xref: /freebsd/sys/netinet6/icmp6.c (revision eb69d1f144a6fcc765d1b9d44a5ae8082353e70b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
69 
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 
73 #include <sys/param.h>
74 #include <sys/domain.h>
75 #include <sys/jail.h>
76 #include <sys/kernel.h>
77 #include <sys/lock.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/signalvar.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sx.h>
86 #include <sys/syslog.h>
87 #include <sys/systm.h>
88 #include <sys/time.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/if_llatbl.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/in_var.h>
101 #include <netinet/ip6.h>
102 #include <netinet/icmp6.h>
103 #include <netinet/tcp_var.h>
104 
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/ip6protosw.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/send.h>
114 
115 extern struct domain inet6domain;
116 
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119 
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123 
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 VNET_DECLARE(struct inpcbhead, ripcb);
126 VNET_DECLARE(int, icmp6errppslim);
127 static VNET_DEFINE(int, icmp6errpps_count) = 0;
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130 
131 #define	V_ripcbinfo			VNET(ripcbinfo)
132 #define	V_ripcb				VNET(ripcb)
133 #define	V_icmp6errppslim		VNET(icmp6errppslim)
134 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
135 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
136 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
137 
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
141 static const char *icmp6_redirect_diag(struct in6_addr *,
142 	struct in6_addr *, struct in6_addr *);
143 static struct mbuf *ni6_input(struct mbuf *, int);
144 static struct mbuf *ni6_nametodns(const char *, int, int);
145 static int ni6_dnsmatch(const char *, int, const char *, int);
146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
147 			  struct ifnet **, struct in6_addr *);
148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
149 				struct ifnet *, int);
150 static int icmp6_notify_error(struct mbuf **, int, int, int);
151 
152 /*
153  * Kernel module interface for updating icmp6stat.  The argument is an index
154  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
155  * general layout of icmp6stat into the caller, it doesn't encode its
156  * location, so that future changes to add, for example, per-CPU stats
157  * support won't cause binary compatibility problems for kernel modules.
158  */
159 void
160 kmod_icmp6stat_inc(int statnum)
161 {
162 
163 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
164 }
165 
166 static void
167 icmp6_errcount(int type, int code)
168 {
169 	switch (type) {
170 	case ICMP6_DST_UNREACH:
171 		switch (code) {
172 		case ICMP6_DST_UNREACH_NOROUTE:
173 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
174 			return;
175 		case ICMP6_DST_UNREACH_ADMIN:
176 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
177 			return;
178 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
179 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
180 			return;
181 		case ICMP6_DST_UNREACH_ADDR:
182 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
183 			return;
184 		case ICMP6_DST_UNREACH_NOPORT:
185 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
186 			return;
187 		}
188 		break;
189 	case ICMP6_PACKET_TOO_BIG:
190 		ICMP6STAT_INC(icp6s_opacket_too_big);
191 		return;
192 	case ICMP6_TIME_EXCEEDED:
193 		switch (code) {
194 		case ICMP6_TIME_EXCEED_TRANSIT:
195 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
196 			return;
197 		case ICMP6_TIME_EXCEED_REASSEMBLY:
198 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
199 			return;
200 		}
201 		break;
202 	case ICMP6_PARAM_PROB:
203 		switch (code) {
204 		case ICMP6_PARAMPROB_HEADER:
205 			ICMP6STAT_INC(icp6s_oparamprob_header);
206 			return;
207 		case ICMP6_PARAMPROB_NEXTHEADER:
208 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
209 			return;
210 		case ICMP6_PARAMPROB_OPTION:
211 			ICMP6STAT_INC(icp6s_oparamprob_option);
212 			return;
213 		}
214 		break;
215 	case ND_REDIRECT:
216 		ICMP6STAT_INC(icp6s_oredirect);
217 		return;
218 	}
219 	ICMP6STAT_INC(icp6s_ounknown);
220 }
221 
222 /*
223  * A wrapper function for icmp6_error() necessary when the erroneous packet
224  * may not contain enough scope zone information.
225  */
226 void
227 icmp6_error2(struct mbuf *m, int type, int code, int param,
228     struct ifnet *ifp)
229 {
230 	struct ip6_hdr *ip6;
231 
232 	if (ifp == NULL)
233 		return;
234 
235 #ifndef PULLDOWN_TEST
236 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
237 #else
238 	if (m->m_len < sizeof(struct ip6_hdr)) {
239 		m = m_pullup(m, sizeof(struct ip6_hdr));
240 		if (m == NULL)
241 			return;
242 	}
243 #endif
244 
245 	ip6 = mtod(m, struct ip6_hdr *);
246 
247 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
248 		return;
249 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
250 		return;
251 
252 	icmp6_error(m, type, code, param);
253 }
254 
255 /*
256  * Generate an error packet of type error in response to bad IP6 packet.
257  */
258 void
259 icmp6_error(struct mbuf *m, int type, int code, int param)
260 {
261 	struct ip6_hdr *oip6, *nip6;
262 	struct icmp6_hdr *icmp6;
263 	u_int preplen;
264 	int off;
265 	int nxt;
266 
267 	ICMP6STAT_INC(icp6s_error);
268 
269 	/* count per-type-code statistics */
270 	icmp6_errcount(type, code);
271 
272 #ifdef M_DECRYPTED	/*not openbsd*/
273 	if (m->m_flags & M_DECRYPTED) {
274 		ICMP6STAT_INC(icp6s_canterror);
275 		goto freeit;
276 	}
277 #endif
278 
279 #ifndef PULLDOWN_TEST
280 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
281 #else
282 	if (m->m_len < sizeof(struct ip6_hdr)) {
283 		m = m_pullup(m, sizeof(struct ip6_hdr));
284 		if (m == NULL)
285 			return;
286 	}
287 #endif
288 	oip6 = mtod(m, struct ip6_hdr *);
289 
290 	/*
291 	 * If the destination address of the erroneous packet is a multicast
292 	 * address, or the packet was sent using link-layer multicast,
293 	 * we should basically suppress sending an error (RFC 2463, Section
294 	 * 2.4).
295 	 * We have two exceptions (the item e.2 in that section):
296 	 * - the Packet Too Big message can be sent for path MTU discovery.
297 	 * - the Parameter Problem Message that can be allowed an icmp6 error
298 	 *   in the option type field.  This check has been done in
299 	 *   ip6_unknown_opt(), so we can just check the type and code.
300 	 */
301 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
302 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
303 	    (type != ICMP6_PACKET_TOO_BIG &&
304 	     (type != ICMP6_PARAM_PROB ||
305 	      code != ICMP6_PARAMPROB_OPTION)))
306 		goto freeit;
307 
308 	/*
309 	 * RFC 2463, 2.4 (e.5): source address check.
310 	 * XXX: the case of anycast source?
311 	 */
312 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
313 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
314 		goto freeit;
315 
316 	/*
317 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
318 	 * don't do it.
319 	 */
320 	nxt = -1;
321 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
322 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
323 		struct icmp6_hdr *icp;
324 
325 #ifndef PULLDOWN_TEST
326 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
327 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
328 #else
329 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
330 			sizeof(*icp));
331 		if (icp == NULL) {
332 			ICMP6STAT_INC(icp6s_tooshort);
333 			return;
334 		}
335 #endif
336 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
337 		    icp->icmp6_type == ND_REDIRECT) {
338 			/*
339 			 * ICMPv6 error
340 			 * Special case: for redirect (which is
341 			 * informational) we must not send icmp6 error.
342 			 */
343 			ICMP6STAT_INC(icp6s_canterror);
344 			goto freeit;
345 		} else {
346 			/* ICMPv6 informational - send the error */
347 		}
348 	} else {
349 		/* non-ICMPv6 - send the error */
350 	}
351 
352 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
353 
354 	/* Finally, do rate limitation check. */
355 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
356 		ICMP6STAT_INC(icp6s_toofreq);
357 		goto freeit;
358 	}
359 
360 	/*
361 	 * OK, ICMP6 can be generated.
362 	 */
363 
364 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
365 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
366 
367 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
368 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
369 	if (m == NULL) {
370 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
371 		return;
372 	}
373 
374 	nip6 = mtod(m, struct ip6_hdr *);
375 	nip6->ip6_src  = oip6->ip6_src;
376 	nip6->ip6_dst  = oip6->ip6_dst;
377 
378 	in6_clearscope(&oip6->ip6_src);
379 	in6_clearscope(&oip6->ip6_dst);
380 
381 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
382 	icmp6->icmp6_type = type;
383 	icmp6->icmp6_code = code;
384 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
385 
386 	/*
387 	 * icmp6_reflect() is designed to be in the input path.
388 	 * icmp6_error() can be called from both input and output path,
389 	 * and if we are in output path rcvif could contain bogus value.
390 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
391 	 * information in ip header (nip6).
392 	 */
393 	m->m_pkthdr.rcvif = NULL;
394 
395 	ICMP6STAT_INC(icp6s_outhist[type]);
396 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
397 
398 	return;
399 
400   freeit:
401 	/*
402 	 * If we can't tell whether or not we can generate ICMP6, free it.
403 	 */
404 	m_freem(m);
405 }
406 
407 /*
408  * Process a received ICMP6 message.
409  */
410 int
411 icmp6_input(struct mbuf **mp, int *offp, int proto)
412 {
413 	struct mbuf *m = *mp, *n;
414 	struct ifnet *ifp;
415 	struct ip6_hdr *ip6, *nip6;
416 	struct icmp6_hdr *icmp6, *nicmp6;
417 	int off = *offp;
418 	int icmp6len = m->m_pkthdr.len - *offp;
419 	int code, sum, noff;
420 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
421 	int ip6len, error;
422 
423 	ifp = m->m_pkthdr.rcvif;
424 
425 #ifndef PULLDOWN_TEST
426 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
427 	/* m might change if M_LOOP.  So, call mtod after this */
428 #endif
429 
430 	/*
431 	 * Locate icmp6 structure in mbuf, and check
432 	 * that not corrupted and of at least minimum length
433 	 */
434 
435 	ip6 = mtod(m, struct ip6_hdr *);
436 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
437 	if (icmp6len < sizeof(struct icmp6_hdr)) {
438 		ICMP6STAT_INC(icp6s_tooshort);
439 		goto freeit;
440 	}
441 
442 	/*
443 	 * Check multicast group membership.
444 	 * Note: SSM filters are not applied for ICMPv6 traffic.
