xref: /freebsd/sys/netinet6/icmp6.c (revision aa24f48b361effe51163877d84f1b70d32b77e04)
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  * 3. 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 #define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/domain.h>
73 #include <sys/jail.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/proc.h>
79 #include <sys/protosw.h>
80 #include <sys/signalvar.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/sx.h>
84 #include <sys/syslog.h>
85 #include <sys/systm.h>
86 #include <sys/time.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_dl.h>
91 #include <net/if_llatbl.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet/tcp_var.h>
102 
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/ip6protosw.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/mld6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/send.h>
112 
113 extern struct domain inet6domain;
114 
115 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
116 VNET_PCPUSTAT_SYSINIT(icmp6stat);
117 
118 #ifdef VIMAGE
119 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
120 #endif /* VIMAGE */
121 
122 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
123 VNET_DECLARE(struct inpcbhead, ripcb);
124 VNET_DECLARE(int, icmp6errppslim);
125 static VNET_DEFINE(int, icmp6errpps_count) = 0;
126 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
127 VNET_DECLARE(int, icmp6_nodeinfo);
128 
129 #define	V_ripcbinfo			VNET(ripcbinfo)
130 #define	V_ripcb				VNET(ripcb)
131 #define	V_icmp6errppslim		VNET(icmp6errppslim)
132 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
133 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
134 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
135 
136 static void icmp6_errcount(int, int);
137 static int icmp6_rip6_input(struct mbuf **, int);
138 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
139 static const char *icmp6_redirect_diag(struct in6_addr *,
140 	struct in6_addr *, struct in6_addr *);
141 static struct mbuf *ni6_input(struct mbuf *, int);
142 static struct mbuf *ni6_nametodns(const char *, int, int);
143 static int ni6_dnsmatch(const char *, int, const char *, int);
144 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
145 			  struct ifnet **, struct in6_addr *);
146 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
147 				struct ifnet *, int);
148 static int icmp6_notify_error(struct mbuf **, int, int, int);
149 
150 /*
151  * Kernel module interface for updating icmp6stat.  The argument is an index
152  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
153  * general layout of icmp6stat into the caller, it doesn't encode its
154  * location, so that future changes to add, for example, per-CPU stats
155  * support won't cause binary compatibility problems for kernel modules.
156  */
157 void
158 kmod_icmp6stat_inc(int statnum)
159 {
160 
161 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
162 }
163 
164 static void
165 icmp6_errcount(int type, int code)
166 {
167 	switch (type) {
168 	case ICMP6_DST_UNREACH:
169 		switch (code) {
170 		case ICMP6_DST_UNREACH_NOROUTE:
171 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
172 			return;
173 		case ICMP6_DST_UNREACH_ADMIN:
174 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
175 			return;
176 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
177 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
178 			return;
179 		case ICMP6_DST_UNREACH_ADDR:
180 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
181 			return;
182 		case ICMP6_DST_UNREACH_NOPORT:
183 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
184 			return;
185 		}
186 		break;
187 	case ICMP6_PACKET_TOO_BIG:
188 		ICMP6STAT_INC(icp6s_opacket_too_big);
189 		return;
190 	case ICMP6_TIME_EXCEEDED:
191 		switch (code) {
192 		case ICMP6_TIME_EXCEED_TRANSIT:
193 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
194 			return;
195 		case ICMP6_TIME_EXCEED_REASSEMBLY:
196 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
197 			return;
198 		}
199 		break;
200 	case ICMP6_PARAM_PROB:
201 		switch (code) {
202 		case ICMP6_PARAMPROB_HEADER:
203 			ICMP6STAT_INC(icp6s_oparamprob_header);
204 			return;
205 		case ICMP6_PARAMPROB_NEXTHEADER:
206 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
207 			return;
208 		case ICMP6_PARAMPROB_OPTION:
209 			ICMP6STAT_INC(icp6s_oparamprob_option);
210 			return;
211 		}
212 		break;
213 	case ND_REDIRECT:
214 		ICMP6STAT_INC(icp6s_oredirect);
215 		return;
216 	}
217 	ICMP6STAT_INC(icp6s_ounknown);
218 }
219 
220 /*
221  * A wrapper function for icmp6_error() necessary when the erroneous packet
222  * may not contain enough scope zone information.
223  */
224 void
225 icmp6_error2(struct mbuf *m, int type, int code, int param,
226     struct ifnet *ifp)
227 {
228 	struct ip6_hdr *ip6;
229 
230 	if (ifp == NULL)
231 		return;
232 
233 #ifndef PULLDOWN_TEST
234 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
235 #else
236 	if (m->m_len < sizeof(struct ip6_hdr)) {
237 		m = m_pullup(m, sizeof(struct ip6_hdr));
238 		if (m == NULL)
239 			return;
240 	}
241 #endif
242 
243 	ip6 = mtod(m, struct ip6_hdr *);
244 
245 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
246 		return;
247 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
248 		return;
249 
250 	icmp6_error(m, type, code, param);
251 }
252 
253 /*
254  * Generate an error packet of type error in response to bad IP6 packet.
255  */
256 void
257 icmp6_error(struct mbuf *m, int type, int code, int param)
258 {
259 	struct ip6_hdr *oip6, *nip6;
260 	struct icmp6_hdr *icmp6;
261 	u_int preplen;
262 	int off;
263 	int nxt;
264 
265 	ICMP6STAT_INC(icp6s_error);
266 
267 	/* count per-type-code statistics */
268 	icmp6_errcount(type, code);
269 
270 #ifdef M_DECRYPTED	/*not openbsd*/
271 	if (m->m_flags & M_DECRYPTED) {
272 		ICMP6STAT_INC(icp6s_canterror);
273 		goto freeit;
274 	}
275 #endif
276 
277 #ifndef PULLDOWN_TEST
278 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
279 #else
280 	if (m->m_len < sizeof(struct ip6_hdr)) {
281 		m = m_pullup(m, sizeof(struct ip6_hdr));
282 		if (m == NULL)
283 			return;
284 	}
285 #endif
286 	oip6 = mtod(m, struct ip6_hdr *);
287 
288 	/*
289 	 * If the destination address of the erroneous packet is a multicast
290 	 * address, or the packet was sent using link-layer multicast,
291 	 * we should basically suppress sending an error (RFC 2463, Section
292 	 * 2.4).
293 	 * We have two exceptions (the item e.2 in that section):
294 	 * - the Packet Too Big message can be sent for path MTU discovery.
295 	 * - the Parameter Problem Message that can be allowed an icmp6 error
296 	 *   in the option type field.  This check has been done in
297 	 *   ip6_unknown_opt(), so we can just check the type and code.
298 	 */
299 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
300 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
301 	    (type != ICMP6_PACKET_TOO_BIG &&
302 	     (type != ICMP6_PARAM_PROB ||
303 	      code != ICMP6_PARAMPROB_OPTION)))
304 		goto freeit;
305 
306 	/*
307 	 * RFC 2463, 2.4 (e.5): source address check.
308 	 * XXX: the case of anycast source?
309 	 */
310 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
311 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
312 		goto freeit;
313 
314 	/*
315 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
316 	 * don't do it.
317 	 */
318 	nxt = -1;
319 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
320 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
321 		struct icmp6_hdr *icp;
322 
323 #ifndef PULLDOWN_TEST
324 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
325 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
326 #else
327 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
328 			sizeof(*icp));
329 		if (icp == NULL) {
330 			ICMP6STAT_INC(icp6s_tooshort);
331 			return;
332 		}
333 #endif
334 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
335 		    icp->icmp6_type == ND_REDIRECT) {
336 			/*
337 			 * ICMPv6 error
338 			 * Special case: for redirect (which is
339 			 * informational) we must not send icmp6 error.
340 			 */
341 			ICMP6STAT_INC(icp6s_canterror);
342 			goto freeit;
343 		} else {
344 			/* ICMPv6 informational - send the error */
345 		}
346 	} else {
347 		/* non-ICMPv6 - send the error */
348 	}
349 
350 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
351 
352 	/* Finally, do rate limitation check. */
353 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
354 		ICMP6STAT_INC(icp6s_toofreq);
355 		goto freeit;
356 	}
357 
358 	/*
359 	 * OK, ICMP6 can be generated.
360 	 */
361 
362 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
363 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
364 
365 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
366 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
367 	if (m == NULL) {
368 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
369 		return;
370 	}
371 
372 	nip6 = mtod(m, struct ip6_hdr *);
373 	nip6->ip6_src  = oip6->ip6_src;
374 	nip6->ip6_dst  = oip6->ip6_dst;
375 
376 	in6_clearscope(&oip6->ip6_src);
377 	in6_clearscope(&oip6->ip6_dst);
378 
379 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
380 	icmp6->icmp6_type = type;
381 	icmp6->icmp6_code = code;
382 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
383 
384 	/*
385 	 * icmp6_reflect() is designed to be in the input path.
386 	 * icmp6_error() can be called from both input and output path,
387 	 * and if we are in output path rcvif could contain bogus value.
388 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
389 	 * information in ip header (nip6).
390 	 */
391 	m->m_pkthdr.rcvif = NULL;
392 
393 	ICMP6STAT_INC(icp6s_outhist[type]);
394 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
395 
396 	return;
397 
398   freeit:
399 	/*
400 	 * If we can't tell whether or not we can generate ICMP6, free it.
401 	 */
402 	m_freem(m);
403 }
404 
405 /*
406  * Process a received ICMP6 message.
407  */
408 int
409 icmp6_input(struct mbuf **mp, int *offp, int proto)
410 {
411 	struct mbuf *m = *mp, *n;
412 	struct ifnet *ifp;
413 	struct ip6_hdr *ip6, *nip6;
414 	struct icmp6_hdr *icmp6, *nicmp6;
415 	int off = *offp;
416 	int icmp6len = m->m_pkthdr.len - *offp;
417 	int code, sum, noff;
418 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
419 	int ip6len, error;
420 
421 	ifp = m->m_pkthdr.rcvif;
422 
423 #ifndef PULLDOWN_TEST
424 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
425 	/* m might change if M_LOOP.  So, call mtod after this */
426 #endif
427 
428 	/*
429 	 * Locate icmp6 structure in mbuf, and check
430 	 * that not corrupted and of at least minimum length
431 	 */
432 
433 	ip6 = mtod(m, struct ip6_hdr *);
434 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
435 	if (icmp6len < sizeof(struct icmp6_hdr)) {
436 		ICMP6STAT_INC(icp6s_tooshort);
437 		goto freeit;
438 	}
439 
440 	/*
441 	 * Check multicast group membership.
