FreeBSD/Linux Kernel Cross Reference
sys/netinet6/icmp6.c
1 /* $FreeBSD$ */
2 /* $KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $ */
3
4 /*-
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 /*-
34 * Copyright (c) 1982, 1986, 1988, 1993
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 4. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
62 */
63
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67
68 #include <sys/param.h>
69 #include <sys/domain.h>
70 #include <sys/kernel.h>
71 #include <sys/lock.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/signalvar.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sx.h>
79 #include <sys/syslog.h>
80 #include <sys/systm.h>
81 #include <sys/time.h>
82
83 #include <net/if.h>
84 #include <net/if_dl.h>
85 #include <net/if_types.h>
86 #include <net/route.h>
87
88 #include <netinet/in.h>
89 #include <netinet/in_pcb.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/tcp_var.h>
94 #include <netinet6/in6_ifattach.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/ip6protosw.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/mld6_var.h>
99 #include <netinet6/nd6.h>
100
101 #ifdef IPSEC
102 #include <netinet6/ipsec.h>
103 #include <netkey/key.h>
104 #endif
105
106 #ifdef FAST_IPSEC
107 #include <netipsec/ipsec.h>
108 #include <netipsec/key.h>
109 #endif
110
111 #include <net/net_osdep.h>
112
113 extern struct domain inet6domain;
114
115 struct icmp6stat icmp6stat;
116
117 extern struct inpcbinfo ripcbinfo;
118 extern struct inpcbhead ripcb;
119 extern int icmp6errppslim;
120 static int icmp6errpps_count = 0;
121 static struct timeval icmp6errppslim_last;
122 extern int icmp6_nodeinfo;
123
124 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
125 static int icmp6_rip6_input __P((struct mbuf **, int));
126 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
127 static const char *icmp6_redirect_diag __P((struct in6_addr *,
128 struct in6_addr *, struct in6_addr *));
129 static struct mbuf *ni6_input __P((struct mbuf *, int));
130 static struct mbuf *ni6_nametodns __P((const char *, int, int));
131 static int ni6_dnsmatch __P((const char *, int, const char *, int));
132 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
133 struct ifnet **, char *));
134 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
135 struct ifnet *, int));
136 static int icmp6_notify_error __P((struct mbuf **, int, int, int));
137
138 #ifdef COMPAT_RFC1885
139 static struct route_in6 icmp6_reflect_rt;
140 #endif
141
142
143 void
144 icmp6_init()
145 {
146 mld6_init();
147 }
148
149 static void
150 icmp6_errcount(stat, type, code)
151 struct icmp6errstat *stat;
152 int type, code;
153 {
154 switch (type) {
155 case ICMP6_DST_UNREACH:
156 switch (code) {
157 case ICMP6_DST_UNREACH_NOROUTE:
158 stat->icp6errs_dst_unreach_noroute++;
159 return;
160 case ICMP6_DST_UNREACH_ADMIN:
161 stat->icp6errs_dst_unreach_admin++;
162 return;
163 case ICMP6_DST_UNREACH_BEYONDSCOPE:
164 stat->icp6errs_dst_unreach_beyondscope++;
165 return;
166 case ICMP6_DST_UNREACH_ADDR:
167 stat->icp6errs_dst_unreach_addr++;
168 return;
169 case ICMP6_DST_UNREACH_NOPORT:
170 stat->icp6errs_dst_unreach_noport++;
171 return;
172 }
173 break;
174 case ICMP6_PACKET_TOO_BIG:
175 stat->icp6errs_packet_too_big++;
176 return;
177 case ICMP6_TIME_EXCEEDED:
178 switch (code) {
179 case ICMP6_TIME_EXCEED_TRANSIT:
180 stat->icp6errs_time_exceed_transit++;
181 return;
182 case ICMP6_TIME_EXCEED_REASSEMBLY:
183 stat->icp6errs_time_exceed_reassembly++;
184 return;
185 }
186 break;
187 case ICMP6_PARAM_PROB:
188 switch (code) {
189 case ICMP6_PARAMPROB_HEADER:
190 stat->icp6errs_paramprob_header++;
191 return;
192 case ICMP6_PARAMPROB_NEXTHEADER:
193 stat->icp6errs_paramprob_nextheader++;
194 return;
195 case ICMP6_PARAMPROB_OPTION:
196 stat->icp6errs_paramprob_option++;
197 return;
198 }
199 break;
200 case ND_REDIRECT:
201 stat->icp6errs_redirect++;
202 return;
203 }
204 stat->icp6errs_unknown++;
205 }
206
207 /*
208 * Generate an error packet of type error in response to bad IP6 packet.
209 */
210 void
211 icmp6_error(m, type, code, param)
212 struct mbuf *m;
213 int type, code, param;
214 {
215 struct ip6_hdr *oip6, *nip6;
216 struct icmp6_hdr *icmp6;
217 u_int preplen;
218 int off;
219 int nxt;
220
221 icmp6stat.icp6s_error++;
222
223 /* count per-type-code statistics */
224 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
225
226 #ifdef M_DECRYPTED /*not openbsd*/
227 if (m->m_flags & M_DECRYPTED) {
228 icmp6stat.icp6s_canterror++;
229 goto freeit;
230 }
231 #endif
232
233 #ifndef PULLDOWN_TEST
234 IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
235 #else
236 if (m->m_len < sizeof(struct ip6_hdr)) {
237 m = m_pullup(m, sizeof(struct ip6_hdr));
238 if (m == NULL)
239 return;
240 }
241 #endif
242 oip6 = mtod(m, struct ip6_hdr *);
243
244 /*
245 * If the destination address of the erroneous packet is a multicast
246 * address, or the packet was sent using link-layer multicast,
247 * we should basically suppress sending an error (RFC 2463, Section
248 * 2.4).
249 * We have two exceptions (the item e.2 in that section):
250 * - the Pakcet Too Big message can be sent for path MTU discovery.
251 * - the Parameter Problem Message that can be allowed an icmp6 error
252 * in the option type field. This check has been done in
253 * ip6_unknown_opt(), so we can just check the type and code.
254 */
255 if ((m->m_flags & (M_BCAST|M_MCAST) ||
256 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
257 (type != ICMP6_PACKET_TOO_BIG &&
258 (type != ICMP6_PARAM_PROB ||
259 code != ICMP6_PARAMPROB_OPTION)))
260 goto freeit;
261
262 /*
263 * RFC 2463, 2.4 (e.5): source address check.
264 * XXX: the case of anycast source?
265 */
266 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
267 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
268 goto freeit;
269
270 /*
271 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
272 * don't do it.
273 */
274 nxt = -1;
275 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
276 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
277 struct icmp6_hdr *icp;
278
279 #ifndef PULLDOWN_TEST
280 IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
281 icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
282 #else
283 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
284 sizeof(*icp));
285 if (icp == NULL) {
286 icmp6stat.icp6s_tooshort++;
287 return;
288 }
289 #endif
290 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
291 icp->icmp6_type == ND_REDIRECT) {
292 /*
293 * ICMPv6 error
294 * Special case: for redirect (which is
295 * informational) we must not send icmp6 error.
296 */
297 icmp6stat.icp6s_canterror++;
298 goto freeit;
299 } else {
300 /* ICMPv6 informational - send the error */
301 }
302 } else {
303 /* non-ICMPv6 - send the error */
304 }
305
306 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
307
308 /* Finally, do rate limitation check. */
309 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
310 icmp6stat.icp6s_toofreq++;
311 goto freeit;
312 }
313
314 /*
315 * OK, ICMP6 can be generated.
316 */
317
318 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
319 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
320
321 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
322 M_PREPEND(m, preplen, M_DONTWAIT);
323 if (m && m->m_len < preplen)
324 m = m_pullup(m, preplen);
325 if (m == NULL) {
326 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
327 return;
328 }
329
330 nip6 = mtod(m, struct ip6_hdr *);
331 nip6->ip6_src = oip6->ip6_src;
332 nip6->ip6_dst = oip6->ip6_dst;
333
334 in6_clearscope(&oip6->ip6_src);
335 in6_clearscope(&oip6->ip6_dst);
336
337 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
338 icmp6->icmp6_type = type;
339 icmp6->icmp6_code = code;
340 icmp6->icmp6_pptr = htonl((u_int32_t)param);
341
342 /*
343 * icmp6_reflect() is designed to be in the input path.
344 * icmp6_error() can be called from both input and outut path,
345 * and if we are in output path rcvif could contain bogus value.
346 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
347 * information in ip header (nip6).
348 */
349 m->m_pkthdr.rcvif = NULL;
350
351 icmp6stat.icp6s_outhist[type]++;
352 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
353
354 return;
355
356 freeit:
357 /*
358 * If we can't tell wheter or not we can generate ICMP6, free it.
359 */
360 m_freem(m);
361 }
362
363 /*
364 * Process a received ICMP6 message.
365 */
366 int
367 icmp6_input(mp, offp, proto)
368 struct mbuf **mp;
369 int *offp, proto;
370 {
371 struct mbuf *m = *mp, *n;
372 struct ip6_hdr *ip6, *nip6;
373 struct icmp6_hdr *icmp6, *nicmp6;
374 int off = *offp;
375 int icmp6len = m->m_pkthdr.len - *offp;
376 int code, sum, noff;
377
378 #ifndef PULLDOWN_TEST
379 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
380 /* m might change if M_LOOP. So, call mtod after this */
381 #endif
382
383 /*
384 * Locate icmp6 structure in mbuf, and check
385 * that not corrupted and of at least minimum length
386 */
387
388 ip6 = mtod(m, struct ip6_hdr *);
389 if (icmp6len < sizeof(struct icmp6_hdr)) {
390 icmp6stat.icp6s_tooshort++;
391 goto freeit;
392 }
393
394 /*
395 * calculate the checksum
396 */
397 #ifndef PULLDOWN_TEST
398 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
399 #else
400 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
401 if (icmp6 == NULL) {
402 icmp6stat.icp6s_tooshort++;
403 return IPPROTO_DONE;
404 }
405 #endif
406 code = icmp6->icmp6_code;
407
408 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
409 nd6log((LOG_ERR,
410 "ICMP6 checksum error(%d|%x) %s\n",
411 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
412 icmp6stat.icp6s_checksum++;
413 goto freeit;
414 }
415
416 if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
417 /*
418 * Deliver very specific ICMP6 type only.
419 * This is important to deilver TOOBIG. Otherwise PMTUD
420 * will not work.
