xref: /netbsd-src/sys/netinet6/icmp6.c (revision aef5eb5f59cdfe8314f1b5f78ac04eb144e44010)
1 /*	$NetBSD: icmp6.c,v 1.252 2022/08/29 09:14:02 knakahara 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 1.252 2022/08/29 09:14:02 knakahara Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_compat_netbsd.h"
69 #include "opt_inet.h"
70 #include "opt_ipsec.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/kmem.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 #include <sys/sysctl.h>
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 #include <net/if_dl.h>
89 #include <net/if_types.h>
90 #include <net/nd.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet/wqinput.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/ip6_private.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/icmp6_private.h>
100 #include <netinet6/mld6_var.h>
101 #include <netinet6/in6_pcb.h>
102 #include <netinet6/in6_ifattach.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/nd6.h>
105 #include <netinet6/scope6_var.h>
106 
107 #ifdef IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/ipsec6.h>
110 #include <netipsec/key.h>
111 #endif
112 
113 #include "faith.h"
114 #if defined(NFAITH) && 0 < NFAITH
115 #include <net/if_faith.h>
116 #endif
117 
118 /* Ensure that non packed structures are the desired size. */
119 __CTASSERT(sizeof(struct icmp6_hdr) == 8);
120 __CTASSERT(sizeof(struct icmp6_nodeinfo) == 16);
121 __CTASSERT(sizeof(struct icmp6_namelookup) == 20);
122 __CTASSERT(sizeof(struct icmp6_router_renum) == 16);
123 
124 __CTASSERT(sizeof(struct nd_router_solicit) == 8);
125 __CTASSERT(sizeof(struct nd_router_advert) == 16);
126 __CTASSERT(sizeof(struct nd_neighbor_solicit) == 24);
127 __CTASSERT(sizeof(struct nd_neighbor_advert) == 24);
128 __CTASSERT(sizeof(struct nd_redirect) == 40);
129 __CTASSERT(sizeof(struct nd_opt_hdr) == 2);
130 __CTASSERT(sizeof(struct nd_opt_route_info) == 8);
131 __CTASSERT(sizeof(struct nd_opt_prefix_info) == 32);
132 __CTASSERT(sizeof(struct nd_opt_rd_hdr) == 8);
133 __CTASSERT(sizeof(struct nd_opt_mtu) == 8);
134 __CTASSERT(sizeof(struct nd_opt_nonce) == 2 + ND_OPT_NONCE_LEN);
135 __CTASSERT(sizeof(struct nd_opt_rdnss) == 8);
136 __CTASSERT(sizeof(struct nd_opt_dnssl) == 8);
137 
138 __CTASSERT(sizeof(struct mld_hdr) == 24);
139 __CTASSERT(sizeof(struct ni_reply_fqdn) == 8);
140 __CTASSERT(sizeof(struct rr_pco_match) == 24);
141 __CTASSERT(sizeof(struct rr_pco_use) == 32);
142 __CTASSERT(sizeof(struct rr_result) == 24);
143 
144 extern struct domain inet6domain;
145 
146 percpu_t *icmp6stat_percpu;
147 
148 extern struct inpcbtable raw6cbtable;
149 extern int icmp6errppslim;
150 static int icmp6errpps_count = 0;
151 static struct timeval icmp6errppslim_last;
152 extern int icmp6_nodeinfo;
153 
154 bool icmp6_dynamic_rt_msg = false;
155 
156 /*
157  * List of callbacks to notify when Path MTU changes are made.
158  */
159 struct icmp6_mtudisc_callback {
160 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
161 	void (*mc_func)(struct in6_addr *);
162 };
163 
164 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
165     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
166 
167 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
168 extern int pmtu_expire;
169 
170 /* XXX do these values make any sense? */
171 static int icmp6_mtudisc_hiwat = 1280;
172 static int icmp6_mtudisc_lowat = 256;
173 
174 /*
175  * keep track of # of redirect routes.
176  */
177 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
178 
179 /* XXX experimental, turned off */
180 static int icmp6_redirect_hiwat = -1;
181 static int icmp6_redirect_lowat = -1;
182 
183 /* Protect mtudisc and redirect stuffs */
184 static kmutex_t icmp6_mtx __cacheline_aligned;
185 
186 static bool icmp6_reflect_pmtu = false;
187 
188 static void icmp6_errcount(u_int, int, int);
189 static int icmp6_rip6_input(struct mbuf **, int);
190 static void icmp6_reflect(struct mbuf *, size_t);
191 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
192 static const char *icmp6_redirect_diag(char *, size_t, struct in6_addr *,
193     struct in6_addr *, struct in6_addr *);
194 static void icmp6_redirect_input(struct mbuf *, int);
195 static struct mbuf *ni6_input(struct mbuf *, int);
196 static struct mbuf *ni6_nametodns(const char *, int, int);
197 static int ni6_dnsmatch(const char *, int, const char *, int);
198 static int ni6_addrs(struct icmp6_nodeinfo *, struct ifnet **, char *,
199     struct psref *);
200 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
201     struct ifnet *, int);
202 static int icmp6_notify_error(struct mbuf *, int, int, int);
203 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
204 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
205 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
206 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
207 
208 /* workqueue-based pr_input */
209 static struct wqinput *icmp6_wqinput;
210 static void _icmp6_input(struct mbuf *m, int off, int proto);
211 
212 void
213 icmp6_init(void)
214 {
215 
216 	sysctl_net_inet6_icmp6_setup(NULL);
217 	mld_init();
218 
219 	mutex_init(&icmp6_mtx, MUTEX_DEFAULT, IPL_NONE);
220 	mutex_enter(&icmp6_mtx);
221 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
222 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
223 	mutex_exit(&icmp6_mtx);
224 
225 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
226 
227 	icmp6_wqinput = wqinput_create("icmp6", _icmp6_input);
228 }
229 
230 static void
231 icmp6_errcount(u_int base, int type, int code)
232 {
233 	switch (type) {
234 	case ICMP6_DST_UNREACH:
235 		switch (code) {
236 		case ICMP6_DST_UNREACH_NOROUTE:
237 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
238 			return;
239 		case ICMP6_DST_UNREACH_ADMIN:
240 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
241 			return;
242 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
243 			ICMP6_STATINC(base +
244 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
245 			return;
246 		case ICMP6_DST_UNREACH_ADDR:
247 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
248 			return;
249 		case ICMP6_DST_UNREACH_NOPORT:
250 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
251 			return;
252 		}
253 		break;
254 	case ICMP6_PACKET_TOO_BIG:
255 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
256 		return;
257 	case ICMP6_TIME_EXCEEDED:
258 		switch (code) {
259 		case ICMP6_TIME_EXCEED_TRANSIT:
260 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
261 			return;
262 		case ICMP6_TIME_EXCEED_REASSEMBLY:
263 			ICMP6_STATINC(base +
264 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
265 			return;
266 		}
267 		break;
268 	case ICMP6_PARAM_PROB:
269 		switch (code) {
270 		case ICMP6_PARAMPROB_HEADER:
271 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
272 			return;
273 		case ICMP6_PARAMPROB_NEXTHEADER:
274 			ICMP6_STATINC(base +
275 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
276 			return;
277 		case ICMP6_PARAMPROB_OPTION:
278 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
279 			return;
280 		}
281 		break;
282 	case ND_REDIRECT:
283 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
284 		return;
285 	}
286 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
287 }
288 
289 /*
290  * Register a Path MTU Discovery callback.
291  */
292 void
293 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
294 {
295 	struct icmp6_mtudisc_callback *mc, *new;
296 
297 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
298 
299 	mutex_enter(&icmp6_mtx);
300 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
301 	     mc = LIST_NEXT(mc, mc_list)) {
302 		if (mc->mc_func == func) {
303 			mutex_exit(&icmp6_mtx);
304 			kmem_free(new, sizeof(*mc));
305 			return;
306 		}
307 	}
308 
309 	new->mc_func = func;
310 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, new, mc_list);
311 	mutex_exit(&icmp6_mtx);
312 }
313 
314 /*
315  * A wrapper function for icmp6_error() necessary when the erroneous packet
316  * may not contain enough scope zone information.
317  */
318 void
319 icmp6_error2(struct mbuf *m, int type, int code, int param,
320 	struct ifnet *ifp, struct in6_addr *src)
321 {
322 	struct ip6_hdr *ip6;
323 
324 	KASSERT(ifp != NULL);
325 
326 	if (m->m_len < sizeof(struct ip6_hdr)) {
327 		m = m_pullup(m, sizeof(struct ip6_hdr));
328 		if (m == NULL)
329 			return;
330 	}
331 
332 	ip6 = mtod(m, struct ip6_hdr *);
333 
334 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
335 		goto out;
336 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
337 		goto out;
338 
339 	*src = ip6->ip6_src;
340 	icmp6_error(m, type, code, param);
341 	return;
342 
343 out:
344 	m_freem(m);
345 }
346 
347 /*
348  * Generate an error packet of type error in response to bad IP6 packet.
349  */
350 void
351 icmp6_error(struct mbuf *m, int type, int code, int param)
352 {
353 	struct ip6_hdr *oip6, *nip6;
354 	struct icmp6_hdr *icmp6;
355 	u_int preplen;
356 	int off;
357 	int nxt;
358 
359 	ICMP6_STATINC(ICMP6_STAT_ERROR);
360 
361 	/* count per-type-code statistics */
362 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
363 
364 	if (m->m_flags & M_DECRYPTED) {
365 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
366 		goto freeit;
367 	}
368 
369 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
370 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
371 		return;
372 	oip6 = mtod(m, struct ip6_hdr *);
373 
374 	/*
375 	 * If the destination address of the erroneous packet is a multicast
376 	 * address, or the packet was sent using link-layer multicast,
377 	 * we should basically suppress sending an error (RFC 2463, Section
378 	 * 2.4).
379 	 * We have two exceptions (the item e.2 in that section):
380 	 * - the Packet Too Big message can be sent for path MTU discovery.
381 	 * - the Parameter Problem Message that can be allowed an icmp6 error
382 	 *   in the option type field.  This check has been done in
383 	 *   ip6_unknown_opt(), so we can just check the type and code.
384 	 */
385 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
386 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
387 	    (type != ICMP6_PACKET_TOO_BIG &&
388 	     (type != ICMP6_PARAM_PROB ||
389 	      code != ICMP6_PARAMPROB_OPTION)))
390 		goto freeit;
391 
392 	/*
393 	 * RFC 2463, 2.4 (e.5): source address check.
394 	 * XXX: the case of anycast source?
395 	 */
396 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
397 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
398 		goto freeit;
399 
400 	/*
401 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
402 	 * don't do it.
403 	 */
404 	nxt = -1;
405 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
406 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
407 		struct icmp6_hdr *icp;
408 
409 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
410 			sizeof(*icp));
411 		if (icp == NULL) {
412 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
413 			return;
414 		}
415 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
416 		    icp->icmp6_type == ND_REDIRECT) {
417 			/*
418 			 * ICMPv6 error
419 			 * Special case: for redirect (which is
420 			 * informational) we must not send icmp6 error.
421 			 */
422 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
423 			goto freeit;
424 		} else {
425 			/* ICMPv6 informational - send the error */
426 		}
427 	} else {
428 		/* non-ICMPv6 - send the error */
429 	}
430 
431 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
432 
433 	/* Finally, do rate limitation check. */
434 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
435 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
436 		goto freeit;
437 	}
438 
439 	/*
440 	 * OK, ICMP6 can be generated.
441 	 */
442 
443 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
444 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
445 
446 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
447 	M_PREPEND(m, preplen, M_DONTWAIT);
448 	if (m && M_UNWRITABLE(m, preplen))
449 		m = m_pullup(m, preplen);
450 	if (m == NULL) {
451 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
452 		return;
453 	}
454 
455 	nip6 = mtod(m, struct ip6_hdr *);
456 	nip6->ip6_src  = oip6->ip6_src;
457 	nip6->ip6_dst  = oip6->ip6_dst;
458 
459 	in6_clearscope(&oip6->ip6_src);
460 	in6_clearscope(&oip6->ip6_dst);
461 
462 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
463 	icmp6->icmp6_type = type;
464 	icmp6->icmp6_code = code;
465 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
466 
467 	/*
468 	 * icmp6_reflect() is designed to be in the input path.
