xref: /freebsd/sys/netinet6/icmp6.c (revision be181ee2a28aa2b4b0e76684bce9f673ef668874)
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/ip6_var.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/mld6_var.h>
113 #include <netinet6/nd6.h>
114 #include <netinet6/send.h>
115 
116 extern ip6proto_ctlinput_t	*ip6_ctlprotox[];
117 
118 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
119 VNET_PCPUSTAT_SYSINIT(icmp6stat);
120 
121 #ifdef VIMAGE
122 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
123 #endif /* VIMAGE */
124 
125 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
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_icmp6errppslim		VNET(icmp6errppslim)
133 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
134 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
135 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
136 
137 static void icmp6_errcount(int, int);
138 static int icmp6_rip6_input(struct mbuf **, int);
139 static void icmp6_reflect(struct mbuf *, size_t);
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);
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 	if (m->m_len < sizeof(struct ip6_hdr)) {
236 		m = m_pullup(m, sizeof(struct ip6_hdr));
237 		if (m == NULL) {
238 			IP6STAT_INC(ip6s_exthdrtoolong);
239 			return;
240 		}
241 	}
242 	ip6 = mtod(m, struct ip6_hdr *);
243 
244 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
245 		return;
246 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
247 		return;
248 
249 	icmp6_error(m, type, code, param);
250 }
251 
252 /*
253  * Generate an error packet of type error in response to bad IP6 packet.
254  */
255 void
256 icmp6_error(struct mbuf *m, int type, int code, int param)
257 {
258 	struct ip6_hdr *oip6, *nip6;
259 	struct icmp6_hdr *icmp6;
260 	struct epoch_tracker et;
261 	u_int preplen;
262 	int off;
263 	int nxt;
264 
265 	ICMP6STAT_INC(icp6s_error);
266 
267 	/* count per-type-code statistics */
268 	icmp6_errcount(type, code);
269 
270 #ifdef M_DECRYPTED	/*not openbsd*/
271 	if (m->m_flags & M_DECRYPTED) {
272 		ICMP6STAT_INC(icp6s_canterror);
273 		goto freeit;
274 	}
275 #endif
276 
277 	if (m->m_len < sizeof(struct ip6_hdr)) {
278 		m = m_pullup(m, sizeof(struct ip6_hdr));
279 		if (m == NULL) {
280 			IP6STAT_INC(ip6s_exthdrtoolong);
281 			return;
282 		}
283 	}
284 	oip6 = mtod(m, struct ip6_hdr *);
285 
286 	/*
287 	 * If the destination address of the erroneous packet is a multicast
288 	 * address, or the packet was sent using link-layer multicast,
289 	 * we should basically suppress sending an error (RFC 2463, Section
290 	 * 2.4).
291 	 * We have two exceptions (the item e.2 in that section):
292 	 * - the Packet Too Big message can be sent for path MTU discovery.
293 	 * - the Parameter Problem Message that can be allowed an icmp6 error
294 	 *   in the option type field.  This check has been done in
295 	 *   ip6_unknown_opt(), so we can just check the type and code.
296 	 */
297 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
298 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
299 	    (type != ICMP6_PACKET_TOO_BIG &&
300 	     (type != ICMP6_PARAM_PROB ||
301 	      code != ICMP6_PARAMPROB_OPTION)))
302 		goto freeit;
303 
304 	/*
305 	 * RFC 2463, 2.4 (e.5): source address check.
306 	 * XXX: the case of anycast source?
307 	 */
308 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
309 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
310 		goto freeit;
311 
312 	/*
313 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
314 	 * don't do it.
315 	 */
316 	nxt = -1;
317 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
318 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
319 		struct icmp6_hdr *icp;
320 
321 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
322 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
323 			if (m == NULL) {
324 				IP6STAT_INC(ip6s_exthdrtoolong);
325 				return;
326 			}
327 		}
328 		oip6 = mtod(m, struct ip6_hdr *);
329 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
330 
331 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
332 		    icp->icmp6_type == ND_REDIRECT) {
333 			/*
334 			 * ICMPv6 error
335 			 * Special case: for redirect (which is
336 			 * informational) we must not send icmp6 error.
337 			 */
338 			ICMP6STAT_INC(icp6s_canterror);
339 			goto freeit;
340 		} else {
341 			/* ICMPv6 informational - send the error */
342 		}
343 	} else {
344 		/* non-ICMPv6 - send the error */
345 	}
346 
347 	/* Finally, do rate limitation check. */
348 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
349 		ICMP6STAT_INC(icp6s_toofreq);
350 		goto freeit;
351 	}
352 
353 	/*
354 	 * OK, ICMP6 can be generated.
355 	 */
356 
357 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
358 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
359 
360 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
361 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
362 	if (m == NULL) {
363 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
364 		return;
365 	}
366 
367 	nip6 = mtod(m, struct ip6_hdr *);
368 	nip6->ip6_src  = oip6->ip6_src;
369 	nip6->ip6_dst  = oip6->ip6_dst;
370 
371 	in6_clearscope(&oip6->ip6_src);
372 	in6_clearscope(&oip6->ip6_dst);
373 
374 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
375 	icmp6->icmp6_type = type;
376 	icmp6->icmp6_code = code;
377 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
378 
379 	ICMP6STAT_INC(icp6s_outhist[type]);
380 	NET_EPOCH_ENTER(et);
381 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
382 	NET_EPOCH_EXIT(et);
383 
384 	return;
385 
386   freeit:
387 	/*
388 	 * If we can't tell whether or not we can generate ICMP6, free it.
389 	 */
390 	m_freem(m);
391 }
392 
393 int
394 icmp6_errmap(const struct icmp6_hdr *icmp6)
395 {
396 
397 	switch (icmp6->icmp6_type) {
398 	case ICMP6_DST_UNREACH:
399 		switch (icmp6->icmp6_code) {
400 		case ICMP6_DST_UNREACH_NOROUTE:
401 		case ICMP6_DST_UNREACH_ADDR:
402 			return (EHOSTUNREACH);
403 		case ICMP6_DST_UNREACH_NOPORT:
404 		case ICMP6_DST_UNREACH_ADMIN:
405 			return (ECONNREFUSED);
406 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
407 			return (ENOPROTOOPT);
408 		default:
409 			return (0);	/* Shouldn't happen. */
410 		}
411 	case ICMP6_PACKET_TOO_BIG:
412 		return (EMSGSIZE);
413 	case ICMP6_TIME_EXCEEDED:
414 		switch (icmp6->icmp6_code) {
415 		case ICMP6_TIME_EXCEED_TRANSIT:
416 			return (EHOSTUNREACH);
417 		case ICMP6_TIME_EXCEED_REASSEMBLY:
418 			return (0);
419 		default:
420 			return (0);	/* Shouldn't happen. */
421 		}
422 	case ICMP6_PARAM_PROB:
423 		switch (icmp6->icmp6_code) {
424 		case ICMP6_PARAMPROB_NEXTHEADER:
425 			return (ECONNREFUSED);
426 		case ICMP6_PARAMPROB_HEADER:
427 		case ICMP6_PARAMPROB_OPTION:
428 			return (ENOPROTOOPT);
429 		default:
430 			return (0);	/* Shouldn't happen. */
431 		}
432 	default:
433 		return (0);
434 	}
435 }
436 
437 /*
438  * Process a received ICMP6 message.
439  */
440 int
441 icmp6_input(struct mbuf **mp, int *offp, int proto)
442 {
443 	struct mbuf *m, *n;
444 	struct ifnet *ifp;
445 	struct ip6_hdr *ip6, *nip6;
446 	struct icmp6_hdr *icmp6, *nicmp6;
447 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
448 	int code, error, icmp6len, ip6len, noff, off, sum;
449 
450 	NET_EPOCH_ASSERT();
451 
452 	m = *mp;
453 	off = *offp;
454 
455 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
456 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
457 		if (m == NULL) {
458 			IP6STAT_INC(ip6s_exthdrtoolong);
459 			*mp = m;
460 			return (IPPROTO_DONE);
461 		}
462 	}
463 
464 	/*
465 	 * Locate icmp6 structure in mbuf, and check
466 	 * that not corrupted and of at least minimum length
467 	 */
468 
469 	icmp6len = m->m_pkthdr.len - off;
470 	if (icmp6len < sizeof(struct icmp6_hdr)) {
471 		ICMP6STAT_INC(icp6s_tooshort);
472 		goto freeit;
473 	}
474 
475 	ip6 = mtod(m, struct ip6_hdr *);
476 	ifp = m->m_pkthdr.rcvif;
477 	/*
478 	 * Check multicast group membership.
479 	 * Note: SSM filters are not applied for ICMPv6 traffic.
480 	 */
481 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
482 		struct in6_multi	*inm;
483 
484 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
485 		if (inm == NULL) {
486 			IP6STAT_INC(ip6s_notmember);
487 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
488 			goto freeit;
489 		}
490 	}
491 
492 	/* Calculate the checksum. */
493 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
494 	code = icmp6->icmp6_code;
495 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
496 		nd6log((LOG_ERR,
497 		    "ICMP6 checksum error(%d|%x) %s\n",
498 		    icmp6->icmp6_type, sum,
499 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
500 		ICMP6STAT_INC(icp6s_checksum);
501 		goto freeit;
502 	}
503 
504 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
505 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
506 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
507 		icmp6_ifstat_inc(ifp, ifs6_in_error);
508 
509 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
510 	switch (icmp6->icmp6_type) {
511 	case ICMP6_DST_UNREACH:
512 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
513 		switch (code) {
514 		case ICMP6_DST_UNREACH_ADMIN:
515 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
516 		case ICMP6_DST_UNREACH_NOROUTE:
517 		case ICMP6_DST_UNREACH_ADDR:
518 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
519 		case ICMP6_DST_UNREACH_NOPORT:
520 			goto deliver;
521 		default:
522 			goto badcode;
523 		}
524 	case ICMP6_PACKET_TOO_BIG:
525 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
526 		/*
527 		 * Validation is made in icmp6_mtudisc_update.