421 */
422 switch (icmp6->icmp6_type) {
423 case ICMP6_DST_UNREACH:
424 case ICMP6_PACKET_TOO_BIG:
425 case ICMP6_TIME_EXCEEDED:
426 break;
427 default:
428 goto freeit;
429 }
430 }
431
432 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
433 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
434 if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
435 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
436
437 switch (icmp6->icmp6_type) {
438 case ICMP6_DST_UNREACH:
439 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
440 switch (code) {
441 case ICMP6_DST_UNREACH_NOROUTE:
442 code = PRC_UNREACH_NET;
443 break;
444 case ICMP6_DST_UNREACH_ADMIN:
445 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
446 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
447 break;
448 case ICMP6_DST_UNREACH_ADDR:
449 code = PRC_HOSTDEAD;
450 break;
451 #ifdef COMPAT_RFC1885
452 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
453 code = PRC_UNREACH_SRCFAIL;
454 break;
455 #else
456 case ICMP6_DST_UNREACH_BEYONDSCOPE:
457 /* I mean "source address was incorrect." */
458 code = PRC_PARAMPROB;
459 break;
460 #endif
461 case ICMP6_DST_UNREACH_NOPORT:
462 code = PRC_UNREACH_PORT;
463 break;
464 default:
465 goto badcode;
466 }
467 goto deliver;
468 break;
469
470 case ICMP6_PACKET_TOO_BIG:
471 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
472
473 /* validation is made in icmp6_mtudisc_update */
474
475 code = PRC_MSGSIZE;
476
477 /*
478 * Updating the path MTU will be done after examining
479 * intermediate extension headers.
480 */
481 goto deliver;
482 break;
483
484 case ICMP6_TIME_EXCEEDED:
485 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
486 switch (code) {
487 case ICMP6_TIME_EXCEED_TRANSIT:
488 case ICMP6_TIME_EXCEED_REASSEMBLY:
489 code += PRC_TIMXCEED_INTRANS;
490 break;
491 default:
492 goto badcode;
493 }
494 goto deliver;
495 break;
496
497 case ICMP6_PARAM_PROB:
498 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
499 switch (code) {
500 case ICMP6_PARAMPROB_NEXTHEADER:
501 code = PRC_UNREACH_PROTOCOL;
502 break;
503 case ICMP6_PARAMPROB_HEADER:
504 case ICMP6_PARAMPROB_OPTION:
505 code = PRC_PARAMPROB;
506 break;
507 default:
508 goto badcode;
509 }
510 goto deliver;
511 break;
512
513 case ICMP6_ECHO_REQUEST:
514 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
515 if (code != 0)
516 goto badcode;
517 if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
518 /* Give up remote */
519 break;
520 }
521 if ((n->m_flags & M_EXT) != 0
522 || n->m_len < off + sizeof(struct icmp6_hdr)) {
523 struct mbuf *n0 = n;
524 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
525 int n0len;
526
527 /*
528 * Prepare an internal mbuf. m_pullup() doesn't
529 * always copy the length we specified.
530 */
531 if (maxlen >= MCLBYTES) {
532 /* Give up remote */
533 m_freem(n0);
534 break;
535 }
536 MGETHDR(n, M_DONTWAIT, n0->m_type);
537 n0len = n0->m_pkthdr.len; /* save for use below */
538 if (n)
539 M_MOVE_PKTHDR(n, n0);
540 if (n && maxlen >= MHLEN) {
541 MCLGET(n, M_DONTWAIT);
542 if ((n->m_flags & M_EXT) == 0) {
543 m_free(n);
544 n = NULL;
545 }
546 }
547 if (n == NULL) {
548 /* Give up remote */
549 m_freem(n0);
550 break;
551 }
552 /*
553 * Copy IPv6 and ICMPv6 only.
554 */
555 nip6 = mtod(n, struct ip6_hdr *);
556 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
557 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
558 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
559 noff = sizeof(struct ip6_hdr);
560 /* new mbuf contains only ipv6+icmpv6 headers */
561 n->m_len = noff + sizeof(struct icmp6_hdr);
562 /*
563 * Adjust mbuf. ip6_plen will be adjusted in
564 * ip6_output().
565 */
566 m_adj(n0, off + sizeof(struct icmp6_hdr));
567 /* recalculate complete packet size */
568 n->m_pkthdr.len = n0len + (noff - off);
569 n->m_next = n0;
570 } else {
571 nip6 = mtod(n, struct ip6_hdr *);
572 nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
573 noff = off;
574 }
575 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
576 nicmp6->icmp6_code = 0;
577 if (n) {
578 icmp6stat.icp6s_reflect++;
579 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
580 icmp6_reflect(n, noff);
581 }
582 break;
583
584 case ICMP6_ECHO_REPLY:
585 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
586 if (code != 0)
587 goto badcode;
588 break;
589
590 case MLD_LISTENER_QUERY:
591 case MLD_LISTENER_REPORT:
592 if (icmp6len < sizeof(struct mld_hdr))
593 goto badlen;
594 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
595 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
596 else
597 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
598 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
599 /* give up local */
600 mld6_input(m, off);
601 m = NULL;
602 goto freeit;
603 }
604 mld6_input(n, off);
605 /* m stays. */
606 break;
607
608 case MLD_LISTENER_DONE:
609 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
610 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */
611 goto badlen;
612 break; /* nothing to be done in kernel */
613
614 case MLD_MTRACE_RESP:
615 case MLD_MTRACE:
616 /* XXX: these two are experimental. not officially defined. */
617 /* XXX: per-interface statistics? */
618 break; /* just pass it to applications */
619
620 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
621 {
622 enum { WRU, FQDN } mode;
623
624 if (!icmp6_nodeinfo)
625 break;
626
627 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
628 mode = WRU;
629 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
630 mode = FQDN;
631 else
632 goto badlen;
633
634 #define hostnamelen strlen(hostname)
635 if (mode == FQDN) {
636 #ifndef PULLDOWN_TEST
637 IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
638 IPPROTO_DONE);
639 #endif
640 n = m_copy(m, 0, M_COPYALL);
641 if (n)
642 n = ni6_input(n, off);
643 /* XXX meaningless if n == NULL */
644 noff = sizeof(struct ip6_hdr);
645 } else {
646 u_char *p;
647 int maxlen, maxhlen;
648
649 /*
650 * XXX: this combination of flags is pointless,
651 * but should we keep this for compatibility?
652 */
653 if ((icmp6_nodeinfo & 5) != 5)
654 break;
655
656 if (code != 0)
657 goto badcode;
658 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
659 if (maxlen >= MCLBYTES) {
660 /* Give up remote */
661 break;
662 }
663 MGETHDR(n, M_DONTWAIT, m->m_type);
664 if (n && maxlen > MHLEN) {
665 MCLGET(n, M_DONTWAIT);
666 if ((n->m_flags & M_EXT) == 0) {
667 m_free(n);
668 n = NULL;
669 }
670 }
671 if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
672 /*
673 * Previous code did a blind M_COPY_PKTHDR
674 * and said "just for rcvif". If true, then
675 * we could tolerate the dup failing (due to
676 * the deep copy of the tag chain). For now
677 * be conservative and just fail.
678 */
679 m_free(n);
680 n = NULL;
681 }
682 if (n == NULL) {
683 /* Give up remote */
684 break;
685 }
686 n->m_pkthdr.rcvif = NULL;
687 n->m_len = 0;
688 maxhlen = M_TRAILINGSPACE(n) - maxlen;
689 if (maxhlen > hostnamelen)
690 maxhlen = hostnamelen;
691 /*
692 * Copy IPv6 and ICMPv6 only.
