xref: /netbsd-src/sys/netinet6/icmp6.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: icmp6.c,v 1.214 2017/11/05 07:03:37 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 1.214 2017/11/05 07:03:37 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/ipsec_var.h>
108 #include <netipsec/ipsec_private.h>
109 #include <netipsec/ipsec6.h>
110 #include <netipsec/key.h>
111 #endif
112 
113 
114 #include "faith.h"
115 #if defined(NFAITH) && 0 < NFAITH
116 #include <net/if_faith.h>
117 #endif
118 
119 #include <net/net_osdep.h>
120 
121 extern struct domain inet6domain;
122 
123 percpu_t *icmp6stat_percpu;
124 
125 extern struct inpcbtable raw6cbtable;
126 extern int icmp6errppslim;
127 static int icmp6errpps_count = 0;
128 static struct timeval icmp6errppslim_last;
129 extern int icmp6_nodeinfo;
130 
131 /*
132  * List of callbacks to notify when Path MTU changes are made.
133  */
134 struct icmp6_mtudisc_callback {
135 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
136 	void (*mc_func)(struct in6_addr *);
137 };
138 
139 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
140     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
141 
142 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
143 extern int pmtu_expire;
144 
145 /* XXX do these values make any sense? */
146 static int icmp6_mtudisc_hiwat = 1280;
147 static int icmp6_mtudisc_lowat = 256;
148 
149 /*
150  * keep track of # of redirect routes.
151  */
152 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
153 
154 /* XXX experimental, turned off */
155 static int icmp6_redirect_hiwat = -1;
156 static int icmp6_redirect_lowat = -1;
157 
158 /* Protect mtudisc and redirect stuffs */
159 static kmutex_t icmp6_mtx __cacheline_aligned;
160 
161 static void icmp6_errcount(u_int, int, int);
162 static int icmp6_rip6_input(struct mbuf **, int);
163 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
164 static const char *icmp6_redirect_diag(char *, size_t, struct in6_addr *,
165 	struct in6_addr *, struct in6_addr *);
166 static struct mbuf *ni6_input(struct mbuf *, int);
167 static struct mbuf *ni6_nametodns(const char *, int, int);
168 static int ni6_dnsmatch(const char *, int, const char *, int);
169 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
170 			  struct ifnet **, char *, struct psref *);
171 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
172 				struct ifnet *, int);
173 static int icmp6_notify_error(struct mbuf *, int, int, int);
174 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
175 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
176 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
177 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
178 
179 /* workqueue-based pr_input */
180 static struct wqinput *icmp6_wqinput;
181 static void _icmp6_input(struct mbuf *m, int off, int proto);
182 
183 void
184 icmp6_init(void)
185 {
186 
187 	sysctl_net_inet6_icmp6_setup(NULL);
188 	mld_init();
189 
190 	mutex_init(&icmp6_mtx, MUTEX_DEFAULT, IPL_NONE);
191 	mutex_enter(&icmp6_mtx);
192 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
193 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
194 	mutex_exit(&icmp6_mtx);
195 
196 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
197 
198 	icmp6_wqinput = wqinput_create("icmp6", _icmp6_input);
199 }
200 
201 static void
202 icmp6_errcount(u_int base, int type, int code)
203 {
204 	switch (type) {
205 	case ICMP6_DST_UNREACH:
206 		switch (code) {
207 		case ICMP6_DST_UNREACH_NOROUTE:
208 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
209 			return;
210 		case ICMP6_DST_UNREACH_ADMIN:
211 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
212 			return;
213 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
214 			ICMP6_STATINC(base +
215 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
216 			return;
217 		case ICMP6_DST_UNREACH_ADDR:
218 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
219 			return;
220 		case ICMP6_DST_UNREACH_NOPORT:
221 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
222 			return;
223 		}
224 		break;
225 	case ICMP6_PACKET_TOO_BIG:
226 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
227 		return;
228 	case ICMP6_TIME_EXCEEDED:
229 		switch (code) {
230 		case ICMP6_TIME_EXCEED_TRANSIT:
231 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
232 			return;
233 		case ICMP6_TIME_EXCEED_REASSEMBLY:
234 			ICMP6_STATINC(base +
235 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
236 			return;
237 		}
238 		break;
239 	case ICMP6_PARAM_PROB:
240 		switch (code) {
241 		case ICMP6_PARAMPROB_HEADER:
242 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
243 			return;
244 		case ICMP6_PARAMPROB_NEXTHEADER:
245 			ICMP6_STATINC(base +
246 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
247 			return;
248 		case ICMP6_PARAMPROB_OPTION:
249 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
250 			return;
251 		}
252 		break;
253 	case ND_REDIRECT:
254 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
255 		return;
256 	}
257 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
258 }
259 
260 /*
261  * Register a Path MTU Discovery callback.
262  */
263 void
264 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
265 {
266 	struct icmp6_mtudisc_callback *mc, *new;
267 
268 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
269 
270 	mutex_enter(&icmp6_mtx);
271 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
272 	     mc = LIST_NEXT(mc, mc_list)) {
273 		if (mc->mc_func == func) {
274 			mutex_exit(&icmp6_mtx);
275 			kmem_free(new, sizeof(*mc));
276 			return;
277 		}
278 	}
279 
280 	new->mc_func = func;
281 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, new, mc_list);
282 	mutex_exit(&icmp6_mtx);
283 }
284 
285 /*
286  * A wrapper function for icmp6_error() necessary when the erroneous packet
287  * may not contain enough scope zone information.
288  */
289 void
290 icmp6_error2(struct mbuf *m, int type, int code, int param,
291 	struct ifnet *ifp)
292 {
293 	struct ip6_hdr *ip6;
294 
295 	if (ifp == NULL)
296 		return;
297 
298 	if (m->m_len < sizeof(struct ip6_hdr)) {
299 		m = m_pullup(m, sizeof(struct ip6_hdr));
300 		if (m == NULL)
301 			return;
302 	}
303 
304 	ip6 = mtod(m, struct ip6_hdr *);
305 
306 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
307 		return;
308 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
309 		return;
310 
311 	icmp6_error(m, type, code, param);
312 }
313 
314 /*
315  * Generate an error packet of type error in response to bad IP6 packet.
316  */
317 void
318 icmp6_error(struct mbuf *m, int type, int code, int param)
319 {
320 	struct ip6_hdr *oip6, *nip6;
321 	struct icmp6_hdr *icmp6;
322 	u_int preplen;
323 	int off;
324 	int nxt;
325 
326 	ICMP6_STATINC(ICMP6_STAT_ERROR);
327 
328 	/* count per-type-code statistics */
329 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
330 
331 	if (m->m_flags & M_DECRYPTED) {
332 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
333 		goto freeit;
334 	}
335 
336 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
337 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
338 		return;
339 	oip6 = mtod(m, struct ip6_hdr *);
340 
341 	/*
342 	 * If the destination address of the erroneous packet is a multicast
343 	 * address, or the packet was sent using link-layer multicast,
344 	 * we should basically suppress sending an error (RFC 2463, Section
345 	 * 2.4).
346 	 * We have two exceptions (the item e.2 in that section):
347 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
348 	 * - the Parameter Problem Message that can be allowed an icmp6 error
349 	 *   in the option type field.  This check has been done in
350 	 *   ip6_unknown_opt(), so we can just check the type and code.
351 	 */
352 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
353 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
354 	    (type != ICMP6_PACKET_TOO_BIG &&
355 	     (type != ICMP6_PARAM_PROB ||
356 	      code != ICMP6_PARAMPROB_OPTION)))
357 		goto freeit;
358 
359 	/*
360 	 * RFC 2463, 2.4 (e.5): source address check.
361 	 * XXX: the case of anycast source?
362 	 */
363 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
364 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
365 		goto freeit;
366 
367 	/*
368 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
369 	 * don't do it.
370 	 */
371 	nxt = -1;
372 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
373 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
374 		struct icmp6_hdr *icp;
375 
376 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
377 			sizeof(*icp));
378 		if (icp == NULL) {
379 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
380 			return;
381 		}
382 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
383 		    icp->icmp6_type == ND_REDIRECT) {
384 			/*
385 			 * ICMPv6 error
386 			 * Special case: for redirect (which is
387 			 * informational) we must not send icmp6 error.
388 			 */
389 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
390 			goto freeit;
391 		} else {
392 			/* ICMPv6 informational - send the error */
393 		}
394 	}
395 #if 0 /* controversial */
396 	else if (off >= 0 && nxt == IPPROTO_ESP) {
397 		/*
398 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
399 		 * don't respond.
400 		 */
401 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
402 		goto freeit;
403 	}
404 #endif
405 	else {
406 		/* non-ICMPv6 - send the error */
407 	}
408 
409 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
410 
411 	/* Finally, do rate limitation check. */
412 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
413 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
414 		goto freeit;
415 	}
416 
417 	/*
418 	 * OK, ICMP6 can be generated.
419 	 */
420 
421 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
422 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
423 
424 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
425 	M_PREPEND(m, preplen, M_DONTWAIT);
426 	if (m && M_UNWRITABLE(m, preplen))
427 		m = m_pullup(m, preplen);
428 	if (m == NULL) {
429 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
430 		return;
431 	}
432 
433 	nip6 = mtod(m, struct ip6_hdr *);
434 	nip6->ip6_src  = oip6->ip6_src;
435 	nip6->ip6_dst  = oip6->ip6_dst;
436 
437 	in6_clearscope(&oip6->ip6_src);
438 	in6_clearscope(&oip6->ip6_dst);
439 
440 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
441 	icmp6->icmp6_type = type;
442 	icmp6->icmp6_code = code;
443 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
444 
445 	/*
446 	 * icmp6_reflect() is designed to be in the input path.
447 	 * icmp6_error() can be called from both input and output path,
448 	 * and if we are in output path rcvif could contain bogus value.
449 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
450 	 * information in ip header (nip6).
451 	 */
452 	m_reset_rcvif(m);
453 
454 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
455 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
456 
457 	return;
458 
459   freeit:
460 	/*
461 	 * If we can't tell whether or not we can generate ICMP6, free it.
462 	 */
463 	m_freem(m);
464 }
465 
466 /*
467  * Process a received ICMP6 message.
468  */
469 static void
470 _icmp6_input(struct mbuf *m, int off, int proto)
471 {
472 	struct mbuf *n;
473 	struct ip6_hdr *ip6, *nip6;
474 	struct icmp6_hdr *icmp6, *nicmp6;
475 	int icmp6len = m->m_pkthdr.len - off;
476 	int code, sum, noff;
477 	struct ifnet *rcvif;
478 	struct psref psref;
479 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
480 
481 	rcvif = m_get_rcvif_psref(m, &psref);
482 	if (__predict_false(rcvif == NULL))
483 		goto freeit;
484 
485 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
486 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
487 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
488 
489 	/*
490 	 * Locate icmp6 structure in mbuf, and check
491 	 * that not corrupted and of at least minimum length
492 	 */
493 
494 	if (icmp6len < sizeof(struct icmp6_hdr)) {
495 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
496 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
497 		goto freeit;
498 	}
499 
500 	if (m->m_len < sizeof(struct ip6_hdr)) {
501 		m = m_pullup(m, sizeof(struct ip6_hdr));
502 		if (m == NULL) {
503 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
504 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
505 			goto freeit;
506 		}
507 	}
508 
509 	ip6 = mtod(m, struct ip6_hdr *);
510 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
511 	if (icmp6 == NULL) {
512 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
513 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
514 		goto freeit;
515 	}
516 	/*
517 	 * Enforce alignment requirements that are violated in
518 	 * some cases, see kern/50766 for details.
