlen < IP6OPT_RTALERT_LEN) {
975 				IP6STAT_INC(ip6s_toosmall);
976 				goto bad;
977 			}
978 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
979 				/* XXX stat */
980 				icmp6_error(m, ICMP6_PARAM_PROB,
981 				    ICMP6_PARAMPROB_HEADER,
982 				    erroff + opt + 1 - opthead);
983 				return (-1);
984 			}
985 			optlen = IP6OPT_RTALERT_LEN;
986 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
987 			*rtalertp = ntohs(rtalert_val);
988 			break;
989 		case IP6OPT_JUMBO:
990 			/* XXX may need check for alignment */
991 			if (hbhlen < IP6OPT_JUMBO_LEN) {
992 				IP6STAT_INC(ip6s_toosmall);
993 				goto bad;
994 			}
995 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
996 				/* XXX stat */
997 				icmp6_error(m, ICMP6_PARAM_PROB,
998 				    ICMP6_PARAMPROB_HEADER,
999 				    erroff + opt + 1 - opthead);
1000 				return (-1);
1001 			}
1002 			optlen = IP6OPT_JUMBO_LEN;
1003 
1004 			/*
1005 			 * IPv6 packets that have non 0 payload length
1006 			 * must not contain a jumbo payload option.
1007 			 */
1008 			ip6 = mtod(m, struct ip6_hdr *);
1009 			if (ip6->ip6_plen) {
1010 				IP6STAT_INC(ip6s_badoptions);
1011 				icmp6_error(m, ICMP6_PARAM_PROB,
1012 				    ICMP6_PARAMPROB_HEADER,
1013 				    erroff + opt - opthead);
1014 				return (-1);
1015 			}
1016 
1017 			/*
1018 			 * We may see jumbolen in unaligned location, so
1019 			 * we'd need to perform bcopy().
1020 			 */
1021 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1022 			jumboplen = (u_int32_t)htonl(jumboplen);
1023 
1024 #if 1
1025 			/*
1026 			 * if there are multiple jumbo payload options,
1027 			 * *plenp will be non-zero and the packet will be
1028 			 * rejected.
1029 			 * the behavior may need some debate in ipngwg -
1030 			 * multiple options does not make sense, however,
1031 			 * there's no explicit mention in specification.
1032 			 */
1033 			if (*plenp != 0) {
1034 				IP6STAT_INC(ip6s_badoptions);
1035 				icmp6_error(m, ICMP6_PARAM_PROB,
1036 				    ICMP6_PARAMPROB_HEADER,
1037 				    erroff + opt + 2 - opthead);
1038 				return (-1);
1039 			}
1040 #endif
1041 
1042 			/*
1043 			 * jumbo payload length must be larger than 65535.
1044 			 */
1045 			if (jumboplen <= IPV6_MAXPACKET) {
1046 				IP6STAT_INC(ip6s_badoptions);
1047 				icmp6_error(m, ICMP6_PARAM_PROB,
1048 				    ICMP6_PARAMPROB_HEADER,
1049 				    erroff + opt + 2 - opthead);
1050 				return (-1);
1051 			}
1052 			*plenp = jumboplen;
1053 
1054 			break;
1055 		default:		/* unknown option */
1056 			if (hbhlen < IP6OPT_MINLEN) {
1057 				IP6STAT_INC(ip6s_toosmall);
1058 				goto bad;
1059 			}
1060 			optlen = ip6_unknown_opt(opt, m,
1061 			    erroff + opt - opthead);
1062 			if (optlen == -1)
1063 				return (-1);
1064 			optlen += 2;
1065 			break;
1066 		}
1067 	}
1068 
1069 	return (0);
1070 
1071   bad:
1072 	m_freem(m);
1073 	return (-1);
1074 }
1075 
1076 /*
1077  * Unknown option processing.