693 */
694 nip6 = mtod(n, struct ip6_hdr *);
695 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
696 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
697 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
698 p = (u_char *)(nicmp6 + 1);
699 bzero(p, 4);
700 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
701 noff = sizeof(struct ip6_hdr);
702 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
703 sizeof(struct icmp6_hdr) + 4 + maxhlen;
704 nicmp6->icmp6_type = ICMP6_WRUREPLY;
705 nicmp6->icmp6_code = 0;
706 }
707 #undef hostnamelen
708 if (n) {
709 icmp6stat.icp6s_reflect++;
710 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
711 icmp6_reflect(n, noff);
712 }
713 break;
714 }
715
716 case ICMP6_WRUREPLY:
717 if (code != 0)
718 goto badcode;
719 break;
720
721 case ND_ROUTER_SOLICIT:
722 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
723 if (code != 0)
724 goto badcode;
725 if (icmp6len < sizeof(struct nd_router_solicit))
726 goto badlen;
727 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
728 /* give up local */
729 nd6_rs_input(m, off, icmp6len);
730 m = NULL;
731 goto freeit;
732 }
733 nd6_rs_input(n, off, icmp6len);
734 /* m stays. */
735 break;
736
737 case ND_ROUTER_ADVERT:
738 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
739 if (code != 0)
740 goto badcode;
741 if (icmp6len < sizeof(struct nd_router_advert))
742 goto badlen;
743 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
744 /* give up local */
745 nd6_ra_input(m, off, icmp6len);
746 m = NULL;
747 goto freeit;
748 }
749 nd6_ra_input(n, off, icmp6len);
750 /* m stays. */
751 break;
752
753 case ND_NEIGHBOR_SOLICIT:
754 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
755 if (code != 0)
756 goto badcode;
757 if (icmp6len < sizeof(struct nd_neighbor_solicit))
758 goto badlen;
759 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
760 /* give up local */
761 nd6_ns_input(m, off, icmp6len);
762 m = NULL;
763 goto freeit;
764 }
765 nd6_ns_input(n, off, icmp6len);
766 /* m stays. */
767 break;
768
769 case ND_NEIGHBOR_ADVERT:
770 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
771 if (code != 0)
772 goto badcode;
773 if (icmp6len < sizeof(struct nd_neighbor_advert))
774 goto badlen;
775 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
776 /* give up local */
777 nd6_na_input(m, off, icmp6len);
778 m = NULL;
779 goto freeit;
780 }
781 nd6_na_input(n, off, icmp6len);
782 /* m stays. */
783 break;
784
785 case ND_REDIRECT:
786 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
787 if (code != 0)
788 goto badcode;
789 if (icmp6len < sizeof(struct nd_redirect))
790 goto badlen;
791 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
792 /* give up local */
793 icmp6_redirect_input(m, off);
794 m = NULL;
795 goto freeit;
796 }
797 icmp6_redirect_input(n, off);
798 /* m stays. */
799 break;
800
801 case ICMP6_ROUTER_RENUMBERING:
802 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
803 code != ICMP6_ROUTER_RENUMBERING_RESULT)
804 goto badcode;
805 if (icmp6len < sizeof(struct icmp6_router_renum))
806 goto badlen;
807 break;
808
809 default:
810 nd6log((LOG_DEBUG,
811 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
812 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
813 ip6_sprintf(&ip6->ip6_dst),
814 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
815 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
816 /* ICMPv6 error: MUST deliver it by spec... */
817 code = PRC_NCMDS;
818 /* deliver */
819 } else {
820 /* ICMPv6 informational: MUST not deliver */
821 break;
822 }
823 deliver:
824 if (icmp6_notify_error(&m, off, icmp6len, code)) {
825 /* In this case, m should've been freed. */
826 return (IPPROTO_DONE);
827 }
828 break;
829
830 badcode:
831 icmp6stat.icp6s_badcode++;
832 break;
833
834 badlen:
835 icmp6stat.icp6s_badlen++;
836 break;
837 }
838
839 /* deliver the packet to appropriate sockets */
840 icmp6_rip6_input(&m, *offp);
841
842 return IPPROTO_DONE;
843
844 freeit:
845 m_freem(m);
846 return IPPROTO_DONE;
847 }
848
849 static int
850 icmp6_notify_error(mp, off, icmp6len, code)
851 struct mbuf **mp;
852 int off, icmp6len, code;
853 {
854 struct mbuf *m = *mp;
855 struct icmp6_hdr *icmp6;
856 struct ip6_hdr *eip6;
857 u_int32_t notifymtu;
858 struct sockaddr_in6 icmp6src, icmp6dst;
859
860 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
861 icmp6stat.icp6s_tooshort++;
862 goto freeit;
863 }
864 #ifndef PULLDOWN_TEST
865 IP6_EXTHDR_CHECK(m, off,
866 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
867 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
868 #else
869 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
870 sizeof(*icmp6) + sizeof(struct ip6_hdr));
871 if (icmp6 == NULL) {
872 icmp6stat.icp6s_tooshort++;
873 return (-1);
874 }
875 #endif
876 eip6 = (struct ip6_hdr *)(icmp6 + 1);
877
878 /* Detect the upper level protocol */
879 {
880 void (*ctlfunc) __P((int, struct sockaddr *, void *));
881 u_int8_t nxt = eip6->ip6_nxt;
882 int eoff = off + sizeof(struct icmp6_hdr) +
883 sizeof(struct ip6_hdr);
884 struct ip6ctlparam ip6cp;
885 struct in6_addr *finaldst = NULL;
886 int icmp6type = icmp6->icmp6_type;
887 struct ip6_frag *fh;
888 struct ip6_rthdr *rth;
889 struct ip6_rthdr0 *rth0;
890 int rthlen;
891
892 while (1) { /* XXX: should avoid infinite loop explicitly? */
893 struct ip6_ext *eh;
894
895 switch (nxt) {
896 case IPPROTO_HOPOPTS:
897 case IPPROTO_DSTOPTS:
898 case IPPROTO_AH:
899 #ifndef PULLDOWN_TEST
900 IP6_EXTHDR_CHECK(m, 0,
901 eoff + sizeof(struct ip6_ext), -1);
902 eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
903 #else
904 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
905 eoff, sizeof(*eh));
906 if (eh == NULL) {
907 icmp6stat.icp6s_tooshort++;
908 return (-1);
909 }
910 #endif
911
912 if (nxt == IPPROTO_AH)
913 eoff += (eh->ip6e_len + 2) << 2;
914 else
915 eoff += (eh->ip6e_len + 1) << 3;
916 nxt = eh->ip6e_nxt;
917 break;
918 case IPPROTO_ROUTING:
919 /*
920 * When the erroneous packet contains a
921 * routing header, we should examine the
922 * header to determine the final destination.
923 * Otherwise, we can't properly update
924 * information that depends on the final
925 * destination (e.g. path MTU).
926 */
927 #ifndef PULLDOWN_TEST
928 IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
929 rth = (struct ip6_rthdr *)
930 (mtod(m, caddr_t) + eoff);
931 #else
932 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
933 eoff, sizeof(*rth));
934 if (rth == NULL) {
935 icmp6stat.icp6s_tooshort++;
936 return (-1);
937 }
938 #endif
939 rthlen = (rth->ip6r_len + 1) << 3;
940 /*
941 * XXX: currently there is no
942 * officially defined type other
943 * than type-0.
944 * Note that if the segment left field
945 * is 0, all intermediate hops must
946 * have been passed.
947 */
948 if (rth->ip6r_segleft &&
949 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
950 int hops;
951
952 #ifndef PULLDOWN_TEST
953 IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
954 rth0 = (struct ip6_rthdr0 *)
955 (mtod(m, caddr_t) + eoff);
956 #else
957 IP6_EXTHDR_GET(rth0,
958 struct ip6_rthdr0 *, m,
959 eoff, rthlen);
960 if (rth0 == NULL) {
961 icmp6stat.icp6s_tooshort++;
962 return (-1);
963 }
964 #endif
965 /* just ignore a bogus header */
966 if ((rth0->ip6r0_len % 2) == 0 &&
967 (hops = rth0->ip6r0_len/2))
968 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
969 }
970 eoff += rthlen;
971 nxt = rth->ip6r_nxt;
972 break;
973 case IPPROTO_FRAGMENT:
974 #ifndef PULLDOWN_TEST
975 IP6_EXTHDR_CHECK(m, 0, eoff +
976 sizeof(struct ip6_frag), -1);
977 fh = (struct ip6_frag *)(mtod(m, caddr_t) +
978 eoff);
979 #else
980 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
981 eoff, sizeof(*fh));
982 if (fh == NULL) {
983 icmp6stat.icp6s_tooshort++;
984 return (-1);
985 }
986 #endif
987 /*
988 * Data after a fragment header is meaningless
989 * unless it is the first fragment, but
990 * we'll go to the notify label for path MTU
991 * discovery.
992 */
993 if (fh->ip6f_offlg & IP6F_OFF_MASK)
994 goto notify;
995
996 eoff += sizeof(struct ip6_frag);
997 nxt = fh->ip6f_nxt;
998 break;
999 default:
1000 /*
1001 * This case includes ESP and the No Next
1002 * Header. In such cases going to the notify
1003 * label does not have any meaning
1004 * (i.e. ctlfunc will be NULL), but we go
1005 * anyway since we might have to update
1006 * path MTU information.
1007 */
1008 goto notify;
1009 }
1010 }
1011 notify:
1012 #ifndef PULLDOWN_TEST
1013 icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1014 #else
1015 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1016 sizeof(*icmp6) + sizeof(struct ip6_hdr));
1017 if (icmp6 == NULL) {
1018 icmp6stat.icp6s_tooshort++;
1019 return (-1);
1020 }
1021 #endif
1022
1023 /*
1024 * retrieve parameters from the inner IPv6 header, and convert
1025 * them into sockaddr structures.
1026 * XXX: there is no guarantee that the source or destination
1027 * addresses of the inner packet are in the same scope as
1028 * the addresses of the icmp packet. But there is no other
1029 * way to determine the zone.
1030 */
1031 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1032
1033 bzero(&icmp6dst, sizeof(icmp6dst));
1034 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1035 icmp6dst.sin6_family = AF_INET6;
1036 if (finaldst == NULL)
1037 icmp6dst.sin6_addr = eip6->ip6_dst;
1038 else
1039 icmp6dst.sin6_addr = *finaldst;
1040 if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1041 &icmp6dst.sin6_scope_id))
1042 goto freeit;
1043 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1044 NULL, NULL)) {
1045 /* should be impossible */
1046 nd6log((LOG_DEBUG,
1047 "icmp6_notify_error: in6_embedscope failed\n"));
1048 goto freeit;
1049 }
1050
1051 /*
1052 * retrieve parameters from the inner IPv6 header, and convert
1053 * them into sockaddr structures.
1054 */
1055 bzero(&icmp6src, sizeof(icmp6src));
1056 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1057 icmp6src.sin6_family = AF_INET6;
1058 icmp6src.sin6_addr = eip6->ip6_src;
1059 if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1060 &icmp6src.sin6_scope_id)) {
1061 goto freeit;
1062 }
1063 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1064 NULL, NULL)) {
1065 /* should be impossible */
1066 nd6log((LOG_DEBUG,
1067 "icmp6_notify_error: in6_embedscope failed\n"));
1068 goto freeit;
1069 }
1070 icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1071
1072 if (finaldst == NULL)
1073 finaldst = &eip6->ip6_dst;
1074 ip6cp.ip6c_m = m;
1075 ip6cp.ip6c_icmp6 = icmp6;
1076 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1077 ip6cp.ip6c_off = eoff;
1078 ip6cp.ip6c_finaldst = finaldst;
1079 ip6cp.ip6c_src = &icmp6src;
1080 ip6cp.ip6c_nxt = nxt;
1081
1082 if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1083 notifymtu = ntohl(icmp6->icmp6_mtu);
1084 ip6cp.ip6c_cmdarg = (void *)¬ifymtu;
1085 icmp6_mtudisc_update(&ip6cp, 1); /*XXX*/
1086 }
1087
1088 ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1089 (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1090 if (ctlfunc) {
1091 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1092 &ip6cp);
1093 }
1094 }
1095 *mp = m;
1096 return (0);
1097
1098 freeit:
1099 m_freem(m);
1100 return (-1);
1101 }
1102
1103 void
1104 icmp6_mtudisc_update(ip6cp, validated)
1105 struct ip6ctlparam *ip6cp;
1106 int validated;
1107 {
1108 struct in6_addr *dst = ip6cp->ip6c_finaldst;
1109 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1110 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */
1111 u_int mtu = ntohl(icmp6->icmp6_mtu);
1112 struct in_conninfo inc;
1113
1114 #if 0
1115 /*
1116 * RFC2460 section 5, last paragraph.
1117 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1118 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1119 * due to packet translator in the middle.
1120 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1121 * special handling.
1122 */
1123 if (mtu < IPV6_MMTU)
1124 return;
1125 #endif
1126
1127 /*
1128 * we reject ICMPv6 too big with abnormally small value.
1129 * XXX what is the good definition of "abnormally small"?
