xref: /freebsd/sys/netinet6/icmp6.c (revision 94942af266ac119ede0ca836f9aa5a5ac0582938)
1 /*	$FreeBSD$	*/
2 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*-
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/domain.h>
70 #include <sys/kernel.h>
71 #include <sys/lock.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/signalvar.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sx.h>
79 #include <sys/syslog.h>
80 #include <sys/systm.h>
81 #include <sys/time.h>
82 
83 #include <net/if.h>
84 #include <net/if_dl.h>
85 #include <net/if_types.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_pcb.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/tcp_var.h>
94 #include <netinet6/in6_ifattach.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/ip6protosw.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/scope6_var.h>
99 #include <netinet6/mld6_var.h>
100 #include <netinet6/nd6.h>
101 
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netkey/key.h>
105 #endif
106 
107 #ifdef FAST_IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/key.h>
110 #endif
111 
112 extern struct domain inet6domain;
113 
114 struct icmp6stat icmp6stat;
115 
116 extern struct inpcbinfo ripcbinfo;
117 extern struct inpcbhead ripcb;
118 extern int icmp6errppslim;
119 static int icmp6errpps_count = 0;
120 static struct timeval icmp6errppslim_last;
121 extern int icmp6_nodeinfo;
122 
123 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
124 static int icmp6_rip6_input __P((struct mbuf **, int));
125 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
126 static const char *icmp6_redirect_diag __P((struct in6_addr *,
127 	struct in6_addr *, struct in6_addr *));
128 static struct mbuf *ni6_input __P((struct mbuf *, int));
129 static struct mbuf *ni6_nametodns __P((const char *, int, int));
130 static int ni6_dnsmatch __P((const char *, int, const char *, int));
131 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
132 			  struct ifnet **, struct in6_addr *));
133 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
134 				struct ifnet *, int));
135 static int icmp6_notify_error __P((struct mbuf **, int, int, int));
136 
137 
138 void
139 icmp6_init()
140 {
141 	mld6_init();
142 }
143 
144 static void
145 icmp6_errcount(stat, type, code)
146 	struct icmp6errstat *stat;
147 	int type, code;
148 {
149 	switch (type) {
150 	case ICMP6_DST_UNREACH:
151 		switch (code) {
152 		case ICMP6_DST_UNREACH_NOROUTE:
153 			stat->icp6errs_dst_unreach_noroute++;
154 			return;
155 		case ICMP6_DST_UNREACH_ADMIN:
156 			stat->icp6errs_dst_unreach_admin++;
157 			return;
158 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
159 			stat->icp6errs_dst_unreach_beyondscope++;
160 			return;
161 		case ICMP6_DST_UNREACH_ADDR:
162 			stat->icp6errs_dst_unreach_addr++;
163 			return;
164 		case ICMP6_DST_UNREACH_NOPORT:
165 			stat->icp6errs_dst_unreach_noport++;
166 			return;
167 		}
168 		break;
169 	case ICMP6_PACKET_TOO_BIG:
170 		stat->icp6errs_packet_too_big++;
171 		return;
172 	case ICMP6_TIME_EXCEEDED:
173 		switch (code) {
174 		case ICMP6_TIME_EXCEED_TRANSIT:
175 			stat->icp6errs_time_exceed_transit++;
176 			return;
177 		case ICMP6_TIME_EXCEED_REASSEMBLY:
178 			stat->icp6errs_time_exceed_reassembly++;
179 			return;
180 		}
181 		break;
182 	case ICMP6_PARAM_PROB:
183 		switch (code) {
184 		case ICMP6_PARAMPROB_HEADER:
185 			stat->icp6errs_paramprob_header++;
186 			return;
187 		case ICMP6_PARAMPROB_NEXTHEADER:
188 			stat->icp6errs_paramprob_nextheader++;
189 			return;
190 		case ICMP6_PARAMPROB_OPTION:
191 			stat->icp6errs_paramprob_option++;
192 			return;
193 		}
194 		break;
195 	case ND_REDIRECT:
196 		stat->icp6errs_redirect++;
197 		return;
198 	}
199 	stat->icp6errs_unknown++;
200 }
201 
202 /*
203  * A wrapper function for icmp6_error() necessary when the erroneous packet
204  * may not contain enough scope zone information.
205  */
206 void
207 icmp6_error2(m, type, code, param, ifp)
208 	struct mbuf *m;
209 	int type, code, param;
210 	struct ifnet *ifp;
211 {
212 	struct ip6_hdr *ip6;
213 
214 	if (ifp == NULL)
215 		return;
216 
217 #ifndef PULLDOWN_TEST
218 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
219 #else
220 	if (m->m_len < sizeof(struct ip6_hdr)) {
221 		m = m_pullup(m, sizeof(struct ip6_hdr));
222 		if (m == NULL)
223 			return;
224 	}
225 #endif
226 
227 	ip6 = mtod(m, struct ip6_hdr *);
228 
229 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
230 		return;
231 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
232 		return;
233 
234 	icmp6_error(m, type, code, param);
235 }
236 
237 /*
238  * Generate an error packet of type error in response to bad IP6 packet.
239  */
240 void
241 icmp6_error(m, type, code, param)
242 	struct mbuf *m;
243 	int type, code, param;
244 {
245 	struct ip6_hdr *oip6, *nip6;
246 	struct icmp6_hdr *icmp6;
247 	u_int preplen;
248 	int off;
249 	int nxt;
250 
251 	icmp6stat.icp6s_error++;
252 
253 	/* count per-type-code statistics */
254 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
255 
256 #ifdef M_DECRYPTED	/*not openbsd*/
257 	if (m->m_flags & M_DECRYPTED) {
258 		icmp6stat.icp6s_canterror++;
259 		goto freeit;
260 	}
261 #endif
262 
263 #ifndef PULLDOWN_TEST
264 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
265 #else
266 	if (m->m_len < sizeof(struct ip6_hdr)) {
267 		m = m_pullup(m, sizeof(struct ip6_hdr));
268 		if (m == NULL)
269 			return;
270 	}
271 #endif
272 	oip6 = mtod(m, struct ip6_hdr *);
273 
274 	/*
275 	 * If the destination address of the erroneous packet is a multicast
276 	 * address, or the packet was sent using link-layer multicast,
277 	 * we should basically suppress sending an error (RFC 2463, Section
278 	 * 2.4).
279 	 * We have two exceptions (the item e.2 in that section):
280 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
281 	 * - the Parameter Problem Message that can be allowed an icmp6 error
282 	 *   in the option type field.  This check has been done in
283 	 *   ip6_unknown_opt(), so we can just check the type and code.
284 	 */
285 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
286 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
287 	    (type != ICMP6_PACKET_TOO_BIG &&
288 	     (type != ICMP6_PARAM_PROB ||
289 	      code != ICMP6_PARAMPROB_OPTION)))
290 		goto freeit;
291 
292 	/*
293 	 * RFC 2463, 2.4 (e.5): source address check.
294 	 * XXX: the case of anycast source?
295 	 */
296 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
297 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
298 		goto freeit;
299 
300 	/*
301 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
302 	 * don't do it.
303 	 */
304 	nxt = -1;
305 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
306 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
307 		struct icmp6_hdr *icp;
308 
309 #ifndef PULLDOWN_TEST
310 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
311 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
312 #else
313 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
314 			sizeof(*icp));
315 		if (icp == NULL) {
316 			icmp6stat.icp6s_tooshort++;
317 			return;
318 		}
319 #endif
320 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
321 		    icp->icmp6_type == ND_REDIRECT) {
322 			/*
323 			 * ICMPv6 error
324 			 * Special case: for redirect (which is
325 			 * informational) we must not send icmp6 error.
326 			 */
327 			icmp6stat.icp6s_canterror++;
328 			goto freeit;
329 		} else {
330 			/* ICMPv6 informational - send the error */
331 		}
332 	} else {
333 		/* non-ICMPv6 - send the error */
334 	}
335 
336 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
337 
338 	/* Finally, do rate limitation check. */
339 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
340 		icmp6stat.icp6s_toofreq++;
341 		goto freeit;
342 	}
343 
344 	/*
345 	 * OK, ICMP6 can be generated.
346 	 */
347 
348 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
349 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
350 
351 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
352 	M_PREPEND(m, preplen, M_DONTWAIT);
353 	if (m && m->m_len < preplen)
354 		m = m_pullup(m, preplen);
355 	if (m == NULL) {
356 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
357 		return;
358 	}
359 
360 	nip6 = mtod(m, struct ip6_hdr *);
361 	nip6->ip6_src  = oip6->ip6_src;
362 	nip6->ip6_dst  = oip6->ip6_dst;
363 
364 	in6_clearscope(&oip6->ip6_src);
365 	in6_clearscope(&oip6->ip6_dst);
366 
367 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
368 	icmp6->icmp6_type = type;
369 	icmp6->icmp6_code = code;
370 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
371 
372 	/*
373 	 * icmp6_reflect() is designed to be in the input path.
374 	 * icmp6_error() can be called from both input and output path,
375 	 * and if we are in output path rcvif could contain bogus value.
376 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
377 	 * information in ip header (nip6).
378 	 */
379 	m->m_pkthdr.rcvif = NULL;
380 
381 	icmp6stat.icp6s_outhist[type]++;
382 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
383 
384 	return;
385 
386   freeit:
387 	/*
388 	 * If we can't tell whether or not we can generate ICMP6, free it.
389 	 */
390 	m_freem(m);
391 }
392 
393 /*
394  * Process a received ICMP6 message.
395  */
396 int
397 icmp6_input(mp, offp, proto)
398 	struct mbuf **mp;
399 	int *offp, proto;
400 {
401 	struct mbuf *m = *mp, *n;
402 	struct ip6_hdr *ip6, *nip6;
403 	struct icmp6_hdr *icmp6, *nicmp6;
404 	int off = *offp;
405 	int icmp6len = m->m_pkthdr.len - *offp;
406 	int code, sum, noff;
407 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
408 
409 #ifndef PULLDOWN_TEST
410 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
411 	/* m might change if M_LOOP.  So, call mtod after this */
412 #endif
413 
414 	/*
415 	 * Locate icmp6 structure in mbuf, and check
416 	 * that not corrupted and of at least minimum length
417 	 */
418 
419 	ip6 = mtod(m, struct ip6_hdr *);
420 	if (icmp6len < sizeof(struct icmp6_hdr)) {
421 		icmp6stat.icp6s_tooshort++;
422 		goto freeit;
423 	}
424 
425 	/*
426 	 * calculate the checksum
427 	 */
428 #ifndef PULLDOWN_TEST
429 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
430 #else
431 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
432 	if (icmp6 == NULL) {
433 		icmp6stat.icp6s_tooshort++;
434 		return IPPROTO_DONE;
435 	}
436 #endif
437 	code = icmp6->icmp6_code;
438 
439 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
440 		nd6log((LOG_ERR,
441 		    "ICMP6 checksum error(%d|%x) %s\n",
442 		    icmp6->icmp6_type, sum,
443 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
444 		icmp6stat.icp6s_checksum++;
445 		goto freeit;
446 	}
447 
448 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
449 		/*
450 		 * Deliver very specific ICMP6 type only.
