xref: /freebsd/sys/netinet6/icmp6.c (revision 0a36787e4c1fa0cf77dcf83be0867178476e372b)
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/route/route_ctl.h>
97 #include <net/route/nhop.h>
98 #include <net/vnet.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp6.h>
105 #include <netinet/tcp_var.h>
106 
107 #include <netinet6/in6_fib.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/ip6protosw.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/mld6_var.h>
114 #include <netinet6/nd6.h>
115 #include <netinet6/send.h>
116 
117 extern struct domain inet6domain;
118 
119 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
120 VNET_PCPUSTAT_SYSINIT(icmp6stat);
121 
122 #ifdef VIMAGE
123 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
124 #endif /* VIMAGE */
125 
126 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
127 VNET_DECLARE(struct inpcbhead, ripcb);
128 VNET_DECLARE(int, icmp6errppslim);
129 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
130 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
131 VNET_DECLARE(int, icmp6_nodeinfo);
132 
133 #define	V_ripcbinfo			VNET(ripcbinfo)
134 #define	V_ripcb				VNET(ripcb)
135 #define	V_icmp6errppslim		VNET(icmp6errppslim)
136 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
137 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
138 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
139 
140 static void icmp6_errcount(int, int);
141 static int icmp6_rip6_input(struct mbuf **, int);
142 static void icmp6_reflect(struct mbuf *, size_t);
143 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
144 static const char *icmp6_redirect_diag(struct in6_addr *,
145 	struct in6_addr *, struct in6_addr *);
146 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
147 static struct mbuf *ni6_nametodns(const char *, int, int);
148 static int ni6_dnsmatch(const char *, int, const char *, int);
149 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
150 			  struct ifnet **, struct in6_addr *);
151 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
152 				struct ifnet *, int);
153 static int icmp6_notify_error(struct mbuf **, int, int, int);
154 
155 /*
156  * Kernel module interface for updating icmp6stat.  The argument is an index
157  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
158  * general layout of icmp6stat into the caller, it doesn't encode its
159  * location, so that future changes to add, for example, per-CPU stats
160  * support won't cause binary compatibility problems for kernel modules.
161  */
162 void
163 kmod_icmp6stat_inc(int statnum)
164 {
165 
166 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
167 }
168 
169 static void
170 icmp6_errcount(int type, int code)
171 {
172 	switch (type) {
173 	case ICMP6_DST_UNREACH:
174 		switch (code) {
175 		case ICMP6_DST_UNREACH_NOROUTE:
176 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
177 			return;
178 		case ICMP6_DST_UNREACH_ADMIN:
179 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
180 			return;
181 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
182 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
183 			return;
184 		case ICMP6_DST_UNREACH_ADDR:
185 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
186 			return;
187 		case ICMP6_DST_UNREACH_NOPORT:
188 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
189 			return;
190 		}
191 		break;
192 	case ICMP6_PACKET_TOO_BIG:
193 		ICMP6STAT_INC(icp6s_opacket_too_big);
194 		return;
195 	case ICMP6_TIME_EXCEEDED:
196 		switch (code) {
197 		case ICMP6_TIME_EXCEED_TRANSIT:
198 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
199 			return;
200 		case ICMP6_TIME_EXCEED_REASSEMBLY:
201 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
202 			return;
203 		}
204 		break;
205 	case ICMP6_PARAM_PROB:
206 		switch (code) {
207 		case ICMP6_PARAMPROB_HEADER:
208 			ICMP6STAT_INC(icp6s_oparamprob_header);
209 			return;
210 		case ICMP6_PARAMPROB_NEXTHEADER:
211 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
212 			return;
213 		case ICMP6_PARAMPROB_OPTION:
214 			ICMP6STAT_INC(icp6s_oparamprob_option);
215 			return;
216 		}
217 		break;
218 	case ND_REDIRECT:
219 		ICMP6STAT_INC(icp6s_oredirect);
220 		return;
221 	}
222 	ICMP6STAT_INC(icp6s_ounknown);
223 }
224 
225 /*
226  * A wrapper function for icmp6_error() necessary when the erroneous packet
227  * may not contain enough scope zone information.
228  */
229 void
230 icmp6_error2(struct mbuf *m, int type, int code, int param,
231     struct ifnet *ifp)
232 {
233 	struct ip6_hdr *ip6;
234 
235 	if (ifp == NULL)
236 		return;
237 
238 	if (m->m_len < sizeof(struct ip6_hdr)) {
239 		m = m_pullup(m, sizeof(struct ip6_hdr));
240 		if (m == NULL) {
241 			IP6STAT_INC(ip6s_exthdrtoolong);
242 			return;
243 		}
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 	struct epoch_tracker et;
264 	u_int preplen;
265 	int off;
266 	int nxt;
267 
268 	ICMP6STAT_INC(icp6s_error);
269 
270 	/* count per-type-code statistics */
271 	icmp6_errcount(type, code);
272 
273 #ifdef M_DECRYPTED	/*not openbsd*/
274 	if (m->m_flags & M_DECRYPTED) {
275 		ICMP6STAT_INC(icp6s_canterror);
276 		goto freeit;
277 	}
278 #endif
279 
280 	if (m->m_len < sizeof(struct ip6_hdr)) {
281 		m = m_pullup(m, sizeof(struct ip6_hdr));
282 		if (m == NULL) {
283 			IP6STAT_INC(ip6s_exthdrtoolong);
284 			return;
285 		}
286 	}
287 	oip6 = mtod(m, struct ip6_hdr *);
288 
289 	/*
290 	 * If the destination address of the erroneous packet is a multicast
291 	 * address, or the packet was sent using link-layer multicast,
292 	 * we should basically suppress sending an error (RFC 2463, Section
293 	 * 2.4).
294 	 * We have two exceptions (the item e.2 in that section):
295 	 * - the Packet Too Big message can be sent for path MTU discovery.
296 	 * - the Parameter Problem Message that can be allowed an icmp6 error
297 	 *   in the option type field.  This check has been done in
298 	 *   ip6_unknown_opt(), so we can just check the type and code.
299 	 */
300 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
301 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
302 	    (type != ICMP6_PACKET_TOO_BIG &&
303 	     (type != ICMP6_PARAM_PROB ||
304 	      code != ICMP6_PARAMPROB_OPTION)))
305 		goto freeit;
306 
307 	/*
308 	 * RFC 2463, 2.4 (e.5): source address check.
309 	 * XXX: the case of anycast source?
310 	 */
311 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
312 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
313 		goto freeit;
314 
315 	/*
316 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
317 	 * don't do it.
318 	 */
319 	nxt = -1;
320 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
321 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
322 		struct icmp6_hdr *icp;
323 
324 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
325 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
326 			if (m == NULL) {
327 				IP6STAT_INC(ip6s_exthdrtoolong);
328 				return;
329 			}
330 		}
331 		oip6 = mtod(m, struct ip6_hdr *);
332 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
333 
334 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
335 		    icp->icmp6_type == ND_REDIRECT) {
336 			/*
337 			 * ICMPv6 error
338 			 * Special case: for redirect (which is
339 			 * informational) we must not send icmp6 error.
340 			 */
341 			ICMP6STAT_INC(icp6s_canterror);
342 			goto freeit;
343 		} else {
344 			/* ICMPv6 informational - send the error */
345 		}
346 	} else {
347 		/* non-ICMPv6 - send the error */
348 	}
349 
350 	/* Finally, do rate limitation check. */
351 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
352 		ICMP6STAT_INC(icp6s_toofreq);
353 		goto freeit;
354 	}
355 
356 	/*
357 	 * OK, ICMP6 can be generated.
358 	 */
359 
360 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
361 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
362 
363 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
364 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
365 	if (m == NULL) {
366 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
367 		return;
368 	}
369 
370 	nip6 = mtod(m, struct ip6_hdr *);
371 	nip6->ip6_src  = oip6->ip6_src;
372 	nip6->ip6_dst  = oip6->ip6_dst;
373 
374 	in6_clearscope(&oip6->ip6_src);
375 	in6_clearscope(&oip6->ip6_dst);
376 
377 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
378 	icmp6->icmp6_type = type;
379 	icmp6->icmp6_code = code;
380 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
381 
382 	ICMP6STAT_INC(icp6s_outhist[type]);
383 	NET_EPOCH_ENTER(et);
384 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
385 	NET_EPOCH_EXIT(et);
386 
387 	return;
388 
389   freeit:
390 	/*
391 	 * If we can't tell whether or not we can generate ICMP6, free it.
392 	 */
393 	m_freem(m);
394 }
395 
396 /*
397  * Process a received ICMP6 message.
398  */
399 int
400 icmp6_input(struct mbuf **mp, int *offp, int proto)
401 {
402 	struct mbuf *m, *n;
403 	struct ifnet *ifp;
404 	struct ip6_hdr *ip6, *nip6;
405 	struct icmp6_hdr *icmp6, *nicmp6;
406 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
407 	int code, error, icmp6len, ip6len, noff, off, sum;
408 
409 	NET_EPOCH_ASSERT();
410 
411 	m = *mp;
412 	off = *offp;
413 
414 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
415 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
416 		if (m == NULL) {
417 			IP6STAT_INC(ip6s_exthdrtoolong);
418 			*mp = m;
419 			return (IPPROTO_DONE);
420 		}
421 	}
422 
423 	/*
424 	 * Locate icmp6 structure in mbuf, and check
425 	 * that not corrupted and of at least minimum length
426 	 */
427 
428 	icmp6len = m->m_pkthdr.len - off;
429 	if (icmp6len < sizeof(struct icmp6_hdr)) {
430 		ICMP6STAT_INC(icp6s_tooshort);
431 		goto freeit;
432 	}
433 
434 	ip6 = mtod(m, struct ip6_hdr *);
435 	ifp = m->m_pkthdr.rcvif;
436 	/*
437 	 * Check multicast group membership.
438 	 * Note: SSM filters are not applied for ICMPv6 traffic.