445 	 */
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
447 		struct in6_multi	*inm;
448 
449 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
450 		if (inm == NULL) {
451 			IP6STAT_INC(ip6s_notmember);
452 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
453 			goto freeit;
454 		}
455 	}
456 
457 	/*
458 	 * calculate the checksum
459 	 */
460 #ifndef PULLDOWN_TEST
461 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
462 #else
463 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
464 	if (icmp6 == NULL) {
465 		ICMP6STAT_INC(icp6s_tooshort);
466 		return IPPROTO_DONE;
467 	}
468 #endif
469 	code = icmp6->icmp6_code;
470 
471 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
472 		nd6log((LOG_ERR,
473 		    "ICMP6 checksum error(%d|%x) %s\n",
474 		    icmp6->icmp6_type, sum,
475 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
476 		ICMP6STAT_INC(icp6s_checksum);
477 		goto freeit;
478 	}
479 
480 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
481 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
482 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
483 		icmp6_ifstat_inc(ifp, ifs6_in_error);
484 
485 	switch (icmp6->icmp6_type) {
486 	case ICMP6_DST_UNREACH:
487 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
488 		switch (code) {
489 		case ICMP6_DST_UNREACH_NOROUTE:
490 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
491 			code = PRC_UNREACH_NET;
492 			break;
493 		case ICMP6_DST_UNREACH_ADMIN:
494 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
495 			code = PRC_UNREACH_ADMIN_PROHIB;
496 			break;
497 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
498 			/* I mean "source address was incorrect." */
499 			code = PRC_PARAMPROB;
500 			break;
501 		case ICMP6_DST_UNREACH_NOPORT:
502 			code = PRC_UNREACH_PORT;
503 			break;
504 		default:
505 			goto badcode;
506 		}
507 		goto deliver;
508 		break;
509 
510 	case ICMP6_PACKET_TOO_BIG:
511 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
512 
513 		/* validation is made in icmp6_mtudisc_update */
514 
515 		code = PRC_MSGSIZE;
516 
517 		/*
518 		 * Updating the path MTU will be done after examining
519 		 * intermediate extension headers.
520 		 */
521 		goto deliver;
522 		break;
523 
524 	case ICMP6_TIME_EXCEEDED:
525 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
526 		switch (code) {
527 		case ICMP6_TIME_EXCEED_TRANSIT:
528 			code = PRC_TIMXCEED_INTRANS;
529 			break;
530 		case ICMP6_TIME_EXCEED_REASSEMBLY:
531 			code = PRC_TIMXCEED_REASS;
532 			break;
533 		default:
534 			goto badcode;
535 		}
536 		goto deliver;
537 		break;
538 
539 	case ICMP6_PARAM_PROB:
540 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
541 		switch (code) {
542 		case ICMP6_PARAMPROB_NEXTHEADER:
543 			code = PRC_UNREACH_PROTOCOL;
544 			break;
545 		case ICMP6_PARAMPROB_HEADER:
546 		case ICMP6_PARAMPROB_OPTION:
547 			code = PRC_PARAMPROB;
548 			break;
549 		default:
550 			goto badcode;
551 		}
552 		goto deliver;
553 		break;
554 
555 	case ICMP6_ECHO_REQUEST:
556 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
557 		if (code != 0)
558 			goto badcode;
559 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
560 			/* Give up remote */
561 			break;
562 		}
563 		if (!M_WRITABLE(n)
564 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
565 			struct mbuf *n0 = n;
566 			int n0len;
567 
568 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
569 			n = m_gethdr(M_NOWAIT, n0->m_type);
570 			if (n == NULL) {
571 				/* Give up remote */
572 				m_freem(n0);
573 				break;
574 			}
575 
576 			m_move_pkthdr(n, n0);	/* FIB copied. */
577 			n0len = n0->m_pkthdr.len;	/* save for use below */
578 			/*
579 			 * Copy IPv6 and ICMPv6 only.
580 			 */
581 			nip6 = mtod(n, struct ip6_hdr *);
582 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
583 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
584 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
585 			noff = sizeof(struct ip6_hdr);
586 			/* new mbuf contains only ipv6+icmpv6 headers */
587 			n->m_len = noff + sizeof(struct icmp6_hdr);
588 			/*
589 			 * Adjust mbuf.  ip6_plen will be adjusted in
590 			 * ip6_output().
591 			 */
592 			m_adj(n0, off + sizeof(struct icmp6_hdr));
593 			/* recalculate complete packet size */
594 			n->m_pkthdr.len = n0len + (noff - off);
595 			n->m_next = n0;
596 		} else {
597 			nip6 = mtod(n, struct ip6_hdr *);
598 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
599 			    sizeof(*nicmp6));
600 			noff = off;
601 		}
602 		if (n) {
603 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
604 			nicmp6->icmp6_code = 0;
605 			ICMP6STAT_INC(icp6s_reflect);
606 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
607 			icmp6_reflect(n, noff);
608 		}
609 		break;
610 
611 	case ICMP6_ECHO_REPLY:
612 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
613 		if (code != 0)
614 			goto badcode;
615 		break;
616 
617 	case MLD_LISTENER_QUERY:
618 	case MLD_LISTENER_REPORT:
619 	case MLD_LISTENER_DONE:
620 	case MLDV2_LISTENER_REPORT:
621 		/*
622 		 * Drop MLD traffic which is not link-local, has a hop limit
623 		 * of greater than 1 hop, or which does not have the
624 		 * IPv6 HBH Router Alert option.
625 		 * As IPv6 HBH options are stripped in ip6_input() we must
626 		 * check an mbuf header flag.
627 		 * XXX Should we also sanity check that these messages
628 		 * were directed to a link-local multicast prefix?
629 		 */
630 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
631 			goto freeit;
632 		if (mld_input(m, off, icmp6len) != 0)
633 			return (IPPROTO_DONE);
634 		/* m stays. */
635 		break;
636 
637 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
638 	    {
639 		enum { WRU, FQDN } mode;
640 
641 		if (!V_icmp6_nodeinfo)
642 			break;
643 
644 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
645 			mode = WRU;
646 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
647 			mode = FQDN;
648 		else
649 			goto badlen;
650 
651 		if (mode == FQDN) {
652 #ifndef PULLDOWN_TEST
653 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
654 			    IPPROTO_DONE);
655 #endif
656 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
657 			if (n)
658 				n = ni6_input(n, off);
659 			/* XXX meaningless if n == NULL */
660 			noff = sizeof(struct ip6_hdr);
661 		} else {
662 			struct prison *pr;
663 			u_char *p;
664 			int maxhlen, hlen;
665 
666 			/*
667 			 * XXX: this combination of flags is pointless,
668 			 * but should we keep this for compatibility?
669 			 */
670 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
671 			    ICMP6_NODEINFO_TMPADDROK)) !=
672 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
673 				break;
674 
675 			if (code != 0)
676 				goto badcode;
677 
678 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
679 			n = m_gethdr(M_NOWAIT, m->m_type);
680 			if (n == NULL) {
681 				/* Give up remote */
682 				break;
683 			}
684 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
685 				/*
686 				 * Previous code did a blind M_COPY_PKTHDR
687 				 * and said "just for rcvif".  If true, then
688 				 * we could tolerate the dup failing (due to
689 				 * the deep copy of the tag chain).  For now
690 				 * be conservative and just fail.
691 				 */
692 				m_free(n);
693 				n = NULL;
694 				break;
695 			}
696 			maxhlen = M_TRAILINGSPACE(n) -
697 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
698 			pr = curthread->td_ucred->cr_prison;
699 			mtx_lock(&pr->pr_mtx);
700 			hlen = strlen(pr->pr_hostname);
701 			if (maxhlen > hlen)
702 				maxhlen = hlen;
703 			/*
704 			 * Copy IPv6 and ICMPv6 only.
