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
63 #define MBUF_PRIVATE /* XXXRW: Optimisation tries to avoid M_EXT mbufs */
64
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67
68 #include <sys/param.h>
69 #include <sys/domain.h>
70 #include <sys/jail.h>
71 #include <sys/kernel.h>
72 #include <sys/lock.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/proc.h>
76 #include <sys/protosw.h>
77 #include <sys/signalvar.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/sx.h>
81 #include <sys/syslog.h>
82 #include <sys/systm.h>
83 #include <sys/time.h>
84
85 #include <net/if.h>
86 #include <net/if_var.h>
87 #include <net/if_dl.h>
88 #include <net/if_llatbl.h>
89 #include <net/if_private.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/route/route_ctl.h>
93 #include <net/route/nhop.h>
94 #include <net/vnet.h>
95
96 #include <netinet/in.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet/tcp_var.h>
102
103 #include <netinet6/in6_fib.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/ip6_var.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/mld6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/send.h>
111
112 extern ip6proto_ctlinput_t *ip6_ctlprotox[];
113
114 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
115 VNET_PCPUSTAT_SYSINIT(icmp6stat);
116 SYSCTL_VNET_PCPUSTAT(_net_inet6_icmp6, ICMPV6CTL_STATS, stats,
117 struct icmp6stat, icmp6stat,
118 "ICMPv6 statistics (struct icmp6stat, netinet/icmp6.h)");
119
120 #ifdef VIMAGE
121 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
122 #endif /* VIMAGE */
123
124 VNET_DEFINE_STATIC(int, icmp6_rediraccept) = 1;
125 #define V_icmp6_rediraccept VNET(icmp6_rediraccept)
126 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRACCEPT, rediraccept,
127 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_rediraccept), 0,
128 "Accept ICMPv6 redirect messages");
129
130 VNET_DEFINE_STATIC(int, icmp6_redirtimeout) = 10 * 60; /* 10 minutes */
131 #define V_icmp6_redirtimeout VNET(icmp6_redirtimeout)
132 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRTIMEOUT, redirtimeout,
133 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_redirtimeout), 0,
134 "Delay in seconds before expiring redirect route");
135
136 VNET_DEFINE_STATIC(int, icmp6_nodeinfo) = 0;
137 #define V_icmp6_nodeinfo VNET(icmp6_nodeinfo)
138 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO, nodeinfo,
139 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_nodeinfo), 0,
140 "Mask of enabled RFC4620 node information query types");
141
142 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
143 #define V_ripcbinfo VNET(ripcbinfo)
144
145 VNET_DECLARE(int, rip_bind_all_fibs);
146 #define V_rip_bind_all_fibs VNET(rip_bind_all_fibs)
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 static void icmp6_mtudisc_update(struct ip6ctlparam *ip6cp);
162
163 /*
164 * Kernel module interface for updating icmp6stat. The argument is an index
165 * into icmp6stat treated as an array of u_quad_t. While this encodes the
166 * general layout of icmp6stat into the caller, it doesn't encode its
167 * location, so that future changes to add, for example, per-CPU stats
168 * support won't cause binary compatibility problems for kernel modules.
169 */
170 void
kmod_icmp6stat_inc(int statnum)171 kmod_icmp6stat_inc(int statnum)
172 {
173
174 counter_u64_add(VNET(icmp6stat)[statnum], 1);
175 }
176
177 static void
icmp6_errcount(int type,int code)178 icmp6_errcount(int type, int code)
179 {
180 switch (type) {
181 case ICMP6_DST_UNREACH:
182 switch (code) {
183 case ICMP6_DST_UNREACH_NOROUTE:
184 ICMP6STAT_INC(icp6s_odst_unreach_noroute);
185 return;
186 case ICMP6_DST_UNREACH_ADMIN:
187 ICMP6STAT_INC(icp6s_odst_unreach_admin);
188 return;
189 case ICMP6_DST_UNREACH_BEYONDSCOPE:
190 ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
191 return;
192 case ICMP6_DST_UNREACH_ADDR:
193 ICMP6STAT_INC(icp6s_odst_unreach_addr);
194 return;
195 case ICMP6_DST_UNREACH_NOPORT:
196 ICMP6STAT_INC(icp6s_odst_unreach_noport);
197 return;
198 }
199 break;
200 case ICMP6_PACKET_TOO_BIG:
201 ICMP6STAT_INC(icp6s_opacket_too_big);
202 return;
203 case ICMP6_TIME_EXCEEDED:
204 switch (code) {
205 case ICMP6_TIME_EXCEED_TRANSIT:
206 ICMP6STAT_INC(icp6s_otime_exceed_transit);
207 return;
208 case ICMP6_TIME_EXCEED_REASSEMBLY:
209 ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
210 return;
211 }
212 break;
213 case ICMP6_PARAM_PROB:
214 switch (code) {
215 case ICMP6_PARAMPROB_HEADER:
216 ICMP6STAT_INC(icp6s_oparamprob_header);
217 return;
218 case ICMP6_PARAMPROB_NEXTHEADER:
219 ICMP6STAT_INC(icp6s_oparamprob_nextheader);
220 return;
221 case ICMP6_PARAMPROB_OPTION:
222 ICMP6STAT_INC(icp6s_oparamprob_option);
223 return;
224 }
225 break;
226 case ND_REDIRECT:
227 ICMP6STAT_INC(icp6s_oredirect);
228 return;
229 }
230 ICMP6STAT_INC(icp6s_ounknown);
231 }
232
233 /*
234 * A wrapper function for icmp6_error() necessary when the erroneous packet
235 * may not contain enough scope zone information.
236 */
237 void
icmp6_error2(struct mbuf * m,int type,int code,int param,struct ifnet * ifp)238 icmp6_error2(struct mbuf *m, int type, int code, int param,
239 struct ifnet *ifp)
240 {
241 struct ip6_hdr *ip6;
242
243 if (ifp == NULL)
244 return;
245
246 if (m->m_len < sizeof(struct ip6_hdr)) {
247 m = m_pullup(m, sizeof(struct ip6_hdr));
248 if (m == NULL) {
249 IP6STAT_INC(ip6s_exthdrtoolong);
250 return;
251 }
252 }
253 ip6 = mtod(m, struct ip6_hdr *);
254
255 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
256 return;
257 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
258 return;
259
260 icmp6_error(m, type, code, param);
261 }
262
263 /*
264 * Generate an error packet of type error in response to bad IP6 packet.
265 */
266 void
icmp6_error(struct mbuf * m,int type,int code,int param)267 icmp6_error(struct mbuf *m, int type, int code, int param)
268 {
269 struct ip6_hdr *oip6, *nip6;
270 struct icmp6_hdr *icmp6;
271 struct epoch_tracker et;
272 u_int preplen;
273 int off;
274 int nxt;
275
276 ICMP6STAT_INC(icp6s_error);
277
278 /* count per-type-code statistics */
279 icmp6_errcount(type, code);
280
281 #ifdef M_DECRYPTED /*not openbsd*/
282 if (m->m_flags & M_DECRYPTED) {
283 ICMP6STAT_INC(icp6s_canterror);
284 goto freeit;
285 }
286 #endif
287
288 if (m->m_len < sizeof(struct ip6_hdr)) {
289 m = m_pullup(m, sizeof(struct ip6_hdr));
290 if (m == NULL) {
291 IP6STAT_INC(ip6s_exthdrtoolong);
292 return;
293 }
294 }
295 oip6 = mtod(m, struct ip6_hdr *);
296
297 /*
298 * If the destination address of the erroneous packet is a multicast
299 * address, or the packet was sent using link-layer multicast,
300 * we should basically suppress sending an error (RFC 2463, Section
301 * 2.4).
302 * We have two exceptions (the item e.2 in that section):
303 * - the Packet Too Big message can be sent for path MTU discovery.
304 * - the Parameter Problem Message that can be allowed an icmp6 error
305 * in the option type field. This check has been done in
306 * ip6_unknown_opt(), so we can just check the type and code.
307 */
308 if ((m->m_flags & (M_BCAST|M_MCAST) ||
309 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
310 (type != ICMP6_PACKET_TOO_BIG &&
311 (type != ICMP6_PARAM_PROB ||
312 code != ICMP6_PARAMPROB_OPTION)))
313 goto freeit;
314
315 /*
316 * RFC 2463, 2.4 (e.5): source address check.
317 * XXX: the case of anycast source?
318 */
319 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
320 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
321 goto freeit;
322
323 /*
324 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
325 * don't do it.
326 */
327 nxt = -1;
328 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
329 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
330 struct icmp6_hdr *icp;
331
332 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
333 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
334 if (m == NULL) {
335 IP6STAT_INC(ip6s_exthdrtoolong);
336 return;
337 }
338 }
339 oip6 = mtod(m, struct ip6_hdr *);
340 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
341
342 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
343 icp->icmp6_type == ND_REDIRECT) {
344 /*
345 * ICMPv6 error
346 * Special case: for redirect (which is
347 * informational) we must not send icmp6 error.
348 */
349 ICMP6STAT_INC(icp6s_canterror);
350 goto freeit;
351 } else {
352 /* ICMPv6 informational - send the error */
353 }
354 } else {
355 /* non-ICMPv6 - send the error */
356 }
357
358 /* Finally, do rate limitation check. */
359 if (icmp6_ratelimit(&oip6->ip6_src, type, code))
360 goto freeit;
361
362 /*
363 * OK, ICMP6 can be generated.
364 */
365
366 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
367 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
368
369 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
370 M_PREPEND(m, preplen, M_NOWAIT); /* FIB is also copied over. */
371 if (m == NULL) {
372 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
373 return;
374 }
375
376 nip6 = mtod(m, struct ip6_hdr *);
377 nip6->ip6_src = oip6->ip6_src;
378 nip6->ip6_dst = oip6->ip6_dst;
379
380 in6_clearscope(&oip6->ip6_src);
381 in6_clearscope(&oip6->ip6_dst);
382
383 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
384 icmp6->icmp6_type = type;
385 icmp6->icmp6_code = code;
386 icmp6->icmp6_pptr = htonl((u_int32_t)param);
387
388 ICMP6STAT_INC2(icp6s_outhist, type);
389 NET_EPOCH_ENTER(et);
390 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
391 NET_EPOCH_EXIT(et);
392
393 return;
394
395 freeit:
396 /*
397 * If we can't tell whether or not we can generate ICMP6, free it.
398 */
399 m_freem(m);
400 }
401
402 int
icmp6_errmap(const struct icmp6_hdr * icmp6)403 icmp6_errmap(const struct icmp6_hdr *icmp6)
404 {
405
406 switch (icmp6->icmp6_type) {
407 case ICMP6_DST_UNREACH:
408 switch (icmp6->icmp6_code) {
409 case ICMP6_DST_UNREACH_NOROUTE:
410 case ICMP6_DST_UNREACH_ADDR:
411 return (EHOSTUNREACH);
412 case ICMP6_DST_UNREACH_NOPORT:
413 case ICMP6_DST_UNREACH_ADMIN:
414 return (ECONNREFUSED);
415 case ICMP6_DST_UNREACH_BEYONDSCOPE:
416 return (ENOPROTOOPT);
417 default:
418 return (0); /* Shouldn't happen. */
419 }
420 case ICMP6_PACKET_TOO_BIG:
421 return (EMSGSIZE);
422 case ICMP6_TIME_EXCEEDED:
423 switch (icmp6->icmp6_code) {
424 case ICMP6_TIME_EXCEED_TRANSIT:
425 return (EHOSTUNREACH);
426 case ICMP6_TIME_EXCEED_REASSEMBLY:
427 return (0);
428 default:
429 return (0); /* Shouldn't happen. */
430 }
431 case ICMP6_PARAM_PROB:
432 switch (icmp6->icmp6_code) {
433 case ICMP6_PARAMPROB_NEXTHEADER:
434 return (ECONNREFUSED);
435 case ICMP6_PARAMPROB_HEADER:
436 case ICMP6_PARAMPROB_OPTION:
437 return (ENOPROTOOPT);
438 default:
439 return (0); /* Shouldn't happen. */
440 }
441 default:
442 return (0);
443 }
444 }
445
446 /*
447 * Process a received ICMP6 message.