439 	 */
440 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
441 		struct in6_multi	*inm;
442 
443 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
444 		if (inm == NULL) {
445 			IP6STAT_INC(ip6s_notmember);
446 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
447 			goto freeit;
448 		}
449 	}
450 
451 	/* Calculate the checksum. */
452 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
453 	code = icmp6->icmp6_code;
454 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
455 		nd6log((LOG_ERR,
456 		    "ICMP6 checksum error(%d|%x) %s\n",
457 		    icmp6->icmp6_type, sum,
458 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
459 		ICMP6STAT_INC(icp6s_checksum);
460 		goto freeit;
461 	}
462 
463 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
464 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
465 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
466 		icmp6_ifstat_inc(ifp, ifs6_in_error);
467 
468 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
469 	switch (icmp6->icmp6_type) {
470 	case ICMP6_DST_UNREACH:
471 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
472 		switch (code) {
473 		case ICMP6_DST_UNREACH_NOROUTE:
474 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
475 			code = PRC_UNREACH_NET;
476 			break;
477 		case ICMP6_DST_UNREACH_ADMIN:
478 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
479 			code = PRC_UNREACH_ADMIN_PROHIB;
480 			break;
481 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
482 			/* I mean "source address was incorrect." */
483 			code = PRC_PARAMPROB;
484 			break;
485 		case ICMP6_DST_UNREACH_NOPORT:
486 			code = PRC_UNREACH_PORT;
487 			break;
488 		default:
489 			goto badcode;
490 		}
491 		goto deliver;
492 		break;
493 
494 	case ICMP6_PACKET_TOO_BIG:
495 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
496 
497 		/* validation is made in icmp6_mtudisc_update */
498 
499 		code = PRC_MSGSIZE;
500 
501 		/*
502 		 * Updating the path MTU will be done after examining
503 		 * intermediate extension headers.
504 		 */
505 		goto deliver;
506 		break;
507 
508 	case ICMP6_TIME_EXCEEDED:
509 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
510 		switch (code) {
511 		case ICMP6_TIME_EXCEED_TRANSIT:
512 			code = PRC_TIMXCEED_INTRANS;
513 			break;
514 		case ICMP6_TIME_EXCEED_REASSEMBLY:
515 			code = PRC_TIMXCEED_REASS;
516 			break;
517 		default:
518 			goto badcode;
519 		}
520 		goto deliver;
521 		break;
522 
523 	case ICMP6_PARAM_PROB:
524 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
525 		switch (code) {
526 		case ICMP6_PARAMPROB_NEXTHEADER:
527 			code = PRC_UNREACH_PROTOCOL;
528 			break;
529 		case ICMP6_PARAMPROB_HEADER:
530 		case ICMP6_PARAMPROB_OPTION:
531 			code = PRC_PARAMPROB;
532 			break;
533 		default:
534 			goto badcode;
535 		}
536 		goto deliver;
537 		break;
538 
539 	case ICMP6_ECHO_REQUEST:
540 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
541 		if (code != 0)
542 			goto badcode;
543 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
544 			/* Give up remote */
545 			break;
546 		}
547 		if (!M_WRITABLE(n)
548 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
549 			struct mbuf *n0 = n;
550 			int n0len;
551 
552 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
553 			n = m_gethdr(M_NOWAIT, n0->m_type);
554 			if (n == NULL) {
555 				/* Give up remote */
556 				m_freem(n0);
557 				break;
558 			}
559 
560 			m_move_pkthdr(n, n0);	/* FIB copied. */
561 			n0len = n0->m_pkthdr.len;	/* save for use below */
562 			/*
563 			 * Copy IPv6 and ICMPv6 only.
564 			 */
565 			nip6 = mtod(n, struct ip6_hdr *);
566 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
567 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
568 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
569 			noff = sizeof(struct ip6_hdr);
570 			/* new mbuf contains only ipv6+icmpv6 headers */
571 			n->m_len = noff + sizeof(struct icmp6_hdr);
572 			/*
573 			 * Adjust mbuf.  ip6_plen will be adjusted in
574 			 * ip6_output().
575 			 */
576 			m_adj(n0, off + sizeof(struct icmp6_hdr));
577 			/* recalculate complete packet size */
578 			n->m_pkthdr.len = n0len + (noff - off);
579 			n->m_next = n0;
580 		} else {
581 			if (n->m_len < off + sizeof(*nicmp6)) {
582 				n = m_pullup(n, off + sizeof(*nicmp6));
583 				if (n == NULL) {
584 					IP6STAT_INC(ip6s_exthdrtoolong);
585 					break;
586 				}
587 			}
588 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
589 			noff = off;
590 		}
591 		if (n) {
592 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
593 			nicmp6->icmp6_code = 0;
594 			ICMP6STAT_INC(icp6s_reflect);
595 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
596 			icmp6_reflect(n, noff);
597 		}
598 		break;
599 
600 	case ICMP6_ECHO_REPLY:
601 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
602 		if (code != 0)
603 			goto badcode;
604 		break;
605 
606 	case MLD_LISTENER_QUERY:
607 	case MLD_LISTENER_REPORT:
608 	case MLD_LISTENER_DONE:
609 	case MLDV2_LISTENER_REPORT:
610 		/*
611 		 * Drop MLD traffic which is not link-local, has a hop limit
612 		 * of greater than 1 hop, or which does not have the
613 		 * IPv6 HBH Router Alert option.
614 		 * As IPv6 HBH options are stripped in ip6_input() we must
615 		 * check an mbuf header flag.
616 		 * XXX Should we also sanity check that these messages
617 		 * were directed to a link-local multicast prefix?
618 		 */
619 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
620 			goto freeit;
621 		if (mld_input(&m, off, icmp6len) != 0) {
622 			*mp = NULL;
623 			return (IPPROTO_DONE);
624 		}
625 		/* m stays. */
626 		break;
627 
628 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
629 	    {
630 		enum { WRU, FQDN } mode;
631 		struct prison *pr;
632 
633 		if (!V_icmp6_nodeinfo)
634 			break;
635 
636 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
637 			mode = WRU;
638 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
639 			mode = FQDN;
640 		else
641 			goto badlen;
642 
643 		pr = NULL;
644 		sx_slock(&allprison_lock);
645 		TAILQ_FOREACH(pr, &allprison, pr_list)
646 			if (pr->pr_vnet == ifp->if_vnet)
647 				break;
648 		sx_sunlock(&allprison_lock);
649 		if (pr == NULL)
650 			pr = curthread->td_ucred->cr_prison;
651 		if (mode == FQDN) {
652 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
653 				m = m_pullup(m, off +
654 				    sizeof(struct icmp6_nodeinfo));
655 				if (m == NULL) {
656 					IP6STAT_INC(ip6s_exthdrtoolong);
657 					*mp = m;
658 					return (IPPROTO_DONE);
659 				}
660 			}
661 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
662 			if (n)
663 				n = ni6_input(n, off, pr);
664 			/* XXX meaningless if n == NULL */
665 			noff = sizeof(struct ip6_hdr);
666 		} else {
667 			u_char *p;
668 			int maxhlen, hlen;
669 
670 			/*
671 			 * XXX: this combination of flags is pointless,
672 			 * but should we keep this for compatibility?
673 			 */
674 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
675 			    ICMP6_NODEINFO_TMPADDROK)) !=
676 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
677 				break;
678 
679 			if (code != 0)
680 				goto badcode;
681 
682 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
683 			n = m_gethdr(M_NOWAIT, m->m_type);
684 			if (n == NULL) {
685 				/* Give up remote */
686 				break;
687 			}
688 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
689 				/*
690 				 * Previous code did a blind M_COPY_PKTHDR
691 				 * and said "just for rcvif".  If true, then
692 				 * we could tolerate the dup failing (due to
693 				 * the deep copy of the tag chain).  For now
694 				 * be conservative and just fail.
695 				 */
696 				m_free(n);
697 				n = NULL;
698 				break;
699 			}
700 			/*
701 			 * Copy IPv6 and ICMPv6 only.