442 	 * Note: SSM filters are not applied for ICMPv6 traffic.
443 	 */
444 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
445 		struct in6_multi	*inm;
446 
447 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
448 		if (inm == NULL) {
449 			IP6STAT_INC(ip6s_notmember);
450 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
451 			goto freeit;
452 		}
453 	}
454 
455 	/*
456 	 * calculate the checksum
457 	 */
458 #ifndef PULLDOWN_TEST
459 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
460 #else
461 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
462 	if (icmp6 == NULL) {
463 		ICMP6STAT_INC(icp6s_tooshort);
464 		return IPPROTO_DONE;
465 	}
466 #endif
467 	code = icmp6->icmp6_code;
468 
469 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
470 		nd6log((LOG_ERR,
471 		    "ICMP6 checksum error(%d|%x) %s\n",
472 		    icmp6->icmp6_type, sum,
473 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
474 		ICMP6STAT_INC(icp6s_checksum);
475 		goto freeit;
476 	}
477 
478 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
479 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
480 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
481 		icmp6_ifstat_inc(ifp, ifs6_in_error);
482 
483 	switch (icmp6->icmp6_type) {
484 	case ICMP6_DST_UNREACH:
485 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
486 		switch (code) {
487 		case ICMP6_DST_UNREACH_NOROUTE:
488 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
489 			code = PRC_UNREACH_NET;
490 			break;
491 		case ICMP6_DST_UNREACH_ADMIN:
492 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
493 			code = PRC_UNREACH_ADMIN_PROHIB;
494 			break;
495 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
496 			/* I mean "source address was incorrect." */
497 			code = PRC_PARAMPROB;
498 			break;
499 		case ICMP6_DST_UNREACH_NOPORT:
500 			code = PRC_UNREACH_PORT;
501 			break;
502 		default:
503 			goto badcode;
504 		}
505 		goto deliver;
506 		break;
507 
508 	case ICMP6_PACKET_TOO_BIG:
509 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
510 
511 		/* validation is made in icmp6_mtudisc_update */
512 
513 		code = PRC_MSGSIZE;
514 
515 		/*
516 		 * Updating the path MTU will be done after examining
517 		 * intermediate extension headers.
518 		 */
519 		goto deliver;
520 		break;
521 
522 	case ICMP6_TIME_EXCEEDED:
523 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
524 		switch (code) {
525 		case ICMP6_TIME_EXCEED_TRANSIT:
526 			code = PRC_TIMXCEED_INTRANS;
527 			break;
528 		case ICMP6_TIME_EXCEED_REASSEMBLY:
529 			code = PRC_TIMXCEED_REASS;
530 			break;
531 		default:
532 			goto badcode;
533 		}
534 		goto deliver;
535 		break;
536 
537 	case ICMP6_PARAM_PROB:
538 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
539 		switch (code) {
540 		case ICMP6_PARAMPROB_NEXTHEADER:
541 			code = PRC_UNREACH_PROTOCOL;
542 			break;
543 		case ICMP6_PARAMPROB_HEADER:
544 		case ICMP6_PARAMPROB_OPTION:
545 			code = PRC_PARAMPROB;
546 			break;
547 		default:
548 			goto badcode;
549 		}
550 		goto deliver;
551 		break;
552 
553 	case ICMP6_ECHO_REQUEST:
554 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
555 		if (code != 0)
556 			goto badcode;
557 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
558 			/* Give up remote */
559 			break;
560 		}
561 		if (!M_WRITABLE(n)
562 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
563 			struct mbuf *n0 = n;
564 			int n0len;
565 
566 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
567 			n = m_gethdr(M_NOWAIT, n0->m_type);
568 			if (n == NULL) {
569 				/* Give up remote */
570 				m_freem(n0);
571 				break;
572 			}
573 
574 			m_move_pkthdr(n, n0);	/* FIB copied. */
575 			n0len = n0->m_pkthdr.len;	/* save for use below */
576 			/*
577 			 * Copy IPv6 and ICMPv6 only.
578 			 */
579 			nip6 = mtod(n, struct ip6_hdr *);
580 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
581 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
582 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
583 			noff = sizeof(struct ip6_hdr);
584 			/* new mbuf contains only ipv6+icmpv6 headers */
585 			n->m_len = noff + sizeof(struct icmp6_hdr);
586 			/*
587 			 * Adjust mbuf.  ip6_plen will be adjusted in
588 			 * ip6_output().
589 			 */
590 			m_adj(n0, off + sizeof(struct icmp6_hdr));
591 			/* recalculate complete packet size */
592 			n->m_pkthdr.len = n0len + (noff - off);
593 			n->m_next = n0;
594 		} else {
595 			nip6 = mtod(n, struct ip6_hdr *);
596 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
597 			    sizeof(*nicmp6));
598 			noff = off;
599 		}
600 		if (n) {
601 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
602 			nicmp6->icmp6_code = 0;
603 			ICMP6STAT_INC(icp6s_reflect);
604 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
605 			icmp6_reflect(n, noff);
606 		}
607 		break;
608 
609 	case ICMP6_ECHO_REPLY:
610 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
611 		if (code != 0)
612 			goto badcode;
613 		break;
614 
615 	case MLD_LISTENER_QUERY:
616 	case MLD_LISTENER_REPORT:
617 	case MLD_LISTENER_DONE:
618 	case MLDV2_LISTENER_REPORT:
619 		/*
620 		 * Drop MLD traffic which is not link-local, has a hop limit
621 		 * of greater than 1 hop, or which does not have the
622 		 * IPv6 HBH Router Alert option.
623 		 * As IPv6 HBH options are stripped in ip6_input() we must
624 		 * check an mbuf header flag.
625 		 * XXX Should we also sanity check that these messages
626 		 * were directed to a link-local multicast prefix?
627 		 */
628 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
629 			goto freeit;
630 		if (mld_input(m, off, icmp6len) != 0)
631 			return (IPPROTO_DONE);
632 		/* m stays. */
633 		break;
634 
635 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
636 	    {
637 		enum { WRU, FQDN } mode;
638 
639 		if (!V_icmp6_nodeinfo)
640 			break;
641 
642 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
643 			mode = WRU;
644 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
645 			mode = FQDN;
646 		else
647 			goto badlen;
648 
649 		if (mode == FQDN) {
650 #ifndef PULLDOWN_TEST
651 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
652 			    IPPROTO_DONE);
653 #endif
654 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
655 			if (n)
656 				n = ni6_input(n, off);
657 			/* XXX meaningless if n == NULL */
658 			noff = sizeof(struct ip6_hdr);
659 		} else {
660 			struct prison *pr;
661 			u_char *p;
662 			int maxhlen, hlen;
663 
664 			/*
665 			 * XXX: this combination of flags is pointless,
666 			 * but should we keep this for compatibility?
667 			 */
668 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
669 			    ICMP6_NODEINFO_TMPADDROK)) !=
670 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
671 				break;
672 
673 			if (code != 0)
674 				goto badcode;
675 
676 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
677 			n = m_gethdr(M_NOWAIT, m->m_type);
678 			if (n == NULL) {
679 				/* Give up remote */
680 				break;
681 			}
682 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
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 				break;
693 			}
694 			maxhlen = M_TRAILINGSPACE(n) -
695 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
696 			pr = curthread->td_ucred->cr_prison;
697 			mtx_lock(&pr->pr_mtx);
698 			hlen = strlen(pr->pr_hostname);
699 			if (maxhlen > hlen)
700 				maxhlen = hlen;
701 			/*
702 			 * Copy IPv6 and ICMPv6 only.
