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 #include <sys/cdefs.h>
64 #define MBUF_PRIVATE /* XXXRW: Optimisation tries to avoid M_EXT mbufs */
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68
69 #include <sys/param.h>
70 #include <sys/domain.h>
71 #include <sys/jail.h>
72 #include <sys/kernel.h>
73 #include <sys/lock.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/protosw.h>
78 #include <sys/signalvar.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sx.h>
82 #include <sys/syslog.h>
83 #include <sys/systm.h>
84 #include <sys/time.h>
85
86 #include <net/if.h>
87 #include <net/if_var.h>
88 #include <net/if_dl.h>
89 #include <net/if_llatbl.h>
90 #include <net/if_private.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 #include <net/route/route_ctl.h>
94 #include <net/route/nhop.h>
95 #include <net/vnet.h>
96
97 #include <netinet/in.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip6.h>
101 #include <netinet/icmp6.h>
102 #include <netinet/tcp_var.h>
103
104 #include <netinet6/in6_fib.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/in6_pcb.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/mld6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/send.h>
112
113 extern ip6proto_ctlinput_t *ip6_ctlprotox[];
114
115 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
116 VNET_PCPUSTAT_SYSINIT(icmp6stat);
117 SYSCTL_VNET_PCPUSTAT(_net_inet6_icmp6, ICMPV6CTL_STATS, stats,
118 struct icmp6stat, icmp6stat,
119 "ICMPv6 statistics (struct icmp6stat, netinet/icmp6.h)");
120
121 #ifdef VIMAGE
122 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
123 #endif /* VIMAGE */
124
125 VNET_DEFINE_STATIC(int, icmp6_rediraccept) = 1;
126 #define V_icmp6_rediraccept VNET(icmp6_rediraccept)
127 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRACCEPT, rediraccept,
128 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_rediraccept), 0,
129 "Accept ICMPv6 redirect messages");
130
131 VNET_DEFINE_STATIC(int, icmp6_redirtimeout) = 10 * 60; /* 10 minutes */
132 #define V_icmp6_redirtimeout VNET(icmp6_redirtimeout)
133 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_REDIRTIMEOUT, redirtimeout,
134 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_redirtimeout), 0,
135 "Delay in seconds before expiring redirect route");
136
137 VNET_DEFINE_STATIC(int, icmp6_nodeinfo) = 0;
138 #define V_icmp6_nodeinfo VNET(icmp6_nodeinfo)
139 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO, nodeinfo,
140 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6_nodeinfo), 0,
141 "Mask of enabled RFC4620 node information query types");
142
143 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
144 #define V_ripcbinfo VNET(ripcbinfo)
145
146 static void icmp6_errcount(int, int);
147 static int icmp6_rip6_input(struct mbuf **, int);
148 static void icmp6_reflect(struct mbuf *, size_t);
149 static const char *icmp6_redirect_diag(struct in6_addr *,
150 struct in6_addr *, struct in6_addr *);
151 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
152 static struct mbuf *ni6_nametodns(const char *, int, int);
153 static int ni6_dnsmatch(const char *, int, const char *, int);
154 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
155 struct ifnet **, struct in6_addr *);
156 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
157 struct ifnet *, int);
158 static int icmp6_notify_error(struct mbuf **, int, int);
159
160 /*
161 * Kernel module interface for updating icmp6stat. The argument is an index
162 * into icmp6stat treated as an array of u_quad_t. While this encodes the
163 * general layout of icmp6stat into the caller, it doesn't encode its
164 * location, so that future changes to add, for example, per-CPU stats
165 * support won't cause binary compatibility problems for kernel modules.
166 */
167 void
kmod_icmp6stat_inc(int statnum)168 kmod_icmp6stat_inc(int statnum)
169 {
170
171 counter_u64_add(VNET(icmp6stat)[statnum], 1);
172 }
173
174 static void
icmp6_errcount(int type,int code)175 icmp6_errcount(int type, int code)
176 {
177 switch (type) {
178 case ICMP6_DST_UNREACH:
179 switch (code) {
180 case ICMP6_DST_UNREACH_NOROUTE:
181 ICMP6STAT_INC(icp6s_odst_unreach_noroute);
182 return;
183 case ICMP6_DST_UNREACH_ADMIN:
184 ICMP6STAT_INC(icp6s_odst_unreach_admin);
185 return;
186 case ICMP6_DST_UNREACH_BEYONDSCOPE:
187 ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
188 return;
189 case ICMP6_DST_UNREACH_ADDR:
190 ICMP6STAT_INC(icp6s_odst_unreach_addr);
191 return;
192 case ICMP6_DST_UNREACH_NOPORT:
193 ICMP6STAT_INC(icp6s_odst_unreach_noport);
194 return;
195 }
196 break;
197 case ICMP6_PACKET_TOO_BIG:
198 ICMP6STAT_INC(icp6s_opacket_too_big);
199 return;
200 case ICMP6_TIME_EXCEEDED:
201 switch (code) {
202 case ICMP6_TIME_EXCEED_TRANSIT:
203 ICMP6STAT_INC(icp6s_otime_exceed_transit);
204 return;
205 case ICMP6_TIME_EXCEED_REASSEMBLY:
206 ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
207 return;
208 }
209 break;
210 case ICMP6_PARAM_PROB:
211 switch (code) {
212 case ICMP6_PARAMPROB_HEADER:
213 ICMP6STAT_INC(icp6s_oparamprob_header);
214 return;
215 case ICMP6_PARAMPROB_NEXTHEADER:
216 ICMP6STAT_INC(icp6s_oparamprob_nextheader);
217 return;
218 case ICMP6_PARAMPROB_OPTION:
219 ICMP6STAT_INC(icp6s_oparamprob_option);
220 return;
221 }
222 break;
223 case ND_REDIRECT:
224 ICMP6STAT_INC(icp6s_oredirect);
225 return;
226 }
227 ICMP6STAT_INC(icp6s_ounknown);
228 }
229
230 /*
231 * A wrapper function for icmp6_error() necessary when the erroneous packet
232 * may not contain enough scope zone information.
233 */
234 void
icmp6_error2(struct mbuf * m,int type,int code,int param,struct ifnet * ifp)235 icmp6_error2(struct mbuf *m, int type, int code, int param,
236 struct ifnet *ifp)
237 {
238 struct ip6_hdr *ip6;
239
240 if (ifp == NULL)
241 return;
242
243 if (m->m_len < sizeof(struct ip6_hdr)) {
244 m = m_pullup(m, sizeof(struct ip6_hdr));
245 if (m == NULL) {
246 IP6STAT_INC(ip6s_exthdrtoolong);
247 return;
248 }
249 }
250 ip6 = mtod(m, struct ip6_hdr *);
251
252 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
253 return;
254 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
255 return;
256
257 icmp6_error(m, type, code, param);
258 }
259
260 /*
261 * Generate an error packet of type error in response to bad IP6 packet.
262 */
263 void
icmp6_error(struct mbuf * m,int type,int code,int param)264 icmp6_error(struct mbuf *m, int type, int code, int param)
265 {
266 struct ip6_hdr *oip6, *nip6;
267 struct icmp6_hdr *icmp6;
268 struct epoch_tracker et;
269 u_int preplen;
270 int off;
271 int nxt;
272
273 ICMP6STAT_INC(icp6s_error);
274
275 /* count per-type-code statistics */
276 icmp6_errcount(type, code);
277
278 #ifdef M_DECRYPTED /*not openbsd*/
279 if (m->m_flags & M_DECRYPTED) {
280 ICMP6STAT_INC(icp6s_canterror);
281 goto freeit;
282 }
283 #endif
284
285 if (m->m_len < sizeof(struct ip6_hdr)) {
286 m = m_pullup(m, sizeof(struct ip6_hdr));
287 if (m == NULL) {
288 IP6STAT_INC(ip6s_exthdrtoolong);
289 return;
290 }
291 }
292 oip6 = mtod(m, struct ip6_hdr *);
293
294 /*
295 * If the destination address of the erroneous packet is a multicast
296 * address, or the packet was sent using link-layer multicast,
297 * we should basically suppress sending an error (RFC 2463, Section
298 * 2.4).
299 * We have two exceptions (the item e.2 in that section):
300 * - the Packet Too Big message can be sent for path MTU discovery.
301 * - the Parameter Problem Message that can be allowed an icmp6 error
302 * in the option type field. This check has been done in
303 * ip6_unknown_opt(), so we can just check the type and code.
304 */
305 if ((m->m_flags & (M_BCAST|M_MCAST) ||
306 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
307 (type != ICMP6_PACKET_TOO_BIG &&
308 (type != ICMP6_PARAM_PROB ||
309 code != ICMP6_PARAMPROB_OPTION)))
310 goto freeit;
311
312 /*
313 * RFC 2463, 2.4 (e.5): source address check.
314 * XXX: the case of anycast source?
315 */
316 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
317 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
318 goto freeit;
319
320 /*
321 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
322 * don't do it.
323 */
324 nxt = -1;
325 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
326 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
327 struct icmp6_hdr *icp;
328
329 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
330 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
331 if (m == NULL) {
332 IP6STAT_INC(ip6s_exthdrtoolong);
333 return;
334 }
335 }
336 oip6 = mtod(m, struct ip6_hdr *);
337 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
338
339 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
340 icp->icmp6_type == ND_REDIRECT) {
341 /*
342 * ICMPv6 error
343 * Special case: for redirect (which is
344 * informational) we must not send icmp6 error.
345 */
346 ICMP6STAT_INC(icp6s_canterror);
347 goto freeit;
348 } else {
349 /* ICMPv6 informational - send the error */
350 }
351 } else {
352 /* non-ICMPv6 - send the error */
353 }
354
355 /* Finally, do rate limitation check. */
356 if (icmp6_ratelimit(&oip6->ip6_src, type, code))
357 goto freeit;
358
359 /*
360 * OK, ICMP6 can be generated.
361 */
362
363 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
364 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
365
366 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
367 M_PREPEND(m, preplen, M_NOWAIT); /* FIB is also copied over. */
368 if (m == NULL) {
369 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
370 return;
371 }
372
373 nip6 = mtod(m, struct ip6_hdr *);
374 nip6->ip6_src = oip6->ip6_src;
375 nip6->ip6_dst = oip6->ip6_dst;
376
377 in6_clearscope(&oip6->ip6_src);
378 in6_clearscope(&oip6->ip6_dst);
379
380 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
381 icmp6->icmp6_type = type;
382 icmp6->icmp6_code = code;
383 icmp6->icmp6_pptr = htonl((u_int32_t)param);
384
385 ICMP6STAT_INC2(icp6s_outhist, type);
386 NET_EPOCH_ENTER(et);
387 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
388 NET_EPOCH_EXIT(et);
389
390 return;
391
392 freeit:
393 /*
394 * If we can't tell whether or not we can generate ICMP6, free it.