448 */
449 int
icmp6_input(struct mbuf ** mp,int * offp,int proto)450 icmp6_input(struct mbuf **mp, int *offp, int proto)
451 {
452 struct mbuf *m, *n;
453 struct ifnet *ifp;
454 struct ip6_hdr *ip6, *nip6;
455 struct icmp6_hdr *icmp6, *nicmp6;
456 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
457 int code, error, icmp6len, ip6len, noff, off, sum;
458
459 NET_EPOCH_ASSERT();
460
461 m = *mp;
462 off = *offp;
463
464 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
465 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
466 if (m == NULL) {
467 IP6STAT_INC(ip6s_exthdrtoolong);
468 *mp = m;
469 return (IPPROTO_DONE);
470 }
471 }
472
473 /*
474 * Locate icmp6 structure in mbuf, and check
475 * that not corrupted and of at least minimum length
476 */
477
478 icmp6len = m->m_pkthdr.len - off;
479 if (icmp6len < sizeof(struct icmp6_hdr)) {
480 ICMP6STAT_INC(icp6s_tooshort);
481 goto freeit;
482 }
483
484 ip6 = mtod(m, struct ip6_hdr *);
485 ifp = m->m_pkthdr.rcvif;
486 /*
487 * Check multicast group membership.
488 * Note: SSM filters are not applied for ICMPv6 traffic.
489 */
490 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
491 struct in6_multi *inm;
492
493 inm = in6m_lookup(ifp, &ip6->ip6_dst);
494 if (inm == NULL) {
495 IP6STAT_INC(ip6s_notmember);
496 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
497 goto freeit;
498 }
499 }
500
501 /* Calculate the checksum. */
502 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
503 code = icmp6->icmp6_code;
504 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
505 nd6log((LOG_ERR,
506 "ICMP6 checksum error(%d|%x) %s\n",
507 icmp6->icmp6_type, sum,
508 ip6_sprintf(ip6bufs, &ip6->ip6_src)));
509 ICMP6STAT_INC(icp6s_checksum);
510 goto freeit;
511 }
512
513 ICMP6STAT_INC2(icp6s_inhist, icmp6->icmp6_type);
514 icmp6_ifstat_inc(ifp, ifs6_in_msg);
515 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
516 icmp6_ifstat_inc(ifp, ifs6_in_error);
517
518 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
519 switch (icmp6->icmp6_type) {
520 case ICMP6_DST_UNREACH:
521 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
522 switch (code) {
523 case ICMP6_DST_UNREACH_ADMIN:
524 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
525 case ICMP6_DST_UNREACH_NOROUTE:
526 case ICMP6_DST_UNREACH_ADDR:
527 case ICMP6_DST_UNREACH_BEYONDSCOPE:
528 case ICMP6_DST_UNREACH_NOPORT:
529 goto deliver;
530 default:
531 goto badcode;
532 }
533 case ICMP6_PACKET_TOO_BIG:
534 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
535 /*
536 * Validation is made in icmp6_mtudisc_update.
537 * Updating the path MTU will be done after examining
538 * intermediate extension headers.
539 */
540 goto deliver;
541 case ICMP6_TIME_EXCEEDED:
542 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
543 switch (code) {
544 case ICMP6_TIME_EXCEED_TRANSIT:
545 case ICMP6_TIME_EXCEED_REASSEMBLY:
546 goto deliver;
547 default:
548 goto badcode;
549 }
550 case ICMP6_PARAM_PROB:
551 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
552 switch (code) {
553 case ICMP6_PARAMPROB_NEXTHEADER:
554 case ICMP6_PARAMPROB_HEADER:
555 case ICMP6_PARAMPROB_OPTION:
556 goto deliver;
557 default:
558 goto badcode;
559 }
560 case ICMP6_ECHO_REQUEST:
561 icmp6_ifstat_inc(ifp, ifs6_in_echo);
562 if (code != 0)
563 goto badcode;
564 if (icmp6_ratelimit(&ip6->ip6_src, ICMP6_ECHO_REPLY, 0))
565 break;
566 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
567 /* Give up remote */
568 break;
569 }
570 if (!M_WRITABLE(n)
571 || n->m_len < off + sizeof(struct icmp6_hdr)) {
572 struct mbuf *n0 = n;
573 int n0len;
574
575 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
576 n = m_gethdr(M_NOWAIT, n0->m_type);
577 if (n == NULL) {
578 /* Give up remote */
579 m_freem(n0);
580 break;
581 }
582
583 m_move_pkthdr(n, n0); /* FIB copied. */
584 n0len = n0->m_pkthdr.len; /* save for use below */
585 /*
586 * Copy IPv6 and ICMPv6 only.
587 */
588 nip6 = mtod(n, struct ip6_hdr *);
589 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
590 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
591 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
592 noff = sizeof(struct ip6_hdr);
593 /* new mbuf contains only ipv6+icmpv6 headers */
594 n->m_len = noff + sizeof(struct icmp6_hdr);
595 /*
596 * Adjust mbuf. ip6_plen will be adjusted in
597 * ip6_output().
598 */
599 m_adj(n0, off + sizeof(struct icmp6_hdr));
600 /* recalculate complete packet size */
601 n->m_pkthdr.len = n0len + (noff - off);
602 n->m_next = n0;
603 } else {
604 if (n->m_len < off + sizeof(*nicmp6)) {
605 n = m_pullup(n, off + sizeof(*nicmp6));
606 if (n == NULL) {
607 IP6STAT_INC(ip6s_exthdrtoolong);
608 break;
609 }
610 }
611 nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
612 noff = off;
613 }
614 if (n) {
615 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
616 nicmp6->icmp6_code = 0;
617 ICMP6STAT_INC(icp6s_reflect);
618 ICMP6STAT_INC2(icp6s_outhist, ICMP6_ECHO_REPLY);
619 icmp6_reflect(n, noff);
620 }
621 break;
622
623 case ICMP6_ECHO_REPLY:
624 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
625 if (code != 0)
626 goto badcode;
627 break;
628
629 case MLD_LISTENER_QUERY:
630 case MLD_LISTENER_REPORT:
631 case MLD_LISTENER_DONE:
632 case MLDV2_LISTENER_REPORT:
633 /*
634 * Drop MLD traffic which is not link-local, has a hop limit
635 * of greater than 1 hop, or which does not have the
636 * IPv6 HBH Router Alert option.
637 * As IPv6 HBH options are stripped in ip6_input() we must
638 * check an mbuf header flag.
639 * XXX Should we also sanity check that these messages
640 * were directed to a link-local multicast prefix?
641 */
642 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
643 goto freeit;
644 if (mld_input(&m, off, icmp6len) != 0) {
645 *mp = NULL;
646 return (IPPROTO_DONE);
647 }
648 /* m stays. */
649 break;
650
651 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
652 {
653 enum { WRU, FQDN } mode;
654 struct prison *pr;
655
656 if (!V_icmp6_nodeinfo)
657 break;
658
659 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
660 mode = WRU;
661 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
662 mode = FQDN;
663 else
664 goto badlen;
665
666 pr = NULL;
667 sx_slock(&allprison_lock);
668 TAILQ_FOREACH(pr, &allprison, pr_list)
669 if (pr->pr_vnet == ifp->if_vnet)
670 break;
671 sx_sunlock(&allprison_lock);
672 if (pr == NULL)
673 pr = curthread->td_ucred->cr_prison;
674 if (mode == FQDN) {
675 if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
676 m = m_pullup(m, off +
677 sizeof(struct icmp6_nodeinfo));
678 if (m == NULL) {
679 IP6STAT_INC(ip6s_exthdrtoolong);
680 *mp = m;
681 return (IPPROTO_DONE);
682 }
683 }
684 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
685 if (n)
686 n = ni6_input(n, off, pr);
687 /* XXX meaningless if n == NULL */
688 noff = sizeof(struct ip6_hdr);
689 } else {
690 u_char *p;
691 int maxhlen, hlen;
692
693 /*
694 * XXX: this combination of flags is pointless,
695 * but should we keep this for compatibility?
696 */
697 if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
698 ICMP6_NODEINFO_TMPADDROK)) !=
699 (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
700 break;
701
702 if (code != 0)
703 goto badcode;
704
705 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
706 n = m_gethdr(M_NOWAIT, m->m_type);
707 if (n == NULL) {
708 /* Give up remote */
709 break;
710 }
711 if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
712 /*
713 * Previous code did a blind M_COPY_PKTHDR
714 * and said "just for rcvif". If true, then
715 * we could tolerate the dup failing (due to
716 * the deep copy of the tag chain). For now
717 * be conservative and just fail.
718 */
719 m_free(n);
720 n = NULL;
721 break;
722 }
723 /*
724 * Copy IPv6 and ICMPv6 only.
