xref: /freebsd/sys/netinet6/icmp6.c (revision f6a3b357e9be4c6423c85eff9a847163a0d307c8)
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 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
128 VNET_DEFINE_STATIC(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, struct prison *);
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 	ICMP6STAT_INC(icp6s_outhist[type]);
387 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
388 
389 	return;
390 
391   freeit:
392 	/*
393 	 * If we can't tell whether or not we can generate ICMP6, free it.
394 	 */
395 	m_freem(m);
396 }
397 
398 /*
399  * Process a received ICMP6 message.
400  */
401 int
402 icmp6_input(struct mbuf **mp, int *offp, int proto)
403 {
404 	struct mbuf *m = *mp, *n;
405 	struct ifnet *ifp;
406 	struct ip6_hdr *ip6, *nip6;
407 	struct icmp6_hdr *icmp6, *nicmp6;
408 	int off = *offp;
409 	int icmp6len = m->m_pkthdr.len - *offp;
410 	int code, sum, noff;
411 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
412 	int ip6len, error;
413 
414 	ifp = m->m_pkthdr.rcvif;
415 
416 #ifndef PULLDOWN_TEST
417 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
418 	/* m might change if M_LOOP.  So, call mtod after this */
419 #endif
420 
421 	/*
422 	 * Locate icmp6 structure in mbuf, and check
423 	 * that not corrupted and of at least minimum length
424 	 */
425 
426 	ip6 = mtod(m, struct ip6_hdr *);
427 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
428 	if (icmp6len < sizeof(struct icmp6_hdr)) {
429 		ICMP6STAT_INC(icp6s_tooshort);
430 		goto freeit;
431 	}
432 
433 	/*
434 	 * Check multicast group membership.
435 	 * Note: SSM filters are not applied for ICMPv6 traffic.
436 	 */
437 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
438 		struct in6_multi	*inm;
439 
440 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
441 		if (inm == NULL) {
442 			IP6STAT_INC(ip6s_notmember);
443 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
444 			goto freeit;
445 		}
446 	}
447 
448 	/*
449 	 * calculate the checksum
450 	 */
451 #ifndef PULLDOWN_TEST
452 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
453 #else
454 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
455 	if (icmp6 == NULL) {
456 		ICMP6STAT_INC(icp6s_tooshort);
457 		return IPPROTO_DONE;
458 	}
459 #endif
460 	code = icmp6->icmp6_code;
461 
462 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
463 		nd6log((LOG_ERR,
464 		    "ICMP6 checksum error(%d|%x) %s\n",
465 		    icmp6->icmp6_type, sum,
466 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
467 		ICMP6STAT_INC(icp6s_checksum);
468 		goto freeit;
469 	}
470 
471 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
472 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
473 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
474 		icmp6_ifstat_inc(ifp, ifs6_in_error);
475 
476 	switch (icmp6->icmp6_type) {
477 	case ICMP6_DST_UNREACH:
478 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
479 		switch (code) {
480 		case ICMP6_DST_UNREACH_NOROUTE:
481 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
482 			code = PRC_UNREACH_NET;
483 			break;
484 		case ICMP6_DST_UNREACH_ADMIN:
485 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
486 			code = PRC_UNREACH_ADMIN_PROHIB;
487 			break;
488 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
489 			/* I mean "source address was incorrect." */
490 			code = PRC_PARAMPROB;
491 			break;
492 		case ICMP6_DST_UNREACH_NOPORT:
493 			code = PRC_UNREACH_PORT;
494 			break;
495 		default:
496 			goto badcode;
497 		}
498 		goto deliver;
499 		break;
500 
501 	case ICMP6_PACKET_TOO_BIG:
502 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
503 
504 		/* validation is made in icmp6_mtudisc_update */
505 
506 		code = PRC_MSGSIZE;
507 
508 		/*
509 		 * Updating the path MTU will be done after examining
510 		 * intermediate extension headers.
511 		 */
512 		goto deliver;
513 		break;
514 
515 	case ICMP6_TIME_EXCEEDED:
516 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
517 		switch (code) {
518 		case ICMP6_TIME_EXCEED_TRANSIT:
519 			code = PRC_TIMXCEED_INTRANS;
520 			break;
521 		case ICMP6_TIME_EXCEED_REASSEMBLY:
522 			code = PRC_TIMXCEED_REASS;
523 			break;
524 		default:
525 			goto badcode;
526 		}
527 		goto deliver;
528 		break;
529 
530 	case ICMP6_PARAM_PROB:
531 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
532 		switch (code) {
533 		case ICMP6_PARAMPROB_NEXTHEADER:
534 			code = PRC_UNREACH_PROTOCOL;
535 			break;
536 		case ICMP6_PARAMPROB_HEADER:
537 		case ICMP6_PARAMPROB_OPTION:
538 			code = PRC_PARAMPROB;
539 			break;
540 		default:
541 			goto badcode;
542 		}
543 		goto deliver;
544 		break;
545 
546 	case ICMP6_ECHO_REQUEST:
547 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
548 		if (code != 0)
549 			goto badcode;
550 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
551 			/* Give up remote */
552 			break;
553 		}
554 		if (!M_WRITABLE(n)
555 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
556 			struct mbuf *n0 = n;
557 			int n0len;
558 
559 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
560 			n = m_gethdr(M_NOWAIT, n0->m_type);
561 			if (n == NULL) {
562 				/* Give up remote */
563 				m_freem(n0);
564 				break;
565 			}
566 
567 			m_move_pkthdr(n, n0);	/* FIB copied. */
568 			n0len = n0->m_pkthdr.len;	/* save for use below */
569 			/*
570 			 * Copy IPv6 and ICMPv6 only.
571 			 */
572 			nip6 = mtod(n, struct ip6_hdr *);
573 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
574 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
575 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
576 			noff = sizeof(struct ip6_hdr);
577 			/* new mbuf contains only ipv6+icmpv6 headers */
578 			n->m_len = noff + sizeof(struct icmp6_hdr);
579 			/*
580 			 * Adjust mbuf.  ip6_plen will be adjusted in
581 			 * ip6_output().
582 			 */
583 			m_adj(n0, off + sizeof(struct icmp6_hdr));
584 			/* recalculate complete packet size */
585 			n->m_pkthdr.len = n0len + (noff - off);
586 			n->m_next = n0;
587 		} else {
588 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
589 			    sizeof(*nicmp6));
590 			noff = off;
591 		}
592 		if (n) {
593 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
594 			nicmp6->icmp6_code = 0;
595 			ICMP6STAT_INC(icp6s_reflect);
596 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
597 			icmp6_reflect(n, noff);
598 		}
599 		break;
600 
601 	case ICMP6_ECHO_REPLY:
602 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
603 		if (code != 0)
604 			goto badcode;
605 		break;
606 
607 	case MLD_LISTENER_QUERY:
608 	case MLD_LISTENER_REPORT:
609 	case MLD_LISTENER_DONE:
610 	case MLDV2_LISTENER_REPORT:
611 		/*
612 		 * Drop MLD traffic which is not link-local, has a hop limit
613 		 * of greater than 1 hop, or which does not have the
614 		 * IPv6 HBH Router Alert option.
615 		 * As IPv6 HBH options are stripped in ip6_input() we must
616 		 * check an mbuf header flag.
617 		 * XXX Should we also sanity check that these messages
618 		 * were directed to a link-local multicast prefix?
