xref: /freebsd/sys/netinet6/icmp6.c (revision aa64588d28258aef88cc33b8043112e8856948d0)
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_DEFINE(struct icmp6stat, icmp6stat);
117 
118 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
119 VNET_DECLARE(struct inpcbhead, ripcb);
120 VNET_DECLARE(int, icmp6errppslim);
121 static VNET_DEFINE(int, icmp6errpps_count) = 0;
122 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
123 VNET_DECLARE(int, icmp6_nodeinfo);
124 
125 #define	V_ripcbinfo			VNET(ripcbinfo)
126 #define	V_ripcb				VNET(ripcb)
127 #define	V_icmp6errppslim		VNET(icmp6errppslim)
128 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
129 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
130 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
131 
132 static void icmp6_errcount(struct icmp6errstat *, int, int);
133 static int icmp6_rip6_input(struct mbuf **, int);
134 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
135 static const char *icmp6_redirect_diag __P((struct in6_addr *,
136 	struct in6_addr *, struct in6_addr *));
137 static struct mbuf *ni6_input(struct mbuf *, int);
138 static struct mbuf *ni6_nametodns(const char *, int, int);
139 static int ni6_dnsmatch(const char *, int, const char *, int);
140 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
141 			  struct ifnet **, struct in6_addr *));
142 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
143 				struct ifnet *, int));
144 static int icmp6_notify_error(struct mbuf **, int, int, int);
145 
146 /*
147  * Kernel module interface for updating icmp6stat.  The argument is an index
148  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
149  * general layout of icmp6stat into the caller, it doesn't encode its
150  * location, so that future changes to add, for example, per-CPU stats
151  * support won't cause binary compatibility problems for kernel modules.
152  */
153 void
154 kmod_icmp6stat_inc(int statnum)
155 {
156 
157 	(*((u_quad_t *)&V_icmp6stat + statnum))++;
158 }
159 
160 static void
161 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
162 {
163 	switch (type) {
164 	case ICMP6_DST_UNREACH:
165 		switch (code) {
166 		case ICMP6_DST_UNREACH_NOROUTE:
167 			stat->icp6errs_dst_unreach_noroute++;
168 			return;
169 		case ICMP6_DST_UNREACH_ADMIN:
170 			stat->icp6errs_dst_unreach_admin++;
171 			return;
172 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
173 			stat->icp6errs_dst_unreach_beyondscope++;
174 			return;
175 		case ICMP6_DST_UNREACH_ADDR:
176 			stat->icp6errs_dst_unreach_addr++;
177 			return;
178 		case ICMP6_DST_UNREACH_NOPORT:
179 			stat->icp6errs_dst_unreach_noport++;
180 			return;
181 		}
182 		break;
183 	case ICMP6_PACKET_TOO_BIG:
184 		stat->icp6errs_packet_too_big++;
185 		return;
186 	case ICMP6_TIME_EXCEEDED:
187 		switch (code) {
188 		case ICMP6_TIME_EXCEED_TRANSIT:
189 			stat->icp6errs_time_exceed_transit++;
190 			return;
191 		case ICMP6_TIME_EXCEED_REASSEMBLY:
192 			stat->icp6errs_time_exceed_reassembly++;
193 			return;
194 		}
195 		break;
196 	case ICMP6_PARAM_PROB:
197 		switch (code) {
198 		case ICMP6_PARAMPROB_HEADER:
199 			stat->icp6errs_paramprob_header++;
200 			return;
201 		case ICMP6_PARAMPROB_NEXTHEADER:
202 			stat->icp6errs_paramprob_nextheader++;
203 			return;
204 		case ICMP6_PARAMPROB_OPTION:
205 			stat->icp6errs_paramprob_option++;
206 			return;
207 		}
208 		break;
209 	case ND_REDIRECT:
210 		stat->icp6errs_redirect++;
211 		return;
212 	}
213 	stat->icp6errs_unknown++;
214 }
215 
216 /*
217  * A wrapper function for icmp6_error() necessary when the erroneous packet
218  * may not contain enough scope zone information.
219  */
220 void
221 icmp6_error2(struct mbuf *m, int type, int code, int param,
222     struct ifnet *ifp)
223 {
224 	struct ip6_hdr *ip6;
225 
226 	if (ifp == NULL)
227 		return;
228 
229 #ifndef PULLDOWN_TEST
230 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
231 #else
232 	if (m->m_len < sizeof(struct ip6_hdr)) {
233 		m = m_pullup(m, sizeof(struct ip6_hdr));
234 		if (m == NULL)
235 			return;
236 	}
237 #endif
238 
239 	ip6 = mtod(m, struct ip6_hdr *);
240 
241 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
242 		return;
243 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
244 		return;
245 
246 	icmp6_error(m, type, code, param);
247 }
248 
249 /*
250  * Generate an error packet of type error in response to bad IP6 packet.
251  */
252 void
253 icmp6_error(struct mbuf *m, int type, int code, int param)
254 {
255 	struct ip6_hdr *oip6, *nip6;
256 	struct icmp6_hdr *icmp6;
257 	u_int preplen;
258 	int off;
259 	int nxt;
260 
261 	ICMP6STAT_INC(icp6s_error);
262 
263 	/* count per-type-code statistics */
264 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code);
265 
266 #ifdef M_DECRYPTED	/*not openbsd*/
267 	if (m->m_flags & M_DECRYPTED) {
268 		ICMP6STAT_INC(icp6s_canterror);
269 		goto freeit;
270 	}
271 #endif
272 
273 #ifndef PULLDOWN_TEST
274 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
275 #else
276 	if (m->m_len < sizeof(struct ip6_hdr)) {
277 		m = m_pullup(m, sizeof(struct ip6_hdr));
278 		if (m == NULL)
279 			return;
280 	}
281 #endif
282 	oip6 = mtod(m, struct ip6_hdr *);
283 
284 	/*
285 	 * If the destination address of the erroneous packet is a multicast
286 	 * address, or the packet was sent using link-layer multicast,
287 	 * we should basically suppress sending an error (RFC 2463, Section
288 	 * 2.4).
289 	 * We have two exceptions (the item e.2 in that section):
290 	 * - the Packet Too Big message can be sent for path MTU discovery.
291 	 * - the Parameter Problem Message that can be allowed an icmp6 error
292 	 *   in the option type field.  This check has been done in
293 	 *   ip6_unknown_opt(), so we can just check the type and code.
294 	 */
295 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
296 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
297 	    (type != ICMP6_PACKET_TOO_BIG &&
298 	     (type != ICMP6_PARAM_PROB ||
299 	      code != ICMP6_PARAMPROB_OPTION)))
300 		goto freeit;
301 
302 	/*
303 	 * RFC 2463, 2.4 (e.5): source address check.
304 	 * XXX: the case of anycast source?
305 	 */
306 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
307 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
308 		goto freeit;
309 
310 	/*
311 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
312 	 * don't do it.
313 	 */
314 	nxt = -1;
315 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
316 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
317 		struct icmp6_hdr *icp;
318 
319 #ifndef PULLDOWN_TEST
320 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
321 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
322 #else
323 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
324 			sizeof(*icp));
325 		if (icp == NULL) {
326 			ICMP6STAT_INC(icp6s_tooshort);
327 			return;
328 		}
329 #endif
330 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
331 		    icp->icmp6_type == ND_REDIRECT) {
332 			/*
333 			 * ICMPv6 error
334 			 * Special case: for redirect (which is
335 			 * informational) we must not send icmp6 error.
336 			 */
337 			ICMP6STAT_INC(icp6s_canterror);
338 			goto freeit;
339 		} else {
340 			/* ICMPv6 informational - send the error */
341 		}
342 	} else {
343 		/* non-ICMPv6 - send the error */
344 	}
345 
346 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
347 
348 	/* Finally, do rate limitation check. */
349 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
350 		ICMP6STAT_INC(icp6s_toofreq);
351 		goto freeit;
352 	}
353 
354 	/*
355 	 * OK, ICMP6 can be generated.
356 	 */
357 
358 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
359 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
360 
361 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
362 	M_PREPEND(m, preplen, M_DONTWAIT);
363 	if (m && m->m_len < preplen)
364 		m = m_pullup(m, preplen);
365 	if (m == NULL) {
366 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
367 		return;
368 	}
369 
370 	nip6 = mtod(m, struct ip6_hdr *);
371 	nip6->ip6_src  = oip6->ip6_src;
372 	nip6->ip6_dst  = oip6->ip6_dst;
373 
374 	in6_clearscope(&oip6->ip6_src);
375 	in6_clearscope(&oip6->ip6_dst);
376 
377 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
378 	icmp6->icmp6_type = type;
379 	icmp6->icmp6_code = code;
380 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
381 
382 	/*
383 	 * icmp6_reflect() is designed to be in the input path.
384 	 * icmp6_error() can be called from both input and output path,
385 	 * and if we are in output path rcvif could contain bogus value.
386 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
387 	 * information in ip header (nip6).
388 	 */
389 	m->m_pkthdr.rcvif = NULL;
390 
391 	ICMP6STAT_INC(icp6s_outhist[type]);
392 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
393 
394 	return;
395 
396   freeit:
397 	/*
398 	 * If we can't tell whether or not we can generate ICMP6, free it.
399 	 */
400 	m_freem(m);
401 }
402 
403 /*
404  * Process a received ICMP6 message.
405  */
406 int
407 icmp6_input(struct mbuf **mp, int *offp, int proto)
408 {
409 	struct mbuf *m = *mp, *n;
410 	struct ifnet *ifp;
411 	struct ip6_hdr *ip6, *nip6;
412 	struct icmp6_hdr *icmp6, *nicmp6;
413 	int off = *offp;
414 	int icmp6len = m->m_pkthdr.len - *offp;
415 	int code, sum, noff;
416 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
417 
418 	ifp = m->m_pkthdr.rcvif;
419 
420 #ifndef PULLDOWN_TEST
421 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
422 	/* m might change if M_LOOP.  So, call mtod after this */
423 #endif
424 
425 	/*
426 	 * Locate icmp6 structure in mbuf, and check
427 	 * that not corrupted and of at least minimum length
428 	 */
429 
430 	ip6 = mtod(m, struct ip6_hdr *);
431 	if (icmp6len < sizeof(struct icmp6_hdr)) {
432 		ICMP6STAT_INC(icp6s_tooshort);
433 		goto freeit;
434 	}
435 
436 	/*
437 	 * Check multicast group membership.
438 	 * Note: SSM filters are not applied for ICMPv6 traffic.
