xref: /netbsd-src/sys/netinet6/icmp6.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: icmp6.c,v 1.209 2017/02/13 07:18:20 ozaki-r 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 1.209 2017/02/13 07:18:20 ozaki-r 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 <netinet/wqinput.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/ip6_private.h>
96 #include <netinet/icmp6.h>
97 #include <netinet6/icmp6_private.h>
98 #include <netinet6/mld6_var.h>
99 #include <netinet6/in6_pcb.h>
100 #include <netinet6/nd6.h>
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/ip6protosw.h>
103 #include <netinet6/scope6_var.h>
104 
105 #ifdef IPSEC
106 #include <netipsec/ipsec.h>
107 #include <netipsec/key.h>
108 #endif
109 
110 
111 #include "faith.h"
112 #if defined(NFAITH) && 0 < NFAITH
113 #include <net/if_faith.h>
114 #endif
115 
116 #include <net/net_osdep.h>
117 
118 extern struct domain inet6domain;
119 
120 percpu_t *icmp6stat_percpu;
121 
122 extern struct inpcbtable raw6cbtable;
123 extern int icmp6errppslim;
124 static int icmp6errpps_count = 0;
125 static struct timeval icmp6errppslim_last;
126 extern int icmp6_nodeinfo;
127 
128 /*
129  * List of callbacks to notify when Path MTU changes are made.
130  */
131 struct icmp6_mtudisc_callback {
132 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
133 	void (*mc_func)(struct in6_addr *);
134 };
135 
136 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
137     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
138 
139 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
140 extern int pmtu_expire;
141 
142 /* XXX do these values make any sense? */
143 static int icmp6_mtudisc_hiwat = 1280;
144 static int icmp6_mtudisc_lowat = 256;
145 
146 /*
147  * keep track of # of redirect routes.
148  */
149 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
150 
151 /* XXX experimental, turned off */
152 static int icmp6_redirect_hiwat = -1;
153 static int icmp6_redirect_lowat = -1;
154 
155 /* Protect mtudisc and redirect stuffs */
156 static kmutex_t icmp6_mtx __cacheline_aligned;
157 
158 static void icmp6_errcount(u_int, int, int);
159 static int icmp6_rip6_input(struct mbuf **, int);
160 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
161 static const char *icmp6_redirect_diag(char *, size_t, struct in6_addr *,
162 	struct in6_addr *, struct in6_addr *);
163 static struct mbuf *ni6_input(struct mbuf *, int);
164 static struct mbuf *ni6_nametodns(const char *, int, int);
165 static int ni6_dnsmatch(const char *, int, const char *, int);
166 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
167 			  struct ifnet **, char *, struct psref *);
168 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
169 				struct ifnet *, int);
170 static int icmp6_notify_error(struct mbuf *, int, int, int);
171 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
172 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
173 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
174 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
175 
176 /* workqueue-based pr_input */
177 static struct wqinput *icmp6_wqinput;
178 static void _icmp6_input(struct mbuf *m, int off, int proto);
179 
180 void
181 icmp6_init(void)
182 {
183 
184 	sysctl_net_inet6_icmp6_setup(NULL);
185 	mld_init();
186 
187 	mutex_init(&icmp6_mtx, MUTEX_DEFAULT, IPL_NONE);
188 	mutex_enter(&icmp6_mtx);
189 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
190 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
191 	mutex_exit(&icmp6_mtx);
192 
193 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
194 
195 	icmp6_wqinput = wqinput_create("icmp6", _icmp6_input);
196 }
197 
198 static void
199 icmp6_errcount(u_int base, int type, int code)
200 {
201 	switch (type) {
202 	case ICMP6_DST_UNREACH:
203 		switch (code) {
204 		case ICMP6_DST_UNREACH_NOROUTE:
205 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
206 			return;
207 		case ICMP6_DST_UNREACH_ADMIN:
208 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
209 			return;
210 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
211 			ICMP6_STATINC(base +
212 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
213 			return;
214 		case ICMP6_DST_UNREACH_ADDR:
215 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
216 			return;
217 		case ICMP6_DST_UNREACH_NOPORT:
218 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
219 			return;
220 		}
221 		break;
222 	case ICMP6_PACKET_TOO_BIG:
223 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
224 		return;
225 	case ICMP6_TIME_EXCEEDED:
226 		switch (code) {
227 		case ICMP6_TIME_EXCEED_TRANSIT:
228 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
229 			return;
230 		case ICMP6_TIME_EXCEED_REASSEMBLY:
231 			ICMP6_STATINC(base +
232 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
233 			return;
234 		}
235 		break;
236 	case ICMP6_PARAM_PROB:
237 		switch (code) {
238 		case ICMP6_PARAMPROB_HEADER:
239 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
240 			return;
241 		case ICMP6_PARAMPROB_NEXTHEADER:
242 			ICMP6_STATINC(base +
243 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
244 			return;
245 		case ICMP6_PARAMPROB_OPTION:
246 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
247 			return;
248 		}
249 		break;
250 	case ND_REDIRECT:
251 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
252 		return;
253 	}
254 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
255 }
256 
257 /*
258  * Register a Path MTU Discovery callback.
259  */
260 void
261 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
262 {
263 	struct icmp6_mtudisc_callback *mc, *new;
264 
265 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
266 
267 	mutex_enter(&icmp6_mtx);
268 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
269 	     mc = LIST_NEXT(mc, mc_list)) {
270 		if (mc->mc_func == func) {
271 			mutex_exit(&icmp6_mtx);
272 			kmem_free(new, sizeof(*mc));
273 			return;
274 		}
275 	}
276 
277 	new->mc_func = func;
278 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, new, mc_list);
279 	mutex_exit(&icmp6_mtx);
280 }
281 
282 /*
283  * A wrapper function for icmp6_error() necessary when the erroneous packet
284  * may not contain enough scope zone information.
285  */
286 void
287 icmp6_error2(struct mbuf *m, int type, int code, int param,
288 	struct ifnet *ifp)
289 {
290 	struct ip6_hdr *ip6;
291 
292 	if (ifp == NULL)
293 		return;
294 
295 	if (m->m_len < sizeof(struct ip6_hdr)) {
296 		m = m_pullup(m, sizeof(struct ip6_hdr));
297 		if (m == NULL)
298 			return;
299 	}
300 
301 	ip6 = mtod(m, struct ip6_hdr *);
302 
303 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
304 		return;
305 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
306 		return;
307 
308 	icmp6_error(m, type, code, param);
309 }
310 
311 /*
312  * Generate an error packet of type error in response to bad IP6 packet.
313  */
314 void
315 icmp6_error(struct mbuf *m, int type, int code, int param)
316 {
317 	struct ip6_hdr *oip6, *nip6;
318 	struct icmp6_hdr *icmp6;
319 	u_int preplen;
320 	int off;
321 	int nxt;
322 
323 	ICMP6_STATINC(ICMP6_STAT_ERROR);
324 
325 	/* count per-type-code statistics */
326 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
327 
328 	if (m->m_flags & M_DECRYPTED) {
329 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
330 		goto freeit;
331 	}
332 
333 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
334 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
335 		return;
336 	oip6 = mtod(m, struct ip6_hdr *);
337 
338 	/*
339 	 * If the destination address of the erroneous packet is a multicast
340 	 * address, or the packet was sent using link-layer multicast,
341 	 * we should basically suppress sending an error (RFC 2463, Section
342 	 * 2.4).
343 	 * We have two exceptions (the item e.2 in that section):
344 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
345 	 * - the Parameter Problem Message that can be allowed an icmp6 error
346 	 *   in the option type field.  This check has been done in
347 	 *   ip6_unknown_opt(), so we can just check the type and code.
348 	 */
349 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
350 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
351 	    (type != ICMP6_PACKET_TOO_BIG &&
352 	     (type != ICMP6_PARAM_PROB ||
353 	      code != ICMP6_PARAMPROB_OPTION)))
354 		goto freeit;
355 
356 	/*
357 	 * RFC 2463, 2.4 (e.5): source address check.
358 	 * XXX: the case of anycast source?
359 	 */
360 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
361 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
362 		goto freeit;
363 
364 	/*
365 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
366 	 * don't do it.
367 	 */
368 	nxt = -1;
369 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
370 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
371 		struct icmp6_hdr *icp;
372 
373 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
374 			sizeof(*icp));
375 		if (icp == NULL) {
376 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
377 			return;
378 		}
379 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
380 		    icp->icmp6_type == ND_REDIRECT) {
381 			/*
382 			 * ICMPv6 error
383 			 * Special case: for redirect (which is
384 			 * informational) we must not send icmp6 error.
385 			 */
386 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
387 			goto freeit;
388 		} else {
389 			/* ICMPv6 informational - send the error */
390 		}
391 	}
392 #if 0 /* controversial */
393 	else if (off >= 0 && nxt == IPPROTO_ESP) {
394 		/*
395 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
396 		 * don't respond.
397 		 */
398 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
399 		goto freeit;
400 	}
401 #endif
402 	else {
403 		/* non-ICMPv6 - send the error */
404 	}
405 
406 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
407 
408 	/* Finally, do rate limitation check. */
409 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
410 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
411 		goto freeit;
412 	}
413 
414 	/*
415 	 * OK, ICMP6 can be generated.
416 	 */
417 
418 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
419 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
420 
421 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
422 	M_PREPEND(m, preplen, M_DONTWAIT);
423 	if (m && M_UNWRITABLE(m, preplen))
424 		m = m_pullup(m, preplen);
425 	if (m == NULL) {
426 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
427 		return;
428 	}
429 
430 	nip6 = mtod(m, struct ip6_hdr *);
431 	nip6->ip6_src  = oip6->ip6_src;
432 	nip6->ip6_dst  = oip6->ip6_dst;
433 
434 	in6_clearscope(&oip6->ip6_src);
435 	in6_clearscope(&oip6->ip6_dst);
436 
437 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
438 	icmp6->icmp6_type = type;
439 	icmp6->icmp6_code = code;
440 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
441 
442 	/*
443 	 * icmp6_reflect() is designed to be in the input path.
444 	 * icmp6_error() can be called from both input and output path,
445 	 * and if we are in output path rcvif could contain bogus value.
446 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
447 	 * information in ip header (nip6).
448 	 */
449 	m_reset_rcvif(m);
450 
451 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
452 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
453 
454 	return;
455 
456   freeit:
457 	/*
458 	 * If we can't tell whether or not we can generate ICMP6, free it.
459 	 */
460 	m_freem(m);
461 }
462 
463 /*
464  * Process a received ICMP6 message.
465  */
466 static void
467 _icmp6_input(struct mbuf *m, int off, int proto)
468 {
469 	struct mbuf *n;
470 	struct ip6_hdr *ip6, *nip6;
471 	struct icmp6_hdr *icmp6, *nicmp6;
472 	int icmp6len = m->m_pkthdr.len - off;
473 	int code, sum, noff;
474 	struct ifnet *rcvif;
475 	struct psref psref;
476 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
477 
478 	rcvif = m_get_rcvif_psref(m, &psref);
479 	if (__predict_false(rcvif == NULL))
480 		goto freeit;
481 
482 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
483 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
484 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
485 
486 	/*
487 	 * Locate icmp6 structure in mbuf, and check
488 	 * that not corrupted and of at least minimum length
489 	 */
490 
491 	if (icmp6len < sizeof(struct icmp6_hdr)) {
492 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
493 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
494 		goto freeit;
495 	}
496 
497 	ip6 = mtod(m, struct ip6_hdr *);
498 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
499 	if (icmp6 == NULL) {
500 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
501 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
502 		goto freeit;
503 	}
504 	/*
505 	 * Enforce alignment requirements that are violated in
506 	 * some cases, see kern/50766 for details.