619 		 */
620 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
621 			goto freeit;
622 		if (mld_input(m, off, icmp6len) != 0)
623 			return (IPPROTO_DONE);
624 		/* m stays. */
625 		break;
626 
627 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
628 	    {
629 		enum { WRU, FQDN } mode;
630 		struct prison *pr;
631 
632 		if (!V_icmp6_nodeinfo)
633 			break;
634 
635 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
636 			mode = WRU;
637 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
638 			mode = FQDN;
639 		else
640 			goto badlen;
641 
642 		pr = NULL;
643 		sx_slock(&allprison_lock);
644 		TAILQ_FOREACH(pr, &allprison, pr_list)
645 			if (pr->pr_vnet == ifp->if_vnet)
646 				break;
647 		sx_sunlock(&allprison_lock);
648 		if (pr == NULL)
649 			pr = curthread->td_ucred->cr_prison;
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, pr);
658 			/* XXX meaningless if n == NULL */
659 			noff = sizeof(struct ip6_hdr);
660 		} else {
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 			/*
695 			 * Copy IPv6 and ICMPv6 only.
696 			 */
697 			nip6 = mtod(n, struct ip6_hdr *);
698 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
699 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
700 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
701 			p = (u_char *)(nicmp6 + 1);
702 			bzero(p, 4);
703 
704 			maxhlen = M_TRAILINGSPACE(n) -
705 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
706 			mtx_lock(&pr->pr_mtx);
707 			hlen = strlen(pr->pr_hostname);
708 			if (maxhlen > hlen)
709 				maxhlen = hlen;
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, struct prison *pr)
1179 {
1180 	struct icmp6_nodeinfo *ni6, *nni6;
1181 	struct mbuf *n = NULL;
1182 	u_int16_t qtype;
1183 	int subjlen;
1184 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1185 	struct ni_reply_fqdn *fqdn;
1186 	int addrs;		/* for NI_QTYPE_NODEADDR */
1187 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1188 	struct in6_addr in6_subj; /* subject address */
1189 	struct ip6_hdr *ip6;
1190 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1191 	char *subj = NULL;
1192 	struct in6_ifaddr *ia6 = NULL;
1193 
1194 	ip6 = mtod(m, struct ip6_hdr *);
1195 #ifndef PULLDOWN_TEST
1196 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1197 #else
1198 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1199 	if (ni6 == NULL) {
1200 		/* m is already reclaimed */
1201 		return (NULL);
1202 	}
1203 #endif
1204 
1205 	/*
1206 	 * Validate IPv6 source address.
1207 	 * The default configuration MUST be to refuse answering queries from
1208 	 * global-scope addresses according to RFC4602.
1209 	 * Notes:
1210 	 *  - it's not very clear what "refuse" means; this implementation
1211 	 *    simply drops it.
1212 	 *  - it's not very easy to identify global-scope (unicast) addresses
1213 	 *    since there are many prefixes for them.  It should be safer
1214 	 *    and in practice sufficient to check "all" but loopback and
1215 	 *    link-local (note that site-local unicast was deprecated and
1216 	 *    ULA is defined as global scope-wise)
1217 	 */
1218 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1219 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1220 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1221 		goto bad;
1222 
1223 	/*
1224 	 * Validate IPv6 destination address.
1225 	 *
1226 	 * The Responder must discard the Query without further processing
1227 	 * unless it is one of the Responder's unicast or anycast addresses, or
1228 	 * a link-local scope multicast address which the Responder has joined.
1229 	 * [RFC4602, Section 5.]
1230 	 */
1231 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1232 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1233 			goto bad;
1234 		/* else it's a link-local multicast, fine */
1235 	} else {		/* unicast or anycast */
1236 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1237 		if (ia6 == NULL)
1238 			goto bad; /* XXX impossible */
1239 
1240 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1241 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1242 			ifa_free(&ia6->ia_ifa);
1243 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1244 				"a temporary address in %s:%d",
1245 			       __FILE__, __LINE__));
1246 			goto bad;
1247 		}
1248 		ifa_free(&ia6->ia_ifa);
1249 	}
1250 
1251 	/* validate query Subject field. */
1252 	qtype = ntohs(ni6->ni_qtype);
1253 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1254 	switch (qtype) {
1255 	case NI_QTYPE_NOOP:
1256 	case NI_QTYPE_SUPTYPES:
1257 		/* 07 draft */
1258 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1259 			break;
1260 		/* FALLTHROUGH */
1261 	case NI_QTYPE_FQDN:
1262 	case NI_QTYPE_NODEADDR:
1263 	case NI_QTYPE_IPV4ADDR:
1264 		switch (ni6->ni_code) {
1265 		case ICMP6_NI_SUBJ_IPV6:
1266 #if ICMP6_NI_SUBJ_IPV6 != 0
1267 		case 0:
1268 #endif
1269 			/*
1270 			 * backward compatibility - try to accept 03 draft
1271 			 * format, where no Subject is present.
1272 			 */
1273 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1274 			    subjlen == 0) {
1275 				oldfqdn++;
1276 				break;
1277 			}
1278 #if ICMP6_NI_SUBJ_IPV6 != 0
1279 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1280 				goto bad;
1281 #endif
1282 
1283 			if (subjlen != sizeof(struct in6_addr))
1284 				goto bad;
1285 
1286 			/*
1287 			 * Validate Subject address.
1288 			 *
1289 			 * Not sure what exactly "address belongs to the node"
1290 			 * means in the spec, is it just unicast, or what?
1291 			 *
1292 			 * At this moment we consider Subject address as
1293 			 * "belong to the node" if the Subject address equals
1294 			 * to the IPv6 destination address; validation for
1295 			 * IPv6 destination address should have done enough
1296 			 * check for us.
1297 			 *
1298 			 * We do not do proxy at this moment.
1299 			 */
1300 			/* m_pulldown instead of copy? */
1301 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1302 			    subjlen, (caddr_t)&in6_subj);
1303 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1304 				goto bad;
1305 
1306 			subj = (char *)&in6_subj;
1307 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1308 				break;
1309 
1310 			/*
1311 			 * XXX if we are to allow other cases, we should really
1312 			 * be careful about scope here.
1313 			 * basically, we should disallow queries toward IPv6
1314 			 * destination X with subject Y,
1315 			 * if scope(X) > scope(Y).
1316 			 * if we allow scope(X) > scope(Y), it will result in
1317 			 * information leakage across scope boundary.
1318 			 */
1319 			goto bad;
1320 
1321 		case ICMP6_NI_SUBJ_FQDN:
1322 			/*
1323 			 * Validate Subject name with gethostname(3).
1324 			 *
1325 			 * The behavior may need some debate, since:
1326 			 * - we are not sure if the node has FQDN as
1327 			 *   hostname (returned by gethostname(3)).
1328 			 * - the code does wildcard match for truncated names.
1329 			 *   however, we are not sure if we want to perform
1330 			 *   wildcard match, if gethostname(3) side has
1331 			 *   truncated hostname.
