xref: /freebsd/sys/netinet6/icmp6.c (revision 7773002178c8dbc52b44e4d705f07706409af8e4)
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 <netinet6/in6_ifattach.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/mld6_var.h>
102 #include <netinet6/nd6.h>
103 
104 #ifdef IPSEC
105 #include <netinet6/ipsec.h>
106 #include <netkey/key.h>
107 #endif
108 
109 #ifdef FAST_IPSEC
110 #include <netipsec/ipsec.h>
111 #include <netipsec/key.h>
112 #endif
113 
114 #include <net/net_osdep.h>
115 
116 extern struct domain inet6domain;
117 
118 struct icmp6stat icmp6stat;
119 
120 extern struct inpcbhead ripcb;
121 extern int icmp6errppslim;
122 static int icmp6errpps_count = 0;
123 static struct timeval icmp6errppslim_last;
124 extern int icmp6_nodeinfo;
125 
126 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
127 static int icmp6_rip6_input __P((struct mbuf **, int));
128 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
129 static const char *icmp6_redirect_diag __P((struct in6_addr *,
130 	struct in6_addr *, struct in6_addr *));
131 static struct mbuf *ni6_input __P((struct mbuf *, int));
132 static struct mbuf *ni6_nametodns __P((const char *, int, int));
133 static int ni6_dnsmatch __P((const char *, int, const char *, int));
134 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
135 			  struct ifnet **, char *));
136 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
137 				struct ifnet *, int));
138 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
139 
140 #ifdef COMPAT_RFC1885
141 static struct route_in6 icmp6_reflect_rt;
142 #endif
143 
144 
145 void
146 icmp6_init()
147 {
148 	mld6_init();
149 }
150 
151 static void
152 icmp6_errcount(stat, type, code)
153 	struct icmp6errstat *stat;
154 	int type, code;
155 {
156 	switch (type) {
157 	case ICMP6_DST_UNREACH:
158 		switch (code) {
159 		case ICMP6_DST_UNREACH_NOROUTE:
160 			stat->icp6errs_dst_unreach_noroute++;
161 			return;
162 		case ICMP6_DST_UNREACH_ADMIN:
163 			stat->icp6errs_dst_unreach_admin++;
164 			return;
165 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
166 			stat->icp6errs_dst_unreach_beyondscope++;
167 			return;
168 		case ICMP6_DST_UNREACH_ADDR:
169 			stat->icp6errs_dst_unreach_addr++;
170 			return;
171 		case ICMP6_DST_UNREACH_NOPORT:
172 			stat->icp6errs_dst_unreach_noport++;
173 			return;
174 		}
175 		break;
176 	case ICMP6_PACKET_TOO_BIG:
177 		stat->icp6errs_packet_too_big++;
178 		return;
179 	case ICMP6_TIME_EXCEEDED:
180 		switch (code) {
181 		case ICMP6_TIME_EXCEED_TRANSIT:
182 			stat->icp6errs_time_exceed_transit++;
183 			return;
184 		case ICMP6_TIME_EXCEED_REASSEMBLY:
185 			stat->icp6errs_time_exceed_reassembly++;
186 			return;
187 		}
188 		break;
189 	case ICMP6_PARAM_PROB:
190 		switch (code) {
191 		case ICMP6_PARAMPROB_HEADER:
192 			stat->icp6errs_paramprob_header++;
193 			return;
194 		case ICMP6_PARAMPROB_NEXTHEADER:
195 			stat->icp6errs_paramprob_nextheader++;
196 			return;
197 		case ICMP6_PARAMPROB_OPTION:
198 			stat->icp6errs_paramprob_option++;
199 			return;
200 		}
201 		break;
202 	case ND_REDIRECT:
203 		stat->icp6errs_redirect++;
204 		return;
205 	}
206 	stat->icp6errs_unknown++;
207 }
208 
209 /*
210  * Generate an error packet of type error in response to bad IP6 packet.
211  */
212 void
213 icmp6_error(m, type, code, param)
214 	struct mbuf *m;
215 	int type, code, param;
216 {
217 	struct ip6_hdr *oip6, *nip6;
218 	struct icmp6_hdr *icmp6;
219 	u_int preplen;
220 	int off;
221 	int nxt;
222 
223 	icmp6stat.icp6s_error++;
224 
225 	/* count per-type-code statistics */
226 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
227 
228 #ifdef M_DECRYPTED	/*not openbsd*/
229 	if (m->m_flags & M_DECRYPTED) {
230 		icmp6stat.icp6s_canterror++;
231 		goto freeit;
232 	}
233 #endif
234 
235 #ifndef PULLDOWN_TEST
236 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
237 #else
238 	if (m->m_len < sizeof(struct ip6_hdr)) {
239 		m = m_pullup(m, sizeof(struct ip6_hdr));
240 		if (m == NULL)
241 			return;
242 	}
243 #endif
244 	oip6 = mtod(m, struct ip6_hdr *);
245 
246 	/*
247 	 * If the destination address of the erroneous packet is a multicast
248 	 * address, or the packet was sent using link-layer multicast,
249 	 * we should basically suppress sending an error (RFC 2463, Section
250 	 * 2.4).
251 	 * We have two exceptions (the item e.2 in that section):
252 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
253 	 * - the Parameter Problem Message that can be allowed an icmp6 error
254 	 *   in the option type field.  This check has been done in
255 	 *   ip6_unknown_opt(), so we can just check the type and code.
256 	 */
257 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
258 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
259 	    (type != ICMP6_PACKET_TOO_BIG &&
260 	     (type != ICMP6_PARAM_PROB ||
261 	      code != ICMP6_PARAMPROB_OPTION)))
262 		goto freeit;
263 
264 	/*
265 	 * RFC 2463, 2.4 (e.5): source address check.
266 	 * XXX: the case of anycast source?
267 	 */
268 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
269 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
270 		goto freeit;
271 
272 	/*
273 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
274 	 * don't do it.
275 	 */
276 	nxt = -1;
277 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
278 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
279 		struct icmp6_hdr *icp;
280 
281 #ifndef PULLDOWN_TEST
282 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
283 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
284 #else
285 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
286 			sizeof(*icp));
287 		if (icp == NULL) {
288 			icmp6stat.icp6s_tooshort++;
289 			return;
290 		}
291 #endif
292 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
293 		    icp->icmp6_type == ND_REDIRECT) {
294 			/*
295 			 * ICMPv6 error
296 			 * Special case: for redirect (which is
297 			 * informational) we must not send icmp6 error.
298 			 */
299 			icmp6stat.icp6s_canterror++;
300 			goto freeit;
301 		} else {
302 			/* ICMPv6 informational - send the error */
303 		}
304 	} else {
305 		/* non-ICMPv6 - send the error */
306 	}
307 
308 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
309 
310 	/* Finally, do rate limitation check. */
311 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
312 		icmp6stat.icp6s_toofreq++;
313 		goto freeit;
314 	}
315 
316 	/*
317 	 * OK, ICMP6 can be generated.
318 	 */
319 
320 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
321 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
322 
323 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
324 	M_PREPEND(m, preplen, M_DONTWAIT);
325 	if (m && m->m_len < preplen)
326 		m = m_pullup(m, preplen);
327 	if (m == NULL) {
328 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
329 		return;
330 	}
331 
332 	nip6 = mtod(m, struct ip6_hdr *);
333 	nip6->ip6_src  = oip6->ip6_src;
334 	nip6->ip6_dst  = oip6->ip6_dst;
335 
336 	in6_clearscope(&oip6->ip6_src);
337 	in6_clearscope(&oip6->ip6_dst);
338 
339 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
340 	icmp6->icmp6_type = type;
341 	icmp6->icmp6_code = code;
342 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
343 
344 	/*
345 	 * icmp6_reflect() is designed to be in the input path.
346 	 * icmp6_error() can be called from both input and outut path,
347 	 * and if we are in output path rcvif could contain bogus value.
348 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
349 	 * information in ip header (nip6).
350 	 */
351 	m->m_pkthdr.rcvif = NULL;
352 
353 	icmp6stat.icp6s_outhist[type]++;
354 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
355 
356 	return;
357 
358   freeit:
359 	/*
360 	 * If we can't tell wheter or not we can generate ICMP6, free it.
361 	 */
362 	m_freem(m);
363 }
364 
365 /*
366  * Process a received ICMP6 message.
367  */
368 int
369 icmp6_input(mp, offp, proto)
370 	struct mbuf **mp;
371 	int *offp, proto;
372 {
373 	struct mbuf *m = *mp, *n;
374 	struct ip6_hdr *ip6, *nip6;
375 	struct icmp6_hdr *icmp6, *nicmp6;
376 	int off = *offp;
377 	int icmp6len = m->m_pkthdr.len - *offp;
378 	int code, sum, noff;
379 
380 #ifndef PULLDOWN_TEST
381 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
382 	/* m might change if M_LOOP.  So, call mtod after this */
383 #endif
384 
385 	/*
386 	 * Locate icmp6 structure in mbuf, and check
387 	 * that not corrupted and of at least minimum length
388 	 */
389 
390 	ip6 = mtod(m, struct ip6_hdr *);
391 	if (icmp6len < sizeof(struct icmp6_hdr)) {
392 		icmp6stat.icp6s_tooshort++;
393 		goto freeit;
394 	}
395 
396 	/*
397 	 * calculate the checksum
398 	 */
399 #ifndef PULLDOWN_TEST
400 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
401 #else
402 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
403 	if (icmp6 == NULL) {
404 		icmp6stat.icp6s_tooshort++;
405 		return IPPROTO_DONE;
406 	}
407 #endif
408 	code = icmp6->icmp6_code;
409 
410 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
411 		nd6log((LOG_ERR,
412 		    "ICMP6 checksum error(%d|%x) %s\n",
413 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
414 		icmp6stat.icp6s_checksum++;
415 		goto freeit;
416 	}
417 
418 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
419 		/*
420 		 * Deliver very specific ICMP6 type only.
421 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
422 		 * will not work.
