xref: /freebsd/sys/netinet6/icmp6.c (revision 830940567b49bb0c08dfaed40418999e76616909)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.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_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/icmp6.h>
99 #include <netinet/tcp_var.h>
100 
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/nd6.h>
108 
109 #ifdef IPSEC
110 #include <netipsec/ipsec.h>
111 #include <netipsec/key.h>
112 #endif
113 
114 extern struct domain inet6domain;
115 
116 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
117 VNET_DECLARE(struct inpcbhead, ripcb);
118 VNET_DECLARE(int, icmp6errppslim);
119 VNET_DECLARE(int, icmp6_nodeinfo);
120 
121 #define	V_ripcbinfo			VNET(ripcbinfo)
122 #define	V_ripcb				VNET(ripcb)
123 #define	V_icmp6errppslim		VNET(icmp6errppslim)
124 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
125 
126 VNET_DEFINE(struct icmp6stat, icmp6stat);
127 static VNET_DEFINE(int, icmp6errpps_count);
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 
130 #define	V_icmp6errpps_count	VNET(icmp6errpps_count)
131 #define	V_icmp6errppslim_last	VNET(icmp6errppslim_last)
132 
133 static void icmp6_errcount(struct icmp6errstat *, int, int);
134 static int icmp6_rip6_input(struct mbuf **, int);
135 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
136 static const char *icmp6_redirect_diag __P((struct in6_addr *,
137 	struct in6_addr *, struct in6_addr *));
138 static struct mbuf *ni6_input(struct mbuf *, int);
139 static struct mbuf *ni6_nametodns(const char *, int, int);
140 static int ni6_dnsmatch(const char *, int, const char *, int);
141 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
142 			  struct ifnet **, struct in6_addr *));
143 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
144 				struct ifnet *, int));
145 static int icmp6_notify_error(struct mbuf **, int, int, int);
146 
147 
148 void
149 icmp6_init(void)
150 {
151 
152 	V_icmp6errpps_count = 0;
153 }
154 
155 /*
156  * Kernel module interface for updating icmp6stat.  The argument is an index
157  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
158  * general layout of icmp6stat into the caller, it doesn't encode its
159  * location, so that future changes to add, for example, per-CPU stats
160  * support won't cause binary compatibility problems for kernel modules.
161  */
162 void
163 kmod_icmp6stat_inc(int statnum)
164 {
165 
166 	(*((u_quad_t *)&V_icmp6stat + statnum))++;
167 }
168 
169 static void
170 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
171 {
172 	switch (type) {
173 	case ICMP6_DST_UNREACH:
174 		switch (code) {
175 		case ICMP6_DST_UNREACH_NOROUTE:
176 			stat->icp6errs_dst_unreach_noroute++;
177 			return;
178 		case ICMP6_DST_UNREACH_ADMIN:
179 			stat->icp6errs_dst_unreach_admin++;
180 			return;
181 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
182 			stat->icp6errs_dst_unreach_beyondscope++;
183 			return;
184 		case ICMP6_DST_UNREACH_ADDR:
185 			stat->icp6errs_dst_unreach_addr++;
186 			return;
187 		case ICMP6_DST_UNREACH_NOPORT:
188 			stat->icp6errs_dst_unreach_noport++;
189 			return;
190 		}
191 		break;
192 	case ICMP6_PACKET_TOO_BIG:
193 		stat->icp6errs_packet_too_big++;
194 		return;
195 	case ICMP6_TIME_EXCEEDED:
196 		switch (code) {
197 		case ICMP6_TIME_EXCEED_TRANSIT:
198 			stat->icp6errs_time_exceed_transit++;
199 			return;
200 		case ICMP6_TIME_EXCEED_REASSEMBLY:
201 			stat->icp6errs_time_exceed_reassembly++;
202 			return;
203 		}
204 		break;
205 	case ICMP6_PARAM_PROB:
206 		switch (code) {
207 		case ICMP6_PARAMPROB_HEADER:
208 			stat->icp6errs_paramprob_header++;
209 			return;
210 		case ICMP6_PARAMPROB_NEXTHEADER:
211 			stat->icp6errs_paramprob_nextheader++;
212 			return;
213 		case ICMP6_PARAMPROB_OPTION:
214 			stat->icp6errs_paramprob_option++;
215 			return;
216 		}
217 		break;
218 	case ND_REDIRECT:
219 		stat->icp6errs_redirect++;
220 		return;
221 	}
222 	stat->icp6errs_unknown++;
223 }
224 
225 /*
226  * A wrapper function for icmp6_error() necessary when the erroneous packet
227  * may not contain enough scope zone information.
228  */
229 void
230 icmp6_error2(struct mbuf *m, int type, int code, int param,
231     struct ifnet *ifp)
232 {
233 	struct ip6_hdr *ip6;
234 
235 	if (ifp == NULL)
236 		return;
237 
238 #ifndef PULLDOWN_TEST
239 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
240 #else
241 	if (m->m_len < sizeof(struct ip6_hdr)) {
242 		m = m_pullup(m, sizeof(struct ip6_hdr));
243 		if (m == NULL)
244 			return;
245 	}
246 #endif
247 
248 	ip6 = mtod(m, struct ip6_hdr *);
249 
250 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
251 		return;
252 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
253 		return;
254 
255 	icmp6_error(m, type, code, param);
256 }
257 
258 /*
259  * Generate an error packet of type error in response to bad IP6 packet.
260  */
261 void
262 icmp6_error(struct mbuf *m, int type, int code, int param)
263 {
264 	struct ip6_hdr *oip6, *nip6;
265 	struct icmp6_hdr *icmp6;
266 	u_int preplen;
267 	int off;
268 	int nxt;
269 
270 	ICMP6STAT_INC(icp6s_error);
271 
272 	/* count per-type-code statistics */
273 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code);
274 
275 #ifdef M_DECRYPTED	/*not openbsd*/
276 	if (m->m_flags & M_DECRYPTED) {
277 		ICMP6STAT_INC(icp6s_canterror);
278 		goto freeit;
279 	}
280 #endif
281 
282 #ifndef PULLDOWN_TEST
283 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
284 #else
285 	if (m->m_len < sizeof(struct ip6_hdr)) {
286 		m = m_pullup(m, sizeof(struct ip6_hdr));
287 		if (m == NULL)
288 			return;
289 	}
290 #endif
291 	oip6 = mtod(m, struct ip6_hdr *);
292 
293 	/*
294 	 * If the destination address of the erroneous packet is a multicast
295 	 * address, or the packet was sent using link-layer multicast,
296 	 * we should basically suppress sending an error (RFC 2463, Section
297 	 * 2.4).
298 	 * We have two exceptions (the item e.2 in that section):
299 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
300 	 * - the Parameter Problem Message that can be allowed an icmp6 error
301 	 *   in the option type field.  This check has been done in
302 	 *   ip6_unknown_opt(), so we can just check the type and code.
303 	 */
304 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
305 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
306 	    (type != ICMP6_PACKET_TOO_BIG &&
307 	     (type != ICMP6_PARAM_PROB ||
308 	      code != ICMP6_PARAMPROB_OPTION)))
309 		goto freeit;
310 
311 	/*
312 	 * RFC 2463, 2.4 (e.5): source address check.
313 	 * XXX: the case of anycast source?
314 	 */
315 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
316 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
317 		goto freeit;
318 
319 	/*
320 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
321 	 * don't do it.
322 	 */
323 	nxt = -1;
324 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
325 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
326 		struct icmp6_hdr *icp;
327 
328 #ifndef PULLDOWN_TEST
329 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
330 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
331 #else
332 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
333 			sizeof(*icp));
334 		if (icp == NULL) {
335 			ICMP6STAT_INC(icp6s_tooshort);
336 			return;
337 		}
338 #endif
339 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
340 		    icp->icmp6_type == ND_REDIRECT) {
341 			/*
342 			 * ICMPv6 error
343 			 * Special case: for redirect (which is
344 			 * informational) we must not send icmp6 error.
345 			 */
346 			ICMP6STAT_INC(icp6s_canterror);
347 			goto freeit;
348 		} else {
349 			/* ICMPv6 informational - send the error */
350 		}
351 	} else {
352 		/* non-ICMPv6 - send the error */
353 	}
354 
355 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
356 
357 	/* Finally, do rate limitation check. */
358 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
359 		ICMP6STAT_INC(icp6s_toofreq);
360 		goto freeit;
361 	}
362 
363 	/*
364 	 * OK, ICMP6 can be generated.
365 	 */
366 
367 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
368 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
369 
370 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
371 	M_PREPEND(m, preplen, M_DONTWAIT);
372 	if (m && m->m_len < preplen)
373 		m = m_pullup(m, preplen);
374 	if (m == NULL) {
375 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
376 		return;
377 	}
378 
379 	nip6 = mtod(m, struct ip6_hdr *);
380 	nip6->ip6_src  = oip6->ip6_src;
381 	nip6->ip6_dst  = oip6->ip6_dst;
382 
383 	in6_clearscope(&oip6->ip6_src);
384 	in6_clearscope(&oip6->ip6_dst);
385 
386 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
387 	icmp6->icmp6_type = type;
388 	icmp6->icmp6_code = code;
389 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
390 
391 	/*
392 	 * icmp6_reflect() is designed to be in the input path.
393 	 * icmp6_error() can be called from both input and output path,
394 	 * and if we are in output path rcvif could contain bogus value.
395 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
396 	 * information in ip header (nip6).
397 	 */
398 	m->m_pkthdr.rcvif = NULL;
399 
400 	ICMP6STAT_INC(icp6s_outhist[type]);
401 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
402 
403 	return;
404 
405   freeit:
406 	/*
407 	 * If we can't tell whether or not we can generate ICMP6, free it.
408 	 */
409 	m_freem(m);
410 }
411 
412 /*
413  * Process a received ICMP6 message.
414  */
415 int
416 icmp6_input(struct mbuf **mp, int *offp, int proto)
417 {
418 	struct mbuf *m = *mp, *n;
419 	struct ifnet *ifp;
420 	struct ip6_hdr *ip6, *nip6;
421 	struct icmp6_hdr *icmp6, *nicmp6;
422 	int off = *offp;
423 	int icmp6len = m->m_pkthdr.len - *offp;
424 	int code, sum, noff;
425 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
426 
427 	ifp = m->m_pkthdr.rcvif;
428 
429 #ifndef PULLDOWN_TEST
430 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
431 	/* m might change if M_LOOP.  So, call mtod after this */
432 #endif
433 
434 	/*
435 	 * Locate icmp6 structure in mbuf, and check
436 	 * that not corrupted and of at least minimum length
437 	 */
438 
439 	ip6 = mtod(m, struct ip6_hdr *);
440 	if (icmp6len < sizeof(struct icmp6_hdr)) {
441 		ICMP6STAT_INC(icp6s_tooshort);
442 		goto freeit;
443 	}
444 
445 	/*
446 	 * Check multicast group membership.
