xref: /netbsd-src/sys/netinet6/icmp6.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: icmp6.c,v 1.196 2016/08/19 12:26:01 roy 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 <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.196 2016/08/19 12:26:01 roy Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_ipsec.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/kmem.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 #include <sys/domain.h>
83 #include <sys/sysctl.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/ip6_private.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/icmp6_private.h>
97 #include <netinet6/mld6_var.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/ip6protosw.h>
102 #include <netinet6/scope6_var.h>
103 
104 #ifdef IPSEC
105 #include <netipsec/ipsec.h>
106 #include <netipsec/key.h>
107 #endif
108 
109 
110 #include "faith.h"
111 #if defined(NFAITH) && 0 < NFAITH
112 #include <net/if_faith.h>
113 #endif
114 
115 #include <net/net_osdep.h>
116 
117 extern struct domain inet6domain;
118 
119 percpu_t *icmp6stat_percpu;
120 
121 extern struct inpcbtable raw6cbtable;
122 extern int icmp6errppslim;
123 static int icmp6errpps_count = 0;
124 static struct timeval icmp6errppslim_last;
125 extern int icmp6_nodeinfo;
126 
127 /*
128  * List of callbacks to notify when Path MTU changes are made.
129  */
130 struct icmp6_mtudisc_callback {
131 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
132 	void (*mc_func)(struct in6_addr *);
133 };
134 
135 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
136     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
137 
138 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
139 extern int pmtu_expire;
140 
141 /* XXX do these values make any sense? */
142 static int icmp6_mtudisc_hiwat = 1280;
143 static int icmp6_mtudisc_lowat = 256;
144 
145 /*
146  * keep track of # of redirect routes.
147  */
148 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
149 
150 /* XXX experimental, turned off */
151 static int icmp6_redirect_hiwat = -1;
152 static int icmp6_redirect_lowat = -1;
153 
154 static void icmp6_errcount(u_int, int, int);
155 static int icmp6_rip6_input(struct mbuf **, int);
156 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
157 static const char *icmp6_redirect_diag(struct in6_addr *,
158 	struct in6_addr *, struct in6_addr *);
159 static struct mbuf *ni6_input(struct mbuf *, int);
160 static struct mbuf *ni6_nametodns(const char *, int, int);
161 static int ni6_dnsmatch(const char *, int, const char *, int);
162 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
163 			  struct ifnet **, char *, struct psref *);
164 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
165 				struct ifnet *, int);
166 static int icmp6_notify_error(struct mbuf *, int, int, int);
167 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
168 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
169 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
170 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
171 
172 
173 void
174 icmp6_init(void)
175 {
176 
177 	sysctl_net_inet6_icmp6_setup(NULL);
178 	mld_init();
179 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
180 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
181 
182 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
183 }
184 
185 static void
186 icmp6_errcount(u_int base, int type, int code)
187 {
188 	switch (type) {
189 	case ICMP6_DST_UNREACH:
190 		switch (code) {
191 		case ICMP6_DST_UNREACH_NOROUTE:
192 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
193 			return;
194 		case ICMP6_DST_UNREACH_ADMIN:
195 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
196 			return;
197 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
198 			ICMP6_STATINC(base +
199 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			ICMP6_STATINC(base +
219 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
220 			return;
221 		}
222 		break;
223 	case ICMP6_PARAM_PROB:
224 		switch (code) {
225 		case ICMP6_PARAMPROB_HEADER:
226 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
227 			return;
228 		case ICMP6_PARAMPROB_NEXTHEADER:
229 			ICMP6_STATINC(base +
230 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
231 			return;
232 		case ICMP6_PARAMPROB_OPTION:
233 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
234 			return;
235 		}
236 		break;
237 	case ND_REDIRECT:
238 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
239 		return;
240 	}
241 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
242 }
243 
244 /*
245  * Register a Path MTU Discovery callback.
246  */
247 void
248 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
249 {
250 	struct icmp6_mtudisc_callback *mc;
251 
252 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
253 	     mc = LIST_NEXT(mc, mc_list)) {
254 		if (mc->mc_func == func)
255 			return;
256 	}
257 
258 	mc = kmem_alloc(sizeof(*mc), KM_SLEEP);
259 	mc->mc_func = func;
260 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
261 }
262 
263 /*
264  * A wrapper function for icmp6_error() necessary when the erroneous packet
265  * may not contain enough scope zone information.
266  */
267 void
268 icmp6_error2(struct mbuf *m, int type, int code, int param,
269 	struct ifnet *ifp)
270 {
271 	struct ip6_hdr *ip6;
272 
273 	if (ifp == NULL)
274 		return;
275 
276 	if (m->m_len < sizeof(struct ip6_hdr)) {
277 		m = m_pullup(m, sizeof(struct ip6_hdr));
278 		if (m == NULL)
279 			return;
280 	}
281 
282 	ip6 = mtod(m, struct ip6_hdr *);
283 
284 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
285 		return;
286 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
287 		return;
288 
289 	icmp6_error(m, type, code, param);
290 }
291 
292 /*
293  * Generate an error packet of type error in response to bad IP6 packet.
294  */
295 void
296 icmp6_error(struct mbuf *m, int type, int code, int param)
297 {
298 	struct ip6_hdr *oip6, *nip6;
299 	struct icmp6_hdr *icmp6;
300 	u_int preplen;
301 	int off;
302 	int nxt;
303 
304 	ICMP6_STATINC(ICMP6_STAT_ERROR);
305 
306 	/* count per-type-code statistics */
307 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
308 
309 	if (m->m_flags & M_DECRYPTED) {
310 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
311 		goto freeit;
312 	}
313 
314 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
315 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
316 		return;
317 	oip6 = mtod(m, struct ip6_hdr *);
318 
319 	/*
320 	 * If the destination address of the erroneous packet is a multicast
321 	 * address, or the packet was sent using link-layer multicast,
322 	 * we should basically suppress sending an error (RFC 2463, Section
323 	 * 2.4).
324 	 * We have two exceptions (the item e.2 in that section):
325 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
326 	 * - the Parameter Problem Message that can be allowed an icmp6 error
327 	 *   in the option type field.  This check has been done in
328 	 *   ip6_unknown_opt(), so we can just check the type and code.
329 	 */
330 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
331 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
332 	    (type != ICMP6_PACKET_TOO_BIG &&
333 	     (type != ICMP6_PARAM_PROB ||
334 	      code != ICMP6_PARAMPROB_OPTION)))
335 		goto freeit;
336 
337 	/*
338 	 * RFC 2463, 2.4 (e.5): source address check.
339 	 * XXX: the case of anycast source?
340 	 */
341 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
342 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
343 		goto freeit;
344 
345 	/*
346 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
347 	 * don't do it.
348 	 */
349 	nxt = -1;
350 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
351 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
352 		struct icmp6_hdr *icp;
353 
354 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
355 			sizeof(*icp));
356 		if (icp == NULL) {
357 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
358 			return;
359 		}
360 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
361 		    icp->icmp6_type == ND_REDIRECT) {
362 			/*
363 			 * ICMPv6 error
364 			 * Special case: for redirect (which is
365 			 * informational) we must not send icmp6 error.
366 			 */
367 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
368 			goto freeit;
369 		} else {
370 			/* ICMPv6 informational - send the error */
371 		}
372 	}
373 #if 0 /* controversial */
374 	else if (off >= 0 && nxt == IPPROTO_ESP) {
375 		/*
376 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
377 		 * don't respond.
378 		 */
379 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
380 		goto freeit;
381 	}
382 #endif
383 	else {
384 		/* non-ICMPv6 - send the error */
385 	}
386 
387 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
388 
389 	/* Finally, do rate limitation check. */
390 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
391 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
392 		goto freeit;
393 	}
394 
395 	/*
396 	 * OK, ICMP6 can be generated.
397 	 */
398 
399 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
400 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
401 
402 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
403 	M_PREPEND(m, preplen, M_DONTWAIT);
404 	if (m && M_UNWRITABLE(m, preplen))
405 		m = m_pullup(m, preplen);
406 	if (m == NULL) {
407 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
408 		return;
409 	}
410 
411 	nip6 = mtod(m, struct ip6_hdr *);
412 	nip6->ip6_src  = oip6->ip6_src;
413 	nip6->ip6_dst  = oip6->ip6_dst;
414 
415 	in6_clearscope(&oip6->ip6_src);
416 	in6_clearscope(&oip6->ip6_dst);
417 
418 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
419 	icmp6->icmp6_type = type;
420 	icmp6->icmp6_code = code;
421 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
422 
423 	/*
424 	 * icmp6_reflect() is designed to be in the input path.
425 	 * icmp6_error() can be called from both input and output path,
426 	 * and if we are in output path rcvif could contain bogus value.
427 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
428 	 * information in ip header (nip6).
429 	 */
430 	m_reset_rcvif(m);
431 
432 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
433 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
434 
435 	return;
436 
437   freeit:
438 	/*
439 	 * If we can't tell whether or not we can generate ICMP6, free it.
440 	 */
441 	m_freem(m);
442 }
443 
444 /*
445  * Process a received ICMP6 message.
446  */
447 int
448 icmp6_input(struct mbuf **mp, int *offp, int proto)
449 {
450 	struct mbuf *m = *mp, *n;
451 	struct ip6_hdr *ip6, *nip6;
452 	struct icmp6_hdr *icmp6, *nicmp6;
453 	int off = *offp;
454 	int icmp6len = m->m_pkthdr.len - *offp;
455 	int code, sum, noff;
456 	struct ifnet *rcvif;
457 	struct psref psref;
458 
459 	rcvif = m_get_rcvif_psref(m, &psref);
460 	if (__predict_false(rcvif == NULL))
461 		goto freeit;
462 
463 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
464 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
465 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
466 
467 	/*
468 	 * Locate icmp6 structure in mbuf, and check
469 	 * that not corrupted and of at least minimum length
470 	 */
471 
472 	if (icmp6len < sizeof(struct icmp6_hdr)) {
473 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
474 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
475 		goto freeit;
476 	}
477 
478 	ip6 = mtod(m, struct ip6_hdr *);
479 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
480 	if (icmp6 == NULL) {
481 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
482 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
483 		goto freeit;
484 	}
485 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
486 
487 	/*
488 	 * calculate the checksum
489 	 */
490 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
491 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
492 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src));
493 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
494 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
495 		goto freeit;
496 	}
497 
498 #if defined(NFAITH) && 0 < NFAITH
499 	if (faithprefix(&ip6->ip6_dst)) {
500 		/*
501 		 * Deliver very specific ICMP6 type only.