439 	 */
440 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
441 		struct in6_multi	*inm;
442 
443 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
444 		if (inm == NULL) {
445 			IP6STAT_INC(ip6s_notmember);
446 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
447 			goto freeit;
448 		}
449 	}
450 
451 	/*
452 	 * calculate the checksum
453 	 */
454 #ifndef PULLDOWN_TEST
455 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
456 #else
457 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
458 	if (icmp6 == NULL) {
459 		ICMP6STAT_INC(icp6s_tooshort);
460 		return IPPROTO_DONE;
461 	}
462 #endif
463 	code = icmp6->icmp6_code;
464 
465 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
466 		nd6log((LOG_ERR,
467 		    "ICMP6 checksum error(%d|%x) %s\n",
468 		    icmp6->icmp6_type, sum,
469 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
470 		ICMP6STAT_INC(icp6s_checksum);
471 		goto freeit;
472 	}
473 
474 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
475 		/*
476 		 * Deliver very specific ICMP6 type only.
477 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
478 		 * will not work.
479 		 */
480 		switch (icmp6->icmp6_type) {
481 		case ICMP6_DST_UNREACH:
482 		case ICMP6_PACKET_TOO_BIG:
483 		case ICMP6_TIME_EXCEEDED:
484 			break;
485 		default:
486 			goto freeit;
487 		}
488 	}
489 
490 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
491 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
492 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
493 		icmp6_ifstat_inc(ifp, ifs6_in_error);
494 
495 	switch (icmp6->icmp6_type) {
496 	case ICMP6_DST_UNREACH:
497 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
498 		switch (code) {
499 		case ICMP6_DST_UNREACH_NOROUTE:
500 			code = PRC_UNREACH_NET;
501 			break;
502 		case ICMP6_DST_UNREACH_ADMIN:
503 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
504 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
505 			break;
506 		case ICMP6_DST_UNREACH_ADDR:
507 			code = PRC_HOSTDEAD;
508 			break;
509 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
510 			/* I mean "source address was incorrect." */
511 			code = PRC_PARAMPROB;
512 			break;
513 		case ICMP6_DST_UNREACH_NOPORT:
514 			code = PRC_UNREACH_PORT;
515 			break;
516 		default:
517 			goto badcode;
518 		}
519 		goto deliver;
520 		break;
521 
522 	case ICMP6_PACKET_TOO_BIG:
523 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
524 
525 		/* validation is made in icmp6_mtudisc_update */
526 
527 		code = PRC_MSGSIZE;
528 
529 		/*
530 		 * Updating the path MTU will be done after examining
531 		 * intermediate extension headers.
532 		 */
533 		goto deliver;
534 		break;
535 
536 	case ICMP6_TIME_EXCEEDED:
537 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
538 		switch (code) {
539 		case ICMP6_TIME_EXCEED_TRANSIT:
540 			code = PRC_TIMXCEED_INTRANS;
541 			break;
542 		case ICMP6_TIME_EXCEED_REASSEMBLY:
543 			code = PRC_TIMXCEED_REASS;
544 			break;
545 		default:
546 			goto badcode;
547 		}
548 		goto deliver;
549 		break;
550 
551 	case ICMP6_PARAM_PROB:
552 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
553 		switch (code) {
554 		case ICMP6_PARAMPROB_NEXTHEADER:
555 			code = PRC_UNREACH_PROTOCOL;
556 			break;
557 		case ICMP6_PARAMPROB_HEADER:
558 		case ICMP6_PARAMPROB_OPTION:
559 			code = PRC_PARAMPROB;
560 			break;
561 		default:
562 			goto badcode;
563 		}
564 		goto deliver;
565 		break;
566 
567 	case ICMP6_ECHO_REQUEST:
568 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
569 		if (code != 0)
570 			goto badcode;
571 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
572 			/* Give up remote */
573 			break;
574 		}
575 		if ((n->m_flags & M_EXT) != 0
576 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
577 			struct mbuf *n0 = n;
578 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
579 			int n0len;
580 
581 			MGETHDR(n, M_DONTWAIT, n0->m_type);
582 			n0len = n0->m_pkthdr.len;	/* save for use below */
583 			if (n)
584 				M_MOVE_PKTHDR(n, n0);
585 			if (n && maxlen >= MHLEN) {
586 				MCLGET(n, M_DONTWAIT);
587 				if ((n->m_flags & M_EXT) == 0) {
588 					m_free(n);
589 					n = NULL;
590 				}
591 			}
592 			if (n == NULL) {
593 				/* Give up remote */
594 				m_freem(n0);
595 				break;
596 			}
597 			/*
598 			 * Copy IPv6 and ICMPv6 only.
599 			 */
600 			nip6 = mtod(n, struct ip6_hdr *);
601 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
602 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
603 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
604 			noff = sizeof(struct ip6_hdr);
605 			/* new mbuf contains only ipv6+icmpv6 headers */
606 			n->m_len = noff + sizeof(struct icmp6_hdr);
607 			/*
608 			 * Adjust mbuf.  ip6_plen will be adjusted in
609 			 * ip6_output().
610 			 */
611 			m_adj(n0, off + sizeof(struct icmp6_hdr));
612 			/* recalculate complete packet size */
613 			n->m_pkthdr.len = n0len + (noff - off);
614 			n->m_next = n0;
615 		} else {
616 			nip6 = mtod(n, struct ip6_hdr *);
617 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
618 			    sizeof(*nicmp6));
619 			noff = off;
620 		}
621 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
622 		nicmp6->icmp6_code = 0;
623 		if (n) {
624 			ICMP6STAT_INC(icp6s_reflect);
625 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
626 			icmp6_reflect(n, noff);
627 		}
628 		break;
629 
630 	case ICMP6_ECHO_REPLY:
631 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
632 		if (code != 0)
633 			goto badcode;
634 		break;
635 
636 	case MLD_LISTENER_QUERY:
637 	case MLD_LISTENER_REPORT:
638 	case MLD_LISTENER_DONE:
639 	case MLDV2_LISTENER_REPORT:
640 		/*
641 		 * Drop MLD traffic which is not link-local, has a hop limit
642 		 * of greater than 1 hop, or which does not have the
643 		 * IPv6 HBH Router Alert option.
644 		 * As IPv6 HBH options are stripped in ip6_input() we must
645 		 * check an mbuf header flag.
646 		 * XXX Should we also sanity check that these messages
647 		 * were directed to a link-local multicast prefix?
648 		 */
649 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
650 			goto freeit;
651 		if (mld_input(m, off, icmp6len) != 0)
652 			return (IPPROTO_DONE);
653 		/* m stays. */
654 		break;
655 
656 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
657 	    {
658 		enum { WRU, FQDN } mode;
659 
660 		if (!V_icmp6_nodeinfo)
661 			break;
662 
663 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
664 			mode = WRU;
665 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
666 			mode = FQDN;
667 		else
668 			goto badlen;
669 
670 		if (mode == FQDN) {
671 #ifndef PULLDOWN_TEST
672 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
673 			    IPPROTO_DONE);
674 #endif
675 			n = m_copy(m, 0, M_COPYALL);
676 			if (n)
677 				n = ni6_input(n, off);
678 			/* XXX meaningless if n == NULL */
679 			noff = sizeof(struct ip6_hdr);
680 		} else {
681 			struct prison *pr;
682 			u_char *p;
683 			int maxlen, maxhlen, hlen;
684 
685 			/*
686 			 * XXX: this combination of flags is pointless,
687 			 * but should we keep this for compatibility?
688 			 */
689 			if ((V_icmp6_nodeinfo & 5) != 5)
690 				break;
691 
692 			if (code != 0)
693 				goto badcode;
694 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
695 			if (maxlen >= MCLBYTES) {
696 				/* Give up remote */
697 				break;
698 			}
699 			MGETHDR(n, M_DONTWAIT, m->m_type);
700 			if (n && maxlen > MHLEN) {
701 				MCLGET(n, M_DONTWAIT);
702 				if ((n->m_flags & M_EXT) == 0) {
703 					m_free(n);
704 					n = NULL;
705 				}
706 			}
707 			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
708 				/*
709 				 * Previous code did a blind M_COPY_PKTHDR
710 				 * and said "just for rcvif".  If true, then
711 				 * we could tolerate the dup failing (due to
712 				 * the deep copy of the tag chain).  For now
713 				 * be conservative and just fail.
714 				 */
715 				m_free(n);
716 				n = NULL;
717 			}
718 			if (n == NULL) {
719 				/* Give up remote */
720 				break;
721 			}
722 			n->m_pkthdr.rcvif = NULL;
723 			n->m_len = 0;
724 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
725 			pr = curthread->td_ucred->cr_prison;
726 			mtx_lock(&pr->pr_mtx);
727 			hlen = strlen(pr->pr_hostname);
728 			if (maxhlen > hlen)
729 				maxhlen = hlen;
730 			/*
731 			 * Copy IPv6 and ICMPv6 only.