423 		 */
424 		switch (icmp6->icmp6_type) {
425 		case ICMP6_DST_UNREACH:
426 		case ICMP6_PACKET_TOO_BIG:
427 		case ICMP6_TIME_EXCEEDED:
428 			break;
429 		default:
430 			goto freeit;
431 		}
432 	}
433 
434 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
435 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
436 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
437 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
438 
439 	switch (icmp6->icmp6_type) {
440 	case ICMP6_DST_UNREACH:
441 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
442 		switch (code) {
443 		case ICMP6_DST_UNREACH_NOROUTE:
444 			code = PRC_UNREACH_NET;
445 			break;
446 		case ICMP6_DST_UNREACH_ADMIN:
447 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
448 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
449 			break;
450 		case ICMP6_DST_UNREACH_ADDR:
451 			code = PRC_HOSTDEAD;
452 			break;
453 #ifdef COMPAT_RFC1885
454 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
455 			code = PRC_UNREACH_SRCFAIL;
456 			break;
457 #else
458 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
459 			/* I mean "source address was incorrect." */
460 			code = PRC_PARAMPROB;
461 			break;
462 #endif
463 		case ICMP6_DST_UNREACH_NOPORT:
464 			code = PRC_UNREACH_PORT;
465 			break;
466 		default:
467 			goto badcode;
468 		}
469 		goto deliver;
470 		break;
471 
472 	case ICMP6_PACKET_TOO_BIG:
473 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
474 		if (code != 0)
475 			goto badcode;
476 
477 		code = PRC_MSGSIZE;
478 
479 		/*
480 		 * Updating the path MTU will be done after examining
481 		 * intermediate extension headers.
482 		 */
483 		goto deliver;
484 		break;
485 
486 	case ICMP6_TIME_EXCEEDED:
487 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
488 		switch (code) {
489 		case ICMP6_TIME_EXCEED_TRANSIT:
490 		case ICMP6_TIME_EXCEED_REASSEMBLY:
491 			code += PRC_TIMXCEED_INTRANS;
492 			break;
493 		default:
494 			goto badcode;
495 		}
496 		goto deliver;
497 		break;
498 
499 	case ICMP6_PARAM_PROB:
500 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
501 		switch (code) {
502 		case ICMP6_PARAMPROB_NEXTHEADER:
503 			code = PRC_UNREACH_PROTOCOL;
504 			break;
505 		case ICMP6_PARAMPROB_HEADER:
506 		case ICMP6_PARAMPROB_OPTION:
507 			code = PRC_PARAMPROB;
508 			break;
509 		default:
510 			goto badcode;
511 		}
512 		goto deliver;
513 		break;
514 
515 	case ICMP6_ECHO_REQUEST:
516 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
517 		if (code != 0)
518 			goto badcode;
519 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
520 			/* Give up remote */
521 			break;
522 		}
523 		if ((n->m_flags & M_EXT) != 0
524 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
525 			struct mbuf *n0 = n;
526 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
527 			int n0len;
528 
529 			/*
530 			 * Prepare an internal mbuf.  m_pullup() doesn't
531 			 * always copy the length we specified.
532 			 */
533 			if (maxlen >= MCLBYTES) {
534 				/* Give up remote */
535 				m_freem(n0);
536 				break;
537 			}
538 			MGETHDR(n, M_DONTWAIT, n0->m_type);
539 			n0len = n0->m_pkthdr.len;	/* save for use below */
540 			if (n)
541 				M_MOVE_PKTHDR(n, n0);
542 			if (n && maxlen >= MHLEN) {
543 				MCLGET(n, M_DONTWAIT);
544 				if ((n->m_flags & M_EXT) == 0) {
545 					m_free(n);
546 					n = NULL;
547 				}
548 			}
549 			if (n == NULL) {
550 				/* Give up remote */
551 				m_freem(n0);
552 				break;
553 			}
554 			/*
555 			 * Copy IPv6 and ICMPv6 only.
556 			 */
557 			nip6 = mtod(n, struct ip6_hdr *);
558 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
559 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
560 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
561 			noff = sizeof(struct ip6_hdr);
562 			/* new mbuf contains only ipv6+icmpv6 headers */
563 			n->m_len = noff + sizeof(struct icmp6_hdr);
564 			/*
565 			 * Adjust mbuf.  ip6_plen will be adjusted in
566 			 * ip6_output().
567 			 */
568 			m_adj(n0, off + sizeof(struct icmp6_hdr));
569 			/* recalculate complete packet size */
570 			n->m_pkthdr.len = n0len + (noff - off);
571 			n->m_next = n0;
572 		} else {
573 			nip6 = mtod(n, struct ip6_hdr *);
574 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
575 			noff = off;
576 		}
577 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
578 		nicmp6->icmp6_code = 0;
579 		if (n) {
580 			icmp6stat.icp6s_reflect++;
581 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
582 			icmp6_reflect(n, noff);
583 		}
584 		break;
585 
586 	case ICMP6_ECHO_REPLY:
587 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
588 		if (code != 0)
589 			goto badcode;
590 		break;
591 
592 	case MLD_LISTENER_QUERY:
593 	case MLD_LISTENER_REPORT:
594 		if (icmp6len < sizeof(struct mld_hdr))
595 			goto badlen;
596 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
597 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
598 		else
599 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
600 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
601 			/* give up local */
602 			mld6_input(m, off);
603 			m = NULL;
604 			goto freeit;
605 		}
606 		mld6_input(n, off);
607 		/* m stays. */
608 		break;
609 
610 	case MLD_LISTENER_DONE:
611 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
612 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
613 			goto badlen;
614 		break;		/* nothing to be done in kernel */
615 
616 	case MLD_MTRACE_RESP:
617 	case MLD_MTRACE:
618 		/* XXX: these two are experimental.  not officially defined. */
619 		/* XXX: per-interface statistics? */
620 		break;		/* just pass it to applications */
621 
622 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
623 	    {
624 		enum { WRU, FQDN } mode;
625 
626 		if (!icmp6_nodeinfo)
627 			break;
628 
629 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
630 			mode = WRU;
631 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
632 			mode = FQDN;
633 		else
634 			goto badlen;
635 
636 #define hostnamelen	strlen(hostname)
637 		if (mode == FQDN) {
638 #ifndef PULLDOWN_TEST
639 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
640 			    IPPROTO_DONE);
641 #endif
642 			n = m_copy(m, 0, M_COPYALL);
643 			if (n)
644 				n = ni6_input(n, off);
645 			/* XXX meaningless if n == NULL */
646 			noff = sizeof(struct ip6_hdr);
647 		} else {
648 			u_char *p;
649 			int maxlen, maxhlen;
650 
651 			if ((icmp6_nodeinfo & 5) != 5)
652 				break;
653 
654 			if (code != 0)
655 				goto badcode;
656 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
657 			if (maxlen >= MCLBYTES) {
658 				/* Give up remote */
659 				break;
660 			}
661 			MGETHDR(n, M_DONTWAIT, m->m_type);
662 			if (n && maxlen > MHLEN) {
663 				MCLGET(n, M_DONTWAIT);
664 				if ((n->m_flags & M_EXT) == 0) {
665 					m_free(n);
666 					n = NULL;
667 				}
668 			}
669 			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
670 				/*
671 				 * Previous code did a blind M_COPY_PKTHDR
672 				 * and said "just for rcvif".  If true, then
673 				 * we could tolerate the dup failing (due to
674 				 * the deep copy of the tag chain).  For now
675 				 * be conservative and just fail.
676 				 */
677 				m_free(n);
678 				n = NULL;
679 			}
680 			if (n == NULL) {
681 				/* Give up remote */
682 				break;
683 			}
684 			n->m_pkthdr.rcvif = NULL;
685 			n->m_len = 0;
686 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
687 			if (maxhlen > hostnamelen)
688 				maxhlen = hostnamelen;
689 			/*
690 			 * Copy IPv6 and ICMPv6 only.
691 			 */
692 			nip6 = mtod(n, struct ip6_hdr *);
693 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
694 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
695 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
696 			p = (u_char *)(nicmp6 + 1);
697 			bzero(p, 4);
698 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
699 			noff = sizeof(struct ip6_hdr);
700 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
701 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
702 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
703 			nicmp6->icmp6_code = 0;
704 		}
705 #undef hostnamelen
706 		if (n) {
707 			icmp6stat.icp6s_reflect++;
708 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
709 			icmp6_reflect(n, noff);
710 		}
711 		break;
712 	    }
713 
714 	case ICMP6_WRUREPLY:
715 		if (code != 0)
716 			goto badcode;
717 		break;
718 
719 	case ND_ROUTER_SOLICIT:
720 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
721 		if (code != 0)
722 			goto badcode;
723 		if (icmp6len < sizeof(struct nd_router_solicit))
724 			goto badlen;
725 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
726 			/* give up local */
727 			nd6_rs_input(m, off, icmp6len);
728 			m = NULL;
729 			goto freeit;
730 		}
731 		nd6_rs_input(n, off, icmp6len);
732 		/* m stays. */
733 		break;
734 
735 	case ND_ROUTER_ADVERT:
736 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
737 		if (code != 0)
738 			goto badcode;
739 		if (icmp6len < sizeof(struct nd_router_advert))
740 			goto badlen;
741 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
742 			/* give up local */
743 			nd6_ra_input(m, off, icmp6len);
744 			m = NULL;
745 			goto freeit;
746 		}
747 		nd6_ra_input(n, off, icmp6len);
748 		/* m stays. */
749 		break;
750 
751 	case ND_NEIGHBOR_SOLICIT:
752 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
753 		if (code != 0)
754 			goto badcode;
755 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
756 			goto badlen;
757 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
758 			/* give up local */
759 			nd6_ns_input(m, off, icmp6len);
760 			m = NULL;
761 			goto freeit;
762 		}
763 		nd6_ns_input(n, off, icmp6len);
764 		/* m stays. */
765 		break;
766 
767 	case ND_NEIGHBOR_ADVERT:
768 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
769 		if (code != 0)
770 			goto badcode;
771 		if (icmp6len < sizeof(struct nd_neighbor_advert))
772 			goto badlen;
773 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
774 			/* give up local */
775 			nd6_na_input(m, off, icmp6len);
776 			m = NULL;
777 			goto freeit;
778 		}
779 		nd6_na_input(n, off, icmp6len);
780 		/* m stays. */
781 		break;
782 
783 	case ND_REDIRECT:
784 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
785 		if (code != 0)
786 			goto badcode;
787 		if (icmp6len < sizeof(struct nd_redirect))
788 			goto badlen;
789 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
790 			/* give up local */
791 			icmp6_redirect_input(m, off);
792 			m = NULL;
793 			goto freeit;
794 		}
795 		icmp6_redirect_input(n, off);
796 		/* m stays. */
797 		break;
798 
