xref: /freebsd/sys/netinet6/icmp6.c (revision a743684e600e6dbe68f9218219686402bdea07e0)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/domain.h>
72 #include <sys/jail.h>
73 #include <sys/kernel.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86 
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/icmp6.h>
99 #include <netinet/tcp_var.h>
100 
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/ip6protosw.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/nd6.h>
108 
109 #ifdef IPSEC
110 #include <netipsec/ipsec.h>
111 #include <netipsec/key.h>
112 #endif
113 
114 extern struct domain inet6domain;
115 
116 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
117 VNET_DECLARE(struct inpcbhead, ripcb);
118 VNET_DECLARE(int, icmp6errppslim);
119 VNET_DECLARE(int, icmp6_nodeinfo);
120 
121 #define	V_ripcbinfo			VNET(ripcbinfo)
122 #define	V_ripcb				VNET(ripcb)
123 #define	V_icmp6errppslim		VNET(icmp6errppslim)
124 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
125 
126 VNET_DEFINE(struct icmp6stat, icmp6stat);
127 static VNET_DEFINE(int, icmp6errpps_count);
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 
130 #define	V_icmp6errpps_count	VNET(icmp6errpps_count)
131 #define	V_icmp6errppslim_last	VNET(icmp6errppslim_last)
132 
133 static void icmp6_errcount(struct icmp6errstat *, int, int);
134 static int icmp6_rip6_input(struct mbuf **, int);
135 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
136 static const char *icmp6_redirect_diag __P((struct in6_addr *,
137 	struct in6_addr *, struct in6_addr *));
138 static struct mbuf *ni6_input(struct mbuf *, int);
139 static struct mbuf *ni6_nametodns(const char *, int, int);
140 static int ni6_dnsmatch(const char *, int, const char *, int);
141 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
142 			  struct ifnet **, struct in6_addr *));
143 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
144 				struct ifnet *, int));
145 static int icmp6_notify_error(struct mbuf **, int, int, int);
146 
147 
148 void
149 icmp6_init(void)
150 {
151 
152 	V_icmp6errpps_count = 0;
153 }
154 
155 static void
156 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
157 {
158 	switch (type) {
159 	case ICMP6_DST_UNREACH:
160 		switch (code) {
161 		case ICMP6_DST_UNREACH_NOROUTE:
162 			stat->icp6errs_dst_unreach_noroute++;
163 			return;
164 		case ICMP6_DST_UNREACH_ADMIN:
165 			stat->icp6errs_dst_unreach_admin++;
166 			return;
167 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
168 			stat->icp6errs_dst_unreach_beyondscope++;
169 			return;
170 		case ICMP6_DST_UNREACH_ADDR:
171 			stat->icp6errs_dst_unreach_addr++;
172 			return;
173 		case ICMP6_DST_UNREACH_NOPORT:
174 			stat->icp6errs_dst_unreach_noport++;
175 			return;
176 		}
177 		break;
178 	case ICMP6_PACKET_TOO_BIG:
179 		stat->icp6errs_packet_too_big++;
180 		return;
181 	case ICMP6_TIME_EXCEEDED:
182 		switch (code) {
183 		case ICMP6_TIME_EXCEED_TRANSIT:
184 			stat->icp6errs_time_exceed_transit++;
185 			return;
186 		case ICMP6_TIME_EXCEED_REASSEMBLY:
187 			stat->icp6errs_time_exceed_reassembly++;
188 			return;
189 		}
190 		break;
191 	case ICMP6_PARAM_PROB:
192 		switch (code) {
193 		case ICMP6_PARAMPROB_HEADER:
194 			stat->icp6errs_paramprob_header++;
195 			return;
196 		case ICMP6_PARAMPROB_NEXTHEADER:
197 			stat->icp6errs_paramprob_nextheader++;
198 			return;
199 		case ICMP6_PARAMPROB_OPTION:
200 			stat->icp6errs_paramprob_option++;
201 			return;
202 		}
203 		break;
204 	case ND_REDIRECT:
205 		stat->icp6errs_redirect++;
206 		return;
207 	}
208 	stat->icp6errs_unknown++;
209 }
210 
211 /*
212  * A wrapper function for icmp6_error() necessary when the erroneous packet
213  * may not contain enough scope zone information.
214  */
215 void
216 icmp6_error2(struct mbuf *m, int type, int code, int param,
217     struct ifnet *ifp)
218 {
219 	struct ip6_hdr *ip6;
220 
221 	if (ifp == NULL)
222 		return;
223 
224 #ifndef PULLDOWN_TEST
225 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
226 #else
227 	if (m->m_len < sizeof(struct ip6_hdr)) {
228 		m = m_pullup(m, sizeof(struct ip6_hdr));
229 		if (m == NULL)
230 			return;
231 	}
232 #endif
233 
234 	ip6 = mtod(m, struct ip6_hdr *);
235 
236 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
237 		return;
238 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
239 		return;
240 
241 	icmp6_error(m, type, code, param);
242 }
243 
244 /*
245  * Generate an error packet of type error in response to bad IP6 packet.
246  */
247 void
248 icmp6_error(struct mbuf *m, int type, int code, int param)
249 {
250 	struct ip6_hdr *oip6, *nip6;
251 	struct icmp6_hdr *icmp6;
252 	u_int preplen;
253 	int off;
254 	int nxt;
255 
256 	ICMP6STAT_INC(icp6s_error);
257 
258 	/* count per-type-code statistics */
259 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code);
260 
261 #ifdef M_DECRYPTED	/*not openbsd*/
262 	if (m->m_flags & M_DECRYPTED) {
263 		ICMP6STAT_INC(icp6s_canterror);
264 		goto freeit;
265 	}
266 #endif
267 
268 #ifndef PULLDOWN_TEST
269 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
270 #else
271 	if (m->m_len < sizeof(struct ip6_hdr)) {
272 		m = m_pullup(m, sizeof(struct ip6_hdr));
273 		if (m == NULL)
274 			return;
275 	}
276 #endif
277 	oip6 = mtod(m, struct ip6_hdr *);
278 
279 	/*
280 	 * If the destination address of the erroneous packet is a multicast
281 	 * address, or the packet was sent using link-layer multicast,
282 	 * we should basically suppress sending an error (RFC 2463, Section
283 	 * 2.4).
284 	 * We have two exceptions (the item e.2 in that section):
285 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
286 	 * - the Parameter Problem Message that can be allowed an icmp6 error
287 	 *   in the option type field.  This check has been done in
288 	 *   ip6_unknown_opt(), so we can just check the type and code.
289 	 */
290 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
291 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
292 	    (type != ICMP6_PACKET_TOO_BIG &&
293 	     (type != ICMP6_PARAM_PROB ||
294 	      code != ICMP6_PARAMPROB_OPTION)))
295 		goto freeit;
296 
297 	/*
298 	 * RFC 2463, 2.4 (e.5): source address check.
299 	 * XXX: the case of anycast source?
300 	 */
301 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
302 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
303 		goto freeit;
304 
305 	/*
306 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
307 	 * don't do it.
308 	 */
309 	nxt = -1;
310 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
311 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
312 		struct icmp6_hdr *icp;
313 
314 #ifndef PULLDOWN_TEST
315 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
316 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
317 #else
318 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
319 			sizeof(*icp));
320 		if (icp == NULL) {
321 			ICMP6STAT_INC(icp6s_tooshort);
322 			return;
323 		}
324 #endif
325 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
326 		    icp->icmp6_type == ND_REDIRECT) {
327 			/*
328 			 * ICMPv6 error
329 			 * Special case: for redirect (which is
330 			 * informational) we must not send icmp6 error.
331 			 */
332 			ICMP6STAT_INC(icp6s_canterror);
333 			goto freeit;
334 		} else {
335 			/* ICMPv6 informational - send the error */
336 		}
337 	} else {
338 		/* non-ICMPv6 - send the error */
339 	}
340 
341 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
342 
343 	/* Finally, do rate limitation check. */
344 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
345 		ICMP6STAT_INC(icp6s_toofreq);
346 		goto freeit;
347 	}
348 
349 	/*
350 	 * OK, ICMP6 can be generated.
351 	 */
352 
353 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
354 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
355 
356 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
357 	M_PREPEND(m, preplen, M_DONTWAIT);
358 	if (m && m->m_len < preplen)
359 		m = m_pullup(m, preplen);
360 	if (m == NULL) {
361 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
362 		return;
363 	}
364 
365 	nip6 = mtod(m, struct ip6_hdr *);
366 	nip6->ip6_src  = oip6->ip6_src;
367 	nip6->ip6_dst  = oip6->ip6_dst;
368 
369 	in6_clearscope(&oip6->ip6_src);
370 	in6_clearscope(&oip6->ip6_dst);
371 
372 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
373 	icmp6->icmp6_type = type;
374 	icmp6->icmp6_code = code;
375 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
376 
377 	/*
378 	 * icmp6_reflect() is designed to be in the input path.
379 	 * icmp6_error() can be called from both input and output path,
380 	 * and if we are in output path rcvif could contain bogus value.
381 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
382 	 * information in ip header (nip6).
383 	 */
384 	m->m_pkthdr.rcvif = NULL;
385 
386 	ICMP6STAT_INC(icp6s_outhist[type]);
387 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
388 
389 	return;
390 
391   freeit:
392 	/*
393 	 * If we can't tell whether or not we can generate ICMP6, free it.
394 	 */
395 	m_freem(m);
396 }
397 
398 /*
399  * Process a received ICMP6 message.
400  */
401 int
402 icmp6_input(struct mbuf **mp, int *offp, int proto)
403 {
404 	struct mbuf *m = *mp, *n;
405 	struct ifnet *ifp;
406 	struct ip6_hdr *ip6, *nip6;
407 	struct icmp6_hdr *icmp6, *nicmp6;
408 	int off = *offp;
409 	int icmp6len = m->m_pkthdr.len - *offp;
410 	int code, sum, noff;
411 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
412 
413 	ifp = m->m_pkthdr.rcvif;
414 
415 #ifndef PULLDOWN_TEST
416 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
417 	/* m might change if M_LOOP.  So, call mtod after this */
418 #endif
419 
420 	/*
421 	 * Locate icmp6 structure in mbuf, and check
422 	 * that not corrupted and of at least minimum length
423 	 */
424 
425 	ip6 = mtod(m, struct ip6_hdr *);
426 	if (icmp6len < sizeof(struct icmp6_hdr)) {
427 		ICMP6STAT_INC(icp6s_tooshort);
428 		goto freeit;
429 	}
430 
431 	/*
432 	 * Check multicast group membership.
433 	 * Note: SSM filters are not applied for ICMPv6 traffic.
