xref: /netbsd-src/sys/netinet6/icmp6.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: icmp6.c,v 1.238 2018/06/01 07:13:35 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 1.238 2018/06/01 07:13:35 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/ipsec6.h>
108 #include <netipsec/key.h>
109 #endif
110 
111 #include "faith.h"
112 #if defined(NFAITH) && 0 < NFAITH
113 #include <net/if_faith.h>
114 #endif
115 
116 extern struct domain inet6domain;
117 
118 percpu_t *icmp6stat_percpu;
119 
120 extern struct inpcbtable raw6cbtable;
121 extern int icmp6errppslim;
122 static int icmp6errpps_count = 0;
123 static struct timeval icmp6errppslim_last;
124 extern int icmp6_nodeinfo;
125 
126 /*
127  * List of callbacks to notify when Path MTU changes are made.
128  */
129 struct icmp6_mtudisc_callback {
130 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
131 	void (*mc_func)(struct in6_addr *);
132 };
133 
134 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
135     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
136 
137 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
138 extern int pmtu_expire;
139 
140 /* XXX do these values make any sense? */
141 static int icmp6_mtudisc_hiwat = 1280;
142 static int icmp6_mtudisc_lowat = 256;
143 
144 /*
145  * keep track of # of redirect routes.
146  */
147 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
148 
149 /* XXX experimental, turned off */
150 static int icmp6_redirect_hiwat = -1;
151 static int icmp6_redirect_lowat = -1;
152 
153 /* Protect mtudisc and redirect stuffs */
154 static kmutex_t icmp6_mtx __cacheline_aligned;
155 
156 static void icmp6_errcount(u_int, int, int);
157 static int icmp6_rip6_input(struct mbuf **, int);
158 static void icmp6_reflect(struct mbuf *, size_t);
159 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
160 static const char *icmp6_redirect_diag(char *, size_t, struct in6_addr *,
161     struct in6_addr *, struct in6_addr *);
162 static void icmp6_redirect_input(struct mbuf *, int);
163 static struct mbuf *ni6_input(struct mbuf *, int);
164 static struct mbuf *ni6_nametodns(const char *, int, int);
165 static int ni6_dnsmatch(const char *, int, const char *, int);
166 static int ni6_addrs(struct icmp6_nodeinfo *, struct ifnet **, char *,
167     struct psref *);
168 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
169     struct ifnet *, int);
170 static int icmp6_notify_error(struct mbuf *, int, int, int);
171 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
172 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
173 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
174 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
175 
176 /* workqueue-based pr_input */
177 static struct wqinput *icmp6_wqinput;
178 static void _icmp6_input(struct mbuf *m, int off, int proto);
179 
180 void
181 icmp6_init(void)
182 {
183 
184 	sysctl_net_inet6_icmp6_setup(NULL);
185 	mld_init();
186 
187 	mutex_init(&icmp6_mtx, MUTEX_DEFAULT, IPL_NONE);
188 	mutex_enter(&icmp6_mtx);
189 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
190 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
191 	mutex_exit(&icmp6_mtx);
192 
193 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
194 
195 	icmp6_wqinput = wqinput_create("icmp6", _icmp6_input);
196 }
197 
198 static void
199 icmp6_errcount(u_int base, int type, int code)
200 {
201 	switch (type) {
202 	case ICMP6_DST_UNREACH:
203 		switch (code) {
204 		case ICMP6_DST_UNREACH_NOROUTE:
205 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
206 			return;
207 		case ICMP6_DST_UNREACH_ADMIN:
208 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
209 			return;
210 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
211 			ICMP6_STATINC(base +
212 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
213 			return;
214 		case ICMP6_DST_UNREACH_ADDR:
215 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
216 			return;
217 		case ICMP6_DST_UNREACH_NOPORT:
218 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
219 			return;
220 		}
221 		break;
222 	case ICMP6_PACKET_TOO_BIG:
223 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
224 		return;
225 	case ICMP6_TIME_EXCEEDED:
226 		switch (code) {
227 		case ICMP6_TIME_EXCEED_TRANSIT:
228 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
229 			return;
230 		case ICMP6_TIME_EXCEED_REASSEMBLY:
231 			ICMP6_STATINC(base +
232 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
233 			return;
234 		}
235 		break;
236 	case ICMP6_PARAM_PROB:
237 		switch (code) {
238 		case ICMP6_PARAMPROB_HEADER:
239 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
240 			return;
241 		case ICMP6_PARAMPROB_NEXTHEADER:
242 			ICMP6_STATINC(base +
243 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
244 			return;
245 		case ICMP6_PARAMPROB_OPTION:
246 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
247 			return;
248 		}
249 		break;
250 	case ND_REDIRECT:
251 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
252 		return;
253 	}
254 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
255 }
256 
257 /*
258  * Register a Path MTU Discovery callback.
259  */
260 void
261 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
262 {
263 	struct icmp6_mtudisc_callback *mc, *new;
264 
265 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
266 
267 	mutex_enter(&icmp6_mtx);
268 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
269 	     mc = LIST_NEXT(mc, mc_list)) {
270 		if (mc->mc_func == func) {
271 			mutex_exit(&icmp6_mtx);
272 			kmem_free(new, sizeof(*mc));
273 			return;
274 		}
275 	}
276 
277 	new->mc_func = func;
278 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, new, mc_list);
279 	mutex_exit(&icmp6_mtx);
280 }
281 
282 /*
283  * A wrapper function for icmp6_error() necessary when the erroneous packet
284  * may not contain enough scope zone information.
285  */
286 void
287 icmp6_error2(struct mbuf *m, int type, int code, int param,
288 	struct ifnet *ifp)
289 {
290 	struct ip6_hdr *ip6;
291 
292 	KASSERT(ifp != NULL);
293 
294 	if (m->m_len < sizeof(struct ip6_hdr)) {
295 		m = m_pullup(m, sizeof(struct ip6_hdr));
296 		if (m == NULL)
297 			return;
298 	}
299 
300 	ip6 = mtod(m, struct ip6_hdr *);
301 
302 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
303 		goto out;
304 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
305 		goto out;
306 
307 	icmp6_error(m, type, code, param);
308 	return;
309 
310 out:
311 	m_freem(m);
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 Packet 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 	} else {
395 		/* non-ICMPv6 - send the error */
396 	}
397 
398 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
399 
400 	/* Finally, do rate limitation check. */
401 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
402 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
403 		goto freeit;
404 	}
405 
406 	/*
407 	 * OK, ICMP6 can be generated.
408 	 */
409 
410 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
411 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
412 
413 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
414 	M_PREPEND(m, preplen, M_DONTWAIT);
415 	if (m && M_UNWRITABLE(m, preplen))
416 		m = m_pullup(m, preplen);
417 	if (m == NULL) {
418 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
419 		return;
420 	}
421 
422 	nip6 = mtod(m, struct ip6_hdr *);
423 	nip6->ip6_src  = oip6->ip6_src;
424 	nip6->ip6_dst  = oip6->ip6_dst;
425 
426 	in6_clearscope(&oip6->ip6_src);
427 	in6_clearscope(&oip6->ip6_dst);
428 
429 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
430 	icmp6->icmp6_type = type;
431 	icmp6->icmp6_code = code;
432 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
433 
434 	/*
435 	 * icmp6_reflect() is designed to be in the input path.
436 	 * icmp6_error() can be called from both input and output path,
437 	 * and if we are in output path rcvif could contain bogus value.
438 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
439 	 * information in ip header (nip6).
440 	 */
441 	m_reset_rcvif(m);
442 
443 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
444 
445 	/* header order: IPv6 - ICMPv6 */
446 	icmp6_reflect(m, sizeof(struct ip6_hdr));
447 
448 	return;
449 
450 freeit:
451 	/*
452 	 * If we can't tell whether or not we can generate ICMP6, free it.
453 	 */
454 	m_freem(m);
455 }
456 
457 /*
458  * Process a received ICMP6 message.
459  */
460 static void
461 _icmp6_input(struct mbuf *m, int off, int proto)
462 {
463 	struct mbuf *n;
464 	struct ip6_hdr *ip6, *nip6;
465 	struct icmp6_hdr *icmp6, *nicmp6;
466 	int icmp6len = m->m_pkthdr.len - off;
467 	int code, sum;
468 	struct ifnet *rcvif;
469 	struct psref psref;
470 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
471 
472 	rcvif = m_get_rcvif_psref(m, &psref);
473 	if (__predict_false(rcvif == NULL))
474 		goto freeit;
475 
476 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
477 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
478 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
479 
480 	/*
481 	 * Locate icmp6 structure in mbuf, and check
482 	 * that not corrupted and of at least minimum length
483 	 */
484 
485 	if (icmp6len < sizeof(struct icmp6_hdr)) {
486 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
487 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
488 		goto freeit;
489 	}
490 
491 	if (m->m_len < sizeof(struct ip6_hdr)) {
492 		m = m_pullup(m, sizeof(struct ip6_hdr));
493 		if (m == NULL) {
494 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
495 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
496 			goto freeit;
497 		}
498 	}
499 
500 	ip6 = mtod(m, struct ip6_hdr *);
501 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
502 	if (icmp6 == NULL) {
503 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
504 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
505 		goto freeit;
506 	}
507 
508 	/*
509 	 * Enforce alignment requirements that are violated in
510 	 * some cases, see kern/50766 for details.
511 	 */
512 	if (IP6_HDR_ALIGNED_P(icmp6) == 0) {
513 		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
514 		if (m == NULL) {
515 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
516 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
517 			goto freeit;
518 		}
519 		ip6 = mtod(m, struct ip6_hdr *);
520 		icmp6 = (struct icmp6_hdr *)(mtod(m, char *) + off);
521 	}
522 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
523 
524 	/*
525 	 * calculate the checksum
526 	 */
527 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
528 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
529 		    icmp6->icmp6_type, sum, IN6_PRINT(ip6buf, &ip6->ip6_src));
530 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
531 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
532 		goto freeit;
533 	}
534 
535 #if defined(NFAITH) && 0 < NFAITH
536 	if (faithprefix(&ip6->ip6_dst)) {
537 		/*
538 		 * Deliver very specific ICMP6 type only.
539 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
540 		 * will not work.
541 		 */
542 		switch (icmp6->icmp6_type) {
543 		case ICMP6_DST_UNREACH:
544 		case ICMP6_PACKET_TOO_BIG:
545 		case ICMP6_TIME_EXCEEDED:
546 			break;
547 		default:
548 			goto freeit;
549 		}
550 	}
551 #endif
552 
553 	code = icmp6->icmp6_code;
554 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
555 
556 	switch (icmp6->icmp6_type) {
557 	case ICMP6_DST_UNREACH:
558 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
559 		switch (code) {
560 		case ICMP6_DST_UNREACH_NOROUTE:
561 			code = PRC_UNREACH_NET;
562 			break;
563 		case ICMP6_DST_UNREACH_ADMIN:
564 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
565 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
566 			break;
567 		case ICMP6_DST_UNREACH_ADDR:
568 			code = PRC_HOSTDEAD;
569 			break;
570 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
571 			/* I mean "source address was incorrect." */
572 			code = PRC_UNREACH_NET;
573 			break;
574 		case ICMP6_DST_UNREACH_NOPORT:
575 			code = PRC_UNREACH_PORT;
576 			break;
577 		default:
578 			goto badcode;
579 		}
580 		goto deliver;
581 
582 	case ICMP6_PACKET_TOO_BIG:
583 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
584 
585 		/*
586 		 * MTU is checked in icmp6_mtudisc.
