xref: /netbsd-src/sys/netinet6/icmp6.c (revision 53b02e147d4ed531c0d2a5ca9b3e8026ba3e99b5)
1 /*	$NetBSD: icmp6.c,v 1.250 2021/02/19 14:52:00 christos 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 1.250 2021/02/19 14:52:00 christos 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 /*
155  * List of callbacks to notify when Path MTU changes are made.
156  */
157 struct icmp6_mtudisc_callback {
158 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
159 	void (*mc_func)(struct in6_addr *);
160 };
161 
162 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
163     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
164 
165 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
166 extern int pmtu_expire;
167 
168 /* XXX do these values make any sense? */
169 static int icmp6_mtudisc_hiwat = 1280;
170 static int icmp6_mtudisc_lowat = 256;
171 
172 /*
173  * keep track of # of redirect routes.
174  */
175 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
176 
177 /* XXX experimental, turned off */
178 static int icmp6_redirect_hiwat = -1;
179 static int icmp6_redirect_lowat = -1;
180 
181 /* Protect mtudisc and redirect stuffs */
182 static kmutex_t icmp6_mtx __cacheline_aligned;
183 
184 static void icmp6_errcount(u_int, int, int);
185 static int icmp6_rip6_input(struct mbuf **, int);
186 static void icmp6_reflect(struct mbuf *, size_t);
187 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
188 static const char *icmp6_redirect_diag(char *, size_t, struct in6_addr *,
189     struct in6_addr *, struct in6_addr *);
190 static void icmp6_redirect_input(struct mbuf *, int);
191 static struct mbuf *ni6_input(struct mbuf *, int);
192 static struct mbuf *ni6_nametodns(const char *, int, int);
193 static int ni6_dnsmatch(const char *, int, const char *, int);
194 static int ni6_addrs(struct icmp6_nodeinfo *, struct ifnet **, char *,
195     struct psref *);
196 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
197     struct ifnet *, int);
198 static int icmp6_notify_error(struct mbuf *, int, int, int);
199 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
200 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
201 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
202 static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
203 
204 /* workqueue-based pr_input */
205 static struct wqinput *icmp6_wqinput;
206 static void _icmp6_input(struct mbuf *m, int off, int proto);
207 
208 void
209 icmp6_init(void)
210 {
211 
212 	sysctl_net_inet6_icmp6_setup(NULL);
213 	mld_init();
214 
215 	mutex_init(&icmp6_mtx, MUTEX_DEFAULT, IPL_NONE);
216 	mutex_enter(&icmp6_mtx);
217 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
218 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
219 	mutex_exit(&icmp6_mtx);
220 
221 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
222 
223 	icmp6_wqinput = wqinput_create("icmp6", _icmp6_input);
224 }
225 
226 static void
227 icmp6_errcount(u_int base, int type, int code)
228 {
229 	switch (type) {
230 	case ICMP6_DST_UNREACH:
231 		switch (code) {
232 		case ICMP6_DST_UNREACH_NOROUTE:
233 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
234 			return;
235 		case ICMP6_DST_UNREACH_ADMIN:
236 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
237 			return;
238 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
239 			ICMP6_STATINC(base +
240 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
241 			return;
242 		case ICMP6_DST_UNREACH_ADDR:
243 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
244 			return;
245 		case ICMP6_DST_UNREACH_NOPORT:
246 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
247 			return;
248 		}
249 		break;
250 	case ICMP6_PACKET_TOO_BIG:
251 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
252 		return;
253 	case ICMP6_TIME_EXCEEDED:
254 		switch (code) {
255 		case ICMP6_TIME_EXCEED_TRANSIT:
256 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
257 			return;
258 		case ICMP6_TIME_EXCEED_REASSEMBLY:
259 			ICMP6_STATINC(base +
260 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
261 			return;
262 		}
263 		break;
264 	case ICMP6_PARAM_PROB:
265 		switch (code) {
266 		case ICMP6_PARAMPROB_HEADER:
267 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
268 			return;
269 		case ICMP6_PARAMPROB_NEXTHEADER:
270 			ICMP6_STATINC(base +
271 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
272 			return;
273 		case ICMP6_PARAMPROB_OPTION:
274 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
275 			return;
276 		}
277 		break;
278 	case ND_REDIRECT:
279 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
280 		return;
281 	}
282 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
283 }
284 
285 /*
286  * Register a Path MTU Discovery callback.
287  */
288 void
289 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
290 {
291 	struct icmp6_mtudisc_callback *mc, *new;
292 
293 	new = kmem_alloc(sizeof(*mc), KM_SLEEP);
294 
295 	mutex_enter(&icmp6_mtx);
296 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
297 	     mc = LIST_NEXT(mc, mc_list)) {
298 		if (mc->mc_func == func) {
299 			mutex_exit(&icmp6_mtx);
300 			kmem_free(new, sizeof(*mc));
301 			return;
302 		}
303 	}
304 
305 	new->mc_func = func;
306 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, new, mc_list);
307 	mutex_exit(&icmp6_mtx);
308 }
309 
310 /*
311  * A wrapper function for icmp6_error() necessary when the erroneous packet
312  * may not contain enough scope zone information.
313  */
314 void
315 icmp6_error2(struct mbuf *m, int type, int code, int param,
316 	struct ifnet *ifp, struct in6_addr *src)
317 {
318 	struct ip6_hdr *ip6;
319 
320 	KASSERT(ifp != NULL);
321 
322 	if (m->m_len < sizeof(struct ip6_hdr)) {
323 		m = m_pullup(m, sizeof(struct ip6_hdr));
324 		if (m == NULL)
325 			return;
326 	}
327 
328 	ip6 = mtod(m, struct ip6_hdr *);
329 
330 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
331 		goto out;
332 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
333 		goto out;
334 
335 	*src = ip6->ip6_src;
336 	icmp6_error(m, type, code, param);
337 	return;
338 
339 out:
340 	m_freem(m);
341 }
342 
343 /*
344  * Generate an error packet of type error in response to bad IP6 packet.
345  */
346 void
347 icmp6_error(struct mbuf *m, int type, int code, int param)
348 {
349 	struct ip6_hdr *oip6, *nip6;
350 	struct icmp6_hdr *icmp6;
351 	u_int preplen;
352 	int off;
353 	int nxt;
354 
355 	ICMP6_STATINC(ICMP6_STAT_ERROR);
356 
357 	/* count per-type-code statistics */
358 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
359 
360 	if (m->m_flags & M_DECRYPTED) {
361 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
362 		goto freeit;
363 	}
364 
365 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
366 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
367 		return;
368 	oip6 = mtod(m, struct ip6_hdr *);
369 
370 	/*
371 	 * If the destination address of the erroneous packet is a multicast
372 	 * address, or the packet was sent using link-layer multicast,
373 	 * we should basically suppress sending an error (RFC 2463, Section
374 	 * 2.4).
375 	 * We have two exceptions (the item e.2 in that section):
376 	 * - the Packet Too Big message can be sent for path MTU discovery.
377 	 * - the Parameter Problem Message that can be allowed an icmp6 error
378 	 *   in the option type field.  This check has been done in
379 	 *   ip6_unknown_opt(), so we can just check the type and code.
380 	 */
381 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
382 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
383 	    (type != ICMP6_PACKET_TOO_BIG &&
384 	     (type != ICMP6_PARAM_PROB ||
385 	      code != ICMP6_PARAMPROB_OPTION)))
386 		goto freeit;
387 
388 	/*
389 	 * RFC 2463, 2.4 (e.5): source address check.
390 	 * XXX: the case of anycast source?
391 	 */
392 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
393 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
394 		goto freeit;
395 
396 	/*
397 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
398 	 * don't do it.
399 	 */
400 	nxt = -1;
401 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
402 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
403 		struct icmp6_hdr *icp;
404 
405 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
406 			sizeof(*icp));
407 		if (icp == NULL) {
408 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
409 			return;
410 		}
411 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
412 		    icp->icmp6_type == ND_REDIRECT) {
413 			/*
414 			 * ICMPv6 error
415 			 * Special case: for redirect (which is
416 			 * informational) we must not send icmp6 error.
417 			 */
418 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
419 			goto freeit;
420 		} else {
421 			/* ICMPv6 informational - send the error */
422 		}
423 	} else {
424 		/* non-ICMPv6 - send the error */
425 	}
426 
427 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
428 
429 	/* Finally, do rate limitation check. */
430 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
431 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
432 		goto freeit;
433 	}
434 
435 	/*
436 	 * OK, ICMP6 can be generated.
437 	 */
438 
439 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
440 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
441 
442 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
443 	M_PREPEND(m, preplen, M_DONTWAIT);
444 	if (m && M_UNWRITABLE(m, preplen))
445 		m = m_pullup(m, preplen);
446 	if (m == NULL) {
447 		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
448 		return;
449 	}
450 
451 	nip6 = mtod(m, struct ip6_hdr *);
452 	nip6->ip6_src  = oip6->ip6_src;
453 	nip6->ip6_dst  = oip6->ip6_dst;
454 
455 	in6_clearscope(&oip6->ip6_src);
456 	in6_clearscope(&oip6->ip6_dst);
457 
458 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
459 	icmp6->icmp6_type = type;
460 	icmp6->icmp6_code = code;
461 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
462 
463 	/*
464 	 * icmp6_reflect() is designed to be in the input path.
465 	 * icmp6_error() can be called from both input and output path,
466 	 * and if we are in output path rcvif could contain bogus value.
467 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
468 	 * information in ip header (nip6).
469 	 */
470 	m_reset_rcvif(m);
471 
472 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
473 
474 	/* header order: IPv6 - ICMPv6 */
475 	icmp6_reflect(m, sizeof(struct ip6_hdr));
476 
477 	return;
478 
479 freeit:
480 	/*
481 	 * If we can't tell whether or not we can generate ICMP6, free it.
482 	 */
483 	m_freem(m);
484 }
485 
486 /*
487  * Process a received ICMP6 message.
488  */
489 static void
490 _icmp6_input(struct mbuf *m, int off, int proto)
491 {
492 	struct mbuf *n;
493 	struct ip6_hdr *ip6, *nip6;
494 	struct icmp6_hdr *icmp6, *nicmp6;
495 	int icmp6len = m->m_pkthdr.len - off;
496 	int code, sum;
497 	struct ifnet *rcvif;
498 	struct psref psref;
499 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
500 
501 	rcvif = m_get_rcvif_psref(m, &psref);
502 	if (__predict_false(rcvif == NULL))
503 		goto freeit;
504 
505 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
506 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
507 	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
508 
509 	/*
510 	 * Locate icmp6 structure in mbuf, and check
511 	 * that not corrupted and of at least minimum length
512 	 */
513 
514 	if (icmp6len < sizeof(struct icmp6_hdr)) {
515 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
516 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
517 		goto freeit;
518 	}
519 
520 	if (m->m_len < sizeof(struct ip6_hdr)) {
521 		m = m_pullup(m, sizeof(struct ip6_hdr));
522 		if (m == NULL) {
523 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
524 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
525 			goto freeit;
526 		}
527 	}
528 
529 	ip6 = mtod(m, struct ip6_hdr *);
530 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
531 	if (icmp6 == NULL) {
532 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
533 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
534 		goto freeit;
535 	}
536 
537 	/*
538 	 * Enforce alignment requirements that are violated in
539 	 * some cases, see kern/50766 for details.