434 	 */
435 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
436 		struct in6_multi	*inm;
437 
438 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
439 		if (inm == NULL) {
440 			IP6STAT_INC(ip6s_notmember);
441 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
442 			goto freeit;
443 		}
444 	}
445 
446 	/*
447 	 * calculate the checksum
448 	 */
449 #ifndef PULLDOWN_TEST
450 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
451 #else
452 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
453 	if (icmp6 == NULL) {
454 		ICMP6STAT_INC(icp6s_tooshort);
455 		return IPPROTO_DONE;
456 	}
457 #endif
458 	code = icmp6->icmp6_code;
459 
460 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
461 		nd6log((LOG_ERR,
462 		    "ICMP6 checksum error(%d|%x) %s\n",
463 		    icmp6->icmp6_type, sum,
464 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
465 		ICMP6STAT_INC(icp6s_checksum);
466 		goto freeit;
467 	}
468 
469 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
470 		/*
471 		 * Deliver very specific ICMP6 type only.
472 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
473 		 * will not work.
474 		 */
475 		switch (icmp6->icmp6_type) {
476 		case ICMP6_DST_UNREACH:
477 		case ICMP6_PACKET_TOO_BIG:
478 		case ICMP6_TIME_EXCEEDED:
479 			break;
480 		default:
481 			goto freeit;
482 		}
483 	}
484 
485 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
486 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
487 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
488 		icmp6_ifstat_inc(ifp, ifs6_in_error);
489 
490 	switch (icmp6->icmp6_type) {
491 	case ICMP6_DST_UNREACH:
492 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
493 		switch (code) {
494 		case ICMP6_DST_UNREACH_NOROUTE:
495 			code = PRC_UNREACH_NET;
496 			break;
497 		case ICMP6_DST_UNREACH_ADMIN:
498 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
499 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
500 			break;
501 		case ICMP6_DST_UNREACH_ADDR:
502 			code = PRC_HOSTDEAD;
503 			break;
504 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
505 			/* I mean "source address was incorrect." */
506 			code = PRC_PARAMPROB;
507 			break;
508 		case ICMP6_DST_UNREACH_NOPORT:
509 			code = PRC_UNREACH_PORT;
510 			break;
511 		default:
512 			goto badcode;
513 		}
514 		goto deliver;
515 		break;
516 
517 	case ICMP6_PACKET_TOO_BIG:
518 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
519 
520 		/* validation is made in icmp6_mtudisc_update */
521 
522 		code = PRC_MSGSIZE;
523 
524 		/*
525 		 * Updating the path MTU will be done after examining
526 		 * intermediate extension headers.
527 		 */
528 		goto deliver;
529 		break;
530 
531 	case ICMP6_TIME_EXCEEDED:
532 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
533 		switch (code) {
534 		case ICMP6_TIME_EXCEED_TRANSIT:
535 			code = PRC_TIMXCEED_INTRANS;
536 			break;
537 		case ICMP6_TIME_EXCEED_REASSEMBLY:
538 			code = PRC_TIMXCEED_REASS;
539 			break;
540 		default:
541 			goto badcode;
542 		}
543 		goto deliver;
544 		break;
545 
546 	case ICMP6_PARAM_PROB:
547 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
548 		switch (code) {
549 		case ICMP6_PARAMPROB_NEXTHEADER:
550 			code = PRC_UNREACH_PROTOCOL;
551 			break;
552 		case ICMP6_PARAMPROB_HEADER:
553 		case ICMP6_PARAMPROB_OPTION:
554 			code = PRC_PARAMPROB;
555 			break;
556 		default:
557 			goto badcode;
558 		}
559 		goto deliver;
560 		break;
561 
562 	case ICMP6_ECHO_REQUEST:
563 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
564 		if (code != 0)
565 			goto badcode;
566 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
567 			/* Give up remote */
568 			break;
569 		}
570 		if ((n->m_flags & M_EXT) != 0
571 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
572 			struct mbuf *n0 = n;
573 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
574 			int n0len;
575 
576 			MGETHDR(n, M_DONTWAIT, n0->m_type);
577 			n0len = n0->m_pkthdr.len;	/* save for use below */
578 			if (n)
579 				M_MOVE_PKTHDR(n, n0);
580 			if (n && maxlen >= MHLEN) {
581 				MCLGET(n, M_DONTWAIT);
582 				if ((n->m_flags & M_EXT) == 0) {
583 					m_free(n);
584 					n = NULL;
585 				}
586 			}
587 			if (n == NULL) {
588 				/* Give up remote */
589 				m_freem(n0);
590 				break;
591 			}
592 			/*
593 			 * Copy IPv6 and ICMPv6 only.
594 			 */
595 			nip6 = mtod(n, struct ip6_hdr *);
596 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
597 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
598 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
599 			noff = sizeof(struct ip6_hdr);
600 			/* new mbuf contains only ipv6+icmpv6 headers */
601 			n->m_len = noff + sizeof(struct icmp6_hdr);
602 			/*
603 			 * Adjust mbuf.  ip6_plen will be adjusted in
604 			 * ip6_output().
605 			 */
606 			m_adj(n0, off + sizeof(struct icmp6_hdr));
607 			/* recalculate complete packet size */
608 			n->m_pkthdr.len = n0len + (noff - off);
609 			n->m_next = n0;
610 		} else {
611 			nip6 = mtod(n, struct ip6_hdr *);
612 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
613 			    sizeof(*nicmp6));
614 			noff = off;
615 		}
616 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
617 		nicmp6->icmp6_code = 0;
618 		if (n) {
619 			ICMP6STAT_INC(icp6s_reflect);
620 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
621 			icmp6_reflect(n, noff);
622 		}
623 		break;
624 
625 	case ICMP6_ECHO_REPLY:
626 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
627 		if (code != 0)
628 			goto badcode;
629 		break;
630 
631 	case MLD_LISTENER_QUERY:
632 	case MLD_LISTENER_REPORT:
633 	case MLD_LISTENER_DONE:
634 	case MLDV2_LISTENER_REPORT:
635 		/*
636 		 * Drop MLD traffic which is not link-local, has a hop limit
637 		 * of greater than 1 hop, or which does not have the
638 		 * IPv6 HBH Router Alert option.
639 		 * As IPv6 HBH options are stripped in ip6_input() we must
640 		 * check an mbuf header flag.
641 		 * XXX Should we also sanity check that these messages
642 		 * were directed to a link-local multicast prefix?
643 		 */
644 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
645 			goto freeit;
646 		if (mld_input(m, off, icmp6len) != 0)
647 			return (IPPROTO_DONE);
648 		/* m stays. */
649 		break;
650 
651 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
652 	    {
653 		enum { WRU, FQDN } mode;
654 
655 		if (!V_icmp6_nodeinfo)
656 			break;
657 
658 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
659 			mode = WRU;
660 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
661 			mode = FQDN;
662 		else
663 			goto badlen;
664 
665 		if (mode == FQDN) {
666 #ifndef PULLDOWN_TEST
667 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
668 			    IPPROTO_DONE);
669 #endif
670 			n = m_copy(m, 0, M_COPYALL);
671 			if (n)
672 				n = ni6_input(n, off);
673 			/* XXX meaningless if n == NULL */
674 			noff = sizeof(struct ip6_hdr);
675 		} else {
676 			struct prison *pr;
677 			u_char *p;
678 			int maxlen, maxhlen, hlen;
679 
680 			/*
681 			 * XXX: this combination of flags is pointless,
682 			 * but should we keep this for compatibility?
683 			 */
684 			if ((V_icmp6_nodeinfo & 5) != 5)
685 				break;
686 
687 			if (code != 0)
688 				goto badcode;
689 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
690 			if (maxlen >= MCLBYTES) {
691 				/* Give up remote */
692 				break;
693 			}
694 			MGETHDR(n, M_DONTWAIT, m->m_type);
695 			if (n && maxlen > MHLEN) {
696 				MCLGET(n, M_DONTWAIT);
697 				if ((n->m_flags & M_EXT) == 0) {
698 					m_free(n);
699 					n = NULL;
700 				}
701 			}
702 			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
703 				/*
704 				 * Previous code did a blind M_COPY_PKTHDR
705 				 * and said "just for rcvif".  If true, then
706 				 * we could tolerate the dup failing (due to
707 				 * the deep copy of the tag chain).  For now
708 				 * be conservative and just fail.
709 				 */
710 				m_free(n);
711 				n = NULL;
712 			}
713 			if (n == NULL) {
714 				/* Give up remote */
715 				break;
716 			}
717 			n->m_pkthdr.rcvif = NULL;
718 			n->m_len = 0;
719 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
720 			pr = curthread->td_ucred->cr_prison;
721 			mtx_lock(&pr->pr_mtx);
722 			hlen = strlen(pr->pr_hostname);
723 			if (maxhlen > hlen)
724 				maxhlen = hlen;
725 			/*
726 			 * Copy IPv6 and ICMPv6 only.
