xref: /openbsd-src/sys/netinet6/icmp6.c (revision dcb03dac24b6a52fa9b433eceeaac7ff6f5182e3)
1 /*	$OpenBSD: icmp6.c,v 1.248 2023/04/05 21:51:47 bluhm 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 "carp.h"
65 #include "pf.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/sysctl.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 
79 #include <net/if.h>
80 #include <net/if_var.h>
81 #include <net/route.h>
82 #include <net/if_dl.h>
83 #include <net/if_types.h>
84 
85 #include <netinet/in.h>
86 #include <netinet/ip.h>
87 #include <netinet6/in6_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/ip6_var.h>
90 #include <netinet/icmp6.h>
91 #include <netinet6/mld6_var.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet6/nd6.h>
94 #include <netinet6/ip6protosw.h>
95 
96 #if NCARP > 0
97 #include <netinet/ip_carp.h>
98 #endif
99 
100 #if NPF > 0
101 #include <net/pfvar.h>
102 #endif
103 
104 struct cpumem *icmp6counters;
105 
106 extern int icmp6errppslim;
107 static int icmp6errpps_count = 0;
108 static struct timeval icmp6errppslim_last;
109 
110 /*
111  * List of callbacks to notify when Path MTU changes are made.
112  */
113 struct icmp6_mtudisc_callback {
114 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
115 	void (*mc_func)(struct sockaddr_in6 *, u_int);
116 };
117 
118 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
119     LIST_HEAD_INITIALIZER(icmp6_mtudisc_callbacks);
120 
121 struct rttimer_queue icmp6_mtudisc_timeout_q;
122 
123 /* XXX do these values make any sense? */
124 static int icmp6_mtudisc_hiwat = 1280;
125 static int icmp6_mtudisc_lowat = 256;
126 
127 /*
128  * keep track of # of redirect routes.
129  */
130 struct rttimer_queue icmp6_redirect_timeout_q;
131 
132 void	icmp6_errcount(int, int);
133 int	icmp6_ratelimit(const struct in6_addr *, const int, const int);
134 const char *icmp6_redirect_diag(struct in6_addr *, struct in6_addr *,
135 	    struct in6_addr *);
136 int	icmp6_notify_error(struct mbuf *, int, int, int);
137 void	icmp6_mtudisc_timeout(struct rtentry *, u_int);
138 
139 void
140 icmp6_init(void)
141 {
142 	mld6_init();
143 	rt_timer_queue_init(&icmp6_mtudisc_timeout_q, ip6_mtudisc_timeout,
144 	    &icmp6_mtudisc_timeout);
145 	rt_timer_queue_init(&icmp6_redirect_timeout_q, icmp6_redirtimeout,
146 	    NULL);
147 	icmp6counters = counters_alloc(icp6s_ncounters);
148 }
149 
150 void
151 icmp6_errcount(int type, int code)
152 {
153 	enum icmp6stat_counters c = icp6s_ounknown;
154 
155 	switch (type) {
156 	case ICMP6_DST_UNREACH:
157 		switch (code) {
158 		case ICMP6_DST_UNREACH_NOROUTE:
159 			c = icp6s_odst_unreach_noroute;
160 			break;
161 		case ICMP6_DST_UNREACH_ADMIN:
162 			c = icp6s_odst_unreach_admin;
163 			break;
164 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
165 			c = icp6s_odst_unreach_beyondscope;
166 			break;
167 		case ICMP6_DST_UNREACH_ADDR:
168 			c = icp6s_odst_unreach_addr;
169 			break;
170 		case ICMP6_DST_UNREACH_NOPORT:
171 			c = icp6s_odst_unreach_noport;
172 			break;
173 		}
174 		break;
175 	case ICMP6_PACKET_TOO_BIG:
176 		c = icp6s_opacket_too_big;
177 		break;
178 	case ICMP6_TIME_EXCEEDED:
179 		switch (code) {
180 		case ICMP6_TIME_EXCEED_TRANSIT:
181 			c = icp6s_otime_exceed_transit;
182 			break;
183 		case ICMP6_TIME_EXCEED_REASSEMBLY:
184 			c = icp6s_otime_exceed_reassembly;
185 			break;
186 		}
187 		break;
188 	case ICMP6_PARAM_PROB:
189 		switch (code) {
190 		case ICMP6_PARAMPROB_HEADER:
191 			c = icp6s_oparamprob_header;
192 			break;
193 		case ICMP6_PARAMPROB_NEXTHEADER:
194 			c = icp6s_oparamprob_nextheader;
195 			break;
196 		case ICMP6_PARAMPROB_OPTION:
197 			c = icp6s_oparamprob_option;
198 			break;
199 		}
200 		break;
201 	case ND_REDIRECT:
202 		c = icp6s_oredirect;
203 		break;
204 	}
205 
206 	icmp6stat_inc(c);
207 }
208 
209 /*
210  * Register a Path MTU Discovery callback.
211  */
212 void
213 icmp6_mtudisc_callback_register(void (*func)(struct sockaddr_in6 *, u_int))
214 {
215 	struct icmp6_mtudisc_callback *mc;
216 
217 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list) {
218 		if (mc->mc_func == func)
219 			return;
220 	}
221 
222 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
223 	if (mc == NULL)
224 		panic("%s", __func__);
225 
226 	mc->mc_func = func;
227 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
228 }
229 
230 struct mbuf *
231 icmp6_do_error(struct mbuf *m, int type, int code, int param)
232 {
233 	struct ip6_hdr *oip6, *nip6;
234 	struct icmp6_hdr *icmp6;
235 	u_int preplen;
236 	int off;
237 	int nxt;
238 
239 	icmp6stat_inc(icp6s_error);
240 
241 	/* count per-type-code statistics */
242 	icmp6_errcount(type, code);
243 
244 	if (m->m_len < sizeof(struct ip6_hdr)) {
245 		m = m_pullup(m, sizeof(struct ip6_hdr));
246 		if (m == NULL)
247 			return (NULL);
248 	}
249 	oip6 = mtod(m, struct ip6_hdr *);
250 
251 	/*
252 	 * If the destination address of the erroneous packet is a multicast
253 	 * address, or the packet was sent using link-layer multicast,
254 	 * we should basically suppress sending an error (RFC 2463, Section
255 	 * 2.4).
256 	 * We have two exceptions (the item e.2 in that section):
257 	 * - the Packet Too Big message can be sent for path MTU discovery.
258 	 * - the Parameter Problem Message that can be allowed an icmp6 error
259 	 *   in the option type field.  This check has been done in
260 	 *   ip6_unknown_opt(), so we can just check the type and code.
261 	 */
262 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
263 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
264 	    (type != ICMP6_PACKET_TOO_BIG &&
265 	     (type != ICMP6_PARAM_PROB ||
266 	      code != ICMP6_PARAMPROB_OPTION)))
267 		goto freeit;
268 
269 	/*
270 	 * RFC 2463, 2.4 (e.5): source address check.
271 	 * XXX: the case of anycast source?
272 	 */
273 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
274 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
275 		goto freeit;
276 
277 	/*
278 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
279 	 * don't do it.
280 	 */
281 	nxt = -1;
282 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
283 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
284 		struct icmp6_hdr *icp;
285 
286 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
287 			sizeof(*icp));
288 		if (icp == NULL) {
289 			icmp6stat_inc(icp6s_tooshort);
290 			return (NULL);
291 		}
292 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
293 		    icp->icmp6_type == ND_REDIRECT) {
294 			/*
295 			 * ICMPv6 error
296 			 * Special case: for redirect (which is
297 			 * informational) we must not send icmp6 error.
298 			 */
299 			icmp6stat_inc(icp6s_canterror);
300 			goto freeit;
301 		} else {
302 			/* ICMPv6 informational - send the error */
303 		}
304 	}
305 	else {
306 		/* non-ICMPv6 - send the error */
307 	}
308 
309 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
310 
311 	/* Finally, do rate limitation check. */
312 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
313 		icmp6stat_inc(icp6s_toofreq);
314 		goto freeit;
315 	}
316 
317 	/*
318 	 * OK, ICMP6 can be generated.