799 	case ICMP6_ROUTER_RENUMBERING:
800 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
801 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
802 			goto badcode;
803 		if (icmp6len < sizeof(struct icmp6_router_renum))
804 			goto badlen;
805 		break;
806 
807 	default:
808 		nd6log((LOG_DEBUG,
809 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
810 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
811 		    ip6_sprintf(&ip6->ip6_dst),
812 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
813 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
814 			/* ICMPv6 error: MUST deliver it by spec... */
815 			code = PRC_NCMDS;
816 			/* deliver */
817 		} else {
818 			/* ICMPv6 informational: MUST not deliver */
819 			break;
820 		}
821 	deliver:
822 		if (icmp6_notify_error(m, off, icmp6len, code)) {
823 			/* In this case, m should've been freed. */
824 			return (IPPROTO_DONE);
825 		}
826 		break;
827 
828 	badcode:
829 		icmp6stat.icp6s_badcode++;
830 		break;
831 
832 	badlen:
833 		icmp6stat.icp6s_badlen++;
834 		break;
835 	}
836 
837 	/* deliver the packet to appropriate sockets */
838 	icmp6_rip6_input(&m, *offp);
839 
840 	return IPPROTO_DONE;
841 
842  freeit:
843 	m_freem(m);
844 	return IPPROTO_DONE;
845 }
846 
847 static int
848 icmp6_notify_error(m, off, icmp6len, code)
849 	struct mbuf *m;
850 	int off, icmp6len, code;
851 {
852 	struct icmp6_hdr *icmp6;
853 	struct ip6_hdr *eip6;
854 	u_int32_t notifymtu;
855 	struct sockaddr_in6 icmp6src, icmp6dst;
856 
857 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
858 		icmp6stat.icp6s_tooshort++;
859 		goto freeit;
860 	}
861 #ifndef PULLDOWN_TEST
862 	IP6_EXTHDR_CHECK(m, off,
863 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
864 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
865 #else
866 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
867 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
868 	if (icmp6 == NULL) {
869 		icmp6stat.icp6s_tooshort++;
870 		return (-1);
871 	}
872 #endif
873 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
874 
875 	/* Detect the upper level protocol */
876 	{
877 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
878 		u_int8_t nxt = eip6->ip6_nxt;
879 		int eoff = off + sizeof(struct icmp6_hdr) +
880 		    sizeof(struct ip6_hdr);
881 		struct ip6ctlparam ip6cp;
882 		struct in6_addr *finaldst = NULL;
883 		int icmp6type = icmp6->icmp6_type;
884 		struct ip6_frag *fh;
885 		struct ip6_rthdr *rth;
886 		struct ip6_rthdr0 *rth0;
887 		int rthlen;
888 
889 		while (1) { /* XXX: should avoid infinite loop explicitly? */
890 			struct ip6_ext *eh;
891 
892 			switch (nxt) {
893 			case IPPROTO_HOPOPTS:
894 			case IPPROTO_DSTOPTS:
895 			case IPPROTO_AH:
896 #ifndef PULLDOWN_TEST
897 				IP6_EXTHDR_CHECK(m, 0,
898 				    eoff + sizeof(struct ip6_ext), -1);
899 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
900 #else
901 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
902 				    eoff, sizeof(*eh));
903 				if (eh == NULL) {
904 					icmp6stat.icp6s_tooshort++;
905 					return (-1);
906 				}
907 #endif
908 
909 				if (nxt == IPPROTO_AH)
910 					eoff += (eh->ip6e_len + 2) << 2;
911 				else
912 					eoff += (eh->ip6e_len + 1) << 3;
913 				nxt = eh->ip6e_nxt;
914 				break;
915 			case IPPROTO_ROUTING:
916 				/*
917 				 * When the erroneous packet contains a
918 				 * routing header, we should examine the
919 				 * header to determine the final destination.
920 				 * Otherwise, we can't properly update
921 				 * information that depends on the final
922 				 * destination (e.g. path MTU).
923 				 */
924 #ifndef PULLDOWN_TEST
925 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
926 				rth = (struct ip6_rthdr *)
927 				    (mtod(m, caddr_t) + eoff);
928 #else
929 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
930 				    eoff, sizeof(*rth));
931 				if (rth == NULL) {
932 					icmp6stat.icp6s_tooshort++;
933 					return (-1);
934 				}
935 #endif
936 				rthlen = (rth->ip6r_len + 1) << 3;
937 				/*
938 				 * XXX: currently there is no
939 				 * officially defined type other
940 				 * than type-0.
941 				 * Note that if the segment left field
942 				 * is 0, all intermediate hops must
943 				 * have been passed.
944 				 */
945 				if (rth->ip6r_segleft &&
946 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
947 					int hops;
948 
949 #ifndef PULLDOWN_TEST
950 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
951 					rth0 = (struct ip6_rthdr0 *)
952 					    (mtod(m, caddr_t) + eoff);
953 #else
954 					IP6_EXTHDR_GET(rth0,
955 					    struct ip6_rthdr0 *, m,
956 					    eoff, rthlen);
957 					if (rth0 == NULL) {
958 						icmp6stat.icp6s_tooshort++;
959 						return (-1);
960 					}
961 #endif
962 					/* just ignore a bogus header */
963 					if ((rth0->ip6r0_len % 2) == 0 &&
964 					    (hops = rth0->ip6r0_len/2))
965 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
966 				}
967 				eoff += rthlen;
968 				nxt = rth->ip6r_nxt;
969 				break;
970 			case IPPROTO_FRAGMENT:
971 #ifndef PULLDOWN_TEST
972 				IP6_EXTHDR_CHECK(m, 0, eoff +
973 				    sizeof(struct ip6_frag), -1);
974 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
975 				    eoff);
976 #else
977 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
978 				    eoff, sizeof(*fh));
979 				if (fh == NULL) {
980 					icmp6stat.icp6s_tooshort++;
981 					return (-1);
982 				}
983 #endif
984 				/*
985 				 * Data after a fragment header is meaningless
986 				 * unless it is the first fragment, but
987 				 * we'll go to the notify label for path MTU
988 				 * discovery.
989 				 */
990 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
991 					goto notify;
992 
993 				eoff += sizeof(struct ip6_frag);
994 				nxt = fh->ip6f_nxt;
995 				break;
996 			default:
997 				/*
998 				 * This case includes ESP and the No Next
999 				 * Header.  In such cases going to the notify
1000 				 * label does not have any meaning
1001 				 * (i.e. ctlfunc will be NULL), but we go
1002 				 * anyway since we might have to update
1003 				 * path MTU information.
1004 				 */
1005 				goto notify;
1006 			}
1007 		}
1008 	  notify:
1009 #ifndef PULLDOWN_TEST
1010 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1011 #else
1012 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1013 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1014 		if (icmp6 == NULL) {
1015 			icmp6stat.icp6s_tooshort++;
1016 			return (-1);
1017 		}
1018 #endif
1019 
1020 		/*
1021 		 * retrieve parameters from the inner IPv6 header, and convert
1022 		 * them into sockaddr structures.
1023 		 * XXX: there is no guarantee that the source or destination
1024 		 * addresses of the inner packet are in the same scope as
1025 		 * the addresses of the icmp packet.  But there is no other
1026 		 * way to determine the zone.
1027 		 */
1028 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1029 
1030 		bzero(&icmp6dst, sizeof(icmp6dst));
1031 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1032 		icmp6dst.sin6_family = AF_INET6;
1033 		if (finaldst == NULL)
1034 			icmp6dst.sin6_addr = eip6->ip6_dst;
1035 		else
1036 			icmp6dst.sin6_addr = *finaldst;
1037 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1038 		    &icmp6dst.sin6_scope_id))
1039 			goto freeit;
1040 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1041 				   NULL, NULL)) {
1042 			/* should be impossbile */
1043 			nd6log((LOG_DEBUG,
1044 			    "icmp6_notify_error: in6_embedscope failed\n"));
1045 			goto freeit;
1046 		}
1047 
1048 		/*
1049 		 * retrieve parameters from the inner IPv6 header, and convert
1050 		 * them into sockaddr structures.
1051 		 */
1052 		bzero(&icmp6src, sizeof(icmp6src));
1053 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1054 		icmp6src.sin6_family = AF_INET6;
1055 		icmp6src.sin6_addr = eip6->ip6_src;
1056 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1057 		    &icmp6src.sin6_scope_id)) {
1058 			goto freeit;
1059 		}
1060 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1061 				   NULL, NULL)) {
1062 			/* should be impossbile */
1063 			nd6log((LOG_DEBUG,
1064 			    "icmp6_notify_error: in6_embedscope failed\n"));
1065 			goto freeit;
1066 		}
1067 		icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1068 
1069 		if (finaldst == NULL)
1070 			finaldst = &eip6->ip6_dst;
1071 		ip6cp.ip6c_m = m;
1072 		ip6cp.ip6c_icmp6 = icmp6;
1073 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1074 		ip6cp.ip6c_off = eoff;
1075 		ip6cp.ip6c_finaldst = finaldst;
1076 		ip6cp.ip6c_src = &icmp6src;
1077 		ip6cp.ip6c_nxt = nxt;
1078 
1079 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1080 			notifymtu = ntohl(icmp6->icmp6_mtu);
1081 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1082 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1083 		}
1084 
1085 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1086 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1087 		if (ctlfunc) {
1088 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1089 			    &ip6cp);
1090 		}
1091 	}
1092 	return (0);
1093 
1094   freeit:
1095 	m_freem(m);
1096 	return (-1);
1097 }
1098 
1099 void
1100 icmp6_mtudisc_update(ip6cp, validated)
1101 	struct ip6ctlparam *ip6cp;
1102 	int validated;
1103 {
1104 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1105 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1106 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1107 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1108 	struct rtentry *rt = NULL;
1109 	struct sockaddr_in6 sin6;
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(&sin6, sizeof(sin6));
1135 	sin6.sin6_family = PF_INET6;
1136 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1137 	sin6.sin6_addr = *dst;
1138 	/* XXX normally, this won't happen */
1139 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1140 		sin6.sin6_addr.s6_addr16[1] =
1141 		    htons(m->m_pkthdr.rcvif->if_index);
1142 	}
1143 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1144 	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1145 		      RTF_CLONING | RTF_PRCLONING);
1146 
1147 	if (rt && (rt->rt_flags & RTF_HOST) &&
1148 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1149 		if (mtu < IPV6_MMTU) {
1150 				/* xxx */
1151 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1152 		} else if (mtu < rt->rt_ifp->if_mtu &&
1153 			   rt->rt_rmx.rmx_mtu > mtu) {
1154 			icmp6stat.icp6s_pmtuchg++;
1155 			rt->rt_rmx.rmx_mtu = mtu;
1156 		}
1157 	}
1158 	if (rt)
1159 		rtfree(rt);
1160 }
1161 
1162 /*
1163  * Process a Node Information Query packet, based on
1164  * draft-ietf-ipngwg-icmp-name-lookups-07.