528 		 * Updating the path MTU will be done after examining
529 		 * intermediate extension headers.
530 		 */
531 		goto deliver;
532 	case ICMP6_TIME_EXCEEDED:
533 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
534 		switch (code) {
535 		case ICMP6_TIME_EXCEED_TRANSIT:
536 		case ICMP6_TIME_EXCEED_REASSEMBLY:
537 			goto deliver;
538 		default:
539 			goto badcode;
540 		}
541 	case ICMP6_PARAM_PROB:
542 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
543 		switch (code) {
544 		case ICMP6_PARAMPROB_NEXTHEADER:
545 		case ICMP6_PARAMPROB_HEADER:
546 		case ICMP6_PARAMPROB_OPTION:
547 			goto deliver;
548 		default:
549 			goto badcode;
550 		}
551 	case ICMP6_ECHO_REQUEST:
552 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
553 		if (code != 0)
554 			goto badcode;
555 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
556 			/* Give up remote */
557 			break;
558 		}
559 		if (!M_WRITABLE(n)
560 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
561 			struct mbuf *n0 = n;
562 			int n0len;
563 
564 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
565 			n = m_gethdr(M_NOWAIT, n0->m_type);
566 			if (n == NULL) {
567 				/* Give up remote */
568 				m_freem(n0);
569 				break;
570 			}
571 
572 			m_move_pkthdr(n, n0);	/* FIB copied. */
573 			n0len = n0->m_pkthdr.len;	/* save for use below */
574 			/*
575 			 * Copy IPv6 and ICMPv6 only.
576 			 */
577 			nip6 = mtod(n, struct ip6_hdr *);
578 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
579 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
580 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
581 			noff = sizeof(struct ip6_hdr);
582 			/* new mbuf contains only ipv6+icmpv6 headers */
583 			n->m_len = noff + sizeof(struct icmp6_hdr);
584 			/*
585 			 * Adjust mbuf.  ip6_plen will be adjusted in
586 			 * ip6_output().
587 			 */
588 			m_adj(n0, off + sizeof(struct icmp6_hdr));
589 			/* recalculate complete packet size */
590 			n->m_pkthdr.len = n0len + (noff - off);
591 			n->m_next = n0;
592 		} else {
593 			if (n->m_len < off + sizeof(*nicmp6)) {
594 				n = m_pullup(n, off + sizeof(*nicmp6));
595 				if (n == NULL) {
596 					IP6STAT_INC(ip6s_exthdrtoolong);
597 					break;
598 				}
599 			}
600 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
601 			noff = off;
602 		}
603 		if (n) {
604 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
605 			nicmp6->icmp6_code = 0;
606 			ICMP6STAT_INC(icp6s_reflect);
607 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
608 			icmp6_reflect(n, noff);
609 		}
610 		break;
611 
612 	case ICMP6_ECHO_REPLY:
613 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
614 		if (code != 0)
615 			goto badcode;
616 		break;
617 
618 	case MLD_LISTENER_QUERY:
619 	case MLD_LISTENER_REPORT:
620 	case MLD_LISTENER_DONE:
621 	case MLDV2_LISTENER_REPORT:
622 		/*
623 		 * Drop MLD traffic which is not link-local, has a hop limit
624 		 * of greater than 1 hop, or which does not have the
625 		 * IPv6 HBH Router Alert option.
626 		 * As IPv6 HBH options are stripped in ip6_input() we must
627 		 * check an mbuf header flag.
628 		 * XXX Should we also sanity check that these messages
629 		 * were directed to a link-local multicast prefix?
630 		 */
631 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
632 			goto freeit;
633 		if (mld_input(&m, off, icmp6len) != 0) {
634 			*mp = NULL;
635 			return (IPPROTO_DONE);
636 		}
637 		/* m stays. */
638 		break;
639 
640 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
641 	    {
642 		enum { WRU, FQDN } mode;
643 		struct prison *pr;
644 
645 		if (!V_icmp6_nodeinfo)
646 			break;
647 
648 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
649 			mode = WRU;
650 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
651 			mode = FQDN;
652 		else
653 			goto badlen;
654 
655 		pr = NULL;
656 		sx_slock(&allprison_lock);
657 		TAILQ_FOREACH(pr, &allprison, pr_list)
658 			if (pr->pr_vnet == ifp->if_vnet)
659 				break;
660 		sx_sunlock(&allprison_lock);
661 		if (pr == NULL)
662 			pr = curthread->td_ucred->cr_prison;
663 		if (mode == FQDN) {
664 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
665 				m = m_pullup(m, off +
666 				    sizeof(struct icmp6_nodeinfo));
667 				if (m == NULL) {
668 					IP6STAT_INC(ip6s_exthdrtoolong);
669 					*mp = m;
670 					return (IPPROTO_DONE);
671 				}
672 			}
673 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
674 			if (n)
675 				n = ni6_input(n, off, pr);
676 			/* XXX meaningless if n == NULL */
677 			noff = sizeof(struct ip6_hdr);
678 		} else {
679 			u_char *p;
680 			int maxhlen, hlen;
681 
682 			/*
683 			 * XXX: this combination of flags is pointless,
684 			 * but should we keep this for compatibility?
685 			 */
686 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
687 			    ICMP6_NODEINFO_TMPADDROK)) !=
688 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
689 				break;
690 
691 			if (code != 0)
692 				goto badcode;
693 
694 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
695 			n = m_gethdr(M_NOWAIT, m->m_type);
696 			if (n == NULL) {
697 				/* Give up remote */
698 				break;
699 			}
700 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
701 				/*
702 				 * Previous code did a blind M_COPY_PKTHDR
703 				 * and said "just for rcvif".  If true, then
704 				 * we could tolerate the dup failing (due to
705 				 * the deep copy of the tag chain).  For now
706 				 * be conservative and just fail.
707 				 */
708 				m_free(n);
709 				n = NULL;
710 				break;
711 			}
712 			/*
713 			 * Copy IPv6 and ICMPv6 only.
714 			 */
715 			nip6 = mtod(n, struct ip6_hdr *);
716 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
717 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
718 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
719 			p = (u_char *)(nicmp6 + 1);
720 			bzero(p, 4);
721 
722 			maxhlen = M_TRAILINGSPACE(n) -
723 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
724 			mtx_lock(&pr->pr_mtx);
725 			hlen = strlen(pr->pr_hostname);
726 			if (maxhlen > hlen)
727 				maxhlen = hlen;
728 			/* meaningless TTL */
729 			bcopy(pr->pr_hostname, p + 4, maxhlen);
730 			mtx_unlock(&pr->pr_mtx);
731 			noff = sizeof(struct ip6_hdr);
732 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
733 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
734 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
735 			nicmp6->icmp6_code = 0;
736 		}
737 		if (n) {
738 			ICMP6STAT_INC(icp6s_reflect);
739 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
740 			icmp6_reflect(n, noff);
741 		}
742 		break;
743 	    }
744 
745 	case ICMP6_WRUREPLY:
746 		if (code != 0)
747 			goto badcode;
748 		break;
749 
750 	case ND_ROUTER_SOLICIT:
751 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
752 		if (code != 0)
753 			goto badcode;
754 		if (icmp6len < sizeof(struct nd_router_solicit))
755 			goto badlen;
756 		if (send_sendso_input_hook != NULL) {
757 			if (m->m_len < off + icmp6len) {
758 				m = m_pullup(m, off + icmp6len);
759 				if (m == NULL) {
760 					IP6STAT_INC(ip6s_exthdrtoolong);
761 					*mp = NULL;
762 					return (IPPROTO_DONE);
763 				}
764 			}
765 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
766 			if (error == 0) {
767 				m = NULL;
768 				goto freeit;
769 			}
770 		}
771 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
772 		nd6_rs_input(m, off, icmp6len);
773 		m = n;
774 		if (m == NULL)
775 			goto freeit;
776 		break;
777 
778 	case ND_ROUTER_ADVERT:
779 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
780 		if (code != 0)
781 			goto badcode;
782 		if (icmp6len < sizeof(struct nd_router_advert))
783 			goto badlen;
784 		if (send_sendso_input_hook != NULL) {
785 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
786 			if (error == 0) {
787 				m = NULL;
788 				goto freeit;
789 			}
790 		}
791 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
792 		nd6_ra_input(m, off, icmp6len);
793 		m = n;
794 		if (m == NULL)
795 			goto freeit;
796 		break;
797 
798 	case ND_NEIGHBOR_SOLICIT:
799 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
800 		if (code != 0)
801 			goto badcode;
802 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
803 			goto badlen;
804 		if (send_sendso_input_hook != NULL) {
805 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
806 			if (error == 0) {
807 				m = NULL;
808 				goto freeit;
809 			}
810 		}
811 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
812 		nd6_ns_input(m, off, icmp6len);
813 		m = n;
814 		if (m == NULL)
815 			goto freeit;
816 		break;
817 
818 	case ND_NEIGHBOR_ADVERT:
819 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
820 		if (code != 0)
821 			goto badcode;
822 		if (icmp6len < sizeof(struct nd_neighbor_advert))
823 			goto badlen;
824 		if (send_sendso_input_hook != NULL) {
825 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
826 			if (error == 0) {
827 				m = NULL;
828 				goto freeit;
829 			}
830 		}
831 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
832 		nd6_na_input(m, off, icmp6len);
833 		m = n;
834 		if (m == NULL)
835 			goto freeit;
836 		break;
837 
838 	case ND_REDIRECT:
839 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
840 		if (code != 0)
841 			goto badcode;
842 		if (icmp6len < sizeof(struct nd_redirect))
843 			goto badlen;
844 		if (send_sendso_input_hook != NULL) {
845 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
846 			if (error == 0) {
847 				m = NULL;
848 				goto freeit;
849 			}
850 		}
851 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
852 		icmp6_redirect_input(m, off);
853 		m = n;