447 	 * Note: SSM filters are not applied for ICMPv6 traffic.
448 	 */
449 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
450 		struct in6_multi	*inm;
451 
452 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
453 		if (inm == NULL) {
454 			IP6STAT_INC(ip6s_notmember);
455 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
456 			goto freeit;
457 		}
458 	}
459 
460 	/*
461 	 * calculate the checksum
462 	 */
463 #ifndef PULLDOWN_TEST
464 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
465 #else
466 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
467 	if (icmp6 == NULL) {
468 		ICMP6STAT_INC(icp6s_tooshort);
469 		return IPPROTO_DONE;
470 	}
471 #endif
472 	code = icmp6->icmp6_code;
473 
474 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
475 		nd6log((LOG_ERR,
476 		    "ICMP6 checksum error(%d|%x) %s\n",
477 		    icmp6->icmp6_type, sum,
478 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
479 		ICMP6STAT_INC(icp6s_checksum);
480 		goto freeit;
481 	}
482 
483 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
484 		/*
485 		 * Deliver very specific ICMP6 type only.
486 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
487 		 * will not work.
488 		 */
489 		switch (icmp6->icmp6_type) {
490 		case ICMP6_DST_UNREACH:
491 		case ICMP6_PACKET_TOO_BIG:
492 		case ICMP6_TIME_EXCEEDED:
493 			break;
494 		default:
495 			goto freeit;
496 		}
497 	}
498 
499 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
500 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
501 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
502 		icmp6_ifstat_inc(ifp, ifs6_in_error);
503 
504 	switch (icmp6->icmp6_type) {
505 	case ICMP6_DST_UNREACH:
506 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
507 		switch (code) {
508 		case ICMP6_DST_UNREACH_NOROUTE:
509 			code = PRC_UNREACH_NET;
510 			break;
511 		case ICMP6_DST_UNREACH_ADMIN:
512 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
513 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
514 			break;
515 		case ICMP6_DST_UNREACH_ADDR:
516 			code = PRC_HOSTDEAD;
517 			break;
518 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
519 			/* I mean "source address was incorrect." */
520 			code = PRC_PARAMPROB;
521 			break;
522 		case ICMP6_DST_UNREACH_NOPORT:
523 			code = PRC_UNREACH_PORT;
524 			break;
525 		default:
526 			goto badcode;
527 		}
528 		goto deliver;
529 		break;
530 
531 	case ICMP6_PACKET_TOO_BIG:
532 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
533 
534 		/* validation is made in icmp6_mtudisc_update */
535 
536 		code = PRC_MSGSIZE;
537 
538 		/*
539 		 * Updating the path MTU will be done after examining
540 		 * intermediate extension headers.
541 		 */
542 		goto deliver;
543 		break;
544 
545 	case ICMP6_TIME_EXCEEDED:
546 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
547 		switch (code) {
548 		case ICMP6_TIME_EXCEED_TRANSIT:
549 			code = PRC_TIMXCEED_INTRANS;
550 			break;
551 		case ICMP6_TIME_EXCEED_REASSEMBLY:
552 			code = PRC_TIMXCEED_REASS;
553 			break;
554 		default:
555 			goto badcode;
556 		}
557 		goto deliver;
558 		break;
559 
560 	case ICMP6_PARAM_PROB:
561 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
562 		switch (code) {
563 		case ICMP6_PARAMPROB_NEXTHEADER:
564 			code = PRC_UNREACH_PROTOCOL;
565 			break;
566 		case ICMP6_PARAMPROB_HEADER:
567 		case ICMP6_PARAMPROB_OPTION:
568 			code = PRC_PARAMPROB;
569 			break;
570 		default:
571 			goto badcode;
572 		}
573 		goto deliver;
574 		break;
575 
576 	case ICMP6_ECHO_REQUEST:
577 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
578 		if (code != 0)
579 			goto badcode;
580 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
581 			/* Give up remote */
582 			break;
583 		}
584 		if ((n->m_flags & M_EXT) != 0
585 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
586 			struct mbuf *n0 = n;
587 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
588 			int n0len;
589 
590 			MGETHDR(n, M_DONTWAIT, n0->m_type);
591 			n0len = n0->m_pkthdr.len;	/* save for use below */
592 			if (n)
593 				M_MOVE_PKTHDR(n, n0);
594 			if (n && maxlen >= MHLEN) {
595 				MCLGET(n, M_DONTWAIT);
596 				if ((n->m_flags & M_EXT) == 0) {
597 					m_free(n);
598 					n = NULL;
599 				}
600 			}
601 			if (n == NULL) {
602 				/* Give up remote */
603 				m_freem(n0);
604 				break;
605 			}
606 			/*
607 			 * Copy IPv6 and ICMPv6 only.
608 			 */
609 			nip6 = mtod(n, struct ip6_hdr *);
610 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
611 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
612 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
613 			noff = sizeof(struct ip6_hdr);
614 			/* new mbuf contains only ipv6+icmpv6 headers */
615 			n->m_len = noff + sizeof(struct icmp6_hdr);
616 			/*
617 			 * Adjust mbuf.  ip6_plen will be adjusted in
618 			 * ip6_output().
619 			 */
620 			m_adj(n0, off + sizeof(struct icmp6_hdr));
621 			/* recalculate complete packet size */
622 			n->m_pkthdr.len = n0len + (noff - off);
623 			n->m_next = n0;
624 		} else {
625 			nip6 = mtod(n, struct ip6_hdr *);
626 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
627 			    sizeof(*nicmp6));
628 			noff = off;
629 		}
630 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
631 		nicmp6->icmp6_code = 0;
632 		if (n) {
633 			ICMP6STAT_INC(icp6s_reflect);
634 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
635 			icmp6_reflect(n, noff);
636 		}
637 		break;
638 
639 	case ICMP6_ECHO_REPLY:
640 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
641 		if (code != 0)
642 			goto badcode;
643 		break;
644 
645 	case MLD_LISTENER_QUERY:
646 	case MLD_LISTENER_REPORT:
647 	case MLD_LISTENER_DONE:
648 	case MLDV2_LISTENER_REPORT:
649 		/*
650 		 * Drop MLD traffic which is not link-local, has a hop limit
651 		 * of greater than 1 hop, or which does not have the
652 		 * IPv6 HBH Router Alert option.
653 		 * As IPv6 HBH options are stripped in ip6_input() we must
654 		 * check an mbuf header flag.
655 		 * XXX Should we also sanity check that these messages
656 		 * were directed to a link-local multicast prefix?
657 		 */
658 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
659 			goto freeit;
660 		if (mld_input(m, off, icmp6len) != 0)
661 			return (IPPROTO_DONE);
662 		/* m stays. */
663 		break;
664 
665 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
666 	    {
667 		enum { WRU, FQDN } mode;
668 
669 		if (!V_icmp6_nodeinfo)
670 			break;
671 
672 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
673 			mode = WRU;
674 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
675 			mode = FQDN;
676 		else
677 			goto badlen;
678 
679 		if (mode == FQDN) {
680 #ifndef PULLDOWN_TEST
681 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
682 			    IPPROTO_DONE);
683 #endif
684 			n = m_copy(m, 0, M_COPYALL);
685 			if (n)
686 				n = ni6_input(n, off);
687 			/* XXX meaningless if n == NULL */
688 			noff = sizeof(struct ip6_hdr);
689 		} else {
690 			struct prison *pr;
691 			u_char *p;
692 			int maxlen, maxhlen, hlen;
693 
694 			/*
695 			 * XXX: this combination of flags is pointless,
696 			 * but should we keep this for compatibility?
697 			 */
698 			if ((V_icmp6_nodeinfo & 5) != 5)
699 				break;
700 
701 			if (code != 0)
702 				goto badcode;
703 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
704 			if (maxlen >= MCLBYTES) {
705 				/* Give up remote */
706 				break;
707 			}
708 			MGETHDR(n, M_DONTWAIT, m->m_type);
709 			if (n && maxlen > MHLEN) {
710 				MCLGET(n, M_DONTWAIT);
711 				if ((n->m_flags & M_EXT) == 0) {
712 					m_free(n);
713 					n = NULL;
714 				}
715 			}
716 			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
717 				/*
718 				 * Previous code did a blind M_COPY_PKTHDR
719 				 * and said "just for rcvif".  If true, then
720 				 * we could tolerate the dup failing (due to
721 				 * the deep copy of the tag chain).  For now
722 				 * be conservative and just fail.
723 				 */
724 				m_free(n);
725 				n = NULL;
726 			}
727 			if (n == NULL) {
728 				/* Give up remote */
729 				break;
730 			}
731 			n->m_pkthdr.rcvif = NULL;
732 			n->m_len = 0;
733 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
734 			pr = curthread->td_ucred->cr_prison;
735 			mtx_lock(&pr->pr_mtx);
736 			hlen = strlen(pr->pr_hostname);
737 			if (maxhlen > hlen)
738 				maxhlen = hlen;
739 			/*
740 			 * Copy IPv6 and ICMPv6 only.