469 	 * icmp6_error() can be called from both input and output path,
470 	 * and if we are in output path rcvif could contain bogus value.
471 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
472 	 * information in ip header (nip6).
473 	 */
474 	m_reset_rcvif(m);
475 
476 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
477 
478 	/* header order: IPv6 - ICMPv6 */
479 	icmp6_reflect(m, sizeof(struct ip6_hdr));
480 
481 	return;
482 
483 freeit:
484 	/*
485 	 * If we can't tell whether or not we can generate ICMP6, free it.
486 	 */
487 	m_freem(m);
488 }
489 
490 /*
491  * Process a received ICMP6 message.
492  */
493 static void
494 _icmp6_input(struct mbuf *m, int off, int proto)
495 {
496 	struct mbuf *n;
497 	struct ip6_hdr *ip6, *nip6;
498 	struct icmp6_hdr *icmp6, *nicmp6;
499 	int icmp6len = m->m_pkthdr.len - off;
500 	int code, sum;
501 	struct ifnet *rcvif;
502 	struct psref psref;
503 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
504 
505 	rcvif = m_get_rcvif_psref(m, &psref);
506 	if (__predict_false(rcvif == NULL))
507 		goto freeit;
508 
509 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
510 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
511 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
512 
513 	/*
514 	 * Locate icmp6 structure in mbuf, and check
515 	 * that not corrupted and of at least minimum length
516 	 */
517 
518 	if (icmp6len < sizeof(struct icmp6_hdr)) {
519 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
520 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
521 		goto freeit;
522 	}
523 
524 	if (m->m_len < sizeof(struct ip6_hdr)) {
525 		m = m_pullup(m, sizeof(struct ip6_hdr));
526 		if (m == NULL) {
527 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
528 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
529 			goto freeit;
530 		}
531 	}
532 
533 	ip6 = mtod(m, struct ip6_hdr *);
534 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
535 	if (icmp6 == NULL) {
536 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
537 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
538 		goto freeit;
539 	}
540 
541 	/*
542 	 * Enforce alignment requirements that are violated in
543 	 * some cases, see kern/50766 for details.
544 	 */
545 	if (ACCESSIBLE_POINTER(icmp6, struct ip6_hdr) == 0) {
546 		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
547 		if (m == NULL) {
548 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
549 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
550 			goto freeit;
551 		}
552 		ip6 = mtod(m, struct ip6_hdr *);
553 		icmp6 = (struct icmp6_hdr *)(mtod(m, char *) + off);
554 	}
555 	KASSERT(ACCESSIBLE_POINTER(icmp6, struct ip6_hdr));
556 
557 	/*
558 	 * calculate the checksum
559 	 */
560 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
561 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
562 		    icmp6->icmp6_type, sum, IN6_PRINT(ip6buf, &ip6->ip6_src));
563 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
564 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
565 		goto freeit;
566 	}
567 
568 #if defined(NFAITH) && 0 < NFAITH
569 	if (faithprefix(&ip6->ip6_dst)) {
570 		/*
571 		 * Deliver very specific ICMP6 type only.
572 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
573 		 * will not work.
574 		 */
575 		switch (icmp6->icmp6_type) {
576 		case ICMP6_DST_UNREACH:
577 		case ICMP6_PACKET_TOO_BIG:
578 		case ICMP6_TIME_EXCEEDED:
579 			break;
580 		default:
581 			goto freeit;
582 		}
583 	}
584 #endif
585 
586 	code = icmp6->icmp6_code;
587 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
588 
589 	switch (icmp6->icmp6_type) {
590 	case ICMP6_DST_UNREACH:
591 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
592 		switch (code) {
593 		case ICMP6_DST_UNREACH_NOROUTE:
594 			code = PRC_UNREACH_NET;
595 			break;
596 		case ICMP6_DST_UNREACH_ADMIN:
597 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
598 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
599 			break;
600 		case ICMP6_DST_UNREACH_ADDR:
601 			code = PRC_HOSTDEAD;
602 			break;
603 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
604 			/* I mean "source address was incorrect." */
605 			code = PRC_UNREACH_NET;
606 			break;
607 		case ICMP6_DST_UNREACH_NOPORT:
608 			code = PRC_UNREACH_PORT;
609 			break;
610 		default:
611 			goto badcode;
612 		}
613 		goto deliver;
614 
615 	case ICMP6_PACKET_TOO_BIG:
616 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
617 
618 		/*
619 		 * MTU is checked in icmp6_mtudisc.
620 		 */
621 		code = PRC_MSGSIZE;
622 
623 		/*
624 		 * Updating the path MTU will be done after examining
625 		 * intermediate extension headers.
626 		 */
627 		goto deliver;
628 
629 	case ICMP6_TIME_EXCEEDED:
630 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
631 		switch (code) {
632 		case ICMP6_TIME_EXCEED_TRANSIT:
633 			code = PRC_TIMXCEED_INTRANS;
634 			break;
635 		case ICMP6_TIME_EXCEED_REASSEMBLY:
636 			code = PRC_TIMXCEED_REASS;
637 			break;
638 		default:
639 			goto badcode;
640 		}
641 		goto deliver;
642 
643 	case ICMP6_PARAM_PROB:
644 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
645 		switch (code) {
646 		case ICMP6_PARAMPROB_NEXTHEADER:
647 			code = PRC_UNREACH_PROTOCOL;
648 			break;
649 		case ICMP6_PARAMPROB_HEADER:
650 		case ICMP6_PARAMPROB_OPTION:
651 			code = PRC_PARAMPROB;
652 			break;
653 		default:
654 			goto badcode;
655 		}
656 		goto deliver;
657 
658 	case ICMP6_ECHO_REQUEST:
659 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
660 		if (code != 0)
661 			goto badcode;
662 		/*
663 		 * Copy mbuf to send to two data paths: userland socket(s),
664 		 * and to the querier (echo reply).
665 		 * m: a copy for socket, n: a copy for querier
666 		 *
667 		 * If the first mbuf is shared, or the first mbuf is too short,
668 		 * copy the first part of the data into a fresh mbuf.
669 		 * Otherwise, we will wrongly overwrite both copies.
670 		 */
671 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
672 			/* Give up local */
673 			n = m;
674 			m = NULL;
675 		} else if (M_UNWRITABLE(n, off + sizeof(struct icmp6_hdr))) {
676 			struct mbuf *n0 = n;
677 
678 			/*
679 			 * Prepare an internal mbuf.  m_pullup() doesn't
680 			 * always copy the length we specified.
681 			 */
682 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
683 				/* Give up local */
684 				n = m;
685 				m = NULL;
686 			}
687 			m_freem(n0);
688 		}
689 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
690 		    sizeof(*nicmp6));
691 		if (nicmp6 == NULL)
692 			goto freeit;
693 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
694 		nicmp6->icmp6_code = 0;
695 		if (n) {
696 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
697 			icmp6s[ICMP6_STAT_REFLECT]++;
698 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
699 			ICMP6_STAT_PUTREF();
700 			icmp6_reflect(n, off);
701 		}
702 		if (!m)
703 			goto freeit;
704 		break;
705 
706 	case ICMP6_ECHO_REPLY:
707 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
708 		if (code != 0)
709 			goto badcode;
710 		break;
711 
712 	case MLD_LISTENER_QUERY:
713 	case MLD_LISTENER_REPORT:
714 		if (icmp6len < sizeof(struct mld_hdr))
715 			goto badlen;
716 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
717 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
718 		else
719 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
720 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
721 			/* give up local */
722 			mld_input(m, off);
723 			m = NULL;
724 			goto freeit;
725 		}
726 		mld_input(n, off);
727 		/* m stays. */
728 		break;
729 
730 	case MLD_LISTENER_DONE:
731 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
732 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
733 			goto badlen;
734 		break;		/* nothing to be done in kernel */
735 
736 	case MLD_MTRACE_RESP:
737 	case MLD_MTRACE:
738 		/* XXX: these two are experimental.  not officially defined. */
739 		/* XXX: per-interface statistics? */
740 		break;		/* just pass it to applications */
741 
742 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
743 	    {
744 		enum { WRU, FQDN } mode;
745 
746 		if (!icmp6_nodeinfo)
747 			break;
748 
749 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
750 			mode = WRU;
751 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
752 			mode = FQDN;
753 		else
754 			goto badlen;
755 
756 		if (mode == FQDN) {
757 			n = m_copypacket(m, M_DONTWAIT);
758 			if (n)
759 				n = ni6_input(n, off);
760 		} else {
761 			u_char *p;
762 			int maxhlen;
763 
764 			if ((icmp6_nodeinfo & 5) != 5)
765 				break;
766 
767 			if (code != 0)
768 				goto badcode;
769 			MGETHDR(n, M_DONTWAIT, m->m_type);
770 			if (n && ICMP6_MAXLEN > MHLEN) {
771 				MCLGET(n, M_DONTWAIT);
772 				if ((n->m_flags & M_EXT) == 0) {
773 					m_free(n);
774 					n = NULL;
775 				}
776 			}
777 			if (n == NULL) {
778 				/* Give up remote */
779 				break;
780 			}
781 			m_reset_rcvif(n);
782 			n->m_len = 0;
783 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
784 			if (maxhlen < 0) {
785 				m_free(n);
786 				break;
787 			}
788 			if (maxhlen > hostnamelen)
789 				maxhlen = hostnamelen;
790 			/*
791 			 * Copy IPv6 and ICMPv6 only.
