1 /* $OpenBSD: icmp6.c,v 1.94 2007/06/01 00:52:38 henning Exp $ */
2 /* $KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei 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 * 3. 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 "faith.h"
65 #include "carp.h"
66
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/protosw.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/time.h>
75 #include <sys/kernel.h>
76 #include <sys/syslog.h>
77 #include <sys/domain.h>
78
79 #include <net/if.h>
80 #include <net/route.h>
81 #include <net/if_dl.h>
82 #include <net/if_types.h>
83
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/ip6_var.h>
90 #include <netinet/icmp6.h>
91 #include <netinet6/mld6_var.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet6/nd6.h>
94 #include <netinet6/in6_ifattach.h>
95 #include <netinet6/ip6protosw.h>
96
97 #if NCARP > 0
98 #include <netinet/ip_carp.h>
99 #endif
100
101 /* inpcb members */
102 #define in6pcb inpcb
103 #define in6p_laddr inp_laddr6
104 #define in6p_faddr inp_faddr6
105 #define in6p_icmp6filt inp_icmp6filt
106 #define in6p_route inp_route
107 #define in6p_socket inp_socket
108 #define in6p_flags inp_flags
109 #define in6p_moptions inp_moptions6
110 #define in6p_outputopts inp_outputopts6
111 #define in6p_ip6 inp_ipv6
112 #define in6p_flowinfo inp_flowinfo
113 #define in6p_sp inp_sp
114 #define in6p_next inp_next
115 #define in6p_prev inp_prev
116 /* macro names */
117 #define sotoin6pcb sotoinpcb
118 /* function names */
119 #define in6_pcbdetach in_pcbdetach
120 #define in6_rtchange in_rtchange
121
122 /*
123 * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
124 * others...
125 */
126 #define in6p_ip6_nxt inp_ipv6.ip6_nxt
127
128 extern struct domain inet6domain;
129 extern struct ip6protosw inet6sw[];
130 extern u_char ip6_protox[];
131
132 struct icmp6stat icmp6stat;
133
134 extern struct inpcbtable rawin6pcbtable;
135 extern int icmp6errppslim;
136 static int icmp6errpps_count = 0;
137 static struct timeval icmp6errppslim_last;
138 extern int icmp6_nodeinfo;
139
140 /*
141 * List of callbacks to notify when Path MTU changes are made.
142 */
143 struct icmp6_mtudisc_callback {
144 LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
145 void (*mc_func)(struct in6_addr *);
146 };
147
148 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
149 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
150
151 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
152 extern int pmtu_expire;
153
154 /* XXX do these values make any sense? */
155 static int icmp6_mtudisc_hiwat = 1280;
156 static int icmp6_mtudisc_lowat = 256;
157
158 /*
159 * keep track of # of redirect routes.
160 */
161 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
162
163 /* XXX experimental, turned off */
164 static int icmp6_redirect_hiwat = -1;
165 static int icmp6_redirect_lowat = -1;
166
167 static void icmp6_errcount(struct icmp6errstat *, int, int);
168 static int icmp6_rip6_input(struct mbuf **, int);
169 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
170 static const char *icmp6_redirect_diag(struct in6_addr *,
171 struct in6_addr *, struct in6_addr *);
172 static struct mbuf *ni6_input(struct mbuf *, int);
173 static struct mbuf *ni6_nametodns(const char *, int, int);
174 static int ni6_dnsmatch(const char *, int, const char *, int);
175 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
176 struct ifnet **, char *);
177 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
178 struct ifnet *, int);
179 static int icmp6_notify_error(struct mbuf *, int, int, int);
180 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
181 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
182 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
183
184 void
185 icmp6_init()
186 {
187 mld6_init();
188 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
189 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
190 }
191
192 static void
193 icmp6_errcount(stat, type, code)
194 struct icmp6errstat *stat;
195 int type, code;
196 {
197 switch (type) {
198 case ICMP6_DST_UNREACH:
199 switch (code) {
200 case ICMP6_DST_UNREACH_NOROUTE:
201 stat->icp6errs_dst_unreach_noroute++;
202 return;
203 case ICMP6_DST_UNREACH_ADMIN:
204 stat->icp6errs_dst_unreach_admin++;
205 return;
206 case ICMP6_DST_UNREACH_BEYONDSCOPE:
207 stat->icp6errs_dst_unreach_beyondscope++;
208 return;
209 case ICMP6_DST_UNREACH_ADDR:
210 stat->icp6errs_dst_unreach_addr++;
211 return;
212 case ICMP6_DST_UNREACH_NOPORT:
213 stat->icp6errs_dst_unreach_noport++;
214 return;
215 }
216 break;
217 case ICMP6_PACKET_TOO_BIG:
218 stat->icp6errs_packet_too_big++;
219 return;
220 case ICMP6_TIME_EXCEEDED:
221 switch (code) {
222 case ICMP6_TIME_EXCEED_TRANSIT:
223 stat->icp6errs_time_exceed_transit++;
224 return;
225 case ICMP6_TIME_EXCEED_REASSEMBLY:
226 stat->icp6errs_time_exceed_reassembly++;
227 return;
228 }
229 break;
230 case ICMP6_PARAM_PROB:
231 switch (code) {
232 case ICMP6_PARAMPROB_HEADER:
233 stat->icp6errs_paramprob_header++;
234 return;
235 case ICMP6_PARAMPROB_NEXTHEADER:
236 stat->icp6errs_paramprob_nextheader++;
237 return;
238 case ICMP6_PARAMPROB_OPTION:
239 stat->icp6errs_paramprob_option++;
240 return;
241 }
242 break;
243 case ND_REDIRECT:
244 stat->icp6errs_redirect++;
245 return;
246 }
247 stat->icp6errs_unknown++;
248 }
249
250 /*
251 * Register a Path MTU Discovery callback.
