1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
32 */
33
34 /*-
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 *
62 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
63 */
64
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD: stable/12/sys/netinet6/icmp6.c 370289 2021-08-11 02:35:48Z kbowling $");
67
68 #define MBUF_PRIVATE /* XXXRW: Optimisation tries to avoid M_EXT mbufs */
69
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72
73 #include <sys/param.h>
74 #include <sys/domain.h>
75 #include <sys/jail.h>
76 #include <sys/kernel.h>
77 #include <sys/lock.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/signalvar.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sx.h>
86 #include <sys/syslog.h>
87 #include <sys/systm.h>
88 #include <sys/time.h>
89
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/if_llatbl.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96 #include <net/vnet.h>
97
98 #include <netinet/in.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/in_var.h>
101 #include <netinet/ip6.h>
102 #include <netinet/icmp6.h>
103 #include <netinet/tcp_var.h>
104
105 #include <netinet6/in6_fib.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/ip6protosw.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 struct domain inet6domain;
116
117 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
118 VNET_PCPUSTAT_SYSINIT(icmp6stat);
119
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123
124 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
125 VNET_DECLARE(struct inpcbhead, ripcb);
126 VNET_DECLARE(int, icmp6errppslim);
127 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
128 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130
131 #define V_ripcbinfo VNET(ripcbinfo)
132 #define V_ripcb VNET(ripcb)
133 #define V_icmp6errppslim VNET(icmp6errppslim)
134 #define V_icmp6errpps_count VNET(icmp6errpps_count)
135 #define V_icmp6errppslim_last VNET(icmp6errppslim_last)
136 #define V_icmp6_nodeinfo VNET(icmp6_nodeinfo)
137
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
141 static const char *icmp6_redirect_diag(struct in6_addr *,
142 struct in6_addr *, struct in6_addr *);
143 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
144 static struct mbuf *ni6_nametodns(const char *, int, int);
145 static int ni6_dnsmatch(const char *, int, const char *, int);
146 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
147 struct ifnet **, struct in6_addr *);
148 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
149 struct ifnet *, int);
150 static int icmp6_notify_error(struct mbuf **, int, int, int);
151
152 /*
153 * Kernel module interface for updating icmp6stat. The argument is an index
154 * into icmp6stat treated as an array of u_quad_t. While this encodes the
155 * general layout of icmp6stat into the caller, it doesn't encode its
156 * location, so that future changes to add, for example, per-CPU stats
157 * support won't cause binary compatibility problems for kernel modules.
158 */
159 void
kmod_icmp6stat_inc(int statnum)160 kmod_icmp6stat_inc(int statnum)
161 {
162
163 counter_u64_add(VNET(icmp6stat)[statnum], 1);
164 }
165
166 static void
icmp6_errcount(int type,int code)167 icmp6_errcount(int type, int code)
168 {
169 switch (type) {
170 case ICMP6_DST_UNREACH:
171 switch (code) {
172 case ICMP6_DST_UNREACH_NOROUTE:
173 ICMP6STAT_INC(icp6s_odst_unreach_noroute);
174 return;
175 case ICMP6_DST_UNREACH_ADMIN:
176 ICMP6STAT_INC(icp6s_odst_unreach_admin);
177 return;
178 case ICMP6_DST_UNREACH_BEYONDSCOPE:
179 ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
180 return;
181 case ICMP6_DST_UNREACH_ADDR:
182 ICMP6STAT_INC(icp6s_odst_unreach_addr);
183 return;
184 case ICMP6_DST_UNREACH_NOPORT:
185 ICMP6STAT_INC(icp6s_odst_unreach_noport);
186 return;
187 }
188 break;
189 case ICMP6_PACKET_TOO_BIG:
190 ICMP6STAT_INC(icp6s_opacket_too_big);
191 return;
192 case ICMP6_TIME_EXCEEDED:
193 switch (code) {
194 case ICMP6_TIME_EXCEED_TRANSIT:
195 ICMP6STAT_INC(icp6s_otime_exceed_transit);
196 return;
197 case ICMP6_TIME_EXCEED_REASSEMBLY:
198 ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
199 return;
200 }
201 break;
202 case ICMP6_PARAM_PROB:
203 switch (code) {
204 case ICMP6_PARAMPROB_HEADER:
205 ICMP6STAT_INC(icp6s_oparamprob_header);
206 return;
207 case ICMP6_PARAMPROB_NEXTHEADER:
208 ICMP6STAT_INC(icp6s_oparamprob_nextheader);
209 return;
210 case ICMP6_PARAMPROB_OPTION:
211 ICMP6STAT_INC(icp6s_oparamprob_option);
212 return;
213 }
214 break;
215 case ND_REDIRECT:
216 ICMP6STAT_INC(icp6s_oredirect);
217 return;
218 }
219 ICMP6STAT_INC(icp6s_ounknown);
220 }
221
222 /*
223 * A wrapper function for icmp6_error() necessary when the erroneous packet
224 * may not contain enough scope zone information.
225 */
226 void
icmp6_error2(struct mbuf * m,int type,int code,int param,struct ifnet * ifp)227 icmp6_error2(struct mbuf *m, int type, int code, int param,
228 struct ifnet *ifp)
229 {
230 struct ip6_hdr *ip6;
231
232 if (ifp == NULL)
233 return;
234
235 if (m->m_len < sizeof(struct ip6_hdr)) {
236 m = m_pullup(m, sizeof(struct ip6_hdr));
237 if (m == NULL) {
238 IP6STAT_INC(ip6s_exthdrtoolong);
239 return;
240 }
241 }
242 ip6 = mtod(m, struct ip6_hdr *);
243
244 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
245 return;
246 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
247 return;
248
249 icmp6_error(m, type, code, param);
250 }
251
252 /*
253 * Generate an error packet of type error in response to bad IP6 packet.
254 */
255 void
icmp6_error(struct mbuf * m,int type,int code,int param)256 icmp6_error(struct mbuf *m, int type, int code, int param)
257 {
258 struct ip6_hdr *oip6, *nip6;
259 struct icmp6_hdr *icmp6;
260 u_int preplen;
261 int off;
262 int nxt;
263
264 ICMP6STAT_INC(icp6s_error);
265
266 /* count per-type-code statistics */
267 icmp6_errcount(type, code);
268
269 #ifdef M_DECRYPTED /*not openbsd*/
270 if (m->m_flags & M_DECRYPTED) {
271 ICMP6STAT_INC(icp6s_canterror);
272 goto freeit;
273 }
274 #endif
275
276 if (m->m_len < sizeof(struct ip6_hdr)) {
277 m = m_pullup(m, sizeof(struct ip6_hdr));
278 if (m == NULL) {
279 IP6STAT_INC(ip6s_exthdrtoolong);
280 return;
281 }
282 }
283 oip6 = mtod(m, struct ip6_hdr *);
284
285 /*
286 * If the destination address of the erroneous packet is a multicast
287 * address, or the packet was sent using link-layer multicast,
288 * we should basically suppress sending an error (RFC 2463, Section
289 * 2.4).
290 * We have two exceptions (the item e.2 in that section):
291 * - the Packet Too Big message can be sent for path MTU discovery.
292 * - the Parameter Problem Message that can be allowed an icmp6 error
293 * in the option type field. This check has been done in
294 * ip6_unknown_opt(), so we can just check the type and code.
295 */
296 if ((m->m_flags & (M_BCAST|M_MCAST) ||
297 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
298 (type != ICMP6_PACKET_TOO_BIG &&
299 (type != ICMP6_PARAM_PROB ||
300 code != ICMP6_PARAMPROB_OPTION)))
301 goto freeit;
302
303 /*
304 * RFC 2463, 2.4 (e.5): source address check.
305 * XXX: the case of anycast source?
306 */
307 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
308 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
309 goto freeit;
310
311 /*
312 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
313 * don't do it.
314 */
315 nxt = -1;
316 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
317 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
318 struct icmp6_hdr *icp;
319
320 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
321 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
322 if (m == NULL) {
323 IP6STAT_INC(ip6s_exthdrtoolong);
324 return;
325 }
326 }
327 oip6 = mtod(m, struct ip6_hdr *);
328 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
329
330 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
331 icp->icmp6_type == ND_REDIRECT) {
332 /*
333 * ICMPv6 error
334 * Special case: for redirect (which is
335 * informational) we must not send icmp6 error.
336 */
337 ICMP6STAT_INC(icp6s_canterror);
338 goto freeit;
339 } else {
340 /* ICMPv6 informational - send the error */
341 }
342 } else {
343 /* non-ICMPv6 - send the error */
344 }
345
346 /* Finally, do rate limitation check. */
347 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
348 ICMP6STAT_INC(icp6s_toofreq);
349 goto freeit;
350 }
351
352 /*
353 * OK, ICMP6 can be generated.
354 */
355
356 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
357 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
358
359 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
360 M_PREPEND(m, preplen, M_NOWAIT); /* FIB is also copied over. */
361 if (m == NULL) {
362 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
363 return;
364 }
365
366 nip6 = mtod(m, struct ip6_hdr *);
367 nip6->ip6_src = oip6->ip6_src;
368 nip6->ip6_dst = oip6->ip6_dst;
369
370 in6_clearscope(&oip6->ip6_src);
371 in6_clearscope(&oip6->ip6_dst);
372
373 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
374 icmp6->icmp6_type = type;
375 icmp6->icmp6_code = code;
376 icmp6->icmp6_pptr = htonl((u_int32_t)param);
377
378 ICMP6STAT_INC(icp6s_outhist[type]);
379 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
380
381 return;
382
383 freeit:
384 /*
385 * If we can't tell whether or not we can generate ICMP6, free it.
386 */
387 m_freem(m);
388 }
389
390 /*
391 * Process a received ICMP6 message.
392 */
393 int
icmp6_input(struct mbuf ** mp,int * offp,int proto)394 icmp6_input(struct mbuf **mp, int *offp, int proto)
395 {
396 struct mbuf *m, *n;
397 struct ifnet *ifp;
398 struct ip6_hdr *ip6, *nip6;
399 struct icmp6_hdr *icmp6, *nicmp6;
400 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
401 int code, error, icmp6len, ip6len, noff, off, sum;
402
403 m = *mp;
404 off = *offp;
405
406 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
407 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
408 if (m == NULL) {
409 IP6STAT_INC(ip6s_exthdrtoolong);
410 *mp = m;
411 return (IPPROTO_DONE);
412 }
413 }
414
415 /*
416 * Locate icmp6 structure in mbuf, and check
417 * that not corrupted and of at least minimum length
418 */
419
420 icmp6len = m->m_pkthdr.len - off;
421 if (icmp6len < sizeof(struct icmp6_hdr)) {
422 ICMP6STAT_INC(icp6s_tooshort);
423 goto freeit;
424 }
425
426 ip6 = mtod(m, struct ip6_hdr *);
427 ifp = m->m_pkthdr.rcvif;
428 /*
429 * Check multicast group membership.
