xref: /freebsd-14.2/sys/netinet6/icmp6.c (revision 2472e352)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 #define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/domain.h>
73 #include <sys/jail.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/proc.h>
79 #include <sys/protosw.h>
80 #include <sys/signalvar.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/sx.h>
84 #include <sys/syslog.h>
85 #include <sys/systm.h>
86 #include <sys/time.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_dl.h>
91 #include <net/if_llatbl.h>
92 #include <net/if_private.h>
93 #include <net/if_types.h>
94 #include <net/route.h>
95 #include <net/route/route_ctl.h>
96 #include <net/route/nhop.h>
97 #include <net/vnet.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_pcb.h>
101 #include <netinet/in_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet/tcp_var.h>
105 
106 #include <netinet6/in6_fib.h>
107 #include <netinet6/in6_ifattach.h>
108 #include <netinet6/in6_pcb.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/nd6.h>
113 #include <netinet6/send.h>
114 
115 extern ip6proto_ctlinput_t	*ip6_ctlprotox[];
116 
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119 SYSCTL_VNET_PCPUSTAT(_net_inet6_icmp6, ICMPV6CTL_STATS, stats,
120     struct icmp6stat, icmp6stat,
121     "ICMPv6 statistics (struct icmp6stat, netinet/icmp6.h)");
122 
123 #ifdef VIMAGE
124 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
125 #endif /* VIMAGE */
126 
127 VNET_DEFINE_STATIC(int, icmp6_rediraccept) = 1;
128 #define	V_icmp6_rediraccept	VNET(icmp6_rediraccept)
129 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRACCEPT, rediraccept,
130     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_rediraccept), 0,
131     "Accept ICMPv6 redirect messages");
132 
133 VNET_DEFINE_STATIC(int, icmp6_redirtimeout) = 10 * 60;	/* 10 minutes */
134 #define	V_icmp6_redirtimeout	VNET(icmp6_redirtimeout)
135 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRTIMEOUT, redirtimeout,
136     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_redirtimeout), 0,
137     "Delay in seconds before expiring redirect route");
138 
139 VNET_DEFINE_STATIC(int, icmp6_nodeinfo) = 0;
140 #define	V_icmp6_nodeinfo	VNET(icmp6_nodeinfo)
141 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO, nodeinfo,
142     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_nodeinfo), 0,
143     "Mask of enabled RFC4620 node information query types");
144 
145 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
146 #define	V_ripcbinfo		VNET(ripcbinfo)
147 
148 static void icmp6_errcount(int, int);
149 static int icmp6_rip6_input(struct mbuf **, int);
150 static void icmp6_reflect(struct mbuf *, size_t);
151 static const char *icmp6_redirect_diag(struct in6_addr *,
152 	struct in6_addr *, struct in6_addr *);
153 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
154 static struct mbuf *ni6_nametodns(const char *, int, int);
155 static int ni6_dnsmatch(const char *, int, const char *, int);
156 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
157 			  struct ifnet **, struct in6_addr *);
158 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
159 				struct ifnet *, int);
160 static int icmp6_notify_error(struct mbuf **, int, int);
161 
162 /*
163  * Kernel module interface for updating icmp6stat.  The argument is an index
164  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
165  * general layout of icmp6stat into the caller, it doesn't encode its
166  * location, so that future changes to add, for example, per-CPU stats
167  * support won't cause binary compatibility problems for kernel modules.
168  */
169 void
kmod_icmp6stat_inc(int statnum)170 kmod_icmp6stat_inc(int statnum)
171 {
172 
173 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
174 }
175 
176 static void
icmp6_errcount(int type,int code)177 icmp6_errcount(int type, int code)
178 {
179 	switch (type) {
180 	case ICMP6_DST_UNREACH:
181 		switch (code) {
182 		case ICMP6_DST_UNREACH_NOROUTE:
183 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
184 			return;
185 		case ICMP6_DST_UNREACH_ADMIN:
186 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
187 			return;
188 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
189 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
190 			return;
191 		case ICMP6_DST_UNREACH_ADDR:
192 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
193 			return;
194 		case ICMP6_DST_UNREACH_NOPORT:
195 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
196 			return;
197 		}
198 		break;
199 	case ICMP6_PACKET_TOO_BIG:
200 		ICMP6STAT_INC(icp6s_opacket_too_big);
201 		return;
202 	case ICMP6_TIME_EXCEEDED:
203 		switch (code) {
204 		case ICMP6_TIME_EXCEED_TRANSIT:
205 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
206 			return;
207 		case ICMP6_TIME_EXCEED_REASSEMBLY:
208 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
209 			return;
210 		}
211 		break;
212 	case ICMP6_PARAM_PROB:
213 		switch (code) {
214 		case ICMP6_PARAMPROB_HEADER:
215 			ICMP6STAT_INC(icp6s_oparamprob_header);
216 			return;
217 		case ICMP6_PARAMPROB_NEXTHEADER:
218 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
219 			return;
220 		case ICMP6_PARAMPROB_OPTION:
221 			ICMP6STAT_INC(icp6s_oparamprob_option);
222 			return;
223 		}
224 		break;
225 	case ND_REDIRECT:
226 		ICMP6STAT_INC(icp6s_oredirect);
227 		return;
228 	}
229 	ICMP6STAT_INC(icp6s_ounknown);
230 }
231 
232 /*
233  * A wrapper function for icmp6_error() necessary when the erroneous packet
234  * may not contain enough scope zone information.
235  */
236 void
icmp6_error2(struct mbuf * m,int type,int code,int param,struct ifnet * ifp)237 icmp6_error2(struct mbuf *m, int type, int code, int param,
238     struct ifnet *ifp)
239 {
240 	struct ip6_hdr *ip6;
241 
242 	if (ifp == NULL)
243 		return;
244 
245 	if (m->m_len < sizeof(struct ip6_hdr)) {
246 		m = m_pullup(m, sizeof(struct ip6_hdr));
247 		if (m == NULL) {
248 			IP6STAT_INC(ip6s_exthdrtoolong);
249 			return;
250 		}
251 	}
252 	ip6 = mtod(m, struct ip6_hdr *);
253 
254 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
255 		return;
256 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
257 		return;
258 
259 	icmp6_error(m, type, code, param);
260 }
261 
262 /*
263  * Generate an error packet of type error in response to bad IP6 packet.
264  */
265 void
icmp6_error(struct mbuf * m,int type,int code,int param)266 icmp6_error(struct mbuf *m, int type, int code, int param)
267 {
268 	struct ip6_hdr *oip6, *nip6;
269 	struct icmp6_hdr *icmp6;
270 	struct epoch_tracker et;
271 	u_int preplen;
272 	int off;
273 	int nxt;
274 
275 	ICMP6STAT_INC(icp6s_error);
276 
277 	/* count per-type-code statistics */
278 	icmp6_errcount(type, code);
279 
280 #ifdef M_DECRYPTED	/*not openbsd*/
281 	if (m->m_flags & M_DECRYPTED) {
282 		ICMP6STAT_INC(icp6s_canterror);
283 		goto freeit;
284 	}
285 #endif
286 
287 	if (m->m_len < sizeof(struct ip6_hdr)) {
288 		m = m_pullup(m, sizeof(struct ip6_hdr));
289 		if (m == NULL) {
290 			IP6STAT_INC(ip6s_exthdrtoolong);
291 			return;
292 		}
293 	}
294 	oip6 = mtod(m, struct ip6_hdr *);
295 
296 	/*
297 	 * If the destination address of the erroneous packet is a multicast
298 	 * address, or the packet was sent using link-layer multicast,
299 	 * we should basically suppress sending an error (RFC 2463, Section
300 	 * 2.4).
301 	 * We have two exceptions (the item e.2 in that section):
302 	 * - the Packet Too Big message can be sent for path MTU discovery.
303 	 * - the Parameter Problem Message that can be allowed an icmp6 error
304 	 *   in the option type field.  This check has been done in
305 	 *   ip6_unknown_opt(), so we can just check the type and code.
306 	 */
307 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
308 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
309 	    (type != ICMP6_PACKET_TOO_BIG &&
310 	     (type != ICMP6_PARAM_PROB ||
311 	      code != ICMP6_PARAMPROB_OPTION)))
312 		goto freeit;
313 
314 	/*
315 	 * RFC 2463, 2.4 (e.5): source address check.
316 	 * XXX: the case of anycast source?
317 	 */
318 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
319 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
320 		goto freeit;
321 
322 	/*
323 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
324 	 * don't do it.
325 	 */
326 	nxt = -1;
327 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
328 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
329 		struct icmp6_hdr *icp;
330 
331 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
332 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
333 			if (m == NULL) {
334 				IP6STAT_INC(ip6s_exthdrtoolong);
335 				return;
336 			}
337 		}
338 		oip6 = mtod(m, struct ip6_hdr *);
339 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
340 
341 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
342 		    icp->icmp6_type == ND_REDIRECT) {
343 			/*
344 			 * ICMPv6 error
345 			 * Special case: for redirect (which is
346 			 * informational) we must not send icmp6 error.
347 			 */
348 			ICMP6STAT_INC(icp6s_canterror);
349 			goto freeit;
350 		} else {
351 			/* ICMPv6 informational - send the error */
352 		}
353 	} else {
354 		/* non-ICMPv6 - send the error */
355 	}
356 
357 	/* Finally, do rate limitation check. */
358 	if (icmp6_ratelimit(&oip6->ip6_src, type, code))
359 		goto freeit;
360 
361 	/*
362 	 * OK, ICMP6 can be generated.
363 	 */
364 
365 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
366 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
367 
368 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
369 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
370 	if (m == NULL) {
371 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
372 		return;
373 	}
374 
375 	nip6 = mtod(m, struct ip6_hdr *);
376 	nip6->ip6_src  = oip6->ip6_src;
377 	nip6->ip6_dst  = oip6->ip6_dst;
378 
379 	in6_clearscope(&oip6->ip6_src);
380 	in6_clearscope(&oip6->ip6_dst);
381 
382 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
383 	icmp6->icmp6_type = type;
384 	icmp6->icmp6_code = code;
385 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
386 
387 	ICMP6STAT_INC(icp6s_outhist[type]);
388 	NET_EPOCH_ENTER(et);
389 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
390 	NET_EPOCH_EXIT(et);
391 
392 	return;
393 
394   freeit:
395 	/*
396 	 * If we can't tell whether or not we can generate ICMP6, free it.
397 	 */
398 	m_freem(m);
399 }
400 
401 int
icmp6_errmap(const struct icmp6_hdr * icmp6)402 icmp6_errmap(const struct icmp6_hdr *icmp6)
403 {
404 
405 	switch (icmp6->icmp6_type) {
406 	case ICMP6_DST_UNREACH:
407 		switch (icmp6->icmp6_code) {
408 		case ICMP6_DST_UNREACH_NOROUTE:
409 		case ICMP6_DST_UNREACH_ADDR:
410 			return (EHOSTUNREACH);
411 		case ICMP6_DST_UNREACH_NOPORT:
412 		case ICMP6_DST_UNREACH_ADMIN:
413 			return (ECONNREFUSED);
414 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
415 			return (ENOPROTOOPT);
416 		default:
417 			return (0);	/* Shouldn't happen. */
418 		}
419 	case ICMP6_PACKET_TOO_BIG:
420 		return (EMSGSIZE);
421 	case ICMP6_TIME_EXCEEDED:
422 		switch (icmp6->icmp6_code) {
423 		case ICMP6_TIME_EXCEED_TRANSIT:
424 			return (EHOSTUNREACH);
425 		case ICMP6_TIME_EXCEED_REASSEMBLY:
426 			return (0);
427 		default:
428 			return (0);	/* Shouldn't happen. */
429 		}
430 	case ICMP6_PARAM_PROB:
431 		switch (icmp6->icmp6_code) {
432 		case ICMP6_PARAMPROB_NEXTHEADER:
433 			return (ECONNREFUSED);
434 		case ICMP6_PARAMPROB_HEADER:
435 		case ICMP6_PARAMPROB_OPTION:
436 			return (ENOPROTOOPT);
437 		default:
438 			return (0);	/* Shouldn't happen. */
439 		}
440 	default:
441 		return (0);
442 	}
443 }
444 
445 /*
446  * Process a received ICMP6 message.
