xref: /freebsd/sys/netinet6/icmp6.c (revision 52f72944b8f5abb2386eae924357dee8aea17d5b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
69 
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 
73 #include <sys/param.h>
74 #include <sys/domain.h>
75 #include <sys/jail.h>
76 #include <sys/kernel.h>
77 #include <sys/lock.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/signalvar.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sx.h>
86 #include <sys/syslog.h>
87 #include <sys/systm.h>
88 #include <sys/time.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/if_llatbl.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/in_var.h>
101 #include <netinet/ip6.h>
102 #include <netinet/icmp6.h>
103 #include <netinet/tcp_var.h>
104 
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/ip6protosw.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/send.h>
114 
115 extern struct domain inet6domain;
116 
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119 
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123 
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 VNET_DECLARE(struct inpcbhead, ripcb);
126 VNET_DECLARE(int, icmp6errppslim);
127 static VNET_DEFINE(int, icmp6errpps_count) = 0;
128 static VNET_DEFINE(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130 
131 #define	V_ripcbinfo			VNET(ripcbinfo)
132 #define	V_ripcb				VNET(ripcb)
133 #define	V_icmp6errppslim		VNET(icmp6errppslim)
134 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
135 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
136 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
137 
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
141 static const char *icmp6_redirect_diag(struct in6_addr *,
142 	struct in6_addr *, struct in6_addr *);
143 static struct mbuf *ni6_input(struct mbuf *, int);
144 static struct mbuf *ni6_nametodns(const char *, int, int);
145 static int ni6_dnsmatch(const char *, int, const char *, int);
146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
147 			  struct ifnet **, struct in6_addr *);
148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
149 				struct ifnet *, int);
150 static int icmp6_notify_error(struct mbuf **, int, int, int);
151 
152 /*
153  * Kernel module interface for updating icmp6stat.  The argument is an index
154  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
155  * general layout of icmp6stat into the caller, it doesn't encode its
156  * location, so that future changes to add, for example, per-CPU stats
157  * support won't cause binary compatibility problems for kernel modules.
158  */
159 void
160 kmod_icmp6stat_inc(int statnum)
161 {
162 
163 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
164 }
165 
166 static void
167 icmp6_errcount(int type, int code)
168 {
169 	switch (type) {
170 	case ICMP6_DST_UNREACH:
171 		switch (code) {
172 		case ICMP6_DST_UNREACH_NOROUTE:
173 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
174 			return;
175 		case ICMP6_DST_UNREACH_ADMIN:
176 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
177 			return;
178 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
179 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
180 			return;
181 		case ICMP6_DST_UNREACH_ADDR:
182 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
183 			return;
184 		case ICMP6_DST_UNREACH_NOPORT:
185 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
186 			return;
187 		}
188 		break;
189 	case ICMP6_PACKET_TOO_BIG:
190 		ICMP6STAT_INC(icp6s_opacket_too_big);
191 		return;
192 	case ICMP6_TIME_EXCEEDED:
193 		switch (code) {
194 		case ICMP6_TIME_EXCEED_TRANSIT:
195 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
196 			return;
197 		case ICMP6_TIME_EXCEED_REASSEMBLY:
198 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
199 			return;
200 		}
201 		break;
202 	case ICMP6_PARAM_PROB:
203 		switch (code) {
204 		case ICMP6_PARAMPROB_HEADER:
205 			ICMP6STAT_INC(icp6s_oparamprob_header);
206 			return;
207 		case ICMP6_PARAMPROB_NEXTHEADER:
208 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
209 			return;
210 		case ICMP6_PARAMPROB_OPTION:
211 			ICMP6STAT_INC(icp6s_oparamprob_option);
212 			return;
213 		}
214 		break;
215 	case ND_REDIRECT:
216 		ICMP6STAT_INC(icp6s_oredirect);
217 		return;
218 	}
219 	ICMP6STAT_INC(icp6s_ounknown);
220 }
221 
222 /*
223  * A wrapper function for icmp6_error() necessary when the erroneous packet
224  * may not contain enough scope zone information.
225  */
226 void
227 icmp6_error2(struct mbuf *m, int type, int code, int param,
228     struct ifnet *ifp)
229 {
230 	struct ip6_hdr *ip6;
231 
232 	if (ifp == NULL)
233 		return;
234 
235 #ifndef PULLDOWN_TEST
236 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
237 #else
238 	if (m->m_len < sizeof(struct ip6_hdr)) {
239 		m = m_pullup(m, sizeof(struct ip6_hdr));
240 		if (m == NULL)
241 			return;
242 	}
243 #endif
244 
245 	ip6 = mtod(m, struct ip6_hdr *);
246 
247 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
248 		return;
249 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
250 		return;
251 
252 	icmp6_error(m, type, code, param);
253 }
254 
255 /*
256  * Generate an error packet of type error in response to bad IP6 packet.
257  */
258 void
259 icmp6_error(struct mbuf *m, int type, int code, int param)
260 {
261 	struct ip6_hdr *oip6, *nip6;
262 	struct icmp6_hdr *icmp6;
263 	u_int preplen;
264 	int off;
265 	int nxt;
266 
267 	ICMP6STAT_INC(icp6s_error);
268 
269 	/* count per-type-code statistics */
270 	icmp6_errcount(type, code);
271 
272 #ifdef M_DECRYPTED	/*not openbsd*/
273 	if (m->m_flags & M_DECRYPTED) {
274 		ICMP6STAT_INC(icp6s_canterror);
275 		goto freeit;
276 	}
277 #endif
278 
279 #ifndef PULLDOWN_TEST
280 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
281 #else
282 	if (m->m_len < sizeof(struct ip6_hdr)) {
283 		m = m_pullup(m, sizeof(struct ip6_hdr));
284 		if (m == NULL)
285 			return;
286 	}
287 #endif
288 	oip6 = mtod(m, struct ip6_hdr *);
289 
290 	/*
291 	 * If the destination address of the erroneous packet is a multicast
292 	 * address, or the packet was sent using link-layer multicast,
293 	 * we should basically suppress sending an error (RFC 2463, Section
294 	 * 2.4).
295 	 * We have two exceptions (the item e.2 in that section):
296 	 * - the Packet Too Big message can be sent for path MTU discovery.
297 	 * - the Parameter Problem Message that can be allowed an icmp6 error
298 	 *   in the option type field.  This check has been done in
299 	 *   ip6_unknown_opt(), so we can just check the type and code.
300 	 */
301 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
302 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
303 	    (type != ICMP6_PACKET_TOO_BIG &&
304 	     (type != ICMP6_PARAM_PROB ||
305 	      code != ICMP6_PARAMPROB_OPTION)))
306 		goto freeit;
307 
308 	/*
309 	 * RFC 2463, 2.4 (e.5): source address check.
310 	 * XXX: the case of anycast source?
311 	 */
312 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
313 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
314 		goto freeit;
315 
316 	/*
317 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
318 	 * don't do it.
319 	 */
320 	nxt = -1;
321 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
322 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
323 		struct icmp6_hdr *icp;
324 
325 #ifndef PULLDOWN_TEST
326 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
327 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
328 #else
329 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
330 			sizeof(*icp));
331 		if (icp == NULL) {
332 			ICMP6STAT_INC(icp6s_tooshort);
333 			return;
334 		}
335 #endif
336 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
337 		    icp->icmp6_type == ND_REDIRECT) {
338 			/*
339 			 * ICMPv6 error
340 			 * Special case: for redirect (which is
341 			 * informational) we must not send icmp6 error.
342 			 */
343 			ICMP6STAT_INC(icp6s_canterror);
344 			goto freeit;
345 		} else {
346 			/* ICMPv6 informational - send the error */
347 		}
348 	} else {
349 		/* non-ICMPv6 - send the error */
350 	}
351 
352 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
353 
354 	/* Finally, do rate limitation check. */
355 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
356 		ICMP6STAT_INC(icp6s_toofreq);
357 		goto freeit;
358 	}
359 
360 	/*
361 	 * OK, ICMP6 can be generated.
362 	 */
363 
364 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
365 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
366 
367 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
368 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
369 	if (m == NULL) {
370 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
371 		return;
372 	}
373 
374 	nip6 = mtod(m, struct ip6_hdr *);
375 	nip6->ip6_src  = oip6->ip6_src;
376 	nip6->ip6_dst  = oip6->ip6_dst;
377 
378 	in6_clearscope(&oip6->ip6_src);
379 	in6_clearscope(&oip6->ip6_dst);
380 
381 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
382 	icmp6->icmp6_type = type;
383 	icmp6->icmp6_code = code;
384 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
385 
386 	/*
387 	 * icmp6_reflect() is designed to be in the input path.
388 	 * icmp6_error() can be called from both input and output path,
389 	 * and if we are in output path rcvif could contain bogus value.
390 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
391 	 * information in ip header (nip6).
392 	 */
393 	m->m_pkthdr.rcvif = NULL;
394 
395 	ICMP6STAT_INC(icp6s_outhist[type]);
396 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
397 
398 	return;
399 
400   freeit:
401 	/*
402 	 * If we can't tell whether or not we can generate ICMP6, free it.
403 	 */
404 	m_freem(m);
405 }
406 
407 /*
408  * Process a received ICMP6 message.
409  */
410 int
411 icmp6_input(struct mbuf **mp, int *offp, int proto)
412 {
413 	struct mbuf *m = *mp, *n;
414 	struct ifnet *ifp;
415 	struct ip6_hdr *ip6, *nip6;
416 	struct icmp6_hdr *icmp6, *nicmp6;
417 	int off = *offp;
418 	int icmp6len = m->m_pkthdr.len - *offp;
419 	int code, sum, noff;
420 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
421 	int ip6len, error;
422 
423 	ifp = m->m_pkthdr.rcvif;
424 
425 #ifndef PULLDOWN_TEST
426 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
427 	/* m might change if M_LOOP.  So, call mtod after this */
428 #endif
429 
430 	/*
431 	 * Locate icmp6 structure in mbuf, and check
432 	 * that not corrupted and of at least minimum length
433 	 */
434 
435 	ip6 = mtod(m, struct ip6_hdr *);
436 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
437 	if (icmp6len < sizeof(struct icmp6_hdr)) {
438 		ICMP6STAT_INC(icp6s_tooshort);
439 		goto freeit;
440 	}
441 
442 	/*
443 	 * Check multicast group membership.
444 	 * Note: SSM filters are not applied for ICMPv6 traffic.