502 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
503 		 * will not work.
504 		 */
505 		switch (icmp6->icmp6_type) {
506 		case ICMP6_DST_UNREACH:
507 		case ICMP6_PACKET_TOO_BIG:
508 		case ICMP6_TIME_EXCEEDED:
509 			break;
510 		default:
511 			goto freeit;
512 		}
513 	}
514 #endif
515 
516 	code = icmp6->icmp6_code;
517 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
518 
519 	switch (icmp6->icmp6_type) {
520 	case ICMP6_DST_UNREACH:
521 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
522 		switch (code) {
523 		case ICMP6_DST_UNREACH_NOROUTE:
524 			code = PRC_UNREACH_NET;
525 			break;
526 		case ICMP6_DST_UNREACH_ADMIN:
527 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
528 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
529 			break;
530 		case ICMP6_DST_UNREACH_ADDR:
531 			code = PRC_HOSTDEAD;
532 			break;
533 #ifdef COMPAT_RFC1885
534 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
535 			code = PRC_UNREACH_SRCFAIL;
536 			break;
537 #else
538 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
539 			/* I mean "source address was incorrect." */
540 			code = PRC_UNREACH_NET;
541 			break;
542 #endif
543 		case ICMP6_DST_UNREACH_NOPORT:
544 			code = PRC_UNREACH_PORT;
545 			break;
546 		default:
547 			goto badcode;
548 		}
549 		goto deliver;
550 
551 	case ICMP6_PACKET_TOO_BIG:
552 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
553 
554 		/*
555 		 * MTU is checked in icmp6_mtudisc.
556 		 */
557 		code = PRC_MSGSIZE;
558 
559 		/*
560 		 * Updating the path MTU will be done after examining
561 		 * intermediate extension headers.
562 		 */
563 		goto deliver;
564 
565 	case ICMP6_TIME_EXCEEDED:
566 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
567 		switch (code) {
568 		case ICMP6_TIME_EXCEED_TRANSIT:
569 			code = PRC_TIMXCEED_INTRANS;
570 			break;
571 		case ICMP6_TIME_EXCEED_REASSEMBLY:
572 			code = PRC_TIMXCEED_REASS;
573 			break;
574 		default:
575 			goto badcode;
576 		}
577 		goto deliver;
578 
579 	case ICMP6_PARAM_PROB:
580 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
581 		switch (code) {
582 		case ICMP6_PARAMPROB_NEXTHEADER:
583 			code = PRC_UNREACH_PROTOCOL;
584 			break;
585 		case ICMP6_PARAMPROB_HEADER:
586 		case ICMP6_PARAMPROB_OPTION:
587 			code = PRC_PARAMPROB;
588 			break;
589 		default:
590 			goto badcode;
591 		}
592 		goto deliver;
593 
594 	case ICMP6_ECHO_REQUEST:
595 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
596 		if (code != 0)
597 			goto badcode;
598 		/*
599 		 * Copy mbuf to send to two data paths: userland socket(s),
600 		 * and to the querier (echo reply).
601 		 * m: a copy for socket, n: a copy for querier
602 		 *
603 		 * If the first mbuf is shared, or the first mbuf is too short,
604 		 * copy the first part of the data into a fresh mbuf.
605 		 * Otherwise, we will wrongly overwrite both copies.
606 		 */
607 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
608 			/* Give up local */
609 			n = m;
610 			m = NULL;
611 		} else if (M_READONLY(n) ||
612 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
613 			struct mbuf *n0 = n;
614 
615 			/*
616 			 * Prepare an internal mbuf.  m_pullup() doesn't
617 			 * always copy the length we specified.
618 			 */
619 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
620 				/* Give up local */
621 				n = m;
622 				m = NULL;
623 			}
624 			m_freem(n0);
625 		}
626 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
627 		    sizeof(*nicmp6));
628 		if (nicmp6 == NULL)
629 			goto freeit;
630 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
631 		nicmp6->icmp6_code = 0;
632 		if (n) {
633 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
634 			icmp6s[ICMP6_STAT_REFLECT]++;
635 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
636 			ICMP6_STAT_PUTREF();
637 			icmp6_reflect(n, off);
638 		}
639 		if (!m)
640 			goto freeit;
641 		break;
642 
643 	case ICMP6_ECHO_REPLY:
644 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
645 		if (code != 0)
646 			goto badcode;
647 		break;
648 
649 	case MLD_LISTENER_QUERY:
650 	case MLD_LISTENER_REPORT:
651 		if (icmp6len < sizeof(struct mld_hdr))
652 			goto badlen;
653 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
654 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
655 		else
656 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
657 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
658 			/* give up local */
659 			mld_input(m, off);
660 			m = NULL;
661 			goto freeit;
662 		}
663 		mld_input(n, off);
664 		/* m stays. */
665 		break;
666 
667 	case MLD_LISTENER_DONE:
668 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
669 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
670 			goto badlen;
671 		break;		/* nothing to be done in kernel */
672 
673 	case MLD_MTRACE_RESP:
674 	case MLD_MTRACE:
675 		/* XXX: these two are experimental.  not officially defined. */
676 		/* XXX: per-interface statistics? */
677 		break;		/* just pass it to applications */
678 
679 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
680 	    {
681 		enum { WRU, FQDN } mode;
682 
683 		if (!icmp6_nodeinfo)
684 			break;
685 
686 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
687 			mode = WRU;
688 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
689 			mode = FQDN;
690 		else
691 			goto badlen;
692 
693 		if (mode == FQDN) {
694 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
695 			if (n)
696 				n = ni6_input(n, off);
697 			/* XXX meaningless if n == NULL */
698 			noff = sizeof(struct ip6_hdr);
699 		} else {
700 			u_char *p;
701 			int maxhlen;
702 
703 			if ((icmp6_nodeinfo & 5) != 5)
704 				break;
705 
706 			if (code != 0)
707 				goto badcode;
708 			MGETHDR(n, M_DONTWAIT, m->m_type);
709 			if (n && ICMP6_MAXLEN > MHLEN) {
710 				MCLGET(n, M_DONTWAIT);
711 				if ((n->m_flags & M_EXT) == 0) {
712 					m_free(n);
713 					n = NULL;
714 				}
715 			}
716 			if (n == NULL) {
717 				/* Give up remote */
718 				break;
719 			}
720 			m_reset_rcvif(n);
721 			n->m_len = 0;
722 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
723 			if (maxhlen < 0)
724 				break;
725 			if (maxhlen > hostnamelen)
726 				maxhlen = hostnamelen;
727 			/*
728 			 * Copy IPv6 and ICMPv6 only.
729 			 */
730 			nip6 = mtod(n, struct ip6_hdr *);
731 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
732 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
733 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
734 			p = (u_char *)(nicmp6 + 1);
735 			memset(p, 0, 4);
736 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
737 			noff = sizeof(struct ip6_hdr);
738 			M_COPY_PKTHDR(n, m); /* just for rcvif */
739 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
740 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
741 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
742 			nicmp6->icmp6_code = 0;
743 		}
744 #undef hostnamelen
745 		if (n) {
746 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
747 			icmp6s[ICMP6_STAT_REFLECT]++;
748 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
749 			ICMP6_STAT_PUTREF();
750 			icmp6_reflect(n, noff);
751 		}
752 		break;
753 	    }
754 
755 	case ICMP6_WRUREPLY:
756 		if (code != 0)
757 			goto badcode;
758 		break;
759 
760 	case ND_ROUTER_SOLICIT:
761 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
762 		if (code != 0)
763 			goto badcode;
764 		if (icmp6len < sizeof(struct nd_router_solicit))
765 			goto badlen;
766 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
767 			/* give up local */
768 			nd6_rs_input(m, off, icmp6len);
769 			m = NULL;
770 			goto freeit;
771 		}
772 		nd6_rs_input(n, off, icmp6len);
773 		/* m stays. */
774 		break;
775 
776 	case ND_ROUTER_ADVERT:
777 		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
778 		if (code != 0)
779 			goto badcode;
780 		if (icmp6len < sizeof(struct nd_router_advert))
781 			goto badlen;
782 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
783 			/* give up local */
784 			nd6_ra_input(m, off, icmp6len);
785 			m = NULL;
786 			goto freeit;
787 		}
788 		nd6_ra_input(n, off, icmp6len);
789 		/* m stays. */
790 		break;
791 
792 	case ND_NEIGHBOR_SOLICIT:
793 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
794 		if (code != 0)
795 			goto badcode;
796 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
797 			goto badlen;
798 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
799 			/* give up local */
800 			nd6_ns_input(m, off, icmp6len);
801 			m = NULL;
802 			goto freeit;
803 		}
804 		nd6_ns_input(n, off, icmp6len);
805 		/* m stays. */
806 		break;
807 
808 	case ND_NEIGHBOR_ADVERT:
809 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
810 		if (code != 0)
811 			goto badcode;
812 		if (icmp6len < sizeof(struct nd_neighbor_advert))
813 			goto badlen;
814 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
815 			/* give up local */
816 			nd6_na_input(m, off, icmp6len);
817 			m = NULL;
818 			goto freeit;
819 		}
820 		nd6_na_input(n, off, icmp6len);
821 		/* m stays. */
822 		break;
823 
824 	case ND_REDIRECT:
825 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
826 		if (code != 0)
827 			goto badcode;
828 		if (icmp6len < sizeof(struct nd_redirect))
829 			goto badlen;
830 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
831 			/* give up local */
832 			icmp6_redirect_input(m, off);
833 			m = NULL;
834 			goto freeit;
835 		}
836 		icmp6_redirect_input(n, off);
837 		/* m stays. */
838 		break;
839 
840 	case ICMP6_ROUTER_RENUMBERING:
841 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
842 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
843 			goto badcode;
844 		if (icmp6len < sizeof(struct icmp6_router_renum))
845 			goto badlen;
846 		break;
847 
848 	default:
849 		nd6log(LOG_DEBUG, "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
850 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
851 		    ip6_sprintf(&ip6->ip6_dst),
852 		    rcvif ? rcvif->if_index : 0);
853 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
854 			/* ICMPv6 error: MUST deliver it by spec... */
855 			code = PRC_NCMDS;
856 			/* deliver */
857 		} else {
858 			/* ICMPv6 informational: MUST not deliver */
859 			break;
860 		}
861 	deliver:
862 		if (icmp6_notify_error(m, off, icmp6len, code)) {
863 			/* In this case, m should've been freed. */
864 			m_put_rcvif_psref(rcvif, &psref);
865 			return (IPPROTO_DONE);
866 		}
867 		break;
868 
869 	badcode:
870 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
871 		break;
872 
873 	badlen:
874 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
875 		break;
876 	}
877 	m_put_rcvif_psref(rcvif, &psref);
878 
879 	/* deliver the packet to appropriate sockets */
880 	icmp6_rip6_input(&m, *offp);
881 
882 	return IPPROTO_DONE;
883 
884  freeit:
885 	m_put_rcvif_psref(rcvif, &psref);
886 	m_freem(m);
887 	return IPPROTO_DONE;
888 }
889 
890 static int
891 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
892 {
893 	struct icmp6_hdr *icmp6;
894 	struct ip6_hdr *eip6;
895 	u_int32_t notifymtu;
896 	struct sockaddr_in6 icmp6src, icmp6dst;
897 
898 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
899 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
900 		goto freeit;
901 	}
902 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
903 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
904 	if (icmp6 == NULL) {
905 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
906 		return (-1);
907 	}
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 		struct ifnet *rcvif;
924 		int s;
925 
926 		while (1) { /* XXX: should avoid infinite loop explicitly? */
927 			struct ip6_ext *eh;
928 
929 			switch (nxt) {
930 			case IPPROTO_HOPOPTS:
931 			case IPPROTO_DSTOPTS:
932 			case IPPROTO_AH:
933 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
934 					       eoff, sizeof(*eh));
935 				if (eh == NULL) {
936 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
937 					return (-1);
938 				}
939 
940 				if (nxt == IPPROTO_AH)
941 					eoff += (eh->ip6e_len + 2) << 2;
942 				else
943 					eoff += (eh->ip6e_len + 1) << 3;
944 				nxt = eh->ip6e_nxt;
945 				break;
946 			case IPPROTO_ROUTING:
947 				/*
948 				 * When the erroneous packet contains a
949 				 * routing header, we should examine the
950 				 * header to determine the final destination.