252 */
253 void
254 icmp6_mtudisc_callback_register(func)
255 void (*func)(struct in6_addr *);
256 {
257 struct icmp6_mtudisc_callback *mc;
258
259 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
260 mc = LIST_NEXT(mc, mc_list)) {
261 if (mc->mc_func == func)
262 return;
263 }
264
265 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
266 if (mc == NULL)
267 panic("icmp6_mtudisc_callback_register");
268
269 mc->mc_func = func;
270 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
271 }
272
273 /*
274 * Generate an error packet of type error in response to bad IP6 packet.
275 */
276 void
277 icmp6_error(m, type, code, param)
278 struct mbuf *m;
279 int type, code, param;
280 {
281 struct ip6_hdr *oip6, *nip6;
282 struct icmp6_hdr *icmp6;
283 u_int preplen;
284 int off;
285 int nxt;
286
287 icmp6stat.icp6s_error++;
288
289 /* count per-type-code statistics */
290 icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
291
292 if (m->m_len < sizeof(struct ip6_hdr)) {
293 m = m_pullup(m, sizeof(struct ip6_hdr));
294 if (m == NULL)
295 return;
296 }
297 oip6 = mtod(m, struct ip6_hdr *);
298
299 /*
300 * If the destination address of the erroneous packet is a multicast
301 * address, or the packet was sent using link-layer multicast,
302 * we should basically suppress sending an error (RFC 2463, Section
303 * 2.4).
304 * We have two exceptions (the item e.2 in that section):
305 * - the Pakcet Too Big message can be sent for path MTU discovery.
306 * - the Parameter Problem Message that can be allowed an icmp6 error
307 * in the option type field. This check has been done in
308 * ip6_unknown_opt(), so we can just check the type and code.
309 */
310 if ((m->m_flags & (M_BCAST|M_MCAST) ||
311 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
312 (type != ICMP6_PACKET_TOO_BIG &&
313 (type != ICMP6_PARAM_PROB ||
314 code != ICMP6_PARAMPROB_OPTION)))
315 goto freeit;
316
317 /*
318 * RFC 2463, 2.4 (e.5): source address check.
319 * XXX: the case of anycast source?
320 */
321 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
322 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
323 goto freeit;
324
325 /*
326 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
327 * don't do it.
328 */
329 nxt = -1;
330 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
331 if (off >= 0 && nxt == IPPROTO_ICMPV6) {
332 struct icmp6_hdr *icp;
333
334 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
335 sizeof(*icp));
336 if (icp == NULL) {
337 icmp6stat.icp6s_tooshort++;
338 return;
339 }
340 if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
341 icp->icmp6_type == ND_REDIRECT) {
342 /*
343 * ICMPv6 error
344 * Special case: for redirect (which is
345 * informational) we must not send icmp6 error.
346 */
347 icmp6stat.icp6s_canterror++;
348 goto freeit;
349 } else {
350 /* ICMPv6 informational - send the error */
351 }
352 }
353 else {
354 /* non-ICMPv6 - send the error */
355 }
356
357 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
358
359 /* Finally, do rate limitation check. */
360 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
361 icmp6stat.icp6s_toofreq++;
362 goto freeit;
363 }
364
365 /*
366 * OK, ICMP6 can be generated.