540 	 */
541 	if (ACCESSIBLE_POINTER(icmp6, struct ip6_hdr) == 0) {
542 		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
543 		if (m == NULL) {
544 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
545 			icmp6_ifstat_inc(rcvif, ifs6_in_error);
546 			goto freeit;
547 		}
548 		ip6 = mtod(m, struct ip6_hdr *);
549 		icmp6 = (struct icmp6_hdr *)(mtod(m, char *) + off);
550 	}
551 	KASSERT(ACCESSIBLE_POINTER(icmp6, struct ip6_hdr));
552 
553 	/*
554 	 * calculate the checksum
555 	 */
556 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
557 		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
558 		    icmp6->icmp6_type, sum, IN6_PRINT(ip6buf, &ip6->ip6_src));
559 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
560 		icmp6_ifstat_inc(rcvif, ifs6_in_error);
561 		goto freeit;
562 	}
563 
564 #if defined(NFAITH) && 0 < NFAITH
565 	if (faithprefix(&ip6->ip6_dst)) {
566 		/*
567 		 * Deliver very specific ICMP6 type only.
568 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
569 		 * will not work.
570 		 */
571 		switch (icmp6->icmp6_type) {
572 		case ICMP6_DST_UNREACH:
573 		case ICMP6_PACKET_TOO_BIG:
574 		case ICMP6_TIME_EXCEEDED:
575 			break;
576 		default:
577 			goto freeit;
578 		}
579 	}
580 #endif
581 
582 	code = icmp6->icmp6_code;
583 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
584 
585 	switch (icmp6->icmp6_type) {
586 	case ICMP6_DST_UNREACH:
587 		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
588 		switch (code) {
589 		case ICMP6_DST_UNREACH_NOROUTE:
590 			code = PRC_UNREACH_NET;
591 			break;
592 		case ICMP6_DST_UNREACH_ADMIN:
593 			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
594 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
595 			break;
596 		case ICMP6_DST_UNREACH_ADDR:
597 			code = PRC_HOSTDEAD;
598 			break;
599 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
600 			/* I mean "source address was incorrect." */
601 			code = PRC_UNREACH_NET;
602 			break;
603 		case ICMP6_DST_UNREACH_NOPORT:
604 			code = PRC_UNREACH_PORT;
605 			break;
606 		default:
607 			goto badcode;
608 		}
609 		goto deliver;
610 
611 	case ICMP6_PACKET_TOO_BIG:
612 		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
613 
614 		/*
615 		 * MTU is checked in icmp6_mtudisc.
616 		 */
617 		code = PRC_MSGSIZE;
618 
619 		/*
620 		 * Updating the path MTU will be done after examining
621 		 * intermediate extension headers.
622 		 */
623 		goto deliver;
624 
625 	case ICMP6_TIME_EXCEEDED:
626 		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
627 		switch (code) {
628 		case ICMP6_TIME_EXCEED_TRANSIT:
629 			code = PRC_TIMXCEED_INTRANS;
630 			break;
631 		case ICMP6_TIME_EXCEED_REASSEMBLY:
632 			code = PRC_TIMXCEED_REASS;
633 			break;
634 		default:
635 			goto badcode;
636 		}
637 		goto deliver;
638 
639 	case ICMP6_PARAM_PROB:
640 		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
641 		switch (code) {
642 		case ICMP6_PARAMPROB_NEXTHEADER:
643 			code = PRC_UNREACH_PROTOCOL;
644 			break;
645 		case ICMP6_PARAMPROB_HEADER:
646 		case ICMP6_PARAMPROB_OPTION:
647 			code = PRC_PARAMPROB;
648 			break;
649 		default:
650 			goto badcode;
651 		}
652 		goto deliver;
653 
654 	case ICMP6_ECHO_REQUEST:
655 		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
656 		if (code != 0)
657 			goto badcode;
658 		/*
659 		 * Copy mbuf to send to two data paths: userland socket(s),
660 		 * and to the querier (echo reply).
661 		 * m: a copy for socket, n: a copy for querier
662 		 *
663 		 * If the first mbuf is shared, or the first mbuf is too short,
664 		 * copy the first part of the data into a fresh mbuf.
665 		 * Otherwise, we will wrongly overwrite both copies.
666 		 */
667 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
668 			/* Give up local */
669 			n = m;
670 			m = NULL;
671 		} else if (M_UNWRITABLE(n, off + sizeof(struct icmp6_hdr))) {
672 			struct mbuf *n0 = n;
673 
674 			/*
675 			 * Prepare an internal mbuf.  m_pullup() doesn't
676 			 * always copy the length we specified.
677 			 */
678 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
679 				/* Give up local */
680 				n = m;
681 				m = NULL;
682 			}
683 			m_freem(n0);
684 		}
685 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
686 		    sizeof(*nicmp6));
687 		if (nicmp6 == NULL)
688 			goto freeit;
689 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
690 		nicmp6->icmp6_code = 0;
691 		if (n) {
692 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
693 			icmp6s[ICMP6_STAT_REFLECT]++;
694 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
695 			ICMP6_STAT_PUTREF();
696 			icmp6_reflect(n, off);
697 		}
698 		if (!m)
699 			goto freeit;
700 		break;
701 
702 	case ICMP6_ECHO_REPLY:
703 		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
704 		if (code != 0)
705 			goto badcode;
706 		break;
707 
708 	case MLD_LISTENER_QUERY:
709 	case MLD_LISTENER_REPORT:
710 		if (icmp6len < sizeof(struct mld_hdr))
711 			goto badlen;
712 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
713 			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
714 		else
715 			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
716 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
717 			/* give up local */
718 			mld_input(m, off);
719 			m = NULL;
720 			goto freeit;
721 		}
722 		mld_input(n, off);
723 		/* m stays. */
724 		break;
725 
726 	case MLD_LISTENER_DONE:
727 		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
728 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
729 			goto badlen;
730 		break;		/* nothing to be done in kernel */
731 
732 	case MLD_MTRACE_RESP:
733 	case MLD_MTRACE:
734 		/* XXX: these two are experimental.  not officially defined. */
735 		/* XXX: per-interface statistics? */
736 		break;		/* just pass it to applications */
737 
738 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
739 	    {
740 		enum { WRU, FQDN } mode;
741 
742 		if (!icmp6_nodeinfo)
743 			break;
744 
745 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
746 			mode = WRU;
747 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
748 			mode = FQDN;
749 		else
750 			goto badlen;
751 
752 		if (mode == FQDN) {
753 			n = m_copypacket(m, M_DONTWAIT);
754 			if (n)
755 				n = ni6_input(n, off);
756 		} else {
757 			u_char *p;
758 			int maxhlen;
759 
760 			if ((icmp6_nodeinfo & 5) != 5)
761 				break;
762 
763 			if (code != 0)
764 				goto badcode;
765 			MGETHDR(n, M_DONTWAIT, m->m_type);
766 			if (n && ICMP6_MAXLEN > MHLEN) {
767 				MCLGET(n, M_DONTWAIT);
768 				if ((n->m_flags & M_EXT) == 0) {
769 					m_free(n);
770 					n = NULL;
771 				}
772 			}
773 			if (n == NULL) {
774 				/* Give up remote */
775 				break;
776 			}
777 			m_reset_rcvif(n);
778 			n->m_len = 0;
779 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
780 			if (maxhlen < 0) {
781 				m_free(n);
782 				break;
783 			}
784 			if (maxhlen > hostnamelen)
785 				maxhlen = hostnamelen;
786 			/*
787 			 * Copy IPv6 and ICMPv6 only.
788 			 */
789 			nip6 = mtod(n, struct ip6_hdr *);
790 			memcpy(nip6, ip6, sizeof(struct ip6_hdr));
791 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
792 			memcpy(nicmp6, icmp6, sizeof(struct icmp6_hdr));
793 
794 			p = (u_char *)(nicmp6 + 1);
795 			memset(p, 0, 4);
796 			memcpy(p + 4, hostname, maxhlen); /* meaningless TTL */
797 
798 			m_copy_pkthdr(n, m);
799 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
800 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
801 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
802 			nicmp6->icmp6_code = 0;
803 		}
804 		if (n) {
805 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
806 			icmp6s[ICMP6_STAT_REFLECT]++;
807 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
808 			ICMP6_STAT_PUTREF();
809 			icmp6_reflect(n, sizeof(struct ip6_hdr));
810 		}
811 		break;
812 	    }
813 
814 	case ICMP6_WRUREPLY:
815 		if (code != 0)
816 			goto badcode;
817 		break;
818 
819 	case ND_ROUTER_SOLICIT:
820 		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
821 		/* FALLTHROUGH */
822 	case ND_ROUTER_ADVERT:
823 		if (icmp6->icmp6_type == ND_ROUTER_ADVERT)
824 			icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
825 		if (code != 0)
826 			goto badcode;
827 		if ((icmp6->icmp6_type == ND_ROUTER_SOLICIT &&
828 		    icmp6len < sizeof(struct nd_router_solicit)) ||
829 		    (icmp6->icmp6_type == ND_ROUTER_ADVERT &&
830 		    icmp6len < sizeof(struct nd_router_advert)))
831 			goto badlen;
832 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
833 			/* give up local */
834 			nd6_rtr_cache(m, off, icmp6len, icmp6->icmp6_type);
835 			m = NULL;
836 			goto freeit;
837 		}
838 		nd6_rtr_cache(n, off, icmp6len, icmp6->icmp6_type);
839 		/* m stays. */
840 		break;
841 
842 	case ND_NEIGHBOR_SOLICIT:
843 		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
844 		if (code != 0)
845 			goto badcode;
846 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
847 			goto badlen;
848 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
849 			/* give up local */
850 			nd6_ns_input(m, off, icmp6len);
851 			m = NULL;
852 			goto freeit;
853 		}
854 		nd6_ns_input(n, off, icmp6len);
855 		/* m stays. */
856 		break;
857 
858 	case ND_NEIGHBOR_ADVERT:
859 		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
860 		if (code != 0)
861 			goto badcode;
862 		if (icmp6len < sizeof(struct nd_neighbor_advert))
863 			goto badlen;
864 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
865 			/* give up local */
866 			nd6_na_input(m, off, icmp6len);
867 			m = NULL;
868 			goto freeit;
869 		}
870 		nd6_na_input(n, off, icmp6len);
871 		/* m stays. */
872 		break;
873 
874 	case ND_REDIRECT:
875 		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
876 		if (code != 0)
877 			goto badcode;
878 		if (icmp6len < sizeof(struct nd_redirect))
879 			goto badlen;
880 		if ((n = m_copypacket(m, M_DONTWAIT)) == NULL) {
881 			/* give up local */
882 			icmp6_redirect_input(m, off);
883 			m = NULL;
884 			goto freeit;
885 		}
886 		icmp6_redirect_input(n, off);
887 		/* m stays. */
888 		break;
889 
890 	case ICMP6_ROUTER_RENUMBERING:
891 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
892 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
893 			goto badcode;
894 		if (icmp6len < sizeof(struct icmp6_router_renum))
895 			goto badlen;
896 		break;
897 
898 	default:
899 		nd6log(LOG_DEBUG,
900 		    "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
901 		    icmp6->icmp6_type,
902 		    IN6_PRINT(ip6buf, &ip6->ip6_src),
903 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst),
904 		    rcvif ? rcvif->if_index : 0);
905 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
906 			/* ICMPv6 error: MUST deliver it by spec... */
907 			code = PRC_NCMDS;
908 			/* deliver */
909 		} else {
910 			/* ICMPv6 informational: MUST not deliver */
911 			break;
912 		}
913 	deliver:
914 		if (icmp6_notify_error(m, off, icmp6len, code)) {
915 			/* In this case, m should've been freed. */
916 			m_put_rcvif_psref(rcvif, &psref);
917 			return;
918 		}
919 		break;
920 
921 	badcode:
922 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
923 		break;
924 
925 	badlen:
926 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
927 		break;
928 	}
929 	m_put_rcvif_psref(rcvif, &psref);
930 
931 	/* deliver the packet to appropriate sockets */
932 	icmp6_rip6_input(&m, off);
933 
934 	return;
935 
936 freeit:
937 	m_put_rcvif_psref(rcvif, &psref);
938 	m_freem(m);
939 	return;
940 }
941 
942 int
943 icmp6_input(struct mbuf **mp, int *offp, int proto)
944 {
945 
946 	wqinput_input(icmp6_wqinput, *mp, *offp, proto);
947 
948 	return IPPROTO_DONE;
949 }
950 
951 static int
952 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
953 {
954 	struct icmp6_hdr *icmp6;
955 	struct ip6_hdr *eip6;
956 	u_int32_t notifymtu;
957 	struct sockaddr_in6 icmp6src, icmp6dst;
958 
959 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
960 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
961 		goto freeit;
962 	}
963 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
964 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
965 	if (icmp6 == NULL) {
966 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
967 		return (-1);
968 	}
969 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
970 
971 	/* Detect the upper level protocol */
972 	{
973 		void *(*ctlfunc)(int, const struct sockaddr *, void *);
974 		u_int8_t nxt = eip6->ip6_nxt;
975 		int eoff = off + sizeof(struct icmp6_hdr) +
976 			sizeof(struct ip6_hdr);
977 		struct ip6ctlparam ip6cp;
978 		struct in6_addr *finaldst = NULL;
979 		int icmp6type = icmp6->icmp6_type;
980 		struct ip6_frag *fh;
981 		struct ip6_rthdr *rth;
982 		struct ifnet *rcvif;
983 		int s;
984 
985 		while (1) { /* XXX: should avoid infinite loop explicitly? */
986 			struct ip6_ext *eh;
987 
988 			switch (nxt) {
989 			case IPPROTO_HOPOPTS:
990 			case IPPROTO_DSTOPTS:
991 			case IPPROTO_AH:
992 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
993 				    eoff, sizeof(*eh));
994 				if (eh == NULL) {
995 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
996 					return (-1);
997 				}
998 
999 				if (nxt == IPPROTO_AH)
1000 					eoff += (eh->ip6e_len + 2) << 2;
1001 				else
1002 					eoff += (eh->ip6e_len + 1) << 3;
1003 				nxt = eh->ip6e_nxt;
1004 				break;
1005 			case IPPROTO_ROUTING:
1006 				/* Ignore the option. */
1007 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1008 				    eoff, sizeof(*rth));
1009 				if (rth == NULL) {
1010 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1011 					return (-1);
1012 				}
1013 
1014 				eoff += (rth->ip6r_len + 1) << 3;
1015 				nxt = rth->ip6r_nxt;
1016 				break;
1017 			case IPPROTO_FRAGMENT:
1018 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1019 				    eoff, sizeof(*fh));
1020 				if (fh == NULL) {
1021 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1022 					return (-1);
1023 				}
1024 				/*
1025 				 * Data after a fragment header is meaningless
1026 				 * unless it is the first fragment, but
1027 				 * we'll go to the notify label for path MTU
1028 				 * discovery.
