xref: /openbsd-src/sys/netinet6/icmp6.c (revision c1a45aed656e7d5627c30c92421893a76f370ccb)
1 /*	$OpenBSD: icmp6.c,v 1.239 2022/04/20 09:38:26 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 /* XXX experimental, turned off */
133 static int icmp6_redirect_lowat = -1;
134 
135 void	icmp6_errcount(int, int);
136 int	icmp6_ratelimit(const struct in6_addr *, const int, const int);
137 const char *icmp6_redirect_diag(struct in6_addr *, struct in6_addr *,
138 	    struct in6_addr *);
139 int	icmp6_notify_error(struct mbuf *, int, int, int);
140 void	icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
141 void	icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
142 
143 void
144 icmp6_init(void)
145 {
146 	mld6_init();
147 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(ip6_mtudisc_timeout);
148 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
149 	icmp6counters = counters_alloc(icp6s_ncounters);
150 }
151 
152 void
153 icmp6_errcount(int type, int code)
154 {
155 	enum icmp6stat_counters c = icp6s_ounknown;
156 
157 	switch (type) {
158 	case ICMP6_DST_UNREACH:
159 		switch (code) {
160 		case ICMP6_DST_UNREACH_NOROUTE:
161 			c = icp6s_odst_unreach_noroute;
162 			break;
163 		case ICMP6_DST_UNREACH_ADMIN:
164 			c = icp6s_odst_unreach_admin;
165 			break;
166 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
167 			c = icp6s_odst_unreach_beyondscope;
168 			break;
169 		case ICMP6_DST_UNREACH_ADDR:
170 			c = icp6s_odst_unreach_addr;
171 			break;
172 		case ICMP6_DST_UNREACH_NOPORT:
173 			c = icp6s_odst_unreach_noport;
174 			break;
175 		}
176 		break;
177 	case ICMP6_PACKET_TOO_BIG:
178 		c = icp6s_opacket_too_big;
179 		break;
180 	case ICMP6_TIME_EXCEEDED:
181 		switch (code) {
182 		case ICMP6_TIME_EXCEED_TRANSIT:
183 			c = icp6s_otime_exceed_transit;
184 			break;
185 		case ICMP6_TIME_EXCEED_REASSEMBLY:
186 			c = icp6s_otime_exceed_reassembly;
187 			break;
188 		}
189 		break;
190 	case ICMP6_PARAM_PROB:
191 		switch (code) {
192 		case ICMP6_PARAMPROB_HEADER:
193 			c = icp6s_oparamprob_header;
194 			break;
195 		case ICMP6_PARAMPROB_NEXTHEADER:
196 			c = icp6s_oparamprob_nextheader;
197 			break;
198 		case ICMP6_PARAMPROB_OPTION:
199 			c = icp6s_oparamprob_option;
200 			break;
201 		}
202 		break;
203 	case ND_REDIRECT:
204 		c = icp6s_oredirect;
205 		break;
206 	}
207 
208 	icmp6stat_inc(c);
209 }
210 
211 /*
212  * Register a Path MTU Discovery callback.
213  */
214 void
215 icmp6_mtudisc_callback_register(void (*func)(struct sockaddr_in6 *, u_int))
216 {
217 	struct icmp6_mtudisc_callback *mc;
218 
219 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list) {
220 		if (mc->mc_func == func)
221 			return;
222 	}
223 
224 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
225 	if (mc == NULL)
226 		panic("%s", __func__);
227 
228 	mc->mc_func = func;
229 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
230 }
231 
232 struct mbuf *
233 icmp6_do_error(struct mbuf *m, int type, int code, int param)
234 {
235 	struct ip6_hdr *oip6, *nip6;
236 	struct icmp6_hdr *icmp6;
237 	u_int preplen;
238 	int off;
239 	int nxt;
240 
241 	icmp6stat_inc(icp6s_error);
242 
243 	/* count per-type-code statistics */
244 	icmp6_errcount(type, code);
245 
246 	if (m->m_len < sizeof(struct ip6_hdr)) {
247 		m = m_pullup(m, sizeof(struct ip6_hdr));
248 		if (m == NULL)
249 			return (NULL);
250 	}
251 	oip6 = mtod(m, struct ip6_hdr *);
252 
253 	/*
254 	 * If the destination address of the erroneous packet is a multicast
255 	 * address, or the packet was sent using link-layer multicast,
256 	 * we should basically suppress sending an error (RFC 2463, Section
257 	 * 2.4).
258 	 * We have two exceptions (the item e.2 in that section):
259 	 * - the Packet Too Big message can be sent for path MTU discovery.
260 	 * - the Parameter Problem Message that can be allowed an icmp6 error
261 	 *   in the option type field.  This check has been done in
262 	 *   ip6_unknown_opt(), so we can just check the type and code.
263 	 */
264 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
265 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
266 	    (type != ICMP6_PACKET_TOO_BIG &&
267 	     (type != ICMP6_PARAM_PROB ||
268 	      code != ICMP6_PARAMPROB_OPTION)))
269 		goto freeit;
270 
271 	/*
272 	 * RFC 2463, 2.4 (e.5): source address check.
273 	 * XXX: the case of anycast source?
274 	 */
275 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
276 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
277 		goto freeit;
278 
279 	/*
280 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
281 	 * don't do it.
282 	 */
283 	nxt = -1;
284 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
285 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
286 		struct icmp6_hdr *icp;
287 
288 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
289 			sizeof(*icp));
290 		if (icp == NULL) {
291 			icmp6stat_inc(icp6s_tooshort);
292 			return (NULL);
293 		}
294 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
295 		    icp->icmp6_type == ND_REDIRECT) {
296 			/*
297 			 * ICMPv6 error
298 			 * Special case: for redirect (which is
299 			 * informational) we must not send icmp6 error.
300 			 */
301 			icmp6stat_inc(icp6s_canterror);
302 			goto freeit;
303 		} else {
304 			/* ICMPv6 informational - send the error */
305 		}
306 	}
307 	else {
308 		/* non-ICMPv6 - send the error */
309 	}
310 
311 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
312 
313 	/* Finally, do rate limitation check. */
314 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
315 		icmp6stat_inc(icp6s_toofreq);
316 		goto freeit;
317 	}
318 
319 	/*
320 	 * OK, ICMP6 can be generated.
321 	 */
322 
323 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
324 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
325 
326 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
327 	M_PREPEND(m, preplen, M_DONTWAIT);
328 	if (m && m->m_len < preplen)
329 		m = m_pullup(m, preplen);
330 	if (m == NULL) {
331 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
332 		return (NULL);
333 	}
334 
335 	nip6 = mtod(m, struct ip6_hdr *);
336 	nip6->ip6_src  = oip6->ip6_src;
337 	nip6->ip6_dst  = oip6->ip6_dst;
338 
339 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_src))
340 		oip6->ip6_src.s6_addr16[1] = 0;
341 	if (IN6_IS_SCOPE_EMBED(&oip6->ip6_dst))
342 		oip6->ip6_dst.s6_addr16[1] = 0;
343 
344 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
345 	icmp6->icmp6_type = type;
346 	icmp6->icmp6_code = code;
347 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
348 
349 	/*
350 	 * icmp6_reflect() is designed to be in the input path.
351 	 * icmp6_error() can be called from both input and output path,
352 	 * and if we are in output path rcvif could contain bogus value.
353 	 * clear m->m_pkthdr.ph_ifidx for safety, we should have enough
354 	 * scope information in ip header (nip6).
355 	 */
356 	m->m_pkthdr.ph_ifidx = 0;
357 
358 	icmp6stat_inc(icp6s_outhist + type);
359 
360 	return (m);
361 
362   freeit:
363 	/*
364 	 * If we can't tell whether or not we can generate ICMP6, free it.
365 	 */
366 	return (m_freem(m));
367 }
368 
369 /*
370  * Generate an error packet of type error in response to bad IP6 packet.
