xref: /openbsd/sys/netinet6/icmp6.c (revision 891d7ab6)
1 /*	$OpenBSD: icmp6.c,v 1.116 2011/04/06 19:23:15 sthen Exp $	*/
2 /*	$KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "faith.h"
65 #include "carp.h"
66 #include "pf.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 #include <sys/domain.h>
79 
80 #include <net/if.h>
81 #include <net/route.h>
82 #include <net/if_dl.h>
83 #include <net/if_types.h>
84 
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/ip.h>
89 #include <netinet/ip6.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/mld6_var.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet6/nd6.h>
95 #include <netinet6/in6_ifattach.h>
96 #include <netinet6/ip6protosw.h>
97 
98 #if NCARP > 0
99 #include <netinet/ip_carp.h>
100 #endif
101 
102 #if NPF > 0
103 #include <net/pfvar.h>
104 #endif
105 
106 /* inpcb members */
107 #define in6pcb		inpcb
108 #define in6p_laddr	inp_laddr6
109 #define in6p_faddr	inp_faddr6
110 #define in6p_icmp6filt	inp_icmp6filt
111 #define in6p_route	inp_route
112 #define in6p_socket	inp_socket
113 #define in6p_flags	inp_flags
114 #define in6p_moptions	inp_moptions6
115 #define in6p_outputopts	inp_outputopts6
116 #define in6p_ip6	inp_ipv6
117 #define in6p_flowinfo	inp_flowinfo
118 #define in6p_sp		inp_sp
119 #define in6p_next	inp_next
120 #define in6p_prev	inp_prev
121 /* macro names */
122 #define sotoin6pcb	sotoinpcb
123 /* function names */
124 #define in6_pcbdetach	in_pcbdetach
125 #define in6_rtchange	in_rtchange
126 
127 /*
128  * for KAME src sync over BSD*'s.  XXX: FreeBSD (>=3) are VERY different from
129  * others...
130  */
131 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
132 
133 extern struct domain inet6domain;
134 extern struct ip6protosw inet6sw[];
135 extern u_char ip6_protox[];
136 
137 struct icmp6stat icmp6stat;
138 
139 extern struct inpcbtable rawin6pcbtable;
140 extern int icmp6errppslim;
141 static int icmp6errpps_count = 0;
142 static struct timeval icmp6errppslim_last;
143 extern int icmp6_nodeinfo;
144 
145 /*
146  * List of callbacks to notify when Path MTU changes are made.
147  */
148 struct icmp6_mtudisc_callback {
149 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
150 	void (*mc_func)(struct in6_addr *);
151 };
152 
153 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
154     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
155 
156 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
157 extern int pmtu_expire;
158 
159 /* XXX do these values make any sense? */
160 static int icmp6_mtudisc_hiwat = 1280;
161 static int icmp6_mtudisc_lowat = 256;
162 
163 /*
164  * keep track of # of redirect routes.
165  */
166 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
167 
168 /* XXX experimental, turned off */
169 static int icmp6_redirect_lowat = -1;
170 
171 void	icmp6_errcount(struct icmp6errstat *, int, int);
172 int	icmp6_rip6_input(struct mbuf **, int);
173 int	icmp6_ratelimit(const struct in6_addr *, const int, const int);
174 const char *icmp6_redirect_diag(struct in6_addr *, struct in6_addr *,
175 	    struct in6_addr *);
176 struct mbuf *ni6_input(struct mbuf *, int);
177 struct mbuf *ni6_nametodns(const char *, int, int);
178 int	ni6_dnsmatch(const char *, int, const char *, int);
179 int	ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, struct ifnet **,
180 	    char *);
181 int	ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
182 	    struct ifnet *, int);
183 int	icmp6_notify_error(struct mbuf *, int, int, int);
184 struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
185 void	icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
186 void	icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
187 
188 void
189 icmp6_init(void)
190 {
191 	mld6_init();
192 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
193 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
194 }
195 
196 void
197 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
198 {
199 	switch (type) {
200 	case ICMP6_DST_UNREACH:
201 		switch (code) {
202 		case ICMP6_DST_UNREACH_NOROUTE:
203 			stat->icp6errs_dst_unreach_noroute++;
204 			return;
205 		case ICMP6_DST_UNREACH_ADMIN:
206 			stat->icp6errs_dst_unreach_admin++;
207 			return;
208 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
209 			stat->icp6errs_dst_unreach_beyondscope++;
210 			return;
211 		case ICMP6_DST_UNREACH_ADDR:
212 			stat->icp6errs_dst_unreach_addr++;
213 			return;
214 		case ICMP6_DST_UNREACH_NOPORT:
215 			stat->icp6errs_dst_unreach_noport++;
216 			return;
217 		}
218 		break;
219 	case ICMP6_PACKET_TOO_BIG:
220 		stat->icp6errs_packet_too_big++;
221 		return;
222 	case ICMP6_TIME_EXCEEDED:
223 		switch (code) {
224 		case ICMP6_TIME_EXCEED_TRANSIT:
225 			stat->icp6errs_time_exceed_transit++;
226 			return;
227 		case ICMP6_TIME_EXCEED_REASSEMBLY:
228 			stat->icp6errs_time_exceed_reassembly++;
229 			return;
230 		}
231 		break;
232 	case ICMP6_PARAM_PROB:
233 		switch (code) {
234 		case ICMP6_PARAMPROB_HEADER:
235 			stat->icp6errs_paramprob_header++;
236 			return;
237 		case ICMP6_PARAMPROB_NEXTHEADER:
238 			stat->icp6errs_paramprob_nextheader++;
239 			return;
240 		case ICMP6_PARAMPROB_OPTION:
241 			stat->icp6errs_paramprob_option++;
242 			return;
243 		}
244 		break;
245 	case ND_REDIRECT:
246 		stat->icp6errs_redirect++;
247 		return;
248 	}
249 	stat->icp6errs_unknown++;
250 }
251 
252 /*
253  * Register a Path MTU Discovery callback.
254  */
255 void
256 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
257 {
258 	struct icmp6_mtudisc_callback *mc;
259 
260 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
261 	     mc = LIST_NEXT(mc, mc_list)) {
262 		if (mc->mc_func == func)
263 			return;
264 	}
265 
266 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
267 	if (mc == NULL)
268 		panic("icmp6_mtudisc_callback_register");
269 
270 	mc->mc_func = func;
271 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
272 }
273 
274 /*
275  * Generate an error packet of type error in response to bad IP6 packet.
276  */
277 void
278 icmp6_error(struct mbuf *m, int type, int code, int param)
279 {
280 	struct ip6_hdr *oip6, *nip6;
281 	struct icmp6_hdr *icmp6;
282 	u_int preplen;
283 	int off;
284 	int nxt;
285 
286 	icmp6stat.icp6s_error++;
287 
288 	/* count per-type-code statistics */
289 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
290 
291 	if (m->m_len < sizeof(struct ip6_hdr)) {
292 		m = m_pullup(m, sizeof(struct ip6_hdr));
293 		if (m == NULL)
294 			return;
295 	}
296 	oip6 = mtod(m, struct ip6_hdr *);
297 
298 	/*
299 	 * If the destination address of the erroneous packet is a multicast
300 	 * address, or the packet was sent using link-layer multicast,
301 	 * we should basically suppress sending an error (RFC 2463, Section
302 	 * 2.4).
303 	 * We have two exceptions (the item e.2 in that section):
304 	 * - the Packet Too Big message can be sent for path MTU discovery.
305 	 * - the Parameter Problem Message that can be allowed an icmp6 error
306 	 *   in the option type field.  This check has been done in
307 	 *   ip6_unknown_opt(), so we can just check the type and code.
308 	 */
309 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
310 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
311 	    (type != ICMP6_PACKET_TOO_BIG &&
312 	     (type != ICMP6_PARAM_PROB ||
313 	      code != ICMP6_PARAMPROB_OPTION)))
314 		goto freeit;
315 
316 	/*
317 	 * RFC 2463, 2.4 (e.5): source address check.
318 	 * XXX: the case of anycast source?
319 	 */
320 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
321 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
322 		goto freeit;
323 
324 	/*
325 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
326 	 * don't do it.
327 	 */
328 	nxt = -1;
329 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
330 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
331 		struct icmp6_hdr *icp;
332 
333 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
334 			sizeof(*icp));
335 		if (icp == NULL) {
336 			icmp6stat.icp6s_tooshort++;
337 			return;
338 		}
339 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
340 		    icp->icmp6_type == ND_REDIRECT) {
341 			/*
342 			 * ICMPv6 error
343 			 * Special case: for redirect (which is
344 			 * informational) we must not send icmp6 error.
345 			 */
346 			icmp6stat.icp6s_canterror++;
347 			goto freeit;
348 		} else {
349 			/* ICMPv6 informational - send the error */
350 		}
351 	}
352 	else {
353 		/* non-ICMPv6 - send the error */
354 	}
355 
356 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
357 
358 	/* Finally, do rate limitation check. */
359 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
360 		icmp6stat.icp6s_toofreq++;
361 		goto freeit;
362 	}
363 
364 	/*
365 	 * OK, ICMP6 can be generated.
366 	 */
367 
368 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
369 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
370 
371 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
372 	M_PREPEND(m, preplen, M_DONTWAIT);
373 	if (m && m->m_len < preplen)
374 		m = m_pullup(m, preplen);
375 	if (m == NULL) {
376 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
377 		return;
378 	}
379 
380 	nip6 = mtod(m, struct ip6_hdr *);
381 	nip6->ip6_src  = oip6->ip6_src;
382 	nip6->ip6_dst  = oip6->ip6_dst;
383 
384 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_src))
385 		oip6->ip6_src.s6_addr16[1] = 0;
386 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_dst))
387 		oip6->ip6_dst.s6_addr16[1] = 0;
388 
389 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
390 	icmp6->icmp6_type = type;
391 	icmp6->icmp6_code = code;
392 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
393 
394 	/*
395 	 * icmp6_reflect() is designed to be in the input path.
396 	 * icmp6_error() can be called from both input and outut path,
397 	 * and if we are in output path rcvif could contain bogus value.
398 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
399 	 * information in ip header (nip6).
400 	 */
401 	m->m_pkthdr.rcvif = NULL;
402 
403 	icmp6stat.icp6s_outhist[type]++;
404 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
405 
406 	return;
407 
408   freeit:
409 	/*
410 	 * If we can't tell wheter or not we can generate ICMP6, free it.