725 */
726 nip6 = mtod(n, struct ip6_hdr *);
727 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
728 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
729 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
730 p = (u_char *)(nicmp6 + 1);
731 bzero(p, 4);
732
733 maxhlen = M_TRAILINGSPACE(n) -
734 (sizeof(*nip6) + sizeof(*nicmp6) + 4);
735 mtx_lock(&pr->pr_mtx);
736 hlen = strlen(pr->pr_hostname);
737 if (maxhlen > hlen)
738 maxhlen = hlen;
739 /* meaningless TTL */
740 bcopy(pr->pr_hostname, p + 4, maxhlen);
741 mtx_unlock(&pr->pr_mtx);
742 noff = sizeof(struct ip6_hdr);
743 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
744 sizeof(struct icmp6_hdr) + 4 + maxhlen;
745 nicmp6->icmp6_type = ICMP6_WRUREPLY;
746 nicmp6->icmp6_code = 0;
747 }
748 if (n) {
749 ICMP6STAT_INC(icp6s_reflect);
750 ICMP6STAT_INC2(icp6s_outhist, ICMP6_WRUREPLY);
751 icmp6_reflect(n, noff);
752 }
753 break;
754 }
755
756 case ICMP6_WRUREPLY:
757 if (code != 0)
758 goto badcode;
759 break;
760
761 case ND_ROUTER_SOLICIT:
762 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
763 if (code != 0)
764 goto badcode;
765 if (icmp6len < sizeof(struct nd_router_solicit))
766 goto badlen;
767 if (send_sendso_input_hook != NULL) {
768 if (m->m_len < off + icmp6len) {
769 m = m_pullup(m, off + icmp6len);
770 if (m == NULL) {
771 IP6STAT_INC(ip6s_exthdrtoolong);
772 *mp = NULL;
773 return (IPPROTO_DONE);
774 }
775 }
776 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
777 if (error == 0) {
778 m = NULL;
779 goto freeit;
780 }
781 }
782 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
783 nd6_rs_input(m, off, icmp6len);
784 m = n;
785 if (m == NULL)
786 goto freeit;
787 break;
788
789 case ND_ROUTER_ADVERT:
790 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
791 if (code != 0)
792 goto badcode;
793 if (icmp6len < sizeof(struct nd_router_advert))
794 goto badlen;
795 if (send_sendso_input_hook != NULL) {
796 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
797 if (error == 0) {
798 m = NULL;
799 goto freeit;
800 }
801 }
802 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
803 nd6_ra_input(m, off, icmp6len);
804 m = n;
805 if (m == NULL)
806 goto freeit;
807 break;
808
809 case ND_NEIGHBOR_SOLICIT:
810 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
811 if (code != 0)
812 goto badcode;
813 if (icmp6len < sizeof(struct nd_neighbor_solicit))
814 goto badlen;
815 if (send_sendso_input_hook != NULL) {
816 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
817 if (error == 0) {
818 m = NULL;
819 goto freeit;
820 }
821 }
822 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
823 nd6_ns_input(m, off, icmp6len);
824 m = n;
825 if (m == NULL)
826 goto freeit;
827 break;
828
829 case ND_NEIGHBOR_ADVERT:
830 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
831 if (code != 0)
832 goto badcode;
833 if (icmp6len < sizeof(struct nd_neighbor_advert))
834 goto badlen;
835 if (send_sendso_input_hook != NULL) {
836 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
837 if (error == 0) {
838 m = NULL;
839 goto freeit;
840 }
841 }
842 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
843 nd6_na_input(m, off, icmp6len);
844 m = n;
845 if (m == NULL)
846 goto freeit;
847 break;
848
849 case ND_REDIRECT:
850 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
851 if (code != 0)
852 goto badcode;
853 if (icmp6len < sizeof(struct nd_redirect))
854 goto badlen;
855 if (send_sendso_input_hook != NULL) {
856 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
857 if (error == 0) {
858 m = NULL;
859 goto freeit;
860 }
861 }
862 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
863 icmp6_redirect_input(m, off);
864 m = n;
865 if (m == NULL)
866 goto freeit;
867 break;
868
869 case ICMP6_ROUTER_RENUMBERING:
870 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
871 code != ICMP6_ROUTER_RENUMBERING_RESULT)
872 goto badcode;
873 if (icmp6len < sizeof(struct icmp6_router_renum))
874 goto badlen;
875 break;
876
877 default:
878 nd6log((LOG_DEBUG,
879 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
880 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
881 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
882 ifp ? ifp->if_index : 0));
883 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
884 /* ICMPv6 error: MUST deliver it by spec... */
885 goto deliver;
886 } else {
887 /* ICMPv6 informational: MUST not deliver */
888 break;
889 }
890 deliver:
891 if (icmp6_notify_error(&m, off, icmp6len) != 0) {
892 /* In this case, m should've been freed. */
893 *mp = NULL;
894 return (IPPROTO_DONE);
895 }
896 break;
897
898 badcode:
899 ICMP6STAT_INC(icp6s_badcode);
900 break;
901
902 badlen:
903 ICMP6STAT_INC(icp6s_badlen);
904 break;
905 }
906
907 /* deliver the packet to appropriate sockets */
908 icmp6_rip6_input(&m, *offp);
909
910 *mp = m;
911 return (IPPROTO_DONE);
912
913 freeit:
914 m_freem(m);
915 *mp = NULL;
916 return (IPPROTO_DONE);
917 }
918
919 static int
icmp6_notify_error(struct mbuf ** mp,int off,int icmp6len)920 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len)
921 {
922 struct mbuf *m;
923 struct icmp6_hdr *icmp6;
924 struct ip6_hdr *eip6;
925 u_int32_t notifymtu;
926 struct sockaddr_in6 icmp6src, icmp6dst;
927
928 m = *mp;
929
930 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
931 ICMP6STAT_INC(icp6s_tooshort);
932 goto freeit;
933 }
934
935 if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
936 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
937 if (m == NULL) {
938 IP6STAT_INC(ip6s_exthdrtoolong);
939 *mp = m;
940 return (-1);
941 }
942 }
943 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
944 eip6 = (struct ip6_hdr *)(icmp6 + 1);
945 bzero(&icmp6dst, sizeof(icmp6dst));
946
947 /* Detect the upper level protocol */
948 {
949 u_int8_t nxt = eip6->ip6_nxt;
950 int eoff = off + sizeof(struct icmp6_hdr) +
951 sizeof(struct ip6_hdr);
952 struct ip6ctlparam ip6cp;
953 int icmp6type = icmp6->icmp6_type;
954 struct ip6_frag *fh;
955 struct ip6_rthdr *rth;
956 struct ip6_rthdr0 *rth0;
957 int rthlen;
958
959 while (1) { /* XXX: should avoid infinite loop explicitly? */
960 struct ip6_ext *eh;
961
962 switch (nxt) {
963 case IPPROTO_HOPOPTS:
964 case IPPROTO_DSTOPTS:
965 case IPPROTO_AH:
966 if (m->m_len < eoff + sizeof(struct ip6_ext)) {
967 m = m_pullup(m, eoff +
968 sizeof(struct ip6_ext));
969 if (m == NULL) {
970 IP6STAT_INC(ip6s_exthdrtoolong);
971 *mp = m;
972 return (-1);
973 }
974 }
975 eh = (struct ip6_ext *)
976 (mtod(m, caddr_t) + eoff);
977 if (nxt == IPPROTO_AH)
978 eoff += (eh->ip6e_len + 2) << 2;
979 else
980 eoff += (eh->ip6e_len + 1) << 3;
981 nxt = eh->ip6e_nxt;
982 break;
983 case IPPROTO_ROUTING:
984 /*
985 * When the erroneous packet contains a
986 * routing header, we should examine the
987 * header to determine the final destination.
988 * Otherwise, we can't properly update
989 * information that depends on the final
990 * destination (e.g. path MTU).
991 */
992 if (m->m_len < eoff + sizeof(*rth)) {
993 m = m_pullup(m, eoff + sizeof(*rth));
994 if (m == NULL) {
995 IP6STAT_INC(ip6s_exthdrtoolong);
996 *mp = m;
997 return (-1);
998 }
999 }
1000 rth = (struct ip6_rthdr *)
1001 (mtod(m, caddr_t) + eoff);
1002 rthlen = (rth->ip6r_len + 1) << 3;
1003 /*
1004 * XXX: currently there is no
1005 * officially defined type other
1006 * than type-0.
1007 * Note that if the segment left field
1008 * is 0, all intermediate hops must
1009 * have been passed.
1010 */
1011 if (rth->ip6r_segleft &&
1012 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1013 int hops;
1014
1015 if (m->m_len < eoff + rthlen) {
1016 m = m_pullup(m, eoff + rthlen);
1017 if (m == NULL) {
1018 IP6STAT_INC(
1019 ip6s_exthdrtoolong);
1020 *mp = m;
1021 return (-1);
1022 }
1023 }
1024 rth0 = (struct ip6_rthdr0 *)
1025 (mtod(m, caddr_t) + eoff);
1026
1027 /* just ignore a bogus header */
1028 if ((rth0->ip6r0_len % 2) == 0 &&
1029 (hops = rth0->ip6r0_len/2))
1030 icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1031 }
1032 eoff += rthlen;
1033 nxt = rth->ip6r_nxt;
1034 break;
1035 case IPPROTO_FRAGMENT:
1036 if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1037 m = m_pullup(m, eoff +
1038 sizeof(struct ip6_frag));
1039 if (m == NULL) {
1040 IP6STAT_INC(ip6s_exthdrtoolong);
1041 *mp = m;
1042 return (-1);
1043 }
1044 }
1045 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1046 eoff);
1047 /*
1048 * Data after a fragment header is meaningless
1049 * unless it is the first fragment, but
1050 * we'll go to the notify label for path MTU
1051 * discovery.
1052 */
1053 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1054 goto notify;
1055
1056 eoff += sizeof(struct ip6_frag);
1057 nxt = fh->ip6f_nxt;
1058 break;
1059 default:
1060 /*
1061 * This case includes ESP and the No Next
1062 * Header. In such cases going to the notify
1063 * label does not have any meaning
1064 * (i.e. ctlfunc will be NULL), but we go
1065 * anyway since we might have to update
1066 * path MTU information.
1067 */
1068 goto notify;
1069 }
1070 }
1071 notify:
1072 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1073
1074 /*
1075 * retrieve parameters from the inner IPv6 header, and convert
1076 * them into sockaddr structures.
1077 * XXX: there is no guarantee that the source or destination
1078 * addresses of the inner packet are in the same scope as
1079 * the addresses of the icmp packet. But there is no other
1080 * way to determine the zone.
1081 */
1082 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1083
1084 /*
1085 * Protocol layers can't do anything useful with unspecified
1086 * addresses.
1087 */
1088 if (IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_src) ||
1089 IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_dst))
1090 goto freeit;
1091
1092 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1093 icmp6dst.sin6_family = AF_INET6;
1094 if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1095 icmp6dst.sin6_addr = eip6->ip6_dst;
1096 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1097 goto freeit;
1098 bzero(&icmp6src, sizeof(icmp6src));
1099 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1100 icmp6src.sin6_family = AF_INET6;
1101 icmp6src.sin6_addr = eip6->ip6_src;
1102 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1103 goto freeit;
1104 icmp6src.sin6_flowinfo =
1105 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1106
1107 ip6cp.ip6c_m = m;
1108 ip6cp.ip6c_icmp6 = icmp6;
1109 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1110 ip6cp.ip6c_off = eoff;
1111 ip6cp.ip6c_finaldst = &icmp6dst;
1112 ip6cp.ip6c_src = &icmp6src;
1113 ip6cp.ip6c_nxt = nxt;
1114
1115 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1116 notifymtu = ntohl(icmp6->icmp6_mtu);
1117 ip6cp.ip6c_cmdarg = (void *)¬ifymtu;
1118 icmp6_mtudisc_update(&ip6cp);
1119 }
1120
1121 if (ip6_ctlprotox[nxt] != NULL)
1122 ip6_ctlprotox[nxt](&ip6cp);
1123 }
1124 *mp = m;
1125 return (0);
1126
1127 freeit:
1128 m_freem(m);
1129 *mp = NULL;
1130 return (-1);
1131 }
1132
1133 static void
icmp6_mtudisc_update(struct ip6ctlparam * ip6cp)1134 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp)
1135 {
1136 struct in6_addr *dst = &ip6cp->ip6c_finaldst->sin6_addr;
1137 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1138 struct mbuf *m = ip6cp->ip6c_m;
1139 u_int mtu = ntohl(icmp6->icmp6_mtu);
1140 struct in_conninfo inc;
1141 uint32_t max_mtu;
1142
1143 if (mtu < IPV6_MMTU)
1144 return;
1145
1146 bzero(&inc, sizeof(inc));
1147 inc.inc_fibnum = M_GETFIB(m);
1148 inc.inc_flags |= INC_ISIPV6;
1149 inc.inc6_faddr = *dst;
1150 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1151 return;
1152
1153 max_mtu = tcp_hc_getmtu(&inc);
1154 if (max_mtu == 0)
1155 max_mtu = tcp_maxmtu6(&inc, NULL);
1156
1157 if (mtu < max_mtu) {
1158 tcp_hc_updatemtu(&inc, mtu);
1159 ICMP6STAT_INC(icp6s_pmtuchg);
1160 }
1161 }
1162
1163 /*
1164 * Process a Node Information Query packet, based on
1165 * draft-ietf-ipngwg-icmp-name-lookups-07.