854 		if (m == NULL)
855 			goto freeit;
856 		break;
857 
858 	case ICMP6_ROUTER_RENUMBERING:
859 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
860 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
861 			goto badcode;
862 		if (icmp6len < sizeof(struct icmp6_router_renum))
863 			goto badlen;
864 		break;
865 
866 	default:
867 		nd6log((LOG_DEBUG,
868 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
869 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
870 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
871 		    ifp ? ifp->if_index : 0));
872 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
873 			/* ICMPv6 error: MUST deliver it by spec... */
874 			goto deliver;
875 		} else {
876 			/* ICMPv6 informational: MUST not deliver */
877 			break;
878 		}
879 	deliver:
880 		if (icmp6_notify_error(&m, off, icmp6len) != 0) {
881 			/* In this case, m should've been freed. */
882 			*mp = NULL;
883 			return (IPPROTO_DONE);
884 		}
885 		break;
886 
887 	badcode:
888 		ICMP6STAT_INC(icp6s_badcode);
889 		break;
890 
891 	badlen:
892 		ICMP6STAT_INC(icp6s_badlen);
893 		break;
894 	}
895 
896 	/* deliver the packet to appropriate sockets */
897 	icmp6_rip6_input(&m, *offp);
898 
899 	*mp = m;
900 	return (IPPROTO_DONE);
901 
902  freeit:
903 	m_freem(m);
904 	*mp = NULL;
905 	return (IPPROTO_DONE);
906 }
907 
908 static int
909 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len)
910 {
911 	struct mbuf *m;
912 	struct icmp6_hdr *icmp6;
913 	struct ip6_hdr *eip6;
914 	u_int32_t notifymtu;
915 	struct sockaddr_in6 icmp6src, icmp6dst;
916 
917 	m = *mp;
918 
919 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
920 		ICMP6STAT_INC(icp6s_tooshort);
921 		goto freeit;
922 	}
923 
924 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
925 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
926 		if (m == NULL) {
927 			IP6STAT_INC(ip6s_exthdrtoolong);
928 			*mp = m;
929 			return (-1);
930 		}
931 	}
932 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
933 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
934 	bzero(&icmp6dst, sizeof(icmp6dst));
935 
936 	/* Detect the upper level protocol */
937 	{
938 		u_int8_t nxt = eip6->ip6_nxt;
939 		int eoff = off + sizeof(struct icmp6_hdr) +
940 		    sizeof(struct ip6_hdr);
941 		struct ip6ctlparam ip6cp;
942 		int icmp6type = icmp6->icmp6_type;
943 		struct ip6_frag *fh;
944 		struct ip6_rthdr *rth;
945 		struct ip6_rthdr0 *rth0;
946 		int rthlen;
947 
948 		while (1) { /* XXX: should avoid infinite loop explicitly? */
949 			struct ip6_ext *eh;
950 
951 			switch (nxt) {
952 			case IPPROTO_HOPOPTS:
953 			case IPPROTO_DSTOPTS:
954 			case IPPROTO_AH:
955 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
956 					m = m_pullup(m, eoff +
957 					    sizeof(struct ip6_ext));
958 					if (m == NULL) {
959 						IP6STAT_INC(ip6s_exthdrtoolong);
960 						*mp = m;
961 						return (-1);
962 					}
963 				}
964 				eh = (struct ip6_ext *)
965 				    (mtod(m, caddr_t) + eoff);
966 				if (nxt == IPPROTO_AH)
967 					eoff += (eh->ip6e_len + 2) << 2;
968 				else
969 					eoff += (eh->ip6e_len + 1) << 3;
970 				nxt = eh->ip6e_nxt;
971 				break;
972 			case IPPROTO_ROUTING:
973 				/*
974 				 * When the erroneous packet contains a
975 				 * routing header, we should examine the
976 				 * header to determine the final destination.
977 				 * Otherwise, we can't properly update
978 				 * information that depends on the final
979 				 * destination (e.g. path MTU).
980 				 */
981 				if (m->m_len < eoff + sizeof(*rth)) {
982 					m = m_pullup(m, eoff + sizeof(*rth));
983 					if (m == NULL) {
984 						IP6STAT_INC(ip6s_exthdrtoolong);
985 						*mp = m;
986 						return (-1);
987 					}
988 				}
989 				rth = (struct ip6_rthdr *)
990 				    (mtod(m, caddr_t) + eoff);
991 				rthlen = (rth->ip6r_len + 1) << 3;
992 				/*
993 				 * XXX: currently there is no
994 				 * officially defined type other
995 				 * than type-0.
996 				 * Note that if the segment left field
997 				 * is 0, all intermediate hops must
998 				 * have been passed.
999 				 */
1000 				if (rth->ip6r_segleft &&
1001 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1002 					int hops;
1003 
1004 					if (m->m_len < eoff + rthlen) {
1005 						m = m_pullup(m, eoff + rthlen);
1006 						if (m == NULL) {
1007 							IP6STAT_INC(
1008 							    ip6s_exthdrtoolong);
1009 							*mp = m;
1010 							return (-1);
1011 						}
1012 					}
1013 					rth0 = (struct ip6_rthdr0 *)
1014 					    (mtod(m, caddr_t) + eoff);
1015 
1016 					/* just ignore a bogus header */
1017 					if ((rth0->ip6r0_len % 2) == 0 &&
1018 					    (hops = rth0->ip6r0_len/2))
1019 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1020 				}
1021 				eoff += rthlen;
1022 				nxt = rth->ip6r_nxt;
1023 				break;
1024 			case IPPROTO_FRAGMENT:
1025 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1026 					m = m_pullup(m, eoff +
1027 					    sizeof(struct ip6_frag));
1028 					if (m == NULL) {
1029 						IP6STAT_INC(ip6s_exthdrtoolong);
1030 						*mp = m;
1031 						return (-1);
1032 					}
1033 				}
1034 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1035 				    eoff);
1036 				/*
1037 				 * Data after a fragment header is meaningless
1038 				 * unless it is the first fragment, but
1039 				 * we'll go to the notify label for path MTU
1040 				 * discovery.
1041 				 */
1042 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1043 					goto notify;
1044 
1045 				eoff += sizeof(struct ip6_frag);
1046 				nxt = fh->ip6f_nxt;
1047 				break;
1048 			default:
1049 				/*
1050 				 * This case includes ESP and the No Next
1051 				 * Header.  In such cases going to the notify
1052 				 * label does not have any meaning
1053 				 * (i.e. ctlfunc will be NULL), but we go
1054 				 * anyway since we might have to update
1055 				 * path MTU information.
1056 				 */
1057 				goto notify;
1058 			}
1059 		}
1060 	  notify:
1061 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1062 
1063 		/*
1064 		 * retrieve parameters from the inner IPv6 header, and convert
1065 		 * them into sockaddr structures.
1066 		 * XXX: there is no guarantee that the source or destination
1067 		 * addresses of the inner packet are in the same scope as
1068 		 * the addresses of the icmp packet.  But there is no other
1069 		 * way to determine the zone.
1070 		 */
1071 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1072 
1073 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1074 		icmp6dst.sin6_family = AF_INET6;
1075 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1076 			icmp6dst.sin6_addr = eip6->ip6_dst;
1077 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1078 			goto freeit;
1079 		bzero(&icmp6src, sizeof(icmp6src));
1080 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1081 		icmp6src.sin6_family = AF_INET6;
1082 		icmp6src.sin6_addr = eip6->ip6_src;
1083 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1084 			goto freeit;
1085 		icmp6src.sin6_flowinfo =
1086 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1087 
1088 		ip6cp.ip6c_m = m;
1089 		ip6cp.ip6c_icmp6 = icmp6;
1090 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1091 		ip6cp.ip6c_off = eoff;
1092 		ip6cp.ip6c_finaldst = &icmp6dst;
1093 		ip6cp.ip6c_src = &icmp6src;
1094 		ip6cp.ip6c_nxt = nxt;
1095 
1096 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1097 			notifymtu = ntohl(icmp6->icmp6_mtu);
1098 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1099 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1100 		}
1101 
1102 		if (ip6_ctlprotox[nxt] != NULL)
1103 			ip6_ctlprotox[nxt](&ip6cp);
1104 	}
1105 	*mp = m;
1106 	return (0);
1107 
1108   freeit:
1109 	m_freem(m);
1110 	*mp = NULL;
1111 	return (-1);
1112 }
1113 
1114 void
1115 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1116 {
1117 	struct in6_addr *dst = &ip6cp->ip6c_finaldst->sin6_addr;
1118 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1119 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1120 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1121 	struct in_conninfo inc;
1122 	uint32_t max_mtu;
1123 
1124 #if 0
1125 	/*
1126 	 * RFC2460 section 5, last paragraph.
1127 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1128 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1129 	 * due to packet translator in the middle.
1130 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1131 	 * special handling.
1132 	 */
1133 	if (mtu < IPV6_MMTU)
1134 		return;
1135 #endif
1136 
1137 	/*
1138 	 * we reject ICMPv6 too big with abnormally small value.
1139 	 * XXX what is the good definition of "abnormally small"?
1140 	 */
1141 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1142 		return;
1143 
1144 	if (!validated)
1145 		return;
1146 
1147 	/*
1148 	 * In case the suggested mtu is less than IPV6_MMTU, we
1149 	 * only need to remember that it was for above mentioned
1150 	 * "alwaysfrag" case.
1151 	 * Try to be as close to the spec as possible.
1152 	 */
1153 	if (mtu < IPV6_MMTU)
1154 		mtu = IPV6_MMTU - 8;
1155 
1156 	bzero(&inc, sizeof(inc));
1157 	inc.inc_fibnum = M_GETFIB(m);
1158 	inc.inc_flags |= INC_ISIPV6;
1159 	inc.inc6_faddr = *dst;
1160 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1161 		return;
1162 
1163 	max_mtu = tcp_hc_getmtu(&inc);
1164 	if (max_mtu == 0)
1165 		max_mtu = tcp_maxmtu6(&inc, NULL);
1166 
1167 	if (mtu < max_mtu) {
1168 		tcp_hc_updatemtu(&inc, mtu);
1169 		ICMP6STAT_INC(icp6s_pmtuchg);
1170 	}
1171 }
1172 
1173 /*
1174  * Process a Node Information Query packet, based on
1175  * draft-ietf-ipngwg-icmp-name-lookups-07.