319 	 */
320 
321 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
322 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
323 
324 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
325 	M_PREPEND(m, preplen, M_DONTWAIT);
326 	if (m && m->m_len < preplen)
327 		m = m_pullup(m, preplen);
328 	if (m == NULL) {
329 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
330 		return (NULL);
331 	}
332 
333 	nip6 = mtod(m, struct ip6_hdr *);
334 	nip6->ip6_src  = oip6->ip6_src;
335 	nip6->ip6_dst  = oip6->ip6_dst;
336 
337 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_src))
338 		oip6->ip6_src.s6_addr16[1] = 0;
339 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_dst))
340 		oip6->ip6_dst.s6_addr16[1] = 0;
341 
342 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
343 	icmp6->icmp6_type = type;
344 	icmp6->icmp6_code = code;
345 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
346 
347 	/*
348 	 * icmp6_reflect() is designed to be in the input path.
349 	 * icmp6_error() can be called from both input and output path,
350 	 * and if we are in output path rcvif could contain bogus value.
351 	 * clear m->m_pkthdr.ph_ifidx for safety, we should have enough
352 	 * scope information in ip header (nip6).
353 	 */
354 	m->m_pkthdr.ph_ifidx = 0;
355 
356 	icmp6stat_inc(icp6s_outhist + type);
357 
358 	return (m);
359 
360   freeit:
361 	/*
362 	 * If we can't tell whether or not we can generate ICMP6, free it.
363 	 */
364 	return (m_freem(m));
365 }
366 
367 /*
368  * Generate an error packet of type error in response to bad IP6 packet.
369  */
370 void
371 icmp6_error(struct mbuf *m, int type, int code, int param)
372 {
373 	struct mbuf	*n;
374 
375 	n = icmp6_do_error(m, type, code, param);
376 	if (n != NULL) {
377 		/* header order: IPv6 - ICMPv6 */
378 		if (!icmp6_reflect(&n, sizeof(struct ip6_hdr), NULL))
379 			ip6_send(n);
380 	}
381 }
382 
383 /*
384  * Process a received ICMP6 message.
385  */
386 int
387 icmp6_input(struct mbuf **mp, int *offp, int proto, int af)
388 {
389 #if NCARP > 0
390 	struct ifnet *ifp;
391 #endif
392 	struct mbuf *m = *mp, *n;
393 	struct ip6_hdr *ip6, *nip6;
394 	struct icmp6_hdr *icmp6, *nicmp6;
395 	int off = *offp;
396 	int icmp6len = m->m_pkthdr.len - off;
397 	int code, sum, noff;
398 	char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
399 
400 	/*
401 	 * Locate icmp6 structure in mbuf, and check
402 	 * that not corrupted and of at least minimum length
403 	 */
404 
405 	ip6 = mtod(m, struct ip6_hdr *);
406 	if (icmp6len < sizeof(struct icmp6_hdr)) {
407 		icmp6stat_inc(icp6s_tooshort);
408 		goto freeit;
409 	}
410 
411 	/*
412 	 * calculate the checksum
413 	 */
414 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
415 	if (icmp6 == NULL) {
416 		icmp6stat_inc(icp6s_tooshort);
417 		return IPPROTO_DONE;
418 	}
419 	code = icmp6->icmp6_code;
420 
421 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
422 		nd6log((LOG_ERR,
423 		    "ICMP6 checksum error(%d|%x) %s\n",
424 		    icmp6->icmp6_type, sum,
425 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src))));
426 		icmp6stat_inc(icp6s_checksum);
427 		goto freeit;
428 	}
429 
430 #if NPF > 0
431 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
432 		switch (icmp6->icmp6_type) {
433 		/*
434 		 * These ICMP6 types map to other connections.  They must be
435 		 * delivered to pr_ctlinput() also for diverted connections.
436 		 */
437 		case ICMP6_DST_UNREACH:
438 		case ICMP6_PACKET_TOO_BIG:
439 		case ICMP6_TIME_EXCEEDED:
440 		case ICMP6_PARAM_PROB:
441 			/*
442 			 * Do not use the divert-to property of the TCP or UDP
443 			 * rule when doing the PCB lookup for the raw socket.
444 			 */
445 			m->m_pkthdr.pf.flags &=~ PF_TAG_DIVERTED;
446 			break;
447 		default:
448 			goto raw;
449 		}
450 	}
451 #endif /* NPF */
452 
453 #if NCARP > 0
454 	ifp = if_get(m->m_pkthdr.ph_ifidx);
455 	if (ifp == NULL)
456 		goto freeit;
457 
458 	if (icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
459 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
460 	    ip6->ip6_dst.s6_addr32, 1)) {
461 		if_put(ifp);
462 		goto freeit;
463 	}
464 
465 	if_put(ifp);
466 #endif
467 	icmp6stat_inc(icp6s_inhist + icmp6->icmp6_type);
468 
469 	switch (icmp6->icmp6_type) {
470 	case ICMP6_DST_UNREACH:
471 		switch (code) {
472 		case ICMP6_DST_UNREACH_NOROUTE:
473 			code = PRC_UNREACH_NET;
474 			break;
475 		case ICMP6_DST_UNREACH_ADMIN:
476 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
477 			break;
478 		case ICMP6_DST_UNREACH_ADDR:
479 			code = PRC_HOSTDEAD;
480 			break;
481 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
482 			/* I mean "source address was incorrect." */
483 			code = PRC_PARAMPROB;
484 			break;
485 		case ICMP6_DST_UNREACH_NOPORT:
486 			code = PRC_UNREACH_PORT;
487 			break;
488 		default:
489 			goto badcode;
490 		}
491 		goto deliver;
492 
493 	case ICMP6_PACKET_TOO_BIG:
494 		/* MTU is checked in icmp6_mtudisc_update. */
495 		code = PRC_MSGSIZE;
496 
497 		/*
498 		 * Updating the path MTU will be done after examining
499 		 * intermediate extension headers.
500 		 */
501 		goto deliver;
502 
503 	case ICMP6_TIME_EXCEEDED:
504 		switch (code) {
505 		case ICMP6_TIME_EXCEED_TRANSIT:
506 			code = PRC_TIMXCEED_INTRANS;
507 			break;
508 		case ICMP6_TIME_EXCEED_REASSEMBLY:
509 			code = PRC_TIMXCEED_REASS;
510 			break;
511 		default:
512 			goto badcode;
513 		}
514 		goto deliver;
515 
516 	case ICMP6_PARAM_PROB:
517 		switch (code) {
518 		case ICMP6_PARAMPROB_NEXTHEADER:
519 			code = PRC_UNREACH_PROTOCOL;
520 			break;
521 		case ICMP6_PARAMPROB_HEADER:
522 		case ICMP6_PARAMPROB_OPTION:
523 			code = PRC_PARAMPROB;
524 			break;
525 		default:
526 			goto badcode;
527 		}
528 		goto deliver;
529 
530 	case ICMP6_ECHO_REQUEST:
531 		if (code != 0)
532 			goto badcode;
533 		/*
534 		 * Copy mbuf to send to two data paths: userland socket(s),
535 		 * and to the querier (echo reply).
536 		 * m: a copy for socket, n: a copy for querier
537 		 */
538 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
539 			/* Give up local */
540 			n = m;
541 			m = *mp = NULL;
542 			goto deliverecho;
543 		}
544 		/*
545 		 * If the first mbuf is shared, or the first mbuf is too short,
546 		 * copy the first part of the data into a fresh mbuf.
547 		 * Otherwise, we will wrongly overwrite both copies.
548 		 */
549 		if ((n->m_flags & M_EXT) != 0 ||
550 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
551 			struct mbuf *n0 = n;
552 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
553 
554 			/*
555 			 * Prepare an internal mbuf.  m_pullup() doesn't
556 			 * always copy the length we specified.