451 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
452 		 * will not work.
453 		 */
454 		switch (icmp6->icmp6_type) {
455 		case ICMP6_DST_UNREACH:
456 		case ICMP6_PACKET_TOO_BIG:
457 		case ICMP6_TIME_EXCEEDED:
458 			break;
459 		default:
460 			goto freeit;
461 		}
462 	}
463 
464 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
465 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
466 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
467 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
468 
469 	switch (icmp6->icmp6_type) {
470 	case ICMP6_DST_UNREACH:
471 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
472 		switch (code) {
473 		case ICMP6_DST_UNREACH_NOROUTE:
474 			code = PRC_UNREACH_NET;
475 			break;
476 		case ICMP6_DST_UNREACH_ADMIN:
477 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
478 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
479 			break;
480 		case ICMP6_DST_UNREACH_ADDR:
481 			code = PRC_HOSTDEAD;
482 			break;
483 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
484 			/* I mean "source address was incorrect." */
485 			code = PRC_PARAMPROB;
486 			break;
487 		case ICMP6_DST_UNREACH_NOPORT:
488 			code = PRC_UNREACH_PORT;
489 			break;
490 		default:
491 			goto badcode;
492 		}
493 		goto deliver;
494 		break;
495 
496 	case ICMP6_PACKET_TOO_BIG:
497 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
498 
499 		/* validation is made in icmp6_mtudisc_update */
500 
501 		code = PRC_MSGSIZE;
502 
503 		/*
504 		 * Updating the path MTU will be done after examining
505 		 * intermediate extension headers.
506 		 */
507 		goto deliver;
508 		break;
509 
510 	case ICMP6_TIME_EXCEEDED:
511 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
512 		switch (code) {
513 		case ICMP6_TIME_EXCEED_TRANSIT:
514 			code = PRC_TIMXCEED_INTRANS;
515 			break;
516 		case ICMP6_TIME_EXCEED_REASSEMBLY:
517 			code = PRC_TIMXCEED_REASS;
518 			break;
519 		default:
520 			goto badcode;
521 		}
522 		goto deliver;
523 		break;
524 
525 	case ICMP6_PARAM_PROB:
526 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
527 		switch (code) {
528 		case ICMP6_PARAMPROB_NEXTHEADER:
529 			code = PRC_UNREACH_PROTOCOL;
530 			break;
531 		case ICMP6_PARAMPROB_HEADER:
532 		case ICMP6_PARAMPROB_OPTION:
533 			code = PRC_PARAMPROB;
534 			break;
535 		default:
536 			goto badcode;
537 		}
538 		goto deliver;
539 		break;
540 
541 	case ICMP6_ECHO_REQUEST:
542 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
543 		if (code != 0)
544 			goto badcode;
545 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
546 			/* Give up remote */
547 			break;
548 		}
549 		if ((n->m_flags & M_EXT) != 0
550 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
551 			struct mbuf *n0 = n;
552 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
553 			int n0len;
554 
555 			MGETHDR(n, M_DONTWAIT, n0->m_type);
556 			n0len = n0->m_pkthdr.len;	/* save for use below */
557 			if (n)
558 				M_MOVE_PKTHDR(n, n0);
559 			if (n && maxlen >= MHLEN) {
560 				MCLGET(n, M_DONTWAIT);
561 				if ((n->m_flags & M_EXT) == 0) {
562 					m_free(n);
563 					n = NULL;
564 				}
565 			}
566 			if (n == NULL) {
567 				/* Give up remote */
568 				m_freem(n0);
569 				break;
570 			}
571 			/*
572 			 * Copy IPv6 and ICMPv6 only.
573 			 */
574 			nip6 = mtod(n, struct ip6_hdr *);
575 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
576 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
577 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
578 			noff = sizeof(struct ip6_hdr);
579 			/* new mbuf contains only ipv6+icmpv6 headers */
580 			n->m_len = noff + sizeof(struct icmp6_hdr);
581 			/*
582 			 * Adjust mbuf.  ip6_plen will be adjusted in
583 			 * ip6_output().
584 			 */
585 			m_adj(n0, off + sizeof(struct icmp6_hdr));
586 			/* recalculate complete packet size */
587 			n->m_pkthdr.len = n0len + (noff - off);
588 			n->m_next = n0;
589 		} else {
590 			nip6 = mtod(n, struct ip6_hdr *);
591 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
592 			    sizeof(*nicmp6));
593 			noff = off;
594 		}
595 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
596 		nicmp6->icmp6_code = 0;
597 		if (n) {
598 			icmp6stat.icp6s_reflect++;
599 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
600 			icmp6_reflect(n, noff);
601 		}
602 		break;
603 
604 	case ICMP6_ECHO_REPLY:
605 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
606 		if (code != 0)
607 			goto badcode;
608 		break;
609 
610 	case MLD_LISTENER_QUERY:
611 	case MLD_LISTENER_REPORT:
612 		if (icmp6len < sizeof(struct mld_hdr))
613 			goto badlen;
614 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
615 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
616 		else
617 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
618 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
619 			/* give up local */
620 			mld6_input(m, off);
621 			m = NULL;
622 			goto freeit;
623 		}
624 		mld6_input(n, off);
625 		/* m stays. */
626 		break;
627 
628 	case MLD_LISTENER_DONE:
629 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
630 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
631 			goto badlen;
632 		break;		/* nothing to be done in kernel */
633 
634 	case MLD_MTRACE_RESP:
635 	case MLD_MTRACE:
636 		/* XXX: these two are experimental.  not officially defined. */
637 		/* XXX: per-interface statistics? */
638 		break;		/* just pass it to applications */
639 
640 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
641 	    {
642 		enum { WRU, FQDN } mode;
643 
644 		if (!icmp6_nodeinfo)
645 			break;
646 
647 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
648 			mode = WRU;
649 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
650 			mode = FQDN;
651 		else
652 			goto badlen;
653 
654 #define hostnamelen	strlen(hostname)
655 		if (mode == FQDN) {
656 #ifndef PULLDOWN_TEST
657 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
658 			    IPPROTO_DONE);
659 #endif
660 			n = m_copy(m, 0, M_COPYALL);
661 			if (n)
662 				n = ni6_input(n, off);
663 			/* XXX meaningless if n == NULL */
664 			noff = sizeof(struct ip6_hdr);
665 		} else {
666 			u_char *p;
667 			int maxlen, maxhlen;
668 
669 			/*
670 			 * XXX: this combination of flags is pointless,
671 			 * but should we keep this for compatibility?
672 			 */
673 			if ((icmp6_nodeinfo & 5) != 5)
674 				break;
675 
676 			if (code != 0)
677 				goto badcode;
678 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
679 			if (maxlen >= MCLBYTES) {
680 				/* Give up remote */
681 				break;
682 			}
683 			MGETHDR(n, M_DONTWAIT, m->m_type);
684 			if (n && maxlen > MHLEN) {
685 				MCLGET(n, M_DONTWAIT);
686 				if ((n->m_flags & M_EXT) == 0) {
687 					m_free(n);
688 					n = NULL;
689 				}
690 			}
691 			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
692 				/*
693 				 * Previous code did a blind M_COPY_PKTHDR
694 				 * and said "just for rcvif".  If true, then
695 				 * we could tolerate the dup failing (due to
696 				 * the deep copy of the tag chain).  For now
697 				 * be conservative and just fail.
698 				 */
699 				m_free(n);
700 				n = NULL;
701 			}
702 			if (n == NULL) {
703 				/* Give up remote */
704 				break;
705 			}
706 			n->m_pkthdr.rcvif = NULL;
707 			n->m_len = 0;
708 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
709 			if (maxhlen > hostnamelen)
710 				maxhlen = hostnamelen;
711 			/*
712 			 * Copy IPv6 and ICMPv6 only.
