xref: /freebsd/sys/netinet6/icmp6.c (revision 0fa02ea5f786ef02befd46f8f083f48c8cd9630b)
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  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 
72 #include <sys/param.h>
73 #include <sys/domain.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86 
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/in_var.h>
95 #include <netinet/ip6.h>
96 #include <netinet/icmp6.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/in6_pcb.h>
100 #include <netinet6/ip6protosw.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/nd6.h>
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #include <netkey/key.h>
108 #endif
109 
110 #ifdef FAST_IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/key.h>
113 #endif
114 
115 #include <net/net_osdep.h>
116 
117 extern struct domain inet6domain;
118 
119 struct icmp6stat icmp6stat;
120 
121 extern struct inpcbhead ripcb;
122 extern int icmp6errppslim;
123 static int icmp6errpps_count = 0;
124 static struct timeval icmp6errppslim_last;
125 extern int icmp6_nodeinfo;
126 
127 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
128 static int icmp6_rip6_input __P((struct mbuf **, int));
129 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
130 static const char *icmp6_redirect_diag __P((struct in6_addr *,
131 	struct in6_addr *, struct in6_addr *));
132 static struct mbuf *ni6_input __P((struct mbuf *, int));
133 static struct mbuf *ni6_nametodns __P((const char *, int, int));
134 static int ni6_dnsmatch __P((const char *, int, const char *, int));
135 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
136 			  struct ifnet **, char *));
137 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
138 				struct ifnet *, int));
139 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
140 
141 #ifdef COMPAT_RFC1885
142 static struct route_in6 icmp6_reflect_rt;
143 #endif
144 
145 
146 void
147 icmp6_init()
148 {
149 	mld6_init();
150 }
151 
152 static void
153 icmp6_errcount(stat, type, code)
154 	struct icmp6errstat *stat;
155 	int type, code;
156 {
157 	switch (type) {
158 	case ICMP6_DST_UNREACH:
159 		switch (code) {
160 		case ICMP6_DST_UNREACH_NOROUTE:
161 			stat->icp6errs_dst_unreach_noroute++;
162 			return;
163 		case ICMP6_DST_UNREACH_ADMIN:
164 			stat->icp6errs_dst_unreach_admin++;
165 			return;
166 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
167 			stat->icp6errs_dst_unreach_beyondscope++;
168 			return;
169 		case ICMP6_DST_UNREACH_ADDR:
170 			stat->icp6errs_dst_unreach_addr++;
171 			return;
172 		case ICMP6_DST_UNREACH_NOPORT:
173 			stat->icp6errs_dst_unreach_noport++;
174 			return;
175 		}
176 		break;
177 	case ICMP6_PACKET_TOO_BIG:
178 		stat->icp6errs_packet_too_big++;
179 		return;
180 	case ICMP6_TIME_EXCEEDED:
181 		switch (code) {
182 		case ICMP6_TIME_EXCEED_TRANSIT:
183 			stat->icp6errs_time_exceed_transit++;
184 			return;
185 		case ICMP6_TIME_EXCEED_REASSEMBLY:
186 			stat->icp6errs_time_exceed_reassembly++;
187 			return;
188 		}
189 		break;
190 	case ICMP6_PARAM_PROB:
191 		switch (code) {
192 		case ICMP6_PARAMPROB_HEADER:
193 			stat->icp6errs_paramprob_header++;
194 			return;
195 		case ICMP6_PARAMPROB_NEXTHEADER:
196 			stat->icp6errs_paramprob_nextheader++;
197 			return;
198 		case ICMP6_PARAMPROB_OPTION:
199 			stat->icp6errs_paramprob_option++;
200 			return;
201 		}
202 		break;
203 	case ND_REDIRECT:
204 		stat->icp6errs_redirect++;
205 		return;
206 	}
207 	stat->icp6errs_unknown++;
208 }
209 
210 /*
211  * Generate an error packet of type error in response to bad IP6 packet.
212  */
213 void
214 icmp6_error(m, type, code, param)
215 	struct mbuf *m;
216 	int type, code, param;
217 {
218 	struct ip6_hdr *oip6, *nip6;
219 	struct icmp6_hdr *icmp6;
220 	u_int preplen;
221 	int off;
222 	int nxt;
223 
224 	icmp6stat.icp6s_error++;
225 
226 	/* count per-type-code statistics */
227 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
228 
229 #ifdef M_DECRYPTED	/*not openbsd*/
230 	if (m->m_flags & M_DECRYPTED) {
231 		icmp6stat.icp6s_canterror++;
232 		goto freeit;
233 	}
234 #endif
235 
236 #ifndef PULLDOWN_TEST
237 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
238 #else
239 	if (m->m_len < sizeof(struct ip6_hdr)) {
240 		m = m_pullup(m, sizeof(struct ip6_hdr));
241 		if (m == NULL)
242 			return;
243 	}
244 #endif
245 	oip6 = mtod(m, struct ip6_hdr *);
246 
247 	/*
248 	 * If the destination address of the erroneous packet is a multicast
249 	 * address, or the packet was sent using link-layer multicast,
250 	 * we should basically suppress sending an error (RFC 2463, Section
251 	 * 2.4).
252 	 * We have two exceptions (the item e.2 in that section):
253 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
254 	 * - the Parameter Problem Message that can be allowed an icmp6 error
255 	 *   in the option type field.  This check has been done in
256 	 *   ip6_unknown_opt(), so we can just check the type and code.
257 	 */
258 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
259 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
260 	    (type != ICMP6_PACKET_TOO_BIG &&
261 	     (type != ICMP6_PARAM_PROB ||
262 	      code != ICMP6_PARAMPROB_OPTION)))
263 		goto freeit;
264 
265 	/*
266 	 * RFC 2463, 2.4 (e.5): source address check.
267 	 * XXX: the case of anycast source?
268 	 */
269 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
270 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
271 		goto freeit;
272 
273 	/*
274 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
275 	 * don't do it.
276 	 */
277 	nxt = -1;
278 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
279 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
280 		struct icmp6_hdr *icp;
281 
282 #ifndef PULLDOWN_TEST
283 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
284 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
285 #else
286 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
287 			sizeof(*icp));
288 		if (icp == NULL) {
289 			icmp6stat.icp6s_tooshort++;
290 			return;
291 		}
292 #endif
293 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
294 		    icp->icmp6_type == ND_REDIRECT) {
295 			/*
296 			 * ICMPv6 error
297 			 * Special case: for redirect (which is
298 			 * informational) we must not send icmp6 error.
299 			 */
300 			icmp6stat.icp6s_canterror++;
301 			goto freeit;
302 		} else {
303 			/* ICMPv6 informational - send the error */
304 		}
305 	} else {
306 		/* non-ICMPv6 - send the error */
307 	}
308 
309 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
310 
311 	/* Finally, do rate limitation check. */
312 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
313 		icmp6stat.icp6s_toofreq++;
314 		goto freeit;
315 	}
316 
317 	/*
318 	 * OK, ICMP6 can be generated.
319 	 */
320 
321 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
322 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
323 
324 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
325 	M_PREPEND(m, preplen, M_DONTWAIT);
326 	if (m && m->m_len < preplen)
327 		m = m_pullup(m, preplen);
328 	if (m == NULL) {
329 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
330 		return;
331 	}
332 
333 	nip6 = mtod(m, struct ip6_hdr *);
334 	nip6->ip6_src  = oip6->ip6_src;
335 	nip6->ip6_dst  = oip6->ip6_dst;
336 
337 	in6_clearscope(&oip6->ip6_src);
338 	in6_clearscope(&oip6->ip6_dst);
339 
340 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
341 	icmp6->icmp6_type = type;
342 	icmp6->icmp6_code = code;
343 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
344 
345 	/*
346 	 * icmp6_reflect() is designed to be in the input path.
347 	 * icmp6_error() can be called from both input and outut path,
348 	 * and if we are in output path rcvif could contain bogus value.
349 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
350 	 * information in ip header (nip6).
351 	 */
352 	m->m_pkthdr.rcvif = NULL;
353 
354 	icmp6stat.icp6s_outhist[type]++;
355 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
356 
357 	return;
358 
359   freeit:
360 	/*
361 	 * If we can't tell wheter or not we can generate ICMP6, free it.
362 	 */
363 	m_freem(m);
364 }
365 
366 /*
367  * Process a received ICMP6 message.
368  */
369 int
370 icmp6_input(mp, offp, proto)
371 	struct mbuf **mp;
372 	int *offp, proto;
373 {
374 	struct mbuf *m = *mp, *n;
375 	struct ip6_hdr *ip6, *nip6;
376 	struct icmp6_hdr *icmp6, *nicmp6;
377 	int off = *offp;
378 	int icmp6len = m->m_pkthdr.len - *offp;
379 	int code, sum, noff;
380 
381 #ifndef PULLDOWN_TEST
382 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
383 	/* m might change if M_LOOP.  So, call mtod after this */
384 #endif
385 
386 	/*
387 	 * Locate icmp6 structure in mbuf, and check
388 	 * that not corrupted and of at least minimum length
389 	 */
390 
391 	ip6 = mtod(m, struct ip6_hdr *);
392 	if (icmp6len < sizeof(struct icmp6_hdr)) {
393 		icmp6stat.icp6s_tooshort++;
394 		goto freeit;
395 	}
396 
397 	/*
398 	 * calculate the checksum
399 	 */
400 #ifndef PULLDOWN_TEST
401 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
402 #else
403 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
404 	if (icmp6 == NULL) {
405 		icmp6stat.icp6s_tooshort++;
406 		return IPPROTO_DONE;
407 	}
408 #endif
409 	code = icmp6->icmp6_code;
410 
411 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
412 		nd6log((LOG_ERR,
413 		    "ICMP6 checksum error(%d|%x) %s\n",
414 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
415 		icmp6stat.icp6s_checksum++;
416 		goto freeit;
417 	}
418 
419 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
420 		/*
421 		 * Deliver very specific ICMP6 type only.
422 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
423 		 * will not work.