411 	 */
412 	m_freem(m);
413 }
414 
415 /*
416  * Process a received ICMP6 message.
417  */
418 int
419 icmp6_input(struct mbuf **mp, int *offp, int proto)
420 {
421 	struct mbuf *m = *mp, *n;
422 	struct ip6_hdr *ip6, *nip6;
423 	struct icmp6_hdr *icmp6, *nicmp6;
424 	int off = *offp;
425 	int icmp6len = m->m_pkthdr.len - *offp;
426 	int code, sum, noff;
427 
428 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
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 	if (icmp6len < sizeof(struct icmp6_hdr)) {
437 		icmp6stat.icp6s_tooshort++;
438 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
439 		goto freeit;
440 	}
441 
442 	/*
443 	 * calculate the checksum
444 	 */
445 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
446 	if (icmp6 == NULL) {
447 		icmp6stat.icp6s_tooshort++;
448 		return IPPROTO_DONE;
449 	}
450 	code = icmp6->icmp6_code;
451 
452 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
453 		nd6log((LOG_ERR,
454 		    "ICMP6 checksum error(%d|%x) %s\n",
455 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
456 		icmp6stat.icp6s_checksum++;
457 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
458 		goto freeit;
459 	}
460 
461 #if defined(NFAITH) && 0 < NFAITH
462 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
463 		/*
464 		 * Deliver very specific ICMP6 type only.
465 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
466 		 * will not work.
467 		 */
468 		switch (icmp6->icmp6_type) {
469 		case ICMP6_DST_UNREACH:
470 		case ICMP6_PACKET_TOO_BIG:
471 		case ICMP6_TIME_EXCEEDED:
472 			break;
473 		default:
474 			goto freeit;
475 		}
476 	}
477 #endif
478 
479 #if NCARP > 0
480 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
481 	    icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
482 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
483 	    ip6->ip6_dst.s6_addr32))
484 		goto freeit;
485 #endif
486 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
487 
488 	switch (icmp6->icmp6_type) {
489 	case ICMP6_DST_UNREACH:
490 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
491 		switch (code) {
492 		case ICMP6_DST_UNREACH_NOROUTE:
493 			code = PRC_UNREACH_NET;
494 			break;
495 		case ICMP6_DST_UNREACH_ADMIN:
496 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
497 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
498 			break;
499 		case ICMP6_DST_UNREACH_ADDR:
500 			code = PRC_HOSTDEAD;
501 			break;
502 #ifdef COMPAT_RFC1885
503 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
504 			code = PRC_UNREACH_SRCFAIL;
505 			break;
506 #else
507 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
508 			/* I mean "source address was incorrect." */
509 			code = PRC_PARAMPROB;
510 			break;
511 #endif
512 		case ICMP6_DST_UNREACH_NOPORT:
513 			code = PRC_UNREACH_PORT;
514 			break;
515 		default:
516 			goto badcode;
517 		}
518 		goto deliver;
519 
520 	case ICMP6_PACKET_TOO_BIG:
521 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
522 
523 		/* MTU is checked in icmp6_mtudisc_update. */
524 		code = PRC_MSGSIZE;
525 
526 		/*
527 		 * Updating the path MTU will be done after examining
528 		 * intermediate extension headers.
529 		 */
530 		goto deliver;
531 
532 	case ICMP6_TIME_EXCEEDED:
533 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
534 		switch (code) {
535 		case ICMP6_TIME_EXCEED_TRANSIT:
536 			code = PRC_TIMXCEED_INTRANS;
537 			break;
538 		case ICMP6_TIME_EXCEED_REASSEMBLY:
539 			code = PRC_TIMXCEED_REASS;
540 			break;
541 		default:
542 			goto badcode;
543 		}
544 		goto deliver;
545 
546 	case ICMP6_PARAM_PROB:
547 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
548 		switch (code) {
549 		case ICMP6_PARAMPROB_NEXTHEADER:
550 			code = PRC_UNREACH_PROTOCOL;
551 			break;
552 		case ICMP6_PARAMPROB_HEADER:
553 		case ICMP6_PARAMPROB_OPTION:
554 			code = PRC_PARAMPROB;
555 			break;
556 		default:
557 			goto badcode;
558 		}
559 		goto deliver;
560 
561 	case ICMP6_ECHO_REQUEST:
562 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
563 		if (code != 0)
564 			goto badcode;
565 		/*
566 		 * Copy mbuf to send to two data paths: userland socket(s),
567 		 * and to the querier (echo reply).
568 		 * m: a copy for socket, n: a copy for querier
569 		 */
570 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
571 			/* Give up local */
572 			n = m;
573 			m = NULL;
574 			goto deliverecho;
575 		}
576 		/*
577 		 * If the first mbuf is shared, or the first mbuf is too short,
578 		 * copy the first part of the data into a fresh mbuf.
579 		 * Otherwise, we will wrongly overwrite both copies.
580 		 */
581 		if ((n->m_flags & M_EXT) != 0 ||
582 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
583 			struct mbuf *n0 = n;
584 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
585 
586 			/*
587 			 * Prepare an internal mbuf.  m_pullup() doesn't
588 			 * always copy the length we specified.
589 			 */
590 			if (maxlen >= MCLBYTES) {
591 				/* Give up remote */
592 				m_freem(n0);
593 				break;
594 			}
595 			MGETHDR(n, M_DONTWAIT, n0->m_type);
596 			if (n && maxlen >= MHLEN) {
597 				MCLGET(n, M_DONTWAIT);
598 				if ((n->m_flags & M_EXT) == 0) {
599 					m_free(n);
600 					n = NULL;
601 				}
602 			}
603 			if (n == NULL) {
604 				/* Give up local */
605 				m_freem(n0);
606 				n = m;
607 				m = NULL;
608 				goto deliverecho;
609 			}
610 			M_MOVE_PKTHDR(n, n0);
611 			/*
612 			 * Copy IPv6 and ICMPv6 only.
613 			 */
614 			nip6 = mtod(n, struct ip6_hdr *);
615 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
616 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
617 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
618 			noff = sizeof(struct ip6_hdr);
619 			n->m_len = noff + sizeof(struct icmp6_hdr);
620 			/*
621 			 * Adjust mbuf.  ip6_plen will be adjusted in
622 			 * ip6_output().
623 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
624 			 */
625 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
626 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
627 			m_adj(n0, off + sizeof(struct icmp6_hdr));
628 			n->m_next = n0;
629 		} else {
630 	 deliverecho:
631 			nip6 = mtod(n, struct ip6_hdr *);
632 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
633 			    sizeof(*nicmp6));
634 			noff = off;
635 		}
636 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
637 		nicmp6->icmp6_code = 0;
638 		if (n) {
639 			icmp6stat.icp6s_reflect++;
640 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
641 			icmp6_reflect(n, noff);
642 		}
643 		if (!m)
644 			goto freeit;
645 		break;
646 
647 	case ICMP6_ECHO_REPLY:
648 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
649 		if (code != 0)
650 			goto badcode;
651 		break;
652 
653 	case MLD_LISTENER_QUERY:
654 	case MLD_LISTENER_REPORT:
655 		if (icmp6len < sizeof(struct mld_hdr))
656 			goto badlen;
657 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
658 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
659 		else
660 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
661 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
662 			/* give up local */
663 			mld6_input(m, off);
664 			m = NULL;
665 			goto freeit;
666 		}
667 		mld6_input(n, off);
668 		/* m stays. */
669 		break;
670 
671 	case MLD_LISTENER_DONE:
672 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
673 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
674 			goto badlen;
675 		break;		/* nothing to be done in kernel */
676 
677 	case MLD_MTRACE_RESP:
678 	case MLD_MTRACE:
679 		/* XXX: these two are experimental.  not officially defined. */
680 		/* XXX: per-interface statistics? */
681 		break;		/* just pass it to applications */
682 
683 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
684 	    {
685 		enum { WRU, FQDN } mode;
686 
687 		if (!icmp6_nodeinfo)
688 			break;
689 
690 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
691 			mode = WRU;
692 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
693 			mode = FQDN;
694 		else
695 			goto badlen;
696 
697 		if (mode == FQDN) {
698 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
699 			if (n)
700 				n = ni6_input(n, off);
701 			/* XXX meaningless if n == NULL */
702 			noff = sizeof(struct ip6_hdr);
703 		} else {
704 			u_char *p;
705 			int maxlen, maxhlen;
706 
707 			if ((icmp6_nodeinfo & 1) == 0)
708 				break;
709 
710 			if (code != 0)
711 				goto badcode;
712 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
713 			if (maxlen >= MCLBYTES) {
714 				/* Give up remote */
715 				break;
716 			}
717 			MGETHDR(n, M_DONTWAIT, m->m_type);
718 			if (n && maxlen > MHLEN) {
719 				MCLGET(n, M_DONTWAIT);
720 				if ((n->m_flags & M_EXT) == 0) {
721 					m_free(n);
722 					n = NULL;
723 				}
724 			}
725 			if (n == NULL) {
726 				/* Give up remote */
727 				break;
728 			}
729 			n->m_pkthdr.rcvif = NULL;
730 			n->m_len = 0;
731 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
732 			if (maxhlen > hostnamelen)
733 				maxhlen = hostnamelen;
734 			/*
735 			 * Copy IPv6 and ICMPv6 only.
