xref: /freebsd/sys/netinet6/icmp6.c (revision 6356dba0b403daa023dec24559ab1f8e602e4f14)
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/kernel.h>
73 #include <sys/lock.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/signalvar.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/sx.h>
81 #include <sys/syslog.h>
82 #include <sys/systm.h>
83 #include <sys/time.h>
84 #include <sys/vimage.h>
85 
86 #include <net/if.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89 #include <net/route.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet/icmp6.h>
96 #include <netinet/tcp_var.h>
97 #include <netinet6/in6_ifattach.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/scope6_var.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/nd6.h>
104 
105 #ifdef IPSEC
106 #include <netipsec/ipsec.h>
107 #include <netipsec/key.h>
108 #endif
109 
110 extern struct domain inet6domain;
111 
112 struct icmp6stat icmp6stat;
113 
114 extern struct inpcbinfo ripcbinfo;
115 extern struct inpcbhead ripcb;
116 extern int icmp6errppslim;
117 static int icmp6errpps_count = 0;
118 static struct timeval icmp6errppslim_last;
119 extern int icmp6_nodeinfo;
120 
121 static void icmp6_errcount(struct icmp6errstat *, int, int);
122 static int icmp6_rip6_input(struct mbuf **, int);
123 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
124 static const char *icmp6_redirect_diag __P((struct in6_addr *,
125 	struct in6_addr *, struct in6_addr *));
126 static struct mbuf *ni6_input(struct mbuf *, int);
127 static struct mbuf *ni6_nametodns(const char *, int, int);
128 static int ni6_dnsmatch(const char *, int, const char *, int);
129 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
130 			  struct ifnet **, struct in6_addr *));
131 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
132 				struct ifnet *, int));
133 static int icmp6_notify_error(struct mbuf **, int, int, int);
134 
135 
136 void
137 icmp6_init(void)
138 {
139 
140 	mld6_init();
141 }
142 
143 static void
144 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
145 {
146 	switch (type) {
147 	case ICMP6_DST_UNREACH:
148 		switch (code) {
149 		case ICMP6_DST_UNREACH_NOROUTE:
150 			stat->icp6errs_dst_unreach_noroute++;
151 			return;
152 		case ICMP6_DST_UNREACH_ADMIN:
153 			stat->icp6errs_dst_unreach_admin++;
154 			return;
155 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
156 			stat->icp6errs_dst_unreach_beyondscope++;
157 			return;
158 		case ICMP6_DST_UNREACH_ADDR:
159 			stat->icp6errs_dst_unreach_addr++;
160 			return;
161 		case ICMP6_DST_UNREACH_NOPORT:
162 			stat->icp6errs_dst_unreach_noport++;
163 			return;
164 		}
165 		break;
166 	case ICMP6_PACKET_TOO_BIG:
167 		stat->icp6errs_packet_too_big++;
168 		return;
169 	case ICMP6_TIME_EXCEEDED:
170 		switch (code) {
171 		case ICMP6_TIME_EXCEED_TRANSIT:
172 			stat->icp6errs_time_exceed_transit++;
173 			return;
174 		case ICMP6_TIME_EXCEED_REASSEMBLY:
175 			stat->icp6errs_time_exceed_reassembly++;
176 			return;
177 		}
178 		break;
179 	case ICMP6_PARAM_PROB:
180 		switch (code) {
181 		case ICMP6_PARAMPROB_HEADER:
182 			stat->icp6errs_paramprob_header++;
183 			return;
184 		case ICMP6_PARAMPROB_NEXTHEADER:
185 			stat->icp6errs_paramprob_nextheader++;
186 			return;
187 		case ICMP6_PARAMPROB_OPTION:
188 			stat->icp6errs_paramprob_option++;
189 			return;
190 		}
191 		break;
192 	case ND_REDIRECT:
193 		stat->icp6errs_redirect++;
194 		return;
195 	}
196 	stat->icp6errs_unknown++;
197 }
198 
199 /*
200  * A wrapper function for icmp6_error() necessary when the erroneous packet
201  * may not contain enough scope zone information.
202  */
203 void
204 icmp6_error2(struct mbuf *m, int type, int code, int param,
205     struct ifnet *ifp)
206 {
207 	struct ip6_hdr *ip6;
208 
209 	if (ifp == NULL)
210 		return;
211 
212 #ifndef PULLDOWN_TEST
213 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
214 #else
215 	if (m->m_len < sizeof(struct ip6_hdr)) {
216 		m = m_pullup(m, sizeof(struct ip6_hdr));
217 		if (m == NULL)
218 			return;
219 	}
220 #endif
221 
222 	ip6 = mtod(m, struct ip6_hdr *);
223 
224 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
225 		return;
226 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
227 		return;
228 
229 	icmp6_error(m, type, code, param);
230 }
231 
232 /*
233  * Generate an error packet of type error in response to bad IP6 packet.
234  */
235 void
236 icmp6_error(struct mbuf *m, int type, int code, int param)
237 {
238 	struct ip6_hdr *oip6, *nip6;
239 	struct icmp6_hdr *icmp6;
240 	u_int preplen;
241 	int off;
242 	int nxt;
243 
244 	V_icmp6stat.icp6s_error++;
245 
246 	/* count per-type-code statistics */
247 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, type, code);
248 
249 #ifdef M_DECRYPTED	/*not openbsd*/
250 	if (m->m_flags & M_DECRYPTED) {
251 		V_icmp6stat.icp6s_canterror++;
252 		goto freeit;
253 	}
254 #endif
255 
256 #ifndef PULLDOWN_TEST
257 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
258 #else
259 	if (m->m_len < sizeof(struct ip6_hdr)) {
260 		m = m_pullup(m, sizeof(struct ip6_hdr));
261 		if (m == NULL)
262 			return;
263 	}
264 #endif
265 	oip6 = mtod(m, struct ip6_hdr *);
266 
267 	/*
268 	 * If the destination address of the erroneous packet is a multicast
269 	 * address, or the packet was sent using link-layer multicast,
270 	 * we should basically suppress sending an error (RFC 2463, Section
271 	 * 2.4).
272 	 * We have two exceptions (the item e.2 in that section):
273 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
274 	 * - the Parameter Problem Message that can be allowed an icmp6 error
275 	 *   in the option type field.  This check has been done in
276 	 *   ip6_unknown_opt(), so we can just check the type and code.
277 	 */
278 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
279 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
280 	    (type != ICMP6_PACKET_TOO_BIG &&
281 	     (type != ICMP6_PARAM_PROB ||
282 	      code != ICMP6_PARAMPROB_OPTION)))
283 		goto freeit;
284 
285 	/*
286 	 * RFC 2463, 2.4 (e.5): source address check.
287 	 * XXX: the case of anycast source?
288 	 */
289 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
290 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
291 		goto freeit;
292 
293 	/*
294 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
295 	 * don't do it.
296 	 */
297 	nxt = -1;
298 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
299 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
300 		struct icmp6_hdr *icp;
301 
302 #ifndef PULLDOWN_TEST
303 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
304 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
305 #else
306 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
307 			sizeof(*icp));
308 		if (icp == NULL) {
309 			V_icmp6stat.icp6s_tooshort++;
310 			return;
311 		}
312 #endif
313 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
314 		    icp->icmp6_type == ND_REDIRECT) {
315 			/*
316 			 * ICMPv6 error
317 			 * Special case: for redirect (which is
318 			 * informational) we must not send icmp6 error.
319 			 */
320 			V_icmp6stat.icp6s_canterror++;
321 			goto freeit;
322 		} else {
323 			/* ICMPv6 informational - send the error */
324 		}
325 	} else {
326 		/* non-ICMPv6 - send the error */
327 	}
328 
329 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
330 
331 	/* Finally, do rate limitation check. */
332 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
333 		V_icmp6stat.icp6s_toofreq++;
334 		goto freeit;
335 	}
336 
337 	/*
338 	 * OK, ICMP6 can be generated.
339 	 */
340 
341 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
342 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
343 
344 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
345 	M_PREPEND(m, preplen, M_DONTWAIT);
346 	if (m && m->m_len < preplen)
347 		m = m_pullup(m, preplen);
348 	if (m == NULL) {
349 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
350 		return;
351 	}
352 
353 	nip6 = mtod(m, struct ip6_hdr *);
354 	nip6->ip6_src  = oip6->ip6_src;
355 	nip6->ip6_dst  = oip6->ip6_dst;
356 
357 	in6_clearscope(&oip6->ip6_src);
358 	in6_clearscope(&oip6->ip6_dst);
359 
360 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
361 	icmp6->icmp6_type = type;
362 	icmp6->icmp6_code = code;
363 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
364 
365 	/*
366 	 * icmp6_reflect() is designed to be in the input path.
367 	 * icmp6_error() can be called from both input and output path,
368 	 * and if we are in output path rcvif could contain bogus value.
369 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
370 	 * information in ip header (nip6).
371 	 */
372 	m->m_pkthdr.rcvif = NULL;
373 
374 	V_icmp6stat.icp6s_outhist[type]++;
375 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
376 
377 	return;
378 
379   freeit:
380 	/*
381 	 * If we can't tell whether or not we can generate ICMP6, free it.
382 	 */
383 	m_freem(m);
384 }
385 
386 /*
387  * Process a received ICMP6 message.
388  */
389 int
390 icmp6_input(struct mbuf **mp, int *offp, int proto)
391 {
392 	struct mbuf *m = *mp, *n;
393 	struct ip6_hdr *ip6, *nip6;
394 	struct icmp6_hdr *icmp6, *nicmp6;
395 	int off = *offp;
396 	int icmp6len = m->m_pkthdr.len - *offp;
397 	int code, sum, noff;
398 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
399 
400 #ifndef PULLDOWN_TEST
401 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
402 	/* m might change if M_LOOP.  So, call mtod after this */
403 #endif
404 
405 	/*
406 	 * Locate icmp6 structure in mbuf, and check
407 	 * that not corrupted and of at least minimum length
408 	 */
409 
410 	ip6 = mtod(m, struct ip6_hdr *);
411 	if (icmp6len < sizeof(struct icmp6_hdr)) {
412 		V_icmp6stat.icp6s_tooshort++;
413 		goto freeit;
414 	}
415 
416 	/*
417 	 * calculate the checksum
418 	 */
419 #ifndef PULLDOWN_TEST
420 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
421 #else
422 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
423 	if (icmp6 == NULL) {
424 		V_icmp6stat.icp6s_tooshort++;
425 		return IPPROTO_DONE;
426 	}
427 #endif
428 	code = icmp6->icmp6_code;
429 
430 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
431 		nd6log((LOG_ERR,
432 		    "ICMP6 checksum error(%d|%x) %s\n",
433 		    icmp6->icmp6_type, sum,
434 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
435 		V_icmp6stat.icp6s_checksum++;
436 		goto freeit;
437 	}
438 
439 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
440 		/*
441 		 * Deliver very specific ICMP6 type only.