792 			 */
793 			nip6 = mtod(n, struct ip6_hdr *);
794 			memcpy(nip6, ip6, sizeof(struct ip6_hdr));
795 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
796 			memcpy(nicmp6, icmp6, sizeof(struct icmp6_hdr));
797 
798 			p = (u_char *)(nicmp6 + 1);
799 			memset(p, 0, 4);
800 			memcpy(p + 4, hostname, maxhlen); /* meaningless TTL */
801 
802 			m_copy_pkthdr(n, m);
803 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
804 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
805 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
806 			nicmp6->icmp6_code = 0;
807 		}
808 		if (n) {
809 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
810 			icmp6s[ICMP6_STAT_REFLECT]++;
811 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
812 			ICMP6_STAT_PUTREF();
813 			icmp6_reflect(n, sizeof(struct ip6_hdr));
814 		}
815 		break;
816 	    }
817 
818 	case ICMP6_WRUREPLY:
819 		if (code != 0)
820 			goto badcode;
821 		break;
822 
823 	case ND_ROUTER_SOLICIT:
824 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
825 		/* FALLTHROUGH */
826 	case ND_ROUTER_ADVERT:
827 		if (icmp6->icmp6_type == ND_ROUTER_ADVERT)
828 			icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
829 		if (code != 0)
830 			goto badcode;
831 		if ((icmp6->icmp6_type == ND_ROUTER_SOLICIT &&
832 		    icmp6len < sizeof(struct nd_router_solicit)) ||
833 		    (icmp6->icmp6_type == ND_ROUTER_ADVERT &&
834 		    icmp6len < sizeof(struct nd_router_advert)))
835 			goto badlen;
836 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
837 			/* give up local */
838 			nd6_rtr_cache(m, off, icmp6len, icmp6->icmp6_type);
839 			m = NULL;
840 			goto freeit;
841 		}
842 		nd6_rtr_cache(n, off, icmp6len, icmp6->icmp6_type);
843 		/* m stays. */
844 		break;
845 
846 	case ND_NEIGHBOR_SOLICIT:
847 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
848 		if (code != 0)
849 			goto badcode;
850 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
851 			goto badlen;
852 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
853 			/* give up local */
854 			nd6_ns_input(m, off, icmp6len);
855 			m = NULL;
856 			goto freeit;
857 		}
858 		nd6_ns_input(n, off, icmp6len);
859 		/* m stays. */
860 		break;
861 
862 	case ND_NEIGHBOR_ADVERT:
863 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
864 		if (code != 0)
865 			goto badcode;
866 		if (icmp6len < sizeof(struct nd_neighbor_advert))
867 			goto badlen;
868 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
869 			/* give up local */
870 			nd6_na_input(m, off, icmp6len);
871 			m = NULL;
872 			goto freeit;
873 		}
874 		nd6_na_input(n, off, icmp6len);
875 		/* m stays. */
876 		break;
877 
878 	case ND_REDIRECT:
879 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
880 		if (code != 0)
881 			goto badcode;
882 		if (icmp6len < sizeof(struct nd_redirect))
883 			goto badlen;
884 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
885 			/* give up local */
886 			icmp6_redirect_input(m, off);
887 			m = NULL;
888 			goto freeit;
889 		}
890 		icmp6_redirect_input(n, off);
891 		/* m stays. */
892 		break;
893 
894 	case ICMP6_ROUTER_RENUMBERING:
895 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
896 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
897 			goto badcode;
898 		if (icmp6len < sizeof(struct icmp6_router_renum))
899 			goto badlen;
900 		break;
901 
902 	default:
903 		nd6log(LOG_DEBUG,
904 		    "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
905 		    icmp6->icmp6_type,
906 		    IN6_PRINT(ip6buf, &ip6->ip6_src),
907 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst),
908 		    rcvif ? rcvif->if_index : 0);
909 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
910 			/* ICMPv6 error: MUST deliver it by spec... */
911 			code = PRC_NCMDS;
912 			/* deliver */
913 		} else {
914 			/* ICMPv6 informational: MUST not deliver */
915 			break;
916 		}
917 	deliver:
918 		if (icmp6_notify_error(m, off, icmp6len, code)) {
919 			/* In this case, m should've been freed. */
920 			m_put_rcvif_psref(rcvif, &psref);
921 			return;
922 		}
923 		break;
924 
925 	badcode:
926 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
927 		break;
928 
929 	badlen:
930 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
931 		break;
932 	}
933 	m_put_rcvif_psref(rcvif, &psref);
934 
935 	/* deliver the packet to appropriate sockets */
936 	icmp6_rip6_input(&m, off);
937 
938 	return;
939 
940 freeit:
941 	m_put_rcvif_psref(rcvif, &psref);
942 	m_freem(m);
943 	return;
944 }
945 
946 int
947 icmp6_input(struct mbuf **mp, int *offp, int proto)
948 {
949 
950 	wqinput_input(icmp6_wqinput, *mp, *offp, proto);
951 
952 	return IPPROTO_DONE;
953 }
954 
955 static int
956 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
957 {
958 	struct icmp6_hdr *icmp6;
959 	struct ip6_hdr *eip6;
960 	u_int32_t notifymtu;
961 	struct sockaddr_in6 icmp6src, icmp6dst;
962 
963 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
964 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
965 		goto freeit;
966 	}
967 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
968 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
969 	if (icmp6 == NULL) {
970 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
971 		return (-1);
972 	}
973 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
974 
975 	/* Detect the upper level protocol */
976 	{
977 		void *(*ctlfunc)(int, const struct sockaddr *, void *);
978 		u_int8_t nxt = eip6->ip6_nxt;
979 		int eoff = off + sizeof(struct icmp6_hdr) +
980 			sizeof(struct ip6_hdr);
981 		struct ip6ctlparam ip6cp;
982 		struct in6_addr *finaldst = NULL;
983 		int icmp6type = icmp6->icmp6_type;
984 		struct ip6_frag *fh;
985 		struct ip6_rthdr *rth;
986 		struct ifnet *rcvif;
987 		int s;
988 
989 		while (1) { /* XXX: should avoid infinite loop explicitly? */
990 			struct ip6_ext *eh;
991 
992 			switch (nxt) {
993 			case IPPROTO_HOPOPTS:
994 			case IPPROTO_DSTOPTS:
995 			case IPPROTO_AH:
996 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
997 				    eoff, sizeof(*eh));
998 				if (eh == NULL) {
999 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1000 					return (-1);
1001 				}
1002 
1003 				if (nxt == IPPROTO_AH)
1004 					eoff += (eh->ip6e_len + 2) << 2;
1005 				else
1006 					eoff += (eh->ip6e_len + 1) << 3;
1007 				nxt = eh->ip6e_nxt;
1008 				break;
1009 			case IPPROTO_ROUTING:
1010 				/* Ignore the option. */
1011 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1012 				    eoff, sizeof(*rth));
1013 				if (rth == NULL) {
1014 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1015 					return (-1);
1016 				}
1017 
1018 				eoff += (rth->ip6r_len + 1) << 3;
1019 				nxt = rth->ip6r_nxt;
1020 				break;
1021 			case IPPROTO_FRAGMENT:
1022 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1023 				    eoff, sizeof(*fh));
1024 				if (fh == NULL) {
1025 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1026 					return (-1);
1027 				}
1028 				/*
1029 				 * Data after a fragment header is meaningless
1030 				 * unless it is the first fragment, but
1031 				 * we'll go to the notify label for path MTU
1032 				 * discovery.
1033 				 */
1034 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1035 					goto notify;
1036 
1037 				eoff += sizeof(struct ip6_frag);
1038 				nxt = fh->ip6f_nxt;
1039 				break;
1040 			default:
1041 				/*
1042 				 * This case includes ESP and the No Next
1043 				 * Header.  In such cases going to the notify
1044 				 * label does not have any meaning
1045 				 * (i.e. ctlfunc will be NULL), but we go
1046 				 * anyway since we might have to update
1047 				 * path MTU information.
1048 				 */
1049 				goto notify;
1050 			}
1051 		}
1052 	  notify:
1053 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1054 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1055 		if (icmp6 == NULL) {
1056 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1057 			return (-1);
1058 		}
1059 
1060 		/*
1061 		 * retrieve parameters from the inner IPv6 header, and convert
1062 		 * them into sockaddr structures.
1063 		 * XXX: there is no guarantee that the source or destination
1064 		 * addresses of the inner packet are in the same scope zone as
1065 		 * the addresses of the icmp packet.  But there is no other
1066 		 * way to determine the zone.
1067 		 */
1068 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1069 
1070 		rcvif = m_get_rcvif(m, &s);
1071 		if (__predict_false(rcvif == NULL))
1072 			goto freeit;
1073 		sockaddr_in6_init(&icmp6dst,
1074 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1075 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1076 			m_put_rcvif(rcvif, &s);
1077 			goto freeit;
1078 		}
1079 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1080 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1081 			m_put_rcvif(rcvif, &s);
1082 			goto freeit;
1083 		}
1084 		m_put_rcvif(rcvif, &s);
1085 
1086 		icmp6src.sin6_flowinfo =
1087 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1088 
1089 		if (finaldst == NULL)
1090 			finaldst = &eip6->ip6_dst;
1091 		ip6cp.ip6c_m = m;
1092 		ip6cp.ip6c_icmp6 = icmp6;
1093 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1094 		ip6cp.ip6c_off = eoff;
1095 		ip6cp.ip6c_finaldst = finaldst;
1096 		ip6cp.ip6c_src = &icmp6src;
1097 		ip6cp.ip6c_nxt = nxt;
1098 
1099 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1100 			notifymtu = ntohl(icmp6->icmp6_mtu);
1101 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1102 		}
1103 
1104 		ctlfunc = inet6sw[ip6_protox[nxt]].pr_ctlinput;
1105 		if (ctlfunc) {
1106 			(void)(*ctlfunc)(code, sin6tosa(&icmp6dst), &ip6cp);
1107 		}
1108 	}
1109 	return (0);
1110 
1111 freeit:
1112 	m_freem(m);
1113 	return (-1);
1114 }
1115 
1116 void
1117 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1118 {
1119 	unsigned long rtcount;
1120 	struct icmp6_mtudisc_callback *mc;
1121 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1122 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1123 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1124 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1125 	struct rtentry *rt = NULL;
1126 	struct sockaddr_in6 sin6;
1127 	struct ifnet *rcvif;
1128 	int s;
1129 
1130 	/*
1131 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1132 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1133 	 * In that one case, the MTU might be less than 1280.
1134 	 */
1135 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1136 		/* is the mtu even sane? */
1137 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1138 			return;
1139 		if (!validated)
1140 			return;
1141 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1142 	}
1143 
1144 	/*
1145 	 * allow non-validated cases if memory is plenty, to make traffic
1146 	 * from non-connected pcb happy.
1147 	 */
1148 	mutex_enter(&icmp6_mtx);
1149 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1150 	if (validated) {
1151 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) {
1152 			mutex_exit(&icmp6_mtx);
1153 			return;
1154 		} else if (0 <= icmp6_mtudisc_lowat &&
1155 		    rtcount > icmp6_mtudisc_lowat) {
1156 			/*
1157 			 * XXX nuke a victim, install the new one.
1158 			 */
1159 		}
1160 	} else {
1161 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) {
1162 			mutex_exit(&icmp6_mtx);
1163 			return;
1164 		}
1165 	}
1166 	mutex_exit(&icmp6_mtx);
1167 
1168 	memset(&sin6, 0, sizeof(sin6));
1169 	sin6.sin6_family = PF_INET6;
1170 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1171 	sin6.sin6_addr = *dst;
1172 	rcvif = m_get_rcvif(m, &s);
1173 	if (__predict_false(rcvif == NULL))
1174 		return;
1175 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1176 		m_put_rcvif(rcvif, &s);
1177 		return;
1178 	}
1179 	m_put_rcvif(rcvif, &s);
1180 
1181 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1182 
1183 	if (rt && (rt->rt_flags & RTF_HOST) &&
1184 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1185 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1186 		if (mtu < rt->rt_ifp->if_mtu) {
1187 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1188 			rt->rt_rmx.rmx_mtu = mtu;
1189 		}
1190 	}
1191 	if (rt) {
1192 		rt_unref(rt);
1193 	}
1194 
1195 	/*
1196 	 * Notify protocols that the MTU for this destination
1197 	 * has changed.
1198 	 */
1199 	mutex_enter(&icmp6_mtx);
1200 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1201 	     mc = LIST_NEXT(mc, mc_list))
1202 		(*mc->mc_func)(&sin6.sin6_addr);
1203 	mutex_exit(&icmp6_mtx);
1204 }
1205 
1206 /*
1207  * Process a Node Information Query packet, based on
1208  * draft-ietf-ipngwg-icmp-name-lookups-07.
1209  *
1210  * Spec incompatibilities:
1211  * - IPv6 Subject address handling
1212  * - IPv4 Subject address handling support missing
1213  * - Proxy reply (answer even if it's not for me)
1214  * - joins NI group address at in6_ifattach() time only, does not cope
1215  *   with hostname changes by sethostname(3)
1216  */
1217 static struct mbuf *
1218 ni6_input(struct mbuf *m, int off)
1219 {
1220 	struct icmp6_nodeinfo *ni6, *nni6;
1221 	struct mbuf *n = NULL;
1222 	u_int16_t qtype;
1223 	int subjlen;
1224 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1225 	struct ni_reply_fqdn *fqdn;
1226 	int addrs;		/* for NI_QTYPE_NODEADDR */
1227 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1228 	struct sockaddr_in6 sin6; /* ip6_dst */
1229 	struct in6_addr in6_subj; /* subject address */
1230 	struct ip6_hdr *ip6;
1231 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1232 	char *subj = NULL;
1233 	struct ifnet *rcvif;
1234 	int s, ss;
1235 	struct ifaddr *ifa;
1236 	struct psref psref;
1237 
1238 	ip6 = mtod(m, struct ip6_hdr *);
1239 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1240 	if (ni6 == NULL) {
1241 		/* m is already reclaimed */
1242 		return NULL;
1243 	}
1244 	KASSERT((m->m_flags & M_PKTHDR) != 0);
1245 
1246 	/*
1247 	 * Validate IPv6 destination address.
1248 	 *
1249 	 * The Responder must discard the Query without further processing
1250 	 * unless it is one of the Responder's unicast or anycast addresses, or
1251 	 * a link-local scope multicast address which the Responder has joined.
1252 	 * [icmp-name-lookups-07, Section 4.]