713 			 */
714 			nip6 = mtod(n, struct ip6_hdr *);
715 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
716 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
717 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
718 			p = (u_char *)(nicmp6 + 1);
719 			bzero(p, 4);
720 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
721 			noff = sizeof(struct ip6_hdr);
722 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
723 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
724 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
725 			nicmp6->icmp6_code = 0;
726 		}
727 #undef hostnamelen
728 		if (n) {
729 			icmp6stat.icp6s_reflect++;
730 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
731 			icmp6_reflect(n, noff);
732 		}
733 		break;
734 	    }
735 
736 	case ICMP6_WRUREPLY:
737 		if (code != 0)
738 			goto badcode;
739 		break;
740 
741 	case ND_ROUTER_SOLICIT:
742 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
743 		if (code != 0)
744 			goto badcode;
745 		if (icmp6len < sizeof(struct nd_router_solicit))
746 			goto badlen;
747 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
748 			/* give up local */
749 			nd6_rs_input(m, off, icmp6len);
750 			m = NULL;
751 			goto freeit;
752 		}
753 		nd6_rs_input(n, off, icmp6len);
754 		/* m stays. */
755 		break;
756 
757 	case ND_ROUTER_ADVERT:
758 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
759 		if (code != 0)
760 			goto badcode;
761 		if (icmp6len < sizeof(struct nd_router_advert))
762 			goto badlen;
763 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
764 			/* give up local */
765 			nd6_ra_input(m, off, icmp6len);
766 			m = NULL;
767 			goto freeit;
768 		}
769 		nd6_ra_input(n, off, icmp6len);
770 		/* m stays. */
771 		break;
772 
773 	case ND_NEIGHBOR_SOLICIT:
774 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
775 		if (code != 0)
776 			goto badcode;
777 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
778 			goto badlen;
779 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
780 			/* give up local */
781 			nd6_ns_input(m, off, icmp6len);
782 			m = NULL;
783 			goto freeit;
784 		}
785 		nd6_ns_input(n, off, icmp6len);
786 		/* m stays. */
787 		break;
788 
789 	case ND_NEIGHBOR_ADVERT:
790 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
791 		if (code != 0)
792 			goto badcode;
793 		if (icmp6len < sizeof(struct nd_neighbor_advert))
794 			goto badlen;
795 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
796 			/* give up local */
797 			nd6_na_input(m, off, icmp6len);
798 			m = NULL;
799 			goto freeit;
800 		}
801 		nd6_na_input(n, off, icmp6len);
802 		/* m stays. */
803 		break;
804 
805 	case ND_REDIRECT:
806 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
807 		if (code != 0)
808 			goto badcode;
809 		if (icmp6len < sizeof(struct nd_redirect))
810 			goto badlen;
811 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
812 			/* give up local */
813 			icmp6_redirect_input(m, off);
814 			m = NULL;
815 			goto freeit;
816 		}
817 		icmp6_redirect_input(n, off);
818 		/* m stays. */
819 		break;
820 
821 	case ICMP6_ROUTER_RENUMBERING:
822 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
823 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
824 			goto badcode;
825 		if (icmp6len < sizeof(struct icmp6_router_renum))
826 			goto badlen;
827 		break;
828 
829 	default:
830 		nd6log((LOG_DEBUG,
831 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
832 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
833 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
834 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
835 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
836 			/* ICMPv6 error: MUST deliver it by spec... */
837 			code = PRC_NCMDS;
838 			/* deliver */
839 		} else {
840 			/* ICMPv6 informational: MUST not deliver */
841 			break;
842 		}
843 	deliver:
844 		if (icmp6_notify_error(&m, off, icmp6len, code)) {
845 			/* In this case, m should've been freed. */
846 			return (IPPROTO_DONE);
847 		}
848 		break;
849 
850 	badcode:
851 		icmp6stat.icp6s_badcode++;
852 		break;
853 
854 	badlen:
855 		icmp6stat.icp6s_badlen++;
856 		break;
857 	}
858 
859 	/* deliver the packet to appropriate sockets */
860 	icmp6_rip6_input(&m, *offp);
861 
862 	return IPPROTO_DONE;
863 
864  freeit:
865 	m_freem(m);
866 	return IPPROTO_DONE;
867 }
868 
869 static int
870 icmp6_notify_error(mp, off, icmp6len, code)
871 	struct mbuf **mp;
872 	int off, icmp6len, code;
873 {
874 	struct mbuf *m = *mp;
875 	struct icmp6_hdr *icmp6;
876 	struct ip6_hdr *eip6;
877 	u_int32_t notifymtu;
878 	struct sockaddr_in6 icmp6src, icmp6dst;
879 
880 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
881 		icmp6stat.icp6s_tooshort++;
882 		goto freeit;
883 	}
884 #ifndef PULLDOWN_TEST
885 	IP6_EXTHDR_CHECK(m, off,
886 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
887 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
888 #else
889 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
890 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
891 	if (icmp6 == NULL) {
892 		icmp6stat.icp6s_tooshort++;
893 		return (-1);
894 	}
895 #endif
896 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
897 
898 	/* Detect the upper level protocol */
899 	{
900 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
901 		u_int8_t nxt = eip6->ip6_nxt;
902 		int eoff = off + sizeof(struct icmp6_hdr) +
903 		    sizeof(struct ip6_hdr);
904 		struct ip6ctlparam ip6cp;
905 		struct in6_addr *finaldst = NULL;
906 		int icmp6type = icmp6->icmp6_type;
907 		struct ip6_frag *fh;
908 		struct ip6_rthdr *rth;
909 		struct ip6_rthdr0 *rth0;
910 		int rthlen;
911 
912 		while (1) { /* XXX: should avoid infinite loop explicitly? */
913 			struct ip6_ext *eh;
914 
915 			switch (nxt) {
916 			case IPPROTO_HOPOPTS:
917 			case IPPROTO_DSTOPTS:
918 			case IPPROTO_AH:
919 #ifndef PULLDOWN_TEST
920 				IP6_EXTHDR_CHECK(m, 0,
921 				    eoff + sizeof(struct ip6_ext), -1);
922 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
923 #else
924 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
925 				    eoff, sizeof(*eh));
926 				if (eh == NULL) {
927 					icmp6stat.icp6s_tooshort++;
928 					return (-1);
929 				}
930 #endif
931 
932 				if (nxt == IPPROTO_AH)
933 					eoff += (eh->ip6e_len + 2) << 2;
934 				else
935 					eoff += (eh->ip6e_len + 1) << 3;
936 				nxt = eh->ip6e_nxt;
937 				break;
938 			case IPPROTO_ROUTING:
939 				/*
940 				 * When the erroneous packet contains a
941 				 * routing header, we should examine the
942 				 * header to determine the final destination.
943 				 * Otherwise, we can't properly update
944 				 * information that depends on the final
945 				 * destination (e.g. path MTU).
946 				 */
947 #ifndef PULLDOWN_TEST
948 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
949 				rth = (struct ip6_rthdr *)
950 				    (mtod(m, caddr_t) + eoff);
951 #else
952 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
953 				    eoff, sizeof(*rth));
954 				if (rth == NULL) {
955 					icmp6stat.icp6s_tooshort++;
956 					return (-1);
957 				}
958 #endif
959 				rthlen = (rth->ip6r_len + 1) << 3;
960 				/*
961 				 * XXX: currently there is no
962 				 * officially defined type other
963 				 * than type-0.
964 				 * Note that if the segment left field
965 				 * is 0, all intermediate hops must
966 				 * have been passed.
967 				 */
968 				if (rth->ip6r_segleft &&
969 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
970 					int hops;
971 
972 #ifndef PULLDOWN_TEST
973 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
974 					rth0 = (struct ip6_rthdr0 *)
975 					    (mtod(m, caddr_t) + eoff);
976 #else
977 					IP6_EXTHDR_GET(rth0,
978 					    struct ip6_rthdr0 *, m,
979 					    eoff, rthlen);
980 					if (rth0 == NULL) {
981 						icmp6stat.icp6s_tooshort++;
982 						return (-1);
983 					}
984 #endif
985 					/* just ignore a bogus header */
986 					if ((rth0->ip6r0_len % 2) == 0 &&
987 					    (hops = rth0->ip6r0_len/2))
988 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
989 				}
990 				eoff += rthlen;
991 				nxt = rth->ip6r_nxt;
992 				break;
993 			case IPPROTO_FRAGMENT:
994 #ifndef PULLDOWN_TEST
995 				IP6_EXTHDR_CHECK(m, 0, eoff +
996 				    sizeof(struct ip6_frag), -1);
997 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
998 				    eoff);
999 #else
1000 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1001 				    eoff, sizeof(*fh));
1002 				if (fh == NULL) {
1003 					icmp6stat.icp6s_tooshort++;
1004 					return (-1);
1005 				}
1006 #endif
1007 				/*
1008 				 * Data after a fragment header is meaningless
1009 				 * unless it is the first fragment, but
1010 				 * we'll go to the notify label for path MTU
1011 				 * discovery.
1012 				 */
1013 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1014 					goto notify;
1015 
1016 				eoff += sizeof(struct ip6_frag);
1017 				nxt = fh->ip6f_nxt;
1018 				break;
1019 			default:
1020 				/*
1021 				 * This case includes ESP and the No Next
1022 				 * Header.  In such cases going to the notify
1023 				 * label does not have any meaning
1024 				 * (i.e. ctlfunc will be NULL), but we go
1025 				 * anyway since we might have to update
1026 				 * path MTU information.
1027 				 */
1028 				goto notify;
1029 			}
1030 		}
1031 	  notify:
1032 #ifndef PULLDOWN_TEST
1033 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1034 #else
1035 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1036 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1037 		if (icmp6 == NULL) {
1038 			icmp6stat.icp6s_tooshort++;
1039 			return (-1);
1040 		}
1041 #endif
1042 
1043 		/*
1044 		 * retrieve parameters from the inner IPv6 header, and convert
1045 		 * them into sockaddr structures.
1046 		 * XXX: there is no guarantee that the source or destination
1047 		 * addresses of the inner packet are in the same scope as
1048 		 * the addresses of the icmp packet.  But there is no other
1049 		 * way to determine the zone.
1050 		 */
1051 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1052 
1053 		bzero(&icmp6dst, sizeof(icmp6dst));
1054 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1055 		icmp6dst.sin6_family = AF_INET6;
1056 		if (finaldst == NULL)
1057 			icmp6dst.sin6_addr = eip6->ip6_dst;
1058 		else
1059 			icmp6dst.sin6_addr = *finaldst;
1060 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1061 			goto freeit;
1062 		bzero(&icmp6src, sizeof(icmp6src));
1063 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1064 		icmp6src.sin6_family = AF_INET6;
1065 		icmp6src.sin6_addr = eip6->ip6_src;
1066 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1067 			goto freeit;
1068 		icmp6src.sin6_flowinfo =
1069 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1070 
1071 		if (finaldst == NULL)
1072 			finaldst = &eip6->ip6_dst;
1073 		ip6cp.ip6c_m = m;
1074 		ip6cp.ip6c_icmp6 = icmp6;
1075 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1076 		ip6cp.ip6c_off = eoff;
1077 		ip6cp.ip6c_finaldst = finaldst;
1078 		ip6cp.ip6c_src = &icmp6src;
1079 		ip6cp.ip6c_nxt = nxt;
1080 
1081 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1082 			notifymtu = ntohl(icmp6->icmp6_mtu);
1083 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1084 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1085 		}
1086 
1087 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1088 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1089 		if (ctlfunc) {
1090 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1091 			    &ip6cp);
1092 		}
1093 	}
1094 	*mp = m;
1095 	return (0);
1096 
1097   freeit:
1098 	m_freem(m);
1099 	return (-1);
1100 }
1101 
1102 void
1103 icmp6_mtudisc_update(ip6cp, validated)
1104 	struct ip6ctlparam *ip6cp;
1105 	int validated;
1106 {
1107 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1108 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1109 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1110 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1111 	struct in_conninfo inc;
1112 
1113 #if 0
1114 	/*
1115 	 * RFC2460 section 5, last paragraph.
1116 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1117 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1118 	 * due to packet translator in the middle.
1119 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1120 	 * special handling.
1121 	 */
1122 	if (mtu < IPV6_MMTU)
1123 		return;
1124 #endif
1125 
1126 	/*
1127 	 * we reject ICMPv6 too big with abnormally small value.
1128 	 * XXX what is the good definition of "abnormally small"?