1130 */
1131 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1132 return;
1133
1134 if (!validated)
1135 return;
1136
1137 bzero(&inc, sizeof(inc));
1138 inc.inc_flags = 1; /* IPv6 */
1139 inc.inc6_faddr = *dst;
1140 /* XXX normally, this won't happen */
1141 if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1142 inc.inc6_faddr.s6_addr16[1] =
1143 htons(m->m_pkthdr.rcvif->if_index);
1144 }
1145
1146 if (mtu < tcp_maxmtu6(&inc)) {
1147 tcp_hc_updatemtu(&inc, mtu);
1148 icmp6stat.icp6s_pmtuchg++;
1149 }
1150 }
1151
1152 /*
1153 * Process a Node Information Query packet, based on
1154 * draft-ietf-ipngwg-icmp-name-lookups-07.
1155 *
1156 * Spec incompatibilities:
1157 * - IPv6 Subject address handling
1158 * - IPv4 Subject address handling support missing
1159 * - Proxy reply (answer even if it's not for me)
1160 * - joins NI group address at in6_ifattach() time only, does not cope
1161 * with hostname changes by sethostname(3)
1162 */
1163 #define hostnamelen strlen(hostname)
1164 static struct mbuf *
1165 ni6_input(m, off)
1166 struct mbuf *m;
1167 int off;
1168 {
1169 struct icmp6_nodeinfo *ni6, *nni6;
1170 struct mbuf *n = NULL;
1171 u_int16_t qtype;
1172 int subjlen;
1173 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1174 struct ni_reply_fqdn *fqdn;
1175 int addrs; /* for NI_QTYPE_NODEADDR */
1176 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1177 struct sockaddr_in6 sin6_sbj; /* subject address */
1178 struct ip6_hdr *ip6;
1179 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1180 char *subj = NULL;
1181 struct in6_ifaddr *ia6 = NULL;
1182
1183 ip6 = mtod(m, struct ip6_hdr *);
1184 #ifndef PULLDOWN_TEST
1185 ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1186 #else
1187 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1188 if (ni6 == NULL) {
1189 /* m is already reclaimed */
1190 return (NULL);
1191 }
1192 #endif
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 ((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 if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1226 goto bad; /* XXX impossible */
1227
1228 if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1229 !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1230 nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1231 "a temporary address in %s:%d",
1232 __FILE__, __LINE__));
1233 goto bad;
1234 }
1235 }
1236
1237 /* validate query Subject field. */
1238 qtype = ntohs(ni6->ni_qtype);
1239 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1240 switch (qtype) {
1241 case NI_QTYPE_NOOP:
1242 case NI_QTYPE_SUPTYPES:
1243 /* 07 draft */
1244 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1245 break;
1246 /* FALLTHROUGH */
1247 case NI_QTYPE_FQDN:
1248 case NI_QTYPE_NODEADDR:
1249 switch (ni6->ni_code) {
1250 case ICMP6_NI_SUBJ_IPV6:
1251 #if ICMP6_NI_SUBJ_IPV6 != 0
1252 case 0:
1253 #endif
1254 /*
1255 * backward compatibility - try to accept 03 draft
1256 * format, where no Subject is present.
1257 */
1258 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1259 subjlen == 0) {
1260 oldfqdn++;
1261 break;
1262 }
1263 #if ICMP6_NI_SUBJ_IPV6 != 0
1264 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1265 goto bad;
1266 #endif
1267
1268 if (subjlen != sizeof(struct in6_addr))
1269 goto bad;
1270
1271 /*
1272 * Validate Subject address.
1273 *
1274 * Not sure what exactly "address belongs to the node"
1275 * means in the spec, is it just unicast, or what?
1276 *
1277 * At this moment we consider Subject address as
1278 * "belong to the node" if the Subject address equals
1279 * to the IPv6 destination address; validation for
1280 * IPv6 destination address should have done enough
1281 * check for us.
1282 *
1283 * We do not do proxy at this moment.
1284 */
1285 /* m_pulldown instead of copy? */
1286 bzero(&sin6_sbj, sizeof(sin6_sbj));
1287 sin6_sbj.sin6_family = AF_INET6;
1288 sin6_sbj.sin6_len = sizeof(sin6_sbj);
1289 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1290 subjlen, (caddr_t)&sin6_sbj.sin6_addr);
1291 if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1292 &sin6_sbj.sin6_addr, &sin6_sbj.sin6_scope_id)) {
1293 goto bad;
1294 }
1295 if (in6_embedscope(&sin6_sbj.sin6_addr, &sin6_sbj,
1296 NULL, NULL))
1297 goto bad; /* XXX should not happen */
1298
1299 subj = (char *)&sin6_sbj;
1300 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1301 &sin6_sbj.sin6_addr))
1302 break;
1303
1304 /*
1305 * XXX if we are to allow other cases, we should really
1306 * be careful about scope here.
1307 * basically, we should disallow queries toward IPv6
1308 * destination X with subject Y,
1309 * if scope(X) > scope(Y).
1310 * if we allow scope(X) > scope(Y), it will result in
1311 * information leakage across scope boundary.
1312 */
1313 goto bad;
1314
1315 case ICMP6_NI_SUBJ_FQDN:
1316 /*
1317 * Validate Subject name with gethostname(3).
1318 *
1319 * The behavior may need some debate, since:
1320 * - we are not sure if the node has FQDN as
1321 * hostname (returned by gethostname(3)).
1322 * - the code does wildcard match for truncated names.
1323 * however, we are not sure if we want to perform
1324 * wildcard match, if gethostname(3) side has
1325 * truncated hostname.
1326 */
1327 n = ni6_nametodns(hostname, hostnamelen, 0);
1328 if (!n || n->m_next || n->m_len == 0)
1329 goto bad;
1330 IP6_EXTHDR_GET(subj, char *, m,
1331 off + sizeof(struct icmp6_nodeinfo), subjlen);
1332 if (subj == NULL)
1333 goto bad;
1334 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1335 n->m_len)) {
1336 goto bad;
1337 }
1338 m_freem(n);
1339 n = NULL;
1340 break;
1341
1342 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1343 default:
1344 goto bad;
1345 }
1346 break;
1347 }
1348
1349 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1350 switch (qtype) {
1351 case NI_QTYPE_FQDN:
1352 if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1353 goto bad;
1354 break;
1355 case NI_QTYPE_NODEADDR:
1356 if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1357 goto bad;
1358 break;
1359 }
1360
1361 /* guess reply length */
1362 switch (qtype) {
1363 case NI_QTYPE_NOOP:
1364 break; /* no reply data */
1365 case NI_QTYPE_SUPTYPES:
1366 replylen += sizeof(u_int32_t);
1367 break;
1368 case NI_QTYPE_FQDN:
1369 /* XXX will append an mbuf */
1370 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1371 break;
1372 case NI_QTYPE_NODEADDR:
1373 addrs = ni6_addrs(ni6, m, &ifp, subj);
1374 if ((replylen += addrs * (sizeof(struct in6_addr) +
1375 sizeof(u_int32_t))) > MCLBYTES)
1376 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1377 break;
1378 default:
1379 /*
1380 * XXX: We must return a reply with the ICMP6 code
1381 * `unknown Qtype' in this case. However we regard the case
1382 * as an FQDN query for backward compatibility.
1383 * Older versions set a random value to this field,
1384 * so it rarely varies in the defined qtypes.
1385 * But the mechanism is not reliable...
1386 * maybe we should obsolete older versions.
1387 */
1388 qtype = NI_QTYPE_FQDN;
1389 /* XXX will append an mbuf */
1390 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1391 oldfqdn++;
1392 break;
1393 }
1394
1395 /* allocate an mbuf to reply. */
1396 MGETHDR(n, M_DONTWAIT, m->m_type);
1397 if (n == NULL) {
1398 m_freem(m);
1399 return (NULL);
1400 }
1401 M_MOVE_PKTHDR(n, m); /* just for recvif */
1402 if (replylen > MHLEN) {
1403 if (replylen > MCLBYTES) {
1404 /*
1405 * XXX: should we try to allocate more? But MCLBYTES
1406 * is probably much larger than IPV6_MMTU...
1407 */
1408 goto bad;
1409 }
1410 MCLGET(n, M_DONTWAIT);
1411 if ((n->m_flags & M_EXT) == 0) {
1412 goto bad;
1413 }
1414 }
1415 n->m_pkthdr.len = n->m_len = replylen;
1416
1417 /* copy mbuf header and IPv6 + Node Information base headers */
1418 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1419 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1420 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1421
1422 /* qtype dependent procedure */
1423 switch (qtype) {
1424 case NI_QTYPE_NOOP:
1425 nni6->ni_code = ICMP6_NI_SUCCESS;
1426 nni6->ni_flags = 0;
1427 break;
1428 case NI_QTYPE_SUPTYPES:
1429 {
1430 u_int32_t v;
1431 nni6->ni_code = ICMP6_NI_SUCCESS;
1432 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1433 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1434 v = (u_int32_t)htonl(0x0000000f);
1435 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1436 break;
1437 }
1438 case NI_QTYPE_FQDN:
1439 nni6->ni_code = ICMP6_NI_SUCCESS;
1440 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1441 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1442 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1443 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1444 /*
1445 * XXX do we really have FQDN in variable "hostname"?
1446 */
1447 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1448 if (n->m_next == NULL)
1449 goto bad;
1450 /* XXX we assume that n->m_next is not a chain */
1451 if (n->m_next->m_next != NULL)
1452 goto bad;
1453 n->m_pkthdr.len += n->m_next->m_len;
1454 break;
1455 case NI_QTYPE_NODEADDR:
1456 {
1457 int lenlim, copied;
1458
1459 nni6->ni_code = ICMP6_NI_SUCCESS;
1460 n->m_pkthdr.len = n->m_len =
1461 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1462 lenlim = M_TRAILINGSPACE(n);
1463 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1464 /* XXX: reset mbuf length */
1465 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1466 sizeof(struct icmp6_nodeinfo) + copied;
1467 break;
1468 }
1469 default:
1470 break; /* XXX impossible! */
1471 }
1472
1473 nni6->ni_type = ICMP6_NI_REPLY;
1474 m_freem(m);
1475 return (n);
1476
1477 bad:
1478 m_freem(m);
1479 if (n)
1480 m_freem(n);
1481 return (NULL);
1482 }
1483 #undef hostnamelen
1484
1485 /*
1486 * make a mbuf with DNS-encoded string. no compression support.
1487 *
1488 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1489 * treated as truncated name (two \0 at the end). this is a wild guess.