1166 *
1167 * Spec incompatibilities:
1168 * - IPv6 Subject address handling
1169 * - IPv4 Subject address handling support missing
1170 * - Proxy reply (answer even if it's not for me)
1171 * - joins NI group address at in6_ifattach() time only, does not cope
1172 * with hostname changes by sethostname(3)
1173 */
1174 static struct mbuf *
ni6_input(struct mbuf * m,int off,struct prison * pr)1175 ni6_input(struct mbuf *m, int off, struct prison *pr)
1176 {
1177 struct icmp6_nodeinfo *ni6, *nni6;
1178 struct mbuf *n = NULL;
1179 u_int16_t qtype;
1180 int subjlen;
1181 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1182 struct ni_reply_fqdn *fqdn;
1183 int addrs; /* for NI_QTYPE_NODEADDR */
1184 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1185 struct in6_addr in6_subj; /* subject address */
1186 struct ip6_hdr *ip6;
1187 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1188 char *subj = NULL;
1189 struct in6_ifaddr *ia6 = NULL;
1190
1191 ip6 = mtod(m, struct ip6_hdr *);
1192 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1193
1194 /*
1195 * Validate IPv6 source address.
1196 * The default configuration MUST be to refuse answering queries from
1197 * global-scope addresses according to RFC4602.
1198 * Notes:
1199 * - it's not very clear what "refuse" means; this implementation
1200 * simply drops it.
1201 * - it's not very easy to identify global-scope (unicast) addresses
1202 * since there are many prefixes for them. It should be safer
1203 * and in practice sufficient to check "all" but loopback and
1204 * link-local (note that site-local unicast was deprecated and
1205 * ULA is defined as global scope-wise)
1206 */
1207 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1208 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1209 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1210 goto bad;
1211
1212 /*
1213 * Validate IPv6 destination address.
1214 *
1215 * The Responder must discard the Query without further processing
1216 * unless it is one of the Responder's unicast or anycast addresses, or
1217 * a link-local scope multicast address which the Responder has joined.
1218 * [RFC4602, Section 5.]
1219 */
1220 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1221 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1222 goto bad;
1223 /* else it's a link-local multicast, fine */
1224 } else { /* unicast or anycast */
1225 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1226 if (ia6 == NULL)
1227 goto bad; /* XXX impossible */
1228
1229 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1230 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1231 nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1232 "a temporary address in %s:%d",
1233 __FILE__, __LINE__));
1234 goto bad;
1235 }
1236 }
1237
1238 /* validate query Subject field. */
1239 qtype = ntohs(ni6->ni_qtype);
1240 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1241 switch (qtype) {
1242 case NI_QTYPE_NOOP:
1243 case NI_QTYPE_SUPTYPES:
1244 /* 07 draft */
1245 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1246 break;
1247 /* FALLTHROUGH */
1248 case NI_QTYPE_FQDN:
1249 case NI_QTYPE_NODEADDR:
1250 case NI_QTYPE_IPV4ADDR:
1251 switch (ni6->ni_code) {
1252 case ICMP6_NI_SUBJ_IPV6:
1253 #if ICMP6_NI_SUBJ_IPV6 != 0
1254 case 0:
1255 #endif
1256 /*
1257 * backward compatibility - try to accept 03 draft
1258 * format, where no Subject is present.
1259 */
1260 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1261 subjlen == 0) {
1262 oldfqdn++;
1263 break;
1264 }
1265 #if ICMP6_NI_SUBJ_IPV6 != 0
1266 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1267 goto bad;
1268 #endif
1269
1270 if (subjlen != sizeof(struct in6_addr))
1271 goto bad;
1272
1273 /*
1274 * Validate Subject address.
1275 *
1276 * Not sure what exactly "address belongs to the node"
1277 * means in the spec, is it just unicast, or what?
1278 *
1279 * At this moment we consider Subject address as
1280 * "belong to the node" if the Subject address equals
1281 * to the IPv6 destination address; validation for
1282 * IPv6 destination address should have done enough
1283 * check for us.
1284 *
1285 * We do not do proxy at this moment.
1286 */
1287 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1288 subjlen, (caddr_t)&in6_subj);
1289 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1290 goto bad;
1291
1292 subj = (char *)&in6_subj;
1293 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1294 break;
1295
1296 /*
1297 * XXX if we are to allow other cases, we should really
1298 * be careful about scope here.
1299 * basically, we should disallow queries toward IPv6
1300 * destination X with subject Y,
1301 * if scope(X) > scope(Y).
1302 * if we allow scope(X) > scope(Y), it will result in
1303 * information leakage across scope boundary.
1304 */
1305 goto bad;
1306
1307 case ICMP6_NI_SUBJ_FQDN:
1308 /*
1309 * Validate Subject name with gethostname(3).
1310 *
1311 * The behavior may need some debate, since:
1312 * - we are not sure if the node has FQDN as
1313 * hostname (returned by gethostname(3)).
1314 * - the code does wildcard match for truncated names.
1315 * however, we are not sure if we want to perform
1316 * wildcard match, if gethostname(3) side has
1317 * truncated hostname.
1318 */
1319 mtx_lock(&pr->pr_mtx);
1320 n = ni6_nametodns(pr->pr_hostname,
1321 strlen(pr->pr_hostname), 0);
1322 mtx_unlock(&pr->pr_mtx);
1323 if (!n || n->m_next || n->m_len == 0)
1324 goto bad;
1325 if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1326 subjlen) {
1327 m = m_pullup(m, off +
1328 sizeof(struct icmp6_nodeinfo) + subjlen);
1329 if (m == NULL) {
1330 IP6STAT_INC(ip6s_exthdrtoolong);
1331 goto bad;
1332 }
1333 }
1334 /* ip6 possibly invalid but not used after. */
1335 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1336 subj = (char *)(mtod(m, caddr_t) + off +
1337 sizeof(struct icmp6_nodeinfo));
1338 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1339 n->m_len)) {
1340 goto bad;
1341 }
1342 m_freem(n);
1343 n = NULL;
1344 break;
1345
1346 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1347 default:
1348 goto bad;
1349 }
1350 break;
1351 }
1352
1353 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1354 switch (qtype) {
1355 case NI_QTYPE_FQDN:
1356 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1357 goto bad;
1358 break;
1359 case NI_QTYPE_NODEADDR:
1360 case NI_QTYPE_IPV4ADDR:
1361 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1362 goto bad;
1363 break;
1364 }
1365
1366 /* guess reply length */
1367 switch (qtype) {
1368 case NI_QTYPE_NOOP:
1369 break; /* no reply data */
1370 case NI_QTYPE_SUPTYPES:
1371 replylen += sizeof(u_int32_t);
1372 break;
1373 case NI_QTYPE_FQDN:
1374 /* XXX will append an mbuf */
1375 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1376 break;
1377 case NI_QTYPE_NODEADDR:
1378 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1379 if ((replylen += addrs * (sizeof(struct in6_addr) +
1380 sizeof(u_int32_t))) > MCLBYTES)
1381 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1382 break;
1383 case NI_QTYPE_IPV4ADDR:
1384 /* unsupported - should respond with unknown Qtype? */
1385 break;
1386 default:
1387 /*
1388 * XXX: We must return a reply with the ICMP6 code
1389 * `unknown Qtype' in this case. However we regard the case
1390 * as an FQDN query for backward compatibility.
1391 * Older versions set a random value to this field,
1392 * so it rarely varies in the defined qtypes.
1393 * But the mechanism is not reliable...
1394 * maybe we should obsolete older versions.
1395 */
1396 qtype = NI_QTYPE_FQDN;
1397 /* XXX will append an mbuf */
1398 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1399 oldfqdn++;
1400 break;
1401 }
1402
1403 /* Allocate an mbuf to reply. */
1404 if (replylen > MCLBYTES) {
1405 /*
1406 * XXX: should we try to allocate more? But MCLBYTES
1407 * is probably much larger than IPV6_MMTU...
1408 */
1409 goto bad;
1410 }
1411 if (replylen > MHLEN)
1412 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1413 else
1414 n = m_gethdr(M_NOWAIT, m->m_type);
1415 if (n == NULL) {
1416 m_freem(m);
1417 return (NULL);
1418 }
1419 m_move_pkthdr(n, m); /* just for recvif and FIB */
1420 n->m_pkthdr.len = n->m_len = replylen;
1421
1422 /* copy mbuf header and IPv6 + Node Information base headers */
1423 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1424 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1425 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1426
1427 /* qtype dependent procedure */
1428 switch (qtype) {
1429 case NI_QTYPE_NOOP:
1430 nni6->ni_code = ICMP6_NI_SUCCESS;
1431 nni6->ni_flags = 0;
1432 break;
1433 case NI_QTYPE_SUPTYPES:
1434 {
1435 u_int32_t v;
1436 nni6->ni_code = ICMP6_NI_SUCCESS;
1437 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1438 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1439 v = (u_int32_t)htonl(0x0000000f);
1440 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1441 break;
1442 }
1443 case NI_QTYPE_FQDN:
1444 nni6->ni_code = ICMP6_NI_SUCCESS;
1445 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1446 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1447 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1448 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1449 /*
1450 * XXX do we really have FQDN in hostname?
1451 */
1452 mtx_lock(&pr->pr_mtx);
1453 n->m_next = ni6_nametodns(pr->pr_hostname,
1454 strlen(pr->pr_hostname), oldfqdn);
1455 mtx_unlock(&pr->pr_mtx);
1456 if (n->m_next == NULL)
1457 goto bad;
1458 /* XXX we assume that n->m_next is not a chain */
1459 if (n->m_next->m_next != NULL)
1460 goto bad;
1461 n->m_pkthdr.len += n->m_next->m_len;
1462 break;
1463 case NI_QTYPE_NODEADDR:
1464 {
1465 int lenlim, copied;
1466
1467 nni6->ni_code = ICMP6_NI_SUCCESS;
1468 n->m_pkthdr.len = n->m_len =
1469 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1470 lenlim = M_TRAILINGSPACE(n);
1471 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1472 /* XXX: reset mbuf length */
1473 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1474 sizeof(struct icmp6_nodeinfo) + copied;
1475 break;
1476 }
1477 default:
1478 break; /* XXX impossible! */
1479 }
1480
1481 nni6->ni_type = ICMP6_NI_REPLY;
1482 m_freem(m);
1483 return (n);
1484
1485 bad:
1486 m_freem(m);
1487 if (n)
1488 m_freem(n);
1489 return (NULL);
1490 }
1491
1492 /*
1493 * make a mbuf with DNS-encoded string. no compression support.
1494 *
1495 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1496 * treated as truncated name (two \0 at the end). this is a wild guess.