557 			 */
558 			if (maxlen >= MCLBYTES) {
559 				/* Give up remote */
560 				m_freem(n0);
561 				break;
562 			}
563 			MGETHDR(n, M_DONTWAIT, n0->m_type);
564 			if (n && maxlen >= MHLEN) {
565 				MCLGET(n, M_DONTWAIT);
566 				if ((n->m_flags & M_EXT) == 0) {
567 					m_free(n);
568 					n = NULL;
569 				}
570 			}
571 			if (n == NULL) {
572 				/* Give up local */
573 				m_freem(n0);
574 				n = m;
575 				m = *mp = NULL;
576 				goto deliverecho;
577 			}
578 			M_MOVE_PKTHDR(n, n0);
579 			/*
580 			 * Copy IPv6 and ICMPv6 only.
581 			 */
582 			nip6 = mtod(n, struct ip6_hdr *);
583 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
584 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
585 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
586 			noff = sizeof(struct ip6_hdr);
587 			n->m_len = noff + sizeof(struct icmp6_hdr);
588 			/*
589 			 * Adjust mbuf.  ip6_plen will be adjusted in
590 			 * ip6_output().
591 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
592 			 */
593 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
594 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
595 			m_adj(n0, off + sizeof(struct icmp6_hdr));
596 			n->m_next = n0;
597 		} else {
598 	 deliverecho:
599 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
600 			    sizeof(*nicmp6));
601 			noff = off;
602 		}
603 		if (n) {
604 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
605 			nicmp6->icmp6_code = 0;
606 			icmp6stat_inc(icp6s_reflect);
607 			icmp6stat_inc(icp6s_outhist + ICMP6_ECHO_REPLY);
608 			if (!icmp6_reflect(&n, noff, NULL))
609 				ip6_send(n);
610 		}
611 		if (!m)
612 			goto freeit;
613 		break;
614 
615 	case ICMP6_ECHO_REPLY:
616 		if (code != 0)
617 			goto badcode;
618 		break;
619 
620 	case MLD_LISTENER_QUERY:
621 	case MLD_LISTENER_REPORT:
622 		if (icmp6len < sizeof(struct mld_hdr))
623 			goto badlen;
624 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
625 			/* give up local */
626 			mld6_input(m, off);
627 			m = NULL;
628 			goto freeit;
629 		}
630 		mld6_input(n, off);
631 		/* m stays. */
632 		break;
633 
634 	case MLD_LISTENER_DONE:
635 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
636 			goto badlen;
637 		break;		/* nothing to be done in kernel */
638 
639 	case MLD_MTRACE_RESP:
640 	case MLD_MTRACE:
641 		/* XXX: these two are experimental.  not officially defined. */
642 		/* XXX: per-interface statistics? */
643 		break;		/* just pass it to applications */
644 
645 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
646 		/* IPv6 Node Information Queries are not supported */
647 		break;
648 	case ICMP6_WRUREPLY:
649 		break;
650 
651 	case ND_ROUTER_SOLICIT:
652 	case ND_ROUTER_ADVERT:
653 		if (code != 0)
654 			goto badcode;
655 		if ((icmp6->icmp6_type == ND_ROUTER_SOLICIT && icmp6len <
656 		    sizeof(struct nd_router_solicit)) ||
657 		    (icmp6->icmp6_type == ND_ROUTER_ADVERT && icmp6len <
658 		    sizeof(struct nd_router_advert)))
659 			goto badlen;
660 
661 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
662 			/* give up local */
663 			nd6_rtr_cache(m, off, icmp6len,
664 			    icmp6->icmp6_type);
665 			m = NULL;
666 			goto freeit;
667 		}
668 		nd6_rtr_cache(n, off, icmp6len, icmp6->icmp6_type);
669 		/* m stays. */
670 		break;
671 
672 	case ND_NEIGHBOR_SOLICIT:
673 		if (code != 0)
674 			goto badcode;
675 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
676 			goto badlen;
677 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
678 			/* give up local */
679 			nd6_ns_input(m, off, icmp6len);
680 			m = NULL;
681 			goto freeit;
682 		}
683 		nd6_ns_input(n, off, icmp6len);
684 		/* m stays. */
685 		break;
686 
687 	case ND_NEIGHBOR_ADVERT:
688 		if (code != 0)
689 			goto badcode;
690 		if (icmp6len < sizeof(struct nd_neighbor_advert))
691 			goto badlen;
692 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
693 			/* give up local */
694 			nd6_na_input(m, off, icmp6len);
695 			m = NULL;
696 			goto freeit;
697 		}
698 		nd6_na_input(n, off, icmp6len);
699 		/* m stays. */
700 		break;
701 
702 	case ND_REDIRECT:
703 		if (code != 0)
704 			goto badcode;
705 		if (icmp6len < sizeof(struct nd_redirect))
706 			goto badlen;
707 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
708 			/* give up local */
709 			icmp6_redirect_input(m, off);
710 			m = NULL;
711 			goto freeit;
712 		}
713 		icmp6_redirect_input(n, off);
714 		/* m stays. */
715 		break;
716 
717 	case ICMP6_ROUTER_RENUMBERING:
718 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
719 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
720 			goto badcode;
721 		if (icmp6len < sizeof(struct icmp6_router_renum))
722 			goto badlen;
723 		break;
724 
725 	default:
726 		nd6log((LOG_DEBUG,
727 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%u)\n",
728 		    icmp6->icmp6_type,
729 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src)),
730 		    inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst)),
731 		    m->m_pkthdr.ph_ifidx));
732 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
733 			/* ICMPv6 error: MUST deliver it by spec... */
734 			code = PRC_NCMDS;
735 			/* deliver */
736 		} else {
737 			/* ICMPv6 informational: MUST not deliver */
738 			break;
739 		}
740 deliver:
741 		if (icmp6_notify_error(m, off, icmp6len, code)) {
742 			/* In this case, m should've been freed. */
743 			return (IPPROTO_DONE);
744 		}
745 		break;
746 
747 badcode:
748 		icmp6stat_inc(icp6s_badcode);
749 		break;
750 
751 badlen:
752 		icmp6stat_inc(icp6s_badlen);
753 		break;
754 	}
755 
756 #if NPF > 0
757 raw:
758 #endif
759 	/* deliver the packet to appropriate sockets */
760 	return rip6_input(mp, offp, proto, af);
761 
762  freeit:
763 	m_freem(m);
764 	return IPPROTO_DONE;
765 }
766 
767 int
768 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
769 {
770 	struct icmp6_hdr *icmp6;
771 	struct ip6_hdr *eip6;
772 	u_int32_t notifymtu;
773 	struct sockaddr_in6 icmp6src, icmp6dst;
774 
775 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
776 		icmp6stat_inc(icp6s_tooshort);
777 		goto freeit;
778 	}
779 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
780 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
781 	if (icmp6 == NULL) {
782 		icmp6stat_inc(icp6s_tooshort);
783 		return (-1);
784 	}
785 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
786 
787 	/* Detect the upper level protocol */
788 	{
789 		void (*ctlfunc)(int, struct sockaddr *, u_int, void *);
790 		u_int8_t nxt = eip6->ip6_nxt;
791 		int eoff = off + sizeof(struct icmp6_hdr) +
792 			sizeof(struct ip6_hdr);
793 		struct ip6ctlparam ip6cp;
794 		struct in6_addr *finaldst = NULL;
795 		int icmp6type = icmp6->icmp6_type;
796 		struct ip6_frag *fh;
797 		struct ip6_rthdr *rth;
798 		struct ip6_rthdr0 *rth0;
799 		int rthlen;
800 
801 		while (1) { /* XXX: should avoid infinite loop explicitly? */
802 			struct ip6_ext *eh;
803 
804 			switch (nxt) {
805 			case IPPROTO_HOPOPTS:
806 			case IPPROTO_DSTOPTS:
807 			case IPPROTO_AH:
808 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
809 					       eoff, sizeof(*eh));
810 				if (eh == NULL) {
811 					icmp6stat_inc(icp6s_tooshort);
812 					return (-1);
813 				}
814 
815 				if (nxt == IPPROTO_AH)
816 					eoff += (eh->ip6e_len + 2) << 2;
817 				else
818 					eoff += (eh->ip6e_len + 1) << 3;
819 				nxt = eh->ip6e_nxt;
820 				break;
821 			case IPPROTO_ROUTING:
822 				/*
823 				 * When the erroneous packet contains a
824 				 * routing header, we should examine the
825 				 * header to determine the final destination.