587 		 */
588 		code = PRC_MSGSIZE;
589 
590 		/*
591 		 * Updating the path MTU will be done after examining
592 		 * intermediate extension headers.
593 		 */
594 		goto deliver;
595 
596 	case ICMP6_TIME_EXCEEDED:
597 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
598 		switch (code) {
599 		case ICMP6_TIME_EXCEED_TRANSIT:
600 			code = PRC_TIMXCEED_INTRANS;
601 			break;
602 		case ICMP6_TIME_EXCEED_REASSEMBLY:
603 			code = PRC_TIMXCEED_REASS;
604 			break;
605 		default:
606 			goto badcode;
607 		}
608 		goto deliver;
609 
610 	case ICMP6_PARAM_PROB:
611 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
612 		switch (code) {
613 		case ICMP6_PARAMPROB_NEXTHEADER:
614 			code = PRC_UNREACH_PROTOCOL;
615 			break;
616 		case ICMP6_PARAMPROB_HEADER:
617 		case ICMP6_PARAMPROB_OPTION:
618 			code = PRC_PARAMPROB;
619 			break;
620 		default:
621 			goto badcode;
622 		}
623 		goto deliver;
624 
625 	case ICMP6_ECHO_REQUEST:
626 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
627 		if (code != 0)
628 			goto badcode;
629 		/*
630 		 * Copy mbuf to send to two data paths: userland socket(s),
631 		 * and to the querier (echo reply).
632 		 * m: a copy for socket, n: a copy for querier
633 		 *
634 		 * If the first mbuf is shared, or the first mbuf is too short,
635 		 * copy the first part of the data into a fresh mbuf.
636 		 * Otherwise, we will wrongly overwrite both copies.
637 		 */
638 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
639 			/* Give up local */
640 			n = m;
641 			m = NULL;
642 		} else if (M_UNWRITABLE(n, off + sizeof(struct icmp6_hdr))) {
643 			struct mbuf *n0 = n;
644 
645 			/*
646 			 * Prepare an internal mbuf.  m_pullup() doesn't
647 			 * always copy the length we specified.
648 			 */
649 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
650 				/* Give up local */
651 				n = m;
652 				m = NULL;
653 			}
654 			m_freem(n0);
655 		}
656 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
657 		    sizeof(*nicmp6));
658 		if (nicmp6 == NULL)
659 			goto freeit;
660 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
661 		nicmp6->icmp6_code = 0;
662 		if (n) {
663 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
664 			icmp6s[ICMP6_STAT_REFLECT]++;
665 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
666 			ICMP6_STAT_PUTREF();
667 			icmp6_reflect(n, off);
668 		}
669 		if (!m)
670 			goto freeit;
671 		break;
672 
673 	case ICMP6_ECHO_REPLY:
674 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
675 		if (code != 0)
676 			goto badcode;
677 		break;
678 
679 	case MLD_LISTENER_QUERY:
680 	case MLD_LISTENER_REPORT:
681 		if (icmp6len < sizeof(struct mld_hdr))
682 			goto badlen;
683 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
684 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
685 		else
686 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
687 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
688 			/* give up local */
689 			mld_input(m, off);
690 			m = NULL;
691 			goto freeit;
692 		}
693 		mld_input(n, off);
694 		/* m stays. */
695 		break;
696 
697 	case MLD_LISTENER_DONE:
698 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
699 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
700 			goto badlen;
701 		break;		/* nothing to be done in kernel */
702 
703 	case MLD_MTRACE_RESP:
704 	case MLD_MTRACE:
705 		/* XXX: these two are experimental.  not officially defined. */
706 		/* XXX: per-interface statistics? */
707 		break;		/* just pass it to applications */
708 
709 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
710 	    {
711 		enum { WRU, FQDN } mode;
712 
713 		if (!icmp6_nodeinfo)
714 			break;
715 
716 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
717 			mode = WRU;
718 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
719 			mode = FQDN;
720 		else
721 			goto badlen;
722 
723 		if (mode == FQDN) {
724 			n = m_copypacket(m, M_DONTWAIT);
725 			if (n)
726 				n = ni6_input(n, off);
727 		} else {
728 			u_char *p;
729 			int maxhlen;
730 
731 			if ((icmp6_nodeinfo & 5) != 5)
732 				break;
733 
734 			if (code != 0)
735 				goto badcode;
736 			MGETHDR(n, M_DONTWAIT, m->m_type);
737 			if (n && ICMP6_MAXLEN > MHLEN) {
738 				MCLGET(n, M_DONTWAIT);
739 				if ((n->m_flags & M_EXT) == 0) {
740 					m_free(n);
741 					n = NULL;
742 				}
743 			}
744 			if (n == NULL) {
745 				/* Give up remote */
746 				break;
747 			}
748 			m_reset_rcvif(n);
749 			n->m_len = 0;
750 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
751 			if (maxhlen < 0) {
752 				m_free(n);
753 				break;
754 			}
755 			if (maxhlen > hostnamelen)
756 				maxhlen = hostnamelen;
757 			/*
758 			 * Copy IPv6 and ICMPv6 only.
759 			 */
760 			nip6 = mtod(n, struct ip6_hdr *);
761 			memcpy(nip6, ip6, sizeof(struct ip6_hdr));
762 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
763 			memcpy(nicmp6, icmp6, sizeof(struct icmp6_hdr));
764 
765 			p = (u_char *)(nicmp6 + 1);
766 			memset(p, 0, 4);
767 			memcpy(p + 4, hostname, maxhlen); /* meaningless TTL */
768 
769 			M_COPY_PKTHDR(n, m);
770 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
771 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
772 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
773 			nicmp6->icmp6_code = 0;
774 		}
775 		if (n) {
776 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
777 			icmp6s[ICMP6_STAT_REFLECT]++;
778 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
779 			ICMP6_STAT_PUTREF();
780 			icmp6_reflect(n, sizeof(struct ip6_hdr));
781 		}
782 		break;
783 	    }
784 
785 	case ICMP6_WRUREPLY:
786 		if (code != 0)
787 			goto badcode;
788 		break;
789 
790 	case ND_ROUTER_SOLICIT:
791 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
792 		if (code != 0)
793 			goto badcode;
794 		if (icmp6len < sizeof(struct nd_router_solicit))
795 			goto badlen;
796 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
797 			/* give up local */
798 			nd6_rs_input(m, off, icmp6len);
799 			m = NULL;
800 			goto freeit;
801 		}
802 		nd6_rs_input(n, off, icmp6len);
803 		/* m stays. */
804 		break;
805 
806 	case ND_ROUTER_ADVERT:
807 		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
808 		if (code != 0)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct nd_router_advert))
811 			goto badlen;
812 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
813 			/* give up local */
814 			nd6_ra_input(m, off, icmp6len);
815 			m = NULL;
816 			goto freeit;
817 		}
818 		nd6_ra_input(n, off, icmp6len);
819 		/* m stays. */
820 		break;
821 
822 	case ND_NEIGHBOR_SOLICIT:
823 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
824 		if (code != 0)
825 			goto badcode;
826 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
827 			goto badlen;
828 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
829 			/* give up local */
830 			nd6_ns_input(m, off, icmp6len);
831 			m = NULL;
832 			goto freeit;
833 		}
834 		nd6_ns_input(n, off, icmp6len);
835 		/* m stays. */
836 		break;
837 
838 	case ND_NEIGHBOR_ADVERT:
839 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
840 		if (code != 0)
841 			goto badcode;
842 		if (icmp6len < sizeof(struct nd_neighbor_advert))
843 			goto badlen;
844 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
845 			/* give up local */
846 			nd6_na_input(m, off, icmp6len);
847 			m = NULL;
848 			goto freeit;
849 		}
850 		nd6_na_input(n, off, icmp6len);
851 		/* m stays. */
852 		break;
853 
854 	case ND_REDIRECT:
855 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
856 		if (code != 0)
857 			goto badcode;
858 		if (icmp6len < sizeof(struct nd_redirect))
859 			goto badlen;
860 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
861 			/* give up local */
862 			icmp6_redirect_input(m, off);
863 			m = NULL;
864 			goto freeit;
865 		}
866 		icmp6_redirect_input(n, off);
867 		/* m stays. */
868 		break;
869 
870 	case ICMP6_ROUTER_RENUMBERING:
871 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
872 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
873 			goto badcode;
874 		if (icmp6len < sizeof(struct icmp6_router_renum))
875 			goto badlen;
876 		break;
877 
878 	default:
879 		nd6log(LOG_DEBUG,
880 		    "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
881 		    icmp6->icmp6_type,
882 		    IN6_PRINT(ip6buf, &ip6->ip6_src),
883 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst),
884 		    rcvif ? rcvif->if_index : 0);
885 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
886 			/* ICMPv6 error: MUST deliver it by spec... */
887 			code = PRC_NCMDS;
888 			/* deliver */
889 		} else {
890 			/* ICMPv6 informational: MUST not deliver */
891 			break;
892 		}
893 	deliver:
894 		if (icmp6_notify_error(m, off, icmp6len, code)) {
895 			/* In this case, m should've been freed. */
896 			m_put_rcvif_psref(rcvif, &psref);
897 			return;
898 		}
899 		break;
900 
901 	badcode:
902 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
903 		break;
904 
905 	badlen:
906 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
907 		break;
908 	}
909 	m_put_rcvif_psref(rcvif, &psref);
910 
911 	/* deliver the packet to appropriate sockets */
912 	icmp6_rip6_input(&m, off);
913 
914 	return;
915 
916 freeit:
917 	m_put_rcvif_psref(rcvif, &psref);
918 	m_freem(m);
919 	return;
920 }
921 
922 int
923 icmp6_input(struct mbuf **mp, int *offp, int proto)
924 {
925 
926 	wqinput_input(icmp6_wqinput, *mp, *offp, proto);
927 
928 	return IPPROTO_DONE;
929 }
930 
931 static int
932 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
933 {
934 	struct icmp6_hdr *icmp6;
935 	struct ip6_hdr *eip6;
936 	u_int32_t notifymtu;
937 	struct sockaddr_in6 icmp6src, icmp6dst;
938 
939 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
940 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
941 		goto freeit;
942 	}
943 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
944 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
945 	if (icmp6 == NULL) {
946 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
947 		return (-1);
948 	}
949 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
950 
951 	/* Detect the upper level protocol */
952 	{
953 		void (*ctlfunc)(int, struct sockaddr *, void *);
954 		u_int8_t nxt = eip6->ip6_nxt;
955 		int eoff = off + sizeof(struct icmp6_hdr) +
956 			sizeof(struct ip6_hdr);
957 		struct ip6ctlparam ip6cp;
958 		struct in6_addr *finaldst = NULL;
959 		int icmp6type = icmp6->icmp6_type;
960 		struct ip6_frag *fh;
961 		struct ip6_rthdr *rth;
962 		struct ifnet *rcvif;
963 		int s;
964 
965 		while (1) { /* XXX: should avoid infinite loop explicitly? */
966 			struct ip6_ext *eh;
967 
968 			switch (nxt) {
969 			case IPPROTO_HOPOPTS:
970 			case IPPROTO_DSTOPTS:
971 			case IPPROTO_AH:
972 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
973 				    eoff, sizeof(*eh));
974 				if (eh == NULL) {
975 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
976 					return (-1);
977 				}
978 
979 				if (nxt == IPPROTO_AH)
980 					eoff += (eh->ip6e_len + 2) << 2;
981 				else
982 					eoff += (eh->ip6e_len + 1) << 3;
983 				nxt = eh->ip6e_nxt;
984 				break;
985 			case IPPROTO_ROUTING:
986 				/* Ignore the option. */
987 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
988 				    eoff, sizeof(*rth));
989 				if (rth == NULL) {
990 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
991 					return (-1);
992 				}
993 
994 				eoff += (rth->ip6r_len + 1) << 3;
995 				nxt = rth->ip6r_nxt;
996 				break;
997 			case IPPROTO_FRAGMENT:
998 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
999 				    eoff, sizeof(*fh));
1000 				if (fh == NULL) {
1001 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1002 					return (-1);
1003 				}
1004 				/*
1005 				 * Data after a fragment header is meaningless
1006 				 * unless it is the first fragment, but
1007 				 * we'll go to the notify label for path MTU
1008 				 * discovery.