1332 			 */
1333 			mtx_lock(&pr->pr_mtx);
1334 			n = ni6_nametodns(pr->pr_hostname,
1335 			    strlen(pr->pr_hostname), 0);
1336 			mtx_unlock(&pr->pr_mtx);
1337 			if (!n || n->m_next || n->m_len == 0)
1338 				goto bad;
1339 			IP6_EXTHDR_GET(subj, char *, m,
1340 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1341 			if (subj == NULL)
1342 				goto bad;
1343 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1344 			    n->m_len)) {
1345 				goto bad;
1346 			}
1347 			m_freem(n);
1348 			n = NULL;
1349 			break;
1350 
1351 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1352 		default:
1353 			goto bad;
1354 		}
1355 		break;
1356 	}
1357 
1358 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1359 	switch (qtype) {
1360 	case NI_QTYPE_FQDN:
1361 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1362 			goto bad;
1363 		break;
1364 	case NI_QTYPE_NODEADDR:
1365 	case NI_QTYPE_IPV4ADDR:
1366 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1367 			goto bad;
1368 		break;
1369 	}
1370 
1371 	/* guess reply length */
1372 	switch (qtype) {
1373 	case NI_QTYPE_NOOP:
1374 		break;		/* no reply data */
1375 	case NI_QTYPE_SUPTYPES:
1376 		replylen += sizeof(u_int32_t);
1377 		break;
1378 	case NI_QTYPE_FQDN:
1379 		/* XXX will append an mbuf */
1380 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1381 		break;
1382 	case NI_QTYPE_NODEADDR:
1383 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1384 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1385 		    sizeof(u_int32_t))) > MCLBYTES)
1386 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1387 		break;
1388 	case NI_QTYPE_IPV4ADDR:
1389 		/* unsupported - should respond with unknown Qtype? */
1390 		break;
1391 	default:
1392 		/*
1393 		 * XXX: We must return a reply with the ICMP6 code
1394 		 * `unknown Qtype' in this case.  However we regard the case
1395 		 * as an FQDN query for backward compatibility.
1396 		 * Older versions set a random value to this field,
1397 		 * so it rarely varies in the defined qtypes.
1398 		 * But the mechanism is not reliable...
1399 		 * maybe we should obsolete older versions.
1400 		 */
1401 		qtype = NI_QTYPE_FQDN;
1402 		/* XXX will append an mbuf */
1403 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1404 		oldfqdn++;
1405 		break;
1406 	}
1407 
1408 	/* Allocate an mbuf to reply. */
1409 	if (replylen > MCLBYTES) {
1410 		/*
1411 		 * XXX: should we try to allocate more? But MCLBYTES
1412 		 * is probably much larger than IPV6_MMTU...
1413 		 */
1414 		goto bad;
1415 	}
1416 	if (replylen > MHLEN)
1417 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1418 	else
1419 		n = m_gethdr(M_NOWAIT, m->m_type);
1420 	if (n == NULL) {
1421 		m_freem(m);
1422 		return (NULL);
1423 	}
1424 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1425 	n->m_pkthdr.len = n->m_len = replylen;
1426 
1427 	/* copy mbuf header and IPv6 + Node Information base headers */
1428 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1429 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1430 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1431 
1432 	/* qtype dependent procedure */
1433 	switch (qtype) {
1434 	case NI_QTYPE_NOOP:
1435 		nni6->ni_code = ICMP6_NI_SUCCESS;
1436 		nni6->ni_flags = 0;
1437 		break;
1438 	case NI_QTYPE_SUPTYPES:
1439 	{
1440 		u_int32_t v;
1441 		nni6->ni_code = ICMP6_NI_SUCCESS;
1442 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1443 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1444 		v = (u_int32_t)htonl(0x0000000f);
1445 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1446 		break;
1447 	}
1448 	case NI_QTYPE_FQDN:
1449 		nni6->ni_code = ICMP6_NI_SUCCESS;
1450 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1451 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1452 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1453 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1454 		/*
1455 		 * XXX do we really have FQDN in hostname?
1456 		 */
1457 		mtx_lock(&pr->pr_mtx);
1458 		n->m_next = ni6_nametodns(pr->pr_hostname,
1459 		    strlen(pr->pr_hostname), oldfqdn);
1460 		mtx_unlock(&pr->pr_mtx);
1461 		if (n->m_next == NULL)
1462 			goto bad;
1463 		/* XXX we assume that n->m_next is not a chain */
1464 		if (n->m_next->m_next != NULL)
1465 			goto bad;
1466 		n->m_pkthdr.len += n->m_next->m_len;
1467 		break;
1468 	case NI_QTYPE_NODEADDR:
1469 	{
1470 		int lenlim, copied;
1471 
1472 		nni6->ni_code = ICMP6_NI_SUCCESS;
1473 		n->m_pkthdr.len = n->m_len =
1474 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1475 		lenlim = M_TRAILINGSPACE(n);
1476 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1477 		/* XXX: reset mbuf length */
1478 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1479 		    sizeof(struct icmp6_nodeinfo) + copied;
1480 		break;
1481 	}
1482 	default:
1483 		break;		/* XXX impossible! */
1484 	}
1485 
1486 	nni6->ni_type = ICMP6_NI_REPLY;
1487 	m_freem(m);
1488 	return (n);
1489 
1490   bad:
1491 	m_freem(m);
1492 	if (n)
1493 		m_freem(n);
1494 	return (NULL);
1495 }
1496 
1497 /*
1498  * make a mbuf with DNS-encoded string.  no compression support.
1499  *
1500  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1501  * treated as truncated name (two \0 at the end).  this is a wild guess.
1502  *
1503  * old - return pascal string if non-zero
1504  */
1505 static struct mbuf *
1506 ni6_nametodns(const char *name, int namelen, int old)
1507 {
1508 	struct mbuf *m;
1509 	char *cp, *ep;
1510 	const char *p, *q;
1511 	int i, len, nterm;
1512 
1513 	if (old)
1514 		len = namelen + 1;
1515 	else
1516 		len = MCLBYTES;
1517 
1518 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1519 	if (len > MLEN)
1520 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1521 	else
1522 		m = m_get(M_NOWAIT, MT_DATA);
1523 	if (m == NULL)
1524 		goto fail;
1525 
1526 	if (old) {
1527 		m->m_len = len;
1528 		*mtod(m, char *) = namelen;
1529 		bcopy(name, mtod(m, char *) + 1, namelen);
1530 		return m;
1531 	} else {
1532 		m->m_len = 0;
1533 		cp = mtod(m, char *);
1534 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1535 
1536 		/* if not certain about my name, return empty buffer */
1537 		if (namelen == 0)
1538 			return m;
1539 
1540 		/*
1541 		 * guess if it looks like shortened hostname, or FQDN.
1542 		 * shortened hostname needs two trailing "\0".
1543 		 */
1544 		i = 0;
1545 		for (p = name; p < name + namelen; p++) {
1546 			if (*p && *p == '.')
1547 				i++;
1548 		}
1549 		if (i < 2)
1550 			nterm = 2;
1551 		else
1552 			nterm = 1;
1553 
1554 		p = name;
1555 		while (cp < ep && p < name + namelen) {
1556 			i = 0;
1557 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1558 				i++;
1559 			/* result does not fit into mbuf */
1560 			if (cp + i + 1 >= ep)
1561 				goto fail;
1562 			/*
1563 			 * DNS label length restriction, RFC1035 page 8.
1564 			 * "i == 0" case is included here to avoid returning
1565 			 * 0-length label on "foo..bar".
1566 			 */
1567 			if (i <= 0 || i >= 64)
1568 				goto fail;
1569 			*cp++ = i;
1570 			bcopy(p, cp, i);
1571 			cp += i;
1572 			p = q;
1573 			if (p < name + namelen && *p == '.')
1574 				p++;
1575 		}
1576 		/* termination */
1577 		if (cp + nterm >= ep)
1578 			goto fail;
1579 		while (nterm-- > 0)
1580 			*cp++ = '\0';
1581 		m->m_len = cp - mtod(m, char *);
1582 		return m;
1583 	}
1584 
1585 	panic("should not reach here");
1586 	/* NOTREACHED */
1587 
1588  fail:
1589 	if (m)
1590 		m_freem(m);
1591 	return NULL;
1592 }
1593 
1594 /*
1595  * check if two DNS-encoded string matches.  takes care of truncated
1596  * form (with \0\0 at the end).  no compression support.