951 				 * Otherwise, we can't properly update
952 				 * information that depends on the final
953 				 * destination (e.g. path MTU).
954 				 */
955 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
956 					       eoff, sizeof(*rth));
957 				if (rth == NULL) {
958 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
959 					return (-1);
960 				}
961 				rthlen = (rth->ip6r_len + 1) << 3;
962 				/*
963 				 * XXX: currently there is no
964 				 * officially defined type other
965 				 * than type-0.
966 				 * Note that if the segment left field
967 				 * is 0, all intermediate hops must
968 				 * have been passed.
969 				 */
970 				if (rth->ip6r_segleft &&
971 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
972 					int hops;
973 
974 					IP6_EXTHDR_GET(rth0,
975 						       struct ip6_rthdr0 *, m,
976 						       eoff, rthlen);
977 					if (rth0 == NULL) {
978 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
979 						return (-1);
980 					}
981 					/* just ignore a bogus header */
982 					if ((rth0->ip6r0_len % 2) == 0 &&
983 					    (hops = rth0->ip6r0_len/2))
984 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
985 				}
986 				eoff += rthlen;
987 				nxt = rth->ip6r_nxt;
988 				break;
989 			case IPPROTO_FRAGMENT:
990 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
991 					       eoff, sizeof(*fh));
992 				if (fh == NULL) {
993 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
994 					return (-1);
995 				}
996 				/*
997 				 * Data after a fragment header is meaningless
998 				 * unless it is the first fragment, but
999 				 * we'll go to the notify label for path MTU
1000 				 * discovery.
1001 				 */
1002 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1003 					goto notify;
1004 
1005 				eoff += sizeof(struct ip6_frag);
1006 				nxt = fh->ip6f_nxt;
1007 				break;
1008 			default:
1009 				/*
1010 				 * This case includes ESP and the No Next
1011 				 * Header.  In such cases going to the notify
1012 				 * label does not have any meaning
1013 				 * (i.e. ctlfunc will be NULL), but we go
1014 				 * anyway since we might have to update
1015 				 * path MTU information.
1016 				 */
1017 				goto notify;
1018 			}
1019 		}
1020 	  notify:
1021 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1022 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1023 		if (icmp6 == NULL) {
1024 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1025 			return (-1);
1026 		}
1027 
1028 		/*
1029 		 * retrieve parameters from the inner IPv6 header, and convert
1030 		 * them into sockaddr structures.
1031 		 * XXX: there is no guarantee that the source or destination
1032 		 * addresses of the inner packet are in the same scope zone as
1033 		 * the addresses of the icmp packet.  But there is no other
1034 		 * way to determine the zone.
1035 		 */
1036 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1037 
1038 		rcvif = m_get_rcvif(m, &s);
1039 		sockaddr_in6_init(&icmp6dst,
1040 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1041 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1042 			m_put_rcvif(rcvif, &s);
1043 			goto freeit;
1044 		}
1045 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1046 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1047 			m_put_rcvif(rcvif, &s);
1048 			goto freeit;
1049 		}
1050 		m_put_rcvif(rcvif, &s);
1051 
1052 		icmp6src.sin6_flowinfo =
1053 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1054 
1055 		if (finaldst == NULL)
1056 			finaldst = &eip6->ip6_dst;
1057 		ip6cp.ip6c_m = m;
1058 		ip6cp.ip6c_icmp6 = icmp6;
1059 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1060 		ip6cp.ip6c_off = eoff;
1061 		ip6cp.ip6c_finaldst = finaldst;
1062 		ip6cp.ip6c_src = &icmp6src;
1063 		ip6cp.ip6c_nxt = nxt;
1064 
1065 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1066 			notifymtu = ntohl(icmp6->icmp6_mtu);
1067 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1068 		}
1069 
1070 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1071 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1072 		if (ctlfunc) {
1073 			(void) (*ctlfunc)(code, sin6tosa(&icmp6dst),
1074 					  &ip6cp);
1075 		}
1076 	}
1077 	return (0);
1078 
1079   freeit:
1080 	m_freem(m);
1081 	return (-1);
1082 }
1083 
1084 void
1085 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1086 {
1087 	unsigned long rtcount;
1088 	struct icmp6_mtudisc_callback *mc;
1089 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1090 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1091 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1092 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1093 	struct rtentry *rt = NULL;
1094 	struct sockaddr_in6 sin6;
1095 	struct ifnet *rcvif;
1096 	int s;
1097 
1098 	/*
1099 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1100 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1101 	 * In that one case, the MTU might be less than 1280.
1102 	 */
1103 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1104 		/* is the mtu even sane? */
1105 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1106 			return;
1107 		if (!validated)
1108 			return;
1109 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1110 	}
1111 
1112 	/*
1113 	 * allow non-validated cases if memory is plenty, to make traffic
1114 	 * from non-connected pcb happy.
1115 	 */
1116 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1117 	if (validated) {
1118 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1119 			return;
1120 		else if (0 <= icmp6_mtudisc_lowat &&
1121 		    rtcount > icmp6_mtudisc_lowat) {
1122 			/*
1123 			 * XXX nuke a victim, install the new one.
1124 			 */
1125 		}
1126 	} else {
1127 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1128 			return;
1129 	}
1130 
1131 	memset(&sin6, 0, sizeof(sin6));
1132 	sin6.sin6_family = PF_INET6;
1133 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1134 	sin6.sin6_addr = *dst;
1135 	s = pserialize_read_enter();
1136 	rcvif = m_get_rcvif(m, &s);
1137 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1138 		m_put_rcvif(rcvif, &s);
1139 		return;
1140 	}
1141 	m_put_rcvif(rcvif, &s);
1142 
1143 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1144 
1145 	if (rt && (rt->rt_flags & RTF_HOST) &&
1146 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1147 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1148 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1149 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1150 			rt->rt_rmx.rmx_mtu = mtu;
1151 		}
1152 	}
1153 	if (rt) {
1154 		rtfree(rt);
1155 	}
1156 
1157 	/*
1158 	 * Notify protocols that the MTU for this destination
1159 	 * has changed.
1160 	 */
1161 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1162 	     mc = LIST_NEXT(mc, mc_list))
1163 		(*mc->mc_func)(&sin6.sin6_addr);
1164 }
1165 
1166 /*
1167  * Process a Node Information Query packet, based on
1168  * draft-ietf-ipngwg-icmp-name-lookups-07.
1169  *
1170  * Spec incompatibilities:
1171  * - IPv6 Subject address handling
1172  * - IPv4 Subject address handling support missing
1173  * - Proxy reply (answer even if it's not for me)
1174  * - joins NI group address at in6_ifattach() time only, does not cope
1175  *   with hostname changes by sethostname(3)
1176  */
1177 static struct mbuf *
1178 ni6_input(struct mbuf *m, int off)
1179 {
1180 	struct icmp6_nodeinfo *ni6, *nni6;
1181 	struct mbuf *n = NULL;
1182 	u_int16_t qtype;
1183 	int subjlen;
1184 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1185 	struct ni_reply_fqdn *fqdn;
1186 	int addrs;		/* for NI_QTYPE_NODEADDR */
1187 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1188 	struct sockaddr_in6 sin6; /* ip6_dst */
1189 	struct in6_addr in6_subj; /* subject address */
1190 	struct ip6_hdr *ip6;
1191 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1192 	char *subj = NULL;
1193 	struct ifnet *rcvif;
1194 	int s, ss;
1195 	struct ifaddr *ifa;
1196 	struct psref psref;
1197 
1198 	ip6 = mtod(m, struct ip6_hdr *);
1199 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1200 	if (ni6 == NULL) {
1201 		/* m is already reclaimed */
1202 		return NULL;
1203 	}
1204 
1205 	/*
1206 	 * Validate IPv6 destination address.
1207 	 *
1208 	 * The Responder must discard the Query without further processing
1209 	 * unless it is one of the Responder's unicast or anycast addresses, or
1210 	 * a link-local scope multicast address which the Responder has joined.
1211 	 * [icmp-name-lookups-07, Section 4.]
1212 	 */
1213 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1214 	/* XXX scopeid */
1215 	ss = pserialize_read_enter();
1216 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1217 	if (ifa != NULL)
1218 		; /* unicast/anycast, fine */
1219 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1220 		; /* link-local multicast, fine */
1221 	else {
1222 		pserialize_read_exit(ss);
1223 		goto bad;
1224 	}
1225 	pserialize_read_exit(ss);
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 	case NI_QTYPE_IPV4ADDR:
1240 		switch (ni6->ni_code) {
1241 		case ICMP6_NI_SUBJ_IPV6:
1242 #if ICMP6_NI_SUBJ_IPV6 != 0
1243 		case 0:
1244 #endif
1245 			/*
1246 			 * backward compatibility - try to accept 03 draft
1247 			 * format, where no Subject is present.