727 			 */
728 			nip6 = mtod(n, struct ip6_hdr *);
729 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
730 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
731 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
732 			p = (u_char *)(nicmp6 + 1);
733 			bzero(p, 4);
734 			/* meaningless TTL */
735 			bcopy(pr->pr_hostname, p + 4, maxhlen);
736 			mtx_unlock(&pr->pr_mtx);
737 			noff = sizeof(struct ip6_hdr);
738 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
739 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
740 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
741 			nicmp6->icmp6_code = 0;
742 		}
743 		if (n) {
744 			ICMP6STAT_INC(icp6s_reflect);
745 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
746 			icmp6_reflect(n, noff);
747 		}
748 		break;
749 	    }
750 
751 	case ICMP6_WRUREPLY:
752 		if (code != 0)
753 			goto badcode;
754 		break;
755 
756 	case ND_ROUTER_SOLICIT:
757 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
758 		if (code != 0)
759 			goto badcode;
760 		if (icmp6len < sizeof(struct nd_router_solicit))
761 			goto badlen;
762 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
763 			/* give up local */
764 			nd6_rs_input(m, off, icmp6len);
765 			m = NULL;
766 			goto freeit;
767 		}
768 		nd6_rs_input(n, off, icmp6len);
769 		/* m stays. */
770 		break;
771 
772 	case ND_ROUTER_ADVERT:
773 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
774 		if (code != 0)
775 			goto badcode;
776 		if (icmp6len < sizeof(struct nd_router_advert))
777 			goto badlen;
778 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
779 			/* give up local */
780 			nd6_ra_input(m, off, icmp6len);
781 			m = NULL;
782 			goto freeit;
783 		}
784 		nd6_ra_input(n, off, icmp6len);
785 		/* m stays. */
786 		break;
787 
788 	case ND_NEIGHBOR_SOLICIT:
789 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
790 		if (code != 0)
791 			goto badcode;
792 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
793 			goto badlen;
794 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
795 			/* give up local */
796 			nd6_ns_input(m, off, icmp6len);
797 			m = NULL;
798 			goto freeit;
799 		}
800 		nd6_ns_input(n, off, icmp6len);
801 		/* m stays. */
802 		break;
803 
804 	case ND_NEIGHBOR_ADVERT:
805 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
806 		if (code != 0)
807 			goto badcode;
808 		if (icmp6len < sizeof(struct nd_neighbor_advert))
809 			goto badlen;
810 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
811 			/* give up local */
812 			nd6_na_input(m, off, icmp6len);
813 			m = NULL;
814 			goto freeit;
815 		}
816 		nd6_na_input(n, off, icmp6len);
817 		/* m stays. */
818 		break;
819 
820 	case ND_REDIRECT:
821 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
822 		if (code != 0)
823 			goto badcode;
824 		if (icmp6len < sizeof(struct nd_redirect))
825 			goto badlen;
826 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
827 			/* give up local */
828 			icmp6_redirect_input(m, off);
829 			m = NULL;
830 			goto freeit;
831 		}
832 		icmp6_redirect_input(n, off);
833 		/* m stays. */
834 		break;
835 
836 	case ICMP6_ROUTER_RENUMBERING:
837 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
838 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
839 			goto badcode;
840 		if (icmp6len < sizeof(struct icmp6_router_renum))
841 			goto badlen;
842 		break;
843 
844 	default:
845 		nd6log((LOG_DEBUG,
846 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
847 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
848 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
849 		    ifp ? ifp->if_index : 0));
850 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
851 			/* ICMPv6 error: MUST deliver it by spec... */
852 			code = PRC_NCMDS;
853 			/* deliver */
854 		} else {
855 			/* ICMPv6 informational: MUST not deliver */
856 			break;
857 		}
858 	deliver:
859 		if (icmp6_notify_error(&m, off, icmp6len, code)) {
860 			/* In this case, m should've been freed. */
861 			return (IPPROTO_DONE);
862 		}
863 		break;
864 
865 	badcode:
866 		ICMP6STAT_INC(icp6s_badcode);
867 		break;
868 
869 	badlen:
870 		ICMP6STAT_INC(icp6s_badlen);
871 		break;
872 	}
873 
874 	/* deliver the packet to appropriate sockets */
875 	icmp6_rip6_input(&m, *offp);
876 
877 	return IPPROTO_DONE;
878 
879  freeit:
880 	m_freem(m);
881 	return IPPROTO_DONE;
882 }
883 
884 static int
885 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
886 {
887 	struct mbuf *m = *mp;
888 	struct icmp6_hdr *icmp6;
889 	struct ip6_hdr *eip6;
890 	u_int32_t notifymtu;
891 	struct sockaddr_in6 icmp6src, icmp6dst;
892 
893 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
894 		ICMP6STAT_INC(icp6s_tooshort);
895 		goto freeit;
896 	}
897 #ifndef PULLDOWN_TEST
898 	IP6_EXTHDR_CHECK(m, off,
899 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
900 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
901 #else
902 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
903 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
904 	if (icmp6 == NULL) {
905 		ICMP6STAT_INC(icp6s_tooshort);
906 		return (-1);
907 	}
908 #endif
909 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
910 
911 	/* Detect the upper level protocol */
912 	{
913 		void (*ctlfunc)(int, struct sockaddr *, void *);
914 		u_int8_t nxt = eip6->ip6_nxt;
915 		int eoff = off + sizeof(struct icmp6_hdr) +
916 		    sizeof(struct ip6_hdr);
917 		struct ip6ctlparam ip6cp;
918 		struct in6_addr *finaldst = NULL;
919 		int icmp6type = icmp6->icmp6_type;
920 		struct ip6_frag *fh;
921 		struct ip6_rthdr *rth;
922 		struct ip6_rthdr0 *rth0;
923 		int rthlen;
924 
925 		while (1) { /* XXX: should avoid infinite loop explicitly? */
926 			struct ip6_ext *eh;
927 
928 			switch (nxt) {
929 			case IPPROTO_HOPOPTS:
930 			case IPPROTO_DSTOPTS:
931 			case IPPROTO_AH:
932 #ifndef PULLDOWN_TEST
933 				IP6_EXTHDR_CHECK(m, 0,
934 				    eoff + sizeof(struct ip6_ext), -1);
935 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
936 #else
937 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
938 				    eoff, sizeof(*eh));
939 				if (eh == NULL) {
940 					ICMP6STAT_INC(icp6s_tooshort);
941 					return (-1);
942 				}
943 #endif
944 
945 				if (nxt == IPPROTO_AH)
946 					eoff += (eh->ip6e_len + 2) << 2;
947 				else
948 					eoff += (eh->ip6e_len + 1) << 3;
949 				nxt = eh->ip6e_nxt;
950 				break;
951 			case IPPROTO_ROUTING:
952 				/*
953 				 * When the erroneous packet contains a
954 				 * routing header, we should examine the
955 				 * header to determine the final destination.
956 				 * Otherwise, we can't properly update
957 				 * information that depends on the final
958 				 * destination (e.g. path MTU).
959 				 */
960 #ifndef PULLDOWN_TEST
961 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
962 				rth = (struct ip6_rthdr *)
963 				    (mtod(m, caddr_t) + eoff);
964 #else
965 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
966 				    eoff, sizeof(*rth));
967 				if (rth == NULL) {
968 					ICMP6STAT_INC(icp6s_tooshort);
969 					return (-1);
970 				}
971 #endif
972 				rthlen = (rth->ip6r_len + 1) << 3;
973 				/*
974 				 * XXX: currently there is no
975 				 * officially defined type other
976 				 * than type-0.
977 				 * Note that if the segment left field
978 				 * is 0, all intermediate hops must
979 				 * have been passed.
980 				 */
981 				if (rth->ip6r_segleft &&
982 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
983 					int hops;
984 
985 #ifndef PULLDOWN_TEST
986 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
987 					rth0 = (struct ip6_rthdr0 *)
988 					    (mtod(m, caddr_t) + eoff);
989 #else
990 					IP6_EXTHDR_GET(rth0,
991 					    struct ip6_rthdr0 *, m,
992 					    eoff, rthlen);
993 					if (rth0 == NULL) {
994 						ICMP6STAT_INC(icp6s_tooshort);
995 						return (-1);
996 					}
997 #endif
998 					/* just ignore a bogus header */
999 					if ((rth0->ip6r0_len % 2) == 0 &&
1000 					    (hops = rth0->ip6r0_len/2))
1001 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1002 				}
1003 				eoff += rthlen;
1004 				nxt = rth->ip6r_nxt;
1005 				break;
1006 			case IPPROTO_FRAGMENT:
1007 #ifndef PULLDOWN_TEST
1008 				IP6_EXTHDR_CHECK(m, 0, eoff +
1009 				    sizeof(struct ip6_frag), -1);
1010 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1011 				    eoff);
1012 #else
1013 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1014 				    eoff, sizeof(*fh));
1015 				if (fh == NULL) {
1016 					ICMP6STAT_INC(icp6s_tooshort);
1017 					return (-1);
1018 				}
1019 #endif
1020 				/*
1021 				 * Data after a fragment header is meaningless
1022 				 * unless it is the first fragment, but
1023 				 * we'll go to the notify label for path MTU
1024 				 * discovery.
1025 				 */
1026 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1027 					goto notify;
1028 
1029 				eoff += sizeof(struct ip6_frag);
1030 				nxt = fh->ip6f_nxt;
1031 				break;
1032 			default:
1033 				/*
1034 				 * This case includes ESP and the No Next
1035 				 * Header.  In such cases going to the notify
1036 				 * label does not have any meaning
1037 				 * (i.e. ctlfunc will be NULL), but we go
1038 				 * anyway since we might have to update
1039 				 * path MTU information.
1040 				 */
1041 				goto notify;
1042 			}
1043 		}
1044 	  notify:
1045 #ifndef PULLDOWN_TEST
1046 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1047 #else
1048 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1049 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1050 		if (icmp6 == NULL) {
1051 			ICMP6STAT_INC(icp6s_tooshort);
1052 			return (-1);
1053 		}
1054 #endif
1055 
1056 		/*
1057 		 * retrieve parameters from the inner IPv6 header, and convert
1058 		 * them into sockaddr structures.
1059 		 * XXX: there is no guarantee that the source or destination
1060 		 * addresses of the inner packet are in the same scope as
1061 		 * the addresses of the icmp packet.  But there is no other
1062 		 * way to determine the zone.
1063 		 */
1064 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1065 
1066 		bzero(&icmp6dst, sizeof(icmp6dst));
1067 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1068 		icmp6dst.sin6_family = AF_INET6;
1069 		if (finaldst == NULL)
1070 			icmp6dst.sin6_addr = eip6->ip6_dst;
1071 		else
1072 			icmp6dst.sin6_addr = *finaldst;
1073 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1074 			goto freeit;
1075 		bzero(&icmp6src, sizeof(icmp6src));
1076 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1077 		icmp6src.sin6_family = AF_INET6;
1078 		icmp6src.sin6_addr = eip6->ip6_src;
1079 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1080 			goto freeit;
1081 		icmp6src.sin6_flowinfo =
1082 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1083 
1084 		if (finaldst == NULL)
1085 			finaldst = &eip6->ip6_dst;
1086 		ip6cp.ip6c_m = m;
1087 		ip6cp.ip6c_icmp6 = icmp6;
1088 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1089 		ip6cp.ip6c_off = eoff;
1090 		ip6cp.ip6c_finaldst = finaldst;
1091 		ip6cp.ip6c_src = &icmp6src;
1092 		ip6cp.ip6c_nxt = nxt;
1093 
1094 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1095 			notifymtu = ntohl(icmp6->icmp6_mtu);
1096 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1097 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1098 		}
1099 
1100 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1101 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1102 		if (ctlfunc) {
1103 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1104 			    &ip6cp);
1105 		}
1106 	}
1107 	*mp = m;
1108 	return (0);
1109 
1110   freeit:
1111 	m_freem(m);
1112 	return (-1);
1113 }
1114 
1115 void
1116 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1117 {
1118 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1119 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1120 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1121 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1122 	struct in_conninfo inc;
1123 
1124 #if 0
1125 	/*
1126 	 * RFC2460 section 5, last paragraph.
1127 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1128 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1129 	 * due to packet translator in the middle.
1130 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1131 	 * special handling.
1132 	 */
1133 	if (mtu < IPV6_MMTU)
1134 		return;
1135 #endif
1136 
1137 	/*
1138 	 * we reject ICMPv6 too big with abnormally small value.
1139 	 * XXX what is the good definition of "abnormally small"?