732 			 */
733 			nip6 = mtod(n, struct ip6_hdr *);
734 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
735 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
736 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
737 			p = (u_char *)(nicmp6 + 1);
738 			bzero(p, 4);
739 			/* meaningless TTL */
740 			bcopy(pr->pr_hostname, p + 4, maxhlen);
741 			mtx_unlock(&pr->pr_mtx);
742 			noff = sizeof(struct ip6_hdr);
743 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
744 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
745 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
746 			nicmp6->icmp6_code = 0;
747 		}
748 		if (n) {
749 			ICMP6STAT_INC(icp6s_reflect);
750 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
751 			icmp6_reflect(n, noff);
752 		}
753 		break;
754 	    }
755 
756 	case ICMP6_WRUREPLY:
757 		if (code != 0)
758 			goto badcode;
759 		break;
760 
761 	case ND_ROUTER_SOLICIT:
762 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
763 		if (code != 0)
764 			goto badcode;
765 		if (icmp6len < sizeof(struct nd_router_solicit))
766 			goto badlen;
767 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
768 			/* give up local */
769 			nd6_rs_input(m, off, icmp6len);
770 			m = NULL;
771 			goto freeit;
772 		}
773 		nd6_rs_input(n, off, icmp6len);
774 		/* m stays. */
775 		break;
776 
777 	case ND_ROUTER_ADVERT:
778 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
779 		if (code != 0)
780 			goto badcode;
781 		if (icmp6len < sizeof(struct nd_router_advert))
782 			goto badlen;
783 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
784 			/* give up local */
785 			nd6_ra_input(m, off, icmp6len);
786 			m = NULL;
787 			goto freeit;
788 		}
789 		nd6_ra_input(n, off, icmp6len);
790 		/* m stays. */
791 		break;
792 
793 	case ND_NEIGHBOR_SOLICIT:
794 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
798 			goto badlen;
799 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800 			/* give up local */
801 			nd6_ns_input(m, off, icmp6len);
802 			m = NULL;
803 			goto freeit;
804 		}
805 		nd6_ns_input(n, off, icmp6len);
806 		/* m stays. */
807 		break;
808 
809 	case ND_NEIGHBOR_ADVERT:
810 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
811 		if (code != 0)
812 			goto badcode;
813 		if (icmp6len < sizeof(struct nd_neighbor_advert))
814 			goto badlen;
815 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
816 			/* give up local */
817 			nd6_na_input(m, off, icmp6len);
818 			m = NULL;
819 			goto freeit;
820 		}
821 		nd6_na_input(n, off, icmp6len);
822 		/* m stays. */
823 		break;
824 
825 	case ND_REDIRECT:
826 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
827 		if (code != 0)
828 			goto badcode;
829 		if (icmp6len < sizeof(struct nd_redirect))
830 			goto badlen;
831 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
832 			/* give up local */
833 			icmp6_redirect_input(m, off);
834 			m = NULL;
835 			goto freeit;
836 		}
837 		icmp6_redirect_input(n, off);
838 		/* m stays. */
839 		break;
840 
841 	case ICMP6_ROUTER_RENUMBERING:
842 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
843 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
844 			goto badcode;
845 		if (icmp6len < sizeof(struct icmp6_router_renum))
846 			goto badlen;
847 		break;
848 
849 	default:
850 		nd6log((LOG_DEBUG,
851 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
852 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
853 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
854 		    ifp ? ifp->if_index : 0));
855 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
856 			/* ICMPv6 error: MUST deliver it by spec... */
857 			code = PRC_NCMDS;
858 			/* deliver */
859 		} else {
860 			/* ICMPv6 informational: MUST not deliver */
861 			break;
862 		}
863 	deliver:
864 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
865 			/* In this case, m should've been freed. */
866 			return (IPPROTO_DONE);
867 		}
868 		break;
869 
870 	badcode:
871 		ICMP6STAT_INC(icp6s_badcode);
872 		break;
873 
874 	badlen:
875 		ICMP6STAT_INC(icp6s_badlen);
876 		break;
877 	}
878 
879 	/* deliver the packet to appropriate sockets */
880 	icmp6_rip6_input(&m, *offp);
881 
882 	return IPPROTO_DONE;
883 
884  freeit:
885 	m_freem(m);
886 	return IPPROTO_DONE;
887 }
888 
889 static int
890 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
891 {
892 	struct mbuf *m = *mp;
893 	struct icmp6_hdr *icmp6;
894 	struct ip6_hdr *eip6;
895 	u_int32_t notifymtu;
896 	struct sockaddr_in6 icmp6src, icmp6dst;
897 
898 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
899 		ICMP6STAT_INC(icp6s_tooshort);
900 		goto freeit;
901 	}
902 #ifndef PULLDOWN_TEST
903 	IP6_EXTHDR_CHECK(m, off,
904 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
905 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
906 #else
907 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
908 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
909 	if (icmp6 == NULL) {
910 		ICMP6STAT_INC(icp6s_tooshort);
911 		return (-1);
912 	}
913 #endif
914 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
915 
916 	/* Detect the upper level protocol */
917 	{
918 		void (*ctlfunc)(int, struct sockaddr *, void *);
919 		u_int8_t nxt = eip6->ip6_nxt;
920 		int eoff = off + sizeof(struct icmp6_hdr) +
921 		    sizeof(struct ip6_hdr);
922 		struct ip6ctlparam ip6cp;
923 		struct in6_addr *finaldst = NULL;
924 		int icmp6type = icmp6->icmp6_type;
925 		struct ip6_frag *fh;
926 		struct ip6_rthdr *rth;
927 		struct ip6_rthdr0 *rth0;
928 		int rthlen;
929 
930 		while (1) { /* XXX: should avoid infinite loop explicitly? */
931 			struct ip6_ext *eh;
932 
933 			switch (nxt) {
934 			case IPPROTO_HOPOPTS:
935 			case IPPROTO_DSTOPTS:
936 			case IPPROTO_AH:
937 #ifndef PULLDOWN_TEST
938 				IP6_EXTHDR_CHECK(m, 0,
939 				    eoff + sizeof(struct ip6_ext), -1);
940 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
941 #else
942 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
943 				    eoff, sizeof(*eh));
944 				if (eh == NULL) {
945 					ICMP6STAT_INC(icp6s_tooshort);
946 					return (-1);
947 				}
948 #endif
949 
950 				if (nxt == IPPROTO_AH)
951 					eoff += (eh->ip6e_len + 2) << 2;
952 				else
953 					eoff += (eh->ip6e_len + 1) << 3;
954 				nxt = eh->ip6e_nxt;
955 				break;
956 			case IPPROTO_ROUTING:
957 				/*
958 				 * When the erroneous packet contains a
959 				 * routing header, we should examine the
960 				 * header to determine the final destination.
961 				 * Otherwise, we can't properly update
962 				 * information that depends on the final
963 				 * destination (e.g. path MTU).
964 				 */
965 #ifndef PULLDOWN_TEST
966 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
967 				rth = (struct ip6_rthdr *)
968 				    (mtod(m, caddr_t) + eoff);
969 #else
970 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
971 				    eoff, sizeof(*rth));
972 				if (rth == NULL) {
973 					ICMP6STAT_INC(icp6s_tooshort);
974 					return (-1);
975 				}
976 #endif
977 				rthlen = (rth->ip6r_len + 1) << 3;
978 				/*
979 				 * XXX: currently there is no
980 				 * officially defined type other
981 				 * than type-0.
982 				 * Note that if the segment left field
983 				 * is 0, all intermediate hops must
984 				 * have been passed.
985 				 */
986 				if (rth->ip6r_segleft &&
987 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
988 					int hops;
989 
990 #ifndef PULLDOWN_TEST
991 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
992 					rth0 = (struct ip6_rthdr0 *)
993 					    (mtod(m, caddr_t) + eoff);
994 #else
995 					IP6_EXTHDR_GET(rth0,
996 					    struct ip6_rthdr0 *, m,
997 					    eoff, rthlen);
998 					if (rth0 == NULL) {
999 						ICMP6STAT_INC(icp6s_tooshort);
1000 						return (-1);
1001 					}
1002 #endif
1003 					/* just ignore a bogus header */
1004 					if ((rth0->ip6r0_len % 2) == 0 &&
1005 					    (hops = rth0->ip6r0_len/2))
1006 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1007 				}
1008 				eoff += rthlen;
1009 				nxt = rth->ip6r_nxt;
1010 				break;
1011 			case IPPROTO_FRAGMENT:
1012 #ifndef PULLDOWN_TEST
1013 				IP6_EXTHDR_CHECK(m, 0, eoff +
1014 				    sizeof(struct ip6_frag), -1);
1015 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1016 				    eoff);
1017 #else
1018 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1019 				    eoff, sizeof(*fh));
1020 				if (fh == NULL) {
1021 					ICMP6STAT_INC(icp6s_tooshort);
1022 					return (-1);
1023 				}
1024 #endif
1025 				/*
1026 				 * Data after a fragment header is meaningless
1027 				 * unless it is the first fragment, but
1028 				 * we'll go to the notify label for path MTU
1029 				 * discovery.
1030 				 */
1031 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1032 					goto notify;
1033 
1034 				eoff += sizeof(struct ip6_frag);
1035 				nxt = fh->ip6f_nxt;
1036 				break;
1037 			default:
1038 				/*
1039 				 * This case includes ESP and the No Next
1040 				 * Header.  In such cases going to the notify
1041 				 * label does not have any meaning
1042 				 * (i.e. ctlfunc will be NULL), but we go
1043 				 * anyway since we might have to update
1044 				 * path MTU information.
1045 				 */
1046 				goto notify;
1047 			}
1048 		}
1049 	  notify:
1050 #ifndef PULLDOWN_TEST
1051 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1052 #else
1053 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1054 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1055 		if (icmp6 == NULL) {
1056 			ICMP6STAT_INC(icp6s_tooshort);
1057 			return (-1);
1058 		}
1059 #endif
1060 
1061 		/*
1062 		 * retrieve parameters from the inner IPv6 header, and convert
1063 		 * them into sockaddr structures.
1064 		 * XXX: there is no guarantee that the source or destination
1065 		 * addresses of the inner packet are in the same scope as
1066 		 * the addresses of the icmp packet.  But there is no other
1067 		 * way to determine the zone.
1068 		 */
1069 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1070 
1071 		bzero(&icmp6dst, sizeof(icmp6dst));
1072 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1073 		icmp6dst.sin6_family = AF_INET6;
1074 		if (finaldst == NULL)
1075 			icmp6dst.sin6_addr = eip6->ip6_dst;
1076 		else
1077 			icmp6dst.sin6_addr = *finaldst;
1078 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1079 			goto freeit;
1080 		bzero(&icmp6src, sizeof(icmp6src));
1081 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1082 		icmp6src.sin6_family = AF_INET6;
1083 		icmp6src.sin6_addr = eip6->ip6_src;
1084 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1085 			goto freeit;
1086 		icmp6src.sin6_flowinfo =
1087 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1088 
1089 		if (finaldst == NULL)
1090 			finaldst = &eip6->ip6_dst;
1091 		ip6cp.ip6c_m = m;
1092 		ip6cp.ip6c_icmp6 = icmp6;
1093 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1094 		ip6cp.ip6c_off = eoff;
1095 		ip6cp.ip6c_finaldst = finaldst;
1096 		ip6cp.ip6c_src = &icmp6src;
1097 		ip6cp.ip6c_nxt = nxt;
1098 
1099 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1100 			notifymtu = ntohl(icmp6->icmp6_mtu);
1101 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1102 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1103 		}
1104 
1105 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1106 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1107 		if (ctlfunc) {
1108 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1109 			    &ip6cp);
1110 		}
1111 	}
1112 	*mp = m;
1113 	return (0);
1114 
1115   freeit:
1116 	m_freem(m);
1117 	return (-1);
1118 }
1119 
1120 void
1121 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1122 {
1123 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1124 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1125 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1126 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1127 	struct in_conninfo inc;
1128 
1129 #if 0
1130 	/*
1131 	 * RFC2460 section 5, last paragraph.
1132 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1133 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1134 	 * due to packet translator in the middle.
1135 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1136 	 * special handling.
1137 	 */
1138 	if (mtu < IPV6_MMTU)
1139 		return;
1140 #endif
1141 
1142 	/*
1143 	 * we reject ICMPv6 too big with abnormally small value.
1144 	 * XXX what is the good definition of "abnormally small"?