736 			 */
737 			nip6 = mtod(n, struct ip6_hdr *);
738 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
739 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
740 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
741 			p = (u_char *)(nicmp6 + 1);
742 			bzero(p, 4);
743 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
744 			noff = sizeof(struct ip6_hdr);
745 			if (m_dup_pkthdr(n, m, M_DONTWAIT)) { /* for rcvif */
746 				m_freem(n);
747 				break;
748 			}
749 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
750 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
751 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
752 			nicmp6->icmp6_code = 0;
753 		}
754 #undef hostnamelen
755 		if (n) {
756 			icmp6stat.icp6s_reflect++;
757 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
758 			icmp6_reflect(n, noff);
759 		}
760 		break;
761 	    }
762 
763 	case ICMP6_WRUREPLY:
764 		if (code != 0)
765 			goto badcode;
766 		break;
767 
768 	case ND_ROUTER_SOLICIT:
769 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
770 		if (code != 0)
771 			goto badcode;
772 		if (icmp6len < sizeof(struct nd_router_solicit))
773 			goto badlen;
774 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
775 			/* give up local */
776 			nd6_rs_input(m, off, icmp6len);
777 			m = NULL;
778 			goto freeit;
779 		}
780 		nd6_rs_input(n, off, icmp6len);
781 		/* m stays. */
782 		break;
783 
784 	case ND_ROUTER_ADVERT:
785 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
786 		if (code != 0)
787 			goto badcode;
788 		if (icmp6len < sizeof(struct nd_router_advert))
789 			goto badlen;
790 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
791 			/* give up local */
792 			nd6_ra_input(m, off, icmp6len);
793 			m = NULL;
794 			goto freeit;
795 		}
796 		nd6_ra_input(n, off, icmp6len);
797 		/* m stays. */
798 		break;
799 
800 	case ND_NEIGHBOR_SOLICIT:
801 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
802 		if (code != 0)
803 			goto badcode;
804 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
805 			goto badlen;
806 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
807 			/* give up local */
808 			nd6_ns_input(m, off, icmp6len);
809 			m = NULL;
810 			goto freeit;
811 		}
812 		nd6_ns_input(n, off, icmp6len);
813 		/* m stays. */
814 		break;
815 
816 	case ND_NEIGHBOR_ADVERT:
817 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
818 		if (code != 0)
819 			goto badcode;
820 		if (icmp6len < sizeof(struct nd_neighbor_advert))
821 			goto badlen;
822 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
823 			/* give up local */
824 			nd6_na_input(m, off, icmp6len);
825 			m = NULL;
826 			goto freeit;
827 		}
828 		nd6_na_input(n, off, icmp6len);
829 		/* m stays. */
830 		break;
831 
832 	case ND_REDIRECT:
833 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
834 		if (code != 0)
835 			goto badcode;
836 		if (icmp6len < sizeof(struct nd_redirect))
837 			goto badlen;
838 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
839 			/* give up local */
840 			icmp6_redirect_input(m, off);
841 			m = NULL;
842 			goto freeit;
843 		}
844 		icmp6_redirect_input(n, off);
845 		/* m stays. */
846 		break;
847 
848 	case ICMP6_ROUTER_RENUMBERING:
849 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
850 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
851 			goto badcode;
852 		if (icmp6len < sizeof(struct icmp6_router_renum))
853 			goto badlen;
854 		break;
855 
856 	default:
857 		nd6log((LOG_DEBUG,
858 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
859 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
860 		    ip6_sprintf(&ip6->ip6_dst),
861 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
862 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
863 			/* ICMPv6 error: MUST deliver it by spec... */
864 			code = PRC_NCMDS;
865 			/* deliver */
866 		} else {
867 			/* ICMPv6 informational: MUST not deliver */
868 			break;
869 		}
870 deliver:
871 		if (icmp6_notify_error(m, off, icmp6len, code)) {
872 			/* In this case, m should've been freed. */
873 			return (IPPROTO_DONE);
874 		}
875 		break;
876 
877 badcode:
878 		icmp6stat.icp6s_badcode++;
879 		break;
880 
881 badlen:
882 		icmp6stat.icp6s_badlen++;
883 		break;
884 	}
885 
886 	/* deliver the packet to appropriate sockets */
887 	icmp6_rip6_input(&m, *offp);
888 
889 	return IPPROTO_DONE;
890 
891  freeit:
892 	m_freem(m);
893 	return IPPROTO_DONE;
894 }
895 
896 int
897 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
898 {
899 	struct icmp6_hdr *icmp6;
900 	struct ip6_hdr *eip6;
901 	u_int32_t notifymtu;
902 	struct sockaddr_in6 icmp6src, icmp6dst;
903 
904 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
905 		icmp6stat.icp6s_tooshort++;
906 		goto freeit;
907 	}
908 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
909 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
910 	if (icmp6 == NULL) {
911 		icmp6stat.icp6s_tooshort++;
912 		return (-1);
913 	}
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 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
938 					       eoff, sizeof(*eh));
939 				if (eh == NULL) {
940 					icmp6stat.icp6s_tooshort++;
941 					return (-1);
942 				}
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 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
960 					       eoff, sizeof(*rth));
961 				if (rth == NULL) {
962 					icmp6stat.icp6s_tooshort++;
963 					return (-1);
964 				}
965 				rthlen = (rth->ip6r_len + 1) << 3;
966 				/*
967 				 * XXX: currently there is no
968 				 * officially defined type other
969 				 * than type-0.
970 				 * Note that if the segment left field
971 				 * is 0, all intermediate hops must
972 				 * have been passed.
973 				 */
974 				if (rth->ip6r_segleft &&
975 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
976 					int hops;
977 
978 					IP6_EXTHDR_GET(rth0,
979 						       struct ip6_rthdr0 *, m,
980 						       eoff, rthlen);
981 					if (rth0 == NULL) {
982 						icmp6stat.icp6s_tooshort++;
983 						return (-1);
984 					}
985 					/* just ignore a bogus header */
986 					if ((rth0->ip6r0_len % 2) == 0 &&
987 					    (hops = rth0->ip6r0_len/2))
988 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
989 				}
990 				eoff += rthlen;
991 				nxt = rth->ip6r_nxt;
992 				break;
993 			case IPPROTO_FRAGMENT:
994 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
995 					       eoff, sizeof(*fh));
996 				if (fh == NULL) {
997 					icmp6stat.icp6s_tooshort++;
998 					return (-1);
999 				}
1000 				/*
1001 				 * Data after a fragment header is meaningless
1002 				 * unless it is the first fragment, but
1003 				 * we'll go to the notify label for path MTU
1004 				 * discovery.
1005 				 */
1006 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1007 					goto notify;
1008 
1009 				eoff += sizeof(struct ip6_frag);
1010 				nxt = fh->ip6f_nxt;
1011 				break;
1012 			default:
1013 				/*
1014 				 * This case includes ESP and the No Next
1015 				 * Header.  In such cases going to the notify
1016 				 * label does not have any meaning
1017 				 * (i.e. ctlfunc will be NULL), but we go
1018 				 * anyway since we might have to update
1019 				 * path MTU information.
1020 				 */
1021 				goto notify;
1022 			}
1023 		}
1024 	  notify:
1025 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1026 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1027 		if (icmp6 == NULL) {
1028 			icmp6stat.icp6s_tooshort++;
1029 			return (-1);
1030 		}
1031 
1032 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1033 		bzero(&icmp6dst, sizeof(icmp6dst));
1034 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1035 		icmp6dst.sin6_family = AF_INET6;
1036 		if (finaldst == NULL)
1037 			icmp6dst.sin6_addr = eip6->ip6_dst;
1038 		else
1039 			icmp6dst.sin6_addr = *finaldst;
1040 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1041 							  &icmp6dst.sin6_addr);
1042 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1043 				   NULL, NULL)) {
1044 			/* should be impossbile */
1045 			nd6log((LOG_DEBUG,
1046 			    "icmp6_notify_error: in6_embedscope failed\n"));
1047 			goto freeit;
1048 		}
1049 
1050 		/*
1051 		 * retrieve parameters from the inner IPv6 header, and convert
1052 		 * them into sockaddr structures.
1053 		 */
1054 		bzero(&icmp6src, sizeof(icmp6src));
1055 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1056 		icmp6src.sin6_family = AF_INET6;
1057 		icmp6src.sin6_addr = eip6->ip6_src;
1058 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1059 							  &icmp6src.sin6_addr);
1060 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1061 				   NULL, NULL)) {
1062 			/* should be impossbile */
1063 			nd6log((LOG_DEBUG,
1064 			    "icmp6_notify_error: in6_embedscope failed\n"));
1065 			goto freeit;
1066 		}
1067 		icmp6src.sin6_flowinfo =
1068 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1069 
1070 		if (finaldst == NULL)
1071 			finaldst = &eip6->ip6_dst;
1072 		ip6cp.ip6c_m = m;
1073 		ip6cp.ip6c_icmp6 = icmp6;
1074 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1075 		ip6cp.ip6c_off = eoff;
1076 		ip6cp.ip6c_finaldst = finaldst;
1077 		ip6cp.ip6c_src = &icmp6src;
1078 		ip6cp.ip6c_nxt = nxt;
1079 #if NPF > 0
1080 		pf_pkt_addr_changed(m);
1081 #endif
1082 
1083 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1084 			notifymtu = ntohl(icmp6->icmp6_mtu);
1085 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1086 		}
1087 
1088 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1089 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1090 		if (ctlfunc) {
1091 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1092 					  &ip6cp);
1093 		}
1094 	}
1095 	return (0);
1096 
1097   freeit:
1098 	m_freem(m);
1099 	return (-1);
1100 }
1101 
1102 void
1103 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1104 {
1105 	unsigned long rtcount;
1106 	struct icmp6_mtudisc_callback *mc;
1107 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1108 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1109 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1110 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1111 	struct rtentry *rt = NULL;
1112 	struct sockaddr_in6 sin6;
1113 
1114 	/*
1115 	 * The MTU may not be less then the minimal IPv6 MTU except for the
1116 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1117 	 */
1118 	if (mtu < IPV6_MMTU - sizeof(struct ip6_frag))
1119 		return;
1120 
1121 	/*
1122 	 * allow non-validated cases if memory is plenty, to make traffic
1123 	 * from non-connected pcb happy.
1124 	 */
1125 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1126 	if (validated) {
1127 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1128 			return;
1129 		else if (0 <= icmp6_mtudisc_lowat &&
1130 		    rtcount > icmp6_mtudisc_lowat) {
1131 			/*
1132 			 * XXX nuke a victim, install the new one.
1133 			 */
1134 		}
1135 	} else {
1136 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1137 			return;
1138 	}
1139 
1140 	bzero(&sin6, sizeof(sin6));
1141 	sin6.sin6_family = PF_INET6;
1142 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1143 	sin6.sin6_addr = *dst;
1144 	/* XXX normally, this won't happen */
1145 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1146 		sin6.sin6_addr.s6_addr16[1] =
1147 		    htons(m->m_pkthdr.rcvif->if_index);
1148 	}
1149 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1150 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1151 
1152 	if (rt && (rt->rt_flags & RTF_HOST) &&
1153 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1154 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1155 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1156 			icmp6stat.icp6s_pmtuchg++;
1157 			rt->rt_rmx.rmx_mtu = mtu;
1158 		}
1159 	}
1160 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1161 		RTFREE(rt);
1162 	}
1163 
1164 	/*
1165 	 * Notify protocols that the MTU for this destination
1166 	 * has changed.