1248 			 */
1249 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1250 			    subjlen == 0) {
1251 				oldfqdn++;
1252 				break;
1253 			}
1254 #if ICMP6_NI_SUBJ_IPV6 != 0
1255 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1256 				goto bad;
1257 #endif
1258 
1259 			if (subjlen != sizeof(sin6.sin6_addr))
1260 				goto bad;
1261 
1262 			/*
1263 			 * Validate Subject address.
1264 			 *
1265 			 * Not sure what exactly "address belongs to the node"
1266 			 * means in the spec, is it just unicast, or what?
1267 			 *
1268 			 * At this moment we consider Subject address as
1269 			 * "belong to the node" if the Subject address equals
1270 			 * to the IPv6 destination address; validation for
1271 			 * IPv6 destination address should have done enough
1272 			 * check for us.
1273 			 *
1274 			 * We do not do proxy at this moment.
1275 			 */
1276 			/* m_pulldown instead of copy? */
1277 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1278 			    subjlen, (void *)&in6_subj);
1279 			rcvif = m_get_rcvif(m, &s);
1280 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1281 				m_put_rcvif(rcvif, &s);
1282 				goto bad;
1283 			}
1284 			m_put_rcvif(rcvif, &s);
1285 
1286 			subj = (char *)&in6_subj;
1287 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
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 	case NI_QTYPE_IPV4ADDR:
1342 		if ((icmp6_nodeinfo & 2) == 0)
1343 			goto bad;
1344 		break;
1345 	}
1346 
1347 	/* guess reply length */
1348 	switch (qtype) {
1349 	case NI_QTYPE_NOOP:
1350 		break;		/* no reply data */
1351 	case NI_QTYPE_SUPTYPES:
1352 		replylen += sizeof(u_int32_t);
1353 		break;
1354 	case NI_QTYPE_FQDN:
1355 		/* XXX will append an mbuf */
1356 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1357 		break;
1358 	case NI_QTYPE_NODEADDR:
1359 		addrs = ni6_addrs(ni6, m, &ifp, subj, &psref);
1360 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1361 					  sizeof(u_int32_t))) > MCLBYTES)
1362 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1363 		break;
1364 	case NI_QTYPE_IPV4ADDR:
1365 		/* unsupported - should respond with unknown Qtype? */
1366 		goto bad;
1367 	default:
1368 		/*
1369 		 * XXX: We must return a reply with the ICMP6 code
1370 		 * `unknown Qtype' in this case.  However we regard the case
1371 		 * as an FQDN query for backward compatibility.
1372 		 * Older versions set a random value to this field,
1373 		 * so it rarely varies in the defined qtypes.
1374 		 * But the mechanism is not reliable...
1375 		 * maybe we should obsolete older versions.
1376 		 */
1377 		qtype = NI_QTYPE_FQDN;
1378 		/* XXX will append an mbuf */
1379 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1380 		oldfqdn++;
1381 		break;
1382 	}
1383 
1384 	/* allocate an mbuf to reply. */
1385 	MGETHDR(n, M_DONTWAIT, m->m_type);
1386 	if (n == NULL) {
1387 		if_put(ifp, &psref);
1388 		m_freem(m);
1389 		return (NULL);
1390 	}
1391 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1392 	if (replylen > MHLEN) {
1393 		if (replylen > MCLBYTES) {
1394 			/*
1395 			 * XXX: should we try to allocate more? But MCLBYTES
1396 			 * is probably much larger than IPV6_MMTU...
1397 			 */
1398 			goto bad;
1399 		}
1400 		MCLGET(n, M_DONTWAIT);
1401 		if ((n->m_flags & M_EXT) == 0) {
1402 			goto bad;
1403 		}
1404 	}
1405 	n->m_pkthdr.len = n->m_len = replylen;
1406 
1407 	/* copy mbuf header and IPv6 + Node Information base headers */
1408 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1409 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1410 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1411 
1412 	/* qtype dependent procedure */
1413 	switch (qtype) {
1414 	case NI_QTYPE_NOOP:
1415 		nni6->ni_code = ICMP6_NI_SUCCESS;
1416 		nni6->ni_flags = 0;
1417 		break;
1418 	case NI_QTYPE_SUPTYPES:
1419 	{
1420 		u_int32_t v;
1421 		nni6->ni_code = ICMP6_NI_SUCCESS;
1422 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1423 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1424 		v = (u_int32_t)htonl(0x0000000f);
1425 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1426 		break;
1427 	}
1428 	case NI_QTYPE_FQDN:
1429 		nni6->ni_code = ICMP6_NI_SUCCESS;
1430 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1431 						sizeof(struct ip6_hdr) +
1432 						sizeof(struct icmp6_nodeinfo));
1433 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1434 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1435 		/*
1436 		 * XXX do we really have FQDN in variable "hostname"?
1437 		 */
1438 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1439 		if (n->m_next == NULL)
1440 			goto bad;
1441 		/* XXX we assume that n->m_next is not a chain */
1442 		if (n->m_next->m_next != NULL)
1443 			goto bad;
1444 		n->m_pkthdr.len += n->m_next->m_len;
1445 		break;
1446 	case NI_QTYPE_NODEADDR:
1447 	{
1448 		int lenlim, copied;
1449 
1450 		nni6->ni_code = ICMP6_NI_SUCCESS;
1451 		n->m_pkthdr.len = n->m_len =
1452 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1453 		lenlim = M_TRAILINGSPACE(n);
1454 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1455 		if_put(ifp, &psref);
1456 		ifp = NULL;
1457 		/* XXX: reset mbuf length */
1458 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1459 			sizeof(struct icmp6_nodeinfo) + copied;
1460 		break;
1461 	}
1462 	default:
1463 		break;		/* XXX impossible! */
1464 	}
1465 
1466 	nni6->ni_type = ICMP6_NI_REPLY;
1467 	m_freem(m);
1468 	return (n);
1469 
1470   bad:
1471 	if_put(ifp, &psref);
1472 	m_freem(m);
1473 	if (n)
1474 		m_freem(n);
1475 	return (NULL);
1476 }
1477 #undef hostnamelen
1478 
1479 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1480 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1481 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1482 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1483 
1484 /*
1485  * make a mbuf with DNS-encoded string.  no compression support.
1486  *
1487  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1488  * treated as truncated name (two \0 at the end).  this is a wild guess.
1489  *
1490  * old - return pascal string if non-zero
1491  */
1492 static struct mbuf *
1493 ni6_nametodns(const char *name, int namelen, int old)
1494 {
1495 	struct mbuf *m;
1496 	char *cp, *ep;
1497 	const char *p, *q;
1498 	int i, len, nterm;
1499 
1500 	if (old)
1501 		len = namelen + 1;
1502 	else
1503 		len = MCLBYTES;
1504 
1505 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1506 	MGET(m, M_DONTWAIT, MT_DATA);
1507 	if (m && len > MLEN) {
1508 		MCLGET(m, M_DONTWAIT);
1509 		if ((m->m_flags & M_EXT) == 0)
1510 			goto fail;
1511 	}
1512 	if (!m)
1513 		goto fail;
1514 	m->m_next = NULL;
1515 
1516 	if (old) {
1517 		m->m_len = len;
1518 		*mtod(m, char *) = namelen;
1519 		bcopy(name, mtod(m, char *) + 1, namelen);
1520 		return m;
1521 	} else {
1522 		m->m_len = 0;
1523 		cp = mtod(m, char *);
1524 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1525 
1526 		/* if not certain about my name, return empty buffer */
1527 		if (namelen == 0)
1528 			return m;
1529 
1530 		/*
1531 		 * guess if it looks like shortened hostname, or FQDN.
1532 		 * shortened hostname needs two trailing "\0".
1533 		 */
1534 		i = 0;
1535 		for (p = name; p < name + namelen; p++) {
1536 			if (*p && *p == '.')
1537 				i++;
1538 		}
1539 		if (i < 2)
1540 			nterm = 2;
1541 		else
1542 			nterm = 1;
1543 
1544 		p = name;
1545 		while (cp < ep && p < name + namelen) {
1546 			i = 0;
1547 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1548 				i++;
1549 			/* result does not fit into mbuf */
1550 			if (cp + i + 1 >= ep)
1551 				goto fail;
1552 			/*
1553 			 * DNS label length restriction, RFC1035 page 8.
1554 			 * "i == 0" case is included here to avoid returning
1555 			 * 0-length label on "foo..bar".
1556 			 */
1557 			if (i <= 0 || i >= 64)
1558 				goto fail;
1559 			*cp++ = i;
1560 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1561 				goto fail;
1562 			while (i > 0) {
1563 				if (!isalnum(*p) && *p != '-')
1564 					goto fail;
1565 				if (isupper(*p)) {
1566 					*cp++ = tolower(*p);
1567 					p++;
1568 				} else
1569 					*cp++ = *p++;
1570 				i--;
1571 			}
1572 			p = q;
1573 			if (p < name + namelen && *p == '.')
1574 				p++;
1575 		}
1576 		/* termination */
1577 		if (cp + nterm >= ep)
1578 			goto fail;
1579 		while (nterm-- > 0)
1580 			*cp++ = '\0';
1581 		m->m_len = cp - mtod(m, char *);
1582 		return m;
1583 	}
1584 
1585 	panic("should not reach here");
1586 	/* NOTREACHED */
1587 
1588  fail:
1589 	if (m)
1590 		m_freem(m);
1591 	return NULL;
1592 }
1593 
1594 /*
1595  * check if two DNS-encoded string matches.  takes care of truncated
1596  * form (with \0\0 at the end).  no compression support.
1597  * XXX upper/lowercase match (see RFC2065)
1598  */
1599 static int
1600 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1601 {
1602 	const char *a0, *b0;
1603 	int l;
1604 
1605 	/* simplest case - need validation? */
1606 	if (alen == blen && memcmp(a, b, alen) == 0)
1607 		return 1;
1608 
1609 	a0 = a;
1610 	b0 = b;
1611 
1612 	/* termination is mandatory */
1613 	if (alen < 2 || blen < 2)
1614 		return 0;
1615 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1616 		return 0;
1617 	alen--;
1618 	blen--;
1619 
1620 	while (a - a0 < alen && b - b0 < blen) {
1621 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1622 			return 0;
1623 
1624 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1625 			return 0;
1626 		/* we don't support compression yet */
1627 		if (a[0] >= 64 || b[0] >= 64)
1628 			return 0;
1629 
1630 		/* truncated case */
1631 		if (a[0] == 0 && a - a0 == alen - 1)
1632 			return 1;
1633 		if (b[0] == 0 && b - b0 == blen - 1)
1634 			return 1;
1635 		if (a[0] == 0 || b[0] == 0)
1636 			return 0;
1637 
1638 		if (a[0] != b[0])
1639 			return 0;
1640 		l = a[0];
1641 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1642 			return 0;
1643 		if (memcmp(a + 1, b + 1, l) != 0)
1644 			return 0;
1645 
1646 		a += 1 + l;
1647 		b += 1 + l;
1648 	}
1649 
1650 	if (a - a0 == alen && b - b0 == blen)
1651 		return 1;
1652 	else
1653 		return 0;
1654 }
1655 
1656 /*
1657  * calculate the number of addresses to be returned in the node info reply.