1029 				 */
1030 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1031 					goto notify;
1032 
1033 				eoff += sizeof(struct ip6_frag);
1034 				nxt = fh->ip6f_nxt;
1035 				break;
1036 			default:
1037 				/*
1038 				 * This case includes ESP and the No Next
1039 				 * Header.  In such cases going to the notify
1040 				 * label does not have any meaning
1041 				 * (i.e. ctlfunc will be NULL), but we go
1042 				 * anyway since we might have to update
1043 				 * path MTU information.
1044 				 */
1045 				goto notify;
1046 			}
1047 		}
1048 	  notify:
1049 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1050 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1051 		if (icmp6 == NULL) {
1052 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1053 			return (-1);
1054 		}
1055 
1056 		/*
1057 		 * retrieve parameters from the inner IPv6 header, and convert
1058 		 * them into sockaddr structures.
1059 		 * XXX: there is no guarantee that the source or destination
1060 		 * addresses of the inner packet are in the same scope zone as
1061 		 * the addresses of the icmp packet.  But there is no other
1062 		 * way to determine the zone.
1063 		 */
1064 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1065 
1066 		rcvif = m_get_rcvif(m, &s);
1067 		if (__predict_false(rcvif == NULL))
1068 			goto freeit;
1069 		sockaddr_in6_init(&icmp6dst,
1070 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1071 		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1072 			m_put_rcvif(rcvif, &s);
1073 			goto freeit;
1074 		}
1075 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1076 		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1077 			m_put_rcvif(rcvif, &s);
1078 			goto freeit;
1079 		}
1080 		m_put_rcvif(rcvif, &s);
1081 
1082 		icmp6src.sin6_flowinfo =
1083 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1084 
1085 		if (finaldst == NULL)
1086 			finaldst = &eip6->ip6_dst;
1087 		ip6cp.ip6c_m = m;
1088 		ip6cp.ip6c_icmp6 = icmp6;
1089 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1090 		ip6cp.ip6c_off = eoff;
1091 		ip6cp.ip6c_finaldst = finaldst;
1092 		ip6cp.ip6c_src = &icmp6src;
1093 		ip6cp.ip6c_nxt = nxt;
1094 
1095 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1096 			notifymtu = ntohl(icmp6->icmp6_mtu);
1097 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1098 		}
1099 
1100 		ctlfunc = inet6sw[ip6_protox[nxt]].pr_ctlinput;
1101 		if (ctlfunc) {
1102 			(void)(*ctlfunc)(code, sin6tosa(&icmp6dst), &ip6cp);
1103 		}
1104 	}
1105 	return (0);
1106 
1107 freeit:
1108 	m_freem(m);
1109 	return (-1);
1110 }
1111 
1112 void
1113 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1114 {
1115 	unsigned long rtcount;
1116 	struct icmp6_mtudisc_callback *mc;
1117 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1118 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1119 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1120 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1121 	struct rtentry *rt = NULL;
1122 	struct sockaddr_in6 sin6;
1123 	struct ifnet *rcvif;
1124 	int s;
1125 
1126 	/*
1127 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1128 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1129 	 * In that one case, the MTU might be less than 1280.
1130 	 */
1131 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1132 		/* is the mtu even sane? */
1133 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1134 			return;
1135 		if (!validated)
1136 			return;
1137 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1138 	}
1139 
1140 	/*
1141 	 * allow non-validated cases if memory is plenty, to make traffic
1142 	 * from non-connected pcb happy.
1143 	 */
1144 	mutex_enter(&icmp6_mtx);
1145 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1146 	if (validated) {
1147 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) {
1148 			mutex_exit(&icmp6_mtx);
1149 			return;
1150 		} else if (0 <= icmp6_mtudisc_lowat &&
1151 		    rtcount > icmp6_mtudisc_lowat) {
1152 			/*
1153 			 * XXX nuke a victim, install the new one.
1154 			 */
1155 		}
1156 	} else {
1157 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) {
1158 			mutex_exit(&icmp6_mtx);
1159 			return;
1160 		}
1161 	}
1162 	mutex_exit(&icmp6_mtx);
1163 
1164 	memset(&sin6, 0, sizeof(sin6));
1165 	sin6.sin6_family = PF_INET6;
1166 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1167 	sin6.sin6_addr = *dst;
1168 	rcvif = m_get_rcvif(m, &s);
1169 	if (__predict_false(rcvif == NULL))
1170 		return;
1171 	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1172 		m_put_rcvif(rcvif, &s);
1173 		return;
1174 	}
1175 	m_put_rcvif(rcvif, &s);
1176 
1177 	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1178 
1179 	if (rt && (rt->rt_flags & RTF_HOST) &&
1180 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1181 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1182 		if (mtu < rt->rt_ifp->if_mtu) {
1183 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1184 			rt->rt_rmx.rmx_mtu = mtu;
1185 		}
1186 	}
1187 	if (rt) {
1188 		rt_unref(rt);
1189 	}
1190 
1191 	/*
1192 	 * Notify protocols that the MTU for this destination
1193 	 * has changed.
1194 	 */
1195 	mutex_enter(&icmp6_mtx);
1196 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1197 	     mc = LIST_NEXT(mc, mc_list))
1198 		(*mc->mc_func)(&sin6.sin6_addr);
1199 	mutex_exit(&icmp6_mtx);
1200 }
1201 
1202 /*
1203  * Process a Node Information Query packet, based on
1204  * draft-ietf-ipngwg-icmp-name-lookups-07.
1205  *
1206  * Spec incompatibilities:
1207  * - IPv6 Subject address handling
1208  * - IPv4 Subject address handling support missing
1209  * - Proxy reply (answer even if it's not for me)
1210  * - joins NI group address at in6_ifattach() time only, does not cope
1211  *   with hostname changes by sethostname(3)
1212  */
1213 static struct mbuf *
1214 ni6_input(struct mbuf *m, int off)
1215 {
1216 	struct icmp6_nodeinfo *ni6, *nni6;
1217 	struct mbuf *n = NULL;
1218 	u_int16_t qtype;
1219 	int subjlen;
1220 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1221 	struct ni_reply_fqdn *fqdn;
1222 	int addrs;		/* for NI_QTYPE_NODEADDR */
1223 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1224 	struct sockaddr_in6 sin6; /* ip6_dst */
1225 	struct in6_addr in6_subj; /* subject address */
1226 	struct ip6_hdr *ip6;
1227 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1228 	char *subj = NULL;
1229 	struct ifnet *rcvif;
1230 	int s, ss;
1231 	struct ifaddr *ifa;
1232 	struct psref psref;
1233 
1234 	ip6 = mtod(m, struct ip6_hdr *);
1235 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1236 	if (ni6 == NULL) {
1237 		/* m is already reclaimed */
1238 		return NULL;
1239 	}
1240 	KASSERT((m->m_flags & M_PKTHDR) != 0);
1241 
1242 	/*
1243 	 * Validate IPv6 destination address.
1244 	 *
1245 	 * The Responder must discard the Query without further processing
1246 	 * unless it is one of the Responder's unicast or anycast addresses, or
1247 	 * a link-local scope multicast address which the Responder has joined.
1248 	 * [icmp-name-lookups-07, Section 4.]
1249 	 */
1250 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1251 	/* XXX scopeid */
1252 	ss = pserialize_read_enter();
1253 	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1254 	if (ifa != NULL) {
1255 		; /* unicast/anycast, fine */
1256 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) {
1257 		; /* link-local multicast, fine */
1258 	} else {
1259 		pserialize_read_exit(ss);
1260 		goto bad;
1261 	}
1262 	pserialize_read_exit(ss);
1263 
1264 	/* validate query Subject field. */
1265 	qtype = ntohs(ni6->ni_qtype);
1266 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1267 	switch (qtype) {
1268 	case NI_QTYPE_NOOP:
1269 	case NI_QTYPE_SUPTYPES:
1270 		/* 07 draft */
1271 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1272 			break;
1273 		/* FALLTHROUGH */
1274 	case NI_QTYPE_FQDN:
1275 	case NI_QTYPE_NODEADDR:
1276 	case NI_QTYPE_IPV4ADDR:
1277 		switch (ni6->ni_code) {
1278 		case ICMP6_NI_SUBJ_IPV6:
1279 #if ICMP6_NI_SUBJ_IPV6 != 0
1280 		case 0:
1281 #endif
1282 			/*
1283 			 * backward compatibility - try to accept 03 draft
1284 			 * format, where no Subject is present.