1597  * XXX upper/lowercase match (see RFC2065)
1598  */
1599 static int
1600 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1601 {
1602 	const char *a0, *b0;
1603 	int l;
1604 
1605 	/* simplest case - need validation? */
1606 	if (alen == blen && bcmp(a, b, alen) == 0)
1607 		return 1;
1608 
1609 	a0 = a;
1610 	b0 = b;
1611 
1612 	/* termination is mandatory */
1613 	if (alen < 2 || blen < 2)
1614 		return 0;
1615 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1616 		return 0;
1617 	alen--;
1618 	blen--;
1619 
1620 	while (a - a0 < alen && b - b0 < blen) {
1621 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1622 			return 0;
1623 
1624 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1625 			return 0;
1626 		/* we don't support compression yet */
1627 		if (a[0] >= 64 || b[0] >= 64)
1628 			return 0;
1629 
1630 		/* truncated case */
1631 		if (a[0] == 0 && a - a0 == alen - 1)
1632 			return 1;
1633 		if (b[0] == 0 && b - b0 == blen - 1)
1634 			return 1;
1635 		if (a[0] == 0 || b[0] == 0)
1636 			return 0;
1637 
1638 		if (a[0] != b[0])
1639 			return 0;
1640 		l = a[0];
1641 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1642 			return 0;
1643 		if (bcmp(a + 1, b + 1, l) != 0)
1644 			return 0;
1645 
1646 		a += 1 + l;
1647 		b += 1 + l;
1648 	}
1649 
1650 	if (a - a0 == alen && b - b0 == blen)
1651 		return 1;
1652 	else
1653 		return 0;
1654 }
1655 
1656 /*
1657  * calculate the number of addresses to be returned in the node info reply.
1658  */
1659 static int
1660 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1661     struct in6_addr *subj)
1662 {
1663 	struct epoch_tracker et;
1664 	struct ifnet *ifp;
1665 	struct in6_ifaddr *ifa6;
1666 	struct ifaddr *ifa;
1667 	int addrs = 0, addrsofif, iffound = 0;
1668 	int niflags = ni6->ni_flags;
1669 
1670 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1671 		switch (ni6->ni_code) {
1672 		case ICMP6_NI_SUBJ_IPV6:
1673 			if (subj == NULL) /* must be impossible... */
1674 				return (0);
1675 			break;
1676 		default:
1677 			/*
1678 			 * XXX: we only support IPv6 subject address for
1679 			 * this Qtype.
1680 			 */
1681 			return (0);
1682 		}
1683 	}
1684 
1685 	NET_EPOCH_ENTER(et);
1686 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1687 		addrsofif = 0;
1688 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1689 			if (ifa->ifa_addr->sa_family != AF_INET6)
1690 				continue;
1691 			ifa6 = (struct in6_ifaddr *)ifa;
1692 
1693 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1694 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1695 				iffound = 1;
1696 
1697 			/*
1698 			 * IPv4-mapped addresses can only be returned by a
1699 			 * Node Information proxy, since they represent
1700 			 * addresses of IPv4-only nodes, which perforce do
1701 			 * not implement this protocol.
1702 			 * [icmp-name-lookups-07, Section 5.4]
1703 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1704 			 * this function at this moment.
1705 			 */
1706 
1707 			/* What do we have to do about ::1? */
1708 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1709 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1710 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1711 					continue;
1712 				break;
1713 			case IPV6_ADDR_SCOPE_SITELOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_GLOBAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1719 					continue;
1720 				break;
1721 			default:
1722 				continue;
1723 			}
1724 
1725 			/*
1726 			 * check if anycast is okay.
1727 			 * XXX: just experimental.  not in the spec.
1728 			 */
1729 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1730 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1731 				continue; /* we need only unicast addresses */
1732 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1733 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1734 				continue;
1735 			}
1736 			addrsofif++; /* count the address */
1737 		}
1738 		if (iffound) {
1739 			*ifpp = ifp;
1740 			NET_EPOCH_EXIT(et);
1741 			return (addrsofif);
1742 		}
1743 
1744 		addrs += addrsofif;
1745 	}
1746 	NET_EPOCH_EXIT(et);
1747 
1748 	return (addrs);
1749 }
1750 
1751 static int
1752 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1753     struct ifnet *ifp0, int resid)
1754 {
1755 	struct epoch_tracker et;
1756 	struct ifnet *ifp;
1757 	struct in6_ifaddr *ifa6;
1758 	struct ifaddr *ifa;
1759 	struct ifnet *ifp_dep = NULL;
1760 	int copied = 0, allow_deprecated = 0;
1761 	u_char *cp = (u_char *)(nni6 + 1);
1762 	int niflags = ni6->ni_flags;
1763 	u_int32_t ltime;
1764 
1765 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1766 		return (0);	/* needless to copy */
1767 
1768 	NET_EPOCH_ENTER(et);
1769 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1770   again:
1771 
1772 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1773 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1774 			if (ifa->ifa_addr->sa_family != AF_INET6)
1775 				continue;
1776 			ifa6 = (struct in6_ifaddr *)ifa;
1777 
1778 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1779 			    allow_deprecated == 0) {
1780 				/*
1781 				 * prefererred address should be put before
1782 				 * deprecated addresses.
1783 				 */
1784 
1785 				/* record the interface for later search */
1786 				if (ifp_dep == NULL)
1787 					ifp_dep = ifp;
1788 
1789 				continue;
1790 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1791 			    allow_deprecated != 0)
1792 				continue; /* we now collect deprecated addrs */
1793 
1794 			/* What do we have to do about ::1? */
1795 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1796 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1797 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1798 					continue;
1799 				break;
1800 			case IPV6_ADDR_SCOPE_SITELOCAL:
1801 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1802 					continue;
1803 				break;
1804 			case IPV6_ADDR_SCOPE_GLOBAL:
1805 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1806 					continue;
1807 				break;
1808 			default:
1809 				continue;
1810 			}
1811 
1812 			/*
1813 			 * check if anycast is okay.
1814 			 * XXX: just experimental.  not in the spec.
1815 			 */
1816 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1817 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1818 				continue;
1819 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1820 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1821 				continue;
1822 			}
1823 
1824 			/* now we can copy the address */
1825 			if (resid < sizeof(struct in6_addr) +
1826 			    sizeof(u_int32_t)) {
1827 				/*
1828 				 * We give up much more copy.
1829 				 * Set the truncate flag and return.
1830 				 */
1831 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1832 				NET_EPOCH_EXIT(et);
1833 				return (copied);
1834 			}
1835 
1836 			/*
1837 			 * Set the TTL of the address.
1838 			 * The TTL value should be one of the following
1839 			 * according to the specification:
1840 			 *
1841 			 * 1. The remaining lifetime of a DHCP lease on the
1842 			 *    address, or
1843 			 * 2. The remaining Valid Lifetime of a prefix from
1844 			 *    which the address was derived through Stateless
1845 			 *    Autoconfiguration.
1846 			 *
1847 			 * Note that we currently do not support stateful
1848 			 * address configuration by DHCPv6, so the former
1849 			 * case can't happen.