703 			 */
704 			nip6 = mtod(n, struct ip6_hdr *);
705 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
706 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
707 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
708 			p = (u_char *)(nicmp6 + 1);
709 			bzero(p, 4);
710 			/* meaningless TTL */
711 			bcopy(pr->pr_hostname, p + 4, maxhlen);
712 			mtx_unlock(&pr->pr_mtx);
713 			noff = sizeof(struct ip6_hdr);
714 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
715 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
716 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
717 			nicmp6->icmp6_code = 0;
718 		}
719 		if (n) {
720 			ICMP6STAT_INC(icp6s_reflect);
721 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
722 			icmp6_reflect(n, noff);
723 		}
724 		break;
725 	    }
726 
727 	case ICMP6_WRUREPLY:
728 		if (code != 0)
729 			goto badcode;
730 		break;
731 
732 	case ND_ROUTER_SOLICIT:
733 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
734 		if (code != 0)
735 			goto badcode;
736 		if (icmp6len < sizeof(struct nd_router_solicit))
737 			goto badlen;
738 		if (send_sendso_input_hook != NULL) {
739 			IP6_EXTHDR_CHECK(m, off, icmp6len, IPPROTO_DONE);
740 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
741 			if (error == 0) {
742 				m = NULL;
743 				goto freeit;
744 			}
745 		}
746 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
747 		nd6_rs_input(m, off, icmp6len);
748 		m = n;
749 		if (m == NULL)
750 			goto freeit;
751 		break;
752 
753 	case ND_ROUTER_ADVERT:
754 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
755 		if (code != 0)
756 			goto badcode;
757 		if (icmp6len < sizeof(struct nd_router_advert))
758 			goto badlen;
759 		if (send_sendso_input_hook != NULL) {
760 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
761 			if (error == 0) {
762 				m = NULL;
763 				goto freeit;
764 			}
765 		}
766 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
767 		nd6_ra_input(m, off, icmp6len);
768 		m = n;
769 		if (m == NULL)
770 			goto freeit;
771 		break;
772 
773 	case ND_NEIGHBOR_SOLICIT:
774 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
775 		if (code != 0)
776 			goto badcode;
777 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
778 			goto badlen;
779 		if (send_sendso_input_hook != NULL) {
780 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
781 			if (error == 0) {
782 				m = NULL;
783 				goto freeit;
784 			}
785 		}
786 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
787 		nd6_ns_input(m, off, icmp6len);
788 		m = n;
789 		if (m == NULL)
790 			goto freeit;
791 		break;
792 
793 	case ND_NEIGHBOR_ADVERT:
794 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_neighbor_advert))
798 			goto badlen;
799 		if (send_sendso_input_hook != NULL) {
800 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
801 			if (error == 0) {
802 				m = NULL;
803 				goto freeit;
804 			}
805 		}
806 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
807 		nd6_na_input(m, off, icmp6len);
808 		m = n;
809 		if (m == NULL)
810 			goto freeit;
811 		break;
812 
813 	case ND_REDIRECT:
814 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
815 		if (code != 0)
816 			goto badcode;
817 		if (icmp6len < sizeof(struct nd_redirect))
818 			goto badlen;
819 		if (send_sendso_input_hook != NULL) {
820 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
821 			if (error == 0) {
822 				m = NULL;
823 				goto freeit;
824 			}
825 		}
826 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
827 		icmp6_redirect_input(m, off);
828 		m = n;
829 		if (m == NULL)
830 			goto freeit;
831 		break;
832 
833 	case ICMP6_ROUTER_RENUMBERING:
834 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
835 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
836 			goto badcode;
837 		if (icmp6len < sizeof(struct icmp6_router_renum))
838 			goto badlen;
839 		break;
840 
841 	default:
842 		nd6log((LOG_DEBUG,
843 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
844 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
845 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
846 		    ifp ? ifp->if_index : 0));
847 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
848 			/* ICMPv6 error: MUST deliver it by spec... */
849 			code = PRC_NCMDS;
850 			/* deliver */
851 		} else {
852 			/* ICMPv6 informational: MUST not deliver */
853 			break;
854 		}
855 	deliver:
856 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
857 			/* In this case, m should've been freed. */
858 			return (IPPROTO_DONE);
859 		}
860 		break;
861 
862 	badcode:
863 		ICMP6STAT_INC(icp6s_badcode);
864 		break;
865 
866 	badlen:
867 		ICMP6STAT_INC(icp6s_badlen);
868 		break;
869 	}
870 
871 	/* deliver the packet to appropriate sockets */
872 	icmp6_rip6_input(&m, *offp);
873 
874 	return IPPROTO_DONE;
875 
876  freeit:
877 	m_freem(m);
878 	return IPPROTO_DONE;
879 }
880 
881 static int
882 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
883 {
884 	struct mbuf *m = *mp;
885 	struct icmp6_hdr *icmp6;
886 	struct ip6_hdr *eip6;
887 	u_int32_t notifymtu;
888 	struct sockaddr_in6 icmp6src, icmp6dst;
889 
890 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
891 		ICMP6STAT_INC(icp6s_tooshort);
892 		goto freeit;
893 	}
894 #ifndef PULLDOWN_TEST
895 	IP6_EXTHDR_CHECK(m, off,
896 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
897 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
898 #else
899 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
900 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
901 	if (icmp6 == NULL) {
902 		ICMP6STAT_INC(icp6s_tooshort);
903 		return (-1);
904 	}
905 #endif
906 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
907 
908 	/* Detect the upper level protocol */
909 	{
910 		void (*ctlfunc)(int, struct sockaddr *, void *);
911 		u_int8_t nxt = eip6->ip6_nxt;
912 		int eoff = off + sizeof(struct icmp6_hdr) +
913 		    sizeof(struct ip6_hdr);
914 		struct ip6ctlparam ip6cp;
915 		struct in6_addr *finaldst = NULL;
916 		int icmp6type = icmp6->icmp6_type;
917 		struct ip6_frag *fh;
918 		struct ip6_rthdr *rth;
919 		struct ip6_rthdr0 *rth0;
920 		int rthlen;
921 
922 		while (1) { /* XXX: should avoid infinite loop explicitly? */
923 			struct ip6_ext *eh;
924 
925 			switch (nxt) {
926 			case IPPROTO_HOPOPTS:
927 			case IPPROTO_DSTOPTS:
928 			case IPPROTO_AH:
929 #ifndef PULLDOWN_TEST
930 				IP6_EXTHDR_CHECK(m, 0,
931 				    eoff + sizeof(struct ip6_ext), -1);
932 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
933 #else
934 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
935 				    eoff, sizeof(*eh));
936 				if (eh == NULL) {
937 					ICMP6STAT_INC(icp6s_tooshort);
938 					return (-1);
939 				}
940 #endif
941 
942 				if (nxt == IPPROTO_AH)
943 					eoff += (eh->ip6e_len + 2) << 2;
944 				else
945 					eoff += (eh->ip6e_len + 1) << 3;
946 				nxt = eh->ip6e_nxt;
947 				break;
948 			case IPPROTO_ROUTING:
949 				/*
950 				 * When the erroneous packet contains a
951 				 * routing header, we should examine the
952 				 * header to determine the final destination.
953 				 * Otherwise, we can't properly update
954 				 * information that depends on the final
955 				 * destination (e.g. path MTU).
956 				 */
957 #ifndef PULLDOWN_TEST
958 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
959 				rth = (struct ip6_rthdr *)
960 				    (mtod(m, caddr_t) + eoff);
961 #else
962 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
963 				    eoff, sizeof(*rth));
964 				if (rth == NULL) {
965 					ICMP6STAT_INC(icp6s_tooshort);
966 					return (-1);
967 				}
968 #endif
969 				rthlen = (rth->ip6r_len + 1) << 3;
970 				/*
971 				 * XXX: currently there is no
972 				 * officially defined type other
973 				 * than type-0.
974 				 * Note that if the segment left field
975 				 * is 0, all intermediate hops must
976 				 * have been passed.
977 				 */
978 				if (rth->ip6r_segleft &&
979 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
980 					int hops;
981 
982 #ifndef PULLDOWN_TEST
983 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
984 					rth0 = (struct ip6_rthdr0 *)
985 					    (mtod(m, caddr_t) + eoff);
986 #else
987 					IP6_EXTHDR_GET(rth0,
988 					    struct ip6_rthdr0 *, m,
989 					    eoff, rthlen);
990 					if (rth0 == NULL) {
991 						ICMP6STAT_INC(icp6s_tooshort);
992 						return (-1);
993 					}
994 #endif
995 					/* just ignore a bogus header */
996 					if ((rth0->ip6r0_len % 2) == 0 &&
997 					    (hops = rth0->ip6r0_len/2))
998 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
999 				}
1000 				eoff += rthlen;
1001 				nxt = rth->ip6r_nxt;
1002 				break;
1003 			case IPPROTO_FRAGMENT:
1004 #ifndef PULLDOWN_TEST
1005 				IP6_EXTHDR_CHECK(m, 0, eoff +
1006 				    sizeof(struct ip6_frag), -1);
1007 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1008 				    eoff);
1009 #else
1010 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1011 				    eoff, sizeof(*fh));
1012 				if (fh == NULL) {
1013 					ICMP6STAT_INC(icp6s_tooshort);
1014 					return (-1);
1015 				}
1016 #endif
1017 				/*
1018 				 * Data after a fragment header is meaningless
1019 				 * unless it is the first fragment, but
1020 				 * we'll go to the notify label for path MTU
1021 				 * discovery.
1022 				 */
1023 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1024 					goto notify;
1025 
1026 				eoff += sizeof(struct ip6_frag);
1027 				nxt = fh->ip6f_nxt;
1028 				break;
1029 			default:
1030 				/*
1031 				 * This case includes ESP and the No Next
1032 				 * Header.  In such cases going to the notify
1033 				 * label does not have any meaning
1034 				 * (i.e. ctlfunc will be NULL), but we go
1035 				 * anyway since we might have to update
1036 				 * path MTU information.
1037 				 */
1038 				goto notify;
1039 			}
1040 		}
1041 	  notify:
1042 #ifndef PULLDOWN_TEST
1043 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1044 #else
1045 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1046 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1047 		if (icmp6 == NULL) {
1048 			ICMP6STAT_INC(icp6s_tooshort);
1049 			return (-1);
1050 		}
1051 #endif
1052 
1053 		/*
1054 		 * retrieve parameters from the inner IPv6 header, and convert
1055 		 * them into sockaddr structures.
1056 		 * XXX: there is no guarantee that the source or destination
1057 		 * addresses of the inner packet are in the same scope as
1058 		 * the addresses of the icmp packet.  But there is no other
1059 		 * way to determine the zone.
1060 		 */
1061 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1062 
1063 		bzero(&icmp6dst, sizeof(icmp6dst));
1064 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1065 		icmp6dst.sin6_family = AF_INET6;
1066 		if (finaldst == NULL)
1067 			icmp6dst.sin6_addr = eip6->ip6_dst;
1068 		else
1069 			icmp6dst.sin6_addr = *finaldst;
1070 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1071 			goto freeit;
1072 		bzero(&icmp6src, sizeof(icmp6src));
1073 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1074 		icmp6src.sin6_family = AF_INET6;
1075 		icmp6src.sin6_addr = eip6->ip6_src;
1076 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1077 			goto freeit;
1078 		icmp6src.sin6_flowinfo =
1079 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1080 
1081 		if (finaldst == NULL)
1082 			finaldst = &eip6->ip6_dst;
1083 		ip6cp.ip6c_m = m;
1084 		ip6cp.ip6c_icmp6 = icmp6;
1085 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1086 		ip6cp.ip6c_off = eoff;
1087 		ip6cp.ip6c_finaldst = finaldst;
1088 		ip6cp.ip6c_src = &icmp6src;
1089 		ip6cp.ip6c_nxt = nxt;
1090 
1091 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1092 			notifymtu = ntohl(icmp6->icmp6_mtu);
1093 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1094 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1095 		}
1096 
1097 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1098 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1099 		if (ctlfunc) {
1100 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1101 			    &ip6cp);
1102 		}
1103 	}
1104 	*mp = m;
1105 	return (0);
1106 
1107   freeit:
1108 	m_freem(m);
1109 	return (-1);
1110 }
1111 
1112 void
1113 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1114 {
1115 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1116 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1117 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1118 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1119 	struct in_conninfo inc;
1120 
1121 #if 0
1122 	/*
1123 	 * RFC2460 section 5, last paragraph.
1124 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1125 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1126 	 * due to packet translator in the middle.
1127 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1128 	 * special handling.
1129 	 */
1130 	if (mtu < IPV6_MMTU)
1131 		return;
1132 #endif
1133 
1134 	/*
1135 	 * we reject ICMPv6 too big with abnormally small value.
1136 	 * XXX what is the good definition of "abnormally small"?
1137 	 */
1138 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1139 		return;
1140 
1141 	if (!validated)
1142 		return;
1143 
1144 	/*
1145 	 * In case the suggested mtu is less than IPV6_MMTU, we
1146 	 * only need to remember that it was for above mentioned
1147 	 * "alwaysfrag" case.
1148 	 * Try to be as close to the spec as possible.