371  */
372 void
373 icmp6_error(struct mbuf *m, int type, int code, int param)
374 {
375 	struct mbuf	*n;
376 
377 	n = icmp6_do_error(m, type, code, param);
378 	if (n != NULL) {
379 		/* header order: IPv6 - ICMPv6 */
380 		if (!icmp6_reflect(&n, sizeof(struct ip6_hdr), NULL))
381 			ip6_send(n);
382 	}
383 }
384 
385 /*
386  * Process a received ICMP6 message.
387  */
388 int
389 icmp6_input(struct mbuf **mp, int *offp, int proto, int af)
390 {
391 #if NCARP > 0
392 	struct ifnet *ifp;
393 #endif
394 	struct mbuf *m = *mp, *n;
395 	struct ip6_hdr *ip6, *nip6;
396 	struct icmp6_hdr *icmp6, *nicmp6;
397 	int off = *offp;
398 	int icmp6len = m->m_pkthdr.len - *offp;
399 	int code, sum, noff;
400 	char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
401 
402 	/*
403 	 * Locate icmp6 structure in mbuf, and check
404 	 * that not corrupted and of at least minimum length
405 	 */
406 
407 	ip6 = mtod(m, struct ip6_hdr *);
408 	if (icmp6len < sizeof(struct icmp6_hdr)) {
409 		icmp6stat_inc(icp6s_tooshort);
410 		goto freeit;
411 	}
412 
413 	/*
414 	 * calculate the checksum
415 	 */
416 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
417 	if (icmp6 == NULL) {
418 		icmp6stat_inc(icp6s_tooshort);
419 		return IPPROTO_DONE;
420 	}
421 	code = icmp6->icmp6_code;
422 
423 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
424 		nd6log((LOG_ERR,
425 		    "ICMP6 checksum error(%d|%x) %s\n",
426 		    icmp6->icmp6_type, sum,
427 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src))));
428 		icmp6stat_inc(icp6s_checksum);
429 		goto freeit;
430 	}
431 
432 #if NPF > 0
433 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
434 		switch (icmp6->icmp6_type) {
435 		/*
436 		 * These ICMP6 types map to other connections.  They must be
437 		 * delivered to pr_ctlinput() also for diverted connections.
438 		 */
439 		case ICMP6_DST_UNREACH:
440 		case ICMP6_PACKET_TOO_BIG:
441 		case ICMP6_TIME_EXCEEDED:
442 		case ICMP6_PARAM_PROB:
443 			/*
444 			 * Do not use the divert-to property of the TCP or UDP
445 			 * rule when doing the PCB lookup for the raw socket.
446 			 */
447 			m->m_pkthdr.pf.flags &=~ PF_TAG_DIVERTED;
448 			break;
449 		default:
450 			goto raw;
451 		}
452 	}
453 #endif /* NPF */
454 
455 #if NCARP > 0
456 	ifp = if_get(m->m_pkthdr.ph_ifidx);
457 	if (ifp == NULL)
458 		goto freeit;
459 
460 	if (icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
461 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
462 	    ip6->ip6_dst.s6_addr32, 1)) {
463 		if_put(ifp);
464 		goto freeit;
465 	}
466 
467 	if_put(ifp);
468 #endif
469 	icmp6stat_inc(icp6s_inhist + icmp6->icmp6_type);
470 
471 	switch (icmp6->icmp6_type) {
472 	case ICMP6_DST_UNREACH:
473 		switch (code) {
474 		case ICMP6_DST_UNREACH_NOROUTE:
475 			code = PRC_UNREACH_NET;
476 			break;
477 		case ICMP6_DST_UNREACH_ADMIN:
478 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
479 			break;
480 		case ICMP6_DST_UNREACH_ADDR:
481 			code = PRC_HOSTDEAD;
482 			break;
483 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
484 			/* I mean "source address was incorrect." */
485 			code = PRC_PARAMPROB;
486 			break;
487 		case ICMP6_DST_UNREACH_NOPORT:
488 			code = PRC_UNREACH_PORT;
489 			break;
490 		default:
491 			goto badcode;
492 		}
493 		goto deliver;
494 
495 	case ICMP6_PACKET_TOO_BIG:
496 		/* MTU is checked in icmp6_mtudisc_update. */
497 		code = PRC_MSGSIZE;
498 
499 		/*
500 		 * Updating the path MTU will be done after examining
501 		 * intermediate extension headers.
502 		 */
503 		goto deliver;
504 
505 	case ICMP6_TIME_EXCEEDED:
506 		switch (code) {
507 		case ICMP6_TIME_EXCEED_TRANSIT:
508 			code = PRC_TIMXCEED_INTRANS;
509 			break;
510 		case ICMP6_TIME_EXCEED_REASSEMBLY:
511 			code = PRC_TIMXCEED_REASS;
512 			break;
513 		default:
514 			goto badcode;
515 		}
516 		goto deliver;
517 
518 	case ICMP6_PARAM_PROB:
519 		switch (code) {
520 		case ICMP6_PARAMPROB_NEXTHEADER:
521 			code = PRC_UNREACH_PROTOCOL;
522 			break;
523 		case ICMP6_PARAMPROB_HEADER:
524 		case ICMP6_PARAMPROB_OPTION:
525 			code = PRC_PARAMPROB;
526 			break;
527 		default:
528 			goto badcode;
529 		}
530 		goto deliver;
531 
532 	case ICMP6_ECHO_REQUEST:
533 		if (code != 0)
534 			goto badcode;
535 		/*
536 		 * Copy mbuf to send to two data paths: userland socket(s),
537 		 * and to the querier (echo reply).
538 		 * m: a copy for socket, n: a copy for querier
539 		 */
540 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
541 			/* Give up local */
542 			n = m;
543 			m = *mp = NULL;
544 			goto deliverecho;
545 		}
546 		/*
547 		 * If the first mbuf is shared, or the first mbuf is too short,
548 		 * copy the first part of the data into a fresh mbuf.
549 		 * Otherwise, we will wrongly overwrite both copies.
550 		 */
551 		if ((n->m_flags & M_EXT) != 0 ||
552 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
553 			struct mbuf *n0 = n;
554 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
555 
556 			/*
557 			 * Prepare an internal mbuf.  m_pullup() doesn't
558 			 * always copy the length we specified.
559 			 */
560 			if (maxlen >= MCLBYTES) {
561 				/* Give up remote */
562 				m_freem(n0);
563 				break;
564 			}
565 			MGETHDR(n, M_DONTWAIT, n0->m_type);
566 			if (n && maxlen >= MHLEN) {
567 				MCLGET(n, M_DONTWAIT);
568 				if ((n->m_flags & M_EXT) == 0) {
569 					m_free(n);
570 					n = NULL;
571 				}
572 			}
573 			if (n == NULL) {
574 				/* Give up local */
575 				m_freem(n0);
576 				n = m;
577 				m = *mp = NULL;
578 				goto deliverecho;
579 			}
580 			M_MOVE_PKTHDR(n, n0);
581 			/*
582 			 * Copy IPv6 and ICMPv6 only.
583 			 */
584 			nip6 = mtod(n, struct ip6_hdr *);
585 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
586 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
587 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
588 			noff = sizeof(struct ip6_hdr);
589 			n->m_len = noff + sizeof(struct icmp6_hdr);
590 			/*
591 			 * Adjust mbuf.  ip6_plen will be adjusted in
592 			 * ip6_output().