741 			 */
742 			nip6 = mtod(n, struct ip6_hdr *);
743 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
744 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
745 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
746 			p = (u_char *)(nicmp6 + 1);
747 			bzero(p, 4);
748 			/* meaningless TTL */
749 			bcopy(pr->pr_hostname, p + 4, maxhlen);
750 			mtx_unlock(&pr->pr_mtx);
751 			noff = sizeof(struct ip6_hdr);
752 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
753 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
754 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
755 			nicmp6->icmp6_code = 0;
756 		}
757 		if (n) {
758 			ICMP6STAT_INC(icp6s_reflect);
759 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
760 			icmp6_reflect(n, noff);
761 		}
762 		break;
763 	    }
764 
765 	case ICMP6_WRUREPLY:
766 		if (code != 0)
767 			goto badcode;
768 		break;
769 
770 	case ND_ROUTER_SOLICIT:
771 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
772 		if (code != 0)
773 			goto badcode;
774 		if (icmp6len < sizeof(struct nd_router_solicit))
775 			goto badlen;
776 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
777 			/* give up local */
778 			nd6_rs_input(m, off, icmp6len);
779 			m = NULL;
780 			goto freeit;
781 		}
782 		nd6_rs_input(n, off, icmp6len);
783 		/* m stays. */
784 		break;
785 
786 	case ND_ROUTER_ADVERT:
787 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
788 		if (code != 0)
789 			goto badcode;
790 		if (icmp6len < sizeof(struct nd_router_advert))
791 			goto badlen;
792 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
793 			/* give up local */
794 			nd6_ra_input(m, off, icmp6len);
795 			m = NULL;
796 			goto freeit;
797 		}
798 		nd6_ra_input(n, off, icmp6len);
799 		/* m stays. */
800 		break;
801 
802 	case ND_NEIGHBOR_SOLICIT:
803 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
804 		if (code != 0)
805 			goto badcode;
806 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
807 			goto badlen;
808 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
809 			/* give up local */
810 			nd6_ns_input(m, off, icmp6len);
811 			m = NULL;
812 			goto freeit;
813 		}
814 		nd6_ns_input(n, off, icmp6len);
815 		/* m stays. */
816 		break;
817 
818 	case ND_NEIGHBOR_ADVERT:
819 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
820 		if (code != 0)
821 			goto badcode;
822 		if (icmp6len < sizeof(struct nd_neighbor_advert))
823 			goto badlen;
824 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
825 			/* give up local */
826 			nd6_na_input(m, off, icmp6len);
827 			m = NULL;
828 			goto freeit;
829 		}
830 		nd6_na_input(n, off, icmp6len);
831 		/* m stays. */
832 		break;
833 
834 	case ND_REDIRECT:
835 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
836 		if (code != 0)
837 			goto badcode;
838 		if (icmp6len < sizeof(struct nd_redirect))
839 			goto badlen;
840 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
841 			/* give up local */
842 			icmp6_redirect_input(m, off);
843 			m = NULL;
844 			goto freeit;
845 		}
846 		icmp6_redirect_input(n, off);
847 		/* m stays. */
848 		break;
849 
850 	case ICMP6_ROUTER_RENUMBERING:
851 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
852 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
853 			goto badcode;
854 		if (icmp6len < sizeof(struct icmp6_router_renum))
855 			goto badlen;
856 		break;
857 
858 	default:
859 		nd6log((LOG_DEBUG,
860 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
861 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
862 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
863 		    ifp ? ifp->if_index : 0));
864 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
865 			/* ICMPv6 error: MUST deliver it by spec... */
866 			code = PRC_NCMDS;
867 			/* deliver */
868 		} else {
869 			/* ICMPv6 informational: MUST not deliver */
870 			break;
871 		}
872 	deliver:
873 		if (icmp6_notify_error(&m, off, icmp6len, code)) {
874 			/* In this case, m should've been freed. */
875 			return (IPPROTO_DONE);
876 		}
877 		break;
878 
879 	badcode:
880 		ICMP6STAT_INC(icp6s_badcode);
881 		break;
882 
883 	badlen:
884 		ICMP6STAT_INC(icp6s_badlen);
885 		break;
886 	}
887 
888 	/* deliver the packet to appropriate sockets */
889 	icmp6_rip6_input(&m, *offp);
890 
891 	return IPPROTO_DONE;
892 
893  freeit:
894 	m_freem(m);
895 	return IPPROTO_DONE;
896 }
897 
898 static int
899 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
900 {
901 	struct mbuf *m = *mp;
902 	struct icmp6_hdr *icmp6;
903 	struct ip6_hdr *eip6;
904 	u_int32_t notifymtu;
905 	struct sockaddr_in6 icmp6src, icmp6dst;
906 
907 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
908 		ICMP6STAT_INC(icp6s_tooshort);
909 		goto freeit;
910 	}
911 #ifndef PULLDOWN_TEST
912 	IP6_EXTHDR_CHECK(m, off,
913 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
914 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
915 #else
916 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
917 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
918 	if (icmp6 == NULL) {
919 		ICMP6STAT_INC(icp6s_tooshort);
920 		return (-1);
921 	}
922 #endif
923 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
924 
925 	/* Detect the upper level protocol */
926 	{
927 		void (*ctlfunc)(int, struct sockaddr *, void *);
928 		u_int8_t nxt = eip6->ip6_nxt;
929 		int eoff = off + sizeof(struct icmp6_hdr) +
930 		    sizeof(struct ip6_hdr);
931 		struct ip6ctlparam ip6cp;
932 		struct in6_addr *finaldst = NULL;
933 		int icmp6type = icmp6->icmp6_type;
934 		struct ip6_frag *fh;
935 		struct ip6_rthdr *rth;
936 		struct ip6_rthdr0 *rth0;
937 		int rthlen;
938 
939 		while (1) { /* XXX: should avoid infinite loop explicitly? */
940 			struct ip6_ext *eh;
941 
942 			switch (nxt) {
943 			case IPPROTO_HOPOPTS:
944 			case IPPROTO_DSTOPTS:
945 			case IPPROTO_AH:
946 #ifndef PULLDOWN_TEST
947 				IP6_EXTHDR_CHECK(m, 0,
948 				    eoff + sizeof(struct ip6_ext), -1);
949 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
950 #else
951 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
952 				    eoff, sizeof(*eh));
953 				if (eh == NULL) {
954 					ICMP6STAT_INC(icp6s_tooshort);
955 					return (-1);
956 				}
957 #endif
958 
959 				if (nxt == IPPROTO_AH)
960 					eoff += (eh->ip6e_len + 2) << 2;
961 				else
962 					eoff += (eh->ip6e_len + 1) << 3;
963 				nxt = eh->ip6e_nxt;
964 				break;
965 			case IPPROTO_ROUTING:
966 				/*
967 				 * When the erroneous packet contains a
968 				 * routing header, we should examine the
969 				 * header to determine the final destination.
970 				 * Otherwise, we can't properly update
971 				 * information that depends on the final
972 				 * destination (e.g. path MTU).
973 				 */
974 #ifndef PULLDOWN_TEST
975 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
976 				rth = (struct ip6_rthdr *)
977 				    (mtod(m, caddr_t) + eoff);
978 #else
979 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
980 				    eoff, sizeof(*rth));
981 				if (rth == NULL) {
982 					ICMP6STAT_INC(icp6s_tooshort);
983 					return (-1);
984 				}
985 #endif
986 				rthlen = (rth->ip6r_len + 1) << 3;
987 				/*
988 				 * XXX: currently there is no
989 				 * officially defined type other
990 				 * than type-0.
991 				 * Note that if the segment left field
992 				 * is 0, all intermediate hops must
993 				 * have been passed.
994 				 */
995 				if (rth->ip6r_segleft &&
996 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
997 					int hops;
998 
999 #ifndef PULLDOWN_TEST
1000 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
1001 					rth0 = (struct ip6_rthdr0 *)
1002 					    (mtod(m, caddr_t) + eoff);
1003 #else
1004 					IP6_EXTHDR_GET(rth0,
1005 					    struct ip6_rthdr0 *, m,
1006 					    eoff, rthlen);
1007 					if (rth0 == NULL) {
1008 						ICMP6STAT_INC(icp6s_tooshort);
1009 						return (-1);
1010 					}
1011 #endif
1012 					/* just ignore a bogus header */
1013 					if ((rth0->ip6r0_len % 2) == 0 &&
1014 					    (hops = rth0->ip6r0_len/2))
1015 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1016 				}
1017 				eoff += rthlen;
1018 				nxt = rth->ip6r_nxt;
1019 				break;
1020 			case IPPROTO_FRAGMENT:
1021 #ifndef PULLDOWN_TEST
1022 				IP6_EXTHDR_CHECK(m, 0, eoff +
1023 				    sizeof(struct ip6_frag), -1);
1024 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1025 				    eoff);
1026 #else
1027 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1028 				    eoff, sizeof(*fh));
1029 				if (fh == NULL) {
1030 					ICMP6STAT_INC(icp6s_tooshort);
1031 					return (-1);
1032 				}
1033 #endif
1034 				/*
1035 				 * Data after a fragment header is meaningless
1036 				 * unless it is the first fragment, but
1037 				 * we'll go to the notify label for path MTU
1038 				 * discovery.
1039 				 */
1040 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1041 					goto notify;
1042 
1043 				eoff += sizeof(struct ip6_frag);
1044 				nxt = fh->ip6f_nxt;
1045 				break;
1046 			default:
1047 				/*
1048 				 * This case includes ESP and the No Next
1049 				 * Header.  In such cases going to the notify
1050 				 * label does not have any meaning
1051 				 * (i.e. ctlfunc will be NULL), but we go
1052 				 * anyway since we might have to update
1053 				 * path MTU information.
1054 				 */
1055 				goto notify;
1056 			}
1057 		}
1058 	  notify:
1059 #ifndef PULLDOWN_TEST
1060 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1061 #else
1062 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1063 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1064 		if (icmp6 == NULL) {
1065 			ICMP6STAT_INC(icp6s_tooshort);
1066 			return (-1);
1067 		}
1068 #endif
1069 
1070 		/*
1071 		 * retrieve parameters from the inner IPv6 header, and convert
1072 		 * them into sockaddr structures.
1073 		 * XXX: there is no guarantee that the source or destination
1074 		 * addresses of the inner packet are in the same scope as
1075 		 * the addresses of the icmp packet.  But there is no other
1076 		 * way to determine the zone.
1077 		 */
1078 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1079 
1080 		bzero(&icmp6dst, sizeof(icmp6dst));
1081 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1082 		icmp6dst.sin6_family = AF_INET6;
1083 		if (finaldst == NULL)
1084 			icmp6dst.sin6_addr = eip6->ip6_dst;
1085 		else
1086 			icmp6dst.sin6_addr = *finaldst;
1087 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1088 			goto freeit;
1089 		bzero(&icmp6src, sizeof(icmp6src));
1090 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1091 		icmp6src.sin6_family = AF_INET6;
1092 		icmp6src.sin6_addr = eip6->ip6_src;
1093 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1094 			goto freeit;
1095 		icmp6src.sin6_flowinfo =
1096 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1097 
1098 		if (finaldst == NULL)
1099 			finaldst = &eip6->ip6_dst;
1100 		ip6cp.ip6c_m = m;
1101 		ip6cp.ip6c_icmp6 = icmp6;
1102 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1103 		ip6cp.ip6c_off = eoff;
1104 		ip6cp.ip6c_finaldst = finaldst;
1105 		ip6cp.ip6c_src = &icmp6src;
1106 		ip6cp.ip6c_nxt = nxt;
1107 
1108 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1109 			notifymtu = ntohl(icmp6->icmp6_mtu);
1110 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1111 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1112 		}
1113 
1114 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1115 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1116 		if (ctlfunc) {
1117 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1118 			    &ip6cp);
1119 		}
1120 	}
1121 	*mp = m;
1122 	return (0);
1123 
1124   freeit:
1125 	m_freem(m);
1126 	return (-1);
1127 }
1128 
1129 void
1130 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1131 {
1132 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1133 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1134 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1135 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1136 	struct in_conninfo inc;
1137 
1138 #if 0
1139 	/*
1140 	 * RFC2460 section 5, last paragraph.
1141 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1142 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1143 	 * due to packet translator in the middle.
1144 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1145 	 * special handling.