442 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
443 		 * will not work.
444 		 */
445 		switch (icmp6->icmp6_type) {
446 		case ICMP6_DST_UNREACH:
447 		case ICMP6_PACKET_TOO_BIG:
448 		case ICMP6_TIME_EXCEEDED:
449 			break;
450 		default:
451 			goto freeit;
452 		}
453 	}
454 
455 	V_icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
456 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
457 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
458 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
459 
460 	switch (icmp6->icmp6_type) {
461 	case ICMP6_DST_UNREACH:
462 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
463 		switch (code) {
464 		case ICMP6_DST_UNREACH_NOROUTE:
465 			code = PRC_UNREACH_NET;
466 			break;
467 		case ICMP6_DST_UNREACH_ADMIN:
468 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
469 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
470 			break;
471 		case ICMP6_DST_UNREACH_ADDR:
472 			code = PRC_HOSTDEAD;
473 			break;
474 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
475 			/* I mean "source address was incorrect." */
476 			code = PRC_PARAMPROB;
477 			break;
478 		case ICMP6_DST_UNREACH_NOPORT:
479 			code = PRC_UNREACH_PORT;
480 			break;
481 		default:
482 			goto badcode;
483 		}
484 		goto deliver;
485 		break;
486 
487 	case ICMP6_PACKET_TOO_BIG:
488 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
489 
490 		/* validation is made in icmp6_mtudisc_update */
491 
492 		code = PRC_MSGSIZE;
493 
494 		/*
495 		 * Updating the path MTU will be done after examining
496 		 * intermediate extension headers.
497 		 */
498 		goto deliver;
499 		break;
500 
501 	case ICMP6_TIME_EXCEEDED:
502 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
503 		switch (code) {
504 		case ICMP6_TIME_EXCEED_TRANSIT:
505 			code = PRC_TIMXCEED_INTRANS;
506 			break;
507 		case ICMP6_TIME_EXCEED_REASSEMBLY:
508 			code = PRC_TIMXCEED_REASS;
509 			break;
510 		default:
511 			goto badcode;
512 		}
513 		goto deliver;
514 		break;
515 
516 	case ICMP6_PARAM_PROB:
517 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
518 		switch (code) {
519 		case ICMP6_PARAMPROB_NEXTHEADER:
520 			code = PRC_UNREACH_PROTOCOL;
521 			break;
522 		case ICMP6_PARAMPROB_HEADER:
523 		case ICMP6_PARAMPROB_OPTION:
524 			code = PRC_PARAMPROB;
525 			break;
526 		default:
527 			goto badcode;
528 		}
529 		goto deliver;
530 		break;
531 
532 	case ICMP6_ECHO_REQUEST:
533 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
534 		if (code != 0)
535 			goto badcode;
536 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
537 			/* Give up remote */
538 			break;
539 		}
540 		if ((n->m_flags & M_EXT) != 0
541 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
542 			struct mbuf *n0 = n;
543 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
544 			int n0len;
545 
546 			MGETHDR(n, M_DONTWAIT, n0->m_type);
547 			n0len = n0->m_pkthdr.len;	/* save for use below */
548 			if (n)
549 				M_MOVE_PKTHDR(n, n0);
550 			if (n && maxlen >= MHLEN) {
551 				MCLGET(n, M_DONTWAIT);
552 				if ((n->m_flags & M_EXT) == 0) {
553 					m_free(n);
554 					n = NULL;
555 				}
556 			}
557 			if (n == NULL) {
558 				/* Give up remote */
559 				m_freem(n0);
560 				break;
561 			}
562 			/*
563 			 * Copy IPv6 and ICMPv6 only.
564 			 */
565 			nip6 = mtod(n, struct ip6_hdr *);
566 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
567 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
568 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
569 			noff = sizeof(struct ip6_hdr);
570 			/* new mbuf contains only ipv6+icmpv6 headers */
571 			n->m_len = noff + sizeof(struct icmp6_hdr);
572 			/*
573 			 * Adjust mbuf.  ip6_plen will be adjusted in
574 			 * ip6_output().
575 			 */
576 			m_adj(n0, off + sizeof(struct icmp6_hdr));
577 			/* recalculate complete packet size */
578 			n->m_pkthdr.len = n0len + (noff - off);
579 			n->m_next = n0;
580 		} else {
581 			nip6 = mtod(n, struct ip6_hdr *);
582 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
583 			    sizeof(*nicmp6));
584 			noff = off;
585 		}
586 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
587 		nicmp6->icmp6_code = 0;
588 		if (n) {
589 			V_icmp6stat.icp6s_reflect++;
590 			V_icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
591 			icmp6_reflect(n, noff);
592 		}
593 		break;
594 
595 	case ICMP6_ECHO_REPLY:
596 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
597 		if (code != 0)
598 			goto badcode;
599 		break;
600 
601 	case MLD_LISTENER_QUERY:
602 	case MLD_LISTENER_REPORT:
603 		if (icmp6len < sizeof(struct mld_hdr))
604 			goto badlen;
605 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
606 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
607 		else
608 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
609 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
610 			/* give up local */
611 			mld6_input(m, off);
612 			m = NULL;
613 			goto freeit;
614 		}
615 		mld6_input(n, off);
616 		/* m stays. */
617 		break;
618 
619 	case MLD_LISTENER_DONE:
620 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
621 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
622 			goto badlen;
623 		break;		/* nothing to be done in kernel */
624 
625 	case MLD_MTRACE_RESP:
626 	case MLD_MTRACE:
627 		/* XXX: these two are experimental.  not officially defined. */
628 		/* XXX: per-interface statistics? */
629 		break;		/* just pass it to applications */
630 
631 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
632 	    {
633 		enum { WRU, FQDN } mode;
634 
635 		if (!V_icmp6_nodeinfo)
636 			break;
637 
638 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
639 			mode = WRU;
640 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
641 			mode = FQDN;
642 		else
643 			goto badlen;
644 
645 #define hostnamelen	strlen(V_hostname)
646 		if (mode == FQDN) {
647 #ifndef PULLDOWN_TEST
648 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
649 			    IPPROTO_DONE);
650 #endif
651 			n = m_copy(m, 0, M_COPYALL);
652 			if (n)
653 				n = ni6_input(n, off);
654 			/* XXX meaningless if n == NULL */
655 			noff = sizeof(struct ip6_hdr);
656 		} else {
657 			u_char *p;
658 			int maxlen, maxhlen;
659 
660 			/*
661 			 * XXX: this combination of flags is pointless,
662 			 * but should we keep this for compatibility?
663 			 */
664 			if ((V_icmp6_nodeinfo & 5) != 5)
665 				break;
666 
667 			if (code != 0)
668 				goto badcode;
669 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
670 			if (maxlen >= MCLBYTES) {
671 				/* Give up remote */
672 				break;
673 			}
674 			MGETHDR(n, M_DONTWAIT, m->m_type);
675 			if (n && maxlen > MHLEN) {
676 				MCLGET(n, M_DONTWAIT);
677 				if ((n->m_flags & M_EXT) == 0) {
678 					m_free(n);
679 					n = NULL;
680 				}
681 			}
682 			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
683 				/*
684 				 * Previous code did a blind M_COPY_PKTHDR
685 				 * and said "just for rcvif".  If true, then
686 				 * we could tolerate the dup failing (due to
687 				 * the deep copy of the tag chain).  For now
688 				 * be conservative and just fail.
689 				 */
690 				m_free(n);
691 				n = NULL;
692 			}
693 			if (n == NULL) {
694 				/* Give up remote */
695 				break;
696 			}
697 			n->m_pkthdr.rcvif = NULL;
698 			n->m_len = 0;
699 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
700 			mtx_lock(&hostname_mtx);
701 			if (maxhlen > hostnamelen)
702 				maxhlen = hostnamelen;
703 			/*
704 			 * Copy IPv6 and ICMPv6 only.