519 	 */
520 	if (IP6_HDR_ALIGNED_P(icmp6) == 0) {
521 		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
522 		if (m == NULL) {
523 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
524 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
525 			goto freeit;
526 		}
527 		ip6 = mtod(m, struct ip6_hdr *);
528 		icmp6 = (struct icmp6_hdr *)(ip6 + 1);
529 	}
530 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
531 
532 	/*
533 	 * calculate the checksum
534 	 */
535 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
536 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
537 		    icmp6->icmp6_type, sum, IN6_PRINT(ip6buf, &ip6->ip6_src));
538 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
539 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
540 		goto freeit;
541 	}
542 
543 #if defined(NFAITH) && 0 < NFAITH
544 	if (faithprefix(&ip6->ip6_dst)) {
545 		/*
546 		 * Deliver very specific ICMP6 type only.
547 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
548 		 * will not work.
549 		 */
550 		switch (icmp6->icmp6_type) {
551 		case ICMP6_DST_UNREACH:
552 		case ICMP6_PACKET_TOO_BIG:
553 		case ICMP6_TIME_EXCEEDED:
554 			break;
555 		default:
556 			goto freeit;
557 		}
558 	}
559 #endif
560 
561 	code = icmp6->icmp6_code;
562 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
563 
564 	switch (icmp6->icmp6_type) {
565 	case ICMP6_DST_UNREACH:
566 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
567 		switch (code) {
568 		case ICMP6_DST_UNREACH_NOROUTE:
569 			code = PRC_UNREACH_NET;
570 			break;
571 		case ICMP6_DST_UNREACH_ADMIN:
572 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
573 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
574 			break;
575 		case ICMP6_DST_UNREACH_ADDR:
576 			code = PRC_HOSTDEAD;
577 			break;
578 #ifdef COMPAT_RFC1885
579 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
580 			code = PRC_UNREACH_SRCFAIL;
581 			break;
582 #else
583 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
584 			/* I mean "source address was incorrect." */
585 			code = PRC_UNREACH_NET;
586 			break;
587 #endif
588 		case ICMP6_DST_UNREACH_NOPORT:
589 			code = PRC_UNREACH_PORT;
590 			break;
591 		default:
592 			goto badcode;
593 		}
594 		goto deliver;
595 
596 	case ICMP6_PACKET_TOO_BIG:
597 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
598 
599 		/*
600 		 * MTU is checked in icmp6_mtudisc.
601 		 */
602 		code = PRC_MSGSIZE;
603 
604 		/*
605 		 * Updating the path MTU will be done after examining
606 		 * intermediate extension headers.
607 		 */
608 		goto deliver;
609 
610 	case ICMP6_TIME_EXCEEDED:
611 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
612 		switch (code) {
613 		case ICMP6_TIME_EXCEED_TRANSIT:
614 			code = PRC_TIMXCEED_INTRANS;
615 			break;
616 		case ICMP6_TIME_EXCEED_REASSEMBLY:
617 			code = PRC_TIMXCEED_REASS;
618 			break;
619 		default:
620 			goto badcode;
621 		}
622 		goto deliver;
623 
624 	case ICMP6_PARAM_PROB:
625 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
626 		switch (code) {
627 		case ICMP6_PARAMPROB_NEXTHEADER:
628 			code = PRC_UNREACH_PROTOCOL;
629 			break;
630 		case ICMP6_PARAMPROB_HEADER:
631 		case ICMP6_PARAMPROB_OPTION:
632 			code = PRC_PARAMPROB;
633 			break;
634 		default:
635 			goto badcode;
636 		}
637 		goto deliver;
638 
639 	case ICMP6_ECHO_REQUEST:
640 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
641 		if (code != 0)
642 			goto badcode;
643 		/*
644 		 * Copy mbuf to send to two data paths: userland socket(s),
645 		 * and to the querier (echo reply).
646 		 * m: a copy for socket, n: a copy for querier
647 		 *
648 		 * If the first mbuf is shared, or the first mbuf is too short,
649 		 * copy the first part of the data into a fresh mbuf.
650 		 * Otherwise, we will wrongly overwrite both copies.
651 		 */
652 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
653 			/* Give up local */
654 			n = m;
655 			m = NULL;
656 		} else if (M_READONLY(n) ||
657 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
658 			struct mbuf *n0 = n;
659 
660 			/*
661 			 * Prepare an internal mbuf.  m_pullup() doesn't
662 			 * always copy the length we specified.
663 			 */
664 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
665 				/* Give up local */
666 				n = m;
667 				m = NULL;
668 			}
669 			m_freem(n0);
670 		}
671 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
672 		    sizeof(*nicmp6));
673 		if (nicmp6 == NULL)
674 			goto freeit;
675 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
676 		nicmp6->icmp6_code = 0;
677 		if (n) {
678 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
679 			icmp6s[ICMP6_STAT_REFLECT]++;
680 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
681 			ICMP6_STAT_PUTREF();
682 			icmp6_reflect(n, off);
683 		}
684 		if (!m)
685 			goto freeit;
686 		break;
687 
688 	case ICMP6_ECHO_REPLY:
689 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
690 		if (code != 0)
691 			goto badcode;
692 		break;
693 
694 	case MLD_LISTENER_QUERY:
695 	case MLD_LISTENER_REPORT:
696 		if (icmp6len < sizeof(struct mld_hdr))
697 			goto badlen;
698 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
699 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
700 		else
701 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
702 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
703 			/* give up local */
704 			mld_input(m, off);
705 			m = NULL;
706 			goto freeit;
707 		}
708 		mld_input(n, off);
709 		/* m stays. */
710 		break;
711 
712 	case MLD_LISTENER_DONE:
713 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
714 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
715 			goto badlen;
716 		break;		/* nothing to be done in kernel */
717 
718 	case MLD_MTRACE_RESP:
719 	case MLD_MTRACE:
720 		/* XXX: these two are experimental.  not officially defined. */
721 		/* XXX: per-interface statistics? */
722 		break;		/* just pass it to applications */
723 
724 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
725 	    {
726 		enum { WRU, FQDN } mode;
727 
728 		if (!icmp6_nodeinfo)
729 			break;
730 
731 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
732 			mode = WRU;
733 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
734 			mode = FQDN;
735 		else
736 			goto badlen;
737 
738 		if (mode == FQDN) {
739 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
740 			if (n)
741 				n = ni6_input(n, off);
742 			/* XXX meaningless if n == NULL */
743 			noff = sizeof(struct ip6_hdr);
744 		} else {
745 			u_char *p;
746 			int maxhlen;
747 
748 			if ((icmp6_nodeinfo & 5) != 5)
749 				break;
750 
751 			if (code != 0)
752 				goto badcode;
753 			MGETHDR(n, M_DONTWAIT, m->m_type);
754 			if (n && ICMP6_MAXLEN > MHLEN) {
755 				MCLGET(n, M_DONTWAIT);
756 				if ((n->m_flags & M_EXT) == 0) {
757 					m_free(n);
758 					n = NULL;
759 				}
760 			}
761 			if (n == NULL) {
762 				/* Give up remote */
763 				break;
764 			}
765 			m_reset_rcvif(n);
766 			n->m_len = 0;
767 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
768 			if (maxhlen < 0)
769 				break;
770 			if (maxhlen > hostnamelen)
771 				maxhlen = hostnamelen;
772 			/*
773 			 * Copy IPv6 and ICMPv6 only.
774 			 */
775 			nip6 = mtod(n, struct ip6_hdr *);
776 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
777 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
778 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
779 			p = (u_char *)(nicmp6 + 1);
780 			memset(p, 0, 4);
781 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
782 			noff = sizeof(struct ip6_hdr);
783 			M_COPY_PKTHDR(n, m); /* just for rcvif */
784 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
785 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
786 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
787 			nicmp6->icmp6_code = 0;
788 		}
789 #undef hostnamelen
790 		if (n) {
791 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
792 			icmp6s[ICMP6_STAT_REFLECT]++;
793 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
794 			ICMP6_STAT_PUTREF();
795 			icmp6_reflect(n, noff);
796 		}
797 		break;
798 	    }
799 
800 	case ICMP6_WRUREPLY:
801 		if (code != 0)
802 			goto badcode;
803 		break;
804 
805 	case ND_ROUTER_SOLICIT:
806 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
807 		if (code != 0)
808 			goto badcode;
809 		if (icmp6len < sizeof(struct nd_router_solicit))
810 			goto badlen;
811 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
812 			/* give up local */
813 			nd6_rs_input(m, off, icmp6len);
814 			m = NULL;
815 			goto freeit;
816 		}
817 		nd6_rs_input(n, off, icmp6len);
818 		/* m stays. */
819 		break;
820 
821 	case ND_ROUTER_ADVERT:
822 		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
823 		if (code != 0)
824 			goto badcode;
825 		if (icmp6len < sizeof(struct nd_router_advert))
826 			goto badlen;
827 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
828 			/* give up local */
829 			nd6_ra_input(m, off, icmp6len);
830 			m = NULL;
831 			goto freeit;
832 		}
833 		nd6_ra_input(n, off, icmp6len);
834 		/* m stays. */
835 		break;
836 
837 	case ND_NEIGHBOR_SOLICIT:
838 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
839 		if (code != 0)
840 			goto badcode;
841 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
842 			goto badlen;
843 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
844 			/* give up local */
845 			nd6_ns_input(m, off, icmp6len);
846 			m = NULL;
847 			goto freeit;
848 		}
849 		nd6_ns_input(n, off, icmp6len);
850 		/* m stays. */
851 		break;
852 
853 	case ND_NEIGHBOR_ADVERT:
854 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
855 		if (code != 0)
856 			goto badcode;
857 		if (icmp6len < sizeof(struct nd_neighbor_advert))
858 			goto badlen;
859 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
860 			/* give up local */
861 			nd6_na_input(m, off, icmp6len);
862 			m = NULL;
863 			goto freeit;
864 		}
865 		nd6_na_input(n, off, icmp6len);
866 		/* m stays. */
867 		break;
868 
869 	case ND_REDIRECT:
870 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
871 		if (code != 0)
872 			goto badcode;
873 		if (icmp6len < sizeof(struct nd_redirect))
874 			goto badlen;
875 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
876 			/* give up local */
877 			icmp6_redirect_input(m, off);
878 			m = NULL;
879 			goto freeit;
880 		}
881 		icmp6_redirect_input(n, off);
882 		/* m stays. */
883 		break;
884 
885 	case ICMP6_ROUTER_RENUMBERING:
886 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
887 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
888 			goto badcode;
889 		if (icmp6len < sizeof(struct icmp6_router_renum))
890 			goto badlen;
891 		break;
892 
893 	default:
894 		nd6log(LOG_DEBUG,
895 		    "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
896 		    icmp6->icmp6_type,
897 		    IN6_PRINT(ip6buf, &ip6->ip6_src),
898 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst),
899 		    rcvif ? rcvif->if_index : 0);
900 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
901 			/* ICMPv6 error: MUST deliver it by spec... */
902 			code = PRC_NCMDS;
903 			/* deliver */
904 		} else {
905 			/* ICMPv6 informational: MUST not deliver */
906 			break;
907 		}
908 	deliver:
909 		if (icmp6_notify_error(m, off, icmp6len, code)) {
910 			/* In this case, m should've been freed. */
911 			m_put_rcvif_psref(rcvif, &psref);
912 			return;
913 		}
914 		break;
915 
916 	badcode:
917 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
918 		break;
919 
920 	badlen:
921 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
922 		break;
923 	}
924 	m_put_rcvif_psref(rcvif, &psref);
925 
926 	/* deliver the packet to appropriate sockets */
927 	icmp6_rip6_input(&m, off);
928 
929 	return;
930 
931  freeit:
932 	m_put_rcvif_psref(rcvif, &psref);
933 	m_freem(m);
934 	return;
935 }
936 
937 int
938 icmp6_input(struct mbuf **mp, int *offp, int proto)
939 {
940 
941 	wqinput_input(icmp6_wqinput, *mp, *offp, proto);
942 
943 	return IPPROTO_DONE;
944 }
945 
946 static int
947 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
948 {
949 	struct icmp6_hdr *icmp6;
950 	struct ip6_hdr *eip6;
951 	u_int32_t notifymtu;
952 	struct sockaddr_in6 icmp6src, icmp6dst;
953 
954 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
955 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
956 		goto freeit;
957 	}
958 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
959 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
960 	if (icmp6 == NULL) {
961 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
962 		return (-1);
963 	}
964 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
965 
966 	/* Detect the upper level protocol */
967 	{
968 		void (*ctlfunc)(int, struct sockaddr *, void *);
969 		u_int8_t nxt = eip6->ip6_nxt;
970 		int eoff = off + sizeof(struct icmp6_hdr) +
971 			sizeof(struct ip6_hdr);
972 		struct ip6ctlparam ip6cp;
973 		struct in6_addr *finaldst = NULL;
974 		int icmp6type = icmp6->icmp6_type;
975 		struct ip6_frag *fh;
976 		struct ip6_rthdr *rth;
977 		struct ip6_rthdr0 *rth0;
978 		int rthlen;
979 		struct ifnet *rcvif;
980 		int s;
981 
982 		while (1) { /* XXX: should avoid infinite loop explicitly? */
983 			struct ip6_ext *eh;
984 
985 			switch (nxt) {
986 			case IPPROTO_HOPOPTS:
987 			case IPPROTO_DSTOPTS:
988 			case IPPROTO_AH:
989 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
990 					       eoff, sizeof(*eh));
991 				if (eh == NULL) {
992 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
993 					return (-1);
994 				}
995 
996 				if (nxt == IPPROTO_AH)
997 					eoff += (eh->ip6e_len + 2) << 2;
998 				else
999 					eoff += (eh->ip6e_len + 1) << 3;
1000 				nxt = eh->ip6e_nxt;
1001 				break;
1002 			case IPPROTO_ROUTING:
1003 				/*
1004 				 * When the erroneous packet contains a
1005 				 * routing header, we should examine the
1006 				 * header to determine the final destination.