447  */
448 int
icmp6_input(struct mbuf ** mp,int * offp,int proto)449 icmp6_input(struct mbuf **mp, int *offp, int proto)
450 {
451 	struct mbuf *m, *n;
452 	struct ifnet *ifp;
453 	struct ip6_hdr *ip6, *nip6;
454 	struct icmp6_hdr *icmp6, *nicmp6;
455 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
456 	int code, error, icmp6len, ip6len, noff, off, sum;
457 
458 	NET_EPOCH_ASSERT();
459 
460 	m = *mp;
461 	off = *offp;
462 
463 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
464 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
465 		if (m == NULL) {
466 			IP6STAT_INC(ip6s_exthdrtoolong);
467 			*mp = m;
468 			return (IPPROTO_DONE);
469 		}
470 	}
471 
472 	/*
473 	 * Locate icmp6 structure in mbuf, and check
474 	 * that not corrupted and of at least minimum length
475 	 */
476 
477 	icmp6len = m->m_pkthdr.len - off;
478 	if (icmp6len < sizeof(struct icmp6_hdr)) {
479 		ICMP6STAT_INC(icp6s_tooshort);
480 		goto freeit;
481 	}
482 
483 	ip6 = mtod(m, struct ip6_hdr *);
484 	ifp = m->m_pkthdr.rcvif;
485 	/*
486 	 * Check multicast group membership.
487 	 * Note: SSM filters are not applied for ICMPv6 traffic.
488 	 */
489 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
490 		struct in6_multi	*inm;
491 
492 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
493 		if (inm == NULL) {
494 			IP6STAT_INC(ip6s_notmember);
495 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
496 			goto freeit;
497 		}
498 	}
499 
500 	/* Calculate the checksum. */
501 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
502 	code = icmp6->icmp6_code;
503 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
504 		nd6log((LOG_ERR,
505 		    "ICMP6 checksum error(%d|%x) %s\n",
506 		    icmp6->icmp6_type, sum,
507 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
508 		ICMP6STAT_INC(icp6s_checksum);
509 		goto freeit;
510 	}
511 
512 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
513 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
514 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
515 		icmp6_ifstat_inc(ifp, ifs6_in_error);
516 
517 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
518 	switch (icmp6->icmp6_type) {
519 	case ICMP6_DST_UNREACH:
520 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
521 		switch (code) {
522 		case ICMP6_DST_UNREACH_ADMIN:
523 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
524 		case ICMP6_DST_UNREACH_NOROUTE:
525 		case ICMP6_DST_UNREACH_ADDR:
526 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
527 		case ICMP6_DST_UNREACH_NOPORT:
528 			goto deliver;
529 		default:
530 			goto badcode;
531 		}
532 	case ICMP6_PACKET_TOO_BIG:
533 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
534 		/*
535 		 * Validation is made in icmp6_mtudisc_update.
536 		 * Updating the path MTU will be done after examining
537 		 * intermediate extension headers.
538 		 */
539 		goto deliver;
540 	case ICMP6_TIME_EXCEEDED:
541 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
542 		switch (code) {
543 		case ICMP6_TIME_EXCEED_TRANSIT:
544 		case ICMP6_TIME_EXCEED_REASSEMBLY:
545 			goto deliver;
546 		default:
547 			goto badcode;
548 		}
549 	case ICMP6_PARAM_PROB:
550 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
551 		switch (code) {
552 		case ICMP6_PARAMPROB_NEXTHEADER:
553 		case ICMP6_PARAMPROB_HEADER:
554 		case ICMP6_PARAMPROB_OPTION:
555 			goto deliver;
556 		default:
557 			goto badcode;
558 		}
559 	case ICMP6_ECHO_REQUEST:
560 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
561 		if (code != 0)
562 			goto badcode;
563 		if (icmp6_ratelimit(&ip6->ip6_src, ICMP6_ECHO_REPLY, 0))
564 			break;
565 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
566 			/* Give up remote */
567 			break;
568 		}
569 		if (!M_WRITABLE(n)
570 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
571 			struct mbuf *n0 = n;
572 			int n0len;
573 
574 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
575 			n = m_gethdr(M_NOWAIT, n0->m_type);
576 			if (n == NULL) {
577 				/* Give up remote */
578 				m_freem(n0);
579 				break;
580 			}
581 
582 			m_move_pkthdr(n, n0);	/* FIB copied. */
583 			n0len = n0->m_pkthdr.len;	/* save for use below */
584 			/*
585 			 * Copy IPv6 and ICMPv6 only.
586 			 */
587 			nip6 = mtod(n, struct ip6_hdr *);
588 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
589 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
590 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
591 			noff = sizeof(struct ip6_hdr);
592 			/* new mbuf contains only ipv6+icmpv6 headers */
593 			n->m_len = noff + sizeof(struct icmp6_hdr);
594 			/*
595 			 * Adjust mbuf.  ip6_plen will be adjusted in
596 			 * ip6_output().
597 			 */
598 			m_adj(n0, off + sizeof(struct icmp6_hdr));
599 			/* recalculate complete packet size */
600 			n->m_pkthdr.len = n0len + (noff - off);
601 			n->m_next = n0;
602 		} else {
603 			if (n->m_len < off + sizeof(*nicmp6)) {
604 				n = m_pullup(n, off + sizeof(*nicmp6));
605 				if (n == NULL) {
606 					IP6STAT_INC(ip6s_exthdrtoolong);
607 					break;
608 				}
609 			}
610 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
611 			noff = off;
612 		}
613 		if (n) {
614 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
615 			nicmp6->icmp6_code = 0;
616 			ICMP6STAT_INC(icp6s_reflect);
617 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
618 			icmp6_reflect(n, noff);
619 		}
620 		break;
621 
622 	case ICMP6_ECHO_REPLY:
623 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
624 		if (code != 0)
625 			goto badcode;
626 		break;
627 
628 	case MLD_LISTENER_QUERY:
629 	case MLD_LISTENER_REPORT:
630 	case MLD_LISTENER_DONE:
631 	case MLDV2_LISTENER_REPORT:
632 		/*
633 		 * Drop MLD traffic which is not link-local, has a hop limit
634 		 * of greater than 1 hop, or which does not have the
635 		 * IPv6 HBH Router Alert option.
636 		 * As IPv6 HBH options are stripped in ip6_input() we must
637 		 * check an mbuf header flag.
638 		 * XXX Should we also sanity check that these messages
639 		 * were directed to a link-local multicast prefix?
640 		 */
641 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
642 			goto freeit;
643 		if (mld_input(&m, off, icmp6len) != 0) {
644 			*mp = NULL;
645 			return (IPPROTO_DONE);
646 		}
647 		/* m stays. */
648 		break;
649 
650 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
651 	    {
652 		enum { WRU, FQDN } mode;
653 		struct prison *pr;
654 
655 		if (!V_icmp6_nodeinfo)
656 			break;
657 
658 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
659 			mode = WRU;
660 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
661 			mode = FQDN;
662 		else
663 			goto badlen;
664 
665 		pr = NULL;
666 		sx_slock(&allprison_lock);
667 		TAILQ_FOREACH(pr, &allprison, pr_list)
668 			if (pr->pr_vnet == ifp->if_vnet)
669 				break;
670 		sx_sunlock(&allprison_lock);
671 		if (pr == NULL)
672 			pr = curthread->td_ucred->cr_prison;
673 		if (mode == FQDN) {
674 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
675 				m = m_pullup(m, off +
676 				    sizeof(struct icmp6_nodeinfo));
677 				if (m == NULL) {
678 					IP6STAT_INC(ip6s_exthdrtoolong);
679 					*mp = m;
680 					return (IPPROTO_DONE);
681 				}
682 			}
683 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
684 			if (n)
685 				n = ni6_input(n, off, pr);
686 			/* XXX meaningless if n == NULL */
687 			noff = sizeof(struct ip6_hdr);
688 		} else {
689 			u_char *p;
690 			int maxhlen, hlen;
691 
692 			/*
693 			 * XXX: this combination of flags is pointless,
694 			 * but should we keep this for compatibility?
695 			 */
696 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
697 			    ICMP6_NODEINFO_TMPADDROK)) !=
698 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
699 				break;
700 
701 			if (code != 0)
702 				goto badcode;
703 
704 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
705 			n = m_gethdr(M_NOWAIT, m->m_type);
706 			if (n == NULL) {
707 				/* Give up remote */
708 				break;
709 			}
710 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
711 				/*
712 				 * Previous code did a blind M_COPY_PKTHDR
713 				 * and said "just for rcvif".  If true, then
714 				 * we could tolerate the dup failing (due to
715 				 * the deep copy of the tag chain).  For now
716 				 * be conservative and just fail.
717 				 */
718 				m_free(n);
719 				n = NULL;
720 				break;
721 			}
722 			/*
723 			 * Copy IPv6 and ICMPv6 only.