593 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
594 			 */
595 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
596 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
597 			m_adj(n0, off + sizeof(struct icmp6_hdr));
598 			n->m_next = n0;
599 		} else {
600 	 deliverecho:
601 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
602 			    sizeof(*nicmp6));
603 			noff = off;
604 		}
605 		if (n) {
606 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
607 			nicmp6->icmp6_code = 0;
608 			icmp6stat_inc(icp6s_reflect);
609 			icmp6stat_inc(icp6s_outhist + ICMP6_ECHO_REPLY);
610 			if (!icmp6_reflect(&n, noff, NULL))
611 				ip6_send(n);
612 		}
613 		if (!m)
614 			goto freeit;
615 		break;
616 
617 	case ICMP6_ECHO_REPLY:
618 		if (code != 0)
619 			goto badcode;
620 		break;
621 
622 	case MLD_LISTENER_QUERY:
623 	case MLD_LISTENER_REPORT:
624 		if (icmp6len < sizeof(struct mld_hdr))
625 			goto badlen;
626 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
627 			/* give up local */
628 			mld6_input(m, off);
629 			m = NULL;
630 			goto freeit;
631 		}
632 		mld6_input(n, off);
633 		/* m stays. */
634 		break;
635 
636 	case MLD_LISTENER_DONE:
637 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
638 			goto badlen;
639 		break;		/* nothing to be done in kernel */
640 
641 	case MLD_MTRACE_RESP:
642 	case MLD_MTRACE:
643 		/* XXX: these two are experimental.  not officially defined. */
644 		/* XXX: per-interface statistics? */
645 		break;		/* just pass it to applications */
646 
647 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
648 		/* IPv6 Node Information Queries are not supported */
649 		break;
650 	case ICMP6_WRUREPLY:
651 		break;
652 
653 	case ND_ROUTER_SOLICIT:
654 	case ND_ROUTER_ADVERT:
655 		if (code != 0)
656 			goto badcode;
657 		if ((icmp6->icmp6_type == ND_ROUTER_SOLICIT && icmp6len <
658 		    sizeof(struct nd_router_solicit)) ||
659 		    (icmp6->icmp6_type == ND_ROUTER_ADVERT && icmp6len <
660 		    sizeof(struct nd_router_advert)))
661 			goto badlen;
662 
663 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
664 			/* give up local */
665 			nd6_rtr_cache(m, off, icmp6len,
666 			    icmp6->icmp6_type);
667 			m = NULL;
668 			goto freeit;
669 		}
670 		nd6_rtr_cache(n, off, icmp6len, icmp6->icmp6_type);
671 		/* m stays. */
672 		break;
673 
674 	case ND_NEIGHBOR_SOLICIT:
675 		if (code != 0)
676 			goto badcode;
677 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
678 			goto badlen;
679 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
680 			/* give up local */
681 			nd6_ns_input(m, off, icmp6len);
682 			m = NULL;
683 			goto freeit;
684 		}
685 		nd6_ns_input(n, off, icmp6len);
686 		/* m stays. */
687 		break;
688 
689 	case ND_NEIGHBOR_ADVERT:
690 		if (code != 0)
691 			goto badcode;
692 		if (icmp6len < sizeof(struct nd_neighbor_advert))
693 			goto badlen;
694 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
695 			/* give up local */
696 			nd6_na_input(m, off, icmp6len);
697 			m = NULL;
698 			goto freeit;
699 		}
700 		nd6_na_input(n, off, icmp6len);
701 		/* m stays. */
702 		break;
703 
704 	case ND_REDIRECT:
705 		if (code != 0)
706 			goto badcode;
707 		if (icmp6len < sizeof(struct nd_redirect))
708 			goto badlen;
709 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
710 			/* give up local */
711 			icmp6_redirect_input(m, off);
712 			m = NULL;
713 			goto freeit;
714 		}
715 		icmp6_redirect_input(n, off);
716 		/* m stays. */
717 		break;
718 
719 	case ICMP6_ROUTER_RENUMBERING:
720 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
721 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
722 			goto badcode;
723 		if (icmp6len < sizeof(struct icmp6_router_renum))
724 			goto badlen;
725 		break;
726 
727 	default:
728 		nd6log((LOG_DEBUG,
729 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%u)\n",
730 		    icmp6->icmp6_type,
731 		    inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src)),
732 		    inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst)),
733 		    m->m_pkthdr.ph_ifidx));
734 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
735 			/* ICMPv6 error: MUST deliver it by spec... */
736 			code = PRC_NCMDS;
737 			/* deliver */
738 		} else {
739 			/* ICMPv6 informational: MUST not deliver */
740 			break;
741 		}
742 deliver:
743 		if (icmp6_notify_error(m, off, icmp6len, code)) {
744 			/* In this case, m should've been freed. */
745 			return (IPPROTO_DONE);
746 		}
747 		break;
748 
749 badcode:
750 		icmp6stat_inc(icp6s_badcode);
751 		break;
752 
753 badlen:
754 		icmp6stat_inc(icp6s_badlen);
755 		break;
756 	}
757 
758 #if NPF > 0
759 raw:
760 #endif
761 	/* deliver the packet to appropriate sockets */
762 	return rip6_input(mp, offp, proto, af);
763 
764  freeit:
765 	m_freem(m);
766 	return IPPROTO_DONE;
767 }
768 
769 int
770 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
771 {
772 	struct icmp6_hdr *icmp6;
773 	struct ip6_hdr *eip6;
774 	u_int32_t notifymtu;
775 	struct sockaddr_in6 icmp6src, icmp6dst;
776 
777 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
778 		icmp6stat_inc(icp6s_tooshort);
779 		goto freeit;
780 	}
781 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
782 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
783 	if (icmp6 == NULL) {
784 		icmp6stat_inc(icp6s_tooshort);
785 		return (-1);
786 	}
787 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
788 
789 	/* Detect the upper level protocol */
790 	{
791 		void (*ctlfunc)(int, struct sockaddr *, u_int, void *);
792 		u_int8_t nxt = eip6->ip6_nxt;
793 		int eoff = off + sizeof(struct icmp6_hdr) +
794 			sizeof(struct ip6_hdr);
795 		struct ip6ctlparam ip6cp;
796 		struct in6_addr *finaldst = NULL;
797 		int icmp6type = icmp6->icmp6_type;
798 		struct ip6_frag *fh;
799 		struct ip6_rthdr *rth;
800 		struct ip6_rthdr0 *rth0;
801 		int rthlen;
802 
803 		while (1) { /* XXX: should avoid infinite loop explicitly? */
804 			struct ip6_ext *eh;
805 
806 			switch (nxt) {
807 			case IPPROTO_HOPOPTS:
808 			case IPPROTO_DSTOPTS:
809 			case IPPROTO_AH:
810 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
811 					       eoff, sizeof(*eh));
812 				if (eh == NULL) {
813 					icmp6stat_inc(icp6s_tooshort);
814 					return (-1);
815 				}
816 
817 				if (nxt == IPPROTO_AH)
818 					eoff += (eh->ip6e_len + 2) << 2;
819 				else
820 					eoff += (eh->ip6e_len + 1) << 3;
821 				nxt = eh->ip6e_nxt;
822 				break;
823 			case IPPROTO_ROUTING:
824 				/*
825 				 * When the erroneous packet contains a
826 				 * routing header, we should examine the
827 				 * header to determine the final destination.
828 				 * Otherwise, we can't properly update
829 				 * information that depends on the final
830 				 * destination (e.g. path MTU).
831 				 */
832 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
833 					       eoff, sizeof(*rth));
834 				if (rth == NULL) {
835 					icmp6stat_inc(icp6s_tooshort);
836 					return (-1);
837 				}
838 				rthlen = (rth->ip6r_len + 1) << 3;
839 				/*
840 				 * XXX: currently there is no
841 				 * officially defined type other
842 				 * than type-0.
843 				 * Note that if the segment left field
844 				 * is 0, all intermediate hops must
845 				 * have been passed.
846 				 */
847 				if (rth->ip6r_segleft &&
848 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
849 					int hops;
850 
851 					IP6_EXTHDR_GET(rth0,
852 						       struct ip6_rthdr0 *, m,
853 						       eoff, rthlen);
854 					if (rth0 == NULL) {
855 						icmp6stat_inc(icp6s_tooshort);
856 						return (-1);
857 					}
858 					/* just ignore a bogus header */
859 					if ((rth0->ip6r0_len % 2) == 0 &&
860 					    (hops = rth0->ip6r0_len/2))
861 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
862 				}
863 				eoff += rthlen;
864 				nxt = rth->ip6r_nxt;
865 				break;
866 			case IPPROTO_FRAGMENT:
867 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
868 					       eoff, sizeof(*fh));
869 				if (fh == NULL) {
870 					icmp6stat_inc(icp6s_tooshort);
871 					return (-1);
872 				}
873 				/*
874 				 * Data after a fragment header is meaningless
875 				 * unless it is the first fragment, but
876 				 * we'll go to the notify label for path MTU
877 				 * discovery.