1007 				 * Otherwise, we can't properly update
1008 				 * information that depends on the final
1009 				 * destination (e.g. path MTU).
1010 				 */
1011 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1012 					       eoff, sizeof(*rth));
1013 				if (rth == NULL) {
1014 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1015 					return (-1);
1016 				}
1017 				rthlen = (rth->ip6r_len + 1) << 3;
1018 				/*
1019 				 * XXX: currently there is no
1020 				 * officially defined type other
1021 				 * than type-0.
1022 				 * Note that if the segment left field
1023 				 * is 0, all intermediate hops must
1024 				 * have been passed.
1025 				 */
1026 				if (rth->ip6r_segleft &&
1027 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1028 					int hops;
1029 
1030 					IP6_EXTHDR_GET(rth0,
1031 						       struct ip6_rthdr0 *, m,
1032 						       eoff, rthlen);
1033 					if (rth0 == NULL) {
1034 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1035 						return (-1);
1036 					}
1037 					/* just ignore a bogus header */
1038 					if ((rth0->ip6r0_len % 2) == 0 &&
1039 					    (hops = rth0->ip6r0_len/2))
1040 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1041 				}
1042 				eoff += rthlen;
1043 				nxt = rth->ip6r_nxt;
1044 				break;
1045 			case IPPROTO_FRAGMENT:
1046 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1047 					       eoff, sizeof(*fh));
1048 				if (fh == NULL) {
1049 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1050 					return (-1);
1051 				}
1052 				/*
1053 				 * Data after a fragment header is meaningless
1054 				 * unless it is the first fragment, but
1055 				 * we'll go to the notify label for path MTU
1056 				 * discovery.
1057 				 */
1058 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1059 					goto notify;
1060 
1061 				eoff += sizeof(struct ip6_frag);
1062 				nxt = fh->ip6f_nxt;
1063 				break;
1064 			default:
1065 				/*
1066 				 * This case includes ESP and the No Next
1067 				 * Header.  In such cases going to the notify
1068 				 * label does not have any meaning
1069 				 * (i.e. ctlfunc will be NULL), but we go
1070 				 * anyway since we might have to update
1071 				 * path MTU information.
1072 				 */
1073 				goto notify;
1074 			}
1075 		}
1076 	  notify:
1077 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1078 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1079 		if (icmp6 == NULL) {
1080 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1081 			return (-1);
1082 		}
1083 
1084 		/*
1085 		 * retrieve parameters from the inner IPv6 header, and convert
1086 		 * them into sockaddr structures.
1087 		 * XXX: there is no guarantee that the source or destination
1088 		 * addresses of the inner packet are in the same scope zone as
1089 		 * the addresses of the icmp packet.  But there is no other
1090 		 * way to determine the zone.
1091 		 */
1092 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1093 
1094 		rcvif = m_get_rcvif(m, &s);
1095 		if (__predict_false(rcvif == NULL))
1096 			goto freeit;
1097 		sockaddr_in6_init(&icmp6dst,
1098 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1099 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1100 			m_put_rcvif(rcvif, &s);
1101 			goto freeit;
1102 		}
1103 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1104 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1105 			m_put_rcvif(rcvif, &s);
1106 			goto freeit;
1107 		}
1108 		m_put_rcvif(rcvif, &s);
1109 
1110 		icmp6src.sin6_flowinfo =
1111 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1112 
1113 		if (finaldst == NULL)
1114 			finaldst = &eip6->ip6_dst;
1115 		ip6cp.ip6c_m = m;
1116 		ip6cp.ip6c_icmp6 = icmp6;
1117 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1118 		ip6cp.ip6c_off = eoff;
1119 		ip6cp.ip6c_finaldst = finaldst;
1120 		ip6cp.ip6c_src = &icmp6src;
1121 		ip6cp.ip6c_nxt = nxt;
1122 
1123 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1124 			notifymtu = ntohl(icmp6->icmp6_mtu);
1125 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1126 		}
1127 
1128 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1129 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1130 		if (ctlfunc) {
1131 			(void) (*ctlfunc)(code, sin6tosa(&icmp6dst),
1132 					  &ip6cp);
1133 		}
1134 	}
1135 	return (0);
1136 
1137   freeit:
1138 	m_freem(m);
1139 	return (-1);
1140 }
1141 
1142 void
1143 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1144 {
1145 	unsigned long rtcount;
1146 	struct icmp6_mtudisc_callback *mc;
1147 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1148 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1149 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1150 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1151 	struct rtentry *rt = NULL;
1152 	struct sockaddr_in6 sin6;
1153 	struct ifnet *rcvif;
1154 	int s;
1155 
1156 	/*
1157 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1158 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1159 	 * In that one case, the MTU might be less than 1280.
1160 	 */
1161 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1162 		/* is the mtu even sane? */
1163 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1164 			return;
1165 		if (!validated)
1166 			return;
1167 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1168 	}
1169 
1170 	/*
1171 	 * allow non-validated cases if memory is plenty, to make traffic
1172 	 * from non-connected pcb happy.
1173 	 */
1174 	mutex_enter(&icmp6_mtx);
1175 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1176 	if (validated) {
1177 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) {
1178 			mutex_exit(&icmp6_mtx);
1179 			return;
1180 		}
1181 		else if (0 <= icmp6_mtudisc_lowat &&
1182 		    rtcount > icmp6_mtudisc_lowat) {
1183 			/*
1184 			 * XXX nuke a victim, install the new one.
1185 			 */
1186 		}
1187 	} else {
1188 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) {
1189 			mutex_exit(&icmp6_mtx);
1190 			return;
1191 		}
1192 	}
1193 	mutex_exit(&icmp6_mtx);
1194 
1195 	memset(&sin6, 0, sizeof(sin6));
1196 	sin6.sin6_family = PF_INET6;
1197 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1198 	sin6.sin6_addr = *dst;
1199 	rcvif = m_get_rcvif(m, &s);
1200 	if (__predict_false(rcvif == NULL))
1201 		return;
1202 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1203 		m_put_rcvif(rcvif, &s);
1204 		return;
1205 	}
1206 	m_put_rcvif(rcvif, &s);
1207 
1208 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1209 
1210 	if (rt && (rt->rt_flags & RTF_HOST) &&
1211 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1212 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1213 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1214 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1215 			rt->rt_rmx.rmx_mtu = mtu;
1216 		}
1217 	}
1218 	if (rt) {
1219 		rt_unref(rt);
1220 	}
1221 
1222 	/*
1223 	 * Notify protocols that the MTU for this destination
1224 	 * has changed.
1225 	 */
1226 	mutex_enter(&icmp6_mtx);
1227 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1228 	     mc = LIST_NEXT(mc, mc_list))
1229 		(*mc->mc_func)(&sin6.sin6_addr);
1230 	mutex_exit(&icmp6_mtx);
1231 }
1232 
1233 /*
1234  * Process a Node Information Query packet, based on
1235  * draft-ietf-ipngwg-icmp-name-lookups-07.
1236  *
1237  * Spec incompatibilities:
1238  * - IPv6 Subject address handling
1239  * - IPv4 Subject address handling support missing
1240  * - Proxy reply (answer even if it's not for me)
1241  * - joins NI group address at in6_ifattach() time only, does not cope
1242  *   with hostname changes by sethostname(3)
1243  */
1244 static struct mbuf *
1245 ni6_input(struct mbuf *m, int off)
1246 {
1247 	struct icmp6_nodeinfo *ni6, *nni6;
1248 	struct mbuf *n = NULL;
1249 	u_int16_t qtype;
1250 	int subjlen;
1251 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1252 	struct ni_reply_fqdn *fqdn;
1253 	int addrs;		/* for NI_QTYPE_NODEADDR */
1254 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1255 	struct sockaddr_in6 sin6; /* ip6_dst */
1256 	struct in6_addr in6_subj; /* subject address */
1257 	struct ip6_hdr *ip6;
1258 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1259 	char *subj = NULL;
1260 	struct ifnet *rcvif;
1261 	int s, ss;
1262 	struct ifaddr *ifa;
1263 	struct psref psref;
1264 
1265 	ip6 = mtod(m, struct ip6_hdr *);
1266 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1267 	if (ni6 == NULL) {
1268 		/* m is already reclaimed */
1269 		return NULL;
1270 	}
1271 
1272 	/*
1273 	 * Validate IPv6 destination address.
1274 	 *
1275 	 * The Responder must discard the Query without further processing
1276 	 * unless it is one of the Responder's unicast or anycast addresses, or
1277 	 * a link-local scope multicast address which the Responder has joined.
1278 	 * [icmp-name-lookups-07, Section 4.]