445 	 */
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
447 		struct in6_multi	*inm;
448 
449 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
450 		if (inm == NULL) {
451 			IP6STAT_INC(ip6s_notmember);
452 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
453 			goto freeit;
454 		}
455 	}
456 
457 	/*
458 	 * calculate the checksum
459 	 */
460 #ifndef PULLDOWN_TEST
461 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
462 #else
463 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
464 	if (icmp6 == NULL) {
465 		ICMP6STAT_INC(icp6s_tooshort);
466 		return IPPROTO_DONE;
467 	}
468 #endif
469 	code = icmp6->icmp6_code;
470 
471 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
472 		nd6log((LOG_ERR,
473 		    "ICMP6 checksum error(%d|%x) %s\n",
474 		    icmp6->icmp6_type, sum,
475 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
476 		ICMP6STAT_INC(icp6s_checksum);
477 		goto freeit;
478 	}
479 
480 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
481 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
482 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
483 		icmp6_ifstat_inc(ifp, ifs6_in_error);
484 
485 	switch (icmp6->icmp6_type) {
486 	case ICMP6_DST_UNREACH:
487 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
488 		switch (code) {
489 		case ICMP6_DST_UNREACH_NOROUTE:
490 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
491 			code = PRC_UNREACH_NET;
492 			break;
493 		case ICMP6_DST_UNREACH_ADMIN:
494 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
495 			code = PRC_UNREACH_ADMIN_PROHIB;
496 			break;
497 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
498 			/* I mean "source address was incorrect." */
499 			code = PRC_PARAMPROB;
500 			break;
501 		case ICMP6_DST_UNREACH_NOPORT:
502 			code = PRC_UNREACH_PORT;
503 			break;
504 		default:
505 			goto badcode;
506 		}
507 		goto deliver;
508 		break;
509 
510 	case ICMP6_PACKET_TOO_BIG:
511 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
512 
513 		/* validation is made in icmp6_mtudisc_update */
514 
515 		code = PRC_MSGSIZE;
516 
517 		/*
518 		 * Updating the path MTU will be done after examining
519 		 * intermediate extension headers.
520 		 */
521 		goto deliver;
522 		break;
523 
524 	case ICMP6_TIME_EXCEEDED:
525 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
526 		switch (code) {
527 		case ICMP6_TIME_EXCEED_TRANSIT:
528 			code = PRC_TIMXCEED_INTRANS;
529 			break;
530 		case ICMP6_TIME_EXCEED_REASSEMBLY:
531 			code = PRC_TIMXCEED_REASS;
532 			break;
533 		default:
534 			goto badcode;
535 		}
536 		goto deliver;
537 		break;
538 
539 	case ICMP6_PARAM_PROB:
540 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
541 		switch (code) {
542 		case ICMP6_PARAMPROB_NEXTHEADER:
543 			code = PRC_UNREACH_PROTOCOL;
544 			break;
545 		case ICMP6_PARAMPROB_HEADER:
546 		case ICMP6_PARAMPROB_OPTION:
547 			code = PRC_PARAMPROB;
548 			break;
549 		default:
550 			goto badcode;
551 		}
552 		goto deliver;
553 		break;
554 
555 	case ICMP6_ECHO_REQUEST:
556 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
557 		if (code != 0)
558 			goto badcode;
559 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
560 			/* Give up remote */
561 			break;
562 		}
563 		if (!M_WRITABLE(n)
564 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
565 			struct mbuf *n0 = n;
566 			int n0len;
567 
568 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
569 			n = m_gethdr(M_NOWAIT, n0->m_type);
570 			if (n == NULL) {
571 				/* Give up remote */
572 				m_freem(n0);
573 				break;
574 			}
575 
576 			m_move_pkthdr(n, n0);	/* FIB copied. */
577 			n0len = n0->m_pkthdr.len;	/* save for use below */
578 			/*
579 			 * Copy IPv6 and ICMPv6 only.
580 			 */
581 			nip6 = mtod(n, struct ip6_hdr *);
582 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
583 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
584 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
585 			noff = sizeof(struct ip6_hdr);
586 			/* new mbuf contains only ipv6+icmpv6 headers */
587 			n->m_len = noff + sizeof(struct icmp6_hdr);
588 			/*
589 			 * Adjust mbuf.  ip6_plen will be adjusted in
590 			 * ip6_output().
591 			 */
592 			m_adj(n0, off + sizeof(struct icmp6_hdr));
593 			/* recalculate complete packet size */
594 			n->m_pkthdr.len = n0len + (noff - off);
595 			n->m_next = n0;
596 		} else {
597 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
598 			    sizeof(*nicmp6));
599 			noff = off;
600 		}
601 		if (n) {
602 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
603 			nicmp6->icmp6_code = 0;
604 			ICMP6STAT_INC(icp6s_reflect);
605 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
606 			icmp6_reflect(n, noff);
607 		}
608 		break;
609 
610 	case ICMP6_ECHO_REPLY:
611 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
612 		if (code != 0)
613 			goto badcode;
614 		break;
615 
616 	case MLD_LISTENER_QUERY:
617 	case MLD_LISTENER_REPORT:
618 	case MLD_LISTENER_DONE:
619 	case MLDV2_LISTENER_REPORT:
620 		/*
621 		 * Drop MLD traffic which is not link-local, has a hop limit
622 		 * of greater than 1 hop, or which does not have the
623 		 * IPv6 HBH Router Alert option.
624 		 * As IPv6 HBH options are stripped in ip6_input() we must
625 		 * check an mbuf header flag.
626 		 * XXX Should we also sanity check that these messages
627 		 * were directed to a link-local multicast prefix?
628 		 */
629 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
630 			goto freeit;
631 		if (mld_input(m, off, icmp6len) != 0)
632 			return (IPPROTO_DONE);
633 		/* m stays. */
634 		break;
635 
636 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
637 	    {
638 		enum { WRU, FQDN } mode;
639 
640 		if (!V_icmp6_nodeinfo)
641 			break;
642 
643 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
644 			mode = WRU;
645 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
646 			mode = FQDN;
647 		else
648 			goto badlen;
649 
650 		if (mode == FQDN) {
651 #ifndef PULLDOWN_TEST
652 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
653 			    IPPROTO_DONE);
654 #endif
655 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
656 			if (n)
657 				n = ni6_input(n, off);
658 			/* XXX meaningless if n == NULL */
659 			noff = sizeof(struct ip6_hdr);
660 		} else {
661 			struct prison *pr;
662 			u_char *p;
663 			int maxhlen, hlen;
664 
665 			/*
666 			 * XXX: this combination of flags is pointless,
667 			 * but should we keep this for compatibility?
668 			 */
669 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
670 			    ICMP6_NODEINFO_TMPADDROK)) !=
671 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
672 				break;
673 
674 			if (code != 0)
675 				goto badcode;
676 
677 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
678 			n = m_gethdr(M_NOWAIT, m->m_type);
679 			if (n == NULL) {
680 				/* Give up remote */
681 				break;
682 			}
683 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
684 				/*
685 				 * Previous code did a blind M_COPY_PKTHDR
686 				 * and said "just for rcvif".  If true, then
687 				 * we could tolerate the dup failing (due to
688 				 * the deep copy of the tag chain).  For now
689 				 * be conservative and just fail.
690 				 */
691 				m_free(n);
692 				n = NULL;
693 				break;
694 			}
695 			maxhlen = M_TRAILINGSPACE(n) -
696 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
697 			pr = curthread->td_ucred->cr_prison;
698 			mtx_lock(&pr->pr_mtx);
699 			hlen = strlen(pr->pr_hostname);
700 			if (maxhlen > hlen)
701 				maxhlen = hlen;
702 			/*
703 			 * Copy IPv6 and ICMPv6 only.