1140 	 */
1141 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1142 		return;
1143 
1144 	if (!validated)
1145 		return;
1146 
1147 	/*
1148 	 * In case the suggested mtu is less than IPV6_MMTU, we
1149 	 * only need to remember that it was for above mentioned
1150 	 * "alwaysfrag" case.
1151 	 * Try to be as close to the spec as possible.
1152 	 */
1153 	if (mtu < IPV6_MMTU)
1154 		mtu = IPV6_MMTU - 8;
1155 
1156 	bzero(&inc, sizeof(inc));
1157 	inc.inc_flags |= INC_ISIPV6;
1158 	inc.inc6_faddr = *dst;
1159 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1160 		return;
1161 
1162 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1163 		tcp_hc_updatemtu(&inc, mtu);
1164 		ICMP6STAT_INC(icp6s_pmtuchg);
1165 	}
1166 }
1167 
1168 /*
1169  * Process a Node Information Query packet, based on
1170  * draft-ietf-ipngwg-icmp-name-lookups-07.
1171  *
1172  * Spec incompatibilities:
1173  * - IPv6 Subject address handling
1174  * - IPv4 Subject address handling support missing
1175  * - Proxy reply (answer even if it's not for me)
1176  * - joins NI group address at in6_ifattach() time only, does not cope
1177  *   with hostname changes by sethostname(3)
1178  */
1179 static struct mbuf *
1180 ni6_input(struct mbuf *m, int off)
1181 {
1182 	struct icmp6_nodeinfo *ni6, *nni6;
1183 	struct mbuf *n = NULL;
1184 	struct prison *pr;
1185 	u_int16_t qtype;
1186 	int subjlen;
1187 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1188 	struct ni_reply_fqdn *fqdn;
1189 	int addrs;		/* for NI_QTYPE_NODEADDR */
1190 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1191 	struct in6_addr in6_subj; /* subject address */
1192 	struct ip6_hdr *ip6;
1193 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1194 	char *subj = NULL;
1195 	struct in6_ifaddr *ia6 = NULL;
1196 
1197 	ip6 = mtod(m, struct ip6_hdr *);
1198 #ifndef PULLDOWN_TEST
1199 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1200 #else
1201 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1202 	if (ni6 == NULL) {
1203 		/* m is already reclaimed */
1204 		return (NULL);
1205 	}
1206 #endif
1207 
1208 	/*
1209 	 * Validate IPv6 source address.
1210 	 * The default configuration MUST be to refuse answering queries from
1211 	 * global-scope addresses according to RFC4602.
1212 	 * Notes:
1213 	 *  - it's not very clear what "refuse" means; this implementation
1214 	 *    simply drops it.
1215 	 *  - it's not very easy to identify global-scope (unicast) addresses
1216 	 *    since there are many prefixes for them.  It should be safer
1217 	 *    and in practice sufficient to check "all" but loopback and
1218 	 *    link-local (note that site-local unicast was deprecated and
1219 	 *    ULA is defined as global scope-wise)
1220 	 */
1221 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1222 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1223 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1224 		goto bad;
1225 
1226 	/*
1227 	 * Validate IPv6 destination address.
1228 	 *
1229 	 * The Responder must discard the Query without further processing
1230 	 * unless it is one of the Responder's unicast or anycast addresses, or
1231 	 * a link-local scope multicast address which the Responder has joined.
1232 	 * [RFC4602, Section 5.]
1233 	 */
1234 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1235 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1236 			goto bad;
1237 		/* else it's a link-local multicast, fine */
1238 	} else {		/* unicast or anycast */
1239 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1240 			goto bad; /* XXX impossible */
1241 
1242 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1243 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1244 			ifa_free(&ia6->ia_ifa);
1245 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1246 				"a temporary address in %s:%d",
1247 			       __FILE__, __LINE__));
1248 			goto bad;
1249 		}
1250 		ifa_free(&ia6->ia_ifa);
1251 	}
1252 
1253 	/* validate query Subject field. */
1254 	qtype = ntohs(ni6->ni_qtype);
1255 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1256 	switch (qtype) {
1257 	case NI_QTYPE_NOOP:
1258 	case NI_QTYPE_SUPTYPES:
1259 		/* 07 draft */
1260 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1261 			break;
1262 		/* FALLTHROUGH */
1263 	case NI_QTYPE_FQDN:
1264 	case NI_QTYPE_NODEADDR:
1265 	case NI_QTYPE_IPV4ADDR:
1266 		switch (ni6->ni_code) {
1267 		case ICMP6_NI_SUBJ_IPV6:
1268 #if ICMP6_NI_SUBJ_IPV6 != 0
1269 		case 0:
1270 #endif
1271 			/*
1272 			 * backward compatibility - try to accept 03 draft
1273 			 * format, where no Subject is present.
1274 			 */
1275 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1276 			    subjlen == 0) {
1277 				oldfqdn++;
1278 				break;
1279 			}
1280 #if ICMP6_NI_SUBJ_IPV6 != 0
1281 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1282 				goto bad;
1283 #endif
1284 
1285 			if (subjlen != sizeof(struct in6_addr))
1286 				goto bad;
1287 
1288 			/*
1289 			 * Validate Subject address.
1290 			 *
1291 			 * Not sure what exactly "address belongs to the node"
1292 			 * means in the spec, is it just unicast, or what?
1293 			 *
1294 			 * At this moment we consider Subject address as
1295 			 * "belong to the node" if the Subject address equals
1296 			 * to the IPv6 destination address; validation for
1297 			 * IPv6 destination address should have done enough
1298 			 * check for us.
1299 			 *
1300 			 * We do not do proxy at this moment.
1301 			 */
1302 			/* m_pulldown instead of copy? */
1303 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1304 			    subjlen, (caddr_t)&in6_subj);
1305 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1306 				goto bad;
1307 
1308 			subj = (char *)&in6_subj;
1309 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1310 				break;
1311 
1312 			/*
1313 			 * XXX if we are to allow other cases, we should really
1314 			 * be careful about scope here.
1315 			 * basically, we should disallow queries toward IPv6
1316 			 * destination X with subject Y,
1317 			 * if scope(X) > scope(Y).
1318 			 * if we allow scope(X) > scope(Y), it will result in
1319 			 * information leakage across scope boundary.
1320 			 */
1321 			goto bad;
1322 
1323 		case ICMP6_NI_SUBJ_FQDN:
1324 			/*
1325 			 * Validate Subject name with gethostname(3).
1326 			 *
1327 			 * The behavior may need some debate, since:
1328 			 * - we are not sure if the node has FQDN as
1329 			 *   hostname (returned by gethostname(3)).
1330 			 * - the code does wildcard match for truncated names.
1331 			 *   however, we are not sure if we want to perform
1332 			 *   wildcard match, if gethostname(3) side has
1333 			 *   truncated hostname.
1334 			 */
1335 			pr = curthread->td_ucred->cr_prison;
1336 			mtx_lock(&pr->pr_mtx);
1337 			n = ni6_nametodns(pr->pr_hostname,
1338 			    strlen(pr->pr_hostname), 0);
1339 			mtx_unlock(&pr->pr_mtx);
1340 			if (!n || n->m_next || n->m_len == 0)
1341 				goto bad;
1342 			IP6_EXTHDR_GET(subj, char *, m,
1343 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1344 			if (subj == NULL)
1345 				goto bad;
1346 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1347 			    n->m_len)) {
1348 				goto bad;
1349 			}
1350 			m_freem(n);
1351 			n = NULL;
1352 			break;
1353 
1354 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1355 		default:
1356 			goto bad;
1357 		}
1358 		break;
1359 	}
1360 
1361 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1362 	switch (qtype) {
1363 	case NI_QTYPE_FQDN:
1364 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1365 			goto bad;
1366 		break;
1367 	case NI_QTYPE_NODEADDR:
1368 	case NI_QTYPE_IPV4ADDR:
1369 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1370 			goto bad;
1371 		break;
1372 	}
1373 
1374 	/* guess reply length */
1375 	switch (qtype) {
1376 	case NI_QTYPE_NOOP:
1377 		break;		/* no reply data */
1378 	case NI_QTYPE_SUPTYPES:
1379 		replylen += sizeof(u_int32_t);
1380 		break;
1381 	case NI_QTYPE_FQDN:
1382 		/* XXX will append an mbuf */
1383 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1384 		break;
1385 	case NI_QTYPE_NODEADDR:
1386 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1387 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1388 		    sizeof(u_int32_t))) > MCLBYTES)
1389 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1390 		break;
1391 	case NI_QTYPE_IPV4ADDR:
1392 		/* unsupported - should respond with unknown Qtype? */
1393 		break;
1394 	default:
1395 		/*
1396 		 * XXX: We must return a reply with the ICMP6 code
1397 		 * `unknown Qtype' in this case.  However we regard the case
1398 		 * as an FQDN query for backward compatibility.
1399 		 * Older versions set a random value to this field,
1400 		 * so it rarely varies in the defined qtypes.
1401 		 * But the mechanism is not reliable...
1402 		 * maybe we should obsolete older versions.
1403 		 */
1404 		qtype = NI_QTYPE_FQDN;
1405 		/* XXX will append an mbuf */
1406 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1407 		oldfqdn++;
1408 		break;
1409 	}
1410 
1411 	/* allocate an mbuf to reply. */
1412 	MGETHDR(n, M_DONTWAIT, m->m_type);
1413 	if (n == NULL) {
1414 		m_freem(m);
1415 		return (NULL);
1416 	}
1417 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1418 	if (replylen > MHLEN) {
1419 		if (replylen > MCLBYTES) {
1420 			/*
1421 			 * XXX: should we try to allocate more? But MCLBYTES
1422 			 * is probably much larger than IPV6_MMTU...
1423 			 */
1424 			goto bad;
1425 		}
1426 		MCLGET(n, M_DONTWAIT);
1427 		if ((n->m_flags & M_EXT) == 0) {
1428 			goto bad;
1429 		}
1430 	}
1431 	n->m_pkthdr.len = n->m_len = replylen;
1432 
1433 	/* copy mbuf header and IPv6 + Node Information base headers */
1434 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1435 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1436 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1437 
1438 	/* qtype dependent procedure */
1439 	switch (qtype) {
1440 	case NI_QTYPE_NOOP:
1441 		nni6->ni_code = ICMP6_NI_SUCCESS;
1442 		nni6->ni_flags = 0;
1443 		break;
1444 	case NI_QTYPE_SUPTYPES:
1445 	{
1446 		u_int32_t v;
1447 		nni6->ni_code = ICMP6_NI_SUCCESS;
1448 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1449 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1450 		v = (u_int32_t)htonl(0x0000000f);
1451 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1452 		break;
1453 	}
1454 	case NI_QTYPE_FQDN:
1455 		nni6->ni_code = ICMP6_NI_SUCCESS;
1456 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1457 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1458 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1459 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1460 		/*
1461 		 * XXX do we really have FQDN in hostname?