1146 	 */
1147 	if (mtu < IPV6_MMTU)
1148 		return;
1149 #endif
1150 
1151 	/*
1152 	 * we reject ICMPv6 too big with abnormally small value.
1153 	 * XXX what is the good definition of "abnormally small"?
1154 	 */
1155 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1156 		return;
1157 
1158 	if (!validated)
1159 		return;
1160 
1161 	/*
1162 	 * In case the suggested mtu is less than IPV6_MMTU, we
1163 	 * only need to remember that it was for above mentioned
1164 	 * "alwaysfrag" case.
1165 	 * Try to be as close to the spec as possible.
1166 	 */
1167 	if (mtu < IPV6_MMTU)
1168 		mtu = IPV6_MMTU - 8;
1169 
1170 	bzero(&inc, sizeof(inc));
1171 	inc.inc_flags |= INC_ISIPV6;
1172 	inc.inc6_faddr = *dst;
1173 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1174 		return;
1175 
1176 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1177 		tcp_hc_updatemtu(&inc, mtu);
1178 		ICMP6STAT_INC(icp6s_pmtuchg);
1179 	}
1180 }
1181 
1182 /*
1183  * Process a Node Information Query packet, based on
1184  * draft-ietf-ipngwg-icmp-name-lookups-07.
1185  *
1186  * Spec incompatibilities:
1187  * - IPv6 Subject address handling
1188  * - IPv4 Subject address handling support missing
1189  * - Proxy reply (answer even if it's not for me)
1190  * - joins NI group address at in6_ifattach() time only, does not cope
1191  *   with hostname changes by sethostname(3)
1192  */
1193 static struct mbuf *
1194 ni6_input(struct mbuf *m, int off)
1195 {
1196 	struct icmp6_nodeinfo *ni6, *nni6;
1197 	struct mbuf *n = NULL;
1198 	struct prison *pr;
1199 	u_int16_t qtype;
1200 	int subjlen;
1201 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1202 	struct ni_reply_fqdn *fqdn;
1203 	int addrs;		/* for NI_QTYPE_NODEADDR */
1204 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1205 	struct in6_addr in6_subj; /* subject address */
1206 	struct ip6_hdr *ip6;
1207 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1208 	char *subj = NULL;
1209 	struct in6_ifaddr *ia6 = NULL;
1210 
1211 	ip6 = mtod(m, struct ip6_hdr *);
1212 #ifndef PULLDOWN_TEST
1213 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1214 #else
1215 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1216 	if (ni6 == NULL) {
1217 		/* m is already reclaimed */
1218 		return (NULL);
1219 	}
1220 #endif
1221 
1222 	/*
1223 	 * Validate IPv6 source address.
1224 	 * The default configuration MUST be to refuse answering queries from
1225 	 * global-scope addresses according to RFC4602.
1226 	 * Notes:
1227 	 *  - it's not very clear what "refuse" means; this implementation
1228 	 *    simply drops it.
1229 	 *  - it's not very easy to identify global-scope (unicast) addresses
1230 	 *    since there are many prefixes for them.  It should be safer
1231 	 *    and in practice sufficient to check "all" but loopback and
1232 	 *    link-local (note that site-local unicast was deprecated and
1233 	 *    ULA is defined as global scope-wise)
1234 	 */
1235 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1236 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1237 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1238 		goto bad;
1239 
1240 	/*
1241 	 * Validate IPv6 destination address.
1242 	 *
1243 	 * The Responder must discard the Query without further processing
1244 	 * unless it is one of the Responder's unicast or anycast addresses, or
1245 	 * a link-local scope multicast address which the Responder has joined.
1246 	 * [RFC4602, Section 5.]
1247 	 */
1248 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1249 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1250 			goto bad;
1251 		/* else it's a link-local multicast, fine */
1252 	} else {		/* unicast or anycast */
1253 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1254 			goto bad; /* XXX impossible */
1255 
1256 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1257 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1258 			ifa_free(&ia6->ia_ifa);
1259 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1260 				"a temporary address in %s:%d",
1261 			       __FILE__, __LINE__));
1262 			goto bad;
1263 		}
1264 		ifa_free(&ia6->ia_ifa);
1265 	}
1266 
1267 	/* validate query Subject field. */
1268 	qtype = ntohs(ni6->ni_qtype);
1269 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1270 	switch (qtype) {
1271 	case NI_QTYPE_NOOP:
1272 	case NI_QTYPE_SUPTYPES:
1273 		/* 07 draft */
1274 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1275 			break;
1276 		/* FALLTHROUGH */
1277 	case NI_QTYPE_FQDN:
1278 	case NI_QTYPE_NODEADDR:
1279 	case NI_QTYPE_IPV4ADDR:
1280 		switch (ni6->ni_code) {
1281 		case ICMP6_NI_SUBJ_IPV6:
1282 #if ICMP6_NI_SUBJ_IPV6 != 0
1283 		case 0:
1284 #endif
1285 			/*
1286 			 * backward compatibility - try to accept 03 draft
1287 			 * format, where no Subject is present.
1288 			 */
1289 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1290 			    subjlen == 0) {
1291 				oldfqdn++;
1292 				break;
1293 			}
1294 #if ICMP6_NI_SUBJ_IPV6 != 0
1295 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1296 				goto bad;
1297 #endif
1298 
1299 			if (subjlen != sizeof(struct in6_addr))
1300 				goto bad;
1301 
1302 			/*
1303 			 * Validate Subject address.
1304 			 *
1305 			 * Not sure what exactly "address belongs to the node"
1306 			 * means in the spec, is it just unicast, or what?
1307 			 *
1308 			 * At this moment we consider Subject address as
1309 			 * "belong to the node" if the Subject address equals
1310 			 * to the IPv6 destination address; validation for
1311 			 * IPv6 destination address should have done enough
1312 			 * check for us.
1313 			 *
1314 			 * We do not do proxy at this moment.
1315 			 */
1316 			/* m_pulldown instead of copy? */
1317 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1318 			    subjlen, (caddr_t)&in6_subj);
1319 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1320 				goto bad;
1321 
1322 			subj = (char *)&in6_subj;
1323 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1324 				break;
1325 
1326 			/*
1327 			 * XXX if we are to allow other cases, we should really
1328 			 * be careful about scope here.
1329 			 * basically, we should disallow queries toward IPv6
1330 			 * destination X with subject Y,
1331 			 * if scope(X) > scope(Y).
1332 			 * if we allow scope(X) > scope(Y), it will result in
1333 			 * information leakage across scope boundary.
1334 			 */
1335 			goto bad;
1336 
1337 		case ICMP6_NI_SUBJ_FQDN:
1338 			/*
1339 			 * Validate Subject name with gethostname(3).
1340 			 *
1341 			 * The behavior may need some debate, since:
1342 			 * - we are not sure if the node has FQDN as
1343 			 *   hostname (returned by gethostname(3)).
1344 			 * - the code does wildcard match for truncated names.
1345 			 *   however, we are not sure if we want to perform
1346 			 *   wildcard match, if gethostname(3) side has
1347 			 *   truncated hostname.
1348 			 */
1349 			pr = curthread->td_ucred->cr_prison;
1350 			mtx_lock(&pr->pr_mtx);
1351 			n = ni6_nametodns(pr->pr_hostname,
1352 			    strlen(pr->pr_hostname), 0);
1353 			mtx_unlock(&pr->pr_mtx);
1354 			if (!n || n->m_next || n->m_len == 0)
1355 				goto bad;
1356 			IP6_EXTHDR_GET(subj, char *, m,
1357 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1358 			if (subj == NULL)
1359 				goto bad;
1360 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1361 			    n->m_len)) {
1362 				goto bad;
1363 			}
1364 			m_freem(n);
1365 			n = NULL;
1366 			break;
1367 
1368 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1369 		default:
1370 			goto bad;
1371 		}
1372 		break;
1373 	}
1374 
1375 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1376 	switch (qtype) {
1377 	case NI_QTYPE_FQDN:
1378 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1379 			goto bad;
1380 		break;
1381 	case NI_QTYPE_NODEADDR:
1382 	case NI_QTYPE_IPV4ADDR:
1383 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1384 			goto bad;
1385 		break;
1386 	}
1387 
1388 	/* guess reply length */
1389 	switch (qtype) {
1390 	case NI_QTYPE_NOOP:
1391 		break;		/* no reply data */
1392 	case NI_QTYPE_SUPTYPES:
1393 		replylen += sizeof(u_int32_t);
1394 		break;
1395 	case NI_QTYPE_FQDN:
1396 		/* XXX will append an mbuf */
1397 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1398 		break;
1399 	case NI_QTYPE_NODEADDR:
1400 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1401 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1402 		    sizeof(u_int32_t))) > MCLBYTES)
1403 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1404 		break;
1405 	case NI_QTYPE_IPV4ADDR:
1406 		/* unsupported - should respond with unknown Qtype? */
1407 		break;
1408 	default:
1409 		/*
1410 		 * XXX: We must return a reply with the ICMP6 code
1411 		 * `unknown Qtype' in this case.  However we regard the case
1412 		 * as an FQDN query for backward compatibility.
1413 		 * Older versions set a random value to this field,
1414 		 * so it rarely varies in the defined qtypes.
1415 		 * But the mechanism is not reliable...
1416 		 * maybe we should obsolete older versions.
1417 		 */
1418 		qtype = NI_QTYPE_FQDN;
1419 		/* XXX will append an mbuf */
1420 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1421 		oldfqdn++;
1422 		break;
1423 	}
1424 
1425 	/* allocate an mbuf to reply. */
1426 	MGETHDR(n, M_DONTWAIT, m->m_type);
1427 	if (n == NULL) {
1428 		m_freem(m);
1429 		return (NULL);
1430 	}
1431 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1432 	if (replylen > MHLEN) {
1433 		if (replylen > MCLBYTES) {
1434 			/*
1435 			 * XXX: should we try to allocate more? But MCLBYTES
1436 			 * is probably much larger than IPV6_MMTU...
1437 			 */
1438 			goto bad;
1439 		}
1440 		MCLGET(n, M_DONTWAIT);
1441 		if ((n->m_flags & M_EXT) == 0) {
1442 			goto bad;
1443 		}
1444 	}
1445 	n->m_pkthdr.len = n->m_len = replylen;
1446 
1447 	/* copy mbuf header and IPv6 + Node Information base headers */
1448 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1449 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1450 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1451 
1452 	/* qtype dependent procedure */
1453 	switch (qtype) {
1454 	case NI_QTYPE_NOOP:
1455 		nni6->ni_code = ICMP6_NI_SUCCESS;
1456 		nni6->ni_flags = 0;
1457 		break;
1458 	case NI_QTYPE_SUPTYPES:
1459 	{
1460 		u_int32_t v;
1461 		nni6->ni_code = ICMP6_NI_SUCCESS;
1462 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1463 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1464 		v = (u_int32_t)htonl(0x0000000f);
1465 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1466 		break;
1467 	}
1468 	case NI_QTYPE_FQDN:
1469 		nni6->ni_code = ICMP6_NI_SUCCESS;
1470 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1471 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1472 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1473 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1474 		/*
1475 		 * XXX do we really have FQDN in hostname?