1165  *
1166  * Spec incompatibilities:
1167  * - IPv6 Subject address handling
1168  * - IPv4 Subject address handling support missing
1169  * - Proxy reply (answer even if it's not for me)
1170  * - joins NI group address at in6_ifattach() time only, does not cope
1171  *   with hostname changes by sethostname(3)
1172  */
1173 #define hostnamelen	strlen(hostname)
1174 static struct mbuf *
1175 ni6_input(m, off)
1176 	struct mbuf *m;
1177 	int off;
1178 {
1179 	struct icmp6_nodeinfo *ni6, *nni6;
1180 	struct mbuf *n = NULL;
1181 	u_int16_t qtype;
1182 	int subjlen;
1183 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1184 	struct ni_reply_fqdn *fqdn;
1185 	int addrs;		/* for NI_QTYPE_NODEADDR */
1186 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1187 	struct sockaddr_in6 sin6_sbj; /* subject address */
1188 	struct sockaddr_in6 sin6_d;
1189 	struct ip6_hdr *ip6;
1190 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1191 	char *subj = NULL;
1192 	struct in6_ifaddr *ia6 = NULL;
1193 
1194 	ip6 = mtod(m, struct ip6_hdr *);
1195 #ifndef PULLDOWN_TEST
1196 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1197 #else
1198 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1199 	if (ni6 == NULL) {
1200 		/* m is already reclaimed */
1201 		return (NULL);
1202 	}
1203 #endif
1204 
1205 	bzero(&sin6_d, sizeof(sin6_d));
1206 	sin6_d.sin6_family = AF_INET6; /* not used, actually */
1207 	sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1208 	sin6_d.sin6_addr = ip6->ip6_dst;
1209 	if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_dst,
1210 	    &sin6_d.sin6_scope_id)) {
1211 		goto bad;
1212 	}
1213 	if (in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL))
1214 		goto bad; /* XXX should not happen */
1215 
1216 	/*
1217 	 * Validate IPv6 destination address.
1218 	 *
1219 	 * The Responder must discard the Query without further processing
1220 	 * unless it is one of the Responder's unicast or anycast addresses, or
1221 	 * a link-local scope multicast address which the Responder has joined.
1222 	 * [icmp-name-lookups-08, Section 4.]
1223 	 */
1224 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1225 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1226 			goto bad;
1227 		/* else it's a link-local multicast, fine */
1228 	} else {		/* unicast or anycast */
1229 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1230 			goto bad; /* XXX impossible */
1231 
1232 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1233 		    !(icmp6_nodeinfo & 4)) {
1234 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1235 				"a temporary address in %s:%d",
1236 			       __FILE__, __LINE__));
1237 			goto bad;
1238 		}
1239 	}
1240 
1241 	/* validate query Subject field. */
1242 	qtype = ntohs(ni6->ni_qtype);
1243 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1244 	switch (qtype) {
1245 	case NI_QTYPE_NOOP:
1246 	case NI_QTYPE_SUPTYPES:
1247 		/* 07 draft */
1248 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1249 			break;
1250 		/* FALLTHROUGH */
1251 	case NI_QTYPE_FQDN:
1252 	case NI_QTYPE_NODEADDR:
1253 		switch (ni6->ni_code) {
1254 		case ICMP6_NI_SUBJ_IPV6:
1255 #if ICMP6_NI_SUBJ_IPV6 != 0
1256 		case 0:
1257 #endif
1258 			/*
1259 			 * backward compatibility - try to accept 03 draft
1260 			 * format, where no Subject is present.
1261 			 */
1262 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1263 			    subjlen == 0) {
1264 				oldfqdn++;
1265 				break;
1266 			}
1267 #if ICMP6_NI_SUBJ_IPV6 != 0
1268 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1269 				goto bad;
1270 #endif
1271 
1272 			if (subjlen != sizeof(struct in6_addr))
1273 				goto bad;
1274 
1275 			/*
1276 			 * Validate Subject address.
1277 			 *
1278 			 * Not sure what exactly "address belongs to the node"
1279 			 * means in the spec, is it just unicast, or what?
1280 			 *
1281 			 * At this moment we consider Subject address as
1282 			 * "belong to the node" if the Subject address equals
1283 			 * to the IPv6 destination address; validation for
1284 			 * IPv6 destination address should have done enough
1285 			 * check for us.
1286 			 *
1287 			 * We do not do proxy at this moment.
1288 			 */
1289 			/* m_pulldown instead of copy? */
1290 			bzero(&sin6_sbj, sizeof(sin6_sbj));
1291 			sin6_sbj.sin6_family = AF_INET6;
1292 			sin6_sbj.sin6_len = sizeof(sin6_sbj);
1293 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1294 			    subjlen, (caddr_t)&sin6_sbj.sin6_addr);
1295 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1296 			    &sin6_sbj.sin6_addr, &sin6_sbj.sin6_scope_id)) {
1297 				goto bad;
1298 			}
1299 			if (in6_embedscope(&sin6_sbj.sin6_addr, &sin6_sbj,
1300 			    NULL, NULL))
1301 				goto bad; /* XXX should not happen */
1302 
1303 			subj = (char *)&sin6_sbj;
1304 			if (SA6_ARE_ADDR_EQUAL(&sin6_sbj, &sin6_d))
1305 				break;
1306 
1307 			/*
1308 			 * XXX if we are to allow other cases, we should really
1309 			 * be careful about scope here.
1310 			 * basically, we should disallow queries toward IPv6
1311 			 * destination X with subject Y,
1312 			 * if scope(X) > scope(Y).
1313 			 * if we allow scope(X) > scope(Y), it will result in
1314 			 * information leakage across scope boundary.
1315 			 */
1316 			goto bad;
1317 
1318 		case ICMP6_NI_SUBJ_FQDN:
1319 			/*
1320 			 * Validate Subject name with gethostname(3).
1321 			 *
1322 			 * The behavior may need some debate, since:
1323 			 * - we are not sure if the node has FQDN as
1324 			 *   hostname (returned by gethostname(3)).
1325 			 * - the code does wildcard match for truncated names.
1326 			 *   however, we are not sure if we want to perform
1327 			 *   wildcard match, if gethostname(3) side has
1328 			 *   truncated hostname.
1329 			 */
1330 			n = ni6_nametodns(hostname, hostnamelen, 0);
1331 			if (!n || n->m_next || n->m_len == 0)
1332 				goto bad;
1333 			IP6_EXTHDR_GET(subj, char *, m,
1334 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1335 			if (subj == NULL)
1336 				goto bad;
1337 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1338 			    n->m_len)) {
1339 				goto bad;
1340 			}
1341 			m_freem(n);
1342 			n = NULL;
1343 			break;
1344 
1345 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1346 		default:
1347 			goto bad;
1348 		}
1349 		break;
1350 	}
1351 
1352 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1353 	switch (qtype) {
1354 	case NI_QTYPE_FQDN:
1355 		if ((icmp6_nodeinfo & 1) == 0)
1356 			goto bad;
1357 		break;
1358 	case NI_QTYPE_NODEADDR:
1359 		if ((icmp6_nodeinfo & 2) == 0)
1360 			goto bad;
1361 		break;
1362 	}
1363 
1364 	/* guess reply length */
1365 	switch (qtype) {
1366 	case NI_QTYPE_NOOP:
1367 		break;		/* no reply data */
1368 	case NI_QTYPE_SUPTYPES:
1369 		replylen += sizeof(u_int32_t);
1370 		break;
1371 	case NI_QTYPE_FQDN:
1372 		/* XXX will append an mbuf */
1373 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1374 		break;
1375 	case NI_QTYPE_NODEADDR:
1376 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1377 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1378 		    sizeof(u_int32_t))) > MCLBYTES)
1379 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1380 		break;
1381 	default:
1382 		/*
1383 		 * XXX: We must return a reply with the ICMP6 code
1384 		 * `unknown Qtype' in this case.  However we regard the case
1385 		 * as an FQDN query for backward compatibility.
1386 		 * Older versions set a random value to this field,
1387 		 * so it rarely varies in the defined qtypes.
1388 		 * But the mechanism is not reliable...
1389 		 * maybe we should obsolete older versions.
1390 		 */
1391 		qtype = NI_QTYPE_FQDN;
1392 		/* XXX will append an mbuf */
1393 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1394 		oldfqdn++;
1395 		break;
1396 	}
1397 
1398 	/* allocate an mbuf to reply. */
1399 	MGETHDR(n, M_DONTWAIT, m->m_type);
1400 	if (n == NULL) {
1401 		m_freem(m);
1402 		return (NULL);
1403 	}
1404 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1405 	if (replylen > MHLEN) {
1406 		if (replylen > MCLBYTES) {
1407 			/*
1408 			 * XXX: should we try to allocate more? But MCLBYTES
1409 			 * is probably much larger than IPV6_MMTU...
1410 			 */
1411 			goto bad;
1412 		}
1413 		MCLGET(n, M_DONTWAIT);
1414 		if ((n->m_flags & M_EXT) == 0) {
1415 			goto bad;
1416 		}
1417 	}
1418 	n->m_pkthdr.len = n->m_len = replylen;
1419 
1420 	/* copy mbuf header and IPv6 + Node Information base headers */
1421 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1422 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1423 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1424 
1425 	/* qtype dependent procedure */
1426 	switch (qtype) {
1427 	case NI_QTYPE_NOOP:
1428 		nni6->ni_code = ICMP6_NI_SUCCESS;
1429 		nni6->ni_flags = 0;
1430 		break;
1431 	case NI_QTYPE_SUPTYPES:
1432 	{
1433 		u_int32_t v;
1434 		nni6->ni_code = ICMP6_NI_SUCCESS;
1435 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1436 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1437 		v = (u_int32_t)htonl(0x0000000f);
1438 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1439 		break;
1440 	}
1441 	case NI_QTYPE_FQDN:
1442 		nni6->ni_code = ICMP6_NI_SUCCESS;
1443 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1444 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1445 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1446 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1447 		/*
1448 		 * XXX do we really have FQDN in variable "hostname"?
1449 		 */
1450 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1451 		if (n->m_next == NULL)
1452 			goto bad;
1453 		/* XXX we assume that n->m_next is not a chain */
1454 		if (n->m_next->m_next != NULL)
1455 			goto bad;
1456 		n->m_pkthdr.len += n->m_next->m_len;
1457 		break;
1458 	case NI_QTYPE_NODEADDR:
1459 	{
1460 		int lenlim, copied;
1461 
1462 		nni6->ni_code = ICMP6_NI_SUCCESS;
1463 		n->m_pkthdr.len = n->m_len =
1464 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1465 		lenlim = M_TRAILINGSPACE(n);
1466 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1467 		/* XXX: reset mbuf length */
1468 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1469 		    sizeof(struct icmp6_nodeinfo) + copied;
1470 		break;
1471 	}
1472 	default:
1473 		break;		/* XXX impossible! */
1474 	}
1475 
1476 	nni6->ni_type = ICMP6_NI_REPLY;
1477 	m_freem(m);
1478 	return (n);
1479 
1480   bad:
1481 	m_freem(m);
1482 	if (n)
1483 		m_freem(n);
1484 	return (NULL);
1485 }
1486 #undef hostnamelen
1487 
1488 /*
1489  * make a mbuf with DNS-encoded string.  no compression support.