702 			 */
703 			nip6 = mtod(n, struct ip6_hdr *);
704 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
705 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
706 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
707 			p = (u_char *)(nicmp6 + 1);
708 			bzero(p, 4);
709 
710 			maxhlen = M_TRAILINGSPACE(n) -
711 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
712 			mtx_lock(&pr->pr_mtx);
713 			hlen = strlen(pr->pr_hostname);
714 			if (maxhlen > hlen)
715 				maxhlen = hlen;
716 			/* meaningless TTL */
717 			bcopy(pr->pr_hostname, p + 4, maxhlen);
718 			mtx_unlock(&pr->pr_mtx);
719 			noff = sizeof(struct ip6_hdr);
720 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
721 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
722 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
723 			nicmp6->icmp6_code = 0;
724 		}
725 		if (n) {
726 			ICMP6STAT_INC(icp6s_reflect);
727 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
728 			icmp6_reflect(n, noff);
729 		}
730 		break;
731 	    }
732 
733 	case ICMP6_WRUREPLY:
734 		if (code != 0)
735 			goto badcode;
736 		break;
737 
738 	case ND_ROUTER_SOLICIT:
739 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
740 		if (code != 0)
741 			goto badcode;
742 		if (icmp6len < sizeof(struct nd_router_solicit))
743 			goto badlen;
744 		if (send_sendso_input_hook != NULL) {
745 			if (m->m_len < off + icmp6len) {
746 				m = m_pullup(m, off + icmp6len);
747 				if (m == NULL) {
748 					IP6STAT_INC(ip6s_exthdrtoolong);
749 					*mp = NULL;
750 					return (IPPROTO_DONE);
751 				}
752 			}
753 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
754 			if (error == 0) {
755 				m = NULL;
756 				goto freeit;
757 			}
758 		}
759 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
760 		nd6_rs_input(m, off, icmp6len);
761 		m = n;
762 		if (m == NULL)
763 			goto freeit;
764 		break;
765 
766 	case ND_ROUTER_ADVERT:
767 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
768 		if (code != 0)
769 			goto badcode;
770 		if (icmp6len < sizeof(struct nd_router_advert))
771 			goto badlen;
772 		if (send_sendso_input_hook != NULL) {
773 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
774 			if (error == 0) {
775 				m = NULL;
776 				goto freeit;
777 			}
778 		}
779 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
780 		nd6_ra_input(m, off, icmp6len);
781 		m = n;
782 		if (m == NULL)
783 			goto freeit;
784 		break;
785 
786 	case ND_NEIGHBOR_SOLICIT:
787 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
788 		if (code != 0)
789 			goto badcode;
790 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
791 			goto badlen;
792 		if (send_sendso_input_hook != NULL) {
793 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
794 			if (error == 0) {
795 				m = NULL;
796 				goto freeit;
797 			}
798 		}
799 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
800 		nd6_ns_input(m, off, icmp6len);
801 		m = n;
802 		if (m == NULL)
803 			goto freeit;
804 		break;
805 
806 	case ND_NEIGHBOR_ADVERT:
807 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
808 		if (code != 0)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct nd_neighbor_advert))
811 			goto badlen;
812 		if (send_sendso_input_hook != NULL) {
813 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
814 			if (error == 0) {
815 				m = NULL;
816 				goto freeit;
817 			}
818 		}
819 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
820 		nd6_na_input(m, off, icmp6len);
821 		m = n;
822 		if (m == NULL)
823 			goto freeit;
824 		break;
825 
826 	case ND_REDIRECT:
827 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
828 		if (code != 0)
829 			goto badcode;
830 		if (icmp6len < sizeof(struct nd_redirect))
831 			goto badlen;
832 		if (send_sendso_input_hook != NULL) {
833 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
834 			if (error == 0) {
835 				m = NULL;
836 				goto freeit;
837 			}
838 		}
839 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
840 		icmp6_redirect_input(m, off);
841 		m = n;
842 		if (m == NULL)
843 			goto freeit;
844 		break;
845 
846 	case ICMP6_ROUTER_RENUMBERING:
847 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
848 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
849 			goto badcode;
850 		if (icmp6len < sizeof(struct icmp6_router_renum))
851 			goto badlen;
852 		break;
853 
854 	default:
855 		nd6log((LOG_DEBUG,
856 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
857 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
858 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
859 		    ifp ? ifp->if_index : 0));
860 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
861 			/* ICMPv6 error: MUST deliver it by spec... */
862 			code = PRC_NCMDS;
863 			/* deliver */
864 		} else {
865 			/* ICMPv6 informational: MUST not deliver */
866 			break;
867 		}
868 	deliver:
869 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
870 			/* In this case, m should've been freed. */
871 			*mp = NULL;
872 			return (IPPROTO_DONE);
873 		}
874 		break;
875 
876 	badcode:
877 		ICMP6STAT_INC(icp6s_badcode);
878 		break;
879 
880 	badlen:
881 		ICMP6STAT_INC(icp6s_badlen);
882 		break;
883 	}
884 
885 	/* deliver the packet to appropriate sockets */
886 	icmp6_rip6_input(&m, *offp);
887 
888 	*mp = m;
889 	return (IPPROTO_DONE);
890 
891  freeit:
892 	m_freem(m);
893 	*mp = NULL;
894 	return (IPPROTO_DONE);
895 }
896 
897 static int
898 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
899 {
900 	struct mbuf *m;
901 	struct icmp6_hdr *icmp6;
902 	struct ip6_hdr *eip6;
903 	u_int32_t notifymtu;
904 	struct sockaddr_in6 icmp6src, icmp6dst;
905 
906 	m = *mp;
907 
908 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
909 		ICMP6STAT_INC(icp6s_tooshort);
910 		goto freeit;
911 	}
912 
913 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
914 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
915 		if (m == NULL) {
916 			IP6STAT_INC(ip6s_exthdrtoolong);
917 			*mp = m;
918 			return (-1);
919 		}
920 	}
921 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
922 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
923 	bzero(&icmp6dst, sizeof(icmp6dst));
924 
925 	/* Detect the upper level protocol */
926 	{
927 		void (*ctlfunc)(int, struct sockaddr *, void *);
928 		u_int8_t nxt = eip6->ip6_nxt;
929 		int eoff = off + sizeof(struct icmp6_hdr) +
930 		    sizeof(struct ip6_hdr);
931 		struct ip6ctlparam ip6cp;
932 		int icmp6type = icmp6->icmp6_type;
933 		struct ip6_frag *fh;
934 		struct ip6_rthdr *rth;
935 		struct ip6_rthdr0 *rth0;
936 		int rthlen;
937 
938 		while (1) { /* XXX: should avoid infinite loop explicitly? */
939 			struct ip6_ext *eh;
940 
941 			switch (nxt) {
942 			case IPPROTO_HOPOPTS:
943 			case IPPROTO_DSTOPTS:
944 			case IPPROTO_AH:
945 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
946 					m = m_pullup(m, eoff +
947 					    sizeof(struct ip6_ext));
948 					if (m == NULL) {
949 						IP6STAT_INC(ip6s_exthdrtoolong);
950 						*mp = m;
951 						return (-1);
952 					}
953 				}
954 				eh = (struct ip6_ext *)
955 				    (mtod(m, caddr_t) + eoff);
956 				if (nxt == IPPROTO_AH)
957 					eoff += (eh->ip6e_len + 2) << 2;
958 				else
959 					eoff += (eh->ip6e_len + 1) << 3;
960 				nxt = eh->ip6e_nxt;
961 				break;
962 			case IPPROTO_ROUTING:
963 				/*
964 				 * When the erroneous packet contains a
965 				 * routing header, we should examine the
966 				 * header to determine the final destination.
967 				 * Otherwise, we can't properly update
968 				 * information that depends on the final
969 				 * destination (e.g. path MTU).
970 				 */
971 				if (m->m_len < eoff + sizeof(*rth)) {
972 					m = m_pullup(m, eoff + sizeof(*rth));
973 					if (m == NULL) {
974 						IP6STAT_INC(ip6s_exthdrtoolong);
975 						*mp = m;
976 						return (-1);
977 					}
978 				}
979 				rth = (struct ip6_rthdr *)
980 				    (mtod(m, caddr_t) + eoff);
981 				rthlen = (rth->ip6r_len + 1) << 3;
982 				/*
983 				 * XXX: currently there is no
984 				 * officially defined type other
985 				 * than type-0.
986 				 * Note that if the segment left field
987 				 * is 0, all intermediate hops must
988 				 * have been passed.
989 				 */
990 				if (rth->ip6r_segleft &&
991 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
992 					int hops;
993 
994 					if (m->m_len < eoff + rthlen) {
995 						m = m_pullup(m, eoff + rthlen);
996 						if (m == NULL) {
997 							IP6STAT_INC(
998 							    ip6s_exthdrtoolong);
999 							*mp = m;
1000 							return (-1);
1001 						}
1002 					}
1003 					rth0 = (struct ip6_rthdr0 *)
1004 					    (mtod(m, caddr_t) + eoff);
1005 
1006 					/* just ignore a bogus header */
1007 					if ((rth0->ip6r0_len % 2) == 0 &&
1008 					    (hops = rth0->ip6r0_len/2))
1009 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1010 				}
1011 				eoff += rthlen;
1012 				nxt = rth->ip6r_nxt;
1013 				break;
1014 			case IPPROTO_FRAGMENT:
1015 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1016 					m = m_pullup(m, eoff +
1017 					    sizeof(struct ip6_frag));
1018 					if (m == NULL) {
1019 						IP6STAT_INC(ip6s_exthdrtoolong);
1020 						*mp = m;
1021 						return (-1);
1022 					}
1023 				}
1024 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1025 				    eoff);
1026 				/*
1027 				 * Data after a fragment header is meaningless
1028 				 * unless it is the first fragment, but
1029 				 * we'll go to the notify label for path MTU
1030 				 * discovery.
1031 				 */
1032 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1033 					goto notify;
1034 
1035 				eoff += sizeof(struct ip6_frag);
1036 				nxt = fh->ip6f_nxt;
1037 				break;
1038 			default:
1039 				/*
1040 				 * This case includes ESP and the No Next
1041 				 * Header.  In such cases going to the notify
1042 				 * label does not have any meaning
1043 				 * (i.e. ctlfunc will be NULL), but we go
1044 				 * anyway since we might have to update
1045 				 * path MTU information.
1046 				 */
1047 				goto notify;
1048 			}
1049 		}
1050 	  notify:
1051 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
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 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1064 		icmp6dst.sin6_family = AF_INET6;
1065 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1066 			icmp6dst.sin6_addr = eip6->ip6_dst;
1067 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1068 			goto freeit;
1069 		bzero(&icmp6src, sizeof(icmp6src));
1070 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1071 		icmp6src.sin6_family = AF_INET6;
1072 		icmp6src.sin6_addr = eip6->ip6_src;
1073 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1074 			goto freeit;
1075 		icmp6src.sin6_flowinfo =
1076 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1077 
1078 		ip6cp.ip6c_m = m;
1079 		ip6cp.ip6c_icmp6 = icmp6;
1080 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1081 		ip6cp.ip6c_off = eoff;
1082 		ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1083 		ip6cp.ip6c_src = &icmp6src;
1084 		ip6cp.ip6c_nxt = nxt;
1085 
1086 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1087 			notifymtu = ntohl(icmp6->icmp6_mtu);
1088 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1089 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1090 		}
1091 
1092 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1093 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1094 		if (ctlfunc) {
1095 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1096 			    &ip6cp);
1097 		}
1098 	}
1099 	*mp = m;
1100 	return (0);
1101 
1102   freeit:
1103 	m_freem(m);
1104 	*mp = NULL;
1105 	return (-1);
1106 }
1107 
1108 void
1109 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1110 {
1111 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1112 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1113 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1114 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1115 	struct in_conninfo inc;
1116 
1117 #if 0
1118 	/*
1119 	 * RFC2460 section 5, last paragraph.
1120 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1121 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1122 	 * due to packet translator in the middle.
1123 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1124 	 * special handling.
1125 	 */
1126 	if (mtu < IPV6_MMTU)
1127 		return;
1128 #endif
1129 
1130 	/*
1131 	 * we reject ICMPv6 too big with abnormally small value.
1132 	 * XXX what is the good definition of "abnormally small"?
1133 	 */
1134 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1135 		return;
1136 
1137 	if (!validated)
1138 		return;
1139 
1140 	/*
1141 	 * In case the suggested mtu is less than IPV6_MMTU, we
1142 	 * only need to remember that it was for above mentioned
1143 	 * "alwaysfrag" case.
1144 	 * Try to be as close to the spec as possible.
1145 	 */
1146 	if (mtu < IPV6_MMTU)
1147 		mtu = IPV6_MMTU - 8;
1148 
1149 	bzero(&inc, sizeof(inc));
1150 	inc.inc_fibnum = M_GETFIB(m);
1151 	inc.inc_flags |= INC_ISIPV6;
1152 	inc.inc6_faddr = *dst;
1153 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1154 		return;
1155 
1156 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1157 		tcp_hc_updatemtu(&inc, mtu);
1158 		ICMP6STAT_INC(icp6s_pmtuchg);
1159 	}
1160 }
1161 
1162 /*
1163  * Process a Node Information Query packet, based on
1164  * draft-ietf-ipngwg-icmp-name-lookups-07.