724 			 */
725 			nip6 = mtod(n, struct ip6_hdr *);
726 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
727 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
728 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
729 			p = (u_char *)(nicmp6 + 1);
730 			bzero(p, 4);
731 
732 			maxhlen = M_TRAILINGSPACE(n) -
733 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
734 			mtx_lock(&pr->pr_mtx);
735 			hlen = strlen(pr->pr_hostname);
736 			if (maxhlen > hlen)
737 				maxhlen = hlen;
738 			/* meaningless TTL */
739 			bcopy(pr->pr_hostname, p + 4, maxhlen);
740 			mtx_unlock(&pr->pr_mtx);
741 			noff = sizeof(struct ip6_hdr);
742 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
743 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
744 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
745 			nicmp6->icmp6_code = 0;
746 		}
747 		if (n) {
748 			ICMP6STAT_INC(icp6s_reflect);
749 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
750 			icmp6_reflect(n, noff);
751 		}
752 		break;
753 	    }
754 
755 	case ICMP6_WRUREPLY:
756 		if (code != 0)
757 			goto badcode;
758 		break;
759 
760 	case ND_ROUTER_SOLICIT:
761 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
762 		if (code != 0)
763 			goto badcode;
764 		if (icmp6len < sizeof(struct nd_router_solicit))
765 			goto badlen;
766 		if (send_sendso_input_hook != NULL) {
767 			if (m->m_len < off + icmp6len) {
768 				m = m_pullup(m, off + icmp6len);
769 				if (m == NULL) {
770 					IP6STAT_INC(ip6s_exthdrtoolong);
771 					*mp = NULL;
772 					return (IPPROTO_DONE);
773 				}
774 			}
775 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
776 			if (error == 0) {
777 				m = NULL;
778 				goto freeit;
779 			}
780 		}
781 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
782 		nd6_rs_input(m, off, icmp6len);
783 		m = n;
784 		if (m == NULL)
785 			goto freeit;
786 		break;
787 
788 	case ND_ROUTER_ADVERT:
789 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
790 		if (code != 0)
791 			goto badcode;
792 		if (icmp6len < sizeof(struct nd_router_advert))
793 			goto badlen;
794 		if (send_sendso_input_hook != NULL) {
795 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
796 			if (error == 0) {
797 				m = NULL;
798 				goto freeit;
799 			}
800 		}
801 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
802 		nd6_ra_input(m, off, icmp6len);
803 		m = n;
804 		if (m == NULL)
805 			goto freeit;
806 		break;
807 
808 	case ND_NEIGHBOR_SOLICIT:
809 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
810 		if (code != 0)
811 			goto badcode;
812 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
813 			goto badlen;
814 		if (send_sendso_input_hook != NULL) {
815 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
816 			if (error == 0) {
817 				m = NULL;
818 				goto freeit;
819 			}
820 		}
821 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
822 		nd6_ns_input(m, off, icmp6len);
823 		m = n;
824 		if (m == NULL)
825 			goto freeit;
826 		break;
827 
828 	case ND_NEIGHBOR_ADVERT:
829 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
830 		if (code != 0)
831 			goto badcode;
832 		if (icmp6len < sizeof(struct nd_neighbor_advert))
833 			goto badlen;
834 		if (send_sendso_input_hook != NULL) {
835 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
836 			if (error == 0) {
837 				m = NULL;
838 				goto freeit;
839 			}
840 		}
841 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
842 		nd6_na_input(m, off, icmp6len);
843 		m = n;
844 		if (m == NULL)
845 			goto freeit;
846 		break;
847 
848 	case ND_REDIRECT:
849 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
850 		if (code != 0)
851 			goto badcode;
852 		if (icmp6len < sizeof(struct nd_redirect))
853 			goto badlen;
854 		if (send_sendso_input_hook != NULL) {
855 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
856 			if (error == 0) {
857 				m = NULL;
858 				goto freeit;
859 			}
860 		}
861 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
862 		icmp6_redirect_input(m, off);
863 		m = n;
864 		if (m == NULL)
865 			goto freeit;
866 		break;
867 
868 	case ICMP6_ROUTER_RENUMBERING:
869 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
870 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
871 			goto badcode;
872 		if (icmp6len < sizeof(struct icmp6_router_renum))
873 			goto badlen;
874 		break;
875 
876 	default:
877 		nd6log((LOG_DEBUG,
878 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
879 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
880 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
881 		    ifp ? ifp->if_index : 0));
882 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
883 			/* ICMPv6 error: MUST deliver it by spec... */
884 			goto deliver;
885 		} else {
886 			/* ICMPv6 informational: MUST not deliver */
887 			break;
888 		}
889 	deliver:
890 		if (icmp6_notify_error(&m, off, icmp6len) != 0) {
891 			/* In this case, m should've been freed. */
892 			*mp = NULL;
893 			return (IPPROTO_DONE);
894 		}
895 		break;
896 
897 	badcode:
898 		ICMP6STAT_INC(icp6s_badcode);
899 		break;
900 
901 	badlen:
902 		ICMP6STAT_INC(icp6s_badlen);
903 		break;
904 	}
905 
906 	/* deliver the packet to appropriate sockets */
907 	icmp6_rip6_input(&m, *offp);
908 
909 	*mp = m;
910 	return (IPPROTO_DONE);
911 
912  freeit:
913 	m_freem(m);
914 	*mp = NULL;
915 	return (IPPROTO_DONE);
916 }
917 
918 static int
icmp6_notify_error(struct mbuf ** mp,int off,int icmp6len)919 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len)
920 {
921 	struct mbuf *m;
922 	struct icmp6_hdr *icmp6;
923 	struct ip6_hdr *eip6;
924 	u_int32_t notifymtu;
925 	struct sockaddr_in6 icmp6src, icmp6dst;
926 
927 	m = *mp;
928 
929 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
930 		ICMP6STAT_INC(icp6s_tooshort);
931 		goto freeit;
932 	}
933 
934 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
935 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
936 		if (m == NULL) {
937 			IP6STAT_INC(ip6s_exthdrtoolong);
938 			*mp = m;
939 			return (-1);
940 		}
941 	}
942 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
943 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
944 	bzero(&icmp6dst, sizeof(icmp6dst));
945 
946 	/* Detect the upper level protocol */
947 	{
948 		u_int8_t nxt = eip6->ip6_nxt;
949 		int eoff = off + sizeof(struct icmp6_hdr) +
950 		    sizeof(struct ip6_hdr);
951 		struct ip6ctlparam ip6cp;
952 		int icmp6type = icmp6->icmp6_type;
953 		struct ip6_frag *fh;
954 		struct ip6_rthdr *rth;
955 		struct ip6_rthdr0 *rth0;
956 		int rthlen;
957 
958 		while (1) { /* XXX: should avoid infinite loop explicitly? */
959 			struct ip6_ext *eh;
960 
961 			switch (nxt) {
962 			case IPPROTO_HOPOPTS:
963 			case IPPROTO_DSTOPTS:
964 			case IPPROTO_AH:
965 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
966 					m = m_pullup(m, eoff +
967 					    sizeof(struct ip6_ext));
968 					if (m == NULL) {
969 						IP6STAT_INC(ip6s_exthdrtoolong);
970 						*mp = m;
971 						return (-1);
972 					}
973 				}
974 				eh = (struct ip6_ext *)
975 				    (mtod(m, caddr_t) + eoff);
976 				if (nxt == IPPROTO_AH)
977 					eoff += (eh->ip6e_len + 2) << 2;
978 				else
979 					eoff += (eh->ip6e_len + 1) << 3;
980 				nxt = eh->ip6e_nxt;
981 				break;
982 			case IPPROTO_ROUTING:
983 				/*
984 				 * When the erroneous packet contains a
985 				 * routing header, we should examine the
986 				 * header to determine the final destination.
987 				 * Otherwise, we can't properly update
988 				 * information that depends on the final
989 				 * destination (e.g. path MTU).
990 				 */
991 				if (m->m_len < eoff + sizeof(*rth)) {
992 					m = m_pullup(m, eoff + sizeof(*rth));
993 					if (m == NULL) {
994 						IP6STAT_INC(ip6s_exthdrtoolong);
995 						*mp = m;
996 						return (-1);
997 					}
998 				}
999 				rth = (struct ip6_rthdr *)
1000 				    (mtod(m, caddr_t) + eoff);
1001 				rthlen = (rth->ip6r_len + 1) << 3;
1002 				/*
1003 				 * XXX: currently there is no
1004 				 * officially defined type other
1005 				 * than type-0.
1006 				 * Note that if the segment left field
1007 				 * is 0, all intermediate hops must
1008 				 * have been passed.
1009 				 */
1010 				if (rth->ip6r_segleft &&
1011 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1012 					int hops;
1013 
1014 					if (m->m_len < eoff + rthlen) {
1015 						m = m_pullup(m, eoff + rthlen);
1016 						if (m == NULL) {
1017 							IP6STAT_INC(
1018 							    ip6s_exthdrtoolong);
1019 							*mp = m;
1020 							return (-1);
1021 						}
1022 					}
1023 					rth0 = (struct ip6_rthdr0 *)
1024 					    (mtod(m, caddr_t) + eoff);
1025 
1026 					/* just ignore a bogus header */
1027 					if ((rth0->ip6r0_len % 2) == 0 &&
1028 					    (hops = rth0->ip6r0_len/2))
1029 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1030 				}
1031 				eoff += rthlen;
1032 				nxt = rth->ip6r_nxt;
1033 				break;
1034 			case IPPROTO_FRAGMENT:
1035 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1036 					m = m_pullup(m, eoff +
1037 					    sizeof(struct ip6_frag));
1038 					if (m == NULL) {
1039 						IP6STAT_INC(ip6s_exthdrtoolong);
1040 						*mp = m;
1041 						return (-1);
1042 					}
1043 				}
1044 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1045 				    eoff);
1046 				/*
1047 				 * Data after a fragment header is meaningless
1048 				 * unless it is the first fragment, but
1049 				 * we'll go to the notify label for path MTU
1050 				 * discovery.
1051 				 */
1052 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1053 					goto notify;
1054 
1055 				eoff += sizeof(struct ip6_frag);
1056 				nxt = fh->ip6f_nxt;
1057 				break;
1058 			default:
1059 				/*
1060 				 * This case includes ESP and the No Next
1061 				 * Header.  In such cases going to the notify
1062 				 * label does not have any meaning
1063 				 * (i.e. ctlfunc will be NULL), but we go
1064 				 * anyway since we might have to update
1065 				 * path MTU information.
1066 				 */
1067 				goto notify;
1068 			}
1069 		}
1070 	  notify:
1071 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1072 
1073 		/*
1074 		 * retrieve parameters from the inner IPv6 header, and convert
1075 		 * them into sockaddr structures.
1076 		 * XXX: there is no guarantee that the source or destination
1077 		 * addresses of the inner packet are in the same scope as
1078 		 * the addresses of the icmp packet.  But there is no other
1079 		 * way to determine the zone.
1080 		 */
1081 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1082 
1083 		/*
1084 		 * Protocol layers can't do anything useful with unspecified
1085 		 * addresses.
1086 		 */
1087 		if (IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_src) ||
1088 		    IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_dst))
1089 			goto freeit;
1090 
1091 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1092 		icmp6dst.sin6_family = AF_INET6;
1093 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1094 			icmp6dst.sin6_addr = eip6->ip6_dst;
1095 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1096 			goto freeit;
1097 		bzero(&icmp6src, sizeof(icmp6src));
1098 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1099 		icmp6src.sin6_family = AF_INET6;
1100 		icmp6src.sin6_addr = eip6->ip6_src;
1101 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1102 			goto freeit;
1103 		icmp6src.sin6_flowinfo =
1104 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1105 
1106 		ip6cp.ip6c_m = m;
1107 		ip6cp.ip6c_icmp6 = icmp6;
1108 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1109 		ip6cp.ip6c_off = eoff;
1110 		ip6cp.ip6c_finaldst = &icmp6dst;
1111 		ip6cp.ip6c_src = &icmp6src;
1112 		ip6cp.ip6c_nxt = nxt;
1113 
1114 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1115 			notifymtu = ntohl(icmp6->icmp6_mtu);
1116 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1117 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1118 		}
1119 
1120 		if (ip6_ctlprotox[nxt] != NULL)
1121 			ip6_ctlprotox[nxt](&ip6cp);
1122 	}
1123 	*mp = m;
1124 	return (0);
1125 
1126   freeit:
1127 	m_freem(m);
1128 	*mp = NULL;
1129 	return (-1);
1130 }
1131 
1132 void
icmp6_mtudisc_update(struct ip6ctlparam * ip6cp,int validated)1133 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1134 {
1135 	struct in6_addr *dst = &ip6cp->ip6c_finaldst->sin6_addr;
1136 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1137 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1138 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1139 	struct in_conninfo inc;
1140 	uint32_t max_mtu;
1141 
1142 #if 0
1143 	/*
1144 	 * RFC2460 section 5, last paragraph.
1145 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1146 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1147 	 * due to packet translator in the middle.
1148 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1149 	 * special handling.
1150 	 */
1151 	if (mtu < IPV6_MMTU)
1152 		return;
1153 #endif
1154 
1155 	/*
1156 	 * we reject ICMPv6 too big with abnormally small value.
1157 	 * XXX what is the good definition of "abnormally small"?
1158 	 */
1159 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1160 		return;
1161 
1162 	if (!validated)
1163 		return;
1164 
1165 	/*
1166 	 * In case the suggested mtu is less than IPV6_MMTU, we
1167 	 * only need to remember that it was for above mentioned
1168 	 * "alwaysfrag" case.
1169 	 * Try to be as close to the spec as possible.