1490  *
1491  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1492  * treated as truncated name (two \0 at the end).  this is a wild guess.
1493  */
1494 static struct mbuf *
1495 ni6_nametodns(name, namelen, old)
1496 	const char *name;
1497 	int namelen;
1498 	int old;	/* return pascal string if non-zero */
1499 {
1500 	struct mbuf *m;
1501 	char *cp, *ep;
1502 	const char *p, *q;
1503 	int i, len, nterm;
1504 
1505 	if (old)
1506 		len = namelen + 1;
1507 	else
1508 		len = MCLBYTES;
1509 
1510 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1511 	MGET(m, M_DONTWAIT, MT_DATA);
1512 	if (m && len > MLEN) {
1513 		MCLGET(m, M_DONTWAIT);
1514 		if ((m->m_flags & M_EXT) == 0)
1515 			goto fail;
1516 	}
1517 	if (!m)
1518 		goto fail;
1519 	m->m_next = NULL;
1520 
1521 	if (old) {
1522 		m->m_len = len;
1523 		*mtod(m, char *) = namelen;
1524 		bcopy(name, mtod(m, char *) + 1, namelen);
1525 		return m;
1526 	} else {
1527 		m->m_len = 0;
1528 		cp = mtod(m, char *);
1529 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1530 
1531 		/* if not certain about my name, return empty buffer */
1532 		if (namelen == 0)
1533 			return m;
1534 
1535 		/*
1536 		 * guess if it looks like shortened hostname, or FQDN.
1537 		 * shortened hostname needs two trailing "\0".
1538 		 */
1539 		i = 0;
1540 		for (p = name; p < name + namelen; p++) {
1541 			if (*p && *p == '.')
1542 				i++;
1543 		}
1544 		if (i < 2)
1545 			nterm = 2;
1546 		else
1547 			nterm = 1;
1548 
1549 		p = name;
1550 		while (cp < ep && p < name + namelen) {
1551 			i = 0;
1552 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1553 				i++;
1554 			/* result does not fit into mbuf */
1555 			if (cp + i + 1 >= ep)
1556 				goto fail;
1557 			/*
1558 			 * DNS label length restriction, RFC1035 page 8.
1559 			 * "i == 0" case is included here to avoid returning
1560 			 * 0-length label on "foo..bar".
1561 			 */
1562 			if (i <= 0 || i >= 64)
1563 				goto fail;
1564 			*cp++ = i;
1565 			bcopy(p, cp, i);
1566 			cp += i;
1567 			p = q;
1568 			if (p < name + namelen && *p == '.')
1569 				p++;
1570 		}
1571 		/* termination */
1572 		if (cp + nterm >= ep)
1573 			goto fail;
1574 		while (nterm-- > 0)
1575 			*cp++ = '\0';
1576 		m->m_len = cp - mtod(m, char *);
1577 		return m;
1578 	}
1579 
1580 	panic("should not reach here");
1581 	/* NOTREACHED */
1582 
1583  fail:
1584 	if (m)
1585 		m_freem(m);
1586 	return NULL;
1587 }
1588 
1589 /*
1590  * check if two DNS-encoded string matches.  takes care of truncated
1591  * form (with \0\0 at the end).  no compression support.
1592  * XXX upper/lowercase match (see RFC2065)
1593  */
1594 static int
1595 ni6_dnsmatch(a, alen, b, blen)
1596 	const char *a;
1597 	int alen;
1598 	const char *b;
1599 	int blen;
1600 {
1601 	const char *a0, *b0;
1602 	int l;
1603 
1604 	/* simplest case - need validation? */
1605 	if (alen == blen && bcmp(a, b, alen) == 0)
1606 		return 1;
1607 
1608 	a0 = a;
1609 	b0 = b;
1610 
1611 	/* termination is mandatory */
1612 	if (alen < 2 || blen < 2)
1613 		return 0;
1614 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1615 		return 0;
1616 	alen--;
1617 	blen--;
1618 
1619 	while (a - a0 < alen && b - b0 < blen) {
1620 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1621 			return 0;
1622 
1623 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1624 			return 0;
1625 		/* we don't support compression yet */
1626 		if (a[0] >= 64 || b[0] >= 64)
1627 			return 0;
1628 
1629 		/* truncated case */
1630 		if (a[0] == 0 && a - a0 == alen - 1)
1631 			return 1;
1632 		if (b[0] == 0 && b - b0 == blen - 1)
1633 			return 1;
1634 		if (a[0] == 0 || b[0] == 0)
1635 			return 0;
1636 
1637 		if (a[0] != b[0])
1638 			return 0;
1639 		l = a[0];
1640 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1641 			return 0;
1642 		if (bcmp(a + 1, b + 1, l) != 0)
1643 			return 0;
1644 
1645 		a += 1 + l;
1646 		b += 1 + l;
1647 	}
1648 
1649 	if (a - a0 == alen && b - b0 == blen)
1650 		return 1;
1651 	else
1652 		return 0;
1653 }
1654 
1655 /*
1656  * calculate the number of addresses to be returned in the node info reply.
1657  */
1658 static int
1659 ni6_addrs(ni6, m, ifpp, subj)
1660 	struct icmp6_nodeinfo *ni6;
1661 	struct mbuf *m;
1662 	struct ifnet **ifpp;
1663 	char *subj;
1664 {
1665 	struct ifnet *ifp;
1666 	struct in6_ifaddr *ifa6;
1667 	struct ifaddr *ifa;
1668 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1669 	int addrs = 0, addrsofif, iffound = 0;
1670 	int niflags = ni6->ni_flags;
1671 
1672 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1673 		switch (ni6->ni_code) {
1674 		case ICMP6_NI_SUBJ_IPV6:
1675 			if (subj == NULL) /* must be impossible... */
1676 				return (0);
1677 			subj_ip6 = (struct sockaddr_in6 *)subj;
1678 			break;
1679 		default:
1680 			/*
1681 			 * XXX: we only support IPv6 subject address for
1682 			 * this Qtype.
1683 			 */
1684 			return (0);
1685 		}
1686 	}
1687 
1688 	IFNET_RLOCK();
1689 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1690 		addrsofif = 0;
1691 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1692 			if (ifa->ifa_addr->sa_family != AF_INET6)
1693 				continue;
1694 			ifa6 = (struct in6_ifaddr *)ifa;
1695 
1696 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1697 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1698 			    &ifa6->ia_addr.sin6_addr))
1699 				iffound = 1;
1700 
1701 			/*
1702 			 * IPv4-mapped addresses can only be returned by a
1703 			 * Node Information proxy, since they represent
1704 			 * addresses of IPv4-only nodes, which perforce do
1705 			 * not implement this protocol.
1706 			 * [icmp-name-lookups-07, Section 5.4]
1707 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1708 			 * this function at this moment.
1709 			 */
1710 
1711 			/* What do we have to do about ::1? */
1712 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1713 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_SITELOCAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1719 					continue;
1720 				break;
1721 			case IPV6_ADDR_SCOPE_GLOBAL:
1722 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1723 					continue;
1724 				break;
1725 			default:
1726 				continue;
1727 			}
1728 
1729 			/*
1730 			 * check if anycast is okay.
1731 			 * XXX: just experimental.  not in the spec.
1732 			 */
1733 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1734 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1735 				continue; /* we need only unicast addresses */
1736 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1737 			    (icmp6_nodeinfo & 4) == 0) {
1738 				continue;
1739 			}
1740 			addrsofif++; /* count the address */
1741 		}
1742 		if (iffound) {
1743 			*ifpp = ifp;
1744 			IFNET_RUNLOCK();
1745 			return (addrsofif);
1746 		}
1747 
1748 		addrs += addrsofif;
1749 	}
1750 	IFNET_RUNLOCK();
1751 
1752 	return (addrs);
1753 }
1754 
1755 static int
1756 ni6_store_addrs(ni6, nni6, ifp0, resid)
1757 	struct icmp6_nodeinfo *ni6, *nni6;
1758 	struct ifnet *ifp0;
1759 	int resid;
1760 {
1761 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1762 	struct in6_ifaddr *ifa6;
1763 	struct ifaddr *ifa;
1764 	struct ifnet *ifp_dep = NULL;
1765 	int copied = 0, allow_deprecated = 0;
1766 	u_char *cp = (u_char *)(nni6 + 1);
1767 	int niflags = ni6->ni_flags;
1768 	u_int32_t ltime;
1769 
1770 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1771 		return (0);	/* needless to copy */
1772 
1773 	IFNET_RLOCK();
1774   again:
1775 
1776 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1777 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1778 		     ifa = ifa->ifa_list.tqe_next) {
1779 			if (ifa->ifa_addr->sa_family != AF_INET6)
1780 				continue;
1781 			ifa6 = (struct in6_ifaddr *)ifa;
1782 
1783 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1784 			    allow_deprecated == 0) {
1785 				/*
1786 				 * prefererred address should be put before
1787 				 * deprecated addresses.
1788 				 */
1789 
1790 				/* record the interface for later search */
1791 				if (ifp_dep == NULL)
1792 					ifp_dep = ifp;
1793 
1794 				continue;
1795 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1796 			    allow_deprecated != 0)
1797 				continue; /* we now collect deprecated addrs */
1798 
1799 			/* What do we have to do about ::1? */
1800 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1801 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1802 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1803 					continue;
1804 				break;
1805 			case IPV6_ADDR_SCOPE_SITELOCAL:
1806 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1807 					continue;
1808 				break;
1809 			case IPV6_ADDR_SCOPE_GLOBAL:
1810 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1811 					continue;
1812 				break;
1813 			default:
1814 				continue;
1815 			}
1816 
1817 			/*
1818 			 * check if anycast is okay.
1819 			 * XXX: just experimental.  not in the spec.
1820 			 */
1821 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1822 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1823 				continue;
1824 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1825 			    (icmp6_nodeinfo & 4) == 0) {
1826 				continue;
1827 			}
1828 
1829 			/* now we can copy the address */
1830 			if (resid < sizeof(struct in6_addr) +
1831 			    sizeof(u_int32_t)) {
1832 				/*
1833 				 * We give up much more copy.