430 * Note: SSM filters are not applied for ICMPv6 traffic.
431 */
432 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
433 struct in6_multi *inm;
434
435 inm = in6m_lookup(ifp, &ip6->ip6_dst);
436 if (inm == NULL) {
437 IP6STAT_INC(ip6s_notmember);
438 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
439 goto freeit;
440 }
441 }
442
443 /* Calculate the checksum. */
444 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
445 code = icmp6->icmp6_code;
446 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
447 nd6log((LOG_ERR,
448 "ICMP6 checksum error(%d|%x) %s\n",
449 icmp6->icmp6_type, sum,
450 ip6_sprintf(ip6bufs, &ip6->ip6_src)));
451 ICMP6STAT_INC(icp6s_checksum);
452 goto freeit;
453 }
454
455 ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
456 icmp6_ifstat_inc(ifp, ifs6_in_msg);
457 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
458 icmp6_ifstat_inc(ifp, ifs6_in_error);
459
460 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
461 switch (icmp6->icmp6_type) {
462 case ICMP6_DST_UNREACH:
463 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
464 switch (code) {
465 case ICMP6_DST_UNREACH_NOROUTE:
466 case ICMP6_DST_UNREACH_ADDR: /* PRC_HOSTDEAD is a DOS */
467 code = PRC_UNREACH_NET;
468 break;
469 case ICMP6_DST_UNREACH_ADMIN:
470 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
471 code = PRC_UNREACH_ADMIN_PROHIB;
472 break;
473 case ICMP6_DST_UNREACH_BEYONDSCOPE:
474 /* I mean "source address was incorrect." */
475 code = PRC_PARAMPROB;
476 break;
477 case ICMP6_DST_UNREACH_NOPORT:
478 code = PRC_UNREACH_PORT;
479 break;
480 default:
481 goto badcode;
482 }
483 goto deliver;
484 break;
485
486 case ICMP6_PACKET_TOO_BIG:
487 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
488
489 /* validation is made in icmp6_mtudisc_update */
490
491 code = PRC_MSGSIZE;
492
493 /*
494 * Updating the path MTU will be done after examining
495 * intermediate extension headers.
496 */
497 goto deliver;
498 break;
499
500 case ICMP6_TIME_EXCEEDED:
501 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
502 switch (code) {
503 case ICMP6_TIME_EXCEED_TRANSIT:
504 code = PRC_TIMXCEED_INTRANS;
505 break;
506 case ICMP6_TIME_EXCEED_REASSEMBLY:
507 code = PRC_TIMXCEED_REASS;
508 break;
509 default:
510 goto badcode;
511 }
512 goto deliver;
513 break;
514
515 case ICMP6_PARAM_PROB:
516 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
517 switch (code) {
518 case ICMP6_PARAMPROB_NEXTHEADER:
519 code = PRC_UNREACH_PROTOCOL;
520 break;
521 case ICMP6_PARAMPROB_HEADER:
522 case ICMP6_PARAMPROB_OPTION:
523 code = PRC_PARAMPROB;
524 break;
525 default:
526 goto badcode;
527 }
528 goto deliver;
529 break;
530
531 case ICMP6_ECHO_REQUEST:
532 icmp6_ifstat_inc(ifp, ifs6_in_echo);
533 if (code != 0)
534 goto badcode;
535 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
536 /* Give up remote */
537 break;
538 }
539 if (!M_WRITABLE(n)
540 || n->m_len < off + sizeof(struct icmp6_hdr)) {
541 struct mbuf *n0 = n;
542 int n0len;
543
544 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
545 n = m_gethdr(M_NOWAIT, n0->m_type);
546 if (n == NULL) {
547 /* Give up remote */
548 m_freem(n0);
549 break;
550 }
551
552 m_move_pkthdr(n, n0); /* FIB copied. */
553 n0len = n0->m_pkthdr.len; /* save for use below */
554 /*
555 * Copy IPv6 and ICMPv6 only.
556 */
557 nip6 = mtod(n, struct ip6_hdr *);
558 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
559 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
560 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
561 noff = sizeof(struct ip6_hdr);
562 /* new mbuf contains only ipv6+icmpv6 headers */
563 n->m_len = noff + sizeof(struct icmp6_hdr);
564 /*
565 * Adjust mbuf. ip6_plen will be adjusted in
566 * ip6_output().
567 */
568 m_adj(n0, off + sizeof(struct icmp6_hdr));
569 /* recalculate complete packet size */
570 n->m_pkthdr.len = n0len + (noff - off);
571 n->m_next = n0;
572 } else {
573 if (n->m_len < off + sizeof(*nicmp6)) {
574 n = m_pullup(n, off + sizeof(*nicmp6));
575 if (n == NULL) {
576 IP6STAT_INC(ip6s_exthdrtoolong);
577 break;
578 }
579 }
580 nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
581 noff = off;
582 }
583 if (n) {
584 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
585 nicmp6->icmp6_code = 0;
586 ICMP6STAT_INC(icp6s_reflect);
587 ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
588 icmp6_reflect(n, noff);
589 }
590 break;
591
592 case ICMP6_ECHO_REPLY:
593 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
594 if (code != 0)
595 goto badcode;
596 break;
597
598 case MLD_LISTENER_QUERY:
599 case MLD_LISTENER_REPORT:
600 case MLD_LISTENER_DONE:
601 case MLDV2_LISTENER_REPORT:
602 /*
603 * Drop MLD traffic which is not link-local, has a hop limit
604 * of greater than 1 hop, or which does not have the
605 * IPv6 HBH Router Alert option.
606 * As IPv6 HBH options are stripped in ip6_input() we must
607 * check an mbuf header flag.
608 * XXX Should we also sanity check that these messages
609 * were directed to a link-local multicast prefix?
610 */
611 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
612 goto freeit;
613 if (mld_input(&m, off, icmp6len) != 0) {
614 *mp = NULL;
615 return (IPPROTO_DONE);
616 }
617 /* m stays. */
618 break;
619
620 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
621 {
622 enum { WRU, FQDN } mode;
623 struct prison *pr;
624
625 if (!V_icmp6_nodeinfo)
626 break;
627
628 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
629 mode = WRU;
630 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
631 mode = FQDN;
632 else
633 goto badlen;
634
635 pr = NULL;
636 sx_slock(&allprison_lock);
637 TAILQ_FOREACH(pr, &allprison, pr_list)
638 if (pr->pr_vnet == ifp->if_vnet)
639 break;
640 sx_sunlock(&allprison_lock);
641 if (pr == NULL)
642 pr = curthread->td_ucred->cr_prison;
643 if (mode == FQDN) {
644 if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
645 m = m_pullup(m, off +
646 sizeof(struct icmp6_nodeinfo));
647 if (m == NULL) {
648 IP6STAT_INC(ip6s_exthdrtoolong);
649 *mp = m;
650 return (IPPROTO_DONE);
651 }
652 }
653 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
654 if (n)
655 n = ni6_input(n, off, pr);
656 /* XXX meaningless if n == NULL */
657 noff = sizeof(struct ip6_hdr);
658 } else {
659 u_char *p;
660 int maxhlen, hlen;
661
662 /*
663 * XXX: this combination of flags is pointless,
664 * but should we keep this for compatibility?
665 */
666 if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
667 ICMP6_NODEINFO_TMPADDROK)) !=
668 (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
669 break;
670
671 if (code != 0)
672 goto badcode;
673
674 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
675 n = m_gethdr(M_NOWAIT, m->m_type);
676 if (n == NULL) {
677 /* Give up remote */
678 break;
679 }
680 if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
681 /*
682 * Previous code did a blind M_COPY_PKTHDR
683 * and said "just for rcvif". If true, then
684 * we could tolerate the dup failing (due to
685 * the deep copy of the tag chain). For now
686 * be conservative and just fail.
687 */
688 m_free(n);
689 n = NULL;
690 break;
691 }
692 /*
693 * Copy IPv6 and ICMPv6 only.
