xref: /xnu-11417.140.69/bsd/netinet6/icmp6.c (revision 43a90889846e00bfb5cf1d255cdc0a701a1e05a4)
1 /*
2  * Copyright (c) 2000-2025 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 
29 /*	$FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.6 2001/07/10 09:44:16 ume Exp $	*/
30 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
31 
32 /*
33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the project nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  */
60 
61 /*
62  * Copyright (c) 1982, 1986, 1988, 1993
63  *	The Regents of the University of California.  All rights reserved.
64  *
65  * Redistribution and use in source and binary forms, with or without
66  * modification, are permitted provided that the following conditions
67  * are met:
68  * 1. Redistributions of source code must retain the above copyright
69  *    notice, this list of conditions and the following disclaimer.
70  * 2. Redistributions in binary form must reproduce the above copyright
71  *    notice, this list of conditions and the following disclaimer in the
72  *    documentation and/or other materials provided with the distribution.
73  * 3. All advertising materials mentioning features or use of this software
74  *    must display the following acknowledgement:
75  *	This product includes software developed by the University of
76  *	California, Berkeley and its contributors.
77  * 4. Neither the name of the University nor the names of its contributors
78  *    may be used to endorse or promote products derived from this software
79  *    without specific prior written permission.
80  *
81  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
82  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
83  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
84  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
85  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
86  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
87  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
88  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
89  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
90  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
91  * SUCH DAMAGE.
92  *
93  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
94  */
95 
96 
97 #include <sys/param.h>
98 #include <sys/systm.h>
99 #include <sys/lock.h>
100 #include <sys/malloc.h>
101 #include <sys/mcache.h>
102 #include <sys/mbuf.h>
103 #include <sys/protosw.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/time.h>
107 #include <sys/kernel.h>
108 #include <sys/syslog.h>
109 #include <sys/domain.h>
110 #include <sys/kauth.h>
111 
112 #include <net/if.h>
113 #include <net/route.h>
114 #include <net/if_dl.h>
115 #include <net/if_types.h>
116 
117 #include <netinet/in.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip6.h>
120 #include <netinet6/ip6_var.h>
121 #include <netinet/icmp6.h>
122 #include <netinet6/mld6_var.h>
123 #include <netinet/in_pcb.h>
124 #include <netinet6/in6_pcb.h>
125 #include <netinet6/in6_var.h>
126 #include <netinet6/nd6.h>
127 #include <netinet6/in6_ifattach.h>
128 #include <netinet6/ip6protosw.h>
129 #include <netinet6/scope6_var.h>
130 
131 #if IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netkey/key.h>
134 #endif
135 
136 #include <net/net_osdep.h>
137 
138 #if NECP
139 #include <net/necp.h>
140 #endif
141 
142 #include <net/sockaddr_utils.h>
143 
144 extern struct ip6protosw *ip6_protox[IPPROTO_MAX];
145 
146 extern uint32_t rip_sendspace;
147 extern uint32_t rip_recvspace;
148 
149 struct icmp6stat icmp6stat;
150 
151 extern struct inpcbhead ripcb;
152 extern int icmp6errppslim;
153 extern int icmp6errppslim_random_incr;
154 extern int icmp6rappslim;
155 static int icmp6errpps_count = 0;
156 static int icmp6rapps_count = 0;
157 static struct timeval icmp6errppslim_last;
158 static struct timeval icmp6rappslim_last;
159 extern int icmp6_nodeinfo;
160 extern struct inpcbinfo ripcbinfo;
161 
162 static void icmp6_errcount(struct icmp6errstat *, int, int);
163 static int icmp6_rip6_input(struct mbuf **, int);
164 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
165 static const char *icmp6_redirect_diag(struct in6_addr *,
166     struct in6_addr *, struct in6_addr *);
167 static struct mbuf *ni6_input(struct mbuf *, int);
168 static struct mbuf *ni6_nametodns(const char *__counted_by(len), uint32_t len, int);
169 static int ni6_dnsmatch(const char *__counted_by(alen), int alen, const char *__counted_by(blen), int blen);
170 static int ni6_addrs(struct icmp6_nodeinfo *,
171     struct ifnet **, char *__indexable);
172 static int ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *__indexable nni6,
173     struct ifnet *ifp0, int resid);
174 static int icmp6_notify_error(struct mbuf *, int, int, int);
175 
176 
177 
178 void
icmp6_init(struct ip6protosw * pp,struct domain * dp)179 icmp6_init(struct ip6protosw *pp, struct domain *dp)
180 {
181 #pragma unused(dp)
182 	static int icmp6_initialized = 0;
183 
184 	/* Also called from ip6_init() without pp */
185 	VERIFY(pp == NULL ||
186 	    (pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
187 
188 	/* This gets called by more than one protocols, so initialize once */
189 	if (!icmp6_initialized) {
190 		icmp6_initialized = 1;
191 		mld_init();
192 		if (icmp6errppslim >= 0 &&
193 		    icmp6errppslim_random_incr > 0 &&
194 		    icmp6errppslim <= INT32_MAX - (icmp6errppslim_random_incr + 1)) {
195 			icmp6errppslim += (random() % icmp6errppslim_random_incr) + 1;
196 		}
197 	}
198 }
199 
200 static void
icmp6_errcount(struct icmp6errstat * stat,int type,int code)201 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
202 {
203 	switch (type) {
204 	case ICMP6_DST_UNREACH:
205 		switch (code) {
206 		case ICMP6_DST_UNREACH_NOROUTE:
207 			stat->icp6errs_dst_unreach_noroute++;
208 			return;
209 		case ICMP6_DST_UNREACH_ADMIN:
210 			stat->icp6errs_dst_unreach_admin++;
211 			return;
212 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
213 			stat->icp6errs_dst_unreach_beyondscope++;
214 			return;
215 		case ICMP6_DST_UNREACH_ADDR:
216 			stat->icp6errs_dst_unreach_addr++;
217 			return;
218 		case ICMP6_DST_UNREACH_NOPORT:
219 			stat->icp6errs_dst_unreach_noport++;
220 			return;
221 		}
222 		break;
223 	case ICMP6_PACKET_TOO_BIG:
224 		stat->icp6errs_packet_too_big++;
225 		return;
226 	case ICMP6_TIME_EXCEEDED:
227 		switch (code) {
228 		case ICMP6_TIME_EXCEED_TRANSIT:
229 			stat->icp6errs_time_exceed_transit++;
230 			return;
231 		case ICMP6_TIME_EXCEED_REASSEMBLY:
232 			stat->icp6errs_time_exceed_reassembly++;
233 			return;
234 		}
235 		break;
236 	case ICMP6_PARAM_PROB:
237 		switch (code) {
238 		case ICMP6_PARAMPROB_HEADER:
239 			stat->icp6errs_paramprob_header++;
240 			return;
241 		case ICMP6_PARAMPROB_NEXTHEADER:
242 			stat->icp6errs_paramprob_nextheader++;
243 			return;
244 		case ICMP6_PARAMPROB_OPTION:
245 			stat->icp6errs_paramprob_option++;
246 			return;
247 		}
248 		break;
249 	case ND_REDIRECT:
250 		stat->icp6errs_redirect++;
251 		return;
252 	}
253 	stat->icp6errs_unknown++;
254 }
255 
256 /*
257  * Generate packet gencount for ICMPv6 for a given error type
258  * and code.
259  */
260 static uint32_t
icmp6_error_packet_gencount(int type,int code)261 icmp6_error_packet_gencount(int type, int code)
262 {
263 	return (PF_INET6 << 24) | (type << 16) | (code << 8);
264 }
265 
266 static int suppress_icmp6_port_unreach = 0;
267 
268 SYSCTL_DECL(_net_inet6_icmp6);
269 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, suppress_icmp6_port_unreach, CTLFLAG_RW | CTLFLAG_LOCKED,
270     &suppress_icmp6_port_unreach, 0,
271     "Suppress ICMPv6 destination unreachable type with code port unreachable");
272 
273 /*
274  * Generate an error packet of type error in response to bad IP6 packet.
275  */
276 void
icmp6_error(struct mbuf * m,int type,int code,int param)277 icmp6_error(struct mbuf *m, int type, int code, int param)
278 {
279 	icmp6_error_flag(m, type, code, param, ICMP6_ERROR_RST_MRCVIF);
280 }
281 
282 void
icmp6_error_flag(struct mbuf * m,int type,int code,int param,int flags)283 icmp6_error_flag(struct mbuf *m, int type, int code, int param, int flags)
284 {
285 	struct ip6_hdr *__single oip6;
286 	struct ip6_hdr *nip6;
287 	struct icmp6_hdr *__single icmp6;
288 	u_int preplen;
289 	int off;
290 
291 	icmp6stat.icp6s_error++;
292 
293 	/* count per-type-code statistics */
294 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
295 
296 	if (suppress_icmp6_port_unreach && type == ICMP6_DST_UNREACH &&
297 	    code == ICMP6_DST_UNREACH_NOPORT) {
298 		goto freeit;
299 	}
300 
301 #ifdef M_DECRYPTED      /*not openbsd*/
302 	if (m->m_flags & M_DECRYPTED) {
303 		icmp6stat.icp6s_canterror++;
304 		goto freeit;
305 	}
306 #endif
307 
308 #ifndef PULLDOWN_TEST
309 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), return );
310 #else
311 	if (m->m_len < sizeof(struct ip6_hdr)) {
312 		m = m_pullup(m, sizeof(struct ip6_hdr));
313 		if (m == NULL) {
314 			return;
315 		}
316 	}
317 #endif
318 	oip6 = mtod(m, struct ip6_hdr *);
319 
320 	/*
321 	 * If the destination address of the erroneous packet is a multicast
322 	 * address, or the packet was sent using link-layer multicast,
323 	 * we should basically suppress sending an error (RFC 2463, Section
324 	 * 2.4).
325 	 * We have two exceptions (the item e.2 in that section):
326 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
327 	 * - the Parameter Problem Message that can be allowed an icmp6 error
328 	 *   in the option type field.  This check has been done in
329 	 *   ip6_unknown_opt(), so we can just check the type and code.
330 	 */
331 	if ((m->m_flags & (M_BCAST | M_MCAST) ||
332 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
333 	    (type != ICMP6_PACKET_TOO_BIG &&
334 	    (type != ICMP6_PARAM_PROB ||
335 	    code != ICMP6_PARAMPROB_OPTION))) {
336 		goto freeit;
337 	}
338 
339 	/*
340 	 * RFC 2463, 2.4 (e.5): source address check.
341 	 * XXX: the case of anycast source?
342 	 */
343 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
344 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) {
345 		goto freeit;
346 	}
347 
348 	/*
349 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
350 	 * don't do it.
351 	 *
352 	 * We want to check for that for all ICMP error types, other than
353 	 * ICMP6_PARAM_PROB when it is being sent in response of first frag
354 	 * with incomplete header.
355 	 * That also includes the case when the first frag has incomplete ICMPv6
356 	 * header. The check below in that case would fail the IP6_EXTHDR_CHECK
357 	 * and would otherwise prevent us from sending the error back.
358 	 */
359 	if (type != ICMP6_PARAM_PROB ||
360 	    code != ICMP6_PARAMPROB_FIRSTFRAG_INCOMP_HDR) {
361 		int nxt = -1;
362 		off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
363 		if (off >= 0 && nxt == IPPROTO_ICMPV6) {
364 			struct icmp6_hdr *__single icp;
365 
366 #ifndef PULLDOWN_TEST
367 			IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), return );
368 			icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
369 #else
370 			IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
371 			    sizeof(*icp));
372 			if (icp == NULL) {
373 				icmp6stat.icp6s_tooshort++;
374 				return;
375 			}
376 #endif
377 			if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
378 			    icp->icmp6_type == ND_REDIRECT) {
379 				/*
380 				 * ICMPv6 error
381 				 * Special case: for redirect (which is
382 				 * informational) we must not send icmp6 error.
383 				 */
384 				icmp6stat.icp6s_canterror++;
385 				goto freeit;
386 			} else {
387 				/* ICMPv6 informational - send the error */
388 			}
389 		} else {
390 			/* non-ICMPv6 - send the error */
391 		}
392 	}
393 
394 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
395 
396 	/*
397 	 * OK, ICMP6 can be generated.
398 	 */
399 
400 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) {
401 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
402 	}
403 
404 	/*
405 	 * To avoid some flavors of port scanning and other attacks,
406 	 * use packet suppression without using any other sort of
407 	 * rate limiting with static bounds.
408 	 * XXX Not setting PKTF_FLOW_ID here because we were concerned
409 	 * about it triggering regression elsewhere outside of network stack
410 	 * where there might be an assumption around flow ID being non-zero.
411 	 * It should be noted though that previously if PKTF_FLOW_ID was not
412 	 * set, PF would have generated flow hash irrespective of ICMPv4/v6
413 	 * type. That doesn't happen now and PF only computes hash for ICMP
414 	 * types that need state creation (which is not true of error types).
415 	 * It would have been a problem because we really want all the ICMP
416 	 * error type packets to share the same flow ID for global suppression.
417 	 */
418 	m->m_pkthdr.comp_gencnt = icmp6_error_packet_gencount(type, code);
419 
420 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
421 	M_PREPEND(m, preplen, M_DONTWAIT, 1);
422 	if (m == NULL) {
423 		nd6log(debug, "ENOBUFS in icmp6_error %d\n", __LINE__);
424 		return;
425 	}
426 
427 	nip6 = mtod(m, struct ip6_hdr *);
428 	nip6->ip6_src  = oip6->ip6_src;
429 	nip6->ip6_dst  = oip6->ip6_dst;
430 
431 	in6_clearscope(&oip6->ip6_src);
432 	in6_clearscope(&oip6->ip6_dst);
433 
434 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
435 	icmp6->icmp6_type = (uint8_t)type;
436 	icmp6->icmp6_code = (uint8_t)code;
437 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
438 
439 	/*
440 	 * icmp6_reflect() is designed to be in the input path.
441 	 * icmp6_error() can be called from both input and output path,
442 	 * and if we are in output path rcvif could contain bogus value.
443 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
444 	 * information in ip header (nip6).
445 	 */
446 	if (flags & ICMP6_ERROR_RST_MRCVIF) {
447 		m->m_pkthdr.rcvif = NULL;
448 	}
449 
450 	icmp6stat.icp6s_outhist[type]++;
451 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
452 
453 	return;
454 
455 freeit:
456 	/*
457 	 * If we can't tell whether or not we can generate ICMP6, free it.
458 	 */
459 	m_freem(m);
460 }
461 
462 /*
463  * Process a received ICMP6 message.