1176  *
1177  * Spec incompatibilities:
1178  * - IPv6 Subject address handling
1179  * - IPv4 Subject address handling support missing
1180  * - Proxy reply (answer even if it's not for me)
1181  * - joins NI group address at in6_ifattach() time only, does not cope
1182  *   with hostname changes by sethostname(3)
1183  */
1184 static struct mbuf *
1185 ni6_input(struct mbuf *m, int off, struct prison *pr)
1186 {
1187 	struct icmp6_nodeinfo *ni6, *nni6;
1188 	struct mbuf *n = NULL;
1189 	u_int16_t qtype;
1190 	int subjlen;
1191 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1192 	struct ni_reply_fqdn *fqdn;
1193 	int addrs;		/* for NI_QTYPE_NODEADDR */
1194 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1195 	struct in6_addr in6_subj; /* subject address */
1196 	struct ip6_hdr *ip6;
1197 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1198 	char *subj = NULL;
1199 	struct in6_ifaddr *ia6 = NULL;
1200 
1201 	ip6 = mtod(m, struct ip6_hdr *);
1202 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1203 
1204 	/*
1205 	 * Validate IPv6 source address.
1206 	 * The default configuration MUST be to refuse answering queries from
1207 	 * global-scope addresses according to RFC4602.
1208 	 * Notes:
1209 	 *  - it's not very clear what "refuse" means; this implementation
1210 	 *    simply drops it.
1211 	 *  - it's not very easy to identify global-scope (unicast) addresses
1212 	 *    since there are many prefixes for them.  It should be safer
1213 	 *    and in practice sufficient to check "all" but loopback and
1214 	 *    link-local (note that site-local unicast was deprecated and
1215 	 *    ULA is defined as global scope-wise)
1216 	 */
1217 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1218 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1219 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1220 		goto bad;
1221 
1222 	/*
1223 	 * Validate IPv6 destination address.
1224 	 *
1225 	 * The Responder must discard the Query without further processing
1226 	 * unless it is one of the Responder's unicast or anycast addresses, or
1227 	 * a link-local scope multicast address which the Responder has joined.
1228 	 * [RFC4602, Section 5.]
1229 	 */
1230 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1231 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1232 			goto bad;
1233 		/* else it's a link-local multicast, fine */
1234 	} else {		/* unicast or anycast */
1235 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1236 		if (ia6 == NULL)
1237 			goto bad; /* XXX impossible */
1238 
1239 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1240 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1241 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1242 				"a temporary address in %s:%d",
1243 			       __FILE__, __LINE__));
1244 			goto bad;
1245 		}
1246 	}
1247 
1248 	/* validate query Subject field. */
1249 	qtype = ntohs(ni6->ni_qtype);
1250 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1251 	switch (qtype) {
1252 	case NI_QTYPE_NOOP:
1253 	case NI_QTYPE_SUPTYPES:
1254 		/* 07 draft */
1255 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1256 			break;
1257 		/* FALLTHROUGH */
1258 	case NI_QTYPE_FQDN:
1259 	case NI_QTYPE_NODEADDR:
1260 	case NI_QTYPE_IPV4ADDR:
1261 		switch (ni6->ni_code) {
1262 		case ICMP6_NI_SUBJ_IPV6:
1263 #if ICMP6_NI_SUBJ_IPV6 != 0
1264 		case 0:
1265 #endif
1266 			/*
1267 			 * backward compatibility - try to accept 03 draft
1268 			 * format, where no Subject is present.
1269 			 */
1270 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1271 			    subjlen == 0) {
1272 				oldfqdn++;
1273 				break;
1274 			}
1275 #if ICMP6_NI_SUBJ_IPV6 != 0
1276 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1277 				goto bad;
1278 #endif
1279 
1280 			if (subjlen != sizeof(struct in6_addr))
1281 				goto bad;
1282 
1283 			/*
1284 			 * Validate Subject address.
1285 			 *
1286 			 * Not sure what exactly "address belongs to the node"
1287 			 * means in the spec, is it just unicast, or what?
1288 			 *
1289 			 * At this moment we consider Subject address as
1290 			 * "belong to the node" if the Subject address equals
1291 			 * to the IPv6 destination address; validation for
1292 			 * IPv6 destination address should have done enough
1293 			 * check for us.
1294 			 *
1295 			 * We do not do proxy at this moment.
1296 			 */
1297 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1298 			    subjlen, (caddr_t)&in6_subj);
1299 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1300 				goto bad;
1301 
1302 			subj = (char *)&in6_subj;
1303 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1304 				break;
1305 
1306 			/*
1307 			 * XXX if we are to allow other cases, we should really
1308 			 * be careful about scope here.
1309 			 * basically, we should disallow queries toward IPv6
1310 			 * destination X with subject Y,
1311 			 * if scope(X) > scope(Y).
1312 			 * if we allow scope(X) > scope(Y), it will result in
1313 			 * information leakage across scope boundary.
1314 			 */
1315 			goto bad;
1316 
1317 		case ICMP6_NI_SUBJ_FQDN:
1318 			/*
1319 			 * Validate Subject name with gethostname(3).
1320 			 *
1321 			 * The behavior may need some debate, since:
1322 			 * - we are not sure if the node has FQDN as
1323 			 *   hostname (returned by gethostname(3)).
1324 			 * - the code does wildcard match for truncated names.
1325 			 *   however, we are not sure if we want to perform
1326 			 *   wildcard match, if gethostname(3) side has
1327 			 *   truncated hostname.
1328 			 */
1329 			mtx_lock(&pr->pr_mtx);
1330 			n = ni6_nametodns(pr->pr_hostname,
1331 			    strlen(pr->pr_hostname), 0);
1332 			mtx_unlock(&pr->pr_mtx);
1333 			if (!n || n->m_next || n->m_len == 0)
1334 				goto bad;
1335 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1336 			    subjlen) {
1337 				m = m_pullup(m, off +
1338 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1339 				if (m == NULL) {
1340 					IP6STAT_INC(ip6s_exthdrtoolong);
1341 					goto bad;
1342 				}
1343 			}
1344 			/* ip6 possibly invalid but not used after. */
1345 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1346 			subj = (char *)(mtod(m, caddr_t) + off +
1347 			    sizeof(struct icmp6_nodeinfo));
1348 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1349 			    n->m_len)) {
1350 				goto bad;
1351 			}
1352 			m_freem(n);
1353 			n = NULL;
1354 			break;
1355 
1356 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1357 		default:
1358 			goto bad;
1359 		}
1360 		break;
1361 	}
1362 
1363 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1364 	switch (qtype) {
1365 	case NI_QTYPE_FQDN:
1366 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1367 			goto bad;
1368 		break;
1369 	case NI_QTYPE_NODEADDR:
1370 	case NI_QTYPE_IPV4ADDR:
1371 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1372 			goto bad;
1373 		break;
1374 	}
1375 
1376 	/* guess reply length */
1377 	switch (qtype) {
1378 	case NI_QTYPE_NOOP:
1379 		break;		/* no reply data */
1380 	case NI_QTYPE_SUPTYPES:
1381 		replylen += sizeof(u_int32_t);
1382 		break;
1383 	case NI_QTYPE_FQDN:
1384 		/* XXX will append an mbuf */
1385 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1386 		break;
1387 	case NI_QTYPE_NODEADDR:
1388 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1389 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1390 		    sizeof(u_int32_t))) > MCLBYTES)
1391 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1392 		break;
1393 	case NI_QTYPE_IPV4ADDR:
1394 		/* unsupported - should respond with unknown Qtype? */
1395 		break;
1396 	default:
1397 		/*
1398 		 * XXX: We must return a reply with the ICMP6 code
1399 		 * `unknown Qtype' in this case.  However we regard the case
1400 		 * as an FQDN query for backward compatibility.
1401 		 * Older versions set a random value to this field,
1402 		 * so it rarely varies in the defined qtypes.
1403 		 * But the mechanism is not reliable...
1404 		 * maybe we should obsolete older versions.
1405 		 */
1406 		qtype = NI_QTYPE_FQDN;
1407 		/* XXX will append an mbuf */
1408 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1409 		oldfqdn++;
1410 		break;
1411 	}
1412 
1413 	/* Allocate an mbuf to reply. */
1414 	if (replylen > MCLBYTES) {
1415 		/*
1416 		 * XXX: should we try to allocate more? But MCLBYTES
1417 		 * is probably much larger than IPV6_MMTU...
1418 		 */
1419 		goto bad;
1420 	}
1421 	if (replylen > MHLEN)
1422 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1423 	else
1424 		n = m_gethdr(M_NOWAIT, m->m_type);
1425 	if (n == NULL) {
1426 		m_freem(m);
1427 		return (NULL);
1428 	}
1429 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1430 	n->m_pkthdr.len = n->m_len = replylen;
1431 
1432 	/* copy mbuf header and IPv6 + Node Information base headers */
1433 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1434 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1435 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1436 
1437 	/* qtype dependent procedure */
1438 	switch (qtype) {
1439 	case NI_QTYPE_NOOP:
1440 		nni6->ni_code = ICMP6_NI_SUCCESS;
1441 		nni6->ni_flags = 0;
1442 		break;
1443 	case NI_QTYPE_SUPTYPES:
1444 	{
1445 		u_int32_t v;
1446 		nni6->ni_code = ICMP6_NI_SUCCESS;
1447 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1448 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1449 		v = (u_int32_t)htonl(0x0000000f);
1450 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1451 		break;
1452 	}
1453 	case NI_QTYPE_FQDN:
1454 		nni6->ni_code = ICMP6_NI_SUCCESS;
1455 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1456 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1457 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1458 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1459 		/*
1460 		 * XXX do we really have FQDN in hostname?