1476 		 */
1477 		pr = curthread->td_ucred->cr_prison;
1478 		mtx_lock(&pr->pr_mtx);
1479 		n->m_next = ni6_nametodns(pr->pr_hostname,
1480 		    strlen(pr->pr_hostname), oldfqdn);
1481 		mtx_unlock(&pr->pr_mtx);
1482 		if (n->m_next == NULL)
1483 			goto bad;
1484 		/* XXX we assume that n->m_next is not a chain */
1485 		if (n->m_next->m_next != NULL)
1486 			goto bad;
1487 		n->m_pkthdr.len += n->m_next->m_len;
1488 		break;
1489 	case NI_QTYPE_NODEADDR:
1490 	{
1491 		int lenlim, copied;
1492 
1493 		nni6->ni_code = ICMP6_NI_SUCCESS;
1494 		n->m_pkthdr.len = n->m_len =
1495 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1496 		lenlim = M_TRAILINGSPACE(n);
1497 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1498 		/* XXX: reset mbuf length */
1499 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1500 		    sizeof(struct icmp6_nodeinfo) + copied;
1501 		break;
1502 	}
1503 	default:
1504 		break;		/* XXX impossible! */
1505 	}
1506 
1507 	nni6->ni_type = ICMP6_NI_REPLY;
1508 	m_freem(m);
1509 	return (n);
1510 
1511   bad:
1512 	m_freem(m);
1513 	if (n)
1514 		m_freem(n);
1515 	return (NULL);
1516 }
1517 
1518 /*
1519  * make a mbuf with DNS-encoded string.  no compression support.
1520  *
1521  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1522  * treated as truncated name (two \0 at the end).  this is a wild guess.
1523  *
1524  * old - return pascal string if non-zero
1525  */
1526 static struct mbuf *
1527 ni6_nametodns(const char *name, int namelen, int old)
1528 {
1529 	struct mbuf *m;
1530 	char *cp, *ep;
1531 	const char *p, *q;
1532 	int i, len, nterm;
1533 
1534 	if (old)
1535 		len = namelen + 1;
1536 	else
1537 		len = MCLBYTES;
1538 
1539 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1540 	MGET(m, M_DONTWAIT, MT_DATA);
1541 	if (m && len > MLEN) {
1542 		MCLGET(m, M_DONTWAIT);
1543 		if ((m->m_flags & M_EXT) == 0)
1544 			goto fail;
1545 	}
1546 	if (!m)
1547 		goto fail;
1548 	m->m_next = NULL;
1549 
1550 	if (old) {
1551 		m->m_len = len;
1552 		*mtod(m, char *) = namelen;
1553 		bcopy(name, mtod(m, char *) + 1, namelen);
1554 		return m;
1555 	} else {
1556 		m->m_len = 0;
1557 		cp = mtod(m, char *);
1558 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1559 
1560 		/* if not certain about my name, return empty buffer */
1561 		if (namelen == 0)
1562 			return m;
1563 
1564 		/*
1565 		 * guess if it looks like shortened hostname, or FQDN.
1566 		 * shortened hostname needs two trailing "\0".
1567 		 */
1568 		i = 0;
1569 		for (p = name; p < name + namelen; p++) {
1570 			if (*p && *p == '.')
1571 				i++;
1572 		}
1573 		if (i < 2)
1574 			nterm = 2;
1575 		else
1576 			nterm = 1;
1577 
1578 		p = name;
1579 		while (cp < ep && p < name + namelen) {
1580 			i = 0;
1581 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1582 				i++;
1583 			/* result does not fit into mbuf */
1584 			if (cp + i + 1 >= ep)
1585 				goto fail;
1586 			/*
1587 			 * DNS label length restriction, RFC1035 page 8.
1588 			 * "i == 0" case is included here to avoid returning
1589 			 * 0-length label on "foo..bar".
1590 			 */
1591 			if (i <= 0 || i >= 64)
1592 				goto fail;
1593 			*cp++ = i;
1594 			bcopy(p, cp, i);
1595 			cp += i;
1596 			p = q;
1597 			if (p < name + namelen && *p == '.')
1598 				p++;
1599 		}
1600 		/* termination */
1601 		if (cp + nterm >= ep)
1602 			goto fail;
1603 		while (nterm-- > 0)
1604 			*cp++ = '\0';
1605 		m->m_len = cp - mtod(m, char *);
1606 		return m;
1607 	}
1608 
1609 	panic("should not reach here");
1610 	/* NOTREACHED */
1611 
1612  fail:
1613 	if (m)
1614 		m_freem(m);
1615 	return NULL;
1616 }
1617 
1618 /*
1619  * check if two DNS-encoded string matches.  takes care of truncated
1620  * form (with \0\0 at the end).  no compression support.
1621  * XXX upper/lowercase match (see RFC2065)
1622  */
1623 static int
1624 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1625 {
1626 	const char *a0, *b0;
1627 	int l;
1628 
1629 	/* simplest case - need validation? */
1630 	if (alen == blen && bcmp(a, b, alen) == 0)
1631 		return 1;
1632 
1633 	a0 = a;
1634 	b0 = b;
1635 
1636 	/* termination is mandatory */
1637 	if (alen < 2 || blen < 2)
1638 		return 0;
1639 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1640 		return 0;
1641 	alen--;
1642 	blen--;
1643 
1644 	while (a - a0 < alen && b - b0 < blen) {
1645 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1646 			return 0;
1647 
1648 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1649 			return 0;
1650 		/* we don't support compression yet */
1651 		if (a[0] >= 64 || b[0] >= 64)
1652 			return 0;
1653 
1654 		/* truncated case */
1655 		if (a[0] == 0 && a - a0 == alen - 1)
1656 			return 1;
1657 		if (b[0] == 0 && b - b0 == blen - 1)
1658 			return 1;
1659 		if (a[0] == 0 || b[0] == 0)
1660 			return 0;
1661 
1662 		if (a[0] != b[0])
1663 			return 0;
1664 		l = a[0];
1665 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1666 			return 0;
1667 		if (bcmp(a + 1, b + 1, l) != 0)
1668 			return 0;
1669 
1670 		a += 1 + l;
1671 		b += 1 + l;
1672 	}
1673 
1674 	if (a - a0 == alen && b - b0 == blen)
1675 		return 1;
1676 	else
1677 		return 0;
1678 }
1679 
1680 /*
1681  * calculate the number of addresses to be returned in the node info reply.
1682  */
1683 static int
1684 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1685     struct in6_addr *subj)
1686 {
1687 	struct ifnet *ifp;
1688 	struct in6_ifaddr *ifa6;
1689 	struct ifaddr *ifa;
1690 	int addrs = 0, addrsofif, iffound = 0;
1691 	int niflags = ni6->ni_flags;
1692 
1693 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1694 		switch (ni6->ni_code) {
1695 		case ICMP6_NI_SUBJ_IPV6:
1696 			if (subj == NULL) /* must be impossible... */
1697 				return (0);
1698 			break;
1699 		default:
1700 			/*
1701 			 * XXX: we only support IPv6 subject address for
1702 			 * this Qtype.
1703 			 */
1704 			return (0);
1705 		}
1706 	}
1707 
1708 	IFNET_RLOCK();
1709 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1710 		addrsofif = 0;
1711 		IF_ADDR_LOCK(ifp);
1712 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1713 			if (ifa->ifa_addr->sa_family != AF_INET6)
1714 				continue;
1715 			ifa6 = (struct in6_ifaddr *)ifa;
1716 
1717 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1718 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1719 				iffound = 1;
1720 
1721 			/*
1722 			 * IPv4-mapped addresses can only be returned by a
1723 			 * Node Information proxy, since they represent
1724 			 * addresses of IPv4-only nodes, which perforce do
1725 			 * not implement this protocol.
1726 			 * [icmp-name-lookups-07, Section 5.4]
1727 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1728 			 * this function at this moment.
1729 			 */
1730 
1731 			/* What do we have to do about ::1? */
1732 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1733 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1734 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1735 					continue;
1736 				break;
1737 			case IPV6_ADDR_SCOPE_SITELOCAL:
1738 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1739 					continue;
1740 				break;
1741 			case IPV6_ADDR_SCOPE_GLOBAL:
1742 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1743 					continue;
1744 				break;
1745 			default:
1746 				continue;
1747 			}
1748 
1749 			/*
1750 			 * check if anycast is okay.
1751 			 * XXX: just experimental.  not in the spec.
1752 			 */
1753 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1754 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1755 				continue; /* we need only unicast addresses */
1756 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1757 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1758 				continue;
1759 			}
1760 			addrsofif++; /* count the address */
1761 		}
1762 		IF_ADDR_UNLOCK(ifp);
1763 		if (iffound) {
1764 			*ifpp = ifp;
1765 			IFNET_RUNLOCK();
1766 			return (addrsofif);
1767 		}
1768 
1769 		addrs += addrsofif;
1770 	}
1771 	IFNET_RUNLOCK();
1772 
1773 	return (addrs);
1774 }
1775 
1776 static int
1777 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1778     struct ifnet *ifp0, int resid)
1779 {
1780 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1781 	struct in6_ifaddr *ifa6;
1782 	struct ifaddr *ifa;
1783 	struct ifnet *ifp_dep = NULL;
1784 	int copied = 0, allow_deprecated = 0;
1785 	u_char *cp = (u_char *)(nni6 + 1);
1786 	int niflags = ni6->ni_flags;
1787 	u_int32_t ltime;
1788 
1789 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1790 		return (0);	/* needless to copy */
1791 
1792 	IFNET_RLOCK();
1793   again:
1794 
1795 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1796 		IF_ADDR_LOCK(ifp);
1797 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1798 			if (ifa->ifa_addr->sa_family != AF_INET6)
1799 				continue;
1800 			ifa6 = (struct in6_ifaddr *)ifa;
1801 
1802 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1803 			    allow_deprecated == 0) {
1804 				/*
1805 				 * prefererred address should be put before
1806 				 * deprecated addresses.
1807 				 */
1808 
1809 				/* record the interface for later search */
1810 				if (ifp_dep == NULL)
1811 					ifp_dep = ifp;
1812 
1813 				continue;
1814 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1815 			    allow_deprecated != 0)
1816 				continue; /* we now collect deprecated addrs */
1817 
1818 			/* What do we have to do about ::1? */
1819 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1820 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1821 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1822 					continue;
1823 				break;
1824 			case IPV6_ADDR_SCOPE_SITELOCAL:
1825 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1826 					continue;
1827 				break;
1828 			case IPV6_ADDR_SCOPE_GLOBAL:
1829 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1830 					continue;
1831 				break;
1832 			default:
1833 				continue;
1834 			}
1835 
1836 			/*
1837 			 * check if anycast is okay.