705 			 */
706 			nip6 = mtod(n, struct ip6_hdr *);
707 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
708 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
709 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
710 			p = (u_char *)(nicmp6 + 1);
711 			bzero(p, 4);
712 			bcopy(V_hostname, p + 4, maxhlen); /* meaningless TTL */
713 			mtx_unlock(&hostname_mtx);
714 			noff = sizeof(struct ip6_hdr);
715 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
716 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
717 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
718 			nicmp6->icmp6_code = 0;
719 		}
720 #undef hostnamelen
721 		if (n) {
722 			V_icmp6stat.icp6s_reflect++;
723 			V_icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
724 			icmp6_reflect(n, noff);
725 		}
726 		break;
727 	    }
728 
729 	case ICMP6_WRUREPLY:
730 		if (code != 0)
731 			goto badcode;
732 		break;
733 
734 	case ND_ROUTER_SOLICIT:
735 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
736 		if (code != 0)
737 			goto badcode;
738 		if (icmp6len < sizeof(struct nd_router_solicit))
739 			goto badlen;
740 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
741 			/* give up local */
742 			nd6_rs_input(m, off, icmp6len);
743 			m = NULL;
744 			goto freeit;
745 		}
746 		nd6_rs_input(n, off, icmp6len);
747 		/* m stays. */
748 		break;
749 
750 	case ND_ROUTER_ADVERT:
751 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
752 		if (code != 0)
753 			goto badcode;
754 		if (icmp6len < sizeof(struct nd_router_advert))
755 			goto badlen;
756 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
757 			/* give up local */
758 			nd6_ra_input(m, off, icmp6len);
759 			m = NULL;
760 			goto freeit;
761 		}
762 		nd6_ra_input(n, off, icmp6len);
763 		/* m stays. */
764 		break;
765 
766 	case ND_NEIGHBOR_SOLICIT:
767 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
768 		if (code != 0)
769 			goto badcode;
770 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
771 			goto badlen;
772 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
773 			/* give up local */
774 			nd6_ns_input(m, off, icmp6len);
775 			m = NULL;
776 			goto freeit;
777 		}
778 		nd6_ns_input(n, off, icmp6len);
779 		/* m stays. */
780 		break;
781 
782 	case ND_NEIGHBOR_ADVERT:
783 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
784 		if (code != 0)
785 			goto badcode;
786 		if (icmp6len < sizeof(struct nd_neighbor_advert))
787 			goto badlen;
788 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
789 			/* give up local */
790 			nd6_na_input(m, off, icmp6len);
791 			m = NULL;
792 			goto freeit;
793 		}
794 		nd6_na_input(n, off, icmp6len);
795 		/* m stays. */
796 		break;
797 
798 	case ND_REDIRECT:
799 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
800 		if (code != 0)
801 			goto badcode;
802 		if (icmp6len < sizeof(struct nd_redirect))
803 			goto badlen;
804 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
805 			/* give up local */
806 			icmp6_redirect_input(m, off);
807 			m = NULL;
808 			goto freeit;
809 		}
810 		icmp6_redirect_input(n, off);
811 		/* m stays. */
812 		break;
813 
814 	case ICMP6_ROUTER_RENUMBERING:
815 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
816 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
817 			goto badcode;
818 		if (icmp6len < sizeof(struct icmp6_router_renum))
819 			goto badlen;
820 		break;
821 
822 	default:
823 		nd6log((LOG_DEBUG,
824 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
825 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
826 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
827 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
828 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
829 			/* ICMPv6 error: MUST deliver it by spec... */
830 			code = PRC_NCMDS;
831 			/* deliver */
832 		} else {
833 			/* ICMPv6 informational: MUST not deliver */
834 			break;
835 		}
836 	deliver:
837 		if (icmp6_notify_error(&m, off, icmp6len, code)) {
838 			/* In this case, m should've been freed. */
839 			return (IPPROTO_DONE);
840 		}
841 		break;
842 
843 	badcode:
844 		V_icmp6stat.icp6s_badcode++;
845 		break;
846 
847 	badlen:
848 		V_icmp6stat.icp6s_badlen++;
849 		break;
850 	}
851 
852 	/* deliver the packet to appropriate sockets */
853 	icmp6_rip6_input(&m, *offp);
854 
855 	return IPPROTO_DONE;
856 
857  freeit:
858 	m_freem(m);
859 	return IPPROTO_DONE;
860 }
861 
862 static int
863 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
864 {
865 	struct mbuf *m = *mp;
866 	struct icmp6_hdr *icmp6;
867 	struct ip6_hdr *eip6;
868 	u_int32_t notifymtu;
869 	struct sockaddr_in6 icmp6src, icmp6dst;
870 
871 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
872 		V_icmp6stat.icp6s_tooshort++;
873 		goto freeit;
874 	}
875 #ifndef PULLDOWN_TEST
876 	IP6_EXTHDR_CHECK(m, off,
877 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
878 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
879 #else
880 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
881 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
882 	if (icmp6 == NULL) {
883 		V_icmp6stat.icp6s_tooshort++;
884 		return (-1);
885 	}
886 #endif
887 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
888 
889 	/* Detect the upper level protocol */
890 	{
891 		void (*ctlfunc)(int, struct sockaddr *, void *);
892 		u_int8_t nxt = eip6->ip6_nxt;
893 		int eoff = off + sizeof(struct icmp6_hdr) +
894 		    sizeof(struct ip6_hdr);
895 		struct ip6ctlparam ip6cp;
896 		struct in6_addr *finaldst = NULL;
897 		int icmp6type = icmp6->icmp6_type;
898 		struct ip6_frag *fh;
899 		struct ip6_rthdr *rth;
900 		struct ip6_rthdr0 *rth0;
901 		int rthlen;
902 
903 		while (1) { /* XXX: should avoid infinite loop explicitly? */
904 			struct ip6_ext *eh;
905 
906 			switch (nxt) {
907 			case IPPROTO_HOPOPTS:
908 			case IPPROTO_DSTOPTS:
909 			case IPPROTO_AH:
910 #ifndef PULLDOWN_TEST
911 				IP6_EXTHDR_CHECK(m, 0,
912 				    eoff + sizeof(struct ip6_ext), -1);
913 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
914 #else
915 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
916 				    eoff, sizeof(*eh));
917 				if (eh == NULL) {
918 					V_icmp6stat.icp6s_tooshort++;
919 					return (-1);
920 				}
921 #endif
922 
923 				if (nxt == IPPROTO_AH)
924 					eoff += (eh->ip6e_len + 2) << 2;
925 				else
926 					eoff += (eh->ip6e_len + 1) << 3;
927 				nxt = eh->ip6e_nxt;
928 				break;
929 			case IPPROTO_ROUTING:
930 				/*
931 				 * When the erroneous packet contains a
932 				 * routing header, we should examine the
933 				 * header to determine the final destination.
934 				 * Otherwise, we can't properly update
935 				 * information that depends on the final
936 				 * destination (e.g. path MTU).
937 				 */
938 #ifndef PULLDOWN_TEST
939 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
940 				rth = (struct ip6_rthdr *)
941 				    (mtod(m, caddr_t) + eoff);
942 #else
943 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
944 				    eoff, sizeof(*rth));
945 				if (rth == NULL) {
946 					V_icmp6stat.icp6s_tooshort++;
947 					return (-1);
948 				}
949 #endif
950 				rthlen = (rth->ip6r_len + 1) << 3;
951 				/*
952 				 * XXX: currently there is no
953 				 * officially defined type other
954 				 * than type-0.
955 				 * Note that if the segment left field
956 				 * is 0, all intermediate hops must
957 				 * have been passed.
958 				 */
959 				if (rth->ip6r_segleft &&
960 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
961 					int hops;
962 
963 #ifndef PULLDOWN_TEST
964 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
965 					rth0 = (struct ip6_rthdr0 *)
966 					    (mtod(m, caddr_t) + eoff);
967 #else
968 					IP6_EXTHDR_GET(rth0,
969 					    struct ip6_rthdr0 *, m,
970 					    eoff, rthlen);
971 					if (rth0 == NULL) {
972 						V_icmp6stat.icp6s_tooshort++;
973 						return (-1);
974 					}
975 #endif
976 					/* just ignore a bogus header */
977 					if ((rth0->ip6r0_len % 2) == 0 &&
978 					    (hops = rth0->ip6r0_len/2))
979 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
980 				}
981 				eoff += rthlen;
982 				nxt = rth->ip6r_nxt;
983 				break;
984 			case IPPROTO_FRAGMENT:
985 #ifndef PULLDOWN_TEST
986 				IP6_EXTHDR_CHECK(m, 0, eoff +
987 				    sizeof(struct ip6_frag), -1);
988 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
989 				    eoff);
990 #else
991 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
992 				    eoff, sizeof(*fh));
993 				if (fh == NULL) {
994 					V_icmp6stat.icp6s_tooshort++;
995 					return (-1);
996 				}
997 #endif
998 				/*
999 				 * Data after a fragment header is meaningless
1000 				 * unless it is the first fragment, but
1001 				 * we'll go to the notify label for path MTU
1002 				 * discovery.
1003 				 */
1004 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1005 					goto notify;
1006 
1007 				eoff += sizeof(struct ip6_frag);
1008 				nxt = fh->ip6f_nxt;
1009 				break;
1010 			default:
1011 				/*
1012 				 * This case includes ESP and the No Next
1013 				 * Header.  In such cases going to the notify
1014 				 * label does not have any meaning
1015 				 * (i.e. ctlfunc will be NULL), but we go
1016 				 * anyway since we might have to update
1017 				 * path MTU information.
1018 				 */
1019 				goto notify;
1020 			}
1021 		}
1022 	  notify:
1023 #ifndef PULLDOWN_TEST
1024 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1025 #else
1026 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1027 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1028 		if (icmp6 == NULL) {
1029 			V_icmp6stat.icp6s_tooshort++;
1030 			return (-1);
1031 		}
1032 #endif
1033 
1034 		/*
1035 		 * retrieve parameters from the inner IPv6 header, and convert
1036 		 * them into sockaddr structures.
1037 		 * XXX: there is no guarantee that the source or destination
1038 		 * addresses of the inner packet are in the same scope as
1039 		 * the addresses of the icmp packet.  But there is no other
1040 		 * way to determine the zone.
1041 		 */
1042 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1043 
1044 		bzero(&icmp6dst, sizeof(icmp6dst));
1045 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1046 		icmp6dst.sin6_family = AF_INET6;
1047 		if (finaldst == NULL)
1048 			icmp6dst.sin6_addr = eip6->ip6_dst;
1049 		else
1050 			icmp6dst.sin6_addr = *finaldst;
1051 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1052 			goto freeit;
1053 		bzero(&icmp6src, sizeof(icmp6src));
1054 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1055 		icmp6src.sin6_family = AF_INET6;
1056 		icmp6src.sin6_addr = eip6->ip6_src;
1057 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1058 			goto freeit;
1059 		icmp6src.sin6_flowinfo =
1060 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1061 
1062 		if (finaldst == NULL)
1063 			finaldst = &eip6->ip6_dst;
1064 		ip6cp.ip6c_m = m;
1065 		ip6cp.ip6c_icmp6 = icmp6;
1066 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1067 		ip6cp.ip6c_off = eoff;
1068 		ip6cp.ip6c_finaldst = finaldst;
1069 		ip6cp.ip6c_src = &icmp6src;
1070 		ip6cp.ip6c_nxt = nxt;
1071 
1072 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1073 			notifymtu = ntohl(icmp6->icmp6_mtu);
1074 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1075 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1076 		}
1077 
1078 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1079 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1080 		if (ctlfunc) {
1081 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1082 			    &ip6cp);
1083 		}
1084 	}
1085 	*mp = m;
1086 	return (0);
1087 
1088   freeit:
1089 	m_freem(m);
1090 	return (-1);
1091 }
1092 
1093 void
1094 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1095 {
1096 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1097 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1098 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1099 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1100 	struct in_conninfo inc;
1101 
1102 #if 0
1103 	/*
1104 	 * RFC2460 section 5, last paragraph.
1105 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1106 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1107 	 * due to packet translator in the middle.
1108 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1109 	 * special handling.
1110 	 */
1111 	if (mtu < IPV6_MMTU)
1112 		return;
1113 #endif
1114 
1115 	/*
1116 	 * we reject ICMPv6 too big with abnormally small value.
1117 	 * XXX what is the good definition of "abnormally small"?