694 */
695 nip6 = mtod(n, struct ip6_hdr *);
696 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
697 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
698 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
699 p = (u_char *)(nicmp6 + 1);
700 bzero(p, 4);
701
702 maxhlen = M_TRAILINGSPACE(n) -
703 (sizeof(*nip6) + sizeof(*nicmp6) + 4);
704 mtx_lock(&pr->pr_mtx);
705 hlen = strlen(pr->pr_hostname);
706 if (maxhlen > hlen)
707 maxhlen = hlen;
708 /* meaningless TTL */
709 bcopy(pr->pr_hostname, p + 4, maxhlen);
710 mtx_unlock(&pr->pr_mtx);
711 noff = sizeof(struct ip6_hdr);
712 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
713 sizeof(struct icmp6_hdr) + 4 + maxhlen;
714 nicmp6->icmp6_type = ICMP6_WRUREPLY;
715 nicmp6->icmp6_code = 0;
716 }
717 if (n) {
718 ICMP6STAT_INC(icp6s_reflect);
719 ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
720 icmp6_reflect(n, noff);
721 }
722 break;
723 }
724
725 case ICMP6_WRUREPLY:
726 if (code != 0)
727 goto badcode;
728 break;
729
730 case ND_ROUTER_SOLICIT:
731 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
732 if (code != 0)
733 goto badcode;
734 if (icmp6len < sizeof(struct nd_router_solicit))
735 goto badlen;
736 if (send_sendso_input_hook != NULL) {
737 if (m->m_len < off + icmp6len) {
738 m = m_pullup(m, off + icmp6len);
739 if (m == NULL) {
740 IP6STAT_INC(ip6s_exthdrtoolong);
741 *mp = NULL;
742 return (IPPROTO_DONE);
743 }
744 }
745 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
746 if (error == 0) {
747 m = NULL;
748 goto freeit;
749 }
750 }
751 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
752 nd6_rs_input(m, off, icmp6len);
753 m = n;
754 if (m == NULL)
755 goto freeit;
756 break;
757
758 case ND_ROUTER_ADVERT:
759 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
760 if (code != 0)
761 goto badcode;
762 if (icmp6len < sizeof(struct nd_router_advert))
763 goto badlen;
764 if (send_sendso_input_hook != NULL) {
765 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
766 if (error == 0) {
767 m = NULL;
768 goto freeit;
769 }
770 }
771 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
772 nd6_ra_input(m, off, icmp6len);
773 m = n;
774 if (m == NULL)
775 goto freeit;
776 break;
777
778 case ND_NEIGHBOR_SOLICIT:
779 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
780 if (code != 0)
781 goto badcode;
782 if (icmp6len < sizeof(struct nd_neighbor_solicit))
783 goto badlen;
784 if (send_sendso_input_hook != NULL) {
785 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
786 if (error == 0) {
787 m = NULL;
788 goto freeit;
789 }
790 }
791 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
792 nd6_ns_input(m, off, icmp6len);
793 m = n;
794 if (m == NULL)
795 goto freeit;
796 break;
797
798 case ND_NEIGHBOR_ADVERT:
799 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
800 if (code != 0)
801 goto badcode;
802 if (icmp6len < sizeof(struct nd_neighbor_advert))
803 goto badlen;
804 if (send_sendso_input_hook != NULL) {
805 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
806 if (error == 0) {
807 m = NULL;
808 goto freeit;
809 }
810 }
811 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
812 nd6_na_input(m, off, icmp6len);
813 m = n;
814 if (m == NULL)
815 goto freeit;
816 break;
817
818 case ND_REDIRECT:
819 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
820 if (code != 0)
821 goto badcode;
822 if (icmp6len < sizeof(struct nd_redirect))
823 goto badlen;
824 if (send_sendso_input_hook != NULL) {
825 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
826 if (error == 0) {
827 m = NULL;
828 goto freeit;
829 }
830 }
831 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
832 icmp6_redirect_input(m, off);
833 m = n;
834 if (m == NULL)
835 goto freeit;
836 break;
837
838 case ICMP6_ROUTER_RENUMBERING:
839 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
840 code != ICMP6_ROUTER_RENUMBERING_RESULT)
841 goto badcode;
842 if (icmp6len < sizeof(struct icmp6_router_renum))
843 goto badlen;
844 break;
845
846 default:
847 nd6log((LOG_DEBUG,
848 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
849 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
850 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
851 ifp ? ifp->if_index : 0));
852 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
853 /* ICMPv6 error: MUST deliver it by spec... */
854 code = PRC_NCMDS;
855 /* deliver */
856 } else {
857 /* ICMPv6 informational: MUST not deliver */
858 break;
859 }
860 deliver:
861 if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
862 /* In this case, m should've been freed. */
863 *mp = NULL;
864 return (IPPROTO_DONE);
865 }
866 break;
867
868 badcode:
869 ICMP6STAT_INC(icp6s_badcode);
870 break;
871
872 badlen:
873 ICMP6STAT_INC(icp6s_badlen);
874 break;
875 }
876
877 /* deliver the packet to appropriate sockets */
878 icmp6_rip6_input(&m, *offp);
879
880 *mp = m;
881 return (IPPROTO_DONE);
882
883 freeit:
884 m_freem(m);
885 *mp = NULL;
886 return (IPPROTO_DONE);
887 }
888
889 static int
icmp6_notify_error(struct mbuf ** mp,int off,int icmp6len,int code)890 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
891 {
892 struct mbuf *m;
893 struct icmp6_hdr *icmp6;
894 struct ip6_hdr *eip6;
895 u_int32_t notifymtu;
896 struct sockaddr_in6 icmp6src, icmp6dst;
897
898 m = *mp;
899
900 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
901 ICMP6STAT_INC(icp6s_tooshort);
902 goto freeit;
903 }
904
905 if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
906 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
907 if (m == NULL) {
908 IP6STAT_INC(ip6s_exthdrtoolong);
909 *mp = m;
910 return (-1);
911 }
912 }
913 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
914 eip6 = (struct ip6_hdr *)(icmp6 + 1);
915 bzero(&icmp6dst, sizeof(icmp6dst));
916
917 /* Detect the upper level protocol */
918 {
919 void (*ctlfunc)(int, struct sockaddr *, void *);
920 u_int8_t nxt = eip6->ip6_nxt;
921 int eoff = off + sizeof(struct icmp6_hdr) +
922 sizeof(struct ip6_hdr);
923 struct ip6ctlparam ip6cp;
924 int icmp6type = icmp6->icmp6_type;
925 struct ip6_frag *fh;
926 struct ip6_rthdr *rth;
927 struct ip6_rthdr0 *rth0;
928 int rthlen;
929
930 while (1) { /* XXX: should avoid infinite loop explicitly? */
931 struct ip6_ext *eh;
932
933 switch (nxt) {
934 case IPPROTO_HOPOPTS:
935 case IPPROTO_DSTOPTS:
936 case IPPROTO_AH:
937 if (m->m_len < eoff + sizeof(struct ip6_ext)) {
938 m = m_pullup(m, eoff +
939 sizeof(struct ip6_ext));
940 if (m == NULL) {
941 IP6STAT_INC(ip6s_exthdrtoolong);
942 *mp = m;
943 return (-1);
944 }
945 }
946 eh = (struct ip6_ext *)
947 (mtod(m, caddr_t) + eoff);
948 if (nxt == IPPROTO_AH)
949 eoff += (eh->ip6e_len + 2) << 2;
950 else
951 eoff += (eh->ip6e_len + 1) << 3;
952 nxt = eh->ip6e_nxt;
953 break;
954 case IPPROTO_ROUTING:
955 /*
956 * When the erroneous packet contains a
957 * routing header, we should examine the
958 * header to determine the final destination.
959 * Otherwise, we can't properly update
960 * information that depends on the final
961 * destination (e.g. path MTU).
962 */
963 if (m->m_len < eoff + sizeof(*rth)) {
964 m = m_pullup(m, eoff + sizeof(*rth));
965 if (m == NULL) {
966 IP6STAT_INC(ip6s_exthdrtoolong);
967 *mp = m;
968 return (-1);
969 }
970 }
971 rth = (struct ip6_rthdr *)
972 (mtod(m, caddr_t) + eoff);
973 rthlen = (rth->ip6r_len + 1) << 3;
974 /*
975 * XXX: currently there is no
976 * officially defined type other
977 * than type-0.
978 * Note that if the segment left field
979 * is 0, all intermediate hops must
980 * have been passed.
981 */
982 if (rth->ip6r_segleft &&
983 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
984 int hops;
985
986 if (m->m_len < eoff + rthlen) {
987 m = m_pullup(m, eoff + rthlen);
988 if (m == NULL) {
989 IP6STAT_INC(
990 ip6s_exthdrtoolong);
991 *mp = m;
992 return (-1);
993 }
994 }
995 rth0 = (struct ip6_rthdr0 *)
996 (mtod(m, caddr_t) + eoff);
997
998 /* just ignore a bogus header */
999 if ((rth0->ip6r0_len % 2) == 0 &&
1000 (hops = rth0->ip6r0_len/2))
1001 icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1002 }
1003 eoff += rthlen;
1004 nxt = rth->ip6r_nxt;
1005 break;
1006 case IPPROTO_FRAGMENT:
1007 if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1008 m = m_pullup(m, eoff +
1009 sizeof(struct ip6_frag));
1010 if (m == NULL) {
1011 IP6STAT_INC(ip6s_exthdrtoolong);
1012 *mp = m;
1013 return (-1);
1014 }
1015 }
1016 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1017 eoff);
1018 /*
1019 * Data after a fragment header is meaningless
1020 * unless it is the first fragment, but
1021 * we'll go to the notify label for path MTU
1022 * discovery.
1023 */
1024 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1025 goto notify;
1026
1027 eoff += sizeof(struct ip6_frag);
1028 nxt = fh->ip6f_nxt;
1029 break;
1030 default:
1031 /*
1032 * This case includes ESP and the No Next
1033 * Header. In such cases going to the notify
1034 * label does not have any meaning
1035 * (i.e. ctlfunc will be NULL), but we go
1036 * anyway since we might have to update
1037 * path MTU information.
1038 */
1039 goto notify;
1040 }
1041 }
1042 notify:
1043 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1044
1045 /*
1046 * retrieve parameters from the inner IPv6 header, and convert
1047 * them into sockaddr structures.
1048 * XXX: there is no guarantee that the source or destination
1049 * addresses of the inner packet are in the same scope as
1050 * the addresses of the icmp packet. But there is no other
1051 * way to determine the zone.
1052 */
1053 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1054
1055 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1056 icmp6dst.sin6_family = AF_INET6;
1057 if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1058 icmp6dst.sin6_addr = eip6->ip6_dst;
1059 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1060 goto freeit;
1061 bzero(&icmp6src, sizeof(icmp6src));
1062 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1063 icmp6src.sin6_family = AF_INET6;
1064 icmp6src.sin6_addr = eip6->ip6_src;
1065 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1066 goto freeit;
1067 icmp6src.sin6_flowinfo =
1068 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1069
1070 ip6cp.ip6c_m = m;
1071 ip6cp.ip6c_icmp6 = icmp6;
1072 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1073 ip6cp.ip6c_off = eoff;
1074 ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1075 ip6cp.ip6c_src = &icmp6src;
1076 ip6cp.ip6c_nxt = nxt;
1077
1078 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1079 notifymtu = ntohl(icmp6->icmp6_mtu);
1080 ip6cp.ip6c_cmdarg = (void *)¬ifymtu;
1081 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/
1082 }
1083
1084 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1085 (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1086 if (ctlfunc) {
1087 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1088 &ip6cp);
1089 }
1090 }
1091 *mp = m;
1092 return (0);
1093
1094 freeit:
1095 m_freem(m);
1096 *mp = NULL;
1097 return (-1);
1098 }
1099
1100 void
icmp6_mtudisc_update(struct ip6ctlparam * ip6cp,int validated)1101 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1102 {
1103 struct in6_addr *dst = ip6cp->ip6c_finaldst;
1104 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1105 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1106 u_int mtu = ntohl(icmp6->icmp6_mtu);
1107 struct in_conninfo inc;
1108
1109 #if 0
1110 /*
1111 * RFC2460 section 5, last paragraph.
1112 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1113 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1114 * due to packet translator in the middle.
1115 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1116 * special handling.
1117 */
1118 if (mtu < IPV6_MMTU)
1119 return;
1120 #endif
1121
1122 /*
1123 * we reject ICMPv6 too big with abnormally small value.
1124 * XXX what is the good definition of "abnormally small"?