1129 	 */
1130 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1131 		return;
1132 
1133 	if (!validated)
1134 		return;
1135 
1136 	bzero(&inc, sizeof(inc));
1137 	inc.inc_flags = 1; /* IPv6 */
1138 	inc.inc6_faddr = *dst;
1139 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1140 		return;
1141 
1142 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1143 		tcp_hc_updatemtu(&inc, mtu);
1144 		icmp6stat.icp6s_pmtuchg++;
1145 	}
1146 }
1147 
1148 /*
1149  * Process a Node Information Query packet, based on
1150  * draft-ietf-ipngwg-icmp-name-lookups-07.
1151  *
1152  * Spec incompatibilities:
1153  * - IPv6 Subject address handling
1154  * - IPv4 Subject address handling support missing
1155  * - Proxy reply (answer even if it's not for me)
1156  * - joins NI group address at in6_ifattach() time only, does not cope
1157  *   with hostname changes by sethostname(3)
1158  */
1159 #define hostnamelen	strlen(hostname)
1160 static struct mbuf *
1161 ni6_input(m, off)
1162 	struct mbuf *m;
1163 	int off;
1164 {
1165 	struct icmp6_nodeinfo *ni6, *nni6;
1166 	struct mbuf *n = NULL;
1167 	u_int16_t qtype;
1168 	int subjlen;
1169 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1170 	struct ni_reply_fqdn *fqdn;
1171 	int addrs;		/* for NI_QTYPE_NODEADDR */
1172 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1173 	struct in6_addr in6_subj; /* subject address */
1174 	struct ip6_hdr *ip6;
1175 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1176 	char *subj = NULL;
1177 	struct in6_ifaddr *ia6 = NULL;
1178 
1179 	ip6 = mtod(m, struct ip6_hdr *);
1180 #ifndef PULLDOWN_TEST
1181 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1182 #else
1183 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1184 	if (ni6 == NULL) {
1185 		/* m is already reclaimed */
1186 		return (NULL);
1187 	}
1188 #endif
1189 
1190 	/*
1191 	 * Validate IPv6 source address.
1192 	 * The default configuration MUST be to refuse answering queries from
1193 	 * global-scope addresses according to RFC4602.
1194 	 * Notes:
1195 	 *  - it's not very clear what "refuse" means; this implementation
1196 	 *    simply drops it.
1197 	 *  - it's not very easy to identify global-scope (unicast) addresses
1198 	 *    since there are many prefixes for them.  It should be safer
1199 	 *    and in practice sufficient to check "all" but loopback and
1200 	 *    link-local (note that site-local unicast was deprecated and
1201 	 *    ULA is defined as global scope-wise)
1202 	 */
1203 	if ((icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1204 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1205 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1206 		goto bad;
1207 
1208 	/*
1209 	 * Validate IPv6 destination address.
1210 	 *
1211 	 * The Responder must discard the Query without further processing
1212 	 * unless it is one of the Responder's unicast or anycast addresses, or
1213 	 * a link-local scope multicast address which the Responder has joined.
1214 	 * [RFC4602, Section 5.]
1215 	 */
1216 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1217 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1218 			goto bad;
1219 		/* else it's a link-local multicast, fine */
1220 	} else {		/* unicast or anycast */
1221 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1222 			goto bad; /* XXX impossible */
1223 
1224 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1225 		    !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1226 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1227 				"a temporary address in %s:%d",
1228 			       __FILE__, __LINE__));
1229 			goto bad;
1230 		}
1231 	}
1232 
1233 	/* validate query Subject field. */
1234 	qtype = ntohs(ni6->ni_qtype);
1235 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1236 	switch (qtype) {
1237 	case NI_QTYPE_NOOP:
1238 	case NI_QTYPE_SUPTYPES:
1239 		/* 07 draft */
1240 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1241 			break;
1242 		/* FALLTHROUGH */
1243 	case NI_QTYPE_FQDN:
1244 	case NI_QTYPE_NODEADDR:
1245 	case NI_QTYPE_IPV4ADDR:
1246 		switch (ni6->ni_code) {
1247 		case ICMP6_NI_SUBJ_IPV6:
1248 #if ICMP6_NI_SUBJ_IPV6 != 0
1249 		case 0:
1250 #endif
1251 			/*
1252 			 * backward compatibility - try to accept 03 draft
1253 			 * format, where no Subject is present.
1254 			 */
1255 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1256 			    subjlen == 0) {
1257 				oldfqdn++;
1258 				break;
1259 			}
1260 #if ICMP6_NI_SUBJ_IPV6 != 0
1261 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1262 				goto bad;
1263 #endif
1264 
1265 			if (subjlen != sizeof(struct in6_addr))
1266 				goto bad;
1267 
1268 			/*
1269 			 * Validate Subject address.
1270 			 *
1271 			 * Not sure what exactly "address belongs to the node"
1272 			 * means in the spec, is it just unicast, or what?
1273 			 *
1274 			 * At this moment we consider Subject address as
1275 			 * "belong to the node" if the Subject address equals
1276 			 * to the IPv6 destination address; validation for
1277 			 * IPv6 destination address should have done enough
1278 			 * check for us.
1279 			 *
1280 			 * We do not do proxy at this moment.
1281 			 */
1282 			/* m_pulldown instead of copy? */
1283 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1284 			    subjlen, (caddr_t)&in6_subj);
1285 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1286 				goto bad;
1287 
1288 			subj = (char *)&in6_subj;
1289 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1290 				break;
1291 
1292 			/*
1293 			 * XXX if we are to allow other cases, we should really
1294 			 * be careful about scope here.
1295 			 * basically, we should disallow queries toward IPv6
1296 			 * destination X with subject Y,
1297 			 * if scope(X) > scope(Y).
1298 			 * if we allow scope(X) > scope(Y), it will result in
1299 			 * information leakage across scope boundary.
1300 			 */
1301 			goto bad;
1302 
1303 		case ICMP6_NI_SUBJ_FQDN:
1304 			/*
1305 			 * Validate Subject name with gethostname(3).
1306 			 *
1307 			 * The behavior may need some debate, since:
1308 			 * - we are not sure if the node has FQDN as
1309 			 *   hostname (returned by gethostname(3)).
1310 			 * - the code does wildcard match for truncated names.
1311 			 *   however, we are not sure if we want to perform
1312 			 *   wildcard match, if gethostname(3) side has
1313 			 *   truncated hostname.
1314 			 */
1315 			n = ni6_nametodns(hostname, hostnamelen, 0);
1316 			if (!n || n->m_next || n->m_len == 0)
1317 				goto bad;
1318 			IP6_EXTHDR_GET(subj, char *, m,
1319 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1320 			if (subj == NULL)
1321 				goto bad;
1322 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1323 			    n->m_len)) {
1324 				goto bad;
1325 			}
1326 			m_freem(n);
1327 			n = NULL;
1328 			break;
1329 
1330 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1331 		default:
1332 			goto bad;
1333 		}
1334 		break;
1335 	}
1336 
1337 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1338 	switch (qtype) {
1339 	case NI_QTYPE_FQDN:
1340 		if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1341 			goto bad;
1342 		break;
1343 	case NI_QTYPE_NODEADDR:
1344 	case NI_QTYPE_IPV4ADDR:
1345 		if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1346 			goto bad;
1347 		break;
1348 	}
1349 
1350 	/* guess reply length */
1351 	switch (qtype) {
1352 	case NI_QTYPE_NOOP:
1353 		break;		/* no reply data */
1354 	case NI_QTYPE_SUPTYPES:
1355 		replylen += sizeof(u_int32_t);
1356 		break;
1357 	case NI_QTYPE_FQDN:
1358 		/* XXX will append an mbuf */
1359 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1360 		break;
1361 	case NI_QTYPE_NODEADDR:
1362 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1363 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1364 		    sizeof(u_int32_t))) > MCLBYTES)
1365 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1366 		break;
1367 	case NI_QTYPE_IPV4ADDR:
1368 		/* unsupported - should respond with unknown Qtype? */
1369 		break;
1370 	default:
1371 		/*
1372 		 * XXX: We must return a reply with the ICMP6 code
1373 		 * `unknown Qtype' in this case.  However we regard the case
1374 		 * as an FQDN query for backward compatibility.
1375 		 * Older versions set a random value to this field,
1376 		 * so it rarely varies in the defined qtypes.
1377 		 * But the mechanism is not reliable...
1378 		 * maybe we should obsolete older versions.
1379 		 */
1380 		qtype = NI_QTYPE_FQDN;
1381 		/* XXX will append an mbuf */
1382 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1383 		oldfqdn++;
1384 		break;
1385 	}
1386 
1387 	/* allocate an mbuf to reply. */
1388 	MGETHDR(n, M_DONTWAIT, m->m_type);
1389 	if (n == NULL) {
1390 		m_freem(m);
1391 		return (NULL);
1392 	}
1393 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1394 	if (replylen > MHLEN) {
1395 		if (replylen > MCLBYTES) {
1396 			/*
1397 			 * XXX: should we try to allocate more? But MCLBYTES
1398 			 * is probably much larger than IPV6_MMTU...
1399 			 */
1400 			goto bad;
1401 		}
1402 		MCLGET(n, M_DONTWAIT);
1403 		if ((n->m_flags & M_EXT) == 0) {
1404 			goto bad;
1405 		}
1406 	}
1407 	n->m_pkthdr.len = n->m_len = replylen;
1408 
1409 	/* copy mbuf header and IPv6 + Node Information base headers */
1410 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1411 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1412 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1413 
1414 	/* qtype dependent procedure */
1415 	switch (qtype) {
1416 	case NI_QTYPE_NOOP:
1417 		nni6->ni_code = ICMP6_NI_SUCCESS;
1418 		nni6->ni_flags = 0;
1419 		break;
1420 	case NI_QTYPE_SUPTYPES:
1421 	{
1422 		u_int32_t v;
1423 		nni6->ni_code = ICMP6_NI_SUCCESS;
1424 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1425 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1426 		v = (u_int32_t)htonl(0x0000000f);
1427 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1428 		break;
1429 	}
1430 	case NI_QTYPE_FQDN:
1431 		nni6->ni_code = ICMP6_NI_SUCCESS;
1432 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1433 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1434 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1435 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1436 		/*
1437 		 * XXX do we really have FQDN in variable "hostname"?