1118 	 */
1119 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1120 		return;
1121 
1122 	if (!validated)
1123 		return;
1124 
1125 	bzero(&inc, sizeof(inc));
1126 	inc.inc_flags = 1; /* IPv6 */
1127 	inc.inc6_faddr = *dst;
1128 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1129 		return;
1130 
1131 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1132 		tcp_hc_updatemtu(&inc, mtu);
1133 		V_icmp6stat.icp6s_pmtuchg++;
1134 	}
1135 }
1136 
1137 /*
1138  * Process a Node Information Query packet, based on
1139  * draft-ietf-ipngwg-icmp-name-lookups-07.
1140  *
1141  * Spec incompatibilities:
1142  * - IPv6 Subject address handling
1143  * - IPv4 Subject address handling support missing
1144  * - Proxy reply (answer even if it's not for me)
1145  * - joins NI group address at in6_ifattach() time only, does not cope
1146  *   with hostname changes by sethostname(3)
1147  */
1148 #define hostnamelen	strlen(V_hostname)
1149 static struct mbuf *
1150 ni6_input(struct mbuf *m, int off)
1151 {
1152 	struct icmp6_nodeinfo *ni6, *nni6;
1153 	struct mbuf *n = NULL;
1154 	u_int16_t qtype;
1155 	int subjlen;
1156 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1157 	struct ni_reply_fqdn *fqdn;
1158 	int addrs;		/* for NI_QTYPE_NODEADDR */
1159 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1160 	struct in6_addr in6_subj; /* subject address */
1161 	struct ip6_hdr *ip6;
1162 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1163 	char *subj = NULL;
1164 	struct in6_ifaddr *ia6 = NULL;
1165 
1166 	ip6 = mtod(m, struct ip6_hdr *);
1167 #ifndef PULLDOWN_TEST
1168 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1169 #else
1170 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1171 	if (ni6 == NULL) {
1172 		/* m is already reclaimed */
1173 		return (NULL);
1174 	}
1175 #endif
1176 
1177 	/*
1178 	 * Validate IPv6 source address.
1179 	 * The default configuration MUST be to refuse answering queries from
1180 	 * global-scope addresses according to RFC4602.
1181 	 * Notes:
1182 	 *  - it's not very clear what "refuse" means; this implementation
1183 	 *    simply drops it.
1184 	 *  - it's not very easy to identify global-scope (unicast) addresses
1185 	 *    since there are many prefixes for them.  It should be safer
1186 	 *    and in practice sufficient to check "all" but loopback and
1187 	 *    link-local (note that site-local unicast was deprecated and
1188 	 *    ULA is defined as global scope-wise)
1189 	 */
1190 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1191 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1192 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1193 		goto bad;
1194 
1195 	/*
1196 	 * Validate IPv6 destination address.
1197 	 *
1198 	 * The Responder must discard the Query without further processing
1199 	 * unless it is one of the Responder's unicast or anycast addresses, or
1200 	 * a link-local scope multicast address which the Responder has joined.
1201 	 * [RFC4602, Section 5.]
1202 	 */
1203 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1204 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1205 			goto bad;
1206 		/* else it's a link-local multicast, fine */
1207 	} else {		/* unicast or anycast */
1208 		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1209 			goto bad; /* XXX impossible */
1210 
1211 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1212 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1213 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1214 				"a temporary address in %s:%d",
1215 			       __FILE__, __LINE__));
1216 			goto bad;
1217 		}
1218 	}
1219 
1220 	/* validate query Subject field. */
1221 	qtype = ntohs(ni6->ni_qtype);
1222 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1223 	switch (qtype) {
1224 	case NI_QTYPE_NOOP:
1225 	case NI_QTYPE_SUPTYPES:
1226 		/* 07 draft */
1227 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1228 			break;
1229 		/* FALLTHROUGH */
1230 	case NI_QTYPE_FQDN:
1231 	case NI_QTYPE_NODEADDR:
1232 	case NI_QTYPE_IPV4ADDR:
1233 		switch (ni6->ni_code) {
1234 		case ICMP6_NI_SUBJ_IPV6:
1235 #if ICMP6_NI_SUBJ_IPV6 != 0
1236 		case 0:
1237 #endif
1238 			/*
1239 			 * backward compatibility - try to accept 03 draft
1240 			 * format, where no Subject is present.
1241 			 */
1242 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1243 			    subjlen == 0) {
1244 				oldfqdn++;
1245 				break;
1246 			}
1247 #if ICMP6_NI_SUBJ_IPV6 != 0
1248 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1249 				goto bad;
1250 #endif
1251 
1252 			if (subjlen != sizeof(struct in6_addr))
1253 				goto bad;
1254 
1255 			/*
1256 			 * Validate Subject address.
1257 			 *
1258 			 * Not sure what exactly "address belongs to the node"
1259 			 * means in the spec, is it just unicast, or what?
1260 			 *
1261 			 * At this moment we consider Subject address as
1262 			 * "belong to the node" if the Subject address equals
1263 			 * to the IPv6 destination address; validation for
1264 			 * IPv6 destination address should have done enough
1265 			 * check for us.
1266 			 *
1267 			 * We do not do proxy at this moment.
1268 			 */
1269 			/* m_pulldown instead of copy? */
1270 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1271 			    subjlen, (caddr_t)&in6_subj);
1272 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1273 				goto bad;
1274 
1275 			subj = (char *)&in6_subj;
1276 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1277 				break;
1278 
1279 			/*
1280 			 * XXX if we are to allow other cases, we should really
1281 			 * be careful about scope here.
1282 			 * basically, we should disallow queries toward IPv6
1283 			 * destination X with subject Y,
1284 			 * if scope(X) > scope(Y).
1285 			 * if we allow scope(X) > scope(Y), it will result in
1286 			 * information leakage across scope boundary.
1287 			 */
1288 			goto bad;
1289 
1290 		case ICMP6_NI_SUBJ_FQDN:
1291 			/*
1292 			 * Validate Subject name with gethostname(3).
1293 			 *
1294 			 * The behavior may need some debate, since:
1295 			 * - we are not sure if the node has FQDN as
1296 			 *   hostname (returned by gethostname(3)).
1297 			 * - the code does wildcard match for truncated names.
1298 			 *   however, we are not sure if we want to perform
1299 			 *   wildcard match, if gethostname(3) side has
1300 			 *   truncated hostname.
1301 			 */
1302 			mtx_lock(&hostname_mtx);
1303 			n = ni6_nametodns(V_hostname, hostnamelen, 0);
1304 			mtx_unlock(&hostname_mtx);
1305 			if (!n || n->m_next || n->m_len == 0)
1306 				goto bad;
1307 			IP6_EXTHDR_GET(subj, char *, m,
1308 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1309 			if (subj == NULL)
1310 				goto bad;
1311 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1312 			    n->m_len)) {
1313 				goto bad;
1314 			}
1315 			m_freem(n);
1316 			n = NULL;
1317 			break;
1318 
1319 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1320 		default:
1321 			goto bad;
1322 		}
1323 		break;
1324 	}
1325 
1326 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1327 	switch (qtype) {
1328 	case NI_QTYPE_FQDN:
1329 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1330 			goto bad;
1331 		break;
1332 	case NI_QTYPE_NODEADDR:
1333 	case NI_QTYPE_IPV4ADDR:
1334 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1335 			goto bad;
1336 		break;
1337 	}
1338 
1339 	/* guess reply length */
1340 	switch (qtype) {
1341 	case NI_QTYPE_NOOP:
1342 		break;		/* no reply data */
1343 	case NI_QTYPE_SUPTYPES:
1344 		replylen += sizeof(u_int32_t);
1345 		break;
1346 	case NI_QTYPE_FQDN:
1347 		/* XXX will append an mbuf */
1348 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1349 		break;
1350 	case NI_QTYPE_NODEADDR:
1351 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1352 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1353 		    sizeof(u_int32_t))) > MCLBYTES)
1354 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1355 		break;
1356 	case NI_QTYPE_IPV4ADDR:
1357 		/* unsupported - should respond with unknown Qtype? */
1358 		break;
1359 	default:
1360 		/*
1361 		 * XXX: We must return a reply with the ICMP6 code
1362 		 * `unknown Qtype' in this case.  However we regard the case
1363 		 * as an FQDN query for backward compatibility.
1364 		 * Older versions set a random value to this field,
1365 		 * so it rarely varies in the defined qtypes.
1366 		 * But the mechanism is not reliable...
1367 		 * maybe we should obsolete older versions.
1368 		 */
1369 		qtype = NI_QTYPE_FQDN;
1370 		/* XXX will append an mbuf */
1371 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1372 		oldfqdn++;
1373 		break;
1374 	}
1375 
1376 	/* allocate an mbuf to reply. */
1377 	MGETHDR(n, M_DONTWAIT, m->m_type);
1378 	if (n == NULL) {
1379 		m_freem(m);
1380 		return (NULL);
1381 	}
1382 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1383 	if (replylen > MHLEN) {
1384 		if (replylen > MCLBYTES) {
1385 			/*
1386 			 * XXX: should we try to allocate more? But MCLBYTES
1387 			 * is probably much larger than IPV6_MMTU...
1388 			 */
1389 			goto bad;
1390 		}
1391 		MCLGET(n, M_DONTWAIT);
1392 		if ((n->m_flags & M_EXT) == 0) {
1393 			goto bad;
1394 		}
1395 	}
1396 	n->m_pkthdr.len = n->m_len = replylen;
1397 
1398 	/* copy mbuf header and IPv6 + Node Information base headers */
1399 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1400 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1401 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1402 
1403 	/* qtype dependent procedure */
1404 	switch (qtype) {
1405 	case NI_QTYPE_NOOP:
1406 		nni6->ni_code = ICMP6_NI_SUCCESS;
1407 		nni6->ni_flags = 0;
1408 		break;
1409 	case NI_QTYPE_SUPTYPES:
1410 	{
1411 		u_int32_t v;
1412 		nni6->ni_code = ICMP6_NI_SUCCESS;
1413 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1414 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1415 		v = (u_int32_t)htonl(0x0000000f);
1416 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1417 		break;
1418 	}
1419 	case NI_QTYPE_FQDN:
1420 		nni6->ni_code = ICMP6_NI_SUCCESS;
1421 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1422 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1423 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1424 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1425 		/*
1426 		 * XXX do we really have FQDN in variable "hostname"?