507 	 */
508 	if (IP6_HDR_ALIGNED_P(icmp6) == 0) {
509 		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
510 		if (m == NULL) {
511 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
512 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
513 			goto freeit;
514 		}
515 		ip6 = mtod(m, struct ip6_hdr *);
516 		icmp6 = (struct icmp6_hdr *)(ip6 + 1);
517 	}
518 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
519 
520 	/*
521 	 * calculate the checksum
522 	 */
523 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
524 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
525 		    icmp6->icmp6_type, sum, IN6_PRINT(ip6buf, &ip6->ip6_src));
526 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
527 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
528 		goto freeit;
529 	}
530 
531 #if defined(NFAITH) && 0 < NFAITH
532 	if (faithprefix(&ip6->ip6_dst)) {
533 		/*
534 		 * Deliver very specific ICMP6 type only.
535 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
536 		 * will not work.
537 		 */
538 		switch (icmp6->icmp6_type) {
539 		case ICMP6_DST_UNREACH:
540 		case ICMP6_PACKET_TOO_BIG:
541 		case ICMP6_TIME_EXCEEDED:
542 			break;
543 		default:
544 			goto freeit;
545 		}
546 	}
547 #endif
548 
549 	code = icmp6->icmp6_code;
550 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
551 
552 	switch (icmp6->icmp6_type) {
553 	case ICMP6_DST_UNREACH:
554 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
555 		switch (code) {
556 		case ICMP6_DST_UNREACH_NOROUTE:
557 			code = PRC_UNREACH_NET;
558 			break;
559 		case ICMP6_DST_UNREACH_ADMIN:
560 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
561 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
562 			break;
563 		case ICMP6_DST_UNREACH_ADDR:
564 			code = PRC_HOSTDEAD;
565 			break;
566 #ifdef COMPAT_RFC1885
567 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
568 			code = PRC_UNREACH_SRCFAIL;
569 			break;
570 #else
571 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
572 			/* I mean "source address was incorrect." */
573 			code = PRC_UNREACH_NET;
574 			break;
575 #endif
576 		case ICMP6_DST_UNREACH_NOPORT:
577 			code = PRC_UNREACH_PORT;
578 			break;
579 		default:
580 			goto badcode;
581 		}
582 		goto deliver;
583 
584 	case ICMP6_PACKET_TOO_BIG:
585 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
586 
587 		/*
588 		 * MTU is checked in icmp6_mtudisc.
589 		 */
590 		code = PRC_MSGSIZE;
591 
592 		/*
593 		 * Updating the path MTU will be done after examining
594 		 * intermediate extension headers.
595 		 */
596 		goto deliver;
597 
598 	case ICMP6_TIME_EXCEEDED:
599 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
600 		switch (code) {
601 		case ICMP6_TIME_EXCEED_TRANSIT:
602 			code = PRC_TIMXCEED_INTRANS;
603 			break;
604 		case ICMP6_TIME_EXCEED_REASSEMBLY:
605 			code = PRC_TIMXCEED_REASS;
606 			break;
607 		default:
608 			goto badcode;
609 		}
610 		goto deliver;
611 
612 	case ICMP6_PARAM_PROB:
613 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
614 		switch (code) {
615 		case ICMP6_PARAMPROB_NEXTHEADER:
616 			code = PRC_UNREACH_PROTOCOL;
617 			break;
618 		case ICMP6_PARAMPROB_HEADER:
619 		case ICMP6_PARAMPROB_OPTION:
620 			code = PRC_PARAMPROB;
621 			break;
622 		default:
623 			goto badcode;
624 		}
625 		goto deliver;
626 
627 	case ICMP6_ECHO_REQUEST:
628 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
629 		if (code != 0)
630 			goto badcode;
631 		/*
632 		 * Copy mbuf to send to two data paths: userland socket(s),
633 		 * and to the querier (echo reply).
634 		 * m: a copy for socket, n: a copy for querier
635 		 *
636 		 * If the first mbuf is shared, or the first mbuf is too short,
637 		 * copy the first part of the data into a fresh mbuf.
638 		 * Otherwise, we will wrongly overwrite both copies.
639 		 */
640 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
641 			/* Give up local */
642 			n = m;
643 			m = NULL;
644 		} else if (M_READONLY(n) ||
645 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
646 			struct mbuf *n0 = n;
647 
648 			/*
649 			 * Prepare an internal mbuf.  m_pullup() doesn't
650 			 * always copy the length we specified.
651 			 */
652 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
653 				/* Give up local */
654 				n = m;
655 				m = NULL;
656 			}
657 			m_freem(n0);
658 		}
659 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
660 		    sizeof(*nicmp6));
661 		if (nicmp6 == NULL)
662 			goto freeit;
663 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
664 		nicmp6->icmp6_code = 0;
665 		if (n) {
666 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
667 			icmp6s[ICMP6_STAT_REFLECT]++;
668 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
669 			ICMP6_STAT_PUTREF();
670 			icmp6_reflect(n, off);
671 		}
672 		if (!m)
673 			goto freeit;
674 		break;
675 
676 	case ICMP6_ECHO_REPLY:
677 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
678 		if (code != 0)
679 			goto badcode;
680 		break;
681 
682 	case MLD_LISTENER_QUERY:
683 	case MLD_LISTENER_REPORT:
684 		if (icmp6len < sizeof(struct mld_hdr))
685 			goto badlen;
686 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
687 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
688 		else
689 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
690 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
691 			/* give up local */
692 			mld_input(m, off);
693 			m = NULL;
694 			goto freeit;
695 		}
696 		mld_input(n, off);
697 		/* m stays. */
698 		break;
699 
700 	case MLD_LISTENER_DONE:
701 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
702 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
703 			goto badlen;
704 		break;		/* nothing to be done in kernel */
705 
706 	case MLD_MTRACE_RESP:
707 	case MLD_MTRACE:
708 		/* XXX: these two are experimental.  not officially defined. */
709 		/* XXX: per-interface statistics? */
710 		break;		/* just pass it to applications */
711 
712 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
713 	    {
714 		enum { WRU, FQDN } mode;
715 
716 		if (!icmp6_nodeinfo)
717 			break;
718 
719 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
720 			mode = WRU;
721 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
722 			mode = FQDN;
723 		else
724 			goto badlen;
725 
726 		if (mode == FQDN) {
727 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
728 			if (n)
729 				n = ni6_input(n, off);
730 			/* XXX meaningless if n == NULL */
731 			noff = sizeof(struct ip6_hdr);
732 		} else {
733 			u_char *p;
734 			int maxhlen;
735 
736 			if ((icmp6_nodeinfo & 5) != 5)
737 				break;
738 
739 			if (code != 0)
740 				goto badcode;
741 			MGETHDR(n, M_DONTWAIT, m->m_type);
742 			if (n && ICMP6_MAXLEN > MHLEN) {
743 				MCLGET(n, M_DONTWAIT);
744 				if ((n->m_flags & M_EXT) == 0) {
745 					m_free(n);
746 					n = NULL;
747 				}
748 			}
749 			if (n == NULL) {
750 				/* Give up remote */
751 				break;
752 			}
753 			m_reset_rcvif(n);
754 			n->m_len = 0;
755 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
756 			if (maxhlen < 0)
757 				break;
758 			if (maxhlen > hostnamelen)
759 				maxhlen = hostnamelen;
760 			/*
761 			 * Copy IPv6 and ICMPv6 only.
762 			 */
763 			nip6 = mtod(n, struct ip6_hdr *);
764 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
765 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
766 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
767 			p = (u_char *)(nicmp6 + 1);
768 			memset(p, 0, 4);
769 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
770 			noff = sizeof(struct ip6_hdr);
771 			M_COPY_PKTHDR(n, m); /* just for rcvif */
772 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
773 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
774 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
775 			nicmp6->icmp6_code = 0;
776 		}
777 #undef hostnamelen
778 		if (n) {
779 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
780 			icmp6s[ICMP6_STAT_REFLECT]++;
781 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
782 			ICMP6_STAT_PUTREF();
783 			icmp6_reflect(n, noff);
784 		}
785 		break;
786 	    }
787 
788 	case ICMP6_WRUREPLY:
789 		if (code != 0)
790 			goto badcode;
791 		break;
792 
793 	case ND_ROUTER_SOLICIT:
794 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_router_solicit))
798 			goto badlen;
799 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800 			/* give up local */
801 			nd6_rs_input(m, off, icmp6len);
802 			m = NULL;
803 			goto freeit;
804 		}
805 		nd6_rs_input(n, off, icmp6len);
806 		/* m stays. */
807 		break;
808 
809 	case ND_ROUTER_ADVERT:
810 		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
811 		if (code != 0)
812 			goto badcode;
813 		if (icmp6len < sizeof(struct nd_router_advert))
814 			goto badlen;
815 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
816 			/* give up local */
817 			nd6_ra_input(m, off, icmp6len);
818 			m = NULL;
819 			goto freeit;
820 		}
821 		nd6_ra_input(n, off, icmp6len);
822 		/* m stays. */
823 		break;
824 
825 	case ND_NEIGHBOR_SOLICIT:
826 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
827 		if (code != 0)
828 			goto badcode;
829 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
830 			goto badlen;
831 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
832 			/* give up local */
833 			nd6_ns_input(m, off, icmp6len);
834 			m = NULL;
835 			goto freeit;
836 		}
837 		nd6_ns_input(n, off, icmp6len);
838 		/* m stays. */
839 		break;
840 
841 	case ND_NEIGHBOR_ADVERT:
842 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
843 		if (code != 0)
844 			goto badcode;
845 		if (icmp6len < sizeof(struct nd_neighbor_advert))
846 			goto badlen;
847 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
848 			/* give up local */
849 			nd6_na_input(m, off, icmp6len);
850 			m = NULL;
851 			goto freeit;
852 		}
853 		nd6_na_input(n, off, icmp6len);
854 		/* m stays. */
855 		break;
856 
857 	case ND_REDIRECT:
858 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
859 		if (code != 0)
860 			goto badcode;
861 		if (icmp6len < sizeof(struct nd_redirect))
862 			goto badlen;
863 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
864 			/* give up local */
865 			icmp6_redirect_input(m, off);
866 			m = NULL;
867 			goto freeit;
868 		}
869 		icmp6_redirect_input(n, off);
870 		/* m stays. */
871 		break;
872 
873 	case ICMP6_ROUTER_RENUMBERING:
874 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
875 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
876 			goto badcode;
877 		if (icmp6len < sizeof(struct icmp6_router_renum))
878 			goto badlen;
879 		break;
880 
881 	default:
882 		nd6log(LOG_DEBUG,
883 		    "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
884 		    icmp6->icmp6_type,
885 		    IN6_PRINT(ip6buf, &ip6->ip6_src),
886 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst),
887 		    rcvif ? rcvif->if_index : 0);
888 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
889 			/* ICMPv6 error: MUST deliver it by spec... */
890 			code = PRC_NCMDS;
891 			/* deliver */
892 		} else {
893 			/* ICMPv6 informational: MUST not deliver */
894 			break;
895 		}
896 	deliver:
897 		if (icmp6_notify_error(m, off, icmp6len, code)) {
898 			/* In this case, m should've been freed. */
899 			m_put_rcvif_psref(rcvif, &psref);
900 			return;
901 		}
902 		break;
903 
904 	badcode:
905 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
906 		break;
907 
908 	badlen:
909 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
910 		break;
911 	}
912 	m_put_rcvif_psref(rcvif, &psref);
913 
914 	/* deliver the packet to appropriate sockets */
915 	icmp6_rip6_input(&m, off);
916 
917 	return;
918 
919  freeit:
920 	m_put_rcvif_psref(rcvif, &psref);
921 	m_freem(m);
922 	return;
923 }
924 
925 int
926 icmp6_input(struct mbuf **mp, int *offp, int proto)
927 {
928 
929 	wqinput_input(icmp6_wqinput, *mp, *offp, proto);
930 
931 	return IPPROTO_DONE;
932 }
933 
934 static int
935 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
936 {
937 	struct icmp6_hdr *icmp6;
938 	struct ip6_hdr *eip6;
939 	u_int32_t notifymtu;
940 	struct sockaddr_in6 icmp6src, icmp6dst;
941 
942 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
943 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
944 		goto freeit;
945 	}
946 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
947 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
948 	if (icmp6 == NULL) {
949 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
950 		return (-1);
951 	}
952 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
953 
954 	/* Detect the upper level protocol */
955 	{
956 		void (*ctlfunc)(int, struct sockaddr *, void *);
957 		u_int8_t nxt = eip6->ip6_nxt;
958 		int eoff = off + sizeof(struct icmp6_hdr) +
959 			sizeof(struct ip6_hdr);
960 		struct ip6ctlparam ip6cp;
961 		struct in6_addr *finaldst = NULL;
962 		int icmp6type = icmp6->icmp6_type;
963 		struct ip6_frag *fh;
964 		struct ip6_rthdr *rth;
965 		struct ip6_rthdr0 *rth0;
966 		int rthlen;
967 		struct ifnet *rcvif;
968 		int s;
969 
970 		while (1) { /* XXX: should avoid infinite loop explicitly? */
971 			struct ip6_ext *eh;
972 
973 			switch (nxt) {
974 			case IPPROTO_HOPOPTS:
975 			case IPPROTO_DSTOPTS:
976 			case IPPROTO_AH:
977 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
978 					       eoff, sizeof(*eh));
979 				if (eh == NULL) {
980 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
981 					return (-1);
982 				}
983 
984 				if (nxt == IPPROTO_AH)
985 					eoff += (eh->ip6e_len + 2) << 2;
986 				else
987 					eoff += (eh->ip6e_len + 1) << 3;
988 				nxt = eh->ip6e_nxt;
989 				break;
990 			case IPPROTO_ROUTING:
991 				/*
992 				 * When the erroneous packet contains a
993 				 * routing header, we should examine the
994 				 * header to determine the final destination.