1009 				 */
1010 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1011 					goto notify;
1012 
1013 				eoff += sizeof(struct ip6_frag);
1014 				nxt = fh->ip6f_nxt;
1015 				break;
1016 			default:
1017 				/*
1018 				 * This case includes ESP and the No Next
1019 				 * Header.  In such cases going to the notify
1020 				 * label does not have any meaning
1021 				 * (i.e. ctlfunc will be NULL), but we go
1022 				 * anyway since we might have to update
1023 				 * path MTU information.
1024 				 */
1025 				goto notify;
1026 			}
1027 		}
1028 	  notify:
1029 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1030 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1031 		if (icmp6 == NULL) {
1032 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1033 			return (-1);
1034 		}
1035 
1036 		/*
1037 		 * retrieve parameters from the inner IPv6 header, and convert
1038 		 * them into sockaddr structures.
1039 		 * XXX: there is no guarantee that the source or destination
1040 		 * addresses of the inner packet are in the same scope zone as
1041 		 * the addresses of the icmp packet.  But there is no other
1042 		 * way to determine the zone.
1043 		 */
1044 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1045 
1046 		rcvif = m_get_rcvif(m, &s);
1047 		if (__predict_false(rcvif == NULL))
1048 			goto freeit;
1049 		sockaddr_in6_init(&icmp6dst,
1050 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1051 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1052 			m_put_rcvif(rcvif, &s);
1053 			goto freeit;
1054 		}
1055 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1056 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1057 			m_put_rcvif(rcvif, &s);
1058 			goto freeit;
1059 		}
1060 		m_put_rcvif(rcvif, &s);
1061 
1062 		icmp6src.sin6_flowinfo =
1063 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1064 
1065 		if (finaldst == NULL)
1066 			finaldst = &eip6->ip6_dst;
1067 		ip6cp.ip6c_m = m;
1068 		ip6cp.ip6c_icmp6 = icmp6;
1069 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1070 		ip6cp.ip6c_off = eoff;
1071 		ip6cp.ip6c_finaldst = finaldst;
1072 		ip6cp.ip6c_src = &icmp6src;
1073 		ip6cp.ip6c_nxt = nxt;
1074 
1075 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1076 			notifymtu = ntohl(icmp6->icmp6_mtu);
1077 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1078 		}
1079 
1080 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1081 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1082 		if (ctlfunc) {
1083 			(void)(*ctlfunc)(code, sin6tosa(&icmp6dst), &ip6cp);
1084 		}
1085 	}
1086 	return (0);
1087 
1088 freeit:
1089 	m_freem(m);
1090 	return (-1);
1091 }
1092 
1093 void
1094 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1095 {
1096 	unsigned long rtcount;
1097 	struct icmp6_mtudisc_callback *mc;
1098 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1099 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1100 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1101 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1102 	struct rtentry *rt = NULL;
1103 	struct sockaddr_in6 sin6;
1104 	struct ifnet *rcvif;
1105 	int s;
1106 
1107 	/*
1108 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1109 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1110 	 * In that one case, the MTU might be less than 1280.
1111 	 */
1112 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1113 		/* is the mtu even sane? */
1114 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1115 			return;
1116 		if (!validated)
1117 			return;
1118 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1119 	}
1120 
1121 	/*
1122 	 * allow non-validated cases if memory is plenty, to make traffic
1123 	 * from non-connected pcb happy.
1124 	 */
1125 	mutex_enter(&icmp6_mtx);
1126 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1127 	if (validated) {
1128 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) {
1129 			mutex_exit(&icmp6_mtx);
1130 			return;
1131 		} else if (0 <= icmp6_mtudisc_lowat &&
1132 		    rtcount > icmp6_mtudisc_lowat) {
1133 			/*
1134 			 * XXX nuke a victim, install the new one.
1135 			 */
1136 		}
1137 	} else {
1138 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) {
1139 			mutex_exit(&icmp6_mtx);
1140 			return;
1141 		}
1142 	}
1143 	mutex_exit(&icmp6_mtx);
1144 
1145 	memset(&sin6, 0, sizeof(sin6));
1146 	sin6.sin6_family = PF_INET6;
1147 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1148 	sin6.sin6_addr = *dst;
1149 	rcvif = m_get_rcvif(m, &s);
1150 	if (__predict_false(rcvif == NULL))
1151 		return;
1152 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1153 		m_put_rcvif(rcvif, &s);
1154 		return;
1155 	}
1156 	m_put_rcvif(rcvif, &s);
1157 
1158 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1159 
1160 	if (rt && (rt->rt_flags & RTF_HOST) &&
1161 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1162 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1163 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1164 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1165 			rt->rt_rmx.rmx_mtu = mtu;
1166 		}
1167 	}
1168 	if (rt) {
1169 		rt_unref(rt);
1170 	}
1171 
1172 	/*
1173 	 * Notify protocols that the MTU for this destination
1174 	 * has changed.
1175 	 */
1176 	mutex_enter(&icmp6_mtx);
1177 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1178 	     mc = LIST_NEXT(mc, mc_list))
1179 		(*mc->mc_func)(&sin6.sin6_addr);
1180 	mutex_exit(&icmp6_mtx);
1181 }
1182 
1183 /*
1184  * Process a Node Information Query packet, based on
1185  * draft-ietf-ipngwg-icmp-name-lookups-07.
1186  *
1187  * Spec incompatibilities:
1188  * - IPv6 Subject address handling
1189  * - IPv4 Subject address handling support missing
1190  * - Proxy reply (answer even if it's not for me)
1191  * - joins NI group address at in6_ifattach() time only, does not cope
1192  *   with hostname changes by sethostname(3)
1193  */
1194 static struct mbuf *
1195 ni6_input(struct mbuf *m, int off)
1196 {
1197 	struct icmp6_nodeinfo *ni6, *nni6;
1198 	struct mbuf *n = NULL;
1199 	u_int16_t qtype;
1200 	int subjlen;
1201 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1202 	struct ni_reply_fqdn *fqdn;
1203 	int addrs;		/* for NI_QTYPE_NODEADDR */
1204 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1205 	struct sockaddr_in6 sin6; /* ip6_dst */
1206 	struct in6_addr in6_subj; /* subject address */
1207 	struct ip6_hdr *ip6;
1208 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1209 	char *subj = NULL;
1210 	struct ifnet *rcvif;
1211 	int s, ss;
1212 	struct ifaddr *ifa;
1213 	struct psref psref;
1214 
1215 	ip6 = mtod(m, struct ip6_hdr *);
1216 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1217 	if (ni6 == NULL) {
1218 		/* m is already reclaimed */
1219 		return NULL;
1220 	}
1221 	KASSERT((m->m_flags & M_PKTHDR) != 0);
1222 
1223 	/*
1224 	 * Validate IPv6 destination address.
1225 	 *
1226 	 * The Responder must discard the Query without further processing
1227 	 * unless it is one of the Responder's unicast or anycast addresses, or
1228 	 * a link-local scope multicast address which the Responder has joined.
1229 	 * [icmp-name-lookups-07, Section 4.]
1230 	 */
1231 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1232 	/* XXX scopeid */
1233 	ss = pserialize_read_enter();
1234 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1235 	if (ifa != NULL) {
1236 		; /* unicast/anycast, fine */
1237 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) {
1238 		; /* link-local multicast, fine */
1239 	} else {
1240 		pserialize_read_exit(ss);
1241 		goto bad;
1242 	}
1243 	pserialize_read_exit(ss);
1244 
1245 	/* validate query Subject field. */
1246 	qtype = ntohs(ni6->ni_qtype);
1247 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1248 	switch (qtype) {
1249 	case NI_QTYPE_NOOP:
1250 	case NI_QTYPE_SUPTYPES:
1251 		/* 07 draft */
1252 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1253 			break;
1254 		/* FALLTHROUGH */
1255 	case NI_QTYPE_FQDN:
1256 	case NI_QTYPE_NODEADDR:
1257 	case NI_QTYPE_IPV4ADDR:
1258 		switch (ni6->ni_code) {
1259 		case ICMP6_NI_SUBJ_IPV6:
1260 #if ICMP6_NI_SUBJ_IPV6 != 0
1261 		case 0:
1262 #endif
1263 			/*
1264 			 * backward compatibility - try to accept 03 draft
1265 			 * format, where no Subject is present.
1266 			 */
1267 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1268 			    subjlen == 0) {
1269 				oldfqdn++;
1270 				break;
1271 			}
1272 #if ICMP6_NI_SUBJ_IPV6 != 0
1273 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1274 				goto bad;
1275 #endif
1276 
1277 			if (subjlen != sizeof(sin6.sin6_addr))
1278 				goto bad;
1279 
1280 			/*
1281 			 * Validate Subject address.
1282 			 *
1283 			 * Not sure what exactly "address belongs to the node"
1284 			 * means in the spec, is it just unicast, or what?
1285 			 *
1286 			 * At this moment we consider Subject address as
1287 			 * "belong to the node" if the Subject address equals
1288 			 * to the IPv6 destination address; validation for
1289 			 * IPv6 destination address should have done enough
1290 			 * check for us.
1291 			 *
1292 			 * We do not do proxy at this moment.
1293 			 */
1294 			/* m_pulldown instead of copy? */
1295 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1296 			    subjlen, (void *)&in6_subj);
1297 			rcvif = m_get_rcvif(m, &s);
1298 			if (__predict_false(rcvif == NULL))
1299 				goto bad;
1300 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1301 				m_put_rcvif(rcvif, &s);
1302 				goto bad;
1303 			}
1304 			m_put_rcvif(rcvif, &s);
1305 
1306 			subj = (char *)&in6_subj;
1307 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1308 				break;
1309 
1310 			/*
1311 			 * XXX if we are to allow other cases, we should really
1312 			 * be careful about scope here.
1313 			 * basically, we should disallow queries toward IPv6
1314 			 * destination X with subject Y, if scope(X) > scope(Y).