1658  */
1659 static int
1660 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1661     struct ifnet **ifpp, char *subj, struct psref *psref)
1662 {
1663 	struct ifnet *ifp;
1664 	struct in6_ifaddr *ia6;
1665 	struct ifaddr *ifa;
1666 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1667 	int addrs = 0, addrsofif, iffound = 0;
1668 	int niflags = ni6->ni_flags;
1669 	int s;
1670 
1671 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1672 		switch (ni6->ni_code) {
1673 		case ICMP6_NI_SUBJ_IPV6:
1674 			if (subj == NULL) /* must be impossible... */
1675 				return (0);
1676 			subj_ip6 = (struct sockaddr_in6 *)subj;
1677 			break;
1678 		default:
1679 			/*
1680 			 * XXX: we only support IPv6 subject address for
1681 			 * this Qtype.
1682 			 */
1683 			return (0);
1684 		}
1685 	}
1686 
1687 	s = pserialize_read_enter();
1688 	IFNET_READER_FOREACH(ifp) {
1689 		addrsofif = 0;
1690 		IFADDR_READER_FOREACH(ifa, ifp) {
1691 			if (ifa->ifa_addr->sa_family != AF_INET6)
1692 				continue;
1693 			ia6 = (struct in6_ifaddr *)ifa;
1694 
1695 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1696 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1697 					       &ia6->ia_addr.sin6_addr))
1698 				iffound = 1;
1699 
1700 			/*
1701 			 * IPv4-mapped addresses can only be returned by a
1702 			 * Node Information proxy, since they represent
1703 			 * addresses of IPv4-only nodes, which perforce do
1704 			 * not implement this protocol.
1705 			 * [icmp-name-lookups-07, Section 5.4]
1706 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1707 			 * this function at this moment.
1708 			 */
1709 
1710 			/* What do we have to do about ::1? */
1711 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1712 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1713 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1714 					continue;
1715 				break;
1716 			case IPV6_ADDR_SCOPE_SITELOCAL:
1717 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1718 					continue;
1719 				break;
1720 			case IPV6_ADDR_SCOPE_GLOBAL:
1721 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1722 					continue;
1723 				break;
1724 			default:
1725 				continue;
1726 			}
1727 
1728 			/*
1729 			 * check if anycast is okay.
1730 			 * XXX: just experimental.  not in the spec.
1731 			 */
1732 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1733 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1734 				continue; /* we need only unicast addresses */
1735 
1736 			addrsofif++; /* count the address */
1737 		}
1738 		if (iffound) {
1739 			if_acquire_NOMPSAFE(ifp, psref);
1740 			pserialize_read_exit(s);
1741 			*ifpp = ifp;
1742 			return (addrsofif);
1743 		}
1744 
1745 		addrs += addrsofif;
1746 	}
1747 	pserialize_read_exit(s);
1748 
1749 	return (addrs);
1750 }
1751 
1752 static int
1753 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1754 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1755 	int resid)
1756 {
1757 	struct ifnet *ifp;
1758 	struct in6_ifaddr *ia6;
1759 	struct ifaddr *ifa;
1760 	struct ifnet *ifp_dep = NULL;
1761 	int copied = 0, allow_deprecated = 0;
1762 	u_char *cp = (u_char *)(nni6 + 1);
1763 	int niflags = ni6->ni_flags;
1764 	u_int32_t ltime;
1765 	int s;
1766 
1767 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1768 		return (0);	/* needless to copy */
1769 
1770 	s = pserialize_read_enter();
1771 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1772   again:
1773 
1774 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1775 	{
1776 		IFADDR_READER_FOREACH(ifa, ifp) {
1777 			if (ifa->ifa_addr->sa_family != AF_INET6)
1778 				continue;
1779 			ia6 = (struct in6_ifaddr *)ifa;
1780 
1781 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1782 			    allow_deprecated == 0) {
1783 				/*
1784 				 * prefererred address should be put before
1785 				 * deprecated addresses.
1786 				 */
1787 
1788 				/* record the interface for later search */
1789 				if (ifp_dep == NULL)
1790 					ifp_dep = ifp;
1791 
1792 				continue;
1793 			}
1794 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1795 				 allow_deprecated != 0)
1796 				continue; /* we now collect deprecated addrs */
1797 
1798 			/* What do we have to do about ::1? */
1799 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1800 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1801 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1802 					continue;
1803 				break;
1804 			case IPV6_ADDR_SCOPE_SITELOCAL:
1805 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1806 					continue;
1807 				break;
1808 			case IPV6_ADDR_SCOPE_GLOBAL:
1809 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1810 					continue;
1811 				break;
1812 			default:
1813 				continue;
1814 			}
1815 
1816 			/*
1817 			 * check if anycast is okay.
1818 			 * XXX: just experimental.  not in the spec.
1819 			 */
1820 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1821 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1822 				continue;
1823 
1824 			/* now we can copy the address */
1825 			if (resid < sizeof(struct in6_addr) +
1826 			    sizeof(u_int32_t)) {
1827 				/*
1828 				 * We give up much more copy.
1829 				 * Set the truncate flag and return.
1830 				 */
1831 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1832 				goto out;
1833 			}
1834 
1835 			/*
1836 			 * Set the TTL of the address.
1837 			 * The TTL value should be one of the following
1838 			 * according to the specification:
1839 			 *
1840 			 * 1. The remaining lifetime of a DHCP lease on the
1841 			 *    address, or
1842 			 * 2. The remaining Valid Lifetime of a prefix from
1843 			 *    which the address was derived through Stateless
1844 			 *    Autoconfiguration.
1845 			 *
1846 			 * Note that we currently do not support stateful
1847 			 * address configuration by DHCPv6, so the former
1848 			 * case can't happen.
1849 			 *
1850 			 * TTL must be 2^31 > TTL >= 0.
1851 			 */
1852 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1853 				ltime = ND6_INFINITE_LIFETIME;
1854 			else {
1855 				if (ia6->ia6_lifetime.ia6t_expire >
1856 				    time_uptime)
1857 					ltime = ia6->ia6_lifetime.ia6t_expire -
1858 					    time_uptime;
1859 				else
1860 					ltime = 0;
1861 			}
1862 			if (ltime > 0x7fffffff)
1863 				ltime = 0x7fffffff;
1864 			ltime = htonl(ltime);
1865 
1866 			bcopy(&ltime, cp, sizeof(u_int32_t));
1867 			cp += sizeof(u_int32_t);
1868 
1869 			/* copy the address itself */
1870 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1871 			      sizeof(struct in6_addr));
1872 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1873 			cp += sizeof(struct in6_addr);
1874 
1875 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1876 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1877 		}
1878 		if (ifp0)	/* we need search only on the specified IF */
1879 			break;
1880 	}
1881 
1882 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1883 		ifp = ifp_dep;
1884 		allow_deprecated = 1;
1885 
1886 		goto again;
1887 	}
1888 out:
1889 	pserialize_read_exit(s);
1890 	return (copied);
1891 }
1892 
1893 /*
1894  * XXX almost dup'ed code with rip6_input.
1895  */
1896 static int
1897 icmp6_rip6_input(struct mbuf **mp, int off)
1898 {
1899 	struct mbuf *m = *mp;
1900 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1901 	struct inpcb_hdr *inph;
1902 	struct in6pcb *in6p;
1903 	struct in6pcb *last = NULL;
1904 	struct sockaddr_in6 rip6src;
1905 	struct icmp6_hdr *icmp6;
1906 	struct mbuf *opts = NULL;
1907 
1908 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1909 	if (icmp6 == NULL) {
1910 		/* m is already reclaimed */
1911 		return IPPROTO_DONE;
1912 	}
1913 
1914 	/*
1915 	 * XXX: the address may have embedded scope zone ID, which should be
1916 	 * hidden from applications.
1917 	 */
1918 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1919 	if (sa6_recoverscope(&rip6src)) {
1920 		m_freem(m);
1921 		return (IPPROTO_DONE);
1922 	}
1923 
1924 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1925 		in6p = (struct in6pcb *)inph;
1926 		if (in6p->in6p_af != AF_INET6)
1927 			continue;
1928 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1929 			continue;
1930 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1931 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1932 			continue;
1933 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1934 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1935 			continue;
1936 		if (in6p->in6p_icmp6filt
1937 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1938 				 in6p->in6p_icmp6filt))
1939 			continue;
1940 		if (last) {
1941 			struct	mbuf *n;
1942 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1943 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1944 					ip6_savecontrol(last, &opts, ip6, n);
1945 				/* strip intermediate headers */
1946 				m_adj(n, off);
1947 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1948 				    sin6tosa(&rip6src), n, opts) == 0) {
1949 					/* should notify about lost packet */
1950 					m_freem(n);
1951 					if (opts)
1952 						m_freem(opts);
1953 				} else
1954 					sorwakeup(last->in6p_socket);
1955 				opts = NULL;
1956 			}
1957 		}
1958 		last = in6p;
1959 	}
1960 	if (last) {
1961 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1962 			ip6_savecontrol(last, &opts, ip6, m);
1963 		/* strip intermediate headers */
1964 		m_adj(m, off);
1965 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1966 		    sin6tosa(&rip6src), m, opts) == 0) {
1967 			m_freem(m);
1968 			if (opts)
1969 				m_freem(opts);
1970 		} else
1971 			sorwakeup(last->in6p_socket);
1972 	} else {
1973 		m_freem(m);
1974 		IP6_STATDEC(IP6_STAT_DELIVERED);
1975 	}
1976 	return IPPROTO_DONE;
1977 }
1978 
1979 /*
1980  * Reflect the ip6 packet back to the source.
1981  * OFF points to the icmp6 header, counted from the top of the mbuf.