1078  * The third argument `off' is the offset from the IPv6 header to the option,
1079  * which is necessary if the IPv6 header the and option header and IPv6 header
1080  * is not contiguous in order to return an ICMPv6 error.
1081  */
1082 int
1083 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1084 {
1085 	struct ip6_hdr *ip6;
1086 
1087 	switch (IP6OPT_TYPE(*optp)) {
1088 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1089 		return ((int)*(optp + 1));
1090 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1091 		m_freem(m);
1092 		return (-1);
1093 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1094 		IP6STAT_INC(ip6s_badoptions);
1095 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1096 		return (-1);
1097 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1098 		IP6STAT_INC(ip6s_badoptions);
1099 		ip6 = mtod(m, struct ip6_hdr *);
1100 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1101 		    (m->m_flags & (M_BCAST|M_MCAST)))
1102 			m_freem(m);
1103 		else
1104 			icmp6_error(m, ICMP6_PARAM_PROB,
1105 				    ICMP6_PARAMPROB_OPTION, off);
1106 		return (-1);
1107 	}
1108 
1109 	m_freem(m);		/* XXX: NOTREACHED */
1110 	return (-1);
1111 }
1112 
1113 /*
1114  * Create the "control" list for this pcb.
1115  * These functions will not modify mbuf chain at all.
1116  *
1117  * With KAME mbuf chain restriction:
1118  * The routine will be called from upper layer handlers like tcp6_input().
1119  * Thus the routine assumes that the caller (tcp6_input) have already
1120  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1121  * very first mbuf on the mbuf chain.
1122  *
1123  * ip6_savecontrol_v4 will handle those options that are possible to be
1124  * set on a v4-mapped socket.
1125  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1126  * options and handle the v6-only ones itself.
1127  */
1128 struct mbuf **
1129 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1130     int *v4only)
1131 {
1132 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1133 
1134 #ifdef SO_TIMESTAMP
1135 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1136 		struct timeval tv;
1137 
1138 		microtime(&tv);
1139 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1140 		    SCM_TIMESTAMP, SOL_SOCKET);
1141 		if (*mp)
1142 			mp = &(*mp)->m_next;
1143 	}
1144 #endif
1145 
1146 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1147 	/* RFC 2292 sec. 5 */
1148 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1149 		struct in6_pktinfo pi6;
1150 
1151 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1152 #ifdef INET
1153 			struct ip *ip;
1154 
1155 			ip = mtod(m, struct ip *);
1156 			pi6.ipi6_addr.s6_addr32[0] = 0;
1157 			pi6.ipi6_addr.s6_addr32[1] = 0;
1158 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1159 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1160 #else
1161 			/* We won't hit this code */
1162 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1163 #endif
1164 		} else {
1165 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1166 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1167 		}
1168 		pi6.ipi6_ifindex =
1169 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1170 
1171 		*mp = sbcreatecontrol((caddr_t) &pi6,
1172 		    sizeof(struct in6_pktinfo),
1173 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1174 		if (*mp)
1175 			mp = &(*mp)->m_next;
1176 	}
1177 
1178 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1179 		int hlim;
1180 
1181 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1182 #ifdef INET
1183 			struct ip *ip;
1184 
1185 			ip = mtod(m, struct ip *);
1186 			hlim = ip->ip_ttl;
1187 #else
1188 			/* We won't hit this code */
1189 			hlim = 0;
1190 #endif
1191 		} else {
1192 			hlim = ip6->ip6_hlim & 0xff;
1193 		}
1194 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1195 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1196 		    IPPROTO_IPV6);
1197 		if (*mp)
1198 			mp = &(*mp)->m_next;
1199 	}
1200 
1201 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1202 		int tclass;
1203 
1204 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1205 #ifdef INET
1206 			struct ip *ip;
1207 
1208 			ip = mtod(m, struct ip *);
1209 			tclass = ip->ip_tos;
1210 #else
1211 			/* We won't hit this code */
1212 			tclass = 0;
1213 #endif
1214 		} else {
1215 			u_int32_t flowinfo;
1216 
1217 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1218 			flowinfo >>= 20;
1219 			tclass = flowinfo & 0xff;
1220 		}
1221 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1222 		    IPV6_TCLASS, IPPROTO_IPV6);
1223 		if (*mp)
1224 			mp = &(*mp)->m_next;
1225 	}
1226 
1227 	if (v4only != NULL) {
1228 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1229 			*v4only = 1;
1230 		} else {
1231 			*v4only = 0;
1232 		}
1233 	}
1234 
1235 	return (mp);
1236 }
1237 
1238 void
1239 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1240 {
1241 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1242 	int v4only = 0;
1243 
1244 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1245 	if (v4only)
1246 		return;
1247 
1248 	/*
1249 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1250 	 * privilege for the option (see ip6_ctloutput), but it might be too
1251 	 * strict, since there might be some hop-by-hop options which can be
1252 	 * returned to normal user.