395 */
396 m_freem(m);
397 }
398
399 int
icmp6_errmap(const struct icmp6_hdr * icmp6)400 icmp6_errmap(const struct icmp6_hdr *icmp6)
401 {
402
403 switch (icmp6->icmp6_type) {
404 case ICMP6_DST_UNREACH:
405 switch (icmp6->icmp6_code) {
406 case ICMP6_DST_UNREACH_NOROUTE:
407 case ICMP6_DST_UNREACH_ADDR:
408 return (EHOSTUNREACH);
409 case ICMP6_DST_UNREACH_NOPORT:
410 case ICMP6_DST_UNREACH_ADMIN:
411 return (ECONNREFUSED);
412 case ICMP6_DST_UNREACH_BEYONDSCOPE:
413 return (ENOPROTOOPT);
414 default:
415 return (0); /* Shouldn't happen. */
416 }
417 case ICMP6_PACKET_TOO_BIG:
418 return (EMSGSIZE);
419 case ICMP6_TIME_EXCEEDED:
420 switch (icmp6->icmp6_code) {
421 case ICMP6_TIME_EXCEED_TRANSIT:
422 return (EHOSTUNREACH);
423 case ICMP6_TIME_EXCEED_REASSEMBLY:
424 return (0);
425 default:
426 return (0); /* Shouldn't happen. */
427 }
428 case ICMP6_PARAM_PROB:
429 switch (icmp6->icmp6_code) {
430 case ICMP6_PARAMPROB_NEXTHEADER:
431 return (ECONNREFUSED);
432 case ICMP6_PARAMPROB_HEADER:
433 case ICMP6_PARAMPROB_OPTION:
434 return (ENOPROTOOPT);
435 default:
436 return (0); /* Shouldn't happen. */
437 }
438 default:
439 return (0);
440 }
441 }
442
443 /*
444 * Process a received ICMP6 message.
445 */
446 int
icmp6_input(struct mbuf ** mp,int * offp,int proto)447 icmp6_input(struct mbuf **mp, int *offp, int proto)
448 {
449 struct mbuf *m, *n;
450 struct ifnet *ifp;
451 struct ip6_hdr *ip6, *nip6;
452 struct icmp6_hdr *icmp6, *nicmp6;
453 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
454 int code, error, icmp6len, ip6len, noff, off, sum;
455
456 NET_EPOCH_ASSERT();
457
458 m = *mp;
459 off = *offp;
460
461 if (m->m_len < off + sizeof(struct icmp6_hdr)) {
462 m = m_pullup(m, off + sizeof(struct icmp6_hdr));
463 if (m == NULL) {
464 IP6STAT_INC(ip6s_exthdrtoolong);
465 *mp = m;
466 return (IPPROTO_DONE);
467 }
468 }
469
470 /*
471 * Locate icmp6 structure in mbuf, and check
472 * that not corrupted and of at least minimum length
473 */
474
475 icmp6len = m->m_pkthdr.len - off;
476 if (icmp6len < sizeof(struct icmp6_hdr)) {
477 ICMP6STAT_INC(icp6s_tooshort);
478 goto freeit;
479 }
480
481 ip6 = mtod(m, struct ip6_hdr *);
482 ifp = m->m_pkthdr.rcvif;
483 /*
484 * Check multicast group membership.
485 * Note: SSM filters are not applied for ICMPv6 traffic.
486 */
487 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
488 struct in6_multi *inm;
489
490 inm = in6m_lookup(ifp, &ip6->ip6_dst);
491 if (inm == NULL) {
492 IP6STAT_INC(ip6s_notmember);
493 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
494 goto freeit;
495 }
496 }
497
498 /* Calculate the checksum. */
499 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
500 code = icmp6->icmp6_code;
501 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
502 nd6log((LOG_ERR,
503 "ICMP6 checksum error(%d|%x) %s\n",
504 icmp6->icmp6_type, sum,
505 ip6_sprintf(ip6bufs, &ip6->ip6_src)));
506 ICMP6STAT_INC(icp6s_checksum);
507 goto freeit;
508 }
509
510 ICMP6STAT_INC2(icp6s_inhist, icmp6->icmp6_type);
511 icmp6_ifstat_inc(ifp, ifs6_in_msg);
512 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
513 icmp6_ifstat_inc(ifp, ifs6_in_error);
514
515 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
516 switch (icmp6->icmp6_type) {
517 case ICMP6_DST_UNREACH:
518 icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
519 switch (code) {
520 case ICMP6_DST_UNREACH_ADMIN:
521 icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
522 case ICMP6_DST_UNREACH_NOROUTE:
523 case ICMP6_DST_UNREACH_ADDR:
524 case ICMP6_DST_UNREACH_BEYONDSCOPE:
525 case ICMP6_DST_UNREACH_NOPORT:
526 goto deliver;
527 default:
528 goto badcode;
529 }
530 case ICMP6_PACKET_TOO_BIG:
531 icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
532 /*
533 * Validation is made in icmp6_mtudisc_update.
534 * Updating the path MTU will be done after examining
535 * intermediate extension headers.
536 */
537 goto deliver;
538 case ICMP6_TIME_EXCEEDED:
539 icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
540 switch (code) {
541 case ICMP6_TIME_EXCEED_TRANSIT:
542 case ICMP6_TIME_EXCEED_REASSEMBLY:
543 goto deliver;
544 default:
545 goto badcode;
546 }
547 case ICMP6_PARAM_PROB:
548 icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
549 switch (code) {
550 case ICMP6_PARAMPROB_NEXTHEADER:
551 case ICMP6_PARAMPROB_HEADER:
552 case ICMP6_PARAMPROB_OPTION:
553 goto deliver;
554 default:
555 goto badcode;
556 }
557 case ICMP6_ECHO_REQUEST:
558 icmp6_ifstat_inc(ifp, ifs6_in_echo);
559 if (code != 0)
560 goto badcode;
561 if (icmp6_ratelimit(&ip6->ip6_src, ICMP6_ECHO_REPLY, 0))
562 break;
563 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
564 /* Give up remote */
565 break;
566 }
567 if (!M_WRITABLE(n)
568 || n->m_len < off + sizeof(struct icmp6_hdr)) {
569 struct mbuf *n0 = n;
570 int n0len;
571
572 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
573 n = m_gethdr(M_NOWAIT, n0->m_type);
574 if (n == NULL) {
575 /* Give up remote */
576 m_freem(n0);
577 break;
578 }
579
580 m_move_pkthdr(n, n0); /* FIB copied. */
581 n0len = n0->m_pkthdr.len; /* save for use below */
582 /*
583 * Copy IPv6 and ICMPv6 only.
584 */
585 nip6 = mtod(n, struct ip6_hdr *);
586 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
587 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
588 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
589 noff = sizeof(struct ip6_hdr);
590 /* new mbuf contains only ipv6+icmpv6 headers */
591 n->m_len = noff + sizeof(struct icmp6_hdr);
592 /*
593 * Adjust mbuf. ip6_plen will be adjusted in
594 * ip6_output().
595 */
596 m_adj(n0, off + sizeof(struct icmp6_hdr));
597 /* recalculate complete packet size */
598 n->m_pkthdr.len = n0len + (noff - off);
599 n->m_next = n0;
600 } else {
601 if (n->m_len < off + sizeof(*nicmp6)) {
602 n = m_pullup(n, off + sizeof(*nicmp6));
603 if (n == NULL) {
604 IP6STAT_INC(ip6s_exthdrtoolong);
605 break;
606 }
607 }
608 nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
609 noff = off;
610 }
611 if (n) {
612 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
613 nicmp6->icmp6_code = 0;
614 ICMP6STAT_INC(icp6s_reflect);
615 ICMP6STAT_INC2(icp6s_outhist, ICMP6_ECHO_REPLY);
616 icmp6_reflect(n, noff);
617 }
618 break;
619
620 case ICMP6_ECHO_REPLY:
621 icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
622 if (code != 0)
623 goto badcode;
624 break;
625
626 case MLD_LISTENER_QUERY:
627 case MLD_LISTENER_REPORT:
628 case MLD_LISTENER_DONE:
629 case MLDV2_LISTENER_REPORT:
630 /*
631 * Drop MLD traffic which is not link-local, has a hop limit
632 * of greater than 1 hop, or which does not have the
633 * IPv6 HBH Router Alert option.
634 * As IPv6 HBH options are stripped in ip6_input() we must
635 * check an mbuf header flag.
636 * XXX Should we also sanity check that these messages
637 * were directed to a link-local multicast prefix?
638 */
639 if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
640 goto freeit;
641 if (mld_input(&m, off, icmp6len) != 0) {
642 *mp = NULL;
643 return (IPPROTO_DONE);
644 }
645 /* m stays. */
646 break;
647
648 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
649 {
650 enum { WRU, FQDN } mode;
651 struct prison *pr;
652
653 if (!V_icmp6_nodeinfo)
654 break;
655
656 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
657 mode = WRU;
658 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
659 mode = FQDN;
660 else
661 goto badlen;
662
663 pr = NULL;
664 sx_slock(&allprison_lock);
665 TAILQ_FOREACH(pr, &allprison, pr_list)
666 if (pr->pr_vnet == ifp->if_vnet)
667 break;
668 sx_sunlock(&allprison_lock);
669 if (pr == NULL)
670 pr = curthread->td_ucred->cr_prison;
671 if (mode == FQDN) {
672 if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
673 m = m_pullup(m, off +
674 sizeof(struct icmp6_nodeinfo));
675 if (m == NULL) {
676 IP6STAT_INC(ip6s_exthdrtoolong);
677 *mp = m;
678 return (IPPROTO_DONE);
679 }
680 }
681 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
682 if (n)
683 n = ni6_input(n, off, pr);
684 /* XXX meaningless if n == NULL */
685 noff = sizeof(struct ip6_hdr);
686 } else {
687 u_char *p;
688 int maxhlen, hlen;
689
690 /*
691 * XXX: this combination of flags is pointless,
692 * but should we keep this for compatibility?
693 */
694 if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
695 ICMP6_NODEINFO_TMPADDROK)) !=
696 (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
697 break;
698
699 if (code != 0)
700 goto badcode;
701
702 CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
703 n = m_gethdr(M_NOWAIT, m->m_type);
704 if (n == NULL) {
705 /* Give up remote */
706 break;
707 }
708 if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
709 /*
710 * Previous code did a blind M_COPY_PKTHDR
711 * and said "just for rcvif". If true, then
712 * we could tolerate the dup failing (due to
713 * the deep copy of the tag chain). For now
714 * be conservative and just fail.
715 */
716 m_free(n);
717 n = NULL;
718 break;
719 }
720 /*
721 * Copy IPv6 and ICMPv6 only.