1838 			 * XXX: just experimental.  not in the spec.
1839 			 */
1840 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1841 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1842 				continue;
1843 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1844 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1845 				continue;
1846 			}
1847 
1848 			/* now we can copy the address */
1849 			if (resid < sizeof(struct in6_addr) +
1850 			    sizeof(u_int32_t)) {
1851 				IF_ADDR_UNLOCK(ifp);
1852 				/*
1853 				 * We give up much more copy.
1854 				 * Set the truncate flag and return.
1855 				 */
1856 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1857 				IFNET_RUNLOCK();
1858 				return (copied);
1859 			}
1860 
1861 			/*
1862 			 * Set the TTL of the address.
1863 			 * The TTL value should be one of the following
1864 			 * according to the specification:
1865 			 *
1866 			 * 1. The remaining lifetime of a DHCP lease on the
1867 			 *    address, or
1868 			 * 2. The remaining Valid Lifetime of a prefix from
1869 			 *    which the address was derived through Stateless
1870 			 *    Autoconfiguration.
1871 			 *
1872 			 * Note that we currently do not support stateful
1873 			 * address configuration by DHCPv6, so the former
1874 			 * case can't happen.
1875 			 */
1876 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1877 				ltime = ND6_INFINITE_LIFETIME;
1878 			else {
1879 				if (ifa6->ia6_lifetime.ia6t_expire >
1880 				    time_second)
1881 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1882 				else
1883 					ltime = 0;
1884 			}
1885 
1886 			bcopy(&ltime, cp, sizeof(u_int32_t));
1887 			cp += sizeof(u_int32_t);
1888 
1889 			/* copy the address itself */
1890 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1891 			    sizeof(struct in6_addr));
1892 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1893 			cp += sizeof(struct in6_addr);
1894 
1895 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1896 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1897 		}
1898 		IF_ADDR_UNLOCK(ifp);
1899 		if (ifp0)	/* we need search only on the specified IF */
1900 			break;
1901 	}
1902 
1903 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1904 		ifp = ifp_dep;
1905 		allow_deprecated = 1;
1906 
1907 		goto again;
1908 	}
1909 
1910 	IFNET_RUNLOCK();
1911 
1912 	return (copied);
1913 }
1914 
1915 /*
1916  * XXX almost dup'ed code with rip6_input.
1917  */
1918 static int
1919 icmp6_rip6_input(struct mbuf **mp, int off)
1920 {
1921 	struct mbuf *m = *mp;
1922 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1923 	struct inpcb *in6p;
1924 	struct inpcb *last = NULL;
1925 	struct sockaddr_in6 fromsa;
1926 	struct icmp6_hdr *icmp6;
1927 	struct mbuf *opts = NULL;
1928 
1929 #ifndef PULLDOWN_TEST
1930 	/* this is assumed to be safe. */
1931 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1932 #else
1933 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1934 	if (icmp6 == NULL) {
1935 		/* m is already reclaimed */
1936 		return (IPPROTO_DONE);
1937 	}
1938 #endif
1939 
1940 	/*
1941 	 * XXX: the address may have embedded scope zone ID, which should be
1942 	 * hidden from applications.
1943 	 */
1944 	bzero(&fromsa, sizeof(fromsa));
1945 	fromsa.sin6_family = AF_INET6;
1946 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1947 	fromsa.sin6_addr = ip6->ip6_src;
1948 	if (sa6_recoverscope(&fromsa)) {
1949 		m_freem(m);
1950 		return (IPPROTO_DONE);
1951 	}
1952 
1953 	INP_INFO_RLOCK(&V_ripcbinfo);
1954 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1955 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1956 			continue;
1957 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1958 			continue;
1959 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1960 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1961 			continue;
1962 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1963 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1964 			continue;
1965 		INP_RLOCK(in6p);
1966 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1967 		    in6p->in6p_icmp6filt)) {
1968 			INP_RUNLOCK(in6p);
1969 			continue;
1970 		}
1971 		if (last != NULL) {
1972 			struct	mbuf *n = NULL;
1973 
1974 			/*
1975 			 * Recent network drivers tend to allocate a single
1976 			 * mbuf cluster, rather than to make a couple of
1977 			 * mbufs without clusters.  Also, since the IPv6 code
1978 			 * path tries to avoid m_pullup(), it is highly
1979 			 * probable that we still have an mbuf cluster here
1980 			 * even though the necessary length can be stored in an
1981 			 * mbuf's internal buffer.
1982 			 * Meanwhile, the default size of the receive socket
1983 			 * buffer for raw sockets is not so large.  This means
1984 			 * the possibility of packet loss is relatively higher
1985 			 * than before.  To avoid this scenario, we copy the
1986 			 * received data to a separate mbuf that does not use
1987 			 * a cluster, if possible.
1988 			 * XXX: it is better to copy the data after stripping
1989 			 * intermediate headers.
1990 			 */
1991 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1992 			    m->m_len <= MHLEN) {
1993 				MGET(n, M_DONTWAIT, m->m_type);
1994 				if (n != NULL) {
1995 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1996 						bcopy(m->m_data, n->m_data,
1997 						      m->m_len);
1998 						n->m_len = m->m_len;
1999 					} else {
2000 						m_free(n);
2001 						n = NULL;
2002 					}
2003 				}
2004 			}
2005 			if (n != NULL ||
2006 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2007 				if (last->inp_flags & INP_CONTROLOPTS)
2008 					ip6_savecontrol(last, n, &opts);
2009 				/* strip intermediate headers */
2010 				m_adj(n, off);
2011 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2012 				if (sbappendaddr_locked(
2013 				    &last->inp_socket->so_rcv,
2014 				    (struct sockaddr *)&fromsa, n, opts)
2015 				    == 0) {
2016 					/* should notify about lost packet */
2017 					m_freem(n);
2018 					if (opts) {
2019 						m_freem(opts);
2020 					}
2021 					SOCKBUF_UNLOCK(
2022 					    &last->inp_socket->so_rcv);
2023 				} else
2024 					sorwakeup_locked(last->inp_socket);
2025 				opts = NULL;
2026 			}
2027 			INP_RUNLOCK(last);
2028 		}
2029 		last = in6p;
2030 	}
2031 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2032 	if (last != NULL) {
2033 		if (last->inp_flags & INP_CONTROLOPTS)
2034 			ip6_savecontrol(last, m, &opts);
2035 		/* strip intermediate headers */
2036 		m_adj(m, off);
2037 
2038 		/* avoid using mbuf clusters if possible (see above) */
2039 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2040 		    m->m_len <= MHLEN) {
2041 			struct mbuf *n;
2042 
2043 			MGET(n, M_DONTWAIT, m->m_type);
2044 			if (n != NULL) {
2045 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2046 					bcopy(m->m_data, n->m_data, m->m_len);
2047 					n->m_len = m->m_len;
2048 
2049 					m_freem(m);
2050 					m = n;
2051 				} else {
2052 					m_freem(n);
2053 					n = NULL;
2054 				}
2055 			}
2056 		}
2057 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2058 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2059 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2060 			m_freem(m);
2061 			if (opts)
2062 				m_freem(opts);
2063 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2064 		} else
2065 			sorwakeup_locked(last->inp_socket);
2066 		INP_RUNLOCK(last);
2067 	} else {
2068 		m_freem(m);
2069 		IP6STAT_DEC(ip6s_delivered);
2070 	}
2071 	return IPPROTO_DONE;
2072 }
2073 
2074 /*
2075  * Reflect the ip6 packet back to the source.
2076  * OFF points to the icmp6 header, counted from the top of the mbuf.
2077  */
2078 void
2079 icmp6_reflect(struct mbuf *m, size_t off)
2080 {
2081 	struct ip6_hdr *ip6;
2082 	struct icmp6_hdr *icmp6;
2083 	struct in6_ifaddr *ia = NULL;
2084 	int plen;
2085 	int type, code;
2086 	struct ifnet *outif = NULL;
2087 	struct in6_addr origdst, src, *srcp = NULL;
2088 
2089 	/* too short to reflect */
2090 	if (off < sizeof(struct ip6_hdr)) {
2091 		nd6log((LOG_DEBUG,
2092 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2093 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2094 		    __FILE__, __LINE__));
2095 		goto bad;
2096 	}
2097 
2098 	/*
2099 	 * If there are extra headers between IPv6 and ICMPv6, strip
2100 	 * off that header first.
2101 	 */
2102 #ifdef DIAGNOSTIC
2103 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2104 		panic("assumption failed in icmp6_reflect");
2105 #endif
2106 	if (off > sizeof(struct ip6_hdr)) {
2107 		size_t l;
2108 		struct ip6_hdr nip6;
2109 
2110 		l = off - sizeof(struct ip6_hdr);
2111 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2112 		m_adj(m, l);
2113 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2114 		if (m->m_len < l) {
2115 			if ((m = m_pullup(m, l)) == NULL)
2116 				return;
2117 		}
2118 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2119 	} else /* off == sizeof(struct ip6_hdr) */ {
2120 		size_t l;
2121 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2122 		if (m->m_len < l) {
2123 			if ((m = m_pullup(m, l)) == NULL)
2124 				return;
2125 		}
2126 	}
2127 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2128 	ip6 = mtod(m, struct ip6_hdr *);
2129 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2130 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2131 	type = icmp6->icmp6_type; /* keep type for statistics */
2132 	code = icmp6->icmp6_code; /* ditto. */
2133 
2134 	origdst = ip6->ip6_dst;
2135 	/*
2136 	 * ip6_input() drops a packet if its src is multicast.
2137 	 * So, the src is never multicast.
2138 	 */
2139 	ip6->ip6_dst = ip6->ip6_src;
2140 
2141 	/*
2142 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2143 	 * use dst as the src for the reply.
2144 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2145 	 * (for example) when we encounter an error while forwarding procedure
2146 	 * destined to a duplicated address of ours.
2147 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2148 	 * procedure of an outgoing packet of our own, in which case we need
2149 	 * to search in the ifaddr list.