1253 	 */
1254 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1255 	/* XXX scopeid */
1256 	ss = pserialize_read_enter();
1257 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1258 	if (ifa != NULL) {
1259 		; /* unicast/anycast, fine */
1260 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) {
1261 		; /* link-local multicast, fine */
1262 	} else {
1263 		pserialize_read_exit(ss);
1264 		goto bad;
1265 	}
1266 	pserialize_read_exit(ss);
1267 
1268 	/* validate query Subject field. */
1269 	qtype = ntohs(ni6->ni_qtype);
1270 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1271 	switch (qtype) {
1272 	case NI_QTYPE_NOOP:
1273 	case NI_QTYPE_SUPTYPES:
1274 		/* 07 draft */
1275 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1276 			break;
1277 		/* FALLTHROUGH */
1278 	case NI_QTYPE_FQDN:
1279 	case NI_QTYPE_NODEADDR:
1280 	case NI_QTYPE_IPV4ADDR:
1281 		switch (ni6->ni_code) {
1282 		case ICMP6_NI_SUBJ_IPV6:
1283 #if ICMP6_NI_SUBJ_IPV6 != 0
1284 		case 0:
1285 #endif
1286 			/*
1287 			 * backward compatibility - try to accept 03 draft
1288 			 * format, where no Subject is present.
1289 			 */
1290 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1291 			    subjlen == 0) {
1292 				oldfqdn++;
1293 				break;
1294 			}
1295 #if ICMP6_NI_SUBJ_IPV6 != 0
1296 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1297 				goto bad;
1298 #endif
1299 
1300 			if (subjlen != sizeof(sin6.sin6_addr))
1301 				goto bad;
1302 
1303 			/*
1304 			 * Validate Subject address.
1305 			 *
1306 			 * Not sure what exactly "address belongs to the node"
1307 			 * means in the spec, is it just unicast, or what?
1308 			 *
1309 			 * At this moment we consider Subject address as
1310 			 * "belong to the node" if the Subject address equals
1311 			 * to the IPv6 destination address; validation for
1312 			 * IPv6 destination address should have done enough
1313 			 * check for us.
1314 			 *
1315 			 * We do not do proxy at this moment.
1316 			 */
1317 			/* m_pulldown instead of copy? */
1318 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1319 			    subjlen, (void *)&in6_subj);
1320 			rcvif = m_get_rcvif(m, &s);
1321 			if (__predict_false(rcvif == NULL))
1322 				goto bad;
1323 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1324 				m_put_rcvif(rcvif, &s);
1325 				goto bad;
1326 			}
1327 			m_put_rcvif(rcvif, &s);
1328 
1329 			subj = (char *)&in6_subj;
1330 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1331 				break;
1332 
1333 			/*
1334 			 * XXX if we are to allow other cases, we should really
1335 			 * be careful about scope here.
1336 			 * basically, we should disallow queries toward IPv6
1337 			 * destination X with subject Y, if scope(X) > scope(Y).
1338 			 * if we allow scope(X) > scope(Y), it will result in
1339 			 * information leakage across scope boundary.
1340 			 */
1341 			goto bad;
1342 
1343 		case ICMP6_NI_SUBJ_FQDN:
1344 			/*
1345 			 * Validate Subject name with gethostname(3).
1346 			 *
1347 			 * The behavior may need some debate, since:
1348 			 * - we are not sure if the node has FQDN as
1349 			 *   hostname (returned by gethostname(3)).
1350 			 * - the code does wildcard match for truncated names.
1351 			 *   however, we are not sure if we want to perform
1352 			 *   wildcard match, if gethostname(3) side has
1353 			 *   truncated hostname.
1354 			 */
1355 			n = ni6_nametodns(hostname, hostnamelen, 0);
1356 			if (!n || n->m_next || n->m_len == 0)
1357 				goto bad;
1358 			IP6_EXTHDR_GET(subj, char *, m,
1359 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1360 			if (subj == NULL)
1361 				goto bad;
1362 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1363 			    n->m_len)) {
1364 				goto bad;
1365 			}
1366 			m_freem(n);
1367 			n = NULL;
1368 			break;
1369 
1370 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1371 		default:
1372 			goto bad;
1373 		}
1374 		break;
1375 	}
1376 
1377 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1378 	switch (qtype) {
1379 	case NI_QTYPE_FQDN:
1380 		if ((icmp6_nodeinfo & 1) == 0)
1381 			goto bad;
1382 		break;
1383 	case NI_QTYPE_NODEADDR:
1384 	case NI_QTYPE_IPV4ADDR:
1385 		if ((icmp6_nodeinfo & 2) == 0)
1386 			goto bad;
1387 		break;
1388 	}
1389 
1390 	/* guess reply length */
1391 	switch (qtype) {
1392 	case NI_QTYPE_NOOP:
1393 		break;		/* no reply data */
1394 	case NI_QTYPE_SUPTYPES:
1395 		replylen += sizeof(u_int32_t);
1396 		break;
1397 	case NI_QTYPE_FQDN:
1398 		/* will append an mbuf */
1399 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1400 		break;
1401 	case NI_QTYPE_NODEADDR:
1402 		addrs = ni6_addrs(ni6, &ifp, subj, &psref);
1403 		replylen += addrs *
1404 		    (sizeof(struct in6_addr) + sizeof(u_int32_t));
1405 		if (replylen > MCLBYTES)
1406 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1407 		break;
1408 	case NI_QTYPE_IPV4ADDR:
1409 		/* unsupported - should respond with unknown Qtype? */
1410 		goto bad;
1411 	default:
1412 		/*
1413 		 * XXX: We must return a reply with the ICMP6 code
1414 		 * `unknown Qtype' in this case.  However we regard the case
1415 		 * as an FQDN query for backward compatibility.
1416 		 * Older versions set a random value to this field,
1417 		 * so it rarely varies in the defined qtypes.
1418 		 * But the mechanism is not reliable...
1419 		 * maybe we should obsolete older versions.
1420 		 */
1421 		qtype = NI_QTYPE_FQDN;
1422 		/* will append an mbuf */
1423 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1424 		oldfqdn++;
1425 		break;
1426 	}
1427 
1428 	/* allocate an mbuf to reply. */
1429 	MGETHDR(n, M_DONTWAIT, m->m_type);
1430 	if (n == NULL) {
1431 		goto bad;
1432 	}
1433 	m_move_pkthdr(n, m);
1434 	if (replylen > MHLEN) {
1435 		if (replylen > MCLBYTES) {
1436 			/*
1437 			 * XXX: should we try to allocate more? But MCLBYTES
1438 			 * is probably much larger than IPV6_MMTU...
1439 			 */
1440 			goto bad;
1441 		}
1442 		MCLGET(n, M_DONTWAIT);
1443 		if ((n->m_flags & M_EXT) == 0) {
1444 			goto bad;
1445 		}
1446 	}
1447 	n->m_pkthdr.len = n->m_len = replylen;
1448 
1449 	/* copy mbuf header and IPv6 + Node Information base headers */
1450 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1451 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1452 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1453 
1454 	/* qtype dependent procedure */
1455 	switch (qtype) {
1456 	case NI_QTYPE_NOOP:
1457 		nni6->ni_code = ICMP6_NI_SUCCESS;
1458 		nni6->ni_flags = 0;
1459 		break;
1460 	case NI_QTYPE_SUPTYPES:
1461 	{
1462 		u_int32_t v;
1463 		nni6->ni_code = ICMP6_NI_SUCCESS;
1464 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1465 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1466 		v = (u_int32_t)htonl(0x0000000f);
1467 		memcpy(nni6 + 1, &v, sizeof(u_int32_t));
1468 		break;
1469 	}
1470 	case NI_QTYPE_FQDN:
1471 		nni6->ni_code = ICMP6_NI_SUCCESS;
1472 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1473 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1474 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1475 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1476 		/*
1477 		 * XXX do we really have FQDN in variable "hostname"?
1478 		 */
1479 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1480 		if (n->m_next == NULL)
1481 			goto bad;
1482 		/* XXX we assume that n->m_next is not a chain */
1483 		if (n->m_next->m_next != NULL)
1484 			goto bad;
1485 		n->m_pkthdr.len += n->m_next->m_len;
1486 		break;
1487 	case NI_QTYPE_NODEADDR:
1488 	{
1489 		int lenlim, copied;
1490 
1491 		nni6->ni_code = ICMP6_NI_SUCCESS;
1492 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1493 		    sizeof(struct icmp6_nodeinfo);
1494 		lenlim = M_TRAILINGSPACE(n);
1495 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1496 		if_put(ifp, &psref);
1497 		ifp = NULL;
1498 		/* update mbuf length */
1499 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1500 		    sizeof(struct icmp6_nodeinfo) + copied;
1501 		break;
1502 	}
1503 	default:
1504 		panic("%s: impossible", __func__);
1505 		break;
1506 	}
1507 
1508 	nni6->ni_type = ICMP6_NI_REPLY;
1509 	m_freem(m);
1510 	return n;
1511 
1512 bad:
1513 	if_put(ifp, &psref);
1514 	m_freem(m);
1515 	if (n)
1516 		m_freem(n);
1517 	return NULL;
1518 }
1519 
1520 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1521 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1522 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1523 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1524 
1525 /*
1526  * make a mbuf with DNS-encoded string.  no compression support.
1527  *
1528  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1529  * treated as truncated name (two \0 at the end).  this is a wild guess.
1530  *
1531  * old - return pascal string if non-zero
1532  */
1533 static struct mbuf *
1534 ni6_nametodns(const char *name, int namelen, int old)
1535 {
1536 	struct mbuf *m;
1537 	char *cp, *ep;
1538 	const char *p, *q;
1539 	int i, len, nterm;
1540 
1541 	if (old)
1542 		len = namelen + 1;
1543 	else
1544 		len = MCLBYTES;
1545 
1546 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1547 	MGET(m, M_DONTWAIT, MT_DATA);
1548 	if (m && len > MLEN) {
1549 		MCLGET(m, M_DONTWAIT);
1550 		if ((m->m_flags & M_EXT) == 0)
1551 			goto fail;
1552 	}
1553 	if (!m)
1554 		goto fail;
1555 	m->m_next = NULL;
1556 
1557 	if (old) {
1558 		m->m_len = len;
1559 		*mtod(m, char *) = namelen;
1560 		memcpy(mtod(m, char *) + 1, name, namelen);
1561 		return m;
1562 	} else {
1563 		m->m_len = 0;
1564 		cp = mtod(m, char *);
1565 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1566 
1567 		/* if not certain about my name, return empty buffer */
1568 		if (namelen == 0)
1569 			return m;
1570 
1571 		/*
1572 		 * guess if it looks like shortened hostname, or FQDN.
1573 		 * shortened hostname needs two trailing "\0".
1574 		 */
1575 		i = 0;
1576 		for (p = name; p < name + namelen; p++) {
1577 			if (*p == '.')
1578 				i++;
1579 		}
1580 		if (i < 2)
1581 			nterm = 2;
1582 		else
1583 			nterm = 1;
1584 
1585 		p = name;
1586 		while (cp < ep && p < name + namelen) {
1587 			i = 0;
1588 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1589 				i++;
1590 			/* result does not fit into mbuf */
1591 			if (cp + i + 1 >= ep)
1592 				goto fail;
1593 			/*
1594 			 * DNS label length restriction, RFC1035 page 8.
1595 			 * "i == 0" case is included here to avoid returning
1596 			 * 0-length label on "foo..bar".
1597 			 */
1598 			if (i <= 0 || i >= 64)
1599 				goto fail;
1600 			*cp++ = i;
1601 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1602 				goto fail;
1603 			while (i > 0) {
1604 				if (!isalnum(*p) && *p != '-')
1605 					goto fail;
1606 				if (isupper(*p)) {
1607 					*cp++ = tolower(*p);
1608 					p++;
1609 				} else
1610 					*cp++ = *p++;
1611 				i--;
1612 			}
1613 			p = q;
1614 			if (p < name + namelen && *p == '.')
