xref: /freebsd/sys/netinet6/icmp6.c (revision d4eeb02986980bf33dd56c41ceb9fc5f180c0d47)
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 ipproto_ctlinput_t	*ip6_ctlprotox[];
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(int, icmp6errppslim);
128 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
129 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
130 VNET_DECLARE(int, icmp6_nodeinfo);
131 
132 #define	V_ripcbinfo			VNET(ripcbinfo)
133 #define	V_icmp6errppslim		VNET(icmp6errppslim)
134 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
135 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
136 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
137 
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static void icmp6_reflect(struct mbuf *, size_t);
141 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
142 static const char *icmp6_redirect_diag(struct in6_addr *,
143 	struct in6_addr *, struct in6_addr *);
144 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
145 static struct mbuf *ni6_nametodns(const char *, int, int);
146 static int ni6_dnsmatch(const char *, int, const char *, int);
147 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
148 			  struct ifnet **, struct in6_addr *);
149 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
150 				struct ifnet *, int);
151 static int icmp6_notify_error(struct mbuf **, int, int, int);
152 
153 /*
154  * Kernel module interface for updating icmp6stat.  The argument is an index
155  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
156  * general layout of icmp6stat into the caller, it doesn't encode its
157  * location, so that future changes to add, for example, per-CPU stats
158  * support won't cause binary compatibility problems for kernel modules.
159  */
160 void
161 kmod_icmp6stat_inc(int statnum)
162 {
163 
164 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
165 }
166 
167 static void
168 icmp6_errcount(int type, int code)
169 {
170 	switch (type) {
171 	case ICMP6_DST_UNREACH:
172 		switch (code) {
173 		case ICMP6_DST_UNREACH_NOROUTE:
174 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
175 			return;
176 		case ICMP6_DST_UNREACH_ADMIN:
177 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
178 			return;
179 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
180 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
181 			return;
182 		case ICMP6_DST_UNREACH_ADDR:
183 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
184 			return;
185 		case ICMP6_DST_UNREACH_NOPORT:
186 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
187 			return;
188 		}
189 		break;
190 	case ICMP6_PACKET_TOO_BIG:
191 		ICMP6STAT_INC(icp6s_opacket_too_big);
192 		return;
193 	case ICMP6_TIME_EXCEEDED:
194 		switch (code) {
195 		case ICMP6_TIME_EXCEED_TRANSIT:
196 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
197 			return;
198 		case ICMP6_TIME_EXCEED_REASSEMBLY:
199 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
200 			return;
201 		}
202 		break;
203 	case ICMP6_PARAM_PROB:
204 		switch (code) {
205 		case ICMP6_PARAMPROB_HEADER:
206 			ICMP6STAT_INC(icp6s_oparamprob_header);
207 			return;
208 		case ICMP6_PARAMPROB_NEXTHEADER:
209 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
210 			return;
211 		case ICMP6_PARAMPROB_OPTION:
212 			ICMP6STAT_INC(icp6s_oparamprob_option);
213 			return;
214 		}
215 		break;
216 	case ND_REDIRECT:
217 		ICMP6STAT_INC(icp6s_oredirect);
218 		return;
219 	}
220 	ICMP6STAT_INC(icp6s_ounknown);
221 }
222 
223 /*
224  * A wrapper function for icmp6_error() necessary when the erroneous packet
225  * may not contain enough scope zone information.
226  */
227 void
228 icmp6_error2(struct mbuf *m, int type, int code, int param,
229     struct ifnet *ifp)
230 {
231 	struct ip6_hdr *ip6;
232 
233 	if (ifp == NULL)
234 		return;
235 
236 	if (m->m_len < sizeof(struct ip6_hdr)) {
237 		m = m_pullup(m, sizeof(struct ip6_hdr));
238 		if (m == NULL) {
239 			IP6STAT_INC(ip6s_exthdrtoolong);
240 			return;
241 		}
242 	}
243 	ip6 = mtod(m, struct ip6_hdr *);
244 
245 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
246 		return;
247 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
248 		return;
249 
250 	icmp6_error(m, type, code, param);
251 }
252 
253 /*
254  * Generate an error packet of type error in response to bad IP6 packet.
255  */
256 void
257 icmp6_error(struct mbuf *m, int type, int code, int param)
258 {
259 	struct ip6_hdr *oip6, *nip6;
260 	struct icmp6_hdr *icmp6;
261 	struct epoch_tracker et;
262 	u_int preplen;
263 	int off;
264 	int nxt;
265 
266 	ICMP6STAT_INC(icp6s_error);
267 
268 	/* count per-type-code statistics */
269 	icmp6_errcount(type, code);
270 
271 #ifdef M_DECRYPTED	/*not openbsd*/
272 	if (m->m_flags & M_DECRYPTED) {
273 		ICMP6STAT_INC(icp6s_canterror);
274 		goto freeit;
275 	}
276 #endif
277 
278 	if (m->m_len < sizeof(struct ip6_hdr)) {
279 		m = m_pullup(m, sizeof(struct ip6_hdr));
280 		if (m == NULL) {
281 			IP6STAT_INC(ip6s_exthdrtoolong);
282 			return;
283 		}
284 	}
285 	oip6 = mtod(m, struct ip6_hdr *);
286 
287 	/*
288 	 * If the destination address of the erroneous packet is a multicast
289 	 * address, or the packet was sent using link-layer multicast,
290 	 * we should basically suppress sending an error (RFC 2463, Section
291 	 * 2.4).
292 	 * We have two exceptions (the item e.2 in that section):
293 	 * - the Packet Too Big message can be sent for path MTU discovery.
294 	 * - the Parameter Problem Message that can be allowed an icmp6 error
295 	 *   in the option type field.  This check has been done in
296 	 *   ip6_unknown_opt(), so we can just check the type and code.
297 	 */
298 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
299 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
300 	    (type != ICMP6_PACKET_TOO_BIG &&
301 	     (type != ICMP6_PARAM_PROB ||
302 	      code != ICMP6_PARAMPROB_OPTION)))
303 		goto freeit;
304 
305 	/*
306 	 * RFC 2463, 2.4 (e.5): source address check.
307 	 * XXX: the case of anycast source?
308 	 */
309 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
310 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
311 		goto freeit;
312 
313 	/*
314 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
315 	 * don't do it.
316 	 */
317 	nxt = -1;
318 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
319 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
320 		struct icmp6_hdr *icp;
321 
322 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
323 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
324 			if (m == NULL) {
325 				IP6STAT_INC(ip6s_exthdrtoolong);
326 				return;
327 			}
328 		}
329 		oip6 = mtod(m, struct ip6_hdr *);
330 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
331 
332 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
333 		    icp->icmp6_type == ND_REDIRECT) {
334 			/*
335 			 * ICMPv6 error
336 			 * Special case: for redirect (which is
337 			 * informational) we must not send icmp6 error.
338 			 */
339 			ICMP6STAT_INC(icp6s_canterror);
340 			goto freeit;
341 		} else {
342 			/* ICMPv6 informational - send the error */
343 		}
344 	} else {
345 		/* non-ICMPv6 - send the error */
346 	}
347 
348 	/* Finally, do rate limitation check. */
349 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
350 		ICMP6STAT_INC(icp6s_toofreq);
351 		goto freeit;
352 	}
353 
354 	/*
355 	 * OK, ICMP6 can be generated.
356 	 */
357 
358 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
359 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
360 
361 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
362 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
363 	if (m == NULL) {
364 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
365 		return;
366 	}
367 
368 	nip6 = mtod(m, struct ip6_hdr *);
369 	nip6->ip6_src  = oip6->ip6_src;
370 	nip6->ip6_dst  = oip6->ip6_dst;
371 
372 	in6_clearscope(&oip6->ip6_src);
373 	in6_clearscope(&oip6->ip6_dst);
374 
375 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
376 	icmp6->icmp6_type = type;
377 	icmp6->icmp6_code = code;
378 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
379 
380 	ICMP6STAT_INC(icp6s_outhist[type]);
381 	NET_EPOCH_ENTER(et);
382 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
383 	NET_EPOCH_EXIT(et);
384 
385 	return;
386 
387   freeit:
388 	/*
389 	 * If we can't tell whether or not we can generate ICMP6, free it.
390 	 */
391 	m_freem(m);
392 }
393 
394 /*
395  * Process a received ICMP6 message.
396  */
397 int
398 icmp6_input(struct mbuf **mp, int *offp, int proto)
399 {
400 	struct mbuf *m, *n;
401 	struct ifnet *ifp;
402 	struct ip6_hdr *ip6, *nip6;
403 	struct icmp6_hdr *icmp6, *nicmp6;
404 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
405 	int code, error, icmp6len, ip6len, noff, off, sum;
406 
407 	NET_EPOCH_ASSERT();
408 
409 	m = *mp;
410 	off = *offp;
411 
412 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
413 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
414 		if (m == NULL) {
415 			IP6STAT_INC(ip6s_exthdrtoolong);
416 			*mp = m;
417 			return (IPPROTO_DONE);
418 		}
419 	}
420 
421 	/*
422 	 * Locate icmp6 structure in mbuf, and check
423 	 * that not corrupted and of at least minimum length
424 	 */
425 
426 	icmp6len = m->m_pkthdr.len - off;
427 	if (icmp6len < sizeof(struct icmp6_hdr)) {
428 		ICMP6STAT_INC(icp6s_tooshort);
429 		goto freeit;
430 	}
431 
432 	ip6 = mtod(m, struct ip6_hdr *);
433 	ifp = m->m_pkthdr.rcvif;
434 	/*
435 	 * Check multicast group membership.
436 	 * Note: SSM filters are not applied for ICMPv6 traffic.