1850 			 */
1851 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1852 				ltime = ND6_INFINITE_LIFETIME;
1853 			else {
1854 				if (ifa6->ia6_lifetime.ia6t_expire >
1855 				    time_uptime)
1856 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1857 				else
1858 					ltime = 0;
1859 			}
1860 
1861 			bcopy(&ltime, cp, sizeof(u_int32_t));
1862 			cp += sizeof(u_int32_t);
1863 
1864 			/* copy the address itself */
1865 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1866 			    sizeof(struct in6_addr));
1867 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1868 			cp += sizeof(struct in6_addr);
1869 
1870 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1871 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1872 		}
1873 		if (ifp0)	/* we need search only on the specified IF */
1874 			break;
1875 	}
1876 
1877 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1878 		ifp = ifp_dep;
1879 		allow_deprecated = 1;
1880 
1881 		goto again;
1882 	}
1883 
1884 	NET_EPOCH_EXIT(et);
1885 
1886 	return (copied);
1887 }
1888 
1889 /*
1890  * XXX almost dup'ed code with rip6_input.
1891  */
1892 static int
1893 icmp6_rip6_input(struct mbuf **mp, int off)
1894 {
1895 	struct mbuf *m = *mp;
1896 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1897 	struct inpcb *inp;
1898 	struct inpcb *last = NULL;
1899 	struct sockaddr_in6 fromsa;
1900 	struct icmp6_hdr *icmp6;
1901 	struct epoch_tracker et;
1902 	struct mbuf *opts = NULL;
1903 
1904 #ifndef PULLDOWN_TEST
1905 	/* this is assumed to be safe. */
1906 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1907 #else
1908 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1909 	if (icmp6 == NULL) {
1910 		/* m is already reclaimed */
1911 		return (IPPROTO_DONE);
1912 	}
1913 #endif
1914 
1915 	/*
1916 	 * XXX: the address may have embedded scope zone ID, which should be
1917 	 * hidden from applications.
1918 	 */
1919 	bzero(&fromsa, sizeof(fromsa));
1920 	fromsa.sin6_family = AF_INET6;
1921 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1922 	fromsa.sin6_addr = ip6->ip6_src;
1923 	if (sa6_recoverscope(&fromsa)) {
1924 		m_freem(m);
1925 		return (IPPROTO_DONE);
1926 	}
1927 
1928 	INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1929 	CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
1930 		if ((inp->inp_vflag & INP_IPV6) == 0)
1931 			continue;
1932 		if (inp->inp_ip_p != IPPROTO_ICMPV6)
1933 			continue;
1934 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1935 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1936 			continue;
1937 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1938 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1939 			continue;
1940 		INP_RLOCK(inp);
1941 		if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1942 			INP_RUNLOCK(inp);
1943 			continue;
1944 		}
1945 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1946 		    inp->in6p_icmp6filt)) {
1947 			INP_RUNLOCK(inp);
1948 			continue;
1949 		}
1950 		if (last != NULL) {
1951 			struct	mbuf *n = NULL;
1952 
1953 			/*
1954 			 * Recent network drivers tend to allocate a single
1955 			 * mbuf cluster, rather than to make a couple of
1956 			 * mbufs without clusters.  Also, since the IPv6 code
1957 			 * path tries to avoid m_pullup(), it is highly
1958 			 * probable that we still have an mbuf cluster here
1959 			 * even though the necessary length can be stored in an
1960 			 * mbuf's internal buffer.
1961 			 * Meanwhile, the default size of the receive socket
1962 			 * buffer for raw sockets is not so large.  This means
1963 			 * the possibility of packet loss is relatively higher
1964 			 * than before.  To avoid this scenario, we copy the
1965 			 * received data to a separate mbuf that does not use
1966 			 * a cluster, if possible.
1967 			 * XXX: it is better to copy the data after stripping
1968 			 * intermediate headers.
1969 			 */
1970 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1971 			    m->m_len <= MHLEN) {
1972 				n = m_get(M_NOWAIT, m->m_type);
1973 				if (n != NULL) {
1974 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1975 						bcopy(m->m_data, n->m_data,
1976 						      m->m_len);
1977 						n->m_len = m->m_len;
1978 					} else {
1979 						m_free(n);
1980 						n = NULL;
1981 					}
1982 				}
1983 			}
1984 			if (n != NULL ||
1985 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1986 				if (last->inp_flags & INP_CONTROLOPTS)
1987 					ip6_savecontrol(last, n, &opts);
1988 				/* strip intermediate headers */
1989 				m_adj(n, off);
1990 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1991 				if (sbappendaddr_locked(
1992 				    &last->inp_socket->so_rcv,
1993 				    (struct sockaddr *)&fromsa, n, opts)
1994 				    == 0) {
1995 					/* should notify about lost packet */
1996 					m_freem(n);
1997 					if (opts) {
1998 						m_freem(opts);
1999 					}
2000 					SOCKBUF_UNLOCK(
2001 					    &last->inp_socket->so_rcv);
2002 				} else
2003 					sorwakeup_locked(last->inp_socket);
2004 				opts = NULL;
2005 			}
2006 			INP_RUNLOCK(last);
2007 		}
2008 		last = inp;
2009 	}
2010 	INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
2011 	if (last != NULL) {
2012 		if (last->inp_flags & INP_CONTROLOPTS)
2013 			ip6_savecontrol(last, m, &opts);
2014 		/* strip intermediate headers */
2015 		m_adj(m, off);
2016 
2017 		/* avoid using mbuf clusters if possible (see above) */
2018 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2019 		    m->m_len <= MHLEN) {
2020 			struct mbuf *n;
2021 
2022 			n = m_get(M_NOWAIT, m->m_type);
2023 			if (n != NULL) {
2024 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2025 					bcopy(m->m_data, n->m_data, m->m_len);
2026 					n->m_len = m->m_len;
2027 
2028 					m_freem(m);
2029 					m = n;
2030 				} else {
2031 					m_freem(n);
2032 					n = NULL;
2033 				}
2034 			}
2035 		}
2036 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2037 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2038 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2039 			m_freem(m);
2040 			if (opts)
2041 				m_freem(opts);
2042 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2043 		} else
2044 			sorwakeup_locked(last->inp_socket);
2045 		INP_RUNLOCK(last);
2046 	} else {
2047 		m_freem(m);
2048 		IP6STAT_DEC(ip6s_delivered);
2049 	}
2050 	return IPPROTO_DONE;
2051 }
2052 
2053 /*
2054  * Reflect the ip6 packet back to the source.
2055  * OFF points to the icmp6 header, counted from the top of the mbuf.
2056  */
2057 void
2058 icmp6_reflect(struct mbuf *m, size_t off)
2059 {
2060 	struct in6_addr src6, *srcp;
2061 	struct ip6_hdr *ip6;
2062 	struct icmp6_hdr *icmp6;
2063 	struct in6_ifaddr *ia = NULL;
2064 	struct ifnet *outif = NULL;
2065 	int plen;
2066 	int type, code, hlim;
2067 
2068 	/* too short to reflect */
2069 	if (off < sizeof(struct ip6_hdr)) {
2070 		nd6log((LOG_DEBUG,
2071 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2072 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2073 		    __FILE__, __LINE__));
2074 		goto bad;
2075 	}
2076 
2077 	/*
2078 	 * If there are extra headers between IPv6 and ICMPv6, strip
2079 	 * off that header first.
2080 	 */
2081 #ifdef DIAGNOSTIC
2082 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2083 		panic("assumption failed in icmp6_reflect");
2084 #endif
2085 	if (off > sizeof(struct ip6_hdr)) {
2086 		size_t l;
2087 		struct ip6_hdr nip6;
2088 
2089 		l = off - sizeof(struct ip6_hdr);
2090 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2091 		m_adj(m, l);
2092 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2093 		if (m->m_len < l) {
2094 			if ((m = m_pullup(m, l)) == NULL)
2095 				return;
2096 		}
2097 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2098 	} else /* off == sizeof(struct ip6_hdr) */ {
2099 		size_t l;
2100 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2101 		if (m->m_len < l) {
2102 			if ((m = m_pullup(m, l)) == NULL)
2103 				return;
2104 		}
2105 	}
2106 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2107 	ip6 = mtod(m, struct ip6_hdr *);
2108 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2109 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2110 	type = icmp6->icmp6_type; /* keep type for statistics */
2111 	code = icmp6->icmp6_code; /* ditto. */
2112 	hlim = 0;
2113 	srcp = NULL;
2114 
2115 	/*
2116 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2117 	 * use dst as the src for the reply.