705 			 */
706 			nip6 = mtod(n, struct ip6_hdr *);
707 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
708 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
709 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
710 			p = (u_char *)(nicmp6 + 1);
711 			bzero(p, 4);
712 			/* meaningless TTL */
713 			bcopy(pr->pr_hostname, p + 4, maxhlen);
714 			mtx_unlock(&pr->pr_mtx);
715 			noff = sizeof(struct ip6_hdr);
716 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
717 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
718 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
719 			nicmp6->icmp6_code = 0;
720 		}
721 		if (n) {
722 			ICMP6STAT_INC(icp6s_reflect);
723 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
724 			icmp6_reflect(n, noff);
725 		}
726 		break;
727 	    }
728 
729 	case ICMP6_WRUREPLY:
730 		if (code != 0)
731 			goto badcode;
732 		break;
733 
734 	case ND_ROUTER_SOLICIT:
735 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
736 		if (code != 0)
737 			goto badcode;
738 		if (icmp6len < sizeof(struct nd_router_solicit))
739 			goto badlen;
740 		if (send_sendso_input_hook != NULL) {
741 			IP6_EXTHDR_CHECK(m, off, icmp6len, IPPROTO_DONE);
742 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
743 			if (error == 0) {
744 				m = NULL;
745 				goto freeit;
746 			}
747 		}
748 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
749 		nd6_rs_input(m, off, icmp6len);
750 		m = n;
751 		if (m == NULL)
752 			goto freeit;
753 		break;
754 
755 	case ND_ROUTER_ADVERT:
756 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
757 		if (code != 0)
758 			goto badcode;
759 		if (icmp6len < sizeof(struct nd_router_advert))
760 			goto badlen;
761 		if (send_sendso_input_hook != NULL) {
762 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
763 			if (error == 0) {
764 				m = NULL;
765 				goto freeit;
766 			}
767 		}
768 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
769 		nd6_ra_input(m, off, icmp6len);
770 		m = n;
771 		if (m == NULL)
772 			goto freeit;
773 		break;
774 
775 	case ND_NEIGHBOR_SOLICIT:
776 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
777 		if (code != 0)
778 			goto badcode;
779 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
780 			goto badlen;
781 		if (send_sendso_input_hook != NULL) {
782 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
783 			if (error == 0) {
784 				m = NULL;
785 				goto freeit;
786 			}
787 		}
788 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
789 		nd6_ns_input(m, off, icmp6len);
790 		m = n;
791 		if (m == NULL)
792 			goto freeit;
793 		break;
794 
795 	case ND_NEIGHBOR_ADVERT:
796 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
797 		if (code != 0)
798 			goto badcode;
799 		if (icmp6len < sizeof(struct nd_neighbor_advert))
800 			goto badlen;
801 		if (send_sendso_input_hook != NULL) {
802 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
803 			if (error == 0) {
804 				m = NULL;
805 				goto freeit;
806 			}
807 		}
808 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
809 		nd6_na_input(m, off, icmp6len);
810 		m = n;
811 		if (m == NULL)
812 			goto freeit;
813 		break;
814 
815 	case ND_REDIRECT:
816 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
817 		if (code != 0)
818 			goto badcode;
819 		if (icmp6len < sizeof(struct nd_redirect))
820 			goto badlen;
821 		if (send_sendso_input_hook != NULL) {
822 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
823 			if (error == 0) {
824 				m = NULL;
825 				goto freeit;
826 			}
827 		}
828 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
829 		icmp6_redirect_input(m, off);
830 		m = n;
831 		if (m == NULL)
832 			goto freeit;
833 		break;
834 
835 	case ICMP6_ROUTER_RENUMBERING:
836 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
837 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
838 			goto badcode;
839 		if (icmp6len < sizeof(struct icmp6_router_renum))
840 			goto badlen;
841 		break;
842 
843 	default:
844 		nd6log((LOG_DEBUG,
845 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
846 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
847 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
848 		    ifp ? ifp->if_index : 0));
849 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
850 			/* ICMPv6 error: MUST deliver it by spec... */
851 			code = PRC_NCMDS;
852 			/* deliver */
853 		} else {
854 			/* ICMPv6 informational: MUST not deliver */
855 			break;
856 		}
857 	deliver:
858 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
859 			/* In this case, m should've been freed. */
860 			return (IPPROTO_DONE);
861 		}
862 		break;
863 
864 	badcode:
865 		ICMP6STAT_INC(icp6s_badcode);
866 		break;
867 
868 	badlen:
869 		ICMP6STAT_INC(icp6s_badlen);
870 		break;
871 	}
872 
873 	/* deliver the packet to appropriate sockets */
874 	icmp6_rip6_input(&m, *offp);
875 
876 	return IPPROTO_DONE;
877 
878  freeit:
879 	m_freem(m);
880 	return IPPROTO_DONE;
881 }
882 
883 static int
884 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
885 {
886 	struct mbuf *m = *mp;
887 	struct icmp6_hdr *icmp6;
888 	struct ip6_hdr *eip6;
889 	u_int32_t notifymtu;
890 	struct sockaddr_in6 icmp6src, icmp6dst;
891 
892 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
893 		ICMP6STAT_INC(icp6s_tooshort);
894 		goto freeit;
895 	}
896 #ifndef PULLDOWN_TEST
897 	IP6_EXTHDR_CHECK(m, off,
898 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
899 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
900 #else
901 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
902 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
903 	if (icmp6 == NULL) {
904 		ICMP6STAT_INC(icp6s_tooshort);
905 		return (-1);
906 	}
907 #endif
908 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
909 
910 	/* Detect the upper level protocol */
911 	{
912 		void (*ctlfunc)(int, struct sockaddr *, void *);
913 		u_int8_t nxt = eip6->ip6_nxt;
914 		int eoff = off + sizeof(struct icmp6_hdr) +
915 		    sizeof(struct ip6_hdr);
916 		struct ip6ctlparam ip6cp;
917 		struct in6_addr *finaldst = NULL;
918 		int icmp6type = icmp6->icmp6_type;
919 		struct ip6_frag *fh;
920 		struct ip6_rthdr *rth;
921 		struct ip6_rthdr0 *rth0;
922 		int rthlen;
923 
924 		while (1) { /* XXX: should avoid infinite loop explicitly? */
925 			struct ip6_ext *eh;
926 
927 			switch (nxt) {
928 			case IPPROTO_HOPOPTS:
929 			case IPPROTO_DSTOPTS:
930 			case IPPROTO_AH:
931 #ifndef PULLDOWN_TEST
932 				IP6_EXTHDR_CHECK(m, 0,
933 				    eoff + sizeof(struct ip6_ext), -1);
934 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
935 #else
936 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
937 				    eoff, sizeof(*eh));
938 				if (eh == NULL) {
939 					ICMP6STAT_INC(icp6s_tooshort);
940 					return (-1);
941 				}
942 #endif
943 
944 				if (nxt == IPPROTO_AH)
945 					eoff += (eh->ip6e_len + 2) << 2;
946 				else
947 					eoff += (eh->ip6e_len + 1) << 3;
948 				nxt = eh->ip6e_nxt;
949 				break;
950 			case IPPROTO_ROUTING:
951 				/*
952 				 * When the erroneous packet contains a
953 				 * routing header, we should examine the
954 				 * header to determine the final destination.
955 				 * Otherwise, we can't properly update
956 				 * information that depends on the final
957 				 * destination (e.g. path MTU).
958 				 */
959 #ifndef PULLDOWN_TEST
960 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
961 				rth = (struct ip6_rthdr *)
962 				    (mtod(m, caddr_t) + eoff);
963 #else
964 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
965 				    eoff, sizeof(*rth));
966 				if (rth == NULL) {
967 					ICMP6STAT_INC(icp6s_tooshort);
968 					return (-1);
969 				}
970 #endif
971 				rthlen = (rth->ip6r_len + 1) << 3;
972 				/*
973 				 * XXX: currently there is no
974 				 * officially defined type other
975 				 * than type-0.
976 				 * Note that if the segment left field
977 				 * is 0, all intermediate hops must
978 				 * have been passed.
979 				 */
980 				if (rth->ip6r_segleft &&
981 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
982 					int hops;
983 
984 #ifndef PULLDOWN_TEST
985 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
986 					rth0 = (struct ip6_rthdr0 *)
987 					    (mtod(m, caddr_t) + eoff);
988 #else
989 					IP6_EXTHDR_GET(rth0,
990 					    struct ip6_rthdr0 *, m,
991 					    eoff, rthlen);
992 					if (rth0 == NULL) {
993 						ICMP6STAT_INC(icp6s_tooshort);
994 						return (-1);
995 					}
996 #endif
997 					/* just ignore a bogus header */
998 					if ((rth0->ip6r0_len % 2) == 0 &&
999 					    (hops = rth0->ip6r0_len/2))
1000 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1001 				}
1002 				eoff += rthlen;
1003 				nxt = rth->ip6r_nxt;
1004 				break;
1005 			case IPPROTO_FRAGMENT:
1006 #ifndef PULLDOWN_TEST
1007 				IP6_EXTHDR_CHECK(m, 0, eoff +
1008 				    sizeof(struct ip6_frag), -1);
1009 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1010 				    eoff);
1011 #else
1012 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1013 				    eoff, sizeof(*fh));
1014 				if (fh == NULL) {
1015 					ICMP6STAT_INC(icp6s_tooshort);
1016 					return (-1);
1017 				}
1018 #endif
1019 				/*
1020 				 * Data after a fragment header is meaningless
1021 				 * unless it is the first fragment, but
1022 				 * we'll go to the notify label for path MTU
1023 				 * discovery.
1024 				 */
1025 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1026 					goto notify;
1027 
1028 				eoff += sizeof(struct ip6_frag);
1029 				nxt = fh->ip6f_nxt;
1030 				break;
1031 			default:
1032 				/*
1033 				 * This case includes ESP and the No Next
1034 				 * Header.  In such cases going to the notify
1035 				 * label does not have any meaning
1036 				 * (i.e. ctlfunc will be NULL), but we go
1037 				 * anyway since we might have to update
1038 				 * path MTU information.
1039 				 */
1040 				goto notify;
1041 			}
1042 		}
1043 	  notify:
1044 #ifndef PULLDOWN_TEST
1045 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1046 #else
1047 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1048 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1049 		if (icmp6 == NULL) {
1050 			ICMP6STAT_INC(icp6s_tooshort);
1051 			return (-1);
1052 		}
1053 #endif
1054 
1055 		/*
1056 		 * retrieve parameters from the inner IPv6 header, and convert
1057 		 * them into sockaddr structures.
1058 		 * XXX: there is no guarantee that the source or destination
1059 		 * addresses of the inner packet are in the same scope as
1060 		 * the addresses of the icmp packet.  But there is no other
1061 		 * way to determine the zone.
1062 		 */
1063 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1064 
1065 		bzero(&icmp6dst, sizeof(icmp6dst));
1066 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1067 		icmp6dst.sin6_family = AF_INET6;
1068 		if (finaldst == NULL)
1069 			icmp6dst.sin6_addr = eip6->ip6_dst;
1070 		else
1071 			icmp6dst.sin6_addr = *finaldst;
1072 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1073 			goto freeit;
1074 		bzero(&icmp6src, sizeof(icmp6src));
1075 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1076 		icmp6src.sin6_family = AF_INET6;
1077 		icmp6src.sin6_addr = eip6->ip6_src;
1078 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1079 			goto freeit;
1080 		icmp6src.sin6_flowinfo =
1081 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1082 
1083 		if (finaldst == NULL)
1084 			finaldst = &eip6->ip6_dst;
1085 		ip6cp.ip6c_m = m;
1086 		ip6cp.ip6c_icmp6 = icmp6;
1087 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1088 		ip6cp.ip6c_off = eoff;
1089 		ip6cp.ip6c_finaldst = finaldst;
1090 		ip6cp.ip6c_src = &icmp6src;
1091 		ip6cp.ip6c_nxt = nxt;
1092 
1093 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1094 			notifymtu = ntohl(icmp6->icmp6_mtu);
1095 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1096 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1097 		}
1098 
1099 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1100 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1101 		if (ctlfunc) {
1102 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1103 			    &ip6cp);
1104 		}
1105 	}
1106 	*mp = m;
1107 	return (0);
1108 
1109   freeit:
1110 	m_freem(m);
1111 	return (-1);
1112 }
1113 
1114 void
1115 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1116 {
1117 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1118 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1119 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1120 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1121 	struct in_conninfo inc;
1122 
1123 #if 0
1124 	/*
1125 	 * RFC2460 section 5, last paragraph.
1126 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1127 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1128 	 * due to packet translator in the middle.
1129 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1130 	 * special handling.
1131 	 */
1132 	if (mtu < IPV6_MMTU)
1133 		return;
1134 #endif
1135 
1136 	/*
1137 	 * we reject ICMPv6 too big with abnormally small value.
1138 	 * XXX what is the good definition of "abnormally small"?
1139 	 */
1140 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1141 		return;
1142 
1143 	if (!validated)
1144 		return;
1145 
1146 	/*
1147 	 * In case the suggested mtu is less than IPV6_MMTU, we
1148 	 * only need to remember that it was for above mentioned
1149 	 * "alwaysfrag" case.
1150 	 * Try to be as close to the spec as possible.