1490 */
1491 static struct mbuf *
1492 ni6_nametodns(name, namelen, old)
1493 const char *name;
1494 int namelen;
1495 int old; /* return pascal string if non-zero */
1496 {
1497 struct mbuf *m;
1498 char *cp, *ep;
1499 const char *p, *q;
1500 int i, len, nterm;
1501
1502 if (old)
1503 len = namelen + 1;
1504 else
1505 len = MCLBYTES;
1506
1507 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1508 MGET(m, M_DONTWAIT, MT_DATA);
1509 if (m && len > MLEN) {
1510 MCLGET(m, M_DONTWAIT);
1511 if ((m->m_flags & M_EXT) == 0)
1512 goto fail;
1513 }
1514 if (!m)
1515 goto fail;
1516 m->m_next = NULL;
1517
1518 if (old) {
1519 m->m_len = len;
1520 *mtod(m, char *) = namelen;
1521 bcopy(name, mtod(m, char *) + 1, namelen);
1522 return m;
1523 } else {
1524 m->m_len = 0;
1525 cp = mtod(m, char *);
1526 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1527
1528 /* if not certain about my name, return empty buffer */
1529 if (namelen == 0)
1530 return m;
1531
1532 /*
1533 * guess if it looks like shortened hostname, or FQDN.
1534 * shortened hostname needs two trailing "\0".
1535 */
1536 i = 0;
1537 for (p = name; p < name + namelen; p++) {
1538 if (*p && *p == '.')
1539 i++;
1540 }
1541 if (i < 2)
1542 nterm = 2;
1543 else
1544 nterm = 1;
1545
1546 p = name;
1547 while (cp < ep && p < name + namelen) {
1548 i = 0;
1549 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1550 i++;
1551 /* result does not fit into mbuf */
1552 if (cp + i + 1 >= ep)
1553 goto fail;
1554 /*
1555 * DNS label length restriction, RFC1035 page 8.
1556 * "i == 0" case is included here to avoid returning
1557 * 0-length label on "foo..bar".
1558 */
1559 if (i <= 0 || i >= 64)
1560 goto fail;
1561 *cp++ = i;
1562 bcopy(p, cp, i);
1563 cp += i;
1564 p = q;
1565 if (p < name + namelen && *p == '.')
1566 p++;
1567 }
1568 /* termination */
1569 if (cp + nterm >= ep)
1570 goto fail;
1571 while (nterm-- > 0)
1572 *cp++ = '\0';
1573 m->m_len = cp - mtod(m, char *);
1574 return m;
1575 }
1576
1577 panic("should not reach here");
1578 /* NOTREACHED */
1579
1580 fail:
1581 if (m)
1582 m_freem(m);
1583 return NULL;
1584 }
1585
1586 /*
1587 * check if two DNS-encoded string matches. takes care of truncated
1588 * form (with \0\0 at the end). no compression support.
1589 * XXX upper/lowercase match (see RFC2065)
1590 */
1591 static int
1592 ni6_dnsmatch(a, alen, b, blen)
1593 const char *a;
1594 int alen;
1595 const char *b;
1596 int blen;
1597 {
1598 const char *a0, *b0;
1599 int l;
1600
1601 /* simplest case - need validation? */
1602 if (alen == blen && bcmp(a, b, alen) == 0)
1603 return 1;
1604
1605 a0 = a;
1606 b0 = b;
1607
1608 /* termination is mandatory */
1609 if (alen < 2 || blen < 2)
1610 return 0;
1611 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1612 return 0;
1613 alen--;
1614 blen--;
1615
1616 while (a - a0 < alen && b - b0 < blen) {
1617 if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1618 return 0;
1619
1620 if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1621 return 0;
1622 /* we don't support compression yet */
1623 if (a[0] >= 64 || b[0] >= 64)
1624 return 0;
1625
1626 /* truncated case */
1627 if (a[0] == 0 && a - a0 == alen - 1)
1628 return 1;
1629 if (b[0] == 0 && b - b0 == blen - 1)
1630 return 1;
1631 if (a[0] == 0 || b[0] == 0)
1632 return 0;
1633
1634 if (a[0] != b[0])
1635 return 0;
1636 l = a[0];
1637 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1638 return 0;
1639 if (bcmp(a + 1, b + 1, l) != 0)
1640 return 0;
1641
1642 a += 1 + l;
1643 b += 1 + l;
1644 }
1645
1646 if (a - a0 == alen && b - b0 == blen)
1647 return 1;
1648 else
1649 return 0;
1650 }
1651
1652 /*
1653 * calculate the number of addresses to be returned in the node info reply.
1654 */
1655 static int
1656 ni6_addrs(ni6, m, ifpp, subj)
1657 struct icmp6_nodeinfo *ni6;
1658 struct mbuf *m;
1659 struct ifnet **ifpp;
1660 char *subj;
1661 {
1662 struct ifnet *ifp;
1663 struct in6_ifaddr *ifa6;
1664 struct ifaddr *ifa;
1665 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1666 int addrs = 0, addrsofif, iffound = 0;
1667 int niflags = ni6->ni_flags;
1668
1669 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1670 switch (ni6->ni_code) {
1671 case ICMP6_NI_SUBJ_IPV6:
1672 if (subj == NULL) /* must be impossible... */
1673 return (0);
1674 subj_ip6 = (struct sockaddr_in6 *)subj;
1675 break;
1676 default:
1677 /*
1678 * XXX: we only support IPv6 subject address for
1679 * this Qtype.
1680 */
1681 return (0);
1682 }
1683 }
1684
1685 IFNET_RLOCK();
1686 for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1687 addrsofif = 0;
1688 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1689 if (ifa->ifa_addr->sa_family != AF_INET6)
1690 continue;
1691 ifa6 = (struct in6_ifaddr *)ifa;
1692
1693 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1694 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1695 &ifa6->ia_addr.sin6_addr))
1696 iffound = 1;
1697
1698 /*
1699 * IPv4-mapped addresses can only be returned by a
1700 * Node Information proxy, since they represent
1701 * addresses of IPv4-only nodes, which perforce do
1702 * not implement this protocol.
1703 * [icmp-name-lookups-07, Section 5.4]
1704 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1705 * this function at this moment.
1706 */
1707
1708 /* What do we have to do about ::1? */
1709 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1710 case IPV6_ADDR_SCOPE_LINKLOCAL:
1711 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1712 continue;
1713 break;
1714 case IPV6_ADDR_SCOPE_SITELOCAL:
1715 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1716 continue;
1717 break;
1718 case IPV6_ADDR_SCOPE_GLOBAL:
1719 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1720 continue;
1721 break;
1722 default:
1723 continue;
1724 }
1725
1726 /*
1727 * check if anycast is okay.
1728 * XXX: just experimental. not in the spec.
1729 */
1730 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1731 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1732 continue; /* we need only unicast addresses */
1733 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1734 (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1735 continue;
1736 }
1737 addrsofif++; /* count the address */
1738 }
1739 if (iffound) {
1740 *ifpp = ifp;
1741 IFNET_RUNLOCK();
1742 return (addrsofif);
1743 }
1744
1745 addrs += addrsofif;
1746 }
1747 IFNET_RUNLOCK();
1748
1749 return (addrs);
1750 }
1751
1752 static int
1753 ni6_store_addrs(ni6, nni6, ifp0, resid)
1754 struct icmp6_nodeinfo *ni6, *nni6;
1755 struct ifnet *ifp0;
1756 int resid;
1757 {
1758 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1759 struct in6_ifaddr *ifa6;
1760 struct ifaddr *ifa;
1761 struct ifnet *ifp_dep = NULL;
1762 int copied = 0, allow_deprecated = 0;
1763 u_char *cp = (u_char *)(nni6 + 1);
1764 int niflags = ni6->ni_flags;
1765 u_int32_t ltime;
1766
1767 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1768 return (0); /* needless to copy */
1769
1770 IFNET_RLOCK();
1771 again:
1772
1773 for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1774 for (ifa = ifp->if_addrlist.tqh_first; ifa;
1775 ifa = ifa->ifa_list.tqe_next) {
1776 if (ifa->ifa_addr->sa_family != AF_INET6)
1777 continue;
1778 ifa6 = (struct in6_ifaddr *)ifa;
1779
1780 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1781 allow_deprecated == 0) {
1782 /*
1783 * prefererred address should be put before
1784 * deprecated addresses.
1785 */
1786
1787 /* record the interface for later search */
1788 if (ifp_dep == NULL)
1789 ifp_dep = ifp;
1790
1791 continue;
1792 } else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1793 allow_deprecated != 0)
1794 continue; /* we now collect deprecated addrs */
1795
1796 /* What do we have to do about ::1? */
1797 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1798 case IPV6_ADDR_SCOPE_LINKLOCAL:
1799 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1800 continue;
1801 break;
1802 case IPV6_ADDR_SCOPE_SITELOCAL:
1803 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1804 continue;
1805 break;
1806 case IPV6_ADDR_SCOPE_GLOBAL:
1807 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1808 continue;
1809 break;
1810 default:
1811 continue;
1812 }
1813
1814 /*
1815 * check if anycast is okay.
1816 * XXX: just experimental. not in the spec.
1817 */
1818 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1819 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1820 continue;
1821 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1822 (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1823 continue;
1824 }
1825
1826 /* now we can copy the address */
1827 if (resid < sizeof(struct in6_addr) +
1828 sizeof(u_int32_t)) {
1829 /*
1830 * We give up much more copy.
1831 * Set the truncate flag and return.
1832 */
1833 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1834 IFNET_RUNLOCK();
1835 return (copied);
1836 }
1837
1838 /*
1839 * Set the TTL of the address.
1840 * The TTL value should be one of the following
1841 * according to the specification:
1842 *
1843 * 1. The remaining lifetime of a DHCP lease on the
1844 * address, or
1845 * 2. The remaining Valid Lifetime of a prefix from
1846 * which the address was derived through Stateless
1847 * Autoconfiguration.
1848 *
1849 * Note that we currently do not support stateful
1850 * address configuration by DHCPv6, so the former
1851 * case can't happen.