1438 		 */
1439 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1440 		if (n->m_next == NULL)
1441 			goto bad;
1442 		/* XXX we assume that n->m_next is not a chain */
1443 		if (n->m_next->m_next != NULL)
1444 			goto bad;
1445 		n->m_pkthdr.len += n->m_next->m_len;
1446 		break;
1447 	case NI_QTYPE_NODEADDR:
1448 	{
1449 		int lenlim, copied;
1450 
1451 		nni6->ni_code = ICMP6_NI_SUCCESS;
1452 		n->m_pkthdr.len = n->m_len =
1453 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1454 		lenlim = M_TRAILINGSPACE(n);
1455 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1456 		/* XXX: reset mbuf length */
1457 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1458 		    sizeof(struct icmp6_nodeinfo) + copied;
1459 		break;
1460 	}
1461 	default:
1462 		break;		/* XXX impossible! */
1463 	}
1464 
1465 	nni6->ni_type = ICMP6_NI_REPLY;
1466 	m_freem(m);
1467 	return (n);
1468 
1469   bad:
1470 	m_freem(m);
1471 	if (n)
1472 		m_freem(n);
1473 	return (NULL);
1474 }
1475 #undef hostnamelen
1476 
1477 /*
1478  * make a mbuf with DNS-encoded string.  no compression support.
1479  *
1480  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1481  * treated as truncated name (two \0 at the end).  this is a wild guess.
1482  */
1483 static struct mbuf *
1484 ni6_nametodns(name, namelen, old)
1485 	const char *name;
1486 	int namelen;
1487 	int old;	/* return pascal string if non-zero */
1488 {
1489 	struct mbuf *m;
1490 	char *cp, *ep;
1491 	const char *p, *q;
1492 	int i, len, nterm;
1493 
1494 	if (old)
1495 		len = namelen + 1;
1496 	else
1497 		len = MCLBYTES;
1498 
1499 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1500 	MGET(m, M_DONTWAIT, MT_DATA);
1501 	if (m && len > MLEN) {
1502 		MCLGET(m, M_DONTWAIT);
1503 		if ((m->m_flags & M_EXT) == 0)
1504 			goto fail;
1505 	}
1506 	if (!m)
1507 		goto fail;
1508 	m->m_next = NULL;
1509 
1510 	if (old) {
1511 		m->m_len = len;
1512 		*mtod(m, char *) = namelen;
1513 		bcopy(name, mtod(m, char *) + 1, namelen);
1514 		return m;
1515 	} else {
1516 		m->m_len = 0;
1517 		cp = mtod(m, char *);
1518 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1519 
1520 		/* if not certain about my name, return empty buffer */
1521 		if (namelen == 0)
1522 			return m;
1523 
1524 		/*
1525 		 * guess if it looks like shortened hostname, or FQDN.
1526 		 * shortened hostname needs two trailing "\0".
1527 		 */
1528 		i = 0;
1529 		for (p = name; p < name + namelen; p++) {
1530 			if (*p && *p == '.')
1531 				i++;
1532 		}
1533 		if (i < 2)
1534 			nterm = 2;
1535 		else
1536 			nterm = 1;
1537 
1538 		p = name;
1539 		while (cp < ep && p < name + namelen) {
1540 			i = 0;
1541 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1542 				i++;
1543 			/* result does not fit into mbuf */
1544 			if (cp + i + 1 >= ep)
1545 				goto fail;
1546 			/*
1547 			 * DNS label length restriction, RFC1035 page 8.
1548 			 * "i == 0" case is included here to avoid returning
1549 			 * 0-length label on "foo..bar".
1550 			 */
1551 			if (i <= 0 || i >= 64)
1552 				goto fail;
1553 			*cp++ = i;
1554 			bcopy(p, cp, i);
1555 			cp += i;
1556 			p = q;
1557 			if (p < name + namelen && *p == '.')
1558 				p++;
1559 		}
1560 		/* termination */
1561 		if (cp + nterm >= ep)
1562 			goto fail;
1563 		while (nterm-- > 0)
1564 			*cp++ = '\0';
1565 		m->m_len = cp - mtod(m, char *);
1566 		return m;
1567 	}
1568 
1569 	panic("should not reach here");
1570 	/* NOTREACHED */
1571 
1572  fail:
1573 	if (m)
1574 		m_freem(m);
1575 	return NULL;
1576 }
1577 
1578 /*
1579  * check if two DNS-encoded string matches.  takes care of truncated
1580  * form (with \0\0 at the end).  no compression support.
1581  * XXX upper/lowercase match (see RFC2065)
1582  */
1583 static int
1584 ni6_dnsmatch(a, alen, b, blen)
1585 	const char *a;
1586 	int alen;
1587 	const char *b;
1588 	int blen;
1589 {
1590 	const char *a0, *b0;
1591 	int l;
1592 
1593 	/* simplest case - need validation? */
1594 	if (alen == blen && bcmp(a, b, alen) == 0)
1595 		return 1;
1596 
1597 	a0 = a;
1598 	b0 = b;
1599 
1600 	/* termination is mandatory */
1601 	if (alen < 2 || blen < 2)
1602 		return 0;
1603 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1604 		return 0;
1605 	alen--;
1606 	blen--;
1607 
1608 	while (a - a0 < alen && b - b0 < blen) {
1609 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1610 			return 0;
1611 
1612 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1613 			return 0;
1614 		/* we don't support compression yet */
1615 		if (a[0] >= 64 || b[0] >= 64)
1616 			return 0;
1617 
1618 		/* truncated case */
1619 		if (a[0] == 0 && a - a0 == alen - 1)
1620 			return 1;
1621 		if (b[0] == 0 && b - b0 == blen - 1)
1622 			return 1;
1623 		if (a[0] == 0 || b[0] == 0)
1624 			return 0;
1625 
1626 		if (a[0] != b[0])
1627 			return 0;
1628 		l = a[0];
1629 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1630 			return 0;
1631 		if (bcmp(a + 1, b + 1, l) != 0)
1632 			return 0;
1633 
1634 		a += 1 + l;
1635 		b += 1 + l;
1636 	}
1637 
1638 	if (a - a0 == alen && b - b0 == blen)
1639 		return 1;
1640 	else
1641 		return 0;
1642 }
1643 
1644 /*
1645  * calculate the number of addresses to be returned in the node info reply.
1646  */
1647 static int
1648 ni6_addrs(ni6, m, ifpp, subj)
1649 	struct icmp6_nodeinfo *ni6;
1650 	struct mbuf *m;
1651 	struct ifnet **ifpp;
1652 	struct in6_addr *subj;
1653 {
1654 	struct ifnet *ifp;
1655 	struct in6_ifaddr *ifa6;
1656 	struct ifaddr *ifa;
1657 	int addrs = 0, addrsofif, iffound = 0;
1658 	int niflags = ni6->ni_flags;
1659 
1660 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1661 		switch (ni6->ni_code) {
1662 		case ICMP6_NI_SUBJ_IPV6:
1663 			if (subj == NULL) /* must be impossible... */
1664 				return (0);
1665 			break;
1666 		default:
1667 			/*
1668 			 * XXX: we only support IPv6 subject address for
1669 			 * this Qtype.
1670 			 */
1671 			return (0);
1672 		}
1673 	}
1674 
1675 	IFNET_RLOCK();
1676 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1677 		addrsofif = 0;
1678 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1679 			if (ifa->ifa_addr->sa_family != AF_INET6)
1680 				continue;
1681 			ifa6 = (struct in6_ifaddr *)ifa;
1682 
1683 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1684 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1685 				iffound = 1;
1686 
1687 			/*
1688 			 * IPv4-mapped addresses can only be returned by a
1689 			 * Node Information proxy, since they represent
1690 			 * addresses of IPv4-only nodes, which perforce do
1691 			 * not implement this protocol.
1692 			 * [icmp-name-lookups-07, Section 5.4]
1693 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1694 			 * this function at this moment.
1695 			 */
1696 
1697 			/* What do we have to do about ::1? */
1698 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1699 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1700 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1701 					continue;
1702 				break;
1703 			case IPV6_ADDR_SCOPE_SITELOCAL:
1704 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1705 					continue;
1706 				break;
1707 			case IPV6_ADDR_SCOPE_GLOBAL:
1708 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1709 					continue;
1710 				break;
1711 			default:
1712 				continue;
1713 			}
1714 
1715 			/*
1716 			 * check if anycast is okay.
1717 			 * XXX: just experimental.  not in the spec.
1718 			 */
1719 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1720 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1721 				continue; /* we need only unicast addresses */
1722 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1723 			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1724 				continue;
1725 			}
1726 			addrsofif++; /* count the address */
1727 		}
1728 		if (iffound) {
1729 			*ifpp = ifp;
1730 			IFNET_RUNLOCK();
1731 			return (addrsofif);
1732 		}
1733 
1734 		addrs += addrsofif;
1735 	}
1736 	IFNET_RUNLOCK();
1737 
1738 	return (addrs);
1739 }
1740 
1741 static int
1742 ni6_store_addrs(ni6, nni6, ifp0, resid)
1743 	struct icmp6_nodeinfo *ni6, *nni6;
1744 	struct ifnet *ifp0;
1745 	int resid;
1746 {
1747 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1748 	struct in6_ifaddr *ifa6;
1749 	struct ifaddr *ifa;
1750 	struct ifnet *ifp_dep = NULL;
1751 	int copied = 0, allow_deprecated = 0;
1752 	u_char *cp = (u_char *)(nni6 + 1);
1753 	int niflags = ni6->ni_flags;
1754 	u_int32_t ltime;
1755 
1756 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1757 		return (0);	/* needless to copy */
1758 
1759 	IFNET_RLOCK();
1760   again:
1761 
1762 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1763 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1764 		     ifa = ifa->ifa_list.tqe_next) {
1765 			if (ifa->ifa_addr->sa_family != AF_INET6)
1766 				continue;
1767 			ifa6 = (struct in6_ifaddr *)ifa;
1768 
1769 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1770 			    allow_deprecated == 0) {
1771 				/*
1772 				 * prefererred address should be put before
1773 				 * deprecated addresses.
1774 				 */
1775 
1776 				/* record the interface for later search */
1777 				if (ifp_dep == NULL)
1778 					ifp_dep = ifp;
1779 
1780 				continue;
1781 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1782 			    allow_deprecated != 0)
1783 				continue; /* we now collect deprecated addrs */
1784 
1785 			/* What do we have to do about ::1? */
1786 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1787 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1788 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1789 					continue;
1790 				break;
1791 			case IPV6_ADDR_SCOPE_SITELOCAL:
1792 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1793 					continue;
1794 				break;
1795 			case IPV6_ADDR_SCOPE_GLOBAL:
1796 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1797 					continue;
1798 				break;
1799 			default:
1800 				continue;
1801 			}
1802 
1803 			/*
1804 			 * check if anycast is okay.
1805 			 * XXX: just experimental.  not in the spec.
1806 			 */
1807 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1808 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1809 				continue;
1810 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1811 			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1812 				continue;
1813 			}
1814 
1815 			/* now we can copy the address */
1816 			if (resid < sizeof(struct in6_addr) +
1817 			    sizeof(u_int32_t)) {
1818 				/*
1819 				 * We give up much more copy.