1982  *
1983  * Note: RFC 1885 required that an echo reply should be truncated if it
1984  * did not fit in with (return) path MTU, and KAME code supported the
1985  * behavior.  However, as a clarification after the RFC, this limitation
1986  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1987  * old behavior, with a (non-default) ifdef block, while the new version of
1988  * the spec was an internet-draft status, and even after the new RFC was
1989  * published.  But it would rather make sense to clean the obsoleted part
1990  * up, and to make the code simpler at this stage.
1991  */
1992 void
1993 icmp6_reflect(struct mbuf *m, size_t off)
1994 {
1995 	struct ip6_hdr *ip6;
1996 	struct icmp6_hdr *icmp6;
1997 	const struct in6_ifaddr *ia;
1998 	const struct ip6aux *ip6a;
1999 	int plen;
2000 	int type, code;
2001 	struct ifnet *outif = NULL;
2002 	struct in6_addr origdst;
2003 	const struct in6_addr *src = NULL;
2004 	struct ifnet *rcvif;
2005 	int s;
2006 
2007 	/* too short to reflect */
2008 	if (off < sizeof(struct ip6_hdr)) {
2009 		nd6log(LOG_DEBUG,
2010 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2011 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2012 		    __FILE__, __LINE__);
2013 		goto bad;
2014 	}
2015 
2016 	/*
2017 	 * If there are extra headers between IPv6 and ICMPv6, strip
2018 	 * off that header first.
2019 	 */
2020 #ifdef DIAGNOSTIC
2021 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2022 		panic("assumption failed in icmp6_reflect");
2023 #endif
2024 	if (off > sizeof(struct ip6_hdr)) {
2025 		size_t l;
2026 		struct ip6_hdr nip6;
2027 
2028 		l = off - sizeof(struct ip6_hdr);
2029 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2030 		m_adj(m, l);
2031 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2032 		if (m->m_len < l) {
2033 			if ((m = m_pullup(m, l)) == NULL)
2034 				return;
2035 		}
2036 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
2037 	} else /* off == sizeof(struct ip6_hdr) */ {
2038 		size_t l;
2039 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2040 		if (m->m_len < l) {
2041 			if ((m = m_pullup(m, l)) == NULL)
2042 				return;
2043 		}
2044 	}
2045 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2046 	ip6 = mtod(m, struct ip6_hdr *);
2047 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2048 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2049 	type = icmp6->icmp6_type; /* keep type for statistics */
2050 	code = icmp6->icmp6_code; /* ditto. */
2051 
2052 	origdst = ip6->ip6_dst;
2053 	/*
2054 	 * ip6_input() drops a packet if its src is multicast.
2055 	 * So, the src is never multicast.
2056 	 */
2057 	ip6->ip6_dst = ip6->ip6_src;
2058 
2059 	/*
2060 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2061 	 * use dst as the src for the reply.
2062 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2063 	 * (for example) when we encounter an error while forwarding procedure
2064 	 * destined to a duplicated address of ours.
2065 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2066 	 * procedure of an outgoing packet of our own, in which case we need
2067 	 * to search in the ifaddr list.
2068 	 */
2069 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2070 		;
2071 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2072 		if ((ip6a->ip6a_flags &
2073 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2074 			src = &ip6a->ip6a_src;
2075 	} else {
2076 		union {
2077 			struct sockaddr_in6 sin6;
2078 			struct sockaddr sa;
2079 		} u;
2080 		int _s;
2081 		struct ifaddr *ifa;
2082 
2083 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2084 
2085 		_s = pserialize_read_enter();
2086 		ifa = ifa_ifwithaddr(&u.sa);
2087 
2088 		if (ifa != NULL) {
2089 			ia = ifatoia6(ifa);
2090 			if ((ia->ia6_flags &
2091 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2092 				src = &ia->ia_addr.sin6_addr;
2093 		}
2094 		pserialize_read_exit(_s);
2095 	}
2096 
2097 	if (src == NULL) {
2098 		int e;
2099 		struct sockaddr_in6 sin6;
2100 		struct route ro;
2101 
2102 		/*
2103 		 * This case matches to multicasts, our anycast, or unicasts
2104 		 * that we do not own.  Select a source address based on the
2105 		 * source address of the erroneous packet.
2106 		 */
2107 		/* zone ID should be embedded */
2108 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2109 
2110 		memset(&ro, 0, sizeof(ro));
2111 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL, &e);
2112 		rtcache_free(&ro);
2113 		if (src == NULL) {
2114 			nd6log(LOG_DEBUG,
2115 			    "source can't be determined: "
2116 			    "dst=%s, error=%d\n",
2117 			    ip6_sprintf(&sin6.sin6_addr), e);
2118 			goto bad;
2119 		}
2120 	}
2121 
2122 	ip6->ip6_src = *src;
2123 	ip6->ip6_flow = 0;
2124 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2125 	ip6->ip6_vfc |= IPV6_VERSION;
2126 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2127 	rcvif = m_get_rcvif(m, &s);
2128 	if (rcvif) {
2129 		/* XXX: This may not be the outgoing interface */
2130 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2131 	} else
2132 		ip6->ip6_hlim = ip6_defhlim;
2133 	m_put_rcvif(rcvif, &s);
2134 
2135 	m->m_pkthdr.csum_flags = 0;
2136 	icmp6->icmp6_cksum = 0;
2137 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2138 					sizeof(struct ip6_hdr), plen);
2139 
2140 	/*
2141 	 * XXX option handling
2142 	 */
2143 
2144 	m->m_flags &= ~(M_BCAST|M_MCAST);
2145 
2146 	/*
2147 	 * To avoid a "too big" situation at an intermediate router
2148 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2149 	 * Note that only echo and node information replies are affected,
2150 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2151 	 */
2152 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif)
2153 	    != 0 && outif)
2154 		icmp6_ifstat_inc(outif, ifs6_out_error);
2155 	if (outif)
2156 		icmp6_ifoutstat_inc(outif, type, code);
2157 
2158 	return;
2159 
2160  bad:
2161 	m_freem(m);
2162 	return;
2163 }
2164 
2165 static const char *
2166 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2167 	struct in6_addr *tgt6)
2168 {
2169 	static char buf[1024];
2170 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2171 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2172 	return buf;
2173 }
2174 
2175 void
2176 icmp6_redirect_input(struct mbuf *m, int off)
2177 {
2178 	struct ifnet *ifp;
2179 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2180 	struct nd_redirect *nd_rd;
2181 	int icmp6len = ntohs(ip6->ip6_plen);
2182 	char *lladdr = NULL;
2183 	int lladdrlen = 0;
2184 	struct rtentry *rt = NULL;
2185 	int is_router;
2186 	int is_onlink;
2187 	struct in6_addr src6 = ip6->ip6_src;
2188 	struct in6_addr redtgt6;
2189 	struct in6_addr reddst6;
2190 	union nd_opts ndopts;
2191 	struct psref psref;
2192 
2193 	ifp = m_get_rcvif_psref(m, &psref);
2194 	if (ifp == NULL)
2195 		goto freeit;
2196 
2197 	/* XXX if we are router, we don't update route by icmp6 redirect */
2198 	if (ip6_forwarding)
2199 		goto freeit;
2200 	if (!icmp6_rediraccept)
2201 		goto freeit;
2202 
2203 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2204 	if (nd_rd == NULL) {
2205 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2206 		m_put_rcvif_psref(ifp, &psref);
2207 		return;
2208 	}
2209 	redtgt6 = nd_rd->nd_rd_target;
2210 	reddst6 = nd_rd->nd_rd_dst;
2211 
2212 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2213 	    in6_setscope(&reddst6, ifp, NULL)) {
2214 		goto freeit;
2215 	}
2216 
2217 	/* validation */
2218 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2219 		nd6log(LOG_ERR,
2220 		    "ICMP6 redirect sent from %s rejected; "
2221 		    "must be from linklocal\n", ip6_sprintf(&src6));
2222 		goto bad;
2223 	}
2224 	if (ip6->ip6_hlim != 255) {
2225 		nd6log(LOG_ERR,
2226 		    "ICMP6 redirect sent from %s rejected; "
2227 		    "hlim=%d (must be 255)\n",
2228 		    ip6_sprintf(&src6), ip6->ip6_hlim);
2229 		goto bad;
2230 	}
2231     {
2232 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2233 	struct sockaddr_in6 sin6;
2234 	struct in6_addr *gw6;
2235 
2236 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2237 	rt = rtalloc1(sin6tosa(&sin6), 0);
2238 	if (rt) {
2239 		if (rt->rt_gateway == NULL ||
2240 		    rt->rt_gateway->sa_family != AF_INET6) {
2241 			nd6log(LOG_ERR,
2242 			    "ICMP6 redirect rejected; no route "
2243 			    "with inet6 gateway found for redirect dst: %s\n",
2244 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2245 			rtfree(rt);
2246 			goto bad;
2247 		}
2248 
2249 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2250 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2251 			nd6log(LOG_ERR,
2252 			    "ICMP6 redirect rejected; "
2253 			    "not equal to gw-for-src=%s (must be same): %s\n",
2254 			    ip6_sprintf(gw6),
2255 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2256 			rtfree(rt);
2257 			goto bad;
2258 		}
2259 	} else {
2260 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2261 		    "no route found for redirect dst: %s\n",
2262 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2263 		goto bad;
2264 	}
2265 	rtfree(rt);
2266 	rt = NULL;
2267     }
2268 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2269 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2270 		    "redirect dst must be unicast: %s\n",
2271 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2272 		goto bad;
2273 	}
2274 
2275 	is_router = is_onlink = 0;
2276 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2277 		is_router = 1;	/* router case */
2278 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2279 		is_onlink = 1;	/* on-link destination case */
2280 	if (!is_router && !is_onlink) {
2281 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2282 		    "neither router case nor onlink case: %s\n",
2283 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2284 		goto bad;
2285 	}
2286 	/* validation passed */
2287 
2288 	icmp6len -= sizeof(*nd_rd);
2289 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2290 	if (nd6_options(&ndopts) < 0) {
2291 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2292 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2293 		/* nd6_options have incremented stats */
2294 		goto freeit;
2295 	}
2296 
2297 	if (ndopts.nd_opts_tgt_lladdr) {
2298 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2299 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2300 	}
2301 
2302 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2303 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2304 		    "(if %d, icmp6 packet %d): %s\n",
2305 		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2306 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2307 		goto bad;
2308 	}
2309 
2310 	/* RFC 2461 8.3 */
2311 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2312 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2313 
2314 	m_put_rcvif_psref(ifp, &psref);
2315 	ifp = NULL;
2316 
2317 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2318 		/* perform rtredirect */
2319 		struct sockaddr_in6 sdst;
2320 		struct sockaddr_in6 sgw;
2321 		struct sockaddr_in6 ssrc;
2322 		unsigned long rtcount;
2323 		struct rtentry *newrt = NULL;
2324 
2325 		/*
2326 		 * do not install redirect route, if the number of entries
2327 		 * is too much (> hiwat).  note that, the node (= host) will
2328 		 * work just fine even if we do not install redirect route
2329 		 * (there will be additional hops, though).