1315 			 * if we allow scope(X) > scope(Y), it will result in
1316 			 * information leakage across scope boundary.
1317 			 */
1318 			goto bad;
1319 
1320 		case ICMP6_NI_SUBJ_FQDN:
1321 			/*
1322 			 * Validate Subject name with gethostname(3).
1323 			 *
1324 			 * The behavior may need some debate, since:
1325 			 * - we are not sure if the node has FQDN as
1326 			 *   hostname (returned by gethostname(3)).
1327 			 * - the code does wildcard match for truncated names.
1328 			 *   however, we are not sure if we want to perform
1329 			 *   wildcard match, if gethostname(3) side has
1330 			 *   truncated hostname.
1331 			 */
1332 			n = ni6_nametodns(hostname, hostnamelen, 0);
1333 			if (!n || n->m_next || n->m_len == 0)
1334 				goto bad;
1335 			IP6_EXTHDR_GET(subj, char *, m,
1336 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1337 			if (subj == NULL)
1338 				goto bad;
1339 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1340 			    n->m_len)) {
1341 				goto bad;
1342 			}
1343 			m_freem(n);
1344 			n = NULL;
1345 			break;
1346 
1347 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1348 		default:
1349 			goto bad;
1350 		}
1351 		break;
1352 	}
1353 
1354 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1355 	switch (qtype) {
1356 	case NI_QTYPE_FQDN:
1357 		if ((icmp6_nodeinfo & 1) == 0)
1358 			goto bad;
1359 		break;
1360 	case NI_QTYPE_NODEADDR:
1361 	case NI_QTYPE_IPV4ADDR:
1362 		if ((icmp6_nodeinfo & 2) == 0)
1363 			goto bad;
1364 		break;
1365 	}
1366 
1367 	/* guess reply length */
1368 	switch (qtype) {
1369 	case NI_QTYPE_NOOP:
1370 		break;		/* no reply data */
1371 	case NI_QTYPE_SUPTYPES:
1372 		replylen += sizeof(u_int32_t);
1373 		break;
1374 	case NI_QTYPE_FQDN:
1375 		/* will append an mbuf */
1376 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1377 		break;
1378 	case NI_QTYPE_NODEADDR:
1379 		addrs = ni6_addrs(ni6, &ifp, subj, &psref);
1380 		replylen += addrs *
1381 		    (sizeof(struct in6_addr) + sizeof(u_int32_t));
1382 		if (replylen > MCLBYTES)
1383 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1384 		break;
1385 	case NI_QTYPE_IPV4ADDR:
1386 		/* unsupported - should respond with unknown Qtype? */
1387 		goto bad;
1388 	default:
1389 		/*
1390 		 * XXX: We must return a reply with the ICMP6 code
1391 		 * `unknown Qtype' in this case.  However we regard the case
1392 		 * as an FQDN query for backward compatibility.
1393 		 * Older versions set a random value to this field,
1394 		 * so it rarely varies in the defined qtypes.
1395 		 * But the mechanism is not reliable...
1396 		 * maybe we should obsolete older versions.
1397 		 */
1398 		qtype = NI_QTYPE_FQDN;
1399 		/* will append an mbuf */
1400 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1401 		oldfqdn++;
1402 		break;
1403 	}
1404 
1405 	/* allocate an mbuf to reply. */
1406 	MGETHDR(n, M_DONTWAIT, m->m_type);
1407 	if (n == NULL) {
1408 		goto bad;
1409 	}
1410 	M_MOVE_PKTHDR(n, m);
1411 	if (replylen > MHLEN) {
1412 		if (replylen > MCLBYTES) {
1413 			/*
1414 			 * XXX: should we try to allocate more? But MCLBYTES
1415 			 * is probably much larger than IPV6_MMTU...
1416 			 */
1417 			goto bad;
1418 		}
1419 		MCLGET(n, M_DONTWAIT);
1420 		if ((n->m_flags & M_EXT) == 0) {
1421 			goto bad;
1422 		}
1423 	}
1424 	n->m_pkthdr.len = n->m_len = replylen;
1425 
1426 	/* copy mbuf header and IPv6 + Node Information base headers */
1427 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1428 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1429 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1430 
1431 	/* qtype dependent procedure */
1432 	switch (qtype) {
1433 	case NI_QTYPE_NOOP:
1434 		nni6->ni_code = ICMP6_NI_SUCCESS;
1435 		nni6->ni_flags = 0;
1436 		break;
1437 	case NI_QTYPE_SUPTYPES:
1438 	{
1439 		u_int32_t v;
1440 		nni6->ni_code = ICMP6_NI_SUCCESS;
1441 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1442 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1443 		v = (u_int32_t)htonl(0x0000000f);
1444 		memcpy(nni6 + 1, &v, sizeof(u_int32_t));
1445 		break;
1446 	}
1447 	case NI_QTYPE_FQDN:
1448 		nni6->ni_code = ICMP6_NI_SUCCESS;
1449 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1450 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1451 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1452 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1453 		/*
1454 		 * XXX do we really have FQDN in variable "hostname"?
1455 		 */
1456 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1457 		if (n->m_next == NULL)
1458 			goto bad;
1459 		/* XXX we assume that n->m_next is not a chain */
1460 		if (n->m_next->m_next != NULL)
1461 			goto bad;
1462 		n->m_pkthdr.len += n->m_next->m_len;
1463 		break;
1464 	case NI_QTYPE_NODEADDR:
1465 	{
1466 		int lenlim, copied;
1467 
1468 		nni6->ni_code = ICMP6_NI_SUCCESS;
1469 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1470 		    sizeof(struct icmp6_nodeinfo);
1471 		lenlim = M_TRAILINGSPACE(n);
1472 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1473 		if_put(ifp, &psref);
1474 		ifp = NULL;
1475 		/* update mbuf length */
1476 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1477 		    sizeof(struct icmp6_nodeinfo) + copied;
1478 		break;
1479 	}
1480 	default:
1481 		panic("%s: impossible", __func__);
1482 		break;
1483 	}
1484 
1485 	nni6->ni_type = ICMP6_NI_REPLY;
1486 	m_freem(m);
1487 	return n;
1488 
1489 bad:
1490 	if_put(ifp, &psref);
1491 	m_freem(m);
1492 	if (n)
1493 		m_freem(n);
1494 	return NULL;
1495 }
1496 
1497 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1498 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1499 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1500 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1501 
1502 /*
1503  * make a mbuf with DNS-encoded string.  no compression support.
1504  *
1505  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1506  * treated as truncated name (two \0 at the end).  this is a wild guess.
1507  *
1508  * old - return pascal string if non-zero
1509  */
1510 static struct mbuf *
1511 ni6_nametodns(const char *name, int namelen, int old)
1512 {
1513 	struct mbuf *m;
1514 	char *cp, *ep;
1515 	const char *p, *q;
1516 	int i, len, nterm;
1517 
1518 	if (old)
1519 		len = namelen + 1;
1520 	else
1521 		len = MCLBYTES;
1522 
1523 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1524 	MGET(m, M_DONTWAIT, MT_DATA);
1525 	if (m && len > MLEN) {
1526 		MCLGET(m, M_DONTWAIT);
1527 		if ((m->m_flags & M_EXT) == 0)
1528 			goto fail;
1529 	}
1530 	if (!m)
1531 		goto fail;
1532 	m->m_next = NULL;
1533 
1534 	if (old) {
1535 		m->m_len = len;
1536 		*mtod(m, char *) = namelen;
1537 		memcpy(mtod(m, char *) + 1, name, namelen);
1538 		return m;
1539 	} else {
1540 		m->m_len = 0;
1541 		cp = mtod(m, char *);
1542 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1543 
1544 		/* if not certain about my name, return empty buffer */
1545 		if (namelen == 0)
1546 			return m;
1547 
1548 		/*
1549 		 * guess if it looks like shortened hostname, or FQDN.
1550 		 * shortened hostname needs two trailing "\0".
1551 		 */
1552 		i = 0;
1553 		for (p = name; p < name + namelen; p++) {
1554 			if (*p == '.')
1555 				i++;
1556 		}
1557 		if (i < 2)
1558 			nterm = 2;
1559 		else
1560 			nterm = 1;
1561 
1562 		p = name;
1563 		while (cp < ep && p < name + namelen) {
1564 			i = 0;
1565 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1566 				i++;
1567 			/* result does not fit into mbuf */
1568 			if (cp + i + 1 >= ep)
1569 				goto fail;
1570 			/*
1571 			 * DNS label length restriction, RFC1035 page 8.
1572 			 * "i == 0" case is included here to avoid returning
1573 			 * 0-length label on "foo..bar".
1574 			 */
1575 			if (i <= 0 || i >= 64)
1576 				goto fail;
1577 			*cp++ = i;
1578 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1579 				goto fail;
1580 			while (i > 0) {
1581 				if (!isalnum(*p) && *p != '-')
1582 					goto fail;
1583 				if (isupper(*p)) {
1584 					*cp++ = tolower(*p);
1585 					p++;
1586 				} else
1587 					*cp++ = *p++;
1588 				i--;
1589 			}
1590 			p = q;
1591 			if (p < name + namelen && *p == '.')
1592 				p++;
1593 		}
1594 		/* termination */
1595 		if (cp + nterm >= ep)
1596 			goto fail;
1597 		while (nterm-- > 0)
1598 			*cp++ = '\0';
1599 		m->m_len = cp - mtod(m, char *);
1600 		return m;
1601 	}
1602 
1603 	panic("should not reach here");
1604 	/* NOTREACHED */
1605 
1606 fail:
1607 	if (m)
1608 		m_freem(m);
1609 	return NULL;
1610 }
1611 
1612 /*
1613  * check if two DNS-encoded string matches.  takes care of truncated
1614  * form (with \0\0 at the end).  no compression support.
1615  * XXX upper/lowercase match (see RFC2065)
1616  */
1617 static int
1618 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1619 {
1620 	const char *a0, *b0;
1621 	int l;
1622 
1623 	/* simplest case - need validation? */
1624 	if (alen == blen && memcmp(a, b, alen) == 0)
1625 		return 1;
1626 
1627 	a0 = a;
1628 	b0 = b;
1629 
1630 	/* termination is mandatory */
1631 	if (alen < 2 || blen < 2)
1632 		return 0;
1633 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1634 		return 0;
1635 	alen--;
1636 	blen--;
1637 
1638 	while (a - a0 < alen && b - b0 < blen) {
1639 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1640 			return 0;
1641 
1642 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1643 			return 0;
1644 		/* we don't support compression yet */
1645 		if (a[0] >= 64 || b[0] >= 64)
1646 			return 0;
1647 
1648 		/* truncated case */
1649 		if (a[0] == 0 && a - a0 == alen - 1)
1650 			return 1;
1651 		if (b[0] == 0 && b - b0 == blen - 1)
1652 			return 1;
1653 		if (a[0] == 0 || b[0] == 0)
1654 			return 0;
1655 
1656 		if (a[0] != b[0])
1657 			return 0;
1658 		l = a[0];
1659 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1660 			return 0;
1661 		if (memcmp(a + 1, b + 1, l) != 0)
1662 			return 0;
1663 
1664 		a += 1 + l;
1665 		b += 1 + l;
1666 	}
1667 
1668 	if (a - a0 == alen && b - b0 == blen)
1669 		return 1;
1670 	else
1671 		return 0;
1672 }
1673 
1674 /*
1675  * calculate the number of addresses to be returned in the node info reply.