1279 	 */
1280 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1281 	/* XXX scopeid */
1282 	ss = pserialize_read_enter();
1283 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1284 	if (ifa != NULL)
1285 		; /* unicast/anycast, fine */
1286 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1287 		; /* link-local multicast, fine */
1288 	else {
1289 		pserialize_read_exit(ss);
1290 		goto bad;
1291 	}
1292 	pserialize_read_exit(ss);
1293 
1294 	/* validate query Subject field. */
1295 	qtype = ntohs(ni6->ni_qtype);
1296 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1297 	switch (qtype) {
1298 	case NI_QTYPE_NOOP:
1299 	case NI_QTYPE_SUPTYPES:
1300 		/* 07 draft */
1301 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1302 			break;
1303 		/* FALLTHROUGH */
1304 	case NI_QTYPE_FQDN:
1305 	case NI_QTYPE_NODEADDR:
1306 	case NI_QTYPE_IPV4ADDR:
1307 		switch (ni6->ni_code) {
1308 		case ICMP6_NI_SUBJ_IPV6:
1309 #if ICMP6_NI_SUBJ_IPV6 != 0
1310 		case 0:
1311 #endif
1312 			/*
1313 			 * backward compatibility - try to accept 03 draft
1314 			 * format, where no Subject is present.
1315 			 */
1316 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1317 			    subjlen == 0) {
1318 				oldfqdn++;
1319 				break;
1320 			}
1321 #if ICMP6_NI_SUBJ_IPV6 != 0
1322 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1323 				goto bad;
1324 #endif
1325 
1326 			if (subjlen != sizeof(sin6.sin6_addr))
1327 				goto bad;
1328 
1329 			/*
1330 			 * Validate Subject address.
1331 			 *
1332 			 * Not sure what exactly "address belongs to the node"
1333 			 * means in the spec, is it just unicast, or what?
1334 			 *
1335 			 * At this moment we consider Subject address as
1336 			 * "belong to the node" if the Subject address equals
1337 			 * to the IPv6 destination address; validation for
1338 			 * IPv6 destination address should have done enough
1339 			 * check for us.
1340 			 *
1341 			 * We do not do proxy at this moment.
1342 			 */
1343 			/* m_pulldown instead of copy? */
1344 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1345 			    subjlen, (void *)&in6_subj);
1346 			rcvif = m_get_rcvif(m, &s);
1347 			if (__predict_false(rcvif == NULL))
1348 				goto bad;
1349 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1350 				m_put_rcvif(rcvif, &s);
1351 				goto bad;
1352 			}
1353 			m_put_rcvif(rcvif, &s);
1354 
1355 			subj = (char *)&in6_subj;
1356 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1357 				break;
1358 
1359 			/*
1360 			 * XXX if we are to allow other cases, we should really
1361 			 * be careful about scope here.
1362 			 * basically, we should disallow queries toward IPv6
1363 			 * destination X with subject Y, if scope(X) > scope(Y).
1364 			 * if we allow scope(X) > scope(Y), it will result in
1365 			 * information leakage across scope boundary.
1366 			 */
1367 			goto bad;
1368 
1369 		case ICMP6_NI_SUBJ_FQDN:
1370 			/*
1371 			 * Validate Subject name with gethostname(3).
1372 			 *
1373 			 * The behavior may need some debate, since:
1374 			 * - we are not sure if the node has FQDN as
1375 			 *   hostname (returned by gethostname(3)).
1376 			 * - the code does wildcard match for truncated names.
1377 			 *   however, we are not sure if we want to perform
1378 			 *   wildcard match, if gethostname(3) side has
1379 			 *   truncated hostname.
1380 			 */
1381 			n = ni6_nametodns(hostname, hostnamelen, 0);
1382 			if (!n || n->m_next || n->m_len == 0)
1383 				goto bad;
1384 			IP6_EXTHDR_GET(subj, char *, m,
1385 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1386 			if (subj == NULL)
1387 				goto bad;
1388 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1389 					n->m_len)) {
1390 				goto bad;
1391 			}
1392 			m_freem(n);
1393 			n = NULL;
1394 			break;
1395 
1396 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1397 		default:
1398 			goto bad;
1399 		}
1400 		break;
1401 	}
1402 
1403 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1404 	switch (qtype) {
1405 	case NI_QTYPE_FQDN:
1406 		if ((icmp6_nodeinfo & 1) == 0)
1407 			goto bad;
1408 		break;
1409 	case NI_QTYPE_NODEADDR:
1410 	case NI_QTYPE_IPV4ADDR:
1411 		if ((icmp6_nodeinfo & 2) == 0)
1412 			goto bad;
1413 		break;
1414 	}
1415 
1416 	/* guess reply length */
1417 	switch (qtype) {
1418 	case NI_QTYPE_NOOP:
1419 		break;		/* no reply data */
1420 	case NI_QTYPE_SUPTYPES:
1421 		replylen += sizeof(u_int32_t);
1422 		break;
1423 	case NI_QTYPE_FQDN:
1424 		/* XXX will append an mbuf */
1425 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1426 		break;
1427 	case NI_QTYPE_NODEADDR:
1428 		addrs = ni6_addrs(ni6, m, &ifp, subj, &psref);
1429 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1430 					  sizeof(u_int32_t))) > MCLBYTES)
1431 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1432 		break;
1433 	case NI_QTYPE_IPV4ADDR:
1434 		/* unsupported - should respond with unknown Qtype? */
1435 		goto bad;
1436 	default:
1437 		/*
1438 		 * XXX: We must return a reply with the ICMP6 code
1439 		 * `unknown Qtype' in this case.  However we regard the case
1440 		 * as an FQDN query for backward compatibility.
1441 		 * Older versions set a random value to this field,
1442 		 * so it rarely varies in the defined qtypes.
1443 		 * But the mechanism is not reliable...
1444 		 * maybe we should obsolete older versions.
1445 		 */
1446 		qtype = NI_QTYPE_FQDN;
1447 		/* XXX will append an mbuf */
1448 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1449 		oldfqdn++;
1450 		break;
1451 	}
1452 
1453 	/* allocate an mbuf to reply. */
1454 	MGETHDR(n, M_DONTWAIT, m->m_type);
1455 	if (n == NULL) {
1456 		if_put(ifp, &psref);
1457 		m_freem(m);
1458 		return (NULL);
1459 	}
1460 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1461 	if (replylen > MHLEN) {
1462 		if (replylen > MCLBYTES) {
1463 			/*
1464 			 * XXX: should we try to allocate more? But MCLBYTES
1465 			 * is probably much larger than IPV6_MMTU...
1466 			 */
1467 			goto bad;
1468 		}
1469 		MCLGET(n, M_DONTWAIT);
1470 		if ((n->m_flags & M_EXT) == 0) {
1471 			goto bad;
1472 		}
1473 	}
1474 	n->m_pkthdr.len = n->m_len = replylen;
1475 
1476 	/* copy mbuf header and IPv6 + Node Information base headers */
1477 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1478 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1479 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1480 
1481 	/* qtype dependent procedure */
1482 	switch (qtype) {
1483 	case NI_QTYPE_NOOP:
1484 		nni6->ni_code = ICMP6_NI_SUCCESS;
1485 		nni6->ni_flags = 0;
1486 		break;
1487 	case NI_QTYPE_SUPTYPES:
1488 	{
1489 		u_int32_t v;
1490 		nni6->ni_code = ICMP6_NI_SUCCESS;
1491 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1492 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1493 		v = (u_int32_t)htonl(0x0000000f);
1494 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1495 		break;
1496 	}
1497 	case NI_QTYPE_FQDN:
1498 		nni6->ni_code = ICMP6_NI_SUCCESS;
1499 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1500 						sizeof(struct ip6_hdr) +
1501 						sizeof(struct icmp6_nodeinfo));
1502 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1503 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1504 		/*
1505 		 * XXX do we really have FQDN in variable "hostname"?
1506 		 */
1507 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1508 		if (n->m_next == NULL)
1509 			goto bad;
1510 		/* XXX we assume that n->m_next is not a chain */
1511 		if (n->m_next->m_next != NULL)
1512 			goto bad;
1513 		n->m_pkthdr.len += n->m_next->m_len;
1514 		break;
1515 	case NI_QTYPE_NODEADDR:
1516 	{
1517 		int lenlim, copied;
1518 
1519 		nni6->ni_code = ICMP6_NI_SUCCESS;
1520 		n->m_pkthdr.len = n->m_len =
1521 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1522 		lenlim = M_TRAILINGSPACE(n);
1523 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1524 		if_put(ifp, &psref);
1525 		ifp = NULL;
1526 		/* XXX: reset mbuf length */
1527 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1528 			sizeof(struct icmp6_nodeinfo) + copied;
1529 		break;
1530 	}
1531 	default:
1532 		break;		/* XXX impossible! */
1533 	}
1534 
1535 	nni6->ni_type = ICMP6_NI_REPLY;
1536 	m_freem(m);
1537 	return (n);
1538 
1539   bad:
1540 	if_put(ifp, &psref);
1541 	m_freem(m);
1542 	if (n)
1543 		m_freem(n);
1544 	return (NULL);
1545 }
1546 #undef hostnamelen
1547 
1548 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1549 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1550 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1551 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1552 
1553 /*
1554  * make a mbuf with DNS-encoded string.  no compression support.
1555  *
1556  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1557  * treated as truncated name (two \0 at the end).  this is a wild guess.
1558  *
1559  * old - return pascal string if non-zero
1560  */
1561 static struct mbuf *
1562 ni6_nametodns(const char *name, int namelen, int old)
1563 {
1564 	struct mbuf *m;
1565 	char *cp, *ep;
1566 	const char *p, *q;
1567 	int i, len, nterm;
1568 
1569 	if (old)
1570 		len = namelen + 1;
1571 	else
1572 		len = MCLBYTES;
1573 
1574 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1575 	MGET(m, M_DONTWAIT, MT_DATA);
1576 	if (m && len > MLEN) {
1577 		MCLGET(m, M_DONTWAIT);
1578 		if ((m->m_flags & M_EXT) == 0)
1579 			goto fail;
1580 	}
1581 	if (!m)
1582 		goto fail;
1583 	m->m_next = NULL;
1584 
1585 	if (old) {
1586 		m->m_len = len;
1587 		*mtod(m, char *) = namelen;
1588 		bcopy(name, mtod(m, char *) + 1, namelen);
1589 		return m;
1590 	} else {
1591 		m->m_len = 0;
1592 		cp = mtod(m, char *);
1593 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1594 
1595 		/* if not certain about my name, return empty buffer */
1596 		if (namelen == 0)
1597 			return m;
1598 
1599 		/*
1600 		 * guess if it looks like shortened hostname, or FQDN.
1601 		 * shortened hostname needs two trailing "\0".
1602 		 */
1603 		i = 0;
1604 		for (p = name; p < name + namelen; p++) {
1605 			if (*p == '.')
1606 				i++;
1607 		}
1608 		if (i < 2)
1609 			nterm = 2;
1610 		else
1611 			nterm = 1;
1612 
1613 		p = name;
1614 		while (cp < ep && p < name + namelen) {
1615 			i = 0;
1616 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1617 				i++;
1618 			/* result does not fit into mbuf */
1619 			if (cp + i + 1 >= ep)
1620 				goto fail;
1621 			/*
1622 			 * DNS label length restriction, RFC1035 page 8.
1623 			 * "i == 0" case is included here to avoid returning
1624 			 * 0-length label on "foo..bar".
1625 			 */
1626 			if (i <= 0 || i >= 64)
1627 				goto fail;
1628 			*cp++ = i;
1629 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1630 				goto fail;
1631 			while (i > 0) {
1632 				if (!isalnum(*p) && *p != '-')
1633 					goto fail;
1634 				if (isupper(*p)) {
1635 					*cp++ = tolower(*p);
1636 					p++;
1637 				} else
1638 					*cp++ = *p++;
1639 				i--;
1640 			}
1641 			p = q;
1642 			if (p < name + namelen && *p == '.')