1461 		 */
1462 		mtx_lock(&pr->pr_mtx);
1463 		n->m_next = ni6_nametodns(pr->pr_hostname,
1464 		    strlen(pr->pr_hostname), oldfqdn);
1465 		mtx_unlock(&pr->pr_mtx);
1466 		if (n->m_next == NULL)
1467 			goto bad;
1468 		/* XXX we assume that n->m_next is not a chain */
1469 		if (n->m_next->m_next != NULL)
1470 			goto bad;
1471 		n->m_pkthdr.len += n->m_next->m_len;
1472 		break;
1473 	case NI_QTYPE_NODEADDR:
1474 	{
1475 		int lenlim, copied;
1476 
1477 		nni6->ni_code = ICMP6_NI_SUCCESS;
1478 		n->m_pkthdr.len = n->m_len =
1479 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1480 		lenlim = M_TRAILINGSPACE(n);
1481 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1482 		/* XXX: reset mbuf length */
1483 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1484 		    sizeof(struct icmp6_nodeinfo) + copied;
1485 		break;
1486 	}
1487 	default:
1488 		break;		/* XXX impossible! */
1489 	}
1490 
1491 	nni6->ni_type = ICMP6_NI_REPLY;
1492 	m_freem(m);
1493 	return (n);
1494 
1495   bad:
1496 	m_freem(m);
1497 	if (n)
1498 		m_freem(n);
1499 	return (NULL);
1500 }
1501 
1502 /*
1503  * make a mbuf with DNS-encoded string.  no compression support.
1504  *
1505  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1506  * treated as truncated name (two \0 at the end).  this is a wild guess.
1507  *
1508  * old - return pascal string if non-zero
1509  */
1510 static struct mbuf *
1511 ni6_nametodns(const char *name, int namelen, int old)
1512 {
1513 	struct mbuf *m;
1514 	char *cp, *ep;
1515 	const char *p, *q;
1516 	int i, len, nterm;
1517 
1518 	if (old)
1519 		len = namelen + 1;
1520 	else
1521 		len = MCLBYTES;
1522 
1523 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1524 	if (len > MLEN)
1525 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1526 	else
1527 		m = m_get(M_NOWAIT, MT_DATA);
1528 	if (m == NULL)
1529 		goto fail;
1530 
1531 	if (old) {
1532 		m->m_len = len;
1533 		*mtod(m, char *) = namelen;
1534 		bcopy(name, mtod(m, char *) + 1, namelen);
1535 		return m;
1536 	} else {
1537 		m->m_len = 0;
1538 		cp = mtod(m, char *);
1539 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1540 
1541 		/* if not certain about my name, return empty buffer */
1542 		if (namelen == 0)
1543 			return m;
1544 
1545 		/*
1546 		 * guess if it looks like shortened hostname, or FQDN.
1547 		 * shortened hostname needs two trailing "\0".
1548 		 */
1549 		i = 0;
1550 		for (p = name; p < name + namelen; p++) {
1551 			if (*p && *p == '.')
1552 				i++;
1553 		}
1554 		if (i < 2)
1555 			nterm = 2;
1556 		else
1557 			nterm = 1;
1558 
1559 		p = name;
1560 		while (cp < ep && p < name + namelen) {
1561 			i = 0;
1562 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1563 				i++;
1564 			/* result does not fit into mbuf */
1565 			if (cp + i + 1 >= ep)
1566 				goto fail;
1567 			/*
1568 			 * DNS label length restriction, RFC1035 page 8.
1569 			 * "i == 0" case is included here to avoid returning
1570 			 * 0-length label on "foo..bar".
1571 			 */
1572 			if (i <= 0 || i >= 64)
1573 				goto fail;
1574 			*cp++ = i;
1575 			bcopy(p, cp, i);
1576 			cp += i;
1577 			p = q;
1578 			if (p < name + namelen && *p == '.')
1579 				p++;
1580 		}
1581 		/* termination */
1582 		if (cp + nterm >= ep)
1583 			goto fail;
1584 		while (nterm-- > 0)
1585 			*cp++ = '\0';
1586 		m->m_len = cp - mtod(m, char *);
1587 		return m;
1588 	}
1589 
1590 	panic("should not reach here");
1591 	/* NOTREACHED */
1592 
1593  fail:
1594 	if (m)
1595 		m_freem(m);
1596 	return NULL;
1597 }
1598 
1599 /*
1600  * check if two DNS-encoded string matches.  takes care of truncated
1601  * form (with \0\0 at the end).  no compression support.
1602  * XXX upper/lowercase match (see RFC2065)
1603  */
1604 static int
1605 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1606 {
1607 	const char *a0, *b0;
1608 	int l;
1609 
1610 	/* simplest case - need validation? */
1611 	if (alen == blen && bcmp(a, b, alen) == 0)
1612 		return 1;
1613 
1614 	a0 = a;
1615 	b0 = b;
1616 
1617 	/* termination is mandatory */
1618 	if (alen < 2 || blen < 2)
1619 		return 0;
1620 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1621 		return 0;
1622 	alen--;
1623 	blen--;
1624 
1625 	while (a - a0 < alen && b - b0 < blen) {
1626 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1627 			return 0;
1628 
1629 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1630 			return 0;
1631 		/* we don't support compression yet */
1632 		if (a[0] >= 64 || b[0] >= 64)
1633 			return 0;
1634 
1635 		/* truncated case */
1636 		if (a[0] == 0 && a - a0 == alen - 1)
1637 			return 1;
1638 		if (b[0] == 0 && b - b0 == blen - 1)
1639 			return 1;
1640 		if (a[0] == 0 || b[0] == 0)
1641 			return 0;
1642 
1643 		if (a[0] != b[0])
1644 			return 0;
1645 		l = a[0];
1646 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1647 			return 0;
1648 		if (bcmp(a + 1, b + 1, l) != 0)
1649 			return 0;
1650 
1651 		a += 1 + l;
1652 		b += 1 + l;
1653 	}
1654 
1655 	if (a - a0 == alen && b - b0 == blen)
1656 		return 1;
1657 	else
1658 		return 0;
1659 }
1660 
1661 /*
1662  * calculate the number of addresses to be returned in the node info reply.
1663  */
1664 static int
1665 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1666     struct in6_addr *subj)
1667 {
1668 	struct ifnet *ifp;
1669 	struct in6_ifaddr *ifa6;
1670 	struct ifaddr *ifa;
1671 	int addrs = 0, addrsofif, iffound = 0;
1672 	int niflags = ni6->ni_flags;
1673 
1674 	NET_EPOCH_ASSERT();
1675 
1676 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1677 		switch (ni6->ni_code) {
1678 		case ICMP6_NI_SUBJ_IPV6:
1679 			if (subj == NULL) /* must be impossible... */
1680 				return (0);
1681 			break;
1682 		default:
1683 			/*
1684 			 * XXX: we only support IPv6 subject address for
1685 			 * this Qtype.
1686 			 */
1687 			return (0);
1688 		}
1689 	}
1690 
1691 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1692 		addrsofif = 0;
1693 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1694 			if (ifa->ifa_addr->sa_family != AF_INET6)
1695 				continue;
1696 			ifa6 = (struct in6_ifaddr *)ifa;
1697 
1698 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1699 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1700 				iffound = 1;
1701 
1702 			/*
1703 			 * IPv4-mapped addresses can only be returned by a
1704 			 * Node Information proxy, since they represent
1705 			 * addresses of IPv4-only nodes, which perforce do
1706 			 * not implement this protocol.
1707 			 * [icmp-name-lookups-07, Section 5.4]
1708 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1709 			 * this function at this moment.
1710 			 */
1711 
1712 			/* What do we have to do about ::1? */
1713 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1714 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1715 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1716 					continue;
1717 				break;
1718 			case IPV6_ADDR_SCOPE_SITELOCAL:
1719 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1720 					continue;
1721 				break;
1722 			case IPV6_ADDR_SCOPE_GLOBAL:
1723 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1724 					continue;
1725 				break;
1726 			default:
1727 				continue;
1728 			}
1729 
1730 			/*
1731 			 * check if anycast is okay.
1732 			 * XXX: just experimental.  not in the spec.
1733 			 */
1734 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1735 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1736 				continue; /* we need only unicast addresses */
1737 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1738 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1739 				continue;
1740 			}
1741 			addrsofif++; /* count the address */
1742 		}
1743 		if (iffound) {
1744 			*ifpp = ifp;
1745 			return (addrsofif);
1746 		}
1747 
1748 		addrs += addrsofif;
1749 	}
1750 
1751 	return (addrs);
1752 }
1753 
1754 static int
1755 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1756     struct ifnet *ifp0, int resid)
1757 {
1758 	struct ifnet *ifp;
1759 	struct in6_ifaddr *ifa6;
1760 	struct ifaddr *ifa;
1761 	struct ifnet *ifp_dep = NULL;
1762 	int copied = 0, allow_deprecated = 0;
1763 	u_char *cp = (u_char *)(nni6 + 1);
1764 	int niflags = ni6->ni_flags;
1765 	u_int32_t ltime;
1766 
1767 	NET_EPOCH_ASSERT();
1768 
1769 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1770 		return (0);	/* needless to copy */
1771 
1772 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1773   again:
1774 
1775 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1776 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1777 			if (ifa->ifa_addr->sa_family != AF_INET6)
1778 				continue;
1779 			ifa6 = (struct in6_ifaddr *)ifa;
1780 
1781 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1782 			    allow_deprecated == 0) {
1783 				/*
1784 				 * prefererred address should be put before
1785 				 * deprecated addresses.
1786 				 */
1787 
1788 				/* record the interface for later search */
1789 				if (ifp_dep == NULL)
1790 					ifp_dep = ifp;
1791 
1792 				continue;
1793 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1794 			    allow_deprecated != 0)
1795 				continue; /* we now collect deprecated addrs */
1796 
1797 			/* What do we have to do about ::1? */
1798 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1799 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1800 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1801 					continue;
1802 				break;
1803 			case IPV6_ADDR_SCOPE_SITELOCAL:
1804 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1805 					continue;
1806 				break;
1807 			case IPV6_ADDR_SCOPE_GLOBAL:
1808 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1809 					continue;
1810 				break;
1811 			default:
1812 				continue;
1813 			}
1814 
1815 			/*
1816 			 * check if anycast is okay.