1676  */
1677 static int
1678 ni6_addrs(struct icmp6_nodeinfo *ni6, struct ifnet **ifpp, char *subj,
1679     struct psref *psref)
1680 {
1681 	struct ifnet *ifp;
1682 	struct in6_ifaddr *ia6;
1683 	struct ifaddr *ifa;
1684 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1685 	int addrs = 0, addrsofif, iffound = 0;
1686 	int niflags = ni6->ni_flags;
1687 	int s;
1688 
1689 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1690 		switch (ni6->ni_code) {
1691 		case ICMP6_NI_SUBJ_IPV6:
1692 			if (subj == NULL) /* must be impossible... */
1693 				return 0;
1694 			subj_ip6 = (struct sockaddr_in6 *)subj;
1695 			break;
1696 		default:
1697 			/*
1698 			 * XXX: we only support IPv6 subject address for
1699 			 * this Qtype.
1700 			 */
1701 			return 0;
1702 		}
1703 	}
1704 
1705 	s = pserialize_read_enter();
1706 	IFNET_READER_FOREACH(ifp) {
1707 		addrsofif = 0;
1708 		IFADDR_READER_FOREACH(ifa, ifp) {
1709 			if (ifa->ifa_addr->sa_family != AF_INET6)
1710 				continue;
1711 			ia6 = (struct in6_ifaddr *)ifa;
1712 
1713 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1714 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1715 			     &ia6->ia_addr.sin6_addr))
1716 				iffound = 1;
1717 
1718 			/*
1719 			 * IPv4-mapped addresses can only be returned by a
1720 			 * Node Information proxy, since they represent
1721 			 * addresses of IPv4-only nodes, which perforce do
1722 			 * not implement this protocol.
1723 			 * [icmp-name-lookups-07, Section 5.4]
1724 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1725 			 * this function at this moment.
1726 			 */
1727 
1728 			/* What do we have to do about ::1? */
1729 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1730 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1731 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1732 					continue;
1733 				break;
1734 			case IPV6_ADDR_SCOPE_SITELOCAL:
1735 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1736 					continue;
1737 				break;
1738 			case IPV6_ADDR_SCOPE_GLOBAL:
1739 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1740 					continue;
1741 				break;
1742 			default:
1743 				continue;
1744 			}
1745 
1746 			/*
1747 			 * check if anycast is okay.
1748 			 * XXX: just experimental.  not in the spec.
1749 			 */
1750 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1751 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1752 				continue; /* we need only unicast addresses */
1753 
1754 			addrsofif++; /* count the address */
1755 		}
1756 		if (iffound) {
1757 			if_acquire(ifp, psref);
1758 			pserialize_read_exit(s);
1759 			*ifpp = ifp;
1760 			return addrsofif;
1761 		}
1762 
1763 		addrs += addrsofif;
1764 	}
1765 	pserialize_read_exit(s);
1766 
1767 	return addrs;
1768 }
1769 
1770 static int
1771 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1772 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1773 	int resid)
1774 {
1775 	struct ifnet *ifp;
1776 	struct in6_ifaddr *ia6;
1777 	struct ifaddr *ifa;
1778 	struct ifnet *ifp_dep = NULL;
1779 	int copied = 0, allow_deprecated = 0;
1780 	u_char *cp = (u_char *)(nni6 + 1);
1781 	int niflags = ni6->ni_flags;
1782 	u_int32_t ltime;
1783 	int s;
1784 
1785 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1786 		return 0;	/* needless to copy */
1787 
1788 	s = pserialize_read_enter();
1789 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1790 again:
1791 
1792 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1793 	{
1794 		IFADDR_READER_FOREACH(ifa, ifp) {
1795 			if (ifa->ifa_addr->sa_family != AF_INET6)
1796 				continue;
1797 			ia6 = (struct in6_ifaddr *)ifa;
1798 
1799 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1800 			    allow_deprecated == 0) {
1801 				/*
1802 				 * prefererred address should be put before
1803 				 * deprecated addresses.
1804 				 */
1805 
1806 				/* record the interface for later search */
1807 				if (ifp_dep == NULL)
1808 					ifp_dep = ifp;
1809 
1810 				continue;
1811 			}
1812 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1813 				 allow_deprecated != 0)
1814 				continue; /* we now collect deprecated addrs */
1815 
1816 			/* What do we have to do about ::1? */
1817 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1818 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1819 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1820 					continue;
1821 				break;
1822 			case IPV6_ADDR_SCOPE_SITELOCAL:
1823 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1824 					continue;
1825 				break;
1826 			case IPV6_ADDR_SCOPE_GLOBAL:
1827 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1828 					continue;
1829 				break;
1830 			default:
1831 				continue;
1832 			}
1833 
1834 			/*
1835 			 * check if anycast is okay.
1836 			 * XXX: just experimental.  not in the spec.
1837 			 */
1838 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1839 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1840 				continue;
1841 
1842 			/* now we can copy the address */
1843 			if (resid < sizeof(struct in6_addr) +
1844 			    sizeof(u_int32_t)) {
1845 				/*
1846 				 * We give up much more copy.
1847 				 * Set the truncate flag and return.
1848 				 */
1849 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1850 				goto out;
1851 			}
1852 
1853 			/*
1854 			 * Set the TTL of the address.
1855 			 * The TTL value should be one of the following
1856 			 * according to the specification:
1857 			 *
1858 			 * 1. The remaining lifetime of a DHCP lease on the
1859 			 *    address, or
1860 			 * 2. The remaining Valid Lifetime of a prefix from
1861 			 *    which the address was derived through Stateless
1862 			 *    Autoconfiguration.
1863 			 *
1864 			 * Note that we currently do not support stateful
1865 			 * address configuration by DHCPv6, so the former
1866 			 * case can't happen.
1867 			 *
1868 			 * TTL must be 2^31 > TTL >= 0.
1869 			 */
1870 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1871 				ltime = ND6_INFINITE_LIFETIME;
1872 			else {
1873 				if (ia6->ia6_lifetime.ia6t_expire >
1874 				    time_uptime)
1875 					ltime = ia6->ia6_lifetime.ia6t_expire -
1876 					    time_uptime;
1877 				else
1878 					ltime = 0;
1879 			}
1880 			if (ltime > 0x7fffffff)
1881 				ltime = 0x7fffffff;
1882 			ltime = htonl(ltime);
1883 
1884 			memcpy(cp, &ltime, sizeof(u_int32_t));
1885 			cp += sizeof(u_int32_t);
1886 
1887 			/* copy the address itself */
1888 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1889 			      sizeof(struct in6_addr));
1890 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1891 			cp += sizeof(struct in6_addr);
1892 
1893 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1894 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1895 		}
1896 		if (ifp0)	/* we need search only on the specified IF */
1897 			break;
1898 	}
1899 
1900 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1901 		ifp = ifp_dep;
1902 		allow_deprecated = 1;
1903 
1904 		goto again;
1905 	}
1906 out:
1907 	pserialize_read_exit(s);
1908 	return copied;
1909 }
1910 
1911 /*
1912  * XXX almost dup'ed code with rip6_input.
1913  */
1914 static int
1915 icmp6_rip6_input(struct mbuf **mp, int off)
1916 {
1917 	struct mbuf *m = *mp;
1918 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1919 	struct inpcb_hdr *inph;
1920 	struct in6pcb *in6p;
1921 	struct in6pcb *last = NULL;
1922 	struct sockaddr_in6 rip6src;
1923 	struct icmp6_hdr *icmp6;
1924 	struct mbuf *n, *opts = NULL;
1925 
1926 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1927 	if (icmp6 == NULL) {
1928 		/* m is already reclaimed */
1929 		return IPPROTO_DONE;
1930 	}
1931 
1932 	/*
1933 	 * XXX: the address may have embedded scope zone ID, which should be
1934 	 * hidden from applications.
1935 	 */
1936 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1937 	if (sa6_recoverscope(&rip6src)) {
1938 		m_freem(m);
1939 		return IPPROTO_DONE;
1940 	}
1941 
1942 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1943 		in6p = (struct in6pcb *)inph;
1944 		if (in6p->in6p_af != AF_INET6)
1945 			continue;
1946 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1947 			continue;
1948 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1949 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1950 			continue;
1951 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1952 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1953 			continue;
1954 		if (in6p->in6p_icmp6filt &&
1955 		    ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1956 		    in6p->in6p_icmp6filt))
1957 			continue;
1958 
1959 		if (last == NULL) {
1960 			;
1961 		}
1962 #ifdef IPSEC
1963 		else if (ipsec_used && ipsec_in_reject(m, last)) {
1964 			/* do not inject data into pcb */
1965 		}
1966 #endif
1967 		else if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1968 			if (last->in6p_flags & IN6P_CONTROLOPTS)
1969 				ip6_savecontrol(last, &opts, ip6, n);
1970 			/* strip intermediate headers */
1971 			m_adj(n, off);
1972 			if (sbappendaddr(&last->in6p_socket->so_rcv,
1973 			    sin6tosa(&rip6src), n, opts) == 0) {
1974 				soroverflow(last->in6p_socket);
1975 				m_freem(n);
1976 				if (opts)
1977 					m_freem(opts);
1978 			} else {
1979 				sorwakeup(last->in6p_socket);
1980 			}
1981 			opts = NULL;
1982 		}
1983 
1984 		last = in6p;
1985 	}
1986 
1987 #ifdef IPSEC
1988 	if (ipsec_used && last && ipsec_in_reject(m, last)) {
1989 		m_freem(m);
1990 		IP6_STATDEC(IP6_STAT_DELIVERED);
1991 		/* do not inject data into pcb */
1992 	} else
1993 #endif
1994 	if (last) {
1995 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1996 			ip6_savecontrol(last, &opts, ip6, m);
1997 		/* strip intermediate headers */
1998 		m_adj(m, off);
1999 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2000 		    sin6tosa(&rip6src), m, opts) == 0) {
2001 			soroverflow(last->in6p_socket);
2002 			m_freem(m);
2003 			if (opts)
2004 				m_freem(opts);
2005 		} else {
2006 			sorwakeup(last->in6p_socket);
2007 		}
2008 	} else {
2009 		m_freem(m);
2010 		IP6_STATDEC(IP6_STAT_DELIVERED);
2011 	}
2012 	return IPPROTO_DONE;
2013 }
2014 
2015 /*
2016  * Reflect the ip6 packet back to the source.
2017  * OFF points to the icmp6 header, counted from the top of the mbuf.
2018  *
2019  * Note: RFC 1885 required that an echo reply should be truncated if it
2020  * did not fit in with (return) path MTU, and KAME code supported the
2021  * behavior.  However, as a clarification after the RFC, this limitation
2022  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2023  * old behavior, with a (non-default) ifdef block, while the new version of
2024  * the spec was an internet-draft status, and even after the new RFC was
2025  * published.  But it would rather make sense to clean the obsoleted part
2026  * up, and to make the code simpler at this stage.