1497 *
1498 * old - return pascal string if non-zero
1499 */
1500 static struct mbuf *
ni6_nametodns(const char * name,int namelen,int old)1501 ni6_nametodns(const char *name, int namelen, int old)
1502 {
1503 struct mbuf *m;
1504 char *cp, *ep;
1505 const char *p, *q;
1506 int i, len, nterm;
1507
1508 if (old)
1509 len = namelen + 1;
1510 else
1511 len = MCLBYTES;
1512
1513 /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1514 if (len > MLEN)
1515 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1516 else
1517 m = m_get(M_NOWAIT, MT_DATA);
1518 if (m == NULL)
1519 goto fail;
1520
1521 if (old) {
1522 m->m_len = len;
1523 *mtod(m, char *) = namelen;
1524 bcopy(name, mtod(m, char *) + 1, namelen);
1525 return m;
1526 } else {
1527 m->m_len = 0;
1528 cp = mtod(m, char *);
1529 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1530
1531 /* if not certain about my name, return empty buffer */
1532 if (namelen == 0)
1533 return m;
1534
1535 /*
1536 * guess if it looks like shortened hostname, or FQDN.
1537 * shortened hostname needs two trailing "\0".
1538 */
1539 i = 0;
1540 for (p = name; p < name + namelen; p++) {
1541 if (*p && *p == '.')
1542 i++;
1543 }
1544 if (i < 2)
1545 nterm = 2;
1546 else
1547 nterm = 1;
1548
1549 p = name;
1550 while (cp < ep && p < name + namelen) {
1551 i = 0;
1552 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1553 i++;
1554 /* result does not fit into mbuf */
1555 if (cp + i + 1 >= ep)
1556 goto fail;
1557 /*
1558 * DNS label length restriction, RFC1035 page 8.
1559 * "i == 0" case is included here to avoid returning
1560 * 0-length label on "foo..bar".
1561 */
1562 if (i <= 0 || i >= 64)
1563 goto fail;
1564 *cp++ = i;
1565 bcopy(p, cp, i);
1566 cp += i;
1567 p = q;
1568 if (p < name + namelen && *p == '.')
1569 p++;
1570 }
1571 /* termination */
1572 if (cp + nterm >= ep)
1573 goto fail;
1574 while (nterm-- > 0)
1575 *cp++ = '\0';
1576 m->m_len = cp - mtod(m, char *);
1577 return m;
1578 }
1579
1580 panic("should not reach here");
1581 /* NOTREACHED */
1582
1583 fail:
1584 if (m)
1585 m_freem(m);
1586 return NULL;
1587 }
1588
1589 /*
1590 * check if two DNS-encoded string matches. takes care of truncated
1591 * form (with \0\0 at the end). no compression support.
1592 * XXX upper/lowercase match (see RFC2065)
1593 */
1594 static int
ni6_dnsmatch(const char * a,int alen,const char * b,int blen)1595 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1596 {
1597 const char *a0, *b0;
1598 int l;
1599
1600 /* simplest case - need validation? */
1601 if (alen == blen && bcmp(a, b, alen) == 0)
1602 return 1;
1603
1604 a0 = a;
1605 b0 = b;
1606
1607 /* termination is mandatory */
1608 if (alen < 2 || blen < 2)
1609 return 0;
1610 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1611 return 0;
1612 alen--;
1613 blen--;
1614
1615 while (a - a0 < alen && b - b0 < blen) {
1616 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1617 return 0;
1618
1619 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1620 return 0;
1621 /* we don't support compression yet */
1622 if (a[0] >= 64 || b[0] >= 64)
1623 return 0;
1624
1625 /* truncated case */
1626 if (a[0] == 0 && a - a0 == alen - 1)
1627 return 1;
1628 if (b[0] == 0 && b - b0 == blen - 1)
1629 return 1;
1630 if (a[0] == 0 || b[0] == 0)
1631 return 0;
1632
1633 if (a[0] != b[0])
1634 return 0;
1635 l = a[0];
1636 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1637 return 0;
1638 if (bcmp(a + 1, b + 1, l) != 0)
1639 return 0;
1640
1641 a += 1 + l;
1642 b += 1 + l;
1643 }
1644
1645 if (a - a0 == alen && b - b0 == blen)
1646 return 1;
1647 else
1648 return 0;
1649 }
1650
1651 /*
1652 * calculate the number of addresses to be returned in the node info reply.
1653 */
1654 static int
ni6_addrs(struct icmp6_nodeinfo * ni6,struct mbuf * m,struct ifnet ** ifpp,struct in6_addr * subj)1655 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1656 struct in6_addr *subj)
1657 {
1658 struct ifnet *ifp;
1659 struct in6_ifaddr *ifa6;
1660 struct ifaddr *ifa;
1661 int addrs = 0, addrsofif, iffound = 0;
1662 int niflags = ni6->ni_flags;
1663
1664 NET_EPOCH_ASSERT();
1665
1666 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1667 switch (ni6->ni_code) {
1668 case ICMP6_NI_SUBJ_IPV6:
1669 if (subj == NULL) /* must be impossible... */
1670 return (0);
1671 break;
1672 default:
1673 /*
1674 * XXX: we only support IPv6 subject address for
1675 * this Qtype.
1676 */
1677 return (0);
1678 }
1679 }
1680
1681 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1682 addrsofif = 0;
1683 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1684 if (ifa->ifa_addr->sa_family != AF_INET6)
1685 continue;
1686 ifa6 = (struct in6_ifaddr *)ifa;
1687
1688 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1689 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1690 iffound = 1;
1691
1692 /*
1693 * IPv4-mapped addresses can only be returned by a
1694 * Node Information proxy, since they represent
1695 * addresses of IPv4-only nodes, which perforce do
1696 * not implement this protocol.
1697 * [icmp-name-lookups-07, Section 5.4]
1698 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1699 * this function at this moment.
1700 */
1701
1702 /* What do we have to do about ::1? */
1703 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1704 case IPV6_ADDR_SCOPE_LINKLOCAL:
1705 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1706 continue;
1707 break;
1708 case IPV6_ADDR_SCOPE_SITELOCAL:
1709 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1710 continue;
1711 break;
1712 case IPV6_ADDR_SCOPE_GLOBAL:
1713 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1714 continue;
1715 break;
1716 default:
1717 continue;
1718 }
1719
1720 /*
1721 * check if anycast is okay.
1722 * XXX: just experimental. not in the spec.
1723 */
1724 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1725 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1726 continue; /* we need only unicast addresses */
1727 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1728 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1729 continue;
1730 }
1731 addrsofif++; /* count the address */
1732 }
1733 if (iffound) {
1734 *ifpp = ifp;
1735 return (addrsofif);
1736 }
1737
1738 addrs += addrsofif;
1739 }
1740
1741 return (addrs);
1742 }
1743
1744 static int
ni6_store_addrs(struct icmp6_nodeinfo * ni6,struct icmp6_nodeinfo * nni6,struct ifnet * ifp0,int resid)1745 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1746 struct ifnet *ifp0, int resid)
1747 {
1748 struct ifnet *ifp;
1749 struct in6_ifaddr *ifa6;
1750 struct ifaddr *ifa;
1751 struct ifnet *ifp_dep = NULL;
1752 int copied = 0, allow_deprecated = 0;
1753 u_char *cp = (u_char *)(nni6 + 1);
1754 int niflags = ni6->ni_flags;
1755 u_int32_t ltime;
1756
1757 NET_EPOCH_ASSERT();
1758
1759 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1760 return (0); /* needless to copy */
1761
1762 ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1763 again:
1764
1765 for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1766 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1767 if (ifa->ifa_addr->sa_family != AF_INET6)
1768 continue;
1769 ifa6 = (struct in6_ifaddr *)ifa;
1770
1771 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1772 allow_deprecated == 0) {
1773 /*
1774 * prefererred address should be put before
1775 * deprecated addresses.
1776 */
1777
1778 /* record the interface for later search */
1779 if (ifp_dep == NULL)
1780 ifp_dep = ifp;
1781
1782 continue;
1783 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1784 allow_deprecated != 0)
1785 continue; /* we now collect deprecated addrs */
1786
1787 /* What do we have to do about ::1? */
1788 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1789 case IPV6_ADDR_SCOPE_LINKLOCAL:
1790 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1791 continue;
1792 break;
1793 case IPV6_ADDR_SCOPE_SITELOCAL:
1794 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1795 continue;
1796 break;
1797 case IPV6_ADDR_SCOPE_GLOBAL:
1798 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1799 continue;
1800 break;
1801 default:
1802 continue;
1803 }
1804
1805 /*
1806 * check if anycast is okay.
1807 * XXX: just experimental. not in the spec.
1808 */
1809 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1810 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1811 continue;
1812 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1813 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1814 continue;
1815 }
1816
1817 /* now we can copy the address */
1818 if (resid < sizeof(struct in6_addr) +
1819 sizeof(u_int32_t)) {
1820 /*
1821 * We give up much more copy.
1822 * Set the truncate flag and return.
1823 */
1824 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1825 return (copied);
1826 }
1827
1828 /*
1829 * Set the TTL of the address.
1830 * The TTL value should be one of the following
1831 * according to the specification:
1832 *
1833 * 1. The remaining lifetime of a DHCP lease on the
1834 * address, or
1835 * 2. The remaining Valid Lifetime of a prefix from
1836 * which the address was derived through Stateless
1837 * Autoconfiguration.
1838 *
1839 * Note that we currently do not support stateful
1840 * address configuration by DHCPv6, so the former
1841 * case can't happen.
1842 */
1843 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1844 ltime = ND6_INFINITE_LIFETIME;
1845 else {
1846 if (ifa6->ia6_lifetime.ia6t_expire >
1847 time_uptime)
1848 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1849 else
1850 ltime = 0;
1851 }
1852
1853 bcopy(<ime, cp, sizeof(u_int32_t));
1854 cp += sizeof(u_int32_t);
1855
1856 /* copy the address itself */
1857 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1858 sizeof(struct in6_addr));
1859 in6_clearscope((struct in6_addr *)cp); /* XXX */
1860 cp += sizeof(struct in6_addr);
1861
1862 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1863 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1864 }
1865 if (ifp0) /* we need search only on the specified IF */
1866 break;
1867 }
1868
1869 if (allow_deprecated == 0 && ifp_dep != NULL) {
1870 ifp = ifp_dep;
1871 allow_deprecated = 1;
1872
1873 goto again;
1874 }
1875
1876 return (copied);
1877 }
1878
1879 static bool
icmp6_rip6_match(const struct inpcb * inp,void * v)1880 icmp6_rip6_match(const struct inpcb *inp, void *v)
1881 {
1882 struct ip6_hdr *ip6 = v;
1883
1884 if ((inp->inp_vflag & INP_IPV6) == 0)
1885 return (false);
1886 if (inp->inp_ip_p != IPPROTO_ICMPV6)
1887 return (false);
1888 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1889 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1890 return (false);
1891 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1892 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1893 return (false);
1894 return (true);
1895 }
1896
1897 /*
1898 * XXX almost dup'ed code with rip6_input.
1899 */
1900 static int
icmp6_rip6_input(struct mbuf ** mp,int off)1901 icmp6_rip6_input(struct mbuf **mp, int off)
1902 {
1903 struct mbuf *n, *m = *mp;
1904 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1905 struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1906 INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1907 struct inpcb *inp;
1908 struct sockaddr_in6 fromsa;
1909 struct icmp6_hdr *icmp6;
1910 struct mbuf *opts = NULL;
1911 int delivered = 0, fib;
1912
1913 /* This is assumed to be safe; icmp6_input() does a pullup. */
1914 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1915
1916 /*
1917 * XXX: the address may have embedded scope zone ID, which should be
1918 * hidden from applications.