1165  *
1166  * Spec incompatibilities:
1167  * - IPv6 Subject address handling
1168  * - IPv4 Subject address handling support missing
1169  * - Proxy reply (answer even if it's not for me)
1170  * - joins NI group address at in6_ifattach() time only, does not cope
1171  *   with hostname changes by sethostname(3)
1172  */
1173 static struct mbuf *
1174 ni6_input(struct mbuf *m, int off, struct prison *pr)
1175 {
1176 	struct icmp6_nodeinfo *ni6, *nni6;
1177 	struct mbuf *n = NULL;
1178 	u_int16_t qtype;
1179 	int subjlen;
1180 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1181 	struct ni_reply_fqdn *fqdn;
1182 	int addrs;		/* for NI_QTYPE_NODEADDR */
1183 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1184 	struct in6_addr in6_subj; /* subject address */
1185 	struct ip6_hdr *ip6;
1186 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1187 	char *subj = NULL;
1188 	struct in6_ifaddr *ia6 = NULL;
1189 
1190 	ip6 = mtod(m, struct ip6_hdr *);
1191 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1192 
1193 	/*
1194 	 * Validate IPv6 source address.
1195 	 * The default configuration MUST be to refuse answering queries from
1196 	 * global-scope addresses according to RFC4602.
1197 	 * Notes:
1198 	 *  - it's not very clear what "refuse" means; this implementation
1199 	 *    simply drops it.
1200 	 *  - it's not very easy to identify global-scope (unicast) addresses
1201 	 *    since there are many prefixes for them.  It should be safer
1202 	 *    and in practice sufficient to check "all" but loopback and
1203 	 *    link-local (note that site-local unicast was deprecated and
1204 	 *    ULA is defined as global scope-wise)
1205 	 */
1206 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1207 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1208 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1209 		goto bad;
1210 
1211 	/*
1212 	 * Validate IPv6 destination address.
1213 	 *
1214 	 * The Responder must discard the Query without further processing
1215 	 * unless it is one of the Responder's unicast or anycast addresses, or
1216 	 * a link-local scope multicast address which the Responder has joined.
1217 	 * [RFC4602, Section 5.]
1218 	 */
1219 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1220 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1221 			goto bad;
1222 		/* else it's a link-local multicast, fine */
1223 	} else {		/* unicast or anycast */
1224 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1225 		if (ia6 == NULL)
1226 			goto bad; /* XXX impossible */
1227 
1228 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1229 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1230 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1231 				"a temporary address in %s:%d",
1232 			       __FILE__, __LINE__));
1233 			goto bad;
1234 		}
1235 	}
1236 
1237 	/* validate query Subject field. */
1238 	qtype = ntohs(ni6->ni_qtype);
1239 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1240 	switch (qtype) {
1241 	case NI_QTYPE_NOOP:
1242 	case NI_QTYPE_SUPTYPES:
1243 		/* 07 draft */
1244 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1245 			break;
1246 		/* FALLTHROUGH */
1247 	case NI_QTYPE_FQDN:
1248 	case NI_QTYPE_NODEADDR:
1249 	case NI_QTYPE_IPV4ADDR:
1250 		switch (ni6->ni_code) {
1251 		case ICMP6_NI_SUBJ_IPV6:
1252 #if ICMP6_NI_SUBJ_IPV6 != 0
1253 		case 0:
1254 #endif
1255 			/*
1256 			 * backward compatibility - try to accept 03 draft
1257 			 * format, where no Subject is present.
1258 			 */
1259 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1260 			    subjlen == 0) {
1261 				oldfqdn++;
1262 				break;
1263 			}
1264 #if ICMP6_NI_SUBJ_IPV6 != 0
1265 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1266 				goto bad;
1267 #endif
1268 
1269 			if (subjlen != sizeof(struct in6_addr))
1270 				goto bad;
1271 
1272 			/*
1273 			 * Validate Subject address.
1274 			 *
1275 			 * Not sure what exactly "address belongs to the node"
1276 			 * means in the spec, is it just unicast, or what?
1277 			 *
1278 			 * At this moment we consider Subject address as
1279 			 * "belong to the node" if the Subject address equals
1280 			 * to the IPv6 destination address; validation for
1281 			 * IPv6 destination address should have done enough
1282 			 * check for us.
1283 			 *
1284 			 * We do not do proxy at this moment.
1285 			 */
1286 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1287 			    subjlen, (caddr_t)&in6_subj);
1288 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1289 				goto bad;
1290 
1291 			subj = (char *)&in6_subj;
1292 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1293 				break;
1294 
1295 			/*
1296 			 * XXX if we are to allow other cases, we should really
1297 			 * be careful about scope here.
1298 			 * basically, we should disallow queries toward IPv6
1299 			 * destination X with subject Y,
1300 			 * if scope(X) > scope(Y).
1301 			 * if we allow scope(X) > scope(Y), it will result in
1302 			 * information leakage across scope boundary.
1303 			 */
1304 			goto bad;
1305 
1306 		case ICMP6_NI_SUBJ_FQDN:
1307 			/*
1308 			 * Validate Subject name with gethostname(3).
1309 			 *
1310 			 * The behavior may need some debate, since:
1311 			 * - we are not sure if the node has FQDN as
1312 			 *   hostname (returned by gethostname(3)).
1313 			 * - the code does wildcard match for truncated names.
1314 			 *   however, we are not sure if we want to perform
1315 			 *   wildcard match, if gethostname(3) side has
1316 			 *   truncated hostname.
1317 			 */
1318 			mtx_lock(&pr->pr_mtx);
1319 			n = ni6_nametodns(pr->pr_hostname,
1320 			    strlen(pr->pr_hostname), 0);
1321 			mtx_unlock(&pr->pr_mtx);
1322 			if (!n || n->m_next || n->m_len == 0)
1323 				goto bad;
1324 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1325 			    subjlen) {
1326 				m = m_pullup(m, off +
1327 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1328 				if (m == NULL) {
1329 					IP6STAT_INC(ip6s_exthdrtoolong);
1330 					goto bad;
1331 				}
1332 			}
1333 			/* ip6 possibly invalid but not used after. */
1334 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1335 			subj = (char *)(mtod(m, caddr_t) + off +
1336 			    sizeof(struct icmp6_nodeinfo));
1337 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1338 			    n->m_len)) {
1339 				goto bad;
1340 			}
1341 			m_freem(n);
1342 			n = NULL;
1343 			break;
1344 
1345 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1346 		default:
1347 			goto bad;
1348 		}
1349 		break;
1350 	}
1351 
1352 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1353 	switch (qtype) {
1354 	case NI_QTYPE_FQDN:
1355 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1356 			goto bad;
1357 		break;
1358 	case NI_QTYPE_NODEADDR:
1359 	case NI_QTYPE_IPV4ADDR:
1360 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1361 			goto bad;
1362 		break;
1363 	}
1364 
1365 	/* guess reply length */
1366 	switch (qtype) {
1367 	case NI_QTYPE_NOOP:
1368 		break;		/* no reply data */
1369 	case NI_QTYPE_SUPTYPES:
1370 		replylen += sizeof(u_int32_t);
1371 		break;
1372 	case NI_QTYPE_FQDN:
1373 		/* XXX will append an mbuf */
1374 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1375 		break;
1376 	case NI_QTYPE_NODEADDR:
1377 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1378 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1379 		    sizeof(u_int32_t))) > MCLBYTES)
1380 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1381 		break;
1382 	case NI_QTYPE_IPV4ADDR:
1383 		/* unsupported - should respond with unknown Qtype? */
1384 		break;
1385 	default:
1386 		/*
1387 		 * XXX: We must return a reply with the ICMP6 code
1388 		 * `unknown Qtype' in this case.  However we regard the case
1389 		 * as an FQDN query for backward compatibility.
1390 		 * Older versions set a random value to this field,
1391 		 * so it rarely varies in the defined qtypes.
1392 		 * But the mechanism is not reliable...
1393 		 * maybe we should obsolete older versions.
1394 		 */
1395 		qtype = NI_QTYPE_FQDN;
1396 		/* XXX will append an mbuf */
1397 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1398 		oldfqdn++;
1399 		break;
1400 	}
1401 
1402 	/* Allocate an mbuf to reply. */
1403 	if (replylen > MCLBYTES) {
1404 		/*
1405 		 * XXX: should we try to allocate more? But MCLBYTES
1406 		 * is probably much larger than IPV6_MMTU...
1407 		 */
1408 		goto bad;
1409 	}
1410 	if (replylen > MHLEN)
1411 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1412 	else
1413 		n = m_gethdr(M_NOWAIT, m->m_type);
1414 	if (n == NULL) {
1415 		m_freem(m);
1416 		return (NULL);
1417 	}
1418 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1419 	n->m_pkthdr.len = n->m_len = replylen;
1420 
1421 	/* copy mbuf header and IPv6 + Node Information base headers */
1422 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1423 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1424 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1425 
1426 	/* qtype dependent procedure */
1427 	switch (qtype) {
1428 	case NI_QTYPE_NOOP:
1429 		nni6->ni_code = ICMP6_NI_SUCCESS;
1430 		nni6->ni_flags = 0;
1431 		break;
1432 	case NI_QTYPE_SUPTYPES:
1433 	{
1434 		u_int32_t v;
1435 		nni6->ni_code = ICMP6_NI_SUCCESS;
1436 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1437 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1438 		v = (u_int32_t)htonl(0x0000000f);
1439 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1440 		break;
1441 	}
1442 	case NI_QTYPE_FQDN:
1443 		nni6->ni_code = ICMP6_NI_SUCCESS;
1444 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1445 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1446 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1447 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1448 		/*
1449 		 * XXX do we really have FQDN in hostname?