878 				 */
879 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
880 					goto notify;
881 
882 				eoff += sizeof(struct ip6_frag);
883 				nxt = fh->ip6f_nxt;
884 				break;
885 			default:
886 				/*
887 				 * This case includes ESP and the No Next
888 				 * Header.  In such cases going to the notify
889 				 * label does not have any meaning
890 				 * (i.e. ctlfunc will be NULL), but we go
891 				 * anyway since we might have to update
892 				 * path MTU information.
893 				 */
894 				goto notify;
895 			}
896 		}
897 	  notify:
898 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
899 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
900 		if (icmp6 == NULL) {
901 			icmp6stat_inc(icp6s_tooshort);
902 			return (-1);
903 		}
904 
905 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
906 		bzero(&icmp6dst, sizeof(icmp6dst));
907 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
908 		icmp6dst.sin6_family = AF_INET6;
909 		if (finaldst == NULL)
910 			icmp6dst.sin6_addr = eip6->ip6_dst;
911 		else
912 			icmp6dst.sin6_addr = *finaldst;
913 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
914 		    &icmp6dst.sin6_addr);
915 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst, NULL)) {
916 			/* should be impossible */
917 			nd6log((LOG_DEBUG,
918 			    "icmp6_notify_error: in6_embedscope failed\n"));
919 			goto freeit;
920 		}
921 
922 		/*
923 		 * retrieve parameters from the inner IPv6 header, and convert
924 		 * them into sockaddr structures.
925 		 */
926 		bzero(&icmp6src, sizeof(icmp6src));
927 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
928 		icmp6src.sin6_family = AF_INET6;
929 		icmp6src.sin6_addr = eip6->ip6_src;
930 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
931 		    &icmp6src.sin6_addr);
932 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src, NULL)) {
933 			/* should be impossible */
934 			nd6log((LOG_DEBUG,
935 			    "icmp6_notify_error: in6_embedscope failed\n"));
936 			goto freeit;
937 		}
938 		icmp6src.sin6_flowinfo =
939 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
940 
941 		if (finaldst == NULL)
942 			finaldst = &eip6->ip6_dst;
943 		ip6cp.ip6c_m = m;
944 		ip6cp.ip6c_icmp6 = icmp6;
945 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
946 		ip6cp.ip6c_off = eoff;
947 		ip6cp.ip6c_finaldst = finaldst;
948 		ip6cp.ip6c_src = &icmp6src;
949 		ip6cp.ip6c_nxt = nxt;
950 #if NPF > 0
951 		pf_pkt_addr_changed(m);
952 #endif
953 
954 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
955 			notifymtu = ntohl(icmp6->icmp6_mtu);
956 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
957 		}
958 
959 		ctlfunc = inet6sw[ip6_protox[nxt]].pr_ctlinput;
960 		if (ctlfunc)
961 			(*ctlfunc)(code, sin6tosa(&icmp6dst),
962 			    m->m_pkthdr.ph_rtableid, &ip6cp);
963 	}
964 	return (0);
965 
966   freeit:
967 	m_freem(m);
968 	return (-1);
969 }
970 
971 void
972 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
973 {
974 	unsigned long rtcount;
975 	struct icmp6_mtudisc_callback *mc;
976 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
977 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
978 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
979 	u_int mtu = ntohl(icmp6->icmp6_mtu);
980 	struct rtentry *rt = NULL;
981 	struct sockaddr_in6 sin6;
982 
983 	if (mtu < IPV6_MMTU)
984 		return;
985 
986 	/*
987 	 * allow non-validated cases if memory is plenty, to make traffic
988 	 * from non-connected pcb happy.
989 	 */
990 	rtcount = rt_timer_queue_count(icmp6_mtudisc_timeout_q);
991 	if (validated) {
992 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
993 			return;
994 		else if (0 <= icmp6_mtudisc_lowat &&
995 		    rtcount > icmp6_mtudisc_lowat) {
996 			/*
997 			 * XXX nuke a victim, install the new one.
998 			 */
999 		}
1000 	} else {
1001 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1002 			return;
1003 	}
1004 
1005 	bzero(&sin6, sizeof(sin6));
1006 	sin6.sin6_family = PF_INET6;
1007 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1008 	sin6.sin6_addr = *dst;
1009 	/* XXX normally, this won't happen */
1010 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1011 		sin6.sin6_addr.s6_addr16[1] = htons(m->m_pkthdr.ph_ifidx);
1012 	}
1013 	sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.ph_ifidx,
1014 	    &sin6.sin6_addr);
1015 
1016 	rt = icmp6_mtudisc_clone(&sin6, m->m_pkthdr.ph_rtableid, 0);
1017 
1018 	if (rt != NULL && ISSET(rt->rt_flags, RTF_HOST) &&
1019 	    !(rt->rt_locks & RTV_MTU) &&
1020 	    (rt->rt_mtu > mtu || rt->rt_mtu == 0)) {
1021 		struct ifnet *ifp;
1022 
1023 		ifp = if_get(rt->rt_ifidx);
1024 		if (ifp != NULL && mtu < ifp->if_mtu) {
1025 			icmp6stat_inc(icp6s_pmtuchg);
1026 			rt->rt_mtu = mtu;
1027 		}
1028 		if_put(ifp);
1029 	}
1030 	rtfree(rt);
1031 
1032 	/*
1033 	 * Notify protocols that the MTU for this destination
1034 	 * has changed.
1035 	 */
1036 	LIST_FOREACH(mc, &icmp6_mtudisc_callbacks, mc_list)
1037 		(*mc->mc_func)(&sin6, m->m_pkthdr.ph_rtableid);
1038 }
1039 
1040 /*
1041  * Reflect the ip6 packet back to the source.
1042  * OFF points to the icmp6 header, counted from the top of the mbuf.
1043  */
1044 int
1045 icmp6_reflect(struct mbuf **mp, size_t off, struct sockaddr *sa)
1046 {
1047 	struct mbuf *m = *mp;
1048 	struct rtentry *rt = NULL;
1049 	struct ip6_hdr *ip6;
1050 	struct icmp6_hdr *icmp6;
1051 	struct in6_addr t, *src = NULL;
1052 	struct sockaddr_in6 sa6_src, sa6_dst;
1053 	u_int rtableid;
1054 	u_int8_t pfflags;
1055 
1056 	CTASSERT(sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) <= MHLEN);
1057 
1058 	/* too short to reflect */
1059 	if (off < sizeof(struct ip6_hdr)) {
1060 		nd6log((LOG_DEBUG,
1061 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1062 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1063 		    __FILE__, __LINE__));
1064 		goto bad;
1065 	}
1066 
1067 	if (m->m_pkthdr.ph_loopcnt++ >= M_MAXLOOP) {
1068 		m_freemp(mp);
1069 		return (ELOOP);
1070 	}
1071 	rtableid = m->m_pkthdr.ph_rtableid;
1072 	pfflags = m->m_pkthdr.pf.flags;
1073 	m_resethdr(m);
1074 	m->m_pkthdr.ph_rtableid = rtableid;
1075 	m->m_pkthdr.pf.flags = pfflags & PF_TAG_GENERATED;
1076 
1077 	/*
1078 	 * If there are extra headers between IPv6 and ICMPv6, strip
1079 	 * off that header first.