424 		 */
425 		switch (icmp6->icmp6_type) {
426 		case ICMP6_DST_UNREACH:
427 		case ICMP6_PACKET_TOO_BIG:
428 		case ICMP6_TIME_EXCEEDED:
429 			break;
430 		default:
431 			goto freeit;
432 		}
433 	}
434 
435 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
436 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
437 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
438 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
439 
440 	switch (icmp6->icmp6_type) {
441 	case ICMP6_DST_UNREACH:
442 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
443 		switch (code) {
444 		case ICMP6_DST_UNREACH_NOROUTE:
445 			code = PRC_UNREACH_NET;
446 			break;
447 		case ICMP6_DST_UNREACH_ADMIN:
448 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
449 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
450 			break;
451 		case ICMP6_DST_UNREACH_ADDR:
452 			code = PRC_HOSTDEAD;
453 			break;
454 #ifdef COMPAT_RFC1885
455 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
456 			code = PRC_UNREACH_SRCFAIL;
457 			break;
458 #else
459 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
460 			/* I mean "source address was incorrect." */
461 			code = PRC_PARAMPROB;
462 			break;
463 #endif
464 		case ICMP6_DST_UNREACH_NOPORT:
465 			code = PRC_UNREACH_PORT;
466 			break;
467 		default:
468 			goto badcode;
469 		}
470 		goto deliver;
471 		break;
472 
473 	case ICMP6_PACKET_TOO_BIG:
474 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
475 		if (code != 0)
476 			goto badcode;
477 
478 		code = PRC_MSGSIZE;
479 
480 		/*
481 		 * Updating the path MTU will be done after examining
482 		 * intermediate extension headers.
483 		 */
484 		goto deliver;
485 		break;
486 
487 	case ICMP6_TIME_EXCEEDED:
488 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
489 		switch (code) {
490 		case ICMP6_TIME_EXCEED_TRANSIT:
491 		case ICMP6_TIME_EXCEED_REASSEMBLY:
492 			code += PRC_TIMXCEED_INTRANS;
493 			break;
494 		default:
495 			goto badcode;
496 		}
497 		goto deliver;
498 		break;
499 
500 	case ICMP6_PARAM_PROB:
501 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
502 		switch (code) {
503 		case ICMP6_PARAMPROB_NEXTHEADER:
504 			code = PRC_UNREACH_PROTOCOL;
505 			break;
506 		case ICMP6_PARAMPROB_HEADER:
507 		case ICMP6_PARAMPROB_OPTION:
508 			code = PRC_PARAMPROB;
509 			break;
510 		default:
511 			goto badcode;
512 		}
513 		goto deliver;
514 		break;
515 
516 	case ICMP6_ECHO_REQUEST:
517 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
518 		if (code != 0)
519 			goto badcode;
520 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
521 			/* Give up remote */
522 			break;
523 		}
524 		if ((n->m_flags & M_EXT) != 0
525 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
526 			struct mbuf *n0 = n;
527 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
528 			int n0len;
529 
530 			/*
531 			 * Prepare an internal mbuf.  m_pullup() doesn't
532 			 * always copy the length we specified.
533 			 */
534 			if (maxlen >= MCLBYTES) {
535 				/* Give up remote */
536 				m_freem(n0);
537 				break;
538 			}
539 			MGETHDR(n, M_DONTWAIT, n0->m_type);
540 			n0len = n0->m_pkthdr.len;	/* save for use below */
541 			if (n)
542 				M_MOVE_PKTHDR(n, n0);
543 			if (n && maxlen >= MHLEN) {
544 				MCLGET(n, M_DONTWAIT);
545 				if ((n->m_flags & M_EXT) == 0) {
546 					m_free(n);
547 					n = NULL;
548 				}
549 			}
550 			if (n == NULL) {
551 				/* Give up remote */
552 				m_freem(n0);
553 				break;
554 			}
555 			/*
556 			 * Copy IPv6 and ICMPv6 only.
557 			 */
558 			nip6 = mtod(n, struct ip6_hdr *);
559 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
560 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
561 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
562 			noff = sizeof(struct ip6_hdr);
563 			/* new mbuf contains only ipv6+icmpv6 headers */
564 			n->m_len = noff + sizeof(struct icmp6_hdr);
565 			/*
566 			 * Adjust mbuf.  ip6_plen will be adjusted in
567 			 * ip6_output().
568 			 */
569 			m_adj(n0, off + sizeof(struct icmp6_hdr));
570 			/* recalculate complete packet size */
571 			n->m_pkthdr.len = n0len + (noff - off);
572 			n->m_next = n0;
573 		} else {
574 			nip6 = mtod(n, struct ip6_hdr *);
575 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
576 			noff = off;
577 		}
578 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
579 		nicmp6->icmp6_code = 0;
580 		if (n) {
581 			icmp6stat.icp6s_reflect++;
582 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
583 			icmp6_reflect(n, noff);
584 		}
585 		break;
586 
587 	case ICMP6_ECHO_REPLY:
588 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
589 		if (code != 0)
590 			goto badcode;
591 		break;
592 
593 	case MLD_LISTENER_QUERY:
594 	case MLD_LISTENER_REPORT:
595 		if (icmp6len < sizeof(struct mld_hdr))
596 			goto badlen;
597 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
598 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
599 		else
600 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
601 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
602 			/* give up local */
603 			mld6_input(m, off);
604 			m = NULL;
605 			goto freeit;
606 		}
607 		mld6_input(n, off);
608 		/* m stays. */
609 		break;
610 
611 	case MLD_LISTENER_DONE:
612 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
613 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
614 			goto badlen;
615 		break;		/* nothing to be done in kernel */
616 
617 	case MLD_MTRACE_RESP:
618 	case MLD_MTRACE:
619 		/* XXX: these two are experimental.  not officially defined. */
620 		/* XXX: per-interface statistics? */
621 		break;		/* just pass it to applications */
622 
623 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
624 	    {
625 		enum { WRU, FQDN } mode;
626 
627 		if (!icmp6_nodeinfo)
628 			break;
629 
630 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
631 			mode = WRU;
632 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
633 			mode = FQDN;
634 		else
635 			goto badlen;
636 
637 #define hostnamelen	strlen(hostname)
638 		if (mode == FQDN) {
639 #ifndef PULLDOWN_TEST
640 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
641 			    IPPROTO_DONE);
642 #endif
643 			n = m_copy(m, 0, M_COPYALL);
644 			if (n)
645 				n = ni6_input(n, off);
646 			/* XXX meaningless if n == NULL */
647 			noff = sizeof(struct ip6_hdr);
648 		} else {
649 			u_char *p;
650 			int maxlen, maxhlen;
651 
652 			if ((icmp6_nodeinfo & 5) != 5)
653 				break;
654 
655 			if (code != 0)
656 				goto badcode;
657 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
658 			if (maxlen >= MCLBYTES) {
659 				/* Give up remote */
660 				break;
661 			}
662 			MGETHDR(n, M_DONTWAIT, m->m_type);
663 			if (n && maxlen > MHLEN) {
664 				MCLGET(n, M_DONTWAIT);
665 				if ((n->m_flags & M_EXT) == 0) {
666 					m_free(n);
667 					n = NULL;
668 				}
669 			}
670 			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
671 				/*
672 				 * Previous code did a blind M_COPY_PKTHDR
673 				 * and said "just for rcvif".  If true, then
674 				 * we could tolerate the dup failing (due to
675 				 * the deep copy of the tag chain).  For now
676 				 * be conservative and just fail.
677 				 */
678 				m_free(n);
679 				n = NULL;
680 			}
681 			if (n == NULL) {
682 				/* Give up remote */
683 				break;
684 			}
685 			n->m_pkthdr.rcvif = NULL;
686 			n->m_len = 0;
687 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
688 			if (maxhlen > hostnamelen)
689 				maxhlen = hostnamelen;
690 			/*
691 			 * Copy IPv6 and ICMPv6 only.
692 			 */
693 			nip6 = mtod(n, struct ip6_hdr *);
694 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
695 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
696 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
697 			p = (u_char *)(nicmp6 + 1);
698 			bzero(p, 4);
699 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
700 			noff = sizeof(struct ip6_hdr);
701 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
702 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
703 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
704 			nicmp6->icmp6_code = 0;
705 		}
706 #undef hostnamelen
707 		if (n) {
708 			icmp6stat.icp6s_reflect++;
709 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
710 			icmp6_reflect(n, noff);
711 		}
712 		break;
713 	    }
714 
715 	case ICMP6_WRUREPLY:
716 		if (code != 0)
717 			goto badcode;
718 		break;
719 
720 	case ND_ROUTER_SOLICIT:
721 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
722 		if (code != 0)
723 			goto badcode;
724 		if (icmp6len < sizeof(struct nd_router_solicit))
725 			goto badlen;
726 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
727 			/* give up local */
728 			nd6_rs_input(m, off, icmp6len);
729 			m = NULL;
730 			goto freeit;
731 		}
732 		nd6_rs_input(n, off, icmp6len);
733 		/* m stays. */
734 		break;
735 
736 	case ND_ROUTER_ADVERT:
737 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
738 		if (code != 0)
739 			goto badcode;
740 		if (icmp6len < sizeof(struct nd_router_advert))
741 			goto badlen;
742 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
743 			/* give up local */
744 			nd6_ra_input(m, off, icmp6len);
745 			m = NULL;
746 			goto freeit;
747 		}
748 		nd6_ra_input(n, off, icmp6len);
749 		/* m stays. */
750 		break;
751 
752 	case ND_NEIGHBOR_SOLICIT:
753 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
754 		if (code != 0)
755 			goto badcode;
756 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
757 			goto badlen;
758 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
759 			/* give up local */
760 			nd6_ns_input(m, off, icmp6len);
761 			m = NULL;
762 			goto freeit;
763 		}
764 		nd6_ns_input(n, off, icmp6len);
765 		/* m stays. */
766 		break;
767 
768 	case ND_NEIGHBOR_ADVERT:
769 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
770 		if (code != 0)
771 			goto badcode;
772 		if (icmp6len < sizeof(struct nd_neighbor_advert))
773 			goto badlen;
774 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
775 			/* give up local */
776 			nd6_na_input(m, off, icmp6len);
777 			m = NULL;
778 			goto freeit;
779 		}
780 		nd6_na_input(n, off, icmp6len);
781 		/* m stays. */
782 		break;
783 
784 	case ND_REDIRECT:
785 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
786 		if (code != 0)
787 			goto badcode;
788 		if (icmp6len < sizeof(struct nd_redirect))
789 			goto badlen;
790 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
791 			/* give up local */
792 			icmp6_redirect_input(m, off);
793 			m = NULL;
794 			goto freeit;
795 		}
796 		icmp6_redirect_input(n, off);
797 		/* m stays. */
798 		break;
799 
800 	case ICMP6_ROUTER_RENUMBERING:
801 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
802 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
803 			goto badcode;
804 		if (icmp6len < sizeof(struct icmp6_router_renum))
805 			goto badlen;
806 		break;
807 
808 	default:
809 		nd6log((LOG_DEBUG,
810 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
811 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
812 		    ip6_sprintf(&ip6->ip6_dst),
813 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
814 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
815 			/* ICMPv6 error: MUST deliver it by spec... */
816 			code = PRC_NCMDS;
817 			/* deliver */
818 		} else {
819 			/* ICMPv6 informational: MUST not deliver */
820 			break;
821 		}
822 	deliver:
823 		if (icmp6_notify_error(m, off, icmp6len, code)) {
824 			/* In this case, m should've been freed. */
825 			return (IPPROTO_DONE);
826 		}
827 		break;
828 
829 	badcode:
830 		icmp6stat.icp6s_badcode++;
831 		break;
832 
833 	badlen:
834 		icmp6stat.icp6s_badlen++;
835 		break;
836 	}
837 
838 	/* deliver the packet to appropriate sockets */
839 	icmp6_rip6_input(&m, *offp);
840 
841 	return IPPROTO_DONE;
842 
843  freeit:
844 	m_freem(m);
845 	return IPPROTO_DONE;
846 }
847 
848 static int
849 icmp6_notify_error(m, off, icmp6len, code)
850 	struct mbuf *m;
851 	int off, icmp6len, code;
852 {
853 	struct icmp6_hdr *icmp6;
854 	struct ip6_hdr *eip6;
855 	u_int32_t notifymtu;
856 	struct sockaddr_in6 icmp6src, icmp6dst;
857 
858 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
859 		icmp6stat.icp6s_tooshort++;
860 		goto freeit;
861 	}
862 #ifndef PULLDOWN_TEST
863 	IP6_EXTHDR_CHECK(m, off,
864 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
865 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
866 #else
867 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
868 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
869 	if (icmp6 == NULL) {
870 		icmp6stat.icp6s_tooshort++;
871 		return (-1);
872 	}
873 #endif
874 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
875 
876 	/* Detect the upper level protocol */
877 	{
878 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
879 		u_int8_t nxt = eip6->ip6_nxt;
880 		int eoff = off + sizeof(struct icmp6_hdr) +
881 		    sizeof(struct ip6_hdr);
882 		struct ip6ctlparam ip6cp;
883 		struct in6_addr *finaldst = NULL;
884 		int icmp6type = icmp6->icmp6_type;
885 		struct ip6_frag *fh;
886 		struct ip6_rthdr *rth;
887 		struct ip6_rthdr0 *rth0;
888 		int rthlen;
889 
890 		while (1) { /* XXX: should avoid infinite loop explicitly? */
891 			struct ip6_ext *eh;
892 
893 			switch (nxt) {
894 			case IPPROTO_HOPOPTS:
895 			case IPPROTO_DSTOPTS:
896 			case IPPROTO_AH:
897 #ifndef PULLDOWN_TEST
898 				IP6_EXTHDR_CHECK(m, 0,
899 				    eoff + sizeof(struct ip6_ext), -1);
900 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
901 #else
902 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
903 				    eoff, sizeof(*eh));
904 				if (eh == NULL) {
905 					icmp6stat.icp6s_tooshort++;
906 					return (-1);
907 				}
908 #endif
909 
910 				if (nxt == IPPROTO_AH)
911 					eoff += (eh->ip6e_len + 2) << 2;
912 				else
913 					eoff += (eh->ip6e_len + 1) << 3;
914 				nxt = eh->ip6e_nxt;
915 				break;
916 			case IPPROTO_ROUTING:
917 				/*
918 				 * When the erroneous packet contains a
919 				 * routing header, we should examine the
920 				 * header to determine the final destination.
921 				 * Otherwise, we can't properly update
922 				 * information that depends on the final
923 				 * destination (e.g. path MTU).
924 				 */
925 #ifndef PULLDOWN_TEST
926 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
927 				rth = (struct ip6_rthdr *)
928 				    (mtod(m, caddr_t) + eoff);
929 #else
930 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
931 				    eoff, sizeof(*rth));
932 				if (rth == NULL) {
933 					icmp6stat.icp6s_tooshort++;
934 					return (-1);
935 				}
936 #endif
937 				rthlen = (rth->ip6r_len + 1) << 3;
938 				/*
939 				 * XXX: currently there is no
940 				 * officially defined type other
941 				 * than type-0.
942 				 * Note that if the segment left field
943 				 * is 0, all intermediate hops must
944 				 * have been passed.
945 				 */
946 				if (rth->ip6r_segleft &&
947 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
948 					int hops;
949 
950 #ifndef PULLDOWN_TEST
951 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
952 					rth0 = (struct ip6_rthdr0 *)
953 					    (mtod(m, caddr_t) + eoff);
954 #else
955 					IP6_EXTHDR_GET(rth0,
956 					    struct ip6_rthdr0 *, m,
957 					    eoff, rthlen);
958 					if (rth0 == NULL) {
959 						icmp6stat.icp6s_tooshort++;
960 						return (-1);
961 					}
962 #endif
963 					/* just ignore a bogus header */
964 					if ((rth0->ip6r0_len % 2) == 0 &&
965 					    (hops = rth0->ip6r0_len/2))
966 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
967 				}
968 				eoff += rthlen;
969 				nxt = rth->ip6r_nxt;
970 				break;
971 			case IPPROTO_FRAGMENT:
972 #ifndef PULLDOWN_TEST
973 				IP6_EXTHDR_CHECK(m, 0, eoff +
974 				    sizeof(struct ip6_frag), -1);
975 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
976 				    eoff);
977 #else
978 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
979 				    eoff, sizeof(*fh));
980 				if (fh == NULL) {
981 					icmp6stat.icp6s_tooshort++;
982 					return (-1);
983 				}
984 #endif
985 				/*
986 				 * Data after a fragment header is meaningless
987 				 * unless it is the first fragment, but
988 				 * we'll go to the notify label for path MTU
989 				 * discovery.
990 				 */
991 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
992 					goto notify;
993 
994 				eoff += sizeof(struct ip6_frag);
995 				nxt = fh->ip6f_nxt;
996 				break;
997 			default:
998 				/*
999 				 * This case includes ESP and the No Next
1000 				 * Header.  In such cases going to the notify
1001 				 * label does not have any meaning
1002 				 * (i.e. ctlfunc will be NULL), but we go
1003 				 * anyway since we might have to update
1004 				 * path MTU information.
1005 				 */
1006 				goto notify;
1007 			}
1008 		}
1009 	  notify:
1010 #ifndef PULLDOWN_TEST
1011 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1012 #else
1013 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1014 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1015 		if (icmp6 == NULL) {
1016 			icmp6stat.icp6s_tooshort++;
1017 			return (-1);
1018 		}
1019 #endif
1020 
1021 		/*
1022 		 * retrieve parameters from the inner IPv6 header, and convert
1023 		 * them into sockaddr structures.
1024 		 * XXX: there is no guarantee that the source or destination
1025 		 * addresses of the inner packet are in the same scope as
1026 		 * the addresses of the icmp packet.  But there is no other
1027 		 * way to determine the zone.
1028 		 */
1029 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1030 
1031 		bzero(&icmp6dst, sizeof(icmp6dst));
1032 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1033 		icmp6dst.sin6_family = AF_INET6;
1034 		if (finaldst == NULL)
1035 			icmp6dst.sin6_addr = eip6->ip6_dst;
1036 		else
1037 			icmp6dst.sin6_addr = *finaldst;
1038 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1039 		    &icmp6dst.sin6_scope_id))
1040 			goto freeit;
1041 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1042 				   NULL, NULL)) {
1043 			/* should be impossbile */
1044 			nd6log((LOG_DEBUG,
1045 			    "icmp6_notify_error: in6_embedscope failed\n"));
1046 			goto freeit;
1047 		}
1048 
1049 		/*
1050 		 * retrieve parameters from the inner IPv6 header, and convert
1051 		 * them into sockaddr structures.
1052 		 */
1053 		bzero(&icmp6src, sizeof(icmp6src));
1054 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1055 		icmp6src.sin6_family = AF_INET6;
1056 		icmp6src.sin6_addr = eip6->ip6_src;
1057 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1058 		    &icmp6src.sin6_scope_id)) {
1059 			goto freeit;
1060 		}
1061 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1062 				   NULL, NULL)) {
1063 			/* should be impossbile */
1064 			nd6log((LOG_DEBUG,
1065 			    "icmp6_notify_error: in6_embedscope failed\n"));
1066 			goto freeit;
1067 		}
1068 		icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1069 
1070 		if (finaldst == NULL)
1071 			finaldst = &eip6->ip6_dst;
1072 		ip6cp.ip6c_m = m;
1073 		ip6cp.ip6c_icmp6 = icmp6;
1074 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1075 		ip6cp.ip6c_off = eoff;
1076 		ip6cp.ip6c_finaldst = finaldst;
1077 		ip6cp.ip6c_src = &icmp6src;
1078 		ip6cp.ip6c_nxt = nxt;
1079 
1080 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1081 			notifymtu = ntohl(icmp6->icmp6_mtu);
1082 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1083 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1084 		}
1085 
1086 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1087 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1088 		if (ctlfunc) {
1089 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1090 			    &ip6cp);
1091 		}
1092 	}
1093 	return (0);
1094 
1095   freeit:
1096 	m_freem(m);
1097 	return (-1);
1098 }
1099 
1100 void
1101 icmp6_mtudisc_update(ip6cp, validated)
1102 	struct ip6ctlparam *ip6cp;
1103 	int validated;
1104 {
1105 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1106 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1107 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1108 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1109 	struct in_conninfo inc;
1110 
1111 #if 0
1112 	/*
1113 	 * RFC2460 section 5, last paragraph.
1114 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1115 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1116 	 * due to packet translator in the middle.
1117 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1118 	 * special handling.
1119 	 */
1120 	if (mtu < IPV6_MMTU)
1121 		return;
1122 #endif
1123 
1124 	/*
1125 	 * we reject ICMPv6 too big with abnormally small value.
1126 	 * XXX what is the good definition of "abnormally small"?
1127 	 */
1128 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1129 		return;
1130 
1131 	if (!validated)
1132 		return;
1133 
1134 	bzero(&inc, sizeof(inc));
1135 	inc.inc_flags = 1; /* IPv6 */
1136 	inc.inc6_faddr = *dst;
1137 	/* XXX normally, this won't happen */
1138 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1139 		inc.inc6_faddr.s6_addr16[1] =
1140 		    htons(m->m_pkthdr.rcvif->if_index);
1141 	}
1142 
1143 	if (mtu >= IPV6_MMTU) {
1144 		tcp_hc_updatemtu(&inc, mtu);
1145 		icmp6stat.icp6s_pmtuchg++;
1146 	}
1147 }
1148 
1149 /*
1150  * Process a Node Information Query packet, based on
1151  * draft-ietf-ipngwg-icmp-name-lookups-07.