1450 		 */
1451 		mtx_lock(&pr->pr_mtx);
1452 		n->m_next = ni6_nametodns(pr->pr_hostname,
1453 		    strlen(pr->pr_hostname), oldfqdn);
1454 		mtx_unlock(&pr->pr_mtx);
1455 		if (n->m_next == NULL)
1456 			goto bad;
1457 		/* XXX we assume that n->m_next is not a chain */
1458 		if (n->m_next->m_next != NULL)
1459 			goto bad;
1460 		n->m_pkthdr.len += n->m_next->m_len;
1461 		break;
1462 	case NI_QTYPE_NODEADDR:
1463 	{
1464 		int lenlim, copied;
1465 
1466 		nni6->ni_code = ICMP6_NI_SUCCESS;
1467 		n->m_pkthdr.len = n->m_len =
1468 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1469 		lenlim = M_TRAILINGSPACE(n);
1470 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1471 		/* XXX: reset mbuf length */
1472 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1473 		    sizeof(struct icmp6_nodeinfo) + copied;
1474 		break;
1475 	}
1476 	default:
1477 		break;		/* XXX impossible! */
1478 	}
1479 
1480 	nni6->ni_type = ICMP6_NI_REPLY;
1481 	m_freem(m);
1482 	return (n);
1483 
1484   bad:
1485 	m_freem(m);
1486 	if (n)
1487 		m_freem(n);
1488 	return (NULL);
1489 }
1490 
1491 /*
1492  * make a mbuf with DNS-encoded string.  no compression support.
1493  *
1494  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1495  * treated as truncated name (two \0 at the end).  this is a wild guess.
1496  *
1497  * old - return pascal string if non-zero
1498  */
1499 static struct mbuf *
1500 ni6_nametodns(const char *name, int namelen, int old)
1501 {
1502 	struct mbuf *m;
1503 	char *cp, *ep;
1504 	const char *p, *q;
1505 	int i, len, nterm;
1506 
1507 	if (old)
1508 		len = namelen + 1;
1509 	else
1510 		len = MCLBYTES;
1511 
1512 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1513 	if (len > MLEN)
1514 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1515 	else
1516 		m = m_get(M_NOWAIT, MT_DATA);
1517 	if (m == NULL)
1518 		goto fail;
1519 
1520 	if (old) {
1521 		m->m_len = len;
1522 		*mtod(m, char *) = namelen;
1523 		bcopy(name, mtod(m, char *) + 1, namelen);
1524 		return m;
1525 	} else {
1526 		m->m_len = 0;
1527 		cp = mtod(m, char *);
1528 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1529 
1530 		/* if not certain about my name, return empty buffer */
1531 		if (namelen == 0)
1532 			return m;
1533 
1534 		/*
1535 		 * guess if it looks like shortened hostname, or FQDN.
1536 		 * shortened hostname needs two trailing "\0".
1537 		 */
1538 		i = 0;
1539 		for (p = name; p < name + namelen; p++) {
1540 			if (*p && *p == '.')
1541 				i++;
1542 		}
1543 		if (i < 2)
1544 			nterm = 2;
1545 		else
1546 			nterm = 1;
1547 
1548 		p = name;
1549 		while (cp < ep && p < name + namelen) {
1550 			i = 0;
1551 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1552 				i++;
1553 			/* result does not fit into mbuf */
1554 			if (cp + i + 1 >= ep)
1555 				goto fail;
1556 			/*
1557 			 * DNS label length restriction, RFC1035 page 8.
1558 			 * "i == 0" case is included here to avoid returning
1559 			 * 0-length label on "foo..bar".
1560 			 */
1561 			if (i <= 0 || i >= 64)
1562 				goto fail;
1563 			*cp++ = i;
1564 			bcopy(p, cp, i);
1565 			cp += i;
1566 			p = q;
1567 			if (p < name + namelen && *p == '.')
1568 				p++;
1569 		}
1570 		/* termination */
1571 		if (cp + nterm >= ep)
1572 			goto fail;
1573 		while (nterm-- > 0)
1574 			*cp++ = '\0';
1575 		m->m_len = cp - mtod(m, char *);
1576 		return m;
1577 	}
1578 
1579 	panic("should not reach here");
1580 	/* NOTREACHED */
1581 
1582  fail:
1583 	if (m)
1584 		m_freem(m);
1585 	return NULL;
1586 }
1587 
1588 /*
1589  * check if two DNS-encoded string matches.  takes care of truncated
1590  * form (with \0\0 at the end).  no compression support.
1591  * XXX upper/lowercase match (see RFC2065)
1592  */
1593 static int
1594 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1595 {
1596 	const char *a0, *b0;
1597 	int l;
1598 
1599 	/* simplest case - need validation? */
1600 	if (alen == blen && bcmp(a, b, alen) == 0)
1601 		return 1;
1602 
1603 	a0 = a;
1604 	b0 = b;
1605 
1606 	/* termination is mandatory */
1607 	if (alen < 2 || blen < 2)
1608 		return 0;
1609 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1610 		return 0;
1611 	alen--;
1612 	blen--;
1613 
1614 	while (a - a0 < alen && b - b0 < blen) {
1615 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1616 			return 0;
1617 
1618 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1619 			return 0;
1620 		/* we don't support compression yet */
1621 		if (a[0] >= 64 || b[0] >= 64)
1622 			return 0;
1623 
1624 		/* truncated case */
1625 		if (a[0] == 0 && a - a0 == alen - 1)
1626 			return 1;
1627 		if (b[0] == 0 && b - b0 == blen - 1)
1628 			return 1;
1629 		if (a[0] == 0 || b[0] == 0)
1630 			return 0;
1631 
1632 		if (a[0] != b[0])
1633 			return 0;
1634 		l = a[0];
1635 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1636 			return 0;
1637 		if (bcmp(a + 1, b + 1, l) != 0)
1638 			return 0;
1639 
1640 		a += 1 + l;
1641 		b += 1 + l;
1642 	}
1643 
1644 	if (a - a0 == alen && b - b0 == blen)
1645 		return 1;
1646 	else
1647 		return 0;
1648 }
1649 
1650 /*
1651  * calculate the number of addresses to be returned in the node info reply.
1652  */
1653 static int
1654 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1655     struct in6_addr *subj)
1656 {
1657 	struct ifnet *ifp;
1658 	struct in6_ifaddr *ifa6;
1659 	struct ifaddr *ifa;
1660 	int addrs = 0, addrsofif, iffound = 0;
1661 	int niflags = ni6->ni_flags;
1662 
1663 	NET_EPOCH_ASSERT();
1664 
1665 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1666 		switch (ni6->ni_code) {
1667 		case ICMP6_NI_SUBJ_IPV6:
1668 			if (subj == NULL) /* must be impossible... */
1669 				return (0);
1670 			break;
1671 		default:
1672 			/*
1673 			 * XXX: we only support IPv6 subject address for
1674 			 * this Qtype.
1675 			 */
1676 			return (0);
1677 		}
1678 	}
1679 
1680 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1681 		addrsofif = 0;
1682 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1683 			if (ifa->ifa_addr->sa_family != AF_INET6)
1684 				continue;
1685 			ifa6 = (struct in6_ifaddr *)ifa;
1686 
1687 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1688 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1689 				iffound = 1;
1690 
1691 			/*
1692 			 * IPv4-mapped addresses can only be returned by a
1693 			 * Node Information proxy, since they represent
1694 			 * addresses of IPv4-only nodes, which perforce do
1695 			 * not implement this protocol.
1696 			 * [icmp-name-lookups-07, Section 5.4]
1697 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1698 			 * this function at this moment.
1699 			 */
1700 
1701 			/* What do we have to do about ::1? */
1702 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1703 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1704 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1705 					continue;
1706 				break;
1707 			case IPV6_ADDR_SCOPE_SITELOCAL:
1708 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1709 					continue;
1710 				break;
1711 			case IPV6_ADDR_SCOPE_GLOBAL:
1712 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1713 					continue;
1714 				break;
1715 			default:
1716 				continue;
1717 			}
1718 
1719 			/*
1720 			 * check if anycast is okay.
1721 			 * XXX: just experimental.  not in the spec.
1722 			 */
1723 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1724 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1725 				continue; /* we need only unicast addresses */
1726 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1727 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1728 				continue;
1729 			}
1730 			addrsofif++; /* count the address */
1731 		}
1732 		if (iffound) {
1733 			*ifpp = ifp;
1734 			return (addrsofif);
1735 		}
1736 
1737 		addrs += addrsofif;
1738 	}
1739 
1740 	return (addrs);
1741 }
1742 
1743 static int
1744 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1745     struct ifnet *ifp0, int resid)
1746 {
1747 	struct ifnet *ifp;
1748 	struct in6_ifaddr *ifa6;
1749 	struct ifaddr *ifa;
1750 	struct ifnet *ifp_dep = NULL;
1751 	int copied = 0, allow_deprecated = 0;
1752 	u_char *cp = (u_char *)(nni6 + 1);
1753 	int niflags = ni6->ni_flags;
1754 	u_int32_t ltime;
1755 
1756 	NET_EPOCH_ASSERT();
1757 
1758 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1759 		return (0);	/* needless to copy */
1760 
1761 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1762   again:
1763 
1764 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1765 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1766 			if (ifa->ifa_addr->sa_family != AF_INET6)
1767 				continue;
1768 			ifa6 = (struct in6_ifaddr *)ifa;
1769 
1770 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1771 			    allow_deprecated == 0) {
1772 				/*
1773 				 * prefererred address should be put before
1774 				 * deprecated addresses.
1775 				 */
1776 
1777 				/* record the interface for later search */
1778 				if (ifp_dep == NULL)
1779 					ifp_dep = ifp;
1780 
1781 				continue;
1782 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1783 			    allow_deprecated != 0)
1784 				continue; /* we now collect deprecated addrs */
1785 
1786 			/* What do we have to do about ::1? */
1787 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1788 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1789 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1790 					continue;
1791 				break;
1792 			case IPV6_ADDR_SCOPE_SITELOCAL:
1793 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1794 					continue;
1795 				break;
1796 			case IPV6_ADDR_SCOPE_GLOBAL:
1797 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1798 					continue;
1799 				break;
1800 			default:
1801 				continue;
1802 			}
1803 
1804 			/*
1805 			 * check if anycast is okay.
1806 			 * XXX: just experimental.  not in the spec.
1807 			 */
1808 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1809 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1810 				continue;
1811 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1812 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1813 				continue;
1814 			}
1815 
1816 			/* now we can copy the address */
1817 			if (resid < sizeof(struct in6_addr) +
1818 			    sizeof(u_int32_t)) {
1819 				/*
1820 				 * We give up much more copy.
1821 				 * Set the truncate flag and return.