704 			 */
705 			nip6 = mtod(n, struct ip6_hdr *);
706 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
707 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
708 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
709 			p = (u_char *)(nicmp6 + 1);
710 			bzero(p, 4);
711 			/* meaningless TTL */
712 			bcopy(pr->pr_hostname, p + 4, maxhlen);
713 			mtx_unlock(&pr->pr_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 		if (n) {
721 			ICMP6STAT_INC(icp6s_reflect);
722 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
723 			icmp6_reflect(n, noff);
724 		}
725 		break;
726 	    }
727 
728 	case ICMP6_WRUREPLY:
729 		if (code != 0)
730 			goto badcode;
731 		break;
732 
733 	case ND_ROUTER_SOLICIT:
734 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
735 		if (code != 0)
736 			goto badcode;
737 		if (icmp6len < sizeof(struct nd_router_solicit))
738 			goto badlen;
739 		if (send_sendso_input_hook != NULL) {
740 			IP6_EXTHDR_CHECK(m, off, icmp6len, IPPROTO_DONE);
741 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
742 			if (error == 0) {
743 				m = NULL;
744 				goto freeit;
745 			}
746 		}
747 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
748 		nd6_rs_input(m, off, icmp6len);
749 		m = n;
750 		if (m == NULL)
751 			goto freeit;
752 		break;
753 
754 	case ND_ROUTER_ADVERT:
755 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
756 		if (code != 0)
757 			goto badcode;
758 		if (icmp6len < sizeof(struct nd_router_advert))
759 			goto badlen;
760 		if (send_sendso_input_hook != NULL) {
761 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
762 			if (error == 0) {
763 				m = NULL;
764 				goto freeit;
765 			}
766 		}
767 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
768 		nd6_ra_input(m, off, icmp6len);
769 		m = n;
770 		if (m == NULL)
771 			goto freeit;
772 		break;
773 
774 	case ND_NEIGHBOR_SOLICIT:
775 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
776 		if (code != 0)
777 			goto badcode;
778 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
779 			goto badlen;
780 		if (send_sendso_input_hook != NULL) {
781 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
782 			if (error == 0) {
783 				m = NULL;
784 				goto freeit;
785 			}
786 		}
787 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
788 		nd6_ns_input(m, off, icmp6len);
789 		m = n;
790 		if (m == NULL)
791 			goto freeit;
792 		break;
793 
794 	case ND_NEIGHBOR_ADVERT:
795 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
796 		if (code != 0)
797 			goto badcode;
798 		if (icmp6len < sizeof(struct nd_neighbor_advert))
799 			goto badlen;
800 		if (send_sendso_input_hook != NULL) {
801 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
802 			if (error == 0) {
803 				m = NULL;
804 				goto freeit;
805 			}
806 		}
807 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
808 		nd6_na_input(m, off, icmp6len);
809 		m = n;
810 		if (m == NULL)
811 			goto freeit;
812 		break;
813 
814 	case ND_REDIRECT:
815 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
816 		if (code != 0)
817 			goto badcode;
818 		if (icmp6len < sizeof(struct nd_redirect))
819 			goto badlen;
820 		if (send_sendso_input_hook != NULL) {
821 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
822 			if (error == 0) {
823 				m = NULL;
824 				goto freeit;
825 			}
826 		}
827 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
828 		icmp6_redirect_input(m, off);
829 		m = n;
830 		if (m == NULL)
831 			goto freeit;
832 		break;
833 
834 	case ICMP6_ROUTER_RENUMBERING:
835 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
836 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
837 			goto badcode;
838 		if (icmp6len < sizeof(struct icmp6_router_renum))
839 			goto badlen;
840 		break;
841 
842 	default:
843 		nd6log((LOG_DEBUG,
844 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
845 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
846 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
847 		    ifp ? ifp->if_index : 0));
848 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
849 			/* ICMPv6 error: MUST deliver it by spec... */
850 			code = PRC_NCMDS;
851 			/* deliver */
852 		} else {
853 			/* ICMPv6 informational: MUST not deliver */
854 			break;
855 		}
856 	deliver:
857 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
858 			/* In this case, m should've been freed. */
859 			return (IPPROTO_DONE);
860 		}
861 		break;
862 
863 	badcode:
864 		ICMP6STAT_INC(icp6s_badcode);
865 		break;
866 
867 	badlen:
868 		ICMP6STAT_INC(icp6s_badlen);
869 		break;
870 	}
871 
872 	/* deliver the packet to appropriate sockets */
873 	icmp6_rip6_input(&m, *offp);
874 
875 	return IPPROTO_DONE;
876 
877  freeit:
878 	m_freem(m);
879 	return IPPROTO_DONE;
880 }
881 
882 static int
883 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
884 {
885 	struct mbuf *m = *mp;
886 	struct icmp6_hdr *icmp6;
887 	struct ip6_hdr *eip6;
888 	u_int32_t notifymtu;
889 	struct sockaddr_in6 icmp6src, icmp6dst;
890 
891 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
892 		ICMP6STAT_INC(icp6s_tooshort);
893 		goto freeit;
894 	}
895 #ifndef PULLDOWN_TEST
896 	IP6_EXTHDR_CHECK(m, off,
897 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
898 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
899 #else
900 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
901 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
902 	if (icmp6 == NULL) {
903 		ICMP6STAT_INC(icp6s_tooshort);
904 		return (-1);
905 	}
906 #endif
907 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
908 
909 	/* Detect the upper level protocol */
910 	{
911 		void (*ctlfunc)(int, struct sockaddr *, void *);
912 		u_int8_t nxt = eip6->ip6_nxt;
913 		int eoff = off + sizeof(struct icmp6_hdr) +
914 		    sizeof(struct ip6_hdr);
915 		struct ip6ctlparam ip6cp;
916 		struct in6_addr *finaldst = NULL;
917 		int icmp6type = icmp6->icmp6_type;
918 		struct ip6_frag *fh;
919 		struct ip6_rthdr *rth;
920 		struct ip6_rthdr0 *rth0;
921 		int rthlen;
922 
923 		while (1) { /* XXX: should avoid infinite loop explicitly? */
924 			struct ip6_ext *eh;
925 
926 			switch (nxt) {
927 			case IPPROTO_HOPOPTS:
928 			case IPPROTO_DSTOPTS:
929 			case IPPROTO_AH:
930 #ifndef PULLDOWN_TEST
931 				IP6_EXTHDR_CHECK(m, 0,
932 				    eoff + sizeof(struct ip6_ext), -1);
933 				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
934 #else
935 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
936 				    eoff, sizeof(*eh));
937 				if (eh == NULL) {
938 					ICMP6STAT_INC(icp6s_tooshort);
939 					return (-1);
940 				}
941 #endif
942 
943 				if (nxt == IPPROTO_AH)
944 					eoff += (eh->ip6e_len + 2) << 2;
945 				else
946 					eoff += (eh->ip6e_len + 1) << 3;
947 				nxt = eh->ip6e_nxt;
948 				break;
949 			case IPPROTO_ROUTING:
950 				/*
951 				 * When the erroneous packet contains a
952 				 * routing header, we should examine the
953 				 * header to determine the final destination.
954 				 * Otherwise, we can't properly update
955 				 * information that depends on the final
956 				 * destination (e.g. path MTU).
957 				 */
958 #ifndef PULLDOWN_TEST
959 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
960 				rth = (struct ip6_rthdr *)
961 				    (mtod(m, caddr_t) + eoff);
962 #else
963 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
964 				    eoff, sizeof(*rth));
965 				if (rth == NULL) {
966 					ICMP6STAT_INC(icp6s_tooshort);
967 					return (-1);
968 				}
969 #endif
970 				rthlen = (rth->ip6r_len + 1) << 3;
971 				/*
972 				 * XXX: currently there is no
973 				 * officially defined type other
974 				 * than type-0.
975 				 * Note that if the segment left field
976 				 * is 0, all intermediate hops must
977 				 * have been passed.
978 				 */
979 				if (rth->ip6r_segleft &&
980 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
981 					int hops;
982 
983 #ifndef PULLDOWN_TEST
984 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
985 					rth0 = (struct ip6_rthdr0 *)
986 					    (mtod(m, caddr_t) + eoff);
987 #else
988 					IP6_EXTHDR_GET(rth0,
989 					    struct ip6_rthdr0 *, m,
990 					    eoff, rthlen);
991 					if (rth0 == NULL) {
992 						ICMP6STAT_INC(icp6s_tooshort);
993 						return (-1);
994 					}
995 #endif
996 					/* just ignore a bogus header */
997 					if ((rth0->ip6r0_len % 2) == 0 &&
998 					    (hops = rth0->ip6r0_len/2))
999 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1000 				}
1001 				eoff += rthlen;
1002 				nxt = rth->ip6r_nxt;
1003 				break;
1004 			case IPPROTO_FRAGMENT:
1005 #ifndef PULLDOWN_TEST
1006 				IP6_EXTHDR_CHECK(m, 0, eoff +
1007 				    sizeof(struct ip6_frag), -1);
1008 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1009 				    eoff);
1010 #else
1011 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1012 				    eoff, sizeof(*fh));
1013 				if (fh == NULL) {
1014 					ICMP6STAT_INC(icp6s_tooshort);
1015 					return (-1);
1016 				}
1017 #endif
1018 				/*
1019 				 * Data after a fragment header is meaningless
1020 				 * unless it is the first fragment, but
1021 				 * we'll go to the notify label for path MTU
1022 				 * discovery.
1023 				 */
1024 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1025 					goto notify;
1026 
1027 				eoff += sizeof(struct ip6_frag);
1028 				nxt = fh->ip6f_nxt;
1029 				break;
1030 			default:
1031 				/*
1032 				 * This case includes ESP and the No Next
1033 				 * Header.  In such cases going to the notify
1034 				 * label does not have any meaning
1035 				 * (i.e. ctlfunc will be NULL), but we go
1036 				 * anyway since we might have to update
1037 				 * path MTU information.
1038 				 */
1039 				goto notify;
1040 			}
1041 		}
1042 	  notify:
1043 #ifndef PULLDOWN_TEST
1044 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1045 #else
1046 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1047 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1048 		if (icmp6 == NULL) {
1049 			ICMP6STAT_INC(icp6s_tooshort);
1050 			return (-1);
1051 		}
1052 #endif
1053 
1054 		/*
1055 		 * retrieve parameters from the inner IPv6 header, and convert
1056 		 * them into sockaddr structures.
1057 		 * XXX: there is no guarantee that the source or destination
1058 		 * addresses of the inner packet are in the same scope as
1059 		 * the addresses of the icmp packet.  But there is no other
1060 		 * way to determine the zone.
1061 		 */
1062 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1063 
1064 		bzero(&icmp6dst, sizeof(icmp6dst));
1065 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1066 		icmp6dst.sin6_family = AF_INET6;
1067 		if (finaldst == NULL)
1068 			icmp6dst.sin6_addr = eip6->ip6_dst;
1069 		else
1070 			icmp6dst.sin6_addr = *finaldst;
1071 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1072 			goto freeit;
1073 		bzero(&icmp6src, sizeof(icmp6src));
1074 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1075 		icmp6src.sin6_family = AF_INET6;
1076 		icmp6src.sin6_addr = eip6->ip6_src;
1077 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1078 			goto freeit;
1079 		icmp6src.sin6_flowinfo =
1080 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1081 
1082 		if (finaldst == NULL)
1083 			finaldst = &eip6->ip6_dst;
1084 		ip6cp.ip6c_m = m;
1085 		ip6cp.ip6c_icmp6 = icmp6;
1086 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1087 		ip6cp.ip6c_off = eoff;
1088 		ip6cp.ip6c_finaldst = finaldst;
1089 		ip6cp.ip6c_src = &icmp6src;
1090 		ip6cp.ip6c_nxt = nxt;
1091 
1092 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1093 			notifymtu = ntohl(icmp6->icmp6_mtu);
1094 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1095 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1096 		}
1097 
1098 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1099 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1100 		if (ctlfunc) {
1101 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1102 			    &ip6cp);
1103 		}
1104 	}
1105 	*mp = m;
1106 	return (0);
1107 
1108   freeit:
1109 	m_freem(m);
1110 	return (-1);
1111 }
1112 
1113 void
1114 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1115 {
1116 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1117 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1118 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1119 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1120 	struct in_conninfo inc;
1121 
1122 #if 0
1123 	/*
1124 	 * RFC2460 section 5, last paragraph.
1125 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1126 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1127 	 * due to packet translator in the middle.
1128 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1129 	 * special handling.
1130 	 */
1131 	if (mtu < IPV6_MMTU)
1132 		return;
1133 #endif
1134 
1135 	/*
1136 	 * we reject ICMPv6 too big with abnormally small value.
1137 	 * XXX what is the good definition of "abnormally small"?
1138 	 */
1139 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1140 		return;
1141 
1142 	if (!validated)
1143 		return;
1144 
1145 	/*
1146 	 * In case the suggested mtu is less than IPV6_MMTU, we
1147 	 * only need to remember that it was for above mentioned
1148 	 * "alwaysfrag" case.
1149 	 * Try to be as close to the spec as possible.