1149 	 */
1150 	if (mtu < IPV6_MMTU)
1151 		mtu = IPV6_MMTU - 8;
1152 
1153 	bzero(&inc, sizeof(inc));
1154 	inc.inc_fibnum = M_GETFIB(m);
1155 	inc.inc_flags |= INC_ISIPV6;
1156 	inc.inc6_faddr = *dst;
1157 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1158 		return;
1159 
1160 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1161 		tcp_hc_updatemtu(&inc, mtu);
1162 		ICMP6STAT_INC(icp6s_pmtuchg);
1163 	}
1164 }
1165 
1166 /*
1167  * Process a Node Information Query packet, based on
1168  * draft-ietf-ipngwg-icmp-name-lookups-07.
1169  *
1170  * Spec incompatibilities:
1171  * - IPv6 Subject address handling
1172  * - IPv4 Subject address handling support missing
1173  * - Proxy reply (answer even if it's not for me)
1174  * - joins NI group address at in6_ifattach() time only, does not cope
1175  *   with hostname changes by sethostname(3)
1176  */
1177 static struct mbuf *
1178 ni6_input(struct mbuf *m, int off)
1179 {
1180 	struct icmp6_nodeinfo *ni6, *nni6;
1181 	struct mbuf *n = NULL;
1182 	struct prison *pr;
1183 	u_int16_t qtype;
1184 	int subjlen;
1185 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1186 	struct ni_reply_fqdn *fqdn;
1187 	int addrs;		/* for NI_QTYPE_NODEADDR */
1188 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1189 	struct in6_addr in6_subj; /* subject address */
1190 	struct ip6_hdr *ip6;
1191 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1192 	char *subj = NULL;
1193 	struct in6_ifaddr *ia6 = NULL;
1194 
1195 	ip6 = mtod(m, struct ip6_hdr *);
1196 #ifndef PULLDOWN_TEST
1197 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1198 #else
1199 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1200 	if (ni6 == NULL) {
1201 		/* m is already reclaimed */
1202 		return (NULL);
1203 	}
1204 #endif
1205 
1206 	/*
1207 	 * Validate IPv6 source address.
1208 	 * The default configuration MUST be to refuse answering queries from
1209 	 * global-scope addresses according to RFC4602.
1210 	 * Notes:
1211 	 *  - it's not very clear what "refuse" means; this implementation
1212 	 *    simply drops it.
1213 	 *  - it's not very easy to identify global-scope (unicast) addresses
1214 	 *    since there are many prefixes for them.  It should be safer
1215 	 *    and in practice sufficient to check "all" but loopback and
1216 	 *    link-local (note that site-local unicast was deprecated and
1217 	 *    ULA is defined as global scope-wise)
1218 	 */
1219 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1220 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1221 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1222 		goto bad;
1223 
1224 	/*
1225 	 * Validate IPv6 destination address.
1226 	 *
1227 	 * The Responder must discard the Query without further processing
1228 	 * unless it is one of the Responder's unicast or anycast addresses, or
1229 	 * a link-local scope multicast address which the Responder has joined.
1230 	 * [RFC4602, Section 5.]
1231 	 */
1232 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1233 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1234 			goto bad;
1235 		/* else it's a link-local multicast, fine */
1236 	} else {		/* unicast or anycast */
1237 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1238 		if (ia6 == NULL)
1239 			goto bad; /* XXX impossible */
1240 
1241 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1242 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1243 			ifa_free(&ia6->ia_ifa);
1244 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1245 				"a temporary address in %s:%d",
1246 			       __FILE__, __LINE__));
1247 			goto bad;
1248 		}
1249 		ifa_free(&ia6->ia_ifa);
1250 	}
1251 
1252 	/* validate query Subject field. */
1253 	qtype = ntohs(ni6->ni_qtype);
1254 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1255 	switch (qtype) {
1256 	case NI_QTYPE_NOOP:
1257 	case NI_QTYPE_SUPTYPES:
1258 		/* 07 draft */
1259 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1260 			break;
1261 		/* FALLTHROUGH */
1262 	case NI_QTYPE_FQDN:
1263 	case NI_QTYPE_NODEADDR:
1264 	case NI_QTYPE_IPV4ADDR:
1265 		switch (ni6->ni_code) {
1266 		case ICMP6_NI_SUBJ_IPV6:
1267 #if ICMP6_NI_SUBJ_IPV6 != 0
1268 		case 0:
1269 #endif
1270 			/*
1271 			 * backward compatibility - try to accept 03 draft
1272 			 * format, where no Subject is present.
1273 			 */
1274 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1275 			    subjlen == 0) {
1276 				oldfqdn++;
1277 				break;
1278 			}
1279 #if ICMP6_NI_SUBJ_IPV6 != 0
1280 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1281 				goto bad;
1282 #endif
1283 
1284 			if (subjlen != sizeof(struct in6_addr))
1285 				goto bad;
1286 
1287 			/*
1288 			 * Validate Subject address.
1289 			 *
1290 			 * Not sure what exactly "address belongs to the node"
1291 			 * means in the spec, is it just unicast, or what?
1292 			 *
1293 			 * At this moment we consider Subject address as
1294 			 * "belong to the node" if the Subject address equals
1295 			 * to the IPv6 destination address; validation for
1296 			 * IPv6 destination address should have done enough
1297 			 * check for us.
1298 			 *
1299 			 * We do not do proxy at this moment.
1300 			 */
1301 			/* m_pulldown instead of copy? */
1302 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1303 			    subjlen, (caddr_t)&in6_subj);
1304 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1305 				goto bad;
1306 
1307 			subj = (char *)&in6_subj;
1308 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1309 				break;
1310 
1311 			/*
1312 			 * XXX if we are to allow other cases, we should really
1313 			 * be careful about scope here.
1314 			 * basically, we should disallow queries toward IPv6
1315 			 * destination X with subject Y,
1316 			 * if scope(X) > scope(Y).
1317 			 * if we allow scope(X) > scope(Y), it will result in
1318 			 * information leakage across scope boundary.
1319 			 */
1320 			goto bad;
1321 
1322 		case ICMP6_NI_SUBJ_FQDN:
1323 			/*
1324 			 * Validate Subject name with gethostname(3).
1325 			 *
1326 			 * The behavior may need some debate, since:
1327 			 * - we are not sure if the node has FQDN as
1328 			 *   hostname (returned by gethostname(3)).
1329 			 * - the code does wildcard match for truncated names.
1330 			 *   however, we are not sure if we want to perform
1331 			 *   wildcard match, if gethostname(3) side has
1332 			 *   truncated hostname.
1333 			 */
1334 			pr = curthread->td_ucred->cr_prison;
1335 			mtx_lock(&pr->pr_mtx);
1336 			n = ni6_nametodns(pr->pr_hostname,
1337 			    strlen(pr->pr_hostname), 0);
1338 			mtx_unlock(&pr->pr_mtx);
1339 			if (!n || n->m_next || n->m_len == 0)
1340 				goto bad;
1341 			IP6_EXTHDR_GET(subj, char *, m,
1342 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1343 			if (subj == NULL)
1344 				goto bad;
1345 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1346 			    n->m_len)) {
1347 				goto bad;
1348 			}
1349 			m_freem(n);
1350 			n = NULL;
1351 			break;
1352 
1353 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1354 		default:
1355 			goto bad;
1356 		}
1357 		break;
1358 	}
1359 
1360 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1361 	switch (qtype) {
1362 	case NI_QTYPE_FQDN:
1363 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1364 			goto bad;
1365 		break;
1366 	case NI_QTYPE_NODEADDR:
1367 	case NI_QTYPE_IPV4ADDR:
1368 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1369 			goto bad;
1370 		break;
1371 	}
1372 
1373 	/* guess reply length */
1374 	switch (qtype) {
1375 	case NI_QTYPE_NOOP:
1376 		break;		/* no reply data */
1377 	case NI_QTYPE_SUPTYPES:
1378 		replylen += sizeof(u_int32_t);
1379 		break;
1380 	case NI_QTYPE_FQDN:
1381 		/* XXX will append an mbuf */
1382 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1383 		break;
1384 	case NI_QTYPE_NODEADDR:
1385 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1386 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1387 		    sizeof(u_int32_t))) > MCLBYTES)
1388 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1389 		break;
1390 	case NI_QTYPE_IPV4ADDR:
1391 		/* unsupported - should respond with unknown Qtype? */
1392 		break;
1393 	default:
1394 		/*
1395 		 * XXX: We must return a reply with the ICMP6 code
1396 		 * `unknown Qtype' in this case.  However we regard the case
1397 		 * as an FQDN query for backward compatibility.
1398 		 * Older versions set a random value to this field,
1399 		 * so it rarely varies in the defined qtypes.
1400 		 * But the mechanism is not reliable...
1401 		 * maybe we should obsolete older versions.
1402 		 */
1403 		qtype = NI_QTYPE_FQDN;
1404 		/* XXX will append an mbuf */
1405 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1406 		oldfqdn++;
1407 		break;
1408 	}
1409 
1410 	/* Allocate an mbuf to reply. */
1411 	if (replylen > MCLBYTES) {
1412 		/*
1413 		 * XXX: should we try to allocate more? But MCLBYTES
1414 		 * is probably much larger than IPV6_MMTU...
1415 		 */
1416 		goto bad;
1417 	}
1418 	if (replylen > MHLEN)
1419 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1420 	else
1421 		n = m_gethdr(M_NOWAIT, m->m_type);
1422 	if (n == NULL) {
1423 		m_freem(m);
1424 		return (NULL);
1425 	}
1426 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1427 	n->m_pkthdr.len = n->m_len = replylen;
1428 
1429 	/* copy mbuf header and IPv6 + Node Information base headers */
1430 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1431 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1432 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1433 
1434 	/* qtype dependent procedure */
1435 	switch (qtype) {
1436 	case NI_QTYPE_NOOP:
1437 		nni6->ni_code = ICMP6_NI_SUCCESS;
1438 		nni6->ni_flags = 0;
1439 		break;
1440 	case NI_QTYPE_SUPTYPES:
1441 	{
1442 		u_int32_t v;
1443 		nni6->ni_code = ICMP6_NI_SUCCESS;
1444 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1445 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1446 		v = (u_int32_t)htonl(0x0000000f);
1447 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1448 		break;
1449 	}
1450 	case NI_QTYPE_FQDN:
1451 		nni6->ni_code = ICMP6_NI_SUCCESS;
1452 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1453 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1454 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1455 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1456 		/*
1457 		 * XXX do we really have FQDN in hostname?