1167 	 */
1168 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1169 	     mc = LIST_NEXT(mc, mc_list))
1170 		(*mc->mc_func)(&sin6.sin6_addr);
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 struct mbuf *
1185 ni6_input(struct mbuf *m, int off)
1186 {
1187 	struct icmp6_nodeinfo *ni6, *nni6;
1188 	struct mbuf *n = NULL;
1189 	u_int16_t qtype;
1190 	int subjlen;
1191 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1192 	struct ni_reply_fqdn *fqdn;
1193 	int addrs;		/* for NI_QTYPE_NODEADDR */
1194 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1195 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1196 	struct ip6_hdr *ip6;
1197 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1198 	char *subj = NULL;
1199 
1200 	ip6 = mtod(m, struct ip6_hdr *);
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 
1207 	/*
1208 	 * Validate IPv6 destination address.
1209 	 *
1210 	 * The Responder must discard the Query without further processing
1211 	 * unless it is one of the Responder's unicast or anycast addresses, or
1212 	 * a link-local scope multicast address which the Responder has joined.
1213 	 * [icmp-name-lookups-07, Section 4.]
1214 	 */
1215 	bzero(&sin6, sizeof(sin6));
1216 	sin6.sin6_family = AF_INET6;
1217 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1218 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1219 	/* XXX scopeid */
1220 	if (ifa_ifwithaddr((struct sockaddr *)&sin6, /* XXX */ 0))
1221 		; /* unicast/anycast, fine */
1222 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1223 		; /* link-local multicast, fine */
1224 	else
1225 		goto bad;
1226 
1227 	/* validate query Subject field. */
1228 	qtype = ntohs(ni6->ni_qtype);
1229 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1230 	switch (qtype) {
1231 	case NI_QTYPE_NOOP:
1232 	case NI_QTYPE_SUPTYPES:
1233 		/* 07 draft */
1234 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1235 			break;
1236 		/* FALLTHROUGH */
1237 	case NI_QTYPE_FQDN:
1238 	case NI_QTYPE_NODEADDR:
1239 		switch (ni6->ni_code) {
1240 		case ICMP6_NI_SUBJ_IPV6:
1241 #if ICMP6_NI_SUBJ_IPV6 != 0
1242 		case 0:
1243 #endif
1244 			/*
1245 			 * backward compatibility - try to accept 03 draft
1246 			 * format, where no Subject is present.
1247 			 */
1248 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1249 			    subjlen == 0) {
1250 				oldfqdn++;
1251 				break;
1252 			}
1253 #if ICMP6_NI_SUBJ_IPV6 != 0
1254 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1255 				goto bad;
1256 #endif
1257 
1258 			if (subjlen != sizeof(sin6.sin6_addr))
1259 				goto bad;
1260 
1261 			/*
1262 			 * Validate Subject address.
1263 			 *
1264 			 * Not sure what exactly "address belongs to the node"
1265 			 * means in the spec, is it just unicast, or what?
1266 			 *
1267 			 * At this moment we consider Subject address as
1268 			 * "belong to the node" if the Subject address equals
1269 			 * to the IPv6 destination address; validation for
1270 			 * IPv6 destination address should have done enough
1271 			 * check for us.
1272 			 *
1273 			 * We do not do proxy at this moment.
1274 			 */
1275 			/* m_pulldown instead of copy? */
1276 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1277 			    subjlen, (caddr_t)&sin6.sin6_addr);
1278 			/* XXX kame scope hack */
1279 			if (IN6_IS_SCOPE_EMBED(&sin6.sin6_addr)) {
1280 				if ((m->m_flags & M_PKTHDR) != 0 &&
1281 				    m->m_pkthdr.rcvif) {
1282 					sin6.sin6_addr.s6_addr16[1] =
1283 					    htons(m->m_pkthdr.rcvif->if_index);
1284 				}
1285 			}
1286 			subj = (char *)&sin6;
1287 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1288 				break;
1289 
1290 			/*
1291 			 * XXX if we are to allow other cases, we should really
1292 			 * be careful about scope here.
1293 			 * basically, we should disallow queries toward IPv6
1294 			 * destination X with subject Y, if scope(X) > scope(Y).
1295 			 * if we allow scope(X) > scope(Y), it will result in
1296 			 * information leakage across scope boundary.
1297 			 */
1298 			goto bad;
1299 
1300 		case ICMP6_NI_SUBJ_FQDN:
1301 			/*
1302 			 * Validate Subject name with gethostname(3).
1303 			 *
1304 			 * The behavior may need some debate, since:
1305 			 * - we are not sure if the node has FQDN as
1306 			 *   hostname (returned by gethostname(3)).
1307 			 * - the code does wildcard match for truncated names.
1308 			 *   however, we are not sure if we want to perform
1309 			 *   wildcard match, if gethostname(3) side has
1310 			 *   truncated hostname.
1311 			 */
1312 			n = ni6_nametodns(hostname, hostnamelen, 0);
1313 			if (!n || n->m_next || n->m_len == 0)
1314 				goto bad;
1315 			IP6_EXTHDR_GET(subj, char *, m,
1316 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1317 			if (subj == NULL)
1318 				goto bad;
1319 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1320 					n->m_len)) {
1321 				goto bad;
1322 			}
1323 			m_freem(n);
1324 			n = NULL;
1325 			break;
1326 
1327 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1328 		default:
1329 			goto bad;
1330 		}
1331 		break;
1332 	}
1333 
1334 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1335 	switch (qtype) {
1336 	case NI_QTYPE_FQDN:
1337 		if ((icmp6_nodeinfo & 1) == 0)
1338 			goto bad;
1339 		break;
1340 	case NI_QTYPE_NODEADDR:
1341 		if ((icmp6_nodeinfo & 2) == 0)
1342 			goto bad;
1343 		break;
1344 	}
1345 
1346 	/* guess reply length */
1347 	switch (qtype) {
1348 	case NI_QTYPE_NOOP:
1349 		break;		/* no reply data */
1350 	case NI_QTYPE_SUPTYPES:
1351 		replylen += sizeof(u_int32_t);
1352 		break;
1353 	case NI_QTYPE_FQDN:
1354 		/* XXX will append an mbuf */
1355 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1356 		break;
1357 	case NI_QTYPE_NODEADDR:
1358 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1359 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1360 					  sizeof(u_int32_t))) > MCLBYTES)
1361 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1362 		break;
1363 	default:
1364 		/*
1365 		 * XXX: We must return a reply with the ICMP6 code
1366 		 * `unknown Qtype' in this case.  However we regard the case
1367 		 * as an FQDN query for backward compatibility.
1368 		 * Older versions set a random value to this field,
1369 		 * so it rarely varies in the defined qtypes.
1370 		 * But the mechanism is not reliable...
1371 		 * maybe we should obsolete older versions.
1372 		 */
1373 		qtype = NI_QTYPE_FQDN;
1374 		/* XXX will append an mbuf */
1375 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1376 		oldfqdn++;
1377 		break;
1378 	}
1379 
1380 	/* allocate an mbuf to reply. */
1381 	MGETHDR(n, M_DONTWAIT, m->m_type);
1382 	if (n == NULL) {
1383 		m_freem(m);
1384 		return (NULL);
1385 	}
1386 
1387 	if (m_dup_pkthdr(n, m, M_DONTWAIT)) /* just for rcvif */
1388 		goto bad;
1389 
1390 	if (replylen > MHLEN) {
1391 		if (replylen > MCLBYTES) {
1392 			/*
1393 			 * XXX: should we try to allocate more? But MCLBYTES
1394 			 * is probably much larger than IPV6_MMTU...
1395 			 */
1396 			goto bad;
1397 		}
1398 		MCLGET(n, M_DONTWAIT);
1399 		if ((n->m_flags & M_EXT) == 0) {
1400 			goto bad;
1401 		}
1402 	}
1403 	n->m_pkthdr.len = n->m_len = replylen;
1404 
1405 	/* copy mbuf header and IPv6 + Node Information base headers */
1406 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1407 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1408 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1409 
1410 	/* qtype dependent procedure */
1411 	switch (qtype) {
1412 	case NI_QTYPE_NOOP:
1413 		nni6->ni_code = ICMP6_NI_SUCCESS;
1414 		nni6->ni_flags = 0;
1415 		break;
1416 	case NI_QTYPE_SUPTYPES:
1417 	{
1418 		u_int32_t v;
1419 		nni6->ni_code = ICMP6_NI_SUCCESS;
1420 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1421 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1422 		v = (u_int32_t)htonl(0x0000000f);
1423 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1424 		break;
1425 	}
1426 	case NI_QTYPE_FQDN:
1427 		nni6->ni_code = ICMP6_NI_SUCCESS;
1428 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1429 						sizeof(struct ip6_hdr) +
1430 						sizeof(struct icmp6_nodeinfo));
1431 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1432 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1433 		/*
1434 		 * XXX do we really have FQDN in variable "hostname"?
1435 		 */
1436 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1437 		if (n->m_next == NULL)
1438 			goto bad;
1439 		/* XXX we assume that n->m_next is not a chain */
1440 		if (n->m_next->m_next != NULL)
1441 			goto bad;
1442 		n->m_pkthdr.len += n->m_next->m_len;
1443 		break;
1444 	case NI_QTYPE_NODEADDR:
1445 	{
1446 		int lenlim, copied;
1447 
1448 		nni6->ni_code = ICMP6_NI_SUCCESS;
1449 		n->m_pkthdr.len = n->m_len =
1450 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1451 		lenlim = M_TRAILINGSPACE(n);
1452 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1453 		/* XXX: reset mbuf length */
1454 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1455 			sizeof(struct icmp6_nodeinfo) + copied;
1456 		break;
1457 	}
1458 	default:
1459 		break;		/* XXX impossible! */
1460 	}
1461 
1462 	nni6->ni_type = ICMP6_NI_REPLY;
1463 	m_freem(m);
1464 	return (n);
1465 
1466   bad:
1467 	m_freem(m);
1468 	if (n)
1469 		m_freem(n);
1470 	return (NULL);
1471 }
1472 #undef hostnamelen
1473 
1474 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1475 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1476 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1477 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1478 
1479 /*
1480  * make a mbuf with DNS-encoded string.  no compression support.
1481  *
1482  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1483  * treated as truncated name (two \0 at the end).  this is a wild guess.