1919 */
1920 bzero(&fromsa, sizeof(fromsa));
1921 fromsa.sin6_family = AF_INET6;
1922 fromsa.sin6_len = sizeof(struct sockaddr_in6);
1923 fromsa.sin6_addr = ip6->ip6_src;
1924 if (sa6_recoverscope(&fromsa)) {
1925 m_freem(m);
1926 *mp = NULL;
1927 return (IPPROTO_DONE);
1928 }
1929
1930 fib = M_GETFIB(m);
1931
1932 while ((inp = inp_next(&inpi)) != NULL) {
1933 if (V_rip_bind_all_fibs == 0 && fib != inp->inp_inc.inc_fibnum)
1934 /*
1935 * Sockets bound to a specific FIB can only receive
1936 * packets from that FIB.
1937 */
1938 continue;
1939 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1940 inp->in6p_icmp6filt))
1941 continue;
1942 /*
1943 * Recent network drivers tend to allocate a single
1944 * mbuf cluster, rather than to make a couple of
1945 * mbufs without clusters. Also, since the IPv6 code
1946 * path tries to avoid m_pullup(), it is highly
1947 * probable that we still have an mbuf cluster here
1948 * even though the necessary length can be stored in an
1949 * mbuf's internal buffer.
1950 * Meanwhile, the default size of the receive socket
1951 * buffer for raw sockets is not so large. This means
1952 * the possibility of packet loss is relatively higher
1953 * than before. To avoid this scenario, we copy the
1954 * received data to a separate mbuf that does not use
1955 * a cluster, if possible.
1956 * XXX: it is better to copy the data after stripping
1957 * intermediate headers.
1958 */
1959 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1960 m->m_len <= MHLEN) {
1961 n = m_get(M_NOWAIT, m->m_type);
1962 if (n != NULL) {
1963 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1964 bcopy(m->m_data, n->m_data, m->m_len);
1965 n->m_len = m->m_len;
1966 } else {
1967 m_free(n);
1968 n = NULL;
1969 }
1970 }
1971 } else
1972 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1973 if (n == NULL)
1974 continue;
1975 if (inp->inp_flags & INP_CONTROLOPTS)
1976 ip6_savecontrol(inp, n, &opts);
1977 /* strip intermediate headers */
1978 m_adj(n, off);
1979 SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1980 if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1981 (struct sockaddr *)&fromsa, n, opts) == 0) {
1982 soroverflow_locked(inp->inp_socket);
1983 m_freem(n);
1984 if (opts)
1985 m_freem(opts);
1986 } else {
1987 sorwakeup_locked(inp->inp_socket);
1988 delivered++;
1989 }
1990 opts = NULL;
1991 }
1992 m_freem(m);
1993 *mp = NULL;
1994 if (delivered == 0)
1995 IP6STAT_DEC(ip6s_delivered);
1996 return (IPPROTO_DONE);
1997 }
1998
1999 /*
2000 * Reflect the ip6 packet back to the source.
2001 * OFF points to the icmp6 header, counted from the top of the mbuf.
2002 */
2003 static void
icmp6_reflect(struct mbuf * m,size_t off)2004 icmp6_reflect(struct mbuf *m, size_t off)
2005 {
2006 struct in6_addr src6, *srcp;
2007 struct ip6_hdr *ip6;
2008 struct icmp6_hdr *icmp6;
2009 struct in6_ifaddr *ia = NULL;
2010 struct ifnet *outif = NULL;
2011 int plen;
2012 int type, code, hlim;
2013
2014 /* too short to reflect */
2015 if (off < sizeof(struct ip6_hdr)) {
2016 nd6log((LOG_DEBUG,
2017 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2018 (u_long)off, (u_long)sizeof(struct ip6_hdr),
2019 __FILE__, __LINE__));
2020 goto bad;
2021 }
2022
2023 /*
2024 * If there are extra headers between IPv6 and ICMPv6, strip
2025 * off that header first.
2026 */
2027 #ifdef DIAGNOSTIC
2028 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2029 panic("assumption failed in icmp6_reflect");
2030 #endif
2031 if (off > sizeof(struct ip6_hdr)) {
2032 size_t l;
2033 struct ip6_hdr nip6;
2034
2035 l = off - sizeof(struct ip6_hdr);
2036 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2037 m_adj(m, l);
2038 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2039 if (m->m_len < l) {
2040 if ((m = m_pullup(m, l)) == NULL)
2041 return;
2042 }
2043 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2044 } else /* off == sizeof(struct ip6_hdr) */ {
2045 size_t l;
2046 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2047 if (m->m_len < l) {
2048 if ((m = m_pullup(m, l)) == NULL)
2049 return;
2050 }
2051 }
2052 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2053 ip6 = mtod(m, struct ip6_hdr *);
2054 ip6->ip6_nxt = IPPROTO_ICMPV6;
2055 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2056 type = icmp6->icmp6_type; /* keep type for statistics */
2057 code = icmp6->icmp6_code; /* ditto. */
2058 hlim = 0;
2059 srcp = NULL;
2060
2061 if (__predict_false(IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src))) {
2062 nd6log((LOG_DEBUG,
2063 "icmp6_reflect: source address is unspecified\n"));
2064 goto bad;
2065 }
2066
2067 /*
2068 * If the incoming packet was addressed directly to us (i.e. unicast),
2069 * use dst as the src for the reply.
2070 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2071 * (for example) when we encounter an error while forwarding procedure
2072 * destined to a duplicated address of ours.
2073 */
2074 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2075 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2076 if (ia != NULL && !(ia->ia6_flags &
2077 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2078 src6 = ia->ia_addr.sin6_addr;
2079 srcp = &src6;
2080
2081 if (m->m_pkthdr.rcvif != NULL) {
2082 /* XXX: This may not be the outgoing interface */
2083 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2084 } else
2085 hlim = V_ip6_defhlim;
2086 }
2087 }
2088
2089 if (srcp == NULL) {
2090 int error;
2091 struct in6_addr dst6;
2092 uint32_t scopeid;
2093
2094 /*
2095 * This case matches to multicasts, our anycast, or unicasts
2096 * that we do not own. Select a source address based on the
2097 * source address of the erroneous packet.
2098 */
2099 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2100 error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2101 scopeid, NULL, &src6, &hlim);
2102
2103 if (error) {
2104 char ip6buf[INET6_ADDRSTRLEN];
2105 nd6log((LOG_DEBUG,
2106 "icmp6_reflect: source can't be determined: "
2107 "dst=%s, error=%d\n",
2108 ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2109 goto bad;
2110 }
2111 srcp = &src6;
2112 }
2113 /*
2114 * ip6_input() drops a packet if its src is multicast.
2115 * So, the src is never multicast.
2116 */
2117 ip6->ip6_dst = ip6->ip6_src;
2118 ip6->ip6_src = *srcp;
2119 ip6->ip6_flow = 0;
2120 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2121 ip6->ip6_vfc |= IPV6_VERSION;
2122 ip6->ip6_nxt = IPPROTO_ICMPV6;
2123 ip6->ip6_hlim = hlim;
2124
2125 icmp6->icmp6_cksum = 0;
2126 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2127 sizeof(struct ip6_hdr), plen);
2128
2129 /*
2130 * XXX option handling
2131 */
2132
2133 m->m_flags &= ~(M_BCAST|M_MCAST);
2134 m->m_pkthdr.rcvif = NULL;
2135 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2136 if (outif)
2137 icmp6_ifoutstat_inc(outif, type, code);
2138
2139 return;
2140
2141 bad:
2142 m_freem(m);
2143 return;
2144 }
2145
2146 static const char *
icmp6_redirect_diag(struct in6_addr * src6,struct in6_addr * dst6,struct in6_addr * tgt6)2147 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2148 struct in6_addr *tgt6)
2149 {
2150 static char buf[1024];
2151 char ip6bufs[INET6_ADDRSTRLEN];
2152 char ip6bufd[INET6_ADDRSTRLEN];
2153 char ip6buft[INET6_ADDRSTRLEN];
2154 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2155 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2156 ip6_sprintf(ip6buft, tgt6));
2157 return buf;
2158 }
2159
2160 void
icmp6_redirect_input(struct mbuf * m,int off)2161 icmp6_redirect_input(struct mbuf *m, int off)
2162 {
2163 struct ifnet *ifp;
2164 struct ip6_hdr *ip6;
2165 struct nd_redirect *nd_rd;
2166 struct in6_addr src6, redtgt6, reddst6;
2167 union nd_opts ndopts;
2168 char ip6buf[INET6_ADDRSTRLEN];
2169 char *lladdr;
2170 int icmp6len, is_onlink, is_router, lladdrlen;
2171
2172 M_ASSERTPKTHDR(m);
2173 KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2174
2175 /* XXX if we are router, we don't update route by icmp6 redirect */
2176 if (V_ip6_forwarding)
2177 goto freeit;
2178 if (!V_icmp6_rediraccept)
2179 goto freeit;
2180
2181 /* RFC 6980: Nodes MUST silently ignore fragments */
2182 if(m->m_flags & M_FRAGMENTED)
2183 goto freeit;
2184
2185 ip6 = mtod(m, struct ip6_hdr *);
2186 icmp6len = ntohs(ip6->ip6_plen);
2187 if (m->m_len < off + icmp6len) {
2188 m = m_pullup(m, off + icmp6len);
2189 if (m == NULL) {
2190 IP6STAT_INC(ip6s_exthdrtoolong);
2191 return;
2192 }
2193 }
2194 ip6 = mtod(m, struct ip6_hdr *);
2195 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2196
2197 ifp = m->m_pkthdr.rcvif;
2198 redtgt6 = nd_rd->nd_rd_target;
2199 reddst6 = nd_rd->nd_rd_dst;
2200
2201 if (in6_setscope(&redtgt6, ifp, NULL) ||
2202 in6_setscope(&reddst6, ifp, NULL)) {
2203 goto freeit;
2204 }
2205
2206 /* validation */
2207 src6 = ip6->ip6_src;
2208 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2209 nd6log((LOG_ERR,
2210 "ICMP6 redirect sent from %s rejected; "
2211 "must be from linklocal\n",
2212 ip6_sprintf(ip6buf, &src6)));
2213 goto bad;
2214 }
2215 if (__predict_false(ip6->ip6_hlim != 255)) {
2216 ICMP6STAT_INC(icp6s_invlhlim);
2217 nd6log((LOG_ERR,
2218 "ICMP6 redirect sent from %s rejected; "
2219 "hlim=%d (must be 255)\n",
2220 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2221 goto bad;
2222 }
2223 {
2224 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2225 struct nhop_object *nh;
2226 struct in6_addr kdst;
2227 uint32_t scopeid;
2228
2229 in6_splitscope(&reddst6, &kdst, &scopeid);
2230 NET_EPOCH_ASSERT();
2231 nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2232 if (nh != NULL) {
2233 struct in6_addr nh_addr;
2234 nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2235 if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2236 nd6log((LOG_ERR,
2237 "ICMP6 redirect rejected; no route "
2238 "with inet6 gateway found for redirect dst: %s\n",
2239 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2240 goto bad;
2241 }
2242
2243 /*
2244 * Embed scope zone id into next hop address.