1643 				p++;
1644 		}
1645 		/* termination */
1646 		if (cp + nterm >= ep)
1647 			goto fail;
1648 		while (nterm-- > 0)
1649 			*cp++ = '\0';
1650 		m->m_len = cp - mtod(m, char *);
1651 		return m;
1652 	}
1653 
1654 	panic("should not reach here");
1655 	/* NOTREACHED */
1656 
1657  fail:
1658 	if (m)
1659 		m_freem(m);
1660 	return NULL;
1661 }
1662 
1663 /*
1664  * check if two DNS-encoded string matches.  takes care of truncated
1665  * form (with \0\0 at the end).  no compression support.
1666  * XXX upper/lowercase match (see RFC2065)
1667  */
1668 static int
1669 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1670 {
1671 	const char *a0, *b0;
1672 	int l;
1673 
1674 	/* simplest case - need validation? */
1675 	if (alen == blen && memcmp(a, b, alen) == 0)
1676 		return 1;
1677 
1678 	a0 = a;
1679 	b0 = b;
1680 
1681 	/* termination is mandatory */
1682 	if (alen < 2 || blen < 2)
1683 		return 0;
1684 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1685 		return 0;
1686 	alen--;
1687 	blen--;
1688 
1689 	while (a - a0 < alen && b - b0 < blen) {
1690 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1691 			return 0;
1692 
1693 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1694 			return 0;
1695 		/* we don't support compression yet */
1696 		if (a[0] >= 64 || b[0] >= 64)
1697 			return 0;
1698 
1699 		/* truncated case */
1700 		if (a[0] == 0 && a - a0 == alen - 1)
1701 			return 1;
1702 		if (b[0] == 0 && b - b0 == blen - 1)
1703 			return 1;
1704 		if (a[0] == 0 || b[0] == 0)
1705 			return 0;
1706 
1707 		if (a[0] != b[0])
1708 			return 0;
1709 		l = a[0];
1710 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1711 			return 0;
1712 		if (memcmp(a + 1, b + 1, l) != 0)
1713 			return 0;
1714 
1715 		a += 1 + l;
1716 		b += 1 + l;
1717 	}
1718 
1719 	if (a - a0 == alen && b - b0 == blen)
1720 		return 1;
1721 	else
1722 		return 0;
1723 }
1724 
1725 /*
1726  * calculate the number of addresses to be returned in the node info reply.
1727  */
1728 static int
1729 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1730     struct ifnet **ifpp, char *subj, struct psref *psref)
1731 {
1732 	struct ifnet *ifp;
1733 	struct in6_ifaddr *ia6;
1734 	struct ifaddr *ifa;
1735 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1736 	int addrs = 0, addrsofif, iffound = 0;
1737 	int niflags = ni6->ni_flags;
1738 	int s;
1739 
1740 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1741 		switch (ni6->ni_code) {
1742 		case ICMP6_NI_SUBJ_IPV6:
1743 			if (subj == NULL) /* must be impossible... */
1744 				return (0);
1745 			subj_ip6 = (struct sockaddr_in6 *)subj;
1746 			break;
1747 		default:
1748 			/*
1749 			 * XXX: we only support IPv6 subject address for
1750 			 * this Qtype.
1751 			 */
1752 			return (0);
1753 		}
1754 	}
1755 
1756 	s = pserialize_read_enter();
1757 	IFNET_READER_FOREACH(ifp) {
1758 		addrsofif = 0;
1759 		IFADDR_READER_FOREACH(ifa, ifp) {
1760 			if (ifa->ifa_addr->sa_family != AF_INET6)
1761 				continue;
1762 			ia6 = (struct in6_ifaddr *)ifa;
1763 
1764 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1765 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1766 					       &ia6->ia_addr.sin6_addr))
1767 				iffound = 1;
1768 
1769 			/*
1770 			 * IPv4-mapped addresses can only be returned by a
1771 			 * Node Information proxy, since they represent
1772 			 * addresses of IPv4-only nodes, which perforce do
1773 			 * not implement this protocol.
1774 			 * [icmp-name-lookups-07, Section 5.4]
1775 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1776 			 * this function at this moment.
1777 			 */
1778 
1779 			/* What do we have to do about ::1? */
1780 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1781 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1782 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1783 					continue;
1784 				break;
1785 			case IPV6_ADDR_SCOPE_SITELOCAL:
1786 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1787 					continue;
1788 				break;
1789 			case IPV6_ADDR_SCOPE_GLOBAL:
1790 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1791 					continue;
1792 				break;
1793 			default:
1794 				continue;
1795 			}
1796 
1797 			/*
1798 			 * check if anycast is okay.
1799 			 * XXX: just experimental.  not in the spec.
1800 			 */
1801 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1802 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1803 				continue; /* we need only unicast addresses */
1804 
1805 			addrsofif++; /* count the address */
1806 		}
1807 		if (iffound) {
1808 			if_acquire(ifp, psref);
1809 			pserialize_read_exit(s);
1810 			*ifpp = ifp;
1811 			return (addrsofif);
1812 		}
1813 
1814 		addrs += addrsofif;
1815 	}
1816 	pserialize_read_exit(s);
1817 
1818 	return (addrs);
1819 }
1820 
1821 static int
1822 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1823 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1824 	int resid)
1825 {
1826 	struct ifnet *ifp;
1827 	struct in6_ifaddr *ia6;
1828 	struct ifaddr *ifa;
1829 	struct ifnet *ifp_dep = NULL;
1830 	int copied = 0, allow_deprecated = 0;
1831 	u_char *cp = (u_char *)(nni6 + 1);
1832 	int niflags = ni6->ni_flags;
1833 	u_int32_t ltime;
1834 	int s;
1835 
1836 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1837 		return (0);	/* needless to copy */
1838 
1839 	s = pserialize_read_enter();
1840 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1841   again:
1842 
1843 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1844 	{
1845 		IFADDR_READER_FOREACH(ifa, ifp) {
1846 			if (ifa->ifa_addr->sa_family != AF_INET6)
1847 				continue;
1848 			ia6 = (struct in6_ifaddr *)ifa;
1849 
1850 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1851 			    allow_deprecated == 0) {
1852 				/*
1853 				 * prefererred address should be put before
1854 				 * deprecated addresses.
1855 				 */
1856 
1857 				/* record the interface for later search */
1858 				if (ifp_dep == NULL)
1859 					ifp_dep = ifp;
1860 
1861 				continue;
1862 			}
1863 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1864 				 allow_deprecated != 0)
1865 				continue; /* we now collect deprecated addrs */
1866 
1867 			/* What do we have to do about ::1? */
1868 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1869 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1870 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1871 					continue;
1872 				break;
1873 			case IPV6_ADDR_SCOPE_SITELOCAL:
1874 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1875 					continue;
1876 				break;
1877 			case IPV6_ADDR_SCOPE_GLOBAL:
1878 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1879 					continue;
1880 				break;
1881 			default:
1882 				continue;
1883 			}
1884 
1885 			/*
1886 			 * check if anycast is okay.
1887 			 * XXX: just experimental.  not in the spec.
1888 			 */
1889 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1890 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1891 				continue;
1892 
1893 			/* now we can copy the address */
1894 			if (resid < sizeof(struct in6_addr) +
1895 			    sizeof(u_int32_t)) {
1896 				/*
1897 				 * We give up much more copy.
1898 				 * Set the truncate flag and return.
1899 				 */
1900 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1901 				goto out;
1902 			}
1903 
1904 			/*
1905 			 * Set the TTL of the address.
1906 			 * The TTL value should be one of the following
1907 			 * according to the specification:
1908 			 *
1909 			 * 1. The remaining lifetime of a DHCP lease on the
1910 			 *    address, or
1911 			 * 2. The remaining Valid Lifetime of a prefix from
1912 			 *    which the address was derived through Stateless
1913 			 *    Autoconfiguration.
1914 			 *
1915 			 * Note that we currently do not support stateful
1916 			 * address configuration by DHCPv6, so the former
1917 			 * case can't happen.
1918 			 *
1919 			 * TTL must be 2^31 > TTL >= 0.
1920 			 */
1921 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1922 				ltime = ND6_INFINITE_LIFETIME;
1923 			else {
1924 				if (ia6->ia6_lifetime.ia6t_expire >
1925 				    time_uptime)
1926 					ltime = ia6->ia6_lifetime.ia6t_expire -
1927 					    time_uptime;
1928 				else
1929 					ltime = 0;
1930 			}
1931 			if (ltime > 0x7fffffff)
1932 				ltime = 0x7fffffff;
1933 			ltime = htonl(ltime);
1934 
1935 			bcopy(&ltime, cp, sizeof(u_int32_t));
1936 			cp += sizeof(u_int32_t);
1937 
1938 			/* copy the address itself */
1939 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1940 			      sizeof(struct in6_addr));
1941 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1942 			cp += sizeof(struct in6_addr);
1943 
1944 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1945 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1946 		}
1947 		if (ifp0)	/* we need search only on the specified IF */
1948 			break;
1949 	}
1950 
1951 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1952 		ifp = ifp_dep;
1953 		allow_deprecated = 1;
1954 
1955 		goto again;
1956 	}
1957 out:
1958 	pserialize_read_exit(s);
1959 	return (copied);
1960 }
1961 
1962 /*
1963  * XXX almost dup'ed code with rip6_input.
1964  */
1965 static int
1966 icmp6_rip6_input(struct mbuf **mp, int off)
1967 {
1968 	struct mbuf *m = *mp;
1969 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1970 	struct inpcb_hdr *inph;
1971 	struct in6pcb *in6p;
1972 	struct in6pcb *last = NULL;
1973 	struct sockaddr_in6 rip6src;
1974 	struct icmp6_hdr *icmp6;
1975 	struct mbuf *opts = NULL;
1976 
1977 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1978 	if (icmp6 == NULL) {
1979 		/* m is already reclaimed */
1980 		return IPPROTO_DONE;
1981 	}
1982 
1983 	/*
1984 	 * XXX: the address may have embedded scope zone ID, which should be
1985 	 * hidden from applications.
1986 	 */
1987 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1988 	if (sa6_recoverscope(&rip6src)) {
1989 		m_freem(m);
1990 		return (IPPROTO_DONE);
1991 	}
1992 
1993 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1994 		in6p = (struct in6pcb *)inph;
1995 		if (in6p->in6p_af != AF_INET6)
1996 			continue;
1997 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1998 			continue;
1999 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2000 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2001 			continue;
2002 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2003 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2004 			continue;
2005 		if (in6p->in6p_icmp6filt
2006 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2007 				 in6p->in6p_icmp6filt))
2008 			continue;
2009 		if (last) {
2010 			struct	mbuf *n;
2011 #ifdef IPSEC
2012 			/*
2013 			 * Check AH/ESP integrity
2014 			 */
2015 			if (!ipsec_used ||
2016 			    (ipsec_used && !ipsec6_in_reject(m, last)))
2017 #endif /* IPSEC */
2018 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2019 				if (last->in6p_flags & IN6P_CONTROLOPTS)
2020 					ip6_savecontrol(last, &opts, ip6, n);
2021 				/* strip intermediate headers */
2022 				m_adj(n, off);
2023 				if (sbappendaddr(&last->in6p_socket->so_rcv,
2024 				    sin6tosa(&rip6src), n, opts) == 0) {
2025 					/* should notify about lost packet */
2026 					m_freem(n);
2027 					if (opts)
2028 						m_freem(opts);
2029 				} else
2030 					sorwakeup(last->in6p_socket);
2031 				opts = NULL;
2032 			}
2033 		}
2034 		last = in6p;
2035 	}
2036 #ifdef IPSEC
2037 	if (ipsec_used && last && ipsec6_in_reject(m, last)) {
2038 		m_freem(m);
2039 		/*
2040 		 * XXX ipsec6_in_reject update stat if there is an error
2041 		 * so we just need to update stats by hand in the case of last is
2042 		 * NULL
2043 		 */
2044 		if (!last)
2045 			IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
2046 			IP6_STATDEC(IP6_STAT_DELIVERED);
2047 			/* do not inject data into pcb */
2048 		} else
2049 #endif /* IPSEC */
2050 	if (last) {
2051 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2052 			ip6_savecontrol(last, &opts, ip6, m);
2053 		/* strip intermediate headers */
2054 		m_adj(m, off);
2055 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2056 		    sin6tosa(&rip6src), m, opts) == 0) {
2057 			m_freem(m);
2058 			if (opts)
2059 				m_freem(opts);
2060 		} else
2061 			sorwakeup(last->in6p_socket);
2062 	} else {
2063 		m_freem(m);
2064 		IP6_STATDEC(IP6_STAT_DELIVERED);
2065 	}
2066 	return IPPROTO_DONE;
2067 }
2068 
2069 /*
2070  * Reflect the ip6 packet back to the source.