464  */
465 int
icmp6_input(struct mbuf ** mp,int * offp,int proto)466 icmp6_input(struct mbuf **mp, int *offp, int proto)
467 {
468 #pragma unused(proto)
469 	mbuf_ref_t m = *mp, n;
470 	ifnet_ref_t ifp;
471 	struct ip6_hdr *ip6, *nip6;
472 	struct icmp6_hdr *__single icmp6, *__single nicmp6;
473 	int off = *offp;
474 	int icmp6len = m->m_pkthdr.len - *offp;
475 	int code, sum, noff, proxy = 0;
476 
477 	ifp = m->m_pkthdr.rcvif;
478 
479 #ifndef PULLDOWN_TEST
480 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), return IPPROTO_DONE);
481 	/* m might change if M_LOOP.  So, call mtod after this */
482 #endif
483 
484 	/* Expect 32-bit aligned data pointer on strict-align platforms */
485 	MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
486 
487 	/*
488 	 * Locate icmp6 structure in mbuf, and check
489 	 * that not corrupted and of at least minimum length
490 	 */
491 	ip6 = mtod(m, struct ip6_hdr *);
492 	if (icmp6len < sizeof(struct icmp6_hdr)) {
493 		icmp6stat.icp6s_tooshort++;
494 		goto freeit;
495 	}
496 
497 #ifndef PULLDOWN_TEST
498 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
499 #else
500 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
501 	if (icmp6 == NULL) {
502 		icmp6stat.icp6s_tooshort++;
503 		return IPPROTO_DONE;
504 	}
505 #endif
506 	code = icmp6->icmp6_code;
507 
508 	/*
509 	 * Early check for RFC 6980
510 	 * Drop certain NDP packets if they came in fragmented
511 	 */
512 	switch (icmp6->icmp6_type) {
513 	case ND_ROUTER_SOLICIT:
514 	case ND_ROUTER_ADVERT:
515 	case ND_NEIGHBOR_SOLICIT:
516 	case ND_NEIGHBOR_ADVERT:
517 	case ND_REDIRECT:
518 		if (m->m_pkthdr.pkt_flags & PKTF_REASSEMBLED) {
519 			icmp6stat.icp6s_rfc6980_drop++;
520 			goto freeit;
521 		}
522 		break;
523 	default:
524 		break;
525 	}
526 
527 	/* Apply rate limit before checksum validation. */
528 	if (icmp6_ratelimit(&ip6->ip6_dst, icmp6->icmp6_type, code)) {
529 		icmp6stat.icp6s_toofreq++;
530 		goto freeit;
531 	}
532 
533 	/*
534 	 * Check multicast group membership.
535 	 * Note: SSM filters are not applied for ICMPv6 traffic.
536 	 */
537 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
538 		struct in6_multi *__single inm;
539 
540 		in6_multihead_lock_shared();
541 		IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, inm);
542 		in6_multihead_lock_done();
543 
544 		if (inm == NULL) {
545 			/*
546 			 * Don't discard if this is a Neighbor Solicitation
547 			 * that needs to be proxied (see check down below.)
548 			 */
549 			if (!(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST)) {
550 				ip6stat.ip6s_notmember++;
551 				in6_ifstat_inc(m->m_pkthdr.rcvif,
552 				    ifs6_in_discard);
553 				goto freeit;
554 			}
555 		} else {
556 			IN6M_REMREF(inm);
557 		}
558 	}
559 
560 	/*
561 	 * calculate the checksum
562 	 */
563 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
564 		nd6log(error,
565 		    "ICMP6 checksum error(%d|%x) %s\n",
566 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src));
567 		icmp6stat.icp6s_checksum++;
568 		goto freeit;
569 	}
570 
571 	if (m->m_pkthdr.pkt_flags & PKTF_PROXY_DST) {
572 		/*
573 		 * This is the special case of proxying NS (dst is either
574 		 * solicited-node multicast or unicast); process it locally
575 		 * but don't deliver it to sockets.  It practically lets us
576 		 * steer the packet to nd6_prproxy_ns_input, where more
577 		 * specific tests and actions will be taken.
578 		 */
579 		switch (icmp6->icmp6_type) {
580 		case ND_NEIGHBOR_SOLICIT:
581 			proxy = 1;
582 			break;
583 		default:
584 			goto freeit;
585 		}
586 	}
587 
588 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
589 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
590 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK) {
591 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
592 	}
593 
594 	switch (icmp6->icmp6_type) {
595 	case ICMP6_DST_UNREACH:
596 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
597 		switch (code) {
598 		case ICMP6_DST_UNREACH_NOROUTE:
599 		case ICMP6_DST_UNREACH_ADDR:    /* PRC_HOSTDEAD is a DOS */
600 			code = PRC_UNREACH_NET;
601 			break;
602 		case ICMP6_DST_UNREACH_ADMIN:
603 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
604 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
605 			break;
606 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
607 			/* I mean "source address was incorrect." */
608 			code = PRC_PARAMPROB;
609 			break;
610 		case ICMP6_DST_UNREACH_NOPORT:
611 			code = PRC_UNREACH_PORT;
612 			break;
613 		default:
614 			goto badcode;
615 		}
616 		goto deliver;
617 
618 	case ICMP6_PACKET_TOO_BIG:
619 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
620 		if (ntohl(icmp6->icmp6_mtu) < IPV6_MMTU) {
621 			icmp6stat.icp6s_badpkttoobig++;
622 			goto freeit;
623 		}
624 
625 		code = PRC_MSGSIZE;
626 
627 		/*
628 		 * Updating the path MTU will be done after examining
629 		 * intermediate extension headers.
630 		 */
631 		goto deliver;
632 
633 	case ICMP6_TIME_EXCEEDED:
634 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
635 		switch (code) {
636 		case ICMP6_TIME_EXCEED_TRANSIT:
637 			code = PRC_TIMXCEED_INTRANS;
638 			break;
639 		case ICMP6_TIME_EXCEED_REASSEMBLY:
640 			code = PRC_TIMXCEED_REASS;
641 			break;
642 		default:
643 			goto badcode;
644 		}
645 		goto deliver;
646 
647 	case ICMP6_PARAM_PROB:
648 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
649 		switch (code) {
650 		case ICMP6_PARAMPROB_NEXTHEADER:
651 			code = PRC_UNREACH_PROTOCOL;
652 			break;
653 		case ICMP6_PARAMPROB_HEADER:
654 		case ICMP6_PARAMPROB_OPTION:
655 			code = PRC_PARAMPROB;
656 			break;
657 		default:
658 			goto badcode;
659 		}
660 		goto deliver;
661 
662 	case ICMP6_ECHO_REQUEST:
663 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
664 		if (code != 0) {
665 			goto badcode;
666 		}
667 
668 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
669 			/* Give up remote */
670 			goto rate_limit_checked;
671 		}
672 		if ((n->m_flags & M_EXT) != 0
673 		    || n->m_len < off + sizeof(struct icmp6_hdr)) {
674 			mbuf_ref_t n0 = n;
675 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
676 
677 			/*
678 			 * Prepare an internal mbuf. m_pullup() doesn't
679 			 * always copy the length we specified.
680 			 */
681 			if (maxlen >= MCLBYTES) {
682 				/* Give up remote */
683 				m_freem(n0);
684 				goto rate_limit_checked;
685 			}
686 			MGETHDR(n, M_DONTWAIT, n0->m_type);     /* MAC-OK */
687 			if (n && maxlen >= MHLEN) {
688 				MCLGET(n, M_DONTWAIT);
689 				if ((n->m_flags & M_EXT) == 0) {
690 					m_free(n);
691 					n = NULL;
692 				}
693 			}
694 			if (n == NULL) {
695 				/* Give up remote */
696 				m_freem(n0);
697 				goto rate_limit_checked;
698 			}
699 			M_COPY_PKTHDR(n, n0);
700 			/*
701 			 * Copy IPv6 and ICMPv6 only.
702 			 */
703 			nip6 = mtod(n, struct ip6_hdr *);
704 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
705 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
706 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
707 			noff = sizeof(struct ip6_hdr);
708 			n->m_pkthdr.len = n->m_len =
709 			    noff + sizeof(struct icmp6_hdr);
710 			/*
711 			 * Adjust mbuf. ip6_plen will be adjusted in
712 			 * ip6_output().
713 			 */
714 			m_adj(n0, off + sizeof(struct icmp6_hdr));
715 			n->m_pkthdr.len += n0->m_pkthdr.len;
716 			n->m_next = n0;
717 			n0->m_flags &= ~M_PKTHDR;
718 		} else {
719 			nip6 = mtod(n, struct ip6_hdr *);
720 			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
721 			    sizeof(*nicmp6));
722 			noff = off;
723 		}
724 		if (nicmp6 == NULL) {
725 			panic("nicmp6 is NULL in %s, which isn't good!", __FUNCTION__);
726 		} else {
727 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
728 			nicmp6->icmp6_code = 0;
729 		}
730 		if (n) {
731 			icmp6stat.icp6s_reflect++;
732 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
733 			icmp6_reflect(n, noff);
734 		}
735 		goto rate_limit_checked;
736 
737 	case ICMP6_ECHO_REPLY:
738 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
739 		if (code != 0) {
740 			goto badcode;
741 		}
742 		break;
743 
744 	case MLD_LISTENER_QUERY:
745 	case MLD_LISTENER_REPORT:
746 
747 		if (icmp6len < sizeof(struct mld_hdr)) {
748 			goto badlen;
749 		}
750 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) { /* XXX: ugly... */
751 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
752 		} else {
753 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
754 		}
755 
756 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
757 			/* give up local */
758 			if (mld_input(m, off, icmp6len) == IPPROTO_DONE) {
759 				m = NULL;
760 			}
761 			goto freeit;
762 		}
763 		if (mld_input(n, off, icmp6len) != IPPROTO_DONE) {
764 			m_freem(n);
765 		}
766 		/* m stays. */
767 		goto rate_limit_checked;
768 
769 	case MLD_LISTENER_DONE:
770 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
771 		if (icmp6len < sizeof(struct mld_hdr)) { /* necessary? */
772 			goto badlen;
773 		}
774 		break;          /* nothing to be done in kernel */
775 
776 	case MLD_MTRACE_RESP:
777 	case MLD_MTRACE:
778 		/* XXX: these two are experimental.  not officially defined. */
779 		/* XXX: per-interface statistics? */
780 		break;          /* just pass it to applications */
781 
782 	case ICMP6_NI_QUERY:
783 		if (!icmp6_nodeinfo) {
784 			break;
785 		}
786 //### LD 10/20 Check fbsd differences here. Not sure we're more advanced or not.
787 		/* By RFC 4620 refuse to answer queries from global scope addresses */
788 		if ((icmp6_nodeinfo & 8) != 8 && in6_addrscope(&ip6->ip6_src) == IPV6_ADDR_SCOPE_GLOBAL) {
789 			break;
790 		}
791 
792 		if (icmp6len < sizeof(struct icmp6_nodeinfo)) {
793 			goto badlen;
794 		}
795 
796 #ifndef PULLDOWN_TEST
797 		IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
798 		    return IPPROTO_DONE);
799 #endif
800 
801 		n = m_copy(m, 0, M_COPYALL);
802 		if (n) {
803 			n = ni6_input(n, off);
804 		}
805 		if (n) {
806 			noff = sizeof(struct ip6_hdr);
807 			icmp6stat.icp6s_reflect++;
808 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
809 			icmp6_reflect(n, noff);
810 		}
811 		goto rate_limit_checked;
812 
813 	case ICMP6_WRUREPLY:
814 		if (code != 0) {
815 			goto badcode;
816 		}
817 		break;
818 
819 	case ND_ROUTER_SOLICIT:
820 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
821 		if (code != 0) {
822 			goto badcode;
823 		}
824 		if (icmp6len < sizeof(struct nd_router_solicit)) {
825 			goto badlen;
826 		}
827 
828 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
829 			/* give up local */
830 			nd6_rs_input(m, off, icmp6len);
831 			m = NULL;
832 			goto freeit;
833 		}
834 		nd6_rs_input(n, off, icmp6len);
835 		/* m stays. */
836 		goto rate_limit_checked;
837 
838 	case ND_ROUTER_ADVERT:
839 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
840 		if (code != 0) {
841 			goto badcode;
842 		}
843 		if (icmp6len < sizeof(struct nd_router_advert)) {
844 			goto badlen;
845 		}
846 
847 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
848 			/* give up local */
849 			nd6_ra_input(m, off, icmp6len);
850 			m = NULL;
851 			goto freeit;
852 		}
853 		nd6_ra_input(n, off, icmp6len);
854 		/* m stays. */
855 		goto rate_limit_checked;
856 
857 	case ND_NEIGHBOR_SOLICIT:
858 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
859 		if (code != 0) {
860 			goto badcode;
861 		}
862 		if (icmp6len < sizeof(struct nd_neighbor_solicit)) {
863 			goto badlen;
864 		}
865 
866 		if (proxy ||
867 		    ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL)) {
868 			/* give up local */
869 			nd6_ns_input(m, off, icmp6len);
870 			m = NULL;
871 			goto freeit;
872 		}
873 		nd6_ns_input(n, off, icmp6len);
874 		/* m stays. */
875 		goto rate_limit_checked;
876 
877 	case ND_NEIGHBOR_ADVERT:
878 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
879 		if (code != 0) {
880 			goto badcode;
881 		}
882 		if (icmp6len < sizeof(struct nd_neighbor_advert)) {
883 			goto badlen;
884 		}
885 
886 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
887 			/* give up local */
888 			nd6_na_input(m, off, icmp6len);
889 			m = NULL;
890 			goto freeit;
891 		}
892 		nd6_na_input(n, off, icmp6len);
893 		/* m stays. */
894 		goto rate_limit_checked;
895 
896 	case ND_REDIRECT:
897 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
898 		if (code != 0) {
899 			goto badcode;
900 		}
901 		if (icmp6len < sizeof(struct nd_redirect)) {
902 			goto badlen;
903 		}
904 
905 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
906 			/* give up local */
907 			icmp6_redirect_input(m, off, icmp6len);
908 			m = NULL;
909 			goto freeit;
910 		}
911 		icmp6_redirect_input(n, off, icmp6len);
912 		/* m stays. */
913 		goto rate_limit_checked;
914 
915 	case ICMP6_ROUTER_RENUMBERING:
916 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
917 		    code != ICMP6_ROUTER_RENUMBERING_RESULT) {
918 			goto badcode;
919 		}
920 		if (icmp6len < sizeof(struct icmp6_router_renum)) {
921 			goto badlen;
922 		}
923 		break;
924 
925 	default:
926 		nd6log(debug,
927 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
928 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
929 		    ip6_sprintf(&ip6->ip6_dst),
930 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0);
931 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
932 			/* ICMPv6 error: MUST deliver it by spec... */
933 			code = PRC_NCMDS;
934 			/* deliver */
935 		} else {
936 			/* ICMPv6 informational: MUST not deliver */
937 			goto rate_limit_checked;
938 		}
939 deliver:
940 		if (icmp6_notify_error(m, off, icmp6len, code)) {
941 			/* In this case, m should've been freed. */
942 			return IPPROTO_DONE;
943 		}
944 		break;
945 
946 badcode:
947 		icmp6stat.icp6s_badcode++;
948 		break;
949 
950 badlen:
951 		icmp6stat.icp6s_badlen++;
952 		break;
953 	}
954 
955 rate_limit_checked:
956 	icmp6_rip6_input(&m, *offp);
957 	return IPPROTO_DONE;
958 
959 freeit:
960 	m_freem(m);
961 	return IPPROTO_DONE;
962 }
963 
964 static int
icmp6_notify_error(struct mbuf * m,int off,int icmp6len,int code)965 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
966 {
967 	struct icmp6_hdr *icmp6;
968 	struct ip6_hdr *__single eip6;
969 	u_int32_t notifymtu;
970 	struct sockaddr_in6 icmp6src, icmp6dst;
971 
972 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
973 		icmp6stat.icp6s_tooshort++;
974 		goto freeit;
975 	}
976 #ifndef PULLDOWN_TEST
977 	IP6_EXTHDR_CHECK(m, off,
978 	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
979 	    return -1);
980 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
981 #else
982 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
983 	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
984 	if (icmp6 == NULL) {
985 		icmp6stat.icp6s_tooshort++;
986 		return -1;
987 	}
988 #endif
989 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
990 	SOCKADDR_ZERO(&icmp6dst, sizeof(icmp6dst));
991 
992 	/* Detect the upper level protocol */
993 	{
994 		void (*ctlfunc)(int, struct sockaddr *, void *, struct ifnet *);
995 		u_int8_t nxt = eip6->ip6_nxt;
996 		int eoff = off + sizeof(struct icmp6_hdr) +
997 		    sizeof(struct ip6_hdr);
998 		struct ip6ctlparam ip6cp;
999 		int icmp6type = icmp6->icmp6_type;
1000 		struct ip6_frag *__single fh;
1001 		struct ip6_rthdr *__single rth;
1002 		struct ip6_rthdr0 *rth0;
1003 		int rthlen;
1004 
1005 		while (1) { /* XXX: should avoid infinite loop explicitly? */
1006 			struct ip6_ext *eh;
1007 
1008 			switch (nxt) {
1009 			case IPPROTO_HOPOPTS:
1010 			case IPPROTO_DSTOPTS:
1011 			case IPPROTO_AH:
1012 #ifndef PULLDOWN_TEST
1013 				IP6_EXTHDR_CHECK(m, 0,
1014 				    eoff + sizeof(struct ip6_ext), return -1);
1015 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
1016 				    + eoff);
1017 #else
1018 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
1019 				    eoff, sizeof(*eh));
1020 				if (eh == NULL) {
1021 					icmp6stat.icp6s_tooshort++;
1022 					return -1;
1023 				}
1024 #endif
1025 
1026 				if (nxt == IPPROTO_AH) {
1027 					eoff += (eh->ip6e_len + 2) << 2;
1028 				} else {
1029 					eoff += (eh->ip6e_len + 1) << 3;
1030 				}
1031 				nxt = eh->ip6e_nxt;
1032 				break;
1033 			case IPPROTO_ROUTING:
1034 				/*
1035 				 * When the erroneous packet contains a
1036 				 * routing header, we should examine the
1037 				 * header to determine the final destination.