2245 */
2246 nh_addr = nh->gw6_sa.sin6_addr;
2247
2248 if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2249 nd6log((LOG_ERR,
2250 "ICMP6 redirect rejected; "
2251 "not equal to gw-for-src=%s (must be same): "
2252 "%s\n",
2253 ip6_sprintf(ip6buf, &nh_addr),
2254 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2255 goto bad;
2256 }
2257 } else {
2258 nd6log((LOG_ERR,
2259 "ICMP6 redirect rejected; "
2260 "no route found for redirect dst: %s\n",
2261 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2262 goto bad;
2263 }
2264 }
2265 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2266 nd6log((LOG_ERR,
2267 "ICMP6 redirect rejected; "
2268 "redirect dst must be unicast: %s\n",
2269 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2270 goto bad;
2271 }
2272
2273 is_router = is_onlink = 0;
2274 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2275 is_router = 1; /* router case */
2276 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2277 is_onlink = 1; /* on-link destination case */
2278 if (!is_router && !is_onlink) {
2279 nd6log((LOG_ERR,
2280 "ICMP6 redirect rejected; "
2281 "neither router case nor onlink case: %s\n",
2282 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2283 goto bad;
2284 }
2285
2286 icmp6len -= sizeof(*nd_rd);
2287 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2288 if (nd6_options(&ndopts) < 0) {
2289 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2290 __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2291 /* nd6_options have incremented stats */
2292 goto freeit;
2293 }
2294
2295 lladdr = NULL;
2296 lladdrlen = 0;
2297 if (ndopts.nd_opts_tgt_lladdr) {
2298 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2299 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2300 }
2301
2302 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2303 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2304 "(if %d, icmp6 packet %d): %s\n",
2305 __func__, ip6_sprintf(ip6buf, &redtgt6),
2306 ifp->if_addrlen, lladdrlen - 2,
2307 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2308 goto bad;
2309 }
2310
2311 /* Validation passed. */
2312
2313 /* RFC 2461 8.3 */
2314 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2315 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2316
2317 /*
2318 * Install a gateway route in the better-router case or an interface
2319 * route in the on-link-destination case.
2320 */
2321 {
2322 struct sockaddr_in6 sdst;
2323 struct sockaddr_in6 sgw;
2324 struct sockaddr_in6 ssrc;
2325 struct sockaddr *gw;
2326 int rt_flags;
2327 u_int fibnum;
2328
2329 bzero(&sdst, sizeof(sdst));
2330 bzero(&ssrc, sizeof(ssrc));
2331 sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2332 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2333 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2334 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2335 rt_flags = 0;
2336 if (is_router) {
2337 bzero(&sgw, sizeof(sgw));
2338 sgw.sin6_family = AF_INET6;
2339 sgw.sin6_len = sizeof(struct sockaddr_in6);
2340 bcopy(&redtgt6, &sgw.sin6_addr,
2341 sizeof(struct in6_addr));
2342 gw = (struct sockaddr *)&sgw;
2343 rt_flags |= RTF_GATEWAY;
2344 } else
2345 gw = ifp->if_addr->ifa_addr;
2346 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2347 rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2348 (struct sockaddr *)&ssrc, ifp, rt_flags,
2349 V_icmp6_redirtimeout);
2350 }
2351
2352 freeit:
2353 m_freem(m);
2354 return;
2355
2356 bad:
2357 ICMP6STAT_INC(icp6s_badredirect);
2358 m_freem(m);
2359 }
2360
2361 void
icmp6_redirect_output(struct mbuf * m0,struct nhop_object * nh)2362 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2363 {
2364 struct ifnet *ifp; /* my outgoing interface */
2365 struct in6_addr ifp_ll6;
2366 struct in6_addr *router_ll6;
2367 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2368 struct mbuf *m = NULL; /* newly allocated one */
2369 struct m_tag *mtag;
2370 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2371 struct nd_redirect *nd_rd;
2372 struct llentry *ln = NULL;
2373 size_t maxlen;
2374 u_char *p;
2375 struct ifnet *outif = NULL;
2376 struct sockaddr_in6 src_sa;
2377
2378 icmp6_errcount(ND_REDIRECT, 0);
2379
2380 /* if we are not router, we don't send icmp6 redirect */
2381 if (!V_ip6_forwarding)
2382 goto fail;
2383
2384 /* sanity check */
2385 if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2386 goto fail;
2387
2388 /*
2389 * Address check:
2390 * the source address must identify a neighbor, and
2391 * the destination address must not be a multicast address
2392 * [RFC 2461, sec 8.2]
2393 */
2394 sip6 = mtod(m0, struct ip6_hdr *);
2395 bzero(&src_sa, sizeof(src_sa));
2396 src_sa.sin6_family = AF_INET6;
2397 src_sa.sin6_len = sizeof(src_sa);
2398 src_sa.sin6_addr = sip6->ip6_src;
2399 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2400 goto fail;
2401 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2402 goto fail; /* what should we do here? */
2403
2404 /* rate limit */
2405 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2406 goto fail;
2407
2408 /*
2409 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2410 * we almost always ask for an mbuf cluster for simplicity.
2411 * (MHLEN < IPV6_MMTU is almost always true)
2412 */
2413 #if IPV6_MMTU >= MCLBYTES
2414 # error assumption failed about IPV6_MMTU and MCLBYTES
2415 #endif
2416 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2417 if (m == NULL)
2418 goto fail;
2419 M_SETFIB(m, M_GETFIB(m0));
2420 maxlen = M_TRAILINGSPACE(m);
2421 maxlen = min(IPV6_MMTU, maxlen);
2422 /* just for safety */
2423 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2424 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2425 goto fail;
2426 }
2427
2428 {
2429 /* get ip6 linklocal address for ifp(my outgoing interface). */
2430 struct in6_ifaddr *ia;
2431 if ((ia = in6ifa_ifpforlinklocal(ifp,
2432 IN6_IFF_NOTREADY|
2433 IN6_IFF_ANYCAST)) == NULL)
2434 goto fail;
2435 bcopy(&ia->ia_addr.sin6_addr, &ifp_ll6, sizeof(ifp_ll6));
2436 ifa_free(&ia->ia_ifa);
2437 }
2438
2439 /* get ip6 linklocal address for the router. */
2440 if (nh->nh_flags & NHF_GATEWAY) {
2441 struct sockaddr_in6 *sin6;
2442 sin6 = &nh->gw6_sa;
2443 router_ll6 = &sin6->sin6_addr;
2444 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2445 router_ll6 = (struct in6_addr *)NULL;
2446 } else
2447 router_ll6 = (struct in6_addr *)NULL;
2448
2449 /* ip6 */
2450 ip6 = mtod(m, struct ip6_hdr *);
2451 ip6->ip6_flow = 0;
2452 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2453 ip6->ip6_vfc |= IPV6_VERSION;
2454 /* ip6->ip6_plen will be set later */
2455 ip6->ip6_nxt = IPPROTO_ICMPV6;
2456 ip6->ip6_hlim = 255;
2457 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2458 bcopy(&ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2459 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2460
2461 /* ND Redirect */
2462 nd_rd = (struct nd_redirect *)(ip6 + 1);
2463 nd_rd->nd_rd_type = ND_REDIRECT;
2464 nd_rd->nd_rd_code = 0;
2465 nd_rd->nd_rd_reserved = 0;
2466 if (nh->nh_flags & NHF_GATEWAY) {
2467 /*
2468 * nd_rd->nd_rd_target must be a link-local address in
2469 * better router cases.
2470 */
2471 if (!router_ll6)
2472 goto fail;
2473 bcopy(router_ll6, &nd_rd->nd_rd_target,
2474 sizeof(nd_rd->nd_rd_target));
2475 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2476 sizeof(nd_rd->nd_rd_dst));
2477 } else {
2478 /* make sure redtgt == reddst */
2479 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2480 sizeof(nd_rd->nd_rd_target));
2481 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2482 sizeof(nd_rd->nd_rd_dst));
2483 }
2484
2485 p = (u_char *)(nd_rd + 1);
2486
2487 if (!router_ll6)
2488 goto nolladdropt;
2489
2490 {
2491 /* target lladdr option */
2492 int len;
2493 struct nd_opt_hdr *nd_opt;
2494 char *lladdr;
2495
2496 ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6, 0), ifp);
2497 if (ln == NULL)
2498 goto nolladdropt;
2499
2500 len = sizeof(*nd_opt) + ifp->if_addrlen;
2501 len = (len + 7) & ~7; /* round by 8 */
2502 /* safety check */
2503 if (len + (p - (u_char *)ip6) > maxlen)
2504 goto nolladdropt;
2505
2506 if (ln->la_flags & LLE_VALID) {
2507 nd_opt = (struct nd_opt_hdr *)p;
2508 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2509 nd_opt->nd_opt_len = len >> 3;
2510 lladdr = (char *)(nd_opt + 1);
2511 bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2512 p += len;
2513 }
2514 }
2515 nolladdropt:
2516 if (ln != NULL)
2517 LLE_RUNLOCK(ln);
2518
2519 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2520
2521 /* just to be safe */
2522 #ifdef M_DECRYPTED /*not openbsd*/
2523 if (m0->m_flags & M_DECRYPTED)
2524 goto noredhdropt;
2525 #endif
2526 if (p - (u_char *)ip6 > maxlen)
2527 goto noredhdropt;
2528
2529 {
2530 /* redirected header option */
2531 int len;
2532 struct nd_opt_rd_hdr *nd_opt_rh;
2533
2534 /*
2535 * compute the maximum size for icmp6 redirect header option.
2536 * XXX room for auth header?
2537 */
2538 len = maxlen - (p - (u_char *)ip6);
2539 len &= ~7;
2540
2541 /* This is just for simplicity. */
2542 if (m0->m_pkthdr.len != m0->m_len) {
2543 if (m0->m_next) {
2544 m_freem(m0->m_next);
2545 m0->m_next = NULL;
2546 }
2547 m0->m_pkthdr.len = m0->m_len;
2548 }
2549
2550 /*
2551 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2552 * about padding/truncate rule for the original IP packet.
2553 * From the discussion on IPv6imp in Feb 1999,
2554 * the consensus was:
2555 * - "attach as much as possible" is the goal
2556 * - pad if not aligned (original size can be guessed by
2557 * original ip6 header)
2558 * Following code adds the padding if it is simple enough,
2559 * and truncates if not.