1462 		 */
1463 		pr = curthread->td_ucred->cr_prison;
1464 		mtx_lock(&pr->pr_mtx);
1465 		n->m_next = ni6_nametodns(pr->pr_hostname,
1466 		    strlen(pr->pr_hostname), oldfqdn);
1467 		mtx_unlock(&pr->pr_mtx);
1468 		if (n->m_next == NULL)
1469 			goto bad;
1470 		/* XXX we assume that n->m_next is not a chain */
1471 		if (n->m_next->m_next != NULL)
1472 			goto bad;
1473 		n->m_pkthdr.len += n->m_next->m_len;
1474 		break;
1475 	case NI_QTYPE_NODEADDR:
1476 	{
1477 		int lenlim, copied;
1478 
1479 		nni6->ni_code = ICMP6_NI_SUCCESS;
1480 		n->m_pkthdr.len = n->m_len =
1481 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1482 		lenlim = M_TRAILINGSPACE(n);
1483 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1484 		/* XXX: reset mbuf length */
1485 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1486 		    sizeof(struct icmp6_nodeinfo) + copied;
1487 		break;
1488 	}
1489 	default:
1490 		break;		/* XXX impossible! */
1491 	}
1492 
1493 	nni6->ni_type = ICMP6_NI_REPLY;
1494 	m_freem(m);
1495 	return (n);
1496 
1497   bad:
1498 	m_freem(m);
1499 	if (n)
1500 		m_freem(n);
1501 	return (NULL);
1502 }
1503 
1504 /*
1505  * make a mbuf with DNS-encoded string.  no compression support.
1506  *
1507  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1508  * treated as truncated name (two \0 at the end).  this is a wild guess.
1509  *
1510  * old - return pascal string if non-zero
1511  */
1512 static struct mbuf *
1513 ni6_nametodns(const char *name, int namelen, int old)
1514 {
1515 	struct mbuf *m;
1516 	char *cp, *ep;
1517 	const char *p, *q;
1518 	int i, len, nterm;
1519 
1520 	if (old)
1521 		len = namelen + 1;
1522 	else
1523 		len = MCLBYTES;
1524 
1525 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1526 	MGET(m, M_DONTWAIT, MT_DATA);
1527 	if (m && len > MLEN) {
1528 		MCLGET(m, M_DONTWAIT);
1529 		if ((m->m_flags & M_EXT) == 0)
1530 			goto fail;
1531 	}
1532 	if (!m)
1533 		goto fail;
1534 	m->m_next = NULL;
1535 
1536 	if (old) {
1537 		m->m_len = len;
1538 		*mtod(m, char *) = namelen;
1539 		bcopy(name, mtod(m, char *) + 1, namelen);
1540 		return m;
1541 	} else {
1542 		m->m_len = 0;
1543 		cp = mtod(m, char *);
1544 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1545 
1546 		/* if not certain about my name, return empty buffer */
1547 		if (namelen == 0)
1548 			return m;
1549 
1550 		/*
1551 		 * guess if it looks like shortened hostname, or FQDN.
1552 		 * shortened hostname needs two trailing "\0".
1553 		 */
1554 		i = 0;
1555 		for (p = name; p < name + namelen; p++) {
1556 			if (*p && *p == '.')
1557 				i++;
1558 		}
1559 		if (i < 2)
1560 			nterm = 2;
1561 		else
1562 			nterm = 1;
1563 
1564 		p = name;
1565 		while (cp < ep && p < name + namelen) {
1566 			i = 0;
1567 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1568 				i++;
1569 			/* result does not fit into mbuf */
1570 			if (cp + i + 1 >= ep)
1571 				goto fail;
1572 			/*
1573 			 * DNS label length restriction, RFC1035 page 8.
1574 			 * "i == 0" case is included here to avoid returning
1575 			 * 0-length label on "foo..bar".
1576 			 */
1577 			if (i <= 0 || i >= 64)
1578 				goto fail;
1579 			*cp++ = i;
1580 			bcopy(p, cp, i);
1581 			cp += i;
1582 			p = q;
1583 			if (p < name + namelen && *p == '.')
1584 				p++;
1585 		}
1586 		/* termination */
1587 		if (cp + nterm >= ep)
1588 			goto fail;
1589 		while (nterm-- > 0)
1590 			*cp++ = '\0';
1591 		m->m_len = cp - mtod(m, char *);
1592 		return m;
1593 	}
1594 
1595 	panic("should not reach here");
1596 	/* NOTREACHED */
1597 
1598  fail:
1599 	if (m)
1600 		m_freem(m);
1601 	return NULL;
1602 }
1603 
1604 /*
1605  * check if two DNS-encoded string matches.  takes care of truncated
1606  * form (with \0\0 at the end).  no compression support.
1607  * XXX upper/lowercase match (see RFC2065)
1608  */
1609 static int
1610 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1611 {
1612 	const char *a0, *b0;
1613 	int l;
1614 
1615 	/* simplest case - need validation? */
1616 	if (alen == blen && bcmp(a, b, alen) == 0)
1617 		return 1;
1618 
1619 	a0 = a;
1620 	b0 = b;
1621 
1622 	/* termination is mandatory */
1623 	if (alen < 2 || blen < 2)
1624 		return 0;
1625 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1626 		return 0;
1627 	alen--;
1628 	blen--;
1629 
1630 	while (a - a0 < alen && b - b0 < blen) {
1631 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1632 			return 0;
1633 
1634 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1635 			return 0;
1636 		/* we don't support compression yet */
1637 		if (a[0] >= 64 || b[0] >= 64)
1638 			return 0;
1639 
1640 		/* truncated case */
1641 		if (a[0] == 0 && a - a0 == alen - 1)
1642 			return 1;
1643 		if (b[0] == 0 && b - b0 == blen - 1)
1644 			return 1;
1645 		if (a[0] == 0 || b[0] == 0)
1646 			return 0;
1647 
1648 		if (a[0] != b[0])
1649 			return 0;
1650 		l = a[0];
1651 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1652 			return 0;
1653 		if (bcmp(a + 1, b + 1, l) != 0)
1654 			return 0;
1655 
1656 		a += 1 + l;
1657 		b += 1 + l;
1658 	}
1659 
1660 	if (a - a0 == alen && b - b0 == blen)
1661 		return 1;
1662 	else
1663 		return 0;
1664 }
1665 
1666 /*
1667  * calculate the number of addresses to be returned in the node info reply.
1668  */
1669 static int
1670 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1671     struct in6_addr *subj)
1672 {
1673 	struct ifnet *ifp;
1674 	struct in6_ifaddr *ifa6;
1675 	struct ifaddr *ifa;
1676 	int addrs = 0, addrsofif, iffound = 0;
1677 	int niflags = ni6->ni_flags;
1678 
1679 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1680 		switch (ni6->ni_code) {
1681 		case ICMP6_NI_SUBJ_IPV6:
1682 			if (subj == NULL) /* must be impossible... */
1683 				return (0);
1684 			break;
1685 		default:
1686 			/*
1687 			 * XXX: we only support IPv6 subject address for
1688 			 * this Qtype.
1689 			 */
1690 			return (0);
1691 		}
1692 	}
1693 
1694 	IFNET_RLOCK();
1695 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1696 		addrsofif = 0;
1697 		IF_ADDR_LOCK(ifp);
1698 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1699 			if (ifa->ifa_addr->sa_family != AF_INET6)
1700 				continue;
1701 			ifa6 = (struct in6_ifaddr *)ifa;
1702 
1703 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1704 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1705 				iffound = 1;
1706 
1707 			/*
1708 			 * IPv4-mapped addresses can only be returned by a
1709 			 * Node Information proxy, since they represent
1710 			 * addresses of IPv4-only nodes, which perforce do
1711 			 * not implement this protocol.
1712 			 * [icmp-name-lookups-07, Section 5.4]
1713 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1714 			 * this function at this moment.
1715 			 */
1716 
1717 			/* What do we have to do about ::1? */
1718 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1719 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1720 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1721 					continue;
1722 				break;
1723 			case IPV6_ADDR_SCOPE_SITELOCAL:
1724 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1725 					continue;
1726 				break;
1727 			case IPV6_ADDR_SCOPE_GLOBAL:
1728 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1729 					continue;
1730 				break;
1731 			default:
1732 				continue;
1733 			}
1734 
1735 			/*
1736 			 * check if anycast is okay.
1737 			 * XXX: just experimental.  not in the spec.
1738 			 */
1739 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1740 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1741 				continue; /* we need only unicast addresses */
1742 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1743 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1744 				continue;
1745 			}
1746 			addrsofif++; /* count the address */
1747 		}
1748 		IF_ADDR_UNLOCK(ifp);
1749 		if (iffound) {
1750 			*ifpp = ifp;
1751 			IFNET_RUNLOCK();
1752 			return (addrsofif);
1753 		}
1754 
1755 		addrs += addrsofif;
1756 	}
1757 	IFNET_RUNLOCK();
1758 
1759 	return (addrs);
1760 }
1761 
1762 static int
1763 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1764     struct ifnet *ifp0, int resid)
1765 {
1766 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1767 	struct in6_ifaddr *ifa6;
1768 	struct ifaddr *ifa;
1769 	struct ifnet *ifp_dep = NULL;
1770 	int copied = 0, allow_deprecated = 0;
1771 	u_char *cp = (u_char *)(nni6 + 1);
1772 	int niflags = ni6->ni_flags;
1773 	u_int32_t ltime;
1774 
1775 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1776 		return (0);	/* needless to copy */
1777 
1778 	IFNET_RLOCK();
1779   again:
1780 
1781 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1782 		IF_ADDR_LOCK(ifp);
1783 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1784 			if (ifa->ifa_addr->sa_family != AF_INET6)
1785 				continue;
1786 			ifa6 = (struct in6_ifaddr *)ifa;
1787 
1788 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1789 			    allow_deprecated == 0) {
1790 				/*
1791 				 * prefererred address should be put before
1792 				 * deprecated addresses.
1793 				 */
1794 
1795 				/* record the interface for later search */
1796 				if (ifp_dep == NULL)
1797 					ifp_dep = ifp;
1798 
1799 				continue;
1800 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1801 			    allow_deprecated != 0)
1802 				continue; /* we now collect deprecated addrs */
1803 
1804 			/* What do we have to do about ::1? */
1805 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1806 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1807 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1808 					continue;
1809 				break;
1810 			case IPV6_ADDR_SCOPE_SITELOCAL:
1811 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1812 					continue;
1813 				break;
1814 			case IPV6_ADDR_SCOPE_GLOBAL:
1815 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1816 					continue;
1817 				break;
1818 			default:
1819 				continue;
1820 			}
1821 
1822 			/*
1823 			 * check if anycast is okay.