1817 			 * XXX: just experimental.  not in the spec.
1818 			 */
1819 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1820 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1821 				continue;
1822 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1823 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1824 				continue;
1825 			}
1826 
1827 			/* now we can copy the address */
1828 			if (resid < sizeof(struct in6_addr) +
1829 			    sizeof(u_int32_t)) {
1830 				/*
1831 				 * We give up much more copy.
1832 				 * Set the truncate flag and return.
1833 				 */
1834 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1835 				return (copied);
1836 			}
1837 
1838 			/*
1839 			 * Set the TTL of the address.
1840 			 * The TTL value should be one of the following
1841 			 * according to the specification:
1842 			 *
1843 			 * 1. The remaining lifetime of a DHCP lease on the
1844 			 *    address, or
1845 			 * 2. The remaining Valid Lifetime of a prefix from
1846 			 *    which the address was derived through Stateless
1847 			 *    Autoconfiguration.
1848 			 *
1849 			 * Note that we currently do not support stateful
1850 			 * address configuration by DHCPv6, so the former
1851 			 * case can't happen.
1852 			 */
1853 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1854 				ltime = ND6_INFINITE_LIFETIME;
1855 			else {
1856 				if (ifa6->ia6_lifetime.ia6t_expire >
1857 				    time_uptime)
1858 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1859 				else
1860 					ltime = 0;
1861 			}
1862 
1863 			bcopy(&ltime, cp, sizeof(u_int32_t));
1864 			cp += sizeof(u_int32_t);
1865 
1866 			/* copy the address itself */
1867 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1868 			    sizeof(struct in6_addr));
1869 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1870 			cp += sizeof(struct in6_addr);
1871 
1872 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1873 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1874 		}
1875 		if (ifp0)	/* we need search only on the specified IF */
1876 			break;
1877 	}
1878 
1879 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1880 		ifp = ifp_dep;
1881 		allow_deprecated = 1;
1882 
1883 		goto again;
1884 	}
1885 
1886 	return (copied);
1887 }
1888 
1889 static bool
1890 icmp6_rip6_match(const struct inpcb *inp, void *v)
1891 {
1892 	struct ip6_hdr *ip6 = v;
1893 
1894 	if ((inp->inp_vflag & INP_IPV6) == 0)
1895 		return (false);
1896 	if (inp->inp_ip_p != IPPROTO_ICMPV6)
1897 		return (false);
1898 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1899 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1900 		return (false);
1901 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1902 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1903 		return (false);
1904 	return (true);
1905 }
1906 
1907 /*
1908  * XXX almost dup'ed code with rip6_input.
1909  */
1910 static int
1911 icmp6_rip6_input(struct mbuf **mp, int off)
1912 {
1913 	struct mbuf *n, *m = *mp;
1914 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1915 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1916 	    INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1917 	struct inpcb *inp;
1918 	struct sockaddr_in6 fromsa;
1919 	struct icmp6_hdr *icmp6;
1920 	struct mbuf *opts = NULL;
1921 	int delivered = 0;
1922 
1923 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1924 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1925 
1926 	/*
1927 	 * XXX: the address may have embedded scope zone ID, which should be
1928 	 * hidden from applications.
1929 	 */
1930 	bzero(&fromsa, sizeof(fromsa));
1931 	fromsa.sin6_family = AF_INET6;
1932 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1933 	fromsa.sin6_addr = ip6->ip6_src;
1934 	if (sa6_recoverscope(&fromsa)) {
1935 		m_freem(m);
1936 		*mp = NULL;
1937 		return (IPPROTO_DONE);
1938 	}
1939 
1940 	while ((inp = inp_next(&inpi)) != NULL) {
1941 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1942 		    inp->in6p_icmp6filt))
1943 			continue;
1944 		/*
1945 		 * Recent network drivers tend to allocate a single
1946 		 * mbuf cluster, rather than to make a couple of
1947 		 * mbufs without clusters.  Also, since the IPv6 code
1948 		 * path tries to avoid m_pullup(), it is highly
1949 		 * probable that we still have an mbuf cluster here
1950 		 * even though the necessary length can be stored in an
1951 		 * mbuf's internal buffer.
1952 		 * Meanwhile, the default size of the receive socket
1953 		 * buffer for raw sockets is not so large.  This means
1954 		 * the possibility of packet loss is relatively higher
1955 		 * than before.  To avoid this scenario, we copy the
1956 		 * received data to a separate mbuf that does not use
1957 		 * a cluster, if possible.
1958 		 * XXX: it is better to copy the data after stripping
1959 		 * intermediate headers.
1960 		 */
1961 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1962 		    m->m_len <= MHLEN) {
1963 			n = m_get(M_NOWAIT, m->m_type);
1964 			if (n != NULL) {
1965 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1966 					bcopy(m->m_data, n->m_data, m->m_len);
1967 					n->m_len = m->m_len;
1968 				} else {
1969 					m_free(n);
1970 					n = NULL;
1971 				}
1972 			}
1973 		} else
1974 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1975 		if (n == NULL)
1976 			continue;
1977 		if (inp->inp_flags & INP_CONTROLOPTS)
1978 			ip6_savecontrol(inp, n, &opts);
1979 		/* strip intermediate headers */
1980 		m_adj(n, off);
1981 		SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1982 		if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1983 		    (struct sockaddr *)&fromsa, n, opts) == 0) {
1984 			soroverflow_locked(inp->inp_socket);
1985 			m_freem(n);
1986 			if (opts)
1987 				m_freem(opts);
1988 		} else {
1989 			sorwakeup_locked(inp->inp_socket);
1990 			delivered++;
1991 		}
1992 		opts = NULL;
1993 	}
1994 	m_freem(m);
1995 	*mp = NULL;
1996 	if (delivered == 0)
1997 		IP6STAT_DEC(ip6s_delivered);
1998 	return (IPPROTO_DONE);
1999 }
2000 
2001 /*
2002  * Reflect the ip6 packet back to the source.
2003  * OFF points to the icmp6 header, counted from the top of the mbuf.
2004  */
2005 static void
2006 icmp6_reflect(struct mbuf *m, size_t off)
2007 {
2008 	struct in6_addr src6, *srcp;
2009 	struct ip6_hdr *ip6;
2010 	struct icmp6_hdr *icmp6;
2011 	struct in6_ifaddr *ia = NULL;
2012 	struct ifnet *outif = NULL;
2013 	int plen;
2014 	int type, code, hlim;
2015 
2016 	/* too short to reflect */
2017 	if (off < sizeof(struct ip6_hdr)) {
2018 		nd6log((LOG_DEBUG,
2019 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2020 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2021 		    __FILE__, __LINE__));
2022 		goto bad;
2023 	}
2024 
2025 	/*
2026 	 * If there are extra headers between IPv6 and ICMPv6, strip
2027 	 * off that header first.
2028 	 */
2029 #ifdef DIAGNOSTIC
2030 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2031 		panic("assumption failed in icmp6_reflect");
2032 #endif
2033 	if (off > sizeof(struct ip6_hdr)) {
2034 		size_t l;
2035 		struct ip6_hdr nip6;
2036 
2037 		l = off - sizeof(struct ip6_hdr);
2038 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2039 		m_adj(m, l);
2040 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2041 		if (m->m_len < l) {
2042 			if ((m = m_pullup(m, l)) == NULL)
2043 				return;
2044 		}
2045 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2046 	} else /* off == sizeof(struct ip6_hdr) */ {
2047 		size_t l;
2048 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2049 		if (m->m_len < l) {
2050 			if ((m = m_pullup(m, l)) == NULL)
2051 				return;
2052 		}
2053 	}
2054 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2055 	ip6 = mtod(m, struct ip6_hdr *);
2056 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2057 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2058 	type = icmp6->icmp6_type; /* keep type for statistics */
2059 	code = icmp6->icmp6_code; /* ditto. */
2060 	hlim = 0;
2061 	srcp = NULL;
2062 
2063 	/*
2064 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2065 	 * use dst as the src for the reply.
2066 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2067 	 * (for example) when we encounter an error while forwarding procedure
2068 	 * destined to a duplicated address of ours.
2069 	 */
2070 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2071 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2072 		if (ia != NULL && !(ia->ia6_flags &
2073 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2074 			src6 = ia->ia_addr.sin6_addr;
2075 			srcp = &src6;
2076 
2077 			if (m->m_pkthdr.rcvif != NULL) {
2078 				/* XXX: This may not be the outgoing interface */
2079 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2080 			} else
2081 				hlim = V_ip6_defhlim;
2082 		}
2083 	}
2084 
2085 	if (srcp == NULL) {
2086 		int error;
2087 		struct in6_addr dst6;
2088 		uint32_t scopeid;
2089 
2090 		/*
2091 		 * This case matches to multicasts, our anycast, or unicasts
2092 		 * that we do not own.  Select a source address based on the
2093 		 * source address of the erroneous packet.
2094 		 */
2095 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2096 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2097 		    scopeid, NULL, &src6, &hlim);
2098 
2099 		if (error) {
2100 			char ip6buf[INET6_ADDRSTRLEN];
2101 			nd6log((LOG_DEBUG,
2102 			    "icmp6_reflect: source can't be determined: "
2103 			    "dst=%s, error=%d\n",
2104 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2105 			goto bad;
2106 		}
2107 		srcp = &src6;
2108 	}
2109 	/*
2110 	 * ip6_input() drops a packet if its src is multicast.
2111 	 * So, the src is never multicast.