1150 	 */
1151 	if (mtu < IPV6_MMTU)
1152 		mtu = IPV6_MMTU - 8;
1153 
1154 	bzero(&inc, sizeof(inc));
1155 	inc.inc_fibnum = M_GETFIB(m);
1156 	inc.inc_flags |= INC_ISIPV6;
1157 	inc.inc6_faddr = *dst;
1158 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1159 		return;
1160 
1161 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1162 		tcp_hc_updatemtu(&inc, mtu);
1163 		ICMP6STAT_INC(icp6s_pmtuchg);
1164 	}
1165 }
1166 
1167 /*
1168  * Process a Node Information Query packet, based on
1169  * draft-ietf-ipngwg-icmp-name-lookups-07.
1170  *
1171  * Spec incompatibilities:
1172  * - IPv6 Subject address handling
1173  * - IPv4 Subject address handling support missing
1174  * - Proxy reply (answer even if it's not for me)
1175  * - joins NI group address at in6_ifattach() time only, does not cope
1176  *   with hostname changes by sethostname(3)
1177  */
1178 static struct mbuf *
1179 ni6_input(struct mbuf *m, int off)
1180 {
1181 	struct icmp6_nodeinfo *ni6, *nni6;
1182 	struct mbuf *n = NULL;
1183 	struct prison *pr;
1184 	u_int16_t qtype;
1185 	int subjlen;
1186 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1187 	struct ni_reply_fqdn *fqdn;
1188 	int addrs;		/* for NI_QTYPE_NODEADDR */
1189 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1190 	struct in6_addr in6_subj; /* subject address */
1191 	struct ip6_hdr *ip6;
1192 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1193 	char *subj = NULL;
1194 	struct in6_ifaddr *ia6 = NULL;
1195 
1196 	ip6 = mtod(m, struct ip6_hdr *);
1197 #ifndef PULLDOWN_TEST
1198 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1199 #else
1200 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1201 	if (ni6 == NULL) {
1202 		/* m is already reclaimed */
1203 		return (NULL);
1204 	}
1205 #endif
1206 
1207 	/*
1208 	 * Validate IPv6 source address.
1209 	 * The default configuration MUST be to refuse answering queries from
1210 	 * global-scope addresses according to RFC4602.
1211 	 * Notes:
1212 	 *  - it's not very clear what "refuse" means; this implementation
1213 	 *    simply drops it.
1214 	 *  - it's not very easy to identify global-scope (unicast) addresses
1215 	 *    since there are many prefixes for them.  It should be safer
1216 	 *    and in practice sufficient to check "all" but loopback and
1217 	 *    link-local (note that site-local unicast was deprecated and
1218 	 *    ULA is defined as global scope-wise)
1219 	 */
1220 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1221 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1222 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1223 		goto bad;
1224 
1225 	/*
1226 	 * Validate IPv6 destination address.
1227 	 *
1228 	 * The Responder must discard the Query without further processing
1229 	 * unless it is one of the Responder's unicast or anycast addresses, or
1230 	 * a link-local scope multicast address which the Responder has joined.
1231 	 * [RFC4602, Section 5.]
1232 	 */
1233 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1234 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1235 			goto bad;
1236 		/* else it's a link-local multicast, fine */
1237 	} else {		/* unicast or anycast */
1238 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1239 		if (ia6 == NULL)
1240 			goto bad; /* XXX impossible */
1241 
1242 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1243 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1244 			ifa_free(&ia6->ia_ifa);
1245 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1246 				"a temporary address in %s:%d",
1247 			       __FILE__, __LINE__));
1248 			goto bad;
1249 		}
1250 		ifa_free(&ia6->ia_ifa);
1251 	}
1252 
1253 	/* validate query Subject field. */
1254 	qtype = ntohs(ni6->ni_qtype);
1255 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1256 	switch (qtype) {
1257 	case NI_QTYPE_NOOP:
1258 	case NI_QTYPE_SUPTYPES:
1259 		/* 07 draft */
1260 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1261 			break;
1262 		/* FALLTHROUGH */
1263 	case NI_QTYPE_FQDN:
1264 	case NI_QTYPE_NODEADDR:
1265 	case NI_QTYPE_IPV4ADDR:
1266 		switch (ni6->ni_code) {
1267 		case ICMP6_NI_SUBJ_IPV6:
1268 #if ICMP6_NI_SUBJ_IPV6 != 0
1269 		case 0:
1270 #endif
1271 			/*
1272 			 * backward compatibility - try to accept 03 draft
1273 			 * format, where no Subject is present.
1274 			 */
1275 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1276 			    subjlen == 0) {
1277 				oldfqdn++;
1278 				break;
1279 			}
1280 #if ICMP6_NI_SUBJ_IPV6 != 0
1281 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1282 				goto bad;
1283 #endif
1284 
1285 			if (subjlen != sizeof(struct in6_addr))
1286 				goto bad;
1287 
1288 			/*
1289 			 * Validate Subject address.
1290 			 *
1291 			 * Not sure what exactly "address belongs to the node"
1292 			 * means in the spec, is it just unicast, or what?
1293 			 *
1294 			 * At this moment we consider Subject address as
1295 			 * "belong to the node" if the Subject address equals
1296 			 * to the IPv6 destination address; validation for
1297 			 * IPv6 destination address should have done enough
1298 			 * check for us.
1299 			 *
1300 			 * We do not do proxy at this moment.
1301 			 */
1302 			/* m_pulldown instead of copy? */
1303 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1304 			    subjlen, (caddr_t)&in6_subj);
1305 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1306 				goto bad;
1307 
1308 			subj = (char *)&in6_subj;
1309 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1310 				break;
1311 
1312 			/*
1313 			 * XXX if we are to allow other cases, we should really
1314 			 * be careful about scope here.
1315 			 * basically, we should disallow queries toward IPv6
1316 			 * destination X with subject Y,
1317 			 * if scope(X) > scope(Y).
1318 			 * if we allow scope(X) > scope(Y), it will result in
1319 			 * information leakage across scope boundary.
1320 			 */
1321 			goto bad;
1322 
1323 		case ICMP6_NI_SUBJ_FQDN:
1324 			/*
1325 			 * Validate Subject name with gethostname(3).
1326 			 *
1327 			 * The behavior may need some debate, since:
1328 			 * - we are not sure if the node has FQDN as
1329 			 *   hostname (returned by gethostname(3)).
1330 			 * - the code does wildcard match for truncated names.
1331 			 *   however, we are not sure if we want to perform
1332 			 *   wildcard match, if gethostname(3) side has
1333 			 *   truncated hostname.
1334 			 */
1335 			pr = curthread->td_ucred->cr_prison;
1336 			mtx_lock(&pr->pr_mtx);
1337 			n = ni6_nametodns(pr->pr_hostname,
1338 			    strlen(pr->pr_hostname), 0);
1339 			mtx_unlock(&pr->pr_mtx);
1340 			if (!n || n->m_next || n->m_len == 0)
1341 				goto bad;
1342 			IP6_EXTHDR_GET(subj, char *, m,
1343 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1344 			if (subj == NULL)
1345 				goto bad;
1346 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1347 			    n->m_len)) {
1348 				goto bad;
1349 			}
1350 			m_freem(n);
1351 			n = NULL;
1352 			break;
1353 
1354 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1355 		default:
1356 			goto bad;
1357 		}
1358 		break;
1359 	}
1360 
1361 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1362 	switch (qtype) {
1363 	case NI_QTYPE_FQDN:
1364 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1365 			goto bad;
1366 		break;
1367 	case NI_QTYPE_NODEADDR:
1368 	case NI_QTYPE_IPV4ADDR:
1369 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1370 			goto bad;
1371 		break;
1372 	}
1373 
1374 	/* guess reply length */
1375 	switch (qtype) {
1376 	case NI_QTYPE_NOOP:
1377 		break;		/* no reply data */
1378 	case NI_QTYPE_SUPTYPES:
1379 		replylen += sizeof(u_int32_t);
1380 		break;
1381 	case NI_QTYPE_FQDN:
1382 		/* XXX will append an mbuf */
1383 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1384 		break;
1385 	case NI_QTYPE_NODEADDR:
1386 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1387 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1388 		    sizeof(u_int32_t))) > MCLBYTES)
1389 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1390 		break;
1391 	case NI_QTYPE_IPV4ADDR:
1392 		/* unsupported - should respond with unknown Qtype? */
1393 		break;
1394 	default:
1395 		/*
1396 		 * XXX: We must return a reply with the ICMP6 code
1397 		 * `unknown Qtype' in this case.  However we regard the case
1398 		 * as an FQDN query for backward compatibility.
1399 		 * Older versions set a random value to this field,
1400 		 * so it rarely varies in the defined qtypes.
1401 		 * But the mechanism is not reliable...
1402 		 * maybe we should obsolete older versions.
1403 		 */
1404 		qtype = NI_QTYPE_FQDN;
1405 		/* XXX will append an mbuf */
1406 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1407 		oldfqdn++;
1408 		break;
1409 	}
1410 
1411 	/* Allocate an mbuf to reply. */
1412 	if (replylen > MCLBYTES) {
1413 		/*
1414 		 * XXX: should we try to allocate more? But MCLBYTES
1415 		 * is probably much larger than IPV6_MMTU...
1416 		 */
1417 		goto bad;
1418 	}
1419 	if (replylen > MHLEN)
1420 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1421 	else
1422 		n = m_gethdr(M_NOWAIT, m->m_type);
1423 	if (n == NULL) {
1424 		m_freem(m);
1425 		return (NULL);
1426 	}
1427 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1428 	n->m_pkthdr.len = n->m_len = replylen;
1429 
1430 	/* copy mbuf header and IPv6 + Node Information base headers */
1431 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1432 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1433 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1434 
1435 	/* qtype dependent procedure */
1436 	switch (qtype) {
1437 	case NI_QTYPE_NOOP:
1438 		nni6->ni_code = ICMP6_NI_SUCCESS;
1439 		nni6->ni_flags = 0;
1440 		break;
1441 	case NI_QTYPE_SUPTYPES:
1442 	{
1443 		u_int32_t v;
1444 		nni6->ni_code = ICMP6_NI_SUCCESS;
1445 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1446 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1447 		v = (u_int32_t)htonl(0x0000000f);
1448 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1449 		break;
1450 	}
1451 	case NI_QTYPE_FQDN:
1452 		nni6->ni_code = ICMP6_NI_SUCCESS;
1453 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1454 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1455 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1456 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1457 		/*
1458 		 * XXX do we really have FQDN in hostname?