1824 			 * XXX: just experimental.  not in the spec.
1825 			 */
1826 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1827 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1828 				continue;
1829 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1830 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1831 				continue;
1832 			}
1833 
1834 			/* now we can copy the address */
1835 			if (resid < sizeof(struct in6_addr) +
1836 			    sizeof(u_int32_t)) {
1837 				IF_ADDR_UNLOCK(ifp);
1838 				/*
1839 				 * We give up much more copy.
1840 				 * Set the truncate flag and return.
1841 				 */
1842 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1843 				IFNET_RUNLOCK();
1844 				return (copied);
1845 			}
1846 
1847 			/*
1848 			 * Set the TTL of the address.
1849 			 * The TTL value should be one of the following
1850 			 * according to the specification:
1851 			 *
1852 			 * 1. The remaining lifetime of a DHCP lease on the
1853 			 *    address, or
1854 			 * 2. The remaining Valid Lifetime of a prefix from
1855 			 *    which the address was derived through Stateless
1856 			 *    Autoconfiguration.
1857 			 *
1858 			 * Note that we currently do not support stateful
1859 			 * address configuration by DHCPv6, so the former
1860 			 * case can't happen.
1861 			 */
1862 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1863 				ltime = ND6_INFINITE_LIFETIME;
1864 			else {
1865 				if (ifa6->ia6_lifetime.ia6t_expire >
1866 				    time_second)
1867 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1868 				else
1869 					ltime = 0;
1870 			}
1871 
1872 			bcopy(&ltime, cp, sizeof(u_int32_t));
1873 			cp += sizeof(u_int32_t);
1874 
1875 			/* copy the address itself */
1876 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1877 			    sizeof(struct in6_addr));
1878 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1879 			cp += sizeof(struct in6_addr);
1880 
1881 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1882 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1883 		}
1884 		IF_ADDR_UNLOCK(ifp);
1885 		if (ifp0)	/* we need search only on the specified IF */
1886 			break;
1887 	}
1888 
1889 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1890 		ifp = ifp_dep;
1891 		allow_deprecated = 1;
1892 
1893 		goto again;
1894 	}
1895 
1896 	IFNET_RUNLOCK();
1897 
1898 	return (copied);
1899 }
1900 
1901 /*
1902  * XXX almost dup'ed code with rip6_input.
1903  */
1904 static int
1905 icmp6_rip6_input(struct mbuf **mp, int off)
1906 {
1907 	struct mbuf *m = *mp;
1908 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1909 	struct inpcb *in6p;
1910 	struct inpcb *last = NULL;
1911 	struct sockaddr_in6 fromsa;
1912 	struct icmp6_hdr *icmp6;
1913 	struct mbuf *opts = NULL;
1914 
1915 #ifndef PULLDOWN_TEST
1916 	/* this is assumed to be safe. */
1917 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1918 #else
1919 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1920 	if (icmp6 == NULL) {
1921 		/* m is already reclaimed */
1922 		return (IPPROTO_DONE);
1923 	}
1924 #endif
1925 
1926 	/*
1927 	 * XXX: the address may have embedded scope zone ID, which should be
1928 	 * hidden from applications.
1929 	 */
1930 	bzero(&fromsa, sizeof(fromsa));
1931 	fromsa.sin6_family = AF_INET6;
1932 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1933 	fromsa.sin6_addr = ip6->ip6_src;
1934 	if (sa6_recoverscope(&fromsa)) {
1935 		m_freem(m);
1936 		return (IPPROTO_DONE);
1937 	}
1938 
1939 	INP_INFO_RLOCK(&V_ripcbinfo);
1940 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1941 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1942 			continue;
1943 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1944 			continue;
1945 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1946 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1947 			continue;
1948 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1949 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1950 			continue;
1951 		INP_RLOCK(in6p);
1952 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1953 		    in6p->in6p_icmp6filt)) {
1954 			INP_RUNLOCK(in6p);
1955 			continue;
1956 		}
1957 		if (last != NULL) {
1958 			struct	mbuf *n = NULL;
1959 
1960 			/*
1961 			 * Recent network drivers tend to allocate a single
1962 			 * mbuf cluster, rather than to make a couple of
1963 			 * mbufs without clusters.  Also, since the IPv6 code
1964 			 * path tries to avoid m_pullup(), it is highly
1965 			 * probable that we still have an mbuf cluster here
1966 			 * even though the necessary length can be stored in an
1967 			 * mbuf's internal buffer.
1968 			 * Meanwhile, the default size of the receive socket
1969 			 * buffer for raw sockets is not so large.  This means
1970 			 * the possibility of packet loss is relatively higher
1971 			 * than before.  To avoid this scenario, we copy the
1972 			 * received data to a separate mbuf that does not use
1973 			 * a cluster, if possible.
1974 			 * XXX: it is better to copy the data after stripping
1975 			 * intermediate headers.
1976 			 */
1977 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1978 			    m->m_len <= MHLEN) {
1979 				MGET(n, M_DONTWAIT, m->m_type);
1980 				if (n != NULL) {
1981 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1982 						bcopy(m->m_data, n->m_data,
1983 						      m->m_len);
1984 						n->m_len = m->m_len;
1985 					} else {
1986 						m_free(n);
1987 						n = NULL;
1988 					}
1989 				}
1990 			}
1991 			if (n != NULL ||
1992 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1993 				if (last->inp_flags & INP_CONTROLOPTS)
1994 					ip6_savecontrol(last, n, &opts);
1995 				/* strip intermediate headers */
1996 				m_adj(n, off);
1997 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1998 				if (sbappendaddr_locked(
1999 				    &last->inp_socket->so_rcv,
2000 				    (struct sockaddr *)&fromsa, n, opts)
2001 				    == 0) {
2002 					/* should notify about lost packet */
2003 					m_freem(n);
2004 					if (opts) {
2005 						m_freem(opts);
2006 					}
2007 					SOCKBUF_UNLOCK(
2008 					    &last->inp_socket->so_rcv);
2009 				} else
2010 					sorwakeup_locked(last->inp_socket);
2011 				opts = NULL;
2012 			}
2013 			INP_RUNLOCK(last);
2014 		}
2015 		last = in6p;
2016 	}
2017 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2018 	if (last != NULL) {
2019 		if (last->inp_flags & INP_CONTROLOPTS)
2020 			ip6_savecontrol(last, m, &opts);
2021 		/* strip intermediate headers */
2022 		m_adj(m, off);
2023 
2024 		/* avoid using mbuf clusters if possible (see above) */
2025 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2026 		    m->m_len <= MHLEN) {
2027 			struct mbuf *n;
2028 
2029 			MGET(n, M_DONTWAIT, m->m_type);
2030 			if (n != NULL) {
2031 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2032 					bcopy(m->m_data, n->m_data, m->m_len);
2033 					n->m_len = m->m_len;
2034 
2035 					m_freem(m);
2036 					m = n;
2037 				} else {
2038 					m_freem(n);
2039 					n = NULL;
2040 				}
2041 			}
2042 		}
2043 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2044 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2045 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2046 			m_freem(m);
2047 			if (opts)
2048 				m_freem(opts);
2049 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2050 		} else
2051 			sorwakeup_locked(last->inp_socket);
2052 		INP_RUNLOCK(last);
2053 	} else {
2054 		m_freem(m);
2055 		IP6STAT_DEC(ip6s_delivered);
2056 	}
2057 	return IPPROTO_DONE;
2058 }
2059 
2060 /*
2061  * Reflect the ip6 packet back to the source.
2062  * OFF points to the icmp6 header, counted from the top of the mbuf.
2063  */
2064 void
2065 icmp6_reflect(struct mbuf *m, size_t off)
2066 {
2067 	struct ip6_hdr *ip6;
2068 	struct icmp6_hdr *icmp6;
2069 	struct in6_ifaddr *ia = NULL;
2070 	int plen;
2071 	int type, code;
2072 	struct ifnet *outif = NULL;
2073 	struct in6_addr origdst, src, *srcp = NULL;
2074 
2075 	/* too short to reflect */
2076 	if (off < sizeof(struct ip6_hdr)) {
2077 		nd6log((LOG_DEBUG,
2078 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2079 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2080 		    __FILE__, __LINE__));
2081 		goto bad;
2082 	}
2083 
2084 	/*
2085 	 * If there are extra headers between IPv6 and ICMPv6, strip
2086 	 * off that header first.
2087 	 */
2088 #ifdef DIAGNOSTIC
2089 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2090 		panic("assumption failed in icmp6_reflect");
2091 #endif
2092 	if (off > sizeof(struct ip6_hdr)) {
2093 		size_t l;
2094 		struct ip6_hdr nip6;
2095 
2096 		l = off - sizeof(struct ip6_hdr);
2097 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2098 		m_adj(m, l);
2099 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2100 		if (m->m_len < l) {
2101 			if ((m = m_pullup(m, l)) == NULL)
2102 				return;
2103 		}
2104 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2105 	} else /* off == sizeof(struct ip6_hdr) */ {
2106 		size_t l;
2107 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2108 		if (m->m_len < l) {
2109 			if ((m = m_pullup(m, l)) == NULL)
2110 				return;
2111 		}
2112 	}
2113 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2114 	ip6 = mtod(m, struct ip6_hdr *);
2115 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2116 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2117 	type = icmp6->icmp6_type; /* keep type for statistics */
2118 	code = icmp6->icmp6_code; /* ditto. */
2119 
2120 	origdst = ip6->ip6_dst;
2121 	/*
2122 	 * ip6_input() drops a packet if its src is multicast.
2123 	 * So, the src is never multicast.
2124 	 */
2125 	ip6->ip6_dst = ip6->ip6_src;
2126 
2127 	/*
2128 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2129 	 * use dst as the src for the reply.
2130 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2131 	 * (for example) when we encounter an error while forwarding procedure
2132 	 * destined to a duplicated address of ours.
2133 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2134 	 * procedure of an outgoing packet of our own, in which case we need
2135 	 * to search in the ifaddr list.
2136 	 */
2137 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2138 		if ((ia = ip6_getdstifaddr(m))) {
2139 			if (!(ia->ia6_flags &
2140 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2141 				srcp = &ia->ia_addr.sin6_addr;
2142 		} else {
2143 			struct sockaddr_in6 d;
2144 
2145 			bzero(&d, sizeof(d));
2146 			d.sin6_family = AF_INET6;
2147 			d.sin6_len = sizeof(d);
2148 			d.sin6_addr = origdst;
2149 			ia = (struct in6_ifaddr *)
2150 			    ifa_ifwithaddr((struct sockaddr *)&d);
2151 			if (ia &&
2152 			    !(ia->ia6_flags &
2153 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2154 				srcp = &ia->ia_addr.sin6_addr;
2155 			}
2156 		}
2157 	}
2158 
2159 	if (srcp == NULL) {
2160 		int e;
2161 		struct sockaddr_in6 sin6;
2162 		struct route_in6 ro;
2163 
2164 		/*
2165 		 * This case matches to multicasts, our anycast, or unicasts
2166 		 * that we do not own.  Select a source address based on the
2167 		 * source address of the erroneous packet.