2071  * OFF points to the icmp6 header, counted from the top of the mbuf.
2072  *
2073  * Note: RFC 1885 required that an echo reply should be truncated if it
2074  * did not fit in with (return) path MTU, and KAME code supported the
2075  * behavior.  However, as a clarification after the RFC, this limitation
2076  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2077  * old behavior, with a (non-default) ifdef block, while the new version of
2078  * the spec was an internet-draft status, and even after the new RFC was
2079  * published.  But it would rather make sense to clean the obsoleted part
2080  * up, and to make the code simpler at this stage.
2081  */
2082 void
2083 icmp6_reflect(struct mbuf *m, size_t off)
2084 {
2085 	struct ip6_hdr *ip6;
2086 	struct icmp6_hdr *icmp6;
2087 	const struct in6_ifaddr *ia;
2088 	const struct ip6aux *ip6a;
2089 	int plen;
2090 	int type, code;
2091 	struct ifnet *outif = NULL;
2092 	struct in6_addr origdst;
2093 	struct ifnet *rcvif;
2094 	int s;
2095 	bool ip6_src_filled = false;
2096 
2097 	/* too short to reflect */
2098 	if (off < sizeof(struct ip6_hdr)) {
2099 		nd6log(LOG_DEBUG,
2100 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2101 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2102 		    __FILE__, __LINE__);
2103 		goto bad;
2104 	}
2105 
2106 	/*
2107 	 * If there are extra headers between IPv6 and ICMPv6, strip
2108 	 * off that header first.
2109 	 */
2110 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
2111 	if (off > sizeof(struct ip6_hdr)) {
2112 		size_t l;
2113 		struct ip6_hdr nip6;
2114 
2115 		l = off - sizeof(struct ip6_hdr);
2116 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2117 		m_adj(m, l);
2118 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2119 		if (m->m_len < l) {
2120 			if ((m = m_pullup(m, l)) == NULL)
2121 				return;
2122 		}
2123 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
2124 	} else /* off == sizeof(struct ip6_hdr) */ {
2125 		size_t l;
2126 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2127 		if (m->m_len < l) {
2128 			if ((m = m_pullup(m, l)) == NULL)
2129 				return;
2130 		}
2131 	}
2132 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2133 	ip6 = mtod(m, struct ip6_hdr *);
2134 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2135 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2136 	type = icmp6->icmp6_type; /* keep type for statistics */
2137 	code = icmp6->icmp6_code; /* ditto. */
2138 
2139 	origdst = ip6->ip6_dst;
2140 	/*
2141 	 * ip6_input() drops a packet if its src is multicast.
2142 	 * So, the src is never multicast.
2143 	 */
2144 	ip6->ip6_dst = ip6->ip6_src;
2145 
2146 	/*
2147 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2148 	 * use dst as the src for the reply.
2149 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2150 	 * (for example) when we encounter an error while forwarding procedure
2151 	 * destined to a duplicated address of ours.
2152 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2153 	 * procedure of an outgoing packet of our own, in which case we need
2154 	 * to search in the ifaddr list.
2155 	 */
2156 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2157 		;
2158 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2159 		if ((ip6a->ip6a_flags &
2160 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2161 			ip6->ip6_src = ip6a->ip6a_src;
2162 			ip6_src_filled = true;
2163 		}
2164 	} else {
2165 		union {
2166 			struct sockaddr_in6 sin6;
2167 			struct sockaddr sa;
2168 		} u;
2169 		int _s;
2170 		struct ifaddr *ifa;
2171 
2172 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2173 
2174 		_s = pserialize_read_enter();
2175 		ifa = ifa_ifwithaddr(&u.sa);
2176 
2177 		if (ifa != NULL) {
2178 			ia = ifatoia6(ifa);
2179 			if ((ia->ia6_flags &
2180 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2181 				ip6->ip6_src = ia->ia_addr.sin6_addr;
2182 				ip6_src_filled = true;
2183 			}
2184 		}
2185 		pserialize_read_exit(_s);
2186 	}
2187 
2188 	if (!ip6_src_filled) {
2189 		int e;
2190 		struct sockaddr_in6 sin6;
2191 		struct route ro;
2192 
2193 		/*
2194 		 * This case matches to multicasts, our anycast, or unicasts
2195 		 * that we do not own.  Select a source address based on the
2196 		 * source address of the erroneous packet.
2197 		 */
2198 		/* zone ID should be embedded */
2199 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2200 
2201 		memset(&ro, 0, sizeof(ro));
2202 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2203 		    &ip6->ip6_src);
2204 		rtcache_free(&ro);
2205 		if (e != 0) {
2206 			char ip6buf[INET6_ADDRSTRLEN];
2207 			nd6log(LOG_DEBUG,
2208 			    "source can't be determined: "
2209 			    "dst=%s, error=%d\n",
2210 			    IN6_PRINT(ip6buf, &sin6.sin6_addr), e);
2211 			goto bad;
2212 		}
2213 	}
2214 
2215 	ip6->ip6_flow = 0;
2216 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2217 	ip6->ip6_vfc |= IPV6_VERSION;
2218 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2219 	rcvif = m_get_rcvif(m, &s);
2220 	if (rcvif) {
2221 		/* XXX: This may not be the outgoing interface */
2222 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2223 	} else
2224 		ip6->ip6_hlim = ip6_defhlim;
2225 	m_put_rcvif(rcvif, &s);
2226 
2227 	m->m_pkthdr.csum_flags = 0;
2228 	icmp6->icmp6_cksum = 0;
2229 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2230 					sizeof(struct ip6_hdr), plen);
2231 
2232 	/*
2233 	 * XXX option handling
2234 	 */
2235 
2236 	m->m_flags &= ~(M_BCAST|M_MCAST);
2237 
2238 	/*
2239 	 * To avoid a "too big" situation at an intermediate router
2240 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2241 	 * Note that only echo and node information replies are affected,
2242 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2243 	 */
2244 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif)
2245 	    != 0 && outif)
2246 		icmp6_ifstat_inc(outif, ifs6_out_error);
2247 	if (outif)
2248 		icmp6_ifoutstat_inc(outif, type, code);
2249 
2250 	return;
2251 
2252  bad:
2253 	m_freem(m);
2254 	return;
2255 }
2256 
2257 static const char *
2258 icmp6_redirect_diag(char *buf, size_t buflen, struct in6_addr *src6, struct in6_addr *dst6,
2259 	struct in6_addr *tgt6)
2260 {
2261 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
2262 	char ip6buft[INET6_ADDRSTRLEN];
2263 
2264 	snprintf(buf, buflen, "(src=%s dst=%s tgt=%s)",
2265 	    IN6_PRINT(ip6bufs, src6), IN6_PRINT(ip6bufd, dst6),
2266 	    IN6_PRINT(ip6buft, tgt6));
2267 	return buf;
2268 }
2269 
2270 void
2271 icmp6_redirect_input(struct mbuf *m, int off)
2272 {
2273 	struct ifnet *ifp;
2274 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2275 	struct nd_redirect *nd_rd;
2276 	int icmp6len = ntohs(ip6->ip6_plen);
2277 	char *lladdr = NULL;
2278 	int lladdrlen = 0;
2279 	struct rtentry *rt = NULL;
2280 	int is_router;
2281 	int is_onlink;
2282 	struct in6_addr src6 = ip6->ip6_src;
2283 	struct in6_addr redtgt6;
2284 	struct in6_addr reddst6;
2285 	union nd_opts ndopts;
2286 	struct psref psref;
2287 	char ip6buf[INET6_ADDRSTRLEN];
2288 	char diagbuf[256];
2289 
2290 	ifp = m_get_rcvif_psref(m, &psref);
2291 	if (ifp == NULL)
2292 		goto freeit;
2293 
2294 	/* XXX if we are router, we don't update route by icmp6 redirect */
2295 	if (ip6_forwarding)
2296 		goto freeit;
2297 	if (!icmp6_rediraccept)
2298 		goto freeit;
2299 
2300 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2301 	if (nd_rd == NULL) {
2302 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2303 		m_put_rcvif_psref(ifp, &psref);
2304 		return;
2305 	}
2306 	redtgt6 = nd_rd->nd_rd_target;
2307 	reddst6 = nd_rd->nd_rd_dst;
2308 
2309 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2310 	    in6_setscope(&reddst6, ifp, NULL)) {
2311 		goto freeit;
2312 	}
2313 
2314 	/* validation */
2315 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2316 		nd6log(LOG_ERR,
2317 		    "ICMP6 redirect sent from %s rejected; "
2318 		    "must be from linklocal\n", IN6_PRINT(ip6buf, &src6));
2319 		goto bad;
2320 	}
2321 	if (ip6->ip6_hlim != 255) {
2322 		nd6log(LOG_ERR,
2323 		    "ICMP6 redirect sent from %s rejected; "
2324 		    "hlim=%d (must be 255)\n",
2325 		    IN6_PRINT(ip6buf, &src6), ip6->ip6_hlim);
2326 		goto bad;
2327 	}
2328     {
2329 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2330 	struct sockaddr_in6 sin6;
2331 	struct in6_addr *gw6;
2332 
2333 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2334 	rt = rtalloc1(sin6tosa(&sin6), 0);
2335 	if (rt) {
2336 		if (rt->rt_gateway == NULL ||
2337 		    rt->rt_gateway->sa_family != AF_INET6) {
2338 			nd6log(LOG_ERR,
2339 			    "ICMP6 redirect rejected; no route "
2340 			    "with inet6 gateway found for redirect dst: %s\n",
2341 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2342 			    &src6, &reddst6, &redtgt6));
2343 			rt_unref(rt);
2344 			goto bad;
2345 		}
2346 
2347 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2348 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2349 			nd6log(LOG_ERR,
2350 			    "ICMP6 redirect rejected; "
2351 			    "not equal to gw-for-src=%s (must be same): %s\n",
2352 			    IN6_PRINT(ip6buf, gw6),
2353 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2354 			    &src6, &reddst6, &redtgt6));
2355 			rt_unref(rt);
2356 			goto bad;
2357 		}
2358 	} else {
2359 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2360 		    "no route found for redirect dst: %s\n",
2361 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2362 		    &src6, &reddst6, &redtgt6));
2363 		goto bad;
2364 	}
2365 	rt_unref(rt);
2366 	rt = NULL;
2367     }
2368 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2369 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2370 		    "redirect dst must be unicast: %s\n",
2371 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2372 		    &src6, &reddst6, &redtgt6));
2373 		goto bad;
2374 	}
2375 
2376 	is_router = is_onlink = 0;
2377 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2378 		is_router = 1;	/* router case */
2379 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2380 		is_onlink = 1;	/* on-link destination case */
2381 	if (!is_router && !is_onlink) {
2382 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2383 		    "neither router case nor onlink case: %s\n",
2384 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2385 		    &src6, &reddst6, &redtgt6));
2386 		goto bad;
2387 	}
2388 	/* validation passed */
2389 
2390 	icmp6len -= sizeof(*nd_rd);
2391 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2392 	if (nd6_options(&ndopts) < 0) {
2393 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2394 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2395 		    &src6, &reddst6, &redtgt6));
2396 		/* nd6_options have incremented stats */
2397 		goto freeit;
2398 	}
2399 
2400 	if (ndopts.nd_opts_tgt_lladdr) {
2401 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2402 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2403 	}
2404 
2405 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2406 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2407 		    "(if %d, icmp6 packet %d): %s\n",
2408 		    IN6_PRINT(ip6buf, &redtgt6),
2409 		    ifp->if_addrlen, lladdrlen - 2,
2410 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2411 		    &src6, &reddst6, &redtgt6));
2412 		goto bad;
2413 	}
2414 
2415 	/* RFC 2461 8.3 */
2416 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2417 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2418 
2419 	m_put_rcvif_psref(ifp, &psref);
2420 	ifp = NULL;
2421 
2422 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2423 		/* perform rtredirect */
2424 		struct sockaddr_in6 sdst;
2425 		struct sockaddr_in6 sgw;
2426 		struct sockaddr_in6 ssrc;
2427 		unsigned long rtcount;
2428 		struct rtentry *newrt = NULL;
2429 
2430 		/*
2431 		 * do not install redirect route, if the number of entries
2432 		 * is too much (> hiwat).  note that, the node (= host) will
2433 		 * work just fine even if we do not install redirect route
2434 		 * (there will be additional hops, though).