722 */
723 nip6 = mtod(n, struct ip6_hdr *);
724 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
725 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
726 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
727 p = (u_char *)(nicmp6 + 1);
728 bzero(p, 4);
729
730 maxhlen = M_TRAILINGSPACE(n) -
731 (sizeof(*nip6) + sizeof(*nicmp6) + 4);
732 mtx_lock(&pr->pr_mtx);
733 hlen = strlen(pr->pr_hostname);
734 if (maxhlen > hlen)
735 maxhlen = hlen;
736 /* meaningless TTL */
737 bcopy(pr->pr_hostname, p + 4, maxhlen);
738 mtx_unlock(&pr->pr_mtx);
739 noff = sizeof(struct ip6_hdr);
740 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
741 sizeof(struct icmp6_hdr) + 4 + maxhlen;
742 nicmp6->icmp6_type = ICMP6_WRUREPLY;
743 nicmp6->icmp6_code = 0;
744 }
745 if (n) {
746 ICMP6STAT_INC(icp6s_reflect);
747 ICMP6STAT_INC2(icp6s_outhist, ICMP6_WRUREPLY);
748 icmp6_reflect(n, noff);
749 }
750 break;
751 }
752
753 case ICMP6_WRUREPLY:
754 if (code != 0)
755 goto badcode;
756 break;
757
758 case ND_ROUTER_SOLICIT:
759 icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
760 if (code != 0)
761 goto badcode;
762 if (icmp6len < sizeof(struct nd_router_solicit))
763 goto badlen;
764 if (send_sendso_input_hook != NULL) {
765 if (m->m_len < off + icmp6len) {
766 m = m_pullup(m, off + icmp6len);
767 if (m == NULL) {
768 IP6STAT_INC(ip6s_exthdrtoolong);
769 *mp = NULL;
770 return (IPPROTO_DONE);
771 }
772 }
773 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
774 if (error == 0) {
775 m = NULL;
776 goto freeit;
777 }
778 }
779 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
780 nd6_rs_input(m, off, icmp6len);
781 m = n;
782 if (m == NULL)
783 goto freeit;
784 break;
785
786 case ND_ROUTER_ADVERT:
787 icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
788 if (code != 0)
789 goto badcode;
790 if (icmp6len < sizeof(struct nd_router_advert))
791 goto badlen;
792 if (send_sendso_input_hook != NULL) {
793 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
794 if (error == 0) {
795 m = NULL;
796 goto freeit;
797 }
798 }
799 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
800 nd6_ra_input(m, off, icmp6len);
801 m = n;
802 if (m == NULL)
803 goto freeit;
804 break;
805
806 case ND_NEIGHBOR_SOLICIT:
807 icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
808 if (code != 0)
809 goto badcode;
810 if (icmp6len < sizeof(struct nd_neighbor_solicit))
811 goto badlen;
812 if (send_sendso_input_hook != NULL) {
813 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
814 if (error == 0) {
815 m = NULL;
816 goto freeit;
817 }
818 }
819 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
820 nd6_ns_input(m, off, icmp6len);
821 m = n;
822 if (m == NULL)
823 goto freeit;
824 break;
825
826 case ND_NEIGHBOR_ADVERT:
827 icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
828 if (code != 0)
829 goto badcode;
830 if (icmp6len < sizeof(struct nd_neighbor_advert))
831 goto badlen;
832 if (send_sendso_input_hook != NULL) {
833 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
834 if (error == 0) {
835 m = NULL;
836 goto freeit;
837 }
838 }
839 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
840 nd6_na_input(m, off, icmp6len);
841 m = n;
842 if (m == NULL)
843 goto freeit;
844 break;
845
846 case ND_REDIRECT:
847 icmp6_ifstat_inc(ifp, ifs6_in_redirect);
848 if (code != 0)
849 goto badcode;
850 if (icmp6len < sizeof(struct nd_redirect))
851 goto badlen;
852 if (send_sendso_input_hook != NULL) {
853 error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
854 if (error == 0) {
855 m = NULL;
856 goto freeit;
857 }
858 }
859 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
860 icmp6_redirect_input(m, off);
861 m = n;
862 if (m == NULL)
863 goto freeit;
864 break;
865
866 case ICMP6_ROUTER_RENUMBERING:
867 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
868 code != ICMP6_ROUTER_RENUMBERING_RESULT)
869 goto badcode;
870 if (icmp6len < sizeof(struct icmp6_router_renum))
871 goto badlen;
872 break;
873
874 default:
875 nd6log((LOG_DEBUG,
876 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
877 icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
878 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
879 ifp ? ifp->if_index : 0));
880 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
881 /* ICMPv6 error: MUST deliver it by spec... */
882 goto deliver;
883 } else {
884 /* ICMPv6 informational: MUST not deliver */
885 break;
886 }
887 deliver:
888 if (icmp6_notify_error(&m, off, icmp6len) != 0) {
889 /* In this case, m should've been freed. */
890 *mp = NULL;
891 return (IPPROTO_DONE);
892 }
893 break;
894
895 badcode:
896 ICMP6STAT_INC(icp6s_badcode);
897 break;
898
899 badlen:
900 ICMP6STAT_INC(icp6s_badlen);
901 break;
902 }
903
904 /* deliver the packet to appropriate sockets */
905 icmp6_rip6_input(&m, *offp);
906
907 *mp = m;
908 return (IPPROTO_DONE);
909
910 freeit:
911 m_freem(m);
912 *mp = NULL;
913 return (IPPROTO_DONE);
914 }
915
916 static int
icmp6_notify_error(struct mbuf ** mp,int off,int icmp6len)917 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len)
918 {
919 struct mbuf *m;
920 struct icmp6_hdr *icmp6;
921 struct ip6_hdr *eip6;
922 u_int32_t notifymtu;
923 struct sockaddr_in6 icmp6src, icmp6dst;
924
925 m = *mp;
926
927 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
928 ICMP6STAT_INC(icp6s_tooshort);
929 goto freeit;
930 }
931
932 if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
933 m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
934 if (m == NULL) {
935 IP6STAT_INC(ip6s_exthdrtoolong);
936 *mp = m;
937 return (-1);
938 }
939 }
940 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
941 eip6 = (struct ip6_hdr *)(icmp6 + 1);
942 bzero(&icmp6dst, sizeof(icmp6dst));
943
944 /* Detect the upper level protocol */
945 {
946 u_int8_t nxt = eip6->ip6_nxt;
947 int eoff = off + sizeof(struct icmp6_hdr) +
948 sizeof(struct ip6_hdr);
949 struct ip6ctlparam ip6cp;
950 int icmp6type = icmp6->icmp6_type;
951 struct ip6_frag *fh;
952 struct ip6_rthdr *rth;
953 struct ip6_rthdr0 *rth0;
954 int rthlen;
955
956 while (1) { /* XXX: should avoid infinite loop explicitly? */
957 struct ip6_ext *eh;
958
959 switch (nxt) {
960 case IPPROTO_HOPOPTS:
961 case IPPROTO_DSTOPTS:
962 case IPPROTO_AH:
963 if (m->m_len < eoff + sizeof(struct ip6_ext)) {
964 m = m_pullup(m, eoff +
965 sizeof(struct ip6_ext));
966 if (m == NULL) {
967 IP6STAT_INC(ip6s_exthdrtoolong);
968 *mp = m;
969 return (-1);
970 }
971 }
972 eh = (struct ip6_ext *)
973 (mtod(m, caddr_t) + eoff);
974 if (nxt == IPPROTO_AH)
975 eoff += (eh->ip6e_len + 2) << 2;
976 else
977 eoff += (eh->ip6e_len + 1) << 3;
978 nxt = eh->ip6e_nxt;
979 break;
980 case IPPROTO_ROUTING:
981 /*
982 * When the erroneous packet contains a
983 * routing header, we should examine the
984 * header to determine the final destination.
985 * Otherwise, we can't properly update
986 * information that depends on the final
987 * destination (e.g. path MTU).
988 */
989 if (m->m_len < eoff + sizeof(*rth)) {
990 m = m_pullup(m, eoff + sizeof(*rth));
991 if (m == NULL) {
992 IP6STAT_INC(ip6s_exthdrtoolong);
993 *mp = m;
994 return (-1);
995 }
996 }
997 rth = (struct ip6_rthdr *)
998 (mtod(m, caddr_t) + eoff);
999 rthlen = (rth->ip6r_len + 1) << 3;
1000 /*
1001 * XXX: currently there is no
1002 * officially defined type other
1003 * than type-0.
1004 * Note that if the segment left field
1005 * is 0, all intermediate hops must
1006 * have been passed.
1007 */
1008 if (rth->ip6r_segleft &&
1009 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1010 int hops;
1011
1012 if (m->m_len < eoff + rthlen) {
1013 m = m_pullup(m, eoff + rthlen);
1014 if (m == NULL) {
1015 IP6STAT_INC(
1016 ip6s_exthdrtoolong);
1017 *mp = m;
1018 return (-1);
1019 }
1020 }
1021 rth0 = (struct ip6_rthdr0 *)
1022 (mtod(m, caddr_t) + eoff);
1023
1024 /* just ignore a bogus header */
1025 if ((rth0->ip6r0_len % 2) == 0 &&
1026 (hops = rth0->ip6r0_len/2))
1027 icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1028 }
1029 eoff += rthlen;
1030 nxt = rth->ip6r_nxt;
1031 break;
1032 case IPPROTO_FRAGMENT:
1033 if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1034 m = m_pullup(m, eoff +
1035 sizeof(struct ip6_frag));
1036 if (m == NULL) {
1037 IP6STAT_INC(ip6s_exthdrtoolong);
1038 *mp = m;
1039 return (-1);
1040 }
1041 }
1042 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1043 eoff);
1044 /*
1045 * Data after a fragment header is meaningless
1046 * unless it is the first fragment, but
1047 * we'll go to the notify label for path MTU
1048 * discovery.
1049 */
1050 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1051 goto notify;
1052
1053 eoff += sizeof(struct ip6_frag);
1054 nxt = fh->ip6f_nxt;
1055 break;
1056 default:
1057 /*
1058 * This case includes ESP and the No Next
1059 * Header. In such cases going to the notify
1060 * label does not have any meaning
1061 * (i.e. ctlfunc will be NULL), but we go
1062 * anyway since we might have to update
1063 * path MTU information.
1064 */
1065 goto notify;
1066 }
1067 }
1068 notify:
1069 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1070
1071 /*
1072 * retrieve parameters from the inner IPv6 header, and convert
1073 * them into sockaddr structures.
1074 * XXX: there is no guarantee that the source or destination
1075 * addresses of the inner packet are in the same scope as
1076 * the addresses of the icmp packet. But there is no other
1077 * way to determine the zone.
1078 */
1079 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1080
1081 /*
1082 * Protocol layers can't do anything useful with unspecified
1083 * addresses.
1084 */
1085 if (IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_src) ||
1086 IN6_IS_ADDR_UNSPECIFIED(&eip6->ip6_dst))
1087 goto freeit;
1088
1089 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1090 icmp6dst.sin6_family = AF_INET6;
1091 if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1092 icmp6dst.sin6_addr = eip6->ip6_dst;
1093 if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1094 goto freeit;
1095 bzero(&icmp6src, sizeof(icmp6src));
1096 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1097 icmp6src.sin6_family = AF_INET6;
1098 icmp6src.sin6_addr = eip6->ip6_src;
1099 if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1100 goto freeit;
1101 icmp6src.sin6_flowinfo =
1102 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1103
1104 ip6cp.ip6c_m = m;
1105 ip6cp.ip6c_icmp6 = icmp6;
1106 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1107 ip6cp.ip6c_off = eoff;
1108 ip6cp.ip6c_finaldst = &icmp6dst;
1109 ip6cp.ip6c_src = &icmp6src;
1110 ip6cp.ip6c_nxt = nxt;
1111
1112 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1113 notifymtu = ntohl(icmp6->icmp6_mtu);
1114 ip6cp.ip6c_cmdarg = (void *)¬ifymtu;
1115 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/
1116 }
1117
1118 if (ip6_ctlprotox[nxt] != NULL)
1119 ip6_ctlprotox[nxt](&ip6cp);
1120 }
1121 *mp = m;
1122 return (0);
1123
1124 freeit:
1125 m_freem(m);
1126 *mp = NULL;
1127 return (-1);
1128 }
1129
1130 void
icmp6_mtudisc_update(struct ip6ctlparam * ip6cp,int validated)1131 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1132 {
1133 struct in6_addr *dst = &ip6cp->ip6c_finaldst->sin6_addr;
1134 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1135 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1136 u_int mtu = ntohl(icmp6->icmp6_mtu);
1137 struct in_conninfo inc;
1138 uint32_t max_mtu;
1139
1140 #if 0
1141 /*
1142 * RFC2460 section 5, last paragraph.