995 				 * Otherwise, we can't properly update
996 				 * information that depends on the final
997 				 * destination (e.g. path MTU).
998 				 */
999 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1000 					       eoff, sizeof(*rth));
1001 				if (rth == NULL) {
1002 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1003 					return (-1);
1004 				}
1005 				rthlen = (rth->ip6r_len + 1) << 3;
1006 				/*
1007 				 * XXX: currently there is no
1008 				 * officially defined type other
1009 				 * than type-0.
1010 				 * Note that if the segment left field
1011 				 * is 0, all intermediate hops must
1012 				 * have been passed.
1013 				 */
1014 				if (rth->ip6r_segleft &&
1015 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1016 					int hops;
1017 
1018 					IP6_EXTHDR_GET(rth0,
1019 						       struct ip6_rthdr0 *, m,
1020 						       eoff, rthlen);
1021 					if (rth0 == NULL) {
1022 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1023 						return (-1);
1024 					}
1025 					/* just ignore a bogus header */
1026 					if ((rth0->ip6r0_len % 2) == 0 &&
1027 					    (hops = rth0->ip6r0_len/2))
1028 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1029 				}
1030 				eoff += rthlen;
1031 				nxt = rth->ip6r_nxt;
1032 				break;
1033 			case IPPROTO_FRAGMENT:
1034 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1035 					       eoff, sizeof(*fh));
1036 				if (fh == NULL) {
1037 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1038 					return (-1);
1039 				}
1040 				/*
1041 				 * Data after a fragment header is meaningless
1042 				 * unless it is the first fragment, but
1043 				 * we'll go to the notify label for path MTU
1044 				 * discovery.
1045 				 */
1046 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1047 					goto notify;
1048 
1049 				eoff += sizeof(struct ip6_frag);
1050 				nxt = fh->ip6f_nxt;
1051 				break;
1052 			default:
1053 				/*
1054 				 * This case includes ESP and the No Next
1055 				 * Header.  In such cases going to the notify
1056 				 * label does not have any meaning
1057 				 * (i.e. ctlfunc will be NULL), but we go
1058 				 * anyway since we might have to update
1059 				 * path MTU information.
1060 				 */
1061 				goto notify;
1062 			}
1063 		}
1064 	  notify:
1065 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1066 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1067 		if (icmp6 == NULL) {
1068 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1069 			return (-1);
1070 		}
1071 
1072 		/*
1073 		 * retrieve parameters from the inner IPv6 header, and convert
1074 		 * them into sockaddr structures.
1075 		 * XXX: there is no guarantee that the source or destination
1076 		 * addresses of the inner packet are in the same scope zone as
1077 		 * the addresses of the icmp packet.  But there is no other
1078 		 * way to determine the zone.
1079 		 */
1080 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1081 
1082 		rcvif = m_get_rcvif(m, &s);
1083 		if (__predict_false(rcvif == NULL))
1084 			goto freeit;
1085 		sockaddr_in6_init(&icmp6dst,
1086 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1087 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1088 			m_put_rcvif(rcvif, &s);
1089 			goto freeit;
1090 		}
1091 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1092 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1093 			m_put_rcvif(rcvif, &s);
1094 			goto freeit;
1095 		}
1096 		m_put_rcvif(rcvif, &s);
1097 
1098 		icmp6src.sin6_flowinfo =
1099 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1100 
1101 		if (finaldst == NULL)
1102 			finaldst = &eip6->ip6_dst;
1103 		ip6cp.ip6c_m = m;
1104 		ip6cp.ip6c_icmp6 = icmp6;
1105 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1106 		ip6cp.ip6c_off = eoff;
1107 		ip6cp.ip6c_finaldst = finaldst;
1108 		ip6cp.ip6c_src = &icmp6src;
1109 		ip6cp.ip6c_nxt = nxt;
1110 
1111 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1112 			notifymtu = ntohl(icmp6->icmp6_mtu);
1113 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1114 		}
1115 
1116 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1117 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1118 		if (ctlfunc) {
1119 			(void) (*ctlfunc)(code, sin6tosa(&icmp6dst),
1120 					  &ip6cp);
1121 		}
1122 	}
1123 	return (0);
1124 
1125   freeit:
1126 	m_freem(m);
1127 	return (-1);
1128 }
1129 
1130 void
1131 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1132 {
1133 	unsigned long rtcount;
1134 	struct icmp6_mtudisc_callback *mc;
1135 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1136 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1137 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1138 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1139 	struct rtentry *rt = NULL;
1140 	struct sockaddr_in6 sin6;
1141 	struct ifnet *rcvif;
1142 	int s;
1143 
1144 	/*
1145 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1146 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1147 	 * In that one case, the MTU might be less than 1280.
1148 	 */
1149 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1150 		/* is the mtu even sane? */
1151 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1152 			return;
1153 		if (!validated)
1154 			return;
1155 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1156 	}
1157 
1158 	/*
1159 	 * allow non-validated cases if memory is plenty, to make traffic
1160 	 * from non-connected pcb happy.
1161 	 */
1162 	mutex_enter(&icmp6_mtx);
1163 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1164 	if (validated) {
1165 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) {
1166 			mutex_exit(&icmp6_mtx);
1167 			return;
1168 		}
1169 		else if (0 <= icmp6_mtudisc_lowat &&
1170 		    rtcount > icmp6_mtudisc_lowat) {
1171 			/*
1172 			 * XXX nuke a victim, install the new one.
1173 			 */
1174 		}
1175 	} else {
1176 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) {
1177 			mutex_exit(&icmp6_mtx);
1178 			return;
1179 		}
1180 	}
1181 	mutex_exit(&icmp6_mtx);
1182 
1183 	memset(&sin6, 0, sizeof(sin6));
1184 	sin6.sin6_family = PF_INET6;
1185 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1186 	sin6.sin6_addr = *dst;
1187 	rcvif = m_get_rcvif(m, &s);
1188 	if (__predict_false(rcvif == NULL))
1189 		return;
1190 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1191 		m_put_rcvif(rcvif, &s);
1192 		return;
1193 	}
1194 	m_put_rcvif(rcvif, &s);
1195 
1196 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1197 
1198 	if (rt && (rt->rt_flags & RTF_HOST) &&
1199 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1200 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1201 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1202 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1203 			rt->rt_rmx.rmx_mtu = mtu;
1204 		}
1205 	}
1206 	if (rt) {
1207 		rt_unref(rt);
1208 	}
1209 
1210 	/*
1211 	 * Notify protocols that the MTU for this destination
1212 	 * has changed.
1213 	 */
1214 	mutex_enter(&icmp6_mtx);
1215 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1216 	     mc = LIST_NEXT(mc, mc_list))
1217 		(*mc->mc_func)(&sin6.sin6_addr);
1218 	mutex_exit(&icmp6_mtx);
1219 }
1220 
1221 /*
1222  * Process a Node Information Query packet, based on
1223  * draft-ietf-ipngwg-icmp-name-lookups-07.
1224  *
1225  * Spec incompatibilities:
1226  * - IPv6 Subject address handling
1227  * - IPv4 Subject address handling support missing
1228  * - Proxy reply (answer even if it's not for me)
1229  * - joins NI group address at in6_ifattach() time only, does not cope
1230  *   with hostname changes by sethostname(3)
1231  */
1232 static struct mbuf *
1233 ni6_input(struct mbuf *m, int off)
1234 {
1235 	struct icmp6_nodeinfo *ni6, *nni6;
1236 	struct mbuf *n = NULL;
1237 	u_int16_t qtype;
1238 	int subjlen;
1239 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1240 	struct ni_reply_fqdn *fqdn;
1241 	int addrs;		/* for NI_QTYPE_NODEADDR */
1242 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1243 	struct sockaddr_in6 sin6; /* ip6_dst */
1244 	struct in6_addr in6_subj; /* subject address */
1245 	struct ip6_hdr *ip6;
1246 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1247 	char *subj = NULL;
1248 	struct ifnet *rcvif;
1249 	int s, ss;
1250 	struct ifaddr *ifa;
1251 	struct psref psref;
1252 
1253 	ip6 = mtod(m, struct ip6_hdr *);
1254 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1255 	if (ni6 == NULL) {
1256 		/* m is already reclaimed */
1257 		return NULL;
1258 	}
1259 
1260 	/*
1261 	 * Validate IPv6 destination address.
1262 	 *
1263 	 * The Responder must discard the Query without further processing
1264 	 * unless it is one of the Responder's unicast or anycast addresses, or
1265 	 * a link-local scope multicast address which the Responder has joined.
1266 	 * [icmp-name-lookups-07, Section 4.]