1427 		 */
1428 		mtx_lock(&hostname_mtx);
1429 		n->m_next = ni6_nametodns(V_hostname, hostnamelen, oldfqdn);
1430 		mtx_unlock(&hostname_mtx);
1431 		if (n->m_next == NULL)
1432 			goto bad;
1433 		/* XXX we assume that n->m_next is not a chain */
1434 		if (n->m_next->m_next != NULL)
1435 			goto bad;
1436 		n->m_pkthdr.len += n->m_next->m_len;
1437 		break;
1438 	case NI_QTYPE_NODEADDR:
1439 	{
1440 		int lenlim, copied;
1441 
1442 		nni6->ni_code = ICMP6_NI_SUCCESS;
1443 		n->m_pkthdr.len = n->m_len =
1444 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1445 		lenlim = M_TRAILINGSPACE(n);
1446 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1447 		/* XXX: reset mbuf length */
1448 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1449 		    sizeof(struct icmp6_nodeinfo) + copied;
1450 		break;
1451 	}
1452 	default:
1453 		break;		/* XXX impossible! */
1454 	}
1455 
1456 	nni6->ni_type = ICMP6_NI_REPLY;
1457 	m_freem(m);
1458 	return (n);
1459 
1460   bad:
1461 	m_freem(m);
1462 	if (n)
1463 		m_freem(n);
1464 	return (NULL);
1465 }
1466 #undef hostnamelen
1467 
1468 /*
1469  * make a mbuf with DNS-encoded string.  no compression support.
1470  *
1471  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1472  * treated as truncated name (two \0 at the end).  this is a wild guess.
1473  *
1474  * old - return pascal string if non-zero
1475  */
1476 static struct mbuf *
1477 ni6_nametodns(const char *name, int namelen, int old)
1478 {
1479 	struct mbuf *m;
1480 	char *cp, *ep;
1481 	const char *p, *q;
1482 	int i, len, nterm;
1483 
1484 	if (old)
1485 		len = namelen + 1;
1486 	else
1487 		len = MCLBYTES;
1488 
1489 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1490 	MGET(m, M_DONTWAIT, MT_DATA);
1491 	if (m && len > MLEN) {
1492 		MCLGET(m, M_DONTWAIT);
1493 		if ((m->m_flags & M_EXT) == 0)
1494 			goto fail;
1495 	}
1496 	if (!m)
1497 		goto fail;
1498 	m->m_next = NULL;
1499 
1500 	if (old) {
1501 		m->m_len = len;
1502 		*mtod(m, char *) = namelen;
1503 		bcopy(name, mtod(m, char *) + 1, namelen);
1504 		return m;
1505 	} else {
1506 		m->m_len = 0;
1507 		cp = mtod(m, char *);
1508 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1509 
1510 		/* if not certain about my name, return empty buffer */
1511 		if (namelen == 0)
1512 			return m;
1513 
1514 		/*
1515 		 * guess if it looks like shortened hostname, or FQDN.
1516 		 * shortened hostname needs two trailing "\0".
1517 		 */
1518 		i = 0;
1519 		for (p = name; p < name + namelen; p++) {
1520 			if (*p && *p == '.')
1521 				i++;
1522 		}
1523 		if (i < 2)
1524 			nterm = 2;
1525 		else
1526 			nterm = 1;
1527 
1528 		p = name;
1529 		while (cp < ep && p < name + namelen) {
1530 			i = 0;
1531 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1532 				i++;
1533 			/* result does not fit into mbuf */
1534 			if (cp + i + 1 >= ep)
1535 				goto fail;
1536 			/*
1537 			 * DNS label length restriction, RFC1035 page 8.
1538 			 * "i == 0" case is included here to avoid returning
1539 			 * 0-length label on "foo..bar".
1540 			 */
1541 			if (i <= 0 || i >= 64)
1542 				goto fail;
1543 			*cp++ = i;
1544 			bcopy(p, cp, i);
1545 			cp += i;
1546 			p = q;
1547 			if (p < name + namelen && *p == '.')
1548 				p++;
1549 		}
1550 		/* termination */
1551 		if (cp + nterm >= ep)
1552 			goto fail;
1553 		while (nterm-- > 0)
1554 			*cp++ = '\0';
1555 		m->m_len = cp - mtod(m, char *);
1556 		return m;
1557 	}
1558 
1559 	panic("should not reach here");
1560 	/* NOTREACHED */
1561 
1562  fail:
1563 	if (m)
1564 		m_freem(m);
1565 	return NULL;
1566 }
1567 
1568 /*
1569  * check if two DNS-encoded string matches.  takes care of truncated
1570  * form (with \0\0 at the end).  no compression support.
1571  * XXX upper/lowercase match (see RFC2065)
1572  */
1573 static int
1574 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1575 {
1576 	const char *a0, *b0;
1577 	int l;
1578 
1579 	/* simplest case - need validation? */
1580 	if (alen == blen && bcmp(a, b, alen) == 0)
1581 		return 1;
1582 
1583 	a0 = a;
1584 	b0 = b;
1585 
1586 	/* termination is mandatory */
1587 	if (alen < 2 || blen < 2)
1588 		return 0;
1589 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1590 		return 0;
1591 	alen--;
1592 	blen--;
1593 
1594 	while (a - a0 < alen && b - b0 < blen) {
1595 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1596 			return 0;
1597 
1598 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1599 			return 0;
1600 		/* we don't support compression yet */
1601 		if (a[0] >= 64 || b[0] >= 64)
1602 			return 0;
1603 
1604 		/* truncated case */
1605 		if (a[0] == 0 && a - a0 == alen - 1)
1606 			return 1;
1607 		if (b[0] == 0 && b - b0 == blen - 1)
1608 			return 1;
1609 		if (a[0] == 0 || b[0] == 0)
1610 			return 0;
1611 
1612 		if (a[0] != b[0])
1613 			return 0;
1614 		l = a[0];
1615 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1616 			return 0;
1617 		if (bcmp(a + 1, b + 1, l) != 0)
1618 			return 0;
1619 
1620 		a += 1 + l;
1621 		b += 1 + l;
1622 	}
1623 
1624 	if (a - a0 == alen && b - b0 == blen)
1625 		return 1;
1626 	else
1627 		return 0;
1628 }
1629 
1630 /*
1631  * calculate the number of addresses to be returned in the node info reply.
1632  */
1633 static int
1634 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1635     struct in6_addr *subj)
1636 {
1637 	struct ifnet *ifp;
1638 	struct in6_ifaddr *ifa6;
1639 	struct ifaddr *ifa;
1640 	int addrs = 0, addrsofif, iffound = 0;
1641 	int niflags = ni6->ni_flags;
1642 
1643 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1644 		switch (ni6->ni_code) {
1645 		case ICMP6_NI_SUBJ_IPV6:
1646 			if (subj == NULL) /* must be impossible... */
1647 				return (0);
1648 			break;
1649 		default:
1650 			/*
1651 			 * XXX: we only support IPv6 subject address for
1652 			 * this Qtype.
1653 			 */
1654 			return (0);
1655 		}
1656 	}
1657 
1658 	IFNET_RLOCK();
1659 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1660 		addrsofif = 0;
1661 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1662 			if (ifa->ifa_addr->sa_family != AF_INET6)
1663 				continue;
1664 			ifa6 = (struct in6_ifaddr *)ifa;
1665 
1666 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1667 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1668 				iffound = 1;
1669 
1670 			/*
1671 			 * IPv4-mapped addresses can only be returned by a
1672 			 * Node Information proxy, since they represent
1673 			 * addresses of IPv4-only nodes, which perforce do
1674 			 * not implement this protocol.
1675 			 * [icmp-name-lookups-07, Section 5.4]
1676 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1677 			 * this function at this moment.
1678 			 */
1679 
1680 			/* What do we have to do about ::1? */
1681 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1682 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1683 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1684 					continue;
1685 				break;
1686 			case IPV6_ADDR_SCOPE_SITELOCAL:
1687 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1688 					continue;
1689 				break;
1690 			case IPV6_ADDR_SCOPE_GLOBAL:
1691 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1692 					continue;
1693 				break;
1694 			default:
1695 				continue;
1696 			}
1697 
1698 			/*
1699 			 * check if anycast is okay.
1700 			 * XXX: just experimental.  not in the spec.
1701 			 */
1702 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1703 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1704 				continue; /* we need only unicast addresses */
1705 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1706 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1707 				continue;
1708 			}
1709 			addrsofif++; /* count the address */
1710 		}
1711 		if (iffound) {
1712 			*ifpp = ifp;
1713 			IFNET_RUNLOCK();
1714 			return (addrsofif);
1715 		}
1716 
1717 		addrs += addrsofif;
1718 	}
1719 	IFNET_RUNLOCK();
1720 
1721 	return (addrs);
1722 }
1723 
1724 static int
1725 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1726     struct ifnet *ifp0, int resid)
1727 {
1728 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1729 	struct in6_ifaddr *ifa6;
1730 	struct ifaddr *ifa;
1731 	struct ifnet *ifp_dep = NULL;
1732 	int copied = 0, allow_deprecated = 0;
1733 	u_char *cp = (u_char *)(nni6 + 1);
1734 	int niflags = ni6->ni_flags;
1735 	u_int32_t ltime;
1736 
1737 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1738 		return (0);	/* needless to copy */
1739 
1740 	IFNET_RLOCK();
1741   again:
1742 
1743 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1744 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1745 		     ifa = ifa->ifa_list.tqe_next) {
1746 			if (ifa->ifa_addr->sa_family != AF_INET6)
1747 				continue;
1748 			ifa6 = (struct in6_ifaddr *)ifa;
1749 
1750 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1751 			    allow_deprecated == 0) {
1752 				/*
1753 				 * prefererred address should be put before
1754 				 * deprecated addresses.
1755 				 */
1756 
1757 				/* record the interface for later search */
1758 				if (ifp_dep == NULL)
1759 					ifp_dep = ifp;
1760 
1761 				continue;
1762 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1763 			    allow_deprecated != 0)
1764 				continue; /* we now collect deprecated addrs */
1765 
1766 			/* What do we have to do about ::1? */
1767 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1768 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1769 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1770 					continue;
1771 				break;
1772 			case IPV6_ADDR_SCOPE_SITELOCAL:
1773 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1774 					continue;
1775 				break;
1776 			case IPV6_ADDR_SCOPE_GLOBAL:
1777 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1778 					continue;
1779 				break;
1780 			default:
1781 				continue;
1782 			}
1783 
1784 			/*
1785 			 * check if anycast is okay.
1786 			 * XXX: just experimental.  not in the spec.