2118 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2119 	 * (for example) when we encounter an error while forwarding procedure
2120 	 * destined to a duplicated address of ours.
2121 	 */
2122 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2123 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2124 		if (ia != NULL && !(ia->ia6_flags &
2125 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2126 			src6 = ia->ia_addr.sin6_addr;
2127 			srcp = &src6;
2128 
2129 			if (m->m_pkthdr.rcvif != NULL) {
2130 				/* XXX: This may not be the outgoing interface */
2131 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2132 			} else
2133 				hlim = V_ip6_defhlim;
2134 		}
2135 		if (ia != NULL)
2136 			ifa_free(&ia->ia_ifa);
2137 	}
2138 
2139 	if (srcp == NULL) {
2140 		int error;
2141 		struct in6_addr dst6;
2142 		uint32_t scopeid;
2143 
2144 		/*
2145 		 * This case matches to multicasts, our anycast, or unicasts
2146 		 * that we do not own.  Select a source address based on the
2147 		 * source address of the erroneous packet.
2148 		 */
2149 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2150 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2151 		    scopeid, NULL, &src6, &hlim);
2152 
2153 		if (error) {
2154 			char ip6buf[INET6_ADDRSTRLEN];
2155 			nd6log((LOG_DEBUG,
2156 			    "icmp6_reflect: source can't be determined: "
2157 			    "dst=%s, error=%d\n",
2158 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2159 			goto bad;
2160 		}
2161 		srcp = &src6;
2162 	}
2163 	/*
2164 	 * ip6_input() drops a packet if its src is multicast.
2165 	 * So, the src is never multicast.
2166 	 */
2167 	ip6->ip6_dst = ip6->ip6_src;
2168 	ip6->ip6_src = *srcp;
2169 	ip6->ip6_flow = 0;
2170 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2171 	ip6->ip6_vfc |= IPV6_VERSION;
2172 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2173 	ip6->ip6_hlim = hlim;
2174 
2175 	icmp6->icmp6_cksum = 0;
2176 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2177 	    sizeof(struct ip6_hdr), plen);
2178 
2179 	/*
2180 	 * XXX option handling
2181 	 */
2182 
2183 	m->m_flags &= ~(M_BCAST|M_MCAST);
2184 	m->m_pkthdr.rcvif = NULL;
2185 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2186 	if (outif)
2187 		icmp6_ifoutstat_inc(outif, type, code);
2188 
2189 	return;
2190 
2191  bad:
2192 	m_freem(m);
2193 	return;
2194 }
2195 
2196 void
2197 icmp6_fasttimo(void)
2198 {
2199 
2200 	mld_fasttimo();
2201 }
2202 
2203 void
2204 icmp6_slowtimo(void)
2205 {
2206 
2207 	mld_slowtimo();
2208 }
2209 
2210 static const char *
2211 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2212     struct in6_addr *tgt6)
2213 {
2214 	static char buf[1024];
2215 	char ip6bufs[INET6_ADDRSTRLEN];
2216 	char ip6bufd[INET6_ADDRSTRLEN];
2217 	char ip6buft[INET6_ADDRSTRLEN];
2218 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2219 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2220 	    ip6_sprintf(ip6buft, tgt6));
2221 	return buf;
2222 }
2223 
2224 void
2225 icmp6_redirect_input(struct mbuf *m, int off)
2226 {
2227 	struct ifnet *ifp;
2228 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2229 	struct nd_redirect *nd_rd;
2230 	int icmp6len = ntohs(ip6->ip6_plen);
2231 	char *lladdr = NULL;
2232 	int lladdrlen = 0;
2233 	int is_router;
2234 	int is_onlink;
2235 	struct in6_addr src6 = ip6->ip6_src;
2236 	struct in6_addr redtgt6;
2237 	struct in6_addr reddst6;
2238 	union nd_opts ndopts;
2239 	char ip6buf[INET6_ADDRSTRLEN];
2240 
2241 	M_ASSERTPKTHDR(m);
2242 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2243 
2244 	ifp = m->m_pkthdr.rcvif;
2245 
2246 	/* XXX if we are router, we don't update route by icmp6 redirect */
2247 	if (V_ip6_forwarding)
2248 		goto freeit;
2249 	if (!V_icmp6_rediraccept)
2250 		goto freeit;
2251 
2252 	/* RFC 6980: Nodes MUST silently ignore fragments */
2253 	if(m->m_flags & M_FRAGMENTED)
2254 		goto freeit;
2255 
2256 #ifndef PULLDOWN_TEST
2257 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2258 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2259 #else
2260 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2261 	if (nd_rd == NULL) {
2262 		ICMP6STAT_INC(icp6s_tooshort);
2263 		return;
2264 	}
2265 #endif
2266 	redtgt6 = nd_rd->nd_rd_target;
2267 	reddst6 = nd_rd->nd_rd_dst;
2268 
2269 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2270 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2271 		goto freeit;
2272 	}
2273 
2274 	/* validation */
2275 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2276 		nd6log((LOG_ERR,
2277 		    "ICMP6 redirect sent from %s rejected; "
2278 		    "must be from linklocal\n",
2279 		    ip6_sprintf(ip6buf, &src6)));
2280 		goto bad;
2281 	}
2282 	if (ip6->ip6_hlim != 255) {
2283 		nd6log((LOG_ERR,
2284 		    "ICMP6 redirect sent from %s rejected; "
2285 		    "hlim=%d (must be 255)\n",
2286 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2287 		goto bad;
2288 	}
2289     {
2290 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2291 	struct nhop6_basic nh6;
2292 	struct in6_addr kdst;
2293 	uint32_t scopeid;
2294 
2295 	in6_splitscope(&reddst6, &kdst, &scopeid);
2296 	if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2297 		if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2298 			nd6log((LOG_ERR,
2299 			    "ICMP6 redirect rejected; no route "
2300 			    "with inet6 gateway found for redirect dst: %s\n",
2301 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2302 			goto bad;
2303 		}
2304 
2305 		/*
2306 		 * Embed scope zone id into next hop address, since
2307 		 * fib6_lookup_nh_basic() returns address without embedded
2308 		 * scope zone id.