1143 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1144 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1145 * due to packet translator in the middle.
1146 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1147 * special handling.
1148 */
1149 if (mtu < IPV6_MMTU)
1150 return;
1151 #endif
1152
1153 /*
1154 * we reject ICMPv6 too big with abnormally small value.
1155 * XXX what is the good definition of "abnormally small"?
1156 */
1157 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1158 return;
1159
1160 if (!validated)
1161 return;
1162
1163 /*
1164 * In case the suggested mtu is less than IPV6_MMTU, we
1165 * only need to remember that it was for above mentioned
1166 * "alwaysfrag" case.
1167 * Try to be as close to the spec as possible.
1168 */
1169 if (mtu < IPV6_MMTU)
1170 mtu = IPV6_MMTU - 8;
1171
1172 bzero(&inc, sizeof(inc));
1173 inc.inc_fibnum = M_GETFIB(m);
1174 inc.inc_flags |= INC_ISIPV6;
1175 inc.inc6_faddr = *dst;
1176 if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1177 return;
1178
1179 max_mtu = tcp_hc_getmtu(&inc);
1180 if (max_mtu == 0)
1181 max_mtu = tcp_maxmtu6(&inc, NULL);
1182
1183 if (mtu < max_mtu) {
1184 tcp_hc_updatemtu(&inc, mtu);
1185 ICMP6STAT_INC(icp6s_pmtuchg);
1186 }
1187 }
1188
1189 /*
1190 * Process a Node Information Query packet, based on
1191 * draft-ietf-ipngwg-icmp-name-lookups-07.
1192 *
1193 * Spec incompatibilities:
1194 * - IPv6 Subject address handling
1195 * - IPv4 Subject address handling support missing
1196 * - Proxy reply (answer even if it's not for me)
1197 * - joins NI group address at in6_ifattach() time only, does not cope
1198 * with hostname changes by sethostname(3)
1199 */
1200 static struct mbuf *
ni6_input(struct mbuf * m,int off,struct prison * pr)1201 ni6_input(struct mbuf *m, int off, struct prison *pr)
1202 {
1203 struct icmp6_nodeinfo *ni6, *nni6;
1204 struct mbuf *n = NULL;
1205 u_int16_t qtype;
1206 int subjlen;
1207 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1208 struct ni_reply_fqdn *fqdn;
1209 int addrs; /* for NI_QTYPE_NODEADDR */
1210 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1211 struct in6_addr in6_subj; /* subject address */
1212 struct ip6_hdr *ip6;
1213 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1214 char *subj = NULL;
1215 struct in6_ifaddr *ia6 = NULL;
1216
1217 ip6 = mtod(m, struct ip6_hdr *);
1218 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1219
1220 /*
1221 * Validate IPv6 source address.
1222 * The default configuration MUST be to refuse answering queries from
1223 * global-scope addresses according to RFC4602.
1224 * Notes:
1225 * - it's not very clear what "refuse" means; this implementation
1226 * simply drops it.
1227 * - it's not very easy to identify global-scope (unicast) addresses
1228 * since there are many prefixes for them. It should be safer
1229 * and in practice sufficient to check "all" but loopback and
1230 * link-local (note that site-local unicast was deprecated and
1231 * ULA is defined as global scope-wise)
1232 */
1233 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1234 !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1235 !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1236 goto bad;
1237
1238 /*
1239 * Validate IPv6 destination address.
1240 *
1241 * The Responder must discard the Query without further processing
1242 * unless it is one of the Responder's unicast or anycast addresses, or
1243 * a link-local scope multicast address which the Responder has joined.
1244 * [RFC4602, Section 5.]
1245 */
1246 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1247 if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1248 goto bad;
1249 /* else it's a link-local multicast, fine */
1250 } else { /* unicast or anycast */
1251 ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1252 if (ia6 == NULL)
1253 goto bad; /* XXX impossible */
1254
1255 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1256 !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1257 nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1258 "a temporary address in %s:%d",
1259 __FILE__, __LINE__));
1260 goto bad;
1261 }
1262 }
1263
1264 /* validate query Subject field. */
1265 qtype = ntohs(ni6->ni_qtype);
1266 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1267 switch (qtype) {
1268 case NI_QTYPE_NOOP:
1269 case NI_QTYPE_SUPTYPES:
1270 /* 07 draft */
1271 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1272 break;
1273 /* FALLTHROUGH */
1274 case NI_QTYPE_FQDN:
1275 case NI_QTYPE_NODEADDR:
1276 case NI_QTYPE_IPV4ADDR:
1277 switch (ni6->ni_code) {
1278 case ICMP6_NI_SUBJ_IPV6:
1279 #if ICMP6_NI_SUBJ_IPV6 != 0
1280 case 0:
1281 #endif
1282 /*
1283 * backward compatibility - try to accept 03 draft
1284 * format, where no Subject is present.
1285 */
1286 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1287 subjlen == 0) {
1288 oldfqdn++;
1289 break;
1290 }
1291 #if ICMP6_NI_SUBJ_IPV6 != 0
1292 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1293 goto bad;
1294 #endif
1295
1296 if (subjlen != sizeof(struct in6_addr))
1297 goto bad;
1298
1299 /*
1300 * Validate Subject address.
1301 *
1302 * Not sure what exactly "address belongs to the node"
1303 * means in the spec, is it just unicast, or what?
1304 *
1305 * At this moment we consider Subject address as
1306 * "belong to the node" if the Subject address equals
1307 * to the IPv6 destination address; validation for
1308 * IPv6 destination address should have done enough
1309 * check for us.
1310 *
1311 * We do not do proxy at this moment.
1312 */
1313 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1314 subjlen, (caddr_t)&in6_subj);
1315 if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1316 goto bad;
1317
1318 subj = (char *)&in6_subj;
1319 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1320 break;
1321
1322 /*
1323 * XXX if we are to allow other cases, we should really
1324 * be careful about scope here.
1325 * basically, we should disallow queries toward IPv6
1326 * destination X with subject Y,
1327 * if scope(X) > scope(Y).
1328 * if we allow scope(X) > scope(Y), it will result in
1329 * information leakage across scope boundary.
1330 */
1331 goto bad;
1332
1333 case ICMP6_NI_SUBJ_FQDN:
1334 /*
1335 * Validate Subject name with gethostname(3).
1336 *
1337 * The behavior may need some debate, since:
1338 * - we are not sure if the node has FQDN as
1339 * hostname (returned by gethostname(3)).
1340 * - the code does wildcard match for truncated names.
1341 * however, we are not sure if we want to perform
1342 * wildcard match, if gethostname(3) side has
1343 * truncated hostname.
1344 */
1345 mtx_lock(&pr->pr_mtx);
1346 n = ni6_nametodns(pr->pr_hostname,
1347 strlen(pr->pr_hostname), 0);
1348 mtx_unlock(&pr->pr_mtx);
1349 if (!n || n->m_next || n->m_len == 0)
1350 goto bad;
1351 if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1352 subjlen) {
1353 m = m_pullup(m, off +
1354 sizeof(struct icmp6_nodeinfo) + subjlen);
1355 if (m == NULL) {
1356 IP6STAT_INC(ip6s_exthdrtoolong);
1357 goto bad;
1358 }
1359 }
1360 /* ip6 possibly invalid but not used after. */
1361 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1362 subj = (char *)(mtod(m, caddr_t) + off +
1363 sizeof(struct icmp6_nodeinfo));
1364 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1365 n->m_len)) {
1366 goto bad;
1367 }
1368 m_freem(n);
1369 n = NULL;
1370 break;
1371
1372 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1373 default:
1374 goto bad;
1375 }
1376 break;
1377 }
1378
1379 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1380 switch (qtype) {
1381 case NI_QTYPE_FQDN:
1382 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1383 goto bad;
1384 break;
1385 case NI_QTYPE_NODEADDR:
1386 case NI_QTYPE_IPV4ADDR:
1387 if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1388 goto bad;
1389 break;
1390 }
1391
1392 /* guess reply length */
1393 switch (qtype) {
1394 case NI_QTYPE_NOOP:
1395 break; /* no reply data */
1396 case NI_QTYPE_SUPTYPES:
1397 replylen += sizeof(u_int32_t);
1398 break;
1399 case NI_QTYPE_FQDN:
1400 /* XXX will append an mbuf */
1401 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1402 break;
1403 case NI_QTYPE_NODEADDR:
1404 addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1405 if ((replylen += addrs * (sizeof(struct in6_addr) +
1406 sizeof(u_int32_t))) > MCLBYTES)
1407 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1408 break;
1409 case NI_QTYPE_IPV4ADDR:
1410 /* unsupported - should respond with unknown Qtype? */
1411 break;
1412 default:
1413 /*
1414 * XXX: We must return a reply with the ICMP6 code
1415 * `unknown Qtype' in this case. However we regard the case
1416 * as an FQDN query for backward compatibility.
1417 * Older versions set a random value to this field,
1418 * so it rarely varies in the defined qtypes.
1419 * But the mechanism is not reliable...
1420 * maybe we should obsolete older versions.
1421 */
1422 qtype = NI_QTYPE_FQDN;
1423 /* XXX will append an mbuf */
1424 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1425 oldfqdn++;
1426 break;
1427 }
1428
1429 /* Allocate an mbuf to reply. */
1430 if (replylen > MCLBYTES) {
1431 /*
1432 * XXX: should we try to allocate more? But MCLBYTES
1433 * is probably much larger than IPV6_MMTU...
1434 */
1435 goto bad;
1436 }
1437 if (replylen > MHLEN)
1438 n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1439 else
1440 n = m_gethdr(M_NOWAIT, m->m_type);
1441 if (n == NULL) {
1442 m_freem(m);
1443 return (NULL);
1444 }
1445 m_move_pkthdr(n, m); /* just for recvif and FIB */
1446 n->m_pkthdr.len = n->m_len = replylen;
1447
1448 /* copy mbuf header and IPv6 + Node Information base headers */
1449 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1450 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1451 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1452
1453 /* qtype dependent procedure */
1454 switch (qtype) {
1455 case NI_QTYPE_NOOP:
1456 nni6->ni_code = ICMP6_NI_SUCCESS;
1457 nni6->ni_flags = 0;
1458 break;
1459 case NI_QTYPE_SUPTYPES:
1460 {
1461 u_int32_t v;
1462 nni6->ni_code = ICMP6_NI_SUCCESS;
1463 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1464 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1465 v = (u_int32_t)htonl(0x0000000f);
1466 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1467 break;
1468 }
1469 case NI_QTYPE_FQDN:
1470 nni6->ni_code = ICMP6_NI_SUCCESS;
1471 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1472 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1473 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1474 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1475 /*
1476 * XXX do we really have FQDN in hostname?