437 	 */
438 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
439 		struct in6_multi	*inm;
440 
441 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
442 		if (inm == NULL) {
443 			IP6STAT_INC(ip6s_notmember);
444 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
445 			goto freeit;
446 		}
447 	}
448 
449 	/* Calculate the checksum. */
450 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
451 	code = icmp6->icmp6_code;
452 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
453 		nd6log((LOG_ERR,
454 		    "ICMP6 checksum error(%d|%x) %s\n",
455 		    icmp6->icmp6_type, sum,
456 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
457 		ICMP6STAT_INC(icp6s_checksum);
458 		goto freeit;
459 	}
460 
461 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
462 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
463 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
464 		icmp6_ifstat_inc(ifp, ifs6_in_error);
465 
466 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
467 	switch (icmp6->icmp6_type) {
468 	case ICMP6_DST_UNREACH:
469 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
470 		switch (code) {
471 		case ICMP6_DST_UNREACH_NOROUTE:
472 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
473 			code = PRC_UNREACH_NET;
474 			break;
475 		case ICMP6_DST_UNREACH_ADMIN:
476 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
477 			code = PRC_UNREACH_ADMIN_PROHIB;
478 			break;
479 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
480 			/* I mean "source address was incorrect." */
481 			code = PRC_PARAMPROB;
482 			break;
483 		case ICMP6_DST_UNREACH_NOPORT:
484 			code = PRC_UNREACH_PORT;
485 			break;
486 		default:
487 			goto badcode;
488 		}
489 		goto deliver;
490 		break;
491 
492 	case ICMP6_PACKET_TOO_BIG:
493 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
494 
495 		/* validation is made in icmp6_mtudisc_update */
496 
497 		code = PRC_MSGSIZE;
498 
499 		/*
500 		 * Updating the path MTU will be done after examining
501 		 * intermediate extension headers.
502 		 */
503 		goto deliver;
504 		break;
505 
506 	case ICMP6_TIME_EXCEEDED:
507 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
508 		switch (code) {
509 		case ICMP6_TIME_EXCEED_TRANSIT:
510 			code = PRC_TIMXCEED_INTRANS;
511 			break;
512 		case ICMP6_TIME_EXCEED_REASSEMBLY:
513 			code = PRC_TIMXCEED_REASS;
514 			break;
515 		default:
516 			goto badcode;
517 		}
518 		goto deliver;
519 		break;
520 
521 	case ICMP6_PARAM_PROB:
522 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
523 		switch (code) {
524 		case ICMP6_PARAMPROB_NEXTHEADER:
525 			code = PRC_UNREACH_PROTOCOL;
526 			break;
527 		case ICMP6_PARAMPROB_HEADER:
528 		case ICMP6_PARAMPROB_OPTION:
529 			code = PRC_PARAMPROB;
530 			break;
531 		default:
532 			goto badcode;
533 		}
534 		goto deliver;
535 		break;
536 
537 	case ICMP6_ECHO_REQUEST:
538 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
539 		if (code != 0)
540 			goto badcode;
541 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
542 			/* Give up remote */
543 			break;
544 		}
545 		if (!M_WRITABLE(n)
546 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
547 			struct mbuf *n0 = n;
548 			int n0len;
549 
550 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
551 			n = m_gethdr(M_NOWAIT, n0->m_type);
552 			if (n == NULL) {
553 				/* Give up remote */
554 				m_freem(n0);
555 				break;
556 			}
557 
558 			m_move_pkthdr(n, n0);	/* FIB copied. */
559 			n0len = n0->m_pkthdr.len;	/* save for use below */
560 			/*
561 			 * Copy IPv6 and ICMPv6 only.
562 			 */
563 			nip6 = mtod(n, struct ip6_hdr *);
564 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
565 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
566 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
567 			noff = sizeof(struct ip6_hdr);
568 			/* new mbuf contains only ipv6+icmpv6 headers */
569 			n->m_len = noff + sizeof(struct icmp6_hdr);
570 			/*
571 			 * Adjust mbuf.  ip6_plen will be adjusted in
572 			 * ip6_output().
573 			 */
574 			m_adj(n0, off + sizeof(struct icmp6_hdr));
575 			/* recalculate complete packet size */
576 			n->m_pkthdr.len = n0len + (noff - off);
577 			n->m_next = n0;
578 		} else {
579 			if (n->m_len < off + sizeof(*nicmp6)) {
580 				n = m_pullup(n, off + sizeof(*nicmp6));
581 				if (n == NULL) {
582 					IP6STAT_INC(ip6s_exthdrtoolong);
583 					break;
584 				}
585 			}
586 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
587 			noff = off;
588 		}
589 		if (n) {
590 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
591 			nicmp6->icmp6_code = 0;
592 			ICMP6STAT_INC(icp6s_reflect);
593 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
594 			icmp6_reflect(n, noff);
595 		}
596 		break;
597 
598 	case ICMP6_ECHO_REPLY:
599 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
600 		if (code != 0)
601 			goto badcode;
602 		break;
603 
604 	case MLD_LISTENER_QUERY:
605 	case MLD_LISTENER_REPORT:
606 	case MLD_LISTENER_DONE:
607 	case MLDV2_LISTENER_REPORT:
608 		/*
609 		 * Drop MLD traffic which is not link-local, has a hop limit
610 		 * of greater than 1 hop, or which does not have the
611 		 * IPv6 HBH Router Alert option.
612 		 * As IPv6 HBH options are stripped in ip6_input() we must
613 		 * check an mbuf header flag.
614 		 * XXX Should we also sanity check that these messages
615 		 * were directed to a link-local multicast prefix?
616 		 */
617 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
618 			goto freeit;
619 		if (mld_input(&m, off, icmp6len) != 0) {
620 			*mp = NULL;
621 			return (IPPROTO_DONE);
622 		}
623 		/* m stays. */
624 		break;
625 
626 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
627 	    {
628 		enum { WRU, FQDN } mode;
629 		struct prison *pr;
630 
631 		if (!V_icmp6_nodeinfo)
632 			break;
633 
634 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
635 			mode = WRU;
636 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
637 			mode = FQDN;
638 		else
639 			goto badlen;
640 
641 		pr = NULL;
642 		sx_slock(&allprison_lock);
643 		TAILQ_FOREACH(pr, &allprison, pr_list)
644 			if (pr->pr_vnet == ifp->if_vnet)
645 				break;
646 		sx_sunlock(&allprison_lock);
647 		if (pr == NULL)
648 			pr = curthread->td_ucred->cr_prison;
649 		if (mode == FQDN) {
650 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
651 				m = m_pullup(m, off +
652 				    sizeof(struct icmp6_nodeinfo));
653 				if (m == NULL) {
654 					IP6STAT_INC(ip6s_exthdrtoolong);
655 					*mp = m;
656 					return (IPPROTO_DONE);
657 				}
658 			}
659 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
660 			if (n)
661 				n = ni6_input(n, off, pr);
662 			/* XXX meaningless if n == NULL */
663 			noff = sizeof(struct ip6_hdr);
664 		} else {
665 			u_char *p;
666 			int maxhlen, hlen;
667 
668 			/*
669 			 * XXX: this combination of flags is pointless,
670 			 * but should we keep this for compatibility?
671 			 */
672 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
673 			    ICMP6_NODEINFO_TMPADDROK)) !=
674 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
675 				break;
676 
677 			if (code != 0)
678 				goto badcode;
679 
680 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
681 			n = m_gethdr(M_NOWAIT, m->m_type);
682 			if (n == NULL) {
683 				/* Give up remote */
684 				break;
685 			}
686 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
687 				/*
688 				 * Previous code did a blind M_COPY_PKTHDR
689 				 * and said "just for rcvif".  If true, then
690 				 * we could tolerate the dup failing (due to
691 				 * the deep copy of the tag chain).  For now
692 				 * be conservative and just fail.
693 				 */
694 				m_free(n);
695 				n = NULL;
696 				break;
697 			}
698 			/*
699 			 * Copy IPv6 and ICMPv6 only.