1459 		 */
1460 		pr = curthread->td_ucred->cr_prison;
1461 		mtx_lock(&pr->pr_mtx);
1462 		n->m_next = ni6_nametodns(pr->pr_hostname,
1463 		    strlen(pr->pr_hostname), oldfqdn);
1464 		mtx_unlock(&pr->pr_mtx);
1465 		if (n->m_next == NULL)
1466 			goto bad;
1467 		/* XXX we assume that n->m_next is not a chain */
1468 		if (n->m_next->m_next != NULL)
1469 			goto bad;
1470 		n->m_pkthdr.len += n->m_next->m_len;
1471 		break;
1472 	case NI_QTYPE_NODEADDR:
1473 	{
1474 		int lenlim, copied;
1475 
1476 		nni6->ni_code = ICMP6_NI_SUCCESS;
1477 		n->m_pkthdr.len = n->m_len =
1478 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1479 		lenlim = M_TRAILINGSPACE(n);
1480 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1481 		/* XXX: reset mbuf length */
1482 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1483 		    sizeof(struct icmp6_nodeinfo) + copied;
1484 		break;
1485 	}
1486 	default:
1487 		break;		/* XXX impossible! */
1488 	}
1489 
1490 	nni6->ni_type = ICMP6_NI_REPLY;
1491 	m_freem(m);
1492 	return (n);
1493 
1494   bad:
1495 	m_freem(m);
1496 	if (n)
1497 		m_freem(n);
1498 	return (NULL);
1499 }
1500 
1501 /*
1502  * make a mbuf with DNS-encoded string.  no compression support.
1503  *
1504  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1505  * treated as truncated name (two \0 at the end).  this is a wild guess.
1506  *
1507  * old - return pascal string if non-zero
1508  */
1509 static struct mbuf *
1510 ni6_nametodns(const char *name, int namelen, int old)
1511 {
1512 	struct mbuf *m;
1513 	char *cp, *ep;
1514 	const char *p, *q;
1515 	int i, len, nterm;
1516 
1517 	if (old)
1518 		len = namelen + 1;
1519 	else
1520 		len = MCLBYTES;
1521 
1522 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1523 	if (len > MLEN)
1524 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1525 	else
1526 		m = m_get(M_NOWAIT, MT_DATA);
1527 	if (m == NULL)
1528 		goto fail;
1529 
1530 	if (old) {
1531 		m->m_len = len;
1532 		*mtod(m, char *) = namelen;
1533 		bcopy(name, mtod(m, char *) + 1, namelen);
1534 		return m;
1535 	} else {
1536 		m->m_len = 0;
1537 		cp = mtod(m, char *);
1538 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1539 
1540 		/* if not certain about my name, return empty buffer */
1541 		if (namelen == 0)
1542 			return m;
1543 
1544 		/*
1545 		 * guess if it looks like shortened hostname, or FQDN.
1546 		 * shortened hostname needs two trailing "\0".
1547 		 */
1548 		i = 0;
1549 		for (p = name; p < name + namelen; p++) {
1550 			if (*p && *p == '.')
1551 				i++;
1552 		}
1553 		if (i < 2)
1554 			nterm = 2;
1555 		else
1556 			nterm = 1;
1557 
1558 		p = name;
1559 		while (cp < ep && p < name + namelen) {
1560 			i = 0;
1561 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1562 				i++;
1563 			/* result does not fit into mbuf */
1564 			if (cp + i + 1 >= ep)
1565 				goto fail;
1566 			/*
1567 			 * DNS label length restriction, RFC1035 page 8.
1568 			 * "i == 0" case is included here to avoid returning
1569 			 * 0-length label on "foo..bar".
1570 			 */
1571 			if (i <= 0 || i >= 64)
1572 				goto fail;
1573 			*cp++ = i;
1574 			bcopy(p, cp, i);
1575 			cp += i;
1576 			p = q;
1577 			if (p < name + namelen && *p == '.')
1578 				p++;
1579 		}
1580 		/* termination */
1581 		if (cp + nterm >= ep)
1582 			goto fail;
1583 		while (nterm-- > 0)
1584 			*cp++ = '\0';
1585 		m->m_len = cp - mtod(m, char *);
1586 		return m;
1587 	}
1588 
1589 	panic("should not reach here");
1590 	/* NOTREACHED */
1591 
1592  fail:
1593 	if (m)
1594 		m_freem(m);
1595 	return NULL;
1596 }
1597 
1598 /*
1599  * check if two DNS-encoded string matches.  takes care of truncated
1600  * form (with \0\0 at the end).  no compression support.
1601  * XXX upper/lowercase match (see RFC2065)
1602  */
1603 static int
1604 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1605 {
1606 	const char *a0, *b0;
1607 	int l;
1608 
1609 	/* simplest case - need validation? */
1610 	if (alen == blen && bcmp(a, b, alen) == 0)
1611 		return 1;
1612 
1613 	a0 = a;
1614 	b0 = b;
1615 
1616 	/* termination is mandatory */
1617 	if (alen < 2 || blen < 2)
1618 		return 0;
1619 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1620 		return 0;
1621 	alen--;
1622 	blen--;
1623 
1624 	while (a - a0 < alen && b - b0 < blen) {
1625 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1626 			return 0;
1627 
1628 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1629 			return 0;
1630 		/* we don't support compression yet */
1631 		if (a[0] >= 64 || b[0] >= 64)
1632 			return 0;
1633 
1634 		/* truncated case */
1635 		if (a[0] == 0 && a - a0 == alen - 1)
1636 			return 1;
1637 		if (b[0] == 0 && b - b0 == blen - 1)
1638 			return 1;
1639 		if (a[0] == 0 || b[0] == 0)
1640 			return 0;
1641 
1642 		if (a[0] != b[0])
1643 			return 0;
1644 		l = a[0];
1645 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1646 			return 0;
1647 		if (bcmp(a + 1, b + 1, l) != 0)
1648 			return 0;
1649 
1650 		a += 1 + l;
1651 		b += 1 + l;
1652 	}
1653 
1654 	if (a - a0 == alen && b - b0 == blen)
1655 		return 1;
1656 	else
1657 		return 0;
1658 }
1659 
1660 /*
1661  * calculate the number of addresses to be returned in the node info reply.
1662  */
1663 static int
1664 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1665     struct in6_addr *subj)
1666 {
1667 	struct ifnet *ifp;
1668 	struct in6_ifaddr *ifa6;
1669 	struct ifaddr *ifa;
1670 	int addrs = 0, addrsofif, iffound = 0;
1671 	int niflags = ni6->ni_flags;
1672 
1673 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1674 		switch (ni6->ni_code) {
1675 		case ICMP6_NI_SUBJ_IPV6:
1676 			if (subj == NULL) /* must be impossible... */
1677 				return (0);
1678 			break;
1679 		default:
1680 			/*
1681 			 * XXX: we only support IPv6 subject address for
1682 			 * this Qtype.
1683 			 */
1684 			return (0);
1685 		}
1686 	}
1687 
1688 	IFNET_RLOCK_NOSLEEP();
1689 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1690 		addrsofif = 0;
1691 		IF_ADDR_RLOCK(ifp);
1692 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1693 			if (ifa->ifa_addr->sa_family != AF_INET6)
1694 				continue;
1695 			ifa6 = (struct in6_ifaddr *)ifa;
1696 
1697 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1698 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1699 				iffound = 1;
1700 
1701 			/*
1702 			 * IPv4-mapped addresses can only be returned by a
1703 			 * Node Information proxy, since they represent
1704 			 * addresses of IPv4-only nodes, which perforce do
1705 			 * not implement this protocol.
1706 			 * [icmp-name-lookups-07, Section 5.4]
1707 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1708 			 * this function at this moment.
1709 			 */
1710 
1711 			/* What do we have to do about ::1? */
1712 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1713 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_SITELOCAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1719 					continue;
1720 				break;
1721 			case IPV6_ADDR_SCOPE_GLOBAL:
1722 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1723 					continue;
1724 				break;
1725 			default:
1726 				continue;
1727 			}
1728 
1729 			/*
1730 			 * check if anycast is okay.
1731 			 * XXX: just experimental.  not in the spec.
1732 			 */
1733 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1734 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1735 				continue; /* we need only unicast addresses */
1736 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1737 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1738 				continue;
1739 			}
1740 			addrsofif++; /* count the address */
1741 		}
1742 		IF_ADDR_RUNLOCK(ifp);
1743 		if (iffound) {
1744 			*ifpp = ifp;
1745 			IFNET_RUNLOCK_NOSLEEP();
1746 			return (addrsofif);
1747 		}
1748 
1749 		addrs += addrsofif;
1750 	}
1751 	IFNET_RUNLOCK_NOSLEEP();
1752 
1753 	return (addrs);
1754 }
1755 
1756 static int
1757 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1758     struct ifnet *ifp0, int resid)
1759 {
1760 	struct ifnet *ifp;
1761 	struct in6_ifaddr *ifa6;
1762 	struct ifaddr *ifa;
1763 	struct ifnet *ifp_dep = NULL;
1764 	int copied = 0, allow_deprecated = 0;
1765 	u_char *cp = (u_char *)(nni6 + 1);
1766 	int niflags = ni6->ni_flags;
1767 	u_int32_t ltime;
1768 
1769 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1770 		return (0);	/* needless to copy */
1771 
1772 	IFNET_RLOCK_NOSLEEP();
1773 	ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1774   again:
1775 
1776 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_link)) {
1777 		IF_ADDR_RLOCK(ifp);
1778 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1779 			if (ifa->ifa_addr->sa_family != AF_INET6)
1780 				continue;
1781 			ifa6 = (struct in6_ifaddr *)ifa;
1782 
1783 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1784 			    allow_deprecated == 0) {
1785 				/*
1786 				 * prefererred address should be put before
1787 				 * deprecated addresses.
1788 				 */
1789 
1790 				/* record the interface for later search */
1791 				if (ifp_dep == NULL)
1792 					ifp_dep = ifp;
1793 
1794 				continue;
1795 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1796 			    allow_deprecated != 0)
1797 				continue; /* we now collect deprecated addrs */
1798 
1799 			/* What do we have to do about ::1? */
1800 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1801 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1802 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1803 					continue;
1804 				break;
1805 			case IPV6_ADDR_SCOPE_SITELOCAL:
1806 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1807 					continue;
1808 				break;
1809 			case IPV6_ADDR_SCOPE_GLOBAL:
1810 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1811 					continue;
1812 				break;
1813 			default:
1814 				continue;
1815 			}
1816 
1817 			/*
1818 			 * check if anycast is okay.
1819 			 * XXX: just experimental.  not in the spec.