1822 				 */
1823 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1824 				return (copied);
1825 			}
1826 
1827 			/*
1828 			 * Set the TTL of the address.
1829 			 * The TTL value should be one of the following
1830 			 * according to the specification:
1831 			 *
1832 			 * 1. The remaining lifetime of a DHCP lease on the
1833 			 *    address, or
1834 			 * 2. The remaining Valid Lifetime of a prefix from
1835 			 *    which the address was derived through Stateless
1836 			 *    Autoconfiguration.
1837 			 *
1838 			 * Note that we currently do not support stateful
1839 			 * address configuration by DHCPv6, so the former
1840 			 * case can't happen.
1841 			 */
1842 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1843 				ltime = ND6_INFINITE_LIFETIME;
1844 			else {
1845 				if (ifa6->ia6_lifetime.ia6t_expire >
1846 				    time_uptime)
1847 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1848 				else
1849 					ltime = 0;
1850 			}
1851 
1852 			bcopy(&ltime, cp, sizeof(u_int32_t));
1853 			cp += sizeof(u_int32_t);
1854 
1855 			/* copy the address itself */
1856 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1857 			    sizeof(struct in6_addr));
1858 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1859 			cp += sizeof(struct in6_addr);
1860 
1861 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1862 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1863 		}
1864 		if (ifp0)	/* we need search only on the specified IF */
1865 			break;
1866 	}
1867 
1868 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1869 		ifp = ifp_dep;
1870 		allow_deprecated = 1;
1871 
1872 		goto again;
1873 	}
1874 
1875 	return (copied);
1876 }
1877 
1878 /*
1879  * XXX almost dup'ed code with rip6_input.
1880  */
1881 static int
1882 icmp6_rip6_input(struct mbuf **mp, int off)
1883 {
1884 	struct mbuf *m = *mp;
1885 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1886 	struct inpcb *inp;
1887 	struct inpcb *last = NULL;
1888 	struct sockaddr_in6 fromsa;
1889 	struct icmp6_hdr *icmp6;
1890 	struct mbuf *opts = NULL;
1891 
1892 	NET_EPOCH_ASSERT();
1893 
1894 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1895 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1896 
1897 	/*
1898 	 * XXX: the address may have embedded scope zone ID, which should be
1899 	 * hidden from applications.
1900 	 */
1901 	bzero(&fromsa, sizeof(fromsa));
1902 	fromsa.sin6_family = AF_INET6;
1903 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1904 	fromsa.sin6_addr = ip6->ip6_src;
1905 	if (sa6_recoverscope(&fromsa)) {
1906 		m_freem(m);
1907 		*mp = NULL;
1908 		return (IPPROTO_DONE);
1909 	}
1910 
1911 	CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
1912 		if ((inp->inp_vflag & INP_IPV6) == 0)
1913 			continue;
1914 		if (inp->inp_ip_p != IPPROTO_ICMPV6)
1915 			continue;
1916 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1917 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1918 			continue;
1919 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1920 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1921 			continue;
1922 		INP_RLOCK(inp);
1923 		if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1924 			INP_RUNLOCK(inp);
1925 			continue;
1926 		}
1927 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1928 		    inp->in6p_icmp6filt)) {
1929 			INP_RUNLOCK(inp);
1930 			continue;
1931 		}
1932 		if (last != NULL) {
1933 			struct	mbuf *n = NULL;
1934 
1935 			/*
1936 			 * Recent network drivers tend to allocate a single
1937 			 * mbuf cluster, rather than to make a couple of
1938 			 * mbufs without clusters.  Also, since the IPv6 code
1939 			 * path tries to avoid m_pullup(), it is highly
1940 			 * probable that we still have an mbuf cluster here
1941 			 * even though the necessary length can be stored in an
1942 			 * mbuf's internal buffer.
1943 			 * Meanwhile, the default size of the receive socket
1944 			 * buffer for raw sockets is not so large.  This means
1945 			 * the possibility of packet loss is relatively higher
1946 			 * than before.  To avoid this scenario, we copy the
1947 			 * received data to a separate mbuf that does not use
1948 			 * a cluster, if possible.
1949 			 * XXX: it is better to copy the data after stripping
1950 			 * intermediate headers.
1951 			 */
1952 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1953 			    m->m_len <= MHLEN) {
1954 				n = m_get(M_NOWAIT, m->m_type);
1955 				if (n != NULL) {
1956 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1957 						bcopy(m->m_data, n->m_data,
1958 						      m->m_len);
1959 						n->m_len = m->m_len;
1960 					} else {
1961 						m_free(n);
1962 						n = NULL;
1963 					}
1964 				}
1965 			}
1966 			if (n != NULL ||
1967 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1968 				if (last->inp_flags & INP_CONTROLOPTS)
1969 					ip6_savecontrol(last, n, &opts);
1970 				/* strip intermediate headers */
1971 				m_adj(n, off);
1972 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1973 				if (sbappendaddr_locked(
1974 				    &last->inp_socket->so_rcv,
1975 				    (struct sockaddr *)&fromsa, n, opts)
1976 				    == 0) {
1977 					/* should notify about lost packet */
1978 					m_freem(n);
1979 					if (opts) {
1980 						m_freem(opts);
1981 					}
1982 					SOCKBUF_UNLOCK(
1983 					    &last->inp_socket->so_rcv);
1984 				} else
1985 					sorwakeup_locked(last->inp_socket);
1986 				opts = NULL;
1987 			}
1988 			INP_RUNLOCK(last);
1989 		}
1990 		last = inp;
1991 	}
1992 	if (last != NULL) {
1993 		if (last->inp_flags & INP_CONTROLOPTS)
1994 			ip6_savecontrol(last, m, &opts);
1995 		/* strip intermediate headers */
1996 		m_adj(m, off);
1997 
1998 		/* avoid using mbuf clusters if possible (see above) */
1999 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2000 		    m->m_len <= MHLEN) {
2001 			struct mbuf *n;
2002 
2003 			n = m_get(M_NOWAIT, m->m_type);
2004 			if (n != NULL) {
2005 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2006 					bcopy(m->m_data, n->m_data, m->m_len);
2007 					n->m_len = m->m_len;
2008 
2009 					m_freem(m);
2010 					m = n;
2011 				} else {
2012 					m_freem(n);
2013 					n = NULL;
2014 				}
2015 			}
2016 		}
2017 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2018 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2019 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2020 			m_freem(m);
2021 			if (opts)
2022 				m_freem(opts);
2023 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2024 		} else
2025 			sorwakeup_locked(last->inp_socket);
2026 		INP_RUNLOCK(last);
2027 	} else {
2028 		m_freem(m);
2029 		IP6STAT_DEC(ip6s_delivered);
2030 	}
2031 	*mp = NULL;
2032 	return (IPPROTO_DONE);
2033 }
2034 
2035 /*
2036  * Reflect the ip6 packet back to the source.
2037  * OFF points to the icmp6 header, counted from the top of the mbuf.
2038  */
2039 static void
2040 icmp6_reflect(struct mbuf *m, size_t off)
2041 {
2042 	struct in6_addr src6, *srcp;
2043 	struct ip6_hdr *ip6;
2044 	struct icmp6_hdr *icmp6;
2045 	struct in6_ifaddr *ia = NULL;
2046 	struct ifnet *outif = NULL;
2047 	int plen;
2048 	int type, code, hlim;
2049 
2050 	/* too short to reflect */
2051 	if (off < sizeof(struct ip6_hdr)) {
2052 		nd6log((LOG_DEBUG,
2053 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2054 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2055 		    __FILE__, __LINE__));
2056 		goto bad;
2057 	}
2058 
2059 	/*
2060 	 * If there are extra headers between IPv6 and ICMPv6, strip
2061 	 * off that header first.
2062 	 */
2063 #ifdef DIAGNOSTIC
2064 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2065 		panic("assumption failed in icmp6_reflect");
2066 #endif
2067 	if (off > sizeof(struct ip6_hdr)) {
2068 		size_t l;
2069 		struct ip6_hdr nip6;
2070 
2071 		l = off - sizeof(struct ip6_hdr);
2072 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2073 		m_adj(m, l);
2074 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2075 		if (m->m_len < l) {
2076 			if ((m = m_pullup(m, l)) == NULL)
2077 				return;
2078 		}
2079 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2080 	} else /* off == sizeof(struct ip6_hdr) */ {
2081 		size_t l;
2082 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2083 		if (m->m_len < l) {
2084 			if ((m = m_pullup(m, l)) == NULL)
2085 				return;
2086 		}
2087 	}
2088 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2089 	ip6 = mtod(m, struct ip6_hdr *);
2090 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2091 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2092 	type = icmp6->icmp6_type; /* keep type for statistics */
2093 	code = icmp6->icmp6_code; /* ditto. */
2094 	hlim = 0;
2095 	srcp = NULL;
2096 
2097 	/*
2098 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2099 	 * use dst as the src for the reply.
2100 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2101 	 * (for example) when we encounter an error while forwarding procedure
2102 	 * destined to a duplicated address of ours.
2103 	 */
2104 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2105 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2106 		if (ia != NULL && !(ia->ia6_flags &
2107 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2108 			src6 = ia->ia_addr.sin6_addr;
2109 			srcp = &src6;
2110 
2111 			if (m->m_pkthdr.rcvif != NULL) {
2112 				/* XXX: This may not be the outgoing interface */
2113 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2114 			} else
2115 				hlim = V_ip6_defhlim;
2116 		}
2117 	}
2118 
2119 	if (srcp == NULL) {
2120 		int error;
2121 		struct in6_addr dst6;
2122 		uint32_t scopeid;
2123 
2124 		/*
2125 		 * This case matches to multicasts, our anycast, or unicasts
2126 		 * that we do not own.  Select a source address based on the
2127 		 * source address of the erroneous packet.
2128 		 */
2129 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2130 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2131 		    scopeid, NULL, &src6, &hlim);
2132 
2133 		if (error) {
2134 			char ip6buf[INET6_ADDRSTRLEN];
2135 			nd6log((LOG_DEBUG,
2136 			    "icmp6_reflect: source can't be determined: "
2137 			    "dst=%s, error=%d\n",
2138 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2139 			goto bad;
2140 		}
2141 		srcp = &src6;
2142 	}
2143 	/*
2144 	 * ip6_input() drops a packet if its src is multicast.
2145 	 * So, the src is never multicast.