1267 	 */
1268 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1269 	/* XXX scopeid */
1270 	ss = pserialize_read_enter();
1271 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1272 	if (ifa != NULL)
1273 		; /* unicast/anycast, fine */
1274 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1275 		; /* link-local multicast, fine */
1276 	else {
1277 		pserialize_read_exit(ss);
1278 		goto bad;
1279 	}
1280 	pserialize_read_exit(ss);
1281 
1282 	/* validate query Subject field. */
1283 	qtype = ntohs(ni6->ni_qtype);
1284 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1285 	switch (qtype) {
1286 	case NI_QTYPE_NOOP:
1287 	case NI_QTYPE_SUPTYPES:
1288 		/* 07 draft */
1289 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1290 			break;
1291 		/* FALLTHROUGH */
1292 	case NI_QTYPE_FQDN:
1293 	case NI_QTYPE_NODEADDR:
1294 	case NI_QTYPE_IPV4ADDR:
1295 		switch (ni6->ni_code) {
1296 		case ICMP6_NI_SUBJ_IPV6:
1297 #if ICMP6_NI_SUBJ_IPV6 != 0
1298 		case 0:
1299 #endif
1300 			/*
1301 			 * backward compatibility - try to accept 03 draft
1302 			 * format, where no Subject is present.
1303 			 */
1304 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1305 			    subjlen == 0) {
1306 				oldfqdn++;
1307 				break;
1308 			}
1309 #if ICMP6_NI_SUBJ_IPV6 != 0
1310 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1311 				goto bad;
1312 #endif
1313 
1314 			if (subjlen != sizeof(sin6.sin6_addr))
1315 				goto bad;
1316 
1317 			/*
1318 			 * Validate Subject address.
1319 			 *
1320 			 * Not sure what exactly "address belongs to the node"
1321 			 * means in the spec, is it just unicast, or what?
1322 			 *
1323 			 * At this moment we consider Subject address as
1324 			 * "belong to the node" if the Subject address equals
1325 			 * to the IPv6 destination address; validation for
1326 			 * IPv6 destination address should have done enough
1327 			 * check for us.
1328 			 *
1329 			 * We do not do proxy at this moment.
1330 			 */
1331 			/* m_pulldown instead of copy? */
1332 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1333 			    subjlen, (void *)&in6_subj);
1334 			rcvif = m_get_rcvif(m, &s);
1335 			if (__predict_false(rcvif == NULL))
1336 				goto bad;
1337 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1338 				m_put_rcvif(rcvif, &s);
1339 				goto bad;
1340 			}
1341 			m_put_rcvif(rcvif, &s);
1342 
1343 			subj = (char *)&in6_subj;
1344 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1345 				break;
1346 
1347 			/*
1348 			 * XXX if we are to allow other cases, we should really
1349 			 * be careful about scope here.
1350 			 * basically, we should disallow queries toward IPv6
1351 			 * destination X with subject Y, if scope(X) > scope(Y).
1352 			 * if we allow scope(X) > scope(Y), it will result in
1353 			 * information leakage across scope boundary.
1354 			 */
1355 			goto bad;
1356 
1357 		case ICMP6_NI_SUBJ_FQDN:
1358 			/*
1359 			 * Validate Subject name with gethostname(3).
1360 			 *
1361 			 * The behavior may need some debate, since:
1362 			 * - we are not sure if the node has FQDN as
1363 			 *   hostname (returned by gethostname(3)).
1364 			 * - the code does wildcard match for truncated names.
1365 			 *   however, we are not sure if we want to perform
1366 			 *   wildcard match, if gethostname(3) side has
1367 			 *   truncated hostname.
1368 			 */
1369 			n = ni6_nametodns(hostname, hostnamelen, 0);
1370 			if (!n || n->m_next || n->m_len == 0)
1371 				goto bad;
1372 			IP6_EXTHDR_GET(subj, char *, m,
1373 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1374 			if (subj == NULL)
1375 				goto bad;
1376 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1377 					n->m_len)) {
1378 				goto bad;
1379 			}
1380 			m_freem(n);
1381 			n = NULL;
1382 			break;
1383 
1384 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1385 		default:
1386 			goto bad;
1387 		}
1388 		break;
1389 	}
1390 
1391 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1392 	switch (qtype) {
1393 	case NI_QTYPE_FQDN:
1394 		if ((icmp6_nodeinfo & 1) == 0)
1395 			goto bad;
1396 		break;
1397 	case NI_QTYPE_NODEADDR:
1398 	case NI_QTYPE_IPV4ADDR:
1399 		if ((icmp6_nodeinfo & 2) == 0)
1400 			goto bad;
1401 		break;
1402 	}
1403 
1404 	/* guess reply length */
1405 	switch (qtype) {
1406 	case NI_QTYPE_NOOP:
1407 		break;		/* no reply data */
1408 	case NI_QTYPE_SUPTYPES:
1409 		replylen += sizeof(u_int32_t);
1410 		break;
1411 	case NI_QTYPE_FQDN:
1412 		/* XXX will append an mbuf */
1413 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1414 		break;
1415 	case NI_QTYPE_NODEADDR:
1416 		addrs = ni6_addrs(ni6, m, &ifp, subj, &psref);
1417 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1418 					  sizeof(u_int32_t))) > MCLBYTES)
1419 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1420 		break;
1421 	case NI_QTYPE_IPV4ADDR:
1422 		/* unsupported - should respond with unknown Qtype? */
1423 		goto bad;
1424 	default:
1425 		/*
1426 		 * XXX: We must return a reply with the ICMP6 code
1427 		 * `unknown Qtype' in this case.  However we regard the case
1428 		 * as an FQDN query for backward compatibility.
1429 		 * Older versions set a random value to this field,
1430 		 * so it rarely varies in the defined qtypes.
1431 		 * But the mechanism is not reliable...
1432 		 * maybe we should obsolete older versions.
1433 		 */
1434 		qtype = NI_QTYPE_FQDN;
1435 		/* XXX will append an mbuf */
1436 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1437 		oldfqdn++;
1438 		break;
1439 	}
1440 
1441 	/* allocate an mbuf to reply. */
1442 	MGETHDR(n, M_DONTWAIT, m->m_type);
1443 	if (n == NULL) {
1444 		if_put(ifp, &psref);
1445 		m_freem(m);
1446 		return (NULL);
1447 	}
1448 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1449 	if (replylen > MHLEN) {
1450 		if (replylen > MCLBYTES) {
1451 			/*
1452 			 * XXX: should we try to allocate more? But MCLBYTES
1453 			 * is probably much larger than IPV6_MMTU...
1454 			 */
1455 			goto bad;
1456 		}
1457 		MCLGET(n, M_DONTWAIT);
1458 		if ((n->m_flags & M_EXT) == 0) {
1459 			goto bad;
1460 		}
1461 	}
1462 	n->m_pkthdr.len = n->m_len = replylen;
1463 
1464 	/* copy mbuf header and IPv6 + Node Information base headers */
1465 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1466 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1467 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1468 
1469 	/* qtype dependent procedure */
1470 	switch (qtype) {
1471 	case NI_QTYPE_NOOP:
1472 		nni6->ni_code = ICMP6_NI_SUCCESS;
1473 		nni6->ni_flags = 0;
1474 		break;
1475 	case NI_QTYPE_SUPTYPES:
1476 	{
1477 		u_int32_t v;
1478 		nni6->ni_code = ICMP6_NI_SUCCESS;
1479 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1480 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1481 		v = (u_int32_t)htonl(0x0000000f);
1482 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1483 		break;
1484 	}
1485 	case NI_QTYPE_FQDN:
1486 		nni6->ni_code = ICMP6_NI_SUCCESS;
1487 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1488 						sizeof(struct ip6_hdr) +
1489 						sizeof(struct icmp6_nodeinfo));
1490 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1491 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1492 		/*
1493 		 * XXX do we really have FQDN in variable "hostname"?
1494 		 */
1495 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1496 		if (n->m_next == NULL)
1497 			goto bad;
1498 		/* XXX we assume that n->m_next is not a chain */
1499 		if (n->m_next->m_next != NULL)
1500 			goto bad;
1501 		n->m_pkthdr.len += n->m_next->m_len;
1502 		break;
1503 	case NI_QTYPE_NODEADDR:
1504 	{
1505 		int lenlim, copied;
1506 
1507 		nni6->ni_code = ICMP6_NI_SUCCESS;
1508 		n->m_pkthdr.len = n->m_len =
1509 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1510 		lenlim = M_TRAILINGSPACE(n);
1511 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1512 		if_put(ifp, &psref);
1513 		ifp = NULL;
1514 		/* XXX: reset mbuf length */
1515 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1516 			sizeof(struct icmp6_nodeinfo) + copied;
1517 		break;
1518 	}
1519 	default:
1520 		break;		/* XXX impossible! */
1521 	}
1522 
1523 	nni6->ni_type = ICMP6_NI_REPLY;
1524 	m_freem(m);
1525 	return (n);
1526 
1527   bad:
1528 	if_put(ifp, &psref);
1529 	m_freem(m);
1530 	if (n)
1531 		m_freem(n);
1532 	return (NULL);
1533 }
1534 #undef hostnamelen
1535 
1536 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1537 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1538 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1539 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1540 
1541 /*
1542  * make a mbuf with DNS-encoded string.  no compression support.
1543  *
1544  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1545  * treated as truncated name (two \0 at the end).  this is a wild guess.
1546  *
1547  * old - return pascal string if non-zero
1548  */
1549 static struct mbuf *
1550 ni6_nametodns(const char *name, int namelen, int old)
1551 {
1552 	struct mbuf *m;
1553 	char *cp, *ep;
1554 	const char *p, *q;
1555 	int i, len, nterm;
1556 
1557 	if (old)
1558 		len = namelen + 1;
1559 	else
1560 		len = MCLBYTES;
1561 
1562 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1563 	MGET(m, M_DONTWAIT, MT_DATA);
1564 	if (m && len > MLEN) {
1565 		MCLGET(m, M_DONTWAIT);
1566 		if ((m->m_flags & M_EXT) == 0)
1567 			goto fail;
1568 	}
1569 	if (!m)
1570 		goto fail;
1571 	m->m_next = NULL;
1572 
1573 	if (old) {
1574 		m->m_len = len;
1575 		*mtod(m, char *) = namelen;
1576 		bcopy(name, mtod(m, char *) + 1, namelen);
1577 		return m;
1578 	} else {
1579 		m->m_len = 0;
1580 		cp = mtod(m, char *);
1581 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1582 
1583 		/* if not certain about my name, return empty buffer */
1584 		if (namelen == 0)
1585 			return m;
1586 
1587 		/*
1588 		 * guess if it looks like shortened hostname, or FQDN.
1589 		 * shortened hostname needs two trailing "\0".
1590 		 */
1591 		i = 0;
1592 		for (p = name; p < name + namelen; p++) {
1593 			if (*p == '.')
1594 				i++;
1595 		}
1596 		if (i < 2)
1597 			nterm = 2;
1598 		else
1599 			nterm = 1;
1600 
1601 		p = name;
1602 		while (cp < ep && p < name + namelen) {
1603 			i = 0;
1604 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1605 				i++;
1606 			/* result does not fit into mbuf */
1607 			if (cp + i + 1 >= ep)
1608 				goto fail;
1609 			/*
1610 			 * DNS label length restriction, RFC1035 page 8.
1611 			 * "i == 0" case is included here to avoid returning
1612 			 * 0-length label on "foo..bar".
1613 			 */
1614 			if (i <= 0 || i >= 64)
1615 				goto fail;
1616 			*cp++ = i;
1617 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1618 				goto fail;
1619 			while (i > 0) {
1620 				if (!isalnum(*p) && *p != '-')
1621 					goto fail;
1622 				if (isupper(*p)) {
1623 					*cp++ = tolower(*p);
1624 					p++;
1625 				} else
1626 					*cp++ = *p++;
1627 				i--;
1628 			}
1629 			p = q;
1630 			if (p < name + namelen && *p == '.')