1615 				p++;
1616 		}
1617 		/* termination */
1618 		if (cp + nterm >= ep)
1619 			goto fail;
1620 		while (nterm-- > 0)
1621 			*cp++ = '\0';
1622 		m->m_len = cp - mtod(m, char *);
1623 		return m;
1624 	}
1625 
1626 	panic("should not reach here");
1627 	/* NOTREACHED */
1628 
1629 fail:
1630 	if (m)
1631 		m_freem(m);
1632 	return NULL;
1633 }
1634 
1635 /*
1636  * check if two DNS-encoded string matches.  takes care of truncated
1637  * form (with \0\0 at the end).  no compression support.
1638  * XXX upper/lowercase match (see RFC2065)
1639  */
1640 static int
1641 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1642 {
1643 	const char *a0, *b0;
1644 	int l;
1645 
1646 	/* simplest case - need validation? */
1647 	if (alen == blen && memcmp(a, b, alen) == 0)
1648 		return 1;
1649 
1650 	a0 = a;
1651 	b0 = b;
1652 
1653 	/* termination is mandatory */
1654 	if (alen < 2 || blen < 2)
1655 		return 0;
1656 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1657 		return 0;
1658 	alen--;
1659 	blen--;
1660 
1661 	while (a - a0 < alen && b - b0 < blen) {
1662 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1663 			return 0;
1664 
1665 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1666 			return 0;
1667 		/* we don't support compression yet */
1668 		if (a[0] >= 64 || b[0] >= 64)
1669 			return 0;
1670 
1671 		/* truncated case */
1672 		if (a[0] == 0 && a - a0 == alen - 1)
1673 			return 1;
1674 		if (b[0] == 0 && b - b0 == blen - 1)
1675 			return 1;
1676 		if (a[0] == 0 || b[0] == 0)
1677 			return 0;
1678 
1679 		if (a[0] != b[0])
1680 			return 0;
1681 		l = a[0];
1682 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1683 			return 0;
1684 		if (memcmp(a + 1, b + 1, l) != 0)
1685 			return 0;
1686 
1687 		a += 1 + l;
1688 		b += 1 + l;
1689 	}
1690 
1691 	if (a - a0 == alen && b - b0 == blen)
1692 		return 1;
1693 	else
1694 		return 0;
1695 }
1696 
1697 /*
1698  * calculate the number of addresses to be returned in the node info reply.
1699  */
1700 static int
1701 ni6_addrs(struct icmp6_nodeinfo *ni6, struct ifnet **ifpp, char *subj,
1702     struct psref *psref)
1703 {
1704 	struct ifnet *ifp;
1705 	struct in6_ifaddr *ia6;
1706 	struct ifaddr *ifa;
1707 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1708 	int addrs = 0, addrsofif, iffound = 0;
1709 	int niflags = ni6->ni_flags;
1710 	int s;
1711 
1712 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1713 		switch (ni6->ni_code) {
1714 		case ICMP6_NI_SUBJ_IPV6:
1715 			if (subj == NULL) /* must be impossible... */
1716 				return 0;
1717 			subj_ip6 = (struct sockaddr_in6 *)subj;
1718 			break;
1719 		default:
1720 			/*
1721 			 * XXX: we only support IPv6 subject address for
1722 			 * this Qtype.
1723 			 */
1724 			return 0;
1725 		}
1726 	}
1727 
1728 	s = pserialize_read_enter();
1729 	IFNET_READER_FOREACH(ifp) {
1730 		addrsofif = 0;
1731 		IFADDR_READER_FOREACH(ifa, ifp) {
1732 			if (ifa->ifa_addr->sa_family != AF_INET6)
1733 				continue;
1734 			ia6 = (struct in6_ifaddr *)ifa;
1735 
1736 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1737 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1738 			     &ia6->ia_addr.sin6_addr))
1739 				iffound = 1;
1740 
1741 			/*
1742 			 * IPv4-mapped addresses can only be returned by a
1743 			 * Node Information proxy, since they represent
1744 			 * addresses of IPv4-only nodes, which perforce do
1745 			 * not implement this protocol.
1746 			 * [icmp-name-lookups-07, Section 5.4]
1747 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1748 			 * this function at this moment.
1749 			 */
1750 
1751 			/* What do we have to do about ::1? */
1752 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1753 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1754 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1755 					continue;
1756 				break;
1757 			case IPV6_ADDR_SCOPE_SITELOCAL:
1758 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1759 					continue;
1760 				break;
1761 			case IPV6_ADDR_SCOPE_GLOBAL:
1762 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1763 					continue;
1764 				break;
1765 			default:
1766 				continue;
1767 			}
1768 
1769 			/*
1770 			 * check if anycast is okay.
1771 			 * XXX: just experimental.  not in the spec.
1772 			 */
1773 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1774 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1775 				continue; /* we need only unicast addresses */
1776 
1777 			addrsofif++; /* count the address */
1778 		}
1779 		if (iffound) {
1780 			if_acquire(ifp, psref);
1781 			pserialize_read_exit(s);
1782 			*ifpp = ifp;
1783 			return addrsofif;
1784 		}
1785 
1786 		addrs += addrsofif;
1787 	}
1788 	pserialize_read_exit(s);
1789 
1790 	return addrs;
1791 }
1792 
1793 static int
1794 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1795 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1796 	int resid)
1797 {
1798 	struct ifnet *ifp;
1799 	struct in6_ifaddr *ia6;
1800 	struct ifaddr *ifa;
1801 	struct ifnet *ifp_dep = NULL;
1802 	int copied = 0, allow_deprecated = 0;
1803 	u_char *cp = (u_char *)(nni6 + 1);
1804 	int niflags = ni6->ni_flags;
1805 	u_int32_t ltime;
1806 	int s;
1807 
1808 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1809 		return 0;	/* needless to copy */
1810 
1811 	s = pserialize_read_enter();
1812 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1813 again:
1814 
1815 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1816 	{
1817 		IFADDR_READER_FOREACH(ifa, ifp) {
1818 			if (ifa->ifa_addr->sa_family != AF_INET6)
1819 				continue;
1820 			ia6 = (struct in6_ifaddr *)ifa;
1821 
1822 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1823 			    allow_deprecated == 0) {
1824 				/*
1825 				 * prefererred address should be put before
1826 				 * deprecated addresses.
1827 				 */
1828 
1829 				/* record the interface for later search */
1830 				if (ifp_dep == NULL)
1831 					ifp_dep = ifp;
1832 
1833 				continue;
1834 			}
1835 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1836 				 allow_deprecated != 0)
1837 				continue; /* we now collect deprecated addrs */
1838 
1839 			/* What do we have to do about ::1? */
1840 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1841 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1842 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1843 					continue;
1844 				break;
1845 			case IPV6_ADDR_SCOPE_SITELOCAL:
1846 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1847 					continue;
1848 				break;
1849 			case IPV6_ADDR_SCOPE_GLOBAL:
1850 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1851 					continue;
1852 				break;
1853 			default:
1854 				continue;
1855 			}
1856 
1857 			/*
1858 			 * check if anycast is okay.
1859 			 * XXX: just experimental.  not in the spec.
1860 			 */
1861 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1862 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1863 				continue;
1864 
1865 			/* now we can copy the address */
1866 			if (resid < sizeof(struct in6_addr) +
1867 			    sizeof(u_int32_t)) {
1868 				/*
1869 				 * We give up much more copy.
1870 				 * Set the truncate flag and return.
1871 				 */
1872 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1873 				goto out;
1874 			}
1875 
1876 			/*
1877 			 * Set the TTL of the address.
1878 			 * The TTL value should be one of the following
1879 			 * according to the specification:
1880 			 *
1881 			 * 1. The remaining lifetime of a DHCP lease on the
1882 			 *    address, or
1883 			 * 2. The remaining Valid Lifetime of a prefix from
1884 			 *    which the address was derived through Stateless
1885 			 *    Autoconfiguration.
1886 			 *
1887 			 * Note that we currently do not support stateful
1888 			 * address configuration by DHCPv6, so the former
1889 			 * case can't happen.
1890 			 *
1891 			 * TTL must be 2^31 > TTL >= 0.
1892 			 */
1893 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1894 				ltime = ND6_INFINITE_LIFETIME;
1895 			else {
1896 				if (ia6->ia6_lifetime.ia6t_expire >
1897 				    time_uptime)
1898 					ltime = ia6->ia6_lifetime.ia6t_expire -
1899 					    time_uptime;
1900 				else
1901 					ltime = 0;
1902 			}
1903 			if (ltime > 0x7fffffff)
1904 				ltime = 0x7fffffff;
1905 			ltime = htonl(ltime);
1906 
1907 			memcpy(cp, &ltime, sizeof(u_int32_t));
1908 			cp += sizeof(u_int32_t);
1909 
1910 			/* copy the address itself */
1911 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1912 			      sizeof(struct in6_addr));
1913 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1914 			cp += sizeof(struct in6_addr);
1915 
1916 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1917 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1918 		}
1919 		if (ifp0)	/* we need search only on the specified IF */
1920 			break;
1921 	}
1922 
1923 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1924 		ifp = ifp_dep;
1925 		allow_deprecated = 1;
1926 
1927 		goto again;
1928 	}
1929 out:
1930 	pserialize_read_exit(s);
1931 	return copied;
1932 }
1933 
1934 /*
1935  * XXX almost dup'ed code with rip6_input.
1936  */
1937 static int
1938 icmp6_rip6_input(struct mbuf **mp, int off)
1939 {
1940 	struct mbuf *m = *mp;
1941 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1942 	struct inpcb_hdr *inph;
1943 	struct in6pcb *in6p;
1944 	struct in6pcb *last = NULL;
1945 	struct sockaddr_in6 rip6src;
1946 	struct icmp6_hdr *icmp6;
1947 	struct mbuf *n, *opts = NULL;
1948 
1949 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1950 	if (icmp6 == NULL) {
1951 		/* m is already reclaimed */
1952 		return IPPROTO_DONE;
1953 	}
1954 
1955 	/*
1956 	 * XXX: the address may have embedded scope zone ID, which should be
1957 	 * hidden from applications.
1958 	 */
1959 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1960 	if (sa6_recoverscope(&rip6src)) {
1961 		m_freem(m);
1962 		return IPPROTO_DONE;
1963 	}
1964 
1965 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1966 		in6p = (struct in6pcb *)inph;
1967 		if (in6p->in6p_af != AF_INET6)
1968 			continue;
1969 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1970 			continue;
1971 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1972 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1973 			continue;
1974 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1975 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1976 			continue;
1977 		if (in6p->in6p_icmp6filt &&
1978 		    ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1979 		    in6p->in6p_icmp6filt))
1980 			continue;
1981 
1982 		if (last == NULL) {
1983 			;
1984 		}
1985 #ifdef IPSEC
1986 		else if (ipsec_used && ipsec_in_reject(m, last)) {
1987 			/* do not inject data into pcb */
1988 		}
1989 #endif
1990 		else if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1991 			if (last->in6p_flags & IN6P_CONTROLOPTS)
1992 				ip6_savecontrol(last, &opts, ip6, n);
1993 			/* strip intermediate headers */
1994 			m_adj(n, off);
1995 			if (sbappendaddr(&last->in6p_socket->so_rcv,
1996 			    sin6tosa(&rip6src), n, opts) == 0) {
1997 				soroverflow(last->in6p_socket);
1998 				m_freem(n);
1999 				if (opts)
2000 					m_freem(opts);
2001 			} else {
2002 				sorwakeup(last->in6p_socket);
2003 			}
2004 			opts = NULL;
2005 		}
2006 
2007 		last = in6p;
2008 	}
2009 
2010 #ifdef IPSEC
2011 	if (ipsec_used && last && ipsec_in_reject(m, last)) {
2012 		m_freem(m);
2013 		IP6_STATDEC(IP6_STAT_DELIVERED);
2014 		/* do not inject data into pcb */
2015 	} else
2016 #endif
2017 	if (last) {
2018 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2019 			ip6_savecontrol(last, &opts, ip6, m);
2020 		/* strip intermediate headers */
2021 		m_adj(m, off);
2022 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2023 		    sin6tosa(&rip6src), m, opts) == 0) {
2024 			soroverflow(last->in6p_socket);
2025 			m_freem(m);
2026 			if (opts)
2027 				m_freem(opts);
2028 		} else {
2029 			sorwakeup(last->in6p_socket);
2030 		}
2031 	} else {
2032 		m_freem(m);
2033 		IP6_STATDEC(IP6_STAT_DELIVERED);
2034 	}
2035 	return IPPROTO_DONE;
2036 }
2037 
2038 /*
2039  * Reflect the ip6 packet back to the source.