1852 */
1853 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1854 ltime = ND6_INFINITE_LIFETIME;
1855 else {
1856 if (ifa6->ia6_lifetime.ia6t_expire >
1857 time_second)
1858 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1859 else
1860 ltime = 0;
1861 }
1862
1863 bcopy(<ime, cp, sizeof(u_int32_t));
1864 cp += sizeof(u_int32_t);
1865
1866 /* copy the address itself */
1867 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1868 sizeof(struct in6_addr));
1869 in6_clearscope((struct in6_addr *)cp); /* XXX */
1870 cp += sizeof(struct in6_addr);
1871
1872 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1873 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1874 }
1875 if (ifp0) /* we need search only on the specified IF */
1876 break;
1877 }
1878
1879 if (allow_deprecated == 0 && ifp_dep != NULL) {
1880 ifp = ifp_dep;
1881 allow_deprecated = 1;
1882
1883 goto again;
1884 }
1885
1886 IFNET_RUNLOCK();
1887
1888 return (copied);
1889 }
1890
1891 /*
1892 * XXX almost dup'ed code with rip6_input.
1893 */
1894 static int
1895 icmp6_rip6_input(mp, off)
1896 struct mbuf **mp;
1897 int off;
1898 {
1899 struct mbuf *m = *mp;
1900 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1901 struct in6pcb *in6p;
1902 struct in6pcb *last = NULL;
1903 struct sockaddr_in6 fromsa;
1904 struct icmp6_hdr *icmp6;
1905 struct mbuf *opts = NULL;
1906
1907 #ifndef PULLDOWN_TEST
1908 /* this is assumed to be safe. */
1909 icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1910 #else
1911 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1912 if (icmp6 == NULL) {
1913 /* m is already reclaimed */
1914 return (IPPROTO_DONE);
1915 }
1916 #endif
1917
1918 bzero(&fromsa, sizeof(fromsa));
1919 fromsa.sin6_len = sizeof(struct sockaddr_in6);
1920 fromsa.sin6_family = AF_INET6;
1921 /* KAME hack: recover scopeid */
1922 (void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1923
1924 INP_INFO_RLOCK(&ripcbinfo);
1925 LIST_FOREACH(in6p, &ripcb, inp_list) {
1926 INP_LOCK(in6p);
1927 if ((in6p->inp_vflag & INP_IPV6) == 0) {
1928 docontinue:
1929 INP_UNLOCK(in6p);
1930 continue;
1931 }
1932 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1933 goto docontinue;
1934 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1935 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1936 goto docontinue;
1937 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1938 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1939 goto docontinue;
1940 if (in6p->in6p_icmp6filt
1941 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1942 in6p->in6p_icmp6filt))
1943 goto docontinue;
1944 if (last) {
1945 struct mbuf *n = NULL;
1946
1947 /*
1948 * Recent network drivers tend to allocate a single
1949 * mbuf cluster, rather than to make a couple of
1950 * mbufs without clusters. Also, since the IPv6 code
1951 * path tries to avoid m_pullup(), it is highly
1952 * probable that we still have an mbuf cluster here
1953 * even though the necessary length can be stored in an
1954 * mbuf's internal buffer.
1955 * Meanwhile, the default size of the receive socket
1956 * buffer for raw sockets is not so large. This means
1957 * the possibility of packet loss is relatively higher
1958 * than before. To avoid this scenario, we copy the
1959 * received data to a separate mbuf that does not use
1960 * a cluster, if possible.
1961 * XXX: it is better to copy the data after stripping
1962 * intermediate headers.
1963 */
1964 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1965 m->m_len <= MHLEN) {
1966 MGET(n, M_DONTWAIT, m->m_type);
1967 if (n != NULL) {
1968 m_dup_pkthdr(n, m, M_NOWAIT);
1969 bcopy(m->m_data, n->m_data, m->m_len);
1970 n->m_len = m->m_len;
1971 }
1972 }
1973 if (n != NULL ||
1974 (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1975 if (last->in6p_flags & IN6P_CONTROLOPTS)
1976 ip6_savecontrol(last, n, &opts);
1977 /* strip intermediate headers */
1978 m_adj(n, off);
1979 if (sbappendaddr(&last->in6p_socket->so_rcv,
1980 (struct sockaddr *)&fromsa, n, opts)
1981 == 0) {
1982 /* should notify about lost packet */
1983 m_freem(n);
1984 if (opts) {
1985 m_freem(opts);
1986 }
1987 } else
1988 sorwakeup(last->in6p_socket);
1989 opts = NULL;
1990 }
1991 INP_UNLOCK(last);
1992 }
1993 last = in6p;
1994 }
1995 if (last) {
1996 if (last->in6p_flags & IN6P_CONTROLOPTS)
1997 ip6_savecontrol(last, m, &opts);
1998 /* strip intermediate headers */
1999 m_adj(m, off);
2000
2001 /* avoid using mbuf clusters if possible (see above) */
2002 if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2003 m->m_len <= MHLEN) {
2004 struct mbuf *n;
2005
2006 MGET(n, M_DONTWAIT, m->m_type);
2007 if (n != NULL) {
2008 m_dup_pkthdr(n, m, M_NOWAIT);
2009 bcopy(m->m_data, n->m_data, m->m_len);
2010 n->m_len = m->m_len;
2011
2012 m_freem(m);
2013 m = n;
2014 }
2015 }
2016 if (sbappendaddr(&last->in6p_socket->so_rcv,
2017 (struct sockaddr *)&fromsa, m, opts) == 0) {
2018 m_freem(m);
2019 if (opts)
2020 m_freem(opts);
2021 } else
2022 sorwakeup(last->in6p_socket);
2023 INP_UNLOCK(last);
2024 } else {
2025 m_freem(m);
2026 ip6stat.ip6s_delivered--;
2027 }
2028 INP_INFO_RUNLOCK(&ripcbinfo);
2029 return IPPROTO_DONE;
2030 }
2031
2032 /*
2033 * Reflect the ip6 packet back to the source.
2034 * OFF points to the icmp6 header, counted from the top of the mbuf.
2035 */
2036 void
2037 icmp6_reflect(m, off)
2038 struct mbuf *m;
2039 size_t off;
2040 {
2041 struct ip6_hdr *ip6;
2042 struct icmp6_hdr *icmp6;
2043 struct in6_ifaddr *ia;
2044 struct in6_addr t, *src = 0;
2045 int plen;
2046 int type, code;
2047 struct ifnet *outif = NULL;
2048 struct sockaddr_in6 sa6_src, sa6_dst;
2049 #ifdef COMPAT_RFC1885
2050 int mtu = IPV6_MMTU;
2051 struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2052 #endif
2053
2054 /* too short to reflect */
2055 if (off < sizeof(struct ip6_hdr)) {
2056 nd6log((LOG_DEBUG,
2057 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2058 (u_long)off, (u_long)sizeof(struct ip6_hdr),
2059 __FILE__, __LINE__));
2060 goto bad;
2061 }
2062
2063 /*
2064 * If there are extra headers between IPv6 and ICMPv6, strip
2065 * off that header first.
2066 */
2067 #ifdef DIAGNOSTIC
2068 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2069 panic("assumption failed in icmp6_reflect");
2070 #endif
2071 if (off > sizeof(struct ip6_hdr)) {
2072 size_t l;
2073 struct ip6_hdr nip6;
2074
2075 l = off - sizeof(struct ip6_hdr);
2076 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2077 m_adj(m, l);
2078 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2079 if (m->m_len < l) {
2080 if ((m = m_pullup(m, l)) == NULL)
2081 return;
2082 }
2083 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2084 } else /* off == sizeof(struct ip6_hdr) */ {
2085 size_t l;
2086 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2087 if (m->m_len < l) {
2088 if ((m = m_pullup(m, l)) == NULL)
2089 return;
2090 }
2091 }
2092 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2093 ip6 = mtod(m, struct ip6_hdr *);
2094 ip6->ip6_nxt = IPPROTO_ICMPV6;
2095 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2096 type = icmp6->icmp6_type; /* keep type for statistics */
2097 code = icmp6->icmp6_code; /* ditto. */
2098
2099 t = ip6->ip6_dst;
2100 /*
2101 * ip6_input() drops a packet if its src is multicast.
2102 * So, the src is never multicast.
2103 */
2104 ip6->ip6_dst = ip6->ip6_src;
2105
2106 /*
2107 * XXX: make sure to embed scope zone information, using
2108 * already embedded IDs or the received interface (if any).
2109 * Note that rcvif may be NULL.
2110 * TODO: scoped routing case (XXX).
2111 */
2112 bzero(&sa6_src, sizeof(sa6_src));
2113 sa6_src.sin6_family = AF_INET6;
2114 sa6_src.sin6_len = sizeof(sa6_src);
2115 sa6_src.sin6_addr = ip6->ip6_dst;
2116 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2117 if (in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL))
2118 goto bad;
2119 bzero(&sa6_dst, sizeof(sa6_dst));
2120 sa6_dst.sin6_family = AF_INET6;
2121 sa6_dst.sin6_len = sizeof(sa6_dst);
2122 sa6_dst.sin6_addr = t;
2123 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2124 if (in6_embedscope(&t, &sa6_dst, NULL, NULL))
2125 goto bad;
2126
2127 #ifdef COMPAT_RFC1885
2128 /*
2129 * xxx guess MTU
2130 * RFC 1885 requires that echo reply should be truncated if it
2131 * does not fit in with (return) path MTU, but the description was
2132 * removed in the new spec.
2133 */
2134 if (icmp6_reflect_rt.ro_rt == 0 ||
2135 ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2136 if (icmp6_reflect_rt.ro_rt) {
2137 RTFREE(icmp6_reflect_rt.ro_rt);
2138 icmp6_reflect_rt.ro_rt = 0;
2139 }
2140 bzero(sin6, sizeof(*sin6));
2141 sin6->sin6_family = PF_INET6;
2142 sin6->sin6_len = sizeof(struct sockaddr_in6);
2143 sin6->sin6_addr = ip6->ip6_dst;
2144
2145 rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2146 }
2147
2148 if (icmp6_reflect_rt.ro_rt == 0)
2149 goto bad;
2150
2151 if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2152 && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2153 mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2154
2155 if (mtu < m->m_pkthdr.len) {
2156 plen -= (m->m_pkthdr.len - mtu);
2157 m_adj(m, mtu - m->m_pkthdr.len);
2158 }
2159 #endif
2160 /*
2161 * If the incoming packet was addressed directly to us (i.e. unicast),
2162 * use dst as the src for the reply.
2163 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2164 * (for example) when we encounter an error while forwarding procedure
2165 * destined to a duplicated address of ours.