1477 */
1478 mtx_lock(&pr->pr_mtx);
1479 n->m_next = ni6_nametodns(pr->pr_hostname,
1480 strlen(pr->pr_hostname), oldfqdn);
1481 mtx_unlock(&pr->pr_mtx);
1482 if (n->m_next == NULL)
1483 goto bad;
1484 /* XXX we assume that n->m_next is not a chain */
1485 if (n->m_next->m_next != NULL)
1486 goto bad;
1487 n->m_pkthdr.len += n->m_next->m_len;
1488 break;
1489 case NI_QTYPE_NODEADDR:
1490 {
1491 int lenlim, copied;
1492
1493 nni6->ni_code = ICMP6_NI_SUCCESS;
1494 n->m_pkthdr.len = n->m_len =
1495 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1496 lenlim = M_TRAILINGSPACE(n);
1497 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1498 /* XXX: reset mbuf length */
1499 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1500 sizeof(struct icmp6_nodeinfo) + copied;
1501 break;
1502 }
1503 default:
1504 break; /* XXX impossible! */
1505 }
1506
1507 nni6->ni_type = ICMP6_NI_REPLY;
1508 m_freem(m);
1509 return (n);
1510
1511 bad:
1512 m_freem(m);
1513 if (n)
1514 m_freem(n);
1515 return (NULL);
1516 }
1517
1518 /*
1519 * make a mbuf with DNS-encoded string. no compression support.
1520 *
1521 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1522 * treated as truncated name (two \0 at the end). this is a wild guess.
1523 *
1524 * old - return pascal string if non-zero
1525 */
1526 static struct mbuf *
ni6_nametodns(const char * name,int namelen,int old)1527 ni6_nametodns(const char *name, int namelen, int old)
1528 {
1529 struct mbuf *m;
1530 char *cp, *ep;
1531 const char *p, *q;
1532 int i, len, nterm;
1533
1534 if (old)
1535 len = namelen + 1;
1536 else
1537 len = MCLBYTES;
1538
1539 /* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1540 if (len > MLEN)
1541 m = m_getcl(M_NOWAIT, MT_DATA, 0);
1542 else
1543 m = m_get(M_NOWAIT, MT_DATA);
1544 if (m == NULL)
1545 goto fail;
1546
1547 if (old) {
1548 m->m_len = len;
1549 *mtod(m, char *) = namelen;
1550 bcopy(name, mtod(m, char *) + 1, namelen);
1551 return m;
1552 } else {
1553 m->m_len = 0;
1554 cp = mtod(m, char *);
1555 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1556
1557 /* if not certain about my name, return empty buffer */
1558 if (namelen == 0)
1559 return m;
1560
1561 /*
1562 * guess if it looks like shortened hostname, or FQDN.
1563 * shortened hostname needs two trailing "\0".
1564 */
1565 i = 0;
1566 for (p = name; p < name + namelen; p++) {
1567 if (*p && *p == '.')
1568 i++;
1569 }
1570 if (i < 2)
1571 nterm = 2;
1572 else
1573 nterm = 1;
1574
1575 p = name;
1576 while (cp < ep && p < name + namelen) {
1577 i = 0;
1578 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1579 i++;
1580 /* result does not fit into mbuf */
1581 if (cp + i + 1 >= ep)
1582 goto fail;
1583 /*
1584 * DNS label length restriction, RFC1035 page 8.
1585 * "i == 0" case is included here to avoid returning
1586 * 0-length label on "foo..bar".
1587 */
1588 if (i <= 0 || i >= 64)
1589 goto fail;
1590 *cp++ = i;
1591 bcopy(p, cp, i);
1592 cp += i;
1593 p = q;
1594 if (p < name + namelen && *p == '.')
1595 p++;
1596 }
1597 /* termination */
1598 if (cp + nterm >= ep)
1599 goto fail;
1600 while (nterm-- > 0)
1601 *cp++ = '\0';
1602 m->m_len = cp - mtod(m, char *);
1603 return m;
1604 }
1605
1606 panic("should not reach here");
1607 /* NOTREACHED */
1608
1609 fail:
1610 if (m)
1611 m_freem(m);
1612 return NULL;
1613 }
1614
1615 /*
1616 * check if two DNS-encoded string matches. takes care of truncated
1617 * form (with \0\0 at the end). no compression support.
1618 * XXX upper/lowercase match (see RFC2065)
1619 */
1620 static int
ni6_dnsmatch(const char * a,int alen,const char * b,int blen)1621 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1622 {
1623 const char *a0, *b0;
1624 int l;
1625
1626 /* simplest case - need validation? */
1627 if (alen == blen && bcmp(a, b, alen) == 0)
1628 return 1;
1629
1630 a0 = a;
1631 b0 = b;
1632
1633 /* termination is mandatory */
1634 if (alen < 2 || blen < 2)
1635 return 0;
1636 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1637 return 0;
1638 alen--;
1639 blen--;
1640
1641 while (a - a0 < alen && b - b0 < blen) {
1642 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1643 return 0;
1644
1645 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1646 return 0;
1647 /* we don't support compression yet */
1648 if (a[0] >= 64 || b[0] >= 64)
1649 return 0;
1650
1651 /* truncated case */
1652 if (a[0] == 0 && a - a0 == alen - 1)
1653 return 1;
1654 if (b[0] == 0 && b - b0 == blen - 1)
1655 return 1;
1656 if (a[0] == 0 || b[0] == 0)
1657 return 0;
1658
1659 if (a[0] != b[0])
1660 return 0;
1661 l = a[0];
1662 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1663 return 0;
1664 if (bcmp(a + 1, b + 1, l) != 0)
1665 return 0;
1666
1667 a += 1 + l;
1668 b += 1 + l;
1669 }
1670
1671 if (a - a0 == alen && b - b0 == blen)
1672 return 1;
1673 else
1674 return 0;
1675 }
1676
1677 /*
1678 * calculate the number of addresses to be returned in the node info reply.
1679 */
1680 static int
ni6_addrs(struct icmp6_nodeinfo * ni6,struct mbuf * m,struct ifnet ** ifpp,struct in6_addr * subj)1681 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1682 struct in6_addr *subj)
1683 {
1684 struct ifnet *ifp;
1685 struct in6_ifaddr *ifa6;
1686 struct ifaddr *ifa;
1687 int addrs = 0, addrsofif, iffound = 0;
1688 int niflags = ni6->ni_flags;
1689
1690 NET_EPOCH_ASSERT();
1691
1692 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1693 switch (ni6->ni_code) {
1694 case ICMP6_NI_SUBJ_IPV6:
1695 if (subj == NULL) /* must be impossible... */
1696 return (0);
1697 break;
1698 default:
1699 /*
1700 * XXX: we only support IPv6 subject address for
1701 * this Qtype.
1702 */
1703 return (0);
1704 }
1705 }
1706
1707 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1708 addrsofif = 0;
1709 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1710 if (ifa->ifa_addr->sa_family != AF_INET6)
1711 continue;
1712 ifa6 = (struct in6_ifaddr *)ifa;
1713
1714 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1715 IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1716 iffound = 1;
1717
1718 /*
1719 * IPv4-mapped addresses can only be returned by a
1720 * Node Information proxy, since they represent
1721 * addresses of IPv4-only nodes, which perforce do
1722 * not implement this protocol.
1723 * [icmp-name-lookups-07, Section 5.4]
1724 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1725 * this function at this moment.
1726 */
1727
1728 /* What do we have to do about ::1? */
1729 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1730 case IPV6_ADDR_SCOPE_LINKLOCAL:
1731 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1732 continue;
1733 break;
1734 case IPV6_ADDR_SCOPE_SITELOCAL:
1735 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1736 continue;
1737 break;
1738 case IPV6_ADDR_SCOPE_GLOBAL:
1739 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1740 continue;
1741 break;
1742 default:
1743 continue;
1744 }
1745
1746 /*
1747 * check if anycast is okay.
1748 * XXX: just experimental. not in the spec.
1749 */
1750 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1751 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1752 continue; /* we need only unicast addresses */
1753 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1754 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1755 continue;
1756 }
1757 addrsofif++; /* count the address */
1758 }
1759 if (iffound) {
1760 *ifpp = ifp;
1761 return (addrsofif);
1762 }
1763
1764 addrs += addrsofif;
1765 }
1766
1767 return (addrs);
1768 }
1769
1770 static int
ni6_store_addrs(struct icmp6_nodeinfo * ni6,struct icmp6_nodeinfo * nni6,struct ifnet * ifp0,int resid)1771 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1772 struct ifnet *ifp0, int resid)
1773 {
1774 struct ifnet *ifp;
1775 struct in6_ifaddr *ifa6;
1776 struct ifaddr *ifa;
1777 struct ifnet *ifp_dep = NULL;
1778 int copied = 0, allow_deprecated = 0;
1779 u_char *cp = (u_char *)(nni6 + 1);
1780 int niflags = ni6->ni_flags;
1781 u_int32_t ltime;
1782
1783 NET_EPOCH_ASSERT();
1784
1785 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1786 return (0); /* needless to copy */
1787
1788 ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1789 again:
1790
1791 for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1792 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1793 if (ifa->ifa_addr->sa_family != AF_INET6)
1794 continue;
1795 ifa6 = (struct in6_ifaddr *)ifa;
1796
1797 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1798 allow_deprecated == 0) {
1799 /*
1800 * prefererred address should be put before
1801 * deprecated addresses.
1802 */
1803
1804 /* record the interface for later search */
1805 if (ifp_dep == NULL)
1806 ifp_dep = ifp;
1807
1808 continue;
1809 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1810 allow_deprecated != 0)
1811 continue; /* we now collect deprecated addrs */
1812
1813 /* What do we have to do about ::1? */
1814 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1815 case IPV6_ADDR_SCOPE_LINKLOCAL:
1816 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1817 continue;
1818 break;
1819 case IPV6_ADDR_SCOPE_SITELOCAL:
1820 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1821 continue;
1822 break;
1823 case IPV6_ADDR_SCOPE_GLOBAL:
1824 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1825 continue;
1826 break;
1827 default:
1828 continue;
1829 }
1830
1831 /*
1832 * check if anycast is okay.
1833 * XXX: just experimental. not in the spec.
1834 */
1835 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1836 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1837 continue;
1838 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1839 (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1840 continue;
1841 }
1842
1843 /* now we can copy the address */
1844 if (resid < sizeof(struct in6_addr) +
1845 sizeof(u_int32_t)) {
1846 /*
1847 * We give up much more copy.
1848 * Set the truncate flag and return.
1849 */
1850 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1851 return (copied);
1852 }
1853
1854 /*
1855 * Set the TTL of the address.
1856 * The TTL value should be one of the following
1857 * according to the specification:
1858 *
1859 * 1. The remaining lifetime of a DHCP lease on the
1860 * address, or
1861 * 2. The remaining Valid Lifetime of a prefix from
1862 * which the address was derived through Stateless
1863 * Autoconfiguration.
1864 *
1865 * Note that we currently do not support stateful
1866 * address configuration by DHCPv6, so the former
1867 * case can't happen.
1868 */
1869 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1870 ltime = ND6_INFINITE_LIFETIME;
1871 else {
1872 if (ifa6->ia6_lifetime.ia6t_expire >
1873 time_uptime)
1874 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1875 else
1876 ltime = 0;
1877 }
1878
1879 bcopy(<ime, cp, sizeof(u_int32_t));
1880 cp += sizeof(u_int32_t);
1881
1882 /* copy the address itself */
1883 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1884 sizeof(struct in6_addr));
1885 in6_clearscope((struct in6_addr *)cp); /* XXX */
1886 cp += sizeof(struct in6_addr);
1887
1888 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1889 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1890 }
1891 if (ifp0) /* we need search only on the specified IF */
1892 break;
1893 }
1894
1895 if (allow_deprecated == 0 && ifp_dep != NULL) {
1896 ifp = ifp_dep;
1897 allow_deprecated = 1;
1898
1899 goto again;
1900 }
1901
1902 return (copied);
1903 }
1904
1905 static bool
icmp6_rip6_match(const struct inpcb * inp,void * v)1906 icmp6_rip6_match(const struct inpcb *inp, void *v)
1907 {
1908 struct ip6_hdr *ip6 = v;
1909
1910 if ((inp->inp_vflag & INP_IPV6) == 0)
1911 return (false);
1912 if (inp->inp_ip_p != IPPROTO_ICMPV6)
1913 return (false);
1914 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1915 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1916 return (false);
1917 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1918 !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1919 return (false);
1920 return (true);
1921 }
1922
1923 /*
1924 * XXX almost dup'ed code with rip6_input.