1484  *
1485  * old - return pascal string if non-zero
1486  */
1487 struct mbuf *
1488 ni6_nametodns(const char *name, int namelen, int old)
1489 {
1490 	struct mbuf *m;
1491 	char *cp, *ep;
1492 	const char *p, *q;
1493 	int i, len, nterm;
1494 
1495 	if (old)
1496 		len = namelen + 1;
1497 	else
1498 		len = MCLBYTES;
1499 
1500 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1501 	MGET(m, M_DONTWAIT, MT_DATA);
1502 	if (m && len > MLEN) {
1503 		MCLGET(m, M_DONTWAIT);
1504 		if ((m->m_flags & M_EXT) == 0)
1505 			goto fail;
1506 	}
1507 	if (!m)
1508 		goto fail;
1509 	m->m_next = NULL;
1510 
1511 	if (old) {
1512 		m->m_len = len;
1513 		*mtod(m, char *) = namelen;
1514 		bcopy(name, mtod(m, char *) + 1, namelen);
1515 		return m;
1516 	} else {
1517 		m->m_len = 0;
1518 		cp = mtod(m, char *);
1519 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1520 
1521 		/* if not certain about my name, return empty buffer */
1522 		if (namelen == 0)
1523 			return m;
1524 
1525 		/*
1526 		 * guess if it looks like shortened hostname, or FQDN.
1527 		 * shortened hostname needs two trailing "\0".
1528 		 */
1529 		i = 0;
1530 		for (p = name; p < name + namelen; p++) {
1531 			if (*p && *p == '.')
1532 				i++;
1533 		}
1534 		if (i < 2)
1535 			nterm = 2;
1536 		else
1537 			nterm = 1;
1538 
1539 		p = name;
1540 		while (cp < ep && p < name + namelen) {
1541 			i = 0;
1542 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1543 				i++;
1544 			/* result does not fit into mbuf */
1545 			if (cp + i + 1 >= ep)
1546 				goto fail;
1547 			/*
1548 			 * DNS label length restriction, RFC1035 page 8.
1549 			 * "i == 0" case is included here to avoid returning
1550 			 * 0-length label on "foo..bar".
1551 			 */
1552 			if (i <= 0 || i >= 64)
1553 				goto fail;
1554 			*cp++ = i;
1555 			if (!isalnum(p[0]) || !isalnum(p[i - 1]))
1556 				goto fail;
1557 			while (i > 0) {
1558 				if (!isalnum(*p) && *p != '-')
1559 					goto fail;
1560 				if (isupper(*p)) {
1561 					*cp++ = tolower(*p);
1562 					p++;
1563 				} else
1564 					*cp++ = *p++;
1565 				i--;
1566 			}
1567 			p = q;
1568 			if (p < name + namelen && *p == '.')
1569 				p++;
1570 		}
1571 		/* termination */
1572 		if (cp + nterm >= ep)
1573 			goto fail;
1574 		while (nterm-- > 0)
1575 			*cp++ = '\0';
1576 		m->m_len = cp - mtod(m, char *);
1577 		return m;
1578 	}
1579 
1580 	panic("should not reach here");
1581 	/* NOTREACHED */
1582 
1583  fail:
1584 	if (m)
1585 		m_freem(m);
1586 	return NULL;
1587 }
1588 
1589 /*
1590  * check if two DNS-encoded string matches.  takes care of truncated
1591  * form (with \0\0 at the end).  no compression support.
1592  * XXX upper/lowercase match (see RFC2065)
1593  */
1594 int
1595 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1596 {
1597 	const char *a0, *b0;
1598 	int l;
1599 
1600 	/* simplest case - need validation? */
1601 	if (alen == blen && bcmp(a, b, alen) == 0)
1602 		return 1;
1603 
1604 	a0 = a;
1605 	b0 = b;
1606 
1607 	/* termination is mandatory */
1608 	if (alen < 2 || blen < 2)
1609 		return 0;
1610 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1611 		return 0;
1612 	alen--;
1613 	blen--;
1614 
1615 	while (a - a0 < alen && b - b0 < blen) {
1616 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1617 			return 0;
1618 
1619 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1620 			return 0;
1621 		/* we don't support compression yet */
1622 		if (a[0] >= 64 || b[0] >= 64)
1623 			return 0;
1624 
1625 		/* truncated case */
1626 		if (a[0] == 0 && a - a0 == alen - 1)
1627 			return 1;
1628 		if (b[0] == 0 && b - b0 == blen - 1)
1629 			return 1;
1630 		if (a[0] == 0 || b[0] == 0)
1631 			return 0;
1632 
1633 		if (a[0] != b[0])
1634 			return 0;
1635 		l = a[0];
1636 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1637 			return 0;
1638 		if (bcmp(a + 1, b + 1, l) != 0)
1639 			return 0;
1640 
1641 		a += 1 + l;
1642 		b += 1 + l;
1643 	}
1644 
1645 	if (a - a0 == alen && b - b0 == blen)
1646 		return 1;
1647 	else
1648 		return 0;
1649 }
1650 
1651 /*
1652  * calculate the number of addresses to be returned in the node info reply.
1653  */
1654 int
1655 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1656     char *subj)
1657 {
1658 	struct ifnet *ifp;
1659 	struct in6_ifaddr *ifa6;
1660 	struct ifaddr *ifa;
1661 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1662 	int addrs = 0, addrsofif, iffound = 0;
1663 	int niflags = ni6->ni_flags;
1664 
1665 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1666 		switch (ni6->ni_code) {
1667 		case ICMP6_NI_SUBJ_IPV6:
1668 			if (subj == NULL) /* must be impossible... */
1669 				return (0);
1670 			subj_ip6 = (struct sockaddr_in6 *)subj;
1671 			break;
1672 		default:
1673 			/*
1674 			 * XXX: we only support IPv6 subject address for
1675 			 * this Qtype.
1676 			 */
1677 			return (0);
1678 		}
1679 	}
1680 
1681 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1682 		addrsofif = 0;
1683 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1684 			if (ifa->ifa_addr->sa_family != AF_INET6)
1685 				continue;
1686 			ifa6 = (struct in6_ifaddr *)ifa;
1687 
1688 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1689 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1690 					       &ifa6->ia_addr.sin6_addr))
1691 				iffound = 1;
1692 
1693 			/*
1694 			 * IPv4-mapped addresses can only be returned by a
1695 			 * Node Information proxy, since they represent
1696 			 * addresses of IPv4-only nodes, which perforce do
1697 			 * not implement this protocol.
1698 			 * [icmp-name-lookups-07, Section 5.4]
1699 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1700 			 * this function at this moment.
1701 			 */
1702 
1703 			/* What do we have to do about ::1? */
1704 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1705 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1706 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1707 					continue;
1708 				break;
1709 			case IPV6_ADDR_SCOPE_SITELOCAL:
1710 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1711 					continue;
1712 				break;
1713 			case IPV6_ADDR_SCOPE_GLOBAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1715 					continue;
1716 				break;
1717 			default:
1718 				continue;
1719 			}
1720 
1721 			/*
1722 			 * check if anycast is okay.
1723 			 * XXX: just experimental.  not in the spec.
1724 			 */
1725 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1726 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1727 				continue; /* we need only unicast addresses */
1728 
1729 			addrsofif++; /* count the address */
1730 		}
1731 		if (iffound) {
1732 			*ifpp = ifp;
1733 			return (addrsofif);
1734 		}
1735 
1736 		addrs += addrsofif;
1737 	}
1738 
1739 	return (addrs);
1740 }
1741 
1742 int
1743 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1744     struct ifnet *ifp0, int resid)
1745 {
1746 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1747 	struct in6_ifaddr *ifa6;
1748 	struct ifaddr *ifa;
1749 	struct ifnet *ifp_dep = NULL;
1750 	int copied = 0, allow_deprecated = 0;
1751 	u_char *cp = (u_char *)(nni6 + 1);
1752 	int niflags = ni6->ni_flags;
1753 	u_int32_t ltime;
1754 
1755 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1756 		return (0);	/* needless to copy */
1757 
1758   again:
1759 
1760 	for (; ifp != TAILQ_END(&ifnet); ifp = TAILQ_NEXT(ifp, if_list)) {
1761 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1762 			if (ifa->ifa_addr->sa_family != AF_INET6)
1763 				continue;
1764 			ifa6 = (struct in6_ifaddr *)ifa;
1765 
1766 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1767 			    allow_deprecated == 0) {
1768 				/*
1769 				 * preferred address should be put before
1770 				 * deprecated addresses.
1771 				 */
1772 
1773 				/* record the interface for later search */
1774 				if (ifp_dep == NULL)
1775 					ifp_dep = ifp;
1776 
1777 				continue;
1778 			}
1779 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1780 				 allow_deprecated != 0)
1781 				continue; /* we now collect deprecated addrs */
1782 
1783 			/* What do we have to do about ::1? */
1784 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1785 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1786 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1787 					continue;
1788 				break;
1789 			case IPV6_ADDR_SCOPE_SITELOCAL:
1790 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1791 					continue;
1792 				break;
1793 			case IPV6_ADDR_SCOPE_GLOBAL:
1794 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1795 					continue;
1796 				break;
1797 			default:
1798 				continue;
1799 			}
1800 
1801 			/*
1802 			 * check if anycast is okay.
1803 			 * XXX: just experimental.  not in the spec.
1804 			 */
1805 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1806 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1807 				continue;
1808 
1809 			/* now we can copy the address */
1810 			if (resid < sizeof(struct in6_addr) +
1811 			    sizeof(u_int32_t)) {
1812 				/*
1813 				 * We give up much more copy.
1814 				 * Set the truncate flag and return.
1815 				 */
1816 				nni6->ni_flags |=
1817 					NI_NODEADDR_FLAG_TRUNCATE;
1818 				return (copied);
1819 			}
1820 
1821 			/*
1822 			 * Set the TTL of the address.
1823 			 * The TTL value should be one of the following
1824 			 * according to the specification:
1825 			 *
1826 			 * 1. The remaining lifetime of a DHCP lease on the
1827 			 *    address, or
1828 			 * 2. The remaining Valid Lifetime of a prefix from
1829 			 *    which the address was derived through Stateless
1830 			 *    Autoconfiguration.
1831 			 *
1832 			 * Note that we currently do not support stateful
1833 			 * address configuration by DHCPv6, so the former
1834 			 * case can't happen.
1835 			 *
1836 			 * TTL must be 2^31 > TTL >= 0.