1151 	 */
1152 	if (mtu < IPV6_MMTU)
1153 		mtu = IPV6_MMTU - 8;
1154 
1155 	bzero(&inc, sizeof(inc));
1156 	inc.inc_fibnum = M_GETFIB(m);
1157 	inc.inc_flags |= INC_ISIPV6;
1158 	inc.inc6_faddr = *dst;
1159 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1160 		return;
1161 
1162 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1163 		tcp_hc_updatemtu(&inc, mtu);
1164 		ICMP6STAT_INC(icp6s_pmtuchg);
1165 	}
1166 }
1167 
1168 /*
1169  * Process a Node Information Query packet, based on
1170  * draft-ietf-ipngwg-icmp-name-lookups-07.
1171  *
1172  * Spec incompatibilities:
1173  * - IPv6 Subject address handling
1174  * - IPv4 Subject address handling support missing
1175  * - Proxy reply (answer even if it's not for me)
1176  * - joins NI group address at in6_ifattach() time only, does not cope
1177  *   with hostname changes by sethostname(3)
1178  */
1179 static struct mbuf *
1180 ni6_input(struct mbuf *m, int off)
1181 {
1182 	struct icmp6_nodeinfo *ni6, *nni6;
1183 	struct mbuf *n = NULL;
1184 	struct prison *pr;
1185 	u_int16_t qtype;
1186 	int subjlen;
1187 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1188 	struct ni_reply_fqdn *fqdn;
1189 	int addrs;		/* for NI_QTYPE_NODEADDR */
1190 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1191 	struct in6_addr in6_subj; /* subject address */
1192 	struct ip6_hdr *ip6;
1193 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1194 	char *subj = NULL;
1195 	struct in6_ifaddr *ia6 = NULL;
1196 
1197 	ip6 = mtod(m, struct ip6_hdr *);
1198 #ifndef PULLDOWN_TEST
1199 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1200 #else
1201 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1202 	if (ni6 == NULL) {
1203 		/* m is already reclaimed */
1204 		return (NULL);
1205 	}
1206 #endif
1207 
1208 	/*
1209 	 * Validate IPv6 source address.
1210 	 * The default configuration MUST be to refuse answering queries from
1211 	 * global-scope addresses according to RFC4602.
1212 	 * Notes:
1213 	 *  - it's not very clear what "refuse" means; this implementation
1214 	 *    simply drops it.
1215 	 *  - it's not very easy to identify global-scope (unicast) addresses
1216 	 *    since there are many prefixes for them.  It should be safer
1217 	 *    and in practice sufficient to check "all" but loopback and
1218 	 *    link-local (note that site-local unicast was deprecated and
1219 	 *    ULA is defined as global scope-wise)
1220 	 */
1221 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1222 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1223 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1224 		goto bad;
1225 
1226 	/*
1227 	 * Validate IPv6 destination address.
1228 	 *
1229 	 * The Responder must discard the Query without further processing
1230 	 * unless it is one of the Responder's unicast or anycast addresses, or
1231 	 * a link-local scope multicast address which the Responder has joined.
1232 	 * [RFC4602, Section 5.]
1233 	 */
1234 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1235 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1236 			goto bad;
1237 		/* else it's a link-local multicast, fine */
1238 	} else {		/* unicast or anycast */
1239 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1240 		if (ia6 == NULL)
1241 			goto bad; /* XXX impossible */
1242 
1243 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1244 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1245 			ifa_free(&ia6->ia_ifa);
1246 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1247 				"a temporary address in %s:%d",
1248 			       __FILE__, __LINE__));
1249 			goto bad;
1250 		}
1251 		ifa_free(&ia6->ia_ifa);
1252 	}
1253 
1254 	/* validate query Subject field. */
1255 	qtype = ntohs(ni6->ni_qtype);
1256 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1257 	switch (qtype) {
1258 	case NI_QTYPE_NOOP:
1259 	case NI_QTYPE_SUPTYPES:
1260 		/* 07 draft */
1261 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1262 			break;
1263 		/* FALLTHROUGH */
1264 	case NI_QTYPE_FQDN:
1265 	case NI_QTYPE_NODEADDR:
1266 	case NI_QTYPE_IPV4ADDR:
1267 		switch (ni6->ni_code) {
1268 		case ICMP6_NI_SUBJ_IPV6:
1269 #if ICMP6_NI_SUBJ_IPV6 != 0
1270 		case 0:
1271 #endif
1272 			/*
1273 			 * backward compatibility - try to accept 03 draft
1274 			 * format, where no Subject is present.
1275 			 */
1276 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1277 			    subjlen == 0) {
1278 				oldfqdn++;
1279 				break;
1280 			}
1281 #if ICMP6_NI_SUBJ_IPV6 != 0
1282 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1283 				goto bad;
1284 #endif
1285 
1286 			if (subjlen != sizeof(struct in6_addr))
1287 				goto bad;
1288 
1289 			/*
1290 			 * Validate Subject address.
1291 			 *
1292 			 * Not sure what exactly "address belongs to the node"
1293 			 * means in the spec, is it just unicast, or what?
1294 			 *
1295 			 * At this moment we consider Subject address as
1296 			 * "belong to the node" if the Subject address equals
1297 			 * to the IPv6 destination address; validation for
1298 			 * IPv6 destination address should have done enough
1299 			 * check for us.
1300 			 *
1301 			 * We do not do proxy at this moment.
1302 			 */
1303 			/* m_pulldown instead of copy? */
1304 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1305 			    subjlen, (caddr_t)&in6_subj);
1306 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1307 				goto bad;
1308 
1309 			subj = (char *)&in6_subj;
1310 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1311 				break;
1312 
1313 			/*
1314 			 * XXX if we are to allow other cases, we should really
1315 			 * be careful about scope here.
1316 			 * basically, we should disallow queries toward IPv6
1317 			 * destination X with subject Y,
1318 			 * if scope(X) > scope(Y).
1319 			 * if we allow scope(X) > scope(Y), it will result in
1320 			 * information leakage across scope boundary.
1321 			 */
1322 			goto bad;
1323 
1324 		case ICMP6_NI_SUBJ_FQDN:
1325 			/*
1326 			 * Validate Subject name with gethostname(3).
1327 			 *
1328 			 * The behavior may need some debate, since:
1329 			 * - we are not sure if the node has FQDN as
1330 			 *   hostname (returned by gethostname(3)).
1331 			 * - the code does wildcard match for truncated names.
1332 			 *   however, we are not sure if we want to perform
1333 			 *   wildcard match, if gethostname(3) side has
1334 			 *   truncated hostname.
1335 			 */
1336 			pr = curthread->td_ucred->cr_prison;
1337 			mtx_lock(&pr->pr_mtx);
1338 			n = ni6_nametodns(pr->pr_hostname,
1339 			    strlen(pr->pr_hostname), 0);
1340 			mtx_unlock(&pr->pr_mtx);
1341 			if (!n || n->m_next || n->m_len == 0)
1342 				goto bad;
1343 			IP6_EXTHDR_GET(subj, char *, m,
1344 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1345 			if (subj == NULL)
1346 				goto bad;
1347 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1348 			    n->m_len)) {
1349 				goto bad;
1350 			}
1351 			m_freem(n);
1352 			n = NULL;
1353 			break;
1354 
1355 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1356 		default:
1357 			goto bad;
1358 		}
1359 		break;
1360 	}
1361 
1362 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1363 	switch (qtype) {
1364 	case NI_QTYPE_FQDN:
1365 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1366 			goto bad;
1367 		break;
1368 	case NI_QTYPE_NODEADDR:
1369 	case NI_QTYPE_IPV4ADDR:
1370 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1371 			goto bad;
1372 		break;
1373 	}
1374 
1375 	/* guess reply length */
1376 	switch (qtype) {
1377 	case NI_QTYPE_NOOP:
1378 		break;		/* no reply data */
1379 	case NI_QTYPE_SUPTYPES:
1380 		replylen += sizeof(u_int32_t);
1381 		break;
1382 	case NI_QTYPE_FQDN:
1383 		/* XXX will append an mbuf */
1384 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1385 		break;
1386 	case NI_QTYPE_NODEADDR:
1387 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1388 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1389 		    sizeof(u_int32_t))) > MCLBYTES)
1390 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1391 		break;
1392 	case NI_QTYPE_IPV4ADDR:
1393 		/* unsupported - should respond with unknown Qtype? */
1394 		break;
1395 	default:
1396 		/*
1397 		 * XXX: We must return a reply with the ICMP6 code
1398 		 * `unknown Qtype' in this case.  However we regard the case
1399 		 * as an FQDN query for backward compatibility.
1400 		 * Older versions set a random value to this field,
1401 		 * so it rarely varies in the defined qtypes.
1402 		 * But the mechanism is not reliable...
1403 		 * maybe we should obsolete older versions.
1404 		 */
1405 		qtype = NI_QTYPE_FQDN;
1406 		/* XXX will append an mbuf */
1407 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1408 		oldfqdn++;
1409 		break;
1410 	}
1411 
1412 	/* Allocate an mbuf to reply. */
1413 	if (replylen > MCLBYTES) {
1414 		/*
1415 		 * XXX: should we try to allocate more? But MCLBYTES
1416 		 * is probably much larger than IPV6_MMTU...
1417 		 */
1418 		goto bad;
1419 	}
1420 	if (replylen > MHLEN)
1421 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1422 	else
1423 		n = m_gethdr(M_NOWAIT, m->m_type);
1424 	if (n == NULL) {
1425 		m_freem(m);
1426 		return (NULL);
1427 	}
1428 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1429 	n->m_pkthdr.len = n->m_len = replylen;
1430 
1431 	/* copy mbuf header and IPv6 + Node Information base headers */
1432 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1433 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1434 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1435 
1436 	/* qtype dependent procedure */
1437 	switch (qtype) {
1438 	case NI_QTYPE_NOOP:
1439 		nni6->ni_code = ICMP6_NI_SUCCESS;
1440 		nni6->ni_flags = 0;
1441 		break;
1442 	case NI_QTYPE_SUPTYPES:
1443 	{
1444 		u_int32_t v;
1445 		nni6->ni_code = ICMP6_NI_SUCCESS;
1446 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1447 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1448 		v = (u_int32_t)htonl(0x0000000f);
1449 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1450 		break;
1451 	}
1452 	case NI_QTYPE_FQDN:
1453 		nni6->ni_code = ICMP6_NI_SUCCESS;
1454 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1455 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1456 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1457 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1458 		/*
1459 		 * XXX do we really have FQDN in hostname?