1145 	 */
1146 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1147 		return;
1148 
1149 	if (!validated)
1150 		return;
1151 
1152 	/*
1153 	 * In case the suggested mtu is less than IPV6_MMTU, we
1154 	 * only need to remember that it was for above mentioned
1155 	 * "alwaysfrag" case.
1156 	 * Try to be as close to the spec as possible.
1157 	 */
1158 	if (mtu < IPV6_MMTU)
1159 		mtu = IPV6_MMTU - 8;
1160 
1161 	bzero(&inc, sizeof(inc));
1162 	inc.inc_flags |= INC_ISIPV6;
1163 	inc.inc6_faddr = *dst;
1164 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1165 		return;
1166 
1167 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1168 		tcp_hc_updatemtu(&inc, mtu);
1169 		ICMP6STAT_INC(icp6s_pmtuchg);
1170 	}
1171 }
1172 
1173 /*
1174  * Process a Node Information Query packet, based on
1175  * draft-ietf-ipngwg-icmp-name-lookups-07.
1176  *
1177  * Spec incompatibilities:
1178  * - IPv6 Subject address handling
1179  * - IPv4 Subject address handling support missing
1180  * - Proxy reply (answer even if it's not for me)
1181  * - joins NI group address at in6_ifattach() time only, does not cope
1182  *   with hostname changes by sethostname(3)
1183  */
1184 static struct mbuf *
1185 ni6_input(struct mbuf *m, int off)
1186 {
1187 	struct icmp6_nodeinfo *ni6, *nni6;
1188 	struct mbuf *n = NULL;
1189 	struct prison *pr;
1190 	u_int16_t qtype;
1191 	int subjlen;
1192 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1193 	struct ni_reply_fqdn *fqdn;
1194 	int addrs;		/* for NI_QTYPE_NODEADDR */
1195 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1196 	struct in6_addr in6_subj; /* subject address */
1197 	struct ip6_hdr *ip6;
1198 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1199 	char *subj = NULL;
1200 	struct in6_ifaddr *ia6 = NULL;
1201 
1202 	ip6 = mtod(m, struct ip6_hdr *);
1203 #ifndef PULLDOWN_TEST
1204 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1205 #else
1206 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1207 	if (ni6 == NULL) {
1208 		/* m is already reclaimed */
1209 		return (NULL);
1210 	}
1211 #endif
1212 
1213 	/*
1214 	 * Validate IPv6 source address.
1215 	 * The default configuration MUST be to refuse answering queries from
1216 	 * global-scope addresses according to RFC4602.
1217 	 * Notes:
1218 	 *  - it's not very clear what "refuse" means; this implementation
1219 	 *    simply drops it.
1220 	 *  - it's not very easy to identify global-scope (unicast) addresses
1221 	 *    since there are many prefixes for them.  It should be safer
1222 	 *    and in practice sufficient to check "all" but loopback and
1223 	 *    link-local (note that site-local unicast was deprecated and
1224 	 *    ULA is defined as global scope-wise)
1225 	 */
1226 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1227 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1228 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1229 		goto bad;
1230 
1231 	/*
1232 	 * Validate IPv6 destination address.
1233 	 *
1234 	 * The Responder must discard the Query without further processing
1235 	 * unless it is one of the Responder's unicast or anycast addresses, or
1236 	 * a link-local scope multicast address which the Responder has joined.
1237 	 * [RFC4602, Section 5.]
1238 	 */
1239 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1240 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1241 			goto bad;
1242 		/* else it's a link-local multicast, fine */
1243 	} else {		/* unicast or anycast */
1244 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1245 			goto bad; /* XXX impossible */
1246 
1247 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1248 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1249 			ifa_free(&ia6->ia_ifa);
1250 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1251 				"a temporary address in %s:%d",
1252 			       __FILE__, __LINE__));
1253 			goto bad;
1254 		}
1255 		ifa_free(&ia6->ia_ifa);
1256 	}
1257 
1258 	/* validate query Subject field. */
1259 	qtype = ntohs(ni6->ni_qtype);
1260 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1261 	switch (qtype) {
1262 	case NI_QTYPE_NOOP:
1263 	case NI_QTYPE_SUPTYPES:
1264 		/* 07 draft */
1265 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1266 			break;
1267 		/* FALLTHROUGH */
1268 	case NI_QTYPE_FQDN:
1269 	case NI_QTYPE_NODEADDR:
1270 	case NI_QTYPE_IPV4ADDR:
1271 		switch (ni6->ni_code) {
1272 		case ICMP6_NI_SUBJ_IPV6:
1273 #if ICMP6_NI_SUBJ_IPV6 != 0
1274 		case 0:
1275 #endif
1276 			/*
1277 			 * backward compatibility - try to accept 03 draft
1278 			 * format, where no Subject is present.
1279 			 */
1280 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1281 			    subjlen == 0) {
1282 				oldfqdn++;
1283 				break;
1284 			}
1285 #if ICMP6_NI_SUBJ_IPV6 != 0
1286 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1287 				goto bad;
1288 #endif
1289 
1290 			if (subjlen != sizeof(struct in6_addr))
1291 				goto bad;
1292 
1293 			/*
1294 			 * Validate Subject address.
1295 			 *
1296 			 * Not sure what exactly "address belongs to the node"
1297 			 * means in the spec, is it just unicast, or what?
1298 			 *
1299 			 * At this moment we consider Subject address as
1300 			 * "belong to the node" if the Subject address equals
1301 			 * to the IPv6 destination address; validation for
1302 			 * IPv6 destination address should have done enough
1303 			 * check for us.
1304 			 *
1305 			 * We do not do proxy at this moment.
1306 			 */
1307 			/* m_pulldown instead of copy? */
1308 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1309 			    subjlen, (caddr_t)&in6_subj);
1310 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1311 				goto bad;
1312 
1313 			subj = (char *)&in6_subj;
1314 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1315 				break;
1316 
1317 			/*
1318 			 * XXX if we are to allow other cases, we should really
1319 			 * be careful about scope here.
1320 			 * basically, we should disallow queries toward IPv6
1321 			 * destination X with subject Y,
1322 			 * if scope(X) > scope(Y).
1323 			 * if we allow scope(X) > scope(Y), it will result in
1324 			 * information leakage across scope boundary.
1325 			 */
1326 			goto bad;
1327 
1328 		case ICMP6_NI_SUBJ_FQDN:
1329 			/*
1330 			 * Validate Subject name with gethostname(3).
1331 			 *
1332 			 * The behavior may need some debate, since:
1333 			 * - we are not sure if the node has FQDN as
1334 			 *   hostname (returned by gethostname(3)).
1335 			 * - the code does wildcard match for truncated names.
1336 			 *   however, we are not sure if we want to perform
1337 			 *   wildcard match, if gethostname(3) side has
1338 			 *   truncated hostname.
1339 			 */
1340 			pr = curthread->td_ucred->cr_prison;
1341 			mtx_lock(&pr->pr_mtx);
1342 			n = ni6_nametodns(pr->pr_hostname,
1343 			    strlen(pr->pr_hostname), 0);
1344 			mtx_unlock(&pr->pr_mtx);
1345 			if (!n || n->m_next || n->m_len == 0)
1346 				goto bad;
1347 			IP6_EXTHDR_GET(subj, char *, m,
1348 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1349 			if (subj == NULL)
1350 				goto bad;
1351 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1352 			    n->m_len)) {
1353 				goto bad;
1354 			}
1355 			m_freem(n);
1356 			n = NULL;
1357 			break;
1358 
1359 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1360 		default:
1361 			goto bad;
1362 		}
1363 		break;
1364 	}
1365 
1366 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1367 	switch (qtype) {
1368 	case NI_QTYPE_FQDN:
1369 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1370 			goto bad;
1371 		break;
1372 	case NI_QTYPE_NODEADDR:
1373 	case NI_QTYPE_IPV4ADDR:
1374 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1375 			goto bad;
1376 		break;
1377 	}
1378 
1379 	/* guess reply length */
1380 	switch (qtype) {
1381 	case NI_QTYPE_NOOP:
1382 		break;		/* no reply data */
1383 	case NI_QTYPE_SUPTYPES:
1384 		replylen += sizeof(u_int32_t);
1385 		break;
1386 	case NI_QTYPE_FQDN:
1387 		/* XXX will append an mbuf */
1388 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1389 		break;
1390 	case NI_QTYPE_NODEADDR:
1391 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1392 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1393 		    sizeof(u_int32_t))) > MCLBYTES)
1394 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1395 		break;
1396 	case NI_QTYPE_IPV4ADDR:
1397 		/* unsupported - should respond with unknown Qtype? */
1398 		break;
1399 	default:
1400 		/*
1401 		 * XXX: We must return a reply with the ICMP6 code
1402 		 * `unknown Qtype' in this case.  However we regard the case
1403 		 * as an FQDN query for backward compatibility.
1404 		 * Older versions set a random value to this field,
1405 		 * so it rarely varies in the defined qtypes.
1406 		 * But the mechanism is not reliable...
1407 		 * maybe we should obsolete older versions.
1408 		 */
1409 		qtype = NI_QTYPE_FQDN;
1410 		/* XXX will append an mbuf */
1411 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1412 		oldfqdn++;
1413 		break;
1414 	}
1415 
1416 	/* allocate an mbuf to reply. */
1417 	MGETHDR(n, M_DONTWAIT, m->m_type);
1418 	if (n == NULL) {
1419 		m_freem(m);
1420 		return (NULL);
1421 	}
1422 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1423 	if (replylen > MHLEN) {
1424 		if (replylen > MCLBYTES) {
1425 			/*
1426 			 * XXX: should we try to allocate more? But MCLBYTES
1427 			 * is probably much larger than IPV6_MMTU...
1428 			 */
1429 			goto bad;
1430 		}
1431 		MCLGET(n, M_DONTWAIT);
1432 		if ((n->m_flags & M_EXT) == 0) {
1433 			goto bad;
1434 		}
1435 	}
1436 	n->m_pkthdr.len = n->m_len = replylen;
1437 
1438 	/* copy mbuf header and IPv6 + Node Information base headers */
1439 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1440 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1441 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1442 
1443 	/* qtype dependent procedure */
1444 	switch (qtype) {
1445 	case NI_QTYPE_NOOP:
1446 		nni6->ni_code = ICMP6_NI_SUCCESS;
1447 		nni6->ni_flags = 0;
1448 		break;
1449 	case NI_QTYPE_SUPTYPES:
1450 	{
1451 		u_int32_t v;
1452 		nni6->ni_code = ICMP6_NI_SUCCESS;
1453 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1454 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1455 		v = (u_int32_t)htonl(0x0000000f);
1456 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1457 		break;
1458 	}
1459 	case NI_QTYPE_FQDN:
1460 		nni6->ni_code = ICMP6_NI_SUCCESS;
1461 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1462 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1463 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1464 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1465 		/*
1466 		 * XXX do we really have FQDN in hostname?