1125 */
1126 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1127 return;
1128
1129 if (!validated)
1130 return;
1131
1132 /*
1133 * In case the suggested mtu is less than IPV6_MMTU, we
1134 * only need to remember that it was for above mentioned
1135 * "alwaysfrag" case.
1136 * Try to be as close to the spec as possible.
1137 */
1138 if (mtu < IPV6_MMTU)
1139 mtu = IPV6_MMTU - 8;
1140
1141 bzero(&inc, sizeof(inc));
1142 inc.inc_fibnum = M_GETFIB(m);
1143 inc.inc_flags |= INC_ISIPV6;
1144 inc.inc6_faddr = *dst;
1145 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1146 return;
1147
1148 if (mtu < tcp_maxmtu6(&inc, NULL)) {
1149 tcp_hc_updatemtu(&inc, mtu);
1150 ICMP6STAT_INC(icp6s_pmtuchg);
1151 }
1152 }
1153
1154 /*
1155 * Process a Node Information Query packet, based on
1156 * draft-ietf-ipngwg-icmp-name-lookups-07.
1157 *
1158 * Spec incompatibilities:
1159 * - IPv6 Subject address handling
1160 * - IPv4 Subject address handling support missing
1161 * - Proxy reply (answer even if it's not for me)
1162 * - joins NI group address at in6_ifattach() time only, does not cope
1163 * with hostname changes by sethostname(3)
1164 */
1165 static struct mbuf *
ni6_input(struct mbuf * m,int off,struct prison * pr)1166 ni6_input(struct mbuf *m, int off, struct prison *pr)
1167 {
1168 struct icmp6_nodeinfo *ni6, *nni6;
1169 struct mbuf *n = NULL;
1170 u_int16_t qtype;
1171 int subjlen;
1172 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1173 struct ni_reply_fqdn *fqdn;
1174 int addrs; /* for NI_QTYPE_NODEADDR */
1175 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1176 struct in6_addr in6_subj; /* subject address */
1177 struct ip6_hdr *ip6;
1178 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1179 char *subj = NULL;
1180 struct in6_ifaddr *ia6 = NULL;
1181
1182 ip6 = mtod(m, struct ip6_hdr *);
1183 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1184
1185 /*
1186 * Validate IPv6 source address.
1187 * The default configuration MUST be to refuse answering queries from
1188 * global-scope addresses according to RFC4602.
1189 * Notes:
1190 * - it's not very clear what "refuse" means; this implementation
1191 * simply drops it.
1192 * - it's not very easy to identify global-scope (unicast) addresses
1193 * since there are many prefixes for them. It should be safer
1194 * and in practice sufficient to check "all" but loopback and
1195 * link-local (note that site-local unicast was deprecated and
1196 * ULA is defined as global scope-wise)
1197 */
1198 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1199 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1200 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1201 goto bad;
1202
1203 /*
1204 * Validate IPv6 destination address.
1205 *
1206 * The Responder must discard the Query without further processing
1207 * unless it is one of the Responder's unicast or anycast addresses, or
1208 * a link-local scope multicast address which the Responder has joined.
1209 * [RFC4602, Section 5.]
1210 */
1211 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1212 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1213 goto bad;
1214 /* else it's a link-local multicast, fine */
1215 } else { /* unicast or anycast */
1216 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1217 if (ia6 == NULL)
1218 goto bad; /* XXX impossible */
1219
1220 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1221 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1222 ifa_free(&ia6->ia_ifa);
1223 nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1224 "a temporary address in %s:%d",
1225 __FILE__, __LINE__));
1226 goto bad;
1227 }
1228 ifa_free(&ia6->ia_ifa);
1229 }
1230
1231 /* validate query Subject field. */
1232 qtype = ntohs(ni6->ni_qtype);
1233 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1234 switch (qtype) {
1235 case NI_QTYPE_NOOP:
1236 case NI_QTYPE_SUPTYPES:
1237 /* 07 draft */
1238 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1239 break;
1240 /* FALLTHROUGH */
1241 case NI_QTYPE_FQDN:
1242 case NI_QTYPE_NODEADDR:
1243 case NI_QTYPE_IPV4ADDR:
1244 switch (ni6->ni_code) {
1245 case ICMP6_NI_SUBJ_IPV6:
1246 #if ICMP6_NI_SUBJ_IPV6 != 0
1247 case 0:
1248 #endif
1249 /*
1250 * backward compatibility - try to accept 03 draft
1251 * format, where no Subject is present.
1252 */
1253 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1254 subjlen == 0) {
1255 oldfqdn++;
1256 break;
1257 }
1258 #if ICMP6_NI_SUBJ_IPV6 != 0
1259 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1260 goto bad;
1261 #endif
1262
1263 if (subjlen != sizeof(struct in6_addr))
1264 goto bad;
1265
1266 /*
1267 * Validate Subject address.
1268 *
1269 * Not sure what exactly "address belongs to the node"
1270 * means in the spec, is it just unicast, or what?
1271 *
1272 * At this moment we consider Subject address as
1273 * "belong to the node" if the Subject address equals
1274 * to the IPv6 destination address; validation for
1275 * IPv6 destination address should have done enough
1276 * check for us.
1277 *
1278 * We do not do proxy at this moment.
1279 */
1280 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1281 subjlen, (caddr_t)&in6_subj);
1282 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1283 goto bad;
1284
1285 subj = (char *)&in6_subj;
1286 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1287 break;
1288
1289 /*
1290 * XXX if we are to allow other cases, we should really
1291 * be careful about scope here.
1292 * basically, we should disallow queries toward IPv6
1293 * destination X with subject Y,
1294 * if scope(X) > scope(Y).
1295 * if we allow scope(X) > scope(Y), it will result in
1296 * information leakage across scope boundary.
1297 */
1298 goto bad;
1299
1300 case ICMP6_NI_SUBJ_FQDN:
1301 /*
1302 * Validate Subject name with gethostname(3).
1303 *
1304 * The behavior may need some debate, since:
1305 * - we are not sure if the node has FQDN as
1306 * hostname (returned by gethostname(3)).
1307 * - the code does wildcard match for truncated names.
1308 * however, we are not sure if we want to perform
1309 * wildcard match, if gethostname(3) side has
1310 * truncated hostname.
1311 */
1312 mtx_lock(&pr->pr_mtx);
1313 n = ni6_nametodns(pr->pr_hostname,
1314 strlen(pr->pr_hostname), 0);
1315 mtx_unlock(&pr->pr_mtx);
1316 if (!n || n->m_next || n->m_len == 0)
1317 goto bad;
1318 if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1319 subjlen) {
1320 m = m_pullup(m, off +
1321 sizeof(struct icmp6_nodeinfo) + subjlen);
1322 if (m == NULL) {
1323 IP6STAT_INC(ip6s_exthdrtoolong);
1324 goto bad;
1325 }
1326 }
1327 /* ip6 possibly invalid but not used after. */
1328 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1329 subj = (char *)(mtod(m, caddr_t) + off +
1330 sizeof(struct icmp6_nodeinfo));
1331 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1332 n->m_len)) {
1333 goto bad;
1334 }
1335 m_freem(n);
1336 n = NULL;
1337 break;
1338
1339 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1340 default:
1341 goto bad;
1342 }
1343 break;
1344 }
1345
1346 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1347 switch (qtype) {
1348 case NI_QTYPE_FQDN:
1349 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1350 goto bad;
1351 break;
1352 case NI_QTYPE_NODEADDR:
1353 case NI_QTYPE_IPV4ADDR:
1354 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1355 goto bad;
1356 break;
1357 }
1358
1359 /* guess reply length */
1360 switch (qtype) {
1361 case NI_QTYPE_NOOP:
1362 break; /* no reply data */
1363 case NI_QTYPE_SUPTYPES:
1364 replylen += sizeof(u_int32_t);
1365 break;
1366 case NI_QTYPE_FQDN:
1367 /* XXX will append an mbuf */
1368 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1369 break;
1370 case NI_QTYPE_NODEADDR:
1371 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1372 if ((replylen += addrs * (sizeof(struct in6_addr) +
1373 sizeof(u_int32_t))) > MCLBYTES)
1374 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1375 break;
1376 case NI_QTYPE_IPV4ADDR:
1377 /* unsupported - should respond with unknown Qtype? */
1378 break;
1379 default:
1380 /*
1381 * XXX: We must return a reply with the ICMP6 code
1382 * `unknown Qtype' in this case. However we regard the case
1383 * as an FQDN query for backward compatibility.
1384 * Older versions set a random value to this field,
1385 * so it rarely varies in the defined qtypes.
1386 * But the mechanism is not reliable...
1387 * maybe we should obsolete older versions.
1388 */
1389 qtype = NI_QTYPE_FQDN;
1390 /* XXX will append an mbuf */
1391 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1392 oldfqdn++;
1393 break;
1394 }
1395
1396 /* Allocate an mbuf to reply. */
1397 if (replylen > MCLBYTES) {
1398 /*
1399 * XXX: should we try to allocate more? But MCLBYTES
1400 * is probably much larger than IPV6_MMTU...
1401 */
1402 goto bad;
1403 }
1404 if (replylen > MHLEN)
1405 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1406 else
1407 n = m_gethdr(M_NOWAIT, m->m_type);
1408 if (n == NULL) {
1409 m_freem(m);
1410 return (NULL);
1411 }
1412 m_move_pkthdr(n, m); /* just for recvif and FIB */
1413 n->m_pkthdr.len = n->m_len = replylen;
1414
1415 /* copy mbuf header and IPv6 + Node Information base headers */
1416 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1417 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1418 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1419
1420 /* qtype dependent procedure */
1421 switch (qtype) {
1422 case NI_QTYPE_NOOP:
1423 nni6->ni_code = ICMP6_NI_SUCCESS;
1424 nni6->ni_flags = 0;
1425 break;
1426 case NI_QTYPE_SUPTYPES:
1427 {
1428 u_int32_t v;
1429 nni6->ni_code = ICMP6_NI_SUCCESS;
1430 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1431 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1432 v = (u_int32_t)htonl(0x0000000f);
1433 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1434 break;
1435 }
1436 case NI_QTYPE_FQDN:
1437 nni6->ni_code = ICMP6_NI_SUCCESS;
1438 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1439 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1440 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1441 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1442 /*
1443 * XXX do we really have FQDN in hostname?