1170 	 */
1171 	if (mtu < IPV6_MMTU)
1172 		mtu = IPV6_MMTU - 8;
1173 
1174 	bzero(&inc, sizeof(inc));
1175 	inc.inc_fibnum = M_GETFIB(m);
1176 	inc.inc_flags |= INC_ISIPV6;
1177 	inc.inc6_faddr = *dst;
1178 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1179 		return;
1180 
1181 	max_mtu = tcp_hc_getmtu(&inc);
1182 	if (max_mtu == 0)
1183 		max_mtu = tcp_maxmtu6(&inc, NULL);
1184 
1185 	if (mtu < max_mtu) {
1186 		tcp_hc_updatemtu(&inc, mtu);
1187 		ICMP6STAT_INC(icp6s_pmtuchg);
1188 	}
1189 }
1190 
1191 /*
1192  * Process a Node Information Query packet, based on
1193  * draft-ietf-ipngwg-icmp-name-lookups-07.
1194  *
1195  * Spec incompatibilities:
1196  * - IPv6 Subject address handling
1197  * - IPv4 Subject address handling support missing
1198  * - Proxy reply (answer even if it's not for me)
1199  * - joins NI group address at in6_ifattach() time only, does not cope
1200  *   with hostname changes by sethostname(3)
1201  */
1202 static struct mbuf *
ni6_input(struct mbuf * m,int off,struct prison * pr)1203 ni6_input(struct mbuf *m, int off, struct prison *pr)
1204 {
1205 	struct icmp6_nodeinfo *ni6, *nni6;
1206 	struct mbuf *n = NULL;
1207 	u_int16_t qtype;
1208 	int subjlen;
1209 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1210 	struct ni_reply_fqdn *fqdn;
1211 	int addrs;		/* for NI_QTYPE_NODEADDR */
1212 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1213 	struct in6_addr in6_subj; /* subject address */
1214 	struct ip6_hdr *ip6;
1215 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1216 	char *subj = NULL;
1217 	struct in6_ifaddr *ia6 = NULL;
1218 
1219 	ip6 = mtod(m, struct ip6_hdr *);
1220 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1221 
1222 	/*
1223 	 * Validate IPv6 source address.
1224 	 * The default configuration MUST be to refuse answering queries from
1225 	 * global-scope addresses according to RFC4602.
1226 	 * Notes:
1227 	 *  - it's not very clear what "refuse" means; this implementation
1228 	 *    simply drops it.
1229 	 *  - it's not very easy to identify global-scope (unicast) addresses
1230 	 *    since there are many prefixes for them.  It should be safer
1231 	 *    and in practice sufficient to check "all" but loopback and
1232 	 *    link-local (note that site-local unicast was deprecated and
1233 	 *    ULA is defined as global scope-wise)
1234 	 */
1235 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1236 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1237 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1238 		goto bad;
1239 
1240 	/*
1241 	 * Validate IPv6 destination address.
1242 	 *
1243 	 * The Responder must discard the Query without further processing
1244 	 * unless it is one of the Responder's unicast or anycast addresses, or
1245 	 * a link-local scope multicast address which the Responder has joined.
1246 	 * [RFC4602, Section 5.]
1247 	 */
1248 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1249 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1250 			goto bad;
1251 		/* else it's a link-local multicast, fine */
1252 	} else {		/* unicast or anycast */
1253 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1254 		if (ia6 == NULL)
1255 			goto bad; /* XXX impossible */
1256 
1257 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1258 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1259 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1260 				"a temporary address in %s:%d",
1261 			       __FILE__, __LINE__));
1262 			goto bad;
1263 		}
1264 	}
1265 
1266 	/* validate query Subject field. */
1267 	qtype = ntohs(ni6->ni_qtype);
1268 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1269 	switch (qtype) {
1270 	case NI_QTYPE_NOOP:
1271 	case NI_QTYPE_SUPTYPES:
1272 		/* 07 draft */
1273 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1274 			break;
1275 		/* FALLTHROUGH */
1276 	case NI_QTYPE_FQDN:
1277 	case NI_QTYPE_NODEADDR:
1278 	case NI_QTYPE_IPV4ADDR:
1279 		switch (ni6->ni_code) {
1280 		case ICMP6_NI_SUBJ_IPV6:
1281 #if ICMP6_NI_SUBJ_IPV6 != 0
1282 		case 0:
1283 #endif
1284 			/*
1285 			 * backward compatibility - try to accept 03 draft
1286 			 * format, where no Subject is present.
1287 			 */
1288 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1289 			    subjlen == 0) {
1290 				oldfqdn++;
1291 				break;
1292 			}
1293 #if ICMP6_NI_SUBJ_IPV6 != 0
1294 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1295 				goto bad;
1296 #endif
1297 
1298 			if (subjlen != sizeof(struct in6_addr))
1299 				goto bad;
1300 
1301 			/*
1302 			 * Validate Subject address.
1303 			 *
1304 			 * Not sure what exactly "address belongs to the node"
1305 			 * means in the spec, is it just unicast, or what?
1306 			 *
1307 			 * At this moment we consider Subject address as
1308 			 * "belong to the node" if the Subject address equals
1309 			 * to the IPv6 destination address; validation for
1310 			 * IPv6 destination address should have done enough
1311 			 * check for us.
1312 			 *
1313 			 * We do not do proxy at this moment.
1314 			 */
1315 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1316 			    subjlen, (caddr_t)&in6_subj);
1317 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1318 				goto bad;
1319 
1320 			subj = (char *)&in6_subj;
1321 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1322 				break;
1323 
1324 			/*
1325 			 * XXX if we are to allow other cases, we should really
1326 			 * be careful about scope here.
1327 			 * basically, we should disallow queries toward IPv6
1328 			 * destination X with subject Y,
1329 			 * if scope(X) > scope(Y).
1330 			 * if we allow scope(X) > scope(Y), it will result in
1331 			 * information leakage across scope boundary.
1332 			 */
1333 			goto bad;
1334 
1335 		case ICMP6_NI_SUBJ_FQDN:
1336 			/*
1337 			 * Validate Subject name with gethostname(3).
1338 			 *
1339 			 * The behavior may need some debate, since:
1340 			 * - we are not sure if the node has FQDN as
1341 			 *   hostname (returned by gethostname(3)).
1342 			 * - the code does wildcard match for truncated names.
1343 			 *   however, we are not sure if we want to perform
1344 			 *   wildcard match, if gethostname(3) side has
1345 			 *   truncated hostname.
1346 			 */
1347 			mtx_lock(&pr->pr_mtx);
1348 			n = ni6_nametodns(pr->pr_hostname,
1349 			    strlen(pr->pr_hostname), 0);
1350 			mtx_unlock(&pr->pr_mtx);
1351 			if (!n || n->m_next || n->m_len == 0)
1352 				goto bad;
1353 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1354 			    subjlen) {
1355 				m = m_pullup(m, off +
1356 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1357 				if (m == NULL) {
1358 					IP6STAT_INC(ip6s_exthdrtoolong);
1359 					goto bad;
1360 				}
1361 			}
1362 			/* ip6 possibly invalid but not used after. */
1363 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1364 			subj = (char *)(mtod(m, caddr_t) + off +
1365 			    sizeof(struct icmp6_nodeinfo));
1366 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1367 			    n->m_len)) {
1368 				goto bad;
1369 			}
1370 			m_freem(n);
1371 			n = NULL;
1372 			break;
1373 
1374 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1375 		default:
1376 			goto bad;
1377 		}
1378 		break;
1379 	}
1380 
1381 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1382 	switch (qtype) {
1383 	case NI_QTYPE_FQDN:
1384 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1385 			goto bad;
1386 		break;
1387 	case NI_QTYPE_NODEADDR:
1388 	case NI_QTYPE_IPV4ADDR:
1389 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1390 			goto bad;
1391 		break;
1392 	}
1393 
1394 	/* guess reply length */
1395 	switch (qtype) {
1396 	case NI_QTYPE_NOOP:
1397 		break;		/* no reply data */
1398 	case NI_QTYPE_SUPTYPES:
1399 		replylen += sizeof(u_int32_t);
1400 		break;
1401 	case NI_QTYPE_FQDN:
1402 		/* XXX will append an mbuf */
1403 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1404 		break;
1405 	case NI_QTYPE_NODEADDR:
1406 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1407 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1408 		    sizeof(u_int32_t))) > MCLBYTES)
1409 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1410 		break;
1411 	case NI_QTYPE_IPV4ADDR:
1412 		/* unsupported - should respond with unknown Qtype? */
1413 		break;
1414 	default:
1415 		/*
1416 		 * XXX: We must return a reply with the ICMP6 code
1417 		 * `unknown Qtype' in this case.  However we regard the case
1418 		 * as an FQDN query for backward compatibility.
1419 		 * Older versions set a random value to this field,
1420 		 * so it rarely varies in the defined qtypes.
1421 		 * But the mechanism is not reliable...
1422 		 * maybe we should obsolete older versions.
1423 		 */
1424 		qtype = NI_QTYPE_FQDN;
1425 		/* XXX will append an mbuf */
1426 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1427 		oldfqdn++;
1428 		break;
1429 	}
1430 
1431 	/* Allocate an mbuf to reply. */
1432 	if (replylen > MCLBYTES) {
1433 		/*
1434 		 * XXX: should we try to allocate more? But MCLBYTES
1435 		 * is probably much larger than IPV6_MMTU...
1436 		 */
1437 		goto bad;
1438 	}
1439 	if (replylen > MHLEN)
1440 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1441 	else
1442 		n = m_gethdr(M_NOWAIT, m->m_type);
1443 	if (n == NULL) {
1444 		m_freem(m);
1445 		return (NULL);
1446 	}
1447 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1448 	n->m_pkthdr.len = n->m_len = replylen;
1449 
1450 	/* copy mbuf header and IPv6 + Node Information base headers */
1451 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1452 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1453 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1454 
1455 	/* qtype dependent procedure */
1456 	switch (qtype) {
1457 	case NI_QTYPE_NOOP:
1458 		nni6->ni_code = ICMP6_NI_SUCCESS;
1459 		nni6->ni_flags = 0;
1460 		break;
1461 	case NI_QTYPE_SUPTYPES:
1462 	{
1463 		u_int32_t v;
1464 		nni6->ni_code = ICMP6_NI_SUCCESS;
1465 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1466 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1467 		v = (u_int32_t)htonl(0x0000000f);
1468 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1469 		break;
1470 	}
1471 	case NI_QTYPE_FQDN:
1472 		nni6->ni_code = ICMP6_NI_SUCCESS;
1473 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1474 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1475 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1476 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1477 		/*
1478 		 * XXX do we really have FQDN in hostname?
1479 		 */
1480 		mtx_lock(&pr->pr_mtx);
1481 		n->m_next = ni6_nametodns(pr->pr_hostname,
1482 		    strlen(pr->pr_hostname), oldfqdn);
1483 		mtx_unlock(&pr->pr_mtx);
1484 		if (n->m_next == NULL)
1485 			goto bad;
1486 		/* XXX we assume that n->m_next is not a chain */
1487 		if (n->m_next->m_next != NULL)
1488 			goto bad;
1489 		n->m_pkthdr.len += n->m_next->m_len;
1490 		break;
1491 	case NI_QTYPE_NODEADDR:
1492 	{
1493 		int lenlim, copied;
1494 
1495 		nni6->ni_code = ICMP6_NI_SUCCESS;
1496 		n->m_pkthdr.len = n->m_len =
1497 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1498 		lenlim = M_TRAILINGSPACE(n);
1499 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1500 		/* XXX: reset mbuf length */
1501 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1502 		    sizeof(struct icmp6_nodeinfo) + copied;
1503 		break;
1504 	}
1505 	default:
1506 		break;		/* XXX impossible! */
1507 	}
1508 
1509 	nni6->ni_type = ICMP6_NI_REPLY;
1510 	m_freem(m);
1511 	return (n);
1512 
1513   bad:
1514 	m_freem(m);
1515 	if (n)
1516 		m_freem(n);
1517 	return (NULL);
1518 }
1519 
1520 /*
1521  * make a mbuf with DNS-encoded string.  no compression support.