2309 		 */
2310 		if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL))
2311 			goto freeit;
2312 
2313 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2314 			nd6log((LOG_ERR,
2315 			    "ICMP6 redirect rejected; "
2316 			    "not equal to gw-for-src=%s (must be same): "
2317 			    "%s\n",
2318 			    ip6_sprintf(ip6buf, &nh6.nh_addr),
2319 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2320 			goto bad;
2321 		}
2322 	} else {
2323 		nd6log((LOG_ERR,
2324 		    "ICMP6 redirect rejected; "
2325 		    "no route found for redirect dst: %s\n",
2326 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2327 		goto bad;
2328 	}
2329     }
2330 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2331 		nd6log((LOG_ERR,
2332 		    "ICMP6 redirect rejected; "
2333 		    "redirect dst must be unicast: %s\n",
2334 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2335 		goto bad;
2336 	}
2337 
2338 	is_router = is_onlink = 0;
2339 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2340 		is_router = 1;	/* router case */
2341 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2342 		is_onlink = 1;	/* on-link destination case */
2343 	if (!is_router && !is_onlink) {
2344 		nd6log((LOG_ERR,
2345 		    "ICMP6 redirect rejected; "
2346 		    "neither router case nor onlink case: %s\n",
2347 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2348 		goto bad;
2349 	}
2350 
2351 	icmp6len -= sizeof(*nd_rd);
2352 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2353 	if (nd6_options(&ndopts) < 0) {
2354 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2355 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2356 		/* nd6_options have incremented stats */
2357 		goto freeit;
2358 	}
2359 
2360 	if (ndopts.nd_opts_tgt_lladdr) {
2361 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2362 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2363 	}
2364 
2365 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2366 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2367 		    "(if %d, icmp6 packet %d): %s\n",
2368 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2369 		    ifp->if_addrlen, lladdrlen - 2,
2370 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2371 		goto bad;
2372 	}
2373 
2374 	/* Validation passed. */
2375 
2376 	/* RFC 2461 8.3 */
2377 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2378 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2379 
2380 	/*
2381 	 * Install a gateway route in the better-router case or an interface
2382 	 * route in the on-link-destination case.
2383 	 */
2384 	{
2385 		struct sockaddr_in6 sdst;
2386 		struct sockaddr_in6 sgw;
2387 		struct sockaddr_in6 ssrc;
2388 		struct sockaddr *gw;
2389 		int rt_flags;
2390 		u_int fibnum;
2391 
2392 		bzero(&sdst, sizeof(sdst));
2393 		bzero(&ssrc, sizeof(ssrc));
2394 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2395 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2396 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2397 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2398 		rt_flags = RTF_HOST;
2399 		if (is_router) {
2400 			bzero(&sgw, sizeof(sgw));
2401 			sgw.sin6_family = AF_INET6;
2402 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2403 			bcopy(&redtgt6, &sgw.sin6_addr,
2404 				sizeof(struct in6_addr));
2405 			gw = (struct sockaddr *)&sgw;
2406 			rt_flags |= RTF_GATEWAY;
2407 		} else
2408 			gw = ifp->if_addr->ifa_addr;
2409 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2410 			in6_rtredirect((struct sockaddr *)&sdst, gw,
2411 			    (struct sockaddr *)NULL, rt_flags,
2412 			    (struct sockaddr *)&ssrc, fibnum);
2413 	}
2414 	/* finally update cached route in each socket via pfctlinput */
2415     {
2416 	struct sockaddr_in6 sdst;
2417 
2418 	bzero(&sdst, sizeof(sdst));
2419 	sdst.sin6_family = AF_INET6;
2420 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2421 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2422 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2423     }
2424 
2425  freeit:
2426 	m_freem(m);
2427 	return;
2428 
2429  bad:
2430 	ICMP6STAT_INC(icp6s_badredirect);
2431 	m_freem(m);
2432 }
2433 
2434 void
2435 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2436 {
2437 	struct ifnet *ifp;	/* my outgoing interface */
2438 	struct in6_addr *ifp_ll6;
2439 	struct in6_addr *router_ll6;
2440 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2441 	struct mbuf *m = NULL;	/* newly allocated one */
2442 	struct m_tag *mtag;
2443 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2444 	struct nd_redirect *nd_rd;
2445 	struct llentry *ln = NULL;
2446 	size_t maxlen;
2447 	u_char *p;
2448 	struct ifnet *outif = NULL;
2449 	struct sockaddr_in6 src_sa;
2450 
2451 	icmp6_errcount(ND_REDIRECT, 0);
2452 
2453 	/* if we are not router, we don't send icmp6 redirect */
2454 	if (!V_ip6_forwarding)
2455 		goto fail;
2456 
2457 	/* sanity check */
2458 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2459 		goto fail;
2460 
2461 	/*
2462 	 * Address check:
2463 	 *  the source address must identify a neighbor, and
2464 	 *  the destination address must not be a multicast address
2465 	 *  [RFC 2461, sec 8.2]
2466 	 */
2467 	sip6 = mtod(m0, struct ip6_hdr *);
2468 	bzero(&src_sa, sizeof(src_sa));
2469 	src_sa.sin6_family = AF_INET6;
2470 	src_sa.sin6_len = sizeof(src_sa);
2471 	src_sa.sin6_addr = sip6->ip6_src;
2472 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2473 		goto fail;
2474 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2475 		goto fail;	/* what should we do here? */
2476 
2477 	/* rate limit */
2478 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2479 		goto fail;
2480 
2481 	/*
2482 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2483 	 * we almost always ask for an mbuf cluster for simplicity.
2484 	 * (MHLEN < IPV6_MMTU is almost always true)
2485 	 */
2486 #if IPV6_MMTU >= MCLBYTES
2487 # error assumption failed about IPV6_MMTU and MCLBYTES
2488 #endif
2489 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2490 	if (m == NULL)
2491 		goto fail;
2492 	M_SETFIB(m, rt->rt_fibnum);
2493 	maxlen = M_TRAILINGSPACE(m);
2494 	maxlen = min(IPV6_MMTU, maxlen);
2495 	/* just for safety */
2496 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2497 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2498 		goto fail;
2499 	}
2500 
2501 	{
2502 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2503 		struct in6_ifaddr *ia;
2504 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2505 						 IN6_IFF_NOTREADY|
2506 						 IN6_IFF_ANYCAST)) == NULL)
2507 			goto fail;
2508 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2509 		/* XXXRW: reference released prematurely. */
2510 		ifa_free(&ia->ia_ifa);
2511 	}
2512 
2513 	/* get ip6 linklocal address for the router. */
2514 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2515 		struct sockaddr_in6 *sin6;
2516 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2517 		router_ll6 = &sin6->sin6_addr;
2518 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2519 			router_ll6 = (struct in6_addr *)NULL;
2520 	} else
2521 		router_ll6 = (struct in6_addr *)NULL;
2522 
2523 	/* ip6 */
2524 	ip6 = mtod(m, struct ip6_hdr *);
2525 	ip6->ip6_flow = 0;
2526 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2527 	ip6->ip6_vfc |= IPV6_VERSION;
2528 	/* ip6->ip6_plen will be set later */
2529 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2530 	ip6->ip6_hlim = 255;
2531 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2532 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2533 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2534 
2535 	/* ND Redirect */
2536 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2537 	nd_rd->nd_rd_type = ND_REDIRECT;
2538 	nd_rd->nd_rd_code = 0;
2539 	nd_rd->nd_rd_reserved = 0;
2540 	if (rt->rt_flags & RTF_GATEWAY) {
2541 		/*
2542 		 * nd_rd->nd_rd_target must be a link-local address in
2543 		 * better router cases.