2027  */
2028 static void
2029 icmp6_reflect(struct mbuf *m, size_t off)
2030 {
2031 	struct ip6_hdr *ip6;
2032 	struct icmp6_hdr *icmp6;
2033 	const struct in6_ifaddr *ia;
2034 	const struct ip6aux *ip6a;
2035 	int plen;
2036 	int type, code;
2037 	struct ifnet *outif = NULL;
2038 	struct in6_addr origdst;
2039 	struct ifnet *rcvif;
2040 	int s;
2041 	bool ip6_src_filled = false;
2042 
2043 	/* too short to reflect */
2044 	if (off < sizeof(struct ip6_hdr)) {
2045 		nd6log(LOG_DEBUG,
2046 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2047 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2048 		    __FILE__, __LINE__);
2049 		goto bad;
2050 	}
2051 
2052 	/*
2053 	 * If there are extra headers between IPv6 and ICMPv6, strip
2054 	 * off that header first.
2055 	 */
2056 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
2057 	if (off > sizeof(struct ip6_hdr)) {
2058 		size_t l;
2059 		struct ip6_hdr nip6;
2060 
2061 		l = off - sizeof(struct ip6_hdr);
2062 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2063 		m_adj(m, l);
2064 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2065 		if (m->m_len < l) {
2066 			if ((m = m_pullup(m, l)) == NULL)
2067 				return;
2068 		}
2069 		memcpy(mtod(m, void *), (void *)&nip6, sizeof(nip6));
2070 	} else {
2071 		size_t l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2072 		if (m->m_len < l) {
2073 			if ((m = m_pullup(m, l)) == NULL)
2074 				return;
2075 		}
2076 	}
2077 
2078 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2079 	ip6 = mtod(m, struct ip6_hdr *);
2080 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2081 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2082 	type = icmp6->icmp6_type; /* keep type for statistics */
2083 	code = icmp6->icmp6_code; /* ditto. */
2084 
2085 	origdst = ip6->ip6_dst;
2086 	/*
2087 	 * ip6_input() drops a packet if its src is multicast.
2088 	 * So, the src is never multicast.
2089 	 */
2090 	ip6->ip6_dst = ip6->ip6_src;
2091 
2092 	/*
2093 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2094 	 * use dst as the src for the reply.
2095 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2096 	 * (for example) when we encounter an error while forwarding procedure
2097 	 * destined to a duplicated address of ours.
2098 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2099 	 * procedure of an outgoing packet of our own, in which case we need
2100 	 * to search in the ifaddr list.
2101 	 */
2102 	if (IN6_IS_ADDR_MULTICAST(&origdst)) {
2103 		;
2104 	} else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2105 		if ((ip6a->ip6a_flags &
2106 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2107 			ip6->ip6_src = ip6a->ip6a_src;
2108 			ip6_src_filled = true;
2109 		}
2110 	} else {
2111 		union {
2112 			struct sockaddr_in6 sin6;
2113 			struct sockaddr sa;
2114 		} u;
2115 		int _s;
2116 		struct ifaddr *ifa;
2117 
2118 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2119 
2120 		_s = pserialize_read_enter();
2121 		ifa = ifa_ifwithaddr(&u.sa);
2122 
2123 		if (ifa != NULL) {
2124 			ia = ifatoia6(ifa);
2125 			if ((ia->ia6_flags &
2126 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2127 				ip6->ip6_src = ia->ia_addr.sin6_addr;
2128 				ip6_src_filled = true;
2129 			}
2130 		}
2131 		pserialize_read_exit(_s);
2132 	}
2133 
2134 	if (!ip6_src_filled) {
2135 		int e;
2136 		struct sockaddr_in6 sin6;
2137 		struct route ro;
2138 
2139 		/*
2140 		 * This case matches to multicasts, our anycast, or unicasts
2141 		 * that we do not own.  Select a source address based on the
2142 		 * source address of the erroneous packet.
2143 		 */
2144 		/* zone ID should be embedded */
2145 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2146 
2147 		memset(&ro, 0, sizeof(ro));
2148 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2149 		    &ip6->ip6_src);
2150 		rtcache_free(&ro);
2151 		if (e != 0) {
2152 			char ip6buf[INET6_ADDRSTRLEN];
2153 			nd6log(LOG_DEBUG,
2154 			    "source can't be determined: "
2155 			    "dst=%s, error=%d\n",
2156 			    IN6_PRINT(ip6buf, &sin6.sin6_addr), e);
2157 			goto bad;
2158 		}
2159 	}
2160 
2161 	ip6->ip6_flow = 0;
2162 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2163 	ip6->ip6_vfc |= IPV6_VERSION;
2164 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2165 	rcvif = m_get_rcvif(m, &s);
2166 	if (rcvif) {
2167 		/* XXX: This may not be the outgoing interface */
2168 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2169 	} else {
2170 		ip6->ip6_hlim = ip6_defhlim;
2171 	}
2172 	m_put_rcvif(rcvif, &s);
2173 
2174 	m->m_pkthdr.csum_flags = 0;
2175 	icmp6->icmp6_cksum = 0;
2176 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2177 	    sizeof(struct ip6_hdr), plen);
2178 
2179 	/*
2180 	 * XXX option handling
2181 	 */
2182 
2183 	m->m_flags &= ~(M_BCAST|M_MCAST);
2184 
2185 	/*
2186 	 * To avoid a "too big" situation at an intermediate router
2187 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2188 	 * Note that only echo and node information replies are affected,
2189 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2190 	 */
2191 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2192 	    outif)
2193 		icmp6_ifstat_inc(outif, ifs6_out_error);
2194 	if (outif)
2195 		icmp6_ifoutstat_inc(outif, type, code);
2196 
2197 	return;
2198 
2199  bad:
2200 	m_freem(m);
2201 	return;
2202 }
2203 
2204 static const char *
2205 icmp6_redirect_diag(char *buf, size_t buflen, struct in6_addr *src6,
2206     struct in6_addr *dst6,  struct in6_addr *tgt6)
2207 {
2208 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
2209 	char ip6buft[INET6_ADDRSTRLEN];
2210 
2211 	snprintf(buf, buflen, "(src=%s dst=%s tgt=%s)",
2212 	    IN6_PRINT(ip6bufs, src6), IN6_PRINT(ip6bufd, dst6),
2213 	    IN6_PRINT(ip6buft, tgt6));
2214 	return buf;
2215 }
2216 
2217 static void
2218 icmp6_redirect_input(struct mbuf *m, int off)
2219 {
2220 	struct ifnet *ifp;
2221 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2222 	struct nd_redirect *nd_rd;
2223 	int icmp6len = m->m_pkthdr.len - off;
2224 	char *lladdr = NULL;
2225 	int lladdrlen = 0;
2226 	struct rtentry *rt = NULL;
2227 	int is_router;
2228 	int is_onlink;
2229 	struct in6_addr src6 = ip6->ip6_src;
2230 	struct in6_addr redtgt6;
2231 	struct in6_addr reddst6;
2232 	union nd_opts ndopts;
2233 	struct psref psref;
2234 	char ip6buf[INET6_ADDRSTRLEN];
2235 	char diagbuf[256];
2236 
2237 	ifp = m_get_rcvif_psref(m, &psref);
2238 	if (ifp == NULL)
2239 		goto freeit;
2240 
2241 	/* XXX if we are router, we don't update route by icmp6 redirect */
2242 	if (ip6_forwarding)
2243 		goto freeit;
2244 	if (!icmp6_rediraccept)
2245 		goto freeit;
2246 
2247 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2248 	if (nd_rd == NULL) {
2249 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2250 		m_put_rcvif_psref(ifp, &psref);
2251 		return;
2252 	}
2253 	redtgt6 = nd_rd->nd_rd_target;
2254 	reddst6 = nd_rd->nd_rd_dst;
2255 
2256 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2257 	    in6_setscope(&reddst6, ifp, NULL)) {
2258 		goto freeit;
2259 	}
2260 
2261 	/* validation */
2262 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2263 		nd6log(LOG_ERR,
2264 		    "ICMP6 redirect sent from %s rejected; "
2265 		    "must be from linklocal\n", IN6_PRINT(ip6buf, &src6));
2266 		goto bad;
2267 	}
2268 	if (ip6->ip6_hlim != 255) {
2269 		nd6log(LOG_ERR,
2270 		    "ICMP6 redirect sent from %s rejected; "
2271 		    "hlim=%d (must be 255)\n",
2272 		    IN6_PRINT(ip6buf, &src6), ip6->ip6_hlim);
2273 		goto bad;
2274 	}
2275 
2276     {
2277 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2278 	struct sockaddr_in6 sin6;
2279 	struct in6_addr *gw6;
2280 
2281 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2282 	rt = rtalloc1(sin6tosa(&sin6), 0);
2283 	if (rt) {
2284 		if (rt->rt_gateway == NULL ||
2285 		    rt->rt_gateway->sa_family != AF_INET6) {
2286 			nd6log(LOG_ERR,
2287 			    "ICMP6 redirect rejected; no route "
2288 			    "with inet6 gateway found for redirect dst: %s\n",
2289 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2290 			    &src6, &reddst6, &redtgt6));
2291 			rt_unref(rt);
2292 			goto bad;
2293 		}
2294 
2295 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2296 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2297 			nd6log(LOG_ERR,
2298 			    "ICMP6 redirect rejected; "
2299 			    "not equal to gw-for-src=%s (must be same): %s\n",
2300 			    IN6_PRINT(ip6buf, gw6),
2301 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2302 			    &src6, &reddst6, &redtgt6));
2303 			rt_unref(rt);
2304 			goto bad;
2305 		}
2306 	} else {
2307 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2308 		    "no route found for redirect dst: %s\n",
2309 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2310 		    &src6, &reddst6, &redtgt6));
2311 		goto bad;
2312 	}
2313 	rt_unref(rt);
2314 	rt = NULL;
2315     }
2316 
2317 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2318 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2319 		    "redirect dst must be unicast: %s\n",
2320 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2321 		    &src6, &reddst6, &redtgt6));
2322 		goto bad;
2323 	}
2324 
2325 	is_router = is_onlink = 0;
2326 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2327 		is_router = 1;	/* router case */
2328 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2329 		is_onlink = 1;	/* on-link destination case */
2330 	if (!is_router && !is_onlink) {
2331 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2332 		    "neither router case nor onlink case: %s\n",
2333 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2334 		    &src6, &reddst6, &redtgt6));
2335 		goto bad;
2336 	}
2337 	/* validation passed */
2338 
2339 	icmp6len -= sizeof(*nd_rd);
2340 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2341 	if (nd6_options(&ndopts) < 0) {
2342 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2343 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2344 		    &src6, &reddst6, &redtgt6));
2345 		/* nd6_options have incremented stats */
2346 		goto freeit;
2347 	}
2348 
2349 	if (ndopts.nd_opts_tgt_lladdr) {
2350 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2351 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2352 	}
2353 
2354 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2355 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2356 		    "(if %d, icmp6 packet %d): %s\n",
2357 		    IN6_PRINT(ip6buf, &redtgt6),
2358 		    ifp->if_addrlen, lladdrlen - 2,
2359 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2360 		    &src6, &reddst6, &redtgt6));
2361 		goto bad;
2362 	}
2363 
2364 	/* RFC 2461 8.3 */
2365 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2366 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2367 
2368 	m_put_rcvif_psref(ifp, &psref);
2369 	ifp = NULL;
2370 
2371 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2372 		/* perform rtredirect */
2373 		struct sockaddr_in6 sdst;
2374 		struct sockaddr_in6 sgw;
2375 		struct sockaddr_in6 ssrc;
2376 		unsigned long rtcount;
2377 		struct rtentry *newrt = NULL;
2378 
2379 		/*
2380 		 * do not install redirect route, if the number of entries
2381 		 * is too much (> hiwat).  note that, the node (= host) will
2382 		 * work just fine even if we do not install redirect route
2383 		 * (there will be additional hops, though).