1444 */
1445 mtx_lock(&pr->pr_mtx);
1446 n->m_next = ni6_nametodns(pr->pr_hostname,
1447 strlen(pr->pr_hostname), oldfqdn);
1448 mtx_unlock(&pr->pr_mtx);
1449 if (n->m_next == NULL)
1450 goto bad;
1451 /* XXX we assume that n->m_next is not a chain */
1452 if (n->m_next->m_next != NULL)
1453 goto bad;
1454 n->m_pkthdr.len += n->m_next->m_len;
1455 break;
1456 case NI_QTYPE_NODEADDR:
1457 {
1458 int lenlim, copied;
1459
1460 nni6->ni_code = ICMP6_NI_SUCCESS;
1461 n->m_pkthdr.len = n->m_len =
1462 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1463 lenlim = M_TRAILINGSPACE(n);
1464 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1465 /* XXX: reset mbuf length */
1466 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1467 sizeof(struct icmp6_nodeinfo) + copied;
1468 break;
1469 }
1470 default:
1471 break; /* XXX impossible! */
1472 }
1473
1474 nni6->ni_type = ICMP6_NI_REPLY;
1475 m_freem(m);
1476 return (n);
1477
1478 bad:
1479 m_freem(m);
1480 if (n)
1481 m_freem(n);
1482 return (NULL);
1483 }
1484
1485 /*
1486 * make a mbuf with DNS-encoded string. no compression support.
1487 *
1488 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1489 * treated as truncated name (two \0 at the end). this is a wild guess.
1490 *
1491 * old - return pascal string if non-zero
1492 */
1493 static struct mbuf *
ni6_nametodns(const char * name,int namelen,int old)1494 ni6_nametodns(const char *name, int namelen, int old)
1495 {
1496 struct mbuf *m;
1497 char *cp, *ep;
1498 const char *p, *q;
1499 int i, len, nterm;
1500
1501 if (old)
1502 len = namelen + 1;
1503 else
1504 len = MCLBYTES;
1505
1506 /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1507 if (len > MLEN)
1508 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1509 else
1510 m = m_get(M_NOWAIT, MT_DATA);
1511 if (m == NULL)
1512 goto fail;
1513
1514 if (old) {
1515 m->m_len = len;
1516 *mtod(m, char *) = namelen;
1517 bcopy(name, mtod(m, char *) + 1, namelen);
1518 return m;
1519 } else {
1520 m->m_len = 0;
1521 cp = mtod(m, char *);
1522 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1523
1524 /* if not certain about my name, return empty buffer */
1525 if (namelen == 0)
1526 return m;
1527
1528 /*
1529 * guess if it looks like shortened hostname, or FQDN.
1530 * shortened hostname needs two trailing "\0".
1531 */
1532 i = 0;
1533 for (p = name; p < name + namelen; p++) {
1534 if (*p && *p == '.')
1535 i++;
1536 }
1537 if (i < 2)
1538 nterm = 2;
1539 else
1540 nterm = 1;
1541
1542 p = name;
1543 while (cp < ep && p < name + namelen) {
1544 i = 0;
1545 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1546 i++;
1547 /* result does not fit into mbuf */
1548 if (cp + i + 1 >= ep)
1549 goto fail;
1550 /*
1551 * DNS label length restriction, RFC1035 page 8.
1552 * "i == 0" case is included here to avoid returning
1553 * 0-length label on "foo..bar".
1554 */
1555 if (i <= 0 || i >= 64)
1556 goto fail;
1557 *cp++ = i;
1558 bcopy(p, cp, i);
1559 cp += i;
1560 p = q;
1561 if (p < name + namelen && *p == '.')
1562 p++;
1563 }
1564 /* termination */
1565 if (cp + nterm >= ep)
1566 goto fail;
1567 while (nterm-- > 0)
1568 *cp++ = '\0';
1569 m->m_len = cp - mtod(m, char *);
1570 return m;
1571 }
1572
1573 panic("should not reach here");
1574 /* NOTREACHED */
1575
1576 fail:
1577 if (m)
1578 m_freem(m);
1579 return NULL;
1580 }
1581
1582 /*
1583 * check if two DNS-encoded string matches. takes care of truncated
1584 * form (with \0\0 at the end). no compression support.
1585 * XXX upper/lowercase match (see RFC2065)
1586 */
1587 static int
ni6_dnsmatch(const char * a,int alen,const char * b,int blen)1588 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1589 {
1590 const char *a0, *b0;
1591 int l;
1592
1593 /* simplest case - need validation? */
1594 if (alen == blen && bcmp(a, b, alen) == 0)
1595 return 1;
1596
1597 a0 = a;
1598 b0 = b;
1599
1600 /* termination is mandatory */
1601 if (alen < 2 || blen < 2)
1602 return 0;
1603 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1604 return 0;
1605 alen--;
1606 blen--;
1607
1608 while (a - a0 < alen && b - b0 < blen) {
1609 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1610 return 0;
1611
1612 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1613 return 0;
1614 /* we don't support compression yet */
1615 if (a[0] >= 64 || b[0] >= 64)
1616 return 0;
1617
1618 /* truncated case */
1619 if (a[0] == 0 && a - a0 == alen - 1)
1620 return 1;
1621 if (b[0] == 0 && b - b0 == blen - 1)
1622 return 1;
1623 if (a[0] == 0 || b[0] == 0)
1624 return 0;
1625
1626 if (a[0] != b[0])
1627 return 0;
1628 l = a[0];
1629 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1630 return 0;
1631 if (bcmp(a + 1, b + 1, l) != 0)
1632 return 0;
1633
1634 a += 1 + l;
1635 b += 1 + l;
1636 }
1637
1638 if (a - a0 == alen && b - b0 == blen)
1639 return 1;
1640 else
1641 return 0;
1642 }
1643
1644 /*
1645 * calculate the number of addresses to be returned in the node info reply.
1646 */
1647 static int
ni6_addrs(struct icmp6_nodeinfo * ni6,struct mbuf * m,struct ifnet ** ifpp,struct in6_addr * subj)1648 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1649 struct in6_addr *subj)
1650 {
1651 struct ifnet *ifp;
1652 struct in6_ifaddr *ifa6;
1653 struct ifaddr *ifa;
1654 int addrs = 0, addrsofif, iffound = 0;
1655 int niflags = ni6->ni_flags;
1656
1657 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1658 switch (ni6->ni_code) {
1659 case ICMP6_NI_SUBJ_IPV6:
1660 if (subj == NULL) /* must be impossible... */
1661 return (0);
1662 break;
1663 default:
1664 /*
1665 * XXX: we only support IPv6 subject address for
1666 * this Qtype.
1667 */
1668 return (0);
1669 }
1670 }
1671
1672 IFNET_RLOCK_NOSLEEP();
1673 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1674 addrsofif = 0;
1675 IF_ADDR_RLOCK(ifp);
1676 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1677 if (ifa->ifa_addr->sa_family != AF_INET6)
1678 continue;
1679 ifa6 = (struct in6_ifaddr *)ifa;
1680
1681 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1682 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1683 iffound = 1;
1684
1685 /*
1686 * IPv4-mapped addresses can only be returned by a
1687 * Node Information proxy, since they represent
1688 * addresses of IPv4-only nodes, which perforce do
1689 * not implement this protocol.
1690 * [icmp-name-lookups-07, Section 5.4]
1691 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1692 * this function at this moment.
1693 */
1694
1695 /* What do we have to do about ::1? */
1696 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1697 case IPV6_ADDR_SCOPE_LINKLOCAL:
1698 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1699 continue;
1700 break;
1701 case IPV6_ADDR_SCOPE_SITELOCAL:
1702 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1703 continue;
1704 break;
1705 case IPV6_ADDR_SCOPE_GLOBAL:
1706 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1707 continue;
1708 break;
1709 default:
1710 continue;
1711 }
1712
1713 /*
1714 * check if anycast is okay.
1715 * XXX: just experimental. not in the spec.
1716 */
1717 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1718 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1719 continue; /* we need only unicast addresses */
1720 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1721 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1722 continue;
1723 }
1724 addrsofif++; /* count the address */
1725 }
1726 IF_ADDR_RUNLOCK(ifp);
1727 if (iffound) {
1728 *ifpp = ifp;
1729 IFNET_RUNLOCK_NOSLEEP();
1730 return (addrsofif);
1731 }
1732
1733 addrs += addrsofif;
1734 }
1735 IFNET_RUNLOCK_NOSLEEP();
1736
1737 return (addrs);
1738 }
1739
1740 static int
ni6_store_addrs(struct icmp6_nodeinfo * ni6,struct icmp6_nodeinfo * nni6,struct ifnet * ifp0,int resid)1741 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1742 struct ifnet *ifp0, int resid)
1743 {
1744 struct ifnet *ifp;
1745 struct in6_ifaddr *ifa6;
1746 struct ifaddr *ifa;
1747 struct ifnet *ifp_dep = NULL;
1748 int copied = 0, allow_deprecated = 0;
1749 u_char *cp = (u_char *)(nni6 + 1);
1750 int niflags = ni6->ni_flags;
1751 u_int32_t ltime;
1752
1753 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1754 return (0); /* needless to copy */
1755
1756 IFNET_RLOCK_NOSLEEP();
1757 ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1758 again:
1759
1760 for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1761 IF_ADDR_RLOCK(ifp);
1762 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1763 if (ifa->ifa_addr->sa_family != AF_INET6)
1764 continue;
1765 ifa6 = (struct in6_ifaddr *)ifa;
1766
1767 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1768 allow_deprecated == 0) {
1769 /*
1770 * prefererred address should be put before
1771 * deprecated addresses.
1772 */
1773
1774 /* record the interface for later search */
1775 if (ifp_dep == NULL)
1776 ifp_dep = ifp;
1777
1778 continue;
1779 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1780 allow_deprecated != 0)
1781 continue; /* we now collect deprecated addrs */
1782
1783 /* What do we have to do about ::1? */
1784 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1785 case IPV6_ADDR_SCOPE_LINKLOCAL:
1786 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1787 continue;
1788 break;
1789 case IPV6_ADDR_SCOPE_SITELOCAL:
1790 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1791 continue;
1792 break;
1793 case IPV6_ADDR_SCOPE_GLOBAL:
1794 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1795 continue;
1796 break;
1797 default:
1798 continue;
1799 }
1800
1801 /*
1802 * check if anycast is okay.
1803 * XXX: just experimental. not in the spec.
1804 */
1805 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1806 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1807 continue;
1808 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1809 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1810 continue;
1811 }
1812
1813 /* now we can copy the address */
1814 if (resid < sizeof(struct in6_addr) +
1815 sizeof(u_int32_t)) {
1816 IF_ADDR_RUNLOCK(ifp);
1817 /*
1818 * We give up much more copy.
1819 * Set the truncate flag and return.