1837 			 */
1838 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1839 				ltime = ND6_INFINITE_LIFETIME;
1840 			else {
1841 				if (ifa6->ia6_lifetime.ia6t_expire >
1842 				    time_second)
1843 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1844 				else
1845 					ltime = 0;
1846 			}
1847 
1848 			bcopy(&ltime, cp, sizeof(u_int32_t));
1849 			cp += sizeof(u_int32_t);
1850 
1851 			/* copy the address itself */
1852 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1853 			      sizeof(struct in6_addr));
1854 			/* XXX: KAME link-local hack; remove ifindex */
1855 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1856 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1857 			cp += sizeof(struct in6_addr);
1858 
1859 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1860 			copied += (sizeof(struct in6_addr) +
1861 				   sizeof(u_int32_t));
1862 		}
1863 		if (ifp0)	/* we need search only on the specified IF */
1864 			break;
1865 	}
1866 
1867 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1868 		ifp = ifp_dep;
1869 		allow_deprecated = 1;
1870 
1871 		goto again;
1872 	}
1873 
1874 	return (copied);
1875 }
1876 
1877 /*
1878  * XXX almost dup'ed code with rip6_input.
1879  */
1880 int
1881 icmp6_rip6_input(struct mbuf **mp, int off)
1882 {
1883 	struct mbuf *m = *mp;
1884 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1885 	struct in6pcb *in6p;
1886 	struct in6pcb *last = NULL;
1887 	struct sockaddr_in6 rip6src;
1888 	struct icmp6_hdr *icmp6;
1889 	struct mbuf *opts = NULL;
1890 
1891 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1892 	if (icmp6 == NULL) {
1893 		/* m is already reclaimed */
1894 		return IPPROTO_DONE;
1895 	}
1896 
1897 	bzero(&rip6src, sizeof(rip6src));
1898 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1899 	rip6src.sin6_family = AF_INET6;
1900 	/* KAME hack: recover scopeid */
1901 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1902 
1903 	CIRCLEQ_FOREACH(in6p, &rawin6pcbtable.inpt_queue, inp_queue) {
1904 		if (!(in6p->in6p_flags & INP_IPV6))
1905 			continue;
1906 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1907 			continue;
1908 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1909 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1910 			continue;
1911 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1912 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1913 			continue;
1914 		if (in6p->in6p_icmp6filt
1915 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1916 				 in6p->in6p_icmp6filt))
1917 			continue;
1918 		if (last) {
1919 			struct	mbuf *n;
1920 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1921 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1922 					ip6_savecontrol(last, n, &opts);
1923 				/* strip intermediate headers */
1924 				m_adj(n, off);
1925 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1926 						 (struct sockaddr *)&rip6src,
1927 						 n, opts) == 0) {
1928 					/* should notify about lost packet */
1929 					m_freem(n);
1930 					if (opts)
1931 						m_freem(opts);
1932 				} else
1933 					sorwakeup(last->in6p_socket);
1934 				opts = NULL;
1935 			}
1936 		}
1937 		last = in6p;
1938 	}
1939 	if (last) {
1940 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1941 			ip6_savecontrol(last, m, &opts);
1942 		/* strip intermediate headers */
1943 		m_adj(m, off);
1944 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1945 				 (struct sockaddr *)&rip6src,
1946 				 m, opts) == 0) {
1947 			m_freem(m);
1948 			if (opts)
1949 				m_freem(opts);
1950 		} else
1951 			sorwakeup(last->in6p_socket);
1952 	} else {
1953 		m_freem(m);
1954 		ip6stat.ip6s_delivered--;
1955 	}
1956 	return IPPROTO_DONE;
1957 }
1958 
1959 /*
1960  * Reflect the ip6 packet back to the source.
1961  * OFF points to the icmp6 header, counted from the top of the mbuf.
1962  *
1963  * Note: RFC 1885 required that an echo reply should be truncated if it
1964  * did not fit in with (return) path MTU, and KAME code supported the
1965  * behavior.  However, as a clarification after the RFC, this limitation
1966  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1967  * old behavior, with a (non-default) ifdef block, while the new version of
1968  * the spec was an internet-draft status, and even after the new RFC was
1969  * published.  But it would rather make sense to clean the obsoleted part
1970  * up, and to make the code simpler at this stage.
1971  */
1972 void
1973 icmp6_reflect(struct mbuf *m, size_t off)
1974 {
1975 	struct ip6_hdr *ip6;
1976 	struct icmp6_hdr *icmp6;
1977 	struct in6_ifaddr *ia;
1978 	struct in6_addr t, *src = 0;
1979 	int plen;
1980 	int type, code;
1981 	struct ifnet *outif = NULL;
1982 	struct sockaddr_in6 sa6_src, sa6_dst;
1983 
1984 	/* too short to reflect */
1985 	if (off < sizeof(struct ip6_hdr)) {
1986 		nd6log((LOG_DEBUG,
1987 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1988 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1989 		    __FILE__, __LINE__));
1990 		goto bad;
1991 	}
1992 
1993 	/*
1994 	 * If there are extra headers between IPv6 and ICMPv6, strip
1995 	 * off that header first.
1996 	 */
1997 #ifdef DIAGNOSTIC
1998 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1999 		panic("assumption failed in icmp6_reflect");
2000 #endif
2001 	if (off > sizeof(struct ip6_hdr)) {
2002 		size_t l;
2003 		struct ip6_hdr nip6;
2004 
2005 		l = off - sizeof(struct ip6_hdr);
2006 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2007 		m_adj(m, l);
2008 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2009 		if (m->m_len < l) {
2010 			if ((m = m_pullup(m, l)) == NULL)
2011 				return;
2012 		}
2013 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2014 	} else /* off == sizeof(struct ip6_hdr) */ {
2015 		size_t l;
2016 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2017 		if (m->m_len < l) {
2018 			if ((m = m_pullup(m, l)) == NULL)
2019 				return;
2020 		}
2021 	}
2022 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2023 	ip6 = mtod(m, struct ip6_hdr *);
2024 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2025 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2026 	type = icmp6->icmp6_type; /* keep type for statistics */
2027 	code = icmp6->icmp6_code; /* ditto. */
2028 
2029 	t = ip6->ip6_dst;
2030 	/*
2031 	 * ip6_input() drops a packet if its src is multicast.
2032 	 * So, the src is never multicast.
2033 	 */
2034 	ip6->ip6_dst = ip6->ip6_src;
2035 
2036 	/*
2037 	 * XXX: make sure to embed scope zone information, using
2038 	 * already embedded IDs or the received interface (if any).
2039 	 * Note that rcvif may be NULL.
2040 	 * TODO: scoped routing case (XXX).
2041 	 */
2042 	bzero(&sa6_src, sizeof(sa6_src));
2043 	sa6_src.sin6_family = AF_INET6;
2044 	sa6_src.sin6_len = sizeof(sa6_src);
2045 	sa6_src.sin6_addr = ip6->ip6_dst;
2046 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2047 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2048 	bzero(&sa6_dst, sizeof(sa6_dst));
2049 	sa6_dst.sin6_family = AF_INET6;
2050 	sa6_dst.sin6_len = sizeof(sa6_dst);
2051 	sa6_dst.sin6_addr = t;
2052 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2053 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2054 
2055 	/*
2056 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2057 	 * use dst as the src for the reply.
2058 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2059 	 * (for example) when we encounter an error while forwarding procedure
2060 	 * destined to a duplicated address of ours.
2061 	 */
2062 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2063 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2064 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2065 			src = &t;
2066 			break;
2067 		}
2068 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2069 		/*
2070 		 * This is the case if the dst is our link-local address
2071 		 * and the sender is also ourselves.
2072 		 */
2073 		src = &t;
2074 	}
2075 
2076 	if (src == 0) {
2077 		int e;
2078 		struct route_in6 ro;
2079 
2080 		/*
2081 		 * This case matches to multicasts, our anycast, or unicasts
2082 		 * that we do not own.  Select a source address based on the
2083 		 * source address of the erroneous packet.