2150 	 */
2151 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2152 		if ((ia = ip6_getdstifaddr(m))) {
2153 			if (!(ia->ia6_flags &
2154 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2155 				srcp = &ia->ia_addr.sin6_addr;
2156 		} else {
2157 			struct sockaddr_in6 d;
2158 
2159 			bzero(&d, sizeof(d));
2160 			d.sin6_family = AF_INET6;
2161 			d.sin6_len = sizeof(d);
2162 			d.sin6_addr = origdst;
2163 			ia = (struct in6_ifaddr *)
2164 			    ifa_ifwithaddr((struct sockaddr *)&d);
2165 			if (ia &&
2166 			    !(ia->ia6_flags &
2167 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2168 				srcp = &ia->ia_addr.sin6_addr;
2169 			}
2170 		}
2171 	}
2172 
2173 	if (srcp == NULL) {
2174 		int e;
2175 		struct sockaddr_in6 sin6;
2176 		struct route_in6 ro;
2177 
2178 		/*
2179 		 * This case matches to multicasts, our anycast, or unicasts
2180 		 * that we do not own.  Select a source address based on the
2181 		 * source address of the erroneous packet.
2182 		 */
2183 		bzero(&sin6, sizeof(sin6));
2184 		sin6.sin6_family = AF_INET6;
2185 		sin6.sin6_len = sizeof(sin6);
2186 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2187 
2188 		bzero(&ro, sizeof(ro));
2189 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2190 		if (ro.ro_rt)
2191 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2192 		if (e) {
2193 			char ip6buf[INET6_ADDRSTRLEN];
2194 			nd6log((LOG_DEBUG,
2195 			    "icmp6_reflect: source can't be determined: "
2196 			    "dst=%s, error=%d\n",
2197 			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2198 			goto bad;
2199 		}
2200 		srcp = &src;
2201 	}
2202 
2203 	ip6->ip6_src = *srcp;
2204 	ip6->ip6_flow = 0;
2205 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2206 	ip6->ip6_vfc |= IPV6_VERSION;
2207 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2208 	if (outif)
2209 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2210 	else if (m->m_pkthdr.rcvif) {
2211 		/* XXX: This may not be the outgoing interface */
2212 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2213 	} else
2214 		ip6->ip6_hlim = V_ip6_defhlim;
2215 
2216 	icmp6->icmp6_cksum = 0;
2217 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2218 	    sizeof(struct ip6_hdr), plen);
2219 
2220 	/*
2221 	 * XXX option handling
2222 	 */
2223 
2224 	m->m_flags &= ~(M_BCAST|M_MCAST);
2225 
2226 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2227 	if (outif)
2228 		icmp6_ifoutstat_inc(outif, type, code);
2229 
2230 	if (ia != NULL)
2231 		ifa_free(&ia->ia_ifa);
2232 	return;
2233 
2234  bad:
2235 	if (ia != NULL)
2236 		ifa_free(&ia->ia_ifa);
2237 	m_freem(m);
2238 	return;
2239 }
2240 
2241 void
2242 icmp6_fasttimo(void)
2243 {
2244 
2245 	mld_fasttimo();
2246 }
2247 
2248 void
2249 icmp6_slowtimo(void)
2250 {
2251 
2252 	mld_slowtimo();
2253 }
2254 
2255 static const char *
2256 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2257     struct in6_addr *tgt6)
2258 {
2259 	static char buf[1024];
2260 	char ip6bufs[INET6_ADDRSTRLEN];
2261 	char ip6bufd[INET6_ADDRSTRLEN];
2262 	char ip6buft[INET6_ADDRSTRLEN];
2263 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2264 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2265 	    ip6_sprintf(ip6buft, tgt6));
2266 	return buf;
2267 }
2268 
2269 void
2270 icmp6_redirect_input(struct mbuf *m, int off)
2271 {
2272 	struct ifnet *ifp;
2273 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2274 	struct nd_redirect *nd_rd;
2275 	int icmp6len = ntohs(ip6->ip6_plen);
2276 	char *lladdr = NULL;
2277 	int lladdrlen = 0;
2278 	u_char *redirhdr = NULL;
2279 	int redirhdrlen = 0;
2280 	struct rtentry *rt = NULL;
2281 	int is_router;
2282 	int is_onlink;
2283 	struct in6_addr src6 = ip6->ip6_src;
2284 	struct in6_addr redtgt6;
2285 	struct in6_addr reddst6;
2286 	union nd_opts ndopts;
2287 	char ip6buf[INET6_ADDRSTRLEN];
2288 
2289 	if (!m)
2290 		return;
2291 
2292 	ifp = m->m_pkthdr.rcvif;
2293 
2294 	if (!ifp)
2295 		return;
2296 
2297 	/* XXX if we are router, we don't update route by icmp6 redirect */
2298 	if (V_ip6_forwarding)
2299 		goto freeit;
2300 	if (!V_icmp6_rediraccept)
2301 		goto freeit;
2302 
2303 #ifndef PULLDOWN_TEST
2304 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2305 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2306 #else
2307 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2308 	if (nd_rd == NULL) {
2309 		ICMP6STAT_INC(icp6s_tooshort);
2310 		return;
2311 	}
2312 #endif
2313 	redtgt6 = nd_rd->nd_rd_target;
2314 	reddst6 = nd_rd->nd_rd_dst;
2315 
2316 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2317 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2318 		goto freeit;
2319 	}
2320 
2321 	/* validation */
2322 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2323 		nd6log((LOG_ERR,
2324 		    "ICMP6 redirect sent from %s rejected; "
2325 		    "must be from linklocal\n",
2326 		    ip6_sprintf(ip6buf, &src6)));
2327 		goto bad;
2328 	}
2329 	if (ip6->ip6_hlim != 255) {
2330 		nd6log((LOG_ERR,
2331 		    "ICMP6 redirect sent from %s rejected; "
2332 		    "hlim=%d (must be 255)\n",
2333 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2334 		goto bad;
2335 	}
2336     {
2337 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2338 	struct sockaddr_in6 sin6;
2339 	struct in6_addr *gw6;
2340 
2341 	bzero(&sin6, sizeof(sin6));
2342 	sin6.sin6_family = AF_INET6;
2343 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2344 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2345 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2346 	if (rt) {
2347 		if (rt->rt_gateway == NULL ||
2348 		    rt->rt_gateway->sa_family != AF_INET6) {
2349 			nd6log((LOG_ERR,
2350 			    "ICMP6 redirect rejected; no route "
2351 			    "with inet6 gateway found for redirect dst: %s\n",
2352 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2353 			RTFREE_LOCKED(rt);
2354 			goto bad;
2355 		}
2356 
2357 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2358 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2359 			nd6log((LOG_ERR,
2360 			    "ICMP6 redirect rejected; "
2361 			    "not equal to gw-for-src=%s (must be same): "
2362 			    "%s\n",
2363 			    ip6_sprintf(ip6buf, gw6),
2364 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2365 			RTFREE_LOCKED(rt);
2366 			goto bad;
2367 		}
2368 	} else {
2369 		nd6log((LOG_ERR,
2370 		    "ICMP6 redirect rejected; "
2371 		    "no route found for redirect dst: %s\n",
2372 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2373 		goto bad;
2374 	}
2375 	RTFREE_LOCKED(rt);
2376 	rt = NULL;
2377     }
2378 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2379 		nd6log((LOG_ERR,
2380 		    "ICMP6 redirect rejected; "
2381 		    "redirect dst must be unicast: %s\n",
2382 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2383 		goto bad;
2384 	}
2385 
2386 	is_router = is_onlink = 0;
2387 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2388 		is_router = 1;	/* router case */
2389 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2390 		is_onlink = 1;	/* on-link destination case */
2391 	if (!is_router && !is_onlink) {
2392 		nd6log((LOG_ERR,
2393 		    "ICMP6 redirect rejected; "
2394 		    "neither router case nor onlink case: %s\n",
2395 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2396 		goto bad;
2397 	}
2398 	/* validation passed */
2399 
2400 	icmp6len -= sizeof(*nd_rd);
2401 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2402 	if (nd6_options(&ndopts) < 0) {
2403 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2404 		    "invalid ND option, rejected: %s\n",
2405 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2406 		/* nd6_options have incremented stats */
2407 		goto freeit;
2408 	}
2409 
2410 	if (ndopts.nd_opts_tgt_lladdr) {
2411 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2412 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2413 	}
2414 
2415 	if (ndopts.nd_opts_rh) {
2416 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2417 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2418 	}
2419 
2420 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2421 		nd6log((LOG_INFO,
2422 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2423 		    "(if %d, icmp6 packet %d): %s\n",
2424 		    ip6_sprintf(ip6buf, &redtgt6),
2425 		    ifp->if_addrlen, lladdrlen - 2,
2426 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2427 		goto bad;
2428 	}
2429 
2430 	/* RFC 2461 8.3 */
2431 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2432 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2433 
2434 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2435 		/* perform rtredirect */
2436 		struct sockaddr_in6 sdst;
2437 		struct sockaddr_in6 sgw;
2438 		struct sockaddr_in6 ssrc;
2439 
2440 		bzero(&sdst, sizeof(sdst));
2441 		bzero(&sgw, sizeof(sgw));
2442 		bzero(&ssrc, sizeof(ssrc));
2443 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2444 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2445 			sizeof(struct sockaddr_in6);
2446 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2447 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2448 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2449 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2450 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2451 		    (struct sockaddr *)&ssrc);
2452 	}
2453 	/* finally update cached route in each socket via pfctlinput */
2454     {
2455 	struct sockaddr_in6 sdst;
2456 
2457 	bzero(&sdst, sizeof(sdst));
2458 	sdst.sin6_family = AF_INET6;
2459 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2460 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2461 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2462 #ifdef IPSEC
2463 	key_sa_routechange((struct sockaddr *)&sdst);
2464 #endif /* IPSEC */
2465     }
2466 
2467  freeit:
2468 	m_freem(m);
2469 	return;
2470 
2471  bad:
2472 	ICMP6STAT_INC(icp6s_badredirect);
2473 	m_freem(m);
2474 }
2475 
2476 void
2477 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2478 {
2479 	struct ifnet *ifp;	/* my outgoing interface */
2480 	struct in6_addr *ifp_ll6;
2481 	struct in6_addr *router_ll6;
2482 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2483 	struct mbuf *m = NULL;	/* newly allocated one */
2484 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2485 	struct nd_redirect *nd_rd;
2486 	struct llentry *ln = NULL;
2487 	size_t maxlen;
2488 	u_char *p;
2489 	struct ifnet *outif = NULL;
2490 	struct sockaddr_in6 src_sa;
2491 
2492 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2493 
2494 	/* if we are not router, we don't send icmp6 redirect */
2495 	if (!V_ip6_forwarding)
2496 		goto fail;
2497 
2498 	/* sanity check */
2499 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2500 		goto fail;
2501 
2502 	/*
2503 	 * Address check:
2504 	 *  the source address must identify a neighbor, and
2505 	 *  the destination address must not be a multicast address
2506 	 *  [RFC 2461, sec 8.2]
2507 	 */
2508 	sip6 = mtod(m0, struct ip6_hdr *);
2509 	bzero(&src_sa, sizeof(src_sa));
2510 	src_sa.sin6_family = AF_INET6;
2511 	src_sa.sin6_len = sizeof(src_sa);
2512 	src_sa.sin6_addr = sip6->ip6_src;
2513 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2514 		goto fail;
2515 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2516 		goto fail;	/* what should we do here? */
2517 
2518 	/* rate limit */
2519 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2520 		goto fail;
2521 
2522 	/*
2523 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2524 	 * we almost always ask for an mbuf cluster for simplicity.