1820 */
1821 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1822 IFNET_RUNLOCK_NOSLEEP();
1823 return (copied);
1824 }
1825
1826 /*
1827 * Set the TTL of the address.
1828 * The TTL value should be one of the following
1829 * according to the specification:
1830 *
1831 * 1. The remaining lifetime of a DHCP lease on the
1832 * address, or
1833 * 2. The remaining Valid Lifetime of a prefix from
1834 * which the address was derived through Stateless
1835 * Autoconfiguration.
1836 *
1837 * Note that we currently do not support stateful
1838 * address configuration by DHCPv6, so the former
1839 * case can't happen.
1840 */
1841 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1842 ltime = ND6_INFINITE_LIFETIME;
1843 else {
1844 if (ifa6->ia6_lifetime.ia6t_expire >
1845 time_uptime)
1846 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1847 else
1848 ltime = 0;
1849 }
1850
1851 bcopy(<ime, cp, sizeof(u_int32_t));
1852 cp += sizeof(u_int32_t);
1853
1854 /* copy the address itself */
1855 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1856 sizeof(struct in6_addr));
1857 in6_clearscope((struct in6_addr *)cp); /* XXX */
1858 cp += sizeof(struct in6_addr);
1859
1860 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1861 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1862 }
1863 IF_ADDR_RUNLOCK(ifp);
1864 if (ifp0) /* we need search only on the specified IF */
1865 break;
1866 }
1867
1868 if (allow_deprecated == 0 && ifp_dep != NULL) {
1869 ifp = ifp_dep;
1870 allow_deprecated = 1;
1871
1872 goto again;
1873 }
1874
1875 IFNET_RUNLOCK_NOSLEEP();
1876
1877 return (copied);
1878 }
1879
1880 /*
1881 * XXX almost dup'ed code with rip6_input.
1882 */
1883 static int
icmp6_rip6_input(struct mbuf ** mp,int off)1884 icmp6_rip6_input(struct mbuf **mp, int off)
1885 {
1886 struct mbuf *m = *mp;
1887 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1888 struct inpcb *inp;
1889 struct inpcb *last = NULL;
1890 struct sockaddr_in6 fromsa;
1891 struct icmp6_hdr *icmp6;
1892 struct epoch_tracker et;
1893 struct mbuf *opts = NULL;
1894
1895 /* This is assumed to be safe; icmp6_input() does a pullup. */
1896 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1897
1898 /*
1899 * XXX: the address may have embedded scope zone ID, which should be
1900 * hidden from applications.
1901 */
1902 bzero(&fromsa, sizeof(fromsa));
1903 fromsa.sin6_family = AF_INET6;
1904 fromsa.sin6_len = sizeof(struct sockaddr_in6);
1905 fromsa.sin6_addr = ip6->ip6_src;
1906 if (sa6_recoverscope(&fromsa)) {
1907 m_freem(m);
1908 *mp = NULL;
1909 return (IPPROTO_DONE);
1910 }
1911
1912 INP_INFO_RLOCK_ET(&V_ripcbinfo, et);
1913 CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
1914 if ((inp->inp_vflag & INP_IPV6) == 0)
1915 continue;
1916 if (inp->inp_ip_p != IPPROTO_ICMPV6)
1917 continue;
1918 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1919 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1920 continue;
1921 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1922 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1923 continue;
1924 INP_RLOCK(inp);
1925 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1926 INP_RUNLOCK(inp);
1927 continue;
1928 }
1929 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1930 inp->in6p_icmp6filt)) {
1931 INP_RUNLOCK(inp);
1932 continue;
1933 }
1934 if (last != NULL) {
1935 struct mbuf *n = NULL;
1936
1937 /*
1938 * Recent network drivers tend to allocate a single
1939 * mbuf cluster, rather than to make a couple of
1940 * mbufs without clusters. Also, since the IPv6 code
1941 * path tries to avoid m_pullup(), it is highly
1942 * probable that we still have an mbuf cluster here
1943 * even though the necessary length can be stored in an
1944 * mbuf's internal buffer.
1945 * Meanwhile, the default size of the receive socket
1946 * buffer for raw sockets is not so large. This means
1947 * the possibility of packet loss is relatively higher
1948 * than before. To avoid this scenario, we copy the
1949 * received data to a separate mbuf that does not use
1950 * a cluster, if possible.
1951 * XXX: it is better to copy the data after stripping
1952 * intermediate headers.
1953 */
1954 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1955 m->m_len <= MHLEN) {
1956 n = m_get(M_NOWAIT, m->m_type);
1957 if (n != NULL) {
1958 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1959 bcopy(m->m_data, n->m_data,
1960 m->m_len);
1961 n->m_len = m->m_len;
1962 } else {
1963 m_free(n);
1964 n = NULL;
1965 }
1966 }
1967 }
1968 if (n != NULL ||
1969 (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1970 if (last->inp_flags & INP_CONTROLOPTS)
1971 ip6_savecontrol(last, n, &opts);
1972 /* strip intermediate headers */
1973 m_adj(n, off);
1974 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1975 if (sbappendaddr_locked(
1976 &last->inp_socket->so_rcv,
1977 (struct sockaddr *)&fromsa, n, opts)
1978 == 0) {
1979 soroverflow_locked(last->inp_socket);
1980 m_freem(n);
1981 if (opts) {
1982 m_freem(opts);
1983 }
1984 } else
1985 sorwakeup_locked(last->inp_socket);
1986 opts = NULL;
1987 }
1988 INP_RUNLOCK(last);
1989 }
1990 last = inp;
1991 }
1992 INP_INFO_RUNLOCK_ET(&V_ripcbinfo, et);
1993 if (last != NULL) {
1994 if (last->inp_flags & INP_CONTROLOPTS)
1995 ip6_savecontrol(last, m, &opts);
1996 /* strip intermediate headers */
1997 m_adj(m, off);
1998
1999 /* avoid using mbuf clusters if possible (see above) */
2000 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2001 m->m_len <= MHLEN) {
2002 struct mbuf *n;
2003
2004 n = m_get(M_NOWAIT, m->m_type);
2005 if (n != NULL) {
2006 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2007 bcopy(m->m_data, n->m_data, m->m_len);
2008 n->m_len = m->m_len;
2009
2010 m_freem(m);
2011 m = n;
2012 } else {
2013 m_freem(n);
2014 n = NULL;
2015 }
2016 }
2017 }
2018 SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2019 if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2020 (struct sockaddr *)&fromsa, m, opts) == 0) {
2021 m_freem(m);
2022 if (opts)
2023 m_freem(opts);
2024 soroverflow_locked(last->inp_socket);
2025 } else
2026 sorwakeup_locked(last->inp_socket);
2027 INP_RUNLOCK(last);
2028 } else {
2029 m_freem(m);
2030 IP6STAT_DEC(ip6s_delivered);
2031 }
2032 *mp = NULL;
2033 return (IPPROTO_DONE);
2034 }
2035
2036 /*
2037 * Reflect the ip6 packet back to the source.
2038 * OFF points to the icmp6 header, counted from the top of the mbuf.
2039 */
2040 void
icmp6_reflect(struct mbuf * m,size_t off)2041 icmp6_reflect(struct mbuf *m, size_t off)
2042 {
2043 struct in6_addr src6, *srcp;
2044 struct ip6_hdr *ip6;
2045 struct icmp6_hdr *icmp6;
2046 struct in6_ifaddr *ia = NULL;
2047 struct ifnet *outif = NULL;
2048 int plen;
2049 int type, code, hlim;
2050
2051 /* too short to reflect */
2052 if (off < sizeof(struct ip6_hdr)) {
2053 nd6log((LOG_DEBUG,
2054 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2055 (u_long)off, (u_long)sizeof(struct ip6_hdr),
2056 __FILE__, __LINE__));
2057 goto bad;
2058 }
2059
2060 /*
2061 * If there are extra headers between IPv6 and ICMPv6, strip
2062 * off that header first.
2063 */
2064 #ifdef DIAGNOSTIC
2065 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2066 panic("assumption failed in icmp6_reflect");
2067 #endif
2068 if (off > sizeof(struct ip6_hdr)) {
2069 size_t l;
2070 struct ip6_hdr nip6;
2071
2072 l = off - sizeof(struct ip6_hdr);
2073 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2074 m_adj(m, l);
2075 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2076 if (m->m_len < l) {
2077 if ((m = m_pullup(m, l)) == NULL)
2078 return;
2079 }
2080 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2081 } else /* off == sizeof(struct ip6_hdr) */ {
2082 size_t l;
2083 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2084 if (m->m_len < l) {
2085 if ((m = m_pullup(m, l)) == NULL)
2086 return;
2087 }
2088 }
2089 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2090 ip6 = mtod(m, struct ip6_hdr *);
2091 ip6->ip6_nxt = IPPROTO_ICMPV6;
2092 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2093 type = icmp6->icmp6_type; /* keep type for statistics */
2094 code = icmp6->icmp6_code; /* ditto. */
2095 hlim = 0;
2096 srcp = NULL;
2097
2098 /*
2099 * If the incoming packet was addressed directly to us (i.e. unicast),
2100 * use dst as the src for the reply.
2101 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2102 * (for example) when we encounter an error while forwarding procedure
2103 * destined to a duplicated address of ours.
2104 */
2105 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2106 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2107 if (ia != NULL && !(ia->ia6_flags &
2108 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2109 src6 = ia->ia_addr.sin6_addr;
2110 srcp = &src6;
2111
2112 if (m->m_pkthdr.rcvif != NULL) {
2113 /* XXX: This may not be the outgoing interface */
2114 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2115 } else
2116 hlim = V_ip6_defhlim;
2117 }
2118 if (ia != NULL)
2119 ifa_free(&ia->ia_ifa);
2120 }
2121
2122 if (srcp == NULL) {
2123 int error;
2124 struct in6_addr dst6;
2125 uint32_t scopeid;
2126
2127 /*
2128 * This case matches to multicasts, our anycast, or unicasts
2129 * that we do not own. Select a source address based on the
2130 * source address of the erroneous packet.
2131 */
2132 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2133 error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2134 scopeid, NULL, &src6, &hlim);
2135
2136 if (error) {
2137 char ip6buf[INET6_ADDRSTRLEN];
2138 nd6log((LOG_DEBUG,
2139 "icmp6_reflect: source can't be determined: "
2140 "dst=%s, error=%d\n",
2141 ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2142 goto bad;
2143 }
2144 srcp = &src6;
2145 }
2146 /*
2147 * ip6_input() drops a packet if its src is multicast.