2112 	 */
2113 	ip6->ip6_dst = ip6->ip6_src;
2114 	ip6->ip6_src = *srcp;
2115 	ip6->ip6_flow = 0;
2116 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2117 	ip6->ip6_vfc |= IPV6_VERSION;
2118 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2119 	ip6->ip6_hlim = hlim;
2120 
2121 	icmp6->icmp6_cksum = 0;
2122 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2123 	    sizeof(struct ip6_hdr), plen);
2124 
2125 	/*
2126 	 * XXX option handling
2127 	 */
2128 
2129 	m->m_flags &= ~(M_BCAST|M_MCAST);
2130 	m->m_pkthdr.rcvif = NULL;
2131 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2132 	if (outif)
2133 		icmp6_ifoutstat_inc(outif, type, code);
2134 
2135 	return;
2136 
2137  bad:
2138 	m_freem(m);
2139 	return;
2140 }
2141 
2142 static const char *
2143 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2144     struct in6_addr *tgt6)
2145 {
2146 	static char buf[1024];
2147 	char ip6bufs[INET6_ADDRSTRLEN];
2148 	char ip6bufd[INET6_ADDRSTRLEN];
2149 	char ip6buft[INET6_ADDRSTRLEN];
2150 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2151 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2152 	    ip6_sprintf(ip6buft, tgt6));
2153 	return buf;
2154 }
2155 
2156 void
2157 icmp6_redirect_input(struct mbuf *m, int off)
2158 {
2159 	struct ifnet *ifp;
2160 	struct ip6_hdr *ip6;
2161 	struct nd_redirect *nd_rd;
2162 	struct in6_addr src6, redtgt6, reddst6;
2163 	union nd_opts ndopts;
2164 	char ip6buf[INET6_ADDRSTRLEN];
2165 	char *lladdr;
2166 	int icmp6len, is_onlink, is_router, lladdrlen;
2167 
2168 	M_ASSERTPKTHDR(m);
2169 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2170 
2171 	/* XXX if we are router, we don't update route by icmp6 redirect */
2172 	if (V_ip6_forwarding)
2173 		goto freeit;
2174 	if (!V_icmp6_rediraccept)
2175 		goto freeit;
2176 
2177 	/* RFC 6980: Nodes MUST silently ignore fragments */
2178 	if(m->m_flags & M_FRAGMENTED)
2179 		goto freeit;
2180 
2181 	ip6 = mtod(m, struct ip6_hdr *);
2182 	icmp6len = ntohs(ip6->ip6_plen);
2183 	if (m->m_len < off + icmp6len) {
2184 		m = m_pullup(m, off + icmp6len);
2185 		if (m == NULL) {
2186 			IP6STAT_INC(ip6s_exthdrtoolong);
2187 			return;
2188 		}
2189 	}
2190 	ip6 = mtod(m, struct ip6_hdr *);
2191 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2192 
2193 	ifp = m->m_pkthdr.rcvif;
2194 	redtgt6 = nd_rd->nd_rd_target;
2195 	reddst6 = nd_rd->nd_rd_dst;
2196 
2197 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2198 	    in6_setscope(&reddst6, ifp, NULL)) {
2199 		goto freeit;
2200 	}
2201 
2202 	/* validation */
2203 	src6 = ip6->ip6_src;
2204 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2205 		nd6log((LOG_ERR,
2206 		    "ICMP6 redirect sent from %s rejected; "
2207 		    "must be from linklocal\n",
2208 		    ip6_sprintf(ip6buf, &src6)));
2209 		goto bad;
2210 	}
2211 	if (__predict_false(ip6->ip6_hlim != 255)) {
2212 		ICMP6STAT_INC(icp6s_invlhlim);
2213 		nd6log((LOG_ERR,
2214 		    "ICMP6 redirect sent from %s rejected; "
2215 		    "hlim=%d (must be 255)\n",
2216 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2217 		goto bad;
2218 	}
2219     {
2220 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2221 	struct nhop_object *nh;
2222 	struct in6_addr kdst;
2223 	uint32_t scopeid;
2224 
2225 	in6_splitscope(&reddst6, &kdst, &scopeid);
2226 	NET_EPOCH_ASSERT();
2227 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2228 	if (nh != NULL) {
2229 		struct in6_addr nh_addr;
2230 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2231 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2232 			nd6log((LOG_ERR,
2233 			    "ICMP6 redirect rejected; no route "
2234 			    "with inet6 gateway found for redirect dst: %s\n",
2235 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2236 			goto bad;
2237 		}
2238 
2239 		/*
2240 		 * Embed scope zone id into next hop address.
2241 		 */
2242 		nh_addr = nh->gw6_sa.sin6_addr;
2243 
2244 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2245 			nd6log((LOG_ERR,
2246 			    "ICMP6 redirect rejected; "
2247 			    "not equal to gw-for-src=%s (must be same): "
2248 			    "%s\n",
2249 			    ip6_sprintf(ip6buf, &nh_addr),
2250 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2251 			goto bad;
2252 		}
2253 	} else {
2254 		nd6log((LOG_ERR,
2255 		    "ICMP6 redirect rejected; "
2256 		    "no route found for redirect dst: %s\n",
2257 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2258 		goto bad;
2259 	}
2260     }
2261 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2262 		nd6log((LOG_ERR,
2263 		    "ICMP6 redirect rejected; "
2264 		    "redirect dst must be unicast: %s\n",
2265 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2266 		goto bad;
2267 	}
2268 
2269 	is_router = is_onlink = 0;
2270 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2271 		is_router = 1;	/* router case */
2272 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2273 		is_onlink = 1;	/* on-link destination case */
2274 	if (!is_router && !is_onlink) {
2275 		nd6log((LOG_ERR,
2276 		    "ICMP6 redirect rejected; "
2277 		    "neither router case nor onlink case: %s\n",
2278 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2279 		goto bad;
2280 	}
2281 
2282 	icmp6len -= sizeof(*nd_rd);
2283 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2284 	if (nd6_options(&ndopts) < 0) {
2285 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2286 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2287 		/* nd6_options have incremented stats */
2288 		goto freeit;
2289 	}
2290 
2291 	lladdr = NULL;
2292 	lladdrlen = 0;
2293 	if (ndopts.nd_opts_tgt_lladdr) {
2294 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2295 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2296 	}
2297 
2298 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2299 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2300 		    "(if %d, icmp6 packet %d): %s\n",
2301 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2302 		    ifp->if_addrlen, lladdrlen - 2,
2303 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2304 		goto bad;
2305 	}
2306 
2307 	/* Validation passed. */
2308 
2309 	/* RFC 2461 8.3 */
2310 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2311 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2312 
2313 	/*
2314 	 * Install a gateway route in the better-router case or an interface
2315 	 * route in the on-link-destination case.
2316 	 */
2317 	{
2318 		struct sockaddr_in6 sdst;
2319 		struct sockaddr_in6 sgw;
2320 		struct sockaddr_in6 ssrc;
2321 		struct sockaddr *gw;
2322 		int rt_flags;
2323 		u_int fibnum;
2324 
2325 		bzero(&sdst, sizeof(sdst));
2326 		bzero(&ssrc, sizeof(ssrc));
2327 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2328 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2329 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2330 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2331 		rt_flags = 0;
2332 		if (is_router) {
2333 			bzero(&sgw, sizeof(sgw));
2334 			sgw.sin6_family = AF_INET6;
2335 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2336 			bcopy(&redtgt6, &sgw.sin6_addr,
2337 				sizeof(struct in6_addr));
2338 			gw = (struct sockaddr *)&sgw;
2339 			rt_flags |= RTF_GATEWAY;
2340 		} else
2341 			gw = ifp->if_addr->ifa_addr;
2342 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2343 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2344 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2345 			    V_icmp6_redirtimeout);
2346 	}
2347 
2348  freeit:
2349 	m_freem(m);
2350 	return;
2351 
2352  bad:
2353 	ICMP6STAT_INC(icp6s_badredirect);
2354 	m_freem(m);
2355 }
2356 
2357 void
2358 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2359 {
2360 	struct ifnet *ifp;	/* my outgoing interface */
2361 	struct in6_addr *ifp_ll6;
2362 	struct in6_addr *router_ll6;
2363 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2364 	struct mbuf *m = NULL;	/* newly allocated one */
2365 	struct m_tag *mtag;
2366 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2367 	struct nd_redirect *nd_rd;
2368 	struct llentry *ln = NULL;
2369 	size_t maxlen;
2370 	u_char *p;
2371 	struct ifnet *outif = NULL;
2372 	struct sockaddr_in6 src_sa;
2373 
2374 	icmp6_errcount(ND_REDIRECT, 0);
2375 
2376 	/* if we are not router, we don't send icmp6 redirect */
2377 	if (!V_ip6_forwarding)
2378 		goto fail;
2379 
2380 	/* sanity check */
2381 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2382 		goto fail;
2383 
2384 	/*
2385 	 * Address check:
2386 	 *  the source address must identify a neighbor, and
2387 	 *  the destination address must not be a multicast address
2388 	 *  [RFC 2461, sec 8.2]
2389 	 */
2390 	sip6 = mtod(m0, struct ip6_hdr *);
2391 	bzero(&src_sa, sizeof(src_sa));
2392 	src_sa.sin6_family = AF_INET6;
2393 	src_sa.sin6_len = sizeof(src_sa);
2394 	src_sa.sin6_addr = sip6->ip6_src;
2395 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2396 		goto fail;
2397 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2398 		goto fail;	/* what should we do here? */
2399 
2400 	/* rate limit */
2401 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2402 		goto fail;
2403 
2404 	/*
2405 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2406 	 * we almost always ask for an mbuf cluster for simplicity.