1080 	 */
1081 	if (off > sizeof(struct ip6_hdr)) {
1082 		size_t l;
1083 		struct ip6_hdr nip6;
1084 
1085 		l = off - sizeof(struct ip6_hdr);
1086 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1087 		m_adj(m, l);
1088 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1089 		if (m->m_len < l) {
1090 			if ((m = *mp = m_pullup(m, l)) == NULL)
1091 				return (EMSGSIZE);
1092 		}
1093 		memcpy(mtod(m, caddr_t), &nip6, sizeof(nip6));
1094 	} else /* off == sizeof(struct ip6_hdr) */ {
1095 		size_t l;
1096 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1097 		if (m->m_len < l) {
1098 			if ((m = *mp = m_pullup(m, l)) == NULL)
1099 				return (EMSGSIZE);
1100 		}
1101 	}
1102 	ip6 = mtod(m, struct ip6_hdr *);
1103 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1104 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1105 
1106 	t = ip6->ip6_dst;
1107 	/*
1108 	 * ip6_input() drops a packet if its src is multicast.
1109 	 * So, the src is never multicast.
1110 	 */
1111 	ip6->ip6_dst = ip6->ip6_src;
1112 
1113 	/*
1114 	 * XXX: make sure to embed scope zone information, using
1115 	 * already embedded IDs or the received interface (if any).
1116 	 * Note that rcvif may be NULL.
1117 	 * TODO: scoped routing case (XXX).
1118 	 */
1119 	bzero(&sa6_src, sizeof(sa6_src));
1120 	sa6_src.sin6_family = AF_INET6;
1121 	sa6_src.sin6_len = sizeof(sa6_src);
1122 	sa6_src.sin6_addr = ip6->ip6_dst;
1123 	bzero(&sa6_dst, sizeof(sa6_dst));
1124 	sa6_dst.sin6_family = AF_INET6;
1125 	sa6_dst.sin6_len = sizeof(sa6_dst);
1126 	sa6_dst.sin6_addr = t;
1127 
1128 	if (sa == NULL) {
1129 		/*
1130 		 * If the incoming packet was addressed directly to us (i.e.
1131 		 * unicast), use dst as the src for the reply. The
1132 		 * IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED case would be VERY rare,
1133 		 * but is possible (for example) when we encounter an error
1134 		 * while forwarding procedure destined to a duplicated address
1135 		 * of ours.
1136 		 */
1137 		rt = rtalloc(sin6tosa(&sa6_dst), 0, rtableid);
1138 		if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL) &&
1139 		    !ISSET(ifatoia6(rt->rt_ifa)->ia6_flags,
1140 		    IN6_IFF_ANYCAST|IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED)) {
1141 			src = &t;
1142 		}
1143 		rtfree(rt);
1144 		rt = NULL;
1145 		sa = sin6tosa(&sa6_src);
1146 	}
1147 
1148 	if (src == NULL) {
1149 		struct in6_ifaddr *ia6;
1150 
1151 		/*
1152 		 * This case matches to multicasts, our anycast, or unicasts
1153 		 * that we do not own.  Select a source address based on the
1154 		 * source address of the erroneous packet.
1155 		 */
1156 		rt = rtalloc(sa, RT_RESOLVE, rtableid);
1157 		if (!rtisvalid(rt)) {
1158 			char addr[INET6_ADDRSTRLEN];
1159 
1160 			nd6log((LOG_DEBUG,
1161 			    "%s: source can't be determined: dst=%s\n",
1162 			    __func__, inet_ntop(AF_INET6, &sa6_src.sin6_addr,
1163 			    addr, sizeof(addr))));
1164 			rtfree(rt);
1165 			goto bad;
1166 		}
1167 		ia6 = in6_ifawithscope(rt->rt_ifa->ifa_ifp, &t, rtableid);
1168 		if (ia6 != NULL)
1169 			src = &ia6->ia_addr.sin6_addr;
1170 		if (src == NULL)
1171 			src = &ifatoia6(rt->rt_ifa)->ia_addr.sin6_addr;
1172 	}
1173 
1174 	ip6->ip6_src = *src;
1175 	rtfree(rt);
1176 
1177 	ip6->ip6_flow = 0;
1178 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1179 	ip6->ip6_vfc |= IPV6_VERSION;
1180 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1181 	ip6->ip6_hlim = ip6_defhlim;
1182 
1183 	icmp6->icmp6_cksum = 0;
1184 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1185 
1186 	/*
1187 	 * XXX option handling
1188 	 */
1189 
1190 	m->m_flags &= ~(M_BCAST|M_MCAST);
1191 	return (0);
1192 
1193  bad:
1194 	m_freemp(mp);
1195 	return (EHOSTUNREACH);
1196 }
1197 
1198 void
1199 icmp6_fasttimo(void)
1200 {
1201 
1202 	mld6_fasttimeo();
1203 }
1204 
1205 const char *
1206 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
1207     struct in6_addr *tgt6)
1208 {
1209 	static char buf[1024]; /* XXX */
1210 	char src[INET6_ADDRSTRLEN];
1211 	char dst[INET6_ADDRSTRLEN];
1212 	char tgt[INET6_ADDRSTRLEN];
1213 
1214 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
1215 		 inet_ntop(AF_INET6, src6, src, sizeof(src)),
1216 		 inet_ntop(AF_INET6, dst6, dst, sizeof(dst)),
1217 		 inet_ntop(AF_INET6, tgt6, tgt, sizeof(tgt)));
1218 	return buf;
1219 }
1220 
1221 void
1222 icmp6_redirect_input(struct mbuf *m, int off)
1223 {
1224 	struct ifnet *ifp;
1225 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1226 	struct nd_redirect *nd_rd;
1227 	int icmp6len = ntohs(ip6->ip6_plen);
1228 	char *lladdr = NULL;
1229 	int lladdrlen = 0;
1230 	struct rtentry *rt = NULL;
1231 	int is_router;
1232 	int is_onlink;
1233 	struct in6_addr src6 = ip6->ip6_src;
1234 	struct in6_addr redtgt6;
1235 	struct in6_addr reddst6;
1236 	union nd_opts ndopts;
1237 	char addr[INET6_ADDRSTRLEN];
1238 
1239 	ifp = if_get(m->m_pkthdr.ph_ifidx);
1240 	if (ifp == NULL)
1241 		return;
1242 
1243 	/* XXX if we are router, we don't update route by icmp6 redirect */
1244 	if (ip6_forwarding)
1245 		goto freeit;
1246 	if (!(ifp->if_xflags & IFXF_AUTOCONF6))
1247 		goto freeit;
1248 
1249 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
1250 	if (nd_rd == NULL) {
1251 		icmp6stat_inc(icp6s_tooshort);
1252 		if_put(ifp);
1253 		return;
1254 	}
1255 	redtgt6 = nd_rd->nd_rd_target;
1256 	reddst6 = nd_rd->nd_rd_dst;
1257 
1258 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1259 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
1260 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
1261 		reddst6.s6_addr16[1] = htons(ifp->if_index);
1262 
1263 	/* validation */
1264 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
1265 		nd6log((LOG_ERR,
1266 			"ICMP6 redirect sent from %s rejected; "
1267 			"must be from linklocal\n",
1268 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr))));
1269 		goto bad;
1270 	}
1271 	if (ip6->ip6_hlim != 255) {
1272 		nd6log((LOG_ERR,
1273 			"ICMP6 redirect sent from %s rejected; "
1274 			"hlim=%d (must be 255)\n",
1275 			inet_ntop(AF_INET6, &src6, addr, sizeof(addr)),
1276 			ip6->ip6_hlim));
1277 		goto bad;
1278 	}
1279 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
1280 		nd6log((LOG_ERR,
1281 			"ICMP6 redirect rejected; "
1282 			"redirect dst must be unicast: %s\n",
1283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1284 		goto bad;
1285 	}
1286     {
1287 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
1288 	struct sockaddr_in6 sin6;
1289 	struct in6_addr *gw6;
1290 
1291 	bzero(&sin6, sizeof(sin6));
1292 	sin6.sin6_family = AF_INET6;