1820 			 */
1821 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1822 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1823 				continue;
1824 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1825 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1826 				continue;
1827 			}
1828 
1829 			/* now we can copy the address */
1830 			if (resid < sizeof(struct in6_addr) +
1831 			    sizeof(u_int32_t)) {
1832 				IF_ADDR_RUNLOCK(ifp);
1833 				/*
1834 				 * We give up much more copy.
1835 				 * Set the truncate flag and return.
1836 				 */
1837 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1838 				IFNET_RUNLOCK_NOSLEEP();
1839 				return (copied);
1840 			}
1841 
1842 			/*
1843 			 * Set the TTL of the address.
1844 			 * The TTL value should be one of the following
1845 			 * according to the specification:
1846 			 *
1847 			 * 1. The remaining lifetime of a DHCP lease on the
1848 			 *    address, or
1849 			 * 2. The remaining Valid Lifetime of a prefix from
1850 			 *    which the address was derived through Stateless
1851 			 *    Autoconfiguration.
1852 			 *
1853 			 * Note that we currently do not support stateful
1854 			 * address configuration by DHCPv6, so the former
1855 			 * case can't happen.
1856 			 */
1857 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1858 				ltime = ND6_INFINITE_LIFETIME;
1859 			else {
1860 				if (ifa6->ia6_lifetime.ia6t_expire >
1861 				    time_uptime)
1862 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1863 				else
1864 					ltime = 0;
1865 			}
1866 
1867 			bcopy(&ltime, cp, sizeof(u_int32_t));
1868 			cp += sizeof(u_int32_t);
1869 
1870 			/* copy the address itself */
1871 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1872 			    sizeof(struct in6_addr));
1873 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1874 			cp += sizeof(struct in6_addr);
1875 
1876 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1877 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1878 		}
1879 		IF_ADDR_RUNLOCK(ifp);
1880 		if (ifp0)	/* we need search only on the specified IF */
1881 			break;
1882 	}
1883 
1884 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1885 		ifp = ifp_dep;
1886 		allow_deprecated = 1;
1887 
1888 		goto again;
1889 	}
1890 
1891 	IFNET_RUNLOCK_NOSLEEP();
1892 
1893 	return (copied);
1894 }
1895 
1896 /*
1897  * XXX almost dup'ed code with rip6_input.
1898  */
1899 static int
1900 icmp6_rip6_input(struct mbuf **mp, int off)
1901 {
1902 	struct mbuf *m = *mp;
1903 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1904 	struct inpcb *in6p;
1905 	struct inpcb *last = NULL;
1906 	struct sockaddr_in6 fromsa;
1907 	struct icmp6_hdr *icmp6;
1908 	struct mbuf *opts = NULL;
1909 
1910 #ifndef PULLDOWN_TEST
1911 	/* this is assumed to be safe. */
1912 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1913 #else
1914 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1915 	if (icmp6 == NULL) {
1916 		/* m is already reclaimed */
1917 		return (IPPROTO_DONE);
1918 	}
1919 #endif
1920 
1921 	/*
1922 	 * XXX: the address may have embedded scope zone ID, which should be
1923 	 * hidden from applications.
1924 	 */
1925 	bzero(&fromsa, sizeof(fromsa));
1926 	fromsa.sin6_family = AF_INET6;
1927 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1928 	fromsa.sin6_addr = ip6->ip6_src;
1929 	if (sa6_recoverscope(&fromsa)) {
1930 		m_freem(m);
1931 		return (IPPROTO_DONE);
1932 	}
1933 
1934 	INP_INFO_RLOCK(&V_ripcbinfo);
1935 	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1936 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1937 			continue;
1938 		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1939 			continue;
1940 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1941 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1942 			continue;
1943 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1944 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1945 			continue;
1946 		INP_RLOCK(in6p);
1947 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1948 		    in6p->in6p_icmp6filt)) {
1949 			INP_RUNLOCK(in6p);
1950 			continue;
1951 		}
1952 		if (last != NULL) {
1953 			struct	mbuf *n = NULL;
1954 
1955 			/*
1956 			 * Recent network drivers tend to allocate a single
1957 			 * mbuf cluster, rather than to make a couple of
1958 			 * mbufs without clusters.  Also, since the IPv6 code
1959 			 * path tries to avoid m_pullup(), it is highly
1960 			 * probable that we still have an mbuf cluster here
1961 			 * even though the necessary length can be stored in an
1962 			 * mbuf's internal buffer.
1963 			 * Meanwhile, the default size of the receive socket
1964 			 * buffer for raw sockets is not so large.  This means
1965 			 * the possibility of packet loss is relatively higher
1966 			 * than before.  To avoid this scenario, we copy the
1967 			 * received data to a separate mbuf that does not use
1968 			 * a cluster, if possible.
1969 			 * XXX: it is better to copy the data after stripping
1970 			 * intermediate headers.
1971 			 */
1972 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1973 			    m->m_len <= MHLEN) {
1974 				n = m_get(M_NOWAIT, m->m_type);
1975 				if (n != NULL) {
1976 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1977 						bcopy(m->m_data, n->m_data,
1978 						      m->m_len);
1979 						n->m_len = m->m_len;
1980 					} else {
1981 						m_free(n);
1982 						n = NULL;
1983 					}
1984 				}
1985 			}
1986 			if (n != NULL ||
1987 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1988 				if (last->inp_flags & INP_CONTROLOPTS)
1989 					ip6_savecontrol(last, n, &opts);
1990 				/* strip intermediate headers */
1991 				m_adj(n, off);
1992 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1993 				if (sbappendaddr_locked(
1994 				    &last->inp_socket->so_rcv,
1995 				    (struct sockaddr *)&fromsa, n, opts)
1996 				    == 0) {
1997 					/* should notify about lost packet */
1998 					m_freem(n);
1999 					if (opts) {
2000 						m_freem(opts);
2001 					}
2002 					SOCKBUF_UNLOCK(
2003 					    &last->inp_socket->so_rcv);
2004 				} else
2005 					sorwakeup_locked(last->inp_socket);
2006 				opts = NULL;
2007 			}
2008 			INP_RUNLOCK(last);
2009 		}
2010 		last = in6p;
2011 	}
2012 	INP_INFO_RUNLOCK(&V_ripcbinfo);
2013 	if (last != NULL) {
2014 		if (last->inp_flags & INP_CONTROLOPTS)
2015 			ip6_savecontrol(last, m, &opts);
2016 		/* strip intermediate headers */
2017 		m_adj(m, off);
2018 
2019 		/* avoid using mbuf clusters if possible (see above) */
2020 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2021 		    m->m_len <= MHLEN) {
2022 			struct mbuf *n;
2023 
2024 			n = m_get(M_NOWAIT, m->m_type);
2025 			if (n != NULL) {
2026 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2027 					bcopy(m->m_data, n->m_data, m->m_len);
2028 					n->m_len = m->m_len;
2029 
2030 					m_freem(m);
2031 					m = n;
2032 				} else {
2033 					m_freem(n);
2034 					n = NULL;
2035 				}
2036 			}
2037 		}
2038 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2039 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2040 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2041 			m_freem(m);
2042 			if (opts)
2043 				m_freem(opts);
2044 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2045 		} else
2046 			sorwakeup_locked(last->inp_socket);
2047 		INP_RUNLOCK(last);
2048 	} else {
2049 		m_freem(m);
2050 		IP6STAT_DEC(ip6s_delivered);
2051 	}
2052 	return IPPROTO_DONE;
2053 }
2054 
2055 /*
2056  * Reflect the ip6 packet back to the source.
2057  * OFF points to the icmp6 header, counted from the top of the mbuf.
2058  */
2059 void
2060 icmp6_reflect(struct mbuf *m, size_t off)
2061 {
2062 	struct in6_addr src6, *srcp;
2063 	struct ip6_hdr *ip6;
2064 	struct icmp6_hdr *icmp6;
2065 	struct in6_ifaddr *ia = NULL;
2066 	struct ifnet *outif = NULL;
2067 	int plen;
2068 	int type, code, hlim;
2069 
2070 	/* too short to reflect */
2071 	if (off < sizeof(struct ip6_hdr)) {
2072 		nd6log((LOG_DEBUG,
2073 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2074 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2075 		    __FILE__, __LINE__));
2076 		goto bad;
2077 	}
2078 
2079 	/*
2080 	 * If there are extra headers between IPv6 and ICMPv6, strip
2081 	 * off that header first.
2082 	 */
2083 #ifdef DIAGNOSTIC
2084 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2085 		panic("assumption failed in icmp6_reflect");
2086 #endif
2087 	if (off > sizeof(struct ip6_hdr)) {
2088 		size_t l;
2089 		struct ip6_hdr nip6;
2090 
2091 		l = off - sizeof(struct ip6_hdr);
2092 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2093 		m_adj(m, l);
2094 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2095 		if (m->m_len < l) {
2096 			if ((m = m_pullup(m, l)) == NULL)
2097 				return;
2098 		}
2099 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2100 	} else /* off == sizeof(struct ip6_hdr) */ {
2101 		size_t l;
2102 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2103 		if (m->m_len < l) {
2104 			if ((m = m_pullup(m, l)) == NULL)
2105 				return;
2106 		}
2107 	}
2108 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2109 	ip6 = mtod(m, struct ip6_hdr *);
2110 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2111 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2112 	type = icmp6->icmp6_type; /* keep type for statistics */
2113 	code = icmp6->icmp6_code; /* ditto. */
2114 	hlim = 0;
2115 	srcp = NULL;
2116 
2117 	/*
2118 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2119 	 * use dst as the src for the reply.
2120 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2121 	 * (for example) when we encounter an error while forwarding procedure
2122 	 * destined to a duplicated address of ours.
2123 	 */
2124 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2125 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2126 		if (ia != NULL && !(ia->ia6_flags &
2127 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2128 			src6 = ia->ia_addr.sin6_addr;
2129 			srcp = &src6;
2130 
2131 			if (m->m_pkthdr.rcvif != NULL) {
2132 				/* XXX: This may not be the outgoing interface */
2133 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2134 			} else
2135 				hlim = V_ip6_defhlim;
2136 		}
2137 		if (ia != NULL)
2138 			ifa_free(&ia->ia_ifa);
2139 	}
2140 
2141 	if (srcp == NULL) {
2142 		int error;
2143 		struct in6_addr dst6;
2144 		uint32_t scopeid;
2145 
2146 		/*
2147 		 * This case matches to multicasts, our anycast, or unicasts
2148 		 * that we do not own.  Select a source address based on the
2149 		 * source address of the erroneous packet.