1458 		 */
1459 		pr = curthread->td_ucred->cr_prison;
1460 		mtx_lock(&pr->pr_mtx);
1461 		n->m_next = ni6_nametodns(pr->pr_hostname,
1462 		    strlen(pr->pr_hostname), oldfqdn);
1463 		mtx_unlock(&pr->pr_mtx);
1464 		if (n->m_next == NULL)
1465 			goto bad;
1466 		/* XXX we assume that n->m_next is not a chain */
1467 		if (n->m_next->m_next != NULL)
1468 			goto bad;
1469 		n->m_pkthdr.len += n->m_next->m_len;
1470 		break;
1471 	case NI_QTYPE_NODEADDR:
1472 	{
1473 		int lenlim, copied;
1474 
1475 		nni6->ni_code = ICMP6_NI_SUCCESS;
1476 		n->m_pkthdr.len = n->m_len =
1477 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1478 		lenlim = M_TRAILINGSPACE(n);
1479 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1480 		/* XXX: reset mbuf length */
1481 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1482 		    sizeof(struct icmp6_nodeinfo) + copied;
1483 		break;
1484 	}
1485 	default:
1486 		break;		/* XXX impossible! */
1487 	}
1488 
1489 	nni6->ni_type = ICMP6_NI_REPLY;
1490 	m_freem(m);
1491 	return (n);
1492 
1493   bad:
1494 	m_freem(m);
1495 	if (n)
1496 		m_freem(n);
1497 	return (NULL);
1498 }
1499 
1500 /*
1501  * make a mbuf with DNS-encoded string.  no compression support.
1502  *
1503  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1504  * treated as truncated name (two \0 at the end).  this is a wild guess.
1505  *
1506  * old - return pascal string if non-zero
1507  */
1508 static struct mbuf *
1509 ni6_nametodns(const char *name, int namelen, int old)
1510 {
1511 	struct mbuf *m;
1512 	char *cp, *ep;
1513 	const char *p, *q;
1514 	int i, len, nterm;
1515 
1516 	if (old)
1517 		len = namelen + 1;
1518 	else
1519 		len = MCLBYTES;
1520 
1521 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1522 	if (len > MLEN)
1523 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1524 	else
1525 		m = m_get(M_NOWAIT, MT_DATA);
1526 	if (m == NULL)
1527 		goto fail;
1528 
1529 	if (old) {
1530 		m->m_len = len;
1531 		*mtod(m, char *) = namelen;
1532 		bcopy(name, mtod(m, char *) + 1, namelen);
1533 		return m;
1534 	} else {
1535 		m->m_len = 0;
1536 		cp = mtod(m, char *);
1537 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1538 
1539 		/* if not certain about my name, return empty buffer */
1540 		if (namelen == 0)
1541 			return m;
1542 
1543 		/*
1544 		 * guess if it looks like shortened hostname, or FQDN.
1545 		 * shortened hostname needs two trailing "\0".
1546 		 */
1547 		i = 0;
1548 		for (p = name; p < name + namelen; p++) {
1549 			if (*p && *p == '.')
1550 				i++;
1551 		}
1552 		if (i < 2)
1553 			nterm = 2;
1554 		else
1555 			nterm = 1;
1556 
1557 		p = name;
1558 		while (cp < ep && p < name + namelen) {
1559 			i = 0;
1560 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1561 				i++;
1562 			/* result does not fit into mbuf */
1563 			if (cp + i + 1 >= ep)
1564 				goto fail;
1565 			/*
1566 			 * DNS label length restriction, RFC1035 page 8.
1567 			 * "i == 0" case is included here to avoid returning
1568 			 * 0-length label on "foo..bar".
1569 			 */
1570 			if (i <= 0 || i >= 64)
1571 				goto fail;
1572 			*cp++ = i;
1573 			bcopy(p, cp, i);
1574 			cp += i;
1575 			p = q;
1576 			if (p < name + namelen && *p == '.')
1577 				p++;
1578 		}
1579 		/* termination */
1580 		if (cp + nterm >= ep)
1581 			goto fail;
1582 		while (nterm-- > 0)
1583 			*cp++ = '\0';
1584 		m->m_len = cp - mtod(m, char *);
1585 		return m;
1586 	}
1587 
1588 	panic("should not reach here");
1589 	/* NOTREACHED */
1590 
1591  fail:
1592 	if (m)
1593 		m_freem(m);
1594 	return NULL;
1595 }
1596 
1597 /*
1598  * check if two DNS-encoded string matches.  takes care of truncated
1599  * form (with \0\0 at the end).  no compression support.
1600  * XXX upper/lowercase match (see RFC2065)
1601  */
1602 static int
1603 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1604 {
1605 	const char *a0, *b0;
1606 	int l;
1607 
1608 	/* simplest case - need validation? */
1609 	if (alen == blen && bcmp(a, b, alen) == 0)
1610 		return 1;
1611 
1612 	a0 = a;
1613 	b0 = b;
1614 
1615 	/* termination is mandatory */
1616 	if (alen < 2 || blen < 2)
1617 		return 0;
1618 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1619 		return 0;
1620 	alen--;
1621 	blen--;
1622 
1623 	while (a - a0 < alen && b - b0 < blen) {
1624 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1625 			return 0;
1626 
1627 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1628 			return 0;
1629 		/* we don't support compression yet */
1630 		if (a[0] >= 64 || b[0] >= 64)
1631 			return 0;
1632 
1633 		/* truncated case */
1634 		if (a[0] == 0 && a - a0 == alen - 1)
1635 			return 1;
1636 		if (b[0] == 0 && b - b0 == blen - 1)
1637 			return 1;
1638 		if (a[0] == 0 || b[0] == 0)
1639 			return 0;
1640 
1641 		if (a[0] != b[0])
1642 			return 0;
1643 		l = a[0];
1644 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1645 			return 0;
1646 		if (bcmp(a + 1, b + 1, l) != 0)
1647 			return 0;
1648 
1649 		a += 1 + l;
1650 		b += 1 + l;
1651 	}
1652 
1653 	if (a - a0 == alen && b - b0 == blen)
1654 		return 1;
1655 	else
1656 		return 0;
1657 }
1658 
1659 /*
1660  * calculate the number of addresses to be returned in the node info reply.
1661  */
1662 static int
1663 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1664     struct in6_addr *subj)
1665 {
1666 	struct ifnet *ifp;
1667 	struct in6_ifaddr *ifa6;
1668 	struct ifaddr *ifa;
1669 	int addrs = 0, addrsofif, iffound = 0;
1670 	int niflags = ni6->ni_flags;
1671 
1672 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1673 		switch (ni6->ni_code) {
1674 		case ICMP6_NI_SUBJ_IPV6:
1675 			if (subj == NULL) /* must be impossible... */
1676 				return (0);
1677 			break;
1678 		default:
1679 			/*
1680 			 * XXX: we only support IPv6 subject address for
1681 			 * this Qtype.
1682 			 */
1683 			return (0);
1684 		}
1685 	}
1686 
1687 	IFNET_RLOCK_NOSLEEP();
1688 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1689 		addrsofif = 0;
1690 		IF_ADDR_RLOCK(ifp);
1691 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1692 			if (ifa->ifa_addr->sa_family != AF_INET6)
1693 				continue;
1694 			ifa6 = (struct in6_ifaddr *)ifa;
1695 
1696 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1697 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1698 				iffound = 1;
1699 
1700 			/*
1701 			 * IPv4-mapped addresses can only be returned by a
1702 			 * Node Information proxy, since they represent
1703 			 * addresses of IPv4-only nodes, which perforce do
1704 			 * not implement this protocol.
1705 			 * [icmp-name-lookups-07, Section 5.4]
1706 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1707 			 * this function at this moment.
1708 			 */
1709 
1710 			/* What do we have to do about ::1? */
1711 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1712 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1713 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1714 					continue;
1715 				break;
1716 			case IPV6_ADDR_SCOPE_SITELOCAL:
1717 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1718 					continue;
1719 				break;
1720 			case IPV6_ADDR_SCOPE_GLOBAL:
1721 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1722 					continue;
1723 				break;
1724 			default:
1725 				continue;
1726 			}
1727 
1728 			/*
1729 			 * check if anycast is okay.
1730 			 * XXX: just experimental.  not in the spec.
1731 			 */
1732 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1733 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1734 				continue; /* we need only unicast addresses */
1735 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1736 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1737 				continue;
1738 			}
1739 			addrsofif++; /* count the address */
1740 		}
1741 		IF_ADDR_RUNLOCK(ifp);
1742 		if (iffound) {
1743 			*ifpp = ifp;
1744 			IFNET_RUNLOCK_NOSLEEP();
1745 			return (addrsofif);
1746 		}
1747 
1748 		addrs += addrsofif;
1749 	}
1750 	IFNET_RUNLOCK_NOSLEEP();
1751 
1752 	return (addrs);
1753 }
1754 
1755 static int
1756 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1757     struct ifnet *ifp0, int resid)
1758 {
1759 	struct ifnet *ifp;
1760 	struct in6_ifaddr *ifa6;
1761 	struct ifaddr *ifa;
1762 	struct ifnet *ifp_dep = NULL;
1763 	int copied = 0, allow_deprecated = 0;
1764 	u_char *cp = (u_char *)(nni6 + 1);
1765 	int niflags = ni6->ni_flags;
1766 	u_int32_t ltime;
1767 
1768 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1769 		return (0);	/* needless to copy */
1770 
1771 	IFNET_RLOCK_NOSLEEP();
1772 	ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1773   again:
1774 
1775 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_link)) {
1776 		IF_ADDR_RLOCK(ifp);
1777 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1778 			if (ifa->ifa_addr->sa_family != AF_INET6)
1779 				continue;
1780 			ifa6 = (struct in6_ifaddr *)ifa;
1781 
1782 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1783 			    allow_deprecated == 0) {
1784 				/*
1785 				 * prefererred address should be put before
1786 				 * deprecated addresses.
1787 				 */
1788 
1789 				/* record the interface for later search */
1790 				if (ifp_dep == NULL)
1791 					ifp_dep = ifp;
1792 
1793 				continue;
1794 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1795 			    allow_deprecated != 0)
1796 				continue; /* we now collect deprecated addrs */
1797 
1798 			/* What do we have to do about ::1? */
1799 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1800 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1801 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1802 					continue;
1803 				break;
1804 			case IPV6_ADDR_SCOPE_SITELOCAL:
1805 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1806 					continue;
1807 				break;
1808 			case IPV6_ADDR_SCOPE_GLOBAL:
1809 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1810 					continue;
1811 				break;
1812 			default:
1813 				continue;
1814 			}
1815 
1816 			/*
1817 			 * check if anycast is okay.