1460 		 */
1461 		pr = curthread->td_ucred->cr_prison;
1462 		mtx_lock(&pr->pr_mtx);
1463 		n->m_next = ni6_nametodns(pr->pr_hostname,
1464 		    strlen(pr->pr_hostname), oldfqdn);
1465 		mtx_unlock(&pr->pr_mtx);
1466 		if (n->m_next == NULL)
1467 			goto bad;
1468 		/* XXX we assume that n->m_next is not a chain */
1469 		if (n->m_next->m_next != NULL)
1470 			goto bad;
1471 		n->m_pkthdr.len += n->m_next->m_len;
1472 		break;
1473 	case NI_QTYPE_NODEADDR:
1474 	{
1475 		int lenlim, copied;
1476 
1477 		nni6->ni_code = ICMP6_NI_SUCCESS;
1478 		n->m_pkthdr.len = n->m_len =
1479 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1480 		lenlim = M_TRAILINGSPACE(n);
1481 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1482 		/* XXX: reset mbuf length */
1483 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1484 		    sizeof(struct icmp6_nodeinfo) + copied;
1485 		break;
1486 	}
1487 	default:
1488 		break;		/* XXX impossible! */
1489 	}
1490 
1491 	nni6->ni_type = ICMP6_NI_REPLY;
1492 	m_freem(m);
1493 	return (n);
1494 
1495   bad:
1496 	m_freem(m);
1497 	if (n)
1498 		m_freem(n);
1499 	return (NULL);
1500 }
1501 
1502 /*
1503  * make a mbuf with DNS-encoded string.  no compression support.
1504  *
1505  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1506  * treated as truncated name (two \0 at the end).  this is a wild guess.
1507  *
1508  * old - return pascal string if non-zero
1509  */
1510 static struct mbuf *
1511 ni6_nametodns(const char *name, int namelen, int old)
1512 {
1513 	struct mbuf *m;
1514 	char *cp, *ep;
1515 	const char *p, *q;
1516 	int i, len, nterm;
1517 
1518 	if (old)
1519 		len = namelen + 1;
1520 	else
1521 		len = MCLBYTES;
1522 
1523 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1524 	if (len > MLEN)
1525 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1526 	else
1527 		m = m_get(M_NOWAIT, MT_DATA);
1528 	if (m == NULL)
1529 		goto fail;
1530 
1531 	if (old) {
1532 		m->m_len = len;
1533 		*mtod(m, char *) = namelen;
1534 		bcopy(name, mtod(m, char *) + 1, namelen);
1535 		return m;
1536 	} else {
1537 		m->m_len = 0;
1538 		cp = mtod(m, char *);
1539 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1540 
1541 		/* if not certain about my name, return empty buffer */
1542 		if (namelen == 0)
1543 			return m;
1544 
1545 		/*
1546 		 * guess if it looks like shortened hostname, or FQDN.
1547 		 * shortened hostname needs two trailing "\0".
1548 		 */
1549 		i = 0;
1550 		for (p = name; p < name + namelen; p++) {
1551 			if (*p && *p == '.')
1552 				i++;
1553 		}
1554 		if (i < 2)
1555 			nterm = 2;
1556 		else
1557 			nterm = 1;
1558 
1559 		p = name;
1560 		while (cp < ep && p < name + namelen) {
1561 			i = 0;
1562 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1563 				i++;
1564 			/* result does not fit into mbuf */
1565 			if (cp + i + 1 >= ep)
1566 				goto fail;
1567 			/*
1568 			 * DNS label length restriction, RFC1035 page 8.
1569 			 * "i == 0" case is included here to avoid returning
1570 			 * 0-length label on "foo..bar".
1571 			 */
1572 			if (i <= 0 || i >= 64)
1573 				goto fail;
1574 			*cp++ = i;
1575 			bcopy(p, cp, i);
1576 			cp += i;
1577 			p = q;
1578 			if (p < name + namelen && *p == '.')
1579 				p++;
1580 		}
1581 		/* termination */
1582 		if (cp + nterm >= ep)
1583 			goto fail;
1584 		while (nterm-- > 0)
1585 			*cp++ = '\0';
1586 		m->m_len = cp - mtod(m, char *);
1587 		return m;
1588 	}
1589 
1590 	panic("should not reach here");
1591 	/* NOTREACHED */
1592 
1593  fail:
1594 	if (m)
1595 		m_freem(m);
1596 	return NULL;
1597 }
1598 
1599 /*
1600  * check if two DNS-encoded string matches.  takes care of truncated
1601  * form (with \0\0 at the end).  no compression support.
1602  * XXX upper/lowercase match (see RFC2065)
1603  */
1604 static int
1605 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1606 {
1607 	const char *a0, *b0;
1608 	int l;
1609 
1610 	/* simplest case - need validation? */
1611 	if (alen == blen && bcmp(a, b, alen) == 0)
1612 		return 1;
1613 
1614 	a0 = a;
1615 	b0 = b;
1616 
1617 	/* termination is mandatory */
1618 	if (alen < 2 || blen < 2)
1619 		return 0;
1620 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1621 		return 0;
1622 	alen--;
1623 	blen--;
1624 
1625 	while (a - a0 < alen && b - b0 < blen) {
1626 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1627 			return 0;
1628 
1629 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1630 			return 0;
1631 		/* we don't support compression yet */
1632 		if (a[0] >= 64 || b[0] >= 64)
1633 			return 0;
1634 
1635 		/* truncated case */
1636 		if (a[0] == 0 && a - a0 == alen - 1)
1637 			return 1;
1638 		if (b[0] == 0 && b - b0 == blen - 1)
1639 			return 1;
1640 		if (a[0] == 0 || b[0] == 0)
1641 			return 0;
1642 
1643 		if (a[0] != b[0])
1644 			return 0;
1645 		l = a[0];
1646 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1647 			return 0;
1648 		if (bcmp(a + 1, b + 1, l) != 0)
1649 			return 0;
1650 
1651 		a += 1 + l;
1652 		b += 1 + l;
1653 	}
1654 
1655 	if (a - a0 == alen && b - b0 == blen)
1656 		return 1;
1657 	else
1658 		return 0;
1659 }
1660 
1661 /*
1662  * calculate the number of addresses to be returned in the node info reply.
1663  */
1664 static int
1665 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1666     struct in6_addr *subj)
1667 {
1668 	struct ifnet *ifp;
1669 	struct in6_ifaddr *ifa6;
1670 	struct ifaddr *ifa;
1671 	int addrs = 0, addrsofif, iffound = 0;
1672 	int niflags = ni6->ni_flags;
1673 
1674 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1675 		switch (ni6->ni_code) {
1676 		case ICMP6_NI_SUBJ_IPV6:
1677 			if (subj == NULL) /* must be impossible... */
1678 				return (0);
1679 			break;
1680 		default:
1681 			/*
1682 			 * XXX: we only support IPv6 subject address for
1683 			 * this Qtype.
1684 			 */
1685 			return (0);
1686 		}
1687 	}
1688 
1689 	IFNET_RLOCK_NOSLEEP();
1690 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1691 		addrsofif = 0;
1692 		IF_ADDR_RLOCK(ifp);
1693 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1694 			if (ifa->ifa_addr->sa_family != AF_INET6)
1695 				continue;
1696 			ifa6 = (struct in6_ifaddr *)ifa;
1697 
1698 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1699 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1700 				iffound = 1;
1701 
1702 			/*
1703 			 * IPv4-mapped addresses can only be returned by a
1704 			 * Node Information proxy, since they represent
1705 			 * addresses of IPv4-only nodes, which perforce do
1706 			 * not implement this protocol.
1707 			 * [icmp-name-lookups-07, Section 5.4]
1708 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1709 			 * this function at this moment.
1710 			 */
1711 
1712 			/* What do we have to do about ::1? */
1713 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1714 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1715 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1716 					continue;
1717 				break;
1718 			case IPV6_ADDR_SCOPE_SITELOCAL:
1719 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1720 					continue;
1721 				break;
1722 			case IPV6_ADDR_SCOPE_GLOBAL:
1723 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1724 					continue;
1725 				break;
1726 			default:
1727 				continue;
1728 			}
1729 
1730 			/*
1731 			 * check if anycast is okay.
1732 			 * XXX: just experimental.  not in the spec.
1733 			 */
1734 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1735 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1736 				continue; /* we need only unicast addresses */
1737 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1738 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1739 				continue;
1740 			}
1741 			addrsofif++; /* count the address */
1742 		}
1743 		IF_ADDR_RUNLOCK(ifp);
1744 		if (iffound) {
1745 			*ifpp = ifp;
1746 			IFNET_RUNLOCK_NOSLEEP();
1747 			return (addrsofif);
1748 		}
1749 
1750 		addrs += addrsofif;
1751 	}
1752 	IFNET_RUNLOCK_NOSLEEP();
1753 
1754 	return (addrs);
1755 }
1756 
1757 static int
1758 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1759     struct ifnet *ifp0, int resid)
1760 {
1761 	struct ifnet *ifp;
1762 	struct in6_ifaddr *ifa6;
1763 	struct ifaddr *ifa;
1764 	struct ifnet *ifp_dep = NULL;
1765 	int copied = 0, allow_deprecated = 0;
1766 	u_char *cp = (u_char *)(nni6 + 1);
1767 	int niflags = ni6->ni_flags;
1768 	u_int32_t ltime;
1769 
1770 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1771 		return (0);	/* needless to copy */
1772 
1773 	IFNET_RLOCK_NOSLEEP();
1774 	ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1775   again:
1776 
1777 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_link)) {
1778 		IF_ADDR_RLOCK(ifp);
1779 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1780 			if (ifa->ifa_addr->sa_family != AF_INET6)
1781 				continue;
1782 			ifa6 = (struct in6_ifaddr *)ifa;
1783 
1784 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1785 			    allow_deprecated == 0) {
1786 				/*
1787 				 * prefererred address should be put before
1788 				 * deprecated addresses.
1789 				 */
1790 
1791 				/* record the interface for later search */
1792 				if (ifp_dep == NULL)
1793 					ifp_dep = ifp;
1794 
1795 				continue;
1796 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1797 			    allow_deprecated != 0)
1798 				continue; /* we now collect deprecated addrs */
1799 
1800 			/* What do we have to do about ::1? */
1801 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1802 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1803 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1804 					continue;
1805 				break;
1806 			case IPV6_ADDR_SCOPE_SITELOCAL:
1807 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1808 					continue;
1809 				break;
1810 			case IPV6_ADDR_SCOPE_GLOBAL:
1811 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1812 					continue;
1813 				break;
1814 			default:
1815 				continue;
1816 			}
1817 
1818 			/*
1819 			 * check if anycast is okay.