700 			 */
701 			nip6 = mtod(n, struct ip6_hdr *);
702 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
703 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
704 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
705 			p = (u_char *)(nicmp6 + 1);
706 			bzero(p, 4);
707 
708 			maxhlen = M_TRAILINGSPACE(n) -
709 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
710 			mtx_lock(&pr->pr_mtx);
711 			hlen = strlen(pr->pr_hostname);
712 			if (maxhlen > hlen)
713 				maxhlen = hlen;
714 			/* meaningless TTL */
715 			bcopy(pr->pr_hostname, p + 4, maxhlen);
716 			mtx_unlock(&pr->pr_mtx);
717 			noff = sizeof(struct ip6_hdr);
718 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
719 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
720 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
721 			nicmp6->icmp6_code = 0;
722 		}
723 		if (n) {
724 			ICMP6STAT_INC(icp6s_reflect);
725 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
726 			icmp6_reflect(n, noff);
727 		}
728 		break;
729 	    }
730 
731 	case ICMP6_WRUREPLY:
732 		if (code != 0)
733 			goto badcode;
734 		break;
735 
736 	case ND_ROUTER_SOLICIT:
737 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
738 		if (code != 0)
739 			goto badcode;
740 		if (icmp6len < sizeof(struct nd_router_solicit))
741 			goto badlen;
742 		if (send_sendso_input_hook != NULL) {
743 			if (m->m_len < off + icmp6len) {
744 				m = m_pullup(m, off + icmp6len);
745 				if (m == NULL) {
746 					IP6STAT_INC(ip6s_exthdrtoolong);
747 					*mp = NULL;
748 					return (IPPROTO_DONE);
749 				}
750 			}
751 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
752 			if (error == 0) {
753 				m = NULL;
754 				goto freeit;
755 			}
756 		}
757 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
758 		nd6_rs_input(m, off, icmp6len);
759 		m = n;
760 		if (m == NULL)
761 			goto freeit;
762 		break;
763 
764 	case ND_ROUTER_ADVERT:
765 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
766 		if (code != 0)
767 			goto badcode;
768 		if (icmp6len < sizeof(struct nd_router_advert))
769 			goto badlen;
770 		if (send_sendso_input_hook != NULL) {
771 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
772 			if (error == 0) {
773 				m = NULL;
774 				goto freeit;
775 			}
776 		}
777 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
778 		nd6_ra_input(m, off, icmp6len);
779 		m = n;
780 		if (m == NULL)
781 			goto freeit;
782 		break;
783 
784 	case ND_NEIGHBOR_SOLICIT:
785 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
786 		if (code != 0)
787 			goto badcode;
788 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
789 			goto badlen;
790 		if (send_sendso_input_hook != NULL) {
791 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
792 			if (error == 0) {
793 				m = NULL;
794 				goto freeit;
795 			}
796 		}
797 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
798 		nd6_ns_input(m, off, icmp6len);
799 		m = n;
800 		if (m == NULL)
801 			goto freeit;
802 		break;
803 
804 	case ND_NEIGHBOR_ADVERT:
805 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
806 		if (code != 0)
807 			goto badcode;
808 		if (icmp6len < sizeof(struct nd_neighbor_advert))
809 			goto badlen;
810 		if (send_sendso_input_hook != NULL) {
811 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
812 			if (error == 0) {
813 				m = NULL;
814 				goto freeit;
815 			}
816 		}
817 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
818 		nd6_na_input(m, off, icmp6len);
819 		m = n;
820 		if (m == NULL)
821 			goto freeit;
822 		break;
823 
824 	case ND_REDIRECT:
825 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
826 		if (code != 0)
827 			goto badcode;
828 		if (icmp6len < sizeof(struct nd_redirect))
829 			goto badlen;
830 		if (send_sendso_input_hook != NULL) {
831 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
832 			if (error == 0) {
833 				m = NULL;
834 				goto freeit;
835 			}
836 		}
837 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
838 		icmp6_redirect_input(m, off);
839 		m = n;
840 		if (m == NULL)
841 			goto freeit;
842 		break;
843 
844 	case ICMP6_ROUTER_RENUMBERING:
845 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
846 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
847 			goto badcode;
848 		if (icmp6len < sizeof(struct icmp6_router_renum))
849 			goto badlen;
850 		break;
851 
852 	default:
853 		nd6log((LOG_DEBUG,
854 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
855 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
856 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
857 		    ifp ? ifp->if_index : 0));
858 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
859 			/* ICMPv6 error: MUST deliver it by spec... */
860 			code = PRC_NCMDS;
861 			/* deliver */
862 		} else {
863 			/* ICMPv6 informational: MUST not deliver */
864 			break;
865 		}
866 	deliver:
867 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
868 			/* In this case, m should've been freed. */
869 			*mp = NULL;
870 			return (IPPROTO_DONE);
871 		}
872 		break;
873 
874 	badcode:
875 		ICMP6STAT_INC(icp6s_badcode);
876 		break;
877 
878 	badlen:
879 		ICMP6STAT_INC(icp6s_badlen);
880 		break;
881 	}
882 
883 	/* deliver the packet to appropriate sockets */
884 	icmp6_rip6_input(&m, *offp);
885 
886 	*mp = m;
887 	return (IPPROTO_DONE);
888 
889  freeit:
890 	m_freem(m);
891 	*mp = NULL;
892 	return (IPPROTO_DONE);
893 }
894 
895 static int
896 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
897 {
898 	struct mbuf *m;
899 	struct icmp6_hdr *icmp6;
900 	struct ip6_hdr *eip6;
901 	u_int32_t notifymtu;
902 	struct sockaddr_in6 icmp6src, icmp6dst;
903 
904 	m = *mp;
905 
906 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
907 		ICMP6STAT_INC(icp6s_tooshort);
908 		goto freeit;
909 	}
910 
911 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
912 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
913 		if (m == NULL) {
914 			IP6STAT_INC(ip6s_exthdrtoolong);
915 			*mp = m;
916 			return (-1);
917 		}
918 	}
919 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
920 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
921 	bzero(&icmp6dst, sizeof(icmp6dst));
922 
923 	/* Detect the upper level protocol */
924 	{
925 		u_int8_t nxt = eip6->ip6_nxt;
926 		int eoff = off + sizeof(struct icmp6_hdr) +
927 		    sizeof(struct ip6_hdr);
928 		struct ip6ctlparam ip6cp;
929 		int icmp6type = icmp6->icmp6_type;
930 		struct ip6_frag *fh;
931 		struct ip6_rthdr *rth;
932 		struct ip6_rthdr0 *rth0;
933 		int rthlen;
934 
935 		while (1) { /* XXX: should avoid infinite loop explicitly? */
936 			struct ip6_ext *eh;
937 
938 			switch (nxt) {
939 			case IPPROTO_HOPOPTS:
940 			case IPPROTO_DSTOPTS:
941 			case IPPROTO_AH:
942 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
943 					m = m_pullup(m, eoff +
944 					    sizeof(struct ip6_ext));
945 					if (m == NULL) {
946 						IP6STAT_INC(ip6s_exthdrtoolong);
947 						*mp = m;
948 						return (-1);
949 					}
950 				}
951 				eh = (struct ip6_ext *)
952 				    (mtod(m, caddr_t) + eoff);
953 				if (nxt == IPPROTO_AH)
954 					eoff += (eh->ip6e_len + 2) << 2;
955 				else
956 					eoff += (eh->ip6e_len + 1) << 3;
957 				nxt = eh->ip6e_nxt;
958 				break;
959 			case IPPROTO_ROUTING:
960 				/*
961 				 * When the erroneous packet contains a
962 				 * routing header, we should examine the
963 				 * header to determine the final destination.
964 				 * Otherwise, we can't properly update
965 				 * information that depends on the final
966 				 * destination (e.g. path MTU).
967 				 */
968 				if (m->m_len < eoff + sizeof(*rth)) {
969 					m = m_pullup(m, eoff + sizeof(*rth));
970 					if (m == NULL) {
971 						IP6STAT_INC(ip6s_exthdrtoolong);
972 						*mp = m;
973 						return (-1);
974 					}
975 				}
976 				rth = (struct ip6_rthdr *)
977 				    (mtod(m, caddr_t) + eoff);
978 				rthlen = (rth->ip6r_len + 1) << 3;
979 				/*
980 				 * XXX: currently there is no
981 				 * officially defined type other
982 				 * than type-0.
983 				 * Note that if the segment left field
984 				 * is 0, all intermediate hops must
985 				 * have been passed.
986 				 */
987 				if (rth->ip6r_segleft &&
988 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
989 					int hops;
990 
991 					if (m->m_len < eoff + rthlen) {
992 						m = m_pullup(m, eoff + rthlen);
993 						if (m == NULL) {
994 							IP6STAT_INC(
995 							    ip6s_exthdrtoolong);
996 							*mp = m;
997 							return (-1);
998 						}
999 					}
1000 					rth0 = (struct ip6_rthdr0 *)
1001 					    (mtod(m, caddr_t) + eoff);
1002 
1003 					/* just ignore a bogus header */
1004 					if ((rth0->ip6r0_len % 2) == 0 &&
1005 					    (hops = rth0->ip6r0_len/2))
1006 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1007 				}
1008 				eoff += rthlen;
1009 				nxt = rth->ip6r_nxt;
1010 				break;
1011 			case IPPROTO_FRAGMENT:
1012 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1013 					m = m_pullup(m, eoff +
1014 					    sizeof(struct ip6_frag));
1015 					if (m == NULL) {
1016 						IP6STAT_INC(ip6s_exthdrtoolong);
1017 						*mp = m;
1018 						return (-1);
1019 					}
1020 				}
1021 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1022 				    eoff);
1023 				/*
1024 				 * Data after a fragment header is meaningless
1025 				 * unless it is the first fragment, but
1026 				 * we'll go to the notify label for path MTU
1027 				 * discovery.
1028 				 */
1029 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1030 					goto notify;
1031 
1032 				eoff += sizeof(struct ip6_frag);
1033 				nxt = fh->ip6f_nxt;
1034 				break;
1035 			default:
1036 				/*
1037 				 * This case includes ESP and the No Next
1038 				 * Header.  In such cases going to the notify
1039 				 * label does not have any meaning
1040 				 * (i.e. ctlfunc will be NULL), but we go
1041 				 * anyway since we might have to update
1042 				 * path MTU information.
1043 				 */
1044 				goto notify;
1045 			}
1046 		}
1047 	  notify:
1048 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1049 
1050 		/*
1051 		 * retrieve parameters from the inner IPv6 header, and convert
1052 		 * them into sockaddr structures.
1053 		 * XXX: there is no guarantee that the source or destination
1054 		 * addresses of the inner packet are in the same scope as
1055 		 * the addresses of the icmp packet.  But there is no other
1056 		 * way to determine the zone.
1057 		 */
1058 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1059 
1060 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1061 		icmp6dst.sin6_family = AF_INET6;
1062 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1063 			icmp6dst.sin6_addr = eip6->ip6_dst;
1064 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1065 			goto freeit;
1066 		bzero(&icmp6src, sizeof(icmp6src));
1067 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1068 		icmp6src.sin6_family = AF_INET6;
1069 		icmp6src.sin6_addr = eip6->ip6_src;
1070 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1071 			goto freeit;
1072 		icmp6src.sin6_flowinfo =
1073 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1074 
1075 		ip6cp.ip6c_m = m;
1076 		ip6cp.ip6c_icmp6 = icmp6;
1077 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1078 		ip6cp.ip6c_off = eoff;
1079 		ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1080 		ip6cp.ip6c_src = &icmp6src;
1081 		ip6cp.ip6c_nxt = nxt;
1082 
1083 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1084 			notifymtu = ntohl(icmp6->icmp6_mtu);
1085 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1086 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1087 		}
1088 
1089 		if (ip6_ctlprotox[nxt] != NULL)
1090 			ip6_ctlprotox[nxt](code, (struct sockaddr *)&icmp6dst,
1091 			    &ip6cp);
1092 	}
1093 	*mp = m;
1094 	return (0);
1095 
1096   freeit:
1097 	m_freem(m);
1098 	*mp = NULL;
1099 	return (-1);
1100 }
1101 
1102 void
1103 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1104 {
1105 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1106 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1107 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1108 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1109 	struct in_conninfo inc;
1110 	uint32_t max_mtu;
1111 
1112 #if 0
1113 	/*
1114 	 * RFC2460 section 5, last paragraph.
1115 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1116 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1117 	 * due to packet translator in the middle.
1118 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1119 	 * special handling.
1120 	 */
1121 	if (mtu < IPV6_MMTU)
1122 		return;
1123 #endif
1124 
1125 	/*
1126 	 * we reject ICMPv6 too big with abnormally small value.
1127 	 * XXX what is the good definition of "abnormally small"?