1038 				 * Otherwise, we can't properly update
1039 				 * information that depends on the final
1040 				 * destination (e.g. path MTU).
1041 				 */
1042 #ifndef PULLDOWN_TEST
1043 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
1044 				    return -1);
1045 				rth = (struct ip6_rthdr *)
1046 				    (mtod(m, caddr_t) + eoff);
1047 #else
1048 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1049 				    eoff, sizeof(*rth));
1050 				if (rth == NULL) {
1051 					icmp6stat.icp6s_tooshort++;
1052 					return -1;
1053 				}
1054 #endif
1055 				rthlen = (rth->ip6r_len + 1) << 3;
1056 				/*
1057 				 * XXX: currently there is no
1058 				 * officially defined type other
1059 				 * than type-0.
1060 				 * Note that if the segment left field
1061 				 * is 0, all intermediate hops must
1062 				 * have been passed.
1063 				 */
1064 				if (rth->ip6r_segleft &&
1065 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1066 					int hops;
1067 
1068 #ifndef PULLDOWN_TEST
1069 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
1070 					    return -1);
1071 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
1072 #else
1073 					IP6_EXTHDR_GET(rth0,
1074 					    struct ip6_rthdr0 *, m,
1075 					    eoff, rthlen);
1076 					if (rth0 == NULL) {
1077 						icmp6stat.icp6s_tooshort++;
1078 						return -1;
1079 					}
1080 #endif
1081 					/* just ignore a bogus header */
1082 					if ((rth0->ip6r0_len % 2) == 0 &&
1083 					    (hops = rth0->ip6r0_len / 2)) {
1084 						icmp6dst.sin6_addr = *((struct in6_addr *)(void *)(rth0 + 1) + (hops - 1));
1085 					}
1086 				}
1087 				eoff += rthlen;
1088 				nxt = rth->ip6r_nxt;
1089 				break;
1090 			case IPPROTO_FRAGMENT:
1091 #ifndef PULLDOWN_TEST
1092 				IP6_EXTHDR_CHECK(m, 0, eoff +
1093 				    sizeof(struct ip6_frag),
1094 				    return -1);
1095 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
1096 				    + eoff);
1097 #else
1098 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1099 				    eoff, sizeof(*fh));
1100 				if (fh == NULL) {
1101 					icmp6stat.icp6s_tooshort++;
1102 					return -1;
1103 				}
1104 #endif
1105 				/*
1106 				 * Data after a fragment header is meaningless
1107 				 * unless it is the first fragment, but
1108 				 * we'll go to the notify label for path MTU
1109 				 * discovery.
1110 				 */
1111 				if (fh->ip6f_offlg & IP6F_OFF_MASK) {
1112 					goto notify;
1113 				}
1114 
1115 				eoff += sizeof(struct ip6_frag);
1116 				nxt = fh->ip6f_nxt;
1117 				break;
1118 			default:
1119 				/*
1120 				 * This case includes ESP and the No Next
1121 				 * Header.  In such cases going to the notify
1122 				 * label does not have any meaning
1123 				 * (i.e. ctlfunc will be NULL), but we go
1124 				 * anyway since we might have to update
1125 				 * path MTU information.
1126 				 */
1127 				goto notify;
1128 			}
1129 		}
1130 notify:
1131 #ifndef PULLDOWN_TEST
1132 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1133 #else
1134 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1135 		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1136 		if (icmp6 == NULL) {
1137 			icmp6stat.icp6s_tooshort++;
1138 			return -1;
1139 		}
1140 #endif
1141 
1142 		/*
1143 		 * retrieve parameters from the inner IPv6 header, and convert
1144 		 * them into sockaddr structures.
1145 		 * XXX: there is no guarantee that the source or destination
1146 		 * addresses of the inner packet are in the same scope as
1147 		 * the addresses of the icmp packet.  But there is no other
1148 		 * way to determine the zone.
1149 		 */
1150 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1151 
1152 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1153 		icmp6dst.sin6_family = AF_INET6;
1154 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr)) {
1155 			icmp6dst.sin6_addr = eip6->ip6_dst;
1156 		}
1157 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, IN6_NULL_IF_EMBEDDED_SCOPE(&icmp6dst.sin6_scope_id))) {
1158 			goto freeit;
1159 		}
1160 		SOCKADDR_ZERO(&icmp6src, sizeof(icmp6src));
1161 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1162 		icmp6src.sin6_family = AF_INET6;
1163 		icmp6src.sin6_addr = eip6->ip6_src;
1164 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, IN6_NULL_IF_EMBEDDED_SCOPE(&icmp6dst.sin6_scope_id))) {
1165 			goto freeit;
1166 		}
1167 		icmp6src.sin6_flowinfo =
1168 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1169 
1170 		ip6cp.ip6c_m = m;
1171 		ip6cp.ip6c_icmp6 = icmp6;
1172 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1173 		ip6cp.ip6c_off = eoff;
1174 		ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1175 		ip6cp.ip6c_src = &icmp6src;
1176 		ip6cp.ip6c_nxt = nxt;
1177 
1178 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1179 			notifymtu = ntohl(icmp6->icmp6_mtu);
1180 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1181 			icmp6_mtudisc_update(&ip6cp, 1);        /*XXX*/
1182 		}
1183 
1184 		ctlfunc = ip6_protox[nxt]->pr_ctlinput;
1185 		if (ctlfunc) {
1186 			LCK_MTX_ASSERT(inet6_domain_mutex, LCK_MTX_ASSERT_OWNED);
1187 
1188 			lck_mtx_unlock(inet6_domain_mutex);
1189 
1190 			(void) (*ctlfunc)(code, SA(&icmp6dst),
1191 			    &ip6cp, m->m_pkthdr.rcvif);
1192 
1193 			lck_mtx_lock(inet6_domain_mutex);
1194 		}
1195 	}
1196 	return 0;
1197 
1198 freeit:
1199 	m_freem(m);
1200 	return -1;
1201 }
1202 
1203 void
icmp6_mtudisc_update(struct ip6ctlparam * ip6cp,int validated)1204 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1205 {
1206 	struct in6_addr *__single dst = ip6cp->ip6c_finaldst;
1207 	struct icmp6_hdr *__single icmp6 = ip6cp->ip6c_icmp6;
1208 	mbuf_ref_t m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1209 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1210 	rtentry_ref_t rt = NULL;
1211 	struct sockaddr_in6 sin6;
1212 	/*
1213 	 * we reject ICMPv6 too big with abnormally small value.
1214 	 * XXX what is the good definition of "abnormally small"?
1215 	 */
1216 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) {
1217 		return;
1218 	}
1219 
1220 	if (!validated) {
1221 		return;
1222 	}
1223 
1224 	/* Limit the MTU to the minimum IPv6 MTU */
1225 	if (mtu < IPV6_MMTU) {
1226 		mtu = IPV6_MMTU;
1227 	}
1228 
1229 	SOCKADDR_ZERO(&sin6, sizeof(sin6));
1230 	sin6.sin6_family = PF_INET6;
1231 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1232 	sin6.sin6_addr = *dst;
1233 	/* XXX normally, this won't happen */
1234 	if (in6_embedded_scope && IN6_IS_ADDR_LINKLOCAL(dst)) {
1235 		sin6.sin6_addr.s6_addr16[1] =
1236 		    htons(m->m_pkthdr.rcvif->if_index);
1237 	}
1238 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1239 	/*
1240 	 * XXX On a side note, for asymmetric data-path
1241 	 * the lookup on receive interace is probably not
1242 	 * what we want to do.
1243 	 * That requires looking at the cached route for the
1244 	 * protocol control block.
1245 	 */
1246 	rt = rtalloc1_scoped(SA(&sin6), 0,
1247 	    RTF_CLONING | RTF_PRCLONING, m->m_pkthdr.rcvif->if_index);
1248 	if (rt != NULL) {
1249 		RT_LOCK(rt);
1250 		if ((rt->rt_flags & RTF_HOST) &&
1251 		    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1252 		    mtu < IN6_LINKMTU(rt->rt_ifp) &&
1253 		    rt->rt_rmx.rmx_mtu > mtu) {
1254 			icmp6stat.icp6s_pmtuchg++;
1255 			rt->rt_rmx.rmx_mtu = mtu;
1256 		}
1257 		RT_UNLOCK(rt);
1258 		rtfree(rt);
1259 	}
1260 }
1261 
1262 /*
1263  * Process a Node Information Query packet, based on
1264  * draft-ietf-ipngwg-icmp-name-lookups-07.
1265  *
1266  * Spec incompatibilities:
1267  * - IPv6 Subject address handling
1268  * - IPv4 Subject address handling support missing
1269  * - Proxy reply (answer even if it's not for me)
1270  * - joins NI group address at in6_ifattach() time only, does not cope
1271  *   with hostname changes by sethostname(3)
1272  */
1273 #define hostnamelen     (uint32_t)strbuflen(hostname, sizeof(hostname))
1274 static struct mbuf *
ni6_input(struct mbuf * m,int off)1275 ni6_input(struct mbuf *m, int off)
1276 {
1277 	struct icmp6_nodeinfo *ni6, *nni6;
1278 	mbuf_ref_t n = NULL;
1279 	u_int16_t qtype;
1280 	int subjlen;
1281 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1282 	struct ni_reply_fqdn *__single fqdn;
1283 	int addrs;              /* for NI_QTYPE_NODEADDR */
1284 	ifnet_ref_t ifp = NULL; /* for NI_QTYPE_NODEADDR */
1285 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1286 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1287 	struct ip6_hdr *__single ip6;
1288 	int oldfqdn = 0;        /* if 1, return pascal string (03 draft) */
1289 	char *subj = NULL;
1290 
1291 	ip6 = mtod(m, struct ip6_hdr *);
1292 #ifndef PULLDOWN_TEST
1293 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1294 #else
1295 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1296 	if (ni6 == NULL) {
1297 		/* m is already reclaimed */
1298 		return NULL;
1299 	}
1300 #endif
1301 
1302 	/*
1303 	 * Validate IPv6 source address.
1304 	 * The default configuration MUST be to refuse answering queries from
1305 	 * global-scope addresses according to RFC4602.
1306 	 * Notes:
1307 	 *  - it's not very clear what "refuse" means; this implementation
1308 	 *    simply drops it.
1309 	 *  - it's not very easy to identify global-scope (unicast) addresses
1310 	 *    since there are many prefixes for them.  It should be safer
1311 	 *    and in practice sufficient to check "all" but loopback and
1312 	 *    link-local (note that site-local unicast was deprecated and
1313 	 *    ULA is defined as global scope-wise)
1314 	 */
1315 	if ((icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1316 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1317 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) {
1318 		goto bad;
1319 	}
1320 
1321 	/*
1322 	 * Validate IPv6 destination address.
1323 	 *
1324 	 * The Responder must discard the Query without further processing
1325 	 * unless it is one of the Responder's unicast or anycast addresses, or
1326 	 * a link-local scope multicast address which the Responder has joined.
1327 	 * [RFC4602, Section 5.]
1328 	 */
1329 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1330 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst) &&
1331 		    !IN6_IS_ADDR_MC_UNICAST_BASED_LINKLOCAL(&ip6->ip6_dst)) {
1332 			goto bad;
1333 		}
1334 		/* else it's a link-local multicast, fine */
1335 	} else {                /* unicast or anycast */
1336 		uint32_t ia6_flags;
1337 
1338 		if (ip6_getdstifaddr_info(m, NULL, &ia6_flags) != 0) {
1339 			goto bad; /* XXX impossible */
1340 		}
1341 		if ((ia6_flags & IN6_IFF_TEMPORARY) &&
1342 		    !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1343 			nd6log(debug, "ni6_input: ignore node info to a temporary address in %s:%d",
1344 			    __func__, __LINE__);
1345 			goto bad;
1346 		}
1347 	}
1348 
1349 	/* validate query Subject field. */
1350 	qtype = ntohs(ni6->ni_qtype);
1351 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1352 	switch (qtype) {
1353 	case NI_QTYPE_NOOP:
1354 	case NI_QTYPE_SUPTYPES:
1355 		/* 07 draft */
1356 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) {
1357 			break;
1358 		}
1359 		OS_FALLTHROUGH;
1360 	case NI_QTYPE_FQDN:
1361 	case NI_QTYPE_NODEADDR:
1362 	case NI_QTYPE_IPV4ADDR:
1363 		switch (ni6->ni_code) {
1364 		case ICMP6_NI_SUBJ_IPV6:
1365 #if ICMP6_NI_SUBJ_IPV6 != 0
1366 		case 0:
1367 #endif
1368 			/*
1369 			 * backward compatibility - try to accept 03 draft
1370 			 * format, where no Subject is present.