2148 * So, the src is never multicast.
2149 */
2150 ip6->ip6_dst = ip6->ip6_src;
2151 ip6->ip6_src = *srcp;
2152 ip6->ip6_flow = 0;
2153 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2154 ip6->ip6_vfc |= IPV6_VERSION;
2155 ip6->ip6_nxt = IPPROTO_ICMPV6;
2156 ip6->ip6_hlim = hlim;
2157
2158 icmp6->icmp6_cksum = 0;
2159 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2160 sizeof(struct ip6_hdr), plen);
2161
2162 /*
2163 * XXX option handling
2164 */
2165
2166 m->m_flags &= ~(M_BCAST|M_MCAST);
2167 m->m_pkthdr.rcvif = NULL;
2168 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2169 if (outif)
2170 icmp6_ifoutstat_inc(outif, type, code);
2171
2172 return;
2173
2174 bad:
2175 m_freem(m);
2176 return;
2177 }
2178
2179 void
icmp6_fasttimo(void)2180 icmp6_fasttimo(void)
2181 {
2182
2183 mld_fasttimo();
2184 }
2185
2186 void
icmp6_slowtimo(void)2187 icmp6_slowtimo(void)
2188 {
2189
2190 mld_slowtimo();
2191 }
2192
2193 static const char *
icmp6_redirect_diag(struct in6_addr * src6,struct in6_addr * dst6,struct in6_addr * tgt6)2194 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2195 struct in6_addr *tgt6)
2196 {
2197 static char buf[1024];
2198 char ip6bufs[INET6_ADDRSTRLEN];
2199 char ip6bufd[INET6_ADDRSTRLEN];
2200 char ip6buft[INET6_ADDRSTRLEN];
2201 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2202 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2203 ip6_sprintf(ip6buft, tgt6));
2204 return buf;
2205 }
2206
2207 void
icmp6_redirect_input(struct mbuf * m,int off)2208 icmp6_redirect_input(struct mbuf *m, int off)
2209 {
2210 struct ifnet *ifp;
2211 struct ip6_hdr *ip6;
2212 struct nd_redirect *nd_rd;
2213 struct in6_addr src6, redtgt6, reddst6;
2214 union nd_opts ndopts;
2215 char ip6buf[INET6_ADDRSTRLEN];
2216 char *lladdr;
2217 int icmp6len, is_onlink, is_router, lladdrlen;
2218
2219 M_ASSERTPKTHDR(m);
2220 KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2221
2222 /* XXX if we are router, we don't update route by icmp6 redirect */
2223 if (V_ip6_forwarding)
2224 goto freeit;
2225 if (!V_icmp6_rediraccept)
2226 goto freeit;
2227
2228 /* RFC 6980: Nodes MUST silently ignore fragments */
2229 if(m->m_flags & M_FRAGMENTED)
2230 goto freeit;
2231
2232 ip6 = mtod(m, struct ip6_hdr *);
2233 icmp6len = ntohs(ip6->ip6_plen);
2234 if (m->m_len < off + icmp6len) {
2235 m = m_pullup(m, off + icmp6len);
2236 if (m == NULL) {
2237 IP6STAT_INC(ip6s_exthdrtoolong);
2238 return;
2239 }
2240 }
2241 ip6 = mtod(m, struct ip6_hdr *);
2242 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2243
2244 ifp = m->m_pkthdr.rcvif;
2245 redtgt6 = nd_rd->nd_rd_target;
2246 reddst6 = nd_rd->nd_rd_dst;
2247
2248 if (in6_setscope(&redtgt6, ifp, NULL) ||
2249 in6_setscope(&reddst6, ifp, NULL)) {
2250 goto freeit;
2251 }
2252
2253 /* validation */
2254 src6 = ip6->ip6_src;
2255 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2256 nd6log((LOG_ERR,
2257 "ICMP6 redirect sent from %s rejected; "
2258 "must be from linklocal\n",
2259 ip6_sprintf(ip6buf, &src6)));
2260 goto bad;
2261 }
2262 if (ip6->ip6_hlim != 255) {
2263 nd6log((LOG_ERR,
2264 "ICMP6 redirect sent from %s rejected; "
2265 "hlim=%d (must be 255)\n",
2266 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2267 goto bad;
2268 }
2269 {
2270 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2271 struct nhop6_basic nh6;
2272 struct in6_addr kdst;
2273 uint32_t scopeid;
2274
2275 in6_splitscope(&reddst6, &kdst, &scopeid);
2276 if (fib6_lookup_nh_basic(ifp->if_fib, &kdst, scopeid, 0, 0,&nh6)==0){
2277 if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2278 nd6log((LOG_ERR,
2279 "ICMP6 redirect rejected; no route "
2280 "with inet6 gateway found for redirect dst: %s\n",
2281 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2282 goto bad;
2283 }
2284
2285 /*
2286 * Embed scope zone id into next hop address, since
2287 * fib6_lookup_nh_basic() returns address without embedded
2288 * scope zone id.
2289 */
2290 if (in6_setscope(&nh6.nh_addr, m->m_pkthdr.rcvif, NULL))
2291 goto freeit;
2292
2293 if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2294 nd6log((LOG_ERR,
2295 "ICMP6 redirect rejected; "
2296 "not equal to gw-for-src=%s (must be same): "
2297 "%s\n",
2298 ip6_sprintf(ip6buf, &nh6.nh_addr),
2299 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2300 goto bad;
2301 }
2302 } else {
2303 nd6log((LOG_ERR,
2304 "ICMP6 redirect rejected; "
2305 "no route found for redirect dst: %s\n",
2306 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2307 goto bad;
2308 }
2309 }
2310 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2311 nd6log((LOG_ERR,
2312 "ICMP6 redirect rejected; "
2313 "redirect dst must be unicast: %s\n",
2314 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2315 goto bad;
2316 }
2317
2318 is_router = is_onlink = 0;
2319 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2320 is_router = 1; /* router case */
2321 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2322 is_onlink = 1; /* on-link destination case */
2323 if (!is_router && !is_onlink) {
2324 nd6log((LOG_ERR,
2325 "ICMP6 redirect rejected; "
2326 "neither router case nor onlink case: %s\n",
2327 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2328 goto bad;
2329 }
2330
2331 icmp6len -= sizeof(*nd_rd);
2332 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2333 if (nd6_options(&ndopts) < 0) {
2334 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2335 __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2336 /* nd6_options have incremented stats */
2337 goto freeit;
2338 }
2339
2340 lladdr = NULL;
2341 lladdrlen = 0;
2342 if (ndopts.nd_opts_tgt_lladdr) {
2343 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2344 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2345 }
2346
2347 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2348 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2349 "(if %d, icmp6 packet %d): %s\n",
2350 __func__, ip6_sprintf(ip6buf, &redtgt6),
2351 ifp->if_addrlen, lladdrlen - 2,
2352 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2353 goto bad;
2354 }
2355
2356 /* Validation passed. */
2357
2358 /* RFC 2461 8.3 */
2359 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2360 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2361
2362 /*
2363 * Install a gateway route in the better-router case or an interface
2364 * route in the on-link-destination case.
2365 */
2366 {
2367 struct sockaddr_in6 sdst;
2368 struct sockaddr_in6 sgw;
2369 struct sockaddr_in6 ssrc;
2370 struct sockaddr *gw;
2371 int rt_flags;
2372 u_int fibnum;
2373
2374 bzero(&sdst, sizeof(sdst));
2375 bzero(&ssrc, sizeof(ssrc));
2376 sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2377 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2378 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2379 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2380 rt_flags = RTF_HOST;
2381 if (is_router) {
2382 bzero(&sgw, sizeof(sgw));
2383 sgw.sin6_family = AF_INET6;
2384 sgw.sin6_len = sizeof(struct sockaddr_in6);
2385 bcopy(&redtgt6, &sgw.sin6_addr,
2386 sizeof(struct in6_addr));
2387 gw = (struct sockaddr *)&sgw;
2388 rt_flags |= RTF_GATEWAY;
2389 } else
2390 gw = ifp->if_addr->ifa_addr;
2391 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2392 in6_rtredirect((struct sockaddr *)&sdst, gw,
2393 (struct sockaddr *)NULL, rt_flags,
2394 (struct sockaddr *)&ssrc, fibnum);
2395 }
2396 /* finally update cached route in each socket via pfctlinput */
2397 {
2398 struct sockaddr_in6 sdst;
2399
2400 bzero(&sdst, sizeof(sdst));
2401 sdst.sin6_family = AF_INET6;
2402 sdst.sin6_len = sizeof(struct sockaddr_in6);
2403 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2404 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2405 }
2406
2407 freeit:
2408 m_freem(m);
2409 return;
2410
2411 bad:
2412 ICMP6STAT_INC(icp6s_badredirect);
2413 m_freem(m);
2414 }
2415
2416 void
icmp6_redirect_output(struct mbuf * m0,struct rtentry * rt)2417 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2418 {
2419 struct ifnet *ifp; /* my outgoing interface */
2420 struct in6_addr *ifp_ll6;
2421 struct in6_addr *router_ll6;
2422 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2423 struct mbuf *m = NULL; /* newly allocated one */
2424 struct m_tag *mtag;
2425 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2426 struct nd_redirect *nd_rd;
2427 struct llentry *ln = NULL;
2428 size_t maxlen;
2429 u_char *p;
2430 struct ifnet *outif = NULL;
2431 struct sockaddr_in6 src_sa;
2432
2433 icmp6_errcount(ND_REDIRECT, 0);
2434
2435 /* if we are not router, we don't send icmp6 redirect */
2436 if (!V_ip6_forwarding)
2437 goto fail;
2438
2439 /* sanity check */
2440 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2441 goto fail;
2442
2443 /*
2444 * Address check:
2445 * the source address must identify a neighbor, and
2446 * the destination address must not be a multicast address
2447 * [RFC 2461, sec 8.2]
2448 */
2449 sip6 = mtod(m0, struct ip6_hdr *);
2450 bzero(&src_sa, sizeof(src_sa));
2451 src_sa.sin6_family = AF_INET6;
2452 src_sa.sin6_len = sizeof(src_sa);
2453 src_sa.sin6_addr = sip6->ip6_src;
2454 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2455 goto fail;
2456 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2457 goto fail; /* what should we do here? */
2458
2459 /* rate limit */
2460 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2461 goto fail;
2462
2463 /*
2464 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2465 * we almost always ask for an mbuf cluster for simplicity.