1285 			 */
1286 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1287 			    subjlen == 0) {
1288 				oldfqdn++;
1289 				break;
1290 			}
1291 #if ICMP6_NI_SUBJ_IPV6 != 0
1292 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1293 				goto bad;
1294 #endif
1295 
1296 			if (subjlen != sizeof(sin6.sin6_addr))
1297 				goto bad;
1298 
1299 			/*
1300 			 * Validate Subject address.
1301 			 *
1302 			 * Not sure what exactly "address belongs to the node"
1303 			 * means in the spec, is it just unicast, or what?
1304 			 *
1305 			 * At this moment we consider Subject address as
1306 			 * "belong to the node" if the Subject address equals
1307 			 * to the IPv6 destination address; validation for
1308 			 * IPv6 destination address should have done enough
1309 			 * check for us.
1310 			 *
1311 			 * We do not do proxy at this moment.
1312 			 */
1313 			/* m_pulldown instead of copy? */
1314 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1315 			    subjlen, (void *)&in6_subj);
1316 			rcvif = m_get_rcvif(m, &s);
1317 			if (__predict_false(rcvif == NULL))
1318 				goto bad;
1319 			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1320 				m_put_rcvif(rcvif, &s);
1321 				goto bad;
1322 			}
1323 			m_put_rcvif(rcvif, &s);
1324 
1325 			subj = (char *)&in6_subj;
1326 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1327 				break;
1328 
1329 			/*
1330 			 * XXX if we are to allow other cases, we should really
1331 			 * be careful about scope here.
1332 			 * basically, we should disallow queries toward IPv6
1333 			 * destination X with subject Y, if scope(X) > scope(Y).
1334 			 * if we allow scope(X) > scope(Y), it will result in
1335 			 * information leakage across scope boundary.
1336 			 */
1337 			goto bad;
1338 
1339 		case ICMP6_NI_SUBJ_FQDN:
1340 			/*
1341 			 * Validate Subject name with gethostname(3).
1342 			 *
1343 			 * The behavior may need some debate, since:
1344 			 * - we are not sure if the node has FQDN as
1345 			 *   hostname (returned by gethostname(3)).
1346 			 * - the code does wildcard match for truncated names.
1347 			 *   however, we are not sure if we want to perform
1348 			 *   wildcard match, if gethostname(3) side has
1349 			 *   truncated hostname.
1350 			 */
1351 			n = ni6_nametodns(hostname, hostnamelen, 0);
1352 			if (!n || n->m_next || n->m_len == 0)
1353 				goto bad;
1354 			IP6_EXTHDR_GET(subj, char *, m,
1355 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1356 			if (subj == NULL)
1357 				goto bad;
1358 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1359 			    n->m_len)) {
1360 				goto bad;
1361 			}
1362 			m_freem(n);
1363 			n = NULL;
1364 			break;
1365 
1366 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1367 		default:
1368 			goto bad;
1369 		}
1370 		break;
1371 	}
1372 
1373 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1374 	switch (qtype) {
1375 	case NI_QTYPE_FQDN:
1376 		if ((icmp6_nodeinfo & 1) == 0)
1377 			goto bad;
1378 		break;
1379 	case NI_QTYPE_NODEADDR:
1380 	case NI_QTYPE_IPV4ADDR:
1381 		if ((icmp6_nodeinfo & 2) == 0)
1382 			goto bad;
1383 		break;
1384 	}
1385 
1386 	/* guess reply length */
1387 	switch (qtype) {
1388 	case NI_QTYPE_NOOP:
1389 		break;		/* no reply data */
1390 	case NI_QTYPE_SUPTYPES:
1391 		replylen += sizeof(u_int32_t);
1392 		break;
1393 	case NI_QTYPE_FQDN:
1394 		/* will append an mbuf */
1395 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1396 		break;
1397 	case NI_QTYPE_NODEADDR:
1398 		addrs = ni6_addrs(ni6, &ifp, subj, &psref);
1399 		replylen += addrs *
1400 		    (sizeof(struct in6_addr) + sizeof(u_int32_t));
1401 		if (replylen > MCLBYTES)
1402 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1403 		break;
1404 	case NI_QTYPE_IPV4ADDR:
1405 		/* unsupported - should respond with unknown Qtype? */
1406 		goto bad;
1407 	default:
1408 		/*
1409 		 * XXX: We must return a reply with the ICMP6 code
1410 		 * `unknown Qtype' in this case.  However we regard the case
1411 		 * as an FQDN query for backward compatibility.
1412 		 * Older versions set a random value to this field,
1413 		 * so it rarely varies in the defined qtypes.
1414 		 * But the mechanism is not reliable...
1415 		 * maybe we should obsolete older versions.
1416 		 */
1417 		qtype = NI_QTYPE_FQDN;
1418 		/* will append an mbuf */
1419 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1420 		oldfqdn++;
1421 		break;
1422 	}
1423 
1424 	/* allocate an mbuf to reply. */
1425 	MGETHDR(n, M_DONTWAIT, m->m_type);
1426 	if (n == NULL) {
1427 		goto bad;
1428 	}
1429 	m_move_pkthdr(n, m);
1430 	if (replylen > MHLEN) {
1431 		if (replylen > MCLBYTES) {
1432 			/*
1433 			 * XXX: should we try to allocate more? But MCLBYTES
1434 			 * is probably much larger than IPV6_MMTU...
1435 			 */
1436 			goto bad;
1437 		}
1438 		MCLGET(n, M_DONTWAIT);
1439 		if ((n->m_flags & M_EXT) == 0) {
1440 			goto bad;
1441 		}
1442 	}
1443 	n->m_pkthdr.len = n->m_len = replylen;
1444 
1445 	/* copy mbuf header and IPv6 + Node Information base headers */
1446 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1447 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1448 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1449 
1450 	/* qtype dependent procedure */
1451 	switch (qtype) {
1452 	case NI_QTYPE_NOOP:
1453 		nni6->ni_code = ICMP6_NI_SUCCESS;
1454 		nni6->ni_flags = 0;
1455 		break;
1456 	case NI_QTYPE_SUPTYPES:
1457 	{
1458 		u_int32_t v;
1459 		nni6->ni_code = ICMP6_NI_SUCCESS;
1460 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1461 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1462 		v = (u_int32_t)htonl(0x0000000f);
1463 		memcpy(nni6 + 1, &v, sizeof(u_int32_t));
1464 		break;
1465 	}
1466 	case NI_QTYPE_FQDN:
1467 		nni6->ni_code = ICMP6_NI_SUCCESS;
1468 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1469 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1470 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1471 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1472 		/*
1473 		 * XXX do we really have FQDN in variable "hostname"?
1474 		 */
1475 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1476 		if (n->m_next == NULL)
1477 			goto bad;
1478 		/* XXX we assume that n->m_next is not a chain */
1479 		if (n->m_next->m_next != NULL)
1480 			goto bad;
1481 		n->m_pkthdr.len += n->m_next->m_len;
1482 		break;
1483 	case NI_QTYPE_NODEADDR:
1484 	{
1485 		int lenlim, copied;
1486 
1487 		nni6->ni_code = ICMP6_NI_SUCCESS;
1488 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1489 		    sizeof(struct icmp6_nodeinfo);
1490 		lenlim = M_TRAILINGSPACE(n);
1491 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1492 		if_put(ifp, &psref);
1493 		ifp = NULL;
1494 		/* update mbuf length */
1495 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1496 		    sizeof(struct icmp6_nodeinfo) + copied;
1497 		break;
1498 	}
1499 	default:
1500 		panic("%s: impossible", __func__);
1501 		break;
1502 	}
1503 
1504 	nni6->ni_type = ICMP6_NI_REPLY;
1505 	m_freem(m);
1506 	return n;
1507 
1508 bad:
1509 	if_put(ifp, &psref);
1510 	m_freem(m);
1511 	if (n)
1512 		m_freem(n);
1513 	return NULL;
1514 }
1515 
1516 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1517 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1518 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1519 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1520 
1521 /*
1522  * make a mbuf with DNS-encoded string.  no compression support.
1523  *
1524  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1525  * treated as truncated name (two \0 at the end).  this is a wild guess.
1526  *
1527  * old - return pascal string if non-zero
1528  */
1529 static struct mbuf *
1530 ni6_nametodns(const char *name, int namelen, int old)
1531 {
1532 	struct mbuf *m;
1533 	char *cp, *ep;
1534 	const char *p, *q;
1535 	int i, len, nterm;
1536 
1537 	if (old)
1538 		len = namelen + 1;
1539 	else
1540 		len = MCLBYTES;
1541 
1542 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1543 	MGET(m, M_DONTWAIT, MT_DATA);
1544 	if (m && len > MLEN) {
1545 		MCLGET(m, M_DONTWAIT);
1546 		if ((m->m_flags & M_EXT) == 0)
1547 			goto fail;
1548 	}
1549 	if (!m)
1550 		goto fail;
1551 	m->m_next = NULL;
1552 
1553 	if (old) {
1554 		m->m_len = len;
1555 		*mtod(m, char *) = namelen;
1556 		memcpy(mtod(m, char *) + 1, name, namelen);
1557 		return m;
1558 	} else {
1559 		m->m_len = 0;
1560 		cp = mtod(m, char *);
1561 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1562 
1563 		/* if not certain about my name, return empty buffer */
1564 		if (namelen == 0)
1565 			return m;
1566 
1567 		/*
1568 		 * guess if it looks like shortened hostname, or FQDN.
1569 		 * shortened hostname needs two trailing "\0".
1570 		 */
1571 		i = 0;
1572 		for (p = name; p < name + namelen; p++) {
1573 			if (*p == '.')
1574 				i++;
1575 		}
1576 		if (i < 2)
1577 			nterm = 2;
1578 		else
1579 			nterm = 1;
1580 
1581 		p = name;
1582 		while (cp < ep && p < name + namelen) {
1583 			i = 0;
1584 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1585 				i++;
1586 			/* result does not fit into mbuf */
1587 			if (cp + i + 1 >= ep)
1588 				goto fail;
1589 			/*
1590 			 * DNS label length restriction, RFC1035 page 8.
1591 			 * "i == 0" case is included here to avoid returning
1592 			 * 0-length label on "foo..bar".
1593 			 */
1594 			if (i <= 0 || i >= 64)
1595 				goto fail;
1596 			*cp++ = i;
1597 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1598 				goto fail;
1599 			while (i > 0) {
1600 				if (!isalnum(*p) && *p != '-')
1601 					goto fail;
1602 				if (isupper(*p)) {
1603 					*cp++ = tolower(*p);
1604 					p++;
1605 				} else
1606 					*cp++ = *p++;
1607 				i--;
1608 			}
1609 			p = q;
1610 			if (p < name + namelen && *p == '.')
1611 				p++;
1612 		}
1613 		/* termination */
1614 		if (cp + nterm >= ep)
1615 			goto fail;
1616 		while (nterm-- > 0)
1617 			*cp++ = '\0';
1618 		m->m_len = cp - mtod(m, char *);
1619 		return m;
1620 	}
1621 
1622 	panic("should not reach here");
1623 	/* NOTREACHED */
1624 
1625 fail:
1626 	if (m)
1627 		m_freem(m);
1628 	return NULL;
1629 }
1630 
1631 /*
1632  * check if two DNS-encoded string matches.  takes care of truncated
1633  * form (with \0\0 at the end).  no compression support.