826 				 * Otherwise, we can't properly update
827 				 * information that depends on the final
828 				 * destination (e.g. path MTU).
829 				 */
830 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
831 					       eoff, sizeof(*rth));
832 				if (rth == NULL) {
833 					icmp6stat_inc(icp6s_tooshort);
834 					return (-1);
835 				}
836 				rthlen = (rth->ip6r_len + 1) << 3;
837 				/*
838 				 * XXX: currently there is no
839 				 * officially defined type other
840 				 * than type-0.
841 				 * Note that if the segment left field
842 				 * is 0, all intermediate hops must
843 				 * have been passed.
844 				 */
845 				if (rth->ip6r_segleft &&
846 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
847 					int hops;
848 
849 					IP6_EXTHDR_GET(rth0,
850 						       struct ip6_rthdr0 *, m,
851 						       eoff, rthlen);
852 					if (rth0 == NULL) {
853 						icmp6stat_inc(icp6s_tooshort);
854 						return (-1);
855 					}
856 					/* just ignore a bogus header */
857 					if ((rth0->ip6r0_len % 2) == 0 &&
858 					    (hops = rth0->ip6r0_len/2))
859 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
860 				}
861 				eoff += rthlen;
862 				nxt = rth->ip6r_nxt;
863 				break;
864 			case IPPROTO_FRAGMENT:
865 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
866 					       eoff, sizeof(*fh));
867 				if (fh == NULL) {
868 					icmp6stat_inc(icp6s_tooshort);
869 					return (-1);
870 				}
871 				/*
872 				 * Data after a fragment header is meaningless
873 				 * unless it is the first fragment, but
874 				 * we'll go to the notify label for path MTU
875 				 * discovery.
876 				 */
877 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
878 					goto notify;
879 
880 				eoff += sizeof(struct ip6_frag);
881 				nxt = fh->ip6f_nxt;
882 				break;
883 			default:
884 				/*
885 				 * This case includes ESP and the No Next
886 				 * Header.  In such cases going to the notify
887 				 * label does not have any meaning
888 				 * (i.e. ctlfunc will be NULL), but we go
889 				 * anyway since we might have to update
890 				 * path MTU information.
891 				 */
892 				goto notify;
893 			}
894 		}
895 	  notify:
896 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
897 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
898 		if (icmp6 == NULL) {
899 			icmp6stat_inc(icp6s_tooshort);
900 			return (-1);
901 		}
902 
903 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
904 		bzero(&icmp6dst, sizeof(icmp6dst));
905 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
906 		icmp6dst.sin6_family = AF_INET6;
907 		if (finaldst == NULL)
908 			icmp6dst.sin6_addr = eip6->ip6_dst;
909 		else
910 			icmp6dst.sin6_addr = *finaldst;
911 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
912 		    &icmp6dst.sin6_addr);
913 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, NULL)) {
914 			/* should be impossible */
915 			nd6log((LOG_DEBUG,
916 			    "icmp6_notify_error: in6_embedscope failed\n"));
917 			goto freeit;
918 		}
919 
920 		/*
921 		 * retrieve parameters from the inner IPv6 header, and convert
922 		 * them into sockaddr structures.
923 		 */
924 		bzero(&icmp6src, sizeof(icmp6src));
925 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
926 		icmp6src.sin6_family = AF_INET6;
927 		icmp6src.sin6_addr = eip6->ip6_src;
928 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
929 		    &icmp6src.sin6_addr);
930 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, NULL)) {
931 			/* should be impossible */
932 			nd6log((LOG_DEBUG,
933 			    "icmp6_notify_error: in6_embedscope failed\n"));
934 			goto freeit;
935 		}
936 		icmp6src.sin6_flowinfo =
937 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
938 
939 		if (finaldst == NULL)
940 			finaldst = &eip6->ip6_dst;
941 		ip6cp.ip6c_m = m;
942 		ip6cp.ip6c_icmp6 = icmp6;
943 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
944 		ip6cp.ip6c_off = eoff;
945 		ip6cp.ip6c_finaldst = finaldst;
946 		ip6cp.ip6c_src = &icmp6src;
947 		ip6cp.ip6c_nxt = nxt;
948 #if NPF > 0
949 		pf_pkt_addr_changed(m);
950 #endif
951 
952 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
953 			notifymtu = ntohl(icmp6->icmp6_mtu);
954 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
955 		}
956 
957 		ctlfunc = inet6sw[ip6_protox[nxt]].pr_ctlinput;
958 		if (ctlfunc)
959 			(*ctlfunc)(code, sin6tosa(&icmp6dst),
960 			    m->m_pkthdr.ph_rtableid, &ip6cp);
961 	}
962 	return (0);
963 
964   freeit:
965 	m_freem(m);
966 	return (-1);
967 }
968 
969 void
970 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
971 {
972 	unsigned long rtcount;
973 	struct icmp6_mtudisc_callback *mc;
974 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
975 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
976 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
977 	u_int mtu = ntohl(icmp6->icmp6_mtu);
978 	struct rtentry *rt = NULL;
979 	struct sockaddr_in6 sin6;
980 
981 	if (mtu < IPV6_MMTU)
982 		return;
983 
984 	/*
985 	 * allow non-validated cases if memory is plenty, to make traffic
986 	 * from non-connected pcb happy.
987 	 */
988 	rtcount = rt_timer_queue_count(&icmp6_mtudisc_timeout_q);
989 	if (validated) {
990 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
991 			return;
992 		else if (0 <= icmp6_mtudisc_lowat &&
993 		    rtcount > icmp6_mtudisc_lowat) {
994 			/*
995 			 * XXX nuke a victim, install the new one.
996 			 */
997 		}
998 	} else {
999 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1000 			return;
1001 	}
1002 
1003 	bzero(&sin6, sizeof(sin6));
1004 	sin6.sin6_family = PF_INET6;
1005 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1006 	sin6.sin6_addr = *dst;
1007 	/* XXX normally, this won't happen */
1008 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1009 		sin6.sin6_addr.s6_addr16[1] = htons(m->m_pkthdr.ph_ifidx);
1010 	}
1011 	sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
1012 	    &sin6.sin6_addr);
1013 
1014 	rt = icmp6_mtudisc_clone(&sin6, m->m_pkthdr.ph_rtableid, 0);
1015 
1016 	if (rt != NULL && ISSET(rt->rt_flags, RTF_HOST) &&
1017 	    !(rt->rt_locks & RTV_MTU) &&
1018 	    (rt->rt_mtu > mtu || rt->rt_mtu == 0)) {
1019 		struct ifnet *ifp;
1020 
1021 		ifp = if_get(rt->rt_ifidx);
1022 		if (ifp != NULL && mtu < ifp->if_mtu) {
1023 			icmp6stat_inc(icp6s_pmtuchg);
1024 			rt->rt_mtu = mtu;
1025 		}
1026 		if_put(ifp);
1027 	}
1028 	rtfree(rt);
1029 
1030 	/*
1031 	 * Notify protocols that the MTU for this destination
1032 	 * has changed.
1033 	 */
1034 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list)
1035 		(*mc->mc_func)(&sin6, m->m_pkthdr.ph_rtableid);
1036 }
1037 
1038 /*
1039  * Reflect the ip6 packet back to the source.
1040  * OFF points to the icmp6 header, counted from the top of the mbuf.