1371 			 */
1372 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1373 			    subjlen == 0) {
1374 				oldfqdn++;
1375 				break;
1376 			}
1377 #if ICMP6_NI_SUBJ_IPV6 != 0
1378 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) {
1379 				goto bad;
1380 			}
1381 #endif
1382 
1383 			if (subjlen != sizeof(struct in6_addr)) {
1384 				goto bad;
1385 			}
1386 
1387 			/*
1388 			 * Validate Subject address.
1389 			 *
1390 			 * Not sure what exactly "address belongs to the node"
1391 			 * means in the spec, is it just unicast, or what?
1392 			 *
1393 			 * At this moment we consider Subject address as
1394 			 * "belong to the node" if the Subject address equals
1395 			 * to the IPv6 destination address; validation for
1396 			 * IPv6 destination address should have done enough
1397 			 * check for us.
1398 			 *
1399 			 * We do not do proxy at this moment.
1400 			 */
1401 			/* m_pulldown instead of copy? */
1402 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1403 			    subjlen, (caddr_t)&sin6.sin6_addr);
1404 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1405 			    &sin6.sin6_addr);
1406 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL,
1407 			    NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&sin6.sin6_scope_id));
1408 			SOCKADDR_ZERO(&sin6_d, sizeof(sin6_d));
1409 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1410 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1411 			sin6_d.sin6_addr = ip6->ip6_dst;
1412 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1413 			    &ip6->ip6_dst);
1414 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL,
1415 			    NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&sin6_d.sin6_scope_id));
1416 			subj = (char*)__SA_UTILS_CONV_TO_BYTES(&sin6);
1417 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d)) {
1418 				break;
1419 			}
1420 
1421 			/*
1422 			 * XXX if we are to allow other cases, we should really
1423 			 * be careful about scope here.
1424 			 * basically, we should disallow queries toward IPv6
1425 			 * destination X with subject Y,
1426 			 * if scope(X) > scope(Y).
1427 			 * if we allow scope(X) > scope(Y), it will result in
1428 			 * information leakage across scope boundary.
1429 			 */
1430 			goto bad;
1431 
1432 		case ICMP6_NI_SUBJ_FQDN:
1433 			/*
1434 			 * Validate Subject name with gethostname(3).
1435 			 *
1436 			 * The behavior may need some debate, since:
1437 			 * - we are not sure if the node has FQDN as
1438 			 *   hostname (returned by gethostname(3)).
1439 			 * - the code does wildcard match for truncated names.
1440 			 *   however, we are not sure if we want to perform
1441 			 *   wildcard match, if gethostname(3) side has
1442 			 *   truncated hostname.
1443 			 */
1444 			lck_mtx_lock(&hostname_lock);
1445 			n = ni6_nametodns(hostname, hostnamelen, 0);
1446 			lck_mtx_unlock(&hostname_lock);
1447 			if (!n || n->m_next || n->m_len == 0) {
1448 				goto bad;
1449 			}
1450 			IP6_EXTHDR_GET(subj, char *, m,
1451 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1452 			if (subj == NULL) {
1453 				goto bad;
1454 			}
1455 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1456 			    n->m_len)) {
1457 				goto bad;
1458 			}
1459 			m_freem(n);
1460 			n = NULL;
1461 			break;
1462 
1463 		case ICMP6_NI_SUBJ_IPV4:        /* XXX: to be implemented? */
1464 		default:
1465 			goto bad;
1466 		}
1467 		break;
1468 	}
1469 
1470 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1471 	switch (qtype) {
1472 	case NI_QTYPE_FQDN:
1473 		if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0) {
1474 			goto bad;
1475 		}
1476 		break;
1477 	case NI_QTYPE_NODEADDR:
1478 	case NI_QTYPE_IPV4ADDR:
1479 		if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0) {
1480 			goto bad;
1481 		}
1482 		break;
1483 	}
1484 
1485 	/* guess reply length */
1486 	switch (qtype) {
1487 	case NI_QTYPE_NOOP:
1488 		break;          /* no reply data */
1489 	case NI_QTYPE_SUPTYPES:
1490 		replylen += sizeof(u_int32_t);
1491 		break;
1492 	case NI_QTYPE_FQDN:
1493 		/* XXX will append an mbuf */
1494 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1495 		break;
1496 	case NI_QTYPE_NODEADDR:
1497 		addrs = ni6_addrs(ni6, &ifp, subj);
1498 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1499 		    sizeof(u_int32_t))) > MCLBYTES) {
1500 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1501 		}
1502 		break;
1503 	case NI_QTYPE_IPV4ADDR:
1504 		/* unsupported - should respond with unknown Qtype? */
1505 		break;
1506 	default:
1507 		/*
1508 		 * XXX: We must return a reply with the ICMP6 code
1509 		 * `unknown Qtype' in this case.  However we regard the case
1510 		 * as an FQDN query for backward compatibility.
1511 		 * Older versions set a random value to this field,
1512 		 * so it rarely varies in the defined qtypes.
1513 		 * But the mechanism is not reliable...
1514 		 * maybe we should obsolete older versions.
1515 		 */
1516 		qtype = NI_QTYPE_FQDN;
1517 		/* XXX will append an mbuf */
1518 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1519 		oldfqdn++;
1520 		break;
1521 	}
1522 
1523 	/* allocate an mbuf to reply. */
1524 	MGETHDR(n, M_DONTWAIT, m->m_type);      /* MAC-OK */
1525 	if (n == NULL) {
1526 		m_freem(m);
1527 		if (ifp != NULL) {
1528 			ifnet_release(ifp);
1529 		}
1530 		return NULL;
1531 	}
1532 	M_COPY_PKTHDR(n, m); /* just for recvif */
1533 	if (replylen > MHLEN) {
1534 		if (replylen > MCLBYTES) {
1535 			/*
1536 			 * XXX: should we try to allocate more? But MCLBYTES
1537 			 * is probably much larger than IPV6_MMTU...
1538 			 */
1539 			goto bad;
1540 		}
1541 		MCLGET(n, M_DONTWAIT);
1542 		if ((n->m_flags & M_EXT) == 0) {
1543 			goto bad;
1544 		}
1545 	}
1546 	n->m_pkthdr.len = n->m_len = replylen;
1547 
1548 	/* copy mbuf header and IPv6 + Node Information base headers */
1549 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1550 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1551 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1552 
1553 	/* qtype dependent procedure */
1554 	switch (qtype) {
1555 	case NI_QTYPE_NOOP:
1556 		nni6->ni_code = ICMP6_NI_SUCCESS;
1557 		nni6->ni_flags = 0;
1558 		break;
1559 	case NI_QTYPE_SUPTYPES:
1560 	{
1561 		u_int32_t v;
1562 		nni6->ni_code = ICMP6_NI_SUCCESS;
1563 		nni6->ni_flags = htons(0x0000); /* raw bitmap */
1564 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1565 		v = (u_int32_t)htonl(0x0000000f);
1566 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1567 		break;
1568 	}
1569 	case NI_QTYPE_FQDN:
1570 		nni6->ni_code = ICMP6_NI_SUCCESS;
1571 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1572 		    sizeof(struct ip6_hdr) +
1573 		    sizeof(struct icmp6_nodeinfo));
1574 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1575 		fqdn->ni_fqdn_ttl = 0;  /* ditto. */
1576 		/*
1577 		 * XXX do we really have FQDN in variable "hostname"?
1578 		 */
1579 		lck_mtx_lock(&hostname_lock);
1580 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1581 		lck_mtx_unlock(&hostname_lock);
1582 		if (n->m_next == NULL) {
1583 			goto bad;
1584 		}
1585 		/* XXX we assume that n->m_next is not a chain */
1586 		if (n->m_next->m_next != NULL) {
1587 			goto bad;
1588 		}
1589 		n->m_pkthdr.len += n->m_next->m_len;
1590 		break;
1591 	case NI_QTYPE_NODEADDR:
1592 	{
1593 		int lenlim, copied;
1594 
1595 		nni6->ni_code = ICMP6_NI_SUCCESS;
1596 		n->m_pkthdr.len = n->m_len =
1597 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1598 		lenlim = (int)M_TRAILINGSPACE(n);
1599 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1600 		/* XXX: reset mbuf length */
1601 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1602 		    sizeof(struct icmp6_nodeinfo) + copied;
1603 		break;
1604 	}
1605 	default:
1606 		break;          /* XXX impossible! */
1607 	}
1608 
1609 	nni6->ni_type = ICMP6_NI_REPLY;
1610 	m_freem(m);
1611 	if (ifp != NULL) {
1612 		ifnet_release(ifp);
1613 	}
1614 	return n;
1615 
1616 bad:
1617 	m_freem(m);
1618 	if (n) {
1619 		m_freem(n);
1620 	}
1621 	if (ifp != NULL) {
1622 		ifnet_release(ifp);
1623 	}
1624 	return NULL;
1625 }
1626 #undef hostnamelen
1627 
1628 /*
1629  * make a mbuf with DNS-encoded string.  no compression support.
1630  *
1631  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1632  * treated as truncated name (two \0 at the end).  this is a wild guess.
1633  */
1634 static struct mbuf *
ni6_nametodns(const char * __counted_by (namelen)name,uint32_t namelen,int old)1635 ni6_nametodns(
1636 	const char *__counted_by(namelen)name,
1637 	uint32_t namelen,
1638 	int old)        /* return pascal string if non-zero */
1639 {
1640 	mbuf_ref_t m;
1641 	char *cp, *ep;
1642 	const char *p, *q;
1643 	int i, nterm;
1644 	uint32_t len;
1645 
1646 	if (old) {
1647 		len = namelen + 1;
1648 	} else {
1649 		len = MCLBYTES;
1650 	}
1651 
1652 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1653 	MGET(m, M_DONTWAIT, MT_DATA);
1654 	if (m && len > MLEN) {
1655 		MCLGET(m, M_DONTWAIT);
1656 		if ((m->m_flags & M_EXT) == 0) {
1657 			goto fail;
1658 		}
1659 	}
1660 	if (!m) {
1661 		goto fail;
1662 	}
1663 	m->m_next = NULL;
1664 
1665 	if (old) {
1666 		m->m_len = len;
1667 		*mtod(m, char *) = (char)namelen;
1668 		bcopy(name, mtod(m, char *) + 1, namelen);
1669 		return m;
1670 	} else {
1671 		m->m_len = 0;
1672 		cp = mtod(m, char *);
1673 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1674 
1675 		/* if not certain about my name, return empty buffer */
1676 		if (namelen == 0) {
1677 			return m;
1678 		}
1679 
1680 		/*
1681 		 * guess if it looks like shortened hostname, or FQDN.
1682 		 * shortened hostname needs two trailing "\0".
1683 		 */
1684 		i = 0;
1685 		for (p = name; p < name + namelen; p++) {
1686 			if (*p && *p == '.') {
1687 				i++;
1688 			}
1689 		}
1690 		if (i < 2) {
1691 			nterm = 2;
1692 		} else {
1693 			nterm = 1;
1694 		}
1695 
1696 		p = name;
1697 		while (cp < ep && p < name + namelen) {
1698 			i = 0;
1699 			for (q = p; q < name + namelen && *q && *q != '.'; q++) {
1700 				i++;
1701 			}
1702 			/* result does not fit into mbuf */
1703 			if (cp + i + 1 >= ep) {
1704 				goto fail;
1705 			}
1706 			/*
1707 			 * DNS label length restriction, RFC1035 page 8.
1708 			 * "i == 0" case is included here to avoid returning
1709 			 * 0-length label on "foo..bar".
1710 			 */
1711 			if (i <= 0 || i >= 64) {
1712 				goto fail;
1713 			}
1714 			*cp++ = (char)i;
1715 			bcopy(p, cp, i);
1716 			cp += i;
1717 			p = q;
1718 			if (p < name + namelen && *p == '.') {
1719 				p++;
1720 			}
1721 		}
1722 		/* termination */
1723 		if (cp + nterm >= ep) {
1724 			goto fail;
1725 		}
1726 		while (nterm-- > 0) {
1727 			*cp++ = '\0';
1728 		}
1729 		m->m_len = (int32_t)(cp - mtod(m, char *));
1730 		return m;
1731 	}
1732 
1733 	panic("should not reach here");
1734 	/* NOTREACHED */
1735 
1736 fail:
1737 	if (m) {
1738 		m_freem(m);
1739 	}
1740 	return NULL;
1741 }
1742 
1743 /*
1744  * check if two DNS-encoded string matches.  takes care of truncated
1745  * form (with \0\0 at the end).  no compression support.
1746  * XXX upper/lowercase match (see RFC2065)
1747  */
1748 static int
ni6_dnsmatch(const char * __counted_by (alen0)a0,int alen0,const char * __counted_by (blen0)b0,int blen0)1749 ni6_dnsmatch(const char *__counted_by(alen0)a0, int alen0, const char *__counted_by(blen0)b0, int blen0)
1750 {
1751 	const char *a = a0, *b = b0;
1752 	int l;
1753 
1754 	/* simplest case - need validation? */
1755 	if (alen0 == blen0 && bcmp(a, b, alen0) == 0) {
1756 		return 1;
1757 	}
1758 
1759 	/* termination is mandatory */
1760 	if (alen0 < 2 || blen0 < 2) {
1761 		return 0;
1762 	}
1763 	if (a0[alen0 - 1] != '\0' || b0[blen0 - 1] != '\0') {
1764 		return 0;
1765 	}
1766 
1767 	const int alen = alen0 - 1;
1768 	const int blen = blen0 - 1;
1769 
1770 	while (a - a0 < alen && b - b0 < blen) {
1771 		if (a - a0 + 1 > alen || b - b0 + 1 > blen) {
1772 			return 0;
1773 		}
1774 
1775 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0) {
1776 			return 0;
1777 		}
1778 		/* we don't support compression yet */
1779 		if (a[0] >= 64 || b[0] >= 64) {
1780 			return 0;
1781 		}
1782 
1783 		/* truncated case */
1784 		if (a[0] == 0 && a - a0 == alen - 1) {
1785 			return 1;
1786 		}
1787 		if (b[0] == 0 && b - b0 == blen - 1) {
1788 			return 1;
1789 		}
1790 		if (a[0] == 0 || b[0] == 0) {
1791 			return 0;
1792 		}
1793 
1794 		if (a[0] != b[0]) {
1795 			return 0;
1796 		}
1797 		l = a[0];
1798 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) {
1799 			return 0;
1800 		}
1801 		if (bcmp(a + 1, b + 1, l) != 0) {
1802 			return 0;
1803 		}
1804 
1805 		a += 1 + l;
1806 		b += 1 + l;
1807 	}
1808 
1809 	if (a - a0 == alen && b - b0 == blen) {
1810 		return 1;
1811 	} else {
1812 		return 0;
1813 	}
1814 }
1815 
1816 /*
1817  * calculate the number of addresses to be returned in the node info reply.