1467 		 */
1468 		pr = curthread->td_ucred->cr_prison;
1469 		mtx_lock(&pr->pr_mtx);
1470 		n->m_next = ni6_nametodns(pr->pr_hostname,
1471 		    strlen(pr->pr_hostname), oldfqdn);
1472 		mtx_unlock(&pr->pr_mtx);
1473 		if (n->m_next == NULL)
1474 			goto bad;
1475 		/* XXX we assume that n->m_next is not a chain */
1476 		if (n->m_next->m_next != NULL)
1477 			goto bad;
1478 		n->m_pkthdr.len += n->m_next->m_len;
1479 		break;
1480 	case NI_QTYPE_NODEADDR:
1481 	{
1482 		int lenlim, copied;
1483 
1484 		nni6->ni_code = ICMP6_NI_SUCCESS;
1485 		n->m_pkthdr.len = n->m_len =
1486 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1487 		lenlim = M_TRAILINGSPACE(n);
1488 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1489 		/* XXX: reset mbuf length */
1490 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1491 		    sizeof(struct icmp6_nodeinfo) + copied;
1492 		break;
1493 	}
1494 	default:
1495 		break;		/* XXX impossible! */
1496 	}
1497 
1498 	nni6->ni_type = ICMP6_NI_REPLY;
1499 	m_freem(m);
1500 	return (n);
1501 
1502   bad:
1503 	m_freem(m);
1504 	if (n)
1505 		m_freem(n);
1506 	return (NULL);
1507 }
1508 
1509 /*
1510  * make a mbuf with DNS-encoded string.  no compression support.
1511  *
1512  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1513  * treated as truncated name (two \0 at the end).  this is a wild guess.
1514  *
1515  * old - return pascal string if non-zero
1516  */
1517 static struct mbuf *
1518 ni6_nametodns(const char *name, int namelen, int old)
1519 {
1520 	struct mbuf *m;
1521 	char *cp, *ep;
1522 	const char *p, *q;
1523 	int i, len, nterm;
1524 
1525 	if (old)
1526 		len = namelen + 1;
1527 	else
1528 		len = MCLBYTES;
1529 
1530 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1531 	MGET(m, M_DONTWAIT, MT_DATA);
1532 	if (m && len > MLEN) {
1533 		MCLGET(m, M_DONTWAIT);
1534 		if ((m->m_flags & M_EXT) == 0)
1535 			goto fail;
1536 	}
1537 	if (!m)
1538 		goto fail;
1539 	m->m_next = NULL;
1540 
1541 	if (old) {
1542 		m->m_len = len;
1543 		*mtod(m, char *) = namelen;
1544 		bcopy(name, mtod(m, char *) + 1, namelen);
1545 		return m;
1546 	} else {
1547 		m->m_len = 0;
1548 		cp = mtod(m, char *);
1549 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1550 
1551 		/* if not certain about my name, return empty buffer */
1552 		if (namelen == 0)
1553 			return m;
1554 
1555 		/*
1556 		 * guess if it looks like shortened hostname, or FQDN.
1557 		 * shortened hostname needs two trailing "\0".
1558 		 */
1559 		i = 0;
1560 		for (p = name; p < name + namelen; p++) {
1561 			if (*p && *p == '.')
1562 				i++;
1563 		}
1564 		if (i < 2)
1565 			nterm = 2;
1566 		else
1567 			nterm = 1;
1568 
1569 		p = name;
1570 		while (cp < ep && p < name + namelen) {
1571 			i = 0;
1572 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1573 				i++;
1574 			/* result does not fit into mbuf */
1575 			if (cp + i + 1 >= ep)
1576 				goto fail;
1577 			/*
1578 			 * DNS label length restriction, RFC1035 page 8.
1579 			 * "i == 0" case is included here to avoid returning
1580 			 * 0-length label on "foo..bar".
1581 			 */
1582 			if (i <= 0 || i >= 64)
1583 				goto fail;
1584 			*cp++ = i;
1585 			bcopy(p, cp, i);
1586 			cp += i;
1587 			p = q;
1588 			if (p < name + namelen && *p == '.')
1589 				p++;
1590 		}
1591 		/* termination */
1592 		if (cp + nterm >= ep)
1593 			goto fail;
1594 		while (nterm-- > 0)
1595 			*cp++ = '\0';
1596 		m->m_len = cp - mtod(m, char *);
1597 		return m;
1598 	}
1599 
1600 	panic("should not reach here");
1601 	/* NOTREACHED */
1602 
1603  fail:
1604 	if (m)
1605 		m_freem(m);
1606 	return NULL;
1607 }
1608 
1609 /*
1610  * check if two DNS-encoded string matches.  takes care of truncated
1611  * form (with \0\0 at the end).  no compression support.
1612  * XXX upper/lowercase match (see RFC2065)
1613  */
1614 static int
1615 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1616 {
1617 	const char *a0, *b0;
1618 	int l;
1619 
1620 	/* simplest case - need validation? */
1621 	if (alen == blen && bcmp(a, b, alen) == 0)
1622 		return 1;
1623 
1624 	a0 = a;
1625 	b0 = b;
1626 
1627 	/* termination is mandatory */
1628 	if (alen < 2 || blen < 2)
1629 		return 0;
1630 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1631 		return 0;
1632 	alen--;
1633 	blen--;
1634 
1635 	while (a - a0 < alen && b - b0 < blen) {
1636 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1637 			return 0;
1638 
1639 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1640 			return 0;
1641 		/* we don't support compression yet */
1642 		if (a[0] >= 64 || b[0] >= 64)
1643 			return 0;
1644 
1645 		/* truncated case */
1646 		if (a[0] == 0 && a - a0 == alen - 1)
1647 			return 1;
1648 		if (b[0] == 0 && b - b0 == blen - 1)
1649 			return 1;
1650 		if (a[0] == 0 || b[0] == 0)
1651 			return 0;
1652 
1653 		if (a[0] != b[0])
1654 			return 0;
1655 		l = a[0];
1656 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1657 			return 0;
1658 		if (bcmp(a + 1, b + 1, l) != 0)
1659 			return 0;
1660 
1661 		a += 1 + l;
1662 		b += 1 + l;
1663 	}
1664 
1665 	if (a - a0 == alen && b - b0 == blen)
1666 		return 1;
1667 	else
1668 		return 0;
1669 }
1670 
1671 /*
1672  * calculate the number of addresses to be returned in the node info reply.
1673  */
1674 static int
1675 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1676     struct in6_addr *subj)
1677 {
1678 	struct ifnet *ifp;
1679 	struct in6_ifaddr *ifa6;
1680 	struct ifaddr *ifa;
1681 	int addrs = 0, addrsofif, iffound = 0;
1682 	int niflags = ni6->ni_flags;
1683 
1684 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1685 		switch (ni6->ni_code) {
1686 		case ICMP6_NI_SUBJ_IPV6:
1687 			if (subj == NULL) /* must be impossible... */
1688 				return (0);
1689 			break;
1690 		default:
1691 			/*
1692 			 * XXX: we only support IPv6 subject address for
1693 			 * this Qtype.
1694 			 */
1695 			return (0);
1696 		}
1697 	}
1698 
1699 	IFNET_RLOCK_NOSLEEP();
1700 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1701 		addrsofif = 0;
1702 		IF_ADDR_LOCK(ifp);
1703 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1704 			if (ifa->ifa_addr->sa_family != AF_INET6)
1705 				continue;
1706 			ifa6 = (struct in6_ifaddr *)ifa;
1707 
1708 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1709 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1710 				iffound = 1;
1711 
1712 			/*
1713 			 * IPv4-mapped addresses can only be returned by a
1714 			 * Node Information proxy, since they represent
1715 			 * addresses of IPv4-only nodes, which perforce do
1716 			 * not implement this protocol.
1717 			 * [icmp-name-lookups-07, Section 5.4]
1718 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1719 			 * this function at this moment.
1720 			 */
1721 
1722 			/* What do we have to do about ::1? */
1723 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1724 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1725 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1726 					continue;
1727 				break;
1728 			case IPV6_ADDR_SCOPE_SITELOCAL:
1729 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1730 					continue;
1731 				break;
1732 			case IPV6_ADDR_SCOPE_GLOBAL:
1733 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1734 					continue;
1735 				break;
1736 			default:
1737 				continue;
1738 			}
1739 
1740 			/*
1741 			 * check if anycast is okay.
1742 			 * XXX: just experimental.  not in the spec.
1743 			 */
1744 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1745 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1746 				continue; /* we need only unicast addresses */
1747 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1748 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1749 				continue;
1750 			}
1751 			addrsofif++; /* count the address */
1752 		}
1753 		IF_ADDR_UNLOCK(ifp);
1754 		if (iffound) {
1755 			*ifpp = ifp;
1756 			IFNET_RUNLOCK_NOSLEEP();
1757 			return (addrsofif);
1758 		}
1759 
1760 		addrs += addrsofif;
1761 	}
1762 	IFNET_RUNLOCK_NOSLEEP();
1763 
1764 	return (addrs);
1765 }
1766 
1767 static int
1768 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1769     struct ifnet *ifp0, int resid)
1770 {
1771 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1772 	struct in6_ifaddr *ifa6;
1773 	struct ifaddr *ifa;
1774 	struct ifnet *ifp_dep = NULL;
1775 	int copied = 0, allow_deprecated = 0;
1776 	u_char *cp = (u_char *)(nni6 + 1);
1777 	int niflags = ni6->ni_flags;
1778 	u_int32_t ltime;
1779 
1780 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1781 		return (0);	/* needless to copy */
1782 
1783 	IFNET_RLOCK_NOSLEEP();
1784   again:
1785 
1786 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1787 		IF_ADDR_LOCK(ifp);
1788 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1789 			if (ifa->ifa_addr->sa_family != AF_INET6)
1790 				continue;
1791 			ifa6 = (struct in6_ifaddr *)ifa;
1792 
1793 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1794 			    allow_deprecated == 0) {
1795 				/*
1796 				 * prefererred address should be put before
1797 				 * deprecated addresses.
1798 				 */
1799 
1800 				/* record the interface for later search */
1801 				if (ifp_dep == NULL)
1802 					ifp_dep = ifp;
1803 
1804 				continue;
1805 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1806 			    allow_deprecated != 0)
1807 				continue; /* we now collect deprecated addrs */
1808 
1809 			/* What do we have to do about ::1? */
1810 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1811 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1812 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1813 					continue;
1814 				break;
1815 			case IPV6_ADDR_SCOPE_SITELOCAL:
1816 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1817 					continue;
1818 				break;
1819 			case IPV6_ADDR_SCOPE_GLOBAL:
1820 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1821 					continue;
1822 				break;
1823 			default:
1824 				continue;
1825 			}
1826 
1827 			/*
1828 			 * check if anycast is okay.