1253 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1254 	 */
1255 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1256 		/*
1257 		 * Check if a hop-by-hop options header is contatined in the
1258 		 * received packet, and if so, store the options as ancillary
1259 		 * data. Note that a hop-by-hop options header must be
1260 		 * just after the IPv6 header, which is assured through the
1261 		 * IPv6 input processing.
1262 		 */
1263 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1264 			struct ip6_hbh *hbh;
1265 			int hbhlen = 0;
1266 #ifdef PULLDOWN_TEST
1267 			struct mbuf *ext;
1268 #endif
1269 
1270 #ifndef PULLDOWN_TEST
1271 			hbh = (struct ip6_hbh *)(ip6 + 1);
1272 			hbhlen = (hbh->ip6h_len + 1) << 3;
1273 #else
1274 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1275 			    ip6->ip6_nxt);
1276 			if (ext == NULL) {
1277 				IP6STAT_INC(ip6s_tooshort);
1278 				return;
1279 			}
1280 			hbh = mtod(ext, struct ip6_hbh *);
1281 			hbhlen = (hbh->ip6h_len + 1) << 3;
1282 			if (hbhlen != ext->m_len) {
1283 				m_freem(ext);
1284 				IP6STAT_INC(ip6s_tooshort);
1285 				return;
1286 			}
1287 #endif
1288 
1289 			/*
1290 			 * XXX: We copy the whole header even if a
1291 			 * jumbo payload option is included, the option which
1292 			 * is to be removed before returning according to
1293 			 * RFC2292.
1294 			 * Note: this constraint is removed in RFC3542
1295 			 */
1296 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1297 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1298 			    IPPROTO_IPV6);
1299 			if (*mp)
1300 				mp = &(*mp)->m_next;
1301 #ifdef PULLDOWN_TEST
1302 			m_freem(ext);
1303 #endif
1304 		}
1305 	}
1306 
1307 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1308 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1309 
1310 		/*
1311 		 * Search for destination options headers or routing
1312 		 * header(s) through the header chain, and stores each
1313 		 * header as ancillary data.
1314 		 * Note that the order of the headers remains in
1315 		 * the chain of ancillary data.
1316 		 */
1317 		while (1) {	/* is explicit loop prevention necessary? */
1318 			struct ip6_ext *ip6e = NULL;
1319 			int elen;
1320 #ifdef PULLDOWN_TEST
1321 			struct mbuf *ext = NULL;
1322 #endif
1323 
1324 			/*
1325 			 * if it is not an extension header, don't try to
1326 			 * pull it from the chain.