1522  *
1523  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1524  * treated as truncated name (two \0 at the end).  this is a wild guess.
1525  *
1526  * old - return pascal string if non-zero
1527  */
1528 static struct mbuf *
ni6_nametodns(const char * name,int namelen,int old)1529 ni6_nametodns(const char *name, int namelen, int old)
1530 {
1531 	struct mbuf *m;
1532 	char *cp, *ep;
1533 	const char *p, *q;
1534 	int i, len, nterm;
1535 
1536 	if (old)
1537 		len = namelen + 1;
1538 	else
1539 		len = MCLBYTES;
1540 
1541 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1542 	if (len > MLEN)
1543 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1544 	else
1545 		m = m_get(M_NOWAIT, MT_DATA);
1546 	if (m == NULL)
1547 		goto fail;
1548 
1549 	if (old) {
1550 		m->m_len = len;
1551 		*mtod(m, char *) = namelen;
1552 		bcopy(name, mtod(m, char *) + 1, namelen);
1553 		return m;
1554 	} else {
1555 		m->m_len = 0;
1556 		cp = mtod(m, char *);
1557 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1558 
1559 		/* if not certain about my name, return empty buffer */
1560 		if (namelen == 0)
1561 			return m;
1562 
1563 		/*
1564 		 * guess if it looks like shortened hostname, or FQDN.
1565 		 * shortened hostname needs two trailing "\0".
1566 		 */
1567 		i = 0;
1568 		for (p = name; p < name + namelen; p++) {
1569 			if (*p && *p == '.')
1570 				i++;
1571 		}
1572 		if (i < 2)
1573 			nterm = 2;
1574 		else
1575 			nterm = 1;
1576 
1577 		p = name;
1578 		while (cp < ep && p < name + namelen) {
1579 			i = 0;
1580 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1581 				i++;
1582 			/* result does not fit into mbuf */
1583 			if (cp + i + 1 >= ep)
1584 				goto fail;
1585 			/*
1586 			 * DNS label length restriction, RFC1035 page 8.
1587 			 * "i == 0" case is included here to avoid returning
1588 			 * 0-length label on "foo..bar".
1589 			 */
1590 			if (i <= 0 || i >= 64)
1591 				goto fail;
1592 			*cp++ = i;
1593 			bcopy(p, cp, i);
1594 			cp += i;
1595 			p = q;
1596 			if (p < name + namelen && *p == '.')
1597 				p++;
1598 		}
1599 		/* termination */
1600 		if (cp + nterm >= ep)
1601 			goto fail;
1602 		while (nterm-- > 0)
1603 			*cp++ = '\0';
1604 		m->m_len = cp - mtod(m, char *);
1605 		return m;
1606 	}
1607 
1608 	panic("should not reach here");
1609 	/* NOTREACHED */
1610 
1611  fail:
1612 	if (m)
1613 		m_freem(m);
1614 	return NULL;
1615 }
1616 
1617 /*
1618  * check if two DNS-encoded string matches.  takes care of truncated
1619  * form (with \0\0 at the end).  no compression support.
1620  * XXX upper/lowercase match (see RFC2065)
1621  */
1622 static int
ni6_dnsmatch(const char * a,int alen,const char * b,int blen)1623 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1624 {
1625 	const char *a0, *b0;
1626 	int l;
1627 
1628 	/* simplest case - need validation? */
1629 	if (alen == blen && bcmp(a, b, alen) == 0)
1630 		return 1;
1631 
1632 	a0 = a;
1633 	b0 = b;
1634 
1635 	/* termination is mandatory */
1636 	if (alen < 2 || blen < 2)
1637 		return 0;
1638 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1639 		return 0;
1640 	alen--;
1641 	blen--;
1642 
1643 	while (a - a0 < alen && b - b0 < blen) {
1644 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1645 			return 0;
1646 
1647 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1648 			return 0;
1649 		/* we don't support compression yet */
1650 		if (a[0] >= 64 || b[0] >= 64)
1651 			return 0;
1652 
1653 		/* truncated case */
1654 		if (a[0] == 0 && a - a0 == alen - 1)
1655 			return 1;
1656 		if (b[0] == 0 && b - b0 == blen - 1)
1657 			return 1;
1658 		if (a[0] == 0 || b[0] == 0)
1659 			return 0;
1660 
1661 		if (a[0] != b[0])
1662 			return 0;
1663 		l = a[0];
1664 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1665 			return 0;
1666 		if (bcmp(a + 1, b + 1, l) != 0)
1667 			return 0;
1668 
1669 		a += 1 + l;
1670 		b += 1 + l;
1671 	}
1672 
1673 	if (a - a0 == alen && b - b0 == blen)
1674 		return 1;
1675 	else
1676 		return 0;
1677 }
1678 
1679 /*
1680  * calculate the number of addresses to be returned in the node info reply.
1681  */
1682 static int
ni6_addrs(struct icmp6_nodeinfo * ni6,struct mbuf * m,struct ifnet ** ifpp,struct in6_addr * subj)1683 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1684     struct in6_addr *subj)
1685 {
1686 	struct ifnet *ifp;
1687 	struct in6_ifaddr *ifa6;
1688 	struct ifaddr *ifa;
1689 	int addrs = 0, addrsofif, iffound = 0;
1690 	int niflags = ni6->ni_flags;
1691 
1692 	NET_EPOCH_ASSERT();
1693 
1694 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1695 		switch (ni6->ni_code) {
1696 		case ICMP6_NI_SUBJ_IPV6:
1697 			if (subj == NULL) /* must be impossible... */
1698 				return (0);
1699 			break;
1700 		default:
1701 			/*
1702 			 * XXX: we only support IPv6 subject address for
1703 			 * this Qtype.
1704 			 */
1705 			return (0);
1706 		}
1707 	}
1708 
1709 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1710 		addrsofif = 0;
1711 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1712 			if (ifa->ifa_addr->sa_family != AF_INET6)
1713 				continue;
1714 			ifa6 = (struct in6_ifaddr *)ifa;
1715 
1716 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1717 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1718 				iffound = 1;
1719 
1720 			/*
1721 			 * IPv4-mapped addresses can only be returned by a
1722 			 * Node Information proxy, since they represent
1723 			 * addresses of IPv4-only nodes, which perforce do
1724 			 * not implement this protocol.
1725 			 * [icmp-name-lookups-07, Section 5.4]
1726 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1727 			 * this function at this moment.
1728 			 */
1729 
1730 			/* What do we have to do about ::1? */
1731 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1732 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1733 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1734 					continue;
1735 				break;
1736 			case IPV6_ADDR_SCOPE_SITELOCAL:
1737 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1738 					continue;
1739 				break;
1740 			case IPV6_ADDR_SCOPE_GLOBAL:
1741 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1742 					continue;
1743 				break;
1744 			default:
1745 				continue;
1746 			}
1747 
1748 			/*
1749 			 * check if anycast is okay.
1750 			 * XXX: just experimental.  not in the spec.
1751 			 */
1752 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1753 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1754 				continue; /* we need only unicast addresses */
1755 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1756 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1757 				continue;
1758 			}
1759 			addrsofif++; /* count the address */
1760 		}
1761 		if (iffound) {
1762 			*ifpp = ifp;
1763 			return (addrsofif);
1764 		}
1765 
1766 		addrs += addrsofif;
1767 	}
1768 
1769 	return (addrs);
1770 }
1771 
1772 static int
ni6_store_addrs(struct icmp6_nodeinfo * ni6,struct icmp6_nodeinfo * nni6,struct ifnet * ifp0,int resid)1773 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1774     struct ifnet *ifp0, int resid)
1775 {
1776 	struct ifnet *ifp;
1777 	struct in6_ifaddr *ifa6;
1778 	struct ifaddr *ifa;
1779 	struct ifnet *ifp_dep = NULL;
1780 	int copied = 0, allow_deprecated = 0;
1781 	u_char *cp = (u_char *)(nni6 + 1);
1782 	int niflags = ni6->ni_flags;
1783 	u_int32_t ltime;
1784 
1785 	NET_EPOCH_ASSERT();
1786 
1787 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1788 		return (0);	/* needless to copy */
1789 
1790 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1791   again:
1792 
1793 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1794 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1795 			if (ifa->ifa_addr->sa_family != AF_INET6)
1796 				continue;
1797 			ifa6 = (struct in6_ifaddr *)ifa;
1798 
1799 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1800 			    allow_deprecated == 0) {
1801 				/*
1802 				 * prefererred address should be put before
1803 				 * deprecated addresses.
1804 				 */
1805 
1806 				/* record the interface for later search */
1807 				if (ifp_dep == NULL)
1808 					ifp_dep = ifp;
1809 
1810 				continue;
1811 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1812 			    allow_deprecated != 0)
1813 				continue; /* we now collect deprecated addrs */
1814 
1815 			/* What do we have to do about ::1? */
1816 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1817 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1818 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1819 					continue;
1820 				break;
1821 			case IPV6_ADDR_SCOPE_SITELOCAL:
1822 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1823 					continue;
1824 				break;
1825 			case IPV6_ADDR_SCOPE_GLOBAL:
1826 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1827 					continue;
1828 				break;
1829 			default:
1830 				continue;
1831 			}
1832 
1833 			/*
1834 			 * check if anycast is okay.
1835 			 * XXX: just experimental.  not in the spec.
1836 			 */
1837 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1838 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1839 				continue;
1840 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1841 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1842 				continue;
1843 			}
1844 
1845 			/* now we can copy the address */
1846 			if (resid < sizeof(struct in6_addr) +
1847 			    sizeof(u_int32_t)) {
1848 				/*
1849 				 * We give up much more copy.
1850 				 * Set the truncate flag and return.
1851 				 */
1852 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1853 				return (copied);
1854 			}
1855 
1856 			/*
1857 			 * Set the TTL of the address.
1858 			 * The TTL value should be one of the following
1859 			 * according to the specification:
1860 			 *
1861 			 * 1. The remaining lifetime of a DHCP lease on the
1862 			 *    address, or
1863 			 * 2. The remaining Valid Lifetime of a prefix from
1864 			 *    which the address was derived through Stateless
1865 			 *    Autoconfiguration.
1866 			 *
1867 			 * Note that we currently do not support stateful
1868 			 * address configuration by DHCPv6, so the former
1869 			 * case can't happen.
1870 			 */
1871 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1872 				ltime = ND6_INFINITE_LIFETIME;
1873 			else {
1874 				if (ifa6->ia6_lifetime.ia6t_expire >
1875 				    time_uptime)
1876 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1877 				else
1878 					ltime = 0;
1879 			}
1880 
1881 			bcopy(&ltime, cp, sizeof(u_int32_t));
1882 			cp += sizeof(u_int32_t);
1883 
1884 			/* copy the address itself */
1885 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1886 			    sizeof(struct in6_addr));
1887 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1888 			cp += sizeof(struct in6_addr);
1889 
1890 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1891 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1892 		}
1893 		if (ifp0)	/* we need search only on the specified IF */
1894 			break;
1895 	}
1896 
1897 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1898 		ifp = ifp_dep;
1899 		allow_deprecated = 1;
1900 
1901 		goto again;
1902 	}
1903 
1904 	return (copied);
1905 }
1906 
1907 static bool
icmp6_rip6_match(const struct inpcb * inp,void * v)1908 icmp6_rip6_match(const struct inpcb *inp, void *v)
1909 {
1910 	struct ip6_hdr *ip6 = v;
1911 
1912 	if ((inp->inp_vflag & INP_IPV6) == 0)
1913 		return (false);
1914 	if (inp->inp_ip_p != IPPROTO_ICMPV6)
1915 		return (false);
1916 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1917 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1918 		return (false);
1919 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1920 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1921 		return (false);
1922 	return (true);
1923 }
1924 
1925 /*
1926  * XXX almost dup'ed code with rip6_input.