2407 	 * (MHLEN < IPV6_MMTU is almost always true)
2408 	 */
2409 #if IPV6_MMTU >= MCLBYTES
2410 # error assumption failed about IPV6_MMTU and MCLBYTES
2411 #endif
2412 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2413 	if (m == NULL)
2414 		goto fail;
2415 	M_SETFIB(m, M_GETFIB(m0));
2416 	maxlen = M_TRAILINGSPACE(m);
2417 	maxlen = min(IPV6_MMTU, maxlen);
2418 	/* just for safety */
2419 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2420 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2421 		goto fail;
2422 	}
2423 
2424 	{
2425 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2426 		struct in6_ifaddr *ia;
2427 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2428 						 IN6_IFF_NOTREADY|
2429 						 IN6_IFF_ANYCAST)) == NULL)
2430 			goto fail;
2431 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2432 		/* XXXRW: reference released prematurely. */
2433 		ifa_free(&ia->ia_ifa);
2434 	}
2435 
2436 	/* get ip6 linklocal address for the router. */
2437 	if (nh->nh_flags & NHF_GATEWAY) {
2438 		struct sockaddr_in6 *sin6;
2439 		sin6 = &nh->gw6_sa;
2440 		router_ll6 = &sin6->sin6_addr;
2441 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2442 			router_ll6 = (struct in6_addr *)NULL;
2443 	} else
2444 		router_ll6 = (struct in6_addr *)NULL;
2445 
2446 	/* ip6 */
2447 	ip6 = mtod(m, struct ip6_hdr *);
2448 	ip6->ip6_flow = 0;
2449 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2450 	ip6->ip6_vfc |= IPV6_VERSION;
2451 	/* ip6->ip6_plen will be set later */
2452 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2453 	ip6->ip6_hlim = 255;
2454 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2455 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2456 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2457 
2458 	/* ND Redirect */
2459 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2460 	nd_rd->nd_rd_type = ND_REDIRECT;
2461 	nd_rd->nd_rd_code = 0;
2462 	nd_rd->nd_rd_reserved = 0;
2463 	if (nh->nh_flags & NHF_GATEWAY) {
2464 		/*
2465 		 * nd_rd->nd_rd_target must be a link-local address in
2466 		 * better router cases.
2467 		 */
2468 		if (!router_ll6)
2469 			goto fail;
2470 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2471 		    sizeof(nd_rd->nd_rd_target));
2472 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2473 		    sizeof(nd_rd->nd_rd_dst));
2474 	} else {
2475 		/* make sure redtgt == reddst */
2476 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2477 		    sizeof(nd_rd->nd_rd_target));
2478 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2479 		    sizeof(nd_rd->nd_rd_dst));
2480 	}
2481 
2482 	p = (u_char *)(nd_rd + 1);
2483 
2484 	if (!router_ll6)
2485 		goto nolladdropt;
2486 
2487 	{
2488 		/* target lladdr option */
2489 		int len;
2490 		struct nd_opt_hdr *nd_opt;
2491 		char *lladdr;
2492 
2493 		ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2494 		if (ln == NULL)
2495 			goto nolladdropt;
2496 
2497 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2498 		len = (len + 7) & ~7;	/* round by 8 */
2499 		/* safety check */
2500 		if (len + (p - (u_char *)ip6) > maxlen)
2501 			goto nolladdropt;
2502 
2503 		if (ln->la_flags & LLE_VALID) {
2504 			nd_opt = (struct nd_opt_hdr *)p;
2505 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2506 			nd_opt->nd_opt_len = len >> 3;
2507 			lladdr = (char *)(nd_opt + 1);
2508 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2509 			p += len;
2510 		}
2511 	}
2512 nolladdropt:
2513 	if (ln != NULL)
2514 		LLE_RUNLOCK(ln);
2515 
2516 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2517 
2518 	/* just to be safe */
2519 #ifdef M_DECRYPTED	/*not openbsd*/
2520 	if (m0->m_flags & M_DECRYPTED)
2521 		goto noredhdropt;
2522 #endif
2523 	if (p - (u_char *)ip6 > maxlen)
2524 		goto noredhdropt;
2525 
2526 	{
2527 		/* redirected header option */
2528 		int len;
2529 		struct nd_opt_rd_hdr *nd_opt_rh;
2530 
2531 		/*
2532 		 * compute the maximum size for icmp6 redirect header option.
2533 		 * XXX room for auth header?
2534 		 */
2535 		len = maxlen - (p - (u_char *)ip6);
2536 		len &= ~7;
2537 
2538 		/* This is just for simplicity. */
2539 		if (m0->m_pkthdr.len != m0->m_len) {
2540 			if (m0->m_next) {
2541 				m_freem(m0->m_next);
2542 				m0->m_next = NULL;
2543 			}
2544 			m0->m_pkthdr.len = m0->m_len;
2545 		}
2546 
2547 		/*
2548 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2549 		 * about padding/truncate rule for the original IP packet.
2550 		 * From the discussion on IPv6imp in Feb 1999,
2551 		 * the consensus was:
2552 		 * - "attach as much as possible" is the goal
2553 		 * - pad if not aligned (original size can be guessed by
2554 		 *   original ip6 header)
2555 		 * Following code adds the padding if it is simple enough,
2556 		 * and truncates if not.
2557 		 */
2558 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2559 			panic("assumption failed in %s:%d", __FILE__,
2560 			    __LINE__);
2561 
2562 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2563 			/* not enough room, truncate */
2564 			m0->m_pkthdr.len = m0->m_len = len -
2565 			    sizeof(*nd_opt_rh);
2566 		} else {
2567 			/* enough room, pad or truncate */
2568 			size_t extra;
2569 
2570 			extra = m0->m_pkthdr.len % 8;
2571 			if (extra) {
2572 				/* pad if easy enough, truncate if not */
2573 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2574 					/* pad */
2575 					m0->m_len += (8 - extra);
2576 					m0->m_pkthdr.len += (8 - extra);
2577 				} else {
2578 					/* truncate */
2579 					m0->m_pkthdr.len -= extra;
2580 					m0->m_len -= extra;
2581 				}
2582 			}
2583 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2584 			m0->m_pkthdr.len = m0->m_len = len -
2585 			    sizeof(*nd_opt_rh);
2586 		}
2587 
2588 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2589 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2590 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2591 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2592 		p += sizeof(*nd_opt_rh);
2593 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2594 
2595 		/* connect m0 to m */
2596 		m_tag_delete_chain(m0, NULL);
2597 		m0->m_flags &= ~M_PKTHDR;
2598 		m->m_next = m0;
2599 		m->m_pkthdr.len = m->m_len + m0->m_len;
2600 		m0 = NULL;
2601 	}
2602 noredhdropt:;
2603 	if (m0) {
2604 		m_freem(m0);
2605 		m0 = NULL;
2606 	}
2607 
2608 	/* XXX: clear embedded link IDs in the inner header */
2609 	in6_clearscope(&sip6->ip6_src);
2610 	in6_clearscope(&sip6->ip6_dst);
2611 	in6_clearscope(&nd_rd->nd_rd_target);
2612 	in6_clearscope(&nd_rd->nd_rd_dst);
2613 
2614 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2615 
2616 	nd_rd->nd_rd_cksum = 0;
2617 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2618 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2619 
2620         if (send_sendso_input_hook != NULL) {
2621 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2622 			M_NOWAIT);
2623 		if (mtag == NULL)
2624 			goto fail;
2625 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2626 		m_tag_prepend(m, mtag);
2627 	}
2628 
2629 	/* send the packet to outside... */
2630 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2631 	if (outif) {
2632 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2633 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2634 	}
2635 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2636 
2637 	return;
2638 
2639 fail:
2640 	if (m)
2641 		m_freem(m);
2642 	if (m0)
2643 		m_freem(m0);
2644 }
2645 
2646 /*
2647  * ICMPv6 socket option processing.
2648  */
2649 int
2650 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2651 {
2652 	int error = 0;
2653 	int optlen;
2654 	struct inpcb *inp = sotoinpcb(so);
2655 	int level, op, optname;
2656 
2657 	if (sopt) {
2658 		level = sopt->sopt_level;
2659 		op = sopt->sopt_dir;
2660 		optname = sopt->sopt_name;
2661 		optlen = sopt->sopt_valsize;
2662 	} else
2663 		level = op = optname = optlen = 0;
2664 
2665 	if (level != IPPROTO_ICMPV6) {
2666 		return EINVAL;
2667 	}
2668 
2669 	switch (op) {
2670 	case PRCO_SETOPT:
2671 		switch (optname) {
2672 		case ICMP6_FILTER:
2673 		    {
2674 			struct icmp6_filter ic6f;
2675 
2676 			if (optlen != sizeof(ic6f)) {
2677 				error = EMSGSIZE;
2678 				break;
2679 			}
2680 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2681 			if (error == 0) {
2682 				INP_WLOCK(inp);
2683 				*inp->in6p_icmp6filt = ic6f;
2684 				INP_WUNLOCK(inp);
2685 			}
2686 			break;
2687 		    }
2688 
2689 		default:
2690 			error = ENOPROTOOPT;
2691 			break;
2692 		}
2693 		break;
2694 
2695 	case PRCO_GETOPT:
2696 		switch (optname) {
2697 		case ICMP6_FILTER:
2698 		    {
2699 			struct icmp6_filter ic6f;
2700 
2701 			INP_RLOCK(inp);
2702 			ic6f = *inp->in6p_icmp6filt;
2703 			INP_RUNLOCK(inp);
2704 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2705 			break;
2706 		    }
2707 
2708 		default:
2709 			error = ENOPROTOOPT;
2710 			break;
2711 		}
2712 		break;
2713 	}
2714 
2715 	return (error);
2716 }
2717 
2718 /*
2719  * Perform rate limit check.
2720  * Returns 0 if it is okay to send the icmp6 packet.
2721  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2722  * limitation.
2723  *
2724  * XXX per-destination/type check necessary?
2725  *
2726  * dst - not used at this moment
2727  * type - not used at this moment
2728  * code - not used at this moment
2729  */
2730 static int
2731 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2732     const int code)
2733 {
2734 	int ret;
2735 
2736 	ret = 0;	/* okay to send */
2737 
2738 	/* PPS limit */
2739 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2740 	    V_icmp6errppslim)) {
2741 		/* The packet is subject to rate limit */
2742 		ret++;
2743 	}
2744 
2745 	return ret;
2746 }
2747