1128 	 */
1129 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1130 		return;
1131 
1132 	if (!validated)
1133 		return;
1134 
1135 	/*
1136 	 * In case the suggested mtu is less than IPV6_MMTU, we
1137 	 * only need to remember that it was for above mentioned
1138 	 * "alwaysfrag" case.
1139 	 * Try to be as close to the spec as possible.
1140 	 */
1141 	if (mtu < IPV6_MMTU)
1142 		mtu = IPV6_MMTU - 8;
1143 
1144 	bzero(&inc, sizeof(inc));
1145 	inc.inc_fibnum = M_GETFIB(m);
1146 	inc.inc_flags |= INC_ISIPV6;
1147 	inc.inc6_faddr = *dst;
1148 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1149 		return;
1150 
1151 	max_mtu = tcp_hc_getmtu(&inc);
1152 	if (max_mtu == 0)
1153 		max_mtu = tcp_maxmtu6(&inc, NULL);
1154 
1155 	if (mtu < max_mtu) {
1156 		tcp_hc_updatemtu(&inc, mtu);
1157 		ICMP6STAT_INC(icp6s_pmtuchg);
1158 	}
1159 }
1160 
1161 /*
1162  * Process a Node Information Query packet, based on
1163  * draft-ietf-ipngwg-icmp-name-lookups-07.
1164  *
1165  * Spec incompatibilities:
1166  * - IPv6 Subject address handling
1167  * - IPv4 Subject address handling support missing
1168  * - Proxy reply (answer even if it's not for me)
1169  * - joins NI group address at in6_ifattach() time only, does not cope
1170  *   with hostname changes by sethostname(3)
1171  */
1172 static struct mbuf *
1173 ni6_input(struct mbuf *m, int off, struct prison *pr)
1174 {
1175 	struct icmp6_nodeinfo *ni6, *nni6;
1176 	struct mbuf *n = NULL;
1177 	u_int16_t qtype;
1178 	int subjlen;
1179 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1180 	struct ni_reply_fqdn *fqdn;
1181 	int addrs;		/* for NI_QTYPE_NODEADDR */
1182 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1183 	struct in6_addr in6_subj; /* subject address */
1184 	struct ip6_hdr *ip6;
1185 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1186 	char *subj = NULL;
1187 	struct in6_ifaddr *ia6 = NULL;
1188 
1189 	ip6 = mtod(m, struct ip6_hdr *);
1190 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1191 
1192 	/*
1193 	 * Validate IPv6 source address.
1194 	 * The default configuration MUST be to refuse answering queries from
1195 	 * global-scope addresses according to RFC4602.
1196 	 * Notes:
1197 	 *  - it's not very clear what "refuse" means; this implementation
1198 	 *    simply drops it.
1199 	 *  - it's not very easy to identify global-scope (unicast) addresses
1200 	 *    since there are many prefixes for them.  It should be safer
1201 	 *    and in practice sufficient to check "all" but loopback and
1202 	 *    link-local (note that site-local unicast was deprecated and
1203 	 *    ULA is defined as global scope-wise)
1204 	 */
1205 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1206 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1207 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1208 		goto bad;
1209 
1210 	/*
1211 	 * Validate IPv6 destination address.
1212 	 *
1213 	 * The Responder must discard the Query without further processing
1214 	 * unless it is one of the Responder's unicast or anycast addresses, or
1215 	 * a link-local scope multicast address which the Responder has joined.
1216 	 * [RFC4602, Section 5.]
1217 	 */
1218 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1219 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1220 			goto bad;
1221 		/* else it's a link-local multicast, fine */
1222 	} else {		/* unicast or anycast */
1223 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1224 		if (ia6 == NULL)
1225 			goto bad; /* XXX impossible */
1226 
1227 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1228 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1229 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1230 				"a temporary address in %s:%d",
1231 			       __FILE__, __LINE__));
1232 			goto bad;
1233 		}
1234 	}
1235 
1236 	/* validate query Subject field. */
1237 	qtype = ntohs(ni6->ni_qtype);
1238 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1239 	switch (qtype) {
1240 	case NI_QTYPE_NOOP:
1241 	case NI_QTYPE_SUPTYPES:
1242 		/* 07 draft */
1243 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1244 			break;
1245 		/* FALLTHROUGH */
1246 	case NI_QTYPE_FQDN:
1247 	case NI_QTYPE_NODEADDR:
1248 	case NI_QTYPE_IPV4ADDR:
1249 		switch (ni6->ni_code) {
1250 		case ICMP6_NI_SUBJ_IPV6:
1251 #if ICMP6_NI_SUBJ_IPV6 != 0
1252 		case 0:
1253 #endif
1254 			/*
1255 			 * backward compatibility - try to accept 03 draft
1256 			 * format, where no Subject is present.
1257 			 */
1258 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1259 			    subjlen == 0) {
1260 				oldfqdn++;
1261 				break;
1262 			}
1263 #if ICMP6_NI_SUBJ_IPV6 != 0
1264 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1265 				goto bad;
1266 #endif
1267 
1268 			if (subjlen != sizeof(struct in6_addr))
1269 				goto bad;
1270 
1271 			/*
1272 			 * Validate Subject address.
1273 			 *
1274 			 * Not sure what exactly "address belongs to the node"
1275 			 * means in the spec, is it just unicast, or what?
1276 			 *
1277 			 * At this moment we consider Subject address as
1278 			 * "belong to the node" if the Subject address equals
1279 			 * to the IPv6 destination address; validation for
1280 			 * IPv6 destination address should have done enough
1281 			 * check for us.
1282 			 *
1283 			 * We do not do proxy at this moment.
1284 			 */
1285 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1286 			    subjlen, (caddr_t)&in6_subj);
1287 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1288 				goto bad;
1289 
1290 			subj = (char *)&in6_subj;
1291 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1292 				break;
1293 
1294 			/*
1295 			 * XXX if we are to allow other cases, we should really
1296 			 * be careful about scope here.
1297 			 * basically, we should disallow queries toward IPv6
1298 			 * destination X with subject Y,
1299 			 * if scope(X) > scope(Y).
1300 			 * if we allow scope(X) > scope(Y), it will result in
1301 			 * information leakage across scope boundary.
1302 			 */
1303 			goto bad;
1304 
1305 		case ICMP6_NI_SUBJ_FQDN:
1306 			/*
1307 			 * Validate Subject name with gethostname(3).
1308 			 *
1309 			 * The behavior may need some debate, since:
1310 			 * - we are not sure if the node has FQDN as
1311 			 *   hostname (returned by gethostname(3)).
1312 			 * - the code does wildcard match for truncated names.
1313 			 *   however, we are not sure if we want to perform
1314 			 *   wildcard match, if gethostname(3) side has
1315 			 *   truncated hostname.
1316 			 */
1317 			mtx_lock(&pr->pr_mtx);
1318 			n = ni6_nametodns(pr->pr_hostname,
1319 			    strlen(pr->pr_hostname), 0);
1320 			mtx_unlock(&pr->pr_mtx);
1321 			if (!n || n->m_next || n->m_len == 0)
1322 				goto bad;
1323 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1324 			    subjlen) {
1325 				m = m_pullup(m, off +
1326 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1327 				if (m == NULL) {
1328 					IP6STAT_INC(ip6s_exthdrtoolong);
1329 					goto bad;
1330 				}
1331 			}
1332 			/* ip6 possibly invalid but not used after. */
1333 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1334 			subj = (char *)(mtod(m, caddr_t) + off +
1335 			    sizeof(struct icmp6_nodeinfo));
1336 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1337 			    n->m_len)) {
1338 				goto bad;
1339 			}
1340 			m_freem(n);
1341 			n = NULL;
1342 			break;
1343 
1344 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1345 		default:
1346 			goto bad;
1347 		}
1348 		break;
1349 	}
1350 
1351 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1352 	switch (qtype) {
1353 	case NI_QTYPE_FQDN:
1354 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1355 			goto bad;
1356 		break;
1357 	case NI_QTYPE_NODEADDR:
1358 	case NI_QTYPE_IPV4ADDR:
1359 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1360 			goto bad;
1361 		break;
1362 	}
1363 
1364 	/* guess reply length */
1365 	switch (qtype) {
1366 	case NI_QTYPE_NOOP:
1367 		break;		/* no reply data */
1368 	case NI_QTYPE_SUPTYPES:
1369 		replylen += sizeof(u_int32_t);
1370 		break;
1371 	case NI_QTYPE_FQDN:
1372 		/* XXX will append an mbuf */
1373 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1374 		break;
1375 	case NI_QTYPE_NODEADDR:
1376 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1377 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1378 		    sizeof(u_int32_t))) > MCLBYTES)
1379 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1380 		break;
1381 	case NI_QTYPE_IPV4ADDR:
1382 		/* unsupported - should respond with unknown Qtype? */
1383 		break;
1384 	default:
1385 		/*
1386 		 * XXX: We must return a reply with the ICMP6 code
1387 		 * `unknown Qtype' in this case.  However we regard the case
1388 		 * as an FQDN query for backward compatibility.
1389 		 * Older versions set a random value to this field,
1390 		 * so it rarely varies in the defined qtypes.
1391 		 * But the mechanism is not reliable...
1392 		 * maybe we should obsolete older versions.
1393 		 */
1394 		qtype = NI_QTYPE_FQDN;
1395 		/* XXX will append an mbuf */
1396 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1397 		oldfqdn++;
1398 		break;
1399 	}
1400 
1401 	/* Allocate an mbuf to reply. */
1402 	if (replylen > MCLBYTES) {
1403 		/*
1404 		 * XXX: should we try to allocate more? But MCLBYTES
1405 		 * is probably much larger than IPV6_MMTU...