1152  *
1153  * Spec incompatibilities:
1154  * - IPv6 Subject address handling
1155  * - IPv4 Subject address handling support missing
1156  * - Proxy reply (answer even if it's not for me)
1157  * - joins NI group address at in6_ifattach() time only, does not cope
1158  *   with hostname changes by sethostname(3)
1159  */
1160 #define hostnamelen	strlen(hostname)
1161 static struct mbuf *
1162 ni6_input(m, off)
1163 	struct mbuf *m;
1164 	int off;
1165 {
1166 	struct icmp6_nodeinfo *ni6, *nni6;
1167 	struct mbuf *n = NULL;
1168 	u_int16_t qtype;
1169 	int subjlen;
1170 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1171 	struct ni_reply_fqdn *fqdn;
1172 	int addrs;		/* for NI_QTYPE_NODEADDR */
1173 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1174 	struct sockaddr_in6 sin6_sbj; /* subject address */
1175 	struct sockaddr_in6 sin6_d;
1176 	struct ip6_hdr *ip6;
1177 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1178 	char *subj = NULL;
1179 	struct in6_ifaddr *ia6 = NULL;
1180 
1181 	ip6 = mtod(m, struct ip6_hdr *);
1182 #ifndef PULLDOWN_TEST
1183 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1184 #else
1185 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1186 	if (ni6 == NULL) {
1187 		/* m is already reclaimed */
1188 		return (NULL);
1189 	}
1190 #endif
1191 
1192 	bzero(&sin6_d, sizeof(sin6_d));
1193 	sin6_d.sin6_family = AF_INET6; /* not used, actually */
1194 	sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1195 	sin6_d.sin6_addr = ip6->ip6_dst;
1196 	if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_dst,
1197 	    &sin6_d.sin6_scope_id)) {
1198 		goto bad;
1199 	}
1200 	if (in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL))
1201 		goto bad; /* XXX should not happen */
1202 
1203 	/*
1204 	 * Validate IPv6 destination address.
1205 	 *
1206 	 * The Responder must discard the Query without further processing
1207 	 * unless it is one of the Responder's unicast or anycast addresses, or
1208 	 * a link-local scope multicast address which the Responder has joined.
1209 	 * [icmp-name-lookups-08, Section 4.]
1210 	 */
1211 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1212 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1213 			goto bad;
1214 		/* else it's a link-local multicast, fine */
1215 	} else {		/* unicast or anycast */
1216 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1217 			goto bad; /* XXX impossible */
1218 
1219 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1220 		    !(icmp6_nodeinfo & 4)) {
1221 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1222 				"a temporary address in %s:%d",
1223 			       __FILE__, __LINE__));
1224 			goto bad;
1225 		}
1226 	}
1227 
1228 	/* validate query Subject field. */
1229 	qtype = ntohs(ni6->ni_qtype);
1230 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1231 	switch (qtype) {
1232 	case NI_QTYPE_NOOP:
1233 	case NI_QTYPE_SUPTYPES:
1234 		/* 07 draft */
1235 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1236 			break;
1237 		/* FALLTHROUGH */
1238 	case NI_QTYPE_FQDN:
1239 	case NI_QTYPE_NODEADDR:
1240 		switch (ni6->ni_code) {
1241 		case ICMP6_NI_SUBJ_IPV6:
1242 #if ICMP6_NI_SUBJ_IPV6 != 0
1243 		case 0:
1244 #endif
1245 			/*
1246 			 * backward compatibility - try to accept 03 draft
1247 			 * format, where no Subject is present.
1248 			 */
1249 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1250 			    subjlen == 0) {
1251 				oldfqdn++;
1252 				break;
1253 			}
1254 #if ICMP6_NI_SUBJ_IPV6 != 0
1255 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1256 				goto bad;
1257 #endif
1258 
1259 			if (subjlen != sizeof(struct in6_addr))
1260 				goto bad;
1261 
1262 			/*
1263 			 * Validate Subject address.
1264 			 *
1265 			 * Not sure what exactly "address belongs to the node"
1266 			 * means in the spec, is it just unicast, or what?
1267 			 *
1268 			 * At this moment we consider Subject address as
1269 			 * "belong to the node" if the Subject address equals
1270 			 * to the IPv6 destination address; validation for
1271 			 * IPv6 destination address should have done enough
1272 			 * check for us.
1273 			 *
1274 			 * We do not do proxy at this moment.
1275 			 */
1276 			/* m_pulldown instead of copy? */
1277 			bzero(&sin6_sbj, sizeof(sin6_sbj));
1278 			sin6_sbj.sin6_family = AF_INET6;
1279 			sin6_sbj.sin6_len = sizeof(sin6_sbj);
1280 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1281 			    subjlen, (caddr_t)&sin6_sbj.sin6_addr);
1282 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1283 			    &sin6_sbj.sin6_addr, &sin6_sbj.sin6_scope_id)) {
1284 				goto bad;
1285 			}
1286 			if (in6_embedscope(&sin6_sbj.sin6_addr, &sin6_sbj,
1287 			    NULL, NULL))
1288 				goto bad; /* XXX should not happen */
1289 
1290 			subj = (char *)&sin6_sbj;
1291 			if (SA6_ARE_ADDR_EQUAL(&sin6_sbj, &sin6_d))
1292 				break;
1293 
1294 			/*
1295 			 * XXX if we are to allow other cases, we should really
1296 			 * be careful about scope here.
1297 			 * basically, we should disallow queries toward IPv6
1298 			 * destination X with subject Y,
1299 			 * if scope(X) > scope(Y).
1300 			 * if we allow scope(X) > scope(Y), it will result in
1301 			 * information leakage across scope boundary.
1302 			 */
1303 			goto bad;
1304 
1305 		case ICMP6_NI_SUBJ_FQDN:
1306 			/*
1307 			 * Validate Subject name with gethostname(3).
1308 			 *
1309 			 * The behavior may need some debate, since:
1310 			 * - we are not sure if the node has FQDN as
1311 			 *   hostname (returned by gethostname(3)).
1312 			 * - the code does wildcard match for truncated names.
1313 			 *   however, we are not sure if we want to perform
1314 			 *   wildcard match, if gethostname(3) side has
1315 			 *   truncated hostname.
1316 			 */
1317 			n = ni6_nametodns(hostname, hostnamelen, 0);
1318 			if (!n || n->m_next || n->m_len == 0)
1319 				goto bad;
1320 			IP6_EXTHDR_GET(subj, char *, m,
1321 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1322 			if (subj == NULL)
1323 				goto bad;
1324 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1325 			    n->m_len)) {
1326 				goto bad;
1327 			}
1328 			m_freem(n);
1329 			n = NULL;
1330 			break;
1331 
1332 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1333 		default:
1334 			goto bad;
1335 		}
1336 		break;
1337 	}
1338 
1339 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1340 	switch (qtype) {
1341 	case NI_QTYPE_FQDN:
1342 		if ((icmp6_nodeinfo & 1) == 0)
1343 			goto bad;
1344 		break;
1345 	case NI_QTYPE_NODEADDR:
1346 		if ((icmp6_nodeinfo & 2) == 0)
1347 			goto bad;
1348 		break;
1349 	}
1350 
1351 	/* guess reply length */
1352 	switch (qtype) {
1353 	case NI_QTYPE_NOOP:
1354 		break;		/* no reply data */
1355 	case NI_QTYPE_SUPTYPES:
1356 		replylen += sizeof(u_int32_t);
1357 		break;
1358 	case NI_QTYPE_FQDN:
1359 		/* XXX will append an mbuf */
1360 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1361 		break;
1362 	case NI_QTYPE_NODEADDR:
1363 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1364 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1365 		    sizeof(u_int32_t))) > MCLBYTES)
1366 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1367 		break;
1368 	default:
1369 		/*
1370 		 * XXX: We must return a reply with the ICMP6 code
1371 		 * `unknown Qtype' in this case.  However we regard the case
1372 		 * as an FQDN query for backward compatibility.
1373 		 * Older versions set a random value to this field,
1374 		 * so it rarely varies in the defined qtypes.
1375 		 * But the mechanism is not reliable...
1376 		 * maybe we should obsolete older versions.
1377 		 */
1378 		qtype = NI_QTYPE_FQDN;
1379 		/* XXX will append an mbuf */
1380 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1381 		oldfqdn++;
1382 		break;
1383 	}
1384 
1385 	/* allocate an mbuf to reply. */
1386 	MGETHDR(n, M_DONTWAIT, m->m_type);
1387 	if (n == NULL) {
1388 		m_freem(m);
1389 		return (NULL);
1390 	}
1391 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1392 	if (replylen > MHLEN) {
1393 		if (replylen > MCLBYTES) {
1394 			/*
1395 			 * XXX: should we try to allocate more? But MCLBYTES
1396 			 * is probably much larger than IPV6_MMTU...
1397 			 */
1398 			goto bad;
1399 		}
1400 		MCLGET(n, M_DONTWAIT);
1401 		if ((n->m_flags & M_EXT) == 0) {
1402 			goto bad;
1403 		}
1404 	}
1405 	n->m_pkthdr.len = n->m_len = replylen;
1406 
1407 	/* copy mbuf header and IPv6 + Node Information base headers */
1408 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1409 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1410 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1411 
1412 	/* qtype dependent procedure */
1413 	switch (qtype) {
1414 	case NI_QTYPE_NOOP:
1415 		nni6->ni_code = ICMP6_NI_SUCCESS;
1416 		nni6->ni_flags = 0;
1417 		break;
1418 	case NI_QTYPE_SUPTYPES:
1419 	{
1420 		u_int32_t v;
1421 		nni6->ni_code = ICMP6_NI_SUCCESS;
1422 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1423 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1424 		v = (u_int32_t)htonl(0x0000000f);
1425 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1426 		break;
1427 	}
1428 	case NI_QTYPE_FQDN:
1429 		nni6->ni_code = ICMP6_NI_SUCCESS;
1430 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1431 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1432 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1433 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1434 		/*
1435 		 * XXX do we really have FQDN in variable "hostname"?