1631 				p++;
1632 		}
1633 		/* termination */
1634 		if (cp + nterm >= ep)
1635 			goto fail;
1636 		while (nterm-- > 0)
1637 			*cp++ = '\0';
1638 		m->m_len = cp - mtod(m, char *);
1639 		return m;
1640 	}
1641 
1642 	panic("should not reach here");
1643 	/* NOTREACHED */
1644 
1645  fail:
1646 	if (m)
1647 		m_freem(m);
1648 	return NULL;
1649 }
1650 
1651 /*
1652  * check if two DNS-encoded string matches.  takes care of truncated
1653  * form (with \0\0 at the end).  no compression support.
1654  * XXX upper/lowercase match (see RFC2065)
1655  */
1656 static int
1657 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1658 {
1659 	const char *a0, *b0;
1660 	int l;
1661 
1662 	/* simplest case - need validation? */
1663 	if (alen == blen && memcmp(a, b, alen) == 0)
1664 		return 1;
1665 
1666 	a0 = a;
1667 	b0 = b;
1668 
1669 	/* termination is mandatory */
1670 	if (alen < 2 || blen < 2)
1671 		return 0;
1672 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1673 		return 0;
1674 	alen--;
1675 	blen--;
1676 
1677 	while (a - a0 < alen && b - b0 < blen) {
1678 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1679 			return 0;
1680 
1681 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1682 			return 0;
1683 		/* we don't support compression yet */
1684 		if (a[0] >= 64 || b[0] >= 64)
1685 			return 0;
1686 
1687 		/* truncated case */
1688 		if (a[0] == 0 && a - a0 == alen - 1)
1689 			return 1;
1690 		if (b[0] == 0 && b - b0 == blen - 1)
1691 			return 1;
1692 		if (a[0] == 0 || b[0] == 0)
1693 			return 0;
1694 
1695 		if (a[0] != b[0])
1696 			return 0;
1697 		l = a[0];
1698 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1699 			return 0;
1700 		if (memcmp(a + 1, b + 1, l) != 0)
1701 			return 0;
1702 
1703 		a += 1 + l;
1704 		b += 1 + l;
1705 	}
1706 
1707 	if (a - a0 == alen && b - b0 == blen)
1708 		return 1;
1709 	else
1710 		return 0;
1711 }
1712 
1713 /*
1714  * calculate the number of addresses to be returned in the node info reply.
1715  */
1716 static int
1717 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1718     struct ifnet **ifpp, char *subj, struct psref *psref)
1719 {
1720 	struct ifnet *ifp;
1721 	struct in6_ifaddr *ia6;
1722 	struct ifaddr *ifa;
1723 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1724 	int addrs = 0, addrsofif, iffound = 0;
1725 	int niflags = ni6->ni_flags;
1726 	int s;
1727 
1728 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1729 		switch (ni6->ni_code) {
1730 		case ICMP6_NI_SUBJ_IPV6:
1731 			if (subj == NULL) /* must be impossible... */
1732 				return (0);
1733 			subj_ip6 = (struct sockaddr_in6 *)subj;
1734 			break;
1735 		default:
1736 			/*
1737 			 * XXX: we only support IPv6 subject address for
1738 			 * this Qtype.
1739 			 */
1740 			return (0);
1741 		}
1742 	}
1743 
1744 	s = pserialize_read_enter();
1745 	IFNET_READER_FOREACH(ifp) {
1746 		addrsofif = 0;
1747 		IFADDR_READER_FOREACH(ifa, ifp) {
1748 			if (ifa->ifa_addr->sa_family != AF_INET6)
1749 				continue;
1750 			ia6 = (struct in6_ifaddr *)ifa;
1751 
1752 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1753 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1754 					       &ia6->ia_addr.sin6_addr))
1755 				iffound = 1;
1756 
1757 			/*
1758 			 * IPv4-mapped addresses can only be returned by a
1759 			 * Node Information proxy, since they represent
1760 			 * addresses of IPv4-only nodes, which perforce do
1761 			 * not implement this protocol.
1762 			 * [icmp-name-lookups-07, Section 5.4]
1763 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1764 			 * this function at this moment.
1765 			 */
1766 
1767 			/* What do we have to do about ::1? */
1768 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1769 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1770 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1771 					continue;
1772 				break;
1773 			case IPV6_ADDR_SCOPE_SITELOCAL:
1774 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1775 					continue;
1776 				break;
1777 			case IPV6_ADDR_SCOPE_GLOBAL:
1778 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1779 					continue;
1780 				break;
1781 			default:
1782 				continue;
1783 			}
1784 
1785 			/*
1786 			 * check if anycast is okay.
1787 			 * XXX: just experimental.  not in the spec.
1788 			 */
1789 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1790 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1791 				continue; /* we need only unicast addresses */
1792 
1793 			addrsofif++; /* count the address */
1794 		}
1795 		if (iffound) {
1796 			if_acquire_NOMPSAFE(ifp, psref);
1797 			pserialize_read_exit(s);
1798 			*ifpp = ifp;
1799 			return (addrsofif);
1800 		}
1801 
1802 		addrs += addrsofif;
1803 	}
1804 	pserialize_read_exit(s);
1805 
1806 	return (addrs);
1807 }
1808 
1809 static int
1810 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1811 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1812 	int resid)
1813 {
1814 	struct ifnet *ifp;
1815 	struct in6_ifaddr *ia6;
1816 	struct ifaddr *ifa;
1817 	struct ifnet *ifp_dep = NULL;
1818 	int copied = 0, allow_deprecated = 0;
1819 	u_char *cp = (u_char *)(nni6 + 1);
1820 	int niflags = ni6->ni_flags;
1821 	u_int32_t ltime;
1822 	int s;
1823 
1824 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1825 		return (0);	/* needless to copy */
1826 
1827 	s = pserialize_read_enter();
1828 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1829   again:
1830 
1831 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1832 	{
1833 		IFADDR_READER_FOREACH(ifa, ifp) {
1834 			if (ifa->ifa_addr->sa_family != AF_INET6)
1835 				continue;
1836 			ia6 = (struct in6_ifaddr *)ifa;
1837 
1838 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1839 			    allow_deprecated == 0) {
1840 				/*
1841 				 * prefererred address should be put before
1842 				 * deprecated addresses.
1843 				 */
1844 
1845 				/* record the interface for later search */
1846 				if (ifp_dep == NULL)
1847 					ifp_dep = ifp;
1848 
1849 				continue;
1850 			}
1851 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1852 				 allow_deprecated != 0)
1853 				continue; /* we now collect deprecated addrs */
1854 
1855 			/* What do we have to do about ::1? */
1856 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1857 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1858 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1859 					continue;
1860 				break;
1861 			case IPV6_ADDR_SCOPE_SITELOCAL:
1862 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1863 					continue;
1864 				break;
1865 			case IPV6_ADDR_SCOPE_GLOBAL:
1866 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1867 					continue;
1868 				break;
1869 			default:
1870 				continue;
1871 			}
1872 
1873 			/*
1874 			 * check if anycast is okay.
1875 			 * XXX: just experimental.  not in the spec.
1876 			 */
1877 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1878 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1879 				continue;
1880 
1881 			/* now we can copy the address */
1882 			if (resid < sizeof(struct in6_addr) +
1883 			    sizeof(u_int32_t)) {
1884 				/*
1885 				 * We give up much more copy.
1886 				 * Set the truncate flag and return.
1887 				 */
1888 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1889 				goto out;
1890 			}
1891 
1892 			/*
1893 			 * Set the TTL of the address.
1894 			 * The TTL value should be one of the following
1895 			 * according to the specification:
1896 			 *
1897 			 * 1. The remaining lifetime of a DHCP lease on the
1898 			 *    address, or
1899 			 * 2. The remaining Valid Lifetime of a prefix from
1900 			 *    which the address was derived through Stateless
1901 			 *    Autoconfiguration.
1902 			 *
1903 			 * Note that we currently do not support stateful
1904 			 * address configuration by DHCPv6, so the former
1905 			 * case can't happen.
1906 			 *
1907 			 * TTL must be 2^31 > TTL >= 0.
1908 			 */
1909 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1910 				ltime = ND6_INFINITE_LIFETIME;
1911 			else {
1912 				if (ia6->ia6_lifetime.ia6t_expire >
1913 				    time_uptime)
1914 					ltime = ia6->ia6_lifetime.ia6t_expire -
1915 					    time_uptime;
1916 				else
1917 					ltime = 0;
1918 			}
1919 			if (ltime > 0x7fffffff)
1920 				ltime = 0x7fffffff;
1921 			ltime = htonl(ltime);
1922 
1923 			bcopy(&ltime, cp, sizeof(u_int32_t));
1924 			cp += sizeof(u_int32_t);
1925 
1926 			/* copy the address itself */
1927 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1928 			      sizeof(struct in6_addr));
1929 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1930 			cp += sizeof(struct in6_addr);
1931 
1932 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1933 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1934 		}
1935 		if (ifp0)	/* we need search only on the specified IF */
1936 			break;
1937 	}
1938 
1939 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1940 		ifp = ifp_dep;
1941 		allow_deprecated = 1;
1942 
1943 		goto again;
1944 	}
1945 out:
1946 	pserialize_read_exit(s);
1947 	return (copied);
1948 }
1949 
1950 /*
1951  * XXX almost dup'ed code with rip6_input.
1952  */
1953 static int
1954 icmp6_rip6_input(struct mbuf **mp, int off)
1955 {
1956 	struct mbuf *m = *mp;
1957 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1958 	struct inpcb_hdr *inph;
1959 	struct in6pcb *in6p;
1960 	struct in6pcb *last = NULL;
1961 	struct sockaddr_in6 rip6src;
1962 	struct icmp6_hdr *icmp6;
1963 	struct mbuf *opts = NULL;
1964 
1965 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1966 	if (icmp6 == NULL) {
1967 		/* m is already reclaimed */
1968 		return IPPROTO_DONE;
1969 	}
1970 
1971 	/*
1972 	 * XXX: the address may have embedded scope zone ID, which should be
1973 	 * hidden from applications.
1974 	 */
1975 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1976 	if (sa6_recoverscope(&rip6src)) {
1977 		m_freem(m);
1978 		return (IPPROTO_DONE);
1979 	}
1980 
1981 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1982 		in6p = (struct in6pcb *)inph;
1983 		if (in6p->in6p_af != AF_INET6)
1984 			continue;
1985 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1986 			continue;
1987 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1988 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1989 			continue;
1990 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1991 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1992 			continue;
1993 		if (in6p->in6p_icmp6filt
1994 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1995 				 in6p->in6p_icmp6filt))
1996 			continue;
1997 		if (last) {
1998 			struct	mbuf *n;
1999 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2000 				if (last->in6p_flags & IN6P_CONTROLOPTS)
2001 					ip6_savecontrol(last, &opts, ip6, n);
2002 				/* strip intermediate headers */
2003 				m_adj(n, off);
2004 				if (sbappendaddr(&last->in6p_socket->so_rcv,
2005 				    sin6tosa(&rip6src), n, opts) == 0) {
2006 					/* should notify about lost packet */
2007 					m_freem(n);
2008 					if (opts)
2009 						m_freem(opts);
2010 				} else
2011 					sorwakeup(last->in6p_socket);
2012 				opts = NULL;
2013 			}
2014 		}
2015 		last = in6p;
2016 	}
2017 	if (last) {
2018 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2019 			ip6_savecontrol(last, &opts, ip6, m);
2020 		/* strip intermediate headers */
2021 		m_adj(m, off);
2022 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2023 		    sin6tosa(&rip6src), m, opts) == 0) {
2024 			m_freem(m);
2025 			if (opts)
2026 				m_freem(opts);
2027 		} else
2028 			sorwakeup(last->in6p_socket);
2029 	} else {
2030 		m_freem(m);
2031 		IP6_STATDEC(IP6_STAT_DELIVERED);
2032 	}
2033 	return IPPROTO_DONE;
2034 }
2035 
2036 /*
2037  * Reflect the ip6 packet back to the source.