1925 */
1926 static int
icmp6_rip6_input(struct mbuf ** mp,int off)1927 icmp6_rip6_input(struct mbuf **mp, int off)
1928 {
1929 struct mbuf *n, *m = *mp;
1930 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1931 struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1932 INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1933 struct inpcb *inp;
1934 struct sockaddr_in6 fromsa;
1935 struct icmp6_hdr *icmp6;
1936 struct mbuf *opts = NULL;
1937 int delivered = 0;
1938
1939 /* This is assumed to be safe; icmp6_input() does a pullup. */
1940 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1941
1942 /*
1943 * XXX: the address may have embedded scope zone ID, which should be
1944 * hidden from applications.
1945 */
1946 bzero(&fromsa, sizeof(fromsa));
1947 fromsa.sin6_family = AF_INET6;
1948 fromsa.sin6_len = sizeof(struct sockaddr_in6);
1949 fromsa.sin6_addr = ip6->ip6_src;
1950 if (sa6_recoverscope(&fromsa)) {
1951 m_freem(m);
1952 *mp = NULL;
1953 return (IPPROTO_DONE);
1954 }
1955
1956 while ((inp = inp_next(&inpi)) != NULL) {
1957 if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1958 inp->in6p_icmp6filt))
1959 continue;
1960 /*
1961 * Recent network drivers tend to allocate a single
1962 * mbuf cluster, rather than to make a couple of
1963 * mbufs without clusters. Also, since the IPv6 code
1964 * path tries to avoid m_pullup(), it is highly
1965 * probable that we still have an mbuf cluster here
1966 * even though the necessary length can be stored in an
1967 * mbuf's internal buffer.
1968 * Meanwhile, the default size of the receive socket
1969 * buffer for raw sockets is not so large. This means
1970 * the possibility of packet loss is relatively higher
1971 * than before. To avoid this scenario, we copy the
1972 * received data to a separate mbuf that does not use
1973 * a cluster, if possible.
1974 * XXX: it is better to copy the data after stripping
1975 * intermediate headers.
1976 */
1977 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1978 m->m_len <= MHLEN) {
1979 n = m_get(M_NOWAIT, m->m_type);
1980 if (n != NULL) {
1981 if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1982 bcopy(m->m_data, n->m_data, m->m_len);
1983 n->m_len = m->m_len;
1984 } else {
1985 m_free(n);
1986 n = NULL;
1987 }
1988 }
1989 } else
1990 n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1991 if (n == NULL)
1992 continue;
1993 if (inp->inp_flags & INP_CONTROLOPTS)
1994 ip6_savecontrol(inp, n, &opts);
1995 /* strip intermediate headers */
1996 m_adj(n, off);
1997 SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1998 if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1999 (struct sockaddr *)&fromsa, n, opts) == 0) {
2000 soroverflow_locked(inp->inp_socket);
2001 m_freem(n);
2002 if (opts)
2003 m_freem(opts);
2004 } else {
2005 sorwakeup_locked(inp->inp_socket);
2006 delivered++;
2007 }
2008 opts = NULL;
2009 }
2010 m_freem(m);
2011 *mp = NULL;
2012 if (delivered == 0)
2013 IP6STAT_DEC(ip6s_delivered);
2014 return (IPPROTO_DONE);
2015 }
2016
2017 /*
2018 * Reflect the ip6 packet back to the source.
2019 * OFF points to the icmp6 header, counted from the top of the mbuf.
2020 */
2021 static void
icmp6_reflect(struct mbuf * m,size_t off)2022 icmp6_reflect(struct mbuf *m, size_t off)
2023 {
2024 struct in6_addr src6, *srcp;
2025 struct ip6_hdr *ip6;
2026 struct icmp6_hdr *icmp6;
2027 struct in6_ifaddr *ia = NULL;
2028 struct ifnet *outif = NULL;
2029 int plen;
2030 int type, code, hlim;
2031
2032 /* too short to reflect */
2033 if (off < sizeof(struct ip6_hdr)) {
2034 nd6log((LOG_DEBUG,
2035 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2036 (u_long)off, (u_long)sizeof(struct ip6_hdr),
2037 __FILE__, __LINE__));
2038 goto bad;
2039 }
2040
2041 /*
2042 * If there are extra headers between IPv6 and ICMPv6, strip
2043 * off that header first.
2044 */
2045 #ifdef DIAGNOSTIC
2046 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2047 panic("assumption failed in icmp6_reflect");
2048 #endif
2049 if (off > sizeof(struct ip6_hdr)) {
2050 size_t l;
2051 struct ip6_hdr nip6;
2052
2053 l = off - sizeof(struct ip6_hdr);
2054 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2055 m_adj(m, l);
2056 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2057 if (m->m_len < l) {
2058 if ((m = m_pullup(m, l)) == NULL)
2059 return;
2060 }
2061 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2062 } else /* off == sizeof(struct ip6_hdr) */ {
2063 size_t l;
2064 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2065 if (m->m_len < l) {
2066 if ((m = m_pullup(m, l)) == NULL)
2067 return;
2068 }
2069 }
2070 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2071 ip6 = mtod(m, struct ip6_hdr *);
2072 ip6->ip6_nxt = IPPROTO_ICMPV6;
2073 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2074 type = icmp6->icmp6_type; /* keep type for statistics */
2075 code = icmp6->icmp6_code; /* ditto. */
2076 hlim = 0;
2077 srcp = NULL;
2078
2079 /*
2080 * If the incoming packet was addressed directly to us (i.e. unicast),
2081 * use dst as the src for the reply.
2082 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2083 * (for example) when we encounter an error while forwarding procedure
2084 * destined to a duplicated address of ours.
2085 */
2086 if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2087 ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2088 if (ia != NULL && !(ia->ia6_flags &
2089 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2090 src6 = ia->ia_addr.sin6_addr;
2091 srcp = &src6;
2092
2093 if (m->m_pkthdr.rcvif != NULL) {
2094 /* XXX: This may not be the outgoing interface */
2095 hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2096 } else
2097 hlim = V_ip6_defhlim;
2098 }
2099 }
2100
2101 if (srcp == NULL) {
2102 int error;
2103 struct in6_addr dst6;
2104 uint32_t scopeid;
2105
2106 /*
2107 * This case matches to multicasts, our anycast, or unicasts
2108 * that we do not own. Select a source address based on the
2109 * source address of the erroneous packet.
2110 */
2111 in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2112 error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2113 scopeid, NULL, &src6, &hlim);
2114
2115 if (error) {
2116 char ip6buf[INET6_ADDRSTRLEN];
2117 nd6log((LOG_DEBUG,
2118 "icmp6_reflect: source can't be determined: "
2119 "dst=%s, error=%d\n",
2120 ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2121 goto bad;
2122 }
2123 srcp = &src6;
2124 }
2125 /*
2126 * ip6_input() drops a packet if its src is multicast.
2127 * So, the src is never multicast.
2128 */
2129 ip6->ip6_dst = ip6->ip6_src;
2130 ip6->ip6_src = *srcp;
2131 ip6->ip6_flow = 0;
2132 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2133 ip6->ip6_vfc |= IPV6_VERSION;
2134 ip6->ip6_nxt = IPPROTO_ICMPV6;
2135 ip6->ip6_hlim = hlim;
2136
2137 icmp6->icmp6_cksum = 0;
2138 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2139 sizeof(struct ip6_hdr), plen);
2140
2141 /*
2142 * XXX option handling
2143 */
2144
2145 m->m_flags &= ~(M_BCAST|M_MCAST);
2146 m->m_pkthdr.rcvif = NULL;
2147 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2148 if (outif)
2149 icmp6_ifoutstat_inc(outif, type, code);
2150
2151 return;
2152
2153 bad:
2154 m_freem(m);
2155 return;
2156 }
2157
2158 static const char *
icmp6_redirect_diag(struct in6_addr * src6,struct in6_addr * dst6,struct in6_addr * tgt6)2159 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2160 struct in6_addr *tgt6)
2161 {
2162 static char buf[1024];
2163 char ip6bufs[INET6_ADDRSTRLEN];
2164 char ip6bufd[INET6_ADDRSTRLEN];
2165 char ip6buft[INET6_ADDRSTRLEN];
2166 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2167 ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2168 ip6_sprintf(ip6buft, tgt6));
2169 return buf;
2170 }
2171
2172 void
icmp6_redirect_input(struct mbuf * m,int off)2173 icmp6_redirect_input(struct mbuf *m, int off)
2174 {
2175 struct ifnet *ifp;
2176 struct ip6_hdr *ip6;
2177 struct nd_redirect *nd_rd;
2178 struct in6_addr src6, redtgt6, reddst6;
2179 union nd_opts ndopts;
2180 char ip6buf[INET6_ADDRSTRLEN];
2181 char *lladdr;
2182 int icmp6len, is_onlink, is_router, lladdrlen;
2183
2184 M_ASSERTPKTHDR(m);
2185 KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2186
2187 /* XXX if we are router, we don't update route by icmp6 redirect */
2188 if (V_ip6_forwarding)
2189 goto freeit;
2190 if (!V_icmp6_rediraccept)
2191 goto freeit;
2192
2193 /* RFC 6980: Nodes MUST silently ignore fragments */
2194 if(m->m_flags & M_FRAGMENTED)
2195 goto freeit;
2196
2197 ip6 = mtod(m, struct ip6_hdr *);
2198 icmp6len = ntohs(ip6->ip6_plen);
2199 if (m->m_len < off + icmp6len) {
2200 m = m_pullup(m, off + icmp6len);
2201 if (m == NULL) {
2202 IP6STAT_INC(ip6s_exthdrtoolong);
2203 return;
2204 }
2205 }
2206 ip6 = mtod(m, struct ip6_hdr *);
2207 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2208
2209 ifp = m->m_pkthdr.rcvif;
2210 redtgt6 = nd_rd->nd_rd_target;
2211 reddst6 = nd_rd->nd_rd_dst;
2212
2213 if (in6_setscope(&redtgt6, ifp, NULL) ||
2214 in6_setscope(&reddst6, ifp, NULL)) {
2215 goto freeit;
2216 }
2217
2218 /* validation */
2219 src6 = ip6->ip6_src;
2220 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2221 nd6log((LOG_ERR,
2222 "ICMP6 redirect sent from %s rejected; "
2223 "must be from linklocal\n",
2224 ip6_sprintf(ip6buf, &src6)));
2225 goto bad;
2226 }
2227 if (__predict_false(ip6->ip6_hlim != 255)) {
2228 ICMP6STAT_INC(icp6s_invlhlim);
2229 nd6log((LOG_ERR,
2230 "ICMP6 redirect sent from %s rejected; "
2231 "hlim=%d (must be 255)\n",
2232 ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2233 goto bad;
2234 }
2235 {
2236 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2237 struct nhop_object *nh;
2238 struct in6_addr kdst;
2239 uint32_t scopeid;
2240
2241 in6_splitscope(&reddst6, &kdst, &scopeid);
2242 NET_EPOCH_ASSERT();
2243 nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2244 if (nh != NULL) {
2245 struct in6_addr nh_addr;
2246 nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2247 if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2248 nd6log((LOG_ERR,
2249 "ICMP6 redirect rejected; no route "
2250 "with inet6 gateway found for redirect dst: %s\n",
2251 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2252 goto bad;
2253 }
2254
2255 /*
2256 * Embed scope zone id into next hop address.