1820 				 * Set the truncate flag and return.
1821 				 */
1822 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1823 				IFNET_RUNLOCK();
1824 				return (copied);
1825 			}
1826 
1827 			/*
1828 			 * Set the TTL of the address.
1829 			 * The TTL value should be one of the following
1830 			 * according to the specification:
1831 			 *
1832 			 * 1. The remaining lifetime of a DHCP lease on the
1833 			 *    address, or
1834 			 * 2. The remaining Valid Lifetime of a prefix from
1835 			 *    which the address was derived through Stateless
1836 			 *    Autoconfiguration.
1837 			 *
1838 			 * Note that we currently do not support stateful
1839 			 * address configuration by DHCPv6, so the former
1840 			 * case can't happen.
1841 			 */
1842 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1843 				ltime = ND6_INFINITE_LIFETIME;
1844 			else {
1845 				if (ifa6->ia6_lifetime.ia6t_expire >
1846 				    time_second)
1847 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1848 				else
1849 					ltime = 0;
1850 			}
1851 
1852 			bcopy(&ltime, cp, sizeof(u_int32_t));
1853 			cp += sizeof(u_int32_t);
1854 
1855 			/* copy the address itself */
1856 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1857 			    sizeof(struct in6_addr));
1858 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1859 			cp += sizeof(struct in6_addr);
1860 
1861 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1862 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1863 		}
1864 		if (ifp0)	/* we need search only on the specified IF */
1865 			break;
1866 	}
1867 
1868 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1869 		ifp = ifp_dep;
1870 		allow_deprecated = 1;
1871 
1872 		goto again;
1873 	}
1874 
1875 	IFNET_RUNLOCK();
1876 
1877 	return (copied);
1878 }
1879 
1880 /*
1881  * XXX almost dup'ed code with rip6_input.
1882  */
1883 static int
1884 icmp6_rip6_input(mp, off)
1885 	struct	mbuf **mp;
1886 	int	off;
1887 {
1888 	struct mbuf *m = *mp;
1889 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1890 	struct in6pcb *in6p;
1891 	struct in6pcb *last = NULL;
1892 	struct sockaddr_in6 fromsa;
1893 	struct icmp6_hdr *icmp6;
1894 	struct mbuf *opts = NULL;
1895 
1896 #ifndef PULLDOWN_TEST
1897 	/* this is assumed to be safe. */
1898 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1899 #else
1900 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1901 	if (icmp6 == NULL) {
1902 		/* m is already reclaimed */
1903 		return (IPPROTO_DONE);
1904 	}
1905 #endif
1906 
1907 	/*
1908 	 * XXX: the address may have embedded scope zone ID, which should be
1909 	 * hidden from applications.
1910 	 */
1911 	bzero(&fromsa, sizeof(fromsa));
1912 	fromsa.sin6_family = AF_INET6;
1913 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1914 	fromsa.sin6_addr = ip6->ip6_src;
1915 	if (sa6_recoverscope(&fromsa)) {
1916 		m_freem(m);
1917 		return (IPPROTO_DONE);
1918 	}
1919 
1920 	INP_INFO_RLOCK(&ripcbinfo);
1921 	LIST_FOREACH(in6p, &ripcb, inp_list) {
1922 		INP_LOCK(in6p);
1923 		if ((in6p->inp_vflag & INP_IPV6) == 0) {
1924 	docontinue:
1925 			INP_UNLOCK(in6p);
1926 			continue;
1927 		}
1928 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1929 			goto docontinue;
1930 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1931 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1932 			goto docontinue;
1933 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1934 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1935 			goto docontinue;
1936 		if (in6p->in6p_icmp6filt
1937 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1938 				 in6p->in6p_icmp6filt))
1939 			goto docontinue;
1940 		if (last) {
1941 			struct	mbuf *n = NULL;
1942 
1943 			/*
1944 			 * Recent network drivers tend to allocate a single
1945 			 * mbuf cluster, rather than to make a couple of
1946 			 * mbufs without clusters.  Also, since the IPv6 code
1947 			 * path tries to avoid m_pullup(), it is highly
1948 			 * probable that we still have an mbuf cluster here
1949 			 * even though the necessary length can be stored in an
1950 			 * mbuf's internal buffer.
1951 			 * Meanwhile, the default size of the receive socket
1952 			 * buffer for raw sockets is not so large.  This means
1953 			 * the possibility of packet loss is relatively higher
1954 			 * than before.  To avoid this scenario, we copy the
1955 			 * received data to a separate mbuf that does not use
1956 			 * a cluster, if possible.
1957 			 * XXX: it is better to copy the data after stripping
1958 			 * intermediate headers.
1959 			 */
1960 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1961 			    m->m_len <= MHLEN) {
1962 				MGET(n, M_DONTWAIT, m->m_type);
1963 				if (n != NULL) {
1964 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1965 						bcopy(m->m_data, n->m_data,
1966 						      m->m_len);
1967 						n->m_len = m->m_len;
1968 					} else {
1969 						m_free(n);
1970 						n = NULL;
1971 					}
1972 				}
1973 			}
1974 			if (n != NULL ||
1975 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1976 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1977 					ip6_savecontrol(last, n, &opts);
1978 				/* strip intermediate headers */
1979 				m_adj(n, off);
1980 				SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
1981 				if (sbappendaddr_locked(
1982 				    &last->in6p_socket->so_rcv,
1983 				    (struct sockaddr *)&fromsa, n, opts)
1984 				    == 0) {
1985 					/* should notify about lost packet */
1986 					m_freem(n);
1987 					if (opts) {
1988 						m_freem(opts);
1989 					}
1990 					SOCKBUF_UNLOCK(
1991 					    &last->in6p_socket->so_rcv);
1992 				} else
1993 					sorwakeup_locked(last->in6p_socket);
1994 				opts = NULL;
1995 			}
1996 			INP_UNLOCK(last);
1997 		}
1998 		last = in6p;
1999 	}
2000 	if (last) {
2001 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2002 			ip6_savecontrol(last, m, &opts);
2003 		/* strip intermediate headers */
2004 		m_adj(m, off);
2005 
2006 		/* avoid using mbuf clusters if possible (see above) */
2007 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2008 		    m->m_len <= MHLEN) {
2009 			struct mbuf *n;
2010 
2011 			MGET(n, M_DONTWAIT, m->m_type);
2012 			if (n != NULL) {
2013 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2014 					bcopy(m->m_data, n->m_data, m->m_len);
2015 					n->m_len = m->m_len;
2016 
2017 					m_freem(m);
2018 					m = n;
2019 				} else {
2020 					m_freem(n);
2021 					n = NULL;
2022 				}
2023 			}
2024 		}
2025 		SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
2026 		if (sbappendaddr_locked(&last->in6p_socket->so_rcv,
2027 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2028 			m_freem(m);
2029 			if (opts)
2030 				m_freem(opts);
2031 			SOCKBUF_UNLOCK(&last->in6p_socket->so_rcv);
2032 		} else
2033 			sorwakeup_locked(last->in6p_socket);
2034 		INP_UNLOCK(last);
2035 	} else {
2036 		m_freem(m);
2037 		ip6stat.ip6s_delivered--;
2038 	}
2039 	INP_INFO_RUNLOCK(&ripcbinfo);
2040 	return IPPROTO_DONE;
2041 }
2042 
2043 /*
2044  * Reflect the ip6 packet back to the source.
2045  * OFF points to the icmp6 header, counted from the top of the mbuf.
2046  */
2047 void
2048 icmp6_reflect(m, off)
2049 	struct	mbuf *m;
2050 	size_t off;
2051 {
2052 	struct ip6_hdr *ip6;
2053 	struct icmp6_hdr *icmp6;
2054 	struct in6_ifaddr *ia;
2055 	int plen;
2056 	int type, code;
2057 	struct ifnet *outif = NULL;
2058 	struct in6_addr origdst, *src = NULL;
2059 
2060 	/* too short to reflect */
2061 	if (off < sizeof(struct ip6_hdr)) {
2062 		nd6log((LOG_DEBUG,
2063 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2064 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2065 		    __FILE__, __LINE__));
2066 		goto bad;
2067 	}
2068 
2069 	/*
2070 	 * If there are extra headers between IPv6 and ICMPv6, strip
2071 	 * off that header first.
2072 	 */
2073 #ifdef DIAGNOSTIC
2074 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2075 		panic("assumption failed in icmp6_reflect");
2076 #endif
2077 	if (off > sizeof(struct ip6_hdr)) {
2078 		size_t l;
2079 		struct ip6_hdr nip6;
2080 
2081 		l = off - sizeof(struct ip6_hdr);
2082 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2083 		m_adj(m, l);
2084 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2085 		if (m->m_len < l) {
2086 			if ((m = m_pullup(m, l)) == NULL)
2087 				return;
2088 		}
2089 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2090 	} else /* off == sizeof(struct ip6_hdr) */ {
2091 		size_t l;
2092 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2093 		if (m->m_len < l) {
2094 			if ((m = m_pullup(m, l)) == NULL)
2095 				return;
2096 		}
2097 	}
2098 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2099 	ip6 = mtod(m, struct ip6_hdr *);
2100 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2101 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2102 	type = icmp6->icmp6_type; /* keep type for statistics */
2103 	code = icmp6->icmp6_code; /* ditto. */
2104 
2105 	origdst = ip6->ip6_dst;
2106 	/*
2107 	 * ip6_input() drops a packet if its src is multicast.
2108 	 * So, the src is never multicast.
2109 	 */
2110 	ip6->ip6_dst = ip6->ip6_src;
2111 
2112 	/*
2113 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2114 	 * use dst as the src for the reply.
2115 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2116 	 * (for example) when we encounter an error while forwarding procedure
2117 	 * destined to a duplicated address of ours.
2118 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2119 	 * procedure of an outgoing packet of our own, in which case we need
2120 	 * to search in the ifaddr list.
2121 	 */
2122 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2123 		if ((ia = ip6_getdstifaddr(m))) {
2124 			if (!(ia->ia6_flags &
2125 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2126 				src = &ia->ia_addr.sin6_addr;
2127 		} else {
2128 			struct sockaddr_in6 d;
2129 
2130 			bzero(&d, sizeof(d));
2131 			d.sin6_family = AF_INET6;
2132 			d.sin6_len = sizeof(d);
2133 			d.sin6_addr = origdst;
2134 			ia = (struct in6_ifaddr *)
2135 			    ifa_ifwithaddr((struct sockaddr *)&d);
2136 			if (ia &&
2137 			    !(ia->ia6_flags &
2138 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2139 				src = &ia->ia_addr.sin6_addr;
2140 			}
2141 		}
2142 	}
2143 
2144 	if (src == NULL) {
2145 		int e;
2146 		struct sockaddr_in6 sin6;
2147 		struct route_in6 ro;
2148 
2149 		/*
2150 		 * This case matches to multicasts, our anycast, or unicasts
2151 		 * that we do not own.  Select a source address based on the
2152 		 * source address of the erroneous packet.