2384 		 */
2385 		mutex_enter(&icmp6_mtx);
2386 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2387 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) {
2388 			mutex_exit(&icmp6_mtx);
2389 			goto freeit;
2390 		}
2391 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) {
2392 			mutex_exit(&icmp6_mtx);
2393 			goto freeit;
2394 		} else if (0 <= icmp6_redirect_lowat &&
2395 		    rtcount > icmp6_redirect_lowat) {
2396 			/*
2397 			 * XXX nuke a victim, install the new one.
2398 			 */
2399 		}
2400 
2401 		memset(&sdst, 0, sizeof(sdst));
2402 		memset(&sgw, 0, sizeof(sgw));
2403 		memset(&ssrc, 0, sizeof(ssrc));
2404 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2405 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2406 		    sizeof(struct sockaddr_in6);
2407 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2408 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2409 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2410 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2411 		    RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc), &newrt);
2412 
2413 		if (newrt) {
2414 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2415 			    icmp6_redirect_timeout_q);
2416 			rt_unref(newrt);
2417 		}
2418 		mutex_exit(&icmp6_mtx);
2419 	}
2420 	/* finally update cached route in each socket via pfctlinput */
2421 	{
2422 		struct sockaddr_in6 sdst;
2423 
2424 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2425 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2426 #if defined(IPSEC)
2427 		if (ipsec_used)
2428 			key_sa_routechange(sin6tosa(&sdst));
2429 #endif
2430 	}
2431 
2432 freeit:
2433 	if (ifp != NULL)
2434 		m_put_rcvif_psref(ifp, &psref);
2435 	m_freem(m);
2436 	return;
2437 
2438 bad:
2439 	m_put_rcvif_psref(ifp, &psref);
2440 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2441 	m_freem(m);
2442 }
2443 
2444 void
2445 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2446 {
2447 	struct ifnet *ifp;	/* my outgoing interface */
2448 	struct in6_addr *ifp_ll6;
2449 	struct in6_addr *nexthop;
2450 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2451 	struct mbuf *m = NULL;	/* newly allocated one */
2452 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2453 	struct nd_redirect *nd_rd;
2454 	size_t maxlen;
2455 	u_char *p;
2456 	struct sockaddr_in6 src_sa;
2457 
2458 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2459 
2460 	/* if we are not router, we don't send icmp6 redirect */
2461 	if (!ip6_forwarding)
2462 		goto fail;
2463 
2464 	/* sanity check */
2465 	KASSERT(m0 != NULL);
2466 	KASSERT(rt != NULL);
2467 
2468 	ifp = rt->rt_ifp;
2469 
2470 	/*
2471 	 * Address check:
2472 	 *  the source address must identify a neighbor, and
2473 	 *  the destination address must not be a multicast address
2474 	 *  [RFC 2461, sec 8.2]
2475 	 */
2476 	sip6 = mtod(m0, struct ip6_hdr *);
2477 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2478 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2479 		goto fail;
2480 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2481 		goto fail;	/* what should we do here? */
2482 
2483 	/* rate limit */
2484 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2485 		goto fail;
2486 
2487 	/*
2488 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2489 	 * we almost always ask for an mbuf cluster for simplicity.
2490 	 * (MHLEN < IPV6_MMTU is almost always true)
2491 	 */
2492 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2493 	if (m && IPV6_MMTU >= MHLEN) {
2494 #if IPV6_MMTU >= MCLBYTES
2495 		MEXTMALLOC(m, IPV6_MMTU, M_NOWAIT);
2496 #else
2497 		MCLGET(m, M_DONTWAIT);
2498 #endif
2499 	}
2500 
2501 	if (!m)
2502 		goto fail;
2503 	m_reset_rcvif(m);
2504 	m->m_len = 0;
2505 	maxlen = M_TRAILINGSPACE(m);
2506 	maxlen = min(IPV6_MMTU, maxlen);
2507 
2508 	/* just for safety */
2509 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct nd_redirect) +
2510 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2511 		goto fail;
2512 	}
2513 
2514 	{
2515 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2516 		struct in6_ifaddr *ia;
2517 		int s = pserialize_read_enter();
2518 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2519 						 IN6_IFF_NOTREADY|
2520 						 IN6_IFF_ANYCAST)) == NULL) {
2521 			pserialize_read_exit(s);
2522 			goto fail;
2523 		}
2524 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2525 		pserialize_read_exit(s);
2526 	}
2527 
2528 	/* get ip6 linklocal address for the router. */
2529 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2530 		struct sockaddr_in6 *sin6;
2531 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2532 		nexthop = &sin6->sin6_addr;
2533 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2534 			nexthop = NULL;
2535 	} else
2536 		nexthop = NULL;
2537 
2538 	/* ip6 */
2539 	ip6 = mtod(m, struct ip6_hdr *);
2540 	ip6->ip6_flow = 0;
2541 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2542 	ip6->ip6_vfc |= IPV6_VERSION;
2543 	/* ip6->ip6_plen will be set later */
2544 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2545 	ip6->ip6_hlim = 255;
2546 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2547 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2548 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2549 
2550 	/* ND Redirect */
2551 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2552 	nd_rd->nd_rd_type = ND_REDIRECT;
2553 	nd_rd->nd_rd_code = 0;
2554 	nd_rd->nd_rd_reserved = 0;
2555 	if (rt->rt_flags & RTF_GATEWAY) {
2556 		/*
2557 		 * nd_rd->nd_rd_target must be a link-local address in
2558 		 * better router cases.
2559 		 */
2560 		if (!nexthop)
2561 			goto fail;
2562 		bcopy(nexthop, &nd_rd->nd_rd_target,
2563 		      sizeof(nd_rd->nd_rd_target));
2564 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2565 		      sizeof(nd_rd->nd_rd_dst));
2566 	} else {
2567 		/* make sure redtgt == reddst */
2568 		nexthop = &sip6->ip6_dst;
2569 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2570 		      sizeof(nd_rd->nd_rd_target));
2571 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2572 		      sizeof(nd_rd->nd_rd_dst));
2573 	}
2574 
2575 	p = (u_char *)(nd_rd + 1);
2576 
2577 	{
2578 		/* target lladdr option */
2579 		struct llentry *ln = NULL;
2580 		int len, pad;
2581 		struct nd_opt_hdr *nd_opt;
2582 		char *lladdr;
2583 
2584 		ln = nd6_lookup(nexthop, ifp, false);
2585 		if (ln == NULL)
2586 			goto nolladdropt;
2587 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2588 		len = (len + 7) & ~7;	/* round by 8 */
2589 		pad = len - (sizeof(*nd_opt) + ifp->if_addrlen);
2590 
2591 		/* safety check */
2592 		if (len + (p - (u_char *)ip6) > maxlen) {
2593 			LLE_RUNLOCK(ln);
2594 			goto nolladdropt;
2595 		}
2596 
2597 		if (ln->la_flags & LLE_VALID) {
2598 			nd_opt = (struct nd_opt_hdr *)p;
2599 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2600 			nd_opt->nd_opt_len = len >> 3;
2601 			lladdr = (char *)(nd_opt + 1);
2602 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2603 			memset(lladdr + ifp->if_addrlen, 0, pad);
2604 			p += len;
2605 		}
2606 		LLE_RUNLOCK(ln);
2607 	}
2608 nolladdropt:
2609 
2610 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2611 
2612 	/* just to be safe */
2613 	if (m0->m_flags & M_DECRYPTED)
2614 		goto noredhdropt;
2615 	if (p - (u_char *)ip6 > maxlen)
2616 		goto noredhdropt;
2617 
2618 	{
2619 		/* redirected header option */
2620 		int len;
2621 		struct nd_opt_rd_hdr *nd_opt_rh;
2622 
2623 		/*
2624 		 * compute the maximum size for icmp6 redirect header option.
2625 		 * XXX room for auth header?
2626 		 */
2627 		len = maxlen - (p - (u_char *)ip6);
2628 		len &= ~7;
2629 
2630 		if (len < sizeof(*nd_opt_rh)) {
2631 			goto noredhdropt;
2632 		}
2633 
2634 		/*
2635 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2636 		 * about padding/truncate rule for the original IP packet.
2637 		 * From the discussion on IPv6imp in Feb 1999,
2638 		 * the consensus was:
2639 		 * - "attach as much as possible" is the goal
2640 		 * - pad if not aligned (original size can be guessed by
2641 		 *   original ip6 header)
2642 		 * Following code adds the padding if it is simple enough,
2643 		 * and truncates if not.
2644 		 */
2645 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2646 			/* not enough room, truncate */
2647 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2648 			    m0->m_pkthdr.len);
2649 		} else {
2650 			/*
2651 			 * enough room, truncate if not aligned.
2652 			 * we don't pad here for simplicity.
2653 			 */
2654 			int extra;
2655 
2656 			extra = m0->m_pkthdr.len % 8;
2657 			if (extra) {
2658 				/* truncate */
2659 				m_adj(m0, -extra);
2660 			}
2661 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2662 		}
2663 
2664 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2665 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2666 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2667 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2668 		p += sizeof(*nd_opt_rh);
2669 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2670 
2671 		/* connect m0 to m */
2672 		m->m_pkthdr.len += m0->m_pkthdr.len;
2673 		m_cat(m, m0);
2674 		m0 = NULL;
2675 	}
2676 noredhdropt:
2677 	if (m0) {
2678 		m_freem(m0);
2679 		m0 = NULL;
2680 	}
2681 
2682 	/* XXX: clear embedded link IDs in the inner header */
2683 	in6_clearscope(&sip6->ip6_src);
2684 	in6_clearscope(&sip6->ip6_dst);
2685 	in6_clearscope(&nd_rd->nd_rd_target);
2686 	in6_clearscope(&nd_rd->nd_rd_dst);
2687 
2688 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2689 
2690 	nd_rd->nd_rd_cksum = 0;
2691 	nd_rd->nd_rd_cksum =
2692 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2693 
2694 	/* send the packet to outside... */
2695 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2696 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2697 
2698 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2699 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2700 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2701 
2702 	return;
2703 
2704 fail:
2705 	if (m)
2706 		m_freem(m);
2707 	if (m0)
2708 		m_freem(m0);
2709 }
2710 
2711 /*
2712  * ICMPv6 socket option processing.