1634  * XXX upper/lowercase match (see RFC2065)
1635  */
1636 static int
1637 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1638 {
1639 	const char *a0, *b0;
1640 	int l;
1641 
1642 	/* simplest case - need validation? */
1643 	if (alen == blen && memcmp(a, b, alen) == 0)
1644 		return 1;
1645 
1646 	a0 = a;
1647 	b0 = b;
1648 
1649 	/* termination is mandatory */
1650 	if (alen < 2 || blen < 2)
1651 		return 0;
1652 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1653 		return 0;
1654 	alen--;
1655 	blen--;
1656 
1657 	while (a - a0 < alen && b - b0 < blen) {
1658 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1659 			return 0;
1660 
1661 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1662 			return 0;
1663 		/* we don't support compression yet */
1664 		if (a[0] >= 64 || b[0] >= 64)
1665 			return 0;
1666 
1667 		/* truncated case */
1668 		if (a[0] == 0 && a - a0 == alen - 1)
1669 			return 1;
1670 		if (b[0] == 0 && b - b0 == blen - 1)
1671 			return 1;
1672 		if (a[0] == 0 || b[0] == 0)
1673 			return 0;
1674 
1675 		if (a[0] != b[0])
1676 			return 0;
1677 		l = a[0];
1678 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1679 			return 0;
1680 		if (memcmp(a + 1, b + 1, l) != 0)
1681 			return 0;
1682 
1683 		a += 1 + l;
1684 		b += 1 + l;
1685 	}
1686 
1687 	if (a - a0 == alen && b - b0 == blen)
1688 		return 1;
1689 	else
1690 		return 0;
1691 }
1692 
1693 /*
1694  * calculate the number of addresses to be returned in the node info reply.
1695  */
1696 static int
1697 ni6_addrs(struct icmp6_nodeinfo *ni6, struct ifnet **ifpp, char *subj,
1698     struct psref *psref)
1699 {
1700 	struct ifnet *ifp;
1701 	struct in6_ifaddr *ia6;
1702 	struct ifaddr *ifa;
1703 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1704 	int addrs = 0, addrsofif, iffound = 0;
1705 	int niflags = ni6->ni_flags;
1706 	int s;
1707 
1708 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1709 		switch (ni6->ni_code) {
1710 		case ICMP6_NI_SUBJ_IPV6:
1711 			if (subj == NULL) /* must be impossible... */
1712 				return 0;
1713 			subj_ip6 = (struct sockaddr_in6 *)subj;
1714 			break;
1715 		default:
1716 			/*
1717 			 * XXX: we only support IPv6 subject address for
1718 			 * this Qtype.
1719 			 */
1720 			return 0;
1721 		}
1722 	}
1723 
1724 	s = pserialize_read_enter();
1725 	IFNET_READER_FOREACH(ifp) {
1726 		addrsofif = 0;
1727 		IFADDR_READER_FOREACH(ifa, ifp) {
1728 			if (ifa->ifa_addr->sa_family != AF_INET6)
1729 				continue;
1730 			ia6 = (struct in6_ifaddr *)ifa;
1731 
1732 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1733 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1734 			     &ia6->ia_addr.sin6_addr))
1735 				iffound = 1;
1736 
1737 			/*
1738 			 * IPv4-mapped addresses can only be returned by a
1739 			 * Node Information proxy, since they represent
1740 			 * addresses of IPv4-only nodes, which perforce do
1741 			 * not implement this protocol.
1742 			 * [icmp-name-lookups-07, Section 5.4]
1743 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1744 			 * this function at this moment.
1745 			 */
1746 
1747 			/* What do we have to do about ::1? */
1748 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1749 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1750 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1751 					continue;
1752 				break;
1753 			case IPV6_ADDR_SCOPE_SITELOCAL:
1754 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1755 					continue;
1756 				break;
1757 			case IPV6_ADDR_SCOPE_GLOBAL:
1758 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1759 					continue;
1760 				break;
1761 			default:
1762 				continue;
1763 			}
1764 
1765 			/*
1766 			 * check if anycast is okay.
1767 			 * XXX: just experimental.  not in the spec.
1768 			 */
1769 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1770 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1771 				continue; /* we need only unicast addresses */
1772 
1773 			addrsofif++; /* count the address */
1774 		}
1775 		if (iffound) {
1776 			if_acquire(ifp, psref);
1777 			pserialize_read_exit(s);
1778 			*ifpp = ifp;
1779 			return addrsofif;
1780 		}
1781 
1782 		addrs += addrsofif;
1783 	}
1784 	pserialize_read_exit(s);
1785 
1786 	return addrs;
1787 }
1788 
1789 static int
1790 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1791 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1792 	int resid)
1793 {
1794 	struct ifnet *ifp;
1795 	struct in6_ifaddr *ia6;
1796 	struct ifaddr *ifa;
1797 	struct ifnet *ifp_dep = NULL;
1798 	int copied = 0, allow_deprecated = 0;
1799 	u_char *cp = (u_char *)(nni6 + 1);
1800 	int niflags = ni6->ni_flags;
1801 	u_int32_t ltime;
1802 	int s;
1803 
1804 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1805 		return 0;	/* needless to copy */
1806 
1807 	s = pserialize_read_enter();
1808 	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1809 again:
1810 
1811 	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1812 	{
1813 		IFADDR_READER_FOREACH(ifa, ifp) {
1814 			if (ifa->ifa_addr->sa_family != AF_INET6)
1815 				continue;
1816 			ia6 = (struct in6_ifaddr *)ifa;
1817 
1818 			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1819 			    allow_deprecated == 0) {
1820 				/*
1821 				 * prefererred address should be put before
1822 				 * deprecated addresses.
1823 				 */
1824 
1825 				/* record the interface for later search */
1826 				if (ifp_dep == NULL)
1827 					ifp_dep = ifp;
1828 
1829 				continue;
1830 			}
1831 			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1832 				 allow_deprecated != 0)
1833 				continue; /* we now collect deprecated addrs */
1834 
1835 			/* What do we have to do about ::1? */
1836 			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1837 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1838 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1839 					continue;
1840 				break;
1841 			case IPV6_ADDR_SCOPE_SITELOCAL:
1842 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1843 					continue;
1844 				break;
1845 			case IPV6_ADDR_SCOPE_GLOBAL:
1846 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1847 					continue;
1848 				break;
1849 			default:
1850 				continue;
1851 			}
1852 
1853 			/*
1854 			 * check if anycast is okay.
1855 			 * XXX: just experimental.  not in the spec.
1856 			 */
1857 			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1858 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1859 				continue;
1860 
1861 			/* now we can copy the address */
1862 			if (resid < sizeof(struct in6_addr) +
1863 			    sizeof(u_int32_t)) {
1864 				/*
1865 				 * We give up much more copy.
1866 				 * Set the truncate flag and return.
1867 				 */
1868 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1869 				goto out;
1870 			}
1871 
1872 			/*
1873 			 * Set the TTL of the address.
1874 			 * The TTL value should be one of the following
1875 			 * according to the specification:
1876 			 *
1877 			 * 1. The remaining lifetime of a DHCP lease on the
1878 			 *    address, or
1879 			 * 2. The remaining Valid Lifetime of a prefix from
1880 			 *    which the address was derived through Stateless
1881 			 *    Autoconfiguration.
1882 			 *
1883 			 * Note that we currently do not support stateful
1884 			 * address configuration by DHCPv6, so the former
1885 			 * case can't happen.
1886 			 *
1887 			 * TTL must be 2^31 > TTL >= 0.
1888 			 */
1889 			if (ia6->ia6_lifetime.ia6t_expire == 0)
1890 				ltime = ND6_INFINITE_LIFETIME;
1891 			else {
1892 				if (ia6->ia6_lifetime.ia6t_expire >
1893 				    time_uptime)
1894 					ltime = ia6->ia6_lifetime.ia6t_expire -
1895 					    time_uptime;
1896 				else
1897 					ltime = 0;
1898 			}
1899 			if (ltime > 0x7fffffff)
1900 				ltime = 0x7fffffff;
1901 			ltime = htonl(ltime);
1902 
1903 			memcpy(cp, &ltime, sizeof(u_int32_t));
1904 			cp += sizeof(u_int32_t);
1905 
1906 			/* copy the address itself */
1907 			bcopy(&ia6->ia_addr.sin6_addr, cp,
1908 			      sizeof(struct in6_addr));
1909 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1910 			cp += sizeof(struct in6_addr);
1911 
1912 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1913 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1914 		}
1915 		if (ifp0)	/* we need search only on the specified IF */
1916 			break;
1917 	}
1918 
1919 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1920 		ifp = ifp_dep;
1921 		allow_deprecated = 1;
1922 
1923 		goto again;
1924 	}
1925 out:
1926 	pserialize_read_exit(s);
1927 	return copied;
1928 }
1929 
1930 /*
1931  * XXX almost dup'ed code with rip6_input.
1932  */
1933 static int
1934 icmp6_rip6_input(struct mbuf **mp, int off)
1935 {
1936 	struct mbuf *m = *mp;
1937 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1938 	struct inpcb_hdr *inph;
1939 	struct in6pcb *in6p;
1940 	struct in6pcb *last = NULL;
1941 	struct sockaddr_in6 rip6src;
1942 	struct icmp6_hdr *icmp6;
1943 	struct mbuf *n, *opts = NULL;
1944 
1945 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1946 	if (icmp6 == NULL) {
1947 		/* m is already reclaimed */
1948 		return IPPROTO_DONE;
1949 	}
1950 
1951 	/*
1952 	 * XXX: the address may have embedded scope zone ID, which should be
1953 	 * hidden from applications.
1954 	 */
1955 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1956 	if (sa6_recoverscope(&rip6src)) {
1957 		m_freem(m);
1958 		return IPPROTO_DONE;
1959 	}
1960 
1961 	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1962 		in6p = (struct in6pcb *)inph;
1963 		if (in6p->in6p_af != AF_INET6)
1964 			continue;
1965 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1966 			continue;
1967 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1968 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1969 			continue;
1970 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1971 		    !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1972 			continue;
1973 		if (in6p->in6p_icmp6filt &&
1974 		    ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1975 		    in6p->in6p_icmp6filt))
1976 			continue;
1977 
1978 		if (last == NULL) {
1979 			;
1980 		}
1981 #ifdef IPSEC
1982 		else if (ipsec_used && ipsec_in_reject(m, last)) {
1983 			/* do not inject data into pcb */
1984 		}
1985 #endif
1986 		else if ((n = m_copypacket(m, M_DONTWAIT)) != NULL) {
1987 			if (last->in6p_flags & IN6P_CONTROLOPTS)
1988 				ip6_savecontrol(last, &opts, ip6, n);
1989 			/* strip intermediate headers */
1990 			m_adj(n, off);
1991 			if (sbappendaddr(&last->in6p_socket->so_rcv,
1992 			    sin6tosa(&rip6src), n, opts) == 0) {
1993 				soroverflow(last->in6p_socket);
1994 				m_freem(n);
1995 				if (opts)
1996 					m_freem(opts);
1997 			} else {
1998 				sorwakeup(last->in6p_socket);
1999 			}
2000 			opts = NULL;
2001 		}
2002 
2003 		last = in6p;
2004 	}
2005 
2006 #ifdef IPSEC
2007 	if (ipsec_used && last && ipsec_in_reject(m, last)) {
2008 		m_freem(m);
2009 		IP6_STATDEC(IP6_STAT_DELIVERED);
2010 		/* do not inject data into pcb */
2011 	} else
2012 #endif
2013 	if (last) {
2014 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2015 			ip6_savecontrol(last, &opts, ip6, m);
2016 		/* strip intermediate headers */
2017 		m_adj(m, off);
2018 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2019 		    sin6tosa(&rip6src), m, opts) == 0) {
2020 			soroverflow(last->in6p_socket);
2021 			m_freem(m);
2022 			if (opts)
2023 				m_freem(opts);
2024 		} else {
2025 			sorwakeup(last->in6p_socket);
2026 		}
2027 	} else {
2028 		m_freem(m);
2029 		IP6_STATDEC(IP6_STAT_DELIVERED);
2030 	}
2031 	return IPPROTO_DONE;
2032 }
2033 
2034 /*
2035  * Reflect the ip6 packet back to the source.