2257 */
2258 nh_addr = nh->gw6_sa.sin6_addr;
2259
2260 if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2261 nd6log((LOG_ERR,
2262 "ICMP6 redirect rejected; "
2263 "not equal to gw-for-src=%s (must be same): "
2264 "%s\n",
2265 ip6_sprintf(ip6buf, &nh_addr),
2266 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2267 goto bad;
2268 }
2269 } else {
2270 nd6log((LOG_ERR,
2271 "ICMP6 redirect rejected; "
2272 "no route found for redirect dst: %s\n",
2273 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2274 goto bad;
2275 }
2276 }
2277 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2278 nd6log((LOG_ERR,
2279 "ICMP6 redirect rejected; "
2280 "redirect dst must be unicast: %s\n",
2281 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2282 goto bad;
2283 }
2284
2285 is_router = is_onlink = 0;
2286 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2287 is_router = 1; /* router case */
2288 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2289 is_onlink = 1; /* on-link destination case */
2290 if (!is_router && !is_onlink) {
2291 nd6log((LOG_ERR,
2292 "ICMP6 redirect rejected; "
2293 "neither router case nor onlink case: %s\n",
2294 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2295 goto bad;
2296 }
2297
2298 icmp6len -= sizeof(*nd_rd);
2299 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2300 if (nd6_options(&ndopts) < 0) {
2301 nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2302 __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2303 /* nd6_options have incremented stats */
2304 goto freeit;
2305 }
2306
2307 lladdr = NULL;
2308 lladdrlen = 0;
2309 if (ndopts.nd_opts_tgt_lladdr) {
2310 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2311 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2312 }
2313
2314 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2315 nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2316 "(if %d, icmp6 packet %d): %s\n",
2317 __func__, ip6_sprintf(ip6buf, &redtgt6),
2318 ifp->if_addrlen, lladdrlen - 2,
2319 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2320 goto bad;
2321 }
2322
2323 /* Validation passed. */
2324
2325 /* RFC 2461 8.3 */
2326 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2327 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2328
2329 /*
2330 * Install a gateway route in the better-router case or an interface
2331 * route in the on-link-destination case.
2332 */
2333 {
2334 struct sockaddr_in6 sdst;
2335 struct sockaddr_in6 sgw;
2336 struct sockaddr_in6 ssrc;
2337 struct sockaddr *gw;
2338 int rt_flags;
2339 u_int fibnum;
2340
2341 bzero(&sdst, sizeof(sdst));
2342 bzero(&ssrc, sizeof(ssrc));
2343 sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2344 sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2345 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2346 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2347 rt_flags = 0;
2348 if (is_router) {
2349 bzero(&sgw, sizeof(sgw));
2350 sgw.sin6_family = AF_INET6;
2351 sgw.sin6_len = sizeof(struct sockaddr_in6);
2352 bcopy(&redtgt6, &sgw.sin6_addr,
2353 sizeof(struct in6_addr));
2354 gw = (struct sockaddr *)&sgw;
2355 rt_flags |= RTF_GATEWAY;
2356 } else
2357 gw = ifp->if_addr->ifa_addr;
2358 for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2359 rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2360 (struct sockaddr *)&ssrc, ifp, rt_flags,
2361 V_icmp6_redirtimeout);
2362 }
2363
2364 freeit:
2365 m_freem(m);
2366 return;
2367
2368 bad:
2369 ICMP6STAT_INC(icp6s_badredirect);
2370 m_freem(m);
2371 }
2372
2373 void
icmp6_redirect_output(struct mbuf * m0,struct nhop_object * nh)2374 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2375 {
2376 struct ifnet *ifp; /* my outgoing interface */
2377 struct in6_addr *ifp_ll6;
2378 struct in6_addr *router_ll6;
2379 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2380 struct mbuf *m = NULL; /* newly allocated one */
2381 struct m_tag *mtag;
2382 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2383 struct nd_redirect *nd_rd;
2384 struct llentry *ln = NULL;
2385 size_t maxlen;
2386 u_char *p;
2387 struct ifnet *outif = NULL;
2388 struct sockaddr_in6 src_sa;
2389
2390 icmp6_errcount(ND_REDIRECT, 0);
2391
2392 /* if we are not router, we don't send icmp6 redirect */
2393 if (!V_ip6_forwarding)
2394 goto fail;
2395
2396 /* sanity check */
2397 if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2398 goto fail;
2399
2400 /*
2401 * Address check:
2402 * the source address must identify a neighbor, and
2403 * the destination address must not be a multicast address
2404 * [RFC 2461, sec 8.2]
2405 */
2406 sip6 = mtod(m0, struct ip6_hdr *);
2407 bzero(&src_sa, sizeof(src_sa));
2408 src_sa.sin6_family = AF_INET6;
2409 src_sa.sin6_len = sizeof(src_sa);
2410 src_sa.sin6_addr = sip6->ip6_src;
2411 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2412 goto fail;
2413 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2414 goto fail; /* what should we do here? */
2415
2416 /* rate limit */
2417 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2418 goto fail;
2419
2420 /*
2421 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2422 * we almost always ask for an mbuf cluster for simplicity.
2423 * (MHLEN < IPV6_MMTU is almost always true)
2424 */
2425 #if IPV6_MMTU >= MCLBYTES
2426 # error assumption failed about IPV6_MMTU and MCLBYTES
2427 #endif
2428 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2429 if (m == NULL)
2430 goto fail;
2431 M_SETFIB(m, M_GETFIB(m0));
2432 maxlen = M_TRAILINGSPACE(m);
2433 maxlen = min(IPV6_MMTU, maxlen);
2434 /* just for safety */
2435 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2436 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2437 goto fail;
2438 }
2439
2440 {
2441 /* get ip6 linklocal address for ifp(my outgoing interface). */
2442 struct in6_ifaddr *ia;
2443 if ((ia = in6ifa_ifpforlinklocal(ifp,
2444 IN6_IFF_NOTREADY|
2445 IN6_IFF_ANYCAST)) == NULL)
2446 goto fail;
2447 ifp_ll6 = &ia->ia_addr.sin6_addr;
2448 /* XXXRW: reference released prematurely. */
2449 ifa_free(&ia->ia_ifa);
2450 }
2451
2452 /* get ip6 linklocal address for the router. */
2453 if (nh->nh_flags & NHF_GATEWAY) {
2454 struct sockaddr_in6 *sin6;
2455 sin6 = &nh->gw6_sa;
2456 router_ll6 = &sin6->sin6_addr;
2457 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2458 router_ll6 = (struct in6_addr *)NULL;
2459 } else
2460 router_ll6 = (struct in6_addr *)NULL;
2461
2462 /* ip6 */
2463 ip6 = mtod(m, struct ip6_hdr *);
2464 ip6->ip6_flow = 0;
2465 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2466 ip6->ip6_vfc |= IPV6_VERSION;
2467 /* ip6->ip6_plen will be set later */
2468 ip6->ip6_nxt = IPPROTO_ICMPV6;
2469 ip6->ip6_hlim = 255;
2470 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2471 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2472 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2473
2474 /* ND Redirect */
2475 nd_rd = (struct nd_redirect *)(ip6 + 1);
2476 nd_rd->nd_rd_type = ND_REDIRECT;
2477 nd_rd->nd_rd_code = 0;
2478 nd_rd->nd_rd_reserved = 0;
2479 if (nh->nh_flags & NHF_GATEWAY) {
2480 /*
2481 * nd_rd->nd_rd_target must be a link-local address in
2482 * better router cases.
2483 */
2484 if (!router_ll6)
2485 goto fail;
2486 bcopy(router_ll6, &nd_rd->nd_rd_target,
2487 sizeof(nd_rd->nd_rd_target));
2488 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2489 sizeof(nd_rd->nd_rd_dst));
2490 } else {
2491 /* make sure redtgt == reddst */
2492 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2493 sizeof(nd_rd->nd_rd_target));
2494 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2495 sizeof(nd_rd->nd_rd_dst));
2496 }
2497
2498 p = (u_char *)(nd_rd + 1);
2499
2500 if (!router_ll6)
2501 goto nolladdropt;
2502
2503 {
2504 /* target lladdr option */
2505 int len;
2506 struct nd_opt_hdr *nd_opt;
2507 char *lladdr;
2508
2509 ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6, 0), ifp);
2510 if (ln == NULL)
2511 goto nolladdropt;
2512
2513 len = sizeof(*nd_opt) + ifp->if_addrlen;
2514 len = (len + 7) & ~7; /* round by 8 */
2515 /* safety check */
2516 if (len + (p - (u_char *)ip6) > maxlen)
2517 goto nolladdropt;
2518
2519 if (ln->la_flags & LLE_VALID) {
2520 nd_opt = (struct nd_opt_hdr *)p;
2521 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2522 nd_opt->nd_opt_len = len >> 3;
2523 lladdr = (char *)(nd_opt + 1);
2524 bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2525 p += len;
2526 }
2527 }
2528 nolladdropt:
2529 if (ln != NULL)
2530 LLE_RUNLOCK(ln);
2531
2532 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2533
2534 /* just to be safe */
2535 #ifdef M_DECRYPTED /*not openbsd*/
2536 if (m0->m_flags & M_DECRYPTED)
2537 goto noredhdropt;
2538 #endif
2539 if (p - (u_char *)ip6 > maxlen)
2540 goto noredhdropt;
2541
2542 {
2543 /* redirected header option */
2544 int len;
2545 struct nd_opt_rd_hdr *nd_opt_rh;
2546
2547 /*
2548 * compute the maximum size for icmp6 redirect header option.
2549 * XXX room for auth header?
2550 */
2551 len = maxlen - (p - (u_char *)ip6);
2552 len &= ~7;
2553
2554 /* This is just for simplicity. */
2555 if (m0->m_pkthdr.len != m0->m_len) {
2556 if (m0->m_next) {
2557 m_freem(m0->m_next);
2558 m0->m_next = NULL;
2559 }
2560 m0->m_pkthdr.len = m0->m_len;
2561 }
2562
2563 /*
2564 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2565 * about padding/truncate rule for the original IP packet.
2566 * From the discussion on IPv6imp in Feb 1999,
2567 * the consensus was:
2568 * - "attach as much as possible" is the goal
2569 * - pad if not aligned (original size can be guessed by
2570 * original ip6 header)
2571 * Following code adds the padding if it is simple enough,
2572 * and truncates if not.