2713  */
2714 int
2715 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2716 {
2717 	int error = 0;
2718 	struct in6pcb *in6p = sotoin6pcb(so);
2719 
2720 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2721 		return rip6_ctloutput(op, so, sopt);
2722 
2723 	switch (op) {
2724 	case PRCO_SETOPT:
2725 		switch (sopt->sopt_name) {
2726 		case ICMP6_FILTER:
2727 		    {
2728 			struct icmp6_filter fil;
2729 
2730 			error = sockopt_get(sopt, &fil, sizeof(fil));
2731 			if (error)
2732 				break;
2733 			memcpy(in6p->in6p_icmp6filt, &fil,
2734 			    sizeof(struct icmp6_filter));
2735 			error = 0;
2736 			break;
2737 		    }
2738 
2739 		default:
2740 			error = ENOPROTOOPT;
2741 			break;
2742 		}
2743 		break;
2744 
2745 	case PRCO_GETOPT:
2746 		switch (sopt->sopt_name) {
2747 		case ICMP6_FILTER:
2748 		    {
2749 			if (in6p->in6p_icmp6filt == NULL) {
2750 				error = EINVAL;
2751 				break;
2752 			}
2753 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2754 			    sizeof(struct icmp6_filter));
2755 			break;
2756 		    }
2757 
2758 		default:
2759 			error = ENOPROTOOPT;
2760 			break;
2761 		}
2762 		break;
2763 	}
2764 
2765 	return error;
2766 }
2767 
2768 /*
2769  * Perform rate limit check.
2770  * Returns 0 if it is okay to send the icmp6 packet.
2771  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2772  * limitation.
2773  *
2774  * XXX per-destination/type check necessary?
2775  */
2776 static int
2777 icmp6_ratelimit(
2778 	const struct in6_addr *dst,	/* not used at this moment */
2779 	const int type,		/* not used at this moment */
2780 	const int code)		/* not used at this moment */
2781 {
2782 	int ret;
2783 
2784 	ret = 0;	/* okay to send */
2785 
2786 	/* PPS limit */
2787 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2788 	    icmp6errppslim)) {
2789 		/* The packet is subject to rate limit */
2790 		ret++;
2791 	}
2792 
2793 	return ret;
2794 }
2795 
2796 static struct rtentry *
2797 icmp6_mtudisc_clone(struct sockaddr *dst)
2798 {
2799 	struct rtentry *rt;
2800 	int    error;
2801 
2802 	rt = rtalloc1(dst, 1);
2803 	if (rt == NULL)
2804 		return NULL;
2805 
2806 	/* If we didn't get a host route, allocate one */
2807 	if ((rt->rt_flags & RTF_HOST) == 0) {
2808 		struct rtentry *nrt;
2809 
2810 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2811 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2812 		if (error) {
2813 			rt_unref(rt);
2814 			return NULL;
2815 		}
2816 		nrt->rt_rmx = rt->rt_rmx;
2817 		rt_unref(rt);
2818 		rt = nrt;
2819 	}
2820 
2821 	mutex_enter(&icmp6_mtx);
2822 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2823 			icmp6_mtudisc_timeout_q);
2824 	mutex_exit(&icmp6_mtx);
2825 
2826 	if (error) {
2827 		rt_unref(rt);
2828 		return NULL;
2829 	}
2830 
2831 	return rt;	/* caller need to call rtfree() */
2832 }
2833 
2834 static void
2835 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2836 {
2837 	struct rtentry *retrt;
2838 
2839 	KASSERT(rt != NULL);
2840 	rt_assert_referenced(rt);
2841 
2842 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2843 	    (RTF_DYNAMIC | RTF_HOST)) {
2844 		rtrequest(RTM_DELETE, rt_getkey(rt),
2845 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2846 		rt_unref(rt);
2847 		rt_free(retrt);
2848 	} else {
2849 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2850 			rt->rt_rmx.rmx_mtu = 0;
2851 	}
2852 }
2853 
2854 static void
2855 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2856 {
2857 	struct rtentry *retrt;
2858 
2859 	KASSERT(rt != NULL);
2860 	rt_assert_referenced(rt);
2861 
2862 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2863 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2864 		printf("%s: RTM_DELETE\n", __func__);
2865 		rtrequest(RTM_DELETE, rt_getkey(rt),
2866 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2867 		rt_unref(rt);
2868 		rt_free(retrt);
2869 	}
2870 }
2871 
2872 /*
2873  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2874  */
2875 static int
2876 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2877 {
2878 	(void)&name;
2879 	(void)&l;
2880 	(void)&oname;
2881 
2882 	if (namelen != 0)
2883 		return (EINVAL);
2884 
2885 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2886 	    /*XXXUNCONST*/
2887 	    __UNCONST(newp), newlen));
2888 }
2889 
2890 static int
2891 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2892 {
2893 
2894 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2895 }
2896 
2897 static int
2898 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2899 {
2900 	int error, tmp;
2901 	struct sysctlnode node;
2902 
2903 	mutex_enter(&icmp6_mtx);
2904 
2905 	node = *rnode;
2906 	node.sysctl_data = &tmp;
2907 	tmp = icmp6_redirtimeout;
2908 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2909 	if (error || newp == NULL)
2910 		goto out;
2911 	if (tmp < 0) {
2912 		error = EINVAL;
2913 		goto out;
2914 	}
2915 	icmp6_redirtimeout = tmp;
2916 
2917 	if (icmp6_redirect_timeout_q != NULL) {
2918 		if (icmp6_redirtimeout == 0) {
2919 			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2920 		} else {
2921 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2922 			    icmp6_redirtimeout);
2923 		}
2924 	} else if (icmp6_redirtimeout > 0) {
2925 		icmp6_redirect_timeout_q =
2926 		    rt_timer_queue_create(icmp6_redirtimeout);
2927 	}
2928 	error = 0;
2929 out:
2930 	mutex_exit(&icmp6_mtx);
2931 	return error;
2932 }
2933 
2934 static void
2935 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2936 {
2937 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2938 
2939 	sysctl_createv(clog, 0, NULL, NULL,
2940 		       CTLFLAG_PERMANENT,
2941 		       CTLTYPE_NODE, "inet6", NULL,
2942 		       NULL, 0, NULL, 0,
2943 		       CTL_NET, PF_INET6, CTL_EOL);
2944 	sysctl_createv(clog, 0, NULL, NULL,
2945 		       CTLFLAG_PERMANENT,
2946 		       CTLTYPE_NODE, "icmp6",
2947 		       SYSCTL_DESCR("ICMPv6 related settings"),
2948 		       NULL, 0, NULL, 0,
2949 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2950 
2951 	sysctl_createv(clog, 0, NULL, NULL,
2952 		       CTLFLAG_PERMANENT,
2953 		       CTLTYPE_STRUCT, "stats",
2954 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2955 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2956 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2957 		       ICMPV6CTL_STATS, CTL_EOL);
2958 	sysctl_createv(clog, 0, NULL, NULL,
2959 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2960 		       CTLTYPE_INT, "rediraccept",
2961 		       SYSCTL_DESCR("Accept and process redirect messages"),
2962 		       NULL, 0, &icmp6_rediraccept, 0,
2963 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2964 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2965 	sysctl_createv(clog, 0, NULL, NULL,
2966 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2967 		       CTLTYPE_INT, "redirtimeout",
2968 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2969 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2970 		       &icmp6_redirtimeout, 0,
2971 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2972 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2973 #if 0 /* obsoleted */
2974 	sysctl_createv(clog, 0, NULL, NULL,
2975 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2976 		       CTLTYPE_INT, "errratelimit", NULL,
2977 		       NULL, 0, &icmp6_errratelimit, 0,
2978 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2979 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2980 #endif
2981 	sysctl_createv(clog, 0, NULL, NULL,
2982 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2983 		       CTLTYPE_INT, "nd6_prune",
2984 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2985 		       NULL, 0, &nd6_prune, 0,
2986 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2987 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2988 	sysctl_createv(clog, 0, NULL, NULL,
2989 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2990 		       CTLTYPE_INT, "nd6_delay",
2991 		       SYSCTL_DESCR("First probe delay time"),
2992 		       NULL, 0, &nd6_delay, 0,
2993 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2994 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2995 	sysctl_createv(clog, 0, NULL, NULL,
2996 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2997 		       CTLTYPE_INT, "nd6_umaxtries",
2998 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2999 		       NULL, 0, &nd6_umaxtries, 0,
3000 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3001 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
3002 	sysctl_createv(clog, 0, NULL, NULL,
3003 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3004 		       CTLTYPE_INT, "nd6_mmaxtries",
3005 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
3006 		       NULL, 0, &nd6_mmaxtries, 0,
3007 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3008 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
3009 	sysctl_createv(clog, 0, NULL, NULL,
3010 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3011 		       CTLTYPE_INT, "nd6_useloopback",
3012 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
3013 		       NULL, 0, &nd6_useloopback, 0,
3014 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3015 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
3016 #if 0 /* obsoleted */
3017 	sysctl_createv(clog, 0, NULL, NULL,
3018 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3019 		       CTLTYPE_INT, "nd6_proxyall", NULL,
3020 		       NULL, 0, &nd6_proxyall, 0,
3021 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3022 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
3023 #endif
3024 	sysctl_createv(clog, 0, NULL, NULL,
3025 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3026 		       CTLTYPE_INT, "nodeinfo",
3027 		       SYSCTL_DESCR("Respond to node information requests"),
3028 		       NULL, 0, &icmp6_nodeinfo, 0,
3029 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3030 		       ICMPV6CTL_NODEINFO, CTL_EOL);
3031 	sysctl_createv(clog, 0, NULL, NULL,
3032 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3033 		       CTLTYPE_INT, "errppslimit",
3034 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
3035 		       NULL, 0, &icmp6errppslim, 0,
3036 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3037 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
3038 	sysctl_createv(clog, 0, NULL, NULL,
3039 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3040 		       CTLTYPE_INT, "nd6_maxnudhint",
3041 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
3042 		       NULL, 0, &nd6_maxnudhint, 0,
3043 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3044 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
3045 	sysctl_createv(clog, 0, NULL, NULL,
3046 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3047 		       CTLTYPE_INT, "mtudisc_hiwat",
3048 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3049 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
3050 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3051 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
3052 	sysctl_createv(clog, 0, NULL, NULL,
3053 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3054 		       CTLTYPE_INT, "mtudisc_lowat",
3055 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3056 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
3057 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3058 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
3059 	sysctl_createv(clog, 0, NULL, NULL,
3060 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3061 		       CTLTYPE_INT, "nd6_debug",
3062 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
3063 		       NULL, 0, &nd6_debug, 0,
3064 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3065 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
3066 	sysctl_createv(clog, 0, NULL, NULL,
3067 		       CTLFLAG_PERMANENT,
3068 		       CTLTYPE_STRUCT, "nd6_drlist",
3069 		       SYSCTL_DESCR("Default router list"),
3070 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3071 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3072 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
3073 	sysctl_createv(clog, 0, NULL, NULL,
3074 		       CTLFLAG_PERMANENT,
3075 		       CTLTYPE_STRUCT, "nd6_prlist",
3076 		       SYSCTL_DESCR("Prefix list"),
3077 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3078 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3079 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3080 	sysctl_createv(clog, 0, NULL, NULL,
3081 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3082 		       CTLTYPE_INT, "maxqueuelen",
3083 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3084 		       NULL, 1, &nd6_maxqueuelen, 0,
3085 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3086 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3087 }
3088 
3089 void
3090 icmp6_statinc(u_int stat)
3091 {
3092 
3093 	KASSERT(stat < ICMP6_NSTATS);
3094 	ICMP6_STATINC(stat);
3095 }
3096