1041  */
1042 int
1043 icmp6_reflect(struct mbuf **mp, size_t off, struct sockaddr *sa)
1044 {
1045 	struct mbuf *m = *mp;
1046 	struct rtentry *rt = NULL;
1047 	struct ip6_hdr *ip6;
1048 	struct icmp6_hdr *icmp6;
1049 	struct in6_addr t, *src = NULL;
1050 	struct sockaddr_in6 sa6_src, sa6_dst;
1051 	u_int rtableid;
1052 	u_int8_t pfflags;
1053 
1054 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
1055 
1056 	/* too short to reflect */
1057 	if (off < sizeof(struct ip6_hdr)) {
1058 		nd6log((LOG_DEBUG,
1059 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1060 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1061 		    __FILE__, __LINE__));
1062 		goto bad;
1063 	}
1064 
1065 	if (m->m_pkthdr.ph_loopcnt++ >= M_MAXLOOP) {
1066 		m_freemp(mp);
1067 		return (ELOOP);
1068 	}
1069 	rtableid = m->m_pkthdr.ph_rtableid;
1070 	pfflags = m->m_pkthdr.pf.flags;
1071 	m_resethdr(m);
1072 	m->m_pkthdr.ph_rtableid = rtableid;
1073 	m->m_pkthdr.pf.flags = pfflags & PF_TAG_GENERATED;
1074 
1075 	/*
1076 	 * If there are extra headers between IPv6 and ICMPv6, strip
1077 	 * off that header first.
1078 	 */
1079 	if (off > sizeof(struct ip6_hdr)) {
1080 		size_t l;
1081 		struct ip6_hdr nip6;
1082 
1083 		l = off - sizeof(struct ip6_hdr);
1084 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1085 		m_adj(m, l);
1086 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1087 		if (m->m_len < l) {
1088 			if ((m = *mp = m_pullup(m, l)) == NULL)
1089 				return (EMSGSIZE);
1090 		}
1091 		memcpy(mtod(m, caddr_t), &nip6, sizeof(nip6));
1092 	} else /* off == sizeof(struct ip6_hdr) */ {
1093 		size_t l;
1094 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1095 		if (m->m_len < l) {
1096 			if ((m = *mp = m_pullup(m, l)) == NULL)
1097 				return (EMSGSIZE);
1098 		}
1099 	}
1100 	ip6 = mtod(m, struct ip6_hdr *);
1101 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1102 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1103 
1104 	t = ip6->ip6_dst;
1105 	/*
1106 	 * ip6_input() drops a packet if its src is multicast.
1107 	 * So, the src is never multicast.
1108 	 */
1109 	ip6->ip6_dst = ip6->ip6_src;
1110 
1111 	/*
1112 	 * XXX: make sure to embed scope zone information, using
1113 	 * already embedded IDs or the received interface (if any).
1114 	 * Note that rcvif may be NULL.
1115 	 * TODO: scoped routing case (XXX).
1116 	 */
1117 	bzero(&sa6_src, sizeof(sa6_src));
1118 	sa6_src.sin6_family = AF_INET6;
1119 	sa6_src.sin6_len = sizeof(sa6_src);
1120 	sa6_src.sin6_addr = ip6->ip6_dst;
1121 	bzero(&sa6_dst, sizeof(sa6_dst));
1122 	sa6_dst.sin6_family = AF_INET6;
1123 	sa6_dst.sin6_len = sizeof(sa6_dst);
1124 	sa6_dst.sin6_addr = t;
1125 
1126 	if (sa == NULL) {
1127 		/*
1128 		 * If the incoming packet was addressed directly to us (i.e.
1129 		 * unicast), use dst as the src for the reply. The
1130 		 * IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED case would be VERY rare,
1131 		 * but is possible (for example) when we encounter an error
1132 		 * while forwarding procedure destined to a duplicated address
1133 		 * of ours.
1134 		 */
1135 		rt = rtalloc(sin6tosa(&sa6_dst), 0, rtableid);
1136 		if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL) &&
1137 		    !ISSET(ifatoia6(rt->rt_ifa)->ia6_flags,
1138 		    IN6_IFF_ANYCAST|IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED)) {
1139 			src = &t;
1140 		}
1141 		rtfree(rt);
1142 		rt = NULL;
1143 		sa = sin6tosa(&sa6_src);
1144 	}
1145 
1146 	if (src == NULL) {
1147 		struct in6_ifaddr *ia6;
1148 
1149 		/*
1150 		 * This case matches to multicasts, our anycast, or unicasts
1151 		 * that we do not own.  Select a source address based on the
1152 		 * source address of the erroneous packet.
1153 		 */
1154 		rt = rtalloc(sa, RT_RESOLVE, rtableid);
1155 		if (!rtisvalid(rt)) {
1156 			char addr[INET6_ADDRSTRLEN];
1157 
1158 			nd6log((LOG_DEBUG,
1159 			    "%s: source can't be determined: dst=%s\n",
1160 			    __func__, inet_ntop(AF_INET6, &sa6_src.sin6_addr,
1161 			    addr, sizeof(addr))));
1162 			rtfree(rt);
1163 			goto bad;
1164 		}
1165 		ia6 = in6_ifawithscope(rt->rt_ifa->ifa_ifp, &t, rtableid);
1166 		if (ia6 != NULL)
1167 			src = &ia6->ia_addr.sin6_addr;
1168 		if (src == NULL)
1169 			src = &ifatoia6(rt->rt_ifa)->ia_addr.sin6_addr;
1170 	}
1171 
1172 	ip6->ip6_src = *src;
1173 	rtfree(rt);
1174 
1175 	ip6->ip6_flow = 0;
1176 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1177 	ip6->ip6_vfc |= IPV6_VERSION;
1178 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1179 	ip6->ip6_hlim = ip6_defhlim;
1180 
1181 	icmp6->icmp6_cksum = 0;
1182 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1183 
1184 	/*
1185 	 * XXX option handling
1186 	 */
1187 
1188 	m->m_flags &= ~(M_BCAST|M_MCAST);
1189 	return (0);
1190 
1191  bad:
1192 	m_freemp(mp);
1193 	return (EHOSTUNREACH);
1194 }
1195 
1196 void
1197 icmp6_fasttimo(void)
1198 {
1199 
1200 	mld6_fasttimeo();
1201 }
1202 
1203 const char *
1204 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
1205     struct in6_addr *tgt6)
1206 {
1207 	static char buf[1024]; /* XXX */
1208 	char src[INET6_ADDRSTRLEN];
1209 	char dst[INET6_ADDRSTRLEN];
1210 	char tgt[INET6_ADDRSTRLEN];
1211 
1212 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
1213 		 inet_ntop(AF_INET6, src6, src, sizeof(src)),
1214 		 inet_ntop(AF_INET6, dst6, dst, sizeof(dst)),
1215 		 inet_ntop(AF_INET6, tgt6, tgt, sizeof(tgt)));
1216 	return buf;
1217 }
1218 
1219 void
1220 icmp6_redirect_input(struct mbuf *m, int off)
1221 {
1222 	struct ifnet *ifp;
1223 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1224 	struct nd_redirect *nd_rd;
1225 	int icmp6len = ntohs(ip6->ip6_plen);
1226 	char *lladdr = NULL;
1227 	int lladdrlen = 0;
1228 	struct rtentry *rt = NULL;
1229 	int is_router;
1230 	int is_onlink;
1231 	struct in6_addr src6 = ip6->ip6_src;
1232 	struct in6_addr redtgt6;
1233 	struct in6_addr reddst6;
1234 	struct nd_opts ndopts;
1235 	char addr[INET6_ADDRSTRLEN];
1236 
1237 	ifp = if_get(m->m_pkthdr.ph_ifidx);
1238 	if (ifp == NULL)
1239 		return;
1240 
1241 	/* XXX if we are router, we don't update route by icmp6 redirect */
1242 	if (ip6_forwarding)
1243 		goto freeit;
1244 	if (!(ifp->if_xflags & IFXF_AUTOCONF6))
1245 		goto freeit;
1246 
1247 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
1248 	if (nd_rd == NULL) {
1249 		icmp6stat_inc(icp6s_tooshort);
1250 		if_put(ifp);
1251 		return;
1252 	}
1253 	redtgt6 = nd_rd->nd_rd_target;
1254 	reddst6 = nd_rd->nd_rd_dst;
1255 
1256 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1257 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
1258 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
1259 		reddst6.s6_addr16[1] = htons(ifp->if_index);
1260 
1261 	/* validation */
1262 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
1263 		nd6log((LOG_ERR,
1264 			"ICMP6 redirect sent from %s rejected; "
1265 			"must be from linklocal\n",
1266 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr))));
1267 		goto bad;
1268 	}
1269 	if (ip6->ip6_hlim != 255) {
1270 		nd6log((LOG_ERR,