1834 				 * Set the truncate flag and return.
1835 				 */
1836 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1837 				IFNET_RUNLOCK();
1838 				return (copied);
1839 			}
1840 
1841 			/*
1842 			 * Set the TTL of the address.
1843 			 * The TTL value should be one of the following
1844 			 * according to the specification:
1845 			 *
1846 			 * 1. The remaining lifetime of a DHCP lease on the
1847 			 *    address, or
1848 			 * 2. The remaining Valid Lifetime of a prefix from
1849 			 *    which the address was derived through Stateless
1850 			 *    Autoconfiguration.
1851 			 *
1852 			 * Note that we currently do not support stateful
1853 			 * address configuration by DHCPv6, so the former
1854 			 * case can't happen.
1855 			 */
1856 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1857 				ltime = ND6_INFINITE_LIFETIME;
1858 			else {
1859 				if (ifa6->ia6_lifetime.ia6t_expire >
1860 				    time_second)
1861 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1862 				else
1863 					ltime = 0;
1864 			}
1865 
1866 			bcopy(&ltime, cp, sizeof(u_int32_t));
1867 			cp += sizeof(u_int32_t);
1868 
1869 			/* copy the address itself */
1870 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1871 			    sizeof(struct in6_addr));
1872 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1873 			cp += sizeof(struct in6_addr);
1874 
1875 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1876 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1877 		}
1878 		if (ifp0)	/* we need search only on the specified IF */
1879 			break;
1880 	}
1881 
1882 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1883 		ifp = ifp_dep;
1884 		allow_deprecated = 1;
1885 
1886 		goto again;
1887 	}
1888 
1889 	IFNET_RUNLOCK();
1890 
1891 	return (copied);
1892 }
1893 
1894 /*
1895  * XXX almost dup'ed code with rip6_input.
1896  */
1897 static int
1898 icmp6_rip6_input(mp, off)
1899 	struct	mbuf **mp;
1900 	int	off;
1901 {
1902 	struct mbuf *m = *mp;
1903 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1904 	struct in6pcb *in6p;
1905 	struct in6pcb *last = NULL;
1906 	struct sockaddr_in6 fromsa;
1907 	struct icmp6_hdr *icmp6;
1908 	struct mbuf *opts = NULL;
1909 
1910 #ifndef PULLDOWN_TEST
1911 	/* this is assumed to be safe. */
1912 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1913 #else
1914 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1915 	if (icmp6 == NULL) {
1916 		/* m is already reclaimed */
1917 		return (IPPROTO_DONE);
1918 	}
1919 #endif
1920 
1921 	bzero(&fromsa, sizeof(fromsa));
1922 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1923 	fromsa.sin6_family = AF_INET6;
1924 	/* KAME hack: recover scopeid */
1925 	(void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1926 
1927 	LIST_FOREACH(in6p, &ripcb, inp_list) {
1928 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1929 			continue;
1930 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1931 			continue;
1932 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1933 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1934 			continue;
1935 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1936 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1937 			continue;
1938 		if (in6p->in6p_icmp6filt
1939 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1940 				 in6p->in6p_icmp6filt))
1941 			continue;
1942 		if (last) {
1943 			struct	mbuf *n = NULL;
1944 
1945 			/*
1946 			 * Recent network drivers tend to allocate a single
1947 			 * mbuf cluster, rather than to make a couple of
1948 			 * mbufs without clusters.  Also, since the IPv6 code
1949 			 * path tries to avoid m_pullup(), it is highly
1950 			 * probable that we still have an mbuf cluster here
1951 			 * even though the necessary length can be stored in an
1952 			 * mbuf's internal buffer.
1953 			 * Meanwhile, the default size of the receive socket
1954 			 * buffer for raw sockets is not so large.  This means
1955 			 * the possibility of packet loss is relatively higher
1956 			 * than before.  To avoid this scenario, we copy the
1957 			 * received data to a separate mbuf that does not use
1958 			 * a cluster, if possible.
1959 			 * XXX: it is better to copy the data after stripping
1960 			 * intermediate headers.
1961 			 */
1962 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1963 			    m->m_len <= MHLEN) {
1964 				MGET(n, M_DONTWAIT, m->m_type);
1965 				if (n != NULL) {
1966 					m_dup_pkthdr(n, m, M_NOWAIT);
1967 					bcopy(m->m_data, n->m_data, m->m_len);
1968 					n->m_len = m->m_len;
1969 				}
1970 			}
1971 			if (n != NULL ||
1972 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1973 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1974 					ip6_savecontrol(last, n, &opts);
1975 				/* strip intermediate headers */
1976 				m_adj(n, off);
1977 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1978 				    (struct sockaddr *)&fromsa, n, opts)
1979 				    == 0) {
1980 					/* should notify about lost packet */
1981 					m_freem(n);
1982 					if (opts) {
1983 						m_freem(opts);
1984 					}
1985 				} else
1986 					sorwakeup(last->in6p_socket);
1987 				opts = NULL;
1988 			}
1989 		}
1990 		last = in6p;
1991 	}
1992 	if (last) {
1993 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1994 			ip6_savecontrol(last, m, &opts);
1995 		/* strip intermediate headers */
1996 		m_adj(m, off);
1997 
1998 		/* avoid using mbuf clusters if possible (see above) */
1999 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2000 		    m->m_len <= MHLEN) {
2001 			struct mbuf *n;
2002 
2003 			MGET(n, M_DONTWAIT, m->m_type);
2004 			if (n != NULL) {
2005 				m_dup_pkthdr(n, m, M_NOWAIT);
2006 				bcopy(m->m_data, n->m_data, m->m_len);
2007 				n->m_len = m->m_len;
2008 
2009 				m_freem(m);
2010 				m = n;
2011 			}
2012 		}
2013 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2014 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2015 			m_freem(m);
2016 			if (opts)
2017 				m_freem(opts);
2018 		} else
2019 			sorwakeup(last->in6p_socket);
2020 	} else {
2021 		m_freem(m);
2022 		ip6stat.ip6s_delivered--;
2023 	}
2024 	return IPPROTO_DONE;
2025 }
2026 
2027 /*
2028  * Reflect the ip6 packet back to the source.
2029  * OFF points to the icmp6 header, counted from the top of the mbuf.
2030  */
2031 void
2032 icmp6_reflect(m, off)
2033 	struct	mbuf *m;
2034 	size_t off;
2035 {
2036 	struct ip6_hdr *ip6;
2037 	struct icmp6_hdr *icmp6;
2038 	struct in6_ifaddr *ia;
2039 	struct in6_addr t, *src = 0;
2040 	int plen;
2041 	int type, code;
2042 	struct ifnet *outif = NULL;
2043 	struct sockaddr_in6 sa6_src, sa6_dst;
2044 #ifdef COMPAT_RFC1885
2045 	int mtu = IPV6_MMTU;
2046 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2047 #endif
2048 
2049 	/* too short to reflect */
2050 	if (off < sizeof(struct ip6_hdr)) {
2051 		nd6log((LOG_DEBUG,
2052 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2053 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2054 		    __FILE__, __LINE__));
2055 		goto bad;
2056 	}
2057 
2058 	/*
2059 	 * If there are extra headers between IPv6 and ICMPv6, strip
2060 	 * off that header first.
2061 	 */
2062 #ifdef DIAGNOSTIC
2063 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2064 		panic("assumption failed in icmp6_reflect");
2065 #endif
2066 	if (off > sizeof(struct ip6_hdr)) {
2067 		size_t l;
2068 		struct ip6_hdr nip6;
2069 
2070 		l = off - sizeof(struct ip6_hdr);
2071 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2072 		m_adj(m, l);
2073 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2074 		if (m->m_len < l) {
2075 			if ((m = m_pullup(m, l)) == NULL)
2076 				return;
2077 		}
2078 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2079 	} else /* off == sizeof(struct ip6_hdr) */ {
2080 		size_t l;
2081 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2082 		if (m->m_len < l) {
2083 			if ((m = m_pullup(m, l)) == NULL)
2084 				return;
2085 		}
2086 	}
2087 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2088 	ip6 = mtod(m, struct ip6_hdr *);
2089 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2090 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2091 	type = icmp6->icmp6_type; /* keep type for statistics */
2092 	code = icmp6->icmp6_code; /* ditto. */
2093 
2094 	t = ip6->ip6_dst;
2095 	/*
2096 	 * ip6_input() drops a packet if its src is multicast.
2097 	 * So, the src is never multicast.
2098 	 */
2099 	ip6->ip6_dst = ip6->ip6_src;
2100 
2101 	/*
2102 	 * XXX: make sure to embed scope zone information, using
2103 	 * already embedded IDs or the received interface (if any).
2104 	 * Note that rcvif may be NULL.
2105 	 * TODO: scoped routing case (XXX).
2106 	 */
2107 	bzero(&sa6_src, sizeof(sa6_src));
2108 	sa6_src.sin6_family = AF_INET6;
2109 	sa6_src.sin6_len = sizeof(sa6_src);
2110 	sa6_src.sin6_addr = ip6->ip6_dst;
2111 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2112 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2113 	bzero(&sa6_dst, sizeof(sa6_dst));
2114 	sa6_dst.sin6_family = AF_INET6;
2115 	sa6_dst.sin6_len = sizeof(sa6_dst);
2116 	sa6_dst.sin6_addr = t;
2117 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2118 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2119 
2120 #ifdef COMPAT_RFC1885
2121 	/*
2122 	 * xxx guess MTU
2123 	 * RFC 1885 requires that echo reply should be truncated if it
2124 	 * does not fit in with (return) path MTU, but the description was
2125 	 * removed in the new spec.
2126 	 */
2127 	if (icmp6_reflect_rt.ro_rt == 0 ||
2128 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2129 		if (icmp6_reflect_rt.ro_rt) {
2130 			RTFREE(icmp6_reflect_rt.ro_rt);
2131 			icmp6_reflect_rt.ro_rt = 0;
2132 		}
2133 		bzero(sin6, sizeof(*sin6));
2134 		sin6->sin6_family = PF_INET6;
2135 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2136 		sin6->sin6_addr = ip6->ip6_dst;
2137 
2138 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2139 			    RTF_PRCLONING);
2140 	}
2141 
2142 	if (icmp6_reflect_rt.ro_rt == 0)
2143 		goto bad;
2144 
2145 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2146 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2147 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2148 
2149 	if (mtu < m->m_pkthdr.len) {
2150 		plen -= (m->m_pkthdr.len - mtu);
2151 		m_adj(m, mtu - m->m_pkthdr.len);
2152 	}
2153 #endif
2154 	/*
2155 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2156 	 * use dst as the src for the reply.