2166 */
2167 for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2168 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2169 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2170 src = &t;
2171 break;
2172 }
2173 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2174 /*
2175 * This is the case if the dst is our link-local address
2176 * and the sender is also ourselves.
2177 */
2178 src = &t;
2179 }
2180
2181 if (src == 0) {
2182 int e;
2183 struct route_in6 ro;
2184
2185 /*
2186 * This case matches to multicasts, our anycast, or unicasts
2187 * that we do not own. Select a source address based on the
2188 * source address of the erroneous packet.
2189 */
2190 bzero(&ro, sizeof(ro));
2191 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2192 if (ro.ro_rt)
2193 RTFREE(ro.ro_rt); /* XXX: we could use this */
2194 if (src == NULL) {
2195 nd6log((LOG_DEBUG,
2196 "icmp6_reflect: source can't be determined: "
2197 "dst=%s, error=%d\n",
2198 ip6_sprintf(&sa6_src.sin6_addr), e));
2199 goto bad;
2200 }
2201 }
2202
2203 ip6->ip6_src = *src;
2204
2205 ip6->ip6_flow = 0;
2206 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2207 ip6->ip6_vfc |= IPV6_VERSION;
2208 ip6->ip6_nxt = IPPROTO_ICMPV6;
2209 if (outif)
2210 ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2211 else if (m->m_pkthdr.rcvif) {
2212 /* XXX: This may not be the outgoing interface */
2213 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2214 } else
2215 ip6->ip6_hlim = ip6_defhlim;
2216
2217 icmp6->icmp6_cksum = 0;
2218 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2219 sizeof(struct ip6_hdr), plen);
2220
2221 /*
2222 * XXX option handling
2223 */
2224
2225 m->m_flags &= ~(M_BCAST|M_MCAST);
2226
2227 #ifdef COMPAT_RFC1885
2228 ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2229 #else
2230 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2231 #endif
2232 if (outif)
2233 icmp6_ifoutstat_inc(outif, type, code);
2234
2235 return;
2236
2237 bad:
2238 m_freem(m);
2239 return;
2240 }
2241
2242 void
2243 icmp6_fasttimo()
2244 {
2245
2246 mld6_fasttimeo();
2247 }
2248
2249 static const char *
2250 icmp6_redirect_diag(src6, dst6, tgt6)
2251 struct in6_addr *src6;
2252 struct in6_addr *dst6;
2253 struct in6_addr *tgt6;
2254 {
2255 static char buf[1024];
2256 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2257 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2258 return buf;
2259 }
2260
2261 void
2262 icmp6_redirect_input(m, off)
2263 struct mbuf *m;
2264 int off;
2265 {
2266 struct ifnet *ifp = m->m_pkthdr.rcvif;
2267 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2268 struct nd_redirect *nd_rd;
2269 int icmp6len = ntohs(ip6->ip6_plen);
2270 char *lladdr = NULL;
2271 int lladdrlen = 0;
2272 u_char *redirhdr = NULL;
2273 int redirhdrlen = 0;
2274 struct rtentry *rt = NULL;
2275 int is_router;
2276 int is_onlink;
2277 struct in6_addr src6 = ip6->ip6_src;
2278 struct in6_addr redtgt6;
2279 struct in6_addr reddst6;
2280 union nd_opts ndopts;
2281
2282 if (!m || !ifp)
2283 return;
2284
2285 /* XXX if we are router, we don't update route by icmp6 redirect */
2286 if (ip6_forwarding)
2287 goto freeit;
2288 if (!icmp6_rediraccept)
2289 goto freeit;
2290
2291 #ifndef PULLDOWN_TEST
2292 IP6_EXTHDR_CHECK(m, off, icmp6len,);
2293 nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2294 #else
2295 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2296 if (nd_rd == NULL) {
2297 icmp6stat.icp6s_tooshort++;
2298 return;
2299 }
2300 #endif
2301 redtgt6 = nd_rd->nd_rd_target;
2302 reddst6 = nd_rd->nd_rd_dst;
2303
2304 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2305 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2306 if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2307 reddst6.s6_addr16[1] = htons(ifp->if_index);
2308
2309 /* validation */
2310 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2311 nd6log((LOG_ERR,
2312 "ICMP6 redirect sent from %s rejected; "
2313 "must be from linklocal\n",
2314 ip6_sprintf(&src6)));
2315 goto bad;
2316 }
2317 if (ip6->ip6_hlim != 255) {
2318 nd6log((LOG_ERR,
2319 "ICMP6 redirect sent from %s rejected; "
2320 "hlim=%d (must be 255)\n",
2321 ip6_sprintf(&src6), ip6->ip6_hlim));
2322 goto bad;
2323 }
2324 {
2325 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2326 struct sockaddr_in6 sin6;
2327 struct in6_addr *gw6;
2328
2329 bzero(&sin6, sizeof(sin6));
2330 sin6.sin6_family = AF_INET6;
2331 sin6.sin6_len = sizeof(struct sockaddr_in6);
2332 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2333 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2334 if (rt) {
2335 if (rt->rt_gateway == NULL ||
2336 rt->rt_gateway->sa_family != AF_INET6) {
2337 nd6log((LOG_ERR,
2338 "ICMP6 redirect rejected; no route "
2339 "with inet6 gateway found for redirect dst: %s\n",
2340 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2341 RTFREE_LOCKED(rt);
2342 goto bad;
2343 }
2344
2345 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2346 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2347 nd6log((LOG_ERR,
2348 "ICMP6 redirect rejected; "
2349 "not equal to gw-for-src=%s (must be same): "
2350 "%s\n",
2351 ip6_sprintf(gw6),
2352 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2353 RTFREE_LOCKED(rt);
2354 goto bad;
2355 }
2356 } else {
2357 nd6log((LOG_ERR,
2358 "ICMP6 redirect rejected; "
2359 "no route found for redirect dst: %s\n",
2360 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2361 goto bad;
2362 }
2363 RTFREE_LOCKED(rt);
2364 rt = NULL;
2365 }
2366 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2367 nd6log((LOG_ERR,
2368 "ICMP6 redirect rejected; "
2369 "redirect dst must be unicast: %s\n",
2370 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2371 goto bad;
2372 }
2373
2374 is_router = is_onlink = 0;
2375 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2376 is_router = 1; /* router case */
2377 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2378 is_onlink = 1; /* on-link destination case */
2379 if (!is_router && !is_onlink) {
2380 nd6log((LOG_ERR,
2381 "ICMP6 redirect rejected; "
2382 "neither router case nor onlink case: %s\n",
2383 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2384 goto bad;
2385 }
2386 /* validation passed */
2387
2388 icmp6len -= sizeof(*nd_rd);
2389 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2390 if (nd6_options(&ndopts) < 0) {
2391 nd6log((LOG_INFO, "icmp6_redirect_input: "
2392 "invalid ND option, rejected: %s\n",
2393 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2394 /* nd6_options have incremented stats */
2395 goto freeit;
2396 }
2397
2398 if (ndopts.nd_opts_tgt_lladdr) {
2399 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2400 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2401 }
2402
2403 if (ndopts.nd_opts_rh) {
2404 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2405 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2406 }
2407
2408 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2409 nd6log((LOG_INFO,
2410 "icmp6_redirect_input: lladdrlen mismatch for %s "
2411 "(if %d, icmp6 packet %d): %s\n",
2412 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2413 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2414 goto bad;
2415 }
2416
2417 /* RFC 2461 8.3 */
2418 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2419 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2420
2421 if (!is_onlink) { /* better router case. perform rtredirect. */
2422 /* perform rtredirect */
2423 struct sockaddr_in6 sdst;
2424 struct sockaddr_in6 sgw;
2425 struct sockaddr_in6 ssrc;
2426
2427 bzero(&sdst, sizeof(sdst));
2428 bzero(&sgw, sizeof(sgw));
2429 bzero(&ssrc, sizeof(ssrc));
2430 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2431 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2432 sizeof(struct sockaddr_in6);
2433 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2434 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2435 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2436 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2437 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2438 (struct sockaddr *)&ssrc);
2439 }
2440 /* finally update cached route in each socket via pfctlinput */
2441 {
2442 struct sockaddr_in6 sdst;
2443
2444 bzero(&sdst, sizeof(sdst));
2445 sdst.sin6_family = AF_INET6;
2446 sdst.sin6_len = sizeof(struct sockaddr_in6);
2447 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2448 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2449 #if defined(IPSEC) || defined(FAST_IPSEC)
2450 key_sa_routechange((struct sockaddr *)&sdst);
2451 #endif
2452 }
2453
2454 freeit:
2455 m_freem(m);
2456 return;
2457
2458 bad:
2459 icmp6stat.icp6s_badredirect++;
2460 m_freem(m);
2461 }
2462
2463 void
2464 icmp6_redirect_output(m0, rt)
2465 struct mbuf *m0;
2466 struct rtentry *rt;
2467 {
2468 struct ifnet *ifp; /* my outgoing interface */
2469 struct in6_addr *ifp_ll6;
2470 struct in6_addr *router_ll6;
2471 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2472 struct mbuf *m = NULL; /* newly allocated one */
2473 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2474 struct nd_redirect *nd_rd;
2475 size_t maxlen;
2476 u_char *p;
2477 struct ifnet *outif = NULL;
2478 struct sockaddr_in6 src_sa;
2479
2480 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2481
2482 /* if we are not router, we don't send icmp6 redirect */
2483 if (!ip6_forwarding || ip6_accept_rtadv)
2484 goto fail;
2485
2486 /* sanity check */
2487 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2488 goto fail;
2489
2490 /*
2491 * Address check:
2492 * the source address must identify a neighbor, and
2493 * the destination address must not be a multicast address
2494 * [RFC 2461, sec 8.2]
2495 */
2496 sip6 = mtod(m0, struct ip6_hdr *);
2497 bzero(&src_sa, sizeof(src_sa));
2498 src_sa.sin6_family = AF_INET6;
2499 src_sa.sin6_len = sizeof(src_sa);
2500 src_sa.sin6_addr = sip6->ip6_src;
2501 /* we don't currently use sin6_scope_id, but eventually use it */
2502 if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2503 goto fail;
2504 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2505 goto fail;
2506 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2507 goto fail; /* what should we do here? */
2508
2509 /* rate limit */
2510 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2511 goto fail;
2512
2513 /*
2514 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2515 * we almost always ask for an mbuf cluster for simplicity.