1406 		 */
1407 		goto bad;
1408 	}
1409 	if (replylen > MHLEN)
1410 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1411 	else
1412 		n = m_gethdr(M_NOWAIT, m->m_type);
1413 	if (n == NULL) {
1414 		m_freem(m);
1415 		return (NULL);
1416 	}
1417 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1418 	n->m_pkthdr.len = n->m_len = replylen;
1419 
1420 	/* copy mbuf header and IPv6 + Node Information base headers */
1421 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1422 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1423 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1424 
1425 	/* qtype dependent procedure */
1426 	switch (qtype) {
1427 	case NI_QTYPE_NOOP:
1428 		nni6->ni_code = ICMP6_NI_SUCCESS;
1429 		nni6->ni_flags = 0;
1430 		break;
1431 	case NI_QTYPE_SUPTYPES:
1432 	{
1433 		u_int32_t v;
1434 		nni6->ni_code = ICMP6_NI_SUCCESS;
1435 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1436 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1437 		v = (u_int32_t)htonl(0x0000000f);
1438 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1439 		break;
1440 	}
1441 	case NI_QTYPE_FQDN:
1442 		nni6->ni_code = ICMP6_NI_SUCCESS;
1443 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1444 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1445 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1446 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1447 		/*
1448 		 * XXX do we really have FQDN in hostname?
1449 		 */
1450 		mtx_lock(&pr->pr_mtx);
1451 		n->m_next = ni6_nametodns(pr->pr_hostname,
1452 		    strlen(pr->pr_hostname), oldfqdn);
1453 		mtx_unlock(&pr->pr_mtx);
1454 		if (n->m_next == NULL)
1455 			goto bad;
1456 		/* XXX we assume that n->m_next is not a chain */
1457 		if (n->m_next->m_next != NULL)
1458 			goto bad;
1459 		n->m_pkthdr.len += n->m_next->m_len;
1460 		break;
1461 	case NI_QTYPE_NODEADDR:
1462 	{
1463 		int lenlim, copied;
1464 
1465 		nni6->ni_code = ICMP6_NI_SUCCESS;
1466 		n->m_pkthdr.len = n->m_len =
1467 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1468 		lenlim = M_TRAILINGSPACE(n);
1469 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1470 		/* XXX: reset mbuf length */
1471 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1472 		    sizeof(struct icmp6_nodeinfo) + copied;
1473 		break;
1474 	}
1475 	default:
1476 		break;		/* XXX impossible! */
1477 	}
1478 
1479 	nni6->ni_type = ICMP6_NI_REPLY;
1480 	m_freem(m);
1481 	return (n);
1482 
1483   bad:
1484 	m_freem(m);
1485 	if (n)
1486 		m_freem(n);
1487 	return (NULL);
1488 }
1489 
1490 /*
1491  * make a mbuf with DNS-encoded string.  no compression support.
1492  *
1493  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1494  * treated as truncated name (two \0 at the end).  this is a wild guess.
1495  *
1496  * old - return pascal string if non-zero
1497  */
1498 static struct mbuf *
1499 ni6_nametodns(const char *name, int namelen, int old)
1500 {
1501 	struct mbuf *m;
1502 	char *cp, *ep;
1503 	const char *p, *q;
1504 	int i, len, nterm;
1505 
1506 	if (old)
1507 		len = namelen + 1;
1508 	else
1509 		len = MCLBYTES;
1510 
1511 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1512 	if (len > MLEN)
1513 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1514 	else
1515 		m = m_get(M_NOWAIT, MT_DATA);
1516 	if (m == NULL)
1517 		goto fail;
1518 
1519 	if (old) {
1520 		m->m_len = len;
1521 		*mtod(m, char *) = namelen;
1522 		bcopy(name, mtod(m, char *) + 1, namelen);
1523 		return m;
1524 	} else {
1525 		m->m_len = 0;
1526 		cp = mtod(m, char *);
1527 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1528 
1529 		/* if not certain about my name, return empty buffer */
1530 		if (namelen == 0)
1531 			return m;
1532 
1533 		/*
1534 		 * guess if it looks like shortened hostname, or FQDN.
1535 		 * shortened hostname needs two trailing "\0".
1536 		 */
1537 		i = 0;
1538 		for (p = name; p < name + namelen; p++) {
1539 			if (*p && *p == '.')
1540 				i++;
1541 		}
1542 		if (i < 2)
1543 			nterm = 2;
1544 		else
1545 			nterm = 1;
1546 
1547 		p = name;
1548 		while (cp < ep && p < name + namelen) {
1549 			i = 0;
1550 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1551 				i++;
1552 			/* result does not fit into mbuf */
1553 			if (cp + i + 1 >= ep)
1554 				goto fail;
1555 			/*
1556 			 * DNS label length restriction, RFC1035 page 8.
1557 			 * "i == 0" case is included here to avoid returning
1558 			 * 0-length label on "foo..bar".
1559 			 */
1560 			if (i <= 0 || i >= 64)
1561 				goto fail;
1562 			*cp++ = i;
1563 			bcopy(p, cp, i);
1564 			cp += i;
1565 			p = q;
1566 			if (p < name + namelen && *p == '.')
1567 				p++;
1568 		}
1569 		/* termination */
1570 		if (cp + nterm >= ep)
1571 			goto fail;
1572 		while (nterm-- > 0)
1573 			*cp++ = '\0';
1574 		m->m_len = cp - mtod(m, char *);
1575 		return m;
1576 	}
1577 
1578 	panic("should not reach here");
1579 	/* NOTREACHED */
1580 
1581  fail:
1582 	if (m)
1583 		m_freem(m);
1584 	return NULL;
1585 }
1586 
1587 /*
1588  * check if two DNS-encoded string matches.  takes care of truncated
1589  * form (with \0\0 at the end).  no compression support.
1590  * XXX upper/lowercase match (see RFC2065)
1591  */
1592 static int
1593 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1594 {
1595 	const char *a0, *b0;
1596 	int l;
1597 
1598 	/* simplest case - need validation? */
1599 	if (alen == blen && bcmp(a, b, alen) == 0)
1600 		return 1;
1601 
1602 	a0 = a;
1603 	b0 = b;
1604 
1605 	/* termination is mandatory */
1606 	if (alen < 2 || blen < 2)
1607 		return 0;
1608 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1609 		return 0;
1610 	alen--;
1611 	blen--;
1612 
1613 	while (a - a0 < alen && b - b0 < blen) {
1614 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1615 			return 0;
1616 
1617 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1618 			return 0;
1619 		/* we don't support compression yet */
1620 		if (a[0] >= 64 || b[0] >= 64)
1621 			return 0;
1622 
1623 		/* truncated case */
1624 		if (a[0] == 0 && a - a0 == alen - 1)
1625 			return 1;
1626 		if (b[0] == 0 && b - b0 == blen - 1)
1627 			return 1;
1628 		if (a[0] == 0 || b[0] == 0)
1629 			return 0;
1630 
1631 		if (a[0] != b[0])
1632 			return 0;
1633 		l = a[0];
1634 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1635 			return 0;
1636 		if (bcmp(a + 1, b + 1, l) != 0)
1637 			return 0;
1638 
1639 		a += 1 + l;
1640 		b += 1 + l;
1641 	}
1642 
1643 	if (a - a0 == alen && b - b0 == blen)
1644 		return 1;
1645 	else
1646 		return 0;
1647 }
1648 
1649 /*
1650  * calculate the number of addresses to be returned in the node info reply.
1651  */
1652 static int
1653 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1654     struct in6_addr *subj)
1655 {
1656 	struct ifnet *ifp;
1657 	struct in6_ifaddr *ifa6;
1658 	struct ifaddr *ifa;
1659 	int addrs = 0, addrsofif, iffound = 0;
1660 	int niflags = ni6->ni_flags;
1661 
1662 	NET_EPOCH_ASSERT();
1663 
1664 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1665 		switch (ni6->ni_code) {
1666 		case ICMP6_NI_SUBJ_IPV6:
1667 			if (subj == NULL) /* must be impossible... */
1668 				return (0);
1669 			break;
1670 		default:
1671 			/*
1672 			 * XXX: we only support IPv6 subject address for
1673 			 * this Qtype.
1674 			 */
1675 			return (0);
1676 		}
1677 	}
1678 
1679 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1680 		addrsofif = 0;
1681 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1682 			if (ifa->ifa_addr->sa_family != AF_INET6)
1683 				continue;
1684 			ifa6 = (struct in6_ifaddr *)ifa;
1685 
1686 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1687 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1688 				iffound = 1;
1689 
1690 			/*
1691 			 * IPv4-mapped addresses can only be returned by a
1692 			 * Node Information proxy, since they represent
1693 			 * addresses of IPv4-only nodes, which perforce do
1694 			 * not implement this protocol.
1695 			 * [icmp-name-lookups-07, Section 5.4]
1696 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1697 			 * this function at this moment.
1698 			 */
1699 
1700 			/* What do we have to do about ::1? */
1701 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1702 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1703 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1704 					continue;
1705 				break;
1706 			case IPV6_ADDR_SCOPE_SITELOCAL:
1707 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1708 					continue;
1709 				break;
1710 			case IPV6_ADDR_SCOPE_GLOBAL:
1711 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1712 					continue;
1713 				break;
1714 			default:
1715 				continue;
1716 			}
1717 
1718 			/*
1719 			 * check if anycast is okay.
1720 			 * XXX: just experimental.  not in the spec.
1721 			 */
1722 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1723 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1724 				continue; /* we need only unicast addresses */
1725 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1726 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1727 				continue;
1728 			}
1729 			addrsofif++; /* count the address */
1730 		}
1731 		if (iffound) {
1732 			*ifpp = ifp;
1733 			return (addrsofif);
1734 		}
1735 
1736 		addrs += addrsofif;
1737 	}
1738 
1739 	return (addrs);
1740 }
1741 
1742 static int
1743 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1744     struct ifnet *ifp0, int resid)
1745 {
1746 	struct ifnet *ifp;
1747 	struct in6_ifaddr *ifa6;
1748 	struct ifaddr *ifa;
1749 	struct ifnet *ifp_dep = NULL;
1750 	int copied = 0, allow_deprecated = 0;
1751 	u_char *cp = (u_char *)(nni6 + 1);
1752 	int niflags = ni6->ni_flags;
1753 	u_int32_t ltime;
1754 
1755 	NET_EPOCH_ASSERT();
1756 
1757 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1758 		return (0);	/* needless to copy */
1759 
1760 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1761   again:
1762 
1763 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1764 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1765 			if (ifa->ifa_addr->sa_family != AF_INET6)
1766 				continue;
1767 			ifa6 = (struct in6_ifaddr *)ifa;
1768 
1769 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1770 			    allow_deprecated == 0) {
1771 				/*
1772 				 * prefererred address should be put before
1773 				 * deprecated addresses.