1327 			 */
1328 			switch (nxt) {
1329 			case IPPROTO_DSTOPTS:
1330 			case IPPROTO_ROUTING:
1331 			case IPPROTO_HOPOPTS:
1332 			case IPPROTO_AH: /* is it possible? */
1333 				break;
1334 			default:
1335 				goto loopend;
1336 			}
1337 
1338 #ifndef PULLDOWN_TEST
1339 			if (off + sizeof(*ip6e) > m->m_len)
1340 				goto loopend;
1341 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1342 			if (nxt == IPPROTO_AH)
1343 				elen = (ip6e->ip6e_len + 2) << 2;
1344 			else
1345 				elen = (ip6e->ip6e_len + 1) << 3;
1346 			if (off + elen > m->m_len)
1347 				goto loopend;
1348 #else
1349 			ext = ip6_pullexthdr(m, off, nxt);
1350 			if (ext == NULL) {
1351 				IP6STAT_INC(ip6s_tooshort);
1352 				return;
1353 			}
1354 			ip6e = mtod(ext, struct ip6_ext *);
1355 			if (nxt == IPPROTO_AH)
1356 				elen = (ip6e->ip6e_len + 2) << 2;
1357 			else
1358 				elen = (ip6e->ip6e_len + 1) << 3;
1359 			if (elen != ext->m_len) {
1360 				m_freem(ext);
1361 				IP6STAT_INC(ip6s_tooshort);
1362 				return;
1363 			}
1364 #endif
1365 
1366 			switch (nxt) {
1367 			case IPPROTO_DSTOPTS:
1368 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1369 					break;
1370 
1371 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1372 				    IS2292(in6p,
1373 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1374 				    IPPROTO_IPV6);
1375 				if (*mp)
1376 					mp = &(*mp)->m_next;
1377 				break;
1378 			case IPPROTO_ROUTING:
1379 				if (!(in6p->inp_flags & IN6P_RTHDR))
1380 					break;
1381 
1382 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1383 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1384 				    IPPROTO_IPV6);
1385 				if (*mp)
1386 					mp = &(*mp)->m_next;
1387 				break;
1388 			case IPPROTO_HOPOPTS:
1389 			case IPPROTO_AH: /* is it possible? */
1390 				break;
1391 
1392 			default:
1393 				/*
1394 				 * other cases have been filtered in the above.
1395 				 * none will visit this case.  here we supply
1396 				 * the code just in case (nxt overwritten or
1397 				 * other cases).
1398 				 */
1399 #ifdef PULLDOWN_TEST
1400 				m_freem(ext);
1401 #endif
1402 				goto loopend;
1403 
1404 			}
1405 
1406 			/* proceed with the next header. */
1407 			off += elen;
1408 			nxt = ip6e->ip6e_nxt;
1409 			ip6e = NULL;
1410 #ifdef PULLDOWN_TEST
1411 			m_freem(ext);
1412 			ext = NULL;
1413 #endif
1414 		}
1415 	  loopend:
1416 		;
1417 	}
1418 
1419 	if (in6p->inp_flags2 & INP_RECVFLOWID) {
1420 		uint32_t flowid, flow_type;
1421 
1422 		flowid = m->m_pkthdr.flowid;
1423 		flow_type = M_HASHTYPE_GET(m);
1424 
1425 		/*
1426 		 * XXX should handle the failure of one or the
1427 		 * other - don't populate both?
1428 		 */
1429 		*mp = sbcreatecontrol((caddr_t) &flowid,
1430 		    sizeof(uint32_t), IPV6_FLOWID, IPPROTO_IPV6);
1431 		if (*mp)
1432 			mp = &(*mp)->m_next;
1433 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1434 		    sizeof(uint32_t), IPV6_FLOWTYPE, IPPROTO_IPV6);
1435 		if (*mp)
1436 			mp = &(*mp)->m_next;
1437 	}
1438 
1439 #ifdef	RSS
1440 	if (in6p->inp_flags2 & INP_RECVRSSBUCKETID) {
1441 		uint32_t flowid, flow_type;
1442 		uint32_t rss_bucketid;
1443 
1444 		flowid = m->m_pkthdr.flowid;
1445 		flow_type = M_HASHTYPE_GET(m);
1446 
1447 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1448 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1449 			   sizeof(uint32_t), IPV6_RSSBUCKETID, IPPROTO_IPV6);
1450 			if (*mp)
1451 				mp = &(*mp)->m_next;
1452 		}
1453 	}
1454 #endif
1455 
1456 }
1457 #undef IS2292
1458 
1459 void
1460 ip6_notify_pmtu(struct inpcb *inp, struct sockaddr_in6 *dst, u_int32_t mtu)
1461 {
1462 	struct socket *so;
1463 	struct mbuf *m_mtu;
1464 	struct ip6_mtuinfo mtuctl;
1465 
1466 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1467 	/*
1468 	 * Notify the error by sending IPV6_PATHMTU ancillary data if
1469 	 * application wanted to know the MTU value.