1927  */
1928 static int
icmp6_rip6_input(struct mbuf ** mp,int off)1929 icmp6_rip6_input(struct mbuf **mp, int off)
1930 {
1931 	struct mbuf *n, *m = *mp;
1932 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1933 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1934 	    INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1935 	struct inpcb *inp;
1936 	struct sockaddr_in6 fromsa;
1937 	struct icmp6_hdr *icmp6;
1938 	struct mbuf *opts = NULL;
1939 	int delivered = 0;
1940 
1941 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1942 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1943 
1944 	/*
1945 	 * XXX: the address may have embedded scope zone ID, which should be
1946 	 * hidden from applications.
1947 	 */
1948 	bzero(&fromsa, sizeof(fromsa));
1949 	fromsa.sin6_family = AF_INET6;
1950 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1951 	fromsa.sin6_addr = ip6->ip6_src;
1952 	if (sa6_recoverscope(&fromsa)) {
1953 		m_freem(m);
1954 		*mp = NULL;
1955 		return (IPPROTO_DONE);
1956 	}
1957 
1958 	while ((inp = inp_next(&inpi)) != NULL) {
1959 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1960 		    inp->in6p_icmp6filt))
1961 			continue;
1962 		/*
1963 		 * Recent network drivers tend to allocate a single
1964 		 * mbuf cluster, rather than to make a couple of
1965 		 * mbufs without clusters.  Also, since the IPv6 code
1966 		 * path tries to avoid m_pullup(), it is highly
1967 		 * probable that we still have an mbuf cluster here
1968 		 * even though the necessary length can be stored in an
1969 		 * mbuf's internal buffer.
1970 		 * Meanwhile, the default size of the receive socket
1971 		 * buffer for raw sockets is not so large.  This means
1972 		 * the possibility of packet loss is relatively higher
1973 		 * than before.  To avoid this scenario, we copy the
1974 		 * received data to a separate mbuf that does not use
1975 		 * a cluster, if possible.
1976 		 * XXX: it is better to copy the data after stripping
1977 		 * intermediate headers.
1978 		 */
1979 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1980 		    m->m_len <= MHLEN) {
1981 			n = m_get(M_NOWAIT, m->m_type);
1982 			if (n != NULL) {
1983 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1984 					bcopy(m->m_data, n->m_data, m->m_len);
1985 					n->m_len = m->m_len;
1986 				} else {
1987 					m_free(n);
1988 					n = NULL;
1989 				}
1990 			}
1991 		} else
1992 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1993 		if (n == NULL)
1994 			continue;
1995 		if (inp->inp_flags & INP_CONTROLOPTS)
1996 			ip6_savecontrol(inp, n, &opts);
1997 		/* strip intermediate headers */
1998 		m_adj(n, off);
1999 		SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
2000 		if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
2001 		    (struct sockaddr *)&fromsa, n, opts) == 0) {
2002 			soroverflow_locked(inp->inp_socket);
2003 			m_freem(n);
2004 			if (opts)
2005 				m_freem(opts);
2006 		} else {
2007 			sorwakeup_locked(inp->inp_socket);
2008 			delivered++;
2009 		}
2010 		opts = NULL;
2011 	}
2012 	m_freem(m);
2013 	*mp = NULL;
2014 	if (delivered == 0)
2015 		IP6STAT_DEC(ip6s_delivered);
2016 	return (IPPROTO_DONE);
2017 }
2018 
2019 /*
2020  * Reflect the ip6 packet back to the source.
2021  * OFF points to the icmp6 header, counted from the top of the mbuf.
2022  */
2023 static void
icmp6_reflect(struct mbuf * m,size_t off)2024 icmp6_reflect(struct mbuf *m, size_t off)
2025 {
2026 	struct in6_addr src6, *srcp;
2027 	struct ip6_hdr *ip6;
2028 	struct icmp6_hdr *icmp6;
2029 	struct in6_ifaddr *ia = NULL;
2030 	struct ifnet *outif = NULL;
2031 	int plen;
2032 	int type, code, hlim;
2033 
2034 	/* too short to reflect */
2035 	if (off < sizeof(struct ip6_hdr)) {
2036 		nd6log((LOG_DEBUG,
2037 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2038 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2039 		    __FILE__, __LINE__));
2040 		goto bad;
2041 	}
2042 
2043 	/*
2044 	 * If there are extra headers between IPv6 and ICMPv6, strip
2045 	 * off that header first.
2046 	 */
2047 #ifdef DIAGNOSTIC
2048 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2049 		panic("assumption failed in icmp6_reflect");
2050 #endif
2051 	if (off > sizeof(struct ip6_hdr)) {
2052 		size_t l;
2053 		struct ip6_hdr nip6;
2054 
2055 		l = off - sizeof(struct ip6_hdr);
2056 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2057 		m_adj(m, l);
2058 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2059 		if (m->m_len < l) {
2060 			if ((m = m_pullup(m, l)) == NULL)
2061 				return;
2062 		}
2063 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2064 	} else /* off == sizeof(struct ip6_hdr) */ {
2065 		size_t l;
2066 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2067 		if (m->m_len < l) {
2068 			if ((m = m_pullup(m, l)) == NULL)
2069 				return;
2070 		}
2071 	}
2072 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2073 	ip6 = mtod(m, struct ip6_hdr *);
2074 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2075 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2076 	type = icmp6->icmp6_type; /* keep type for statistics */
2077 	code = icmp6->icmp6_code; /* ditto. */
2078 	hlim = 0;
2079 	srcp = NULL;
2080 
2081 	/*
2082 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2083 	 * use dst as the src for the reply.
2084 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2085 	 * (for example) when we encounter an error while forwarding procedure
2086 	 * destined to a duplicated address of ours.
2087 	 */
2088 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2089 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2090 		if (ia != NULL && !(ia->ia6_flags &
2091 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2092 			src6 = ia->ia_addr.sin6_addr;
2093 			srcp = &src6;
2094 
2095 			if (m->m_pkthdr.rcvif != NULL) {
2096 				/* XXX: This may not be the outgoing interface */
2097 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2098 			} else
2099 				hlim = V_ip6_defhlim;
2100 		}
2101 	}
2102 
2103 	if (srcp == NULL) {
2104 		int error;
2105 		struct in6_addr dst6;
2106 		uint32_t scopeid;
2107 
2108 		/*
2109 		 * This case matches to multicasts, our anycast, or unicasts
2110 		 * that we do not own.  Select a source address based on the
2111 		 * source address of the erroneous packet.
2112 		 */
2113 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2114 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2115 		    scopeid, NULL, &src6, &hlim);
2116 
2117 		if (error) {
2118 			char ip6buf[INET6_ADDRSTRLEN];
2119 			nd6log((LOG_DEBUG,
2120 			    "icmp6_reflect: source can't be determined: "
2121 			    "dst=%s, error=%d\n",
2122 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2123 			goto bad;
2124 		}
2125 		srcp = &src6;
2126 	}
2127 	/*
2128 	 * ip6_input() drops a packet if its src is multicast.
2129 	 * So, the src is never multicast.
2130 	 */
2131 	ip6->ip6_dst = ip6->ip6_src;
2132 	ip6->ip6_src = *srcp;
2133 	ip6->ip6_flow = 0;
2134 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2135 	ip6->ip6_vfc |= IPV6_VERSION;
2136 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2137 	ip6->ip6_hlim = hlim;
2138 
2139 	icmp6->icmp6_cksum = 0;
2140 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2141 	    sizeof(struct ip6_hdr), plen);
2142 
2143 	/*
2144 	 * XXX option handling
2145 	 */
2146 
2147 	m->m_flags &= ~(M_BCAST|M_MCAST);
2148 	m->m_pkthdr.rcvif = NULL;
2149 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2150 	if (outif)
2151 		icmp6_ifoutstat_inc(outif, type, code);
2152 
2153 	return;
2154 
2155  bad:
2156 	m_freem(m);
2157 	return;
2158 }
2159 
2160 static const char *
icmp6_redirect_diag(struct in6_addr * src6,struct in6_addr * dst6,struct in6_addr * tgt6)2161 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2162     struct in6_addr *tgt6)
2163 {
2164 	static char buf[1024];
2165 	char ip6bufs[INET6_ADDRSTRLEN];
2166 	char ip6bufd[INET6_ADDRSTRLEN];
2167 	char ip6buft[INET6_ADDRSTRLEN];
2168 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2169 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2170 	    ip6_sprintf(ip6buft, tgt6));
2171 	return buf;
2172 }
2173 
2174 void
icmp6_redirect_input(struct mbuf * m,int off)2175 icmp6_redirect_input(struct mbuf *m, int off)
2176 {
2177 	struct ifnet *ifp;
2178 	struct ip6_hdr *ip6;
2179 	struct nd_redirect *nd_rd;
2180 	struct in6_addr src6, redtgt6, reddst6;
2181 	union nd_opts ndopts;
2182 	char ip6buf[INET6_ADDRSTRLEN];
2183 	char *lladdr;
2184 	int icmp6len, is_onlink, is_router, lladdrlen;
2185 
2186 	M_ASSERTPKTHDR(m);
2187 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2188 
2189 	/* XXX if we are router, we don't update route by icmp6 redirect */
2190 	if (V_ip6_forwarding)
2191 		goto freeit;
2192 	if (!V_icmp6_rediraccept)
2193 		goto freeit;
2194 
2195 	/* RFC 6980: Nodes MUST silently ignore fragments */
2196 	if(m->m_flags & M_FRAGMENTED)
2197 		goto freeit;
2198 
2199 	ip6 = mtod(m, struct ip6_hdr *);
2200 	icmp6len = ntohs(ip6->ip6_plen);
2201 	if (m->m_len < off + icmp6len) {
2202 		m = m_pullup(m, off + icmp6len);
2203 		if (m == NULL) {
2204 			IP6STAT_INC(ip6s_exthdrtoolong);
2205 			return;
2206 		}
2207 	}
2208 	ip6 = mtod(m, struct ip6_hdr *);
2209 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2210 
2211 	ifp = m->m_pkthdr.rcvif;
2212 	redtgt6 = nd_rd->nd_rd_target;
2213 	reddst6 = nd_rd->nd_rd_dst;
2214 
2215 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2216 	    in6_setscope(&reddst6, ifp, NULL)) {
2217 		goto freeit;
2218 	}
2219 
2220 	/* validation */
2221 	src6 = ip6->ip6_src;
2222 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2223 		nd6log((LOG_ERR,
2224 		    "ICMP6 redirect sent from %s rejected; "
2225 		    "must be from linklocal\n",
2226 		    ip6_sprintf(ip6buf, &src6)));
2227 		goto bad;
2228 	}
2229 	if (__predict_false(ip6->ip6_hlim != 255)) {
2230 		ICMP6STAT_INC(icp6s_invlhlim);
2231 		nd6log((LOG_ERR,
2232 		    "ICMP6 redirect sent from %s rejected; "
2233 		    "hlim=%d (must be 255)\n",
2234 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2235 		goto bad;
2236 	}
2237     {
2238 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2239 	struct nhop_object *nh;
2240 	struct in6_addr kdst;
2241 	uint32_t scopeid;
2242 
2243 	in6_splitscope(&reddst6, &kdst, &scopeid);
2244 	NET_EPOCH_ASSERT();
2245 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2246 	if (nh != NULL) {
2247 		struct in6_addr nh_addr;
2248 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2249 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2250 			nd6log((LOG_ERR,
2251 			    "ICMP6 redirect rejected; no route "
2252 			    "with inet6 gateway found for redirect dst: %s\n",
2253 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2254 			goto bad;
2255 		}
2256 
2257 		/*
2258 		 * Embed scope zone id into next hop address.