1293 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1294 	memcpy(&sin6.sin6_addr, &reddst6, sizeof(reddst6));
1295 	rt = rtalloc(sin6tosa(&sin6), 0, m->m_pkthdr.ph_rtableid);
1296 	if (rt) {
1297 		if (rt->rt_gateway == NULL ||
1298 		    rt->rt_gateway->sa_family != AF_INET6) {
1299 			nd6log((LOG_ERR,
1300 			    "ICMP6 redirect rejected; no route "
1301 			    "with inet6 gateway found for redirect dst: %s\n",
1302 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1303 			rtfree(rt);
1304 			goto bad;
1305 		}
1306 
1307 		gw6 = &(satosin6(rt->rt_gateway)->sin6_addr);
1308 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
1309 			nd6log((LOG_ERR,
1310 				"ICMP6 redirect rejected; "
1311 				"not equal to gw-for-src=%s (must be same): "
1312 				"%s\n",
1313 				inet_ntop(AF_INET6, gw6, addr, sizeof(addr)),
1314 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1315 			rtfree(rt);
1316 			goto bad;
1317 		}
1318 	} else {
1319 		nd6log((LOG_ERR,
1320 			"ICMP6 redirect rejected; "
1321 			"no route found for redirect dst: %s\n",
1322 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1323 		goto bad;
1324 	}
1325 	rtfree(rt);
1326 	rt = NULL;
1327     }
1328 
1329 	is_router = is_onlink = 0;
1330 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
1331 		is_router = 1;	/* router case */
1332 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
1333 		is_onlink = 1;	/* on-link destination case */
1334 	if (!is_router && !is_onlink) {
1335 		nd6log((LOG_ERR,
1336 			"ICMP6 redirect rejected; "
1337 			"neither router case nor onlink case: %s\n",
1338 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1339 		goto bad;
1340 	}
1341 	/* validation passed */
1342 
1343 	icmp6len -= sizeof(*nd_rd);
1344 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
1345 	if (nd6_options(&ndopts) < 0) {
1346 		nd6log((LOG_INFO, "icmp6_redirect_input: "
1347 			"invalid ND option, rejected: %s\n",
1348 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1349 		/* nd6_options have incremented stats */
1350 		goto freeit;
1351 	}
1352 
1353 	if (ndopts.nd_opts_tgt_lladdr) {
1354 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
1355 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
1356 	}
1357 
1358 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
1359 		nd6log((LOG_INFO,
1360 			"icmp6_redirect_input: lladdrlen mismatch for %s "
1361 			"(if %d, icmp6 packet %d): %s\n",
1362 			inet_ntop(AF_INET6, &redtgt6, addr, sizeof(addr)),
1363 			ifp->if_addrlen, lladdrlen - 2,
1364 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
1365 		goto bad;
1366 	}
1367 
1368 	/* RFC 2461 8.3 */
1369 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
1370 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
1371 
1372 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
1373 		/* perform rtredirect */
1374 		struct sockaddr_in6 sdst;
1375 		struct sockaddr_in6 sgw;
1376 		struct sockaddr_in6 ssrc;
1377 		unsigned long rtcount;
1378 		struct rtentry *newrt = NULL;
1379 
1380 		/*
1381 		 * do not install redirect route, if the number of entries
1382 		 * is too much (> hiwat).  note that, the node (= host) will
1383 		 * work just fine even if we do not install redirect route
1384 		 * (there will be additional hops, though).
1385 		 */
1386 		rtcount = rt_timer_queue_count(icmp6_redirect_timeout_q);
1387 		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
1388 			goto freeit;
1389 		else if (0 <= icmp6_redirect_lowat &&
1390 		    rtcount > icmp6_redirect_lowat) {
1391 			/*
1392 			 * XXX nuke a victim, install the new one.
1393 			 */
1394 		}
1395 
1396 		bzero(&sdst, sizeof(sdst));
1397 		bzero(&sgw, sizeof(sgw));
1398 		bzero(&ssrc, sizeof(ssrc));
1399 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
1400 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
1401 			sizeof(struct sockaddr_in6);
1402 		memcpy(&sgw.sin6_addr, &redtgt6, sizeof(struct in6_addr));
1403 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1404 		memcpy(&ssrc.sin6_addr, &src6, sizeof(struct in6_addr));
1405 		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), sin6tosa(&ssrc),
1406 		    &newrt, m->m_pkthdr.ph_rtableid);
1407 		if (newrt != NULL && icmp6_redirtimeout > 0) {
1408 			rt_timer_add(newrt, icmp6_redirect_timeout,
1409 			    icmp6_redirect_timeout_q, m->m_pkthdr.ph_rtableid);
1410 		}
1411 		rtfree(newrt);
1412 	}
1413 	/* finally update cached route in each socket via pfctlinput */
1414 	{
1415 		struct sockaddr_in6 sdst;
1416 
1417 		bzero(&sdst, sizeof(sdst));
1418 		sdst.sin6_family = AF_INET6;
1419 		sdst.sin6_len = sizeof(struct sockaddr_in6);
1420 		memcpy(&sdst.sin6_addr, &reddst6, sizeof(struct in6_addr));
1421 		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
1422 	}
1423 
1424  freeit:
1425 	if_put(ifp);
1426 	m_freem(m);
1427 	return;
1428 
1429  bad:
1430 	if_put(ifp);
1431 	icmp6stat_inc(icp6s_badredirect);
1432 	m_freem(m);
1433 }
1434 
1435 void
1436 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
1437 {
1438 	struct ifnet *ifp = NULL;
1439 	struct in6_addr *ifp_ll6;
1440 	struct in6_addr *nexthop;
1441 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
1442 	struct mbuf *m = NULL;	/* newly allocated one */
1443 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
1444 	struct nd_redirect *nd_rd;
1445 	size_t maxlen;
1446 	u_char *p;
1447 	struct sockaddr_in6 src_sa;
1448 
1449 	icmp6_errcount(ND_REDIRECT, 0);
1450 
1451 	/* if we are not router, we don't send icmp6 redirect */
1452 	if (!ip6_forwarding)
1453 		goto fail;
1454 
1455 	/* sanity check */
1456 	if (m0 == NULL || !rtisvalid(rt))
1457 		goto fail;
1458 
1459 	ifp = if_get(rt->rt_ifidx);
1460 	if (ifp == NULL)
1461 		goto fail;
1462 
1463 	/*
1464 	 * Address check:
1465 	 *  the source address must identify a neighbor, and
1466 	 *  the destination address must not be a multicast address
1467 	 *  [RFC 2461, sec 8.2]
1468 	 */
1469 	sip6 = mtod(m0, struct ip6_hdr *);
1470 	bzero(&src_sa, sizeof(src_sa));
1471 	src_sa.sin6_family = AF_INET6;
1472 	src_sa.sin6_len = sizeof(src_sa);
1473 	src_sa.sin6_addr = sip6->ip6_src;
1474 	/* we don't currently use sin6_scope_id, but eventually use it */
1475 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp->if_index, &sip6->ip6_src);
1476 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
1477 		goto fail;
1478 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
1479 		goto fail;	/* what should we do here? */
1480 
1481 	/* rate limit */
1482 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
1483 		goto fail;
1484 
1485 	/*
1486 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
1487 	 * we almost always ask for an mbuf cluster for simplicity.