2038  * OFF points to the icmp6 header, counted from the top of the mbuf.
2039  *
2040  * Note: RFC 1885 required that an echo reply should be truncated if it
2041  * did not fit in with (return) path MTU, and KAME code supported the
2042  * behavior.  However, as a clarification after the RFC, this limitation
2043  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2044  * old behavior, with a (non-default) ifdef block, while the new version of
2045  * the spec was an internet-draft status, and even after the new RFC was
2046  * published.  But it would rather make sense to clean the obsoleted part
2047  * up, and to make the code simpler at this stage.
2048  */
2049 void
2050 icmp6_reflect(struct mbuf *m, size_t off)
2051 {
2052 	struct ip6_hdr *ip6;
2053 	struct icmp6_hdr *icmp6;
2054 	const struct in6_ifaddr *ia;
2055 	const struct ip6aux *ip6a;
2056 	int plen;
2057 	int type, code;
2058 	struct ifnet *outif = NULL;
2059 	struct in6_addr origdst;
2060 	struct ifnet *rcvif;
2061 	int s;
2062 	bool ip6_src_filled = false;
2063 
2064 	/* too short to reflect */
2065 	if (off < sizeof(struct ip6_hdr)) {
2066 		nd6log(LOG_DEBUG,
2067 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2068 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2069 		    __FILE__, __LINE__);
2070 		goto bad;
2071 	}
2072 
2073 	/*
2074 	 * If there are extra headers between IPv6 and ICMPv6, strip
2075 	 * off that header first.
2076 	 */
2077 #ifdef DIAGNOSTIC
2078 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2079 		panic("assumption failed in icmp6_reflect");
2080 #endif
2081 	if (off > sizeof(struct ip6_hdr)) {
2082 		size_t l;
2083 		struct ip6_hdr nip6;
2084 
2085 		l = off - sizeof(struct ip6_hdr);
2086 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2087 		m_adj(m, l);
2088 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2089 		if (m->m_len < l) {
2090 			if ((m = m_pullup(m, l)) == NULL)
2091 				return;
2092 		}
2093 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
2094 	} else /* off == sizeof(struct ip6_hdr) */ {
2095 		size_t l;
2096 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2097 		if (m->m_len < l) {
2098 			if ((m = m_pullup(m, l)) == NULL)
2099 				return;
2100 		}
2101 	}
2102 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2103 	ip6 = mtod(m, struct ip6_hdr *);
2104 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2105 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2106 	type = icmp6->icmp6_type; /* keep type for statistics */
2107 	code = icmp6->icmp6_code; /* ditto. */
2108 
2109 	origdst = ip6->ip6_dst;
2110 	/*
2111 	 * ip6_input() drops a packet if its src is multicast.
2112 	 * So, the src is never multicast.
2113 	 */
2114 	ip6->ip6_dst = ip6->ip6_src;
2115 
2116 	/*
2117 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2118 	 * use dst as the src for the reply.
2119 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2120 	 * (for example) when we encounter an error while forwarding procedure
2121 	 * destined to a duplicated address of ours.
2122 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2123 	 * procedure of an outgoing packet of our own, in which case we need
2124 	 * to search in the ifaddr list.
2125 	 */
2126 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2127 		;
2128 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2129 		if ((ip6a->ip6a_flags &
2130 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2131 			ip6->ip6_src = ip6a->ip6a_src;
2132 			ip6_src_filled = true;
2133 		}
2134 	} else {
2135 		union {
2136 			struct sockaddr_in6 sin6;
2137 			struct sockaddr sa;
2138 		} u;
2139 		int _s;
2140 		struct ifaddr *ifa;
2141 
2142 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2143 
2144 		_s = pserialize_read_enter();
2145 		ifa = ifa_ifwithaddr(&u.sa);
2146 
2147 		if (ifa != NULL) {
2148 			ia = ifatoia6(ifa);
2149 			if ((ia->ia6_flags &
2150 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2151 				ip6->ip6_src = ia->ia_addr.sin6_addr;
2152 				ip6_src_filled = true;
2153 			}
2154 		}
2155 		pserialize_read_exit(_s);
2156 	}
2157 
2158 	if (!ip6_src_filled) {
2159 		int e;
2160 		struct sockaddr_in6 sin6;
2161 		struct route ro;
2162 
2163 		/*
2164 		 * This case matches to multicasts, our anycast, or unicasts
2165 		 * that we do not own.  Select a source address based on the
2166 		 * source address of the erroneous packet.
2167 		 */
2168 		/* zone ID should be embedded */
2169 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2170 
2171 		memset(&ro, 0, sizeof(ro));
2172 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2173 		    &ip6->ip6_src);
2174 		rtcache_free(&ro);
2175 		if (e != 0) {
2176 			char ip6buf[INET6_ADDRSTRLEN];
2177 			nd6log(LOG_DEBUG,
2178 			    "source can't be determined: "
2179 			    "dst=%s, error=%d\n",
2180 			    IN6_PRINT(ip6buf, &sin6.sin6_addr), e);
2181 			goto bad;
2182 		}
2183 	}
2184 
2185 	ip6->ip6_flow = 0;
2186 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2187 	ip6->ip6_vfc |= IPV6_VERSION;
2188 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2189 	rcvif = m_get_rcvif(m, &s);
2190 	if (rcvif) {
2191 		/* XXX: This may not be the outgoing interface */
2192 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2193 	} else
2194 		ip6->ip6_hlim = ip6_defhlim;
2195 	m_put_rcvif(rcvif, &s);
2196 
2197 	m->m_pkthdr.csum_flags = 0;
2198 	icmp6->icmp6_cksum = 0;
2199 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2200 					sizeof(struct ip6_hdr), plen);
2201 
2202 	/*
2203 	 * XXX option handling
2204 	 */
2205 
2206 	m->m_flags &= ~(M_BCAST|M_MCAST);
2207 
2208 	/*
2209 	 * To avoid a "too big" situation at an intermediate router
2210 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2211 	 * Note that only echo and node information replies are affected,
2212 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2213 	 */
2214 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif)
2215 	    != 0 && outif)
2216 		icmp6_ifstat_inc(outif, ifs6_out_error);
2217 	if (outif)
2218 		icmp6_ifoutstat_inc(outif, type, code);
2219 
2220 	return;
2221 
2222  bad:
2223 	m_freem(m);
2224 	return;
2225 }
2226 
2227 static const char *
2228 icmp6_redirect_diag(char *buf, size_t buflen, struct in6_addr *src6, struct in6_addr *dst6,
2229 	struct in6_addr *tgt6)
2230 {
2231 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
2232 	char ip6buft[INET6_ADDRSTRLEN];
2233 
2234 	snprintf(buf, buflen, "(src=%s dst=%s tgt=%s)",
2235 	    IN6_PRINT(ip6bufs, src6), IN6_PRINT(ip6bufd, dst6),
2236 	    IN6_PRINT(ip6buft, tgt6));
2237 	return buf;
2238 }
2239 
2240 void
2241 icmp6_redirect_input(struct mbuf *m, int off)
2242 {
2243 	struct ifnet *ifp;
2244 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2245 	struct nd_redirect *nd_rd;
2246 	int icmp6len = ntohs(ip6->ip6_plen);
2247 	char *lladdr = NULL;
2248 	int lladdrlen = 0;
2249 	struct rtentry *rt = NULL;
2250 	int is_router;
2251 	int is_onlink;
2252 	struct in6_addr src6 = ip6->ip6_src;
2253 	struct in6_addr redtgt6;
2254 	struct in6_addr reddst6;
2255 	union nd_opts ndopts;
2256 	struct psref psref;
2257 	char ip6buf[INET6_ADDRSTRLEN];
2258 	char diagbuf[256];
2259 
2260 	ifp = m_get_rcvif_psref(m, &psref);
2261 	if (ifp == NULL)
2262 		goto freeit;
2263 
2264 	/* XXX if we are router, we don't update route by icmp6 redirect */
2265 	if (ip6_forwarding)
2266 		goto freeit;
2267 	if (!icmp6_rediraccept)
2268 		goto freeit;
2269 
2270 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2271 	if (nd_rd == NULL) {
2272 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2273 		m_put_rcvif_psref(ifp, &psref);
2274 		return;
2275 	}
2276 	redtgt6 = nd_rd->nd_rd_target;
2277 	reddst6 = nd_rd->nd_rd_dst;
2278 
2279 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2280 	    in6_setscope(&reddst6, ifp, NULL)) {
2281 		goto freeit;
2282 	}
2283 
2284 	/* validation */
2285 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2286 		nd6log(LOG_ERR,
2287 		    "ICMP6 redirect sent from %s rejected; "
2288 		    "must be from linklocal\n", IN6_PRINT(ip6buf, &src6));
2289 		goto bad;
2290 	}
2291 	if (ip6->ip6_hlim != 255) {
2292 		nd6log(LOG_ERR,
2293 		    "ICMP6 redirect sent from %s rejected; "
2294 		    "hlim=%d (must be 255)\n",
2295 		    IN6_PRINT(ip6buf, &src6), ip6->ip6_hlim);
2296 		goto bad;
2297 	}
2298     {
2299 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2300 	struct sockaddr_in6 sin6;
2301 	struct in6_addr *gw6;
2302 
2303 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2304 	rt = rtalloc1(sin6tosa(&sin6), 0);
2305 	if (rt) {
2306 		if (rt->rt_gateway == NULL ||
2307 		    rt->rt_gateway->sa_family != AF_INET6) {
2308 			nd6log(LOG_ERR,
2309 			    "ICMP6 redirect rejected; no route "
2310 			    "with inet6 gateway found for redirect dst: %s\n",
2311 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2312 			    &src6, &reddst6, &redtgt6));
2313 			rt_unref(rt);
2314 			goto bad;
2315 		}
2316 
2317 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2318 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2319 			nd6log(LOG_ERR,
2320 			    "ICMP6 redirect rejected; "
2321 			    "not equal to gw-for-src=%s (must be same): %s\n",
2322 			    IN6_PRINT(ip6buf, gw6),
2323 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2324 			    &src6, &reddst6, &redtgt6));
2325 			rt_unref(rt);
2326 			goto bad;
2327 		}
2328 	} else {
2329 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2330 		    "no route found for redirect dst: %s\n",
2331 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2332 		    &src6, &reddst6, &redtgt6));
2333 		goto bad;
2334 	}
2335 	rt_unref(rt);
2336 	rt = NULL;
2337     }
2338 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2339 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2340 		    "redirect dst must be unicast: %s\n",
2341 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2342 		    &src6, &reddst6, &redtgt6));
2343 		goto bad;
2344 	}
2345 
2346 	is_router = is_onlink = 0;
2347 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2348 		is_router = 1;	/* router case */
2349 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2350 		is_onlink = 1;	/* on-link destination case */
2351 	if (!is_router && !is_onlink) {
2352 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2353 		    "neither router case nor onlink case: %s\n",
2354 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2355 		    &src6, &reddst6, &redtgt6));
2356 		goto bad;
2357 	}
2358 	/* validation passed */
2359 
2360 	icmp6len -= sizeof(*nd_rd);
2361 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2362 	if (nd6_options(&ndopts) < 0) {
2363 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2364 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2365 		    &src6, &reddst6, &redtgt6));
2366 		/* nd6_options have incremented stats */
2367 		goto freeit;
2368 	}
2369 
2370 	if (ndopts.nd_opts_tgt_lladdr) {
2371 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2372 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2373 	}
2374 
2375 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2376 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2377 		    "(if %d, icmp6 packet %d): %s\n",
2378 		    IN6_PRINT(ip6buf, &redtgt6),
2379 		    ifp->if_addrlen, lladdrlen - 2,
2380 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2381 		    &src6, &reddst6, &redtgt6));
2382 		goto bad;
2383 	}
2384 
2385 	/* RFC 2461 8.3 */
2386 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2387 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2388 
2389 	m_put_rcvif_psref(ifp, &psref);
2390 	ifp = NULL;
2391 
2392 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2393 		/* perform rtredirect */
2394 		struct sockaddr_in6 sdst;
2395 		struct sockaddr_in6 sgw;
2396 		struct sockaddr_in6 ssrc;
2397 		unsigned long rtcount;
2398 		struct rtentry *newrt = NULL;
2399 
2400 		/*
2401 		 * do not install redirect route, if the number of entries
2402 		 * is too much (> hiwat).  note that, the node (= host) will
2403 		 * work just fine even if we do not install redirect route
2404 		 * (there will be additional hops, though).