1787 			 */
1788 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1789 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1790 				continue;
1791 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1792 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1793 				continue;
1794 			}
1795 
1796 			/* now we can copy the address */
1797 			if (resid < sizeof(struct in6_addr) +
1798 			    sizeof(u_int32_t)) {
1799 				/*
1800 				 * We give up much more copy.
1801 				 * Set the truncate flag and return.
1802 				 */
1803 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1804 				IFNET_RUNLOCK();
1805 				return (copied);
1806 			}
1807 
1808 			/*
1809 			 * Set the TTL of the address.
1810 			 * The TTL value should be one of the following
1811 			 * according to the specification:
1812 			 *
1813 			 * 1. The remaining lifetime of a DHCP lease on the
1814 			 *    address, or
1815 			 * 2. The remaining Valid Lifetime of a prefix from
1816 			 *    which the address was derived through Stateless
1817 			 *    Autoconfiguration.
1818 			 *
1819 			 * Note that we currently do not support stateful
1820 			 * address configuration by DHCPv6, so the former
1821 			 * case can't happen.
1822 			 */
1823 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1824 				ltime = ND6_INFINITE_LIFETIME;
1825 			else {
1826 				if (ifa6->ia6_lifetime.ia6t_expire >
1827 				    time_second)
1828 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1829 				else
1830 					ltime = 0;
1831 			}
1832 
1833 			bcopy(&ltime, cp, sizeof(u_int32_t));
1834 			cp += sizeof(u_int32_t);
1835 
1836 			/* copy the address itself */
1837 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1838 			    sizeof(struct in6_addr));
1839 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1840 			cp += sizeof(struct in6_addr);
1841 
1842 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1843 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1844 		}
1845 		if (ifp0)	/* we need search only on the specified IF */
1846 			break;
1847 	}
1848 
1849 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1850 		ifp = ifp_dep;
1851 		allow_deprecated = 1;
1852 
1853 		goto again;
1854 	}
1855 
1856 	IFNET_RUNLOCK();
1857 
1858 	return (copied);
1859 }
1860 
1861 /*
1862  * XXX almost dup'ed code with rip6_input.
1863  */
1864 static int
1865 icmp6_rip6_input(struct mbuf **mp, int off)
1866 {
1867 	struct mbuf *m = *mp;
1868 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1869 	struct in6pcb *in6p;
1870 	struct in6pcb *last = NULL;
1871 	struct sockaddr_in6 fromsa;
1872 	struct icmp6_hdr *icmp6;
1873 	struct mbuf *opts = NULL;
1874 
1875 #ifndef PULLDOWN_TEST
1876 	/* this is assumed to be safe. */
1877 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1878 #else
1879 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1880 	if (icmp6 == NULL) {
1881 		/* m is already reclaimed */
1882 		return (IPPROTO_DONE);
1883 	}
1884 #endif
1885 
1886 	/*
1887 	 * XXX: the address may have embedded scope zone ID, which should be
1888 	 * hidden from applications.
1889 	 */
1890 	bzero(&fromsa, sizeof(fromsa));
1891 	fromsa.sin6_family = AF_INET6;
1892 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1893 	fromsa.sin6_addr = ip6->ip6_src;
1894 	if (sa6_recoverscope(&fromsa)) {
1895 		m_freem(m);
1896 		return (IPPROTO_DONE);
1897 	}
1898 
1899 	INP_INFO_RLOCK(&V_ripcbinfo);
1900 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1901 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1902 			continue;
1903 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1904 			continue;
1905 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1906 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1907 			continue;
1908 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1909 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1910 			continue;
1911 		INP_RLOCK(in6p);
1912 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1913 		    in6p->in6p_icmp6filt)) {
1914 			INP_RUNLOCK(in6p);
1915 			continue;
1916 		}
1917 		if (last) {
1918 			struct	mbuf *n = NULL;
1919 
1920 			/*
1921 			 * Recent network drivers tend to allocate a single
1922 			 * mbuf cluster, rather than to make a couple of
1923 			 * mbufs without clusters.  Also, since the IPv6 code
1924 			 * path tries to avoid m_pullup(), it is highly
1925 			 * probable that we still have an mbuf cluster here
1926 			 * even though the necessary length can be stored in an
1927 			 * mbuf's internal buffer.
1928 			 * Meanwhile, the default size of the receive socket
1929 			 * buffer for raw sockets is not so large.  This means
1930 			 * the possibility of packet loss is relatively higher
1931 			 * than before.  To avoid this scenario, we copy the
1932 			 * received data to a separate mbuf that does not use
1933 			 * a cluster, if possible.
1934 			 * XXX: it is better to copy the data after stripping
1935 			 * intermediate headers.
1936 			 */
1937 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1938 			    m->m_len <= MHLEN) {
1939 				MGET(n, M_DONTWAIT, m->m_type);
1940 				if (n != NULL) {
1941 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1942 						bcopy(m->m_data, n->m_data,
1943 						      m->m_len);
1944 						n->m_len = m->m_len;
1945 					} else {
1946 						m_free(n);
1947 						n = NULL;
1948 					}
1949 				}
1950 			}
1951 			if (n != NULL ||
1952 			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1953 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1954 					ip6_savecontrol(last, n, &opts);
1955 				/* strip intermediate headers */
1956 				m_adj(n, off);
1957 				SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
1958 				if (sbappendaddr_locked(
1959 				    &last->in6p_socket->so_rcv,
1960 				    (struct sockaddr *)&fromsa, n, opts)
1961 				    == 0) {
1962 					/* should notify about lost packet */
1963 					m_freem(n);
1964 					if (opts) {
1965 						m_freem(opts);
1966 					}
1967 					SOCKBUF_UNLOCK(
1968 					    &last->in6p_socket->so_rcv);
1969 				} else
1970 					sorwakeup_locked(last->in6p_socket);
1971 				opts = NULL;
1972 			}
1973 			INP_RUNLOCK(last);
1974 		}
1975 		last = in6p;
1976 	}
1977 	INP_INFO_RUNLOCK(&V_ripcbinfo);
1978 	if (last) {
1979 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1980 			ip6_savecontrol(last, m, &opts);
1981 		/* strip intermediate headers */
1982 		m_adj(m, off);
1983 
1984 		/* avoid using mbuf clusters if possible (see above) */
1985 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1986 		    m->m_len <= MHLEN) {
1987 			struct mbuf *n;
1988 
1989 			MGET(n, M_DONTWAIT, m->m_type);
1990 			if (n != NULL) {
1991 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1992 					bcopy(m->m_data, n->m_data, m->m_len);
1993 					n->m_len = m->m_len;
1994 
1995 					m_freem(m);
1996 					m = n;
1997 				} else {
1998 					m_freem(n);
1999 					n = NULL;
2000 				}
2001 			}
2002 		}
2003 		SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
2004 		if (sbappendaddr_locked(&last->in6p_socket->so_rcv,
2005 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2006 			m_freem(m);
2007 			if (opts)
2008 				m_freem(opts);
2009 			SOCKBUF_UNLOCK(&last->in6p_socket->so_rcv);
2010 		} else
2011 			sorwakeup_locked(last->in6p_socket);
2012 		INP_RUNLOCK(last);
2013 	} else {
2014 		m_freem(m);
2015 		V_ip6stat.ip6s_delivered--;
2016 	}
2017 	return IPPROTO_DONE;
2018 }
2019 
2020 /*
2021  * Reflect the ip6 packet back to the source.
2022  * OFF points to the icmp6 header, counted from the top of the mbuf.
2023  */
2024 void
2025 icmp6_reflect(struct mbuf *m, size_t off)
2026 {
2027 	struct ip6_hdr *ip6;
2028 	struct icmp6_hdr *icmp6;
2029 	struct in6_ifaddr *ia;
2030 	int plen;
2031 	int type, code;
2032 	struct ifnet *outif = NULL;
2033 	struct in6_addr origdst, *src = NULL;
2034 
2035 	/* too short to reflect */
2036 	if (off < sizeof(struct ip6_hdr)) {
2037 		nd6log((LOG_DEBUG,
2038 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2039 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2040 		    __FILE__, __LINE__));
2041 		goto bad;
2042 	}
2043 
2044 	/*
2045 	 * If there are extra headers between IPv6 and ICMPv6, strip
2046 	 * off that header first.
2047 	 */
2048 #ifdef DIAGNOSTIC
2049 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2050 		panic("assumption failed in icmp6_reflect");
2051 #endif
2052 	if (off > sizeof(struct ip6_hdr)) {
2053 		size_t l;
2054 		struct ip6_hdr nip6;
2055 
2056 		l = off - sizeof(struct ip6_hdr);
2057 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2058 		m_adj(m, l);
2059 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2060 		if (m->m_len < l) {
2061 			if ((m = m_pullup(m, l)) == NULL)
2062 				return;
2063 		}
2064 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2065 	} else /* off == sizeof(struct ip6_hdr) */ {
2066 		size_t l;
2067 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2068 		if (m->m_len < l) {
2069 			if ((m = m_pullup(m, l)) == NULL)
2070 				return;
2071 		}
2072 	}
2073 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2074 	ip6 = mtod(m, struct ip6_hdr *);
2075 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2076 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2077 	type = icmp6->icmp6_type; /* keep type for statistics */
2078 	code = icmp6->icmp6_code; /* ditto. */
2079 
2080 	origdst = ip6->ip6_dst;
2081 	/*
2082 	 * ip6_input() drops a packet if its src is multicast.
2083 	 * So, the src is never multicast.
2084 	 */
2085 	ip6->ip6_dst = ip6->ip6_src;
2086 
2087 	/*
2088 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2089 	 * use dst as the src for the reply.
2090 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2091 	 * (for example) when we encounter an error while forwarding procedure
2092 	 * destined to a duplicated address of ours.
2093 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2094 	 * procedure of an outgoing packet of our own, in which case we need
2095 	 * to search in the ifaddr list.
2096 	 */
2097 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2098 		if ((ia = ip6_getdstifaddr(m))) {
2099 			if (!(ia->ia6_flags &
2100 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2101 				src = &ia->ia_addr.sin6_addr;
2102 		} else {
2103 			struct sockaddr_in6 d;
2104 
2105 			bzero(&d, sizeof(d));
2106 			d.sin6_family = AF_INET6;
2107 			d.sin6_len = sizeof(d);
2108 			d.sin6_addr = origdst;
2109 			ia = (struct in6_ifaddr *)
2110 			    ifa_ifwithaddr((struct sockaddr *)&d);
2111 			if (ia &&
2112 			    !(ia->ia6_flags &
2113 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2114 				src = &ia->ia_addr.sin6_addr;
2115 			}
2116 		}
2117 	}
2118 
2119 	if (src == NULL) {
2120 		int e;
2121 		struct sockaddr_in6 sin6;
2122 		struct route_in6 ro;
2123 
2124 		/*
2125 		 * This case matches to multicasts, our anycast, or unicasts
2126 		 * that we do not own.  Select a source address based on the
2127 		 * source address of the erroneous packet.