2168 		 */
2169 		bzero(&sin6, sizeof(sin6));
2170 		sin6.sin6_family = AF_INET6;
2171 		sin6.sin6_len = sizeof(sin6);
2172 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2173 
2174 		bzero(&ro, sizeof(ro));
2175 		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &src);
2176 		if (ro.ro_rt)
2177 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2178 		if (e) {
2179 			char ip6buf[INET6_ADDRSTRLEN];
2180 			nd6log((LOG_DEBUG,
2181 			    "icmp6_reflect: source can't be determined: "
2182 			    "dst=%s, error=%d\n",
2183 			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2184 			goto bad;
2185 		}
2186 		srcp = &src;
2187 	}
2188 
2189 	ip6->ip6_src = *srcp;
2190 	ip6->ip6_flow = 0;
2191 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2192 	ip6->ip6_vfc |= IPV6_VERSION;
2193 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2194 	if (outif)
2195 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2196 	else if (m->m_pkthdr.rcvif) {
2197 		/* XXX: This may not be the outgoing interface */
2198 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2199 	} else
2200 		ip6->ip6_hlim = V_ip6_defhlim;
2201 
2202 	icmp6->icmp6_cksum = 0;
2203 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2204 	    sizeof(struct ip6_hdr), plen);
2205 
2206 	/*
2207 	 * XXX option handling
2208 	 */
2209 
2210 	m->m_flags &= ~(M_BCAST|M_MCAST);
2211 
2212 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2213 	if (outif)
2214 		icmp6_ifoutstat_inc(outif, type, code);
2215 
2216 	if (ia != NULL)
2217 		ifa_free(&ia->ia_ifa);
2218 	return;
2219 
2220  bad:
2221 	if (ia != NULL)
2222 		ifa_free(&ia->ia_ifa);
2223 	m_freem(m);
2224 	return;
2225 }
2226 
2227 void
2228 icmp6_fasttimo(void)
2229 {
2230 
2231 	mld_fasttimo();
2232 }
2233 
2234 void
2235 icmp6_slowtimo(void)
2236 {
2237 
2238 	mld_slowtimo();
2239 }
2240 
2241 static const char *
2242 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2243     struct in6_addr *tgt6)
2244 {
2245 	static char buf[1024];
2246 	char ip6bufs[INET6_ADDRSTRLEN];
2247 	char ip6bufd[INET6_ADDRSTRLEN];
2248 	char ip6buft[INET6_ADDRSTRLEN];
2249 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2250 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2251 	    ip6_sprintf(ip6buft, tgt6));
2252 	return buf;
2253 }
2254 
2255 void
2256 icmp6_redirect_input(struct mbuf *m, int off)
2257 {
2258 	struct ifnet *ifp;
2259 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2260 	struct nd_redirect *nd_rd;
2261 	int icmp6len = ntohs(ip6->ip6_plen);
2262 	char *lladdr = NULL;
2263 	int lladdrlen = 0;
2264 	u_char *redirhdr = NULL;
2265 	int redirhdrlen = 0;
2266 	struct rtentry *rt = NULL;
2267 	int is_router;
2268 	int is_onlink;
2269 	struct in6_addr src6 = ip6->ip6_src;
2270 	struct in6_addr redtgt6;
2271 	struct in6_addr reddst6;
2272 	union nd_opts ndopts;
2273 	char ip6buf[INET6_ADDRSTRLEN];
2274 
2275 	if (!m)
2276 		return;
2277 
2278 	ifp = m->m_pkthdr.rcvif;
2279 
2280 	if (!ifp)
2281 		return;
2282 
2283 	/* XXX if we are router, we don't update route by icmp6 redirect */
2284 	if (V_ip6_forwarding)
2285 		goto freeit;
2286 	if (!V_icmp6_rediraccept)
2287 		goto freeit;
2288 
2289 #ifndef PULLDOWN_TEST
2290 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2291 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2292 #else
2293 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2294 	if (nd_rd == NULL) {
2295 		ICMP6STAT_INC(icp6s_tooshort);
2296 		return;
2297 	}
2298 #endif
2299 	redtgt6 = nd_rd->nd_rd_target;
2300 	reddst6 = nd_rd->nd_rd_dst;
2301 
2302 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2303 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2304 		goto freeit;
2305 	}
2306 
2307 	/* validation */
2308 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2309 		nd6log((LOG_ERR,
2310 		    "ICMP6 redirect sent from %s rejected; "
2311 		    "must be from linklocal\n",
2312 		    ip6_sprintf(ip6buf, &src6)));
2313 		goto bad;
2314 	}
2315 	if (ip6->ip6_hlim != 255) {
2316 		nd6log((LOG_ERR,
2317 		    "ICMP6 redirect sent from %s rejected; "
2318 		    "hlim=%d (must be 255)\n",
2319 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2320 		goto bad;
2321 	}
2322     {
2323 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2324 	struct sockaddr_in6 sin6;
2325 	struct in6_addr *gw6;
2326 
2327 	bzero(&sin6, sizeof(sin6));
2328 	sin6.sin6_family = AF_INET6;
2329 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2330 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2331 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2332 	if (rt) {
2333 		if (rt->rt_gateway == NULL ||
2334 		    rt->rt_gateway->sa_family != AF_INET6) {
2335 			nd6log((LOG_ERR,
2336 			    "ICMP6 redirect rejected; no route "
2337 			    "with inet6 gateway found for redirect dst: %s\n",
2338 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2339 			RTFREE_LOCKED(rt);
2340 			goto bad;
2341 		}
2342 
2343 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2344 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2345 			nd6log((LOG_ERR,
2346 			    "ICMP6 redirect rejected; "
2347 			    "not equal to gw-for-src=%s (must be same): "
2348 			    "%s\n",
2349 			    ip6_sprintf(ip6buf, gw6),
2350 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2351 			RTFREE_LOCKED(rt);
2352 			goto bad;
2353 		}
2354 	} else {
2355 		nd6log((LOG_ERR,
2356 		    "ICMP6 redirect rejected; "
2357 		    "no route found for redirect dst: %s\n",
2358 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2359 		goto bad;
2360 	}
2361 	RTFREE_LOCKED(rt);
2362 	rt = NULL;
2363     }
2364 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2365 		nd6log((LOG_ERR,
2366 		    "ICMP6 redirect rejected; "
2367 		    "redirect dst must be unicast: %s\n",
2368 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2369 		goto bad;
2370 	}
2371 
2372 	is_router = is_onlink = 0;
2373 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2374 		is_router = 1;	/* router case */
2375 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2376 		is_onlink = 1;	/* on-link destination case */
2377 	if (!is_router && !is_onlink) {
2378 		nd6log((LOG_ERR,
2379 		    "ICMP6 redirect rejected; "
2380 		    "neither router case nor onlink case: %s\n",
2381 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2382 		goto bad;
2383 	}
2384 	/* validation passed */
2385 
2386 	icmp6len -= sizeof(*nd_rd);
2387 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2388 	if (nd6_options(&ndopts) < 0) {
2389 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2390 		    "invalid ND option, rejected: %s\n",
2391 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2392 		/* nd6_options have incremented stats */
2393 		goto freeit;
2394 	}
2395 
2396 	if (ndopts.nd_opts_tgt_lladdr) {
2397 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2398 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2399 	}
2400 
2401 	if (ndopts.nd_opts_rh) {
2402 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2403 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2404 	}
2405 
2406 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2407 		nd6log((LOG_INFO,
2408 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2409 		    "(if %d, icmp6 packet %d): %s\n",
2410 		    ip6_sprintf(ip6buf, &redtgt6),
2411 		    ifp->if_addrlen, lladdrlen - 2,
2412 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2413 		goto bad;
2414 	}
2415 
2416 	/* RFC 2461 8.3 */
2417 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2418 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2419 
2420 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2421 		/* perform rtredirect */
2422 		struct sockaddr_in6 sdst;
2423 		struct sockaddr_in6 sgw;
2424 		struct sockaddr_in6 ssrc;
2425 
2426 		bzero(&sdst, sizeof(sdst));
2427 		bzero(&sgw, sizeof(sgw));
2428 		bzero(&ssrc, sizeof(ssrc));
2429 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2430 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2431 			sizeof(struct sockaddr_in6);
2432 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2433 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2434 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2435 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2436 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2437 		    (struct sockaddr *)&ssrc);
2438 	}
2439 	/* finally update cached route in each socket via pfctlinput */
2440     {
2441 	struct sockaddr_in6 sdst;
2442 
2443 	bzero(&sdst, sizeof(sdst));
2444 	sdst.sin6_family = AF_INET6;
2445 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2446 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2447 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2448 #ifdef IPSEC
2449 	key_sa_routechange((struct sockaddr *)&sdst);
2450 #endif /* IPSEC */
2451     }
2452 
2453  freeit:
2454 	m_freem(m);
2455 	return;
2456 
2457  bad:
2458 	ICMP6STAT_INC(icp6s_badredirect);
2459 	m_freem(m);
2460 }
2461 
2462 void
2463 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2464 {
2465 	struct ifnet *ifp;	/* my outgoing interface */
2466 	struct in6_addr *ifp_ll6;
2467 	struct in6_addr *router_ll6;
2468 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2469 	struct mbuf *m = NULL;	/* newly allocated one */
2470 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2471 	struct nd_redirect *nd_rd;
2472 	struct llentry *ln = NULL;
2473 	size_t maxlen;
2474 	u_char *p;
2475 	struct ifnet *outif = NULL;
2476 	struct sockaddr_in6 src_sa;
2477 
2478 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2479 
2480 	/* if we are not router, we don't send icmp6 redirect */
2481 	if (!V_ip6_forwarding)
2482 		goto fail;
2483 
2484 	/* sanity check */
2485 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2486 		goto fail;
2487 
2488 	/*
2489 	 * Address check:
2490 	 *  the source address must identify a neighbor, and
2491 	 *  the destination address must not be a multicast address
2492 	 *  [RFC 2461, sec 8.2]
2493 	 */
2494 	sip6 = mtod(m0, struct ip6_hdr *);
2495 	bzero(&src_sa, sizeof(src_sa));
2496 	src_sa.sin6_family = AF_INET6;
2497 	src_sa.sin6_len = sizeof(src_sa);
2498 	src_sa.sin6_addr = sip6->ip6_src;
2499 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2500 		goto fail;
2501 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2502 		goto fail;	/* what should we do here? */
2503 
2504 	/* rate limit */
2505 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2506 		goto fail;
2507 
2508 	/*
2509 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2510 	 * we almost always ask for an mbuf cluster for simplicity.