1818 			 * XXX: just experimental.  not in the spec.
1819 			 */
1820 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1821 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1822 				continue;
1823 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1824 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1825 				continue;
1826 			}
1827 
1828 			/* now we can copy the address */
1829 			if (resid < sizeof(struct in6_addr) +
1830 			    sizeof(u_int32_t)) {
1831 				IF_ADDR_RUNLOCK(ifp);
1832 				/*
1833 				 * We give up much more copy.
1834 				 * Set the truncate flag and return.
1835 				 */
1836 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1837 				IFNET_RUNLOCK_NOSLEEP();
1838 				return (copied);
1839 			}
1840 
1841 			/*
1842 			 * Set the TTL of the address.
1843 			 * The TTL value should be one of the following
1844 			 * according to the specification:
1845 			 *
1846 			 * 1. The remaining lifetime of a DHCP lease on the
1847 			 *    address, or
1848 			 * 2. The remaining Valid Lifetime of a prefix from
1849 			 *    which the address was derived through Stateless
1850 			 *    Autoconfiguration.
1851 			 *
1852 			 * Note that we currently do not support stateful
1853 			 * address configuration by DHCPv6, so the former
1854 			 * case can't happen.
1855 			 */
1856 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1857 				ltime = ND6_INFINITE_LIFETIME;
1858 			else {
1859 				if (ifa6->ia6_lifetime.ia6t_expire >
1860 				    time_uptime)
1861 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1862 				else
1863 					ltime = 0;
1864 			}
1865 
1866 			bcopy(&ltime, cp, sizeof(u_int32_t));
1867 			cp += sizeof(u_int32_t);
1868 
1869 			/* copy the address itself */
1870 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1871 			    sizeof(struct in6_addr));
1872 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1873 			cp += sizeof(struct in6_addr);
1874 
1875 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1876 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1877 		}
1878 		IF_ADDR_RUNLOCK(ifp);
1879 		if (ifp0)	/* we need search only on the specified IF */
1880 			break;
1881 	}
1882 
1883 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1884 		ifp = ifp_dep;
1885 		allow_deprecated = 1;
1886 
1887 		goto again;
1888 	}
1889 
1890 	IFNET_RUNLOCK_NOSLEEP();
1891 
1892 	return (copied);
1893 }
1894 
1895 /*
1896  * XXX almost dup'ed code with rip6_input.
1897  */
1898 static int
1899 icmp6_rip6_input(struct mbuf **mp, int off)
1900 {
1901 	struct mbuf *m = *mp;
1902 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1903 	struct inpcb *in6p;
1904 	struct inpcb *last = NULL;
1905 	struct sockaddr_in6 fromsa;
1906 	struct icmp6_hdr *icmp6;
1907 	struct mbuf *opts = NULL;
1908 
1909 #ifndef PULLDOWN_TEST
1910 	/* this is assumed to be safe. */
1911 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1912 #else
1913 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1914 	if (icmp6 == NULL) {
1915 		/* m is already reclaimed */
1916 		return (IPPROTO_DONE);
1917 	}
1918 #endif
1919 
1920 	/*
1921 	 * XXX: the address may have embedded scope zone ID, which should be
1922 	 * hidden from applications.
1923 	 */
1924 	bzero(&fromsa, sizeof(fromsa));
1925 	fromsa.sin6_family = AF_INET6;
1926 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1927 	fromsa.sin6_addr = ip6->ip6_src;
1928 	if (sa6_recoverscope(&fromsa)) {
1929 		m_freem(m);
1930 		return (IPPROTO_DONE);
1931 	}
1932 
1933 	INP_INFO_RLOCK(&V_ripcbinfo);
1934 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1935 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1936 			continue;
1937 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1938 			continue;
1939 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1940 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1941 			continue;
1942 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1943 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1944 			continue;
1945 		INP_RLOCK(in6p);
1946 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1947 		    in6p->in6p_icmp6filt)) {
1948 			INP_RUNLOCK(in6p);
1949 			continue;
1950 		}
1951 		if (last != NULL) {
1952 			struct	mbuf *n = NULL;
1953 
1954 			/*
1955 			 * Recent network drivers tend to allocate a single
1956 			 * mbuf cluster, rather than to make a couple of
1957 			 * mbufs without clusters.  Also, since the IPv6 code
1958 			 * path tries to avoid m_pullup(), it is highly
1959 			 * probable that we still have an mbuf cluster here
1960 			 * even though the necessary length can be stored in an
1961 			 * mbuf's internal buffer.
1962 			 * Meanwhile, the default size of the receive socket
1963 			 * buffer for raw sockets is not so large.  This means
1964 			 * the possibility of packet loss is relatively higher
1965 			 * than before.  To avoid this scenario, we copy the
1966 			 * received data to a separate mbuf that does not use
1967 			 * a cluster, if possible.
1968 			 * XXX: it is better to copy the data after stripping
1969 			 * intermediate headers.
1970 			 */
1971 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1972 			    m->m_len <= MHLEN) {
1973 				n = m_get(M_NOWAIT, m->m_type);
1974 				if (n != NULL) {
1975 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1976 						bcopy(m->m_data, n->m_data,
1977 						      m->m_len);
1978 						n->m_len = m->m_len;
1979 					} else {
1980 						m_free(n);
1981 						n = NULL;
1982 					}
1983 				}
1984 			}
1985 			if (n != NULL ||
1986 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1987 				if (last->inp_flags & INP_CONTROLOPTS)
1988 					ip6_savecontrol(last, n, &opts);
1989 				/* strip intermediate headers */
1990 				m_adj(n, off);
1991 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1992 				if (sbappendaddr_locked(
1993 				    &last->inp_socket->so_rcv,
1994 				    (struct sockaddr *)&fromsa, n, opts)
1995 				    == 0) {
1996 					/* should notify about lost packet */
1997 					m_freem(n);
1998 					if (opts) {
1999 						m_freem(opts);
2000 					}
2001 					SOCKBUF_UNLOCK(
2002 					    &last->inp_socket->so_rcv);
2003 				} else
2004 					sorwakeup_locked(last->inp_socket);
2005 				opts = NULL;
2006 			}
2007 			INP_RUNLOCK(last);
2008 		}
2009 		last = in6p;
2010 	}
2011 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2012 	if (last != NULL) {
2013 		if (last->inp_flags & INP_CONTROLOPTS)
2014 			ip6_savecontrol(last, m, &opts);
2015 		/* strip intermediate headers */
2016 		m_adj(m, off);
2017 
2018 		/* avoid using mbuf clusters if possible (see above) */
2019 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2020 		    m->m_len <= MHLEN) {
2021 			struct mbuf *n;
2022 
2023 			n = m_get(M_NOWAIT, m->m_type);
2024 			if (n != NULL) {
2025 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2026 					bcopy(m->m_data, n->m_data, m->m_len);
2027 					n->m_len = m->m_len;
2028 
2029 					m_freem(m);
2030 					m = n;
2031 				} else {
2032 					m_freem(n);
2033 					n = NULL;
2034 				}
2035 			}
2036 		}
2037 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2038 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2039 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2040 			m_freem(m);
2041 			if (opts)
2042 				m_freem(opts);
2043 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2044 		} else
2045 			sorwakeup_locked(last->inp_socket);
2046 		INP_RUNLOCK(last);
2047 	} else {
2048 		m_freem(m);
2049 		IP6STAT_DEC(ip6s_delivered);
2050 	}
2051 	return IPPROTO_DONE;
2052 }
2053 
2054 /*
2055  * Reflect the ip6 packet back to the source.
2056  * OFF points to the icmp6 header, counted from the top of the mbuf.
2057  */
2058 void
2059 icmp6_reflect(struct mbuf *m, size_t off)
2060 {
2061 	struct in6_addr src6, *srcp;
2062 	struct ip6_hdr *ip6;
2063 	struct icmp6_hdr *icmp6;
2064 	struct in6_ifaddr *ia = NULL;
2065 	struct ifnet *outif = NULL;
2066 	int plen;
2067 	int type, code, hlim;
2068 
2069 	/* too short to reflect */
2070 	if (off < sizeof(struct ip6_hdr)) {
2071 		nd6log((LOG_DEBUG,
2072 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2073 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2074 		    __FILE__, __LINE__));
2075 		goto bad;
2076 	}
2077 
2078 	/*
2079 	 * If there are extra headers between IPv6 and ICMPv6, strip
2080 	 * off that header first.
2081 	 */
2082 #ifdef DIAGNOSTIC
2083 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2084 		panic("assumption failed in icmp6_reflect");
2085 #endif
2086 	if (off > sizeof(struct ip6_hdr)) {
2087 		size_t l;
2088 		struct ip6_hdr nip6;
2089 
2090 		l = off - sizeof(struct ip6_hdr);
2091 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2092 		m_adj(m, l);
2093 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2094 		if (m->m_len < l) {
2095 			if ((m = m_pullup(m, l)) == NULL)
2096 				return;
2097 		}
2098 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2099 	} else /* off == sizeof(struct ip6_hdr) */ {
2100 		size_t l;
2101 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2102 		if (m->m_len < l) {
2103 			if ((m = m_pullup(m, l)) == NULL)
2104 				return;
2105 		}
2106 	}
2107 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2108 	ip6 = mtod(m, struct ip6_hdr *);
2109 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2110 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2111 	type = icmp6->icmp6_type; /* keep type for statistics */
2112 	code = icmp6->icmp6_code; /* ditto. */
2113 	hlim = 0;
2114 	srcp = NULL;
2115 
2116 	/*
2117 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2118 	 * use dst as the src for the reply.
2119 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2120 	 * (for example) when we encounter an error while forwarding procedure
2121 	 * destined to a duplicated address of ours.
2122 	 */
2123 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2124 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2125 		if (ia != NULL && !(ia->ia6_flags &
2126 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2127 			src6 = ia->ia_addr.sin6_addr;
2128 			srcp = &src6;
2129 
2130 			if (m->m_pkthdr.rcvif != NULL) {
2131 				/* XXX: This may not be the outgoing interface */
2132 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2133 			} else
2134 				hlim = V_ip6_defhlim;
2135 		}
2136 		if (ia != NULL)
2137 			ifa_free(&ia->ia_ifa);
2138 	}
2139 
2140 	if (srcp == NULL) {
2141 		int error;
2142 		struct in6_addr dst6;
2143 		uint32_t scopeid;
2144 
2145 		/*
2146 		 * This case matches to multicasts, our anycast, or unicasts
2147 		 * that we do not own.  Select a source address based on the
2148 		 * source address of the erroneous packet.