2128 		 */
2129 		bzero(&sin6, sizeof(sin6));
2130 		sin6.sin6_family = AF_INET6;
2131 		sin6.sin6_len = sizeof(sin6);
2132 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2133 
2134 		bzero(&ro, sizeof(ro));
2135 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2136 		if (ro.ro_rt)
2137 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2138 		if (src == NULL) {
2139 			char ip6buf[INET6_ADDRSTRLEN];
2140 			nd6log((LOG_DEBUG,
2141 			    "icmp6_reflect: source can't be determined: "
2142 			    "dst=%s, error=%d\n",
2143 			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2144 			goto bad;
2145 		}
2146 	}
2147 
2148 	ip6->ip6_src = *src;
2149 	ip6->ip6_flow = 0;
2150 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2151 	ip6->ip6_vfc |= IPV6_VERSION;
2152 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2153 	if (outif)
2154 		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2155 	else if (m->m_pkthdr.rcvif) {
2156 		/* XXX: This may not be the outgoing interface */
2157 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2158 	} else
2159 		ip6->ip6_hlim = V_ip6_defhlim;
2160 
2161 	icmp6->icmp6_cksum = 0;
2162 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2163 	    sizeof(struct ip6_hdr), plen);
2164 
2165 	/*
2166 	 * XXX option handling
2167 	 */
2168 
2169 	m->m_flags &= ~(M_BCAST|M_MCAST);
2170 
2171 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2172 	if (outif)
2173 		icmp6_ifoutstat_inc(outif, type, code);
2174 
2175 	return;
2176 
2177  bad:
2178 	m_freem(m);
2179 	return;
2180 }
2181 
2182 void
2183 icmp6_fasttimo(void)
2184 {
2185 
2186 	return;
2187 }
2188 
2189 static const char *
2190 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2191     struct in6_addr *tgt6)
2192 {
2193 	static char buf[1024];
2194 	char ip6bufs[INET6_ADDRSTRLEN];
2195 	char ip6bufd[INET6_ADDRSTRLEN];
2196 	char ip6buft[INET6_ADDRSTRLEN];
2197 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2198 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2199 	    ip6_sprintf(ip6buft, tgt6));
2200 	return buf;
2201 }
2202 
2203 void
2204 icmp6_redirect_input(struct mbuf *m, int off)
2205 {
2206 	struct ifnet *ifp;
2207 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2208 	struct nd_redirect *nd_rd;
2209 	int icmp6len = ntohs(ip6->ip6_plen);
2210 	char *lladdr = NULL;
2211 	int lladdrlen = 0;
2212 	u_char *redirhdr = NULL;
2213 	int redirhdrlen = 0;
2214 	struct rtentry *rt = NULL;
2215 	int is_router;
2216 	int is_onlink;
2217 	struct in6_addr src6 = ip6->ip6_src;
2218 	struct in6_addr redtgt6;
2219 	struct in6_addr reddst6;
2220 	union nd_opts ndopts;
2221 	char ip6buf[INET6_ADDRSTRLEN];
2222 
2223 	if (!m)
2224 		return;
2225 
2226 	ifp = m->m_pkthdr.rcvif;
2227 
2228 	if (!ifp)
2229 		return;
2230 
2231 	/* XXX if we are router, we don't update route by icmp6 redirect */
2232 	if (V_ip6_forwarding)
2233 		goto freeit;
2234 	if (!V_icmp6_rediraccept)
2235 		goto freeit;
2236 
2237 #ifndef PULLDOWN_TEST
2238 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2239 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2240 #else
2241 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2242 	if (nd_rd == NULL) {
2243 		V_icmp6stat.icp6s_tooshort++;
2244 		return;
2245 	}
2246 #endif
2247 	redtgt6 = nd_rd->nd_rd_target;
2248 	reddst6 = nd_rd->nd_rd_dst;
2249 
2250 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2251 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2252 		goto freeit;
2253 	}
2254 
2255 	/* validation */
2256 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2257 		nd6log((LOG_ERR,
2258 		    "ICMP6 redirect sent from %s rejected; "
2259 		    "must be from linklocal\n",
2260 		    ip6_sprintf(ip6buf, &src6)));
2261 		goto bad;
2262 	}
2263 	if (ip6->ip6_hlim != 255) {
2264 		nd6log((LOG_ERR,
2265 		    "ICMP6 redirect sent from %s rejected; "
2266 		    "hlim=%d (must be 255)\n",
2267 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2268 		goto bad;
2269 	}
2270     {
2271 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2272 	struct sockaddr_in6 sin6;
2273 	struct in6_addr *gw6;
2274 
2275 	bzero(&sin6, sizeof(sin6));
2276 	sin6.sin6_family = AF_INET6;
2277 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2278 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2279 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2280 	if (rt) {
2281 		if (rt->rt_gateway == NULL ||
2282 		    rt->rt_gateway->sa_family != AF_INET6) {
2283 			nd6log((LOG_ERR,
2284 			    "ICMP6 redirect rejected; no route "
2285 			    "with inet6 gateway found for redirect dst: %s\n",
2286 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2287 			RTFREE_LOCKED(rt);
2288 			goto bad;
2289 		}
2290 
2291 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2292 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2293 			nd6log((LOG_ERR,
2294 			    "ICMP6 redirect rejected; "
2295 			    "not equal to gw-for-src=%s (must be same): "
2296 			    "%s\n",
2297 			    ip6_sprintf(ip6buf, gw6),
2298 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299 			RTFREE_LOCKED(rt);
2300 			goto bad;
2301 		}
2302 	} else {
2303 		nd6log((LOG_ERR,
2304 		    "ICMP6 redirect rejected; "
2305 		    "no route found for redirect dst: %s\n",
2306 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2307 		goto bad;
2308 	}
2309 	RTFREE_LOCKED(rt);
2310 	rt = NULL;
2311     }
2312 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2313 		nd6log((LOG_ERR,
2314 		    "ICMP6 redirect rejected; "
2315 		    "redirect dst must be unicast: %s\n",
2316 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2317 		goto bad;
2318 	}
2319 
2320 	is_router = is_onlink = 0;
2321 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2322 		is_router = 1;	/* router case */
2323 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2324 		is_onlink = 1;	/* on-link destination case */
2325 	if (!is_router && !is_onlink) {
2326 		nd6log((LOG_ERR,
2327 		    "ICMP6 redirect rejected; "
2328 		    "neither router case nor onlink case: %s\n",
2329 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2330 		goto bad;
2331 	}
2332 	/* validation passed */
2333 
2334 	icmp6len -= sizeof(*nd_rd);
2335 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2336 	if (nd6_options(&ndopts) < 0) {
2337 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2338 		    "invalid ND option, rejected: %s\n",
2339 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2340 		/* nd6_options have incremented stats */
2341 		goto freeit;
2342 	}
2343 
2344 	if (ndopts.nd_opts_tgt_lladdr) {
2345 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2346 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2347 	}
2348 
2349 	if (ndopts.nd_opts_rh) {
2350 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2351 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2352 	}
2353 
2354 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2355 		nd6log((LOG_INFO,
2356 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2357 		    "(if %d, icmp6 packet %d): %s\n",
2358 		    ip6_sprintf(ip6buf, &redtgt6),
2359 		    ifp->if_addrlen, lladdrlen - 2,
2360 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2361 		goto bad;
2362 	}
2363 
2364 	/* RFC 2461 8.3 */
2365 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2366 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2367 
2368 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2369 		/* perform rtredirect */
2370 		struct sockaddr_in6 sdst;
2371 		struct sockaddr_in6 sgw;
2372 		struct sockaddr_in6 ssrc;
2373 
2374 		bzero(&sdst, sizeof(sdst));
2375 		bzero(&sgw, sizeof(sgw));
2376 		bzero(&ssrc, sizeof(ssrc));
2377 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2378 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2379 			sizeof(struct sockaddr_in6);
2380 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2381 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2382 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2383 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2384 		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2385 		    (struct sockaddr *)&ssrc);
2386 	}
2387 	/* finally update cached route in each socket via pfctlinput */
2388     {
2389 	struct sockaddr_in6 sdst;
2390 
2391 	bzero(&sdst, sizeof(sdst));
2392 	sdst.sin6_family = AF_INET6;
2393 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2394 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2395 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2396 #ifdef IPSEC
2397 	key_sa_routechange((struct sockaddr *)&sdst);
2398 #endif /* IPSEC */
2399     }
2400 
2401  freeit:
2402 	m_freem(m);
2403 	return;
2404 
2405  bad:
2406 	V_icmp6stat.icp6s_badredirect++;
2407 	m_freem(m);
2408 }
2409 
2410 void
2411 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2412 {
2413 	struct ifnet *ifp;	/* my outgoing interface */
2414 	struct in6_addr *ifp_ll6;
2415 	struct in6_addr *router_ll6;
2416 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2417 	struct mbuf *m = NULL;	/* newly allocated one */
2418 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2419 	struct nd_redirect *nd_rd;
2420 	size_t maxlen;
2421 	u_char *p;
2422 	struct ifnet *outif = NULL;
2423 	struct sockaddr_in6 src_sa;
2424 
2425 	icmp6_errcount(&V_icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2426 
2427 	/* if we are not router, we don't send icmp6 redirect */
2428 	if (!V_ip6_forwarding)
2429 		goto fail;
2430 
2431 	/* sanity check */
2432 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2433 		goto fail;
2434 
2435 	/*
2436 	 * Address check:
2437 	 *  the source address must identify a neighbor, and
2438 	 *  the destination address must not be a multicast address
2439 	 *  [RFC 2461, sec 8.2]
2440 	 */
2441 	sip6 = mtod(m0, struct ip6_hdr *);
2442 	bzero(&src_sa, sizeof(src_sa));
2443 	src_sa.sin6_family = AF_INET6;
2444 	src_sa.sin6_len = sizeof(src_sa);
2445 	src_sa.sin6_addr = sip6->ip6_src;
2446 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2447 		goto fail;
2448 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2449 		goto fail;	/* what should we do here? */
2450 
2451 	/* rate limit */
2452 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2453 		goto fail;
2454 
2455 	/*
2456 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2457 	 * we almost always ask for an mbuf cluster for simplicity.