1470 	 * NOTE: we notify disconnected sockets, because some udp
1471 	 * applications keep sending sockets disconnected.
1472 	 * NOTE: our implementation doesn't notify connected sockets that has
1473 	 * foreign address that is different than given destination addresses
1474 	 * (this is permitted by RFC 3542).
1475 	 */
1476 	if ((inp->inp_flags & IN6P_MTU) == 0 || (
1477 	    !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1478 	    !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &dst->sin6_addr)))
1479 		return;
1480 
1481 	mtuctl.ip6m_mtu = mtu;
1482 	mtuctl.ip6m_addr = *dst;
1483 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1484 		return;
1485 
1486 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1487 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1488 		return;
1489 
1490 	so =  inp->inp_socket;
1491 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1492 	    == 0) {
1493 		m_freem(m_mtu);
1494 		/* XXX: should count statistics */
1495 	} else
1496 		sorwakeup(so);
1497 }
1498 
1499 #ifdef PULLDOWN_TEST
1500 /*
1501  * pull single extension header from mbuf chain.  returns single mbuf that
1502  * contains the result, or NULL on error.
1503  */
1504 static struct mbuf *
1505 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1506 {
1507 	struct ip6_ext ip6e;
1508 	size_t elen;
1509 	struct mbuf *n;
1510 
1511 #ifdef DIAGNOSTIC
1512 	switch (nxt) {
1513 	case IPPROTO_DSTOPTS:
1514 	case IPPROTO_ROUTING:
1515 	case IPPROTO_HOPOPTS:
1516 	case IPPROTO_AH: /* is it possible? */
1517 		break;
1518 	default:
1519 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1520 	}
1521 #endif
1522 
1523 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1524 	if (nxt == IPPROTO_AH)
1525 		elen = (ip6e.ip6e_len + 2) << 2;
1526 	else
1527 		elen = (ip6e.ip6e_len + 1) << 3;
1528 
1529 	if (elen > MLEN)
1530 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1531 	else
1532 		n = m_get(M_NOWAIT, MT_DATA);
1533 	if (n == NULL)
1534 		return NULL;
1535 
1536 	m_copydata(m, off, elen, mtod(n, caddr_t));
1537 	n->m_len = elen;
1538 	return n;
1539 }
1540 #endif
1541 
1542 /*
1543  * Get pointer to the previous header followed by the header
1544  * currently processed.
1545  * XXX: This function supposes that
1546  *	M includes all headers,
1547  *	the next header field and the header length field of each header
1548  *	are valid, and
1549  *	the sum of each header length equals to OFF.
1550  * Because of these assumptions, this function must be called very
1551  * carefully. Moreover, it will not be used in the near future when
1552  * we develop `neater' mechanism to process extension headers.