2153 		 */
2154 		bzero(&sin6, sizeof(sin6));
2155 		sin6.sin6_family = AF_INET6;
2156 		sin6.sin6_len = sizeof(sin6);
2157 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2158 
2159 		bzero(&ro, sizeof(ro));
2160 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2161 		if (ro.ro_rt)
2162 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2163 		if (src == NULL) {
2164 			char ip6buf[INET6_ADDRSTRLEN];
2165 			nd6log((LOG_DEBUG,
2166 			    "icmp6_reflect: source can't be determined: "
2167 			    "dst=%s, error=%d\n",
2168 			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2169 			goto bad;
2170 		}
2171 	}
2172 
2173 	ip6->ip6_src = *src;
2174 	ip6->ip6_flow = 0;
2175 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2176 	ip6->ip6_vfc |= IPV6_VERSION;
2177 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2178 	if (outif)
2179 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2180 	else if (m->m_pkthdr.rcvif) {
2181 		/* XXX: This may not be the outgoing interface */
2182 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2183 	} else
2184 		ip6->ip6_hlim = ip6_defhlim;
2185 
2186 	icmp6->icmp6_cksum = 0;
2187 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2188 	    sizeof(struct ip6_hdr), plen);
2189 
2190 	/*
2191 	 * XXX option handling
2192 	 */
2193 
2194 	m->m_flags &= ~(M_BCAST|M_MCAST);
2195 
2196 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2197 	if (outif)
2198 		icmp6_ifoutstat_inc(outif, type, code);
2199 
2200 	return;
2201 
2202  bad:
2203 	m_freem(m);
2204 	return;
2205 }
2206 
2207 void
2208 icmp6_fasttimo()
2209 {
2210 
2211 	return;
2212 }
2213 
2214 static const char *
2215 icmp6_redirect_diag(src6, dst6, tgt6)
2216 	struct in6_addr *src6;
2217 	struct in6_addr *dst6;
2218 	struct in6_addr *tgt6;
2219 {
2220 	static char buf[1024];
2221 	char ip6bufs[INET6_ADDRSTRLEN];
2222 	char ip6bufd[INET6_ADDRSTRLEN];
2223 	char ip6buft[INET6_ADDRSTRLEN];
2224 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2225 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2226 	    ip6_sprintf(ip6buft, tgt6));
2227 	return buf;
2228 }
2229 
2230 void
2231 icmp6_redirect_input(m, off)
2232 	struct mbuf *m;
2233 	int off;
2234 {
2235 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2236 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2237 	struct nd_redirect *nd_rd;
2238 	int icmp6len = ntohs(ip6->ip6_plen);
2239 	char *lladdr = NULL;
2240 	int lladdrlen = 0;
2241 	u_char *redirhdr = NULL;
2242 	int redirhdrlen = 0;
2243 	struct rtentry *rt = NULL;
2244 	int is_router;
2245 	int is_onlink;
2246 	struct in6_addr src6 = ip6->ip6_src;
2247 	struct in6_addr redtgt6;
2248 	struct in6_addr reddst6;
2249 	union nd_opts ndopts;
2250 	char ip6buf[INET6_ADDRSTRLEN];
2251 
2252 	if (!m || !ifp)
2253 		return;
2254 
2255 	/* XXX if we are router, we don't update route by icmp6 redirect */
2256 	if (ip6_forwarding)
2257 		goto freeit;
2258 	if (!icmp6_rediraccept)
2259 		goto freeit;
2260 
2261 #ifndef PULLDOWN_TEST
2262 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2263 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2264 #else
2265 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2266 	if (nd_rd == NULL) {
2267 		icmp6stat.icp6s_tooshort++;
2268 		return;
2269 	}
2270 #endif
2271 	redtgt6 = nd_rd->nd_rd_target;
2272 	reddst6 = nd_rd->nd_rd_dst;
2273 
2274 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2275 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2276 		goto freeit;
2277 	}
2278 
2279 	/* validation */
2280 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2281 		nd6log((LOG_ERR,
2282 		    "ICMP6 redirect sent from %s rejected; "
2283 		    "must be from linklocal\n",
2284 		    ip6_sprintf(ip6buf, &src6)));
2285 		goto bad;
2286 	}
2287 	if (ip6->ip6_hlim != 255) {
2288 		nd6log((LOG_ERR,
2289 		    "ICMP6 redirect sent from %s rejected; "
2290 		    "hlim=%d (must be 255)\n",
2291 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2292 		goto bad;
2293 	}
2294     {
2295 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2296 	struct sockaddr_in6 sin6;
2297 	struct in6_addr *gw6;
2298 
2299 	bzero(&sin6, sizeof(sin6));
2300 	sin6.sin6_family = AF_INET6;
2301 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2302 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2303 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2304 	if (rt) {
2305 		if (rt->rt_gateway == NULL ||
2306 		    rt->rt_gateway->sa_family != AF_INET6) {
2307 			nd6log((LOG_ERR,
2308 			    "ICMP6 redirect rejected; no route "
2309 			    "with inet6 gateway found for redirect dst: %s\n",
2310 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2311 			RTFREE_LOCKED(rt);
2312 			goto bad;
2313 		}
2314 
2315 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2316 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2317 			nd6log((LOG_ERR,
2318 			    "ICMP6 redirect rejected; "
2319 			    "not equal to gw-for-src=%s (must be same): "
2320 			    "%s\n",
2321 			    ip6_sprintf(ip6buf, gw6),
2322 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2323 			RTFREE_LOCKED(rt);
2324 			goto bad;
2325 		}
2326 	} else {
2327 		nd6log((LOG_ERR,
2328 		    "ICMP6 redirect rejected; "
2329 		    "no route found for redirect dst: %s\n",
2330 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2331 		goto bad;
2332 	}
2333 	RTFREE_LOCKED(rt);
2334 	rt = NULL;
2335     }
2336 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2337 		nd6log((LOG_ERR,
2338 		    "ICMP6 redirect rejected; "
2339 		    "redirect dst must be unicast: %s\n",
2340 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2341 		goto bad;
2342 	}
2343 
2344 	is_router = is_onlink = 0;
2345 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2346 		is_router = 1;	/* router case */
2347 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2348 		is_onlink = 1;	/* on-link destination case */
2349 	if (!is_router && !is_onlink) {
2350 		nd6log((LOG_ERR,
2351 		    "ICMP6 redirect rejected; "
2352 		    "neither router case nor onlink case: %s\n",
2353 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2354 		goto bad;
2355 	}
2356 	/* validation passed */
2357 
2358 	icmp6len -= sizeof(*nd_rd);
2359 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2360 	if (nd6_options(&ndopts) < 0) {
2361 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2362 		    "invalid ND option, rejected: %s\n",
2363 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2364 		/* nd6_options have incremented stats */
2365 		goto freeit;
2366 	}
2367 
2368 	if (ndopts.nd_opts_tgt_lladdr) {
2369 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2370 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2371 	}
2372 
2373 	if (ndopts.nd_opts_rh) {
2374 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2375 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2376 	}
2377 
2378 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2379 		nd6log((LOG_INFO,
2380 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2381 		    "(if %d, icmp6 packet %d): %s\n",
2382 		    ip6_sprintf(ip6buf, &redtgt6),
2383 		    ifp->if_addrlen, lladdrlen - 2,
2384 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2385 		goto bad;
2386 	}
2387 
2388 	/* RFC 2461 8.3 */
2389 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2390 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2391 
2392 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2393 		/* perform rtredirect */
2394 		struct sockaddr_in6 sdst;
2395 		struct sockaddr_in6 sgw;
2396 		struct sockaddr_in6 ssrc;
2397 
2398 		bzero(&sdst, sizeof(sdst));
2399 		bzero(&sgw, sizeof(sgw));
2400 		bzero(&ssrc, sizeof(ssrc));
2401 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2402 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2403 			sizeof(struct sockaddr_in6);
2404 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2405 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2406 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2407 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2408 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2409 		    (struct sockaddr *)&ssrc);
2410 	}
2411 	/* finally update cached route in each socket via pfctlinput */
2412     {
2413 	struct sockaddr_in6 sdst;
2414 
2415 	bzero(&sdst, sizeof(sdst));
2416 	sdst.sin6_family = AF_INET6;
2417 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2418 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2419 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2420 #if defined(IPSEC) || defined(FAST_IPSEC)
2421 	key_sa_routechange((struct sockaddr *)&sdst);
2422 #endif
2423     }
2424 
2425  freeit:
2426 	m_freem(m);
2427 	return;
2428 
2429  bad:
2430 	icmp6stat.icp6s_badredirect++;
2431 	m_freem(m);
2432 }
2433 
2434 void
2435 icmp6_redirect_output(m0, rt)
2436 	struct mbuf *m0;
2437 	struct rtentry *rt;
2438 {
2439 	struct ifnet *ifp;	/* my outgoing interface */
2440 	struct in6_addr *ifp_ll6;
2441 	struct in6_addr *router_ll6;
2442 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2443 	struct mbuf *m = NULL;	/* newly allocated one */
2444 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2445 	struct nd_redirect *nd_rd;
2446 	size_t maxlen;
2447 	u_char *p;
2448 	struct ifnet *outif = NULL;
2449 	struct sockaddr_in6 src_sa;
2450 
2451 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2452 
2453 	/* if we are not router, we don't send icmp6 redirect */
2454 	if (!ip6_forwarding)
2455 		goto fail;
2456 
2457 	/* sanity check */
2458 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2459 		goto fail;
2460 
2461 	/*
2462 	 * Address check:
2463 	 *  the source address must identify a neighbor, and
2464 	 *  the destination address must not be a multicast address
2465 	 *  [RFC 2461, sec 8.2]
2466 	 */
2467 	sip6 = mtod(m0, struct ip6_hdr *);
2468 	bzero(&src_sa, sizeof(src_sa));
2469 	src_sa.sin6_family = AF_INET6;
2470 	src_sa.sin6_len = sizeof(src_sa);
2471 	src_sa.sin6_addr = sip6->ip6_src;
2472 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2473 		goto fail;
2474 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2475 		goto fail;	/* what should we do here? */
2476 
2477 	/* rate limit */
2478 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2479 		goto fail;
2480 
2481 	/*
2482 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2483 	 * we almost always ask for an mbuf cluster for simplicity.