2330 		 */
2331 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2332 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
2333 			goto freeit;
2334 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2335 			goto freeit;
2336 		else if (0 <= icmp6_redirect_lowat &&
2337 		    rtcount > icmp6_redirect_lowat) {
2338 			/*
2339 			 * XXX nuke a victim, install the new one.
2340 			 */
2341 		}
2342 
2343 		memset(&sdst, 0, sizeof(sdst));
2344 		memset(&sgw, 0, sizeof(sgw));
2345 		memset(&ssrc, 0, sizeof(ssrc));
2346 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2347 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2348 			sizeof(struct sockaddr_in6);
2349 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2350 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2351 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2352 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2353 		           RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc),
2354 			   &newrt);
2355 
2356 		if (newrt) {
2357 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2358 			    icmp6_redirect_timeout_q);
2359 			rtfree(newrt);
2360 		}
2361 	}
2362 	/* finally update cached route in each socket via pfctlinput */
2363 	{
2364 		struct sockaddr_in6 sdst;
2365 
2366 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2367 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2368 #if defined(IPSEC)
2369 		if (ipsec_used)
2370 			key_sa_routechange(sin6tosa(&sdst));
2371 #endif
2372 	}
2373 
2374  freeit:
2375 	if (ifp != NULL)
2376 		m_put_rcvif_psref(ifp, &psref);
2377 	m_freem(m);
2378 	return;
2379 
2380  bad:
2381 	m_put_rcvif_psref(ifp, &psref);
2382 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2383 	m_freem(m);
2384 }
2385 
2386 void
2387 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2388 {
2389 	struct ifnet *ifp;	/* my outgoing interface */
2390 	struct in6_addr *ifp_ll6;
2391 	struct in6_addr *nexthop;
2392 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2393 	struct mbuf *m = NULL;	/* newly allocated one */
2394 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2395 	struct nd_redirect *nd_rd;
2396 	size_t maxlen;
2397 	u_char *p;
2398 	struct sockaddr_in6 src_sa;
2399 
2400 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2401 
2402 	/* if we are not router, we don't send icmp6 redirect */
2403 	if (!ip6_forwarding)
2404 		goto fail;
2405 
2406 	/* sanity check */
2407 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2408 		goto fail;
2409 
2410 	/*
2411 	 * Address check:
2412 	 *  the source address must identify a neighbor, and
2413 	 *  the destination address must not be a multicast address
2414 	 *  [RFC 2461, sec 8.2]
2415 	 */
2416 	sip6 = mtod(m0, struct ip6_hdr *);
2417 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2418 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2419 		goto fail;
2420 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2421 		goto fail;	/* what should we do here? */
2422 
2423 	/* rate limit */
2424 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2425 		goto fail;
2426 
2427 	/*
2428 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2429 	 * we almost always ask for an mbuf cluster for simplicity.
2430 	 * (MHLEN < IPV6_MMTU is almost always true)
2431 	 */
2432 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2433 	if (m && IPV6_MMTU >= MHLEN) {
2434 #if IPV6_MMTU >= MCLBYTES
2435 		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2436 #else
2437 		MCLGET(m, M_DONTWAIT);
2438 #endif
2439 	}
2440 
2441 	if (!m)
2442 		goto fail;
2443 	m_reset_rcvif(m);
2444 	m->m_len = 0;
2445 	maxlen = M_TRAILINGSPACE(m);
2446 	maxlen = min(IPV6_MMTU, maxlen);
2447 	/* just for safety */
2448 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2449 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2450 		goto fail;
2451 	}
2452 
2453 	{
2454 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2455 		struct in6_ifaddr *ia;
2456 		int s = pserialize_read_enter();
2457 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2458 						 IN6_IFF_NOTREADY|
2459 						 IN6_IFF_ANYCAST)) == NULL) {
2460 			pserialize_read_exit(s);
2461 			goto fail;
2462 		}
2463 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2464 		pserialize_read_exit(s);
2465 	}
2466 
2467 	/* get ip6 linklocal address for the router. */
2468 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2469 		struct sockaddr_in6 *sin6;
2470 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2471 		nexthop = &sin6->sin6_addr;
2472 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2473 			nexthop = NULL;
2474 	} else
2475 		nexthop = NULL;
2476 
2477 	/* ip6 */
2478 	ip6 = mtod(m, struct ip6_hdr *);
2479 	ip6->ip6_flow = 0;
2480 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2481 	ip6->ip6_vfc |= IPV6_VERSION;
2482 	/* ip6->ip6_plen will be set later */
2483 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2484 	ip6->ip6_hlim = 255;
2485 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2486 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2487 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2488 
2489 	/* ND Redirect */
2490 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2491 	nd_rd->nd_rd_type = ND_REDIRECT;
2492 	nd_rd->nd_rd_code = 0;
2493 	nd_rd->nd_rd_reserved = 0;
2494 	if (rt->rt_flags & RTF_GATEWAY) {
2495 		/*
2496 		 * nd_rd->nd_rd_target must be a link-local address in
2497 		 * better router cases.
2498 		 */
2499 		if (!nexthop)
2500 			goto fail;
2501 		bcopy(nexthop, &nd_rd->nd_rd_target,
2502 		      sizeof(nd_rd->nd_rd_target));
2503 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2504 		      sizeof(nd_rd->nd_rd_dst));
2505 	} else {
2506 		/* make sure redtgt == reddst */
2507 		nexthop = &sip6->ip6_dst;
2508 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2509 		      sizeof(nd_rd->nd_rd_target));
2510 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2511 		      sizeof(nd_rd->nd_rd_dst));
2512 	}
2513 
2514 	p = (u_char *)(nd_rd + 1);
2515 
2516 	{
2517 		/* target lladdr option */
2518 		struct llentry *ln = NULL;
2519 		int len;
2520 		struct nd_opt_hdr *nd_opt;
2521 		char *lladdr;
2522 
2523 		ln = nd6_lookup(nexthop, ifp, false);
2524 		if (ln == NULL)
2525 			goto nolladdropt;
2526 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2527 		len = (len + 7) & ~7;	/* round by 8 */
2528 		/* safety check */
2529 		if (len + (p - (u_char *)ip6) > maxlen) {
2530 			LLE_RUNLOCK(ln);
2531 			goto nolladdropt;
2532 		}
2533 		if (ln->la_flags & LLE_VALID) {
2534 			nd_opt = (struct nd_opt_hdr *)p;
2535 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2536 			nd_opt->nd_opt_len = len >> 3;
2537 			lladdr = (char *)(nd_opt + 1);
2538 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2539 			p += len;
2540 		}
2541 		LLE_RUNLOCK(ln);
2542 	}
2543   nolladdropt:;
2544 
2545 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2546 
2547 	/* just to be safe */
2548 	if (m0->m_flags & M_DECRYPTED)
2549 		goto noredhdropt;
2550 	if (p - (u_char *)ip6 > maxlen)
2551 		goto noredhdropt;
2552 
2553 	{
2554 		/* redirected header option */
2555 		int len;
2556 		struct nd_opt_rd_hdr *nd_opt_rh;
2557 
2558 		/*
2559 		 * compute the maximum size for icmp6 redirect header option.
2560 		 * XXX room for auth header?
2561 		 */
2562 		len = maxlen - (p - (u_char *)ip6);
2563 		len &= ~7;
2564 
2565 		/*
2566 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2567 		 * about padding/truncate rule for the original IP packet.
2568 		 * From the discussion on IPv6imp in Feb 1999,
2569 		 * the consensus was:
2570 		 * - "attach as much as possible" is the goal
2571 		 * - pad if not aligned (original size can be guessed by
2572 		 *   original ip6 header)
2573 		 * Following code adds the padding if it is simple enough,
2574 		 * and truncates if not.
2575 		 */
2576 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2577 			/* not enough room, truncate */
2578 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2579 			    m0->m_pkthdr.len);
2580 		} else {
2581 			/*
2582 			 * enough room, truncate if not aligned.
2583 			 * we don't pad here for simplicity.
2584 			 */
2585 			size_t extra;
2586 
2587 			extra = m0->m_pkthdr.len % 8;
2588 			if (extra) {
2589 				/* truncate */
2590 				m_adj(m0, -extra);
2591 			}
2592 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2593 		}
2594 
2595 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2596 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2597 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2598 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2599 		p += sizeof(*nd_opt_rh);
2600 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2601 
2602 		/* connect m0 to m */
2603 		m->m_pkthdr.len += m0->m_pkthdr.len;
2604 		m_cat(m, m0);
2605 		m0 = NULL;
2606 	}
2607 noredhdropt:
2608 	if (m0) {
2609 		m_freem(m0);
2610 		m0 = NULL;
2611 	}
2612 
2613 	/* XXX: clear embedded link IDs in the inner header */
2614 	in6_clearscope(&sip6->ip6_src);
2615 	in6_clearscope(&sip6->ip6_dst);
2616 	in6_clearscope(&nd_rd->nd_rd_target);
2617 	in6_clearscope(&nd_rd->nd_rd_dst);
2618 
2619 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2620 
2621 	nd_rd->nd_rd_cksum = 0;
2622 	nd_rd->nd_rd_cksum
2623 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2624 
2625 	/* send the packet to outside... */
2626 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2627 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2628 
2629 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2630 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2631 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2632 
2633 	return;
2634 
2635 fail:
2636 	if (m)
2637 		m_freem(m);
2638 	if (m0)
2639 		m_freem(m0);
2640 }
2641 
2642 /*
2643  * ICMPv6 socket option processing.