2405 		 */
2406 		mutex_enter(&icmp6_mtx);
2407 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2408 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) {
2409 			mutex_exit(&icmp6_mtx);
2410 			goto freeit;
2411 		}
2412 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) {
2413 			mutex_exit(&icmp6_mtx);
2414 			goto freeit;
2415 		} else if (0 <= icmp6_redirect_lowat &&
2416 		    rtcount > icmp6_redirect_lowat) {
2417 			/*
2418 			 * XXX nuke a victim, install the new one.
2419 			 */
2420 		}
2421 
2422 		memset(&sdst, 0, sizeof(sdst));
2423 		memset(&sgw, 0, sizeof(sgw));
2424 		memset(&ssrc, 0, sizeof(ssrc));
2425 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2426 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2427 			sizeof(struct sockaddr_in6);
2428 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2429 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2430 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2431 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2432 		           RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc),
2433 			   &newrt);
2434 
2435 		if (newrt) {
2436 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2437 			    icmp6_redirect_timeout_q);
2438 			rt_unref(newrt);
2439 		}
2440 		mutex_exit(&icmp6_mtx);
2441 	}
2442 	/* finally update cached route in each socket via pfctlinput */
2443 	{
2444 		struct sockaddr_in6 sdst;
2445 
2446 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2447 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2448 #if defined(IPSEC)
2449 		if (ipsec_used)
2450 			key_sa_routechange(sin6tosa(&sdst));
2451 #endif
2452 	}
2453 
2454  freeit:
2455 	if (ifp != NULL)
2456 		m_put_rcvif_psref(ifp, &psref);
2457 	m_freem(m);
2458 	return;
2459 
2460  bad:
2461 	m_put_rcvif_psref(ifp, &psref);
2462 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2463 	m_freem(m);
2464 }
2465 
2466 void
2467 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2468 {
2469 	struct ifnet *ifp;	/* my outgoing interface */
2470 	struct in6_addr *ifp_ll6;
2471 	struct in6_addr *nexthop;
2472 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2473 	struct mbuf *m = NULL;	/* newly allocated one */
2474 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2475 	struct nd_redirect *nd_rd;
2476 	size_t maxlen;
2477 	u_char *p;
2478 	struct sockaddr_in6 src_sa;
2479 
2480 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2481 
2482 	/* if we are not router, we don't send icmp6 redirect */
2483 	if (!ip6_forwarding)
2484 		goto fail;
2485 
2486 	/* sanity check */
2487 	KASSERT(m0 != NULL);
2488 	KASSERT(rt != NULL);
2489 
2490 	ifp = rt->rt_ifp;
2491 
2492 	/*
2493 	 * Address check:
2494 	 *  the source address must identify a neighbor, and
2495 	 *  the destination address must not be a multicast address
2496 	 *  [RFC 2461, sec 8.2]
2497 	 */
2498 	sip6 = mtod(m0, struct ip6_hdr *);
2499 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2500 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2501 		goto fail;
2502 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2503 		goto fail;	/* what should we do here? */
2504 
2505 	/* rate limit */
2506 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2507 		goto fail;
2508 
2509 	/*
2510 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2511 	 * we almost always ask for an mbuf cluster for simplicity.
2512 	 * (MHLEN < IPV6_MMTU is almost always true)
2513 	 */
2514 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2515 	if (m && IPV6_MMTU >= MHLEN) {
2516 #if IPV6_MMTU >= MCLBYTES
2517 		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2518 #else
2519 		MCLGET(m, M_DONTWAIT);
2520 #endif
2521 	}
2522 
2523 	if (!m)
2524 		goto fail;
2525 	m_reset_rcvif(m);
2526 	m->m_len = 0;
2527 	maxlen = M_TRAILINGSPACE(m);
2528 	maxlen = min(IPV6_MMTU, maxlen);
2529 	/* just for safety */
2530 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2531 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2532 		goto fail;
2533 	}
2534 
2535 	{
2536 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2537 		struct in6_ifaddr *ia;
2538 		int s = pserialize_read_enter();
2539 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2540 						 IN6_IFF_NOTREADY|
2541 						 IN6_IFF_ANYCAST)) == NULL) {
2542 			pserialize_read_exit(s);
2543 			goto fail;
2544 		}
2545 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2546 		pserialize_read_exit(s);
2547 	}
2548 
2549 	/* get ip6 linklocal address for the router. */
2550 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2551 		struct sockaddr_in6 *sin6;
2552 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2553 		nexthop = &sin6->sin6_addr;
2554 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2555 			nexthop = NULL;
2556 	} else
2557 		nexthop = NULL;
2558 
2559 	/* ip6 */
2560 	ip6 = mtod(m, struct ip6_hdr *);
2561 	ip6->ip6_flow = 0;
2562 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2563 	ip6->ip6_vfc |= IPV6_VERSION;
2564 	/* ip6->ip6_plen will be set later */
2565 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2566 	ip6->ip6_hlim = 255;
2567 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2568 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2569 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2570 
2571 	/* ND Redirect */
2572 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2573 	nd_rd->nd_rd_type = ND_REDIRECT;
2574 	nd_rd->nd_rd_code = 0;
2575 	nd_rd->nd_rd_reserved = 0;
2576 	if (rt->rt_flags & RTF_GATEWAY) {
2577 		/*
2578 		 * nd_rd->nd_rd_target must be a link-local address in
2579 		 * better router cases.
2580 		 */
2581 		if (!nexthop)
2582 			goto fail;
2583 		bcopy(nexthop, &nd_rd->nd_rd_target,
2584 		      sizeof(nd_rd->nd_rd_target));
2585 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2586 		      sizeof(nd_rd->nd_rd_dst));
2587 	} else {
2588 		/* make sure redtgt == reddst */
2589 		nexthop = &sip6->ip6_dst;
2590 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2591 		      sizeof(nd_rd->nd_rd_target));
2592 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2593 		      sizeof(nd_rd->nd_rd_dst));
2594 	}
2595 
2596 	p = (u_char *)(nd_rd + 1);
2597 
2598 	{
2599 		/* target lladdr option */
2600 		struct llentry *ln = NULL;
2601 		int len;
2602 		struct nd_opt_hdr *nd_opt;
2603 		char *lladdr;
2604 
2605 		ln = nd6_lookup(nexthop, ifp, false);
2606 		if (ln == NULL)
2607 			goto nolladdropt;
2608 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2609 		len = (len + 7) & ~7;	/* round by 8 */
2610 		/* safety check */
2611 		if (len + (p - (u_char *)ip6) > maxlen) {
2612 			LLE_RUNLOCK(ln);
2613 			goto nolladdropt;
2614 		}
2615 		if (ln->la_flags & LLE_VALID) {
2616 			nd_opt = (struct nd_opt_hdr *)p;
2617 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2618 			nd_opt->nd_opt_len = len >> 3;
2619 			lladdr = (char *)(nd_opt + 1);
2620 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2621 			p += len;
2622 		}
2623 		LLE_RUNLOCK(ln);
2624 	}
2625   nolladdropt:;
2626 
2627 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2628 
2629 	/* just to be safe */
2630 	if (m0->m_flags & M_DECRYPTED)
2631 		goto noredhdropt;
2632 	if (p - (u_char *)ip6 > maxlen)
2633 		goto noredhdropt;
2634 
2635 	{
2636 		/* redirected header option */
2637 		int len;
2638 		struct nd_opt_rd_hdr *nd_opt_rh;
2639 
2640 		/*
2641 		 * compute the maximum size for icmp6 redirect header option.
2642 		 * XXX room for auth header?
2643 		 */
2644 		len = maxlen - (p - (u_char *)ip6);
2645 		len &= ~7;
2646 
2647 		/*
2648 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2649 		 * about padding/truncate rule for the original IP packet.
2650 		 * From the discussion on IPv6imp in Feb 1999,
2651 		 * the consensus was:
2652 		 * - "attach as much as possible" is the goal
2653 		 * - pad if not aligned (original size can be guessed by
2654 		 *   original ip6 header)
2655 		 * Following code adds the padding if it is simple enough,
2656 		 * and truncates if not.
2657 		 */
2658 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2659 			/* not enough room, truncate */
2660 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2661 			    m0->m_pkthdr.len);
2662 		} else {
2663 			/*
2664 			 * enough room, truncate if not aligned.
2665 			 * we don't pad here for simplicity.
2666 			 */
2667 			size_t extra;
2668 
2669 			extra = m0->m_pkthdr.len % 8;
2670 			if (extra) {
2671 				/* truncate */
2672 				m_adj(m0, -extra);
2673 			}
2674 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2675 		}
2676 
2677 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2678 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2679 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2680 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2681 		p += sizeof(*nd_opt_rh);
2682 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2683 
2684 		/* connect m0 to m */
2685 		m->m_pkthdr.len += m0->m_pkthdr.len;
2686 		m_cat(m, m0);
2687 		m0 = NULL;
2688 	}
2689 noredhdropt:
2690 	if (m0) {
2691 		m_freem(m0);
2692 		m0 = NULL;
2693 	}
2694 
2695 	/* XXX: clear embedded link IDs in the inner header */
2696 	in6_clearscope(&sip6->ip6_src);
2697 	in6_clearscope(&sip6->ip6_dst);
2698 	in6_clearscope(&nd_rd->nd_rd_target);
2699 	in6_clearscope(&nd_rd->nd_rd_dst);
2700 
2701 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2702 
2703 	nd_rd->nd_rd_cksum = 0;
2704 	nd_rd->nd_rd_cksum
2705 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2706 
2707 	/* send the packet to outside... */
2708 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2709 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2710 
2711 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2712 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2713 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2714 
2715 	return;
2716 
2717 fail:
2718 	if (m)
2719 		m_freem(m);
2720 	if (m0)
2721 		m_freem(m0);
2722 }
2723 
2724 /*
2725  * ICMPv6 socket option processing.