1436 		 */
1437 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1438 		if (n->m_next == NULL)
1439 			goto bad;
1440 		/* XXX we assume that n->m_next is not a chain */
1441 		if (n->m_next->m_next != NULL)
1442 			goto bad;
1443 		n->m_pkthdr.len += n->m_next->m_len;
1444 		break;
1445 	case NI_QTYPE_NODEADDR:
1446 	{
1447 		int lenlim, copied;
1448 
1449 		nni6->ni_code = ICMP6_NI_SUCCESS;
1450 		n->m_pkthdr.len = n->m_len =
1451 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1452 		lenlim = M_TRAILINGSPACE(n);
1453 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1454 		/* XXX: reset mbuf length */
1455 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1456 		    sizeof(struct icmp6_nodeinfo) + copied;
1457 		break;
1458 	}
1459 	default:
1460 		break;		/* XXX impossible! */
1461 	}
1462 
1463 	nni6->ni_type = ICMP6_NI_REPLY;
1464 	m_freem(m);
1465 	return (n);
1466 
1467   bad:
1468 	m_freem(m);
1469 	if (n)
1470 		m_freem(n);
1471 	return (NULL);
1472 }
1473 #undef hostnamelen
1474 
1475 /*
1476  * make a mbuf with DNS-encoded string.  no compression support.
1477  *
1478  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1479  * treated as truncated name (two \0 at the end).  this is a wild guess.
1480  */
1481 static struct mbuf *
1482 ni6_nametodns(name, namelen, old)
1483 	const char *name;
1484 	int namelen;
1485 	int old;	/* return pascal string if non-zero */
1486 {
1487 	struct mbuf *m;
1488 	char *cp, *ep;
1489 	const char *p, *q;
1490 	int i, len, nterm;
1491 
1492 	if (old)
1493 		len = namelen + 1;
1494 	else
1495 		len = MCLBYTES;
1496 
1497 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1498 	MGET(m, M_DONTWAIT, MT_DATA);
1499 	if (m && len > MLEN) {
1500 		MCLGET(m, M_DONTWAIT);
1501 		if ((m->m_flags & M_EXT) == 0)
1502 			goto fail;
1503 	}
1504 	if (!m)
1505 		goto fail;
1506 	m->m_next = NULL;
1507 
1508 	if (old) {
1509 		m->m_len = len;
1510 		*mtod(m, char *) = namelen;
1511 		bcopy(name, mtod(m, char *) + 1, namelen);
1512 		return m;
1513 	} else {
1514 		m->m_len = 0;
1515 		cp = mtod(m, char *);
1516 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1517 
1518 		/* if not certain about my name, return empty buffer */
1519 		if (namelen == 0)
1520 			return m;
1521 
1522 		/*
1523 		 * guess if it looks like shortened hostname, or FQDN.
1524 		 * shortened hostname needs two trailing "\0".
1525 		 */
1526 		i = 0;
1527 		for (p = name; p < name + namelen; p++) {
1528 			if (*p && *p == '.')
1529 				i++;
1530 		}
1531 		if (i < 2)
1532 			nterm = 2;
1533 		else
1534 			nterm = 1;
1535 
1536 		p = name;
1537 		while (cp < ep && p < name + namelen) {
1538 			i = 0;
1539 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1540 				i++;
1541 			/* result does not fit into mbuf */
1542 			if (cp + i + 1 >= ep)
1543 				goto fail;
1544 			/*
1545 			 * DNS label length restriction, RFC1035 page 8.
1546 			 * "i == 0" case is included here to avoid returning
1547 			 * 0-length label on "foo..bar".
1548 			 */
1549 			if (i <= 0 || i >= 64)
1550 				goto fail;
1551 			*cp++ = i;
1552 			bcopy(p, cp, i);
1553 			cp += i;
1554 			p = q;
1555 			if (p < name + namelen && *p == '.')
1556 				p++;
1557 		}
1558 		/* termination */
1559 		if (cp + nterm >= ep)
1560 			goto fail;
1561 		while (nterm-- > 0)
1562 			*cp++ = '\0';
1563 		m->m_len = cp - mtod(m, char *);
1564 		return m;
1565 	}
1566 
1567 	panic("should not reach here");
1568 	/* NOTREACHED */
1569 
1570  fail:
1571 	if (m)
1572 		m_freem(m);
1573 	return NULL;
1574 }
1575 
1576 /*
1577  * check if two DNS-encoded string matches.  takes care of truncated
1578  * form (with \0\0 at the end).  no compression support.
1579  * XXX upper/lowercase match (see RFC2065)
1580  */
1581 static int
1582 ni6_dnsmatch(a, alen, b, blen)
1583 	const char *a;
1584 	int alen;
1585 	const char *b;
1586 	int blen;
1587 {
1588 	const char *a0, *b0;
1589 	int l;
1590 
1591 	/* simplest case - need validation? */
1592 	if (alen == blen && bcmp(a, b, alen) == 0)
1593 		return 1;
1594 
1595 	a0 = a;
1596 	b0 = b;
1597 
1598 	/* termination is mandatory */
1599 	if (alen < 2 || blen < 2)
1600 		return 0;
1601 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1602 		return 0;
1603 	alen--;
1604 	blen--;
1605 
1606 	while (a - a0 < alen && b - b0 < blen) {
1607 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1608 			return 0;
1609 
1610 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1611 			return 0;
1612 		/* we don't support compression yet */
1613 		if (a[0] >= 64 || b[0] >= 64)
1614 			return 0;
1615 
1616 		/* truncated case */
1617 		if (a[0] == 0 && a - a0 == alen - 1)
1618 			return 1;
1619 		if (b[0] == 0 && b - b0 == blen - 1)
1620 			return 1;
1621 		if (a[0] == 0 || b[0] == 0)
1622 			return 0;
1623 
1624 		if (a[0] != b[0])
1625 			return 0;
1626 		l = a[0];
1627 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1628 			return 0;
1629 		if (bcmp(a + 1, b + 1, l) != 0)
1630 			return 0;
1631 
1632 		a += 1 + l;
1633 		b += 1 + l;
1634 	}
1635 
1636 	if (a - a0 == alen && b - b0 == blen)
1637 		return 1;
1638 	else
1639 		return 0;
1640 }
1641 
1642 /*
1643  * calculate the number of addresses to be returned in the node info reply.
1644  */
1645 static int
1646 ni6_addrs(ni6, m, ifpp, subj)
1647 	struct icmp6_nodeinfo *ni6;
1648 	struct mbuf *m;
1649 	struct ifnet **ifpp;
1650 	char *subj;
1651 {
1652 	struct ifnet *ifp;
1653 	struct in6_ifaddr *ifa6;
1654 	struct ifaddr *ifa;
1655 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1656 	int addrs = 0, addrsofif, iffound = 0;
1657 	int niflags = ni6->ni_flags;
1658 
1659 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1660 		switch (ni6->ni_code) {
1661 		case ICMP6_NI_SUBJ_IPV6:
1662 			if (subj == NULL) /* must be impossible... */
1663 				return (0);
1664 			subj_ip6 = (struct sockaddr_in6 *)subj;
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_ip6->sin6_addr,
1685 			    &ifa6->ia_addr.sin6_addr))
1686 				iffound = 1;
1687 
1688 			/*
1689 			 * IPv4-mapped addresses can only be returned by a
1690 			 * Node Information proxy, since they represent
1691 			 * addresses of IPv4-only nodes, which perforce do
1692 			 * not implement this protocol.
1693 			 * [icmp-name-lookups-07, Section 5.4]
1694 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1695 			 * this function at this moment.
1696 			 */
1697 
1698 			/* What do we have to do about ::1? */
1699 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1700 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1701 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1702 					continue;
1703 				break;
1704 			case IPV6_ADDR_SCOPE_SITELOCAL:
1705 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1706 					continue;
1707 				break;
1708 			case IPV6_ADDR_SCOPE_GLOBAL:
1709 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1710 					continue;
1711 				break;
1712 			default:
1713 				continue;
1714 			}
1715 
1716 			/*
1717 			 * check if anycast is okay.
1718 			 * XXX: just experimental.  not in the spec.
1719 			 */
1720 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1721 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1722 				continue; /* we need only unicast addresses */
1723 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1724 			    (icmp6_nodeinfo & 4) == 0) {
1725 				continue;
1726 			}
1727 			addrsofif++; /* count the address */
1728 		}
1729 		if (iffound) {
1730 			*ifpp = ifp;
1731 			IFNET_RUNLOCK();
1732 			return (addrsofif);
1733 		}
1734 
1735 		addrs += addrsofif;
1736 	}
1737 	IFNET_RUNLOCK();
1738 
1739 	return (addrs);
1740 }
1741 
1742 static int
1743 ni6_store_addrs(ni6, nni6, ifp0, resid)
1744 	struct icmp6_nodeinfo *ni6, *nni6;
1745 	struct ifnet *ifp0;
1746 	int resid;
1747 {
1748 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1749 	struct in6_ifaddr *ifa6;
1750 	struct ifaddr *ifa;
1751 	struct ifnet *ifp_dep = NULL;
1752 	int copied = 0, allow_deprecated = 0;
1753 	u_char *cp = (u_char *)(nni6 + 1);
1754 	int niflags = ni6->ni_flags;
1755 	u_int32_t ltime;
1756 
1757 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1758 		return (0);	/* needless to copy */
1759 
1760 	IFNET_RLOCK();
1761   again:
1762 
1763 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1764 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1765 		     ifa = ifa->ifa_list.tqe_next) {
1766 			if (ifa->ifa_addr->sa_family != AF_INET6)
1767 				continue;
1768 			ifa6 = (struct in6_ifaddr *)ifa;
1769 
1770 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1771 			    allow_deprecated == 0) {
1772 				/*
1773 				 * prefererred address should be put before
1774 				 * deprecated addresses.
1775 				 */
1776 
1777 				/* record the interface for later search */
1778 				if (ifp_dep == NULL)
1779 					ifp_dep = ifp;
1780 
1781 				continue;
1782 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1783 			    allow_deprecated != 0)
1784 				continue; /* we now collect deprecated addrs */
1785 
1786 			/* What do we have to do about ::1? */
1787 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1788 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1789 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1790 					continue;
1791 				break;
1792 			case IPV6_ADDR_SCOPE_SITELOCAL:
1793 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1794 					continue;
1795 				break;
1796 			case IPV6_ADDR_SCOPE_GLOBAL:
1797 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1798 					continue;
1799 				break;
1800 			default:
1801 				continue;
1802 			}
1803 
1804 			/*
1805 			 * check if anycast is okay.
1806 			 * XXX: just experimental.  not in the spec.
1807 			 */
1808 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1809 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1810 				continue;
1811 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1812 			    (icmp6_nodeinfo & 4) == 0) {
1813 				continue;
1814 			}
1815 
1816 			/* now we can copy the address */
1817 			if (resid < sizeof(struct in6_addr) +
1818 			    sizeof(u_int32_t)) {
1819 				/*
1820 				 * We give up much more copy.