1829 			 * XXX: just experimental.  not in the spec.
1830 			 */
1831 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1832 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1833 				continue;
1834 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1835 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1836 				continue;
1837 			}
1838 
1839 			/* now we can copy the address */
1840 			if (resid < sizeof(struct in6_addr) +
1841 			    sizeof(u_int32_t)) {
1842 				IF_ADDR_UNLOCK(ifp);
1843 				/*
1844 				 * We give up much more copy.
1845 				 * Set the truncate flag and return.
1846 				 */
1847 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1848 				IFNET_RUNLOCK_NOSLEEP();
1849 				return (copied);
1850 			}
1851 
1852 			/*
1853 			 * Set the TTL of the address.
1854 			 * The TTL value should be one of the following
1855 			 * according to the specification:
1856 			 *
1857 			 * 1. The remaining lifetime of a DHCP lease on the
1858 			 *    address, or
1859 			 * 2. The remaining Valid Lifetime of a prefix from
1860 			 *    which the address was derived through Stateless
1861 			 *    Autoconfiguration.
1862 			 *
1863 			 * Note that we currently do not support stateful
1864 			 * address configuration by DHCPv6, so the former
1865 			 * case can't happen.
1866 			 */
1867 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1868 				ltime = ND6_INFINITE_LIFETIME;
1869 			else {
1870 				if (ifa6->ia6_lifetime.ia6t_expire >
1871 				    time_second)
1872 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1873 				else
1874 					ltime = 0;
1875 			}
1876 
1877 			bcopy(&ltime, cp, sizeof(u_int32_t));
1878 			cp += sizeof(u_int32_t);
1879 
1880 			/* copy the address itself */
1881 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1882 			    sizeof(struct in6_addr));
1883 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1884 			cp += sizeof(struct in6_addr);
1885 
1886 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1887 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1888 		}
1889 		IF_ADDR_UNLOCK(ifp);
1890 		if (ifp0)	/* we need search only on the specified IF */
1891 			break;
1892 	}
1893 
1894 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1895 		ifp = ifp_dep;
1896 		allow_deprecated = 1;
1897 
1898 		goto again;
1899 	}
1900 
1901 	IFNET_RUNLOCK_NOSLEEP();
1902 
1903 	return (copied);
1904 }
1905 
1906 /*
1907  * XXX almost dup'ed code with rip6_input.
1908  */
1909 static int
1910 icmp6_rip6_input(struct mbuf **mp, int off)
1911 {
1912 	struct mbuf *m = *mp;
1913 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1914 	struct inpcb *in6p;
1915 	struct inpcb *last = NULL;
1916 	struct sockaddr_in6 fromsa;
1917 	struct icmp6_hdr *icmp6;
1918 	struct mbuf *opts = NULL;
1919 
1920 #ifndef PULLDOWN_TEST
1921 	/* this is assumed to be safe. */
1922 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1923 #else
1924 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1925 	if (icmp6 == NULL) {
1926 		/* m is already reclaimed */
1927 		return (IPPROTO_DONE);
1928 	}
1929 #endif
1930 
1931 	/*
1932 	 * XXX: the address may have embedded scope zone ID, which should be
1933 	 * hidden from applications.
1934 	 */
1935 	bzero(&fromsa, sizeof(fromsa));
1936 	fromsa.sin6_family = AF_INET6;
1937 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1938 	fromsa.sin6_addr = ip6->ip6_src;
1939 	if (sa6_recoverscope(&fromsa)) {
1940 		m_freem(m);
1941 		return (IPPROTO_DONE);
1942 	}
1943 
1944 	INP_INFO_RLOCK(&V_ripcbinfo);
1945 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1946 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1947 			continue;
1948 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1949 			continue;
1950 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1951 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1952 			continue;
1953 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1954 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1955 			continue;
1956 		INP_RLOCK(in6p);
1957 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1958 		    in6p->in6p_icmp6filt)) {
1959 			INP_RUNLOCK(in6p);
1960 			continue;
1961 		}
1962 		if (last != NULL) {
1963 			struct	mbuf *n = NULL;
1964 
1965 			/*
1966 			 * Recent network drivers tend to allocate a single
1967 			 * mbuf cluster, rather than to make a couple of
1968 			 * mbufs without clusters.  Also, since the IPv6 code
1969 			 * path tries to avoid m_pullup(), it is highly
1970 			 * probable that we still have an mbuf cluster here
1971 			 * even though the necessary length can be stored in an
1972 			 * mbuf's internal buffer.
1973 			 * Meanwhile, the default size of the receive socket
1974 			 * buffer for raw sockets is not so large.  This means
1975 			 * the possibility of packet loss is relatively higher
1976 			 * than before.  To avoid this scenario, we copy the
1977 			 * received data to a separate mbuf that does not use
1978 			 * a cluster, if possible.
1979 			 * XXX: it is better to copy the data after stripping
1980 			 * intermediate headers.
1981 			 */
1982 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1983 			    m->m_len <= MHLEN) {
1984 				MGET(n, M_DONTWAIT, m->m_type);
1985 				if (n != NULL) {
1986 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1987 						bcopy(m->m_data, n->m_data,
1988 						      m->m_len);
1989 						n->m_len = m->m_len;
1990 					} else {
1991 						m_free(n);
1992 						n = NULL;
1993 					}
1994 				}
1995 			}
1996 			if (n != NULL ||
1997 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1998 				if (last->inp_flags & INP_CONTROLOPTS)
1999 					ip6_savecontrol(last, n, &opts);
2000 				/* strip intermediate headers */
2001 				m_adj(n, off);
2002 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2003 				if (sbappendaddr_locked(
2004 				    &last->inp_socket->so_rcv,
2005 				    (struct sockaddr *)&fromsa, n, opts)
2006 				    == 0) {
2007 					/* should notify about lost packet */
2008 					m_freem(n);
2009 					if (opts) {
2010 						m_freem(opts);
2011 					}
2012 					SOCKBUF_UNLOCK(
2013 					    &last->inp_socket->so_rcv);
2014 				} else
2015 					sorwakeup_locked(last->inp_socket);
2016 				opts = NULL;
2017 			}
2018 			INP_RUNLOCK(last);
2019 		}
2020 		last = in6p;
2021 	}
2022 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2023 	if (last != NULL) {
2024 		if (last->inp_flags & INP_CONTROLOPTS)
2025 			ip6_savecontrol(last, m, &opts);
2026 		/* strip intermediate headers */
2027 		m_adj(m, off);
2028 
2029 		/* avoid using mbuf clusters if possible (see above) */
2030 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2031 		    m->m_len <= MHLEN) {
2032 			struct mbuf *n;
2033 
2034 			MGET(n, M_DONTWAIT, m->m_type);
2035 			if (n != NULL) {
2036 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2037 					bcopy(m->m_data, n->m_data, m->m_len);
2038 					n->m_len = m->m_len;
2039 
2040 					m_freem(m);
2041 					m = n;
2042 				} else {
2043 					m_freem(n);
2044 					n = NULL;
2045 				}
2046 			}
2047 		}
2048 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2049 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2050 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2051 			m_freem(m);
2052 			if (opts)
2053 				m_freem(opts);
2054 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2055 		} else
2056 			sorwakeup_locked(last->inp_socket);
2057 		INP_RUNLOCK(last);
2058 	} else {
2059 		m_freem(m);
2060 		IP6STAT_DEC(ip6s_delivered);
2061 	}
2062 	return IPPROTO_DONE;
2063 }
2064 
2065 /*
2066  * Reflect the ip6 packet back to the source.
2067  * OFF points to the icmp6 header, counted from the top of the mbuf.
2068  */
2069 void
2070 icmp6_reflect(struct mbuf *m, size_t off)
2071 {
2072 	struct ip6_hdr *ip6;
2073 	struct icmp6_hdr *icmp6;
2074 	struct in6_ifaddr *ia = NULL;
2075 	int plen;
2076 	int type, code;
2077 	struct ifnet *outif = NULL;
2078 	struct in6_addr origdst, src, *srcp = NULL;
2079 
2080 	/* too short to reflect */
2081 	if (off < sizeof(struct ip6_hdr)) {
2082 		nd6log((LOG_DEBUG,
2083 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2084 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2085 		    __FILE__, __LINE__));
2086 		goto bad;
2087 	}
2088 
2089 	/*
2090 	 * If there are extra headers between IPv6 and ICMPv6, strip
2091 	 * off that header first.
2092 	 */
2093 #ifdef DIAGNOSTIC
2094 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2095 		panic("assumption failed in icmp6_reflect");
2096 #endif
2097 	if (off > sizeof(struct ip6_hdr)) {
2098 		size_t l;
2099 		struct ip6_hdr nip6;
2100 
2101 		l = off - sizeof(struct ip6_hdr);
2102 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2103 		m_adj(m, l);
2104 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2105 		if (m->m_len < l) {
2106 			if ((m = m_pullup(m, l)) == NULL)
2107 				return;
2108 		}
2109 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2110 	} else /* off == sizeof(struct ip6_hdr) */ {
2111 		size_t l;
2112 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2113 		if (m->m_len < l) {
2114 			if ((m = m_pullup(m, l)) == NULL)
2115 				return;
2116 		}
2117 	}
2118 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2119 	ip6 = mtod(m, struct ip6_hdr *);
2120 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2121 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2122 	type = icmp6->icmp6_type; /* keep type for statistics */
2123 	code = icmp6->icmp6_code; /* ditto. */
2124 
2125 	origdst = ip6->ip6_dst;
2126 	/*
2127 	 * ip6_input() drops a packet if its src is multicast.
2128 	 * So, the src is never multicast.
2129 	 */
2130 	ip6->ip6_dst = ip6->ip6_src;
2131 
2132 	/*
2133 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2134 	 * use dst as the src for the reply.
2135 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2136 	 * (for example) when we encounter an error while forwarding procedure
2137 	 * destined to a duplicated address of ours.
2138 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2139 	 * procedure of an outgoing packet of our own, in which case we need
2140 	 * to search in the ifaddr list.