2484 	 * (MHLEN < IPV6_MMTU is almost always true)
2485 	 */
2486 #if IPV6_MMTU >= MCLBYTES
2487 # error assumption failed about IPV6_MMTU and MCLBYTES
2488 #endif
2489 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2490 	if (m && IPV6_MMTU >= MHLEN)
2491 		MCLGET(m, M_DONTWAIT);
2492 	if (!m)
2493 		goto fail;
2494 	m->m_pkthdr.rcvif = NULL;
2495 	m->m_len = 0;
2496 	maxlen = M_TRAILINGSPACE(m);
2497 	maxlen = min(IPV6_MMTU, maxlen);
2498 	/* just for safety */
2499 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2500 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2501 		goto fail;
2502 	}
2503 
2504 	{
2505 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2506 		struct in6_ifaddr *ia;
2507 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2508 						 IN6_IFF_NOTREADY|
2509 						 IN6_IFF_ANYCAST)) == NULL)
2510 			goto fail;
2511 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2512 	}
2513 
2514 	/* get ip6 linklocal address for the router. */
2515 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2516 		struct sockaddr_in6 *sin6;
2517 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2518 		router_ll6 = &sin6->sin6_addr;
2519 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2520 			router_ll6 = (struct in6_addr *)NULL;
2521 	} else
2522 		router_ll6 = (struct in6_addr *)NULL;
2523 
2524 	/* ip6 */
2525 	ip6 = mtod(m, struct ip6_hdr *);
2526 	ip6->ip6_flow = 0;
2527 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2528 	ip6->ip6_vfc |= IPV6_VERSION;
2529 	/* ip6->ip6_plen will be set later */
2530 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2531 	ip6->ip6_hlim = 255;
2532 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2533 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2534 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2535 
2536 	/* ND Redirect */
2537 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2538 	nd_rd->nd_rd_type = ND_REDIRECT;
2539 	nd_rd->nd_rd_code = 0;
2540 	nd_rd->nd_rd_reserved = 0;
2541 	if (rt->rt_flags & RTF_GATEWAY) {
2542 		/*
2543 		 * nd_rd->nd_rd_target must be a link-local address in
2544 		 * better router cases.
2545 		 */
2546 		if (!router_ll6)
2547 			goto fail;
2548 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2549 		    sizeof(nd_rd->nd_rd_target));
2550 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2551 		    sizeof(nd_rd->nd_rd_dst));
2552 	} else {
2553 		/* make sure redtgt == reddst */
2554 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2555 		    sizeof(nd_rd->nd_rd_target));
2556 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2557 		    sizeof(nd_rd->nd_rd_dst));
2558 	}
2559 
2560 	p = (u_char *)(nd_rd + 1);
2561 
2562 	if (!router_ll6)
2563 		goto nolladdropt;
2564 
2565 	{
2566 		/* target lladdr option */
2567 		struct rtentry *rt_router = NULL;
2568 		int len;
2569 		struct sockaddr_dl *sdl;
2570 		struct nd_opt_hdr *nd_opt;
2571 		char *lladdr;
2572 
2573 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2574 		if (!rt_router)
2575 			goto nolladdropt;
2576 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2577 		len = (len + 7) & ~7;	/* round by 8 */
2578 		/* safety check */
2579 		if (len + (p - (u_char *)ip6) > maxlen)
2580 			goto nolladdropt;
2581 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2582 		    (rt_router->rt_flags & RTF_LLINFO) &&
2583 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2584 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2585 		    sdl->sdl_alen) {
2586 			nd_opt = (struct nd_opt_hdr *)p;
2587 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2588 			nd_opt->nd_opt_len = len >> 3;
2589 			lladdr = (char *)(nd_opt + 1);
2590 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2591 			p += len;
2592 		}
2593 	}
2594 nolladdropt:;
2595 
2596 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2597 
2598 	/* just to be safe */
2599 #ifdef M_DECRYPTED	/*not openbsd*/
2600 	if (m0->m_flags & M_DECRYPTED)
2601 		goto noredhdropt;
2602 #endif
2603 	if (p - (u_char *)ip6 > maxlen)
2604 		goto noredhdropt;
2605 
2606 	{
2607 		/* redirected header option */
2608 		int len;
2609 		struct nd_opt_rd_hdr *nd_opt_rh;
2610 
2611 		/*
2612 		 * compute the maximum size for icmp6 redirect header option.
2613 		 * XXX room for auth header?
2614 		 */
2615 		len = maxlen - (p - (u_char *)ip6);
2616 		len &= ~7;
2617 
2618 		/* This is just for simplicity. */
2619 		if (m0->m_pkthdr.len != m0->m_len) {
2620 			if (m0->m_next) {
2621 				m_freem(m0->m_next);
2622 				m0->m_next = NULL;
2623 			}
2624 			m0->m_pkthdr.len = m0->m_len;
2625 		}
2626 
2627 		/*
2628 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2629 		 * about padding/truncate rule for the original IP packet.
2630 		 * From the discussion on IPv6imp in Feb 1999,
2631 		 * the consensus was:
2632 		 * - "attach as much as possible" is the goal
2633 		 * - pad if not aligned (original size can be guessed by
2634 		 *   original ip6 header)
2635 		 * Following code adds the padding if it is simple enough,
2636 		 * and truncates if not.
2637 		 */
2638 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2639 			panic("assumption failed in %s:%d", __FILE__,
2640 			    __LINE__);
2641 
2642 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2643 			/* not enough room, truncate */
2644 			m0->m_pkthdr.len = m0->m_len = len -
2645 			    sizeof(*nd_opt_rh);
2646 		} else {
2647 			/* enough room, pad or truncate */
2648 			size_t extra;
2649 
2650 			extra = m0->m_pkthdr.len % 8;
2651 			if (extra) {
2652 				/* pad if easy enough, truncate if not */
2653 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2654 					/* pad */
2655 					m0->m_len += (8 - extra);
2656 					m0->m_pkthdr.len += (8 - extra);
2657 				} else {
2658 					/* truncate */
2659 					m0->m_pkthdr.len -= extra;
2660 					m0->m_len -= extra;
2661 				}
2662 			}
2663 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2664 			m0->m_pkthdr.len = m0->m_len = len -
2665 			    sizeof(*nd_opt_rh);
2666 		}
2667 
2668 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2669 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2670 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2671 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2672 		p += sizeof(*nd_opt_rh);
2673 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2674 
2675 		/* connect m0 to m */
2676 		m_tag_delete_chain(m0, NULL);
2677 		m0->m_flags &= ~M_PKTHDR;
2678 		m->m_next = m0;
2679 		m->m_pkthdr.len = m->m_len + m0->m_len;
2680 		m0 = NULL;
2681 	}
2682 noredhdropt:;
2683 	if (m0) {
2684 		m_freem(m0);
2685 		m0 = NULL;
2686 	}
2687 
2688 	/* XXX: clear embedded link IDs in the inner header */
2689 	in6_clearscope(&sip6->ip6_src);
2690 	in6_clearscope(&sip6->ip6_dst);
2691 	in6_clearscope(&nd_rd->nd_rd_target);
2692 	in6_clearscope(&nd_rd->nd_rd_dst);
2693 
2694 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2695 
2696 	nd_rd->nd_rd_cksum = 0;
2697 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2698 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2699 
2700 	/* send the packet to outside... */
2701 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2702 	if (outif) {
2703 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2704 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2705 	}
2706 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2707 
2708 	return;
2709 
2710 fail:
2711 	if (m)
2712 		m_freem(m);
2713 	if (m0)
2714 		m_freem(m0);
2715 }
2716 
2717 /*
2718  * ICMPv6 socket option processing.
2719  */
2720 int
2721 icmp6_ctloutput(so, sopt)
2722 	struct socket *so;
2723 	struct sockopt *sopt;
2724 {
2725 	int error = 0;
2726 	int optlen;
2727 	struct inpcb *inp = sotoinpcb(so);
2728 	int level, op, optname;
2729 
2730 	if (sopt) {
2731 		level = sopt->sopt_level;
2732 		op = sopt->sopt_dir;
2733 		optname = sopt->sopt_name;
2734 		optlen = sopt->sopt_valsize;
2735 	} else
2736 		level = op = optname = optlen = 0;
2737 
2738 	if (level != IPPROTO_ICMPV6) {
2739 		return EINVAL;
2740 	}
2741 
2742 	switch (op) {
2743 	case PRCO_SETOPT:
2744 		switch (optname) {
2745 		case ICMP6_FILTER:
2746 		    {
2747 			struct icmp6_filter *p;
2748 
2749 			if (optlen != sizeof(*p)) {
2750 				error = EMSGSIZE;
2751 				break;
2752 			}
2753 			if (inp->in6p_icmp6filt == NULL) {
2754 				error = EINVAL;
2755 				break;
2756 			}
2757 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2758 				optlen);
2759 			break;
2760 		    }
2761 
2762 		default:
2763 			error = ENOPROTOOPT;
2764 			break;
2765 		}
2766 		break;
2767 
2768 	case PRCO_GETOPT:
2769 		switch (optname) {
2770 		case ICMP6_FILTER:
2771 		    {
2772 			if (inp->in6p_icmp6filt == NULL) {
2773 				error = EINVAL;
2774 				break;
2775 			}
2776 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2777 				sizeof(struct icmp6_filter));
2778 			break;
2779 		    }
2780 
2781 		default:
2782 			error = ENOPROTOOPT;
2783 			break;
2784 		}
2785 		break;
2786 	}
2787 
2788 	return (error);
2789 }
2790 
2791 /*
2792  * Perform rate limit check.
2793  * Returns 0 if it is okay to send the icmp6 packet.
2794  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2795  * limitation.
2796  *
2797  * XXX per-destination/type check necessary?
2798  */
2799 static int
2800 icmp6_ratelimit(dst, type, code)
2801 	const struct in6_addr *dst;	/* not used at this moment */
2802 	const int type;			/* not used at this moment */
2803 	const int code;			/* not used at this moment */
2804 {
2805 	int ret;
2806 
2807 	ret = 0;	/* okay to send */
2808 
2809 	/* PPS limit */
2810 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2811 	    icmp6errppslim)) {
2812 		/* The packet is subject to rate limit */
2813 		ret++;
2814 	}
2815 
2816 	return ret;
2817 }
2818