1821 				 * Set the truncate flag and return.
1822 				 */
1823 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1824 				IFNET_RUNLOCK();
1825 				return (copied);
1826 			}
1827 
1828 			/*
1829 			 * Set the TTL of the address.
1830 			 * The TTL value should be one of the following
1831 			 * according to the specification:
1832 			 *
1833 			 * 1. The remaining lifetime of a DHCP lease on the
1834 			 *    address, or
1835 			 * 2. The remaining Valid Lifetime of a prefix from
1836 			 *    which the address was derived through Stateless
1837 			 *    Autoconfiguration.
1838 			 *
1839 			 * Note that we currently do not support stateful
1840 			 * address configuration by DHCPv6, so the former
1841 			 * case can't happen.
1842 			 */
1843 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1844 				ltime = ND6_INFINITE_LIFETIME;
1845 			else {
1846 				if (ifa6->ia6_lifetime.ia6t_expire >
1847 				    time_second)
1848 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1849 				else
1850 					ltime = 0;
1851 			}
1852 
1853 			bcopy(&ltime, cp, sizeof(u_int32_t));
1854 			cp += sizeof(u_int32_t);
1855 
1856 			/* copy the address itself */
1857 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1858 			    sizeof(struct in6_addr));
1859 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1860 			cp += sizeof(struct in6_addr);
1861 
1862 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1863 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1864 		}
1865 		if (ifp0)	/* we need search only on the specified IF */
1866 			break;
1867 	}
1868 
1869 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1870 		ifp = ifp_dep;
1871 		allow_deprecated = 1;
1872 
1873 		goto again;
1874 	}
1875 
1876 	IFNET_RUNLOCK();
1877 
1878 	return (copied);
1879 }
1880 
1881 /*
1882  * XXX almost dup'ed code with rip6_input.
1883  */
1884 static int
1885 icmp6_rip6_input(mp, off)
1886 	struct	mbuf **mp;
1887 	int	off;
1888 {
1889 	struct mbuf *m = *mp;
1890 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1891 	struct in6pcb *in6p;
1892 	struct in6pcb *last = NULL;
1893 	struct sockaddr_in6 fromsa;
1894 	struct icmp6_hdr *icmp6;
1895 	struct mbuf *opts = NULL;
1896 
1897 #ifndef PULLDOWN_TEST
1898 	/* this is assumed to be safe. */
1899 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1900 #else
1901 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1902 	if (icmp6 == NULL) {
1903 		/* m is already reclaimed */
1904 		return (IPPROTO_DONE);
1905 	}
1906 #endif
1907 
1908 	bzero(&fromsa, sizeof(fromsa));
1909 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1910 	fromsa.sin6_family = AF_INET6;
1911 	/* KAME hack: recover scopeid */
1912 	(void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1913 
1914 	LIST_FOREACH(in6p, &ripcb, inp_list) {
1915 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1916 			continue;
1917 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1918 			continue;
1919 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1920 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1921 			continue;
1922 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1923 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1924 			continue;
1925 		if (in6p->in6p_icmp6filt
1926 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1927 				 in6p->in6p_icmp6filt))
1928 			continue;
1929 		if (last) {
1930 			struct	mbuf *n = NULL;
1931 
1932 			/*
1933 			 * Recent network drivers tend to allocate a single
1934 			 * mbuf cluster, rather than to make a couple of
1935 			 * mbufs without clusters.  Also, since the IPv6 code
1936 			 * path tries to avoid m_pullup(), it is highly
1937 			 * probable that we still have an mbuf cluster here
1938 			 * even though the necessary length can be stored in an
1939 			 * mbuf's internal buffer.
1940 			 * Meanwhile, the default size of the receive socket
1941 			 * buffer for raw sockets is not so large.  This means
1942 			 * the possibility of packet loss is relatively higher
1943 			 * than before.  To avoid this scenario, we copy the
1944 			 * received data to a separate mbuf that does not use
1945 			 * a cluster, if possible.
1946 			 * XXX: it is better to copy the data after stripping
1947 			 * intermediate headers.
1948 			 */
1949 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1950 			    m->m_len <= MHLEN) {
1951 				MGET(n, M_DONTWAIT, m->m_type);
1952 				if (n != NULL) {
1953 					m_dup_pkthdr(n, m, M_NOWAIT);
1954 					bcopy(m->m_data, n->m_data, m->m_len);
1955 					n->m_len = m->m_len;
1956 				}
1957 			}
1958 			if (n != NULL ||
1959 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1960 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1961 					ip6_savecontrol(last, n, &opts);
1962 				/* strip intermediate headers */
1963 				m_adj(n, off);
1964 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1965 				    (struct sockaddr *)&fromsa, n, opts)
1966 				    == 0) {
1967 					/* should notify about lost packet */
1968 					m_freem(n);
1969 					if (opts) {
1970 						m_freem(opts);
1971 					}
1972 				} else
1973 					sorwakeup(last->in6p_socket);
1974 				opts = NULL;
1975 			}
1976 		}
1977 		last = in6p;
1978 	}
1979 	if (last) {
1980 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1981 			ip6_savecontrol(last, m, &opts);
1982 		/* strip intermediate headers */
1983 		m_adj(m, off);
1984 
1985 		/* avoid using mbuf clusters if possible (see above) */
1986 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1987 		    m->m_len <= MHLEN) {
1988 			struct mbuf *n;
1989 
1990 			MGET(n, M_DONTWAIT, m->m_type);
1991 			if (n != NULL) {
1992 				m_dup_pkthdr(n, m, M_NOWAIT);
1993 				bcopy(m->m_data, n->m_data, m->m_len);
1994 				n->m_len = m->m_len;
1995 
1996 				m_freem(m);
1997 				m = n;
1998 			}
1999 		}
2000 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2001 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2002 			m_freem(m);
2003 			if (opts)
2004 				m_freem(opts);
2005 		} else
2006 			sorwakeup(last->in6p_socket);
2007 	} else {
2008 		m_freem(m);
2009 		ip6stat.ip6s_delivered--;
2010 	}
2011 	return IPPROTO_DONE;
2012 }
2013 
2014 /*
2015  * Reflect the ip6 packet back to the source.
2016  * OFF points to the icmp6 header, counted from the top of the mbuf.
2017  */
2018 void
2019 icmp6_reflect(m, off)
2020 	struct	mbuf *m;
2021 	size_t off;
2022 {
2023 	struct ip6_hdr *ip6;
2024 	struct icmp6_hdr *icmp6;
2025 	struct in6_ifaddr *ia;
2026 	struct in6_addr t, *src = 0;
2027 	int plen;
2028 	int type, code;
2029 	struct ifnet *outif = NULL;
2030 	struct sockaddr_in6 sa6_src, sa6_dst;
2031 #ifdef COMPAT_RFC1885
2032 	int mtu = IPV6_MMTU;
2033 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2034 #endif
2035 
2036 	/* too short to reflect */
2037 	if (off < sizeof(struct ip6_hdr)) {
2038 		nd6log((LOG_DEBUG,
2039 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2040 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2041 		    __FILE__, __LINE__));
2042 		goto bad;
2043 	}
2044 
2045 	/*
2046 	 * If there are extra headers between IPv6 and ICMPv6, strip
2047 	 * off that header first.
2048 	 */
2049 #ifdef DIAGNOSTIC
2050 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2051 		panic("assumption failed in icmp6_reflect");
2052 #endif
2053 	if (off > sizeof(struct ip6_hdr)) {
2054 		size_t l;
2055 		struct ip6_hdr nip6;
2056 
2057 		l = off - sizeof(struct ip6_hdr);
2058 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2059 		m_adj(m, l);
2060 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2061 		if (m->m_len < l) {
2062 			if ((m = m_pullup(m, l)) == NULL)
2063 				return;
2064 		}
2065 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2066 	} else /* off == sizeof(struct ip6_hdr) */ {
2067 		size_t l;
2068 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2069 		if (m->m_len < l) {
2070 			if ((m = m_pullup(m, l)) == NULL)
2071 				return;
2072 		}
2073 	}
2074 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2075 	ip6 = mtod(m, struct ip6_hdr *);
2076 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2077 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2078 	type = icmp6->icmp6_type; /* keep type for statistics */
2079 	code = icmp6->icmp6_code; /* ditto. */
2080 
2081 	t = ip6->ip6_dst;
2082 	/*
2083 	 * ip6_input() drops a packet if its src is multicast.
2084 	 * So, the src is never multicast.
2085 	 */
2086 	ip6->ip6_dst = ip6->ip6_src;
2087 
2088 	/*
2089 	 * XXX: make sure to embed scope zone information, using
2090 	 * already embedded IDs or the received interface (if any).
2091 	 * Note that rcvif may be NULL.
2092 	 * TODO: scoped routing case (XXX).
2093 	 */
2094 	bzero(&sa6_src, sizeof(sa6_src));
2095 	sa6_src.sin6_family = AF_INET6;
2096 	sa6_src.sin6_len = sizeof(sa6_src);
2097 	sa6_src.sin6_addr = ip6->ip6_dst;
2098 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2099 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2100 	bzero(&sa6_dst, sizeof(sa6_dst));
2101 	sa6_dst.sin6_family = AF_INET6;
2102 	sa6_dst.sin6_len = sizeof(sa6_dst);
2103 	sa6_dst.sin6_addr = t;
2104 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2105 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2106 
2107 #ifdef COMPAT_RFC1885
2108 	/*
2109 	 * xxx guess MTU
2110 	 * RFC 1885 requires that echo reply should be truncated if it
2111 	 * does not fit in with (return) path MTU, but the description was
2112 	 * removed in the new spec.
2113 	 */
2114 	if (icmp6_reflect_rt.ro_rt == 0 ||
2115 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2116 		if (icmp6_reflect_rt.ro_rt) {
2117 			RTFREE(icmp6_reflect_rt.ro_rt);
2118 			icmp6_reflect_rt.ro_rt = 0;
2119 		}
2120 		bzero(sin6, sizeof(*sin6));
2121 		sin6->sin6_family = PF_INET6;
2122 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2123 		sin6->sin6_addr = ip6->ip6_dst;
2124 
2125 		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2126 	}
2127 
2128 	if (icmp6_reflect_rt.ro_rt == 0)
2129 		goto bad;
2130 
2131 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2132 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2133 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2134 
2135 	if (mtu < m->m_pkthdr.len) {
2136 		plen -= (m->m_pkthdr.len - mtu);
2137 		m_adj(m, mtu - m->m_pkthdr.len);
2138 	}
2139 #endif
2140 	/*
2141 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2142 	 * use dst as the src for the reply.