1818  */
1819 static int
ni6_addrs(struct icmp6_nodeinfo * ni6,struct ifnet ** ifpp,char * __indexable subj)1820 ni6_addrs(struct icmp6_nodeinfo *ni6, struct ifnet **ifpp, char *__indexable subj)
1821 {
1822 	ifnet_ref_t ifp;
1823 	struct in6_ifaddr *__single ifa6;
1824 	struct ifaddr *__single ifa;
1825 	struct sockaddr_in6 *__single subj_ip6 = NULL; /* XXX pedant */
1826 	int addrs = 0, addrsofif, iffound = 0;
1827 	int niflags = ni6->ni_flags;
1828 
1829 	if (ifpp != NULL) {
1830 		*ifpp = NULL;
1831 	}
1832 
1833 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1834 		switch (ni6->ni_code) {
1835 		case ICMP6_NI_SUBJ_IPV6:
1836 			if (subj == NULL) { /* must be impossible... */
1837 				return 0;
1838 			}
1839 			subj_ip6 = SIN6(subj);
1840 			break;
1841 		default:
1842 			/*
1843 			 * XXX: we only support IPv6 subject address for
1844 			 * this Qtype.
1845 			 */
1846 			return 0;
1847 		}
1848 	}
1849 
1850 	ifnet_head_lock_shared();
1851 	TAILQ_FOREACH(ifp, &ifnet_head, if_list) {
1852 		addrsofif = 0;
1853 		ifnet_lock_shared(ifp);
1854 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1855 		{
1856 			IFA_LOCK(ifa);
1857 			if (ifa->ifa_addr->sa_family != AF_INET6) {
1858 				IFA_UNLOCK(ifa);
1859 				continue;
1860 			}
1861 			ifa6 = ifatoia6(ifa);
1862 
1863 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1864 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1865 			    &ifa6->ia_addr.sin6_addr)) {
1866 				iffound = 1;
1867 			}
1868 
1869 			/*
1870 			 * IPv4-mapped addresses can only be returned by a
1871 			 * Node Information proxy, since they represent
1872 			 * addresses of IPv4-only nodes, which perforce do
1873 			 * not implement this protocol.
1874 			 * [icmp-name-lookups-07, Section 5.4]
1875 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1876 			 * this function at this moment.
1877 			 */
1878 
1879 			/* What do we have to do about ::1? */
1880 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1881 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1882 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) {
1883 					IFA_UNLOCK(ifa);
1884 					continue;
1885 				}
1886 				break;
1887 			case IPV6_ADDR_SCOPE_SITELOCAL:
1888 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) {
1889 					IFA_UNLOCK(ifa);
1890 					continue;
1891 				}
1892 				break;
1893 			case IPV6_ADDR_SCOPE_GLOBAL:
1894 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) {
1895 					IFA_UNLOCK(ifa);
1896 					continue;
1897 				}
1898 				break;
1899 			default:
1900 				IFA_UNLOCK(ifa);
1901 				continue;
1902 			}
1903 
1904 			/*
1905 			 * check if anycast is okay.
1906 			 * XXX: just experimental.  not in the spec.
1907 			 */
1908 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1909 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) {
1910 				IFA_UNLOCK(ifa);
1911 				continue; /* we need only unicast addresses */
1912 			}
1913 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1914 			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1915 				IFA_UNLOCK(ifa);
1916 				continue;
1917 			}
1918 			addrsofif++; /* count the address */
1919 			IFA_UNLOCK(ifa);
1920 		}
1921 		ifnet_lock_done(ifp);
1922 		if (iffound) {
1923 			if (ifpp != NULL) {
1924 				*ifpp = ifp;
1925 				ifnet_reference(ifp);
1926 			}
1927 			ifnet_head_done();
1928 			return addrsofif;
1929 		}
1930 
1931 		addrs += addrsofif;
1932 	}
1933 	ifnet_head_done();
1934 
1935 	return addrs;
1936 }
1937 
1938 static int
ni6_store_addrs(struct icmp6_nodeinfo * ni6,struct icmp6_nodeinfo * __indexable nni6,struct ifnet * ifp0,int resid)1939 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *__indexable nni6,
1940     struct ifnet *ifp0, int resid)
1941 {
1942 	ifnet_ref_t ifp = ifp0;
1943 	struct in6_ifaddr *__single ifa6;
1944 	struct ifaddr *__single ifa;
1945 	ifnet_ref_t ifp_dep = NULL;
1946 	int copied = 0, allow_deprecated = 0;
1947 	u_char *cp = (u_char *)(nni6 + 1);
1948 	int niflags = ni6->ni_flags;
1949 	u_int32_t ltime;
1950 	uint64_t now = net_uptime();
1951 
1952 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) {
1953 		return 0;     /* needless to copy */
1954 	}
1955 again:
1956 
1957 	ifnet_head_lock_shared();
1958 	if (ifp == NULL) {
1959 		ifp = TAILQ_FIRST(&ifnet_head);
1960 	}
1961 
1962 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1963 		ifnet_lock_shared(ifp);
1964 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1965 		    ifa = ifa->ifa_list.tqe_next) {
1966 			struct in6_addrlifetime_i *__single lt;
1967 
1968 			IFA_LOCK(ifa);
1969 			if (ifa->ifa_addr->sa_family != AF_INET6) {
1970 				IFA_UNLOCK(ifa);
1971 				continue;
1972 			}
1973 			ifa6 = ifatoia6(ifa);
1974 
1975 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1976 			    allow_deprecated == 0) {
1977 				/*
1978 				 * prefererred address should be put before
1979 				 * deprecated addresses.
1980 				 */
1981 
1982 				/* record the interface for later search */
1983 				if (ifp_dep == NULL) {
1984 					ifp_dep = ifp;
1985 				}
1986 
1987 				IFA_UNLOCK(ifa);
1988 				continue;
1989 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1990 			    allow_deprecated != 0) {
1991 				IFA_UNLOCK(ifa);
1992 				continue; /* we now collect deprecated addrs */
1993 			}
1994 			/* What do we have to do about ::1? */
1995 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1996 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1997 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) {
1998 					IFA_UNLOCK(ifa);
1999 					continue;
2000 				}
2001 				break;
2002 			case IPV6_ADDR_SCOPE_SITELOCAL:
2003 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) {
2004 					IFA_UNLOCK(ifa);
2005 					continue;
2006 				}
2007 				break;
2008 			case IPV6_ADDR_SCOPE_GLOBAL:
2009 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) {
2010 					IFA_UNLOCK(ifa);
2011 					continue;
2012 				}
2013 				break;
2014 			default:
2015 				IFA_UNLOCK(ifa);
2016 				continue;
2017 			}
2018 
2019 			/*
2020 			 * check if anycast is okay.
2021 			 * XXX: just experimental.  not in the spec.
2022 			 */
2023 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
2024 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) {
2025 				IFA_UNLOCK(ifa);
2026 				continue;
2027 			}
2028 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
2029 			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
2030 				IFA_UNLOCK(ifa);
2031 				continue;
2032 			}
2033 
2034 			/* now we can copy the address */
2035 			if (resid < sizeof(struct in6_addr) +
2036 			    sizeof(u_int32_t)) {
2037 				IFA_UNLOCK(ifa);
2038 				/*
2039 				 * We give up much more copy.
2040 				 * Set the truncate flag and return.
2041 				 */
2042 				nni6->ni_flags |=
2043 				    NI_NODEADDR_FLAG_TRUNCATE;
2044 				ifnet_lock_done(ifp);
2045 				ifnet_head_done();
2046 				return copied;
2047 			}
2048 
2049 			/*
2050 			 * Set the TTL of the address.
2051 			 * The TTL value should be one of the following
2052 			 * according to the specification:
2053 			 *
2054 			 * 1. The remaining lifetime of a DHCP lease on the
2055 			 *    address, or
2056 			 * 2. The remaining Valid Lifetime of a prefix from
2057 			 *    which the address was derived through Stateless
2058 			 *    Autoconfiguration.
2059 			 *
2060 			 * Note that we currently do not support stateful
2061 			 * address configuration by DHCPv6, so the former
2062 			 * case can't happen.
2063 			 */
2064 			lt = &ifa6->ia6_lifetime;
2065 			if (lt->ia6ti_expire == 0) {
2066 				ltime = ND6_INFINITE_LIFETIME;
2067 			} else {
2068 				if (lt->ia6ti_expire > now) {
2069 					ltime = htonl((uint32_t)(lt->ia6ti_expire - now));
2070 				} else {
2071 					ltime = 0;
2072 				}
2073 			}
2074 
2075 			bcopy(&ltime, cp, sizeof(u_int32_t));
2076 			cp += sizeof(u_int32_t);
2077 
2078 			/* copy the address itself */
2079 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
2080 			    sizeof(struct in6_addr));
2081 			/* XXX: KAME link-local hack; remove ifindex */
2082 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr)) {
2083 				((struct in6_addr *)(void *)cp)->s6_addr16[1] = 0;
2084 			}
2085 			cp += sizeof(struct in6_addr);
2086 
2087 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
2088 			copied += (sizeof(struct in6_addr) +
2089 			    sizeof(u_int32_t));
2090 			IFA_UNLOCK(ifa);
2091 		}
2092 		ifnet_lock_done(ifp);
2093 		if (ifp0) {     /* we need search only on the specified IF */
2094 			break;
2095 		}
2096 	}
2097 	ifnet_head_done();
2098 
2099 	if (allow_deprecated == 0 && ifp_dep != NULL) {
2100 		ifp = ifp_dep;
2101 		allow_deprecated = 1;
2102 
2103 		goto again;
2104 	}
2105 
2106 	return copied;
2107 }
2108 
2109 /*
2110  * XXX almost dup'ed code with rip6_input.
2111  */
2112 static int
icmp6_rip6_input(struct mbuf ** mp,int off)2113 icmp6_rip6_input(struct mbuf **mp, int off)
2114 {
2115 	mbuf_ref_t m = *mp;
2116 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2117 	struct in6pcb *__single in6p;
2118 	struct in6pcb *__single last = NULL;
2119 	struct sockaddr_in6 rip6src;
2120 	struct icmp6_hdr *icmp6;
2121 	mbuf_ref_t opts = NULL;
2122 	int ret = 0;
2123 	ifnet_ref_t ifp = m->m_pkthdr.rcvif;
2124 
2125 #ifndef PULLDOWN_TEST
2126 	/* this is assumed to be safe. */
2127 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
2128 #else
2129 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
2130 	if (icmp6 == NULL) {
2131 		/* m is already reclaimed */
2132 		return IPPROTO_DONE;
2133 	}
2134 #endif
2135 
2136 	/*
2137 	 * XXX: the address may have embedded scope zone ID, which should be
2138 	 * hidden from applications.
2139 	 */
2140 	SOCKADDR_ZERO(&rip6src, sizeof(rip6src));
2141 	rip6src.sin6_family = AF_INET6;
2142 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
2143 	rip6src.sin6_addr = ip6->ip6_src;
2144 	if (!in6_embedded_scope) {
2145 		rip6src.sin6_scope_id = IN6_IS_SCOPE_EMBED(&rip6src.sin6_addr) ? ip6_input_getsrcifscope(m) : IFSCOPE_NONE;
2146 	}
2147 	if (sa6_recoverscope(&rip6src, TRUE)) {
2148 		return IPPROTO_DONE;
2149 	}
2150 
2151 	lck_rw_lock_shared(&ripcbinfo.ipi_lock);
2152 	LIST_FOREACH(in6p, &ripcb, inp_list)
2153 	{
2154 		if ((in6p->inp_vflag & INP_IPV6) == 0) {
2155 			continue;
2156 		}
2157 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6) {
2158 			continue;
2159 		}
2160 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2161 		    !in6_are_addr_equal_scoped(&in6p->in6p_laddr, &ip6->ip6_dst, in6p->inp_lifscope, ifp->if_index)) {
2162 			continue;
2163 		}
2164 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2165 		    !in6_are_addr_equal_scoped(&in6p->in6p_faddr, &ip6->ip6_src, in6p->inp_fifscope, ifp->if_index)) {
2166 			continue;
2167 		}
2168 		if (in6p->in6p_icmp6filt
2169 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2170 		    in6p->in6p_icmp6filt)) {
2171 			continue;
2172 		}
2173 
2174 		if (inp_restricted_recv(in6p, ifp)) {
2175 			continue;
2176 		}
2177 
2178 		if (last) {
2179 			mbuf_ref_t n;
2180 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2181 				if ((last->in6p_flags & INP_CONTROLOPTS) != 0 ||
2182 				    SOFLOW_ENABLED(last->in6p_socket) ||
2183 				    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
2184 					ret = ip6_savecontrol(last, n, &opts);
2185 					if (ret != 0) {
2186 						m_freem(n);
2187 						m_freem(opts);
2188 						last = in6p;
2189 						continue;
2190 					}
2191 				}
2192 				/* strip intermediate headers */
2193 				m_adj(n, off);
2194 				so_recv_data_stat(last->in6p_socket, m, 0);
2195 				if (sbappendaddr(&last->in6p_socket->so_rcv,
2196 				    SA(&rip6src),
2197 				    n, opts, NULL) != 0) {
2198 					sorwakeup(last->in6p_socket);
2199 				}
2200 				opts = NULL;
2201 			}
2202 		}
2203 		last = in6p;
2204 	}
2205 	if (last) {
2206 		if ((last->in6p_flags & INP_CONTROLOPTS) != 0 ||
2207 		    SOFLOW_ENABLED(last->in6p_socket) ||
2208 		    SO_RECV_CONTROL_OPTS(last->inp_socket)) {
2209 			ret = ip6_savecontrol(last, m, &opts);
2210 			if (ret != 0) {
2211 				goto error;
2212 			}
2213 		}
2214 		/* strip intermediate headers */
2215 		m_adj(m, off);
2216 		so_recv_data_stat(last->in6p_socket, m, 0);
2217 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2218 		    SA(&rip6src), m, opts, NULL) != 0) {
2219 			sorwakeup(last->in6p_socket);
2220 		}
2221 	} else {
2222 		goto error;
2223 	}
2224 	lck_rw_done(&ripcbinfo.ipi_lock);
2225 	return IPPROTO_DONE;
2226 
2227 error:
2228 	lck_rw_done(&ripcbinfo.ipi_lock);
2229 	m_freem(m);
2230 	m_freem(opts);
2231 	ip6stat.ip6s_delivered--;
2232 	return IPPROTO_DONE;
2233 }
2234 
2235 /*
2236  * Reflect the ip6 packet back to the source.