2036  * OFF points to the icmp6 header, counted from the top of the mbuf.
2037  *
2038  * Note: RFC 1885 required that an echo reply should be truncated if it
2039  * did not fit in with (return) path MTU, and KAME code supported the
2040  * behavior.  However, as a clarification after the RFC, this limitation
2041  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2042  * old behavior, with a (non-default) ifdef block, while the new version of
2043  * the spec was an internet-draft status, and even after the new RFC was
2044  * published.  But it would rather make sense to clean the obsoleted part
2045  * up, and to make the code simpler at this stage.
2046  */
2047 static void
2048 icmp6_reflect(struct mbuf *m, size_t off)
2049 {
2050 	struct ip6_hdr *ip6;
2051 	struct icmp6_hdr *icmp6;
2052 	const struct in6_ifaddr *ia;
2053 	const struct ip6aux *ip6a;
2054 	int plen;
2055 	int type, code;
2056 	struct ifnet *outif = NULL;
2057 	struct in6_addr origdst;
2058 	struct ifnet *rcvif;
2059 	int s;
2060 	bool ip6_src_filled = false;
2061 
2062 	/* too short to reflect */
2063 	if (off < sizeof(struct ip6_hdr)) {
2064 		nd6log(LOG_DEBUG,
2065 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2066 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2067 		    __FILE__, __LINE__);
2068 		goto bad;
2069 	}
2070 
2071 	/*
2072 	 * If there are extra headers between IPv6 and ICMPv6, strip
2073 	 * off that header first.
2074 	 */
2075 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
2076 	if (off > sizeof(struct ip6_hdr)) {
2077 		size_t l;
2078 		struct ip6_hdr nip6;
2079 
2080 		l = off - sizeof(struct ip6_hdr);
2081 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2082 		m_adj(m, l);
2083 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2084 		if (m->m_len < l) {
2085 			if ((m = m_pullup(m, l)) == NULL)
2086 				return;
2087 		}
2088 		memcpy(mtod(m, void *), (void *)&nip6, sizeof(nip6));
2089 	} else {
2090 		size_t l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2091 		if (m->m_len < l) {
2092 			if ((m = m_pullup(m, l)) == NULL)
2093 				return;
2094 		}
2095 	}
2096 
2097 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2098 	ip6 = mtod(m, struct ip6_hdr *);
2099 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2100 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2101 	type = icmp6->icmp6_type; /* keep type for statistics */
2102 	code = icmp6->icmp6_code; /* ditto. */
2103 
2104 	origdst = ip6->ip6_dst;
2105 	/*
2106 	 * ip6_input() drops a packet if its src is multicast.
2107 	 * So, the src is never multicast.
2108 	 */
2109 	ip6->ip6_dst = ip6->ip6_src;
2110 
2111 	/*
2112 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2113 	 * use dst as the src for the reply.
2114 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2115 	 * (for example) when we encounter an error while forwarding procedure
2116 	 * destined to a duplicated address of ours.
2117 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2118 	 * procedure of an outgoing packet of our own, in which case we need
2119 	 * to search in the ifaddr list.
2120 	 */
2121 	if (IN6_IS_ADDR_MULTICAST(&origdst)) {
2122 		;
2123 	} else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2124 		if ((ip6a->ip6a_flags &
2125 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2126 			ip6->ip6_src = ip6a->ip6a_src;
2127 			ip6_src_filled = true;
2128 		}
2129 	} else {
2130 		union {
2131 			struct sockaddr_in6 sin6;
2132 			struct sockaddr sa;
2133 		} u;
2134 		int _s;
2135 		struct ifaddr *ifa;
2136 
2137 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2138 
2139 		_s = pserialize_read_enter();
2140 		ifa = ifa_ifwithaddr(&u.sa);
2141 
2142 		if (ifa != NULL) {
2143 			ia = ifatoia6(ifa);
2144 			if ((ia->ia6_flags &
2145 				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2146 				ip6->ip6_src = ia->ia_addr.sin6_addr;
2147 				ip6_src_filled = true;
2148 			}
2149 		}
2150 		pserialize_read_exit(_s);
2151 	}
2152 
2153 	if (!ip6_src_filled) {
2154 		int e;
2155 		struct sockaddr_in6 sin6;
2156 		struct route ro;
2157 
2158 		/*
2159 		 * This case matches to multicasts, our anycast, or unicasts
2160 		 * that we do not own.  Select a source address based on the
2161 		 * source address of the erroneous packet.
2162 		 */
2163 		/* zone ID should be embedded */
2164 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2165 
2166 		memset(&ro, 0, sizeof(ro));
2167 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2168 		    &ip6->ip6_src);
2169 		rtcache_free(&ro);
2170 		if (e != 0) {
2171 			char ip6buf[INET6_ADDRSTRLEN];
2172 			nd6log(LOG_DEBUG,
2173 			    "source can't be determined: "
2174 			    "dst=%s, error=%d\n",
2175 			    IN6_PRINT(ip6buf, &sin6.sin6_addr), e);
2176 			goto bad;
2177 		}
2178 	}
2179 
2180 	ip6->ip6_flow = 0;
2181 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2182 	ip6->ip6_vfc |= IPV6_VERSION;
2183 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2184 	rcvif = m_get_rcvif(m, &s);
2185 	if (rcvif) {
2186 		/* XXX: This may not be the outgoing interface */
2187 		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2188 	} else {
2189 		ip6->ip6_hlim = ip6_defhlim;
2190 	}
2191 	m_put_rcvif(rcvif, &s);
2192 
2193 	m->m_pkthdr.csum_flags = 0;
2194 	icmp6->icmp6_cksum = 0;
2195 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2196 	    sizeof(struct ip6_hdr), plen);
2197 
2198 	/*
2199 	 * XXX option handling
2200 	 */
2201 
2202 	m->m_flags &= ~(M_BCAST|M_MCAST);
2203 
2204 	/*
2205 	 * To avoid a "too big" situation at an intermediate router
2206 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2207 	 * Note that only echo and node information replies are affected,
2208 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2209 	 */
2210 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2211 	    outif)
2212 		icmp6_ifstat_inc(outif, ifs6_out_error);
2213 	if (outif)
2214 		icmp6_ifoutstat_inc(outif, type, code);
2215 
2216 	return;
2217 
2218  bad:
2219 	m_freem(m);
2220 	return;
2221 }
2222 
2223 static const char *
2224 icmp6_redirect_diag(char *buf, size_t buflen, struct in6_addr *src6,
2225     struct in6_addr *dst6,  struct in6_addr *tgt6)
2226 {
2227 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
2228 	char ip6buft[INET6_ADDRSTRLEN];
2229 
2230 	snprintf(buf, buflen, "(src=%s dst=%s tgt=%s)",
2231 	    IN6_PRINT(ip6bufs, src6), IN6_PRINT(ip6bufd, dst6),
2232 	    IN6_PRINT(ip6buft, tgt6));
2233 	return buf;
2234 }
2235 
2236 static void
2237 icmp6_redirect_input(struct mbuf *m, int off)
2238 {
2239 	struct ifnet *ifp;
2240 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2241 	struct nd_redirect *nd_rd;
2242 	int icmp6len = m->m_pkthdr.len - off;
2243 	char *lladdr = NULL;
2244 	int lladdrlen = 0;
2245 	struct rtentry *rt = NULL;
2246 	int is_router;
2247 	int is_onlink;
2248 	struct in6_addr src6 = ip6->ip6_src;
2249 	struct in6_addr redtgt6;
2250 	struct in6_addr reddst6;
2251 	union nd_opts ndopts;
2252 	struct psref psref;
2253 	char ip6buf[INET6_ADDRSTRLEN];
2254 	char diagbuf[256];
2255 
2256 	ifp = m_get_rcvif_psref(m, &psref);
2257 	if (ifp == NULL)
2258 		goto freeit;
2259 
2260 	/* XXX if we are router, we don't update route by icmp6 redirect */
2261 	if (ip6_forwarding)
2262 		goto freeit;
2263 	if (!icmp6_rediraccept)
2264 		goto freeit;
2265 
2266 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2267 	if (nd_rd == NULL) {
2268 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2269 		m_put_rcvif_psref(ifp, &psref);
2270 		return;
2271 	}
2272 	redtgt6 = nd_rd->nd_rd_target;
2273 	reddst6 = nd_rd->nd_rd_dst;
2274 
2275 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2276 	    in6_setscope(&reddst6, ifp, NULL)) {
2277 		goto freeit;
2278 	}
2279 
2280 	/* validation */
2281 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2282 		nd6log(LOG_ERR,
2283 		    "ICMP6 redirect sent from %s rejected; "
2284 		    "must be from linklocal\n", IN6_PRINT(ip6buf, &src6));
2285 		goto bad;
2286 	}
2287 	if (ip6->ip6_hlim != 255) {
2288 		nd6log(LOG_ERR,
2289 		    "ICMP6 redirect sent from %s rejected; "
2290 		    "hlim=%d (must be 255)\n",
2291 		    IN6_PRINT(ip6buf, &src6), ip6->ip6_hlim);
2292 		goto bad;
2293 	}
2294 
2295     {
2296 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2297 	struct sockaddr_in6 sin6;
2298 	struct in6_addr *gw6;
2299 
2300 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2301 	rt = rtalloc1(sin6tosa(&sin6), 0);
2302 	if (rt) {
2303 		if (rt->rt_gateway == NULL ||
2304 		    rt->rt_gateway->sa_family != AF_INET6) {
2305 			nd6log(LOG_ERR,
2306 			    "ICMP6 redirect rejected; no route "
2307 			    "with inet6 gateway found for redirect dst: %s\n",
2308 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2309 			    &src6, &reddst6, &redtgt6));
2310 			rt_unref(rt);
2311 			goto bad;
2312 		}
2313 
2314 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2315 		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2316 			nd6log(LOG_ERR,
2317 			    "ICMP6 redirect rejected; "
2318 			    "not equal to gw-for-src=%s (must be same): %s\n",
2319 			    IN6_PRINT(ip6buf, gw6),
2320 			    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2321 			    &src6, &reddst6, &redtgt6));
2322 			rt_unref(rt);
2323 			goto bad;
2324 		}
2325 	} else {
2326 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2327 		    "no route found for redirect dst: %s\n",
2328 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2329 		    &src6, &reddst6, &redtgt6));
2330 		goto bad;
2331 	}
2332 	rt_unref(rt);
2333 	rt = NULL;
2334     }
2335 
2336 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2337 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2338 		    "redirect dst must be unicast: %s\n",
2339 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2340 		    &src6, &reddst6, &redtgt6));
2341 		goto bad;
2342 	}
2343 
2344 	is_router = is_onlink = 0;
2345 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2346 		is_router = 1;	/* router case */
2347 	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2348 		is_onlink = 1;	/* on-link destination case */
2349 	if (!is_router && !is_onlink) {
2350 		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2351 		    "neither router case nor onlink case: %s\n",
2352 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2353 		    &src6, &reddst6, &redtgt6));
2354 		goto bad;
2355 	}
2356 	/* validation passed */
2357 
2358 	icmp6len -= sizeof(*nd_rd);
2359 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2360 	if (nd6_options(&ndopts) < 0) {
2361 		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2362 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2363 		    &src6, &reddst6, &redtgt6));
2364 		/* nd6_options have incremented stats */
2365 		goto freeit;
2366 	}
2367 
2368 	if (ndopts.nd_opts_tgt_lladdr) {
2369 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2370 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2371 	}
2372 
2373 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2374 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2375 		    "(if %d, icmp6 packet %d): %s\n",
2376 		    IN6_PRINT(ip6buf, &redtgt6),
2377 		    ifp->if_addrlen, lladdrlen - 2,
2378 		    icmp6_redirect_diag(diagbuf, sizeof(diagbuf),
2379 		    &src6, &reddst6, &redtgt6));
2380 		goto bad;
2381 	}
2382 
2383 	/* RFC 2461 8.3 */
2384 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2385 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2386 
2387 	m_put_rcvif_psref(ifp, &psref);
2388 	ifp = NULL;
2389 
2390 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2391 		/* perform rtredirect */
2392 		struct sockaddr_in6 sdst;
2393 		struct sockaddr_in6 sgw;
2394 		struct sockaddr_in6 ssrc;
2395 		unsigned long rtcount;
2396 		struct rtentry *newrt = NULL;
2397 
2398 		/*
2399 		 * do not install redirect route, if the number of entries
2400 		 * is too much (> hiwat).  note that, the node (= host) will
2401 		 * work just fine even if we do not install redirect route
2402 		 * (there will be additional hops, though).