2141 	 */
2142 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2143 		if ((ia = ip6_getdstifaddr(m))) {
2144 			if (!(ia->ia6_flags &
2145 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2146 				srcp = &ia->ia_addr.sin6_addr;
2147 		} else {
2148 			struct sockaddr_in6 d;
2149 
2150 			bzero(&d, sizeof(d));
2151 			d.sin6_family = AF_INET6;
2152 			d.sin6_len = sizeof(d);
2153 			d.sin6_addr = origdst;
2154 			ia = (struct in6_ifaddr *)
2155 			    ifa_ifwithaddr((struct sockaddr *)&d);
2156 			if (ia &&
2157 			    !(ia->ia6_flags &
2158 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2159 				srcp = &ia->ia_addr.sin6_addr;
2160 			}
2161 		}
2162 	}
2163 
2164 	if ((srcp != NULL) &&
2165 	    (in6_addrscope(srcp) != in6_addrscope(&ip6->ip6_src)))
2166 		srcp = NULL;
2167 
2168 	if (srcp == NULL) {
2169 		int e;
2170 		struct sockaddr_in6 sin6;
2171 		struct route_in6 ro;
2172 
2173 		/*
2174 		 * This case matches to multicasts, our anycast, or unicasts
2175 		 * that we do not own.  Select a source address based on the
2176 		 * source address of the erroneous packet.
2177 		 */
2178 		bzero(&sin6, sizeof(sin6));
2179 		sin6.sin6_family = AF_INET6;
2180 		sin6.sin6_len = sizeof(sin6);
2181 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2182 
2183 		bzero(&ro, sizeof(ro));
2184 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2185 		if (ro.ro_rt)
2186 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2187 		if (e) {
2188 			char ip6buf[INET6_ADDRSTRLEN];
2189 			nd6log((LOG_DEBUG,
2190 			    "icmp6_reflect: source can't be determined: "
2191 			    "dst=%s, error=%d\n",
2192 			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2193 			goto bad;
2194 		}
2195 		srcp = &src;
2196 	}
2197 
2198 	ip6->ip6_src = *srcp;
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 = V_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 
2221 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2222 	if (outif)
2223 		icmp6_ifoutstat_inc(outif, type, code);
2224 
2225 	if (ia != NULL)
2226 		ifa_free(&ia->ia_ifa);
2227 	return;
2228 
2229  bad:
2230 	if (ia != NULL)
2231 		ifa_free(&ia->ia_ifa);
2232 	m_freem(m);
2233 	return;
2234 }
2235 
2236 void
2237 icmp6_fasttimo(void)
2238 {
2239 
2240 	mld_fasttimo();
2241 }
2242 
2243 void
2244 icmp6_slowtimo(void)
2245 {
2246 
2247 	mld_slowtimo();
2248 }
2249 
2250 static const char *
2251 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2252     struct in6_addr *tgt6)
2253 {
2254 	static char buf[1024];
2255 	char ip6bufs[INET6_ADDRSTRLEN];
2256 	char ip6bufd[INET6_ADDRSTRLEN];
2257 	char ip6buft[INET6_ADDRSTRLEN];
2258 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2259 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2260 	    ip6_sprintf(ip6buft, tgt6));
2261 	return buf;
2262 }
2263 
2264 void
2265 icmp6_redirect_input(struct mbuf *m, int off)
2266 {
2267 	struct ifnet *ifp;
2268 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2269 	struct nd_redirect *nd_rd;
2270 	int icmp6len = ntohs(ip6->ip6_plen);
2271 	char *lladdr = NULL;
2272 	int lladdrlen = 0;
2273 	u_char *redirhdr = NULL;
2274 	int redirhdrlen = 0;
2275 	struct rtentry *rt = NULL;
2276 	int is_router;
2277 	int is_onlink;
2278 	struct in6_addr src6 = ip6->ip6_src;
2279 	struct in6_addr redtgt6;
2280 	struct in6_addr reddst6;
2281 	union nd_opts ndopts;
2282 	char ip6buf[INET6_ADDRSTRLEN];
2283 
2284 	if (!m)
2285 		return;
2286 
2287 	ifp = m->m_pkthdr.rcvif;
2288 
2289 	if (!ifp)
2290 		return;
2291 
2292 	/* XXX if we are router, we don't update route by icmp6 redirect */
2293 	if (V_ip6_forwarding)
2294 		goto freeit;
2295 	if (!V_icmp6_rediraccept)
2296 		goto freeit;
2297 
2298 #ifndef PULLDOWN_TEST
2299 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2300 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2301 #else
2302 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2303 	if (nd_rd == NULL) {
2304 		ICMP6STAT_INC(icp6s_tooshort);
2305 		return;
2306 	}
2307 #endif
2308 	redtgt6 = nd_rd->nd_rd_target;
2309 	reddst6 = nd_rd->nd_rd_dst;
2310 
2311 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2312 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2313 		goto freeit;
2314 	}
2315 
2316 	/* validation */
2317 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2318 		nd6log((LOG_ERR,
2319 		    "ICMP6 redirect sent from %s rejected; "
2320 		    "must be from linklocal\n",
2321 		    ip6_sprintf(ip6buf, &src6)));
2322 		goto bad;
2323 	}
2324 	if (ip6->ip6_hlim != 255) {
2325 		nd6log((LOG_ERR,
2326 		    "ICMP6 redirect sent from %s rejected; "
2327 		    "hlim=%d (must be 255)\n",
2328 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2329 		goto bad;
2330 	}
2331     {
2332 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2333 	struct sockaddr_in6 sin6;
2334 	struct in6_addr *gw6;
2335 
2336 	bzero(&sin6, sizeof(sin6));
2337 	sin6.sin6_family = AF_INET6;
2338 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2339 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2340 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2341 	if (rt) {
2342 		if (rt->rt_gateway == NULL ||
2343 		    rt->rt_gateway->sa_family != AF_INET6) {
2344 			nd6log((LOG_ERR,
2345 			    "ICMP6 redirect rejected; no route "
2346 			    "with inet6 gateway found for redirect dst: %s\n",
2347 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2348 			RTFREE_LOCKED(rt);
2349 			goto bad;
2350 		}
2351 
2352 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2353 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2354 			nd6log((LOG_ERR,
2355 			    "ICMP6 redirect rejected; "
2356 			    "not equal to gw-for-src=%s (must be same): "
2357 			    "%s\n",
2358 			    ip6_sprintf(ip6buf, gw6),
2359 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2360 			RTFREE_LOCKED(rt);
2361 			goto bad;
2362 		}
2363 	} else {
2364 		nd6log((LOG_ERR,
2365 		    "ICMP6 redirect rejected; "
2366 		    "no route found for redirect dst: %s\n",
2367 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2368 		goto bad;
2369 	}
2370 	RTFREE_LOCKED(rt);
2371 	rt = NULL;
2372     }
2373 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2374 		nd6log((LOG_ERR,
2375 		    "ICMP6 redirect rejected; "
2376 		    "redirect dst must be unicast: %s\n",
2377 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2378 		goto bad;
2379 	}
2380 
2381 	is_router = is_onlink = 0;
2382 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2383 		is_router = 1;	/* router case */
2384 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2385 		is_onlink = 1;	/* on-link destination case */
2386 	if (!is_router && !is_onlink) {
2387 		nd6log((LOG_ERR,
2388 		    "ICMP6 redirect rejected; "
2389 		    "neither router case nor onlink case: %s\n",
2390 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2391 		goto bad;
2392 	}
2393 	/* validation passed */
2394 
2395 	icmp6len -= sizeof(*nd_rd);
2396 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2397 	if (nd6_options(&ndopts) < 0) {
2398 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2399 		    "invalid ND option, rejected: %s\n",
2400 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2401 		/* nd6_options have incremented stats */
2402 		goto freeit;
2403 	}
2404 
2405 	if (ndopts.nd_opts_tgt_lladdr) {
2406 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2407 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2408 	}
2409 
2410 	if (ndopts.nd_opts_rh) {
2411 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2412 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2413 	}
2414 
2415 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2416 		nd6log((LOG_INFO,
2417 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2418 		    "(if %d, icmp6 packet %d): %s\n",
2419 		    ip6_sprintf(ip6buf, &redtgt6),
2420 		    ifp->if_addrlen, lladdrlen - 2,
2421 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2422 		goto bad;
2423 	}
2424 
2425 	/* RFC 2461 8.3 */
2426 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2427 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2428 
2429 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2430 		/* perform rtredirect */
2431 		struct sockaddr_in6 sdst;
2432 		struct sockaddr_in6 sgw;
2433 		struct sockaddr_in6 ssrc;
2434 
2435 		bzero(&sdst, sizeof(sdst));
2436 		bzero(&sgw, sizeof(sgw));
2437 		bzero(&ssrc, sizeof(ssrc));
2438 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2439 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2440 			sizeof(struct sockaddr_in6);
2441 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2442 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2443 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2444 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2445 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2446 		    (struct sockaddr *)&ssrc);
2447 	}
2448 	/* finally update cached route in each socket via pfctlinput */
2449     {
2450 	struct sockaddr_in6 sdst;
2451 
2452 	bzero(&sdst, sizeof(sdst));
2453 	sdst.sin6_family = AF_INET6;
2454 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2455 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2456 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2457 #ifdef IPSEC
2458 	key_sa_routechange((struct sockaddr *)&sdst);
2459 #endif /* IPSEC */
2460     }
2461 
2462  freeit:
2463 	m_freem(m);
2464 	return;
2465 
2466  bad:
2467 	ICMP6STAT_INC(icp6s_badredirect);
2468 	m_freem(m);
2469 }
2470 
2471 void
2472 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2473 {
2474 	struct ifnet *ifp;	/* my outgoing interface */
2475 	struct in6_addr *ifp_ll6;
2476 	struct in6_addr *router_ll6;
2477 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2478 	struct mbuf *m = NULL;	/* newly allocated one */
2479 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2480 	struct nd_redirect *nd_rd;
2481 	struct llentry *ln = NULL;
2482 	size_t maxlen;
2483 	u_char *p;
2484 	struct ifnet *outif = NULL;
2485 	struct sockaddr_in6 src_sa;
2486 
2487 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2488 
2489 	/* if we are not router, we don't send icmp6 redirect */
2490 	if (!V_ip6_forwarding)
2491 		goto fail;
2492 
2493 	/* sanity check */
2494 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2495 		goto fail;
2496 
2497 	/*
2498 	 * Address check:
2499 	 *  the source address must identify a neighbor, and
2500 	 *  the destination address must not be a multicast address
2501 	 *  [RFC 2461, sec 8.2]
2502 	 */
2503 	sip6 = mtod(m0, struct ip6_hdr *);
2504 	bzero(&src_sa, sizeof(src_sa));
2505 	src_sa.sin6_family = AF_INET6;
2506 	src_sa.sin6_len = sizeof(src_sa);
2507 	src_sa.sin6_addr = sip6->ip6_src;
2508 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2509 		goto fail;
2510 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2511 		goto fail;	/* what should we do here? */
2512 
2513 	/* rate limit */
2514 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2515 		goto fail;
2516 
2517 	/*
2518 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2519 	 * we almost always ask for an mbuf cluster for simplicity.