1820 			 * XXX: just experimental.  not in the spec.
1821 			 */
1822 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1823 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1824 				continue;
1825 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1826 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1827 				continue;
1828 			}
1829 
1830 			/* now we can copy the address */
1831 			if (resid < sizeof(struct in6_addr) +
1832 			    sizeof(u_int32_t)) {
1833 				IF_ADDR_RUNLOCK(ifp);
1834 				/*
1835 				 * We give up much more copy.
1836 				 * Set the truncate flag and return.
1837 				 */
1838 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1839 				IFNET_RUNLOCK_NOSLEEP();
1840 				return (copied);
1841 			}
1842 
1843 			/*
1844 			 * Set the TTL of the address.
1845 			 * The TTL value should be one of the following
1846 			 * according to the specification:
1847 			 *
1848 			 * 1. The remaining lifetime of a DHCP lease on the
1849 			 *    address, or
1850 			 * 2. The remaining Valid Lifetime of a prefix from
1851 			 *    which the address was derived through Stateless
1852 			 *    Autoconfiguration.
1853 			 *
1854 			 * Note that we currently do not support stateful
1855 			 * address configuration by DHCPv6, so the former
1856 			 * case can't happen.
1857 			 */
1858 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1859 				ltime = ND6_INFINITE_LIFETIME;
1860 			else {
1861 				if (ifa6->ia6_lifetime.ia6t_expire >
1862 				    time_uptime)
1863 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1864 				else
1865 					ltime = 0;
1866 			}
1867 
1868 			bcopy(&ltime, cp, sizeof(u_int32_t));
1869 			cp += sizeof(u_int32_t);
1870 
1871 			/* copy the address itself */
1872 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1873 			    sizeof(struct in6_addr));
1874 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1875 			cp += sizeof(struct in6_addr);
1876 
1877 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1878 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1879 		}
1880 		IF_ADDR_RUNLOCK(ifp);
1881 		if (ifp0)	/* we need search only on the specified IF */
1882 			break;
1883 	}
1884 
1885 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1886 		ifp = ifp_dep;
1887 		allow_deprecated = 1;
1888 
1889 		goto again;
1890 	}
1891 
1892 	IFNET_RUNLOCK_NOSLEEP();
1893 
1894 	return (copied);
1895 }
1896 
1897 /*
1898  * XXX almost dup'ed code with rip6_input.
1899  */
1900 static int
1901 icmp6_rip6_input(struct mbuf **mp, int off)
1902 {
1903 	struct mbuf *m = *mp;
1904 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1905 	struct inpcb *in6p;
1906 	struct inpcb *last = NULL;
1907 	struct sockaddr_in6 fromsa;
1908 	struct icmp6_hdr *icmp6;
1909 	struct mbuf *opts = NULL;
1910 
1911 #ifndef PULLDOWN_TEST
1912 	/* this is assumed to be safe. */
1913 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1914 #else
1915 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1916 	if (icmp6 == NULL) {
1917 		/* m is already reclaimed */
1918 		return (IPPROTO_DONE);
1919 	}
1920 #endif
1921 
1922 	/*
1923 	 * XXX: the address may have embedded scope zone ID, which should be
1924 	 * hidden from applications.
1925 	 */
1926 	bzero(&fromsa, sizeof(fromsa));
1927 	fromsa.sin6_family = AF_INET6;
1928 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1929 	fromsa.sin6_addr = ip6->ip6_src;
1930 	if (sa6_recoverscope(&fromsa)) {
1931 		m_freem(m);
1932 		return (IPPROTO_DONE);
1933 	}
1934 
1935 	INP_INFO_RLOCK(&V_ripcbinfo);
1936 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1937 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1938 			continue;
1939 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1940 			continue;
1941 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1942 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1943 			continue;
1944 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1945 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1946 			continue;
1947 		INP_RLOCK(in6p);
1948 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1949 		    in6p->in6p_icmp6filt)) {
1950 			INP_RUNLOCK(in6p);
1951 			continue;
1952 		}
1953 		if (last != NULL) {
1954 			struct	mbuf *n = NULL;
1955 
1956 			/*
1957 			 * Recent network drivers tend to allocate a single
1958 			 * mbuf cluster, rather than to make a couple of
1959 			 * mbufs without clusters.  Also, since the IPv6 code
1960 			 * path tries to avoid m_pullup(), it is highly
1961 			 * probable that we still have an mbuf cluster here
1962 			 * even though the necessary length can be stored in an
1963 			 * mbuf's internal buffer.
1964 			 * Meanwhile, the default size of the receive socket
1965 			 * buffer for raw sockets is not so large.  This means
1966 			 * the possibility of packet loss is relatively higher
1967 			 * than before.  To avoid this scenario, we copy the
1968 			 * received data to a separate mbuf that does not use
1969 			 * a cluster, if possible.
1970 			 * XXX: it is better to copy the data after stripping
1971 			 * intermediate headers.
1972 			 */
1973 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1974 			    m->m_len <= MHLEN) {
1975 				n = m_get(M_NOWAIT, m->m_type);
1976 				if (n != NULL) {
1977 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1978 						bcopy(m->m_data, n->m_data,
1979 						      m->m_len);
1980 						n->m_len = m->m_len;
1981 					} else {
1982 						m_free(n);
1983 						n = NULL;
1984 					}
1985 				}
1986 			}
1987 			if (n != NULL ||
1988 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1989 				if (last->inp_flags & INP_CONTROLOPTS)
1990 					ip6_savecontrol(last, n, &opts);
1991 				/* strip intermediate headers */
1992 				m_adj(n, off);
1993 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1994 				if (sbappendaddr_locked(
1995 				    &last->inp_socket->so_rcv,
1996 				    (struct sockaddr *)&fromsa, n, opts)
1997 				    == 0) {
1998 					/* should notify about lost packet */
1999 					m_freem(n);
2000 					if (opts) {
2001 						m_freem(opts);
2002 					}
2003 					SOCKBUF_UNLOCK(
2004 					    &last->inp_socket->so_rcv);
2005 				} else
2006 					sorwakeup_locked(last->inp_socket);
2007 				opts = NULL;
2008 			}
2009 			INP_RUNLOCK(last);
2010 		}
2011 		last = in6p;
2012 	}
2013 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2014 	if (last != NULL) {
2015 		if (last->inp_flags & INP_CONTROLOPTS)
2016 			ip6_savecontrol(last, m, &opts);
2017 		/* strip intermediate headers */
2018 		m_adj(m, off);
2019 
2020 		/* avoid using mbuf clusters if possible (see above) */
2021 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2022 		    m->m_len <= MHLEN) {
2023 			struct mbuf *n;
2024 
2025 			n = m_get(M_NOWAIT, m->m_type);
2026 			if (n != NULL) {
2027 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2028 					bcopy(m->m_data, n->m_data, m->m_len);
2029 					n->m_len = m->m_len;
2030 
2031 					m_freem(m);
2032 					m = n;
2033 				} else {
2034 					m_freem(n);
2035 					n = NULL;
2036 				}
2037 			}
2038 		}
2039 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2040 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2041 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2042 			m_freem(m);
2043 			if (opts)
2044 				m_freem(opts);
2045 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2046 		} else
2047 			sorwakeup_locked(last->inp_socket);
2048 		INP_RUNLOCK(last);
2049 	} else {
2050 		m_freem(m);
2051 		IP6STAT_DEC(ip6s_delivered);
2052 	}
2053 	return IPPROTO_DONE;
2054 }
2055 
2056 /*
2057  * Reflect the ip6 packet back to the source.
2058  * OFF points to the icmp6 header, counted from the top of the mbuf.
2059  */
2060 void
2061 icmp6_reflect(struct mbuf *m, size_t off)
2062 {
2063 	struct in6_addr src6, *srcp;
2064 	struct ip6_hdr *ip6;
2065 	struct icmp6_hdr *icmp6;
2066 	struct in6_ifaddr *ia = NULL;
2067 	struct ifnet *outif = NULL;
2068 	int plen;
2069 	int type, code, hlim;
2070 
2071 	/* too short to reflect */
2072 	if (off < sizeof(struct ip6_hdr)) {
2073 		nd6log((LOG_DEBUG,
2074 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2075 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2076 		    __FILE__, __LINE__));
2077 		goto bad;
2078 	}
2079 
2080 	/*
2081 	 * If there are extra headers between IPv6 and ICMPv6, strip
2082 	 * off that header first.
2083 	 */
2084 #ifdef DIAGNOSTIC
2085 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2086 		panic("assumption failed in icmp6_reflect");
2087 #endif
2088 	if (off > sizeof(struct ip6_hdr)) {
2089 		size_t l;
2090 		struct ip6_hdr nip6;
2091 
2092 		l = off - sizeof(struct ip6_hdr);
2093 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2094 		m_adj(m, l);
2095 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2096 		if (m->m_len < l) {
2097 			if ((m = m_pullup(m, l)) == NULL)
2098 				return;
2099 		}
2100 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2101 	} else /* off == sizeof(struct ip6_hdr) */ {
2102 		size_t l;
2103 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2104 		if (m->m_len < l) {
2105 			if ((m = m_pullup(m, l)) == NULL)
2106 				return;
2107 		}
2108 	}
2109 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2110 	ip6 = mtod(m, struct ip6_hdr *);
2111 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2112 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2113 	type = icmp6->icmp6_type; /* keep type for statistics */
2114 	code = icmp6->icmp6_code; /* ditto. */
2115 	hlim = 0;
2116 	srcp = NULL;
2117 
2118 	/*
2119 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2120 	 * use dst as the src for the reply.
2121 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2122 	 * (for example) when we encounter an error while forwarding procedure
2123 	 * destined to a duplicated address of ours.
2124 	 */
2125 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2126 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2127 		if (ia != NULL && !(ia->ia6_flags &
2128 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2129 			src6 = ia->ia_addr.sin6_addr;
2130 			srcp = &src6;
2131 
2132 			if (m->m_pkthdr.rcvif != NULL) {
2133 				/* XXX: This may not be the outgoing interface */
2134 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2135 			} else
2136 				hlim = V_ip6_defhlim;
2137 		}
2138 		if (ia != NULL)
2139 			ifa_free(&ia->ia_ifa);
2140 	}
2141 
2142 	if (srcp == NULL) {
2143 		int error;
2144 		struct in6_addr dst6;
2145 		uint32_t scopeid;
2146 
2147 		/*
2148 		 * This case matches to multicasts, our anycast, or unicasts
2149 		 * that we do not own.  Select a source address based on the
2150 		 * source address of the erroneous packet.