2084 		 */
2085 		bzero(&ro, sizeof(ro));
2086 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2087 		if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2088 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2089 		}
2090 		if (src == NULL) {
2091 			nd6log((LOG_DEBUG,
2092 			    "icmp6_reflect: source can't be determined: "
2093 			    "dst=%s, error=%d\n",
2094 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2095 			goto bad;
2096 		}
2097 	}
2098 
2099 	ip6->ip6_src = *src;
2100 
2101 	ip6->ip6_flow = 0;
2102 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2103 	ip6->ip6_vfc |= IPV6_VERSION;
2104 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2105 	if (m->m_pkthdr.rcvif) {
2106 		/* XXX: This may not be the outgoing interface */
2107 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2108 	} else
2109 		ip6->ip6_hlim = ip6_defhlim;
2110 
2111 	icmp6->icmp6_cksum = 0;
2112 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2113 					sizeof(struct ip6_hdr), plen);
2114 
2115 	/*
2116 	 * XXX option handling
2117 	 */
2118 
2119 	m->m_flags &= ~(M_BCAST|M_MCAST);
2120 
2121 	/*
2122 	 * To avoid a "too big" situation at an intermediate router
2123 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2124 	 * Note that only echo and node information replies are affected,
2125 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2126 	 */
2127 #if NPF > 0
2128 	pf_pkt_addr_changed(m);
2129 #endif
2130 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif, NULL) != 0 &&
2131 	    outif)
2132 		icmp6_ifstat_inc(outif, ifs6_out_error);
2133 
2134 	if (outif)
2135 		icmp6_ifoutstat_inc(outif, type, code);
2136 
2137 	return;
2138 
2139  bad:
2140 	m_freem(m);
2141 	return;
2142 }
2143 
2144 void
2145 icmp6_fasttimo(void)
2146 {
2147 
2148 	mld6_fasttimeo();
2149 }
2150 
2151 const char *
2152 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2153     struct in6_addr *tgt6)
2154 {
2155 	static char buf[1024];
2156 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2157 		 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2158 	return buf;
2159 }
2160 
2161 void
2162 icmp6_redirect_input(struct mbuf *m, int off)
2163 {
2164 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2165 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2166 	struct nd_redirect *nd_rd;
2167 	int icmp6len = ntohs(ip6->ip6_plen);
2168 	char *lladdr = NULL;
2169 	int lladdrlen = 0;
2170 	struct rtentry *rt = NULL;
2171 	int is_router;
2172 	int is_onlink;
2173 	struct in6_addr src6 = ip6->ip6_src;
2174 	struct in6_addr redtgt6;
2175 	struct in6_addr reddst6;
2176 	union nd_opts ndopts;
2177 
2178 	if (!ifp)
2179 		return;
2180 
2181 	/* XXX if we are router, we don't update route by icmp6 redirect */
2182 	if (ip6_forwarding)
2183 		goto freeit;
2184 	if (!icmp6_rediraccept)
2185 		goto freeit;
2186 
2187 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2188 	if (nd_rd == NULL) {
2189 		icmp6stat.icp6s_tooshort++;
2190 		return;
2191 	}
2192 	redtgt6 = nd_rd->nd_rd_target;
2193 	reddst6 = nd_rd->nd_rd_dst;
2194 
2195 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2196 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2197 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2198 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2199 
2200 	/* validation */
2201 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2202 		nd6log((LOG_ERR,
2203 			"ICMP6 redirect sent from %s rejected; "
2204 			"must be from linklocal\n", ip6_sprintf(&src6)));
2205 		goto bad;
2206 	}
2207 	if (ip6->ip6_hlim != 255) {
2208 		nd6log((LOG_ERR,
2209 			"ICMP6 redirect sent from %s rejected; "
2210 			"hlim=%d (must be 255)\n",
2211 			ip6_sprintf(&src6), ip6->ip6_hlim));
2212 		goto bad;
2213 	}
2214     {
2215 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2216 	struct sockaddr_in6 sin6;
2217 	struct in6_addr *gw6;
2218 
2219 	bzero(&sin6, sizeof(sin6));
2220 	sin6.sin6_family = AF_INET6;
2221 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2222 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2223 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0);
2224 	if (rt) {
2225 		if (rt->rt_gateway == NULL ||
2226 		    rt->rt_gateway->sa_family != AF_INET6) {
2227 			nd6log((LOG_ERR,
2228 			    "ICMP6 redirect rejected; no route "
2229 			    "with inet6 gateway found for redirect dst: %s\n",
2230 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2231 			RTFREE(rt);
2232 			goto bad;
2233 		}
2234 
2235 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2236 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2237 			nd6log((LOG_ERR,
2238 				"ICMP6 redirect rejected; "
2239 				"not equal to gw-for-src=%s (must be same): "
2240 				"%s\n",
2241 				ip6_sprintf(gw6),
2242 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2243 			RTFREE(rt);
2244 			goto bad;
2245 		}
2246 	} else {
2247 		nd6log((LOG_ERR,
2248 			"ICMP6 redirect rejected; "
2249 			"no route found for redirect dst: %s\n",
2250 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2251 		goto bad;
2252 	}
2253 	RTFREE(rt);
2254 	rt = NULL;
2255     }
2256 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2257 		nd6log((LOG_ERR,
2258 			"ICMP6 redirect rejected; "
2259 			"redirect dst must be unicast: %s\n",
2260 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2261 		goto bad;
2262 	}
2263 
2264 	is_router = is_onlink = 0;
2265 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2266 		is_router = 1;	/* router case */
2267 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2268 		is_onlink = 1;	/* on-link destination case */
2269 	if (!is_router && !is_onlink) {
2270 		nd6log((LOG_ERR,
2271 			"ICMP6 redirect rejected; "
2272 			"neither router case nor onlink case: %s\n",
2273 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2274 		goto bad;
2275 	}
2276 	/* validation passed */
2277 
2278 	icmp6len -= sizeof(*nd_rd);
2279 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2280 	if (nd6_options(&ndopts) < 0) {
2281 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2282 			"invalid ND option, rejected: %s\n",
2283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		/* nd6_options have incremented stats */
2285 		goto freeit;
2286 	}
2287 
2288 	if (ndopts.nd_opts_tgt_lladdr) {
2289 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2290 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2291 	}
2292 
2293 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2294 		nd6log((LOG_INFO,
2295 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2296 			"(if %d, icmp6 packet %d): %s\n",
2297 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2298 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299 		goto bad;
2300 	}
2301 
2302 	/* RFC 2461 8.3 */
2303 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2304 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2305 
2306 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2307 		/* perform rtredirect */
2308 		struct sockaddr_in6 sdst;
2309 		struct sockaddr_in6 sgw;
2310 		struct sockaddr_in6 ssrc;
2311 		unsigned long rtcount;
2312 		struct rtentry *newrt = NULL;
2313 
2314 		/*
2315 		 * do not install redirect route, if the number of entries
2316 		 * is too much (> hiwat).  note that, the node (= host) will
2317 		 * work just fine even if we do not install redirect route
2318 		 * (there will be additional hops, though).
2319 		 */
2320 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2321 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
2322 			goto freeit;
2323 		else if (0 <= icmp6_redirect_lowat &&
2324 		    rtcount > icmp6_redirect_lowat) {
2325 			/*
2326 			 * XXX nuke a victim, install the new one.
2327 			 */
2328 		}
2329 
2330 		bzero(&sdst, sizeof(sdst));
2331 		bzero(&sgw, sizeof(sgw));
2332 		bzero(&ssrc, sizeof(ssrc));
2333 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2334 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2335 			sizeof(struct sockaddr_in6);
2336 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2337 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2338 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2339 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2340 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2341 			   (struct sockaddr *)&ssrc,
2342 			   &newrt, /* XXX */ 0);
2343 
2344 		if (newrt) {
2345 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2346 			    icmp6_redirect_timeout_q, /* XXX */ 0);
2347 			rtfree(newrt);
2348 		}
2349 	}
2350 	/* finally update cached route in each socket via pfctlinput */
2351 	{
2352 		struct sockaddr_in6 sdst;
2353 
2354 		bzero(&sdst, sizeof(sdst));
2355 		sdst.sin6_family = AF_INET6;
2356 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2357 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2358 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2359 	}
2360 
2361  freeit:
2362 	m_freem(m);
2363 	return;
2364 
2365  bad:
2366 	icmp6stat.icp6s_badredirect++;
2367 	m_freem(m);
2368 }
2369 
2370 void
2371 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2372 {
2373 	struct ifnet *ifp;	/* my outgoing interface */
2374 	struct in6_addr *ifp_ll6;
2375 	struct in6_addr *nexthop;
2376 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2377 	struct mbuf *m = NULL;	/* newly allocated one */
2378 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2379 	struct nd_redirect *nd_rd;
2380 	size_t maxlen;
2381 	u_char *p;
2382 	struct sockaddr_in6 src_sa;
2383 
2384 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2385 
2386 	/* if we are not router, we don't send icmp6 redirect */
2387 	if (!ip6_forwarding)
2388 		goto fail;
2389 
2390 	/* sanity check */
2391 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2392 		goto fail;
2393 
2394 	/*
2395 	 * Address check:
2396 	 *  the source address must identify a neighbor, and
2397 	 *  the destination address must not be a multicast address
2398 	 *  [RFC 2461, sec 8.2]
2399 	 */
2400 	sip6 = mtod(m0, struct ip6_hdr *);
2401 	bzero(&src_sa, sizeof(src_sa));
2402 	src_sa.sin6_family = AF_INET6;
2403 	src_sa.sin6_len = sizeof(src_sa);
2404 	src_sa.sin6_addr = sip6->ip6_src;
2405 	/* we don't currently use sin6_scope_id, but eventually use it */
2406 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2407 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2408 		goto fail;
2409 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2410 		goto fail;	/* what should we do here? */
2411 
2412 	/* rate limit */
2413 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2414 		goto fail;
2415 
2416 	/*
2417 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2418 	 * we almost always ask for an mbuf cluster for simplicity.
2419 	 * (MHLEN < IPV6_MMTU is almost always true)
2420 	 */
2421 #if IPV6_MMTU >= MCLBYTES
2422 # error assumption failed about IPV6_MMTU and MCLBYTES
2423 #endif
2424 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2425 	if (m && IPV6_MMTU >= MHLEN)
2426 		MCLGET(m, M_DONTWAIT);
2427 	if (!m)
2428 		goto fail;
2429 	m->m_pkthdr.rcvif = NULL;
2430 	m->m_len = 0;
2431 	maxlen = M_TRAILINGSPACE(m);
2432 	maxlen = min(IPV6_MMTU, maxlen);
2433 	/* just for safety */
2434 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2435 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2436 		goto fail;
2437 	}
2438 
2439 	{
2440 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2441 		struct in6_ifaddr *ia;
2442 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2443 						 IN6_IFF_NOTREADY|
2444 						 IN6_IFF_ANYCAST)) == NULL)
2445 			goto fail;
2446 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2447 	}
2448 
2449 	/* get ip6 linklocal address for the router. */
2450 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2451 		struct sockaddr_in6 *sin6;
2452 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2453 		nexthop = &sin6->sin6_addr;
2454 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2455 			nexthop = NULL;
2456 	} else
2457 		nexthop = NULL;
2458 
2459 	/* ip6 */
2460 	ip6 = mtod(m, struct ip6_hdr *);
2461 	ip6->ip6_flow = 0;
2462 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2463 	ip6->ip6_vfc |= IPV6_VERSION;
2464 	/* ip6->ip6_plen will be set later */
2465 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2466 	ip6->ip6_hlim = 255;
2467 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2468 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2469 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2470 
2471 	/* ND Redirect */
2472 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2473 	nd_rd->nd_rd_type = ND_REDIRECT;
2474 	nd_rd->nd_rd_code = 0;
2475 	nd_rd->nd_rd_reserved = 0;
2476 	if (rt->rt_flags & RTF_GATEWAY) {
2477 		/*
2478 		 * nd_rd->nd_rd_target must be a link-local address in
2479 		 * better router cases.