2435 		 */
2436 		mutex_enter(&icmp6_mtx);
2437 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2438 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) {
2439 			mutex_exit(&icmp6_mtx);
2440 			goto freeit;
2441 		}
2442 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) {
2443 			mutex_exit(&icmp6_mtx);
2444 			goto freeit;
2445 		} else if (0 <= icmp6_redirect_lowat &&
2446 		    rtcount > icmp6_redirect_lowat) {
2447 			/*
2448 			 * XXX nuke a victim, install the new one.
2449 			 */
2450 		}
2451 
2452 		memset(&sdst, 0, sizeof(sdst));
2453 		memset(&sgw, 0, sizeof(sgw));
2454 		memset(&ssrc, 0, sizeof(ssrc));
2455 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2456 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2457 			sizeof(struct sockaddr_in6);
2458 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2459 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2460 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2461 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2462 		           RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc),
2463 			   &newrt);
2464 
2465 		if (newrt) {
2466 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2467 			    icmp6_redirect_timeout_q);
2468 			rt_unref(newrt);
2469 		}
2470 		mutex_exit(&icmp6_mtx);
2471 	}
2472 	/* finally update cached route in each socket via pfctlinput */
2473 	{
2474 		struct sockaddr_in6 sdst;
2475 
2476 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2477 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2478 #if defined(IPSEC)
2479 		if (ipsec_used)
2480 			key_sa_routechange(sin6tosa(&sdst));
2481 #endif
2482 	}
2483 
2484  freeit:
2485 	if (ifp != NULL)
2486 		m_put_rcvif_psref(ifp, &psref);
2487 	m_freem(m);
2488 	return;
2489 
2490  bad:
2491 	m_put_rcvif_psref(ifp, &psref);
2492 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2493 	m_freem(m);
2494 }
2495 
2496 void
2497 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2498 {
2499 	struct ifnet *ifp;	/* my outgoing interface */
2500 	struct in6_addr *ifp_ll6;
2501 	struct in6_addr *nexthop;
2502 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2503 	struct mbuf *m = NULL;	/* newly allocated one */
2504 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2505 	struct nd_redirect *nd_rd;
2506 	size_t maxlen;
2507 	u_char *p;
2508 	struct sockaddr_in6 src_sa;
2509 
2510 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2511 
2512 	/* if we are not router, we don't send icmp6 redirect */
2513 	if (!ip6_forwarding)
2514 		goto fail;
2515 
2516 	/* sanity check */
2517 	KASSERT(m0 != NULL);
2518 	KASSERT(rt != NULL);
2519 
2520 	ifp = rt->rt_ifp;
2521 
2522 	/*
2523 	 * Address check:
2524 	 *  the source address must identify a neighbor, and
2525 	 *  the destination address must not be a multicast address
2526 	 *  [RFC 2461, sec 8.2]
2527 	 */
2528 	sip6 = mtod(m0, struct ip6_hdr *);
2529 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2530 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2531 		goto fail;
2532 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2533 		goto fail;	/* what should we do here? */
2534 
2535 	/* rate limit */
2536 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2537 		goto fail;
2538 
2539 	/*
2540 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2541 	 * we almost always ask for an mbuf cluster for simplicity.
2542 	 * (MHLEN < IPV6_MMTU is almost always true)
2543 	 */
2544 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2545 	if (m && IPV6_MMTU >= MHLEN) {
2546 #if IPV6_MMTU >= MCLBYTES
2547 		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2548 #else
2549 		MCLGET(m, M_DONTWAIT);
2550 #endif
2551 	}
2552 
2553 	if (!m)
2554 		goto fail;
2555 	m_reset_rcvif(m);
2556 	m->m_len = 0;
2557 	maxlen = M_TRAILINGSPACE(m);
2558 	maxlen = min(IPV6_MMTU, maxlen);
2559 	/* just for safety */
2560 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2561 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2562 		goto fail;
2563 	}
2564 
2565 	{
2566 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2567 		struct in6_ifaddr *ia;
2568 		int s = pserialize_read_enter();
2569 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2570 						 IN6_IFF_NOTREADY|
2571 						 IN6_IFF_ANYCAST)) == NULL) {
2572 			pserialize_read_exit(s);
2573 			goto fail;
2574 		}
2575 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2576 		pserialize_read_exit(s);
2577 	}
2578 
2579 	/* get ip6 linklocal address for the router. */
2580 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2581 		struct sockaddr_in6 *sin6;
2582 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2583 		nexthop = &sin6->sin6_addr;
2584 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2585 			nexthop = NULL;
2586 	} else
2587 		nexthop = NULL;
2588 
2589 	/* ip6 */
2590 	ip6 = mtod(m, struct ip6_hdr *);
2591 	ip6->ip6_flow = 0;
2592 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2593 	ip6->ip6_vfc |= IPV6_VERSION;
2594 	/* ip6->ip6_plen will be set later */
2595 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2596 	ip6->ip6_hlim = 255;
2597 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2598 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2599 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2600 
2601 	/* ND Redirect */
2602 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2603 	nd_rd->nd_rd_type = ND_REDIRECT;
2604 	nd_rd->nd_rd_code = 0;
2605 	nd_rd->nd_rd_reserved = 0;
2606 	if (rt->rt_flags & RTF_GATEWAY) {
2607 		/*
2608 		 * nd_rd->nd_rd_target must be a link-local address in
2609 		 * better router cases.
2610 		 */
2611 		if (!nexthop)
2612 			goto fail;
2613 		bcopy(nexthop, &nd_rd->nd_rd_target,
2614 		      sizeof(nd_rd->nd_rd_target));
2615 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2616 		      sizeof(nd_rd->nd_rd_dst));
2617 	} else {
2618 		/* make sure redtgt == reddst */
2619 		nexthop = &sip6->ip6_dst;
2620 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2621 		      sizeof(nd_rd->nd_rd_target));
2622 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2623 		      sizeof(nd_rd->nd_rd_dst));
2624 	}
2625 
2626 	p = (u_char *)(nd_rd + 1);
2627 
2628 	{
2629 		/* target lladdr option */
2630 		struct llentry *ln = NULL;
2631 		int len;
2632 		struct nd_opt_hdr *nd_opt;
2633 		char *lladdr;
2634 
2635 		ln = nd6_lookup(nexthop, ifp, false);
2636 		if (ln == NULL)
2637 			goto nolladdropt;
2638 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2639 		len = (len + 7) & ~7;	/* round by 8 */
2640 		/* safety check */
2641 		if (len + (p - (u_char *)ip6) > maxlen) {
2642 			LLE_RUNLOCK(ln);
2643 			goto nolladdropt;
2644 		}
2645 		if (ln->la_flags & LLE_VALID) {
2646 			nd_opt = (struct nd_opt_hdr *)p;
2647 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2648 			nd_opt->nd_opt_len = len >> 3;
2649 			lladdr = (char *)(nd_opt + 1);
2650 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2651 			p += len;
2652 		}
2653 		LLE_RUNLOCK(ln);
2654 	}
2655   nolladdropt:;
2656 
2657 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2658 
2659 	/* just to be safe */
2660 	if (m0->m_flags & M_DECRYPTED)
2661 		goto noredhdropt;
2662 	if (p - (u_char *)ip6 > maxlen)
2663 		goto noredhdropt;
2664 
2665 	{
2666 		/* redirected header option */
2667 		int len;
2668 		struct nd_opt_rd_hdr *nd_opt_rh;
2669 
2670 		/*
2671 		 * compute the maximum size for icmp6 redirect header option.
2672 		 * XXX room for auth header?
2673 		 */
2674 		len = maxlen - (p - (u_char *)ip6);
2675 		len &= ~7;
2676 
2677 		/*
2678 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2679 		 * about padding/truncate rule for the original IP packet.
2680 		 * From the discussion on IPv6imp in Feb 1999,
2681 		 * the consensus was:
2682 		 * - "attach as much as possible" is the goal
2683 		 * - pad if not aligned (original size can be guessed by
2684 		 *   original ip6 header)
2685 		 * Following code adds the padding if it is simple enough,
2686 		 * and truncates if not.
2687 		 */
2688 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2689 			/* not enough room, truncate */
2690 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2691 			    m0->m_pkthdr.len);
2692 		} else {
2693 			/*
2694 			 * enough room, truncate if not aligned.
2695 			 * we don't pad here for simplicity.
2696 			 */
2697 			size_t extra;
2698 
2699 			extra = m0->m_pkthdr.len % 8;
2700 			if (extra) {
2701 				/* truncate */
2702 				m_adj(m0, -extra);
2703 			}
2704 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2705 		}
2706 
2707 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2708 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2709 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2710 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2711 		p += sizeof(*nd_opt_rh);
2712 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2713 
2714 		/* connect m0 to m */
2715 		m->m_pkthdr.len += m0->m_pkthdr.len;
2716 		m_cat(m, m0);
2717 		m0 = NULL;
2718 	}
2719 noredhdropt:
2720 	if (m0) {
2721 		m_freem(m0);
2722 		m0 = NULL;
2723 	}
2724 
2725 	/* XXX: clear embedded link IDs in the inner header */
2726 	in6_clearscope(&sip6->ip6_src);
2727 	in6_clearscope(&sip6->ip6_dst);
2728 	in6_clearscope(&nd_rd->nd_rd_target);
2729 	in6_clearscope(&nd_rd->nd_rd_dst);
2730 
2731 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2732 
2733 	nd_rd->nd_rd_cksum = 0;
2734 	nd_rd->nd_rd_cksum
2735 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2736 
2737 	/* send the packet to outside... */
2738 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2739 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2740 
2741 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2742 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2743 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2744 
2745 	return;
2746 
2747 fail:
2748 	if (m)
2749 		m_freem(m);
2750 	if (m0)
2751 		m_freem(m0);
2752 }
2753 
2754 /*
2755  * ICMPv6 socket option processing.