2520 	 * (MHLEN < IPV6_MMTU is almost always true)
2521 	 */
2522 #if IPV6_MMTU >= MCLBYTES
2523 # error assumption failed about IPV6_MMTU and MCLBYTES
2524 #endif
2525 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2526 	if (m && IPV6_MMTU >= MHLEN)
2527 		MCLGET(m, M_DONTWAIT);
2528 	if (!m)
2529 		goto fail;
2530 	m->m_pkthdr.rcvif = NULL;
2531 	m->m_len = 0;
2532 	maxlen = M_TRAILINGSPACE(m);
2533 	maxlen = min(IPV6_MMTU, maxlen);
2534 	/* just for safety */
2535 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2536 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2537 		goto fail;
2538 	}
2539 
2540 	{
2541 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2542 		struct in6_ifaddr *ia;
2543 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2544 						 IN6_IFF_NOTREADY|
2545 						 IN6_IFF_ANYCAST)) == NULL)
2546 			goto fail;
2547 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2548 		/* XXXRW: reference released prematurely. */
2549 		ifa_free(&ia->ia_ifa);
2550 	}
2551 
2552 	/* get ip6 linklocal address for the router. */
2553 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2554 		struct sockaddr_in6 *sin6;
2555 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2556 		router_ll6 = &sin6->sin6_addr;
2557 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2558 			router_ll6 = (struct in6_addr *)NULL;
2559 	} else
2560 		router_ll6 = (struct in6_addr *)NULL;
2561 
2562 	/* ip6 */
2563 	ip6 = mtod(m, struct ip6_hdr *);
2564 	ip6->ip6_flow = 0;
2565 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2566 	ip6->ip6_vfc |= IPV6_VERSION;
2567 	/* ip6->ip6_plen will be set later */
2568 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2569 	ip6->ip6_hlim = 255;
2570 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2571 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2572 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2573 
2574 	/* ND Redirect */
2575 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2576 	nd_rd->nd_rd_type = ND_REDIRECT;
2577 	nd_rd->nd_rd_code = 0;
2578 	nd_rd->nd_rd_reserved = 0;
2579 	if (rt->rt_flags & RTF_GATEWAY) {
2580 		/*
2581 		 * nd_rd->nd_rd_target must be a link-local address in
2582 		 * better router cases.
2583 		 */
2584 		if (!router_ll6)
2585 			goto fail;
2586 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2587 		    sizeof(nd_rd->nd_rd_target));
2588 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2589 		    sizeof(nd_rd->nd_rd_dst));
2590 	} else {
2591 		/* make sure redtgt == reddst */
2592 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2593 		    sizeof(nd_rd->nd_rd_target));
2594 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2595 		    sizeof(nd_rd->nd_rd_dst));
2596 	}
2597 
2598 	p = (u_char *)(nd_rd + 1);
2599 
2600 	if (!router_ll6)
2601 		goto nolladdropt;
2602 
2603 	{
2604 		/* target lladdr option */
2605 		int len;
2606 		struct nd_opt_hdr *nd_opt;
2607 		char *lladdr;
2608 
2609 		IF_AFDATA_LOCK(ifp);
2610 		ln = nd6_lookup(router_ll6, 0, ifp);
2611 		IF_AFDATA_UNLOCK(ifp);
2612 		if (ln == NULL)
2613 			goto nolladdropt;
2614 
2615 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2616 		len = (len + 7) & ~7;	/* round by 8 */
2617 		/* safety check */
2618 		if (len + (p - (u_char *)ip6) > maxlen)
2619 			goto nolladdropt;
2620 
2621 		if (ln->la_flags & LLE_VALID) {
2622 			nd_opt = (struct nd_opt_hdr *)p;
2623 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2624 			nd_opt->nd_opt_len = len >> 3;
2625 			lladdr = (char *)(nd_opt + 1);
2626 			bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2627 			p += len;
2628 		}
2629 	}
2630 nolladdropt:
2631 	if (ln != NULL)
2632 		LLE_RUNLOCK(ln);
2633 
2634 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2635 
2636 	/* just to be safe */
2637 #ifdef M_DECRYPTED	/*not openbsd*/
2638 	if (m0->m_flags & M_DECRYPTED)
2639 		goto noredhdropt;
2640 #endif
2641 	if (p - (u_char *)ip6 > maxlen)
2642 		goto noredhdropt;
2643 
2644 	{
2645 		/* redirected header option */
2646 		int len;
2647 		struct nd_opt_rd_hdr *nd_opt_rh;
2648 
2649 		/*
2650 		 * compute the maximum size for icmp6 redirect header option.
2651 		 * XXX room for auth header?
2652 		 */
2653 		len = maxlen - (p - (u_char *)ip6);
2654 		len &= ~7;
2655 
2656 		/* This is just for simplicity. */
2657 		if (m0->m_pkthdr.len != m0->m_len) {
2658 			if (m0->m_next) {
2659 				m_freem(m0->m_next);
2660 				m0->m_next = NULL;
2661 			}
2662 			m0->m_pkthdr.len = m0->m_len;
2663 		}
2664 
2665 		/*
2666 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2667 		 * about padding/truncate rule for the original IP packet.
2668 		 * From the discussion on IPv6imp in Feb 1999,
2669 		 * the consensus was:
2670 		 * - "attach as much as possible" is the goal
2671 		 * - pad if not aligned (original size can be guessed by
2672 		 *   original ip6 header)
2673 		 * Following code adds the padding if it is simple enough,
2674 		 * and truncates if not.
2675 		 */
2676 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2677 			panic("assumption failed in %s:%d", __FILE__,
2678 			    __LINE__);
2679 
2680 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2681 			/* not enough room, truncate */
2682 			m0->m_pkthdr.len = m0->m_len = len -
2683 			    sizeof(*nd_opt_rh);
2684 		} else {
2685 			/* enough room, pad or truncate */
2686 			size_t extra;
2687 
2688 			extra = m0->m_pkthdr.len % 8;
2689 			if (extra) {
2690 				/* pad if easy enough, truncate if not */
2691 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2692 					/* pad */
2693 					m0->m_len += (8 - extra);
2694 					m0->m_pkthdr.len += (8 - extra);
2695 				} else {
2696 					/* truncate */
2697 					m0->m_pkthdr.len -= extra;
2698 					m0->m_len -= extra;
2699 				}
2700 			}
2701 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2702 			m0->m_pkthdr.len = m0->m_len = len -
2703 			    sizeof(*nd_opt_rh);
2704 		}
2705 
2706 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2707 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2708 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2709 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2710 		p += sizeof(*nd_opt_rh);
2711 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2712 
2713 		/* connect m0 to m */
2714 		m_tag_delete_chain(m0, NULL);
2715 		m0->m_flags &= ~M_PKTHDR;
2716 		m->m_next = m0;
2717 		m->m_pkthdr.len = m->m_len + m0->m_len;
2718 		m0 = NULL;
2719 	}
2720 noredhdropt:;
2721 	if (m0) {
2722 		m_freem(m0);
2723 		m0 = NULL;
2724 	}
2725 
2726 	/* XXX: clear embedded link IDs in the inner header */
2727 	in6_clearscope(&sip6->ip6_src);
2728 	in6_clearscope(&sip6->ip6_dst);
2729 	in6_clearscope(&nd_rd->nd_rd_target);
2730 	in6_clearscope(&nd_rd->nd_rd_dst);
2731 
2732 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2733 
2734 	nd_rd->nd_rd_cksum = 0;
2735 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2736 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2737 
2738 	/* send the packet to outside... */
2739 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2740 	if (outif) {
2741 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2742 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2743 	}
2744 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2745 
2746 	return;
2747 
2748 fail:
2749 	if (m)
2750 		m_freem(m);
2751 	if (m0)
2752 		m_freem(m0);
2753 }
2754 
2755 /*
2756  * ICMPv6 socket option processing.
2757  */
2758 int
2759 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2760 {
2761 	int error = 0;
2762 	int optlen;
2763 	struct inpcb *inp = sotoinpcb(so);
2764 	int level, op, optname;
2765 
2766 	if (sopt) {
2767 		level = sopt->sopt_level;
2768 		op = sopt->sopt_dir;
2769 		optname = sopt->sopt_name;
2770 		optlen = sopt->sopt_valsize;
2771 	} else
2772 		level = op = optname = optlen = 0;
2773 
2774 	if (level != IPPROTO_ICMPV6) {
2775 		return EINVAL;
2776 	}
2777 
2778 	switch (op) {
2779 	case PRCO_SETOPT:
2780 		switch (optname) {
2781 		case ICMP6_FILTER:
2782 		    {
2783 			struct icmp6_filter ic6f;
2784 
2785 			if (optlen != sizeof(ic6f)) {
2786 				error = EMSGSIZE;
2787 				break;
2788 			}
2789 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2790 			if (error == 0) {
2791 				INP_WLOCK(inp);
2792 				*inp->in6p_icmp6filt = ic6f;
2793 				INP_WUNLOCK(inp);
2794 			}
2795 			break;
2796 		    }
2797 
2798 		default:
2799 			error = ENOPROTOOPT;
2800 			break;
2801 		}
2802 		break;
2803 
2804 	case PRCO_GETOPT:
2805 		switch (optname) {
2806 		case ICMP6_FILTER:
2807 		    {
2808 			struct icmp6_filter ic6f;
2809 
2810 			INP_RLOCK(inp);
2811 			ic6f = *inp->in6p_icmp6filt;
2812 			INP_RUNLOCK(inp);
2813 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2814 			break;
2815 		    }
2816 
2817 		default:
2818 			error = ENOPROTOOPT;
2819 			break;
2820 		}
2821 		break;
2822 	}
2823 
2824 	return (error);
2825 }
2826 
2827 /*
2828  * Perform rate limit check.
2829  * Returns 0 if it is okay to send the icmp6 packet.
2830  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2831  * limitation.
2832  *
2833  * XXX per-destination/type check necessary?
2834  *
2835  * dst - not used at this moment
2836  * type - not used at this moment
2837  * code - not used at this moment
2838  */
2839 static int
2840 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2841     const int code)
2842 {
2843 	int ret;
2844 
2845 	ret = 0;	/* okay to send */
2846 
2847 	/* PPS limit */
2848 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2849 	    V_icmp6errppslim)) {
2850 		/* The packet is subject to rate limit */
2851 		ret++;
2852 	}
2853 
2854 	return ret;
2855 }
2856