2157 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2158 	 * (for example) when we encounter an error while forwarding procedure
2159 	 * destined to a duplicated address of ours.
2160 	 */
2161 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2162 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2163 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2164 			src = &t;
2165 			break;
2166 		}
2167 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2168 		/*
2169 		 * This is the case if the dst is our link-local address
2170 		 * and the sender is also ourselves.
2171 		 */
2172 		src = &t;
2173 	}
2174 
2175 	if (src == 0) {
2176 		int e;
2177 		struct route_in6 ro;
2178 
2179 		/*
2180 		 * This case matches to multicasts, our anycast, or unicasts
2181 		 * that we do not own.  Select a source address based on the
2182 		 * source address of the erroneous packet.
2183 		 */
2184 		bzero(&ro, sizeof(ro));
2185 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2186 		if (ro.ro_rt)
2187 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2188 		if (src == NULL) {
2189 			nd6log((LOG_DEBUG,
2190 			    "icmp6_reflect: source can't be determined: "
2191 			    "dst=%s, error=%d\n",
2192 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2193 			goto bad;
2194 		}
2195 	}
2196 
2197 	ip6->ip6_src = *src;
2198 
2199 	ip6->ip6_flow = 0;
2200 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2201 	ip6->ip6_vfc |= IPV6_VERSION;
2202 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2203 	if (outif)
2204 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2205 	else if (m->m_pkthdr.rcvif) {
2206 		/* XXX: This may not be the outgoing interface */
2207 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2208 	} else
2209 		ip6->ip6_hlim = ip6_defhlim;
2210 
2211 	icmp6->icmp6_cksum = 0;
2212 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2213 	    sizeof(struct ip6_hdr), plen);
2214 
2215 	/*
2216 	 * XXX option handling
2217 	 */
2218 
2219 	m->m_flags &= ~(M_BCAST|M_MCAST);
2220 #ifdef IPSEC
2221 	/* Don't lookup socket */
2222 	(void)ipsec_setsocket(m, NULL);
2223 #endif /* IPSEC */
2224 
2225 #ifdef COMPAT_RFC1885
2226 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2227 #else
2228 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2229 #endif
2230 	if (outif)
2231 		icmp6_ifoutstat_inc(outif, type, code);
2232 
2233 	return;
2234 
2235  bad:
2236 	m_freem(m);
2237 	return;
2238 }
2239 
2240 void
2241 icmp6_fasttimo()
2242 {
2243 
2244 	mld6_fasttimeo();
2245 }
2246 
2247 static const char *
2248 icmp6_redirect_diag(src6, dst6, tgt6)
2249 	struct in6_addr *src6;
2250 	struct in6_addr *dst6;
2251 	struct in6_addr *tgt6;
2252 {
2253 	static char buf[1024];
2254 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2255 	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2256 	return buf;
2257 }
2258 
2259 void
2260 icmp6_redirect_input(m, off)
2261 	struct mbuf *m;
2262 	int off;
2263 {
2264 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2265 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2266 	struct nd_redirect *nd_rd;
2267 	int icmp6len = ntohs(ip6->ip6_plen);
2268 	char *lladdr = NULL;
2269 	int lladdrlen = 0;
2270 	u_char *redirhdr = NULL;
2271 	int redirhdrlen = 0;
2272 	struct rtentry *rt = NULL;
2273 	int is_router;
2274 	int is_onlink;
2275 	struct in6_addr src6 = ip6->ip6_src;
2276 	struct in6_addr redtgt6;
2277 	struct in6_addr reddst6;
2278 	union nd_opts ndopts;
2279 
2280 	if (!m || !ifp)
2281 		return;
2282 
2283 	/* XXX if we are router, we don't update route by icmp6 redirect */
2284 	if (ip6_forwarding)
2285 		goto freeit;
2286 	if (!icmp6_rediraccept)
2287 		goto freeit;
2288 
2289 #ifndef PULLDOWN_TEST
2290 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2291 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2292 #else
2293 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2294 	if (nd_rd == NULL) {
2295 		icmp6stat.icp6s_tooshort++;
2296 		return;
2297 	}
2298 #endif
2299 	redtgt6 = nd_rd->nd_rd_target;
2300 	reddst6 = nd_rd->nd_rd_dst;
2301 
2302 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2303 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2304 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2305 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2306 
2307 	/* validation */
2308 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2309 		nd6log((LOG_ERR,
2310 		    "ICMP6 redirect sent from %s rejected; "
2311 		    "must be from linklocal\n",
2312 		    ip6_sprintf(&src6)));
2313 		goto bad;
2314 	}
2315 	if (ip6->ip6_hlim != 255) {
2316 		nd6log((LOG_ERR,
2317 		    "ICMP6 redirect sent from %s rejected; "
2318 		    "hlim=%d (must be 255)\n",
2319 		    ip6_sprintf(&src6), ip6->ip6_hlim));
2320 		goto bad;
2321 	}
2322     {
2323 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2324 	struct sockaddr_in6 sin6;
2325 	struct in6_addr *gw6;
2326 
2327 	bzero(&sin6, sizeof(sin6));
2328 	sin6.sin6_family = AF_INET6;
2329 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2330 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2331 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2332 	if (rt) {
2333 		if (rt->rt_gateway == NULL ||
2334 		    rt->rt_gateway->sa_family != AF_INET6) {
2335 			nd6log((LOG_ERR,
2336 			    "ICMP6 redirect rejected; no route "
2337 			    "with inet6 gateway found for redirect dst: %s\n",
2338 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2339 			RTFREE_LOCKED(rt);
2340 			goto bad;
2341 		}
2342 
2343 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2344 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2345 			nd6log((LOG_ERR,
2346 			    "ICMP6 redirect rejected; "
2347 			    "not equal to gw-for-src=%s (must be same): "
2348 			    "%s\n",
2349 			    ip6_sprintf(gw6),
2350 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2351 			RTFREE_LOCKED(rt);
2352 			goto bad;
2353 		}
2354 	} else {
2355 		nd6log((LOG_ERR,
2356 		    "ICMP6 redirect rejected; "
2357 		    "no route found for redirect dst: %s\n",
2358 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2359 		goto bad;
2360 	}
2361 	RTFREE_LOCKED(rt);
2362 	rt = NULL;
2363     }
2364 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2365 		nd6log((LOG_ERR,
2366 		    "ICMP6 redirect rejected; "
2367 		    "redirect dst must be unicast: %s\n",
2368 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2369 		goto bad;
2370 	}
2371 
2372 	is_router = is_onlink = 0;
2373 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2374 		is_router = 1;	/* router case */
2375 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2376 		is_onlink = 1;	/* on-link destination case */
2377 	if (!is_router && !is_onlink) {
2378 		nd6log((LOG_ERR,
2379 		    "ICMP6 redirect rejected; "
2380 		    "neither router case nor onlink case: %s\n",
2381 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2382 		goto bad;
2383 	}
2384 	/* validation passed */
2385 
2386 	icmp6len -= sizeof(*nd_rd);
2387 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2388 	if (nd6_options(&ndopts) < 0) {
2389 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2390 		    "invalid ND option, rejected: %s\n",
2391 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2392 		/* nd6_options have incremented stats */
2393 		goto freeit;
2394 	}
2395 
2396 	if (ndopts.nd_opts_tgt_lladdr) {
2397 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2398 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2399 	}
2400 
2401 	if (ndopts.nd_opts_rh) {
2402 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2403 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2404 	}
2405 
2406 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2407 		nd6log((LOG_INFO,
2408 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2409 		    "(if %d, icmp6 packet %d): %s\n",
2410 		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2411 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2412 		goto bad;
2413 	}
2414 
2415 	/* RFC 2461 8.3 */
2416 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2417 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2418 
2419 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2420 		/* perform rtredirect */
2421 		struct sockaddr_in6 sdst;
2422 		struct sockaddr_in6 sgw;
2423 		struct sockaddr_in6 ssrc;
2424 
2425 		bzero(&sdst, sizeof(sdst));
2426 		bzero(&sgw, sizeof(sgw));
2427 		bzero(&ssrc, sizeof(ssrc));
2428 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2429 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2430 			sizeof(struct sockaddr_in6);
2431 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2432 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2433 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2434 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2435 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2436 		    (struct sockaddr *)&ssrc);
2437 	}
2438 	/* finally update cached route in each socket via pfctlinput */
2439     {
2440 	struct sockaddr_in6 sdst;
2441 
2442 	bzero(&sdst, sizeof(sdst));
2443 	sdst.sin6_family = AF_INET6;
2444 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2445 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2446 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2447 #if defined(IPSEC) || defined(FAST_IPSEC)
2448 	key_sa_routechange((struct sockaddr *)&sdst);
2449 #endif
2450     }
2451 
2452  freeit:
2453 	m_freem(m);
2454 	return;
2455 
2456  bad:
2457 	icmp6stat.icp6s_badredirect++;
2458 	m_freem(m);
2459 }
2460 
2461 void
2462 icmp6_redirect_output(m0, rt)
2463 	struct mbuf *m0;
2464 	struct rtentry *rt;
2465 {
2466 	struct ifnet *ifp;	/* my outgoing interface */
2467 	struct in6_addr *ifp_ll6;
2468 	struct in6_addr *router_ll6;
2469 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2470 	struct mbuf *m = NULL;	/* newly allocated one */
2471 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2472 	struct nd_redirect *nd_rd;
2473 	size_t maxlen;
2474 	u_char *p;
2475 	struct ifnet *outif = NULL;
2476 	struct sockaddr_in6 src_sa;
2477 
2478 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2479 
2480 	/* if we are not router, we don't send icmp6 redirect */
2481 	if (!ip6_forwarding || ip6_accept_rtadv)
2482 		goto fail;
2483 
2484 	/* sanity check */
2485 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2486 		goto fail;
2487 
2488 	/*
2489 	 * Address check:
2490 	 *  the source address must identify a neighbor, and
2491 	 *  the destination address must not be a multicast address
2492 	 *  [RFC 2461, sec 8.2]
2493 	 */
2494 	sip6 = mtod(m0, struct ip6_hdr *);
2495 	bzero(&src_sa, sizeof(src_sa));
2496 	src_sa.sin6_family = AF_INET6;
2497 	src_sa.sin6_len = sizeof(src_sa);
2498 	src_sa.sin6_addr = sip6->ip6_src;
2499 	/* we don't currently use sin6_scope_id, but eventually use it */
2500 	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2501 		goto fail;
2502 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2503 		goto fail;
2504 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2505 		goto fail;	/* what should we do here? */
2506 
2507 	/* rate limit */
2508 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2509 		goto fail;
2510 
2511 	/*
2512 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2513 	 * we almost always ask for an mbuf cluster for simplicity.