2237  * OFF points to the icmp6 header, counted from the top of the mbuf.
2238  */
2239 void
icmp6_reflect(struct mbuf * m,size_t off)2240 icmp6_reflect(struct mbuf *m, size_t off)
2241 {
2242 	mbuf_ref_t m_ip6hdr = m;
2243 	struct ip6_hdr *ip6;
2244 	struct icmp6_hdr *icmp6;
2245 	struct in6_ifaddr *__single ia;
2246 	struct in6_addr t, src_storage;
2247 	struct in6_addr *__single src = 0;
2248 	int plen;
2249 	int type, code;
2250 	ifnet_ref_t outif = NULL;
2251 	struct sockaddr_in6 sa6_src, sa6_dst;
2252 	struct nd_ifinfo *__single ndi = NULL;
2253 	u_int32_t oflow;
2254 	uint32_t sifscope = IFSCOPE_NONE;
2255 	uint32_t fifscope = IFSCOPE_NONE;
2256 	uint32_t tifscope;
2257 	struct ip6_out_args ip6oa;
2258 
2259 	bzero(&ip6oa, sizeof(ip6oa));
2260 	ip6oa.ip6oa_boundif = IFSCOPE_NONE;
2261 	ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
2262 	    IP6OAF_INTCOPROC_ALLOWED | IP6OAF_AWDL_UNRESTRICTED |
2263 	    IP6OAF_MANAGEMENT_ALLOWED | IPOAF_ULTRA_CONSTRAINED_ALLOWED;
2264 	ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
2265 	ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
2266 
2267 	if (!(m->m_pkthdr.pkt_flags & PKTF_LOOP) && m->m_pkthdr.rcvif != NULL) {
2268 		ip6oa.ip6oa_boundif = m->m_pkthdr.rcvif->if_index;
2269 		ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
2270 	}
2271 
2272 	/* too short to reflect */
2273 	if (off < sizeof(struct ip6_hdr)) {
2274 		nd6log(debug,
2275 		    "sanity fail: off=%x, sizeof(ip6)=%x in %s:%d\n",
2276 		    (u_int32_t)off, (u_int32_t)sizeof(struct ip6_hdr),
2277 		    __func__, __LINE__);
2278 		goto bad;
2279 	}
2280 
2281 	/*
2282 	 * If there are extra headers between IPv6 and ICMPv6, strip
2283 	 * off that header first.
2284 	 */
2285 	if (off > sizeof(struct ip6_hdr)) {
2286 		size_t l;
2287 		struct ip6_hdr nip6;
2288 
2289 		l = off - sizeof(struct ip6_hdr);
2290 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2291 		m_adj(m, (int)l);
2292 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2293 		if (m->m_len < l) {
2294 			if ((m_ip6hdr = m_pulldown(m, 0, (int)l, NULL)) == NULL) {
2295 				return;
2296 			}
2297 		}
2298 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2299 	} else { /* off == sizeof(struct ip6_hdr) */
2300 		size_t l;
2301 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2302 		if (m->m_len < l) {
2303 			if ((m_ip6hdr = m_pulldown(m, 0, (int)l, NULL)) == NULL) {
2304 				return;
2305 			}
2306 		}
2307 	}
2308 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2309 	ip6 = mtod(m_ip6hdr, struct ip6_hdr *);
2310 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2311 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2312 	type = icmp6->icmp6_type; /* keep type for statistics */
2313 	code = icmp6->icmp6_code; /* ditto. */
2314 	if (m->m_pkthdr.rcvif != NULL) {
2315 		fifscope = m->m_pkthdr.rcvif->if_index;
2316 		sifscope = m->m_pkthdr.rcvif->if_index;
2317 	} else if (m->m_pkthdr.pkt_flags & PKTF_IFAINFO) {
2318 		ip6_getsrcifaddr_info(m, &fifscope, NULL);
2319 		ip6_getdstifaddr_info(m, &sifscope, NULL);
2320 	} else if (m->m_pkthdr.pkt_ext_flags & PKTF_EXT_OUTPUT_SCOPE) {
2321 		fifscope = ip6_output_getsrcifscope(m);
2322 		sifscope = ip6_output_getdstifscope(m);
2323 	}
2324 	fifscope = IN6_IS_SCOPE_EMBED(&ip6->ip6_src) ? fifscope : IFSCOPE_NONE;
2325 	sifscope = IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) ? sifscope : IFSCOPE_NONE;
2326 	if (!in6_embedded_scope) {
2327 		m->m_pkthdr.pkt_flags &= ~PKTF_IFAINFO;
2328 	}
2329 
2330 	t = ip6->ip6_dst;
2331 	/*
2332 	 * ip6_input() drops a packet if its src is multicast.
2333 	 * So, the src is never multicast.
2334 	 */
2335 	ip6->ip6_dst = ip6->ip6_src;
2336 	/*
2337 	 * XXX: make sure to embed scope zone information, using
2338 	 * already embedded IDs or the received interface (if any).
2339 	 * Note that rcvif may be NULL.
2340 	 */
2341 	SOCKADDR_ZERO(&sa6_src, sizeof(sa6_src));
2342 	sa6_src.sin6_family = AF_INET6;
2343 	sa6_src.sin6_len = sizeof(sa6_src);
2344 	sa6_src.sin6_addr = ip6->ip6_dst;
2345 	if (!in6_embedded_scope) {
2346 		sa6_src.sin6_scope_id = fifscope;
2347 	}
2348 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2349 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL, NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&sa6_src.sin6_scope_id));
2350 	in6_verify_ifscope(&ip6->ip6_dst, sa6_src.sin6_scope_id);
2351 	ip6_output_setdstifscope(m, sa6_src.sin6_scope_id, NULL);
2352 
2353 	if (!in6_embedded_scope &&
2354 	    (ip6oa.ip6oa_flags & IP6OAF_BOUND_IF) == 0 &&
2355 	    IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
2356 		ip6oa.ip6oa_boundif = sa6_src.sin6_scope_id;
2357 		ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
2358 	}
2359 
2360 	SOCKADDR_ZERO(&sa6_dst, sizeof(sa6_dst));
2361 	sa6_dst.sin6_family = AF_INET6;
2362 	sa6_dst.sin6_len = sizeof(sa6_dst);
2363 	sa6_dst.sin6_addr = t;
2364 	if (!in6_embedded_scope) {
2365 		sa6_dst.sin6_scope_id = sifscope;
2366 	}
2367 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2368 	in6_embedscope(&t, &sa6_dst, NULL, NULL, NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&sa6_dst.sin6_scope_id));
2369 	tifscope = sa6_dst.sin6_scope_id;
2370 	/*
2371 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2372 	 * use dst as the src for the reply.
2373 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2374 	 * (for example) when we encounter an error while forwarding procedure
2375 	 * destined to a duplicated address of ours.
2376 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2377 	 * procedure of an outgoing packet of our own, in which case we need
2378 	 * to search in the ifaddr list.
2379 	 */
2380 	lck_rw_lock_shared(&in6_ifaddr_rwlock);
2381 	TAILQ_FOREACH(ia, IN6ADDR_HASH(&t), ia6_hash) {
2382 		IFA_LOCK(&ia->ia_ifa);
2383 		if (in6_are_addr_equal_scoped(&t, &ia->ia_addr.sin6_addr, tifscope, ia->ia_addr.sin6_scope_id) &&
2384 		    (ia->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY)) == 0) {
2385 			IFA_UNLOCK(&ia->ia_ifa);
2386 			src = &t;
2387 			sifscope = tifscope;
2388 			break;
2389 		}
2390 		IFA_UNLOCK(&ia->ia_ifa);
2391 	}
2392 	lck_rw_done(&in6_ifaddr_rwlock);
2393 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) &&
2394 	    ((m->m_flags & M_LOOP) || (m->m_pkthdr.pkt_flags & PKTF_LOOP))) {
2395 		/*
2396 		 * This is the case if the dst is our link-local address
2397 		 * and the sender is also ourselves.  Here we test for both
2398 		 * M_LOOP and PKTF_LOOP, since the former may have been set
2399 		 * in ip6_output() and that we get here as part of callling
2400 		 * ip6_process_hopopts().  See comments in <sys/mbuf.h>
2401 		 */
2402 		src = &t;
2403 		sifscope = tifscope;
2404 	}
2405 
2406 	if (src == NULL) {
2407 		int e;
2408 		struct sockaddr_in6 sin6;
2409 		struct route_in6 ro;
2410 
2411 		/*
2412 		 * This case matches to multicasts, our anycast, or unicasts
2413 		 * that we do not own.  Select a source address based on the
2414 		 * source address of the erroneous packet.
2415 		 */
2416 		SOCKADDR_ZERO(&sin6, sizeof(sin6));
2417 		sin6.sin6_family = AF_INET6;
2418 		sin6.sin6_len = sizeof(sin6);
2419 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2420 		if (!in6_embedded_scope) {
2421 			sin6.sin6_scope_id = fifscope;
2422 		}
2423 
2424 		bzero(&ro, sizeof(ro));
2425 		/*
2426 		 * in6_selectsrc() might return outif with its reference held
2427 		 * even in the error case, so we always need to release it
2428 		 * if non-NULL.
2429 		 */
2430 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, &outif,
2431 		    &src_storage, ip6oa.ip6oa_boundif, &e);
2432 		ROUTE_RELEASE(&ro);
2433 		if (src == NULL) {
2434 			nd6log(debug,
2435 			    "icmp6_reflect: source can't be determined: "
2436 			    "dst=%s, error=%d\n",
2437 			    ip6_sprintf(&sa6_src.sin6_addr), e);
2438 			goto bad;
2439 		}
2440 		if (outif != NULL) {
2441 			sifscope = outif->if_index;
2442 		}
2443 	}
2444 	oflow = ip6->ip6_flow; /* Save for later */
2445 	ip6->ip6_src = *src;
2446 	ip6->ip6_flow = 0;
2447 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2448 	ip6->ip6_vfc |= IPV6_VERSION;
2449 	ip6_output_setsrcifscope(m, sifscope, NULL);
2450 	in6_verify_ifscope(&ip6->ip6_src, sifscope);
2451 
2452 	if (icmp6->icmp6_type == ICMP6_ECHO_REPLY && icmp6->icmp6_code == 0) {
2453 		ip6->ip6_flow |= (oflow & htonl(0x0ff00000));
2454 	}
2455 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2456 	if (outif != NULL && (ndi = ND_IFINFO(outif)) != NULL &&
2457 	    ndi->initialized) {
2458 		lck_mtx_lock(&ndi->lock);
2459 		ip6->ip6_hlim = ndi->chlim;
2460 		lck_mtx_unlock(&ndi->lock);
2461 	}
2462 	if (m->m_pkthdr.rcvif != NULL &&
2463 	    (ndi = ND_IFINFO(m->m_pkthdr.rcvif)) != NULL &&
2464 	    ndi->initialized) {
2465 		/* XXX: This may not be the outgoing interface */
2466 		lck_mtx_lock(&ndi->lock);
2467 		ip6->ip6_hlim = ndi->chlim;
2468 		lck_mtx_unlock(&ndi->lock);
2469 	} else {
2470 		ip6->ip6_hlim = (uint8_t)ip6_defhlim;
2471 	}
2472 	/* Use the same traffic class as in the request to match IPv4 */
2473 	icmp6->icmp6_cksum = 0;
2474 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2475 	    sizeof(struct ip6_hdr), plen);
2476 
2477 	/*
2478 	 * XXX option handling
2479 	 */
2480 	m->m_flags &= ~(M_BCAST | M_MCAST);
2481 
2482 	if (outif != NULL) {
2483 		ifnet_release(outif);
2484 		outif = NULL;
2485 	}
2486 
2487 	m->m_pkthdr.csum_data = 0;
2488 	m->m_pkthdr.csum_flags = 0;
2489 	ip6_output(m, NULL, NULL, IPV6_OUTARGS, NULL, &outif, &ip6oa);
2490 	if (outif != NULL) {
2491 		icmp6_ifoutstat_inc(outif, type, code);
2492 		ifnet_release(outif);
2493 	}
2494 	return;
2495 
2496 bad:
2497 	m_freem(m);
2498 	if (outif != NULL) {
2499 		ifnet_release(outif);
2500 	}
2501 	return;
2502 }
2503 
2504 static const char *
icmp6_redirect_diag(struct in6_addr * src6,struct in6_addr * dst6,struct in6_addr * tgt6)2505 icmp6_redirect_diag(struct in6_addr *src6,
2506     struct in6_addr *dst6,
2507     struct in6_addr *tgt6)
2508 {
2509 	static char buf[1024];
2510 	return tsnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2511 	           ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2512 }
2513 
2514 void
icmp6_redirect_input(struct mbuf * m,int off,int icmp6len)2515 icmp6_redirect_input(struct mbuf *m, int off, int icmp6len)
2516 {
2517 	ifnet_ref_t ifp = NULL;
2518 	struct ip6_hdr *__single ip6 = NULL;
2519 	struct nd_redirect *nd_rd = NULL;
2520 	char *lladdr = NULL;
2521 	int lladdrlen = 0;
2522 	char *redirhdr = NULL;
2523 	int redirhdrlen = 0;
2524 	rtentry_ref_t rt = NULL;
2525 	int is_router = 0;
2526 	int is_onlink = 0;
2527 	struct in6_addr src6 = {};
2528 	struct in6_addr redtgt6 = {};
2529 	struct in6_addr reddst6 = {};
2530 	uint32_t src_ifscope = IFSCOPE_NONE, redtgt_ifscope  = IFSCOPE_NONE, reddst_ifscope = IFSCOPE_NONE;
2531 	union nd_opts ndopts = {};
2532 
2533 	if (m == NULL) {
2534 		return;
2535 	}
2536 
2537 	ifp = m->m_pkthdr.rcvif;
2538 	if (ifp == NULL) {
2539 		goto freeit;
2540 	}
2541 
2542 	/*
2543 	 * If we are an advertising router on this interface,
2544 	 * don't update route by icmp6 redirect.