2403 		 */
2404 		mutex_enter(&icmp6_mtx);
2405 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2406 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) {
2407 			mutex_exit(&icmp6_mtx);
2408 			goto freeit;
2409 		}
2410 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) {
2411 			mutex_exit(&icmp6_mtx);
2412 			goto freeit;
2413 		} else if (0 <= icmp6_redirect_lowat &&
2414 		    rtcount > icmp6_redirect_lowat) {
2415 			/*
2416 			 * XXX nuke a victim, install the new one.
2417 			 */
2418 		}
2419 
2420 		memset(&sdst, 0, sizeof(sdst));
2421 		memset(&sgw, 0, sizeof(sgw));
2422 		memset(&ssrc, 0, sizeof(ssrc));
2423 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2424 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2425 		    sizeof(struct sockaddr_in6);
2426 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2427 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2428 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2429 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2430 		    RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc), &newrt);
2431 
2432 		if (newrt) {
2433 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2434 			    icmp6_redirect_timeout_q);
2435 			rt_unref(newrt);
2436 		}
2437 		mutex_exit(&icmp6_mtx);
2438 	}
2439 	/* finally update cached route in each socket via pfctlinput */
2440 	{
2441 		struct sockaddr_in6 sdst;
2442 
2443 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2444 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2445 #if defined(IPSEC)
2446 		if (ipsec_used)
2447 			key_sa_routechange(sin6tosa(&sdst));
2448 #endif
2449 	}
2450 
2451 freeit:
2452 	if (ifp != NULL)
2453 		m_put_rcvif_psref(ifp, &psref);
2454 	m_freem(m);
2455 	return;
2456 
2457 bad:
2458 	m_put_rcvif_psref(ifp, &psref);
2459 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2460 	m_freem(m);
2461 }
2462 
2463 void
2464 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2465 {
2466 	struct ifnet *ifp;	/* my outgoing interface */
2467 	struct in6_addr *ifp_ll6;
2468 	struct in6_addr *nexthop;
2469 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2470 	struct mbuf *m = NULL;	/* newly allocated one */
2471 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2472 	struct nd_redirect *nd_rd;
2473 	size_t maxlen;
2474 	u_char *p;
2475 	struct sockaddr_in6 src_sa;
2476 
2477 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2478 
2479 	/* if we are not router, we don't send icmp6 redirect */
2480 	if (!ip6_forwarding)
2481 		goto fail;
2482 
2483 	/* sanity check */
2484 	KASSERT(m0 != NULL);
2485 	KASSERT(rt != NULL);
2486 
2487 	ifp = rt->rt_ifp;
2488 
2489 	/*
2490 	 * Address check:
2491 	 *  the source address must identify a neighbor, and
2492 	 *  the destination address must not be a multicast address
2493 	 *  [RFC 2461, sec 8.2]
2494 	 */
2495 	sip6 = mtod(m0, struct ip6_hdr *);
2496 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2497 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2498 		goto fail;
2499 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2500 		goto fail;	/* what should we do here? */
2501 
2502 	/* rate limit */
2503 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2504 		goto fail;
2505 
2506 	/*
2507 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2508 	 * we almost always ask for an mbuf cluster for simplicity.
2509 	 * (MHLEN < IPV6_MMTU is almost always true)
2510 	 */
2511 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2512 	if (m && IPV6_MMTU >= MHLEN) {
2513 #if IPV6_MMTU >= MCLBYTES
2514 		MEXTMALLOC(m, IPV6_MMTU, M_NOWAIT);
2515 #else
2516 		MCLGET(m, M_DONTWAIT);
2517 #endif
2518 	}
2519 
2520 	if (!m)
2521 		goto fail;
2522 	m_reset_rcvif(m);
2523 	m->m_len = 0;
2524 	maxlen = M_TRAILINGSPACE(m);
2525 	maxlen = uimin(IPV6_MMTU, maxlen);
2526 
2527 	/* just for safety */
2528 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct nd_redirect) +
2529 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2530 		goto fail;
2531 	}
2532 
2533 	{
2534 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2535 		struct in6_ifaddr *ia;
2536 		int s = pserialize_read_enter();
2537 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2538 						 IN6_IFF_NOTREADY|
2539 						 IN6_IFF_ANYCAST)) == NULL) {
2540 			pserialize_read_exit(s);
2541 			goto fail;
2542 		}
2543 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2544 		pserialize_read_exit(s);
2545 	}
2546 
2547 	/* get ip6 linklocal address for the router. */
2548 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2549 		struct sockaddr_in6 *sin6;
2550 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2551 		nexthop = &sin6->sin6_addr;
2552 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2553 			nexthop = NULL;
2554 	} else
2555 		nexthop = NULL;
2556 
2557 	/* ip6 */
2558 	ip6 = mtod(m, struct ip6_hdr *);
2559 	ip6->ip6_flow = 0;
2560 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2561 	ip6->ip6_vfc |= IPV6_VERSION;
2562 	/* ip6->ip6_plen will be set later */
2563 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2564 	ip6->ip6_hlim = 255;
2565 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2566 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2567 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2568 
2569 	/* ND Redirect */
2570 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2571 	nd_rd->nd_rd_type = ND_REDIRECT;
2572 	nd_rd->nd_rd_code = 0;
2573 	nd_rd->nd_rd_reserved = 0;
2574 	if (rt->rt_flags & RTF_GATEWAY) {
2575 		/*
2576 		 * nd_rd->nd_rd_target must be a link-local address in
2577 		 * better router cases.
2578 		 */
2579 		if (!nexthop)
2580 			goto fail;
2581 		bcopy(nexthop, &nd_rd->nd_rd_target,
2582 		      sizeof(nd_rd->nd_rd_target));
2583 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2584 		      sizeof(nd_rd->nd_rd_dst));
2585 	} else {
2586 		/* make sure redtgt == reddst */
2587 		nexthop = &sip6->ip6_dst;
2588 		bcopy(&sip6->ip6_dst, &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 	}
2593 
2594 	p = (u_char *)(nd_rd + 1);
2595 
2596 	{
2597 		/* target lladdr option */
2598 		struct llentry *ln = NULL;
2599 		int len, pad;
2600 		struct nd_opt_hdr *nd_opt;
2601 		char *lladdr;
2602 
2603 		ln = nd6_lookup(nexthop, ifp, false);
2604 		if (ln == NULL)
2605 			goto nolladdropt;
2606 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2607 		len = (len + 7) & ~7;	/* round by 8 */
2608 		pad = len - (sizeof(*nd_opt) + ifp->if_addrlen);
2609 
2610 		/* safety check */
2611 		if (len + (p - (u_char *)ip6) > maxlen) {
2612 			LLE_RUNLOCK(ln);
2613 			goto nolladdropt;
2614 		}
2615 
2616 		if (ln->la_flags & LLE_VALID) {
2617 			nd_opt = (struct nd_opt_hdr *)p;
2618 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2619 			nd_opt->nd_opt_len = len >> 3;
2620 			lladdr = (char *)(nd_opt + 1);
2621 			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2622 			memset(lladdr + ifp->if_addrlen, 0, pad);
2623 			p += len;
2624 		}
2625 		LLE_RUNLOCK(ln);
2626 	}
2627 nolladdropt:
2628 
2629 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2630 
2631 	/* just to be safe */
2632 	if (m0->m_flags & M_DECRYPTED)
2633 		goto noredhdropt;
2634 	if (p - (u_char *)ip6 > maxlen)
2635 		goto noredhdropt;
2636 
2637 	{
2638 		/* redirected header option */
2639 		int len;
2640 		struct nd_opt_rd_hdr *nd_opt_rh;
2641 
2642 		/*
2643 		 * compute the maximum size for icmp6 redirect header option.
2644 		 * XXX room for auth header?
2645 		 */
2646 		len = maxlen - (p - (u_char *)ip6);
2647 		len &= ~7;
2648 
2649 		if (len < sizeof(*nd_opt_rh)) {
2650 			goto noredhdropt;
2651 		}
2652 
2653 		/*
2654 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2655 		 * about padding/truncate rule for the original IP packet.
2656 		 * From the discussion on IPv6imp in Feb 1999,
2657 		 * the consensus was:
2658 		 * - "attach as much as possible" is the goal
2659 		 * - pad if not aligned (original size can be guessed by
2660 		 *   original ip6 header)
2661 		 * Following code adds the padding if it is simple enough,
2662 		 * and truncates if not.
2663 		 */
2664 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2665 			/* not enough room, truncate */
2666 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2667 			    m0->m_pkthdr.len);
2668 		} else {
2669 			/*
2670 			 * enough room, truncate if not aligned.
2671 			 * we don't pad here for simplicity.
2672 			 */
2673 			int extra;
2674 
2675 			extra = m0->m_pkthdr.len % 8;
2676 			if (extra) {
2677 				/* truncate */
2678 				m_adj(m0, -extra);
2679 			}
2680 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2681 		}
2682 
2683 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2684 		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2685 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2686 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2687 		p += sizeof(*nd_opt_rh);
2688 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2689 
2690 		/* connect m0 to m */
2691 		m->m_pkthdr.len += m0->m_pkthdr.len;
2692 		m_cat(m, m0);
2693 		m0 = NULL;
2694 	}
2695 noredhdropt:
2696 	if (m0) {
2697 		m_freem(m0);
2698 		m0 = NULL;
2699 	}
2700 
2701 	/* XXX: clear embedded link IDs in the inner header */
2702 	in6_clearscope(&sip6->ip6_src);
2703 	in6_clearscope(&sip6->ip6_dst);
2704 	in6_clearscope(&nd_rd->nd_rd_target);
2705 	in6_clearscope(&nd_rd->nd_rd_dst);
2706 
2707 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2708 
2709 	nd_rd->nd_rd_cksum = 0;
2710 	nd_rd->nd_rd_cksum =
2711 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2712 
2713 	/* send the packet to outside... */
2714 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2715 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2716 
2717 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2718 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2719 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2720 
2721 	return;
2722 
2723 fail:
2724 	if (m)
2725 		m_freem(m);
2726 	if (m0)
2727 		m_freem(m0);
2728 }
2729 
2730 /*
2731  * ICMPv6 socket option processing.