1488 	 * (MHLEN < IPV6_MMTU is almost always true)
1489 	 */
1490 #if IPV6_MMTU >= MCLBYTES
1491 # error assumption failed about IPV6_MMTU and MCLBYTES
1492 #endif
1493 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
1494 	if (m && IPV6_MMTU >= MHLEN)
1495 		MCLGET(m, M_DONTWAIT);
1496 	if (!m)
1497 		goto fail;
1498 	m->m_pkthdr.ph_ifidx = 0;
1499 	m->m_len = 0;
1500 	maxlen = m_trailingspace(m);
1501 	maxlen = min(IPV6_MMTU, maxlen);
1502 	/* just for safety */
1503 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
1504 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
1505 		goto fail;
1506 	}
1507 
1508 	{
1509 		/* get ip6 linklocal address for ifp(my outgoing interface). */
1510 		struct in6_ifaddr *ia6;
1511 		if ((ia6 = in6ifa_ifpforlinklocal(ifp, IN6_IFF_TENTATIVE|
1512 		    IN6_IFF_DUPLICATED|IN6_IFF_ANYCAST)) == NULL)
1513 			goto fail;
1514 		ifp_ll6 = &ia6->ia_addr.sin6_addr;
1515 	}
1516 
1517 	/* get ip6 linklocal address for the router. */
1518 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
1519 		struct sockaddr_in6 *sin6;
1520 		sin6 = satosin6(rt->rt_gateway);
1521 		nexthop = &sin6->sin6_addr;
1522 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
1523 			nexthop = NULL;
1524 	} else
1525 		nexthop = NULL;
1526 
1527 	/* ip6 */
1528 	ip6 = mtod(m, struct ip6_hdr *);
1529 	ip6->ip6_flow = 0;
1530 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1531 	ip6->ip6_vfc |= IPV6_VERSION;
1532 	/* ip6->ip6_plen will be set later */
1533 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1534 	ip6->ip6_hlim = 255;
1535 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
1536 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
1537 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
1538 
1539 	/* ND Redirect */
1540 	nd_rd = (struct nd_redirect *)(ip6 + 1);
1541 	nd_rd->nd_rd_type = ND_REDIRECT;
1542 	nd_rd->nd_rd_code = 0;
1543 	nd_rd->nd_rd_reserved = 0;
1544 	if (rt->rt_flags & RTF_GATEWAY) {
1545 		/*
1546 		 * nd_rd->nd_rd_target must be a link-local address in
1547 		 * better router cases.
1548 		 */
1549 		if (!nexthop)
1550 			goto fail;
1551 		bcopy(nexthop, &nd_rd->nd_rd_target,
1552 		      sizeof(nd_rd->nd_rd_target));
1553 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1554 		      sizeof(nd_rd->nd_rd_dst));
1555 	} else {
1556 		/* make sure redtgt == reddst */
1557 		nexthop = &sip6->ip6_dst;
1558 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
1559 		      sizeof(nd_rd->nd_rd_target));
1560 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
1561 		      sizeof(nd_rd->nd_rd_dst));
1562 	}
1563 
1564 	p = (u_char *)(nd_rd + 1);
1565 
1566 	{
1567 		/* target lladdr option */
1568 		struct rtentry *nrt;
1569 		int len;
1570 		struct sockaddr_dl *sdl;
1571 		struct nd_opt_hdr *nd_opt;
1572 		char *lladdr;
1573 
1574 		len = sizeof(*nd_opt) + ifp->if_addrlen;
1575 		len = (len + 7) & ~7;	/* round by 8 */
1576 		/* safety check */
1577 		if (len + (p - (u_char *)ip6) > maxlen)
1578 			goto nolladdropt;
1579 		nrt = nd6_lookup(nexthop, 0, ifp, ifp->if_rdomain);
1580 		if ((nrt != NULL) &&
1581 		    (nrt->rt_flags & (RTF_GATEWAY|RTF_LLINFO)) == RTF_LLINFO &&
1582 		    (nrt->rt_gateway->sa_family == AF_LINK) &&
1583 		    (sdl = satosdl(nrt->rt_gateway)) &&
1584 		    sdl->sdl_alen) {
1585 			nd_opt = (struct nd_opt_hdr *)p;
1586 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1587 			nd_opt->nd_opt_len = len >> 3;
1588 			lladdr = (char *)(nd_opt + 1);
1589 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
1590 			p += len;
1591 		}
1592 		rtfree(nrt);
1593 	}
1594   nolladdropt:;
1595 
1596 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1597 
1598 	/* just to be safe */
1599 	if (p - (u_char *)ip6 > maxlen)
1600 		goto noredhdropt;
1601 
1602 	{
1603 		/* redirected header option */
1604 		int len;
1605 		struct nd_opt_rd_hdr *nd_opt_rh;
1606 
1607 		/*
1608 		 * compute the maximum size for icmp6 redirect header option.
1609 		 * XXX room for auth header?
1610 		 */
1611 		len = maxlen - (p - (u_char *)ip6);
1612 		len &= ~7;
1613 
1614 		/*
1615 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
1616 		 * about padding/truncate rule for the original IP packet.
1617 		 * From the discussion on IPv6imp in Feb 1999,
1618 		 * the consensus was:
1619 		 * - "attach as much as possible" is the goal
1620 		 * - pad if not aligned (original size can be guessed by
1621 		 *   original ip6 header)
1622 		 * Following code adds the padding if it is simple enough,
1623 		 * and truncates if not.
1624 		 */
1625 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
1626 			/* not enough room, truncate */
1627 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
1628 			    m0->m_pkthdr.len);
1629 		} else {
1630 			/*
1631 			 * enough room, truncate if not aligned.
1632 			 * we don't pad here for simplicity.
1633 			 */
1634 			size_t extra;
1635 
1636 			extra = m0->m_pkthdr.len % 8;
1637 			if (extra) {
1638 				/* truncate */
1639 				m_adj(m0, -extra);
1640 			}
1641 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
1642 		}
1643 
1644 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
1645 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
1646 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
1647 		nd_opt_rh->nd_opt_rh_len = len >> 3;
1648 		p += sizeof(*nd_opt_rh);
1649 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
1650 
1651 		/* connect m0 to m */
1652 		m->m_pkthdr.len += m0->m_pkthdr.len;
1653 		m_cat(m, m0);
1654 		m0 = NULL;
1655 	}
1656 noredhdropt:
1657 	m_freem(m0);
1658 	m0 = NULL;
1659 
1660 	sip6 = mtod(m, struct ip6_hdr *);
1661 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
1662 		sip6->ip6_src.s6_addr16[1] = 0;
1663 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
1664 		sip6->ip6_dst.s6_addr16[1] = 0;
1665 #if 0
1666 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1667 		ip6->ip6_src.s6_addr16[1] = 0;
1668 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1669 		ip6->ip6_dst.s6_addr16[1] = 0;
1670 #endif
1671 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
1672 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
1673 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
1674 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
1675 
1676 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
1677 
1678 	nd_rd->nd_rd_cksum = 0;
1679 	m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT;
1680 
1681 	/* send the packet to outside... */
1682 	ip6_output(m, NULL, NULL, 0, NULL, NULL);
1683 
1684 	icmp6stat_inc(icp6s_outhist + ND_REDIRECT);
1685 
1686 	if_put(ifp);
1687 	return;
1688 
1689 fail:
1690 	if_put(ifp);
1691 	m_freem(m);
1692 	m_freem(m0);
1693 }
1694 
1695 /*
1696  * ICMPv6 socket option processing.
1697  */
1698 int
1699 icmp6_ctloutput(int op, struct socket *so, int level, int optname,
1700     struct mbuf *m)
1701 {
1702 	int error = 0;
1703 	struct inpcb *in6p = sotoinpcb(so);
1704 
1705 	if (level != IPPROTO_ICMPV6)
1706 		return EINVAL;
1707 
1708 	switch (op) {
1709 	case PRCO_SETOPT:
1710 		switch (optname) {
1711 		case ICMP6_FILTER:
1712 		    {
1713 			struct icmp6_filter *p;
1714 
1715 			if (m == NULL || m->m_len != sizeof(*p)) {
1716 				error = EMSGSIZE;
1717 				break;
1718 			}
1719 			p = mtod(m, struct icmp6_filter *);
1720 			if (!p || !in6p->inp_icmp6filt) {
1721 				error = EINVAL;
1722 				break;
1723 			}
1724 			bcopy(p, in6p->inp_icmp6filt,
1725 				sizeof(struct icmp6_filter));
1726 			error = 0;
1727 			break;
1728 		    }
1729 
1730 		default:
1731 			error = ENOPROTOOPT;
1732 			break;
1733 		}
1734 		break;
1735 
1736 	case PRCO_GETOPT:
1737 		switch (optname) {
1738 		case ICMP6_FILTER:
1739 		    {
1740 			struct icmp6_filter *p;
1741 
1742 			if (!in6p->inp_icmp6filt) {
1743 				error = EINVAL;
1744 				break;
1745 			}
1746 			m->m_len = sizeof(struct icmp6_filter);
1747 			p = mtod(m, struct icmp6_filter *);
1748 			bcopy(in6p->inp_icmp6filt, p,
1749 				sizeof(struct icmp6_filter));
1750 			error = 0;
1751 			break;
1752 		    }
1753 
1754 		default:
1755 			error = ENOPROTOOPT;
1756 			break;
1757 		}
1758 		break;
1759 	}
1760 
1761 	return (error);
1762 }
1763 
1764 /*
1765  * Perform rate limit check.