2040  * OFF points to the icmp6 header, counted from the top of the mbuf.
2041  *
2042  * Note: RFC 1885 required that an echo reply should be truncated if it
2043  * did not fit in with (return) path MTU, and KAME code supported the
2044  * behavior.  However, as a clarification after the RFC, this limitation
2045  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2046  * old behavior, with a (non-default) ifdef block, while the new version of
2047  * the spec was an internet-draft status, and even after the new RFC was
2048  * published.  But it would rather make sense to clean the obsoleted part
2049  * up, and to make the code simpler at this stage.
2050  */
2051 static void
2052 icmp6_reflect(struct mbuf *m, size_t off)
2053 {
2054 	struct ip6_hdr *ip6;
2055 	struct icmp6_hdr *icmp6;
2056 	const struct in6_ifaddr *ia;
2057 	const struct ip6aux *ip6a;
2058 	int plen;
2059 	int type, code;
2060 	struct ifnet *outif = NULL;
2061 	struct in6_addr origdst;
2062 	struct ifnet *rcvif;
2063 	int s;
2064 	bool ip6_src_filled = false;
2065 	int flags;
2066 
2067 	/* too short to reflect */
2068 	if (off < sizeof(struct ip6_hdr)) {
2069 		nd6log(LOG_DEBUG,
2070 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2071 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2072 		    __FILE__, __LINE__);
2073 		goto bad;
2074 	}
2075 
2076 	/*
2077 	 * If there are extra headers between IPv6 and ICMPv6, strip
2078 	 * off that header first.
2079 	 */
2080 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
2081 	if (off > sizeof(struct ip6_hdr)) {
2082 		size_t l;
2083 		struct ip6_hdr nip6;
2084 
2085 		l = off - sizeof(struct ip6_hdr);
2086 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2087 		m_adj(m, l);
2088 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2089 		if (m->m_len < l) {
2090 			if ((m = m_pullup(m, l)) == NULL)
2091 				return;
2092 		}
2093 		memcpy(mtod(m, void *), (void *)&nip6, sizeof(nip6));
2094 	} else {
2095 		size_t l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2096 		if (m->m_len < l) {
2097 			if ((m = m_pullup(m, l)) == NULL)
2098 				return;
2099 		}
2100 	}
2101 
2102 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2103 	ip6 = mtod(m, struct ip6_hdr *);
2104 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2105 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2106 	type = icmp6->icmp6_type; /* keep type for statistics */
2107 	code = icmp6->icmp6_code; /* ditto. */
2108 
2109 	origdst = ip6->ip6_dst;
2110 	/*
2111 	 * ip6_input() drops a packet if its src is multicast.
2112 	 * So, the src is never multicast.
2113 	 */
2114 	ip6->ip6_dst = ip6->ip6_src;
2115 
2116 	/*
2117 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2118 	 * use dst as the src for the reply.
2119 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2120 	 * (for example) when we encounter an error while forwarding procedure
2121 	 * destined to a duplicated address of ours.
2122 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2123 	 * procedure of an outgoing packet of our own, in which case we need
2124 	 * to search in the ifaddr list.
2125 	 */
2126 	if (IN6_IS_ADDR_MULTICAST(&origdst)) {
2127 		;
2128 	} else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2129 		if ((ip6a->ip6a_flags &
2130 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2131 			ip6->ip6_src = ip6a->ip6a_src;
2132 			ip6_src_filled = true;
2133 		}
2134 	} else {
2135 		union {
2136 			struct sockaddr_in6 sin6;
2137 			struct sockaddr sa;
2138 		} u;
2139 		int _s;
2140 		struct ifaddr *ifa;
2141 
2142 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2143 
2144 		_s = pserialize_read_enter();
2145 		ifa = ifa_ifwithaddr(&u.sa);
2146 
2147 		if (ifa != NULL) {
2148 			ia = ifatoia6(ifa);
2149 			if ((ia->ia6_flags &
2150 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2151 				ip6->ip6_src = ia->ia_addr.sin6_addr;
2152 				ip6_src_filled = true;
2153 			}
2154 		}
2155 		pserialize_read_exit(_s);
2156 	}
2157 
2158 	if (!ip6_src_filled) {
2159 		int e;
2160 		struct sockaddr_in6 sin6;
2161 		struct route ro;
2162 
2163 		/*
2164 		 * This case matches to multicasts, our anycast, or unicasts
2165 		 * that we do not own.  Select a source address based on the
2166 		 * source address of the erroneous packet.
2167 		 */
2168 		/* zone ID should be embedded */
2169 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2170 
2171 		memset(&ro, 0, sizeof(ro));
2172 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2173 		    &ip6->ip6_src);
2174 		rtcache_free(&ro);
2175 		if (e != 0) {
2176 			char ip6buf[INET6_ADDRSTRLEN];
2177 			nd6log(LOG_DEBUG,
2178 			    "source can't be determined: "
2179 			    "dst=%s, error=%d\n",
2180 			    IN6_PRINT(ip6buf, &sin6.sin6_addr), e);
2181 			goto bad;
2182 		}
2183 	}
2184 
2185 	ip6->ip6_flow = 0;
2186 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2187 	ip6->ip6_vfc |= IPV6_VERSION;
2188 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2189 	rcvif = m_get_rcvif(m, &s);
2190 	if (rcvif) {
2191 		/* XXX: This may not be the outgoing interface */
2192 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2193 	} else {
2194 		ip6->ip6_hlim = ip6_defhlim;
2195 	}
2196 	m_put_rcvif(rcvif, &s);
2197 
2198 	m->m_pkthdr.csum_flags = 0;
2199 	icmp6->icmp6_cksum = 0;
2200 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2201 	    sizeof(struct ip6_hdr), plen);
2202 
2203 	/*
2204 	 * XXX option handling
2205 	 */
2206 
2207 	m->m_flags &= ~(M_BCAST|M_MCAST);
2208 
2209 	/*
2210 	 * Note for icmp6_reflect_pmtu == false
2211 	 * To avoid a "too big" situation at an intermediate router
2212 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2213 	 * Note that only echo and node information replies are affected,
2214 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2215 	 */
2216 	flags = icmp6_reflect_pmtu ? 0 : IPV6_MINMTU;
2217 	if (ip6_output(m, NULL, NULL, flags, NULL, NULL, &outif) != 0 &&
2218 	    outif)
2219 		icmp6_ifstat_inc(outif, ifs6_out_error);
2220 	if (outif)
2221 		icmp6_ifoutstat_inc(outif, type, code);
2222 
2223 	return;
2224 
2225  bad:
2226 	m_freem(m);
2227 	return;
2228 }
2229 
2230 static const char *
2231 icmp6_redirect_diag(char *buf, size_t buflen, struct in6_addr *src6,
2232     struct in6_addr *dst6,  struct in6_addr *tgt6)
2233 {
2234 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
2235 	char ip6buft[INET6_ADDRSTRLEN];
2236 
2237 	snprintf(buf, buflen, "(src=%s dst=%s tgt=%s)",
2238 	    IN6_PRINT(ip6bufs, src6), IN6_PRINT(ip6bufd, dst6),
2239 	    IN6_PRINT(ip6buft, tgt6));
2240 	return buf;
2241 }
2242 
2243 static void
2244 icmp6_redirect_input(struct mbuf *m, int off)
2245 {
2246 	struct ifnet *ifp;
2247 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2248 	struct nd_redirect *nd_rd;
2249 	int icmp6len = m->m_pkthdr.len - off;
2250 	char *lladdr = NULL;
2251 	int lladdrlen = 0;
2252 	struct rtentry *rt = NULL;
2253 	int is_router;
2254 	int is_onlink;
2255 	struct in6_addr src6 = ip6->ip6_src;
2256 	struct in6_addr redtgt6;
2257 	struct in6_addr reddst6;
2258 	union nd_opts ndopts;
2259 	struct psref psref;
2260 	char ip6buf[INET6_ADDRSTRLEN];
2261 	char diagbuf[256];
2262 
2263 	ifp = m_get_rcvif_psref(m, &psref);
2264 	if (ifp == NULL)
2265 		goto freeit;
2266 
2267 	/* XXX if we are router, we don't update route by icmp6 redirect */
2268 	if (ip6_forwarding)
2269 		goto freeit;
2270 	if (!icmp6_rediraccept)
2271 		goto freeit;
2272 
2273 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2274 	if (nd_rd == NULL) {
2275 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2276 		m_put_rcvif_psref(ifp, &psref);
2277 		return;
2278 	}
2279 	redtgt6 = nd_rd->nd_rd_target;
2280 	reddst6 = nd_rd->nd_rd_dst;
2281 
2282 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2283 	    in6_setscope(&reddst6, ifp, NULL)) {
2284 		goto freeit;
2285 	}
2286 
2287 	/* validation */
2288 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2289 		nd6log(LOG_ERR,
2290 		    "ICMP6 redirect sent from %s rejected; "
2291 		    "must be from linklocal\n", IN6_PRINT(ip6buf, &src6));
2292 		goto bad;
2293 	}
2294 	if (ip6->ip6_hlim != 255) {
2295 		nd6log(LOG_ERR,
2296 		    "ICMP6 redirect sent from %s rejected; "
2297 		    "hlim=%d (must be 255)\n",
2298 		    IN6_PRINT(ip6buf, &src6), ip6->ip6_hlim);
2299 		goto bad;
2300 	}
2301 
2302     {
2303 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2304 	struct sockaddr_in6 sin6;
2305 	struct in6_addr *gw6;
2306 
2307 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2308 	rt = rtalloc1(sin6tosa(&sin6), 0);
2309 	if (rt) {
2310 		if (rt->rt_gateway == NULL ||
2311 		    rt->rt_gateway->sa_family != AF_INET6) {
2312 			nd6log(LOG_ERR,
2313 			    "ICMP6 redirect rejected; no route "
2314 			    "with inet6 gateway found for redirect dst: %s\n",
2315 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2316 			    &src6, &reddst6, &redtgt6));
2317 			rt_unref(rt);
2318 			goto bad;
2319 		}
2320 
2321 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2322 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2323 			nd6log(LOG_ERR,
2324 			    "ICMP6 redirect rejected; "
2325 			    "not equal to gw-for-src=%s (must be same): %s\n",
2326 			    IN6_PRINT(ip6buf, gw6),
2327 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2328 			    &src6, &reddst6, &redtgt6));
2329 			rt_unref(rt);
2330 			goto bad;
2331 		}
2332 	} else {
2333 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2334 		    "no route found for redirect dst: %s\n",
2335 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2336 		    &src6, &reddst6, &redtgt6));
2337 		goto bad;
2338 	}
2339 	rt_unref(rt);
2340 	rt = NULL;
2341     }
2342 
2343 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2344 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2345 		    "redirect dst must be unicast: %s\n",
2346 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2347 		    &src6, &reddst6, &redtgt6));
2348 		goto bad;
2349 	}
2350 
2351 	is_router = is_onlink = 0;
2352 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2353 		is_router = 1;	/* router case */
2354 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2355 		is_onlink = 1;	/* on-link destination case */
2356 	if (!is_router && !is_onlink) {
2357 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2358 		    "neither router case nor onlink case: %s\n",
2359 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2360 		    &src6, &reddst6, &redtgt6));
2361 		goto bad;
2362 	}
2363 	/* validation passed */
2364 
2365 	icmp6len -= sizeof(*nd_rd);
2366 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2367 	if (nd6_options(&ndopts) < 0) {
2368 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2369 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2370 		    &src6, &reddst6, &redtgt6));
2371 		/* nd6_options have incremented stats */
2372 		goto freeit;
2373 	}
2374 
2375 	if (ndopts.nd_opts_tgt_lladdr) {
2376 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2377 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2378 	}
2379 
2380 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2381 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2382 		    "(if %d, icmp6 packet %d): %s\n",
2383 		    IN6_PRINT(ip6buf, &redtgt6),
2384 		    ifp->if_addrlen, lladdrlen - 2,
2385 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2386 		    &src6, &reddst6, &redtgt6));
2387 		goto bad;
2388 	}
2389 
2390 	/* RFC 2461 8.3 */
2391 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2392 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2393 
2394 	m_put_rcvif_psref(ifp, &psref);
2395 	ifp = NULL;
2396 
2397 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2398 		/* perform rtredirect */
2399 		struct sockaddr_in6 sdst;
2400 		struct sockaddr_in6 sgw;
2401 		struct sockaddr_in6 ssrc;
2402 		unsigned long rtcount;
2403 		struct rtentry *newrt = NULL;
2404 
2405 		/*
2406 		 * do not install redirect route, if the number of entries
2407 		 * is too much (> hiwat).  note that, the node (= host) will
2408 		 * work just fine even if we do not install redirect route
2409 		 * (there will be additional hops, though).