1774 				 */
1775 
1776 				/* record the interface for later search */
1777 				if (ifp_dep == NULL)
1778 					ifp_dep = ifp;
1779 
1780 				continue;
1781 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1782 			    allow_deprecated != 0)
1783 				continue; /* we now collect deprecated addrs */
1784 
1785 			/* What do we have to do about ::1? */
1786 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1787 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1788 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1789 					continue;
1790 				break;
1791 			case IPV6_ADDR_SCOPE_SITELOCAL:
1792 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1793 					continue;
1794 				break;
1795 			case IPV6_ADDR_SCOPE_GLOBAL:
1796 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1797 					continue;
1798 				break;
1799 			default:
1800 				continue;
1801 			}
1802 
1803 			/*
1804 			 * check if anycast is okay.
1805 			 * XXX: just experimental.  not in the spec.
1806 			 */
1807 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1808 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1809 				continue;
1810 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1811 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1812 				continue;
1813 			}
1814 
1815 			/* now we can copy the address */
1816 			if (resid < sizeof(struct in6_addr) +
1817 			    sizeof(u_int32_t)) {
1818 				/*
1819 				 * We give up much more copy.
1820 				 * Set the truncate flag and return.
1821 				 */
1822 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1823 				return (copied);
1824 			}
1825 
1826 			/*
1827 			 * Set the TTL of the address.
1828 			 * The TTL value should be one of the following
1829 			 * according to the specification:
1830 			 *
1831 			 * 1. The remaining lifetime of a DHCP lease on the
1832 			 *    address, or
1833 			 * 2. The remaining Valid Lifetime of a prefix from
1834 			 *    which the address was derived through Stateless
1835 			 *    Autoconfiguration.
1836 			 *
1837 			 * Note that we currently do not support stateful
1838 			 * address configuration by DHCPv6, so the former
1839 			 * case can't happen.
1840 			 */
1841 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1842 				ltime = ND6_INFINITE_LIFETIME;
1843 			else {
1844 				if (ifa6->ia6_lifetime.ia6t_expire >
1845 				    time_uptime)
1846 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1847 				else
1848 					ltime = 0;
1849 			}
1850 
1851 			bcopy(&ltime, cp, sizeof(u_int32_t));
1852 			cp += sizeof(u_int32_t);
1853 
1854 			/* copy the address itself */
1855 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1856 			    sizeof(struct in6_addr));
1857 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1858 			cp += sizeof(struct in6_addr);
1859 
1860 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1861 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1862 		}
1863 		if (ifp0)	/* we need search only on the specified IF */
1864 			break;
1865 	}
1866 
1867 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1868 		ifp = ifp_dep;
1869 		allow_deprecated = 1;
1870 
1871 		goto again;
1872 	}
1873 
1874 	return (copied);
1875 }
1876 
1877 static bool
1878 icmp6_rip6_match(const struct inpcb *inp, void *v)
1879 {
1880 	struct ip6_hdr *ip6 = v;
1881 
1882 	if ((inp->inp_vflag & INP_IPV6) == 0)
1883 		return (false);
1884 	if (inp->inp_ip_p != IPPROTO_ICMPV6)
1885 		return (false);
1886 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1887 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1888 		return (false);
1889 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1890 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1891 		return (false);
1892 	return (true);
1893 }
1894 
1895 /*
1896  * XXX almost dup'ed code with rip6_input.
1897  */
1898 static int
1899 icmp6_rip6_input(struct mbuf **mp, int off)
1900 {
1901 	struct mbuf *n, *m = *mp;
1902 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1903 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1904 	    INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1905 	struct inpcb *inp;
1906 	struct sockaddr_in6 fromsa;
1907 	struct icmp6_hdr *icmp6;
1908 	struct mbuf *opts = NULL;
1909 	int delivered = 0;
1910 
1911 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1912 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1913 
1914 	/*
1915 	 * XXX: the address may have embedded scope zone ID, which should be
1916 	 * hidden from applications.
1917 	 */
1918 	bzero(&fromsa, sizeof(fromsa));
1919 	fromsa.sin6_family = AF_INET6;
1920 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1921 	fromsa.sin6_addr = ip6->ip6_src;
1922 	if (sa6_recoverscope(&fromsa)) {
1923 		m_freem(m);
1924 		*mp = NULL;
1925 		return (IPPROTO_DONE);
1926 	}
1927 
1928 	while ((inp = inp_next(&inpi)) != NULL) {
1929 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1930 		    inp->in6p_icmp6filt))
1931 			continue;
1932 		/*
1933 		 * Recent network drivers tend to allocate a single
1934 		 * mbuf cluster, rather than to make a couple of
1935 		 * mbufs without clusters.  Also, since the IPv6 code
1936 		 * path tries to avoid m_pullup(), it is highly
1937 		 * probable that we still have an mbuf cluster here
1938 		 * even though the necessary length can be stored in an
1939 		 * mbuf's internal buffer.
1940 		 * Meanwhile, the default size of the receive socket
1941 		 * buffer for raw sockets is not so large.  This means
1942 		 * the possibility of packet loss is relatively higher
1943 		 * than before.  To avoid this scenario, we copy the
1944 		 * received data to a separate mbuf that does not use
1945 		 * a cluster, if possible.
1946 		 * XXX: it is better to copy the data after stripping
1947 		 * intermediate headers.
1948 		 */
1949 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1950 		    m->m_len <= MHLEN) {
1951 			n = m_get(M_NOWAIT, m->m_type);
1952 			if (n != NULL) {
1953 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1954 					bcopy(m->m_data, n->m_data, m->m_len);
1955 					n->m_len = m->m_len;
1956 				} else {
1957 					m_free(n);
1958 					n = NULL;
1959 				}
1960 			}
1961 		} else
1962 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1963 		if (n == NULL)
1964 			continue;
1965 		if (inp->inp_flags & INP_CONTROLOPTS)
1966 			ip6_savecontrol(inp, n, &opts);
1967 		/* strip intermediate headers */
1968 		m_adj(n, off);
1969 		SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1970 		if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1971 		    (struct sockaddr *)&fromsa, n, opts) == 0) {
1972 			soroverflow_locked(inp->inp_socket);
1973 			m_freem(n);
1974 			if (opts)
1975 				m_freem(opts);
1976 		} else {
1977 			sorwakeup_locked(inp->inp_socket);
1978 			delivered++;
1979 		}
1980 		opts = NULL;
1981 	}
1982 	m_freem(m);
1983 	*mp = NULL;
1984 	if (delivered == 0)
1985 		IP6STAT_DEC(ip6s_delivered);
1986 	return (IPPROTO_DONE);
1987 }
1988 
1989 /*
1990  * Reflect the ip6 packet back to the source.
1991  * OFF points to the icmp6 header, counted from the top of the mbuf.
1992  */
1993 static void
1994 icmp6_reflect(struct mbuf *m, size_t off)
1995 {
1996 	struct in6_addr src6, *srcp;
1997 	struct ip6_hdr *ip6;
1998 	struct icmp6_hdr *icmp6;
1999 	struct in6_ifaddr *ia = NULL;
2000 	struct ifnet *outif = NULL;
2001 	int plen;
2002 	int type, code, hlim;
2003 
2004 	/* too short to reflect */
2005 	if (off < sizeof(struct ip6_hdr)) {
2006 		nd6log((LOG_DEBUG,
2007 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2008 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2009 		    __FILE__, __LINE__));
2010 		goto bad;
2011 	}
2012 
2013 	/*
2014 	 * If there are extra headers between IPv6 and ICMPv6, strip
2015 	 * off that header first.
2016 	 */
2017 #ifdef DIAGNOSTIC
2018 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2019 		panic("assumption failed in icmp6_reflect");
2020 #endif
2021 	if (off > sizeof(struct ip6_hdr)) {
2022 		size_t l;
2023 		struct ip6_hdr nip6;
2024 
2025 		l = off - sizeof(struct ip6_hdr);
2026 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2027 		m_adj(m, l);
2028 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2029 		if (m->m_len < l) {
2030 			if ((m = m_pullup(m, l)) == NULL)
2031 				return;
2032 		}
2033 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2034 	} else /* off == sizeof(struct ip6_hdr) */ {
2035 		size_t l;
2036 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2037 		if (m->m_len < l) {
2038 			if ((m = m_pullup(m, l)) == NULL)
2039 				return;
2040 		}
2041 	}
2042 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2043 	ip6 = mtod(m, struct ip6_hdr *);
2044 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2045 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2046 	type = icmp6->icmp6_type; /* keep type for statistics */
2047 	code = icmp6->icmp6_code; /* ditto. */
2048 	hlim = 0;
2049 	srcp = NULL;
2050 
2051 	/*
2052 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2053 	 * use dst as the src for the reply.
2054 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2055 	 * (for example) when we encounter an error while forwarding procedure
2056 	 * destined to a duplicated address of ours.
2057 	 */
2058 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2059 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2060 		if (ia != NULL && !(ia->ia6_flags &
2061 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2062 			src6 = ia->ia_addr.sin6_addr;
2063 			srcp = &src6;
2064 
2065 			if (m->m_pkthdr.rcvif != NULL) {
2066 				/* XXX: This may not be the outgoing interface */
2067 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2068 			} else
2069 				hlim = V_ip6_defhlim;
2070 		}
2071 	}
2072 
2073 	if (srcp == NULL) {
2074 		int error;
2075 		struct in6_addr dst6;
2076 		uint32_t scopeid;
2077 
2078 		/*
2079 		 * This case matches to multicasts, our anycast, or unicasts
2080 		 * that we do not own.  Select a source address based on the
2081 		 * source address of the erroneous packet.
2082 		 */
2083 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2084 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2085 		    scopeid, NULL, &src6, &hlim);
2086 
2087 		if (error) {
2088 			char ip6buf[INET6_ADDRSTRLEN];
2089 			nd6log((LOG_DEBUG,
2090 			    "icmp6_reflect: source can't be determined: "
2091 			    "dst=%s, error=%d\n",
2092 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2093 			goto bad;
2094 		}
2095 		srcp = &src6;
2096 	}
2097 	/*
2098 	 * ip6_input() drops a packet if its src is multicast.