2149 		 */
2150 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2151 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2152 		    scopeid, NULL, &src6, &hlim);
2153 
2154 		if (error) {
2155 			char ip6buf[INET6_ADDRSTRLEN];
2156 			nd6log((LOG_DEBUG,
2157 			    "icmp6_reflect: source can't be determined: "
2158 			    "dst=%s, error=%d\n",
2159 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2160 			goto bad;
2161 		}
2162 		srcp = &src6;
2163 	}
2164 	/*
2165 	 * ip6_input() drops a packet if its src is multicast.
2166 	 * So, the src is never multicast.
2167 	 */
2168 	ip6->ip6_dst = ip6->ip6_src;
2169 	ip6->ip6_src = *srcp;
2170 	ip6->ip6_flow = 0;
2171 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2172 	ip6->ip6_vfc |= IPV6_VERSION;
2173 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2174 	ip6->ip6_hlim = hlim;
2175 
2176 	icmp6->icmp6_cksum = 0;
2177 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2178 	    sizeof(struct ip6_hdr), plen);
2179 
2180 	/*
2181 	 * XXX option handling
2182 	 */
2183 
2184 	m->m_flags &= ~(M_BCAST|M_MCAST);
2185 
2186 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2187 	if (outif)
2188 		icmp6_ifoutstat_inc(outif, type, code);
2189 
2190 	return;
2191 
2192  bad:
2193 	m_freem(m);
2194 	return;
2195 }
2196 
2197 void
2198 icmp6_fasttimo(void)
2199 {
2200 
2201 	mld_fasttimo();
2202 }
2203 
2204 void
2205 icmp6_slowtimo(void)
2206 {
2207 
2208 	mld_slowtimo();
2209 }
2210 
2211 static const char *
2212 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2213     struct in6_addr *tgt6)
2214 {
2215 	static char buf[1024];
2216 	char ip6bufs[INET6_ADDRSTRLEN];
2217 	char ip6bufd[INET6_ADDRSTRLEN];
2218 	char ip6buft[INET6_ADDRSTRLEN];
2219 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2220 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2221 	    ip6_sprintf(ip6buft, tgt6));
2222 	return buf;
2223 }
2224 
2225 void
2226 icmp6_redirect_input(struct mbuf *m, int off)
2227 {
2228 	struct ifnet *ifp;
2229 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2230 	struct nd_redirect *nd_rd;
2231 	int icmp6len = ntohs(ip6->ip6_plen);
2232 	char *lladdr = NULL;
2233 	int lladdrlen = 0;
2234 	int is_router;
2235 	int is_onlink;
2236 	struct in6_addr src6 = ip6->ip6_src;
2237 	struct in6_addr redtgt6;
2238 	struct in6_addr reddst6;
2239 	union nd_opts ndopts;
2240 	char ip6buf[INET6_ADDRSTRLEN];
2241 
2242 	M_ASSERTPKTHDR(m);
2243 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2244 
2245 	ifp = m->m_pkthdr.rcvif;
2246 
2247 	/* XXX if we are router, we don't update route by icmp6 redirect */
2248 	if (V_ip6_forwarding)
2249 		goto freeit;
2250 	if (!V_icmp6_rediraccept)
2251 		goto freeit;
2252 
2253 #ifndef PULLDOWN_TEST
2254 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2255 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2256 #else
2257 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2258 	if (nd_rd == NULL) {
2259 		ICMP6STAT_INC(icp6s_tooshort);
2260 		return;
2261 	}
2262 #endif
2263 	redtgt6 = nd_rd->nd_rd_target;
2264 	reddst6 = nd_rd->nd_rd_dst;
2265 
2266 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2267 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2268 		goto freeit;
2269 	}
2270 
2271 	/* validation */
2272 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2273 		nd6log((LOG_ERR,
2274 		    "ICMP6 redirect sent from %s rejected; "
2275 		    "must be from linklocal\n",
2276 		    ip6_sprintf(ip6buf, &src6)));
2277 		goto bad;
2278 	}
2279 	if (ip6->ip6_hlim != 255) {
2280 		nd6log((LOG_ERR,
2281 		    "ICMP6 redirect sent from %s rejected; "
2282 		    "hlim=%d (must be 255)\n",
2283 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2284 		goto bad;
2285 	}
2286     {
2287 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2288 	struct nhop6_basic nh6;
2289 	struct in6_addr kdst;
2290 	uint32_t scopeid;
2291 
2292 	in6_splitscope(&reddst6, &kdst, &scopeid);
2293 	if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2294 		if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2295 			nd6log((LOG_ERR,
2296 			    "ICMP6 redirect rejected; no route "
2297 			    "with inet6 gateway found for redirect dst: %s\n",
2298 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299 			goto bad;
2300 		}
2301 
2302 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2303 			nd6log((LOG_ERR,
2304 			    "ICMP6 redirect rejected; "
2305 			    "not equal to gw-for-src=%s (must be same): "
2306 			    "%s\n",
2307 			    ip6_sprintf(ip6buf, &nh6.nh_addr),
2308 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2309 			goto bad;
2310 		}
2311 	} else {
2312 		nd6log((LOG_ERR,
2313 		    "ICMP6 redirect rejected; "
2314 		    "no route found for redirect dst: %s\n",
2315 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2316 		goto bad;
2317 	}
2318     }
2319 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2320 		nd6log((LOG_ERR,
2321 		    "ICMP6 redirect rejected; "
2322 		    "redirect dst must be unicast: %s\n",
2323 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2324 		goto bad;
2325 	}
2326 
2327 	is_router = is_onlink = 0;
2328 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2329 		is_router = 1;	/* router case */
2330 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2331 		is_onlink = 1;	/* on-link destination case */
2332 	if (!is_router && !is_onlink) {
2333 		nd6log((LOG_ERR,
2334 		    "ICMP6 redirect rejected; "
2335 		    "neither router case nor onlink case: %s\n",
2336 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2337 		goto bad;
2338 	}
2339 
2340 	icmp6len -= sizeof(*nd_rd);
2341 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2342 	if (nd6_options(&ndopts) < 0) {
2343 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2344 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2345 		/* nd6_options have incremented stats */
2346 		goto freeit;
2347 	}
2348 
2349 	if (ndopts.nd_opts_tgt_lladdr) {
2350 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2351 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2352 	}
2353 
2354 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2355 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2356 		    "(if %d, icmp6 packet %d): %s\n",
2357 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2358 		    ifp->if_addrlen, lladdrlen - 2,
2359 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2360 		goto bad;
2361 	}
2362 
2363 	/* Validation passed. */
2364 
2365 	/* RFC 2461 8.3 */
2366 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2367 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2368 
2369 	/*
2370 	 * Install a gateway route in the better-router case or an interface
2371 	 * route in the on-link-destination case.
2372 	 */
2373 	{
2374 		struct sockaddr_in6 sdst;
2375 		struct sockaddr_in6 sgw;
2376 		struct sockaddr_in6 ssrc;
2377 		struct sockaddr *gw;
2378 		int rt_flags;
2379 		u_int fibnum;
2380 
2381 		bzero(&sdst, sizeof(sdst));
2382 		bzero(&ssrc, sizeof(ssrc));
2383 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2384 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2385 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2386 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2387 		rt_flags = RTF_HOST;
2388 		if (is_router) {
2389 			bzero(&sgw, sizeof(sgw));
2390 			sgw.sin6_family = AF_INET6;
2391 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2392 			bcopy(&redtgt6, &sgw.sin6_addr,
2393 				sizeof(struct in6_addr));
2394 			gw = (struct sockaddr *)&sgw;
2395 			rt_flags |= RTF_GATEWAY;
2396 		} else
2397 			gw = ifp->if_addr->ifa_addr;
2398 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2399 			in6_rtredirect((struct sockaddr *)&sdst, gw,
2400 			    (struct sockaddr *)NULL, rt_flags,
2401 			    (struct sockaddr *)&ssrc, fibnum);
2402 	}
2403 	/* finally update cached route in each socket via pfctlinput */
2404     {
2405 	struct sockaddr_in6 sdst;
2406 
2407 	bzero(&sdst, sizeof(sdst));
2408 	sdst.sin6_family = AF_INET6;
2409 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2410 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2411 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2412     }
2413 
2414  freeit:
2415 	m_freem(m);
2416 	return;
2417 
2418  bad:
2419 	ICMP6STAT_INC(icp6s_badredirect);
2420 	m_freem(m);
2421 }
2422 
2423 void
2424 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2425 {
2426 	struct ifnet *ifp;	/* my outgoing interface */
2427 	struct in6_addr *ifp_ll6;
2428 	struct in6_addr *router_ll6;
2429 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2430 	struct mbuf *m = NULL;	/* newly allocated one */
2431 	struct m_tag *mtag;
2432 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2433 	struct nd_redirect *nd_rd;
2434 	struct llentry *ln = NULL;
2435 	size_t maxlen;
2436 	u_char *p;
2437 	struct ifnet *outif = NULL;
2438 	struct sockaddr_in6 src_sa;
2439 
2440 	icmp6_errcount(ND_REDIRECT, 0);
2441 
2442 	/* if we are not router, we don't send icmp6 redirect */
2443 	if (!V_ip6_forwarding)
2444 		goto fail;
2445 
2446 	/* sanity check */
2447 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2448 		goto fail;
2449 
2450 	/*
2451 	 * Address check:
2452 	 *  the source address must identify a neighbor, and
2453 	 *  the destination address must not be a multicast address
2454 	 *  [RFC 2461, sec 8.2]
2455 	 */
2456 	sip6 = mtod(m0, struct ip6_hdr *);
2457 	bzero(&src_sa, sizeof(src_sa));
2458 	src_sa.sin6_family = AF_INET6;
2459 	src_sa.sin6_len = sizeof(src_sa);
2460 	src_sa.sin6_addr = sip6->ip6_src;
2461 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2462 		goto fail;
2463 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2464 		goto fail;	/* what should we do here? */
2465 
2466 	/* rate limit */
2467 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2468 		goto fail;
2469 
2470 	/*
2471 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2472 	 * we almost always ask for an mbuf cluster for simplicity.