2410 		 */
2411 		mutex_enter(&icmp6_mtx);
2412 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2413 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) {
2414 			mutex_exit(&icmp6_mtx);
2415 			goto freeit;
2416 		}
2417 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) {
2418 			mutex_exit(&icmp6_mtx);
2419 			goto freeit;
2420 		} else if (0 <= icmp6_redirect_lowat &&
2421 		    rtcount > icmp6_redirect_lowat) {
2422 			/*
2423 			 * XXX nuke a victim, install the new one.
2424 			 */
2425 		}
2426 
2427 		memset(&sdst, 0, sizeof(sdst));
2428 		memset(&sgw, 0, sizeof(sgw));
2429 		memset(&ssrc, 0, sizeof(ssrc));
2430 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2431 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2432 		    sizeof(struct sockaddr_in6);
2433 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2434 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2435 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2436 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2437 		    RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc), &newrt);
2438 
2439 		if (newrt) {
2440 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2441 			    icmp6_redirect_timeout_q);
2442 			rt_unref(newrt);
2443 		}
2444 		mutex_exit(&icmp6_mtx);
2445 	}
2446 	/* finally update cached route in each socket via pfctlinput */
2447 	{
2448 		struct sockaddr_in6 sdst;
2449 
2450 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2451 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2452 #if defined(IPSEC)
2453 		if (ipsec_used)
2454 			key_sa_routechange(sin6tosa(&sdst));
2455 #endif
2456 	}
2457 
2458 freeit:
2459 	if (ifp != NULL)
2460 		m_put_rcvif_psref(ifp, &psref);
2461 	m_freem(m);
2462 	return;
2463 
2464 bad:
2465 	m_put_rcvif_psref(ifp, &psref);
2466 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2467 	m_freem(m);
2468 }
2469 
2470 void
2471 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2472 {
2473 	struct ifnet *ifp;	/* my outgoing interface */
2474 	struct in6_addr *ifp_ll6;
2475 	struct in6_addr *nexthop;
2476 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2477 	struct mbuf *m = NULL;	/* newly allocated one */
2478 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2479 	struct nd_redirect *nd_rd;
2480 	size_t maxlen;
2481 	u_char *p;
2482 	struct sockaddr_in6 src_sa;
2483 
2484 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2485 
2486 	/* if we are not router, we don't send icmp6 redirect */
2487 	if (!ip6_forwarding)
2488 		goto fail;
2489 
2490 	/* sanity check */
2491 	KASSERT(m0 != NULL);
2492 	KASSERT(rt != NULL);
2493 
2494 	ifp = rt->rt_ifp;
2495 
2496 	/*
2497 	 * Address check:
2498 	 *  the source address must identify a neighbor, and
2499 	 *  the destination address must not be a multicast address
2500 	 *  [RFC 2461, sec 8.2]
2501 	 */
2502 	sip6 = mtod(m0, struct ip6_hdr *);
2503 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2504 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2505 		goto fail;
2506 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2507 		goto fail;	/* what should we do here? */
2508 
2509 	/* rate limit */
2510 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2511 		goto fail;
2512 
2513 	/*
2514 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2515 	 * we almost always ask for an mbuf cluster for simplicity.
2516 	 * (MHLEN < IPV6_MMTU is almost always true)
2517 	 */
2518 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2519 	if (m && IPV6_MMTU >= MHLEN) {
2520 #if IPV6_MMTU >= MCLBYTES
2521 		MEXTMALLOC(m, IPV6_MMTU, M_NOWAIT);
2522 #else
2523 		MCLGET(m, M_DONTWAIT);
2524 #endif
2525 	}
2526 
2527 	if (!m)
2528 		goto fail;
2529 	m_reset_rcvif(m);
2530 	m->m_len = 0;
2531 	maxlen = M_TRAILINGSPACE(m);
2532 	maxlen = uimin(IPV6_MMTU, maxlen);
2533 
2534 	/* just for safety */
2535 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct nd_redirect) +
2536 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2537 		goto fail;
2538 	}
2539 
2540 	{
2541 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2542 		struct in6_ifaddr *ia;
2543 		int s = pserialize_read_enter();
2544 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2545 						 IN6_IFF_NOTREADY|
2546 						 IN6_IFF_ANYCAST)) == NULL) {
2547 			pserialize_read_exit(s);
2548 			goto fail;
2549 		}
2550 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2551 		pserialize_read_exit(s);
2552 	}
2553 
2554 	/* get ip6 linklocal address for the router. */
2555 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2556 		struct sockaddr_in6 *sin6;
2557 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2558 		nexthop = &sin6->sin6_addr;
2559 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2560 			nexthop = NULL;
2561 	} else
2562 		nexthop = NULL;
2563 
2564 	/* ip6 */
2565 	ip6 = mtod(m, struct ip6_hdr *);
2566 	ip6->ip6_flow = 0;
2567 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2568 	ip6->ip6_vfc |= IPV6_VERSION;
2569 	/* ip6->ip6_plen will be set later */
2570 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2571 	ip6->ip6_hlim = 255;
2572 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2573 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2574 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2575 
2576 	/* ND Redirect */
2577 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2578 	nd_rd->nd_rd_type = ND_REDIRECT;
2579 	nd_rd->nd_rd_code = 0;
2580 	nd_rd->nd_rd_reserved = 0;
2581 	if (rt->rt_flags & RTF_GATEWAY) {
2582 		/*
2583 		 * nd_rd->nd_rd_target must be a link-local address in
2584 		 * better router cases.
2585 		 */
2586 		if (!nexthop)
2587 			goto fail;
2588 		bcopy(nexthop, &nd_rd->nd_rd_target,
2589 		      sizeof(nd_rd->nd_rd_target));
2590 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2591 		      sizeof(nd_rd->nd_rd_dst));
2592 	} else {
2593 		/* make sure redtgt == reddst */
2594 		nexthop = &sip6->ip6_dst;
2595 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2596 		      sizeof(nd_rd->nd_rd_target));
2597 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2598 		      sizeof(nd_rd->nd_rd_dst));
2599 	}
2600 
2601 	p = (u_char *)(nd_rd + 1);
2602 
2603 	{
2604 		/* target lladdr option */
2605 		struct llentry *ln = NULL;
2606 		int len, pad;
2607 		struct nd_opt_hdr *nd_opt;
2608 		char *lladdr;
2609 
2610 		ln = nd6_lookup(nexthop, ifp, false);
2611 		if (ln == NULL)
2612 			goto nolladdropt;
2613 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2614 		len = (len + 7) & ~7;	/* round by 8 */
2615 		pad = len - (sizeof(*nd_opt) + ifp->if_addrlen);
2616 
2617 		/* safety check */
2618 		if (len + (p - (u_char *)ip6) > maxlen) {
2619 			LLE_RUNLOCK(ln);
2620 			goto nolladdropt;
2621 		}
2622 
2623 		if (ln->la_flags & LLE_VALID) {
2624 			nd_opt = (struct nd_opt_hdr *)p;
2625 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2626 			nd_opt->nd_opt_len = len >> 3;
2627 			lladdr = (char *)(nd_opt + 1);
2628 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2629 			memset(lladdr + ifp->if_addrlen, 0, pad);
2630 			p += len;
2631 		}
2632 		LLE_RUNLOCK(ln);
2633 	}
2634 nolladdropt:
2635 
2636 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2637 
2638 	/* just to be safe */
2639 	if (m0->m_flags & M_DECRYPTED)
2640 		goto noredhdropt;
2641 	if (p - (u_char *)ip6 > maxlen)
2642 		goto noredhdropt;
2643 
2644 	{
2645 		/* redirected header option */
2646 		int len;
2647 		struct nd_opt_rd_hdr *nd_opt_rh;
2648 
2649 		/*
2650 		 * compute the maximum size for icmp6 redirect header option.
2651 		 * XXX room for auth header?
2652 		 */
2653 		len = maxlen - (p - (u_char *)ip6);
2654 		len &= ~7;
2655 
2656 		if (len < sizeof(*nd_opt_rh)) {
2657 			goto noredhdropt;
2658 		}
2659 
2660 		/*
2661 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2662 		 * about padding/truncate rule for the original IP packet.
2663 		 * From the discussion on IPv6imp in Feb 1999,
2664 		 * the consensus was:
2665 		 * - "attach as much as possible" is the goal
2666 		 * - pad if not aligned (original size can be guessed by
2667 		 *   original ip6 header)
2668 		 * Following code adds the padding if it is simple enough,
2669 		 * and truncates if not.
2670 		 */
2671 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2672 			/* not enough room, truncate */
2673 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2674 			    m0->m_pkthdr.len);
2675 		} else {
2676 			/*
2677 			 * enough room, truncate if not aligned.
2678 			 * we don't pad here for simplicity.
2679 			 */
2680 			int extra;
2681 
2682 			extra = m0->m_pkthdr.len % 8;
2683 			if (extra) {
2684 				/* truncate */
2685 				m_adj(m0, -extra);
2686 			}
2687 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2688 		}
2689 
2690 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2691 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2692 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2693 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2694 		p += sizeof(*nd_opt_rh);
2695 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2696 
2697 		/* connect m0 to m */
2698 		m->m_pkthdr.len += m0->m_pkthdr.len;
2699 		m_cat(m, m0);
2700 		m0 = NULL;
2701 	}
2702 noredhdropt:
2703 	if (m0) {
2704 		m_freem(m0);
2705 		m0 = NULL;
2706 	}
2707 
2708 	/* XXX: clear embedded link IDs in the inner header */
2709 	in6_clearscope(&sip6->ip6_src);
2710 	in6_clearscope(&sip6->ip6_dst);
2711 	in6_clearscope(&nd_rd->nd_rd_target);
2712 	in6_clearscope(&nd_rd->nd_rd_dst);
2713 
2714 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2715 
2716 	nd_rd->nd_rd_cksum = 0;
2717 	nd_rd->nd_rd_cksum =
2718 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2719 
2720 	/* send the packet to outside... */
2721 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2722 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2723 
2724 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2725 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2726 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2727 
2728 	return;
2729 
2730 fail:
2731 	if (m)
2732 		m_freem(m);
2733 	if (m0)
2734 		m_freem(m0);
2735 }
2736 
2737 /*
2738  * ICMPv6 socket option processing.
2739  */
2740 int
2741 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2742 {
2743 	int error = 0;
2744 	struct in6pcb *in6p = sotoin6pcb(so);
2745 
2746 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2747 		return rip6_ctloutput(op, so, sopt);
2748 
2749 	switch (op) {
2750 	case PRCO_SETOPT:
2751 		switch (sopt->sopt_name) {
2752 		case ICMP6_FILTER:
2753 		    {
2754 			struct icmp6_filter fil;
2755 
2756 			error = sockopt_get(sopt, &fil, sizeof(fil));
2757 			if (error)
2758 				break;
2759 			memcpy(in6p->in6p_icmp6filt, &fil,
2760 			    sizeof(struct icmp6_filter));
2761 			error = 0;
2762 			break;
2763 		    }
2764 
2765 		default:
2766 			error = ENOPROTOOPT;
2767 			break;
2768 		}
2769 		break;
2770 
2771 	case PRCO_GETOPT:
2772 		switch (sopt->sopt_name) {
2773 		case ICMP6_FILTER:
2774 		    {
2775 			if (in6p->in6p_icmp6filt == NULL) {
2776 				error = EINVAL;
2777 				break;
2778 			}
2779 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2780 			    sizeof(struct icmp6_filter));
2781 			break;
2782 		    }
2783 
2784 		default:
2785 			error = ENOPROTOOPT;
2786 			break;
2787 		}
2788 		break;
2789 	}
2790 
2791 	return error;
2792 }
2793 
2794 /*
2795  * Perform rate limit check.