2099 	 * So, the src is never multicast.
2100 	 */
2101 	ip6->ip6_dst = ip6->ip6_src;
2102 	ip6->ip6_src = *srcp;
2103 	ip6->ip6_flow = 0;
2104 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2105 	ip6->ip6_vfc |= IPV6_VERSION;
2106 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2107 	ip6->ip6_hlim = hlim;
2108 
2109 	icmp6->icmp6_cksum = 0;
2110 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2111 	    sizeof(struct ip6_hdr), plen);
2112 
2113 	/*
2114 	 * XXX option handling
2115 	 */
2116 
2117 	m->m_flags &= ~(M_BCAST|M_MCAST);
2118 	m->m_pkthdr.rcvif = NULL;
2119 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2120 	if (outif)
2121 		icmp6_ifoutstat_inc(outif, type, code);
2122 
2123 	return;
2124 
2125  bad:
2126 	m_freem(m);
2127 	return;
2128 }
2129 
2130 static const char *
2131 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2132     struct in6_addr *tgt6)
2133 {
2134 	static char buf[1024];
2135 	char ip6bufs[INET6_ADDRSTRLEN];
2136 	char ip6bufd[INET6_ADDRSTRLEN];
2137 	char ip6buft[INET6_ADDRSTRLEN];
2138 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2139 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2140 	    ip6_sprintf(ip6buft, tgt6));
2141 	return buf;
2142 }
2143 
2144 void
2145 icmp6_redirect_input(struct mbuf *m, int off)
2146 {
2147 	struct ifnet *ifp;
2148 	struct ip6_hdr *ip6;
2149 	struct nd_redirect *nd_rd;
2150 	struct in6_addr src6, redtgt6, reddst6;
2151 	union nd_opts ndopts;
2152 	char ip6buf[INET6_ADDRSTRLEN];
2153 	char *lladdr;
2154 	int icmp6len, is_onlink, is_router, lladdrlen;
2155 
2156 	M_ASSERTPKTHDR(m);
2157 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2158 
2159 	/* XXX if we are router, we don't update route by icmp6 redirect */
2160 	if (V_ip6_forwarding)
2161 		goto freeit;
2162 	if (!V_icmp6_rediraccept)
2163 		goto freeit;
2164 
2165 	/* RFC 6980: Nodes MUST silently ignore fragments */
2166 	if(m->m_flags & M_FRAGMENTED)
2167 		goto freeit;
2168 
2169 	ip6 = mtod(m, struct ip6_hdr *);
2170 	icmp6len = ntohs(ip6->ip6_plen);
2171 	if (m->m_len < off + icmp6len) {
2172 		m = m_pullup(m, off + icmp6len);
2173 		if (m == NULL) {
2174 			IP6STAT_INC(ip6s_exthdrtoolong);
2175 			return;
2176 		}
2177 	}
2178 	ip6 = mtod(m, struct ip6_hdr *);
2179 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2180 
2181 	ifp = m->m_pkthdr.rcvif;
2182 	redtgt6 = nd_rd->nd_rd_target;
2183 	reddst6 = nd_rd->nd_rd_dst;
2184 
2185 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2186 	    in6_setscope(&reddst6, ifp, NULL)) {
2187 		goto freeit;
2188 	}
2189 
2190 	/* validation */
2191 	src6 = ip6->ip6_src;
2192 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2193 		nd6log((LOG_ERR,
2194 		    "ICMP6 redirect sent from %s rejected; "
2195 		    "must be from linklocal\n",
2196 		    ip6_sprintf(ip6buf, &src6)));
2197 		goto bad;
2198 	}
2199 	if (__predict_false(ip6->ip6_hlim != 255)) {
2200 		ICMP6STAT_INC(icp6s_invlhlim);
2201 		nd6log((LOG_ERR,
2202 		    "ICMP6 redirect sent from %s rejected; "
2203 		    "hlim=%d (must be 255)\n",
2204 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2205 		goto bad;
2206 	}
2207     {
2208 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2209 	struct nhop_object *nh;
2210 	struct in6_addr kdst;
2211 	uint32_t scopeid;
2212 
2213 	in6_splitscope(&reddst6, &kdst, &scopeid);
2214 	NET_EPOCH_ASSERT();
2215 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2216 	if (nh != NULL) {
2217 		struct in6_addr nh_addr;
2218 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2219 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2220 			nd6log((LOG_ERR,
2221 			    "ICMP6 redirect rejected; no route "
2222 			    "with inet6 gateway found for redirect dst: %s\n",
2223 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2224 			goto bad;
2225 		}
2226 
2227 		/*
2228 		 * Embed scope zone id into next hop address.
2229 		 */
2230 		nh_addr = nh->gw6_sa.sin6_addr;
2231 
2232 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2233 			nd6log((LOG_ERR,
2234 			    "ICMP6 redirect rejected; "
2235 			    "not equal to gw-for-src=%s (must be same): "
2236 			    "%s\n",
2237 			    ip6_sprintf(ip6buf, &nh_addr),
2238 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2239 			goto bad;
2240 		}
2241 	} else {
2242 		nd6log((LOG_ERR,
2243 		    "ICMP6 redirect rejected; "
2244 		    "no route found for redirect dst: %s\n",
2245 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2246 		goto bad;
2247 	}
2248     }
2249 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2250 		nd6log((LOG_ERR,
2251 		    "ICMP6 redirect rejected; "
2252 		    "redirect dst must be unicast: %s\n",
2253 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2254 		goto bad;
2255 	}
2256 
2257 	is_router = is_onlink = 0;
2258 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2259 		is_router = 1;	/* router case */
2260 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2261 		is_onlink = 1;	/* on-link destination case */
2262 	if (!is_router && !is_onlink) {
2263 		nd6log((LOG_ERR,
2264 		    "ICMP6 redirect rejected; "
2265 		    "neither router case nor onlink case: %s\n",
2266 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2267 		goto bad;
2268 	}
2269 
2270 	icmp6len -= sizeof(*nd_rd);
2271 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2272 	if (nd6_options(&ndopts) < 0) {
2273 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2274 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2275 		/* nd6_options have incremented stats */
2276 		goto freeit;
2277 	}
2278 
2279 	lladdr = NULL;
2280 	lladdrlen = 0;
2281 	if (ndopts.nd_opts_tgt_lladdr) {
2282 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2283 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2284 	}
2285 
2286 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2287 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2288 		    "(if %d, icmp6 packet %d): %s\n",
2289 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2290 		    ifp->if_addrlen, lladdrlen - 2,
2291 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2292 		goto bad;
2293 	}
2294 
2295 	/* Validation passed. */
2296 
2297 	/* RFC 2461 8.3 */
2298 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2299 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2300 
2301 	/*
2302 	 * Install a gateway route in the better-router case or an interface
2303 	 * route in the on-link-destination case.
2304 	 */
2305 	{
2306 		struct sockaddr_in6 sdst;
2307 		struct sockaddr_in6 sgw;
2308 		struct sockaddr_in6 ssrc;
2309 		struct sockaddr *gw;
2310 		int rt_flags;
2311 		u_int fibnum;
2312 
2313 		bzero(&sdst, sizeof(sdst));
2314 		bzero(&ssrc, sizeof(ssrc));
2315 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2316 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2317 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2318 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2319 		rt_flags = 0;
2320 		if (is_router) {
2321 			bzero(&sgw, sizeof(sgw));
2322 			sgw.sin6_family = AF_INET6;
2323 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2324 			bcopy(&redtgt6, &sgw.sin6_addr,
2325 				sizeof(struct in6_addr));
2326 			gw = (struct sockaddr *)&sgw;
2327 			rt_flags |= RTF_GATEWAY;
2328 		} else
2329 			gw = ifp->if_addr->ifa_addr;
2330 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2331 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2332 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2333 			    V_icmp6_redirtimeout);
2334 	}
2335 
2336  freeit:
2337 	m_freem(m);
2338 	return;
2339 
2340  bad:
2341 	ICMP6STAT_INC(icp6s_badredirect);
2342 	m_freem(m);
2343 }
2344 
2345 void
2346 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2347 {
2348 	struct ifnet *ifp;	/* my outgoing interface */
2349 	struct in6_addr *ifp_ll6;
2350 	struct in6_addr *router_ll6;
2351 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2352 	struct mbuf *m = NULL;	/* newly allocated one */
2353 	struct m_tag *mtag;
2354 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2355 	struct nd_redirect *nd_rd;
2356 	struct llentry *ln = NULL;
2357 	size_t maxlen;
2358 	u_char *p;
2359 	struct ifnet *outif = NULL;
2360 	struct sockaddr_in6 src_sa;
2361 
2362 	icmp6_errcount(ND_REDIRECT, 0);
2363 
2364 	/* if we are not router, we don't send icmp6 redirect */
2365 	if (!V_ip6_forwarding)
2366 		goto fail;
2367 
2368 	/* sanity check */
2369 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2370 		goto fail;
2371 
2372 	/*
2373 	 * Address check:
2374 	 *  the source address must identify a neighbor, and
2375 	 *  the destination address must not be a multicast address
2376 	 *  [RFC 2461, sec 8.2]
2377 	 */
2378 	sip6 = mtod(m0, struct ip6_hdr *);
2379 	bzero(&src_sa, sizeof(src_sa));
2380 	src_sa.sin6_family = AF_INET6;
2381 	src_sa.sin6_len = sizeof(src_sa);
2382 	src_sa.sin6_addr = sip6->ip6_src;
2383 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2384 		goto fail;
2385 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2386 		goto fail;	/* what should we do here? */
2387 
2388 	/* rate limit */
2389 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2390 		goto fail;
2391 
2392 	/*
2393 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2394 	 * we almost always ask for an mbuf cluster for simplicity.