2473 	 * (MHLEN < IPV6_MMTU is almost always true)
2474 	 */
2475 #if IPV6_MMTU >= MCLBYTES
2476 # error assumption failed about IPV6_MMTU and MCLBYTES
2477 #endif
2478 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2479 	if (m == NULL)
2480 		goto fail;
2481 	M_SETFIB(m, rt->rt_fibnum);
2482 	maxlen = M_TRAILINGSPACE(m);
2483 	maxlen = min(IPV6_MMTU, maxlen);
2484 	/* just for safety */
2485 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2486 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2487 		goto fail;
2488 	}
2489 
2490 	{
2491 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2492 		struct in6_ifaddr *ia;
2493 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2494 						 IN6_IFF_NOTREADY|
2495 						 IN6_IFF_ANYCAST)) == NULL)
2496 			goto fail;
2497 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2498 		/* XXXRW: reference released prematurely. */
2499 		ifa_free(&ia->ia_ifa);
2500 	}
2501 
2502 	/* get ip6 linklocal address for the router. */
2503 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2504 		struct sockaddr_in6 *sin6;
2505 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2506 		router_ll6 = &sin6->sin6_addr;
2507 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2508 			router_ll6 = (struct in6_addr *)NULL;
2509 	} else
2510 		router_ll6 = (struct in6_addr *)NULL;
2511 
2512 	/* ip6 */
2513 	ip6 = mtod(m, struct ip6_hdr *);
2514 	ip6->ip6_flow = 0;
2515 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2516 	ip6->ip6_vfc |= IPV6_VERSION;
2517 	/* ip6->ip6_plen will be set later */
2518 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2519 	ip6->ip6_hlim = 255;
2520 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2521 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2522 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2523 
2524 	/* ND Redirect */
2525 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2526 	nd_rd->nd_rd_type = ND_REDIRECT;
2527 	nd_rd->nd_rd_code = 0;
2528 	nd_rd->nd_rd_reserved = 0;
2529 	if (rt->rt_flags & RTF_GATEWAY) {
2530 		/*
2531 		 * nd_rd->nd_rd_target must be a link-local address in
2532 		 * better router cases.
2533 		 */
2534 		if (!router_ll6)
2535 			goto fail;
2536 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2537 		    sizeof(nd_rd->nd_rd_target));
2538 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2539 		    sizeof(nd_rd->nd_rd_dst));
2540 	} else {
2541 		/* make sure redtgt == reddst */
2542 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2543 		    sizeof(nd_rd->nd_rd_target));
2544 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2545 		    sizeof(nd_rd->nd_rd_dst));
2546 	}
2547 
2548 	p = (u_char *)(nd_rd + 1);
2549 
2550 	if (!router_ll6)
2551 		goto nolladdropt;
2552 
2553 	{
2554 		/* target lladdr option */
2555 		int len;
2556 		struct nd_opt_hdr *nd_opt;
2557 		char *lladdr;
2558 
2559 		IF_AFDATA_RLOCK(ifp);
2560 		ln = nd6_lookup(router_ll6, 0, ifp);
2561 		IF_AFDATA_RUNLOCK(ifp);
2562 		if (ln == NULL)
2563 			goto nolladdropt;
2564 
2565 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2566 		len = (len + 7) & ~7;	/* round by 8 */
2567 		/* safety check */
2568 		if (len + (p - (u_char *)ip6) > maxlen)
2569 			goto nolladdropt;
2570 
2571 		if (ln->la_flags & LLE_VALID) {
2572 			nd_opt = (struct nd_opt_hdr *)p;
2573 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2574 			nd_opt->nd_opt_len = len >> 3;
2575 			lladdr = (char *)(nd_opt + 1);
2576 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2577 			p += len;
2578 		}
2579 	}
2580 nolladdropt:
2581 	if (ln != NULL)
2582 		LLE_RUNLOCK(ln);
2583 
2584 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2585 
2586 	/* just to be safe */
2587 #ifdef M_DECRYPTED	/*not openbsd*/
2588 	if (m0->m_flags & M_DECRYPTED)
2589 		goto noredhdropt;
2590 #endif
2591 	if (p - (u_char *)ip6 > maxlen)
2592 		goto noredhdropt;
2593 
2594 	{
2595 		/* redirected header option */
2596 		int len;
2597 		struct nd_opt_rd_hdr *nd_opt_rh;
2598 
2599 		/*
2600 		 * compute the maximum size for icmp6 redirect header option.
2601 		 * XXX room for auth header?
2602 		 */
2603 		len = maxlen - (p - (u_char *)ip6);
2604 		len &= ~7;
2605 
2606 		/* This is just for simplicity. */
2607 		if (m0->m_pkthdr.len != m0->m_len) {
2608 			if (m0->m_next) {
2609 				m_freem(m0->m_next);
2610 				m0->m_next = NULL;
2611 			}
2612 			m0->m_pkthdr.len = m0->m_len;
2613 		}
2614 
2615 		/*
2616 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2617 		 * about padding/truncate rule for the original IP packet.
2618 		 * From the discussion on IPv6imp in Feb 1999,
2619 		 * the consensus was:
2620 		 * - "attach as much as possible" is the goal
2621 		 * - pad if not aligned (original size can be guessed by
2622 		 *   original ip6 header)
2623 		 * Following code adds the padding if it is simple enough,
2624 		 * and truncates if not.
2625 		 */
2626 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2627 			panic("assumption failed in %s:%d", __FILE__,
2628 			    __LINE__);
2629 
2630 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2631 			/* not enough room, truncate */
2632 			m0->m_pkthdr.len = m0->m_len = len -
2633 			    sizeof(*nd_opt_rh);
2634 		} else {
2635 			/* enough room, pad or truncate */
2636 			size_t extra;
2637 
2638 			extra = m0->m_pkthdr.len % 8;
2639 			if (extra) {
2640 				/* pad if easy enough, truncate if not */
2641 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2642 					/* pad */
2643 					m0->m_len += (8 - extra);
2644 					m0->m_pkthdr.len += (8 - extra);
2645 				} else {
2646 					/* truncate */
2647 					m0->m_pkthdr.len -= extra;
2648 					m0->m_len -= extra;
2649 				}
2650 			}
2651 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2652 			m0->m_pkthdr.len = m0->m_len = len -
2653 			    sizeof(*nd_opt_rh);
2654 		}
2655 
2656 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2657 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2658 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2659 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2660 		p += sizeof(*nd_opt_rh);
2661 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2662 
2663 		/* connect m0 to m */
2664 		m_tag_delete_chain(m0, NULL);
2665 		m0->m_flags &= ~M_PKTHDR;
2666 		m->m_next = m0;
2667 		m->m_pkthdr.len = m->m_len + m0->m_len;
2668 		m0 = NULL;
2669 	}
2670 noredhdropt:;
2671 	if (m0) {
2672 		m_freem(m0);
2673 		m0 = NULL;
2674 	}
2675 
2676 	/* XXX: clear embedded link IDs in the inner header */
2677 	in6_clearscope(&sip6->ip6_src);
2678 	in6_clearscope(&sip6->ip6_dst);
2679 	in6_clearscope(&nd_rd->nd_rd_target);
2680 	in6_clearscope(&nd_rd->nd_rd_dst);
2681 
2682 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2683 
2684 	nd_rd->nd_rd_cksum = 0;
2685 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2686 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2687 
2688         if (send_sendso_input_hook != NULL) {
2689 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2690 			M_NOWAIT);
2691 		if (mtag == NULL)
2692 			goto fail;
2693 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2694 		m_tag_prepend(m, mtag);
2695 	}
2696 
2697 	/* send the packet to outside... */
2698 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2699 	if (outif) {
2700 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2701 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2702 	}
2703 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2704 
2705 	return;
2706 
2707 fail:
2708 	if (m)
2709 		m_freem(m);
2710 	if (m0)
2711 		m_freem(m0);
2712 }
2713 
2714 /*
2715  * ICMPv6 socket option processing.
2716  */
2717 int
2718 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2719 {
2720 	int error = 0;
2721 	int optlen;
2722 	struct inpcb *inp = sotoinpcb(so);
2723 	int level, op, optname;
2724 
2725 	if (sopt) {
2726 		level = sopt->sopt_level;
2727 		op = sopt->sopt_dir;
2728 		optname = sopt->sopt_name;
2729 		optlen = sopt->sopt_valsize;
2730 	} else
2731 		level = op = optname = optlen = 0;
2732 
2733 	if (level != IPPROTO_ICMPV6) {
2734 		return EINVAL;
2735 	}
2736 
2737 	switch (op) {
2738 	case PRCO_SETOPT:
2739 		switch (optname) {
2740 		case ICMP6_FILTER:
2741 		    {
2742 			struct icmp6_filter ic6f;
2743 
2744 			if (optlen != sizeof(ic6f)) {
2745 				error = EMSGSIZE;
2746 				break;
2747 			}
2748 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2749 			if (error == 0) {
2750 				INP_WLOCK(inp);
2751 				*inp->in6p_icmp6filt = ic6f;
2752 				INP_WUNLOCK(inp);
2753 			}
2754 			break;
2755 		    }
2756 
2757 		default:
2758 			error = ENOPROTOOPT;
2759 			break;
2760 		}
2761 		break;
2762 
2763 	case PRCO_GETOPT:
2764 		switch (optname) {
2765 		case ICMP6_FILTER:
2766 		    {
2767 			struct icmp6_filter ic6f;
2768 
2769 			INP_RLOCK(inp);
2770 			ic6f = *inp->in6p_icmp6filt;
2771 			INP_RUNLOCK(inp);
2772 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2773 			break;
2774 		    }
2775 
2776 		default:
2777 			error = ENOPROTOOPT;
2778 			break;
2779 		}
2780 		break;
2781 	}
2782 
2783 	return (error);
2784 }
2785 
2786 /*
2787  * Perform rate limit check.
2788  * Returns 0 if it is okay to send the icmp6 packet.
2789  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2790  * limitation.
2791  *
2792  * XXX per-destination/type check necessary?
2793  *
2794  * dst - not used at this moment
2795  * type - not used at this moment
2796  * code - not used at this moment
2797  */
2798 static int
2799 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2800     const int code)
2801 {
2802 	int ret;
2803 
2804 	ret = 0;	/* okay to send */
2805 
2806 	/* PPS limit */
2807 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2808 	    V_icmp6errppslim)) {
2809 		/* The packet is subject to rate limit */
2810 		ret++;
2811 	}
2812 
2813 	return ret;
2814 }
2815