2573 */
2574 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2575 panic("assumption failed in %s:%d", __FILE__,
2576 __LINE__);
2577
2578 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2579 /* not enough room, truncate */
2580 m0->m_pkthdr.len = m0->m_len = len -
2581 sizeof(*nd_opt_rh);
2582 } else {
2583 /* enough room, pad or truncate */
2584 size_t extra;
2585
2586 extra = m0->m_pkthdr.len % 8;
2587 if (extra) {
2588 /* pad if easy enough, truncate if not */
2589 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2590 /* pad */
2591 m0->m_len += (8 - extra);
2592 m0->m_pkthdr.len += (8 - extra);
2593 } else {
2594 /* truncate */
2595 m0->m_pkthdr.len -= extra;
2596 m0->m_len -= extra;
2597 }
2598 }
2599 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2600 m0->m_pkthdr.len = m0->m_len = len -
2601 sizeof(*nd_opt_rh);
2602 }
2603
2604 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2605 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2606 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2607 nd_opt_rh->nd_opt_rh_len = len >> 3;
2608 p += sizeof(*nd_opt_rh);
2609 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2610
2611 /* connect m0 to m */
2612 m_tag_delete_chain(m0, NULL);
2613 m0->m_flags &= ~M_PKTHDR;
2614 m->m_next = m0;
2615 m->m_pkthdr.len = m->m_len + m0->m_len;
2616 m0 = NULL;
2617 }
2618 noredhdropt:;
2619 if (m0) {
2620 m_freem(m0);
2621 m0 = NULL;
2622 }
2623
2624 /* XXX: clear embedded link IDs in the inner header */
2625 in6_clearscope(&sip6->ip6_src);
2626 in6_clearscope(&sip6->ip6_dst);
2627 in6_clearscope(&nd_rd->nd_rd_target);
2628 in6_clearscope(&nd_rd->nd_rd_dst);
2629
2630 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2631
2632 nd_rd->nd_rd_cksum = 0;
2633 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2634 sizeof(*ip6), ntohs(ip6->ip6_plen));
2635
2636 if (send_sendso_input_hook != NULL) {
2637 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2638 M_NOWAIT);
2639 if (mtag == NULL)
2640 goto fail;
2641 *(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2642 m_tag_prepend(m, mtag);
2643 }
2644
2645 /* send the packet to outside... */
2646 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2647 if (outif) {
2648 icmp6_ifstat_inc(outif, ifs6_out_msg);
2649 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2650 }
2651 ICMP6STAT_INC2(icp6s_outhist, ND_REDIRECT);
2652
2653 return;
2654
2655 fail:
2656 if (m)
2657 m_freem(m);
2658 if (m0)
2659 m_freem(m0);
2660 }
2661
2662 /*
2663 * ICMPv6 socket option processing.
2664 */
2665 int
icmp6_ctloutput(struct socket * so,struct sockopt * sopt)2666 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2667 {
2668 int error = 0;
2669 int optlen;
2670 struct inpcb *inp = sotoinpcb(so);
2671 int level, op, optname;
2672
2673 if (sopt) {
2674 level = sopt->sopt_level;
2675 op = sopt->sopt_dir;
2676 optname = sopt->sopt_name;
2677 optlen = sopt->sopt_valsize;
2678 } else
2679 level = op = optname = optlen = 0;
2680
2681 if (level != IPPROTO_ICMPV6) {
2682 return EINVAL;
2683 }
2684
2685 switch (op) {
2686 case PRCO_SETOPT:
2687 switch (optname) {
2688 case ICMP6_FILTER:
2689 {
2690 struct icmp6_filter ic6f;
2691
2692 if (optlen != sizeof(ic6f)) {
2693 error = EMSGSIZE;
2694 break;
2695 }
2696 error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2697 if (error == 0) {
2698 INP_WLOCK(inp);
2699 *inp->in6p_icmp6filt = ic6f;
2700 INP_WUNLOCK(inp);
2701 }
2702 break;
2703 }
2704
2705 default:
2706 error = ENOPROTOOPT;
2707 break;
2708 }
2709 break;
2710
2711 case PRCO_GETOPT:
2712 switch (optname) {
2713 case ICMP6_FILTER:
2714 {
2715 struct icmp6_filter ic6f;
2716
2717 INP_RLOCK(inp);
2718 ic6f = *inp->in6p_icmp6filt;
2719 INP_RUNLOCK(inp);
2720 error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2721 break;
2722 }
2723
2724 default:
2725 error = ENOPROTOOPT;
2726 break;
2727 }
2728 break;
2729 }
2730
2731 return (error);
2732 }
2733
2734 static int sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS);
2735 VNET_DEFINE_STATIC(u_int, icmp6errppslim) = 100;
2736 #define V_icmp6errppslim VNET(icmp6errppslim)
2737 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ERRPPSLIMIT, errppslimit,
2738 CTLTYPE_UINT | CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6errppslim), 0,
2739 &sysctl_icmp6lim_and_jitter, "IU",
2740 "Maximum number of ICMPv6 error/reply messages per second");
2741
2742 VNET_DEFINE_STATIC(int, icmp6lim_curr_jitter) = 0;
2743 #define V_icmp6lim_curr_jitter VNET(icmp6lim_curr_jitter)
2744
2745 VNET_DEFINE_STATIC(u_int, icmp6lim_jitter) = 8;
2746 #define V_icmp6lim_jitter VNET(icmp6lim_jitter)
2747 SYSCTL_PROC(_net_inet6_icmp6, OID_AUTO, icmp6lim_jitter, CTLTYPE_UINT |
2748 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_jitter), 0,
2749 &sysctl_icmp6lim_and_jitter, "IU",
2750 "Random errppslimit jitter adjustment limit");
2751
2752 VNET_DEFINE_STATIC(int, icmp6lim_output) = 1;
2753 #define V_icmp6lim_output VNET(icmp6lim_output)
2754 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, icmp6lim_output,
2755 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(icmp6lim_output), 0,
2756 "Enable logging of ICMPv6 response rate limiting");
2757
2758 typedef enum {
2759 RATELIM_PARAM_PROB = 0,
2760 RATELIM_TOO_BIG,
2761 RATELIM_UNREACH,
2762 RATELIM_TEXCEED,
2763 RATELIM_REDIR,
2764 RATELIM_REPLY,
2765 RATELIM_OTHER,
2766 RATELIM_MAX
2767 } ratelim_which;
2768
2769 static const char *icmp6_rate_descrs[RATELIM_MAX] = {
2770 [RATELIM_PARAM_PROB] = "bad IPv6 header",
2771 [RATELIM_TOO_BIG] = "packet too big",
2772 [RATELIM_UNREACH] = "destination unreachable",
2773 [RATELIM_TEXCEED] = "time exceeded",
2774 [RATELIM_REPLY] = "echo reply",
2775 [RATELIM_REDIR] = "neighbor discovery redirect",
2776 [RATELIM_OTHER] = "(other)",
2777 };
2778
2779 static void
icmp6lim_new_jitter(void)2780 icmp6lim_new_jitter(void)
2781 {
2782 /*
2783 * Adjust limit +/- to jitter the measurement to deny a side-channel
2784 * port scan as in https://dl.acm.org/doi/10.1145/3372297.3417280
2785 */
2786 if (V_icmp6lim_jitter > 0)
2787 V_icmp6lim_curr_jitter =
2788 arc4random_uniform(V_icmp6lim_jitter * 2 + 1) -
2789 V_icmp6lim_jitter;
2790 }
2791
2792 static int
sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)2793 sysctl_icmp6lim_and_jitter(SYSCTL_HANDLER_ARGS)
2794 {
2795 uint32_t new;
2796 int error;
2797 bool lim;
2798
2799 MPASS(oidp->oid_arg1 == &VNET_NAME(icmp6errppslim) ||
2800 oidp->oid_arg1 == &VNET_NAME(icmp6lim_jitter));
2801
2802 lim = (oidp->oid_arg1 == &VNET_NAME(icmp6errppslim));
2803 new = lim ? V_icmp6errppslim : V_icmp6lim_jitter;
2804 error = sysctl_handle_int(oidp, &new, 0, req);
2805 if (error == 0 && req->newptr) {
2806 if (lim) {
2807 if (new != 0 && new <= V_icmp6lim_jitter)
2808 error = EINVAL;
2809 else
2810 V_icmp6errppslim = new;
2811 } else {
2812 if (new >= V_icmp6errppslim)
2813 error = EINVAL;
2814 else {
2815 V_icmp6lim_jitter = new;
2816 icmp6lim_new_jitter();
2817 }
2818 }
2819 }
2820 MPASS(V_icmp6errppslim + V_icmp6lim_curr_jitter >= 0);
2821
2822 return (error);
2823 }
2824
2825
2826 VNET_DEFINE_STATIC(struct counter_rate, icmp6_rates[RATELIM_MAX]);
2827 #define V_icmp6_rates VNET(icmp6_rates)
2828
2829 static void
icmp6_ratelimit_init(void)2830 icmp6_ratelimit_init(void)
2831 {
2832
2833 for (int i = 0; i < RATELIM_MAX; i++) {
2834 V_icmp6_rates[i].cr_rate = counter_u64_alloc(M_WAITOK);
2835 V_icmp6_rates[i].cr_ticks = ticks;
2836 }
2837 icmp6lim_new_jitter();
2838 }
2839 VNET_SYSINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
2840 icmp6_ratelimit_init, NULL);
2841
2842 #ifdef VIMAGE
2843 static void
icmp6_ratelimit_uninit(void)2844 icmp6_ratelimit_uninit(void)
2845 {
2846
2847 for (int i = 0; i < RATELIM_MAX; i++)
2848 counter_u64_free(V_icmp6_rates[i].cr_rate);
2849 }
2850 VNET_SYSUNINIT(icmp6_ratelimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
2851 icmp6_ratelimit_uninit, NULL);
2852 #endif
2853
2854 /*
2855 * Perform rate limit check.
2856 * Returns 0 if it is okay to send the icmp6 packet.
2857 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2858 * limitation.
2859 *
2860 * XXX per-destination/type check necessary?
2861 *
2862 * dst - not used at this moment
2863 * code - not used at this moment
2864 */
2865 int
icmp6_ratelimit(const struct in6_addr * dst,const int type,const int code)2866 icmp6_ratelimit(const struct in6_addr *dst, const int type, const int code)
2867 {
2868 ratelim_which which;
2869 int64_t pps;
2870
2871 if (V_icmp6errppslim == 0)
2872 return (0);
2873
2874 switch (type) {
2875 case ICMP6_PARAM_PROB:
2876 which = RATELIM_PARAM_PROB;
2877 break;
2878 case ICMP6_PACKET_TOO_BIG:
2879 which = RATELIM_TOO_BIG;
2880 break;
2881 case ICMP6_DST_UNREACH:
2882 which = RATELIM_UNREACH;
2883 break;
2884 case ICMP6_TIME_EXCEEDED:
2885 which = RATELIM_TEXCEED;
2886 break;
2887 case ND_REDIRECT:
2888 which = RATELIM_REDIR;
2889 break;
2890 case ICMP6_ECHO_REPLY:
2891 which = RATELIM_REPLY;
2892 break;
2893 default:
2894 which = RATELIM_OTHER;
2895 break;
2896 };
2897
2898 pps = counter_ratecheck(&V_icmp6_rates[which], V_icmp6errppslim +
2899 V_icmp6lim_curr_jitter);
2900 if (pps > 0) {
2901 if (V_icmp6lim_output)
2902 log(LOG_NOTICE, "Limiting ICMPv6 %s output from %jd "
2903 "to %d packets/sec\n", icmp6_rate_descrs[which],
2904 (intmax_t )pps, V_icmp6errppslim +
2905 V_icmp6lim_curr_jitter);
2906 icmp6lim_new_jitter();
2907 }
2908 if (pps == -1) {
2909 ICMP6STAT_INC(icp6s_toofreq);
2910 return (-1);
2911 }
2912
2913 return (0);
2914 }
2915