2796  * Returns 0 if it is okay to send the icmp6 packet.
2797  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2798  * limitation.
2799  *
2800  * XXX per-destination/type check necessary?
2801  */
2802 static int
2803 icmp6_ratelimit(
2804 	const struct in6_addr *dst,	/* not used at this moment */
2805 	const int type,		/* not used at this moment */
2806 	const int code)		/* not used at this moment */
2807 {
2808 	int ret;
2809 
2810 	ret = 0;	/* okay to send */
2811 
2812 	/* PPS limit */
2813 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2814 	    icmp6errppslim)) {
2815 		/* The packet is subject to rate limit */
2816 		ret++;
2817 	}
2818 
2819 	return ret;
2820 }
2821 
2822 static struct rtentry *
2823 icmp6_mtudisc_clone(struct sockaddr *dst)
2824 {
2825 	struct rtentry *rt;
2826 	int    error;
2827 
2828 	rt = rtalloc1(dst, 1);
2829 	if (rt == NULL)
2830 		return NULL;
2831 
2832 	/* If we didn't get a host route, allocate one */
2833 	if ((rt->rt_flags & RTF_HOST) == 0) {
2834 		struct rtentry *nrt;
2835 
2836 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2837 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2838 		if (error) {
2839 			rt_unref(rt);
2840 			return NULL;
2841 		}
2842 		nrt->rt_rmx = rt->rt_rmx;
2843 		rt_newmsg_dynamic(RTM_ADD, nrt);
2844 		rt_unref(rt);
2845 		rt = nrt;
2846 	}
2847 
2848 	mutex_enter(&icmp6_mtx);
2849 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2850 			icmp6_mtudisc_timeout_q);
2851 	mutex_exit(&icmp6_mtx);
2852 
2853 	if (error) {
2854 		rt_unref(rt);
2855 		return NULL;
2856 	}
2857 
2858 	return rt;	/* caller need to call rtfree() */
2859 }
2860 
2861 static void
2862 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2863 {
2864 	struct rtentry *retrt;
2865 
2866 	KASSERT(rt != NULL);
2867 	rt_assert_referenced(rt);
2868 
2869 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2870 	    (RTF_DYNAMIC | RTF_HOST)) {
2871 		rtrequest(RTM_DELETE, rt_getkey(rt),
2872 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2873 		rt_newmsg_dynamic(RTM_DELETE, retrt);
2874 		rt_unref(rt);
2875 		rt_free(retrt);
2876 	} else {
2877 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2878 			rt->rt_rmx.rmx_mtu = 0;
2879 	}
2880 }
2881 
2882 static void
2883 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2884 {
2885 	struct rtentry *retrt;
2886 
2887 	KASSERT(rt != NULL);
2888 	rt_assert_referenced(rt);
2889 
2890 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2891 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2892 		rtrequest(RTM_DELETE, rt_getkey(rt),
2893 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2894 		rt_newmsg_dynamic(RTM_DELETE, retrt);
2895 		rt_unref(rt);
2896 		rt_free(retrt);
2897 	}
2898 }
2899 
2900 #ifdef COMPAT_90
2901 /*
2902  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2903  */
2904 static int
2905 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2906 {
2907 	(void)&name;
2908 	(void)&l;
2909 	(void)&oname;
2910 
2911 	if (namelen != 0)
2912 		return (EINVAL);
2913 
2914 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2915 	    /*XXXUNCONST*/
2916 	    __UNCONST(newp), newlen));
2917 }
2918 #endif
2919 
2920 static int
2921 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2922 {
2923 
2924 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2925 }
2926 
2927 static int
2928 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2929 {
2930 	int error, tmp;
2931 	struct sysctlnode node;
2932 
2933 	mutex_enter(&icmp6_mtx);
2934 
2935 	node = *rnode;
2936 	node.sysctl_data = &tmp;
2937 	tmp = icmp6_redirtimeout;
2938 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2939 	if (error || newp == NULL)
2940 		goto out;
2941 	if (tmp < 0) {
2942 		error = EINVAL;
2943 		goto out;
2944 	}
2945 	icmp6_redirtimeout = tmp;
2946 
2947 	if (icmp6_redirect_timeout_q != NULL) {
2948 		if (icmp6_redirtimeout == 0) {
2949 			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2950 		} else {
2951 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2952 			    icmp6_redirtimeout);
2953 		}
2954 	} else if (icmp6_redirtimeout > 0) {
2955 		icmp6_redirect_timeout_q =
2956 		    rt_timer_queue_create(icmp6_redirtimeout);
2957 	}
2958 	error = 0;
2959 out:
2960 	mutex_exit(&icmp6_mtx);
2961 	return error;
2962 }
2963 
2964 static void
2965 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2966 {
2967 
2968 	sysctl_createv(clog, 0, NULL, NULL,
2969 		       CTLFLAG_PERMANENT,
2970 		       CTLTYPE_NODE, "inet6", NULL,
2971 		       NULL, 0, NULL, 0,
2972 		       CTL_NET, PF_INET6, CTL_EOL);
2973 	sysctl_createv(clog, 0, NULL, NULL,
2974 		       CTLFLAG_PERMANENT,
2975 		       CTLTYPE_NODE, "icmp6",
2976 		       SYSCTL_DESCR("ICMPv6 related settings"),
2977 		       NULL, 0, NULL, 0,
2978 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2979 
2980 	sysctl_createv(clog, 0, NULL, NULL,
2981 		       CTLFLAG_PERMANENT,
2982 		       CTLTYPE_STRUCT, "stats",
2983 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2984 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2985 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2986 		       ICMPV6CTL_STATS, CTL_EOL);
2987 	sysctl_createv(clog, 0, NULL, NULL,
2988 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2989 		       CTLTYPE_INT, "rediraccept",
2990 		       SYSCTL_DESCR("Accept and process redirect messages"),
2991 		       NULL, 0, &icmp6_rediraccept, 0,
2992 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2993 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2994 	sysctl_createv(clog, 0, NULL, NULL,
2995 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2996 		       CTLTYPE_INT, "redirtimeout",
2997 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2998 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2999 		       &icmp6_redirtimeout, 0,
3000 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3001 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
3002 #if 0 /* obsoleted */
3003 	sysctl_createv(clog, 0, NULL, NULL,
3004 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3005 		       CTLTYPE_INT, "errratelimit", NULL,
3006 		       NULL, 0, &icmp6_errratelimit, 0,
3007 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3008 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
3009 #endif
3010 	sysctl_createv(clog, 0, NULL, NULL,
3011 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3012 		       CTLTYPE_INT, "nd6_prune",
3013 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
3014 		       NULL, 0, &nd6_prune, 0,
3015 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3016 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
3017 	sysctl_createv(clog, 0, NULL, NULL,
3018 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3019 		       CTLTYPE_INT, "nd6_delay",
3020 		       SYSCTL_DESCR("First probe delay time"),
3021 		       NULL, 0, &nd6_nd_domain.nd_delay, 0,
3022 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3023 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
3024 	sysctl_createv(clog, 0, NULL, NULL,
3025 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3026 		       CTLTYPE_INT, "nd6_mmaxtries",
3027 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
3028 		       NULL, 0, &nd6_nd_domain.nd_mmaxtries, 0,
3029 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3030 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
3031 	sysctl_createv(clog, 0, NULL, NULL,
3032 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3033 		       CTLTYPE_INT, "nd6_umaxtries",
3034 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
3035 		       NULL, 0, &nd6_nd_domain.nd_umaxtries, 0,
3036 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3037 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
3038 	sysctl_createv(clog, 0, NULL, NULL,
3039 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3040 		       CTLTYPE_INT, "nd6_maxnudhint",
3041 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
3042 		       NULL, 0, &nd6_nd_domain.nd_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, "maxqueuelen",
3048 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3049 		       NULL, 1, &nd6_nd_domain.nd_maxqueuelen, 0,
3050 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3051 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3052 	sysctl_createv(clog, 0, NULL, NULL,
3053 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3054 		       CTLTYPE_INT, "nd6_useloopback",
3055 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
3056 		       NULL, 0, &nd6_useloopback, 0,
3057 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3058 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
3059 #if 0 /* obsoleted */
3060 	sysctl_createv(clog, 0, NULL, NULL,
3061 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3062 		       CTLTYPE_INT, "nd6_proxyall", NULL,
3063 		       NULL, 0, &nd6_proxyall, 0,
3064 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3065 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
3066 #endif
3067 	sysctl_createv(clog, 0, NULL, NULL,
3068 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3069 		       CTLTYPE_INT, "nodeinfo",
3070 		       SYSCTL_DESCR("Respond to node information requests"),
3071 		       NULL, 0, &icmp6_nodeinfo, 0,
3072 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3073 		       ICMPV6CTL_NODEINFO, CTL_EOL);
3074 	sysctl_createv(clog, 0, NULL, NULL,
3075 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3076 		       CTLTYPE_INT, "errppslimit",
3077 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
3078 		       NULL, 0, &icmp6errppslim, 0,
3079 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3080 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
3081 	sysctl_createv(clog, 0, NULL, NULL,
3082 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3083 		       CTLTYPE_INT, "mtudisc_hiwat",
3084 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3085 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
3086 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3087 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
3088 	sysctl_createv(clog, 0, NULL, NULL,
3089 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3090 		       CTLTYPE_INT, "mtudisc_lowat",
3091 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3092 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
3093 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3094 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
3095 	sysctl_createv(clog, 0, NULL, NULL,
3096 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3097 		       CTLTYPE_INT, "nd6_debug",
3098 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
3099 		       NULL, 0, &nd6_debug, 0,
3100 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3101 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
3102 #ifdef COMPAT_90
3103 	sysctl_createv(clog, 0, NULL, NULL,
3104 		       CTLFLAG_PERMANENT,
3105 		       CTLTYPE_STRUCT, "nd6_drlist",
3106 		       SYSCTL_DESCR("Default router list"),
3107 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3108 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3109 		       OICMPV6CTL_ND6_DRLIST, CTL_EOL);
3110 	sysctl_createv(clog, 0, NULL, NULL,
3111 		       CTLFLAG_PERMANENT,
3112 		       CTLTYPE_STRUCT, "nd6_prlist",
3113 		       SYSCTL_DESCR("Prefix list"),
3114 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3115 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3116 		       OICMPV6CTL_ND6_PRLIST, CTL_EOL);
3117 #endif
3118 	sysctl_createv(clog, 0, NULL, NULL,
3119 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3120 		       CTLTYPE_BOOL, "reflect_pmtu",
3121 		       SYSCTL_DESCR("Use path MTU Discovery for icmpv6 reflect"),
3122 		       NULL, 0, &icmp6_reflect_pmtu, 0,
3123 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3124 		       ICMPV6CTL_REFLECT_PMTU, CTL_EOL);
3125 	sysctl_createv(clog, 0, NULL, NULL,
3126 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3127 		       CTLTYPE_BOOL, "dynamic_rt_msg",
3128 		       SYSCTL_DESCR("Send routing message for RTF_DYNAMIC"),
3129 		       NULL, 0, &icmp6_dynamic_rt_msg, 0,
3130 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3131 		       ICMPV6CTL_DYNAMIC_RT_MSG, CTL_EOL);
3132 }
3133 
3134 void
3135 icmp6_statinc(u_int stat)
3136 {
3137 
3138 	KASSERT(stat < ICMP6_NSTATS);
3139 	ICMP6_STATINC(stat);
3140 }
3141