2545 	 */
2546 	if (ifp->if_ipv6_router_mode == IPV6_ROUTER_MODE_EXCLUSIVE) {
2547 		goto freeit;
2548 	}
2549 	if (!icmp6_rediraccept) {
2550 		goto freeit;
2551 	}
2552 
2553 	ip6 = mtod(m, struct ip6_hdr *);
2554 	src6 = ip6->ip6_src;
2555 	src_ifscope = in6_addr2scopeid(ifp, &src6);
2556 #ifndef PULLDOWN_TEST
2557 	IP6_EXTHDR_CHECK(m, off, icmp6len, return );
2558 	nd_rd = (struct nd_redirect *)(mtod(m, caddr_t) + off);
2559 #else
2560 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2561 	if (nd_rd == NULL) {
2562 		icmp6stat.icp6s_tooshort++;
2563 		goto freeit;
2564 	}
2565 #endif
2566 	ip6 = mtod(m, struct ip6_hdr *);
2567 
2568 	redtgt6 = nd_rd->nd_rd_target;
2569 	reddst6 = nd_rd->nd_rd_dst;
2570 
2571 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, &redtgt_ifscope) ||
2572 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, &reddst_ifscope)) {
2573 		goto freeit;
2574 	}
2575 
2576 	/* validation */
2577 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2578 		nd6log(error,
2579 		    "ICMP6 redirect sent from %s rejected; "
2580 		    "must be from linklocal\n", ip6_sprintf(&src6));
2581 		goto bad;
2582 	}
2583 	if (ip6->ip6_hlim != IPV6_MAXHLIM) {
2584 		nd6log(error,
2585 		    "ICMP6 redirect sent from %s rejected; "
2586 		    "hlim=%d (must be 255)\n",
2587 		    ip6_sprintf(&src6), ip6->ip6_hlim);
2588 		goto bad;
2589 	}
2590 	{
2591 		/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2592 		struct sockaddr_in6 sin6;
2593 		struct in6_addr *__single gw6;
2594 
2595 		SOCKADDR_ZERO(&sin6, sizeof(sin6));
2596 		sin6.sin6_family = AF_INET6;
2597 		sin6.sin6_len = sizeof(struct sockaddr_in6);
2598 		if (!in6_embedded_scope) {
2599 			sin6.sin6_scope_id = reddst_ifscope;
2600 		}
2601 		bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2602 		rt = rtalloc1_scoped(SA(&sin6), 0, 0, ifp->if_index);
2603 		if (rt) {
2604 			RT_LOCK(rt);
2605 			if (rt->rt_gateway == NULL ||
2606 			    rt->rt_gateway->sa_family != AF_INET6) {
2607 				nd6log(error,
2608 				    "ICMP6 redirect rejected; no route "
2609 				    "with inet6 gateway found for redirect dst: %s\n",
2610 				    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2611 				RT_UNLOCK(rt);
2612 				rtfree(rt);
2613 				goto bad;
2614 			}
2615 
2616 			gw6 = &((SIN6(rt->rt_gateway))->sin6_addr);
2617 			if (!in6_are_addr_equal_scoped(&src6, gw6, src_ifscope, (SIN6(rt->rt_gateway))->sin6_scope_id)) {
2618 				nd6log(error,
2619 				    "ICMP6 redirect rejected; "
2620 				    "not equal to gw-for-src=%s (must be same): "
2621 				    "%s\n",
2622 				    ip6_sprintf(gw6),
2623 				    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2624 				RT_UNLOCK(rt);
2625 				rtfree(rt);
2626 				goto bad;
2627 			}
2628 		} else {
2629 			nd6log(error,
2630 			    "ICMP6 redirect rejected; "
2631 			    "no route found for redirect dst: %s\n",
2632 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2633 			goto bad;
2634 		}
2635 		RT_UNLOCK(rt);
2636 		rtfree(rt);
2637 		rt = NULL;
2638 	}
2639 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2640 		nd6log(error,
2641 		    "ICMP6 redirect rejected; "
2642 		    "redirect dst must be unicast: %s\n",
2643 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2644 		goto bad;
2645 	}
2646 
2647 	is_router = is_onlink = 0;
2648 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) {
2649 		is_router = 1;  /* router case */
2650 	}
2651 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) {
2652 		is_onlink = 1;  /* on-link destination case */
2653 	}
2654 	if (!is_router && !is_onlink) {
2655 		nd6log(error,
2656 		    "ICMP6 redirect rejected; "
2657 		    "neither router case nor onlink case: %s\n",
2658 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2659 		goto bad;
2660 	}
2661 	/* validation passed */
2662 
2663 	icmp6len -= sizeof(*nd_rd);
2664 
2665 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2666 	if (nd6_options(&ndopts) < 0) {
2667 		nd6log(info, "icmp6_redirect_input: "
2668 		    "invalid ND option, rejected: %s\n",
2669 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2670 		/* nd6_options have incremented stats */
2671 		goto freeit;
2672 	}
2673 
2674 	if (ndopts.nd_opts_tgt_lladdr) {
2675 		ND_OPT_LLADDR(ndopts.nd_opts_tgt_lladdr, nd_opt_len, lladdr, lladdrlen);
2676 	}
2677 
2678 	if (ndopts.nd_opts_rh) {
2679 		ND_OPT_LLADDR(ndopts.nd_opts_rh, nd_opt_rh_len, redirhdr, redirhdrlen);
2680 	}
2681 
2682 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2683 		nd6log(info,
2684 		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2685 		    "(if %d, icmp6 packet %d): %s\n",
2686 		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2687 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2688 		goto bad;
2689 	}
2690 
2691 	/* RFC 2461 8.3 */
2692 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2693 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER, NULL);
2694 
2695 	if (!is_onlink) {       /* better router case.  perform rtredirect. */
2696 		/* perform rtredirect */
2697 		struct sockaddr_in6 sdst;
2698 		struct sockaddr_in6 sgw;
2699 		struct sockaddr_in6 ssrc;
2700 
2701 		SOCKADDR_ZERO(&sdst, sizeof(sdst));
2702 		SOCKADDR_ZERO(&sgw, sizeof(sgw));
2703 		SOCKADDR_ZERO(&ssrc, sizeof(ssrc));
2704 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2705 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2706 		    sizeof(struct sockaddr_in6);
2707 		if (!in6_embedded_scope) {
2708 			sdst.sin6_scope_id = reddst_ifscope;
2709 			sgw.sin6_scope_id = redtgt_ifscope;
2710 			ssrc.sin6_scope_id = src_ifscope;
2711 		}
2712 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2713 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2714 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2715 
2716 		rtredirect(ifp, SA(&sdst),
2717 		    SA(&sgw), NULL, RTF_GATEWAY | RTF_HOST,
2718 		    SA(&ssrc), NULL);
2719 	}
2720 	/* finally update cached route in each socket via pfctlinput */
2721 	{
2722 		struct sockaddr_in6 sdst;
2723 
2724 		SOCKADDR_ZERO(&sdst, sizeof(sdst));
2725 		sdst.sin6_family = AF_INET6;
2726 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2727 		if (!in6_embedded_scope) {
2728 			sdst.sin6_scope_id = reddst_ifscope;
2729 		}
2730 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2731 
2732 		pfctlinput(PRC_REDIRECT_HOST, SA(&sdst));
2733 #if IPSEC
2734 		key_sa_routechange(SA(&sdst));
2735 #endif
2736 	}
2737 
2738 freeit:
2739 	m_freem(m);
2740 	return;
2741 
2742 bad:
2743 	icmp6stat.icp6s_badredirect++;
2744 	m_freem(m);
2745 }
2746 
2747 void
icmp6_redirect_output(struct mbuf * m0,struct rtentry * rt)2748 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2749 {
2750 	ifnet_ref_t ifp;      /* my outgoing interface */
2751 	struct in6_addr ifp_ll6;
2752 	struct in6_addr *__single router_ll6;
2753 	struct ip6_hdr *__single sip6;   /* m0 as struct ip6_hdr */
2754 	mbuf_ref_t m = NULL;  /* newly allocated one */
2755 	struct ip6_hdr *ip6;             /* m as struct ip6_hdr */
2756 	struct nd_redirect *nd_rd;
2757 	size_t maxlen;
2758 	u_char *p;
2759 	ifnet_ref_t outif = NULL;
2760 	struct sockaddr_in6 src_sa;
2761 	struct ip6_out_args ip6oa;
2762 
2763 	bzero(&ip6oa, sizeof(ip6oa));
2764 	ip6oa.ip6oa_boundif = IFSCOPE_NONE;
2765 	ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR;
2766 	ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
2767 	ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
2768 
2769 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2770 
2771 	if (rt != NULL) {
2772 		RT_LOCK(rt);
2773 	}
2774 
2775 	/* sanity check */
2776 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) {
2777 		goto fail;
2778 	}
2779 
2780 	/*
2781 	 * If we are not a router to begin with, or not an advertising
2782 	 * router on this interface, don't send icmp6 redirect.
2783 	 */
2784 	if (!ip6_forwarding || ifp->if_ipv6_router_mode != IPV6_ROUTER_MODE_EXCLUSIVE) {
2785 		goto fail;
2786 	}
2787 
2788 	/*
2789 	 * Address check:
2790 	 *  the source address must identify a neighbor, and
2791 	 *  the destination address must not be a multicast address
2792 	 *  [RFC 2461, sec 8.2]
2793 	 */
2794 	sip6 = mtod(m0, struct ip6_hdr *);
2795 	SOCKADDR_ZERO(&src_sa, sizeof(src_sa));
2796 	src_sa.sin6_family = AF_INET6;
2797 	src_sa.sin6_len = sizeof(src_sa);
2798 	src_sa.sin6_addr = sip6->ip6_src;
2799 	/* we don't currently use sin6_scope_id, but eventually use it */
2800 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2801 
2802 	RT_UNLOCK(rt);
2803 	if (nd6_is_addr_neighbor(&src_sa, ifp, 0) == 0) {
2804 		/* already unlocked */
2805 		rt = NULL;
2806 		goto fail;
2807 	}
2808 	RT_LOCK(rt);
2809 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) {
2810 		goto fail;      /* what should we do here? */
2811 	}
2812 	/* rate limit */
2813 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) {
2814 		goto fail;
2815 	}
2816 
2817 	/*
2818 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2819 	 * we almost always ask for an mbuf cluster for simplicity.
2820 	 * (MHLEN < IPV6_MMTU is almost always true)
2821 	 */
2822 #if IPV6_MMTU >= MCLBYTES
2823 # error assumption failed about IPV6_MMTU and MCLBYTES
2824 #endif
2825 	MGETHDR(m, M_DONTWAIT, MT_HEADER);      /* MAC-OK */
2826 	if (m && IPV6_MMTU >= MHLEN) {
2827 		MCLGET(m, M_DONTWAIT);
2828 	}
2829 	if (!m) {
2830 		goto fail;
2831 	}
2832 	m->m_pkthdr.rcvif = NULL;
2833 	m->m_len = 0;
2834 	maxlen = M_TRAILINGSPACE(m);
2835 	maxlen = MIN(IPV6_MMTU, maxlen);
2836 	/* just for safety */
2837 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2838 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2839 		goto fail;
2840 	}
2841 
2842 	{
2843 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2844 		struct in6_ifaddr *ia;
2845 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2846 		    IN6_IFF_NOTREADY |
2847 		    IN6_IFF_ANYCAST)) == NULL) {
2848 			goto fail;
2849 		}
2850 		IFA_LOCK(&ia->ia_ifa);
2851 		ifp_ll6 = ia->ia_addr.sin6_addr;
2852 		IFA_UNLOCK(&ia->ia_ifa);
2853 		ifa_remref(&ia->ia_ifa);
2854 	}
2855 
2856 	/* get ip6 linklocal address for the router. */
2857 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2858 		struct sockaddr_in6 *__single sin6;
2859 		sin6 = SIN6(rt->rt_gateway);
2860 		router_ll6 = &sin6->sin6_addr;
2861 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6)) {
2862 			router_ll6 = (struct in6_addr *)NULL;
2863 		}
2864 	} else {
2865 		router_ll6 = (struct in6_addr *)NULL;
2866 	}
2867 
2868 	/* ip6 */
2869 	ip6 = mtod(m, struct ip6_hdr *);
2870 	ip6->ip6_flow = 0;
2871 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2872 	ip6->ip6_vfc |= IPV6_VERSION;
2873 	/* ip6->ip6_plen will be set later */
2874 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2875 	ip6->ip6_hlim = IPV6_MAXHLIM;
2876 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2877 	bcopy(&ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2878 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2879 
2880 	/* ND Redirect */
2881 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2882 	nd_rd->nd_rd_type = ND_REDIRECT;
2883 	nd_rd->nd_rd_code = 0;
2884 	nd_rd->nd_rd_reserved = 0;
2885 	if (rt->rt_flags & RTF_GATEWAY) {
2886 		/*
2887 		 * nd_rd->nd_rd_target must be a link-local address in
2888 		 * better router cases.
2889 		 */
2890 		if (!router_ll6) {
2891 			goto fail;
2892 		}
2893 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2894 		    sizeof(nd_rd->nd_rd_target));
2895 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2896 		    sizeof(nd_rd->nd_rd_dst));
2897 	} else {
2898 		/* make sure redtgt == reddst */
2899 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2900 		    sizeof(nd_rd->nd_rd_target));
2901 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2902 		    sizeof(nd_rd->nd_rd_dst));
2903 	}
2904 	RT_UNLOCK(rt);
2905 	rt = NULL;
2906 
2907 	p = (u_char *)(nd_rd + 1);
2908 
2909 	if (!router_ll6) {
2910 		goto nolladdropt;
2911 	}
2912 
2913 	{
2914 		/* target lladdr option */
2915 		rtentry_ref_t rt_router = NULL;
2916 		int len;
2917 		struct sockaddr_dl *__single sdl;
2918 		struct nd_opt_hdr *nd_opt;
2919 		char *lladdr;
2920 
2921 		/* Callee returns a locked route upon success */
2922 		rt_router = nd6_lookup(router_ll6, 0, ifp, 0);
2923 		if (!rt_router) {
2924 			goto nolladdropt;
2925 		}
2926 		RT_LOCK_ASSERT_HELD(rt_router);
2927 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2928 		len = (len + 7) & ~7;   /* round by 8 */
2929 		/* safety check */
2930 		if (len + (p - (u_char *)ip6) > maxlen) {
2931 			RT_REMREF_LOCKED(rt_router);
2932 			RT_UNLOCK(rt_router);
2933 			goto nolladdropt;
2934 		}
2935 
2936 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2937 		    (rt_router->rt_flags & RTF_LLINFO) &&
2938 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2939 		    (sdl = SDL(rt_router->rt_gateway)) && sdl->sdl_alen) {
2940 			nd_opt = (struct nd_opt_hdr *)p;
2941 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2942 			nd_opt->nd_opt_len = (uint8_t)(len >> 3);
2943 			lladdr = (char *)(nd_opt + 1);
2944 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2945 			p += len;
2946 		}
2947 		RT_REMREF_LOCKED(rt_router);
2948 		RT_UNLOCK(rt_router);
2949 	}
2950 
2951 nolladdropt:;
2952 
2953 	m->m_pkthdr.len = m->m_len = (int32_t)(p - (u_char *)ip6);
2954 
2955 	/* just to be safe */
2956 #ifdef M_DECRYPTED      /*not openbsd*/
2957 	if (m0->m_flags & M_DECRYPTED) {
2958 		goto noredhdropt;
2959 	}
2960 #endif
2961 	if (p - (u_char *)ip6 > maxlen) {
2962 		goto noredhdropt;
2963 	}
2964 
2965 	{
2966 		/* redirected header option */
2967 		int len;
2968 		struct nd_opt_rd_hdr *nd_opt_rh;
2969 
2970 		/*
2971 		 * compute the maximum size for icmp6 redirect header option.