2544 		 */
2545 		if (!router_ll6)
2546 			goto fail;
2547 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2548 		    sizeof(nd_rd->nd_rd_target));
2549 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2550 		    sizeof(nd_rd->nd_rd_dst));
2551 	} else {
2552 		/* make sure redtgt == reddst */
2553 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2554 		    sizeof(nd_rd->nd_rd_target));
2555 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2556 		    sizeof(nd_rd->nd_rd_dst));
2557 	}
2558 
2559 	p = (u_char *)(nd_rd + 1);
2560 
2561 	if (!router_ll6)
2562 		goto nolladdropt;
2563 
2564 	{
2565 		/* target lladdr option */
2566 		struct epoch_tracker et;
2567 		int len;
2568 		struct nd_opt_hdr *nd_opt;
2569 		char *lladdr;
2570 
2571 		NET_EPOCH_ENTER(et);
2572 		ln = nd6_lookup(router_ll6, 0, ifp);
2573 		NET_EPOCH_EXIT(et);
2574 		if (ln == NULL)
2575 			goto nolladdropt;
2576 
2577 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2578 		len = (len + 7) & ~7;	/* round by 8 */
2579 		/* safety check */
2580 		if (len + (p - (u_char *)ip6) > maxlen)
2581 			goto nolladdropt;
2582 
2583 		if (ln->la_flags & LLE_VALID) {
2584 			nd_opt = (struct nd_opt_hdr *)p;
2585 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2586 			nd_opt->nd_opt_len = len >> 3;
2587 			lladdr = (char *)(nd_opt + 1);
2588 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2589 			p += len;
2590 		}
2591 	}
2592 nolladdropt:
2593 	if (ln != NULL)
2594 		LLE_RUNLOCK(ln);
2595 
2596 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2597 
2598 	/* just to be safe */
2599 #ifdef M_DECRYPTED	/*not openbsd*/
2600 	if (m0->m_flags & M_DECRYPTED)
2601 		goto noredhdropt;
2602 #endif
2603 	if (p - (u_char *)ip6 > maxlen)
2604 		goto noredhdropt;
2605 
2606 	{
2607 		/* redirected header option */
2608 		int len;
2609 		struct nd_opt_rd_hdr *nd_opt_rh;
2610 
2611 		/*
2612 		 * compute the maximum size for icmp6 redirect header option.
2613 		 * XXX room for auth header?
2614 		 */
2615 		len = maxlen - (p - (u_char *)ip6);
2616 		len &= ~7;
2617 
2618 		/* This is just for simplicity. */
2619 		if (m0->m_pkthdr.len != m0->m_len) {
2620 			if (m0->m_next) {
2621 				m_freem(m0->m_next);
2622 				m0->m_next = NULL;
2623 			}
2624 			m0->m_pkthdr.len = m0->m_len;
2625 		}
2626 
2627 		/*
2628 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2629 		 * about padding/truncate rule for the original IP packet.
2630 		 * From the discussion on IPv6imp in Feb 1999,
2631 		 * the consensus was:
2632 		 * - "attach as much as possible" is the goal
2633 		 * - pad if not aligned (original size can be guessed by
2634 		 *   original ip6 header)
2635 		 * Following code adds the padding if it is simple enough,
2636 		 * and truncates if not.
2637 		 */
2638 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2639 			panic("assumption failed in %s:%d", __FILE__,
2640 			    __LINE__);
2641 
2642 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2643 			/* not enough room, truncate */
2644 			m0->m_pkthdr.len = m0->m_len = len -
2645 			    sizeof(*nd_opt_rh);
2646 		} else {
2647 			/* enough room, pad or truncate */
2648 			size_t extra;
2649 
2650 			extra = m0->m_pkthdr.len % 8;
2651 			if (extra) {
2652 				/* pad if easy enough, truncate if not */
2653 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2654 					/* pad */
2655 					m0->m_len += (8 - extra);
2656 					m0->m_pkthdr.len += (8 - extra);
2657 				} else {
2658 					/* truncate */
2659 					m0->m_pkthdr.len -= extra;
2660 					m0->m_len -= extra;
2661 				}
2662 			}
2663 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2664 			m0->m_pkthdr.len = m0->m_len = len -
2665 			    sizeof(*nd_opt_rh);
2666 		}
2667 
2668 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2669 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2670 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2671 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2672 		p += sizeof(*nd_opt_rh);
2673 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2674 
2675 		/* connect m0 to m */
2676 		m_tag_delete_chain(m0, NULL);
2677 		m0->m_flags &= ~M_PKTHDR;
2678 		m->m_next = m0;
2679 		m->m_pkthdr.len = m->m_len + m0->m_len;
2680 		m0 = NULL;
2681 	}
2682 noredhdropt:;
2683 	if (m0) {
2684 		m_freem(m0);
2685 		m0 = NULL;
2686 	}
2687 
2688 	/* XXX: clear embedded link IDs in the inner header */
2689 	in6_clearscope(&sip6->ip6_src);
2690 	in6_clearscope(&sip6->ip6_dst);
2691 	in6_clearscope(&nd_rd->nd_rd_target);
2692 	in6_clearscope(&nd_rd->nd_rd_dst);
2693 
2694 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2695 
2696 	nd_rd->nd_rd_cksum = 0;
2697 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2698 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2699 
2700         if (send_sendso_input_hook != NULL) {
2701 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2702 			M_NOWAIT);
2703 		if (mtag == NULL)
2704 			goto fail;
2705 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2706 		m_tag_prepend(m, mtag);
2707 	}
2708 
2709 	/* send the packet to outside... */
2710 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2711 	if (outif) {
2712 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2713 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2714 	}
2715 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2716 
2717 	return;
2718 
2719 fail:
2720 	if (m)
2721 		m_freem(m);
2722 	if (m0)
2723 		m_freem(m0);
2724 }
2725 
2726 /*
2727  * ICMPv6 socket option processing.
2728  */
2729 int
2730 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2731 {
2732 	int error = 0;
2733 	int optlen;
2734 	struct inpcb *inp = sotoinpcb(so);
2735 	int level, op, optname;
2736 
2737 	if (sopt) {
2738 		level = sopt->sopt_level;
2739 		op = sopt->sopt_dir;
2740 		optname = sopt->sopt_name;
2741 		optlen = sopt->sopt_valsize;
2742 	} else
2743 		level = op = optname = optlen = 0;
2744 
2745 	if (level != IPPROTO_ICMPV6) {
2746 		return EINVAL;
2747 	}
2748 
2749 	switch (op) {
2750 	case PRCO_SETOPT:
2751 		switch (optname) {
2752 		case ICMP6_FILTER:
2753 		    {
2754 			struct icmp6_filter ic6f;
2755 
2756 			if (optlen != sizeof(ic6f)) {
2757 				error = EMSGSIZE;
2758 				break;
2759 			}
2760 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2761 			if (error == 0) {
2762 				INP_WLOCK(inp);
2763 				*inp->in6p_icmp6filt = ic6f;
2764 				INP_WUNLOCK(inp);
2765 			}
2766 			break;
2767 		    }
2768 
2769 		default:
2770 			error = ENOPROTOOPT;
2771 			break;
2772 		}
2773 		break;
2774 
2775 	case PRCO_GETOPT:
2776 		switch (optname) {
2777 		case ICMP6_FILTER:
2778 		    {
2779 			struct icmp6_filter ic6f;
2780 
2781 			INP_RLOCK(inp);
2782 			ic6f = *inp->in6p_icmp6filt;
2783 			INP_RUNLOCK(inp);
2784 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2785 			break;
2786 		    }
2787 
2788 		default:
2789 			error = ENOPROTOOPT;
2790 			break;
2791 		}
2792 		break;
2793 	}
2794 
2795 	return (error);
2796 }
2797 
2798 /*
2799  * Perform rate limit check.
2800  * Returns 0 if it is okay to send the icmp6 packet.
2801  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2802  * limitation.
2803  *
2804  * XXX per-destination/type check necessary?
2805  *
2806  * dst - not used at this moment
2807  * type - not used at this moment
2808  * code - not used at this moment
2809  */
2810 static int
2811 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2812     const int code)
2813 {
2814 	int ret;
2815 
2816 	ret = 0;	/* okay to send */
2817 
2818 	/* PPS limit */
2819 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2820 	    V_icmp6errppslim)) {
2821 		/* The packet is subject to rate limit */
2822 		ret++;
2823 	}
2824 
2825 	return ret;
2826 }
2827