2511 	 * (MHLEN < IPV6_MMTU is almost always true)
2512 	 */
2513 #if IPV6_MMTU >= MCLBYTES
2514 # error assumption failed about IPV6_MMTU and MCLBYTES
2515 #endif
2516 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2517 	if (m && IPV6_MMTU >= MHLEN)
2518 		MCLGET(m, M_DONTWAIT);
2519 	if (!m)
2520 		goto fail;
2521 	m->m_pkthdr.rcvif = NULL;
2522 	m->m_len = 0;
2523 	maxlen = M_TRAILINGSPACE(m);
2524 	maxlen = min(IPV6_MMTU, maxlen);
2525 	/* just for safety */
2526 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2527 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2528 		goto fail;
2529 	}
2530 
2531 	{
2532 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2533 		struct in6_ifaddr *ia;
2534 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2535 						 IN6_IFF_NOTREADY|
2536 						 IN6_IFF_ANYCAST)) == NULL)
2537 			goto fail;
2538 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2539 		/* XXXRW: reference released prematurely. */
2540 		ifa_free(&ia->ia_ifa);
2541 	}
2542 
2543 	/* get ip6 linklocal address for the router. */
2544 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2545 		struct sockaddr_in6 *sin6;
2546 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2547 		router_ll6 = &sin6->sin6_addr;
2548 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2549 			router_ll6 = (struct in6_addr *)NULL;
2550 	} else
2551 		router_ll6 = (struct in6_addr *)NULL;
2552 
2553 	/* ip6 */
2554 	ip6 = mtod(m, struct ip6_hdr *);
2555 	ip6->ip6_flow = 0;
2556 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2557 	ip6->ip6_vfc |= IPV6_VERSION;
2558 	/* ip6->ip6_plen will be set later */
2559 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2560 	ip6->ip6_hlim = 255;
2561 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2562 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2563 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2564 
2565 	/* ND Redirect */
2566 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2567 	nd_rd->nd_rd_type = ND_REDIRECT;
2568 	nd_rd->nd_rd_code = 0;
2569 	nd_rd->nd_rd_reserved = 0;
2570 	if (rt->rt_flags & RTF_GATEWAY) {
2571 		/*
2572 		 * nd_rd->nd_rd_target must be a link-local address in
2573 		 * better router cases.
2574 		 */
2575 		if (!router_ll6)
2576 			goto fail;
2577 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2578 		    sizeof(nd_rd->nd_rd_target));
2579 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2580 		    sizeof(nd_rd->nd_rd_dst));
2581 	} else {
2582 		/* make sure redtgt == reddst */
2583 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2584 		    sizeof(nd_rd->nd_rd_target));
2585 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2586 		    sizeof(nd_rd->nd_rd_dst));
2587 	}
2588 
2589 	p = (u_char *)(nd_rd + 1);
2590 
2591 	if (!router_ll6)
2592 		goto nolladdropt;
2593 
2594 	{
2595 		/* target lladdr option */
2596 		int len;
2597 		struct nd_opt_hdr *nd_opt;
2598 		char *lladdr;
2599 
2600 		IF_AFDATA_LOCK(ifp);
2601 		ln = nd6_lookup(router_ll6, 0, ifp);
2602 		IF_AFDATA_UNLOCK(ifp);
2603 		if (ln == NULL)
2604 			goto nolladdropt;
2605 
2606 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2607 		len = (len + 7) & ~7;	/* round by 8 */
2608 		/* safety check */
2609 		if (len + (p - (u_char *)ip6) > maxlen)
2610 			goto nolladdropt;
2611 
2612 		if (ln->la_flags & LLE_VALID) {
2613 			nd_opt = (struct nd_opt_hdr *)p;
2614 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2615 			nd_opt->nd_opt_len = len >> 3;
2616 			lladdr = (char *)(nd_opt + 1);
2617 			bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen);
2618 			p += len;
2619 		}
2620 	}
2621 nolladdropt:
2622 	if (ln != NULL)
2623 		LLE_RUNLOCK(ln);
2624 
2625 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2626 
2627 	/* just to be safe */
2628 #ifdef M_DECRYPTED	/*not openbsd*/
2629 	if (m0->m_flags & M_DECRYPTED)
2630 		goto noredhdropt;
2631 #endif
2632 	if (p - (u_char *)ip6 > maxlen)
2633 		goto noredhdropt;
2634 
2635 	{
2636 		/* redirected header option */
2637 		int len;
2638 		struct nd_opt_rd_hdr *nd_opt_rh;
2639 
2640 		/*
2641 		 * compute the maximum size for icmp6 redirect header option.
2642 		 * XXX room for auth header?
2643 		 */
2644 		len = maxlen - (p - (u_char *)ip6);
2645 		len &= ~7;
2646 
2647 		/* This is just for simplicity. */
2648 		if (m0->m_pkthdr.len != m0->m_len) {
2649 			if (m0->m_next) {
2650 				m_freem(m0->m_next);
2651 				m0->m_next = NULL;
2652 			}
2653 			m0->m_pkthdr.len = m0->m_len;
2654 		}
2655 
2656 		/*
2657 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2658 		 * about padding/truncate rule for the original IP packet.
2659 		 * From the discussion on IPv6imp in Feb 1999,
2660 		 * the consensus was:
2661 		 * - "attach as much as possible" is the goal
2662 		 * - pad if not aligned (original size can be guessed by
2663 		 *   original ip6 header)
2664 		 * Following code adds the padding if it is simple enough,
2665 		 * and truncates if not.
2666 		 */
2667 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2668 			panic("assumption failed in %s:%d", __FILE__,
2669 			    __LINE__);
2670 
2671 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2672 			/* not enough room, truncate */
2673 			m0->m_pkthdr.len = m0->m_len = len -
2674 			    sizeof(*nd_opt_rh);
2675 		} else {
2676 			/* enough room, pad or truncate */
2677 			size_t extra;
2678 
2679 			extra = m0->m_pkthdr.len % 8;
2680 			if (extra) {
2681 				/* pad if easy enough, truncate if not */
2682 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2683 					/* pad */
2684 					m0->m_len += (8 - extra);
2685 					m0->m_pkthdr.len += (8 - extra);
2686 				} else {
2687 					/* truncate */
2688 					m0->m_pkthdr.len -= extra;
2689 					m0->m_len -= extra;
2690 				}
2691 			}
2692 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2693 			m0->m_pkthdr.len = m0->m_len = len -
2694 			    sizeof(*nd_opt_rh);
2695 		}
2696 
2697 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2698 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2699 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2700 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2701 		p += sizeof(*nd_opt_rh);
2702 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2703 
2704 		/* connect m0 to m */
2705 		m_tag_delete_chain(m0, NULL);
2706 		m0->m_flags &= ~M_PKTHDR;
2707 		m->m_next = m0;
2708 		m->m_pkthdr.len = m->m_len + m0->m_len;
2709 		m0 = NULL;
2710 	}
2711 noredhdropt:;
2712 	if (m0) {
2713 		m_freem(m0);
2714 		m0 = NULL;
2715 	}
2716 
2717 	/* XXX: clear embedded link IDs in the inner header */
2718 	in6_clearscope(&sip6->ip6_src);
2719 	in6_clearscope(&sip6->ip6_dst);
2720 	in6_clearscope(&nd_rd->nd_rd_target);
2721 	in6_clearscope(&nd_rd->nd_rd_dst);
2722 
2723 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2724 
2725 	nd_rd->nd_rd_cksum = 0;
2726 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2727 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2728 
2729 	/* send the packet to outside... */
2730 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2731 	if (outif) {
2732 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2733 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2734 	}
2735 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2736 
2737 	return;
2738 
2739 fail:
2740 	if (m)
2741 		m_freem(m);
2742 	if (m0)
2743 		m_freem(m0);
2744 }
2745 
2746 /*
2747  * ICMPv6 socket option processing.
2748  */
2749 int
2750 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2751 {
2752 	int error = 0;
2753 	int optlen;
2754 	struct inpcb *inp = sotoinpcb(so);
2755 	int level, op, optname;
2756 
2757 	if (sopt) {
2758 		level = sopt->sopt_level;
2759 		op = sopt->sopt_dir;
2760 		optname = sopt->sopt_name;
2761 		optlen = sopt->sopt_valsize;
2762 	} else
2763 		level = op = optname = optlen = 0;
2764 
2765 	if (level != IPPROTO_ICMPV6) {
2766 		return EINVAL;
2767 	}
2768 
2769 	switch (op) {
2770 	case PRCO_SETOPT:
2771 		switch (optname) {
2772 		case ICMP6_FILTER:
2773 		    {
2774 			struct icmp6_filter ic6f;
2775 
2776 			if (optlen != sizeof(ic6f)) {
2777 				error = EMSGSIZE;
2778 				break;
2779 			}
2780 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2781 			if (error == 0) {
2782 				INP_WLOCK(inp);
2783 				*inp->in6p_icmp6filt = ic6f;
2784 				INP_WUNLOCK(inp);
2785 			}
2786 			break;
2787 		    }
2788 
2789 		default:
2790 			error = ENOPROTOOPT;
2791 			break;
2792 		}
2793 		break;
2794 
2795 	case PRCO_GETOPT:
2796 		switch (optname) {
2797 		case ICMP6_FILTER:
2798 		    {
2799 			struct icmp6_filter ic6f;
2800 
2801 			INP_RLOCK(inp);
2802 			ic6f = *inp->in6p_icmp6filt;
2803 			INP_RUNLOCK(inp);
2804 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2805 			break;
2806 		    }
2807 
2808 		default:
2809 			error = ENOPROTOOPT;
2810 			break;
2811 		}
2812 		break;
2813 	}
2814 
2815 	return (error);
2816 }
2817 
2818 /*
2819  * Perform rate limit check.
2820  * Returns 0 if it is okay to send the icmp6 packet.
2821  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2822  * limitation.
2823  *
2824  * XXX per-destination/type check necessary?
2825  *
2826  * dst - not used at this moment
2827  * type - not used at this moment
2828  * code - not used at this moment
2829  */
2830 static int
2831 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2832     const int code)
2833 {
2834 	int ret;
2835 
2836 	ret = 0;	/* okay to send */
2837 
2838 	/* PPS limit */
2839 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2840 	    V_icmp6errppslim)) {
2841 		/* The packet is subject to rate limit */
2842 		ret++;
2843 	}
2844 
2845 	return ret;
2846 }
2847