1766  * Returns 0 if it is okay to send the icmp6 packet.
1767  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
1768  * limitation.
1769  *
1770  * XXX per-destination/type check necessary?
1771  *
1772  * dst - not used at this moment
1773  * type - not used at this moment
1774  * code - not used at this moment
1775  */
1776 int
1777 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
1778 {
1779 	/* PPS limit */
1780 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
1781 	    icmp6errppslim))
1782 		return 1;	/* The packet is subject to rate limit */
1783 	return 0;		/* okay to send */
1784 }
1785 
1786 struct rtentry *
1787 icmp6_mtudisc_clone(struct sockaddr_in6 *dst, u_int rtableid, int ipsec)
1788 {
1789 	struct rtentry *rt;
1790 	int    error;
1791 
1792 	rt = rtalloc(sin6tosa(dst), RT_RESOLVE, rtableid);
1793 
1794 	/* Check if the route is actually usable */
1795 	if (!rtisvalid(rt))
1796 		goto bad;
1797 	/* IPsec needs the route only for PMTU, it can use reject for that */
1798 	if (!ipsec && (rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)))
1799 		goto bad;
1800 
1801 	/*
1802 	 * No PMTU for local routes and permanent neighbors,
1803 	 * ARP and NDP use the same expire timer as the route.
1804 	 */
1805 	if (ISSET(rt->rt_flags, RTF_LOCAL) ||
1806 	    (ISSET(rt->rt_flags, RTF_LLINFO) && rt->rt_expire == 0))
1807 		goto bad;
1808 
1809 	/* If we didn't get a host route, allocate one */
1810 	if ((rt->rt_flags & RTF_HOST) == 0) {
1811 		struct rtentry *nrt;
1812 		struct rt_addrinfo info;
1813 		struct sockaddr_rtlabel sa_rl;
1814 
1815 		memset(&info, 0, sizeof(info));
1816 		info.rti_ifa = rt->rt_ifa;
1817 		info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC;
1818 		info.rti_info[RTAX_DST] = sin6tosa(dst);
1819 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1820 		info.rti_info[RTAX_LABEL] =
1821 		    rtlabel_id2sa(rt->rt_labelid, &sa_rl);
1822 
1823 		error = rtrequest(RTM_ADD, &info, rt->rt_priority, &nrt,
1824 		    rtableid);
1825 		if (error)
1826 			goto bad;
1827 		nrt->rt_rmx = rt->rt_rmx;
1828 		rtfree(rt);
1829 		rt = nrt;
1830 		rtm_send(rt, RTM_ADD, 0, rtableid);
1831 	}
1832 	error = rt_timer_add(rt, icmp6_mtudisc_timeout, icmp6_mtudisc_timeout_q,
1833 	    rtableid);
1834 	if (error)
1835 		goto bad;
1836 
1837 	return (rt);
1838 bad:
1839 	rtfree(rt);
1840 	return (NULL);
1841 }
1842 
1843 void
1844 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1845 {
1846 	struct ifnet *ifp;
1847 
1848 	NET_ASSERT_LOCKED();
1849 
1850 	ifp = if_get(rt->rt_ifidx);
1851 	if (ifp == NULL)
1852 		return;
1853 
1854 	if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) {
1855 		rtdeletemsg(rt, ifp, r->rtt_tableid);
1856 	} else {
1857 		if (!(rt->rt_locks & RTV_MTU))
1858 			rt->rt_mtu = 0;
1859 	}
1860 
1861 	if_put(ifp);
1862 }
1863 
1864 void
1865 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1866 {
1867 	struct ifnet *ifp;
1868 
1869 	NET_ASSERT_LOCKED();
1870 
1871 	ifp = if_get(rt->rt_ifidx);
1872 	if (ifp == NULL)
1873 		return;
1874 
1875 	if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) {
1876 		rtdeletemsg(rt, ifp, r->rtt_tableid);
1877 	}
1878 
1879 	if_put(ifp);
1880 }
1881 
1882 const struct sysctl_bounded_args icmpv6ctl_vars[] = {
1883 	{ ICMPV6CTL_ND6_DELAY, &nd6_delay, 0, INT_MAX },
1884 	{ ICMPV6CTL_ND6_UMAXTRIES, &nd6_umaxtries, 0, INT_MAX },
1885 	{ ICMPV6CTL_ND6_MMAXTRIES, &nd6_mmaxtries, 0, INT_MAX },
1886 	{ ICMPV6CTL_ERRPPSLIMIT, &icmp6errppslim, -1, 1000 },
1887 	{ ICMPV6CTL_ND6_MAXNUDHINT, &nd6_maxnudhint, 0, INT_MAX },
1888 	{ ICMPV6CTL_MTUDISC_HIWAT, &icmp6_mtudisc_hiwat, -1, INT_MAX },
1889 	{ ICMPV6CTL_MTUDISC_LOWAT, &icmp6_mtudisc_lowat, -1, INT_MAX },
1890 	{ ICMPV6CTL_ND6_DEBUG, &nd6_debug, 0, 1 },
1891 };
1892 
1893 int
1894 icmp6_sysctl_icmp6stat(void *oldp, size_t *oldlenp, void *newp)
1895 {
1896 	struct icmp6stat *icmp6stat;
1897 	int ret;
1898 
1899 	CTASSERT(sizeof(*icmp6stat) == icp6s_ncounters * sizeof(uint64_t));
1900 	icmp6stat = malloc(sizeof(*icmp6stat), M_TEMP, M_WAITOK|M_ZERO);
1901 	counters_read(icmp6counters, (uint64_t *)icmp6stat, icp6s_ncounters);
1902 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1903 	    icmp6stat, sizeof(*icmp6stat));
1904 	free(icmp6stat, M_TEMP, sizeof(*icmp6stat));
1905 
1906 	return (ret);
1907 }
1908 
1909 int
1910 icmp6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1911     void *newp, size_t newlen)
1912 {
1913 	int error;
1914 
1915 	/* All sysctl names at this level are terminal. */
1916 	if (namelen != 1)
1917 		return (ENOTDIR);
1918 
1919 	switch (name[0]) {
1920 	case ICMPV6CTL_REDIRTIMEOUT:
1921 		NET_LOCK();
1922 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1923 		    &icmp6_redirtimeout, 0, INT_MAX);
1924 		rt_timer_queue_change(icmp6_redirect_timeout_q,
1925 		    icmp6_redirtimeout);
1926 		NET_UNLOCK();
1927 		break;
1928 
1929 	case ICMPV6CTL_STATS:
1930 		error = icmp6_sysctl_icmp6stat(oldp, oldlenp, newp);
1931 		break;
1932 
1933 	default:
1934 		NET_LOCK();
1935 		error = sysctl_bounded_arr(icmpv6ctl_vars,
1936 		    nitems(icmpv6ctl_vars), name, namelen, oldp, oldlenp, newp,
1937 		    newlen);
1938 		NET_UNLOCK();
1939 		break;
1940 	}
1941 
1942 	return (error);
1943 }
1944