2395 	 * (MHLEN < IPV6_MMTU is almost always true)
2396 	 */
2397 #if IPV6_MMTU >= MCLBYTES
2398 # error assumption failed about IPV6_MMTU and MCLBYTES
2399 #endif
2400 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2401 	if (m == NULL)
2402 		goto fail;
2403 	M_SETFIB(m, M_GETFIB(m0));
2404 	maxlen = M_TRAILINGSPACE(m);
2405 	maxlen = min(IPV6_MMTU, maxlen);
2406 	/* just for safety */
2407 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2408 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2409 		goto fail;
2410 	}
2411 
2412 	{
2413 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2414 		struct in6_ifaddr *ia;
2415 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2416 						 IN6_IFF_NOTREADY|
2417 						 IN6_IFF_ANYCAST)) == NULL)
2418 			goto fail;
2419 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2420 		/* XXXRW: reference released prematurely. */
2421 		ifa_free(&ia->ia_ifa);
2422 	}
2423 
2424 	/* get ip6 linklocal address for the router. */
2425 	if (nh->nh_flags & NHF_GATEWAY) {
2426 		struct sockaddr_in6 *sin6;
2427 		sin6 = &nh->gw6_sa;
2428 		router_ll6 = &sin6->sin6_addr;
2429 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2430 			router_ll6 = (struct in6_addr *)NULL;
2431 	} else
2432 		router_ll6 = (struct in6_addr *)NULL;
2433 
2434 	/* ip6 */
2435 	ip6 = mtod(m, struct ip6_hdr *);
2436 	ip6->ip6_flow = 0;
2437 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2438 	ip6->ip6_vfc |= IPV6_VERSION;
2439 	/* ip6->ip6_plen will be set later */
2440 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2441 	ip6->ip6_hlim = 255;
2442 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2443 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2444 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2445 
2446 	/* ND Redirect */
2447 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2448 	nd_rd->nd_rd_type = ND_REDIRECT;
2449 	nd_rd->nd_rd_code = 0;
2450 	nd_rd->nd_rd_reserved = 0;
2451 	if (nh->nh_flags & NHF_GATEWAY) {
2452 		/*
2453 		 * nd_rd->nd_rd_target must be a link-local address in
2454 		 * better router cases.
2455 		 */
2456 		if (!router_ll6)
2457 			goto fail;
2458 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2459 		    sizeof(nd_rd->nd_rd_target));
2460 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2461 		    sizeof(nd_rd->nd_rd_dst));
2462 	} else {
2463 		/* make sure redtgt == reddst */
2464 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2465 		    sizeof(nd_rd->nd_rd_target));
2466 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2467 		    sizeof(nd_rd->nd_rd_dst));
2468 	}
2469 
2470 	p = (u_char *)(nd_rd + 1);
2471 
2472 	if (!router_ll6)
2473 		goto nolladdropt;
2474 
2475 	{
2476 		/* target lladdr option */
2477 		int len;
2478 		struct nd_opt_hdr *nd_opt;
2479 		char *lladdr;
2480 
2481 		ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2482 		if (ln == NULL)
2483 			goto nolladdropt;
2484 
2485 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2486 		len = (len + 7) & ~7;	/* round by 8 */
2487 		/* safety check */
2488 		if (len + (p - (u_char *)ip6) > maxlen)
2489 			goto nolladdropt;
2490 
2491 		if (ln->la_flags & LLE_VALID) {
2492 			nd_opt = (struct nd_opt_hdr *)p;
2493 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2494 			nd_opt->nd_opt_len = len >> 3;
2495 			lladdr = (char *)(nd_opt + 1);
2496 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2497 			p += len;
2498 		}
2499 	}
2500 nolladdropt:
2501 	if (ln != NULL)
2502 		LLE_RUNLOCK(ln);
2503 
2504 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2505 
2506 	/* just to be safe */
2507 #ifdef M_DECRYPTED	/*not openbsd*/
2508 	if (m0->m_flags & M_DECRYPTED)
2509 		goto noredhdropt;
2510 #endif
2511 	if (p - (u_char *)ip6 > maxlen)
2512 		goto noredhdropt;
2513 
2514 	{
2515 		/* redirected header option */
2516 		int len;
2517 		struct nd_opt_rd_hdr *nd_opt_rh;
2518 
2519 		/*
2520 		 * compute the maximum size for icmp6 redirect header option.
2521 		 * XXX room for auth header?
2522 		 */
2523 		len = maxlen - (p - (u_char *)ip6);
2524 		len &= ~7;
2525 
2526 		/* This is just for simplicity. */
2527 		if (m0->m_pkthdr.len != m0->m_len) {
2528 			if (m0->m_next) {
2529 				m_freem(m0->m_next);
2530 				m0->m_next = NULL;
2531 			}
2532 			m0->m_pkthdr.len = m0->m_len;
2533 		}
2534 
2535 		/*
2536 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2537 		 * about padding/truncate rule for the original IP packet.
2538 		 * From the discussion on IPv6imp in Feb 1999,
2539 		 * the consensus was:
2540 		 * - "attach as much as possible" is the goal
2541 		 * - pad if not aligned (original size can be guessed by
2542 		 *   original ip6 header)
2543 		 * Following code adds the padding if it is simple enough,
2544 		 * and truncates if not.
2545 		 */
2546 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2547 			panic("assumption failed in %s:%d", __FILE__,
2548 			    __LINE__);
2549 
2550 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2551 			/* not enough room, truncate */
2552 			m0->m_pkthdr.len = m0->m_len = len -
2553 			    sizeof(*nd_opt_rh);
2554 		} else {
2555 			/* enough room, pad or truncate */
2556 			size_t extra;
2557 
2558 			extra = m0->m_pkthdr.len % 8;
2559 			if (extra) {
2560 				/* pad if easy enough, truncate if not */
2561 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2562 					/* pad */
2563 					m0->m_len += (8 - extra);
2564 					m0->m_pkthdr.len += (8 - extra);
2565 				} else {
2566 					/* truncate */
2567 					m0->m_pkthdr.len -= extra;
2568 					m0->m_len -= extra;
2569 				}
2570 			}
2571 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2572 			m0->m_pkthdr.len = m0->m_len = len -
2573 			    sizeof(*nd_opt_rh);
2574 		}
2575 
2576 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2577 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2578 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2579 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2580 		p += sizeof(*nd_opt_rh);
2581 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2582 
2583 		/* connect m0 to m */
2584 		m_tag_delete_chain(m0, NULL);
2585 		m0->m_flags &= ~M_PKTHDR;
2586 		m->m_next = m0;
2587 		m->m_pkthdr.len = m->m_len + m0->m_len;
2588 		m0 = NULL;
2589 	}
2590 noredhdropt:;
2591 	if (m0) {
2592 		m_freem(m0);
2593 		m0 = NULL;
2594 	}
2595 
2596 	/* XXX: clear embedded link IDs in the inner header */
2597 	in6_clearscope(&sip6->ip6_src);
2598 	in6_clearscope(&sip6->ip6_dst);
2599 	in6_clearscope(&nd_rd->nd_rd_target);
2600 	in6_clearscope(&nd_rd->nd_rd_dst);
2601 
2602 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2603 
2604 	nd_rd->nd_rd_cksum = 0;
2605 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2606 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2607 
2608         if (send_sendso_input_hook != NULL) {
2609 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2610 			M_NOWAIT);
2611 		if (mtag == NULL)
2612 			goto fail;
2613 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2614 		m_tag_prepend(m, mtag);
2615 	}
2616 
2617 	/* send the packet to outside... */
2618 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2619 	if (outif) {
2620 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2621 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2622 	}
2623 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2624 
2625 	return;
2626 
2627 fail:
2628 	if (m)
2629 		m_freem(m);
2630 	if (m0)
2631 		m_freem(m0);
2632 }
2633 
2634 /*
2635  * ICMPv6 socket option processing.
2636  */
2637 int
2638 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2639 {
2640 	int error = 0;
2641 	int optlen;
2642 	struct inpcb *inp = sotoinpcb(so);
2643 	int level, op, optname;
2644 
2645 	if (sopt) {
2646 		level = sopt->sopt_level;
2647 		op = sopt->sopt_dir;
2648 		optname = sopt->sopt_name;
2649 		optlen = sopt->sopt_valsize;
2650 	} else
2651 		level = op = optname = optlen = 0;
2652 
2653 	if (level != IPPROTO_ICMPV6) {
2654 		return EINVAL;
2655 	}
2656 
2657 	switch (op) {
2658 	case PRCO_SETOPT:
2659 		switch (optname) {
2660 		case ICMP6_FILTER:
2661 		    {
2662 			struct icmp6_filter ic6f;
2663 
2664 			if (optlen != sizeof(ic6f)) {
2665 				error = EMSGSIZE;
2666 				break;
2667 			}
2668 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2669 			if (error == 0) {
2670 				INP_WLOCK(inp);
2671 				*inp->in6p_icmp6filt = ic6f;
2672 				INP_WUNLOCK(inp);
2673 			}
2674 			break;
2675 		    }
2676 
2677 		default:
2678 			error = ENOPROTOOPT;
2679 			break;
2680 		}
2681 		break;
2682 
2683 	case PRCO_GETOPT:
2684 		switch (optname) {
2685 		case ICMP6_FILTER:
2686 		    {
2687 			struct icmp6_filter ic6f;
2688 
2689 			INP_RLOCK(inp);
2690 			ic6f = *inp->in6p_icmp6filt;
2691 			INP_RUNLOCK(inp);
2692 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2693 			break;
2694 		    }
2695 
2696 		default:
2697 			error = ENOPROTOOPT;
2698 			break;
2699 		}
2700 		break;
2701 	}
2702 
2703 	return (error);
2704 }
2705 
2706 /*
2707  * Perform rate limit check.
2708  * Returns 0 if it is okay to send the icmp6 packet.
2709  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2710  * limitation.
2711  *
2712  * XXX per-destination/type check necessary?
2713  *
2714  * dst - not used at this moment
2715  * type - not used at this moment
2716  * code - not used at this moment
2717  */
2718 static int
2719 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2720     const int code)
2721 {
2722 	int ret;
2723 
2724 	ret = 0;	/* okay to send */
2725 
2726 	/* PPS limit */
2727 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2728 	    V_icmp6errppslim)) {
2729 		/* The packet is subject to rate limit */
2730 		ret++;
2731 	}
2732 
2733 	return ret;
2734 }
2735