1271 			"ICMP6 redirect sent from %s rejected; "
1272 			"hlim=%d (must be 255)\n",
1273 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr)),
1274 			ip6->ip6_hlim));
1275 		goto bad;
1276 	}
1277 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
1278 		nd6log((LOG_ERR,
1279 			"ICMP6 redirect rejected; "
1280 			"redirect dst must be unicast: %s\n",
1281 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1282 		goto bad;
1283 	}
1284     {
1285 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
1286 	struct sockaddr_in6 sin6;
1287 	struct in6_addr *gw6;
1288 
1289 	bzero(&sin6, sizeof(sin6));
1290 	sin6.sin6_family = AF_INET6;
1291 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1292 	memcpy(&sin6.sin6_addr, &reddst6, sizeof(reddst6));
1293 	rt = rtalloc(sin6tosa(&sin6), 0, m->m_pkthdr.ph_rtableid);
1294 	if (rt) {
1295 		if (rt->rt_gateway == NULL ||
1296 		    rt->rt_gateway->sa_family != AF_INET6) {
1297 			nd6log((LOG_ERR,
1298 			    "ICMP6 redirect rejected; no route "
1299 			    "with inet6 gateway found for redirect dst: %s\n",
1300 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1301 			rtfree(rt);
1302 			goto bad;
1303 		}
1304 
1305 		gw6 = &(satosin6(rt->rt_gateway)->sin6_addr);
1306 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
1307 			nd6log((LOG_ERR,
1308 				"ICMP6 redirect rejected; "
1309 				"not equal to gw-for-src=%s (must be same): "
1310 				"%s\n",
1311 				inet_ntop(AF_INET6, gw6, addr, sizeof(addr)),
1312 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1313 			rtfree(rt);
1314 			goto bad;
1315 		}
1316 	} else {
1317 		nd6log((LOG_ERR,
1318 			"ICMP6 redirect rejected; "
1319 			"no route found for redirect dst: %s\n",
1320 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1321 		goto bad;
1322 	}
1323 	rtfree(rt);
1324 	rt = NULL;
1325     }
1326 
1327 	is_router = is_onlink = 0;
1328 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1329 		is_router = 1;	/* router case */
1330 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
1331 		is_onlink = 1;	/* on-link destination case */
1332 	if (!is_router && !is_onlink) {
1333 		nd6log((LOG_ERR,
1334 			"ICMP6 redirect rejected; "
1335 			"neither router case nor onlink case: %s\n",
1336 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1337 		goto bad;
1338 	}
1339 	/* validation passed */
1340 
1341 	icmp6len -= sizeof(*nd_rd);
1342 	if (nd6_options(nd_rd + 1, icmp6len, &ndopts) < 0) {
1343 		nd6log((LOG_INFO, "icmp6_redirect_input: "
1344 			"invalid ND option, rejected: %s\n",
1345 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1346 		/* nd6_options have incremented stats */
1347 		goto freeit;
1348 	}
1349 
1350 	if (ndopts.nd_opts_tgt_lladdr) {
1351 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
1352 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
1353 	}
1354 
1355 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
1356 		nd6log((LOG_INFO,
1357 			"icmp6_redirect_input: lladdrlen mismatch for %s "
1358 			"(if %d, icmp6 packet %d): %s\n",
1359 			inet_ntop(AF_INET6, &redtgt6, addr, sizeof(addr)),
1360 			ifp->if_addrlen, lladdrlen - 2,
1361 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1362 		goto bad;
1363 	}
1364 
1365 	/* RFC 2461 8.3 */
1366 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
1367 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
1368 
1369 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
1370 		/* perform rtredirect */
1371 		struct sockaddr_in6 sdst;
1372 		struct sockaddr_in6 sgw;
1373 		struct sockaddr_in6 ssrc;
1374 		unsigned long rtcount;
1375 		struct rtentry *newrt = NULL;
1376 
1377 		/*
1378 		 * do not install redirect route, if the number of entries
1379 		 * is too much (> hiwat).  note that, the node (= host) will
1380 		 * work just fine even if we do not install redirect route
1381 		 * (there will be additional hops, though).
1382 		 */
1383 		rtcount = rt_timer_queue_count(&icmp6_redirect_timeout_q);
1384 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
1385 			goto freeit;
1386 
1387 		bzero(&sdst, sizeof(sdst));
1388 		bzero(&sgw, sizeof(sgw));
1389 		bzero(&ssrc, sizeof(ssrc));
1390 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
1391 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
1392 			sizeof(struct sockaddr_in6);
1393 		memcpy(&sgw.sin6_addr, &redtgt6, sizeof(struct in6_addr));
1394 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1395 		memcpy(&ssrc.sin6_addr, &src6, sizeof(struct in6_addr));
1396 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), sin6tosa(&ssrc),
1397 		    &newrt, m->m_pkthdr.ph_rtableid);
1398 		if (newrt != NULL && icmp6_redirtimeout > 0) {
1399 			rt_timer_add(newrt, &icmp6_redirect_timeout_q,
1400 			    m->m_pkthdr.ph_rtableid);
1401 		}
1402 		rtfree(newrt);
1403 	}
1404 	/* finally update cached route in each socket via pfctlinput */
1405 	{
1406 		struct sockaddr_in6 sdst;
1407 
1408 		bzero(&sdst, sizeof(sdst));
1409 		sdst.sin6_family = AF_INET6;
1410 		sdst.sin6_len = sizeof(struct sockaddr_in6);
1411 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1412 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
1413 	}
1414 
1415  freeit:
1416 	if_put(ifp);
1417 	m_freem(m);
1418 	return;
1419 
1420  bad:
1421 	if_put(ifp);
1422 	icmp6stat_inc(icp6s_badredirect);
1423 	m_freem(m);
1424 }
1425 
1426 void
1427 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
1428 {
1429 	struct ifnet *ifp = NULL;
1430 	struct in6_addr *ifp_ll6;
1431 	struct in6_addr *nexthop;
1432 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
1433 	struct mbuf *m = NULL;	/* newly allocated one */
1434 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
1435 	struct nd_redirect *nd_rd;
1436 	size_t maxlen;
1437 	u_char *p;
1438 	struct sockaddr_in6 src_sa;
1439 
1440 	icmp6_errcount(ND_REDIRECT, 0);
1441 
1442 	/* if we are not router, we don't send icmp6 redirect */
1443 	if (!ip6_forwarding)
1444 		goto fail;
1445 
1446 	/* sanity check */
1447 	if (m0 == NULL || !rtisvalid(rt))
1448 		goto fail;
1449 
1450 	ifp = if_get(rt->rt_ifidx);
1451 	if (ifp == NULL)
1452 		goto fail;
1453 
1454 	/*
1455 	 * Address check:
1456 	 *  the source address must identify a neighbor, and
1457 	 *  the destination address must not be a multicast address
1458 	 *  [RFC 2461, sec 8.2]
1459 	 */
1460 	sip6 = mtod(m0, struct ip6_hdr *);
1461 	bzero(&src_sa, sizeof(src_sa));
1462 	src_sa.sin6_family = AF_INET6;
1463 	src_sa.sin6_len = sizeof(src_sa);
1464 	src_sa.sin6_addr = sip6->ip6_src;
1465 	/* we don't currently use sin6_scope_id, but eventually use it */
1466 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp->if_index, &sip6->ip6_src);
1467 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
1468 		goto fail;
1469 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
1470 		goto fail;	/* what should we do here? */
1471 
1472 	/* rate limit */
1473 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
1474 		goto fail;
1475 
1476 	/*
1477 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
1478 	 * we almost always ask for an mbuf cluster for simplicity.