2151 		 */
2152 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2153 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2154 		    scopeid, NULL, &src6, &hlim);
2155 
2156 		if (error) {
2157 			char ip6buf[INET6_ADDRSTRLEN];
2158 			nd6log((LOG_DEBUG,
2159 			    "icmp6_reflect: source can't be determined: "
2160 			    "dst=%s, error=%d\n",
2161 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2162 			goto bad;
2163 		}
2164 		srcp = &src6;
2165 	}
2166 	/*
2167 	 * ip6_input() drops a packet if its src is multicast.
2168 	 * So, the src is never multicast.
2169 	 */
2170 	ip6->ip6_dst = ip6->ip6_src;
2171 	ip6->ip6_src = *srcp;
2172 	ip6->ip6_flow = 0;
2173 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2174 	ip6->ip6_vfc |= IPV6_VERSION;
2175 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2176 	ip6->ip6_hlim = hlim;
2177 
2178 	icmp6->icmp6_cksum = 0;
2179 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2180 	    sizeof(struct ip6_hdr), plen);
2181 
2182 	/*
2183 	 * XXX option handling
2184 	 */
2185 
2186 	m->m_flags &= ~(M_BCAST|M_MCAST);
2187 
2188 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2189 	if (outif)
2190 		icmp6_ifoutstat_inc(outif, type, code);
2191 
2192 	return;
2193 
2194  bad:
2195 	m_freem(m);
2196 	return;
2197 }
2198 
2199 void
2200 icmp6_fasttimo(void)
2201 {
2202 
2203 	mld_fasttimo();
2204 }
2205 
2206 void
2207 icmp6_slowtimo(void)
2208 {
2209 
2210 	mld_slowtimo();
2211 }
2212 
2213 static const char *
2214 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2215     struct in6_addr *tgt6)
2216 {
2217 	static char buf[1024];
2218 	char ip6bufs[INET6_ADDRSTRLEN];
2219 	char ip6bufd[INET6_ADDRSTRLEN];
2220 	char ip6buft[INET6_ADDRSTRLEN];
2221 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2222 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2223 	    ip6_sprintf(ip6buft, tgt6));
2224 	return buf;
2225 }
2226 
2227 void
2228 icmp6_redirect_input(struct mbuf *m, int off)
2229 {
2230 	struct ifnet *ifp;
2231 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2232 	struct nd_redirect *nd_rd;
2233 	int icmp6len = ntohs(ip6->ip6_plen);
2234 	char *lladdr = NULL;
2235 	int lladdrlen = 0;
2236 	int is_router;
2237 	int is_onlink;
2238 	struct in6_addr src6 = ip6->ip6_src;
2239 	struct in6_addr redtgt6;
2240 	struct in6_addr reddst6;
2241 	union nd_opts ndopts;
2242 	char ip6buf[INET6_ADDRSTRLEN];
2243 
2244 	M_ASSERTPKTHDR(m);
2245 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2246 
2247 	ifp = m->m_pkthdr.rcvif;
2248 
2249 	/* XXX if we are router, we don't update route by icmp6 redirect */
2250 	if (V_ip6_forwarding)
2251 		goto freeit;
2252 	if (!V_icmp6_rediraccept)
2253 		goto freeit;
2254 
2255 	/* RFC 6980: Nodes MUST silently ignore fragments */
2256 	if(m->m_flags & M_FRAGMENTED)
2257 		goto freeit;
2258 
2259 #ifndef PULLDOWN_TEST
2260 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2261 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2262 #else
2263 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2264 	if (nd_rd == NULL) {
2265 		ICMP6STAT_INC(icp6s_tooshort);
2266 		return;
2267 	}
2268 #endif
2269 	redtgt6 = nd_rd->nd_rd_target;
2270 	reddst6 = nd_rd->nd_rd_dst;
2271 
2272 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2273 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2274 		goto freeit;
2275 	}
2276 
2277 	/* validation */
2278 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2279 		nd6log((LOG_ERR,
2280 		    "ICMP6 redirect sent from %s rejected; "
2281 		    "must be from linklocal\n",
2282 		    ip6_sprintf(ip6buf, &src6)));
2283 		goto bad;
2284 	}
2285 	if (ip6->ip6_hlim != 255) {
2286 		nd6log((LOG_ERR,
2287 		    "ICMP6 redirect sent from %s rejected; "
2288 		    "hlim=%d (must be 255)\n",
2289 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2290 		goto bad;
2291 	}
2292     {
2293 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2294 	struct nhop6_basic nh6;
2295 	struct in6_addr kdst;
2296 	uint32_t scopeid;
2297 
2298 	in6_splitscope(&reddst6, &kdst, &scopeid);
2299 	if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2300 		if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2301 			nd6log((LOG_ERR,
2302 			    "ICMP6 redirect rejected; no route "
2303 			    "with inet6 gateway found for redirect dst: %s\n",
2304 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2305 			goto bad;
2306 		}
2307 
2308 		/*
2309 		 * Embed scope zone id into next hop address, since
2310 		 * fib6_lookup_nh_basic() returns address without embedded
2311 		 * scope zone id.
2312 		 */
2313 		if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL))
2314 			goto freeit;
2315 
2316 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2317 			nd6log((LOG_ERR,
2318 			    "ICMP6 redirect rejected; "
2319 			    "not equal to gw-for-src=%s (must be same): "
2320 			    "%s\n",
2321 			    ip6_sprintf(ip6buf, &nh6.nh_addr),
2322 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2323 			goto bad;
2324 		}
2325 	} else {
2326 		nd6log((LOG_ERR,
2327 		    "ICMP6 redirect rejected; "
2328 		    "no route found for redirect dst: %s\n",
2329 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2330 		goto bad;
2331 	}
2332     }
2333 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2334 		nd6log((LOG_ERR,
2335 		    "ICMP6 redirect rejected; "
2336 		    "redirect dst must be unicast: %s\n",
2337 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2338 		goto bad;
2339 	}
2340 
2341 	is_router = is_onlink = 0;
2342 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2343 		is_router = 1;	/* router case */
2344 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2345 		is_onlink = 1;	/* on-link destination case */
2346 	if (!is_router && !is_onlink) {
2347 		nd6log((LOG_ERR,
2348 		    "ICMP6 redirect rejected; "
2349 		    "neither router case nor onlink case: %s\n",
2350 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2351 		goto bad;
2352 	}
2353 
2354 	icmp6len -= sizeof(*nd_rd);
2355 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2356 	if (nd6_options(&ndopts) < 0) {
2357 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2358 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2359 		/* nd6_options have incremented stats */
2360 		goto freeit;
2361 	}
2362 
2363 	if (ndopts.nd_opts_tgt_lladdr) {
2364 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2365 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2366 	}
2367 
2368 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2369 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2370 		    "(if %d, icmp6 packet %d): %s\n",
2371 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2372 		    ifp->if_addrlen, lladdrlen - 2,
2373 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2374 		goto bad;
2375 	}
2376 
2377 	/* Validation passed. */
2378 
2379 	/* RFC 2461 8.3 */
2380 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2381 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2382 
2383 	/*
2384 	 * Install a gateway route in the better-router case or an interface
2385 	 * route in the on-link-destination case.
2386 	 */
2387 	{
2388 		struct sockaddr_in6 sdst;
2389 		struct sockaddr_in6 sgw;
2390 		struct sockaddr_in6 ssrc;
2391 		struct sockaddr *gw;
2392 		int rt_flags;
2393 		u_int fibnum;
2394 
2395 		bzero(&sdst, sizeof(sdst));
2396 		bzero(&ssrc, sizeof(ssrc));
2397 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2398 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2399 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2400 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2401 		rt_flags = RTF_HOST;
2402 		if (is_router) {
2403 			bzero(&sgw, sizeof(sgw));
2404 			sgw.sin6_family = AF_INET6;
2405 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2406 			bcopy(&redtgt6, &sgw.sin6_addr,
2407 				sizeof(struct in6_addr));
2408 			gw = (struct sockaddr *)&sgw;
2409 			rt_flags |= RTF_GATEWAY;
2410 		} else
2411 			gw = ifp->if_addr->ifa_addr;
2412 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2413 			in6_rtredirect((struct sockaddr *)&sdst, gw,
2414 			    (struct sockaddr *)NULL, rt_flags,
2415 			    (struct sockaddr *)&ssrc, fibnum);
2416 	}
2417 	/* finally update cached route in each socket via pfctlinput */
2418     {
2419 	struct sockaddr_in6 sdst;
2420 
2421 	bzero(&sdst, sizeof(sdst));
2422 	sdst.sin6_family = AF_INET6;
2423 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2424 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2425 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2426     }
2427 
2428  freeit:
2429 	m_freem(m);
2430 	return;
2431 
2432  bad:
2433 	ICMP6STAT_INC(icp6s_badredirect);
2434 	m_freem(m);
2435 }
2436 
2437 void
2438 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2439 {
2440 	struct ifnet *ifp;	/* my outgoing interface */
2441 	struct in6_addr *ifp_ll6;
2442 	struct in6_addr *router_ll6;
2443 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2444 	struct mbuf *m = NULL;	/* newly allocated one */
2445 	struct m_tag *mtag;
2446 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2447 	struct nd_redirect *nd_rd;
2448 	struct llentry *ln = NULL;
2449 	size_t maxlen;
2450 	u_char *p;
2451 	struct ifnet *outif = NULL;
2452 	struct sockaddr_in6 src_sa;
2453 
2454 	icmp6_errcount(ND_REDIRECT, 0);
2455 
2456 	/* if we are not router, we don't send icmp6 redirect */
2457 	if (!V_ip6_forwarding)
2458 		goto fail;
2459 
2460 	/* sanity check */
2461 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2462 		goto fail;
2463 
2464 	/*
2465 	 * Address check:
2466 	 *  the source address must identify a neighbor, and
2467 	 *  the destination address must not be a multicast address
2468 	 *  [RFC 2461, sec 8.2]
2469 	 */
2470 	sip6 = mtod(m0, struct ip6_hdr *);
2471 	bzero(&src_sa, sizeof(src_sa));
2472 	src_sa.sin6_family = AF_INET6;
2473 	src_sa.sin6_len = sizeof(src_sa);
2474 	src_sa.sin6_addr = sip6->ip6_src;
2475 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2476 		goto fail;
2477 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2478 		goto fail;	/* what should we do here? */
2479 
2480 	/* rate limit */
2481 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2482 		goto fail;
2483 
2484 	/*
2485 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2486 	 * we almost always ask for an mbuf cluster for simplicity.