2458 	 * (MHLEN < IPV6_MMTU is almost always true)
2459 	 */
2460 #if IPV6_MMTU >= MCLBYTES
2461 # error assumption failed about IPV6_MMTU and MCLBYTES
2462 #endif
2463 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2464 	if (m && IPV6_MMTU >= MHLEN)
2465 		MCLGET(m, M_DONTWAIT);
2466 	if (!m)
2467 		goto fail;
2468 	m->m_pkthdr.rcvif = NULL;
2469 	m->m_len = 0;
2470 	maxlen = M_TRAILINGSPACE(m);
2471 	maxlen = min(IPV6_MMTU, maxlen);
2472 	/* just for safety */
2473 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2474 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2475 		goto fail;
2476 	}
2477 
2478 	{
2479 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2480 		struct in6_ifaddr *ia;
2481 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2482 						 IN6_IFF_NOTREADY|
2483 						 IN6_IFF_ANYCAST)) == NULL)
2484 			goto fail;
2485 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2486 	}
2487 
2488 	/* get ip6 linklocal address for the router. */
2489 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2490 		struct sockaddr_in6 *sin6;
2491 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2492 		router_ll6 = &sin6->sin6_addr;
2493 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2494 			router_ll6 = (struct in6_addr *)NULL;
2495 	} else
2496 		router_ll6 = (struct in6_addr *)NULL;
2497 
2498 	/* ip6 */
2499 	ip6 = mtod(m, struct ip6_hdr *);
2500 	ip6->ip6_flow = 0;
2501 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2502 	ip6->ip6_vfc |= IPV6_VERSION;
2503 	/* ip6->ip6_plen will be set later */
2504 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2505 	ip6->ip6_hlim = 255;
2506 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2507 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2508 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2509 
2510 	/* ND Redirect */
2511 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2512 	nd_rd->nd_rd_type = ND_REDIRECT;
2513 	nd_rd->nd_rd_code = 0;
2514 	nd_rd->nd_rd_reserved = 0;
2515 	if (rt->rt_flags & RTF_GATEWAY) {
2516 		/*
2517 		 * nd_rd->nd_rd_target must be a link-local address in
2518 		 * better router cases.
2519 		 */
2520 		if (!router_ll6)
2521 			goto fail;
2522 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2523 		    sizeof(nd_rd->nd_rd_target));
2524 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2525 		    sizeof(nd_rd->nd_rd_dst));
2526 	} else {
2527 		/* make sure redtgt == reddst */
2528 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2529 		    sizeof(nd_rd->nd_rd_target));
2530 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2531 		    sizeof(nd_rd->nd_rd_dst));
2532 	}
2533 
2534 	p = (u_char *)(nd_rd + 1);
2535 
2536 	if (!router_ll6)
2537 		goto nolladdropt;
2538 
2539 	{
2540 		/* target lladdr option */
2541 		struct rtentry *rt_router = NULL;
2542 		int len;
2543 		struct sockaddr_dl *sdl;
2544 		struct nd_opt_hdr *nd_opt;
2545 		char *lladdr;
2546 
2547 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2548 		if (!rt_router)
2549 			goto nolladdropt;
2550 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2551 		len = (len + 7) & ~7;	/* round by 8 */
2552 		/* safety check */
2553 		if (len + (p - (u_char *)ip6) > maxlen)
2554 			goto nolladdropt;
2555 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2556 		    (rt_router->rt_flags & RTF_LLINFO) &&
2557 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2558 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2559 		    sdl->sdl_alen) {
2560 			nd_opt = (struct nd_opt_hdr *)p;
2561 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2562 			nd_opt->nd_opt_len = len >> 3;
2563 			lladdr = (char *)(nd_opt + 1);
2564 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2565 			p += len;
2566 		}
2567 	}
2568 nolladdropt:;
2569 
2570 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2571 
2572 	/* just to be safe */
2573 #ifdef M_DECRYPTED	/*not openbsd*/
2574 	if (m0->m_flags & M_DECRYPTED)
2575 		goto noredhdropt;
2576 #endif
2577 	if (p - (u_char *)ip6 > maxlen)
2578 		goto noredhdropt;
2579 
2580 	{
2581 		/* redirected header option */
2582 		int len;
2583 		struct nd_opt_rd_hdr *nd_opt_rh;
2584 
2585 		/*
2586 		 * compute the maximum size for icmp6 redirect header option.
2587 		 * XXX room for auth header?
2588 		 */
2589 		len = maxlen - (p - (u_char *)ip6);
2590 		len &= ~7;
2591 
2592 		/* This is just for simplicity. */
2593 		if (m0->m_pkthdr.len != m0->m_len) {
2594 			if (m0->m_next) {
2595 				m_freem(m0->m_next);
2596 				m0->m_next = NULL;
2597 			}
2598 			m0->m_pkthdr.len = m0->m_len;
2599 		}
2600 
2601 		/*
2602 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2603 		 * about padding/truncate rule for the original IP packet.
2604 		 * From the discussion on IPv6imp in Feb 1999,
2605 		 * the consensus was:
2606 		 * - "attach as much as possible" is the goal
2607 		 * - pad if not aligned (original size can be guessed by
2608 		 *   original ip6 header)
2609 		 * Following code adds the padding if it is simple enough,
2610 		 * and truncates if not.
2611 		 */
2612 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2613 			panic("assumption failed in %s:%d", __FILE__,
2614 			    __LINE__);
2615 
2616 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2617 			/* not enough room, truncate */
2618 			m0->m_pkthdr.len = m0->m_len = len -
2619 			    sizeof(*nd_opt_rh);
2620 		} else {
2621 			/* enough room, pad or truncate */
2622 			size_t extra;
2623 
2624 			extra = m0->m_pkthdr.len % 8;
2625 			if (extra) {
2626 				/* pad if easy enough, truncate if not */
2627 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2628 					/* pad */
2629 					m0->m_len += (8 - extra);
2630 					m0->m_pkthdr.len += (8 - extra);
2631 				} else {
2632 					/* truncate */
2633 					m0->m_pkthdr.len -= extra;
2634 					m0->m_len -= extra;
2635 				}
2636 			}
2637 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2638 			m0->m_pkthdr.len = m0->m_len = len -
2639 			    sizeof(*nd_opt_rh);
2640 		}
2641 
2642 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2643 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2644 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2645 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2646 		p += sizeof(*nd_opt_rh);
2647 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2648 
2649 		/* connect m0 to m */
2650 		m_tag_delete_chain(m0, NULL);
2651 		m0->m_flags &= ~M_PKTHDR;
2652 		m->m_next = m0;
2653 		m->m_pkthdr.len = m->m_len + m0->m_len;
2654 		m0 = NULL;
2655 	}
2656 noredhdropt:;
2657 	if (m0) {
2658 		m_freem(m0);
2659 		m0 = NULL;
2660 	}
2661 
2662 	/* XXX: clear embedded link IDs in the inner header */
2663 	in6_clearscope(&sip6->ip6_src);
2664 	in6_clearscope(&sip6->ip6_dst);
2665 	in6_clearscope(&nd_rd->nd_rd_target);
2666 	in6_clearscope(&nd_rd->nd_rd_dst);
2667 
2668 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2669 
2670 	nd_rd->nd_rd_cksum = 0;
2671 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2672 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2673 
2674 	/* send the packet to outside... */
2675 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2676 	if (outif) {
2677 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2678 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2679 	}
2680 	V_icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2681 
2682 	return;
2683 
2684 fail:
2685 	if (m)
2686 		m_freem(m);
2687 	if (m0)
2688 		m_freem(m0);
2689 }
2690 
2691 /*
2692  * ICMPv6 socket option processing.
2693  */
2694 int
2695 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2696 {
2697 	int error = 0;
2698 	int optlen;
2699 	struct inpcb *inp = sotoinpcb(so);
2700 	int level, op, optname;
2701 
2702 	if (sopt) {
2703 		level = sopt->sopt_level;
2704 		op = sopt->sopt_dir;
2705 		optname = sopt->sopt_name;
2706 		optlen = sopt->sopt_valsize;
2707 	} else
2708 		level = op = optname = optlen = 0;
2709 
2710 	if (level != IPPROTO_ICMPV6) {
2711 		return EINVAL;
2712 	}
2713 
2714 	switch (op) {
2715 	case PRCO_SETOPT:
2716 		switch (optname) {
2717 		case ICMP6_FILTER:
2718 		    {
2719 			struct icmp6_filter ic6f;
2720 
2721 			if (optlen != sizeof(ic6f)) {
2722 				error = EMSGSIZE;
2723 				break;
2724 			}
2725 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2726 			if (error == 0) {
2727 				INP_WLOCK(inp);
2728 				*inp->in6p_icmp6filt = ic6f;
2729 				INP_WUNLOCK(inp);
2730 			}
2731 			break;
2732 		    }
2733 
2734 		default:
2735 			error = ENOPROTOOPT;
2736 			break;
2737 		}
2738 		break;
2739 
2740 	case PRCO_GETOPT:
2741 		switch (optname) {
2742 		case ICMP6_FILTER:
2743 		    {
2744 			struct icmp6_filter ic6f;
2745 
2746 			INP_RLOCK(inp);
2747 			ic6f = *inp->in6p_icmp6filt;
2748 			INP_RUNLOCK(inp);
2749 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2750 			break;
2751 		    }
2752 
2753 		default:
2754 			error = ENOPROTOOPT;
2755 			break;
2756 		}
2757 		break;
2758 	}
2759 
2760 	return (error);
2761 }
2762 
2763 /*
2764  * Perform rate limit check.
2765  * Returns 0 if it is okay to send the icmp6 packet.
2766  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2767  * limitation.
2768  *
2769  * XXX per-destination/type check necessary?
2770  *
2771  * dst - not used at this moment
2772  * type - not used at this moment
2773  * code - not used at this moment
2774  */
2775 static int
2776 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2777     const int code)
2778 {
2779 	int ret;
2780 
2781 	ret = 0;	/* okay to send */
2782 
2783 	/* PPS limit */
2784 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2785 	    V_icmp6errppslim)) {
2786 		/* The packet is subject to rate limit */
2787 		ret++;
2788 	}
2789 
2790 	return ret;
2791 }
2792