1479 	 * (MHLEN < IPV6_MMTU is almost always true)
1480 	 */
1481 #if IPV6_MMTU >= MCLBYTES
1482 # error assumption failed about IPV6_MMTU and MCLBYTES
1483 #endif
1484 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
1485 	if (m && IPV6_MMTU >= MHLEN)
1486 		MCLGET(m, M_DONTWAIT);
1487 	if (!m)
1488 		goto fail;
1489 	m->m_pkthdr.ph_ifidx = 0;
1490 	m->m_len = 0;
1491 	maxlen = m_trailingspace(m);
1492 	maxlen = min(IPV6_MMTU, maxlen);
1493 	/* just for safety */
1494 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
1495 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
1496 		goto fail;
1497 	}
1498 
1499 	{
1500 		/* get ip6 linklocal address for ifp(my outgoing interface). */
1501 		struct in6_ifaddr *ia6;
1502 		if ((ia6 = in6ifa_ifpforlinklocal(ifp, IN6_IFF_TENTATIVE|
1503 		    IN6_IFF_DUPLICATED|IN6_IFF_ANYCAST)) == NULL)
1504 			goto fail;
1505 		ifp_ll6 = &ia6->ia_addr.sin6_addr;
1506 	}
1507 
1508 	/* get ip6 linklocal address for the router. */
1509 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
1510 		struct sockaddr_in6 *sin6;
1511 		sin6 = satosin6(rt->rt_gateway);
1512 		nexthop = &sin6->sin6_addr;
1513 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
1514 			nexthop = NULL;
1515 	} else
1516 		nexthop = NULL;
1517 
1518 	/* ip6 */
1519 	ip6 = mtod(m, struct ip6_hdr *);
1520 	ip6->ip6_flow = 0;
1521 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1522 	ip6->ip6_vfc |= IPV6_VERSION;
1523 	/* ip6->ip6_plen will be set later */
1524 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1525 	ip6->ip6_hlim = 255;
1526 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
1527 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
1528 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
1529 
1530 	/* ND Redirect */
1531 	nd_rd = (struct nd_redirect *)(ip6 + 1);
1532 	nd_rd->nd_rd_type = ND_REDIRECT;
1533 	nd_rd->nd_rd_code = 0;
1534 	nd_rd->nd_rd_reserved = 0;
1535 	if (rt->rt_flags & RTF_GATEWAY) {
1536 		/*
1537 		 * nd_rd->nd_rd_target must be a link-local address in
1538 		 * better router cases.
1539 		 */
1540 		if (!nexthop)
1541 			goto fail;
1542 		bcopy(nexthop, &nd_rd->nd_rd_target,
1543 		      sizeof(nd_rd->nd_rd_target));
1544 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1545 		      sizeof(nd_rd->nd_rd_dst));
1546 	} else {
1547 		/* make sure redtgt == reddst */
1548 		nexthop = &sip6->ip6_dst;
1549 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
1550 		      sizeof(nd_rd->nd_rd_target));
1551 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1552 		      sizeof(nd_rd->nd_rd_dst));
1553 	}
1554 
1555 	p = (u_char *)(nd_rd + 1);
1556 
1557 	{
1558 		/* target lladdr option */
1559 		struct rtentry *nrt;
1560 		int len;
1561 		struct sockaddr_dl *sdl;
1562 		struct nd_opt_hdr *nd_opt;
1563 		char *lladdr;
1564 
1565 		len = sizeof(*nd_opt) + ifp->if_addrlen;
1566 		len = (len + 7) & ~7;	/* round by 8 */
1567 		/* safety check */
1568 		if (len + (p - (u_char *)ip6) > maxlen)
1569 			goto nolladdropt;
1570 		nrt = nd6_lookup(nexthop, 0, ifp, ifp->if_rdomain);
1571 		if ((nrt != NULL) &&
1572 		    (nrt->rt_flags & (RTF_GATEWAY|RTF_LLINFO)) == RTF_LLINFO &&
1573 		    (nrt->rt_gateway->sa_family == AF_LINK) &&
1574 		    (sdl = satosdl(nrt->rt_gateway)) &&
1575 		    sdl->sdl_alen) {
1576 			nd_opt = (struct nd_opt_hdr *)p;
1577 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1578 			nd_opt->nd_opt_len = len >> 3;
1579 			lladdr = (char *)(nd_opt + 1);
1580 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
1581 			p += len;
1582 		}
1583 		rtfree(nrt);
1584 	}
1585   nolladdropt:;
1586 
1587 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1588 
1589 	/* just to be safe */
1590 	if (p - (u_char *)ip6 > maxlen)
1591 		goto noredhdropt;
1592 
1593 	{
1594 		/* redirected header option */
1595 		int len;
1596 		struct nd_opt_rd_hdr *nd_opt_rh;
1597 
1598 		/*
1599 		 * compute the maximum size for icmp6 redirect header option.
1600 		 * XXX room for auth header?
1601 		 */
1602 		len = maxlen - (p - (u_char *)ip6);
1603 		len &= ~7;
1604 
1605 		/*
1606 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
1607 		 * about padding/truncate rule for the original IP packet.
1608 		 * From the discussion on IPv6imp in Feb 1999,
1609 		 * the consensus was:
1610 		 * - "attach as much as possible" is the goal
1611 		 * - pad if not aligned (original size can be guessed by
1612 		 *   original ip6 header)
1613 		 * Following code adds the padding if it is simple enough,
1614 		 * and truncates if not.
1615 		 */
1616 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
1617 			/* not enough room, truncate */
1618 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
1619 			    m0->m_pkthdr.len);
1620 		} else {
1621 			/*
1622 			 * enough room, truncate if not aligned.
1623 			 * we don't pad here for simplicity.
1624 			 */
1625 			size_t extra;
1626 
1627 			extra = m0->m_pkthdr.len % 8;
1628 			if (extra) {
1629 				/* truncate */
1630 				m_adj(m0, -extra);
1631 			}
1632 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
1633 		}
1634 
1635 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
1636 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
1637 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
1638 		nd_opt_rh->nd_opt_rh_len = len >> 3;
1639 		p += sizeof(*nd_opt_rh);
1640 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1641 
1642 		/* connect m0 to m */
1643 		m->m_pkthdr.len += m0->m_pkthdr.len;
1644 		m_cat(m, m0);
1645 		m0 = NULL;
1646 	}
1647 noredhdropt:
1648 	m_freem(m0);
1649 	m0 = NULL;
1650 
1651 	sip6 = mtod(m, struct ip6_hdr *);
1652 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
1653 		sip6->ip6_src.s6_addr16[1] = 0;
1654 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
1655 		sip6->ip6_dst.s6_addr16[1] = 0;
1656 #if 0
1657 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1658 		ip6->ip6_src.s6_addr16[1] = 0;
1659 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1660 		ip6->ip6_dst.s6_addr16[1] = 0;
1661 #endif
1662 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
1663 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
1664 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
1665 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
1666 
1667 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
1668 
1669 	nd_rd->nd_rd_cksum = 0;
1670 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1671 
1672 	/* send the packet to outside... */
1673 	ip6_output(m, NULL, NULL, 0, NULL, NULL);
1674 
1675 	icmp6stat_inc(icp6s_outhist + ND_REDIRECT);
1676 
1677 	if_put(ifp);
1678 	return;
1679 
1680 fail:
1681 	if_put(ifp);
1682 	m_freem(m);
1683 	m_freem(m0);
1684 }
1685 
1686 /*
1687  * ICMPv6 socket option processing.