2487 	 * (MHLEN < IPV6_MMTU is almost always true)
2488 	 */
2489 #if IPV6_MMTU >= MCLBYTES
2490 # error assumption failed about IPV6_MMTU and MCLBYTES
2491 #endif
2492 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2493 	if (m == NULL)
2494 		goto fail;
2495 	M_SETFIB(m, rt->rt_fibnum);
2496 	maxlen = M_TRAILINGSPACE(m);
2497 	maxlen = min(IPV6_MMTU, maxlen);
2498 	/* just for safety */
2499 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2500 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2501 		goto fail;
2502 	}
2503 
2504 	{
2505 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2506 		struct in6_ifaddr *ia;
2507 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2508 						 IN6_IFF_NOTREADY|
2509 						 IN6_IFF_ANYCAST)) == NULL)
2510 			goto fail;
2511 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2512 		/* XXXRW: reference released prematurely. */
2513 		ifa_free(&ia->ia_ifa);
2514 	}
2515 
2516 	/* get ip6 linklocal address for the router. */
2517 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2518 		struct sockaddr_in6 *sin6;
2519 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2520 		router_ll6 = &sin6->sin6_addr;
2521 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2522 			router_ll6 = (struct in6_addr *)NULL;
2523 	} else
2524 		router_ll6 = (struct in6_addr *)NULL;
2525 
2526 	/* ip6 */
2527 	ip6 = mtod(m, struct ip6_hdr *);
2528 	ip6->ip6_flow = 0;
2529 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2530 	ip6->ip6_vfc |= IPV6_VERSION;
2531 	/* ip6->ip6_plen will be set later */
2532 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2533 	ip6->ip6_hlim = 255;
2534 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2535 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2536 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2537 
2538 	/* ND Redirect */
2539 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2540 	nd_rd->nd_rd_type = ND_REDIRECT;
2541 	nd_rd->nd_rd_code = 0;
2542 	nd_rd->nd_rd_reserved = 0;
2543 	if (rt->rt_flags & RTF_GATEWAY) {
2544 		/*
2545 		 * nd_rd->nd_rd_target must be a link-local address in
2546 		 * better router cases.
2547 		 */
2548 		if (!router_ll6)
2549 			goto fail;
2550 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2551 		    sizeof(nd_rd->nd_rd_target));
2552 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2553 		    sizeof(nd_rd->nd_rd_dst));
2554 	} else {
2555 		/* make sure redtgt == reddst */
2556 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2557 		    sizeof(nd_rd->nd_rd_target));
2558 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2559 		    sizeof(nd_rd->nd_rd_dst));
2560 	}
2561 
2562 	p = (u_char *)(nd_rd + 1);
2563 
2564 	if (!router_ll6)
2565 		goto nolladdropt;
2566 
2567 	{
2568 		/* target lladdr option */
2569 		int len;
2570 		struct nd_opt_hdr *nd_opt;
2571 		char *lladdr;
2572 
2573 		IF_AFDATA_RLOCK(ifp);
2574 		ln = nd6_lookup(router_ll6, 0, ifp);
2575 		IF_AFDATA_RUNLOCK(ifp);
2576 		if (ln == NULL)
2577 			goto nolladdropt;
2578 
2579 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2580 		len = (len + 7) & ~7;	/* round by 8 */
2581 		/* safety check */
2582 		if (len + (p - (u_char *)ip6) > maxlen)
2583 			goto nolladdropt;
2584 
2585 		if (ln->la_flags & LLE_VALID) {
2586 			nd_opt = (struct nd_opt_hdr *)p;
2587 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2588 			nd_opt->nd_opt_len = len >> 3;
2589 			lladdr = (char *)(nd_opt + 1);
2590 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2591 			p += len;
2592 		}
2593 	}
2594 nolladdropt:
2595 	if (ln != NULL)
2596 		LLE_RUNLOCK(ln);
2597 
2598 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2599 
2600 	/* just to be safe */
2601 #ifdef M_DECRYPTED	/*not openbsd*/
2602 	if (m0->m_flags & M_DECRYPTED)
2603 		goto noredhdropt;
2604 #endif
2605 	if (p - (u_char *)ip6 > maxlen)
2606 		goto noredhdropt;
2607 
2608 	{
2609 		/* redirected header option */
2610 		int len;
2611 		struct nd_opt_rd_hdr *nd_opt_rh;
2612 
2613 		/*
2614 		 * compute the maximum size for icmp6 redirect header option.
2615 		 * XXX room for auth header?
2616 		 */
2617 		len = maxlen - (p - (u_char *)ip6);
2618 		len &= ~7;
2619 
2620 		/* This is just for simplicity. */
2621 		if (m0->m_pkthdr.len != m0->m_len) {
2622 			if (m0->m_next) {
2623 				m_freem(m0->m_next);
2624 				m0->m_next = NULL;
2625 			}
2626 			m0->m_pkthdr.len = m0->m_len;
2627 		}
2628 
2629 		/*
2630 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2631 		 * about padding/truncate rule for the original IP packet.
2632 		 * From the discussion on IPv6imp in Feb 1999,
2633 		 * the consensus was:
2634 		 * - "attach as much as possible" is the goal
2635 		 * - pad if not aligned (original size can be guessed by
2636 		 *   original ip6 header)
2637 		 * Following code adds the padding if it is simple enough,
2638 		 * and truncates if not.
2639 		 */
2640 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2641 			panic("assumption failed in %s:%d", __FILE__,
2642 			    __LINE__);
2643 
2644 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2645 			/* not enough room, truncate */
2646 			m0->m_pkthdr.len = m0->m_len = len -
2647 			    sizeof(*nd_opt_rh);
2648 		} else {
2649 			/* enough room, pad or truncate */
2650 			size_t extra;
2651 
2652 			extra = m0->m_pkthdr.len % 8;
2653 			if (extra) {
2654 				/* pad if easy enough, truncate if not */
2655 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2656 					/* pad */
2657 					m0->m_len += (8 - extra);
2658 					m0->m_pkthdr.len += (8 - extra);
2659 				} else {
2660 					/* truncate */
2661 					m0->m_pkthdr.len -= extra;
2662 					m0->m_len -= extra;
2663 				}
2664 			}
2665 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2666 			m0->m_pkthdr.len = m0->m_len = len -
2667 			    sizeof(*nd_opt_rh);
2668 		}
2669 
2670 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2671 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2672 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2673 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2674 		p += sizeof(*nd_opt_rh);
2675 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2676 
2677 		/* connect m0 to m */
2678 		m_tag_delete_chain(m0, NULL);
2679 		m0->m_flags &= ~M_PKTHDR;
2680 		m->m_next = m0;
2681 		m->m_pkthdr.len = m->m_len + m0->m_len;
2682 		m0 = NULL;
2683 	}
2684 noredhdropt:;
2685 	if (m0) {
2686 		m_freem(m0);
2687 		m0 = NULL;
2688 	}
2689 
2690 	/* XXX: clear embedded link IDs in the inner header */
2691 	in6_clearscope(&sip6->ip6_src);
2692 	in6_clearscope(&sip6->ip6_dst);
2693 	in6_clearscope(&nd_rd->nd_rd_target);
2694 	in6_clearscope(&nd_rd->nd_rd_dst);
2695 
2696 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2697 
2698 	nd_rd->nd_rd_cksum = 0;
2699 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2700 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2701 
2702         if (send_sendso_input_hook != NULL) {
2703 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2704 			M_NOWAIT);
2705 		if (mtag == NULL)
2706 			goto fail;
2707 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2708 		m_tag_prepend(m, mtag);
2709 	}
2710 
2711 	/* send the packet to outside... */
2712 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2713 	if (outif) {
2714 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2715 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2716 	}
2717 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2718 
2719 	return;
2720 
2721 fail:
2722 	if (m)
2723 		m_freem(m);
2724 	if (m0)
2725 		m_freem(m0);
2726 }
2727 
2728 /*
2729  * ICMPv6 socket option processing.
2730  */
2731 int
2732 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2733 {
2734 	int error = 0;
2735 	int optlen;
2736 	struct inpcb *inp = sotoinpcb(so);
2737 	int level, op, optname;
2738 
2739 	if (sopt) {
2740 		level = sopt->sopt_level;
2741 		op = sopt->sopt_dir;
2742 		optname = sopt->sopt_name;
2743 		optlen = sopt->sopt_valsize;
2744 	} else
2745 		level = op = optname = optlen = 0;
2746 
2747 	if (level != IPPROTO_ICMPV6) {
2748 		return EINVAL;
2749 	}
2750 
2751 	switch (op) {
2752 	case PRCO_SETOPT:
2753 		switch (optname) {
2754 		case ICMP6_FILTER:
2755 		    {
2756 			struct icmp6_filter ic6f;
2757 
2758 			if (optlen != sizeof(ic6f)) {
2759 				error = EMSGSIZE;
2760 				break;
2761 			}
2762 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2763 			if (error == 0) {
2764 				INP_WLOCK(inp);
2765 				*inp->in6p_icmp6filt = ic6f;
2766 				INP_WUNLOCK(inp);
2767 			}
2768 			break;
2769 		    }
2770 
2771 		default:
2772 			error = ENOPROTOOPT;
2773 			break;
2774 		}
2775 		break;
2776 
2777 	case PRCO_GETOPT:
2778 		switch (optname) {
2779 		case ICMP6_FILTER:
2780 		    {
2781 			struct icmp6_filter ic6f;
2782 
2783 			INP_RLOCK(inp);
2784 			ic6f = *inp->in6p_icmp6filt;
2785 			INP_RUNLOCK(inp);
2786 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2787 			break;
2788 		    }
2789 
2790 		default:
2791 			error = ENOPROTOOPT;
2792 			break;
2793 		}
2794 		break;
2795 	}
2796 
2797 	return (error);
2798 }
2799 
2800 /*
2801  * Perform rate limit check.
2802  * Returns 0 if it is okay to send the icmp6 packet.
2803  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2804  * limitation.
2805  *
2806  * XXX per-destination/type check necessary?
2807  *
2808  * dst - not used at this moment
2809  * type - not used at this moment
2810  * code - not used at this moment
2811  */
2812 static int
2813 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2814     const int code)
2815 {
2816 	int ret;
2817 
2818 	ret = 0;	/* okay to send */
2819 
2820 	/* PPS limit */
2821 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2822 	    V_icmp6errppslim)) {
2823 		/* The packet is subject to rate limit */
2824 		ret++;
2825 	}
2826 
2827 	return ret;
2828 }
2829