2732  */
2733 int
2734 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2735 {
2736 	int error = 0;
2737 	struct in6pcb *in6p = sotoin6pcb(so);
2738 
2739 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2740 		return rip6_ctloutput(op, so, sopt);
2741 
2742 	switch (op) {
2743 	case PRCO_SETOPT:
2744 		switch (sopt->sopt_name) {
2745 		case ICMP6_FILTER:
2746 		    {
2747 			struct icmp6_filter fil;
2748 
2749 			error = sockopt_get(sopt, &fil, sizeof(fil));
2750 			if (error)
2751 				break;
2752 			memcpy(in6p->in6p_icmp6filt, &fil,
2753 			    sizeof(struct icmp6_filter));
2754 			error = 0;
2755 			break;
2756 		    }
2757 
2758 		default:
2759 			error = ENOPROTOOPT;
2760 			break;
2761 		}
2762 		break;
2763 
2764 	case PRCO_GETOPT:
2765 		switch (sopt->sopt_name) {
2766 		case ICMP6_FILTER:
2767 		    {
2768 			if (in6p->in6p_icmp6filt == NULL) {
2769 				error = EINVAL;
2770 				break;
2771 			}
2772 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2773 			    sizeof(struct icmp6_filter));
2774 			break;
2775 		    }
2776 
2777 		default:
2778 			error = ENOPROTOOPT;
2779 			break;
2780 		}
2781 		break;
2782 	}
2783 
2784 	return error;
2785 }
2786 
2787 /*
2788  * Perform rate limit check.
2789  * Returns 0 if it is okay to send the icmp6 packet.
2790  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2791  * limitation.
2792  *
2793  * XXX per-destination/type check necessary?
2794  */
2795 static int
2796 icmp6_ratelimit(
2797 	const struct in6_addr *dst,	/* not used at this moment */
2798 	const int type,		/* not used at this moment */
2799 	const int code)		/* not used at this moment */
2800 {
2801 	int ret;
2802 
2803 	ret = 0;	/* okay to send */
2804 
2805 	/* PPS limit */
2806 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2807 	    icmp6errppslim)) {
2808 		/* The packet is subject to rate limit */
2809 		ret++;
2810 	}
2811 
2812 	return ret;
2813 }
2814 
2815 static struct rtentry *
2816 icmp6_mtudisc_clone(struct sockaddr *dst)
2817 {
2818 	struct rtentry *rt;
2819 	int    error;
2820 
2821 	rt = rtalloc1(dst, 1);
2822 	if (rt == NULL)
2823 		return NULL;
2824 
2825 	/* If we didn't get a host route, allocate one */
2826 	if ((rt->rt_flags & RTF_HOST) == 0) {
2827 		struct rtentry *nrt;
2828 
2829 		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2830 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2831 		if (error) {
2832 			rt_unref(rt);
2833 			return NULL;
2834 		}
2835 		nrt->rt_rmx = rt->rt_rmx;
2836 		rt_unref(rt);
2837 		rt = nrt;
2838 	}
2839 
2840 	mutex_enter(&icmp6_mtx);
2841 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2842 			icmp6_mtudisc_timeout_q);
2843 	mutex_exit(&icmp6_mtx);
2844 
2845 	if (error) {
2846 		rt_unref(rt);
2847 		return NULL;
2848 	}
2849 
2850 	return rt;	/* caller need to call rtfree() */
2851 }
2852 
2853 static void
2854 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2855 {
2856 	struct rtentry *retrt;
2857 
2858 	KASSERT(rt != NULL);
2859 	rt_assert_referenced(rt);
2860 
2861 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2862 	    (RTF_DYNAMIC | RTF_HOST)) {
2863 		rtrequest(RTM_DELETE, rt_getkey(rt),
2864 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2865 		rt_unref(rt);
2866 		rt_free(retrt);
2867 	} else {
2868 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2869 			rt->rt_rmx.rmx_mtu = 0;
2870 	}
2871 }
2872 
2873 static void
2874 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2875 {
2876 	struct rtentry *retrt;
2877 
2878 	KASSERT(rt != NULL);
2879 	rt_assert_referenced(rt);
2880 
2881 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2882 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2883 		rtrequest(RTM_DELETE, rt_getkey(rt),
2884 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &retrt);
2885 		rt_unref(rt);
2886 		rt_free(retrt);
2887 	}
2888 }
2889 
2890 #ifdef COMPAT_90
2891 /*
2892  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2893  */
2894 static int
2895 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2896 {
2897 	(void)&name;
2898 	(void)&l;
2899 	(void)&oname;
2900 
2901 	if (namelen != 0)
2902 		return (EINVAL);
2903 
2904 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2905 	    /*XXXUNCONST*/
2906 	    __UNCONST(newp), newlen));
2907 }
2908 #endif
2909 
2910 static int
2911 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2912 {
2913 
2914 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2915 }
2916 
2917 static int
2918 sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2919 {
2920 	int error, tmp;
2921 	struct sysctlnode node;
2922 
2923 	mutex_enter(&icmp6_mtx);
2924 
2925 	node = *rnode;
2926 	node.sysctl_data = &tmp;
2927 	tmp = icmp6_redirtimeout;
2928 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2929 	if (error || newp == NULL)
2930 		goto out;
2931 	if (tmp < 0) {
2932 		error = EINVAL;
2933 		goto out;
2934 	}
2935 	icmp6_redirtimeout = tmp;
2936 
2937 	if (icmp6_redirect_timeout_q != NULL) {
2938 		if (icmp6_redirtimeout == 0) {
2939 			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2940 		} else {
2941 			rt_timer_queue_change(icmp6_redirect_timeout_q,
2942 			    icmp6_redirtimeout);
2943 		}
2944 	} else if (icmp6_redirtimeout > 0) {
2945 		icmp6_redirect_timeout_q =
2946 		    rt_timer_queue_create(icmp6_redirtimeout);
2947 	}
2948 	error = 0;
2949 out:
2950 	mutex_exit(&icmp6_mtx);
2951 	return error;
2952 }
2953 
2954 static void
2955 sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2956 {
2957 
2958 	sysctl_createv(clog, 0, NULL, NULL,
2959 		       CTLFLAG_PERMANENT,
2960 		       CTLTYPE_NODE, "inet6", NULL,
2961 		       NULL, 0, NULL, 0,
2962 		       CTL_NET, PF_INET6, CTL_EOL);
2963 	sysctl_createv(clog, 0, NULL, NULL,
2964 		       CTLFLAG_PERMANENT,
2965 		       CTLTYPE_NODE, "icmp6",
2966 		       SYSCTL_DESCR("ICMPv6 related settings"),
2967 		       NULL, 0, NULL, 0,
2968 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2969 
2970 	sysctl_createv(clog, 0, NULL, NULL,
2971 		       CTLFLAG_PERMANENT,
2972 		       CTLTYPE_STRUCT, "stats",
2973 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2974 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2975 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2976 		       ICMPV6CTL_STATS, CTL_EOL);
2977 	sysctl_createv(clog, 0, NULL, NULL,
2978 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2979 		       CTLTYPE_INT, "rediraccept",
2980 		       SYSCTL_DESCR("Accept and process redirect messages"),
2981 		       NULL, 0, &icmp6_rediraccept, 0,
2982 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2983 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2984 	sysctl_createv(clog, 0, NULL, NULL,
2985 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2986 		       CTLTYPE_INT, "redirtimeout",
2987 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2988 		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2989 		       &icmp6_redirtimeout, 0,
2990 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2991 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2992 #if 0 /* obsoleted */
2993 	sysctl_createv(clog, 0, NULL, NULL,
2994 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2995 		       CTLTYPE_INT, "errratelimit", NULL,
2996 		       NULL, 0, &icmp6_errratelimit, 0,
2997 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2998 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2999 #endif
3000 	sysctl_createv(clog, 0, NULL, NULL,
3001 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3002 		       CTLTYPE_INT, "nd6_prune",
3003 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
3004 		       NULL, 0, &nd6_prune, 0,
3005 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3006 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
3007 	sysctl_createv(clog, 0, NULL, NULL,
3008 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3009 		       CTLTYPE_INT, "nd6_delay",
3010 		       SYSCTL_DESCR("First probe delay time"),
3011 		       NULL, 0, &nd6_nd_domain.nd_delay, 0,
3012 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3013 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
3014 	sysctl_createv(clog, 0, NULL, NULL,
3015 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3016 		       CTLTYPE_INT, "nd6_mmaxtries",
3017 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
3018 		       NULL, 0, &nd6_nd_domain.nd_mmaxtries, 0,
3019 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3020 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
3021 	sysctl_createv(clog, 0, NULL, NULL,
3022 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3023 		       CTLTYPE_INT, "nd6_umaxtries",
3024 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
3025 		       NULL, 0, &nd6_nd_domain.nd_umaxtries, 0,
3026 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3027 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
3028 	sysctl_createv(clog, 0, NULL, NULL,
3029 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3030 		       CTLTYPE_INT, "nd6_maxnudhint",
3031 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
3032 		       NULL, 0, &nd6_nd_domain.nd_maxnudhint, 0,
3033 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3034 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
3035 	sysctl_createv(clog, 0, NULL, NULL,
3036 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3037 		       CTLTYPE_INT, "maxqueuelen",
3038 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3039 		       NULL, 1, &nd6_nd_domain.nd_maxqueuelen, 0,
3040 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3041 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3042 	sysctl_createv(clog, 0, NULL, NULL,
3043 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3044 		       CTLTYPE_INT, "nd6_useloopback",
3045 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
3046 		       NULL, 0, &nd6_useloopback, 0,
3047 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3048 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
3049 #if 0 /* obsoleted */
3050 	sysctl_createv(clog, 0, NULL, NULL,
3051 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3052 		       CTLTYPE_INT, "nd6_proxyall", NULL,
3053 		       NULL, 0, &nd6_proxyall, 0,
3054 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3055 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
3056 #endif
3057 	sysctl_createv(clog, 0, NULL, NULL,
3058 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3059 		       CTLTYPE_INT, "nodeinfo",
3060 		       SYSCTL_DESCR("Respond to node information requests"),
3061 		       NULL, 0, &icmp6_nodeinfo, 0,
3062 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3063 		       ICMPV6CTL_NODEINFO, CTL_EOL);
3064 	sysctl_createv(clog, 0, NULL, NULL,
3065 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3066 		       CTLTYPE_INT, "errppslimit",
3067 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
3068 		       NULL, 0, &icmp6errppslim, 0,
3069 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3070 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
3071 	sysctl_createv(clog, 0, NULL, NULL,
3072 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3073 		       CTLTYPE_INT, "mtudisc_hiwat",
3074 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3075 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
3076 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3077 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
3078 	sysctl_createv(clog, 0, NULL, NULL,
3079 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3080 		       CTLTYPE_INT, "mtudisc_lowat",
3081 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
3082 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
3083 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3084 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
3085 	sysctl_createv(clog, 0, NULL, NULL,
3086 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3087 		       CTLTYPE_INT, "nd6_debug",
3088 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
3089 		       NULL, 0, &nd6_debug, 0,
3090 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3091 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
3092 #ifdef COMPAT_90
3093 	sysctl_createv(clog, 0, NULL, NULL,
3094 		       CTLFLAG_PERMANENT,
3095 		       CTLTYPE_STRUCT, "nd6_drlist",
3096 		       SYSCTL_DESCR("Default router list"),
3097 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3098 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3099 		       OICMPV6CTL_ND6_DRLIST, CTL_EOL);
3100 	sysctl_createv(clog, 0, NULL, NULL,
3101 		       CTLFLAG_PERMANENT,
3102 		       CTLTYPE_STRUCT, "nd6_prlist",
3103 		       SYSCTL_DESCR("Prefix list"),
3104 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3105 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3106 		       OICMPV6CTL_ND6_PRLIST, CTL_EOL);
3107 #endif
3108 }
3109 
3110 void
3111 icmp6_statinc(u_int stat)
3112 {
3113 
3114 	KASSERT(stat < ICMP6_NSTATS);
3115 	ICMP6_STATINC(stat);
3116 }
3117