2480 		 */
2481 		if (!nexthop)
2482 			goto fail;
2483 		bcopy(nexthop, &nd_rd->nd_rd_target,
2484 		      sizeof(nd_rd->nd_rd_target));
2485 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2486 		      sizeof(nd_rd->nd_rd_dst));
2487 	} else {
2488 		/* make sure redtgt == reddst */
2489 		nexthop = &sip6->ip6_dst;
2490 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2491 		      sizeof(nd_rd->nd_rd_target));
2492 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2493 		      sizeof(nd_rd->nd_rd_dst));
2494 	}
2495 
2496 	p = (u_char *)(nd_rd + 1);
2497 
2498 	{
2499 		/* target lladdr option */
2500 		struct rtentry *rt_nexthop = NULL;
2501 		int len;
2502 		struct sockaddr_dl *sdl;
2503 		struct nd_opt_hdr *nd_opt;
2504 		char *lladdr;
2505 
2506 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2507 		if (!rt_nexthop)
2508 			goto nolladdropt;
2509 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2510 		len = (len + 7) & ~7;	/* round by 8 */
2511 		/* safety check */
2512 		if (len + (p - (u_char *)ip6) > maxlen)
2513 			goto nolladdropt;
2514 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2515 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2516 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2517 		    (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2518 		    sdl->sdl_alen) {
2519 			nd_opt = (struct nd_opt_hdr *)p;
2520 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2521 			nd_opt->nd_opt_len = len >> 3;
2522 			lladdr = (char *)(nd_opt + 1);
2523 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2524 			p += len;
2525 		}
2526 	}
2527   nolladdropt:;
2528 
2529 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2530 
2531 	/* just to be safe */
2532 	if (p - (u_char *)ip6 > maxlen)
2533 		goto noredhdropt;
2534 
2535 	{
2536 		/* redirected header option */
2537 		int len;
2538 		struct nd_opt_rd_hdr *nd_opt_rh;
2539 
2540 		/*
2541 		 * compute the maximum size for icmp6 redirect header option.
2542 		 * XXX room for auth header?
2543 		 */
2544 		len = maxlen - (p - (u_char *)ip6);
2545 		len &= ~7;
2546 
2547 		/*
2548 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2549 		 * about padding/truncate rule for the original IP packet.
2550 		 * From the discussion on IPv6imp in Feb 1999,
2551 		 * the consensus was:
2552 		 * - "attach as much as possible" is the goal
2553 		 * - pad if not aligned (original size can be guessed by
2554 		 *   original ip6 header)
2555 		 * Following code adds the padding if it is simple enough,
2556 		 * and truncates if not.
2557 		 */
2558 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2559 			/* not enough room, truncate */
2560 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2561 			    m0->m_pkthdr.len);
2562 		} else {
2563 			/*
2564 			 * enough room, truncate if not aligned.
2565 			 * we don't pad here for simplicity.
2566 			 */
2567 			size_t extra;
2568 
2569 			extra = m0->m_pkthdr.len % 8;
2570 			if (extra) {
2571 				/* truncate */
2572 				m_adj(m0, -extra);
2573 			}
2574 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2575 		}
2576 
2577 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2578 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2579 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2580 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2581 		p += sizeof(*nd_opt_rh);
2582 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2583 
2584 		/* connect m0 to m */
2585 		m->m_pkthdr.len += m0->m_pkthdr.len;
2586 		m_cat(m, m0);
2587 		m0 = NULL;
2588 	}
2589 noredhdropt:
2590 	if (m0) {
2591 		m_freem(m0);
2592 		m0 = NULL;
2593 	}
2594 
2595 	sip6 = mtod(m, struct ip6_hdr *);
2596 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2597 		sip6->ip6_src.s6_addr16[1] = 0;
2598 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2599 		sip6->ip6_dst.s6_addr16[1] = 0;
2600 #if 0
2601 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2602 		ip6->ip6_src.s6_addr16[1] = 0;
2603 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2604 		ip6->ip6_dst.s6_addr16[1] = 0;
2605 #endif
2606 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2607 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2608 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2609 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2610 
2611 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2612 
2613 	nd_rd->nd_rd_cksum = 0;
2614 	nd_rd->nd_rd_cksum
2615 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2616 
2617 	/* send the packet to outside... */
2618 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2619 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2620 
2621 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2622 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2623 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2624 
2625 	return;
2626 
2627 fail:
2628 	if (m)
2629 		m_freem(m);
2630 	if (m0)
2631 		m_freem(m0);
2632 }
2633 
2634 /* NRL PCB */
2635 #define sotoin6pcb	sotoinpcb
2636 #define in6pcb		inpcb
2637 #define in6p_icmp6filt	inp_icmp6filt
2638 
2639 /*
2640  * ICMPv6 socket option processing.
2641  */
2642 int
2643 icmp6_ctloutput(int op, struct socket *so, int level, int optname,
2644     struct mbuf **mp)
2645 {
2646 	int error = 0;
2647 	struct in6pcb *in6p = sotoin6pcb(so);
2648 	struct mbuf *m = *mp;
2649 
2650 	if (level != IPPROTO_ICMPV6) {
2651 		if (op == PRCO_SETOPT && m)
2652 			(void)m_free(m);
2653 		return EINVAL;
2654 	}
2655 
2656 	switch (op) {
2657 	case PRCO_SETOPT:
2658 		switch (optname) {
2659 		case ICMP6_FILTER:
2660 		    {
2661 			struct icmp6_filter *p;
2662 
2663 			if (m == NULL || m->m_len != sizeof(*p)) {
2664 				error = EMSGSIZE;
2665 				break;
2666 			}
2667 			p = mtod(m, struct icmp6_filter *);
2668 			if (!p || !in6p->in6p_icmp6filt) {
2669 				error = EINVAL;
2670 				break;
2671 			}
2672 			bcopy(p, in6p->in6p_icmp6filt,
2673 				sizeof(struct icmp6_filter));
2674 			error = 0;
2675 			break;
2676 		    }
2677 
2678 		default:
2679 			error = ENOPROTOOPT;
2680 			break;
2681 		}
2682 		if (m)
2683 			m_freem(m);
2684 		break;
2685 
2686 	case PRCO_GETOPT:
2687 		switch (optname) {
2688 		case ICMP6_FILTER:
2689 		    {
2690 			struct icmp6_filter *p;
2691 
2692 			if (!in6p->in6p_icmp6filt) {
2693 				error = EINVAL;
2694 				break;
2695 			}
2696 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2697 			m->m_len = sizeof(struct icmp6_filter);
2698 			p = mtod(m, struct icmp6_filter *);
2699 			bcopy(in6p->in6p_icmp6filt, p,
2700 				sizeof(struct icmp6_filter));
2701 			error = 0;
2702 			break;
2703 		    }
2704 
2705 		default:
2706 			error = ENOPROTOOPT;
2707 			break;
2708 		}
2709 		break;
2710 	}
2711 
2712 	return (error);
2713 }
2714 
2715 /* NRL PCB */
2716 #undef sotoin6pcb
2717 #undef in6pcb
2718 #undef in6p_icmp6filt
2719 
2720 /*
2721  * Perform rate limit check.
2722  * Returns 0 if it is okay to send the icmp6 packet.
2723  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2724  * limitation.
2725  *
2726  * XXX per-destination/type check necessary?
2727  *
2728  * dst - not used at this moment
2729  * type - not used at this moment
2730  * code - not used at this moment
2731  */
2732 int
2733 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
2734 {
2735 	int ret;
2736 
2737 	ret = 0;	/* okay to send */
2738 
2739 	/* PPS limit */
2740 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2741 	    icmp6errppslim)) {
2742 		/* The packet is subject to rate limit */
2743 		ret++;
2744 	}
2745 
2746 	return ret;
2747 }
2748 
2749 struct rtentry *
2750 icmp6_mtudisc_clone(struct sockaddr *dst)
2751 {
2752 	struct rtentry *rt;
2753 	int    error;
2754 
2755 	rt = rtalloc1(dst, RT_REPORT, 0);
2756 	if (rt == 0)
2757 		return NULL;
2758 
2759 	/* If we didn't get a host route, allocate one */
2760 	if ((rt->rt_flags & RTF_HOST) == 0) {
2761 		struct rt_addrinfo info;
2762 		struct rtentry *nrt;
2763 
2764 		bzero(&info, sizeof(info));
2765 		info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC;
2766 		info.rti_info[RTAX_DST] = dst;
2767 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
2768 		error = rtrequest1(RTM_ADD, &info, rt->rt_priority, &nrt, 0);
2769 		if (error) {
2770 			rtfree(rt);
2771 			return NULL;
2772 		}
2773 		nrt->rt_rmx = rt->rt_rmx;
2774 		rtfree(rt);
2775 		rt = nrt;
2776 	}
2777 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2778 			icmp6_mtudisc_timeout_q, /* XXX */ 0);
2779 	if (error) {
2780 		rtfree(rt);
2781 		return NULL;
2782 	}
2783 
2784 	return rt;	/* caller need to call rtfree() */
2785 }
2786 
2787 void
2788 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2789 {
2790 	if (rt == NULL)
2791 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2792 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2793 	    (RTF_DYNAMIC | RTF_HOST)) {
2794 		struct rt_addrinfo info;
2795 
2796 		bzero(&info, sizeof(info));
2797 		info.rti_flags = rt->rt_flags;
2798 		info.rti_info[RTAX_DST] = rt_key(rt);
2799 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
2800 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
2801 		rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL,
2802 		    r->rtt_tableid);
2803 	} else {
2804 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2805 			rt->rt_rmx.rmx_mtu = 0;
2806 	}
2807 }
2808 
2809 void
2810 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2811 {
2812 	if (rt == NULL)
2813 		panic("icmp6_redirect_timeout: bad route to timeout");
2814 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2815 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2816 		struct rt_addrinfo info;
2817 
2818 		bzero(&info, sizeof(info));
2819 		info.rti_flags = rt->rt_flags;
2820 		info.rti_info[RTAX_DST] = rt_key(rt);
2821 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
2822 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
2823 		rtrequest1(RTM_DELETE, &info, rt->rt_priority, NULL,
2824 		    r->rtt_tableid);
2825 	}
2826 }
2827 
2828 #include <uvm/uvm_extern.h>
2829 #include <sys/proc.h>
2830 #include <sys/sysctl.h>
2831 
2832 int *icmpv6ctl_vars[ICMPV6CTL_MAXID] = ICMPV6CTL_VARS;
2833 
2834 int
2835 icmp6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
2836     void *newp, size_t newlen)
2837 {
2838 	/* All sysctl names at this level are terminal. */
2839 	if (namelen != 1)
2840 		return ENOTDIR;
2841 
2842 	switch (name[0]) {
2843 
2844 	case ICMPV6CTL_STATS:
2845 		return sysctl_rdstruct(oldp, oldlenp, newp,
2846 				&icmp6stat, sizeof(icmp6stat));
2847 	case ICMPV6CTL_ND6_DRLIST:
2848 	case ICMPV6CTL_ND6_PRLIST:
2849 		return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen);
2850 	default:
2851 		if (name[0] < ICMPV6CTL_MAXID)
2852 			return (sysctl_int_arr(icmpv6ctl_vars, name, namelen,
2853 			    oldp, oldlenp, newp, newlen));
2854 		return ENOPROTOOPT;
2855 	}
2856 	/* NOTREACHED */
2857 }
2858