2972 		 * XXX room for auth header?
2973 		 */
2974 		len = (int)(maxlen - (p - (u_char *)ip6));
2975 		len &= ~7;
2976 
2977 		/* This is just for simplicity. */
2978 		if (m0->m_pkthdr.len != m0->m_len) {
2979 			if (m0->m_next) {
2980 				m_freem(m0->m_next);
2981 				m0->m_next = NULL;
2982 			}
2983 			m0->m_pkthdr.len = m0->m_len;
2984 		}
2985 
2986 		/*
2987 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2988 		 * about padding/truncate rule for the original IP packet.
2989 		 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2990 		 * - "attach as much as possible" is the goal
2991 		 * - pad if not aligned (original size can be guessed by original
2992 		 *   ip6 header)
2993 		 * Following code adds the padding if it is simple enough,
2994 		 * and truncates if not.
2995 		 */
2996 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len) {
2997 			panic("assumption failed in %s:%d", __func__, __LINE__);
2998 		}
2999 
3000 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
3001 			/* not enough room, truncate */
3002 			m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
3003 		} else {
3004 			/* enough room, pad or truncate */
3005 			size_t extra;
3006 
3007 			extra = m0->m_pkthdr.len % 8;
3008 			if (extra) {
3009 				/* pad if easy enough, truncate if not */
3010 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
3011 					/* pad */
3012 					m0->m_len += (8 - extra);
3013 					m0->m_pkthdr.len += (8 - extra);
3014 				} else {
3015 					/* truncate */
3016 					m0->m_pkthdr.len -= extra;
3017 					m0->m_len -= extra;
3018 				}
3019 			}
3020 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
3021 			m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
3022 		}
3023 
3024 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
3025 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
3026 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
3027 		nd_opt_rh->nd_opt_rh_len = (uint8_t)(len >> 3);
3028 		p += sizeof(*nd_opt_rh);
3029 		m->m_pkthdr.len = m->m_len = (int32_t)(p - (u_char *)ip6);
3030 
3031 		/* connect m0 to m */
3032 		m->m_next = m0;
3033 		m->m_pkthdr.len = m->m_len + m0->m_len;
3034 	}
3035 noredhdropt:;
3036 
3037 	/* XXX: clear embedded link IDs in the inner header */
3038 	in6_clearscope(&sip6->ip6_src);
3039 	in6_clearscope(&sip6->ip6_dst);
3040 	in6_clearscope(&nd_rd->nd_rd_target);
3041 	in6_clearscope(&nd_rd->nd_rd_dst);
3042 
3043 	ip6->ip6_plen = htons((uint16_t)(m->m_pkthdr.len - sizeof(struct ip6_hdr)));
3044 
3045 	nd_rd->nd_rd_cksum = 0;
3046 	nd_rd->nd_rd_cksum
3047 	        = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
3048 
3049 	/* send the packet to outside... */
3050 	ip6oa.ip6oa_boundif = ifp->if_index;
3051 	ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
3052 	ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
3053 
3054 	ip6_output_setsrcifscope(m, ifp->if_index, NULL);
3055 	ip6_output_setdstifscope(m, ifp->if_index, NULL);
3056 
3057 	ip6_output(m, NULL, NULL, IPV6_OUTARGS, NULL, &outif, &ip6oa);
3058 	if (outif) {
3059 		icmp6_ifstat_inc(outif, ifs6_out_msg);
3060 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
3061 		ifnet_release(outif);
3062 	}
3063 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
3064 
3065 	return;
3066 
3067 fail:
3068 	if (rt != NULL) {
3069 		RT_UNLOCK(rt);
3070 	}
3071 	if (m) {
3072 		m_freem(m);
3073 	}
3074 	if (m0) {
3075 		m_freem(m0);
3076 	}
3077 }
3078 
3079 /*
3080  * ICMPv6 socket option processing.
3081  */
3082 int
icmp6_ctloutput(struct socket * so,struct sockopt * sopt)3083 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
3084 {
3085 	int error = 0;
3086 	size_t optlen;
3087 	struct inpcb *__single inp = sotoinpcb(so);
3088 	int level, op, optname;
3089 
3090 	if (sopt) {
3091 		level = sopt->sopt_level;
3092 		op = sopt->sopt_dir;
3093 		optname = sopt->sopt_name;
3094 		optlen = sopt->sopt_valsize;
3095 	} else {
3096 		level = op = optname = optlen = 0;
3097 	}
3098 
3099 	if (level != IPPROTO_ICMPV6) {
3100 		return EINVAL;
3101 	}
3102 
3103 	switch (op) {
3104 	case PRCO_SETOPT:
3105 		switch (optname) {
3106 		case ICMP6_FILTER:
3107 		{
3108 			struct icmp6_filter *__single p;
3109 
3110 			if (optlen != 0 && optlen != sizeof(*p)) {
3111 				error = EMSGSIZE;
3112 				break;
3113 			}
3114 			if (inp->in6p_icmp6filt == NULL) {
3115 				error = EINVAL;
3116 				break;
3117 			}
3118 
3119 			if (optlen == 0) {
3120 				/* According to RFC 3542, an installed filter can be
3121 				 * cleared by issuing a setsockopt for ICMP6_FILTER
3122 				 * with a zero length.
3123 				 */
3124 				ICMP6_FILTER_SETPASSALL(inp->in6p_icmp6filt);
3125 			} else {
3126 				error = sooptcopyin(sopt, __unsafe_forge_bidi_indexable(void*, inp->in6p_icmp6filt, optlen), optlen,
3127 				    optlen);
3128 			}
3129 			break;
3130 		}
3131 
3132 		default:
3133 			error = ENOPROTOOPT;
3134 			break;
3135 		}
3136 		break;
3137 
3138 	case PRCO_GETOPT:
3139 		switch (optname) {
3140 		case ICMP6_FILTER:
3141 		{
3142 			if (inp->in6p_icmp6filt == NULL) {
3143 				error = EINVAL;
3144 				break;
3145 			}
3146 			size_t copylen = MIN(sizeof(struct icmp6_filter), optlen);
3147 			error = sooptcopyout(sopt, __unsafe_forge_bidi_indexable(void*, inp->in6p_icmp6filt, optlen), copylen);
3148 			break;
3149 		}
3150 
3151 		default:
3152 			error = ENOPROTOOPT;
3153 			break;
3154 		}
3155 		break;
3156 	}
3157 
3158 	return error;
3159 }
3160 
3161 /*
3162  * ICMPv6 socket datagram option processing.
3163  */
3164 int
icmp6_dgram_ctloutput(struct socket * so,struct sockopt * sopt)3165 icmp6_dgram_ctloutput(struct socket *so, struct sockopt *sopt)
3166 {
3167 	/*
3168 	 * For { SOCK_RAW, IPPROTO_ICMPV6 } the pr_ctloutput is
3169 	 * rip6_ctloutput() and not icmp6_ctloutput()
3170 	 */
3171 	if (kauth_cred_issuser(so->so_cred)) {
3172 		return rip6_ctloutput(so, sopt);
3173 	}
3174 
3175 	/* Allow <SOL_SOCKET,SO_BINDTODEVICE> at this level */
3176 	if (sopt->sopt_level == SOL_SOCKET) {
3177 		if (sopt->sopt_name == SO_BINDTODEVICE) {
3178 			return ip6_ctloutput(so, sopt);
3179 		}
3180 		return EINVAL;
3181 	} else if (sopt->sopt_level == IPPROTO_ICMPV6) {
3182 		switch (sopt->sopt_name) {
3183 		case ICMP6_FILTER:
3184 			return icmp6_ctloutput(so, sopt);
3185 		default:
3186 			return EPERM;
3187 		}
3188 	} else if (sopt->sopt_level != IPPROTO_IPV6) {
3189 		return EINVAL;
3190 	}
3191 
3192 	switch (sopt->sopt_name) {
3193 	case IPV6_UNICAST_HOPS:
3194 	case IPV6_CHECKSUM:
3195 	case IPV6_V6ONLY:
3196 	case IPV6_USE_MIN_MTU:
3197 	case IPV6_RECVRTHDR:
3198 	case IPV6_RECVPKTINFO:
3199 	case IPV6_RECVHOPLIMIT:
3200 	case IPV6_PATHMTU:
3201 	case IPV6_PKTINFO:
3202 	case IPV6_HOPLIMIT:
3203 	case IPV6_HOPOPTS:
3204 	case IPV6_DSTOPTS:
3205 	case IPV6_MULTICAST_IF:
3206 	case IPV6_MULTICAST_HOPS:
3207 	case IPV6_MULTICAST_LOOP:
3208 	case IPV6_JOIN_GROUP:
3209 	case IPV6_LEAVE_GROUP:
3210 	case IPV6_PORTRANGE:
3211 	case IPV6_IPSEC_POLICY:
3212 	case IPV6_RECVTCLASS:
3213 	case IPV6_TCLASS:
3214 	case IPV6_2292PKTOPTIONS:
3215 	case IPV6_2292PKTINFO:
3216 	case IPV6_2292HOPLIMIT:
3217 	case IPV6_2292HOPOPTS:
3218 	case IPV6_2292DSTOPTS:
3219 	case IPV6_2292RTHDR:
3220 	case IPV6_BOUND_IF:
3221 	case IPV6_NO_IFT_CELLULAR:
3222 		return ip6_ctloutput(so, sopt);
3223 
3224 	default:
3225 		return EPERM;
3226 	}
3227 }
3228 
3229 __private_extern__ int
icmp6_dgram_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)3230 icmp6_dgram_send(struct socket *so, int flags, struct mbuf *m,
3231     struct sockaddr *nam, struct mbuf *control, struct proc *p)
3232 {
3233 #pragma unused(flags, p)
3234 	int error = 0;
3235 	struct inpcb *__single inp = sotoinpcb(so);
3236 	struct icmp6_hdr *__single icmp6;
3237 
3238 	if (inp == NULL
3239 #if NECP
3240 	    || (necp_socket_should_use_flow_divert(inp))
3241 #endif /* NECP */
3242 	    ) {
3243 		error = (inp == NULL ? EINVAL : EPROTOTYPE);
3244 		goto bad;
3245 	}
3246 
3247 	if (kauth_cred_issuser(so->so_cred)) {
3248 		return rip6_output(m, so, SIN6(nam), control, 0);
3249 	}
3250 
3251 	/*
3252 	 * For an ICMPv6 packet, we should know its type and code
3253 	 */
3254 	if (SOCK_PROTO(so) == IPPROTO_ICMPV6) {
3255 		if (m->m_len < sizeof(struct icmp6_hdr) &&
3256 		    (m = m_pullup(m, sizeof(struct icmp6_hdr))) == NULL) {
3257 			error = ENOBUFS;
3258 			goto bad;
3259 		}
3260 		icmp6 = mtod(m, struct icmp6_hdr *);
3261 
3262 		/*
3263 		 * Allow only to send echo request and node information request
3264 		 * See RFC 2463 for Echo Request Message format
3265 		 */
3266 		if ((icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
3267 		    icmp6->icmp6_code == 0) ||
3268 		    (icmp6->icmp6_type == ICMP6_NI_QUERY &&
3269 		    (icmp6->icmp6_code == ICMP6_NI_SUBJ_IPV6 ||
3270 		    icmp6->icmp6_code == ICMP6_NI_SUBJ_FQDN))) {
3271 			/* Good */
3272 			;
3273 		} else {
3274 			error = EPERM;
3275 			goto bad;
3276 		}
3277 	}
3278 
3279 	so_update_tx_data_stats(so, 1, m->m_pkthdr.len);
3280 
3281 	return rip6_output(m, so, SIN6(nam), control, 0);
3282 bad:
3283 	VERIFY(error != 0);
3284 
3285 	if (m != NULL) {
3286 		m_freem(m);
3287 	}
3288 	if (control != NULL) {
3289 		m_freem(control);
3290 	}
3291 
3292 	return error;
3293 }
3294 
3295 /* Like rip6_attach but without root privilege enforcement */
3296 __private_extern__ int
icmp6_dgram_attach(struct socket * so,int proto,struct proc * p)3297 icmp6_dgram_attach(struct socket *so, int proto, struct proc *p)
3298 {
3299 	struct inpcb *__single inp;
3300 	int error;
3301 
3302 	inp = sotoinpcb(so);
3303 	if (inp) {
3304 		panic("icmp6_dgram_attach");
3305 	}
3306 
3307 	if (proto != IPPROTO_ICMPV6) {
3308 		return EINVAL;
3309 	}
3310 
3311 	error = soreserve(so, rip_sendspace, rip_recvspace);
3312 	if (error) {
3313 		return error;
3314 	}
3315 	error = in_pcballoc(so, &ripcbinfo, p);
3316 	if (error) {
3317 		return error;
3318 	}
3319 	inp = (struct inpcb *)so->so_pcb;
3320 	inp->inp_vflag |= INP_IPV6;
3321 	inp->in6p_ip6_nxt = IPPROTO_ICMPV6;
3322 	inp->in6p_hops = -1;    /* use kernel default */
3323 	inp->in6p_cksum = -1;
3324 	inp->in6p_icmp6filt = kalloc_type(struct icmp6_filter,
3325 	    Z_WAITOK | Z_NOFAIL);
3326 	ICMP6_FILTER_SETPASSALL(inp->in6p_icmp6filt);
3327 	return 0;
3328 }
3329 
3330 
3331 /*
3332  * Perform rate limit check.
3333  * Returns 0 if it is okay to send the icmp6 packet.
3334  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
3335  * limitation.
3336  *
3337  * XXX per-destination check necessary?
3338  */
3339 static int
icmp6_ratelimit(__unused const struct in6_addr * dst,const int type,__unused const int code)3340 icmp6_ratelimit(
3341 	__unused const struct in6_addr *dst,    /* not used at this moment */
3342 	const int type,
3343 	__unused const int code)
3344 {
3345 	int ret;
3346 
3347 	ret = 0;        /* okay to send */
3348 
3349 	/* PPS limit */
3350 	if (type == ND_ROUTER_ADVERT) {
3351 		if (!ppsratecheck(&icmp6rappslim_last, &icmp6rapps_count,
3352 		    icmp6rappslim)) {
3353 			ret++;
3354 		}
3355 	} else if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
3356 	    icmp6errppslim)) {
3357 		/*
3358 		 * We add some randomness here to still generate ICMPv6 error
3359 		 * post icmp6errppslim limit with a probability that goes down
3360 		 * with increased value of icmp6errpps_count.
3361 		 */
3362 		if (icmp6errpps_count > 0 && icmp6errppslim > 0 &&
3363 		    icmp6errpps_count > icmp6errppslim &&
3364 		    (random() % (icmp6errpps_count - icmp6errppslim)) != 0) {
3365 			/* The packet is subject to rate limit */
3366 			ret++;
3367 		}
3368 	}
3369 
3370 	return ret;
3371 }
3372