2259 		 */
2260 		nh_addr = nh->gw6_sa.sin6_addr;
2261 
2262 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2263 			nd6log((LOG_ERR,
2264 			    "ICMP6 redirect rejected; "
2265 			    "not equal to gw-for-src=%s (must be same): "
2266 			    "%s\n",
2267 			    ip6_sprintf(ip6buf, &nh_addr),
2268 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2269 			goto bad;
2270 		}
2271 	} else {
2272 		nd6log((LOG_ERR,
2273 		    "ICMP6 redirect rejected; "
2274 		    "no route found for redirect dst: %s\n",
2275 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2276 		goto bad;
2277 	}
2278     }
2279 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2280 		nd6log((LOG_ERR,
2281 		    "ICMP6 redirect rejected; "
2282 		    "redirect dst must be unicast: %s\n",
2283 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		goto bad;
2285 	}
2286 
2287 	is_router = is_onlink = 0;
2288 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2289 		is_router = 1;	/* router case */
2290 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2291 		is_onlink = 1;	/* on-link destination case */
2292 	if (!is_router && !is_onlink) {
2293 		nd6log((LOG_ERR,
2294 		    "ICMP6 redirect rejected; "
2295 		    "neither router case nor onlink case: %s\n",
2296 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2297 		goto bad;
2298 	}
2299 
2300 	icmp6len -= sizeof(*nd_rd);
2301 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2302 	if (nd6_options(&ndopts) < 0) {
2303 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2304 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2305 		/* nd6_options have incremented stats */
2306 		goto freeit;
2307 	}
2308 
2309 	lladdr = NULL;
2310 	lladdrlen = 0;
2311 	if (ndopts.nd_opts_tgt_lladdr) {
2312 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2313 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2314 	}
2315 
2316 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2317 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2318 		    "(if %d, icmp6 packet %d): %s\n",
2319 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2320 		    ifp->if_addrlen, lladdrlen - 2,
2321 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2322 		goto bad;
2323 	}
2324 
2325 	/* Validation passed. */
2326 
2327 	/* RFC 2461 8.3 */
2328 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2329 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2330 
2331 	/*
2332 	 * Install a gateway route in the better-router case or an interface
2333 	 * route in the on-link-destination case.
2334 	 */
2335 	{
2336 		struct sockaddr_in6 sdst;
2337 		struct sockaddr_in6 sgw;
2338 		struct sockaddr_in6 ssrc;
2339 		struct sockaddr *gw;
2340 		int rt_flags;
2341 		u_int fibnum;
2342 
2343 		bzero(&sdst, sizeof(sdst));
2344 		bzero(&ssrc, sizeof(ssrc));
2345 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2346 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2347 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2348 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2349 		rt_flags = 0;
2350 		if (is_router) {
2351 			bzero(&sgw, sizeof(sgw));
2352 			sgw.sin6_family = AF_INET6;
2353 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2354 			bcopy(&redtgt6, &sgw.sin6_addr,
2355 				sizeof(struct in6_addr));
2356 			gw = (struct sockaddr *)&sgw;
2357 			rt_flags |= RTF_GATEWAY;
2358 		} else
2359 			gw = ifp->if_addr->ifa_addr;
2360 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2361 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2362 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2363 			    V_icmp6_redirtimeout);
2364 	}
2365 
2366  freeit:
2367 	m_freem(m);
2368 	return;
2369 
2370  bad:
2371 	ICMP6STAT_INC(icp6s_badredirect);
2372 	m_freem(m);
2373 }
2374 
2375 void
icmp6_redirect_output(struct mbuf * m0,struct nhop_object * nh)2376 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2377 {
2378 	struct ifnet *ifp;	/* my outgoing interface */
2379 	struct in6_addr *ifp_ll6;
2380 	struct in6_addr *router_ll6;
2381 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2382 	struct mbuf *m = NULL;	/* newly allocated one */
2383 	struct m_tag *mtag;
2384 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2385 	struct nd_redirect *nd_rd;
2386 	struct llentry *ln = NULL;
2387 	size_t maxlen;
2388 	u_char *p;
2389 	struct ifnet *outif = NULL;
2390 	struct sockaddr_in6 src_sa;
2391 
2392 	icmp6_errcount(ND_REDIRECT, 0);
2393 
2394 	/* if we are not router, we don't send icmp6 redirect */
2395 	if (!V_ip6_forwarding)
2396 		goto fail;
2397 
2398 	/* sanity check */
2399 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2400 		goto fail;
2401 
2402 	/*
2403 	 * Address check:
2404 	 *  the source address must identify a neighbor, and
2405 	 *  the destination address must not be a multicast address
2406 	 *  [RFC 2461, sec 8.2]
2407 	 */
2408 	sip6 = mtod(m0, struct ip6_hdr *);
2409 	bzero(&src_sa, sizeof(src_sa));
2410 	src_sa.sin6_family = AF_INET6;
2411 	src_sa.sin6_len = sizeof(src_sa);
2412 	src_sa.sin6_addr = sip6->ip6_src;
2413 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2414 		goto fail;
2415 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2416 		goto fail;	/* what should we do here? */
2417 
2418 	/* rate limit */
2419 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2420 		goto fail;
2421 
2422 	/*
2423 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2424 	 * we almost always ask for an mbuf cluster for simplicity.
2425 	 * (MHLEN < IPV6_MMTU is almost always true)
2426 	 */
2427 #if IPV6_MMTU >= MCLBYTES
2428 # error assumption failed about IPV6_MMTU and MCLBYTES
2429 #endif
2430 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2431 	if (m == NULL)
2432 		goto fail;
2433 	M_SETFIB(m, M_GETFIB(m0));
2434 	maxlen = M_TRAILINGSPACE(m);
2435 	maxlen = min(IPV6_MMTU, maxlen);
2436 	/* just for safety */
2437 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2438 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2439 		goto fail;
2440 	}
2441 
2442 	{
2443 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2444 		struct in6_ifaddr *ia;
2445 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2446 						 IN6_IFF_NOTREADY|
2447 						 IN6_IFF_ANYCAST)) == NULL)
2448 			goto fail;
2449 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2450 		/* XXXRW: reference released prematurely. */
2451 		ifa_free(&ia->ia_ifa);
2452 	}
2453 
2454 	/* get ip6 linklocal address for the router. */
2455 	if (nh->nh_flags & NHF_GATEWAY) {
2456 		struct sockaddr_in6 *sin6;
2457 		sin6 = &nh->gw6_sa;
2458 		router_ll6 = &sin6->sin6_addr;
2459 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2460 			router_ll6 = (struct in6_addr *)NULL;
2461 	} else
2462 		router_ll6 = (struct in6_addr *)NULL;
2463 
2464 	/* ip6 */
2465 	ip6 = mtod(m, struct ip6_hdr *);
2466 	ip6->ip6_flow = 0;
2467 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2468 	ip6->ip6_vfc |= IPV6_VERSION;
2469 	/* ip6->ip6_plen will be set later */
2470 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2471 	ip6->ip6_hlim = 255;
2472 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2473 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2474 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2475 
2476 	/* ND Redirect */
2477 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2478 	nd_rd->nd_rd_type = ND_REDIRECT;
2479 	nd_rd->nd_rd_code = 0;
2480 	nd_rd->nd_rd_reserved = 0;
2481 	if (nh->nh_flags & NHF_GATEWAY) {
2482 		/*
2483 		 * nd_rd->nd_rd_target must be a link-local address in
2484 		 * better router cases.
2485 		 */
2486 		if (!router_ll6)
2487 			goto fail;
2488 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2489 		    sizeof(nd_rd->nd_rd_target));
2490 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2491 		    sizeof(nd_rd->nd_rd_dst));
2492 	} else {
2493 		/* make sure redtgt == reddst */
2494 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2495 		    sizeof(nd_rd->nd_rd_target));
2496 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2497 		    sizeof(nd_rd->nd_rd_dst));
2498 	}
2499 
2500 	p = (u_char *)(nd_rd + 1);
2501 
2502 	if (!router_ll6)
2503 		goto nolladdropt;
2504 
2505 	{
2506 		/* target lladdr option */
2507 		int len;
2508 		struct nd_opt_hdr *nd_opt;
2509 		char *lladdr;
2510 
2511 		ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2512 		if (ln == NULL)
2513 			goto nolladdropt;
2514 
2515 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2516 		len = (len + 7) & ~7;	/* round by 8 */
2517 		/* safety check */
2518 		if (len + (p - (u_char *)ip6) > maxlen)
2519 			goto nolladdropt;
2520 
2521 		if (ln->la_flags & LLE_VALID) {
2522 			nd_opt = (struct nd_opt_hdr *)p;
2523 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2524 			nd_opt->nd_opt_len = len >> 3;
2525 			lladdr = (char *)(nd_opt + 1);
2526 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2527 			p += len;
2528 		}
2529 	}
2530 nolladdropt:
2531 	if (ln != NULL)
2532 		LLE_RUNLOCK(ln);
2533 
2534 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2535 
2536 	/* just to be safe */
2537 #ifdef M_DECRYPTED	/*not openbsd*/
2538 	if (m0->m_flags & M_DECRYPTED)
2539 		goto noredhdropt;
2540 #endif
2541 	if (p - (u_char *)ip6 > maxlen)
2542 		goto noredhdropt;
2543 
2544 	{
2545 		/* redirected header option */
2546 		int len;
2547 		struct nd_opt_rd_hdr *nd_opt_rh;
2548 
2549 		/*
2550 		 * compute the maximum size for icmp6 redirect header option.
2551 		 * XXX room for auth header?
2552 		 */
2553 		len = maxlen - (p - (u_char *)ip6);
2554 		len &= ~7;
2555 
2556 		/* This is just for simplicity. */
2557 		if (m0->m_pkthdr.len != m0->m_len) {
2558 			if (m0->m_next) {
2559 				m_freem(m0->m_next);
2560 				m0->m_next = NULL;
2561 			}
2562 			m0->m_pkthdr.len = m0->m_len;
2563 		}
2564 
2565 		/*
2566 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2567 		 * about padding/truncate rule for the original IP packet.
2568 		 * From the discussion on IPv6imp in Feb 1999,
2569 		 * the consensus was:
2570 		 * - "attach as much as possible" is the goal
2571 		 * - pad if not aligned (original size can be guessed by
2572 		 *   original ip6 header)
2573 		 * Following code adds the padding if it is simple enough,
2574 		 * and truncates if not.