2525 	 * (MHLEN < IPV6_MMTU is almost always true)
2526 	 */
2527 #if IPV6_MMTU >= MCLBYTES
2528 # error assumption failed about IPV6_MMTU and MCLBYTES
2529 #endif
2530 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2531 	if (m && IPV6_MMTU >= MHLEN)
2532 		MCLGET(m, M_DONTWAIT);
2533 	if (!m)
2534 		goto fail;
2535 	m->m_pkthdr.rcvif = NULL;
2536 	m->m_len = 0;
2537 	maxlen = M_TRAILINGSPACE(m);
2538 	maxlen = min(IPV6_MMTU, maxlen);
2539 	/* just for safety */
2540 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2541 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2542 		goto fail;
2543 	}
2544 
2545 	{
2546 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2547 		struct in6_ifaddr *ia;
2548 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2549 						 IN6_IFF_NOTREADY|
2550 						 IN6_IFF_ANYCAST)) == NULL)
2551 			goto fail;
2552 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2553 		/* XXXRW: reference released prematurely. */
2554 		ifa_free(&ia->ia_ifa);
2555 	}
2556 
2557 	/* get ip6 linklocal address for the router. */
2558 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2559 		struct sockaddr_in6 *sin6;
2560 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2561 		router_ll6 = &sin6->sin6_addr;
2562 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2563 			router_ll6 = (struct in6_addr *)NULL;
2564 	} else
2565 		router_ll6 = (struct in6_addr *)NULL;
2566 
2567 	/* ip6 */
2568 	ip6 = mtod(m, struct ip6_hdr *);
2569 	ip6->ip6_flow = 0;
2570 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2571 	ip6->ip6_vfc |= IPV6_VERSION;
2572 	/* ip6->ip6_plen will be set later */
2573 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2574 	ip6->ip6_hlim = 255;
2575 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2576 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2577 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2578 
2579 	/* ND Redirect */
2580 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2581 	nd_rd->nd_rd_type = ND_REDIRECT;
2582 	nd_rd->nd_rd_code = 0;
2583 	nd_rd->nd_rd_reserved = 0;
2584 	if (rt->rt_flags & RTF_GATEWAY) {
2585 		/*
2586 		 * nd_rd->nd_rd_target must be a link-local address in
2587 		 * better router cases.
2588 		 */
2589 		if (!router_ll6)
2590 			goto fail;
2591 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2592 		    sizeof(nd_rd->nd_rd_target));
2593 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2594 		    sizeof(nd_rd->nd_rd_dst));
2595 	} else {
2596 		/* make sure redtgt == reddst */
2597 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2598 		    sizeof(nd_rd->nd_rd_target));
2599 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2600 		    sizeof(nd_rd->nd_rd_dst));
2601 	}
2602 
2603 	p = (u_char *)(nd_rd + 1);
2604 
2605 	if (!router_ll6)
2606 		goto nolladdropt;
2607 
2608 	{
2609 		/* target lladdr option */
2610 		int len;
2611 		struct nd_opt_hdr *nd_opt;
2612 		char *lladdr;
2613 
2614 		IF_AFDATA_LOCK(ifp);
2615 		ln = nd6_lookup(router_ll6, 0, ifp);
2616 		IF_AFDATA_UNLOCK(ifp);
2617 		if (ln == NULL)
2618 			goto nolladdropt;
2619 
2620 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2621 		len = (len + 7) & ~7;	/* round by 8 */
2622 		/* safety check */
2623 		if (len + (p - (u_char *)ip6) > maxlen)
2624 			goto nolladdropt;
2625 
2626 		if (ln->la_flags & LLE_VALID) {
2627 			nd_opt = (struct nd_opt_hdr *)p;
2628 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2629 			nd_opt->nd_opt_len = len >> 3;
2630 			lladdr = (char *)(nd_opt + 1);
2631 			bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2632 			p += len;
2633 		}
2634 	}
2635 nolladdropt:
2636 	if (ln != NULL)
2637 		LLE_RUNLOCK(ln);
2638 
2639 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2640 
2641 	/* just to be safe */
2642 #ifdef M_DECRYPTED	/*not openbsd*/
2643 	if (m0->m_flags & M_DECRYPTED)
2644 		goto noredhdropt;
2645 #endif
2646 	if (p - (u_char *)ip6 > maxlen)
2647 		goto noredhdropt;
2648 
2649 	{
2650 		/* redirected header option */
2651 		int len;
2652 		struct nd_opt_rd_hdr *nd_opt_rh;
2653 
2654 		/*
2655 		 * compute the maximum size for icmp6 redirect header option.
2656 		 * XXX room for auth header?
2657 		 */
2658 		len = maxlen - (p - (u_char *)ip6);
2659 		len &= ~7;
2660 
2661 		/* This is just for simplicity. */
2662 		if (m0->m_pkthdr.len != m0->m_len) {
2663 			if (m0->m_next) {
2664 				m_freem(m0->m_next);
2665 				m0->m_next = NULL;
2666 			}
2667 			m0->m_pkthdr.len = m0->m_len;
2668 		}
2669 
2670 		/*
2671 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2672 		 * about padding/truncate rule for the original IP packet.
2673 		 * From the discussion on IPv6imp in Feb 1999,
2674 		 * the consensus was:
2675 		 * - "attach as much as possible" is the goal
2676 		 * - pad if not aligned (original size can be guessed by
2677 		 *   original ip6 header)
2678 		 * Following code adds the padding if it is simple enough,
2679 		 * and truncates if not.
2680 		 */
2681 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2682 			panic("assumption failed in %s:%d", __FILE__,
2683 			    __LINE__);
2684 
2685 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2686 			/* not enough room, truncate */
2687 			m0->m_pkthdr.len = m0->m_len = len -
2688 			    sizeof(*nd_opt_rh);
2689 		} else {
2690 			/* enough room, pad or truncate */
2691 			size_t extra;
2692 
2693 			extra = m0->m_pkthdr.len % 8;
2694 			if (extra) {
2695 				/* pad if easy enough, truncate if not */
2696 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2697 					/* pad */
2698 					m0->m_len += (8 - extra);
2699 					m0->m_pkthdr.len += (8 - extra);
2700 				} else {
2701 					/* truncate */
2702 					m0->m_pkthdr.len -= extra;
2703 					m0->m_len -= extra;
2704 				}
2705 			}
2706 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2707 			m0->m_pkthdr.len = m0->m_len = len -
2708 			    sizeof(*nd_opt_rh);
2709 		}
2710 
2711 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2712 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2713 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2714 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2715 		p += sizeof(*nd_opt_rh);
2716 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2717 
2718 		/* connect m0 to m */
2719 		m_tag_delete_chain(m0, NULL);
2720 		m0->m_flags &= ~M_PKTHDR;
2721 		m->m_next = m0;
2722 		m->m_pkthdr.len = m->m_len + m0->m_len;
2723 		m0 = NULL;
2724 	}
2725 noredhdropt:;
2726 	if (m0) {
2727 		m_freem(m0);
2728 		m0 = NULL;
2729 	}
2730 
2731 	/* XXX: clear embedded link IDs in the inner header */
2732 	in6_clearscope(&sip6->ip6_src);
2733 	in6_clearscope(&sip6->ip6_dst);
2734 	in6_clearscope(&nd_rd->nd_rd_target);
2735 	in6_clearscope(&nd_rd->nd_rd_dst);
2736 
2737 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2738 
2739 	nd_rd->nd_rd_cksum = 0;
2740 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2741 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2742 
2743 	/* send the packet to outside... */
2744 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2745 	if (outif) {
2746 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2747 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2748 	}
2749 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2750 
2751 	return;
2752 
2753 fail:
2754 	if (m)
2755 		m_freem(m);
2756 	if (m0)
2757 		m_freem(m0);
2758 }
2759 
2760 /*
2761  * ICMPv6 socket option processing.
2762  */
2763 int
2764 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2765 {
2766 	int error = 0;
2767 	int optlen;
2768 	struct inpcb *inp = sotoinpcb(so);
2769 	int level, op, optname;
2770 
2771 	if (sopt) {
2772 		level = sopt->sopt_level;
2773 		op = sopt->sopt_dir;
2774 		optname = sopt->sopt_name;
2775 		optlen = sopt->sopt_valsize;
2776 	} else
2777 		level = op = optname = optlen = 0;
2778 
2779 	if (level != IPPROTO_ICMPV6) {
2780 		return EINVAL;
2781 	}
2782 
2783 	switch (op) {
2784 	case PRCO_SETOPT:
2785 		switch (optname) {
2786 		case ICMP6_FILTER:
2787 		    {
2788 			struct icmp6_filter ic6f;
2789 
2790 			if (optlen != sizeof(ic6f)) {
2791 				error = EMSGSIZE;
2792 				break;
2793 			}
2794 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2795 			if (error == 0) {
2796 				INP_WLOCK(inp);
2797 				*inp->in6p_icmp6filt = ic6f;
2798 				INP_WUNLOCK(inp);
2799 			}
2800 			break;
2801 		    }
2802 
2803 		default:
2804 			error = ENOPROTOOPT;
2805 			break;
2806 		}
2807 		break;
2808 
2809 	case PRCO_GETOPT:
2810 		switch (optname) {
2811 		case ICMP6_FILTER:
2812 		    {
2813 			struct icmp6_filter ic6f;
2814 
2815 			INP_RLOCK(inp);
2816 			ic6f = *inp->in6p_icmp6filt;
2817 			INP_RUNLOCK(inp);
2818 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2819 			break;
2820 		    }
2821 
2822 		default:
2823 			error = ENOPROTOOPT;
2824 			break;
2825 		}
2826 		break;
2827 	}
2828 
2829 	return (error);
2830 }
2831 
2832 /*
2833  * Perform rate limit check.
2834  * Returns 0 if it is okay to send the icmp6 packet.
2835  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2836  * limitation.
2837  *
2838  * XXX per-destination/type check necessary?
2839  *
2840  * dst - not used at this moment
2841  * type - not used at this moment
2842  * code - not used at this moment
2843  */
2844 static int
2845 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2846     const int code)
2847 {
2848 	int ret;
2849 
2850 	ret = 0;	/* okay to send */
2851 
2852 	/* PPS limit */
2853 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2854 	    V_icmp6errppslim)) {
2855 		/* The packet is subject to rate limit */
2856 		ret++;
2857 	}
2858 
2859 	return ret;
2860 }
2861