2146 	 */
2147 	ip6->ip6_dst = ip6->ip6_src;
2148 	ip6->ip6_src = *srcp;
2149 	ip6->ip6_flow = 0;
2150 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2151 	ip6->ip6_vfc |= IPV6_VERSION;
2152 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2153 	ip6->ip6_hlim = hlim;
2154 
2155 	icmp6->icmp6_cksum = 0;
2156 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2157 	    sizeof(struct ip6_hdr), plen);
2158 
2159 	/*
2160 	 * XXX option handling
2161 	 */
2162 
2163 	m->m_flags &= ~(M_BCAST|M_MCAST);
2164 	m->m_pkthdr.rcvif = NULL;
2165 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2166 	if (outif)
2167 		icmp6_ifoutstat_inc(outif, type, code);
2168 
2169 	return;
2170 
2171  bad:
2172 	m_freem(m);
2173 	return;
2174 }
2175 
2176 void
2177 icmp6_fasttimo(void)
2178 {
2179 
2180 	mld_fasttimo();
2181 }
2182 
2183 void
2184 icmp6_slowtimo(void)
2185 {
2186 
2187 	mld_slowtimo();
2188 }
2189 
2190 static const char *
2191 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2192     struct in6_addr *tgt6)
2193 {
2194 	static char buf[1024];
2195 	char ip6bufs[INET6_ADDRSTRLEN];
2196 	char ip6bufd[INET6_ADDRSTRLEN];
2197 	char ip6buft[INET6_ADDRSTRLEN];
2198 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2199 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2200 	    ip6_sprintf(ip6buft, tgt6));
2201 	return buf;
2202 }
2203 
2204 void
2205 icmp6_redirect_input(struct mbuf *m, int off)
2206 {
2207 	struct ifnet *ifp;
2208 	struct ip6_hdr *ip6;
2209 	struct nd_redirect *nd_rd;
2210 	struct in6_addr src6, redtgt6, reddst6;
2211 	union nd_opts ndopts;
2212 	char ip6buf[INET6_ADDRSTRLEN];
2213 	char *lladdr;
2214 	int icmp6len, is_onlink, is_router, lladdrlen;
2215 
2216 	M_ASSERTPKTHDR(m);
2217 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2218 
2219 	/* XXX if we are router, we don't update route by icmp6 redirect */
2220 	if (V_ip6_forwarding)
2221 		goto freeit;
2222 	if (!V_icmp6_rediraccept)
2223 		goto freeit;
2224 
2225 	/* RFC 6980: Nodes MUST silently ignore fragments */
2226 	if(m->m_flags & M_FRAGMENTED)
2227 		goto freeit;
2228 
2229 	ip6 = mtod(m, struct ip6_hdr *);
2230 	icmp6len = ntohs(ip6->ip6_plen);
2231 	if (m->m_len < off + icmp6len) {
2232 		m = m_pullup(m, off + icmp6len);
2233 		if (m == NULL) {
2234 			IP6STAT_INC(ip6s_exthdrtoolong);
2235 			return;
2236 		}
2237 	}
2238 	ip6 = mtod(m, struct ip6_hdr *);
2239 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2240 
2241 	ifp = m->m_pkthdr.rcvif;
2242 	redtgt6 = nd_rd->nd_rd_target;
2243 	reddst6 = nd_rd->nd_rd_dst;
2244 
2245 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2246 	    in6_setscope(&reddst6, ifp, NULL)) {
2247 		goto freeit;
2248 	}
2249 
2250 	/* validation */
2251 	src6 = ip6->ip6_src;
2252 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2253 		nd6log((LOG_ERR,
2254 		    "ICMP6 redirect sent from %s rejected; "
2255 		    "must be from linklocal\n",
2256 		    ip6_sprintf(ip6buf, &src6)));
2257 		goto bad;
2258 	}
2259 	if (__predict_false(ip6->ip6_hlim != 255)) {
2260 		ICMP6STAT_INC(icp6s_invlhlim);
2261 		nd6log((LOG_ERR,
2262 		    "ICMP6 redirect sent from %s rejected; "
2263 		    "hlim=%d (must be 255)\n",
2264 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2265 		goto bad;
2266 	}
2267     {
2268 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2269 	struct nhop_object *nh;
2270 	struct in6_addr kdst;
2271 	uint32_t scopeid;
2272 
2273 	in6_splitscope(&reddst6, &kdst, &scopeid);
2274 	NET_EPOCH_ASSERT();
2275 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2276 	if (nh != NULL) {
2277 		struct in6_addr nh_addr;
2278 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2279 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2280 			nd6log((LOG_ERR,
2281 			    "ICMP6 redirect rejected; no route "
2282 			    "with inet6 gateway found for redirect dst: %s\n",
2283 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 			goto bad;
2285 		}
2286 
2287 		/*
2288 		 * Embed scope zone id into next hop address.
2289 		 */
2290 		nh_addr = nh->gw6_sa.sin6_addr;
2291 
2292 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2293 			nd6log((LOG_ERR,
2294 			    "ICMP6 redirect rejected; "
2295 			    "not equal to gw-for-src=%s (must be same): "
2296 			    "%s\n",
2297 			    ip6_sprintf(ip6buf, &nh_addr),
2298 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299 			goto bad;
2300 		}
2301 	} else {
2302 		nd6log((LOG_ERR,
2303 		    "ICMP6 redirect rejected; "
2304 		    "no route found for redirect dst: %s\n",
2305 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2306 		goto bad;
2307 	}
2308     }
2309 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2310 		nd6log((LOG_ERR,
2311 		    "ICMP6 redirect rejected; "
2312 		    "redirect dst must be unicast: %s\n",
2313 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2314 		goto bad;
2315 	}
2316 
2317 	is_router = is_onlink = 0;
2318 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2319 		is_router = 1;	/* router case */
2320 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2321 		is_onlink = 1;	/* on-link destination case */
2322 	if (!is_router && !is_onlink) {
2323 		nd6log((LOG_ERR,
2324 		    "ICMP6 redirect rejected; "
2325 		    "neither router case nor onlink case: %s\n",
2326 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2327 		goto bad;
2328 	}
2329 
2330 	icmp6len -= sizeof(*nd_rd);
2331 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2332 	if (nd6_options(&ndopts) < 0) {
2333 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2334 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2335 		/* nd6_options have incremented stats */
2336 		goto freeit;
2337 	}
2338 
2339 	lladdr = NULL;
2340 	lladdrlen = 0;
2341 	if (ndopts.nd_opts_tgt_lladdr) {
2342 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2343 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2344 	}
2345 
2346 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2347 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2348 		    "(if %d, icmp6 packet %d): %s\n",
2349 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2350 		    ifp->if_addrlen, lladdrlen - 2,
2351 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2352 		goto bad;
2353 	}
2354 
2355 	/* Validation passed. */
2356 
2357 	/* RFC 2461 8.3 */
2358 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2359 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2360 
2361 	/*
2362 	 * Install a gateway route in the better-router case or an interface
2363 	 * route in the on-link-destination case.
2364 	 */
2365 	{
2366 		struct sockaddr_in6 sdst;
2367 		struct sockaddr_in6 sgw;
2368 		struct sockaddr_in6 ssrc;
2369 		struct sockaddr *gw;
2370 		int rt_flags;
2371 		u_int fibnum;
2372 
2373 		bzero(&sdst, sizeof(sdst));
2374 		bzero(&ssrc, sizeof(ssrc));
2375 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2376 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2377 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2378 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2379 		rt_flags = 0;
2380 		if (is_router) {
2381 			bzero(&sgw, sizeof(sgw));
2382 			sgw.sin6_family = AF_INET6;
2383 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2384 			bcopy(&redtgt6, &sgw.sin6_addr,
2385 				sizeof(struct in6_addr));
2386 			gw = (struct sockaddr *)&sgw;
2387 			rt_flags |= RTF_GATEWAY;
2388 		} else
2389 			gw = ifp->if_addr->ifa_addr;
2390 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2391 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2392 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2393 			    V_icmp6_redirtimeout);
2394 	}
2395 	/* finally update cached route in each socket via pfctlinput */
2396     {
2397 	struct sockaddr_in6 sdst;
2398 
2399 	bzero(&sdst, sizeof(sdst));
2400 	sdst.sin6_family = AF_INET6;
2401 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2402 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2403 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2404     }
2405 
2406  freeit:
2407 	m_freem(m);
2408 	return;
2409 
2410  bad:
2411 	ICMP6STAT_INC(icp6s_badredirect);
2412 	m_freem(m);
2413 }
2414 
2415 void
2416 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2417 {
2418 	struct ifnet *ifp;	/* my outgoing interface */
2419 	struct in6_addr *ifp_ll6;
2420 	struct in6_addr *router_ll6;
2421 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2422 	struct mbuf *m = NULL;	/* newly allocated one */
2423 	struct m_tag *mtag;
2424 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2425 	struct nd_redirect *nd_rd;
2426 	struct llentry *ln = NULL;
2427 	size_t maxlen;
2428 	u_char *p;
2429 	struct ifnet *outif = NULL;
2430 	struct sockaddr_in6 src_sa;
2431 
2432 	icmp6_errcount(ND_REDIRECT, 0);
2433 
2434 	/* if we are not router, we don't send icmp6 redirect */
2435 	if (!V_ip6_forwarding)
2436 		goto fail;
2437 
2438 	/* sanity check */
2439 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2440 		goto fail;
2441 
2442 	/*
2443 	 * Address check:
2444 	 *  the source address must identify a neighbor, and
2445 	 *  the destination address must not be a multicast address
2446 	 *  [RFC 2461, sec 8.2]
2447 	 */
2448 	sip6 = mtod(m0, struct ip6_hdr *);
2449 	bzero(&src_sa, sizeof(src_sa));
2450 	src_sa.sin6_family = AF_INET6;
2451 	src_sa.sin6_len = sizeof(src_sa);
2452 	src_sa.sin6_addr = sip6->ip6_src;
2453 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2454 		goto fail;
2455 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2456 		goto fail;	/* what should we do here? */
2457 
2458 	/* rate limit */
2459 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2460 		goto fail;
2461 
2462 	/*
2463 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2464 	 * we almost always ask for an mbuf cluster for simplicity.