2644  */
2645 int
2646 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2647 {
2648 	int error = 0;
2649 	struct in6pcb *in6p = sotoin6pcb(so);
2650 
2651 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2652 		return rip6_ctloutput(op, so, sopt);
2653 
2654 	switch (op) {
2655 	case PRCO_SETOPT:
2656 		switch (sopt->sopt_name) {
2657 		case ICMP6_FILTER:
2658 		    {
2659 			struct icmp6_filter fil;
2660 
2661 			error = sockopt_get(sopt, &fil, sizeof(fil));
2662 			if (error)
2663 				break;
2664 			memcpy(in6p->in6p_icmp6filt, &fil,
2665 			    sizeof(struct icmp6_filter));
2666 			error = 0;
2667 			break;
2668 		    }
2669 
2670 		default:
2671 			error = ENOPROTOOPT;
2672 			break;
2673 		}
2674 		break;
2675 
2676 	case PRCO_GETOPT:
2677 		switch (sopt->sopt_name) {
2678 		case ICMP6_FILTER:
2679 		    {
2680 			if (in6p->in6p_icmp6filt == NULL) {
2681 				error = EINVAL;
2682 				break;
2683 			}
2684 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2685 			    sizeof(struct icmp6_filter));
2686 			break;
2687 		    }
2688 
2689 		default:
2690 			error = ENOPROTOOPT;
2691 			break;
2692 		}
2693 		break;
2694 	}
2695 
2696 	return (error);
2697 }
2698 
2699 /*
2700  * Perform rate limit check.
2701  * Returns 0 if it is okay to send the icmp6 packet.
2702  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2703  * limitation.
2704  *
2705  * XXX per-destination/type check necessary?
2706  */
2707 static int
2708 icmp6_ratelimit(
2709 	const struct in6_addr *dst,	/* not used at this moment */
2710 	const int type,		/* not used at this moment */
2711 	const int code)		/* not used at this moment */
2712 {
2713 	int ret;
2714 
2715 	ret = 0;	/* okay to send */
2716 
2717 	/* PPS limit */
2718 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2719 	    icmp6errppslim)) {
2720 		/* The packet is subject to rate limit */
2721 		ret++;
2722 	}
2723 
2724 	return ret;
2725 }
2726 
2727 static struct rtentry *
2728 icmp6_mtudisc_clone(struct sockaddr *dst)
2729 {
2730 	struct rtentry *rt;
2731 	int    error;
2732 
2733 	rt = rtalloc1(dst, 1);
2734 	if (rt == 0)
2735 		return NULL;
2736 
2737 	/* If we didn't get a host route, allocate one */
2738 	if ((rt->rt_flags & RTF_HOST) == 0) {
2739 		struct rtentry *nrt;
2740 
2741 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2742 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2743 		if (error) {
2744 			rtfree(rt);
2745 			return NULL;
2746 		}
2747 		nrt->rt_rmx = rt->rt_rmx;
2748 		rtfree(rt);
2749 		rt = nrt;
2750 	}
2751 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2752 			icmp6_mtudisc_timeout_q);
2753 	if (error) {
2754 		rtfree(rt);
2755 		return NULL;
2756 	}
2757 
2758 	return rt;	/* caller need to call rtfree() */
2759 }
2760 
2761 static void
2762 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2763 {
2764 
2765 	KASSERT(rt != NULL);
2766 	rt_assert_referenced(rt);
2767 
2768 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2769 	    (RTF_DYNAMIC | RTF_HOST)) {
2770 		rtrequest(RTM_DELETE, rt_getkey(rt),
2771 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2772 	} else {
2773 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2774 			rt->rt_rmx.rmx_mtu = 0;
2775 	}
2776 }
2777 
2778 static void
2779 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2780 {
2781 
2782 	KASSERT(rt != NULL);
2783 	rt_assert_referenced(rt);
2784 
2785 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2786 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2787 		rtrequest(RTM_DELETE, rt_getkey(rt),
2788 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2789 	}
2790 }
2791 
2792 /*
2793  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2794  */
2795 static int
2796 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2797 {
2798 	(void)&name;
2799 	(void)&l;
2800 	(void)&oname;
2801 
2802 	if (namelen != 0)
2803 		return (EINVAL);
2804 
2805 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2806 	    /*XXXUNCONST*/
2807 	    __UNCONST(newp), newlen));
2808 }
2809 
2810 static int
2811 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2812 {
2813 
2814 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2815 }
2816 
2817 static int
2818 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2819 {
2820 	int error, tmp;
2821 	struct sysctlnode node;
2822 
2823 	node = *rnode;
2824 	node.sysctl_data = &tmp;
2825 	tmp = icmp6_redirtimeout;
2826 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2827 	if (error || newp == NULL)
2828 		return error;
2829 	if (tmp < 0)
2830 		return EINVAL;
2831 	icmp6_redirtimeout = tmp;
2832 
2833 	if (icmp6_redirect_timeout_q != NULL) {
2834 		if (icmp6_redirtimeout == 0) {
2835 			rt_timer_queue_destroy(icmp6_redirect_timeout_q,
2836 			    true);
2837 		} else {
2838 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2839 			    icmp6_redirtimeout);
2840 		}
2841 	} else if (icmp6_redirtimeout > 0) {
2842 		icmp6_redirect_timeout_q =
2843 		    rt_timer_queue_create(icmp6_redirtimeout);
2844 	}
2845 
2846 	return 0;
2847 }
2848 
2849 static void
2850 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2851 {
2852 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2853 
2854 	sysctl_createv(clog, 0, NULL, NULL,
2855 		       CTLFLAG_PERMANENT,
2856 		       CTLTYPE_NODE, "inet6", NULL,
2857 		       NULL, 0, NULL, 0,
2858 		       CTL_NET, PF_INET6, CTL_EOL);
2859 	sysctl_createv(clog, 0, NULL, NULL,
2860 		       CTLFLAG_PERMANENT,
2861 		       CTLTYPE_NODE, "icmp6",
2862 		       SYSCTL_DESCR("ICMPv6 related settings"),
2863 		       NULL, 0, NULL, 0,
2864 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2865 
2866 	sysctl_createv(clog, 0, NULL, NULL,
2867 		       CTLFLAG_PERMANENT,
2868 		       CTLTYPE_STRUCT, "stats",
2869 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2870 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2871 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2872 		       ICMPV6CTL_STATS, CTL_EOL);
2873 	sysctl_createv(clog, 0, NULL, NULL,
2874 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2875 		       CTLTYPE_INT, "rediraccept",
2876 		       SYSCTL_DESCR("Accept and process redirect messages"),
2877 		       NULL, 0, &icmp6_rediraccept, 0,
2878 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2879 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2880 	sysctl_createv(clog, 0, NULL, NULL,
2881 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2882 		       CTLTYPE_INT, "redirtimeout",
2883 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2884 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2885 		       &icmp6_redirtimeout, 0,
2886 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2887 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2888 #if 0 /* obsoleted */
2889 	sysctl_createv(clog, 0, NULL, NULL,
2890 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2891 		       CTLTYPE_INT, "errratelimit", NULL,
2892 		       NULL, 0, &icmp6_errratelimit, 0,
2893 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2894 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2895 #endif
2896 	sysctl_createv(clog, 0, NULL, NULL,
2897 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2898 		       CTLTYPE_INT, "nd6_prune",
2899 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2900 		       NULL, 0, &nd6_prune, 0,
2901 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2902 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2903 	sysctl_createv(clog, 0, NULL, NULL,
2904 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2905 		       CTLTYPE_INT, "nd6_delay",
2906 		       SYSCTL_DESCR("First probe delay time"),
2907 		       NULL, 0, &nd6_delay, 0,
2908 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2909 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2910 	sysctl_createv(clog, 0, NULL, NULL,
2911 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2912 		       CTLTYPE_INT, "nd6_umaxtries",
2913 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2914 		       NULL, 0, &nd6_umaxtries, 0,
2915 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2916 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2917 	sysctl_createv(clog, 0, NULL, NULL,
2918 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2919 		       CTLTYPE_INT, "nd6_mmaxtries",
2920 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2921 		       NULL, 0, &nd6_mmaxtries, 0,
2922 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2923 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2924 	sysctl_createv(clog, 0, NULL, NULL,
2925 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2926 		       CTLTYPE_INT, "nd6_useloopback",
2927 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2928 		       NULL, 0, &nd6_useloopback, 0,
2929 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2930 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2931 #if 0 /* obsoleted */
2932 	sysctl_createv(clog, 0, NULL, NULL,
2933 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2934 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2935 		       NULL, 0, &nd6_proxyall, 0,
2936 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2937 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2938 #endif
2939 	sysctl_createv(clog, 0, NULL, NULL,
2940 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2941 		       CTLTYPE_INT, "nodeinfo",
2942 		       SYSCTL_DESCR("Respond to node information requests"),
2943 		       NULL, 0, &icmp6_nodeinfo, 0,
2944 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2945 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2946 	sysctl_createv(clog, 0, NULL, NULL,
2947 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2948 		       CTLTYPE_INT, "errppslimit",
2949 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2950 		       NULL, 0, &icmp6errppslim, 0,
2951 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2952 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2953 	sysctl_createv(clog, 0, NULL, NULL,
2954 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2955 		       CTLTYPE_INT, "nd6_maxnudhint",
2956 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2957 		       NULL, 0, &nd6_maxnudhint, 0,
2958 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2959 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2960 	sysctl_createv(clog, 0, NULL, NULL,
2961 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2962 		       CTLTYPE_INT, "mtudisc_hiwat",
2963 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2964 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2965 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2966 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2967 	sysctl_createv(clog, 0, NULL, NULL,
2968 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2969 		       CTLTYPE_INT, "mtudisc_lowat",
2970 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2971 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2972 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2973 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2974 	sysctl_createv(clog, 0, NULL, NULL,
2975 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2976 		       CTLTYPE_INT, "nd6_debug",
2977 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2978 		       NULL, 0, &nd6_debug, 0,
2979 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2980 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2981 	sysctl_createv(clog, 0, NULL, NULL,
2982 		       CTLFLAG_PERMANENT,
2983 		       CTLTYPE_STRUCT, "nd6_drlist",
2984 		       SYSCTL_DESCR("Default router list"),
2985 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2986 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2987 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2988 	sysctl_createv(clog, 0, NULL, NULL,
2989 		       CTLFLAG_PERMANENT,
2990 		       CTLTYPE_STRUCT, "nd6_prlist",
2991 		       SYSCTL_DESCR("Prefix list"),
2992 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2993 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2994 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2995 	sysctl_createv(clog, 0, NULL, NULL,
2996 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2997 		       CTLTYPE_INT, "maxqueuelen",
2998 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2999 		       NULL, 1, &nd6_maxqueuelen, 0,
3000 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3001 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3002 }
3003 
3004 void
3005 icmp6_statinc(u_int stat)
3006 {
3007 
3008 	KASSERT(stat < ICMP6_NSTATS);
3009 	ICMP6_STATINC(stat);
3010 }
3011