2516 * (MHLEN < IPV6_MMTU is almost always true)
2517 */
2518 #if IPV6_MMTU >= MCLBYTES
2519 # error assumption failed about IPV6_MMTU and MCLBYTES
2520 #endif
2521 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2522 if (m && IPV6_MMTU >= MHLEN)
2523 MCLGET(m, M_DONTWAIT);
2524 if (!m)
2525 goto fail;
2526 m->m_pkthdr.rcvif = NULL;
2527 m->m_len = 0;
2528 maxlen = M_TRAILINGSPACE(m);
2529 maxlen = min(IPV6_MMTU, maxlen);
2530 /* just for safety */
2531 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2532 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2533 goto fail;
2534 }
2535
2536 {
2537 /* get ip6 linklocal address for ifp(my outgoing interface). */
2538 struct in6_ifaddr *ia;
2539 if ((ia = in6ifa_ifpforlinklocal(ifp,
2540 IN6_IFF_NOTREADY|
2541 IN6_IFF_ANYCAST)) == NULL)
2542 goto fail;
2543 ifp_ll6 = &ia->ia_addr.sin6_addr;
2544 }
2545
2546 /* get ip6 linklocal address for the router. */
2547 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2548 struct sockaddr_in6 *sin6;
2549 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2550 router_ll6 = &sin6->sin6_addr;
2551 if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2552 router_ll6 = (struct in6_addr *)NULL;
2553 } else
2554 router_ll6 = (struct in6_addr *)NULL;
2555
2556 /* ip6 */
2557 ip6 = mtod(m, struct ip6_hdr *);
2558 ip6->ip6_flow = 0;
2559 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2560 ip6->ip6_vfc |= IPV6_VERSION;
2561 /* ip6->ip6_plen will be set later */
2562 ip6->ip6_nxt = IPPROTO_ICMPV6;
2563 ip6->ip6_hlim = 255;
2564 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2565 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2566 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2567
2568 /* ND Redirect */
2569 nd_rd = (struct nd_redirect *)(ip6 + 1);
2570 nd_rd->nd_rd_type = ND_REDIRECT;
2571 nd_rd->nd_rd_code = 0;
2572 nd_rd->nd_rd_reserved = 0;
2573 if (rt->rt_flags & RTF_GATEWAY) {
2574 /*
2575 * nd_rd->nd_rd_target must be a link-local address in
2576 * better router cases.
2577 */
2578 if (!router_ll6)
2579 goto fail;
2580 bcopy(router_ll6, &nd_rd->nd_rd_target,
2581 sizeof(nd_rd->nd_rd_target));
2582 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2583 sizeof(nd_rd->nd_rd_dst));
2584 } else {
2585 /* make sure redtgt == reddst */
2586 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2587 sizeof(nd_rd->nd_rd_target));
2588 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2589 sizeof(nd_rd->nd_rd_dst));
2590 }
2591
2592 p = (u_char *)(nd_rd + 1);
2593
2594 if (!router_ll6)
2595 goto nolladdropt;
2596
2597 {
2598 /* target lladdr option */
2599 struct rtentry *rt_router = NULL;
2600 int len;
2601 struct sockaddr_dl *sdl;
2602 struct nd_opt_hdr *nd_opt;
2603 char *lladdr;
2604
2605 rt_router = nd6_lookup(router_ll6, 0, ifp);
2606 if (!rt_router)
2607 goto nolladdropt;
2608 len = sizeof(*nd_opt) + ifp->if_addrlen;
2609 len = (len + 7) & ~7; /* round by 8 */
2610 /* safety check */
2611 if (len + (p - (u_char *)ip6) > maxlen)
2612 goto nolladdropt;
2613 if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2614 (rt_router->rt_flags & RTF_LLINFO) &&
2615 (rt_router->rt_gateway->sa_family == AF_LINK) &&
2616 (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2617 sdl->sdl_alen) {
2618 nd_opt = (struct nd_opt_hdr *)p;
2619 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2620 nd_opt->nd_opt_len = len >> 3;
2621 lladdr = (char *)(nd_opt + 1);
2622 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2623 p += len;
2624 }
2625 }
2626 nolladdropt:;
2627
2628 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2629
2630 /* just to be safe */
2631 #ifdef M_DECRYPTED /*not openbsd*/
2632 if (m0->m_flags & M_DECRYPTED)
2633 goto noredhdropt;
2634 #endif
2635 if (p - (u_char *)ip6 > maxlen)
2636 goto noredhdropt;
2637
2638 {
2639 /* redirected header option */
2640 int len;
2641 struct nd_opt_rd_hdr *nd_opt_rh;
2642
2643 /*
2644 * compute the maximum size for icmp6 redirect header option.
2645 * XXX room for auth header?
2646 */
2647 len = maxlen - (p - (u_char *)ip6);
2648 len &= ~7;
2649
2650 /* This is just for simplicity. */
2651 if (m0->m_pkthdr.len != m0->m_len) {
2652 if (m0->m_next) {
2653 m_freem(m0->m_next);
2654 m0->m_next = NULL;
2655 }
2656 m0->m_pkthdr.len = m0->m_len;
2657 }
2658
2659 /*
2660 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2661 * about padding/truncate rule for the original IP packet.
2662 * From the discussion on IPv6imp in Feb 1999,
2663 * the consensus was:
2664 * - "attach as much as possible" is the goal
2665 * - pad if not aligned (original size can be guessed by
2666 * original ip6 header)
2667 * Following code adds the padding if it is simple enough,
2668 * and truncates if not.
2669 */
2670 if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2671 panic("assumption failed in %s:%d", __FILE__,
2672 __LINE__);
2673
2674 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2675 /* not enough room, truncate */
2676 m0->m_pkthdr.len = m0->m_len = len -
2677 sizeof(*nd_opt_rh);
2678 } else {
2679 /* enough room, pad or truncate */
2680 size_t extra;
2681
2682 extra = m0->m_pkthdr.len % 8;
2683 if (extra) {
2684 /* pad if easy enough, truncate if not */
2685 if (8 - extra <= M_TRAILINGSPACE(m0)) {
2686 /* pad */
2687 m0->m_len += (8 - extra);
2688 m0->m_pkthdr.len += (8 - extra);
2689 } else {
2690 /* truncate */
2691 m0->m_pkthdr.len -= extra;
2692 m0->m_len -= extra;
2693 }
2694 }
2695 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2696 m0->m_pkthdr.len = m0->m_len = len -
2697 sizeof(*nd_opt_rh);
2698 }
2699
2700 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2701 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2702 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2703 nd_opt_rh->nd_opt_rh_len = len >> 3;
2704 p += sizeof(*nd_opt_rh);
2705 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2706
2707 /* connect m0 to m */
2708 m_tag_delete_chain(m0, NULL);
2709 m0->m_flags &= ~M_PKTHDR;
2710 m->m_next = m0;
2711 m->m_pkthdr.len = m->m_len + m0->m_len;
2712 m0 = NULL;
2713 }
2714 noredhdropt:;
2715 if (m0) {
2716 m_freem(m0);
2717 m0 = NULL;
2718 }
2719
2720 /* XXX: clear embedded link IDs in the inner header */
2721 in6_clearscope(&sip6->ip6_src);
2722 in6_clearscope(&sip6->ip6_dst);
2723 in6_clearscope(&nd_rd->nd_rd_target);
2724 in6_clearscope(&nd_rd->nd_rd_dst);
2725
2726 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2727
2728 nd_rd->nd_rd_cksum = 0;
2729 nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2730 sizeof(*ip6), ntohs(ip6->ip6_plen));
2731
2732 /* send the packet to outside... */
2733 ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2734 if (outif) {
2735 icmp6_ifstat_inc(outif, ifs6_out_msg);
2736 icmp6_ifstat_inc(outif, ifs6_out_redirect);
2737 }
2738 icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2739
2740 return;
2741
2742 fail:
2743 if (m)
2744 m_freem(m);
2745 if (m0)
2746 m_freem(m0);
2747 }
2748
2749 /*
2750 * ICMPv6 socket option processing.
2751 */
2752 int
2753 icmp6_ctloutput(so, sopt)
2754 struct socket *so;
2755 struct sockopt *sopt;
2756 {
2757 int error = 0;
2758 int optlen;
2759 struct inpcb *inp = sotoinpcb(so);
2760 int level, op, optname;
2761
2762 if (sopt) {
2763 level = sopt->sopt_level;
2764 op = sopt->sopt_dir;
2765 optname = sopt->sopt_name;
2766 optlen = sopt->sopt_valsize;
2767 } else
2768 level = op = optname = optlen = 0;
2769
2770 if (level != IPPROTO_ICMPV6) {
2771 return EINVAL;
2772 }
2773
2774 switch (op) {
2775 case PRCO_SETOPT:
2776 switch (optname) {
2777 case ICMP6_FILTER:
2778 {
2779 struct icmp6_filter *p;
2780
2781 if (optlen != sizeof(*p)) {
2782 error = EMSGSIZE;
2783 break;
2784 }
2785 if (inp->in6p_icmp6filt == NULL) {
2786 error = EINVAL;
2787 break;
2788 }
2789 error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2790 optlen);
2791 break;
2792 }
2793
2794 default:
2795 error = ENOPROTOOPT;
2796 break;
2797 }
2798 break;
2799
2800 case PRCO_GETOPT:
2801 switch (optname) {
2802 case ICMP6_FILTER:
2803 {
2804 if (inp->in6p_icmp6filt == NULL) {
2805 error = EINVAL;
2806 break;
2807 }
2808 error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2809 sizeof(struct icmp6_filter));
2810 break;
2811 }
2812
2813 default:
2814 error = ENOPROTOOPT;
2815 break;
2816 }
2817 break;
2818 }
2819
2820 return (error);
2821 }
2822
2823 /*
2824 * Perform rate limit check.
2825 * Returns 0 if it is okay to send the icmp6 packet.
2826 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2827 * limitation.
2828 *
2829 * XXX per-destination/type check necessary?
2830 */
2831 static int
2832 icmp6_ratelimit(dst, type, code)
2833 const struct in6_addr *dst; /* not used at this moment */
2834 const int type; /* not used at this moment */
2835 const int code; /* not used at this moment */
2836 {
2837 int ret;
2838
2839 ret = 0; /* okay to send */
2840
2841 /* PPS limit */
2842 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2843 icmp6errppslim)) {
2844 /* The packet is subject to rate limit */
2845 ret++;
2846 }
2847
2848 return ret;
2849 }
Cache object: 71f0584800adeba0e25ffd768991278f
|