2560 */
2561 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2562 panic("assumption failed in %s:%d", __FILE__,
2563 __LINE__);
2564
2565 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2566 /* not enough room, truncate */
2567 m0->m_pkthdr.len = m0->m_len = len -
2568 sizeof(*nd_opt_rh);
2569 } else {
2570 /* enough room, pad or truncate */
2571 size_t extra;
2572
2573 extra = m0->m_pkthdr.len % 8;
2574 if (extra) {
2575 /* pad if easy enough, truncate if not */
2576 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2577 /* pad */
2578 bzero(m0->m_data + m0->m_len, 8 - extra);
2579 m0->m_len += (8 - extra);
2580 m0->m_pkthdr.len += (8 - extra);
2581 } else {
2582 /* truncate */
2583 m0->m_pkthdr.len -= extra;
2584 m0->m_len -= extra;
2585 }
2586 }
2587 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2588 m0->m_pkthdr.len = m0->m_len = len -
2589 sizeof(*nd_opt_rh);
2590 }
2591
2592 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2593 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2594 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2595 nd_opt_rh->nd_opt_rh_len = len >> 3;
2596 p += sizeof(*nd_opt_rh);
2597 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2598
2599 /* connect m0 to m */
2600 m_tag_delete_chain(m0, NULL);
2601 m0->m_flags &= ~M_PKTHDR;
2602 m->m_next = m0;
2603 m->m_pkthdr.len = m->m_len + m0->m_len;
2604 m0 = NULL;
2605 }
2606 noredhdropt:;
2607 if (m0) {
2608 m_freem(m0);
2609 m0 = NULL;
2610 }
2611
2612 /* XXX: clear embedded link IDs in the inner header */
2613 in6_clearscope(&sip6->ip6_src);
2614 in6_clearscope(&sip6->ip6_dst);
2615 in6_clearscope(&nd_rd->nd_rd_target);
2616 in6_clearscope(&nd_rd->nd_rd_dst);
2617
2618 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2619
2620 nd_rd->nd_rd_cksum = 0;
2621 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2622 sizeof(*ip6), ntohs(ip6->ip6_plen));
2623
2624 if (send_sendso_input_hook != NULL) {
2625 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2626 M_NOWAIT);
2627 if (mtag == NULL)
2628 goto fail;
2629 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2630 m_tag_prepend(m, mtag);
2631 }
2632
2633 /* send the packet to outside... */
2634 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2635 if (outif) {
2636 icmp6_ifstat_inc(outif, ifs6_out_msg);
2637 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2638 }
2639 ICMP6STAT_INC2(icp6s_outhist, ND_REDIRECT);
2640
2641 return;
2642
2643 fail:
2644 if (m)
2645 m_freem(m);
2646 if (m0)
2647 m_freem(m0);
2648 }
2649
2650 /*
2651 * ICMPv6 socket option processing.
2652 */
2653 int
icmp6_ctloutput(struct socket * so,struct sockopt * sopt)2654 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2655 {
2656 int error = 0;
2657 int optlen;
2658 struct inpcb *inp = sotoinpcb(so);
2659 int level, op, optname;
2660
2661 if (sopt) {
2662 level = sopt->sopt_level;
2663 op = sopt->sopt_dir;
2664 optname = sopt->sopt_name;
2665 optlen = sopt->sopt_valsize;
2666 } else
2667 level = op = optname = optlen = 0;
2668
2669 if (level != IPPROTO_ICMPV6) {
2670 return EINVAL;
2671 }
2672
2673 switch (op) {
2674 case SOPT_SET:
2675 switch (optname) {
2676 case ICMP6_FILTER:
2677 {
2678 struct icmp6_filter ic6f;
2679
2680 if (optlen != sizeof(ic6f)) {
2681 error = EMSGSIZE;
2682 break;
2683 }
2684 error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2685 if (error == 0) {
2686 INP_WLOCK(inp);
2687 *inp->in6p_icmp6filt = ic6f;
2688 INP_WUNLOCK(inp);
2689 }
2690 break;
2691 }
2692
2693 default:
2694 error = ENOPROTOOPT;
2695 break;
2696 }
2697 break;
2698
2699 case SOPT_GET:
2700 switch (optname) {
2701 case ICMP6_FILTER:
2702 {
2703 struct icmp6_filter ic6f;
2704
2705 INP_RLOCK(inp);
2706 ic6f = *inp->in6p_icmp6filt;
2707 INP_RUNLOCK(inp);
2708 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2709 break;
2710 }
2711
2712 default:
2713 error = ENOPROTOOPT;
2714 break;
2715 }
2716 break;
2717 }
2718
2719 return (error);
2720 }
2721
2722 static int sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS);
2723 VNET_DEFINE_STATIC(u_int, icmp6errppslim) = 100;
2724 #define V_icmp6errppslim VNET(icmp6errppslim)
2725 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ERRPPSLIMIT, errppslimit,
2726 CTLTYPE_UINT | CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6errppslim), 0,
2727 &sysctl_icmp6lim_and_jitter, "IU",
2728 "Maximum number of ICMPv6 error/reply messages per second");
2729
2730 typedef enum {
2731 RATELIM_PARAM_PROB = 0,
2732 RATELIM_TOO_BIG,
2733 RATELIM_UNREACH,
2734 RATELIM_TEXCEED,
2735 RATELIM_REDIR,
2736 RATELIM_REPLY,
2737 RATELIM_OTHER,
2738 RATELIM_MAX
2739 } ratelim_which;
2740
2741 static const char *icmp6_rate_descrs[RATELIM_MAX] = {
2742 [RATELIM_PARAM_PROB] = "bad IPv6 header",
2743 [RATELIM_TOO_BIG] = "packet too big",
2744 [RATELIM_UNREACH] = "destination unreachable",
2745 [RATELIM_TEXCEED] = "time exceeded",
2746 [RATELIM_REPLY] = "echo reply",
2747 [RATELIM_REDIR] = "neighbor discovery redirect",
2748 [RATELIM_OTHER] = "(other)",
2749 };
2750
2751 VNET_DEFINE_STATIC(int, icmp6lim_curr_jitter[RATELIM_MAX]) = {0};
2752 #define V_icmp6lim_curr_jitter VNET(icmp6lim_curr_jitter)
2753
2754 VNET_DEFINE_STATIC(u_int, icmp6lim_jitter) = 8;
2755 #define V_icmp6lim_jitter VNET(icmp6lim_jitter)
2756 SYSCTL_PROC(_net_inet6_icmp6, OID_AUTO, icmp6lim_jitter, CTLTYPE_UINT |
2757 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_jitter), 0,
2758 &sysctl_icmp6lim_and_jitter, "IU",
2759 "Random errppslimit jitter adjustment limit");
2760
2761 VNET_DEFINE_STATIC(int, icmp6lim_output) = 1;
2762 #define V_icmp6lim_output VNET(icmp6lim_output)
2763 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, icmp6lim_output,
2764 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_output), 0,
2765 "Enable logging of ICMPv6 response rate limiting");
2766
2767 static void
icmp6lim_new_jitter(int which)2768 icmp6lim_new_jitter(int which)
2769 {
2770 /*
2771 * Adjust limit +/- to jitter the measurement to deny a side-channel
2772 * port scan as in https://dl.acm.org/doi/10.1145/3372297.3417280
2773 */
2774 KASSERT(which >= 0 && which < RATELIM_MAX,
2775 ("%s: which %d", __func__, which));
2776 if (V_icmp6lim_jitter > 0)
2777 V_icmp6lim_curr_jitter[which] =
2778 arc4random_uniform(V_icmp6lim_jitter * 2 + 1) -
2779 V_icmp6lim_jitter;
2780 }
2781
2782 static int
sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)2783 sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)
2784 {
2785 uint32_t new;
2786 int error;
2787 bool lim;
2788
2789 MPASS(oidp->oid_arg1 == &VNET_NAME(icmp6errppslim) ||
2790 oidp->oid_arg1 == &VNET_NAME(icmp6lim_jitter));
2791
2792 lim = (oidp->oid_arg1 == &VNET_NAME(icmp6errppslim));
2793 new = lim ? V_icmp6errppslim : V_icmp6lim_jitter;
2794 error = sysctl_handle_int(oidp, &new, 0, req);
2795 if (error == 0 && req->newptr) {
2796 if (lim) {
2797 if (new != 0 && new <= V_icmp6lim_jitter)
2798 error = EINVAL;
2799 else
2800 V_icmp6errppslim = new;
2801 } else {
2802 if (new >= V_icmp6errppslim)
2803 error = EINVAL;
2804 else {
2805 V_icmp6lim_jitter = new;
2806 for (int i = 0; i < RATELIM_MAX; i++) {
2807 icmp6lim_new_jitter(i);
2808 }
2809 }
2810 }
2811 }
2812 MPASS(V_icmp6errppslim == 0 || V_icmp6errppslim > V_icmp6lim_jitter);
2813
2814 return (error);
2815 }
2816
2817
2818 VNET_DEFINE_STATIC(struct counter_rate *, icmp6_rates[RATELIM_MAX]);
2819 #define V_icmp6_rates VNET(icmp6_rates)
2820
2821 static void
icmp6_ratelimit_init(void)2822 icmp6_ratelimit_init(void)
2823 {
2824
2825 for (int i = 0; i < RATELIM_MAX; i++) {
2826 V_icmp6_rates[i] = counter_rate_alloc(M_WAITOK, 1);
2827 icmp6lim_new_jitter(i);
2828 }
2829 }
2830 VNET_SYSINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
2831 icmp6_ratelimit_init, NULL);
2832
2833 #ifdef VIMAGE
2834 static void
icmp6_ratelimit_uninit(void)2835 icmp6_ratelimit_uninit(void)
2836 {
2837
2838 for (int i = 0; i < RATELIM_MAX; i++)
2839 counter_rate_free(V_icmp6_rates[i]);
2840 }
2841 VNET_SYSUNINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
2842 icmp6_ratelimit_uninit, NULL);
2843 #endif
2844
2845 /*
2846 * Perform rate limit check.
2847 * Returns 0 if it is okay to send the icmp6 packet.
2848 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2849 * limitation.
2850 *
2851 * XXX per-destination/type check necessary?
2852 *
2853 * dst - not used at this moment
2854 * code - not used at this moment
2855 */
2856 int
icmp6_ratelimit(const struct in6_addr * dst,const int type,const int code)2857 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
2858 {
2859 ratelim_which which;
2860 int64_t pps;
2861
2862 if (V_icmp6errppslim == 0)
2863 return (0);
2864
2865 switch (type) {
2866 case ICMP6_PARAM_PROB:
2867 which = RATELIM_PARAM_PROB;
2868 break;
2869 case ICMP6_PACKET_TOO_BIG:
2870 which = RATELIM_TOO_BIG;
2871 break;
2872 case ICMP6_DST_UNREACH:
2873 which = RATELIM_UNREACH;
2874 break;
2875 case ICMP6_TIME_EXCEEDED:
2876 which = RATELIM_TEXCEED;
2877 break;
2878 case ND_REDIRECT:
2879 which = RATELIM_REDIR;
2880 break;
2881 case ICMP6_ECHO_REPLY:
2882 which = RATELIM_REPLY;
2883 break;
2884 default:
2885 which = RATELIM_OTHER;
2886 break;
2887 };
2888
2889 pps = counter_ratecheck(V_icmp6_rates[which], V_icmp6errppslim +
2890 V_icmp6lim_curr_jitter[which]);
2891 if (pps > 0) {
2892 if (V_icmp6lim_output)
2893 log(LOG_NOTICE, "Limiting ICMPv6 %s output from %jd "
2894 "to %d packets/sec\n", icmp6_rate_descrs[which],
2895 (intmax_t )pps, V_icmp6errppslim +
2896 V_icmp6lim_curr_jitter[which]);
2897 icmp6lim_new_jitter(which);
2898 }
2899 if (pps == -1) {
2900 ICMP6STAT_INC(icp6s_toofreq);
2901 return (-1);
2902 }
2903
2904 return (0);
2905 }
2906