2143 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2144 	 * (for example) when we encounter an error while forwarding procedure
2145 	 * destined to a duplicated address of ours.
2146 	 */
2147 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2148 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2149 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2150 			src = &t;
2151 			break;
2152 		}
2153 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2154 		/*
2155 		 * This is the case if the dst is our link-local address
2156 		 * and the sender is also ourselves.
2157 		 */
2158 		src = &t;
2159 	}
2160 
2161 	if (src == 0) {
2162 		int e;
2163 		struct route_in6 ro;
2164 
2165 		/*
2166 		 * This case matches to multicasts, our anycast, or unicasts
2167 		 * that we do not own.  Select a source address based on the
2168 		 * source address of the erroneous packet.
2169 		 */
2170 		bzero(&ro, sizeof(ro));
2171 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2172 		if (ro.ro_rt)
2173 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2174 		if (src == NULL) {
2175 			nd6log((LOG_DEBUG,
2176 			    "icmp6_reflect: source can't be determined: "
2177 			    "dst=%s, error=%d\n",
2178 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2179 			goto bad;
2180 		}
2181 	}
2182 
2183 	ip6->ip6_src = *src;
2184 
2185 	ip6->ip6_flow = 0;
2186 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2187 	ip6->ip6_vfc |= IPV6_VERSION;
2188 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2189 	if (outif)
2190 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2191 	else if (m->m_pkthdr.rcvif) {
2192 		/* XXX: This may not be the outgoing interface */
2193 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2194 	} else
2195 		ip6->ip6_hlim = ip6_defhlim;
2196 
2197 	icmp6->icmp6_cksum = 0;
2198 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2199 	    sizeof(struct ip6_hdr), plen);
2200 
2201 	/*
2202 	 * XXX option handling
2203 	 */
2204 
2205 	m->m_flags &= ~(M_BCAST|M_MCAST);
2206 #ifdef IPSEC
2207 	/* Don't lookup socket */
2208 	(void)ipsec_setsocket(m, NULL);
2209 #endif /* IPSEC */
2210 
2211 #ifdef COMPAT_RFC1885
2212 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2213 #else
2214 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2215 #endif
2216 	if (outif)
2217 		icmp6_ifoutstat_inc(outif, type, code);
2218 
2219 	return;
2220 
2221  bad:
2222 	m_freem(m);
2223 	return;
2224 }
2225 
2226 void
2227 icmp6_fasttimo()
2228 {
2229 
2230 	mld6_fasttimeo();
2231 }
2232 
2233 static const char *
2234 icmp6_redirect_diag(src6, dst6, tgt6)
2235 	struct in6_addr *src6;
2236 	struct in6_addr *dst6;
2237 	struct in6_addr *tgt6;
2238 {
2239 	static char buf[1024];
2240 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2241 	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2242 	return buf;
2243 }
2244 
2245 void
2246 icmp6_redirect_input(m, off)
2247 	struct mbuf *m;
2248 	int off;
2249 {
2250 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2251 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2252 	struct nd_redirect *nd_rd;
2253 	int icmp6len = ntohs(ip6->ip6_plen);
2254 	char *lladdr = NULL;
2255 	int lladdrlen = 0;
2256 	u_char *redirhdr = NULL;
2257 	int redirhdrlen = 0;
2258 	struct rtentry *rt = NULL;
2259 	int is_router;
2260 	int is_onlink;
2261 	struct in6_addr src6 = ip6->ip6_src;
2262 	struct in6_addr redtgt6;
2263 	struct in6_addr reddst6;
2264 	union nd_opts ndopts;
2265 
2266 	if (!m || !ifp)
2267 		return;
2268 
2269 	/* XXX if we are router, we don't update route by icmp6 redirect */
2270 	if (ip6_forwarding)
2271 		goto freeit;
2272 	if (!icmp6_rediraccept)
2273 		goto freeit;
2274 
2275 #ifndef PULLDOWN_TEST
2276 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2277 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2278 #else
2279 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2280 	if (nd_rd == NULL) {
2281 		icmp6stat.icp6s_tooshort++;
2282 		return;
2283 	}
2284 #endif
2285 	redtgt6 = nd_rd->nd_rd_target;
2286 	reddst6 = nd_rd->nd_rd_dst;
2287 
2288 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2289 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2290 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2291 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2292 
2293 	/* validation */
2294 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2295 		nd6log((LOG_ERR,
2296 		    "ICMP6 redirect sent from %s rejected; "
2297 		    "must be from linklocal\n",
2298 		    ip6_sprintf(&src6)));
2299 		goto bad;
2300 	}
2301 	if (ip6->ip6_hlim != 255) {
2302 		nd6log((LOG_ERR,
2303 		    "ICMP6 redirect sent from %s rejected; "
2304 		    "hlim=%d (must be 255)\n",
2305 		    ip6_sprintf(&src6), ip6->ip6_hlim));
2306 		goto bad;
2307 	}
2308     {
2309 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2310 	struct sockaddr_in6 sin6;
2311 	struct in6_addr *gw6;
2312 
2313 	bzero(&sin6, sizeof(sin6));
2314 	sin6.sin6_family = AF_INET6;
2315 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2316 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2317 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2318 	if (rt) {
2319 		if (rt->rt_gateway == NULL ||
2320 		    rt->rt_gateway->sa_family != AF_INET6) {
2321 			nd6log((LOG_ERR,
2322 			    "ICMP6 redirect rejected; no route "
2323 			    "with inet6 gateway found for redirect dst: %s\n",
2324 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2325 			RTFREE_LOCKED(rt);
2326 			goto bad;
2327 		}
2328 
2329 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2330 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2331 			nd6log((LOG_ERR,
2332 			    "ICMP6 redirect rejected; "
2333 			    "not equal to gw-for-src=%s (must be same): "
2334 			    "%s\n",
2335 			    ip6_sprintf(gw6),
2336 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2337 			RTFREE_LOCKED(rt);
2338 			goto bad;
2339 		}
2340 	} else {
2341 		nd6log((LOG_ERR,
2342 		    "ICMP6 redirect rejected; "
2343 		    "no route found for redirect dst: %s\n",
2344 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2345 		goto bad;
2346 	}
2347 	RTFREE_LOCKED(rt);
2348 	rt = NULL;
2349     }
2350 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2351 		nd6log((LOG_ERR,
2352 		    "ICMP6 redirect rejected; "
2353 		    "redirect dst must be unicast: %s\n",
2354 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2355 		goto bad;
2356 	}
2357 
2358 	is_router = is_onlink = 0;
2359 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2360 		is_router = 1;	/* router case */
2361 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2362 		is_onlink = 1;	/* on-link destination case */
2363 	if (!is_router && !is_onlink) {
2364 		nd6log((LOG_ERR,
2365 		    "ICMP6 redirect rejected; "
2366 		    "neither router case nor onlink case: %s\n",
2367 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2368 		goto bad;
2369 	}
2370 	/* validation passed */
2371 
2372 	icmp6len -= sizeof(*nd_rd);
2373 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2374 	if (nd6_options(&ndopts) < 0) {
2375 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2376 		    "invalid ND option, rejected: %s\n",
2377 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2378 		/* nd6_options have incremented stats */
2379 		goto freeit;
2380 	}
2381 
2382 	if (ndopts.nd_opts_tgt_lladdr) {
2383 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2384 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2385 	}
2386 
2387 	if (ndopts.nd_opts_rh) {
2388 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2389 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2390 	}
2391 
2392 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2393 		nd6log((LOG_INFO,
2394 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2395 		    "(if %d, icmp6 packet %d): %s\n",
2396 		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2397 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2398 		goto bad;
2399 	}
2400 
2401 	/* RFC 2461 8.3 */
2402 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2403 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2404 
2405 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2406 		/* perform rtredirect */
2407 		struct sockaddr_in6 sdst;
2408 		struct sockaddr_in6 sgw;
2409 		struct sockaddr_in6 ssrc;
2410 
2411 		bzero(&sdst, sizeof(sdst));
2412 		bzero(&sgw, sizeof(sgw));
2413 		bzero(&ssrc, sizeof(ssrc));
2414 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2415 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2416 			sizeof(struct sockaddr_in6);
2417 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2418 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2419 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2420 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2421 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2422 		    (struct sockaddr *)&ssrc);
2423 	}
2424 	/* finally update cached route in each socket via pfctlinput */
2425     {
2426 	struct sockaddr_in6 sdst;
2427 
2428 	bzero(&sdst, sizeof(sdst));
2429 	sdst.sin6_family = AF_INET6;
2430 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2431 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2432 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2433 #if defined(IPSEC) || defined(FAST_IPSEC)
2434 	key_sa_routechange((struct sockaddr *)&sdst);
2435 #endif
2436     }
2437 
2438  freeit:
2439 	m_freem(m);
2440 	return;
2441 
2442  bad:
2443 	icmp6stat.icp6s_badredirect++;
2444 	m_freem(m);
2445 }
2446 
2447 void
2448 icmp6_redirect_output(m0, rt)
2449 	struct mbuf *m0;
2450 	struct rtentry *rt;
2451 {
2452 	struct ifnet *ifp;	/* my outgoing interface */
2453 	struct in6_addr *ifp_ll6;
2454 	struct in6_addr *router_ll6;
2455 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2456 	struct mbuf *m = NULL;	/* newly allocated one */
2457 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2458 	struct nd_redirect *nd_rd;
2459 	size_t maxlen;
2460 	u_char *p;
2461 	struct ifnet *outif = NULL;
2462 	struct sockaddr_in6 src_sa;
2463 
2464 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2465 
2466 	/* if we are not router, we don't send icmp6 redirect */
2467 	if (!ip6_forwarding || ip6_accept_rtadv)
2468 		goto fail;
2469 
2470 	/* sanity check */
2471 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2472 		goto fail;
2473 
2474 	/*
2475 	 * Address check:
2476 	 *  the source address must identify a neighbor, and
2477 	 *  the destination address must not be a multicast address
2478 	 *  [RFC 2461, sec 8.2]
2479 	 */
2480 	sip6 = mtod(m0, struct ip6_hdr *);
2481 	bzero(&src_sa, sizeof(src_sa));
2482 	src_sa.sin6_family = AF_INET6;
2483 	src_sa.sin6_len = sizeof(src_sa);
2484 	src_sa.sin6_addr = sip6->ip6_src;
2485 	/* we don't currently use sin6_scope_id, but eventually use it */
2486 	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2487 		goto fail;
2488 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2489 		goto fail;
2490 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2491 		goto fail;	/* what should we do here? */
2492 
2493 	/* rate limit */
2494 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2495 		goto fail;
2496 
2497 	/*
2498 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2499 	 * we almost always ask for an mbuf cluster for simplicity.