2514 	 * (MHLEN < IPV6_MMTU is almost always true)
2515 	 */
2516 #if IPV6_MMTU >= MCLBYTES
2517 # error assumption failed about IPV6_MMTU and MCLBYTES
2518 #endif
2519 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2520 	if (m && IPV6_MMTU >= MHLEN)
2521 		MCLGET(m, M_DONTWAIT);
2522 	if (!m)
2523 		goto fail;
2524 	m->m_pkthdr.rcvif = NULL;
2525 	m->m_len = 0;
2526 	maxlen = M_TRAILINGSPACE(m);
2527 	maxlen = min(IPV6_MMTU, maxlen);
2528 	/* just for safety */
2529 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2530 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2531 		goto fail;
2532 	}
2533 
2534 	{
2535 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2536 		struct in6_ifaddr *ia;
2537 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2538 						 IN6_IFF_NOTREADY|
2539 						 IN6_IFF_ANYCAST)) == NULL)
2540 			goto fail;
2541 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2542 	}
2543 
2544 	/* get ip6 linklocal address for the router. */
2545 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2546 		struct sockaddr_in6 *sin6;
2547 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2548 		router_ll6 = &sin6->sin6_addr;
2549 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2550 			router_ll6 = (struct in6_addr *)NULL;
2551 	} else
2552 		router_ll6 = (struct in6_addr *)NULL;
2553 
2554 	/* ip6 */
2555 	ip6 = mtod(m, struct ip6_hdr *);
2556 	ip6->ip6_flow = 0;
2557 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2558 	ip6->ip6_vfc |= IPV6_VERSION;
2559 	/* ip6->ip6_plen will be set later */
2560 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2561 	ip6->ip6_hlim = 255;
2562 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2563 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2564 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2565 
2566 	/* ND Redirect */
2567 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2568 	nd_rd->nd_rd_type = ND_REDIRECT;
2569 	nd_rd->nd_rd_code = 0;
2570 	nd_rd->nd_rd_reserved = 0;
2571 	if (rt->rt_flags & RTF_GATEWAY) {
2572 		/*
2573 		 * nd_rd->nd_rd_target must be a link-local address in
2574 		 * better router cases.
2575 		 */
2576 		if (!router_ll6)
2577 			goto fail;
2578 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2579 		    sizeof(nd_rd->nd_rd_target));
2580 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2581 		    sizeof(nd_rd->nd_rd_dst));
2582 	} else {
2583 		/* make sure redtgt == reddst */
2584 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2585 		    sizeof(nd_rd->nd_rd_target));
2586 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2587 		    sizeof(nd_rd->nd_rd_dst));
2588 	}
2589 
2590 	p = (u_char *)(nd_rd + 1);
2591 
2592 	if (!router_ll6)
2593 		goto nolladdropt;
2594 
2595 	{
2596 		/* target lladdr option */
2597 		struct rtentry *rt_router = NULL;
2598 		int len;
2599 		struct sockaddr_dl *sdl;
2600 		struct nd_opt_hdr *nd_opt;
2601 		char *lladdr;
2602 
2603 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2604 		if (!rt_router)
2605 			goto nolladdropt;
2606 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2607 		len = (len + 7) & ~7;	/* round by 8 */
2608 		/* safety check */
2609 		if (len + (p - (u_char *)ip6) > maxlen)
2610 			goto nolladdropt;
2611 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2612 		    (rt_router->rt_flags & RTF_LLINFO) &&
2613 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2614 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2615 		    sdl->sdl_alen) {
2616 			nd_opt = (struct nd_opt_hdr *)p;
2617 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2618 			nd_opt->nd_opt_len = len >> 3;
2619 			lladdr = (char *)(nd_opt + 1);
2620 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2621 			p += len;
2622 		}
2623 	}
2624 nolladdropt:;
2625 
2626 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2627 
2628 	/* just to be safe */
2629 #ifdef M_DECRYPTED	/*not openbsd*/
2630 	if (m0->m_flags & M_DECRYPTED)
2631 		goto noredhdropt;
2632 #endif
2633 	if (p - (u_char *)ip6 > maxlen)
2634 		goto noredhdropt;
2635 
2636 	{
2637 		/* redirected header option */
2638 		int len;
2639 		struct nd_opt_rd_hdr *nd_opt_rh;
2640 
2641 		/*
2642 		 * compute the maximum size for icmp6 redirect header option.
2643 		 * XXX room for auth header?
2644 		 */
2645 		len = maxlen - (p - (u_char *)ip6);
2646 		len &= ~7;
2647 
2648 		/* This is just for simplicity. */
2649 		if (m0->m_pkthdr.len != m0->m_len) {
2650 			if (m0->m_next) {
2651 				m_freem(m0->m_next);
2652 				m0->m_next = NULL;
2653 			}
2654 			m0->m_pkthdr.len = m0->m_len;
2655 		}
2656 
2657 		/*
2658 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2659 		 * about padding/truncate rule for the original IP packet.
2660 		 * From the discussion on IPv6imp in Feb 1999,
2661 		 * the consensus was:
2662 		 * - "attach as much as possible" is the goal
2663 		 * - pad if not aligned (original size can be guessed by
2664 		 *   original ip6 header)
2665 		 * Following code adds the padding if it is simple enough,
2666 		 * and truncates if not.
2667 		 */
2668 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2669 			panic("assumption failed in %s:%d", __FILE__,
2670 			    __LINE__);
2671 
2672 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2673 			/* not enough room, truncate */
2674 			m0->m_pkthdr.len = m0->m_len = len -
2675 			    sizeof(*nd_opt_rh);
2676 		} else {
2677 			/* enough room, pad or truncate */
2678 			size_t extra;
2679 
2680 			extra = m0->m_pkthdr.len % 8;
2681 			if (extra) {
2682 				/* pad if easy enough, truncate if not */
2683 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2684 					/* pad */
2685 					m0->m_len += (8 - extra);
2686 					m0->m_pkthdr.len += (8 - extra);
2687 				} else {
2688 					/* truncate */
2689 					m0->m_pkthdr.len -= extra;
2690 					m0->m_len -= extra;
2691 				}
2692 			}
2693 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2694 			m0->m_pkthdr.len = m0->m_len = len -
2695 			    sizeof(*nd_opt_rh);
2696 		}
2697 
2698 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2699 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2700 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2701 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2702 		p += sizeof(*nd_opt_rh);
2703 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2704 
2705 		/* connect m0 to m */
2706 		m_tag_delete_chain(m0, NULL);
2707 		m0->m_flags &= ~M_PKTHDR;
2708 		m->m_next = m0;
2709 		m->m_pkthdr.len = m->m_len + m0->m_len;
2710 		m0 = NULL;
2711 	}
2712 noredhdropt:;
2713 	if (m0) {
2714 		m_freem(m0);
2715 		m0 = NULL;
2716 	}
2717 
2718 	/* XXX: clear embedded link IDs in the inner header */
2719 	in6_clearscope(&sip6->ip6_src);
2720 	in6_clearscope(&sip6->ip6_dst);
2721 	in6_clearscope(&nd_rd->nd_rd_target);
2722 	in6_clearscope(&nd_rd->nd_rd_dst);
2723 
2724 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2725 
2726 	nd_rd->nd_rd_cksum = 0;
2727 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2728 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2729 
2730 	/* send the packet to outside... */
2731 #ifdef IPSEC
2732 	/* Don't lookup socket */
2733 	(void)ipsec_setsocket(m, NULL);
2734 #endif /* IPSEC */
2735 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2736 	if (outif) {
2737 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2738 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2739 	}
2740 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2741 
2742 	return;
2743 
2744 fail:
2745 	if (m)
2746 		m_freem(m);
2747 	if (m0)
2748 		m_freem(m0);
2749 }
2750 
2751 /*
2752  * ICMPv6 socket option processing.
2753  */
2754 int
2755 icmp6_ctloutput(so, sopt)
2756 	struct socket *so;
2757 	struct sockopt *sopt;
2758 {
2759 	int error = 0;
2760 	int optlen;
2761 	struct inpcb *inp = sotoinpcb(so);
2762 	int level, op, optname;
2763 
2764 	if (sopt) {
2765 		level = sopt->sopt_level;
2766 		op = sopt->sopt_dir;
2767 		optname = sopt->sopt_name;
2768 		optlen = sopt->sopt_valsize;
2769 	} else
2770 		level = op = optname = optlen = 0;
2771 
2772 	if (level != IPPROTO_ICMPV6) {
2773 		return EINVAL;
2774 	}
2775 
2776 	switch (op) {
2777 	case PRCO_SETOPT:
2778 		switch (optname) {
2779 		case ICMP6_FILTER:
2780 		    {
2781 			struct icmp6_filter *p;
2782 
2783 			if (optlen != sizeof(*p)) {
2784 				error = EMSGSIZE;
2785 				break;
2786 			}
2787 			if (inp->in6p_icmp6filt == NULL) {
2788 				error = EINVAL;
2789 				break;
2790 			}
2791 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2792 				optlen);
2793 			break;
2794 		    }
2795 
2796 		default:
2797 			error = ENOPROTOOPT;
2798 			break;
2799 		}
2800 		break;
2801 
2802 	case PRCO_GETOPT:
2803 		switch (optname) {
2804 		case ICMP6_FILTER:
2805 		    {
2806 			if (inp->in6p_icmp6filt == NULL) {
2807 				error = EINVAL;
2808 				break;
2809 			}
2810 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2811 				sizeof(struct icmp6_filter));
2812 			break;
2813 		    }
2814 
2815 		default:
2816 			error = ENOPROTOOPT;
2817 			break;
2818 		}
2819 		break;
2820 	}
2821 
2822 	return (error);
2823 }
2824 
2825 /*
2826  * Perform rate limit check.
2827  * Returns 0 if it is okay to send the icmp6 packet.
2828  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2829  * limitation.
2830  *
2831  * XXX per-destination/type check necessary?
2832  */
2833 static int
2834 icmp6_ratelimit(dst, type, code)
2835 	const struct in6_addr *dst;	/* not used at this moment */
2836 	const int type;			/* not used at this moment */
2837 	const int code;			/* not used at this moment */
2838 {
2839 	int ret;
2840 
2841 	ret = 0;	/* okay to send */
2842 
2843 	/* PPS limit */
2844 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2845 	    icmp6errppslim)) {
2846 		/* The packet is subject to rate limit */
2847 		ret++;
2848 	}
2849 
2850 	return ret;
2851 }
2852