1688  */
1689 int
1690 icmp6_ctloutput(int op, struct socket *so, int level, int optname,
1691     struct mbuf *m)
1692 {
1693 	int error = 0;
1694 	struct inpcb *in6p = sotoinpcb(so);
1695 
1696 	if (level != IPPROTO_ICMPV6)
1697 		return EINVAL;
1698 
1699 	switch (op) {
1700 	case PRCO_SETOPT:
1701 		switch (optname) {
1702 		case ICMP6_FILTER:
1703 		    {
1704 			struct icmp6_filter *p;
1705 
1706 			if (m == NULL || m->m_len != sizeof(*p)) {
1707 				error = EMSGSIZE;
1708 				break;
1709 			}
1710 			p = mtod(m, struct icmp6_filter *);
1711 			if (!p || !in6p->inp_icmp6filt) {
1712 				error = EINVAL;
1713 				break;
1714 			}
1715 			bcopy(p, in6p->inp_icmp6filt,
1716 				sizeof(struct icmp6_filter));
1717 			error = 0;
1718 			break;
1719 		    }
1720 
1721 		default:
1722 			error = ENOPROTOOPT;
1723 			break;
1724 		}
1725 		break;
1726 
1727 	case PRCO_GETOPT:
1728 		switch (optname) {
1729 		case ICMP6_FILTER:
1730 		    {
1731 			struct icmp6_filter *p;
1732 
1733 			if (!in6p->inp_icmp6filt) {
1734 				error = EINVAL;
1735 				break;
1736 			}
1737 			m->m_len = sizeof(struct icmp6_filter);
1738 			p = mtod(m, struct icmp6_filter *);
1739 			bcopy(in6p->inp_icmp6filt, p,
1740 				sizeof(struct icmp6_filter));
1741 			error = 0;
1742 			break;
1743 		    }
1744 
1745 		default:
1746 			error = ENOPROTOOPT;
1747 			break;
1748 		}
1749 		break;
1750 	}
1751 
1752 	return (error);
1753 }
1754 
1755 /*
1756  * Perform rate limit check.
1757  * Returns 0 if it is okay to send the icmp6 packet.
1758  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
1759  * limitation.
1760  *
1761  * XXX per-destination/type check necessary?
1762  *
1763  * dst - not used at this moment
1764  * type - not used at this moment
1765  * code - not used at this moment
1766  */
1767 int
1768 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
1769 {
1770 	/* PPS limit */
1771 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
1772 	    icmp6errppslim))
1773 		return 1;	/* The packet is subject to rate limit */
1774 	return 0;		/* okay to send */
1775 }
1776 
1777 struct rtentry *
1778 icmp6_mtudisc_clone(struct sockaddr_in6 *dst, u_int rtableid, int ipsec)
1779 {
1780 	struct rtentry *rt;
1781 	int    error;
1782 
1783 	rt = rtalloc(sin6tosa(dst), RT_RESOLVE, rtableid);
1784 
1785 	/* Check if the route is actually usable */
1786 	if (!rtisvalid(rt))
1787 		goto bad;
1788 	/* IPsec needs the route only for PMTU, it can use reject for that */
1789 	if (!ipsec && (rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)))
1790 		goto bad;
1791 
1792 	/*
1793 	 * No PMTU for local routes and permanent neighbors,
1794 	 * ARP and NDP use the same expire timer as the route.
1795 	 */
1796 	if (ISSET(rt->rt_flags, RTF_LOCAL) ||
1797 	    (ISSET(rt->rt_flags, RTF_LLINFO) && rt->rt_expire == 0))
1798 		goto bad;
1799 
1800 	/* If we didn't get a host route, allocate one */
1801 	if ((rt->rt_flags & RTF_HOST) == 0) {
1802 		struct rtentry *nrt;
1803 		struct rt_addrinfo info;
1804 		struct sockaddr_rtlabel sa_rl;
1805 
1806 		memset(&info, 0, sizeof(info));
1807 		info.rti_ifa = rt->rt_ifa;
1808 		info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC;
1809 		info.rti_info[RTAX_DST] = sin6tosa(dst);
1810 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1811 		info.rti_info[RTAX_LABEL] =
1812 		    rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1813 
1814 		error = rtrequest(RTM_ADD, &info, rt->rt_priority, &nrt,
1815 		    rtableid);
1816 		if (error)
1817 			goto bad;
1818 		nrt->rt_rmx = rt->rt_rmx;
1819 		rtfree(rt);
1820 		rt = nrt;
1821 		rtm_send(rt, RTM_ADD, 0, rtableid);
1822 	}
1823 	error = rt_timer_add(rt, &icmp6_mtudisc_timeout_q, rtableid);
1824 	if (error)
1825 		goto bad;
1826 
1827 	return (rt);
1828 bad:
1829 	rtfree(rt);
1830 	return (NULL);
1831 }
1832 
1833 void
1834 icmp6_mtudisc_timeout(struct rtentry *rt, u_int rtableid)
1835 {
1836 	struct ifnet *ifp;
1837 
1838 	NET_ASSERT_LOCKED();
1839 
1840 	ifp = if_get(rt->rt_ifidx);
1841 	if (ifp == NULL)
1842 		return;
1843 
1844 	if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) {
1845 		rtdeletemsg(rt, ifp, rtableid);
1846 	} else {
1847 		if (!(rt->rt_locks & RTV_MTU))
1848 			rt->rt_mtu = 0;
1849 	}
1850 
1851 	if_put(ifp);
1852 }
1853 
1854 const struct sysctl_bounded_args icmpv6ctl_vars[] = {
1855 	{ ICMPV6CTL_ND6_DELAY, &nd6_delay, 0, INT_MAX },
1856 	{ ICMPV6CTL_ND6_UMAXTRIES, &nd6_umaxtries, 0, INT_MAX },
1857 	{ ICMPV6CTL_ND6_MMAXTRIES, &nd6_mmaxtries, 0, INT_MAX },
1858 	{ ICMPV6CTL_ERRPPSLIMIT, &icmp6errppslim, -1, 1000 },
1859 	{ ICMPV6CTL_ND6_MAXNUDHINT, &nd6_maxnudhint, 0, INT_MAX },
1860 	{ ICMPV6CTL_MTUDISC_HIWAT, &icmp6_mtudisc_hiwat, -1, INT_MAX },
1861 	{ ICMPV6CTL_MTUDISC_LOWAT, &icmp6_mtudisc_lowat, -1, INT_MAX },
1862 	{ ICMPV6CTL_ND6_DEBUG, &nd6_debug, 0, 1 },
1863 };
1864 
1865 int
1866 icmp6_sysctl_icmp6stat(void *oldp, size_t *oldlenp, void *newp)
1867 {
1868 	struct icmp6stat *icmp6stat;
1869 	int ret;
1870 
1871 	CTASSERT(sizeof(*icmp6stat) == icp6s_ncounters * sizeof(uint64_t));
1872 	icmp6stat = malloc(sizeof(*icmp6stat), M_TEMP, M_WAITOK|M_ZERO);
1873 	counters_read(icmp6counters, (uint64_t *)icmp6stat, icp6s_ncounters);
1874 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1875 	    icmp6stat, sizeof(*icmp6stat));
1876 	free(icmp6stat, M_TEMP, sizeof(*icmp6stat));
1877 
1878 	return (ret);
1879 }
1880 
1881 int
1882 icmp6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1883     void *newp, size_t newlen)
1884 {
1885 	int error;
1886 
1887 	/* All sysctl names at this level are terminal. */
1888 	if (namelen != 1)
1889 		return (ENOTDIR);
1890 
1891 	switch (name[0]) {
1892 	case ICMPV6CTL_REDIRTIMEOUT:
1893 		NET_LOCK();
1894 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1895 		    &icmp6_redirtimeout, 0, INT_MAX);
1896 		rt_timer_queue_change(&icmp6_redirect_timeout_q,
1897 		    icmp6_redirtimeout);
1898 		NET_UNLOCK();
1899 		break;
1900 
1901 	case ICMPV6CTL_STATS:
1902 		error = icmp6_sysctl_icmp6stat(oldp, oldlenp, newp);
1903 		break;
1904 
1905 	case ICMPV6CTL_ND6_QUEUED:
1906 		error = sysctl_rdint(oldp, oldlenp, newp,
1907 		    atomic_load_int(&ln_hold_total));
1908 		break;
1909 
1910 	default:
1911 		NET_LOCK();
1912 		error = sysctl_bounded_arr(icmpv6ctl_vars,
1913 		    nitems(icmpv6ctl_vars), name, namelen, oldp, oldlenp, newp,
1914 		    newlen);
1915 		NET_UNLOCK();
1916 		break;
1917 	}
1918 
1919 	return (error);
1920 }
1921