2500 	 * (MHLEN < IPV6_MMTU is almost always true)
2501 	 */
2502 #if IPV6_MMTU >= MCLBYTES
2503 # error assumption failed about IPV6_MMTU and MCLBYTES
2504 #endif
2505 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2506 	if (m && IPV6_MMTU >= MHLEN)
2507 		MCLGET(m, M_DONTWAIT);
2508 	if (!m)
2509 		goto fail;
2510 	m->m_pkthdr.rcvif = NULL;
2511 	m->m_len = 0;
2512 	maxlen = M_TRAILINGSPACE(m);
2513 	maxlen = min(IPV6_MMTU, maxlen);
2514 	/* just for safety */
2515 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2516 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2517 		goto fail;
2518 	}
2519 
2520 	{
2521 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2522 		struct in6_ifaddr *ia;
2523 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2524 						 IN6_IFF_NOTREADY|
2525 						 IN6_IFF_ANYCAST)) == NULL)
2526 			goto fail;
2527 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2528 	}
2529 
2530 	/* get ip6 linklocal address for the router. */
2531 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2532 		struct sockaddr_in6 *sin6;
2533 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2534 		router_ll6 = &sin6->sin6_addr;
2535 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2536 			router_ll6 = (struct in6_addr *)NULL;
2537 	} else
2538 		router_ll6 = (struct in6_addr *)NULL;
2539 
2540 	/* ip6 */
2541 	ip6 = mtod(m, struct ip6_hdr *);
2542 	ip6->ip6_flow = 0;
2543 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2544 	ip6->ip6_vfc |= IPV6_VERSION;
2545 	/* ip6->ip6_plen will be set later */
2546 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2547 	ip6->ip6_hlim = 255;
2548 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2549 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2550 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2551 
2552 	/* ND Redirect */
2553 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2554 	nd_rd->nd_rd_type = ND_REDIRECT;
2555 	nd_rd->nd_rd_code = 0;
2556 	nd_rd->nd_rd_reserved = 0;
2557 	if (rt->rt_flags & RTF_GATEWAY) {
2558 		/*
2559 		 * nd_rd->nd_rd_target must be a link-local address in
2560 		 * better router cases.
2561 		 */
2562 		if (!router_ll6)
2563 			goto fail;
2564 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2565 		    sizeof(nd_rd->nd_rd_target));
2566 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2567 		    sizeof(nd_rd->nd_rd_dst));
2568 	} else {
2569 		/* make sure redtgt == reddst */
2570 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2571 		    sizeof(nd_rd->nd_rd_target));
2572 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2573 		    sizeof(nd_rd->nd_rd_dst));
2574 	}
2575 
2576 	p = (u_char *)(nd_rd + 1);
2577 
2578 	if (!router_ll6)
2579 		goto nolladdropt;
2580 
2581 	{
2582 		/* target lladdr option */
2583 		struct rtentry *rt_router = NULL;
2584 		int len;
2585 		struct sockaddr_dl *sdl;
2586 		struct nd_opt_hdr *nd_opt;
2587 		char *lladdr;
2588 
2589 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2590 		if (!rt_router)
2591 			goto nolladdropt;
2592 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2593 		len = (len + 7) & ~7;	/* round by 8 */
2594 		/* safety check */
2595 		if (len + (p - (u_char *)ip6) > maxlen)
2596 			goto nolladdropt;
2597 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2598 		    (rt_router->rt_flags & RTF_LLINFO) &&
2599 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2600 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2601 		    sdl->sdl_alen) {
2602 			nd_opt = (struct nd_opt_hdr *)p;
2603 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2604 			nd_opt->nd_opt_len = len >> 3;
2605 			lladdr = (char *)(nd_opt + 1);
2606 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2607 			p += len;
2608 		}
2609 	}
2610 nolladdropt:;
2611 
2612 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2613 
2614 	/* just to be safe */
2615 #ifdef M_DECRYPTED	/*not openbsd*/
2616 	if (m0->m_flags & M_DECRYPTED)
2617 		goto noredhdropt;
2618 #endif
2619 	if (p - (u_char *)ip6 > maxlen)
2620 		goto noredhdropt;
2621 
2622 	{
2623 		/* redirected header option */
2624 		int len;
2625 		struct nd_opt_rd_hdr *nd_opt_rh;
2626 
2627 		/*
2628 		 * compute the maximum size for icmp6 redirect header option.
2629 		 * XXX room for auth header?
2630 		 */
2631 		len = maxlen - (p - (u_char *)ip6);
2632 		len &= ~7;
2633 
2634 		/* This is just for simplicity. */
2635 		if (m0->m_pkthdr.len != m0->m_len) {
2636 			if (m0->m_next) {
2637 				m_freem(m0->m_next);
2638 				m0->m_next = NULL;
2639 			}
2640 			m0->m_pkthdr.len = m0->m_len;
2641 		}
2642 
2643 		/*
2644 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2645 		 * about padding/truncate rule for the original IP packet.
2646 		 * From the discussion on IPv6imp in Feb 1999,
2647 		 * the consensus was:
2648 		 * - "attach as much as possible" is the goal
2649 		 * - pad if not aligned (original size can be guessed by
2650 		 *   original ip6 header)
2651 		 * Following code adds the padding if it is simple enough,
2652 		 * and truncates if not.
2653 		 */
2654 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2655 			panic("assumption failed in %s:%d", __FILE__,
2656 			    __LINE__);
2657 
2658 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2659 			/* not enough room, truncate */
2660 			m0->m_pkthdr.len = m0->m_len = len -
2661 			    sizeof(*nd_opt_rh);
2662 		} else {
2663 			/* enough room, pad or truncate */
2664 			size_t extra;
2665 
2666 			extra = m0->m_pkthdr.len % 8;
2667 			if (extra) {
2668 				/* pad if easy enough, truncate if not */
2669 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2670 					/* pad */
2671 					m0->m_len += (8 - extra);
2672 					m0->m_pkthdr.len += (8 - extra);
2673 				} else {
2674 					/* truncate */
2675 					m0->m_pkthdr.len -= extra;
2676 					m0->m_len -= extra;
2677 				}
2678 			}
2679 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2680 			m0->m_pkthdr.len = m0->m_len = len -
2681 			    sizeof(*nd_opt_rh);
2682 		}
2683 
2684 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2685 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2686 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2687 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2688 		p += sizeof(*nd_opt_rh);
2689 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2690 
2691 		/* connect m0 to m */
2692 		m_tag_delete_chain(m0, NULL);
2693 		m0->m_flags &= ~M_PKTHDR;
2694 		m->m_next = m0;
2695 		m->m_pkthdr.len = m->m_len + m0->m_len;
2696 		m0 = NULL;
2697 	}
2698 noredhdropt:;
2699 	if (m0) {
2700 		m_freem(m0);
2701 		m0 = NULL;
2702 	}
2703 
2704 	/* XXX: clear embedded link IDs in the inner header */
2705 	in6_clearscope(&sip6->ip6_src);
2706 	in6_clearscope(&sip6->ip6_dst);
2707 	in6_clearscope(&nd_rd->nd_rd_target);
2708 	in6_clearscope(&nd_rd->nd_rd_dst);
2709 
2710 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2711 
2712 	nd_rd->nd_rd_cksum = 0;
2713 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2714 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2715 
2716 	/* send the packet to outside... */
2717 #ifdef IPSEC
2718 	/* Don't lookup socket */
2719 	(void)ipsec_setsocket(m, NULL);
2720 #endif /* IPSEC */
2721 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2722 	if (outif) {
2723 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2724 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2725 	}
2726 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2727 
2728 	return;
2729 
2730 fail:
2731 	if (m)
2732 		m_freem(m);
2733 	if (m0)
2734 		m_freem(m0);
2735 }
2736 
2737 /*
2738  * ICMPv6 socket option processing.
2739  */
2740 int
2741 icmp6_ctloutput(so, sopt)
2742 	struct socket *so;
2743 	struct sockopt *sopt;
2744 {
2745 	int error = 0;
2746 	int optlen;
2747 	struct inpcb *inp = sotoinpcb(so);
2748 	int level, op, optname;
2749 
2750 	if (sopt) {
2751 		level = sopt->sopt_level;
2752 		op = sopt->sopt_dir;
2753 		optname = sopt->sopt_name;
2754 		optlen = sopt->sopt_valsize;
2755 	} else
2756 		level = op = optname = optlen = 0;
2757 
2758 	if (level != IPPROTO_ICMPV6) {
2759 		return EINVAL;
2760 	}
2761 
2762 	switch (op) {
2763 	case PRCO_SETOPT:
2764 		switch (optname) {
2765 		case ICMP6_FILTER:
2766 		    {
2767 			struct icmp6_filter *p;
2768 
2769 			if (optlen != sizeof(*p)) {
2770 				error = EMSGSIZE;
2771 				break;
2772 			}
2773 			if (inp->in6p_icmp6filt == NULL) {
2774 				error = EINVAL;
2775 				break;
2776 			}
2777 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2778 				optlen);
2779 			break;
2780 		    }
2781 
2782 		default:
2783 			error = ENOPROTOOPT;
2784 			break;
2785 		}
2786 		break;
2787 
2788 	case PRCO_GETOPT:
2789 		switch (optname) {
2790 		case ICMP6_FILTER:
2791 		    {
2792 			if (inp->in6p_icmp6filt == NULL) {
2793 				error = EINVAL;
2794 				break;
2795 			}
2796 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2797 				sizeof(struct icmp6_filter));
2798 			break;
2799 		    }
2800 
2801 		default:
2802 			error = ENOPROTOOPT;
2803 			break;
2804 		}
2805 		break;
2806 	}
2807 
2808 	return (error);
2809 }
2810 
2811 /*
2812  * Perform rate limit check.
2813  * Returns 0 if it is okay to send the icmp6 packet.
2814  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2815  * limitation.
2816  *
2817  * XXX per-destination/type check necessary?
2818  */
2819 static int
2820 icmp6_ratelimit(dst, type, code)
2821 	const struct in6_addr *dst;	/* not used at this moment */
2822 	const int type;			/* not used at this moment */
2823 	const int code;			/* not used at this moment */
2824 {
2825 	int ret;
2826 
2827 	ret = 0;	/* okay to send */
2828 
2829 	/* PPS limit */
2830 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2831 	    icmp6errppslim)) {
2832 		/* The packet is subject to rate limit */
2833 		ret++;
2834 	}
2835 
2836 	return ret;
2837 }
2838