2756  */
2757 int
2758 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2759 {
2760 	int error = 0;
2761 	struct in6pcb *in6p = sotoin6pcb(so);
2762 
2763 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2764 		return rip6_ctloutput(op, so, sopt);
2765 
2766 	switch (op) {
2767 	case PRCO_SETOPT:
2768 		switch (sopt->sopt_name) {
2769 		case ICMP6_FILTER:
2770 		    {
2771 			struct icmp6_filter fil;
2772 
2773 			error = sockopt_get(sopt, &fil, sizeof(fil));
2774 			if (error)
2775 				break;
2776 			memcpy(in6p->in6p_icmp6filt, &fil,
2777 			    sizeof(struct icmp6_filter));
2778 			error = 0;
2779 			break;
2780 		    }
2781 
2782 		default:
2783 			error = ENOPROTOOPT;
2784 			break;
2785 		}
2786 		break;
2787 
2788 	case PRCO_GETOPT:
2789 		switch (sopt->sopt_name) {
2790 		case ICMP6_FILTER:
2791 		    {
2792 			if (in6p->in6p_icmp6filt == NULL) {
2793 				error = EINVAL;
2794 				break;
2795 			}
2796 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2797 			    sizeof(struct icmp6_filter));
2798 			break;
2799 		    }
2800 
2801 		default:
2802 			error = ENOPROTOOPT;
2803 			break;
2804 		}
2805 		break;
2806 	}
2807 
2808 	return (error);
2809 }
2810 
2811 /*
2812  * Perform rate limit check.
2813  * Returns 0 if it is okay to send the icmp6 packet.
2814  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2815  * limitation.
2816  *
2817  * XXX per-destination/type check necessary?
2818  */
2819 static int
2820 icmp6_ratelimit(
2821 	const struct in6_addr *dst,	/* not used at this moment */
2822 	const int type,		/* not used at this moment */
2823 	const int code)		/* not used at this moment */
2824 {
2825 	int ret;
2826 
2827 	ret = 0;	/* okay to send */
2828 
2829 	/* PPS limit */
2830 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2831 	    icmp6errppslim)) {
2832 		/* The packet is subject to rate limit */
2833 		ret++;
2834 	}
2835 
2836 	return ret;
2837 }
2838 
2839 static struct rtentry *
2840 icmp6_mtudisc_clone(struct sockaddr *dst)
2841 {
2842 	struct rtentry *rt;
2843 	int    error;
2844 
2845 	rt = rtalloc1(dst, 1);
2846 	if (rt == 0)
2847 		return NULL;
2848 
2849 	/* If we didn't get a host route, allocate one */
2850 	if ((rt->rt_flags & RTF_HOST) == 0) {
2851 		struct rtentry *nrt;
2852 
2853 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2854 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2855 		if (error) {
2856 			rt_unref(rt);
2857 			return NULL;
2858 		}
2859 		nrt->rt_rmx = rt->rt_rmx;
2860 		rt_unref(rt);
2861 		rt = nrt;
2862 	}
2863 
2864 	mutex_enter(&icmp6_mtx);
2865 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2866 			icmp6_mtudisc_timeout_q);
2867 	mutex_exit(&icmp6_mtx);
2868 
2869 	if (error) {
2870 		rt_unref(rt);
2871 		return NULL;
2872 	}
2873 
2874 	return rt;	/* caller need to call rtfree() */
2875 }
2876 
2877 static void
2878 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2879 {
2880 
2881 	KASSERT(rt != NULL);
2882 	rt_assert_referenced(rt);
2883 
2884 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2885 	    (RTF_DYNAMIC | RTF_HOST)) {
2886 		rtrequest(RTM_DELETE, rt_getkey(rt),
2887 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2888 	} else {
2889 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2890 			rt->rt_rmx.rmx_mtu = 0;
2891 	}
2892 }
2893 
2894 static void
2895 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2896 {
2897 
2898 	KASSERT(rt != NULL);
2899 	rt_assert_referenced(rt);
2900 
2901 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2902 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2903 		rtrequest(RTM_DELETE, rt_getkey(rt),
2904 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2905 	}
2906 }
2907 
2908 /*
2909  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2910  */
2911 static int
2912 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2913 {
2914 	(void)&name;
2915 	(void)&l;
2916 	(void)&oname;
2917 
2918 	if (namelen != 0)
2919 		return (EINVAL);
2920 
2921 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2922 	    /*XXXUNCONST*/
2923 	    __UNCONST(newp), newlen));
2924 }
2925 
2926 static int
2927 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2928 {
2929 
2930 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2931 }
2932 
2933 static int
2934 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2935 {
2936 	int error, tmp;
2937 	struct sysctlnode node;
2938 
2939 	mutex_enter(&icmp6_mtx);
2940 
2941 	node = *rnode;
2942 	node.sysctl_data = &tmp;
2943 	tmp = icmp6_redirtimeout;
2944 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2945 	if (error || newp == NULL)
2946 		goto out;
2947 	if (tmp < 0) {
2948 		error = EINVAL;
2949 		goto out;
2950 	}
2951 	icmp6_redirtimeout = tmp;
2952 
2953 	if (icmp6_redirect_timeout_q != NULL) {
2954 		if (icmp6_redirtimeout == 0) {
2955 			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2956 		} else {
2957 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2958 			    icmp6_redirtimeout);
2959 		}
2960 	} else if (icmp6_redirtimeout > 0) {
2961 		icmp6_redirect_timeout_q =
2962 		    rt_timer_queue_create(icmp6_redirtimeout);
2963 	}
2964 	error = 0;
2965 out:
2966 	mutex_exit(&icmp6_mtx);
2967 	return error;
2968 }
2969 
2970 static void
2971 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2972 {
2973 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2974 
2975 	sysctl_createv(clog, 0, NULL, NULL,
2976 		       CTLFLAG_PERMANENT,
2977 		       CTLTYPE_NODE, "inet6", NULL,
2978 		       NULL, 0, NULL, 0,
2979 		       CTL_NET, PF_INET6, CTL_EOL);
2980 	sysctl_createv(clog, 0, NULL, NULL,
2981 		       CTLFLAG_PERMANENT,
2982 		       CTLTYPE_NODE, "icmp6",
2983 		       SYSCTL_DESCR("ICMPv6 related settings"),
2984 		       NULL, 0, NULL, 0,
2985 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2986 
2987 	sysctl_createv(clog, 0, NULL, NULL,
2988 		       CTLFLAG_PERMANENT,
2989 		       CTLTYPE_STRUCT, "stats",
2990 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2991 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2992 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2993 		       ICMPV6CTL_STATS, CTL_EOL);
2994 	sysctl_createv(clog, 0, NULL, NULL,
2995 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2996 		       CTLTYPE_INT, "rediraccept",
2997 		       SYSCTL_DESCR("Accept and process redirect messages"),
2998 		       NULL, 0, &icmp6_rediraccept, 0,
2999 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3000 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
3001 	sysctl_createv(clog, 0, NULL, NULL,
3002 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3003 		       CTLTYPE_INT, "redirtimeout",
3004 		       SYSCTL_DESCR("Redirect generated route lifetime"),
3005 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
3006 		       &icmp6_redirtimeout, 0,
3007 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3008 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
3009 #if 0 /* obsoleted */
3010 	sysctl_createv(clog, 0, NULL, NULL,
3011 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3012 		       CTLTYPE_INT, "errratelimit", NULL,
3013 		       NULL, 0, &icmp6_errratelimit, 0,
3014 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3015 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
3016 #endif
3017 	sysctl_createv(clog, 0, NULL, NULL,
3018 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3019 		       CTLTYPE_INT, "nd6_prune",
3020 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
3021 		       NULL, 0, &nd6_prune, 0,
3022 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3023 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
3024 	sysctl_createv(clog, 0, NULL, NULL,
3025 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3026 		       CTLTYPE_INT, "nd6_delay",
3027 		       SYSCTL_DESCR("First probe delay time"),
3028 		       NULL, 0, &nd6_delay, 0,
3029 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3030 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
3031 	sysctl_createv(clog, 0, NULL, NULL,
3032 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3033 		       CTLTYPE_INT, "nd6_umaxtries",
3034 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
3035 		       NULL, 0, &nd6_umaxtries, 0,
3036 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3037 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
3038 	sysctl_createv(clog, 0, NULL, NULL,
3039 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3040 		       CTLTYPE_INT, "nd6_mmaxtries",
3041 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
3042 		       NULL, 0, &nd6_mmaxtries, 0,
3043 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3044 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
3045 	sysctl_createv(clog, 0, NULL, NULL,
3046 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3047 		       CTLTYPE_INT, "nd6_useloopback",
3048 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
3049 		       NULL, 0, &nd6_useloopback, 0,
3050 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3051 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
3052 #if 0 /* obsoleted */
3053 	sysctl_createv(clog, 0, NULL, NULL,
3054 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3055 		       CTLTYPE_INT, "nd6_proxyall", NULL,
3056 		       NULL, 0, &nd6_proxyall, 0,
3057 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3058 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
3059 #endif
3060 	sysctl_createv(clog, 0, NULL, NULL,
3061 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3062 		       CTLTYPE_INT, "nodeinfo",
3063 		       SYSCTL_DESCR("Respond to node information requests"),
3064 		       NULL, 0, &icmp6_nodeinfo, 0,
3065 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3066 		       ICMPV6CTL_NODEINFO, CTL_EOL);
3067 	sysctl_createv(clog, 0, NULL, NULL,
3068 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3069 		       CTLTYPE_INT, "errppslimit",
3070 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
3071 		       NULL, 0, &icmp6errppslim, 0,
3072 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3073 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
3074 	sysctl_createv(clog, 0, NULL, NULL,
3075 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3076 		       CTLTYPE_INT, "nd6_maxnudhint",
3077 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
3078 		       NULL, 0, &nd6_maxnudhint, 0,
3079 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3080 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
3081 	sysctl_createv(clog, 0, NULL, NULL,
3082 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3083 		       CTLTYPE_INT, "mtudisc_hiwat",
3084 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3085 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
3086 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3087 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
3088 	sysctl_createv(clog, 0, NULL, NULL,
3089 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3090 		       CTLTYPE_INT, "mtudisc_lowat",
3091 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3092 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
3093 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3094 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
3095 	sysctl_createv(clog, 0, NULL, NULL,
3096 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3097 		       CTLTYPE_INT, "nd6_debug",
3098 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
3099 		       NULL, 0, &nd6_debug, 0,
3100 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3101 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
3102 	sysctl_createv(clog, 0, NULL, NULL,
3103 		       CTLFLAG_PERMANENT,
3104 		       CTLTYPE_STRUCT, "nd6_drlist",
3105 		       SYSCTL_DESCR("Default router list"),
3106 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3107 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3108 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
3109 	sysctl_createv(clog, 0, NULL, NULL,
3110 		       CTLFLAG_PERMANENT,
3111 		       CTLTYPE_STRUCT, "nd6_prlist",
3112 		       SYSCTL_DESCR("Prefix list"),
3113 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3114 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3115 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3116 	sysctl_createv(clog, 0, NULL, NULL,
3117 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3118 		       CTLTYPE_INT, "maxqueuelen",
3119 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3120 		       NULL, 1, &nd6_maxqueuelen, 0,
3121 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3122 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3123 }
3124 
3125 void
3126 icmp6_statinc(u_int stat)
3127 {
3128 
3129 	KASSERT(stat < ICMP6_NSTATS);
3130 	ICMP6_STATINC(stat);
3131 }
3132