2575 		 */
2576 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2577 			panic("assumption failed in %s:%d", __FILE__,
2578 			    __LINE__);
2579 
2580 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2581 			/* not enough room, truncate */
2582 			m0->m_pkthdr.len = m0->m_len = len -
2583 			    sizeof(*nd_opt_rh);
2584 		} else {
2585 			/* enough room, pad or truncate */
2586 			size_t extra;
2587 
2588 			extra = m0->m_pkthdr.len % 8;
2589 			if (extra) {
2590 				/* pad if easy enough, truncate if not */
2591 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2592 					/* pad */
2593 					m0->m_len += (8 - extra);
2594 					m0->m_pkthdr.len += (8 - extra);
2595 				} else {
2596 					/* truncate */
2597 					m0->m_pkthdr.len -= extra;
2598 					m0->m_len -= extra;
2599 				}
2600 			}
2601 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2602 			m0->m_pkthdr.len = m0->m_len = len -
2603 			    sizeof(*nd_opt_rh);
2604 		}
2605 
2606 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2607 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2608 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2609 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2610 		p += sizeof(*nd_opt_rh);
2611 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2612 
2613 		/* connect m0 to m */
2614 		m_tag_delete_chain(m0, NULL);
2615 		m0->m_flags &= ~M_PKTHDR;
2616 		m->m_next = m0;
2617 		m->m_pkthdr.len = m->m_len + m0->m_len;
2618 		m0 = NULL;
2619 	}
2620 noredhdropt:;
2621 	if (m0) {
2622 		m_freem(m0);
2623 		m0 = NULL;
2624 	}
2625 
2626 	/* XXX: clear embedded link IDs in the inner header */
2627 	in6_clearscope(&sip6->ip6_src);
2628 	in6_clearscope(&sip6->ip6_dst);
2629 	in6_clearscope(&nd_rd->nd_rd_target);
2630 	in6_clearscope(&nd_rd->nd_rd_dst);
2631 
2632 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2633 
2634 	nd_rd->nd_rd_cksum = 0;
2635 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2636 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2637 
2638         if (send_sendso_input_hook != NULL) {
2639 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2640 			M_NOWAIT);
2641 		if (mtag == NULL)
2642 			goto fail;
2643 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2644 		m_tag_prepend(m, mtag);
2645 	}
2646 
2647 	/* send the packet to outside... */
2648 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2649 	if (outif) {
2650 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2651 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2652 	}
2653 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2654 
2655 	return;
2656 
2657 fail:
2658 	if (m)
2659 		m_freem(m);
2660 	if (m0)
2661 		m_freem(m0);
2662 }
2663 
2664 /*
2665  * ICMPv6 socket option processing.
2666  */
2667 int
icmp6_ctloutput(struct socket * so,struct sockopt * sopt)2668 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2669 {
2670 	int error = 0;
2671 	int optlen;
2672 	struct inpcb *inp = sotoinpcb(so);
2673 	int level, op, optname;
2674 
2675 	if (sopt) {
2676 		level = sopt->sopt_level;
2677 		op = sopt->sopt_dir;
2678 		optname = sopt->sopt_name;
2679 		optlen = sopt->sopt_valsize;
2680 	} else
2681 		level = op = optname = optlen = 0;
2682 
2683 	if (level != IPPROTO_ICMPV6) {
2684 		return EINVAL;
2685 	}
2686 
2687 	switch (op) {
2688 	case PRCO_SETOPT:
2689 		switch (optname) {
2690 		case ICMP6_FILTER:
2691 		    {
2692 			struct icmp6_filter ic6f;
2693 
2694 			if (optlen != sizeof(ic6f)) {
2695 				error = EMSGSIZE;
2696 				break;
2697 			}
2698 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2699 			if (error == 0) {
2700 				INP_WLOCK(inp);
2701 				*inp->in6p_icmp6filt = ic6f;
2702 				INP_WUNLOCK(inp);
2703 			}
2704 			break;
2705 		    }
2706 
2707 		default:
2708 			error = ENOPROTOOPT;
2709 			break;
2710 		}
2711 		break;
2712 
2713 	case PRCO_GETOPT:
2714 		switch (optname) {
2715 		case ICMP6_FILTER:
2716 		    {
2717 			struct icmp6_filter ic6f;
2718 
2719 			INP_RLOCK(inp);
2720 			ic6f = *inp->in6p_icmp6filt;
2721 			INP_RUNLOCK(inp);
2722 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2723 			break;
2724 		    }
2725 
2726 		default:
2727 			error = ENOPROTOOPT;
2728 			break;
2729 		}
2730 		break;
2731 	}
2732 
2733 	return (error);
2734 }
2735 
2736 static int sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS);
2737 VNET_DEFINE_STATIC(u_int, icmp6errppslim) = 100;
2738 #define	V_icmp6errppslim	VNET(icmp6errppslim)
2739 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ERRPPSLIMIT, errppslimit,
2740     CTLTYPE_UINT | CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6errppslim), 0,
2741     &sysctl_icmp6lim_and_jitter, "IU",
2742     "Maximum number of ICMPv6 error/reply messages per second");
2743 
2744 VNET_DEFINE_STATIC(int, icmp6lim_curr_jitter) = 0;
2745 #define	V_icmp6lim_curr_jitter	VNET(icmp6lim_curr_jitter)
2746 
2747 VNET_DEFINE_STATIC(u_int, icmp6lim_jitter) = 8;
2748 #define	V_icmp6lim_jitter	VNET(icmp6lim_jitter)
2749 SYSCTL_PROC(_net_inet6_icmp6, OID_AUTO, icmp6lim_jitter, CTLTYPE_UINT |
2750     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_jitter), 0,
2751     &sysctl_icmp6lim_and_jitter, "IU",
2752     "Random errppslimit jitter adjustment limit");
2753 
2754 VNET_DEFINE_STATIC(int, icmp6lim_output) = 1;
2755 #define	V_icmp6lim_output	VNET(icmp6lim_output)
2756 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, icmp6lim_output,
2757     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_output), 0,
2758     "Enable logging of ICMPv6 response rate limiting");
2759 
2760 typedef enum {
2761 	RATELIM_PARAM_PROB = 0,
2762 	RATELIM_TOO_BIG,
2763 	RATELIM_UNREACH,
2764 	RATELIM_TEXCEED,
2765 	RATELIM_REDIR,
2766 	RATELIM_REPLY,
2767 	RATELIM_OTHER,
2768 	RATELIM_MAX
2769 } ratelim_which;
2770 
2771 static const char *icmp6_rate_descrs[RATELIM_MAX] = {
2772 	[RATELIM_PARAM_PROB] = "bad IPv6 header",
2773 	[RATELIM_TOO_BIG] = "packet too big",
2774 	[RATELIM_UNREACH] = "destination unreachable",
2775 	[RATELIM_TEXCEED] = "time exceeded",
2776 	[RATELIM_REPLY] = "echo reply",
2777 	[RATELIM_REDIR] = "neighbor discovery redirect",
2778 	[RATELIM_OTHER] = "(other)",
2779 };
2780 
2781 static void
icmp6lim_new_jitter(void)2782 icmp6lim_new_jitter(void)
2783 {
2784 	/*
2785 	 * Adjust limit +/- to jitter the measurement to deny a side-channel
2786 	 * port scan as in https://dl.acm.org/doi/10.1145/3372297.3417280
2787 	 */
2788 	if (V_icmp6lim_jitter > 0)
2789 		V_icmp6lim_curr_jitter =
2790 		    arc4random_uniform(V_icmp6lim_jitter * 2 + 1) -
2791 		    V_icmp6lim_jitter;
2792 }
2793 
2794 static int
sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)2795 sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)
2796 {
2797 	uint32_t new;
2798 	int error;
2799 	bool lim;
2800 
2801 	MPASS(oidp->oid_arg1 == &VNET_NAME(icmp6errppslim) ||
2802 	    oidp->oid_arg1 == &VNET_NAME(icmp6lim_jitter));
2803 
2804 	lim = (oidp->oid_arg1 == &VNET_NAME(icmp6errppslim));
2805 	new = lim ? V_icmp6errppslim : V_icmp6lim_jitter;
2806 	error = sysctl_handle_int(oidp, &new, 0, req);
2807 	if (error == 0 && req->newptr) {
2808 		if (lim) {
2809 			if (new != 0 && new <= V_icmp6lim_jitter)
2810 				error = EINVAL;
2811 			else
2812 				V_icmp6errppslim = new;
2813 		} else {
2814 			if (new >= V_icmp6errppslim)
2815 				error = EINVAL;
2816 			else {
2817 				V_icmp6lim_jitter = new;
2818 				icmp6lim_new_jitter();
2819 			}
2820 		}
2821 	}
2822 	MPASS(V_icmp6errppslim + V_icmp6lim_curr_jitter >= 0);
2823 
2824 	return (error);
2825 }
2826 
2827 
2828 VNET_DEFINE_STATIC(struct counter_rate, icmp6_rates[RATELIM_MAX]);
2829 #define	V_icmp6_rates	VNET(icmp6_rates)
2830 
2831 static void
icmp6_ratelimit_init(void)2832 icmp6_ratelimit_init(void)
2833 {
2834 
2835 	for (int i = 0; i < RATELIM_MAX; i++) {
2836 		V_icmp6_rates[i].cr_rate = counter_u64_alloc(M_WAITOK);
2837 		V_icmp6_rates[i].cr_ticks = ticks;
2838 	}
2839 	icmp6lim_new_jitter();
2840 }
2841 VNET_SYSINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
2842     icmp6_ratelimit_init, NULL);
2843 
2844 #ifdef VIMAGE
2845 static void
icmp6_ratelimit_uninit(void)2846 icmp6_ratelimit_uninit(void)
2847 {
2848 
2849 	for (int i = 0; i < RATELIM_MAX; i++)
2850 		counter_u64_free(V_icmp6_rates[i].cr_rate);
2851 }
2852 VNET_SYSUNINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
2853     icmp6_ratelimit_uninit, NULL);
2854 #endif
2855 
2856 /*
2857  * Perform rate limit check.
2858  * Returns 0 if it is okay to send the icmp6 packet.
2859  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2860  * limitation.
2861  *
2862  * XXX per-destination/type check necessary?
2863  *
2864  * dst - not used at this moment
2865  * code - not used at this moment
2866  */
2867 int
icmp6_ratelimit(const struct in6_addr * dst,const int type,const int code)2868 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
2869 {
2870 	ratelim_which which;
2871 	int64_t pps;
2872 
2873 	if (V_icmp6errppslim == 0)
2874 		return (0);
2875 
2876 	switch (type) {
2877 	case ICMP6_PARAM_PROB:
2878 		which = RATELIM_PARAM_PROB;
2879 		break;
2880 	case ICMP6_PACKET_TOO_BIG:
2881 		which = RATELIM_TOO_BIG;
2882 		break;
2883 	case ICMP6_DST_UNREACH:
2884 		which = RATELIM_UNREACH;
2885 		break;
2886 	case ICMP6_TIME_EXCEEDED:
2887 		which = RATELIM_TEXCEED;
2888 		break;
2889 	case ND_REDIRECT:
2890 		which = RATELIM_REDIR;
2891 		break;
2892 	case ICMP6_ECHO_REPLY:
2893 		which = RATELIM_REPLY;
2894 		break;
2895 	default:
2896 		which = RATELIM_OTHER;
2897 		break;
2898 	};
2899 
2900 	pps = counter_ratecheck(&V_icmp6_rates[which], V_icmp6errppslim +
2901 	    V_icmp6lim_curr_jitter);
2902 	if (pps > 0) {
2903 		if (V_icmp6lim_output)
2904 			log(LOG_NOTICE, "Limiting ICMPv6 %s output from %jd "
2905 			    "to %d packets/sec\n", icmp6_rate_descrs[which],
2906 			    (intmax_t )pps, V_icmp6errppslim +
2907 			    V_icmp6lim_curr_jitter);
2908 		icmp6lim_new_jitter();
2909 	}
2910 	if (pps == -1) {
2911 		ICMP6STAT_INC(icp6s_toofreq);
2912 		return (-1);
2913 	}
2914 
2915 	return (0);
2916 }
2917