2150 		 */
2151 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2152 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2153 		    scopeid, NULL, &src6, &hlim);
2154 
2155 		if (error) {
2156 			char ip6buf[INET6_ADDRSTRLEN];
2157 			nd6log((LOG_DEBUG,
2158 			    "icmp6_reflect: source can't be determined: "
2159 			    "dst=%s, error=%d\n",
2160 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2161 			goto bad;
2162 		}
2163 		srcp = &src6;
2164 	}
2165 	/*
2166 	 * ip6_input() drops a packet if its src is multicast.
2167 	 * So, the src is never multicast.
2168 	 */
2169 	ip6->ip6_dst = ip6->ip6_src;
2170 	ip6->ip6_src = *srcp;
2171 	ip6->ip6_flow = 0;
2172 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2173 	ip6->ip6_vfc |= IPV6_VERSION;
2174 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2175 	ip6->ip6_hlim = hlim;
2176 
2177 	icmp6->icmp6_cksum = 0;
2178 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2179 	    sizeof(struct ip6_hdr), plen);
2180 
2181 	/*
2182 	 * XXX option handling
2183 	 */
2184 
2185 	m->m_flags &= ~(M_BCAST|M_MCAST);
2186 
2187 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2188 	if (outif)
2189 		icmp6_ifoutstat_inc(outif, type, code);
2190 
2191 	return;
2192 
2193  bad:
2194 	m_freem(m);
2195 	return;
2196 }
2197 
2198 void
2199 icmp6_fasttimo(void)
2200 {
2201 
2202 	mld_fasttimo();
2203 }
2204 
2205 void
2206 icmp6_slowtimo(void)
2207 {
2208 
2209 	mld_slowtimo();
2210 }
2211 
2212 static const char *
2213 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2214     struct in6_addr *tgt6)
2215 {
2216 	static char buf[1024];
2217 	char ip6bufs[INET6_ADDRSTRLEN];
2218 	char ip6bufd[INET6_ADDRSTRLEN];
2219 	char ip6buft[INET6_ADDRSTRLEN];
2220 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2221 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2222 	    ip6_sprintf(ip6buft, tgt6));
2223 	return buf;
2224 }
2225 
2226 void
2227 icmp6_redirect_input(struct mbuf *m, int off)
2228 {
2229 	struct ifnet *ifp;
2230 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2231 	struct nd_redirect *nd_rd;
2232 	int icmp6len = ntohs(ip6->ip6_plen);
2233 	char *lladdr = NULL;
2234 	int lladdrlen = 0;
2235 	int is_router;
2236 	int is_onlink;
2237 	struct in6_addr src6 = ip6->ip6_src;
2238 	struct in6_addr redtgt6;
2239 	struct in6_addr reddst6;
2240 	union nd_opts ndopts;
2241 	char ip6buf[INET6_ADDRSTRLEN];
2242 
2243 	M_ASSERTPKTHDR(m);
2244 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2245 
2246 	ifp = m->m_pkthdr.rcvif;
2247 
2248 	/* XXX if we are router, we don't update route by icmp6 redirect */
2249 	if (V_ip6_forwarding)
2250 		goto freeit;
2251 	if (!V_icmp6_rediraccept)
2252 		goto freeit;
2253 
2254 	/* RFC 6980: Nodes MUST silently ignore fragments */
2255 	if(m->m_flags & M_FRAGMENTED)
2256 		goto freeit;
2257 
2258 #ifndef PULLDOWN_TEST
2259 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2260 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2261 #else
2262 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2263 	if (nd_rd == NULL) {
2264 		ICMP6STAT_INC(icp6s_tooshort);
2265 		return;
2266 	}
2267 #endif
2268 	redtgt6 = nd_rd->nd_rd_target;
2269 	reddst6 = nd_rd->nd_rd_dst;
2270 
2271 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2272 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2273 		goto freeit;
2274 	}
2275 
2276 	/* validation */
2277 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2278 		nd6log((LOG_ERR,
2279 		    "ICMP6 redirect sent from %s rejected; "
2280 		    "must be from linklocal\n",
2281 		    ip6_sprintf(ip6buf, &src6)));
2282 		goto bad;
2283 	}
2284 	if (ip6->ip6_hlim != 255) {
2285 		nd6log((LOG_ERR,
2286 		    "ICMP6 redirect sent from %s rejected; "
2287 		    "hlim=%d (must be 255)\n",
2288 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2289 		goto bad;
2290 	}
2291     {
2292 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2293 	struct nhop6_basic nh6;
2294 	struct in6_addr kdst;
2295 	uint32_t scopeid;
2296 
2297 	in6_splitscope(&reddst6, &kdst, &scopeid);
2298 	if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2299 		if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2300 			nd6log((LOG_ERR,
2301 			    "ICMP6 redirect rejected; no route "
2302 			    "with inet6 gateway found for redirect dst: %s\n",
2303 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2304 			goto bad;
2305 		}
2306 
2307 		/*
2308 		 * Embed scope zone id into next hop address, since
2309 		 * fib6_lookup_nh_basic() returns address without embedded
2310 		 * scope zone id.
2311 		 */
2312 		if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL))
2313 			goto freeit;
2314 
2315 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2316 			nd6log((LOG_ERR,
2317 			    "ICMP6 redirect rejected; "
2318 			    "not equal to gw-for-src=%s (must be same): "
2319 			    "%s\n",
2320 			    ip6_sprintf(ip6buf, &nh6.nh_addr),
2321 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2322 			goto bad;
2323 		}
2324 	} else {
2325 		nd6log((LOG_ERR,
2326 		    "ICMP6 redirect rejected; "
2327 		    "no route found for redirect dst: %s\n",
2328 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2329 		goto bad;
2330 	}
2331     }
2332 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2333 		nd6log((LOG_ERR,
2334 		    "ICMP6 redirect rejected; "
2335 		    "redirect dst must be unicast: %s\n",
2336 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2337 		goto bad;
2338 	}
2339 
2340 	is_router = is_onlink = 0;
2341 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2342 		is_router = 1;	/* router case */
2343 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2344 		is_onlink = 1;	/* on-link destination case */
2345 	if (!is_router && !is_onlink) {
2346 		nd6log((LOG_ERR,
2347 		    "ICMP6 redirect rejected; "
2348 		    "neither router case nor onlink case: %s\n",
2349 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2350 		goto bad;
2351 	}
2352 
2353 	icmp6len -= sizeof(*nd_rd);
2354 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2355 	if (nd6_options(&ndopts) < 0) {
2356 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2357 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2358 		/* nd6_options have incremented stats */
2359 		goto freeit;
2360 	}
2361 
2362 	if (ndopts.nd_opts_tgt_lladdr) {
2363 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2364 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2365 	}
2366 
2367 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2368 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2369 		    "(if %d, icmp6 packet %d): %s\n",
2370 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2371 		    ifp->if_addrlen, lladdrlen - 2,
2372 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2373 		goto bad;
2374 	}
2375 
2376 	/* Validation passed. */
2377 
2378 	/* RFC 2461 8.3 */
2379 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2380 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2381 
2382 	/*
2383 	 * Install a gateway route in the better-router case or an interface
2384 	 * route in the on-link-destination case.
2385 	 */
2386 	{
2387 		struct sockaddr_in6 sdst;
2388 		struct sockaddr_in6 sgw;
2389 		struct sockaddr_in6 ssrc;
2390 		struct sockaddr *gw;
2391 		int rt_flags;
2392 		u_int fibnum;
2393 
2394 		bzero(&sdst, sizeof(sdst));
2395 		bzero(&ssrc, sizeof(ssrc));
2396 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2397 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2398 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2399 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2400 		rt_flags = RTF_HOST;
2401 		if (is_router) {
2402 			bzero(&sgw, sizeof(sgw));
2403 			sgw.sin6_family = AF_INET6;
2404 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2405 			bcopy(&redtgt6, &sgw.sin6_addr,
2406 				sizeof(struct in6_addr));
2407 			gw = (struct sockaddr *)&sgw;
2408 			rt_flags |= RTF_GATEWAY;
2409 		} else
2410 			gw = ifp->if_addr->ifa_addr;
2411 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2412 			in6_rtredirect((struct sockaddr *)&sdst, gw,
2413 			    (struct sockaddr *)NULL, rt_flags,
2414 			    (struct sockaddr *)&ssrc, fibnum);
2415 	}
2416 	/* finally update cached route in each socket via pfctlinput */
2417     {
2418 	struct sockaddr_in6 sdst;
2419 
2420 	bzero(&sdst, sizeof(sdst));
2421 	sdst.sin6_family = AF_INET6;
2422 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2423 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2424 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2425     }
2426 
2427  freeit:
2428 	m_freem(m);
2429 	return;
2430 
2431  bad:
2432 	ICMP6STAT_INC(icp6s_badredirect);
2433 	m_freem(m);
2434 }
2435 
2436 void
2437 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2438 {
2439 	struct ifnet *ifp;	/* my outgoing interface */
2440 	struct in6_addr *ifp_ll6;
2441 	struct in6_addr *router_ll6;
2442 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2443 	struct mbuf *m = NULL;	/* newly allocated one */
2444 	struct m_tag *mtag;
2445 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2446 	struct nd_redirect *nd_rd;
2447 	struct llentry *ln = NULL;
2448 	size_t maxlen;
2449 	u_char *p;
2450 	struct ifnet *outif = NULL;
2451 	struct sockaddr_in6 src_sa;
2452 
2453 	icmp6_errcount(ND_REDIRECT, 0);
2454 
2455 	/* if we are not router, we don't send icmp6 redirect */
2456 	if (!V_ip6_forwarding)
2457 		goto fail;
2458 
2459 	/* sanity check */
2460 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2461 		goto fail;
2462 
2463 	/*
2464 	 * Address check:
2465 	 *  the source address must identify a neighbor, and
2466 	 *  the destination address must not be a multicast address
2467 	 *  [RFC 2461, sec 8.2]
2468 	 */
2469 	sip6 = mtod(m0, struct ip6_hdr *);
2470 	bzero(&src_sa, sizeof(src_sa));
2471 	src_sa.sin6_family = AF_INET6;
2472 	src_sa.sin6_len = sizeof(src_sa);
2473 	src_sa.sin6_addr = sip6->ip6_src;
2474 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2475 		goto fail;
2476 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2477 		goto fail;	/* what should we do here? */
2478 
2479 	/* rate limit */
2480 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2481 		goto fail;
2482 
2483 	/*
2484 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2485 	 * we almost always ask for an mbuf cluster for simplicity.