2465 	 * (MHLEN < IPV6_MMTU is almost always true)
2466 	 */
2467 #if IPV6_MMTU >= MCLBYTES
2468 # error assumption failed about IPV6_MMTU and MCLBYTES
2469 #endif
2470 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2471 	if (m == NULL)
2472 		goto fail;
2473 	M_SETFIB(m, M_GETFIB(m0));
2474 	maxlen = M_TRAILINGSPACE(m);
2475 	maxlen = min(IPV6_MMTU, maxlen);
2476 	/* just for safety */
2477 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2478 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2479 		goto fail;
2480 	}
2481 
2482 	{
2483 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2484 		struct in6_ifaddr *ia;
2485 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2486 						 IN6_IFF_NOTREADY|
2487 						 IN6_IFF_ANYCAST)) == NULL)
2488 			goto fail;
2489 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2490 		/* XXXRW: reference released prematurely. */
2491 		ifa_free(&ia->ia_ifa);
2492 	}
2493 
2494 	/* get ip6 linklocal address for the router. */
2495 	if (nh->nh_flags & NHF_GATEWAY) {
2496 		struct sockaddr_in6 *sin6;
2497 		sin6 = &nh->gw6_sa;
2498 		router_ll6 = &sin6->sin6_addr;
2499 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2500 			router_ll6 = (struct in6_addr *)NULL;
2501 	} else
2502 		router_ll6 = (struct in6_addr *)NULL;
2503 
2504 	/* ip6 */
2505 	ip6 = mtod(m, struct ip6_hdr *);
2506 	ip6->ip6_flow = 0;
2507 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2508 	ip6->ip6_vfc |= IPV6_VERSION;
2509 	/* ip6->ip6_plen will be set later */
2510 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2511 	ip6->ip6_hlim = 255;
2512 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2513 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2514 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2515 
2516 	/* ND Redirect */
2517 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2518 	nd_rd->nd_rd_type = ND_REDIRECT;
2519 	nd_rd->nd_rd_code = 0;
2520 	nd_rd->nd_rd_reserved = 0;
2521 	if (nh->nh_flags & NHF_GATEWAY) {
2522 		/*
2523 		 * nd_rd->nd_rd_target must be a link-local address in
2524 		 * better router cases.
2525 		 */
2526 		if (!router_ll6)
2527 			goto fail;
2528 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2529 		    sizeof(nd_rd->nd_rd_target));
2530 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2531 		    sizeof(nd_rd->nd_rd_dst));
2532 	} else {
2533 		/* make sure redtgt == reddst */
2534 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2535 		    sizeof(nd_rd->nd_rd_target));
2536 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2537 		    sizeof(nd_rd->nd_rd_dst));
2538 	}
2539 
2540 	p = (u_char *)(nd_rd + 1);
2541 
2542 	if (!router_ll6)
2543 		goto nolladdropt;
2544 
2545 	{
2546 		/* target lladdr option */
2547 		int len;
2548 		struct nd_opt_hdr *nd_opt;
2549 		char *lladdr;
2550 
2551 		ln = nd6_lookup(router_ll6, 0, ifp);
2552 		if (ln == NULL)
2553 			goto nolladdropt;
2554 
2555 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2556 		len = (len + 7) & ~7;	/* round by 8 */
2557 		/* safety check */
2558 		if (len + (p - (u_char *)ip6) > maxlen)
2559 			goto nolladdropt;
2560 
2561 		if (ln->la_flags & LLE_VALID) {
2562 			nd_opt = (struct nd_opt_hdr *)p;
2563 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2564 			nd_opt->nd_opt_len = len >> 3;
2565 			lladdr = (char *)(nd_opt + 1);
2566 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2567 			p += len;
2568 		}
2569 	}
2570 nolladdropt:
2571 	if (ln != NULL)
2572 		LLE_RUNLOCK(ln);
2573 
2574 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2575 
2576 	/* just to be safe */
2577 #ifdef M_DECRYPTED	/*not openbsd*/
2578 	if (m0->m_flags & M_DECRYPTED)
2579 		goto noredhdropt;
2580 #endif
2581 	if (p - (u_char *)ip6 > maxlen)
2582 		goto noredhdropt;
2583 
2584 	{
2585 		/* redirected header option */
2586 		int len;
2587 		struct nd_opt_rd_hdr *nd_opt_rh;
2588 
2589 		/*
2590 		 * compute the maximum size for icmp6 redirect header option.
2591 		 * XXX room for auth header?
2592 		 */
2593 		len = maxlen - (p - (u_char *)ip6);
2594 		len &= ~7;
2595 
2596 		/* This is just for simplicity. */
2597 		if (m0->m_pkthdr.len != m0->m_len) {
2598 			if (m0->m_next) {
2599 				m_freem(m0->m_next);
2600 				m0->m_next = NULL;
2601 			}
2602 			m0->m_pkthdr.len = m0->m_len;
2603 		}
2604 
2605 		/*
2606 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2607 		 * about padding/truncate rule for the original IP packet.
2608 		 * From the discussion on IPv6imp in Feb 1999,
2609 		 * the consensus was:
2610 		 * - "attach as much as possible" is the goal
2611 		 * - pad if not aligned (original size can be guessed by
2612 		 *   original ip6 header)
2613 		 * Following code adds the padding if it is simple enough,
2614 		 * and truncates if not.
2615 		 */
2616 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2617 			panic("assumption failed in %s:%d", __FILE__,
2618 			    __LINE__);
2619 
2620 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2621 			/* not enough room, truncate */
2622 			m0->m_pkthdr.len = m0->m_len = len -
2623 			    sizeof(*nd_opt_rh);
2624 		} else {
2625 			/* enough room, pad or truncate */
2626 			size_t extra;
2627 
2628 			extra = m0->m_pkthdr.len % 8;
2629 			if (extra) {
2630 				/* pad if easy enough, truncate if not */
2631 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2632 					/* pad */
2633 					m0->m_len += (8 - extra);
2634 					m0->m_pkthdr.len += (8 - extra);
2635 				} else {
2636 					/* truncate */
2637 					m0->m_pkthdr.len -= extra;
2638 					m0->m_len -= extra;
2639 				}
2640 			}
2641 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2642 			m0->m_pkthdr.len = m0->m_len = len -
2643 			    sizeof(*nd_opt_rh);
2644 		}
2645 
2646 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2647 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2648 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2649 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2650 		p += sizeof(*nd_opt_rh);
2651 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2652 
2653 		/* connect m0 to m */
2654 		m_tag_delete_chain(m0, NULL);
2655 		m0->m_flags &= ~M_PKTHDR;
2656 		m->m_next = m0;
2657 		m->m_pkthdr.len = m->m_len + m0->m_len;
2658 		m0 = NULL;
2659 	}
2660 noredhdropt:;
2661 	if (m0) {
2662 		m_freem(m0);
2663 		m0 = NULL;
2664 	}
2665 
2666 	/* XXX: clear embedded link IDs in the inner header */
2667 	in6_clearscope(&sip6->ip6_src);
2668 	in6_clearscope(&sip6->ip6_dst);
2669 	in6_clearscope(&nd_rd->nd_rd_target);
2670 	in6_clearscope(&nd_rd->nd_rd_dst);
2671 
2672 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2673 
2674 	nd_rd->nd_rd_cksum = 0;
2675 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2676 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2677 
2678         if (send_sendso_input_hook != NULL) {
2679 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2680 			M_NOWAIT);
2681 		if (mtag == NULL)
2682 			goto fail;
2683 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2684 		m_tag_prepend(m, mtag);
2685 	}
2686 
2687 	/* send the packet to outside... */
2688 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2689 	if (outif) {
2690 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2691 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2692 	}
2693 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2694 
2695 	return;
2696 
2697 fail:
2698 	if (m)
2699 		m_freem(m);
2700 	if (m0)
2701 		m_freem(m0);
2702 }
2703 
2704 /*
2705  * ICMPv6 socket option processing.
2706  */
2707 int
2708 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2709 {
2710 	int error = 0;
2711 	int optlen;
2712 	struct inpcb *inp = sotoinpcb(so);
2713 	int level, op, optname;
2714 
2715 	if (sopt) {
2716 		level = sopt->sopt_level;
2717 		op = sopt->sopt_dir;
2718 		optname = sopt->sopt_name;
2719 		optlen = sopt->sopt_valsize;
2720 	} else
2721 		level = op = optname = optlen = 0;
2722 
2723 	if (level != IPPROTO_ICMPV6) {
2724 		return EINVAL;
2725 	}
2726 
2727 	switch (op) {
2728 	case PRCO_SETOPT:
2729 		switch (optname) {
2730 		case ICMP6_FILTER:
2731 		    {
2732 			struct icmp6_filter ic6f;
2733 
2734 			if (optlen != sizeof(ic6f)) {
2735 				error = EMSGSIZE;
2736 				break;
2737 			}
2738 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2739 			if (error == 0) {
2740 				INP_WLOCK(inp);
2741 				*inp->in6p_icmp6filt = ic6f;
2742 				INP_WUNLOCK(inp);
2743 			}
2744 			break;
2745 		    }
2746 
2747 		default:
2748 			error = ENOPROTOOPT;
2749 			break;
2750 		}
2751 		break;
2752 
2753 	case PRCO_GETOPT:
2754 		switch (optname) {
2755 		case ICMP6_FILTER:
2756 		    {
2757 			struct icmp6_filter ic6f;
2758 
2759 			INP_RLOCK(inp);
2760 			ic6f = *inp->in6p_icmp6filt;
2761 			INP_RUNLOCK(inp);
2762 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2763 			break;
2764 		    }
2765 
2766 		default:
2767 			error = ENOPROTOOPT;
2768 			break;
2769 		}
2770 		break;
2771 	}
2772 
2773 	return (error);
2774 }
2775 
2776 /*
2777  * Perform rate limit check.
2778  * Returns 0 if it is okay to send the icmp6 packet.
2779  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2780  * limitation.
2781  *
2782  * XXX per-destination/type check necessary?
2783  *
2784  * dst - not used at this moment
2785  * type - not used at this moment
2786  * code - not used at this moment
2787  */
2788 static int
2789 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2790     const int code)
2791 {
2792 	int ret;
2793 
2794 	ret = 0;	/* okay to send */
2795 
2796 	/* PPS limit */
2797 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2798 	    V_icmp6errppslim)) {
2799 		/* The packet is subject to rate limit */
2800 		ret++;
2801 	}
2802 
2803 	return ret;
2804 }
2805