1553  */
1554 char *
1555 ip6_get_prevhdr(const struct mbuf *m, int off)
1556 {
1557 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1558 
1559 	if (off == sizeof(struct ip6_hdr))
1560 		return (&ip6->ip6_nxt);
1561 	else {
1562 		int len, nxt;
1563 		struct ip6_ext *ip6e = NULL;
1564 
1565 		nxt = ip6->ip6_nxt;
1566 		len = sizeof(struct ip6_hdr);
1567 		while (len < off) {
1568 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1569 
1570 			switch (nxt) {
1571 			case IPPROTO_FRAGMENT:
1572 				len += sizeof(struct ip6_frag);
1573 				break;
1574 			case IPPROTO_AH:
1575 				len += (ip6e->ip6e_len + 2) << 2;
1576 				break;
1577 			default:
1578 				len += (ip6e->ip6e_len + 1) << 3;
1579 				break;
1580 			}
1581 			nxt = ip6e->ip6e_nxt;
1582 		}
1583 		if (ip6e)
1584 			return (&ip6e->ip6e_nxt);
1585 		else
1586 			return NULL;
1587 	}
1588 }
1589 
1590 /*
1591  * get next header offset.  m will be retained.
1592  */
1593 int
1594 ip6_nexthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1595 {
1596 	struct ip6_hdr ip6;
1597 	struct ip6_ext ip6e;
1598 	struct ip6_frag fh;
1599 
1600 	/* just in case */
1601 	if (m == NULL)
1602 		panic("ip6_nexthdr: m == NULL");
1603 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1604 		return -1;
1605 
1606 	switch (proto) {
1607 	case IPPROTO_IPV6:
1608 		if (m->m_pkthdr.len < off + sizeof(ip6))
1609 			return -1;
1610 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1611 		if (nxtp)
1612 			*nxtp = ip6.ip6_nxt;
1613 		off += sizeof(ip6);
1614 		return off;
1615 
1616 	case IPPROTO_FRAGMENT:
1617 		/*
1618 		 * terminate parsing if it is not the first fragment,
1619 		 * it does not make sense to parse through it.
1620 		 */
1621 		if (m->m_pkthdr.len < off + sizeof(fh))
1622 			return -1;
1623 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1624 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1625 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1626 			return -1;
1627 		if (nxtp)
1628 			*nxtp = fh.ip6f_nxt;
1629 		off += sizeof(struct ip6_frag);
1630 		return off;
1631 
1632 	case IPPROTO_AH:
1633 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1634 			return -1;
1635 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1636 		if (nxtp)
1637 			*nxtp = ip6e.ip6e_nxt;
1638 		off += (ip6e.ip6e_len + 2) << 2;
1639 		return off;
1640 
1641 	case IPPROTO_HOPOPTS:
1642 	case IPPROTO_ROUTING:
1643 	case IPPROTO_DSTOPTS:
1644 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1645 			return -1;
1646 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1647 		if (nxtp)
1648 			*nxtp = ip6e.ip6e_nxt;
1649 		off += (ip6e.ip6e_len + 1) << 3;
1650 		return off;
1651 
1652 	case IPPROTO_NONE:
1653 	case IPPROTO_ESP:
1654 	case IPPROTO_IPCOMP:
1655 		/* give up */
1656 		return -1;
1657 
1658 	default:
1659 		return -1;
1660 	}
1661 
1662 	/* NOTREACHED */
1663 }
1664 
1665 /*
1666  * get offset for the last header in the chain.  m will be kept untainted.
1667  */
1668 int
1669 ip6_lasthdr(const struct mbuf *m, int off, int proto, int *nxtp)
1670 {
1671 	int newoff;
1672 	int nxt;
1673 
1674 	if (!nxtp) {
1675 		nxt = -1;
1676 		nxtp = &nxt;
1677 	}
1678 	while (1) {
1679 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1680 		if (newoff < 0)
1681 			return off;
1682 		else if (newoff < off)
1683 			return -1;	/* invalid */
1684 		else if (newoff == off)
1685 			return newoff;
1686 
1687 		off = newoff;
1688 		proto = *nxtp;
1689 	}
1690 }
1691 
1692 /*
1693  * System control for IP6
1694  */
1695 
1696 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1697 	0,		0,		0,		0,
1698 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1699 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1700 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1701 	0,		0,		0,		0,
1702 	ENOPROTOOPT
1703 };
1704