2486 	 * (MHLEN < IPV6_MMTU is almost always true)
2487 	 */
2488 #if IPV6_MMTU >= MCLBYTES
2489 # error assumption failed about IPV6_MMTU and MCLBYTES
2490 #endif
2491 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2492 	if (m == NULL)
2493 		goto fail;
2494 	M_SETFIB(m, rt->rt_fibnum);
2495 	maxlen = M_TRAILINGSPACE(m);
2496 	maxlen = min(IPV6_MMTU, maxlen);
2497 	/* just for safety */
2498 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2499 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2500 		goto fail;
2501 	}
2502 
2503 	{
2504 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2505 		struct in6_ifaddr *ia;
2506 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2507 						 IN6_IFF_NOTREADY|
2508 						 IN6_IFF_ANYCAST)) == NULL)
2509 			goto fail;
2510 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2511 		/* XXXRW: reference released prematurely. */
2512 		ifa_free(&ia->ia_ifa);
2513 	}
2514 
2515 	/* get ip6 linklocal address for the router. */
2516 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2517 		struct sockaddr_in6 *sin6;
2518 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2519 		router_ll6 = &sin6->sin6_addr;
2520 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2521 			router_ll6 = (struct in6_addr *)NULL;
2522 	} else
2523 		router_ll6 = (struct in6_addr *)NULL;
2524 
2525 	/* ip6 */
2526 	ip6 = mtod(m, struct ip6_hdr *);
2527 	ip6->ip6_flow = 0;
2528 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2529 	ip6->ip6_vfc |= IPV6_VERSION;
2530 	/* ip6->ip6_plen will be set later */
2531 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2532 	ip6->ip6_hlim = 255;
2533 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2534 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2535 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2536 
2537 	/* ND Redirect */
2538 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2539 	nd_rd->nd_rd_type = ND_REDIRECT;
2540 	nd_rd->nd_rd_code = 0;
2541 	nd_rd->nd_rd_reserved = 0;
2542 	if (rt->rt_flags & RTF_GATEWAY) {
2543 		/*
2544 		 * nd_rd->nd_rd_target must be a link-local address in
2545 		 * better router cases.
2546 		 */
2547 		if (!router_ll6)
2548 			goto fail;
2549 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2550 		    sizeof(nd_rd->nd_rd_target));
2551 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2552 		    sizeof(nd_rd->nd_rd_dst));
2553 	} else {
2554 		/* make sure redtgt == reddst */
2555 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2556 		    sizeof(nd_rd->nd_rd_target));
2557 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2558 		    sizeof(nd_rd->nd_rd_dst));
2559 	}
2560 
2561 	p = (u_char *)(nd_rd + 1);
2562 
2563 	if (!router_ll6)
2564 		goto nolladdropt;
2565 
2566 	{
2567 		/* target lladdr option */
2568 		int len;
2569 		struct nd_opt_hdr *nd_opt;
2570 		char *lladdr;
2571 
2572 		IF_AFDATA_RLOCK(ifp);
2573 		ln = nd6_lookup(router_ll6, 0, ifp);
2574 		IF_AFDATA_RUNLOCK(ifp);
2575 		if (ln == NULL)
2576 			goto nolladdropt;
2577 
2578 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2579 		len = (len + 7) & ~7;	/* round by 8 */
2580 		/* safety check */
2581 		if (len + (p - (u_char *)ip6) > maxlen)
2582 			goto nolladdropt;
2583 
2584 		if (ln->la_flags & LLE_VALID) {
2585 			nd_opt = (struct nd_opt_hdr *)p;
2586 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2587 			nd_opt->nd_opt_len = len >> 3;
2588 			lladdr = (char *)(nd_opt + 1);
2589 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2590 			p += len;
2591 		}
2592 	}
2593 nolladdropt:
2594 	if (ln != NULL)
2595 		LLE_RUNLOCK(ln);
2596 
2597 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2598 
2599 	/* just to be safe */
2600 #ifdef M_DECRYPTED	/*not openbsd*/
2601 	if (m0->m_flags & M_DECRYPTED)
2602 		goto noredhdropt;
2603 #endif
2604 	if (p - (u_char *)ip6 > maxlen)
2605 		goto noredhdropt;
2606 
2607 	{
2608 		/* redirected header option */
2609 		int len;
2610 		struct nd_opt_rd_hdr *nd_opt_rh;
2611 
2612 		/*
2613 		 * compute the maximum size for icmp6 redirect header option.
2614 		 * XXX room for auth header?
2615 		 */
2616 		len = maxlen - (p - (u_char *)ip6);
2617 		len &= ~7;
2618 
2619 		/* This is just for simplicity. */
2620 		if (m0->m_pkthdr.len != m0->m_len) {
2621 			if (m0->m_next) {
2622 				m_freem(m0->m_next);
2623 				m0->m_next = NULL;
2624 			}
2625 			m0->m_pkthdr.len = m0->m_len;
2626 		}
2627 
2628 		/*
2629 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2630 		 * about padding/truncate rule for the original IP packet.
2631 		 * From the discussion on IPv6imp in Feb 1999,
2632 		 * the consensus was:
2633 		 * - "attach as much as possible" is the goal
2634 		 * - pad if not aligned (original size can be guessed by
2635 		 *   original ip6 header)
2636 		 * Following code adds the padding if it is simple enough,
2637 		 * and truncates if not.
2638 		 */
2639 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2640 			panic("assumption failed in %s:%d", __FILE__,
2641 			    __LINE__);
2642 
2643 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2644 			/* not enough room, truncate */
2645 			m0->m_pkthdr.len = m0->m_len = len -
2646 			    sizeof(*nd_opt_rh);
2647 		} else {
2648 			/* enough room, pad or truncate */
2649 			size_t extra;
2650 
2651 			extra = m0->m_pkthdr.len % 8;
2652 			if (extra) {
2653 				/* pad if easy enough, truncate if not */
2654 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2655 					/* pad */
2656 					m0->m_len += (8 - extra);
2657 					m0->m_pkthdr.len += (8 - extra);
2658 				} else {
2659 					/* truncate */
2660 					m0->m_pkthdr.len -= extra;
2661 					m0->m_len -= extra;
2662 				}
2663 			}
2664 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2665 			m0->m_pkthdr.len = m0->m_len = len -
2666 			    sizeof(*nd_opt_rh);
2667 		}
2668 
2669 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2670 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2671 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2672 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2673 		p += sizeof(*nd_opt_rh);
2674 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2675 
2676 		/* connect m0 to m */
2677 		m_tag_delete_chain(m0, NULL);
2678 		m0->m_flags &= ~M_PKTHDR;
2679 		m->m_next = m0;
2680 		m->m_pkthdr.len = m->m_len + m0->m_len;
2681 		m0 = NULL;
2682 	}
2683 noredhdropt:;
2684 	if (m0) {
2685 		m_freem(m0);
2686 		m0 = NULL;
2687 	}
2688 
2689 	/* XXX: clear embedded link IDs in the inner header */
2690 	in6_clearscope(&sip6->ip6_src);
2691 	in6_clearscope(&sip6->ip6_dst);
2692 	in6_clearscope(&nd_rd->nd_rd_target);
2693 	in6_clearscope(&nd_rd->nd_rd_dst);
2694 
2695 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2696 
2697 	nd_rd->nd_rd_cksum = 0;
2698 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2699 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2700 
2701         if (send_sendso_input_hook != NULL) {
2702 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2703 			M_NOWAIT);
2704 		if (mtag == NULL)
2705 			goto fail;
2706 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2707 		m_tag_prepend(m, mtag);
2708 	}
2709 
2710 	/* send the packet to outside... */
2711 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2712 	if (outif) {
2713 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2714 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2715 	}
2716 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2717 
2718 	return;
2719 
2720 fail:
2721 	if (m)
2722 		m_freem(m);
2723 	if (m0)
2724 		m_freem(m0);
2725 }
2726 
2727 /*
2728  * ICMPv6 socket option processing.
2729  */
2730 int
2731 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2732 {
2733 	int error = 0;
2734 	int optlen;
2735 	struct inpcb *inp = sotoinpcb(so);
2736 	int level, op, optname;
2737 
2738 	if (sopt) {
2739 		level = sopt->sopt_level;
2740 		op = sopt->sopt_dir;
2741 		optname = sopt->sopt_name;
2742 		optlen = sopt->sopt_valsize;
2743 	} else
2744 		level = op = optname = optlen = 0;
2745 
2746 	if (level != IPPROTO_ICMPV6) {
2747 		return EINVAL;
2748 	}
2749 
2750 	switch (op) {
2751 	case PRCO_SETOPT:
2752 		switch (optname) {
2753 		case ICMP6_FILTER:
2754 		    {
2755 			struct icmp6_filter ic6f;
2756 
2757 			if (optlen != sizeof(ic6f)) {
2758 				error = EMSGSIZE;
2759 				break;
2760 			}
2761 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2762 			if (error == 0) {
2763 				INP_WLOCK(inp);
2764 				*inp->in6p_icmp6filt = ic6f;
2765 				INP_WUNLOCK(inp);
2766 			}
2767 			break;
2768 		    }
2769 
2770 		default:
2771 			error = ENOPROTOOPT;
2772 			break;
2773 		}
2774 		break;
2775 
2776 	case PRCO_GETOPT:
2777 		switch (optname) {
2778 		case ICMP6_FILTER:
2779 		    {
2780 			struct icmp6_filter ic6f;
2781 
2782 			INP_RLOCK(inp);
2783 			ic6f = *inp->in6p_icmp6filt;
2784 			INP_RUNLOCK(inp);
2785 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2786 			break;
2787 		    }
2788 
2789 		default:
2790 			error = ENOPROTOOPT;
2791 			break;
2792 		}
2793 		break;
2794 	}
2795 
2796 	return (error);
2797 }
2798 
2799 /*
2800  * Perform rate limit check.
2801  * Returns 0 if it is okay to send the icmp6 packet.
2802  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2803  * limitation.
2804  *
2805  * XXX per-destination/type check necessary?
2806  *
2807  * dst - not used at this moment
2808  * type - not used at this moment
2809  * code - not used at this moment
2810  */
2811 static int
2812 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2813     const int code)
2814 {
2815 	int ret;
2816 
2817 	ret = 0;	/* okay to send */
2818 
2819 	/* PPS limit */
2820 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2821 	    V_icmp6errppslim)) {
2822 		/* The packet is subject to rate limit */
2823 		ret++;
2824 	}
2825 
2826 	return ret;
2827 }
2828