2726  */
2727 int
2728 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2729 {
2730 	int error = 0;
2731 	struct in6pcb *in6p = sotoin6pcb(so);
2732 
2733 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2734 		return rip6_ctloutput(op, so, sopt);
2735 
2736 	switch (op) {
2737 	case PRCO_SETOPT:
2738 		switch (sopt->sopt_name) {
2739 		case ICMP6_FILTER:
2740 		    {
2741 			struct icmp6_filter fil;
2742 
2743 			error = sockopt_get(sopt, &fil, sizeof(fil));
2744 			if (error)
2745 				break;
2746 			memcpy(in6p->in6p_icmp6filt, &fil,
2747 			    sizeof(struct icmp6_filter));
2748 			error = 0;
2749 			break;
2750 		    }
2751 
2752 		default:
2753 			error = ENOPROTOOPT;
2754 			break;
2755 		}
2756 		break;
2757 
2758 	case PRCO_GETOPT:
2759 		switch (sopt->sopt_name) {
2760 		case ICMP6_FILTER:
2761 		    {
2762 			if (in6p->in6p_icmp6filt == NULL) {
2763 				error = EINVAL;
2764 				break;
2765 			}
2766 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2767 			    sizeof(struct icmp6_filter));
2768 			break;
2769 		    }
2770 
2771 		default:
2772 			error = ENOPROTOOPT;
2773 			break;
2774 		}
2775 		break;
2776 	}
2777 
2778 	return (error);
2779 }
2780 
2781 /*
2782  * Perform rate limit check.
2783  * Returns 0 if it is okay to send the icmp6 packet.
2784  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2785  * limitation.
2786  *
2787  * XXX per-destination/type check necessary?
2788  */
2789 static int
2790 icmp6_ratelimit(
2791 	const struct in6_addr *dst,	/* not used at this moment */
2792 	const int type,		/* not used at this moment */
2793 	const int code)		/* not used at this moment */
2794 {
2795 	int ret;
2796 
2797 	ret = 0;	/* okay to send */
2798 
2799 	/* PPS limit */
2800 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2801 	    icmp6errppslim)) {
2802 		/* The packet is subject to rate limit */
2803 		ret++;
2804 	}
2805 
2806 	return ret;
2807 }
2808 
2809 static struct rtentry *
2810 icmp6_mtudisc_clone(struct sockaddr *dst)
2811 {
2812 	struct rtentry *rt;
2813 	int    error;
2814 
2815 	rt = rtalloc1(dst, 1);
2816 	if (rt == 0)
2817 		return NULL;
2818 
2819 	/* If we didn't get a host route, allocate one */
2820 	if ((rt->rt_flags & RTF_HOST) == 0) {
2821 		struct rtentry *nrt;
2822 
2823 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2824 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2825 		if (error) {
2826 			rt_unref(rt);
2827 			return NULL;
2828 		}
2829 		nrt->rt_rmx = rt->rt_rmx;
2830 		rt_unref(rt);
2831 		rt = nrt;
2832 	}
2833 
2834 	mutex_enter(&icmp6_mtx);
2835 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2836 			icmp6_mtudisc_timeout_q);
2837 	mutex_exit(&icmp6_mtx);
2838 
2839 	if (error) {
2840 		rt_unref(rt);
2841 		return NULL;
2842 	}
2843 
2844 	return rt;	/* caller need to call rtfree() */
2845 }
2846 
2847 static void
2848 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2849 {
2850 
2851 	KASSERT(rt != NULL);
2852 	rt_assert_referenced(rt);
2853 
2854 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2855 	    (RTF_DYNAMIC | RTF_HOST)) {
2856 		rtrequest(RTM_DELETE, rt_getkey(rt),
2857 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2858 	} else {
2859 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2860 			rt->rt_rmx.rmx_mtu = 0;
2861 	}
2862 }
2863 
2864 static void
2865 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2866 {
2867 
2868 	KASSERT(rt != NULL);
2869 	rt_assert_referenced(rt);
2870 
2871 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2872 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2873 		rtrequest(RTM_DELETE, rt_getkey(rt),
2874 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2875 	}
2876 }
2877 
2878 /*
2879  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2880  */
2881 static int
2882 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2883 {
2884 	(void)&name;
2885 	(void)&l;
2886 	(void)&oname;
2887 
2888 	if (namelen != 0)
2889 		return (EINVAL);
2890 
2891 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2892 	    /*XXXUNCONST*/
2893 	    __UNCONST(newp), newlen));
2894 }
2895 
2896 static int
2897 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2898 {
2899 
2900 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2901 }
2902 
2903 static int
2904 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2905 {
2906 	int error, tmp;
2907 	struct sysctlnode node;
2908 
2909 	mutex_enter(&icmp6_mtx);
2910 
2911 	node = *rnode;
2912 	node.sysctl_data = &tmp;
2913 	tmp = icmp6_redirtimeout;
2914 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2915 	if (error || newp == NULL)
2916 		goto out;
2917 	if (tmp < 0) {
2918 		error = EINVAL;
2919 		goto out;
2920 	}
2921 	icmp6_redirtimeout = tmp;
2922 
2923 	if (icmp6_redirect_timeout_q != NULL) {
2924 		if (icmp6_redirtimeout == 0) {
2925 			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2926 		} else {
2927 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2928 			    icmp6_redirtimeout);
2929 		}
2930 	} else if (icmp6_redirtimeout > 0) {
2931 		icmp6_redirect_timeout_q =
2932 		    rt_timer_queue_create(icmp6_redirtimeout);
2933 	}
2934 	error = 0;
2935 out:
2936 	mutex_exit(&icmp6_mtx);
2937 	return error;
2938 }
2939 
2940 static void
2941 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2942 {
2943 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2944 
2945 	sysctl_createv(clog, 0, NULL, NULL,
2946 		       CTLFLAG_PERMANENT,
2947 		       CTLTYPE_NODE, "inet6", NULL,
2948 		       NULL, 0, NULL, 0,
2949 		       CTL_NET, PF_INET6, CTL_EOL);
2950 	sysctl_createv(clog, 0, NULL, NULL,
2951 		       CTLFLAG_PERMANENT,
2952 		       CTLTYPE_NODE, "icmp6",
2953 		       SYSCTL_DESCR("ICMPv6 related settings"),
2954 		       NULL, 0, NULL, 0,
2955 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2956 
2957 	sysctl_createv(clog, 0, NULL, NULL,
2958 		       CTLFLAG_PERMANENT,
2959 		       CTLTYPE_STRUCT, "stats",
2960 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2961 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2962 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2963 		       ICMPV6CTL_STATS, CTL_EOL);
2964 	sysctl_createv(clog, 0, NULL, NULL,
2965 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2966 		       CTLTYPE_INT, "rediraccept",
2967 		       SYSCTL_DESCR("Accept and process redirect messages"),
2968 		       NULL, 0, &icmp6_rediraccept, 0,
2969 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2970 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2971 	sysctl_createv(clog, 0, NULL, NULL,
2972 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2973 		       CTLTYPE_INT, "redirtimeout",
2974 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2975 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2976 		       &icmp6_redirtimeout, 0,
2977 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2978 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2979 #if 0 /* obsoleted */
2980 	sysctl_createv(clog, 0, NULL, NULL,
2981 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2982 		       CTLTYPE_INT, "errratelimit", NULL,
2983 		       NULL, 0, &icmp6_errratelimit, 0,
2984 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2985 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2986 #endif
2987 	sysctl_createv(clog, 0, NULL, NULL,
2988 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2989 		       CTLTYPE_INT, "nd6_prune",
2990 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2991 		       NULL, 0, &nd6_prune, 0,
2992 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2993 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2994 	sysctl_createv(clog, 0, NULL, NULL,
2995 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2996 		       CTLTYPE_INT, "nd6_delay",
2997 		       SYSCTL_DESCR("First probe delay time"),
2998 		       NULL, 0, &nd6_delay, 0,
2999 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3000 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
3001 	sysctl_createv(clog, 0, NULL, NULL,
3002 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3003 		       CTLTYPE_INT, "nd6_umaxtries",
3004 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
3005 		       NULL, 0, &nd6_umaxtries, 0,
3006 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3007 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
3008 	sysctl_createv(clog, 0, NULL, NULL,
3009 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3010 		       CTLTYPE_INT, "nd6_mmaxtries",
3011 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
3012 		       NULL, 0, &nd6_mmaxtries, 0,
3013 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3014 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
3015 	sysctl_createv(clog, 0, NULL, NULL,
3016 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3017 		       CTLTYPE_INT, "nd6_useloopback",
3018 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
3019 		       NULL, 0, &nd6_useloopback, 0,
3020 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3021 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
3022 #if 0 /* obsoleted */
3023 	sysctl_createv(clog, 0, NULL, NULL,
3024 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3025 		       CTLTYPE_INT, "nd6_proxyall", NULL,
3026 		       NULL, 0, &nd6_proxyall, 0,
3027 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3028 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
3029 #endif
3030 	sysctl_createv(clog, 0, NULL, NULL,
3031 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3032 		       CTLTYPE_INT, "nodeinfo",
3033 		       SYSCTL_DESCR("Respond to node information requests"),
3034 		       NULL, 0, &icmp6_nodeinfo, 0,
3035 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3036 		       ICMPV6CTL_NODEINFO, CTL_EOL);
3037 	sysctl_createv(clog, 0, NULL, NULL,
3038 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3039 		       CTLTYPE_INT, "errppslimit",
3040 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
3041 		       NULL, 0, &icmp6errppslim, 0,
3042 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3043 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
3044 	sysctl_createv(clog, 0, NULL, NULL,
3045 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3046 		       CTLTYPE_INT, "nd6_maxnudhint",
3047 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
3048 		       NULL, 0, &nd6_maxnudhint, 0,
3049 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3050 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
3051 	sysctl_createv(clog, 0, NULL, NULL,
3052 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3053 		       CTLTYPE_INT, "mtudisc_hiwat",
3054 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3055 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
3056 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3057 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
3058 	sysctl_createv(clog, 0, NULL, NULL,
3059 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3060 		       CTLTYPE_INT, "mtudisc_lowat",
3061 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3062 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
3063 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3064 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
3065 	sysctl_createv(clog, 0, NULL, NULL,
3066 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3067 		       CTLTYPE_INT, "nd6_debug",
3068 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
3069 		       NULL, 0, &nd6_debug, 0,
3070 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3071 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
3072 	sysctl_createv(clog, 0, NULL, NULL,
3073 		       CTLFLAG_PERMANENT,
3074 		       CTLTYPE_STRUCT, "nd6_drlist",
3075 		       SYSCTL_DESCR("Default router list"),
3076 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3077 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3078 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
3079 	sysctl_createv(clog, 0, NULL, NULL,
3080 		       CTLFLAG_PERMANENT,
3081 		       CTLTYPE_STRUCT, "nd6_prlist",
3082 		       SYSCTL_DESCR("Prefix list"),
3083 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3084 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3085 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3086 	sysctl_createv(clog, 0, NULL, NULL,
3087 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3088 		       CTLTYPE_INT, "maxqueuelen",
3089 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3090 		       NULL, 1, &nd6_maxqueuelen, 0,
3091 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3092 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3093 }
3094 
3095 void
3096 icmp6_statinc(u_int stat)
3097 {
3098 
3099 	KASSERT(stat < ICMP6_NSTATS);
3100 	ICMP6_STATINC(stat);
3101 }
3102