2466 * (MHLEN < IPV6_MMTU is almost always true)
2467 */
2468 #if IPV6_MMTU >= MCLBYTES
2469 # error assumption failed about IPV6_MMTU and MCLBYTES
2470 #endif
2471 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2472 if (m == NULL)
2473 goto fail;
2474 M_SETFIB(m, rt->rt_fibnum);
2475 maxlen = M_TRAILINGSPACE(m);
2476 maxlen = min(IPV6_MMTU, maxlen);
2477 /* just for safety */
2478 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2479 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2480 goto fail;
2481 }
2482
2483 {
2484 /* get ip6 linklocal address for ifp(my outgoing interface). */
2485 struct in6_ifaddr *ia;
2486 if ((ia = in6ifa_ifpforlinklocal(ifp,
2487 IN6_IFF_NOTREADY|
2488 IN6_IFF_ANYCAST)) == NULL)
2489 goto fail;
2490 ifp_ll6 = &ia->ia_addr.sin6_addr;
2491 /* XXXRW: reference released prematurely. */
2492 ifa_free(&ia->ia_ifa);
2493 }
2494
2495 /* get ip6 linklocal address for the router. */
2496 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2497 struct sockaddr_in6 *sin6;
2498 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2499 router_ll6 = &sin6->sin6_addr;
2500 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2501 router_ll6 = (struct in6_addr *)NULL;
2502 } else
2503 router_ll6 = (struct in6_addr *)NULL;
2504
2505 /* ip6 */
2506 ip6 = mtod(m, struct ip6_hdr *);
2507 ip6->ip6_flow = 0;
2508 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2509 ip6->ip6_vfc |= IPV6_VERSION;
2510 /* ip6->ip6_plen will be set later */
2511 ip6->ip6_nxt = IPPROTO_ICMPV6;
2512 ip6->ip6_hlim = 255;
2513 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2514 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2515 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2516
2517 /* ND Redirect */
2518 nd_rd = (struct nd_redirect *)(ip6 + 1);
2519 nd_rd->nd_rd_type = ND_REDIRECT;
2520 nd_rd->nd_rd_code = 0;
2521 nd_rd->nd_rd_reserved = 0;
2522 if (rt->rt_flags & RTF_GATEWAY) {
2523 /*
2524 * nd_rd->nd_rd_target must be a link-local address in
2525 * better router cases.
2526 */
2527 if (!router_ll6)
2528 goto fail;
2529 bcopy(router_ll6, &nd_rd->nd_rd_target,
2530 sizeof(nd_rd->nd_rd_target));
2531 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2532 sizeof(nd_rd->nd_rd_dst));
2533 } else {
2534 /* make sure redtgt == reddst */
2535 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2536 sizeof(nd_rd->nd_rd_target));
2537 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2538 sizeof(nd_rd->nd_rd_dst));
2539 }
2540
2541 p = (u_char *)(nd_rd + 1);
2542
2543 if (!router_ll6)
2544 goto nolladdropt;
2545
2546 {
2547 /* target lladdr option */
2548 int len;
2549 struct nd_opt_hdr *nd_opt;
2550 char *lladdr;
2551
2552 IF_AFDATA_RLOCK(ifp);
2553 ln = nd6_lookup(router_ll6, 0, ifp);
2554 IF_AFDATA_RUNLOCK(ifp);
2555 if (ln == NULL)
2556 goto nolladdropt;
2557
2558 len = sizeof(*nd_opt) + ifp->if_addrlen;
2559 len = (len + 7) & ~7; /* round by 8 */
2560 /* safety check */
2561 if (len + (p - (u_char *)ip6) > maxlen)
2562 goto nolladdropt;
2563
2564 if (ln->la_flags & LLE_VALID) {
2565 nd_opt = (struct nd_opt_hdr *)p;
2566 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2567 nd_opt->nd_opt_len = len >> 3;
2568 lladdr = (char *)(nd_opt + 1);
2569 bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2570 p += len;
2571 }
2572 }
2573 nolladdropt:
2574 if (ln != NULL)
2575 LLE_RUNLOCK(ln);
2576
2577 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2578
2579 /* just to be safe */
2580 #ifdef M_DECRYPTED /*not openbsd*/
2581 if (m0->m_flags & M_DECRYPTED)
2582 goto noredhdropt;
2583 #endif
2584 if (p - (u_char *)ip6 > maxlen)
2585 goto noredhdropt;
2586
2587 {
2588 /* redirected header option */
2589 int len;
2590 struct nd_opt_rd_hdr *nd_opt_rh;
2591
2592 /*
2593 * compute the maximum size for icmp6 redirect header option.
2594 * XXX room for auth header?
2595 */
2596 len = maxlen - (p - (u_char *)ip6);
2597 len &= ~7;
2598
2599 /* This is just for simplicity. */
2600 if (m0->m_pkthdr.len != m0->m_len) {
2601 if (m0->m_next) {
2602 m_freem(m0->m_next);
2603 m0->m_next = NULL;
2604 }
2605 m0->m_pkthdr.len = m0->m_len;
2606 }
2607
2608 /*
2609 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2610 * about padding/truncate rule for the original IP packet.
2611 * From the discussion on IPv6imp in Feb 1999,
2612 * the consensus was:
2613 * - "attach as much as possible" is the goal
2614 * - pad if not aligned (original size can be guessed by
2615 * original ip6 header)
2616 * Following code adds the padding if it is simple enough,
2617 * and truncates if not.
2618 */
2619 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2620 panic("assumption failed in %s:%d", __FILE__,
2621 __LINE__);
2622
2623 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2624 /* not enough room, truncate */
2625 m0->m_pkthdr.len = m0->m_len = len -
2626 sizeof(*nd_opt_rh);
2627 } else {
2628 /* enough room, pad or truncate */
2629 size_t extra;
2630
2631 extra = m0->m_pkthdr.len % 8;
2632 if (extra) {
2633 /* pad if easy enough, truncate if not */
2634 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2635 /* pad */
2636 m0->m_len += (8 - extra);
2637 m0->m_pkthdr.len += (8 - extra);
2638 } else {
2639 /* truncate */
2640 m0->m_pkthdr.len -= extra;
2641 m0->m_len -= extra;
2642 }
2643 }
2644 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2645 m0->m_pkthdr.len = m0->m_len = len -
2646 sizeof(*nd_opt_rh);
2647 }
2648
2649 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2650 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2651 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2652 nd_opt_rh->nd_opt_rh_len = len >> 3;
2653 p += sizeof(*nd_opt_rh);
2654 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2655
2656 /* connect m0 to m */
2657 m_tag_delete_chain(m0, NULL);
2658 m0->m_flags &= ~M_PKTHDR;
2659 m->m_next = m0;
2660 m->m_pkthdr.len = m->m_len + m0->m_len;
2661 m0 = NULL;
2662 }
2663 noredhdropt:;
2664 if (m0) {
2665 m_freem(m0);
2666 m0 = NULL;
2667 }
2668
2669 /* XXX: clear embedded link IDs in the inner header */
2670 in6_clearscope(&sip6->ip6_src);
2671 in6_clearscope(&sip6->ip6_dst);
2672 in6_clearscope(&nd_rd->nd_rd_target);
2673 in6_clearscope(&nd_rd->nd_rd_dst);
2674
2675 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2676
2677 nd_rd->nd_rd_cksum = 0;
2678 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2679 sizeof(*ip6), ntohs(ip6->ip6_plen));
2680
2681 if (send_sendso_input_hook != NULL) {
2682 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2683 M_NOWAIT);
2684 if (mtag == NULL)
2685 goto fail;
2686 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2687 m_tag_prepend(m, mtag);
2688 }
2689
2690 /* send the packet to outside... */
2691 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2692 if (outif) {
2693 icmp6_ifstat_inc(outif, ifs6_out_msg);
2694 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2695 }
2696 ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2697
2698 return;
2699
2700 fail:
2701 if (m)
2702 m_freem(m);
2703 if (m0)
2704 m_freem(m0);
2705 }
2706
2707 /*
2708 * ICMPv6 socket option processing.
2709 */
2710 int
icmp6_ctloutput(struct socket * so,struct sockopt * sopt)2711 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2712 {
2713 int error = 0;
2714 int optlen;
2715 struct inpcb *inp = sotoinpcb(so);
2716 int level, op, optname;
2717
2718 if (sopt) {
2719 level = sopt->sopt_level;
2720 op = sopt->sopt_dir;
2721 optname = sopt->sopt_name;
2722 optlen = sopt->sopt_valsize;
2723 } else
2724 level = op = optname = optlen = 0;
2725
2726 if (level != IPPROTO_ICMPV6) {
2727 return EINVAL;
2728 }
2729
2730 switch (op) {
2731 case PRCO_SETOPT:
2732 switch (optname) {
2733 case ICMP6_FILTER:
2734 {
2735 struct icmp6_filter ic6f;
2736
2737 if (optlen != sizeof(ic6f)) {
2738 error = EMSGSIZE;
2739 break;
2740 }
2741 error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2742 if (error == 0) {
2743 INP_WLOCK(inp);
2744 *inp->in6p_icmp6filt = ic6f;
2745 INP_WUNLOCK(inp);
2746 }
2747 break;
2748 }
2749
2750 default:
2751 error = ENOPROTOOPT;
2752 break;
2753 }
2754 break;
2755
2756 case PRCO_GETOPT:
2757 switch (optname) {
2758 case ICMP6_FILTER:
2759 {
2760 struct icmp6_filter ic6f;
2761
2762 INP_RLOCK(inp);
2763 ic6f = *inp->in6p_icmp6filt;
2764 INP_RUNLOCK(inp);
2765 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2766 break;
2767 }
2768
2769 default:
2770 error = ENOPROTOOPT;
2771 break;
2772 }
2773 break;
2774 }
2775
2776 return (error);
2777 }
2778
2779 /*
2780 * Perform rate limit check.
2781 * Returns 0 if it is okay to send the icmp6 packet.
2782 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2783 * limitation.
2784 *
2785 * XXX per-destination/type check necessary?
2786 *
2787 * dst - not used at this moment
2788 * type - not used at this moment
2789 * code - not used at this moment
2790 */
2791 static int
icmp6_ratelimit(const struct in6_addr * dst,const int type,const int code)2792 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2793 const int code)
2794 {
2795 int ret;
2796
2797 ret = 0; /* okay to send */
2798
2799 /* PPS limit */
2800 if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2801 V_icmp6errppslim)) {
2802 /* The packet is subject to rate limit */
2803 ret++;
2804 }
2805
2806 return ret;
2807 }
2808