367 */
368
369 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
370 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
371
372 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
373 M_PREPEND(m, preplen, M_DONTWAIT);
374 if (m && m->m_len < preplen)
375 m = m_pullup(m, preplen);
376 if (m == NULL) {
377 nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
378 return;
379 }
380
381 nip6 = mtod(m, struct ip6_hdr *);
382 nip6->ip6_src = oip6->ip6_src;
383 nip6->ip6_dst = oip6->ip6_dst;
384
385 if (IN6_IS_SCOPE_EMBED(&oip6->ip6_src))
386 oip6->ip6_src.s6_addr16[1] = 0;
387 if (IN6_IS_SCOPE_EMBED(&oip6->ip6_dst))
388 oip6->ip6_dst.s6_addr16[1] = 0;
389
390 icmp6 = (struct icmp6_hdr *)(nip6 + 1);
391 icmp6->icmp6_type = type;
392 icmp6->icmp6_code = code;
393 icmp6->icmp6_pptr = htonl((u_int32_t)param);
394
395 /*
396 * icmp6_reflect() is designed to be in the input path.
397 * icmp6_error() can be called from both input and outut path,
398 * and if we are in output path rcvif could contain bogus value.
399 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
400 * information in ip header (nip6).
401 */
402 m->m_pkthdr.rcvif = NULL;
403
404 icmp6stat.icp6s_outhist[type]++;
405 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
406
407 return;
408
409 freeit:
410 /*
411 * If we can't tell wheter or not we can generate ICMP6, free it.
412 */
413 m_freem(m);
414 }
415
416 /*
417 * Process a received ICMP6 message.
418 */
419 int
420 icmp6_input(mp, offp, proto)
421 struct mbuf **mp;
422 int *offp, proto;
423 {
424 struct mbuf *m = *mp, *n;
425 struct ip6_hdr *ip6, *nip6;
426 struct icmp6_hdr *icmp6, *nicmp6;
427 int off = *offp;
428 int icmp6len = m->m_pkthdr.len - *offp;
429 int code, sum, noff;
430
431 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
432
433 /*
434 * Locate icmp6 structure in mbuf, and check
435 * that not corrupted and of at least minimum length
436 */
437
438 ip6 = mtod(m, struct ip6_hdr *);
439 if (icmp6len < sizeof(struct icmp6_hdr)) {
440 icmp6stat.icp6s_tooshort++;
441 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
442 goto freeit;
443 }
444
445 /*
446 * calculate the checksum
447 */
448 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
449 if (icmp6 == NULL) {
450 icmp6stat.icp6s_tooshort++;
451 return IPPROTO_DONE;
452 }
453 code = icmp6->icmp6_code;
454
455 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
456 nd6log((LOG_ERR,
457 "ICMP6 checksum error(%d|%x) %s\n",
458 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
459 icmp6stat.icp6s_checksum++;
460 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
461 goto freeit;
462 }
463
464 #if defined(NFAITH) && 0 < NFAITH
465 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
466 /*
467 * Deliver very specific ICMP6 type only.
468 * This is important to deliver TOOBIG. Otherwise PMTUD
469 * will not work.
470 */
471 switch (icmp6->icmp6_type) {
472 case ICMP6_DST_UNREACH:
473 case ICMP6_PACKET_TOO_BIG:
474 case ICMP6_TIME_EXCEEDED:
475 break;
476 default:
477 goto freeit;
478 }
479 }
480 #endif
481
482 #if NCARP > 0
483 if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
484 m->m_pkthdr.rcvif->if_flags & IFF_LINK0 &&
485 icmp6->icmp6_type == ICMP6_ECHO_REQUEST &&
486 carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
487 ip6->ip6_dst.s6_addr32))
488 goto freeit;
489 #endif
490 icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
491
492 switch (icmp6->icmp6_type) {
493 case ICMP6_DST_UNREACH:
494 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
495 switch (code) {
496 case ICMP6_DST_UNREACH_NOROUTE:
497 code = PRC_UNREACH_NET;
498 break;
499 case ICMP6_DST_UNREACH_ADMIN:
500 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
501 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
502 break;
503 case ICMP6_DST_UNREACH_ADDR:
504 code = PRC_HOSTDEAD;
505 break;
506 #ifdef COMPAT_RFC1885
507 case ICMP6_DST_UNREACH_NOTNEIGHBOR:
508 code = PRC_UNREACH_SRCFAIL;
509 break;
510 #else
511 case ICMP6_DST_UNREACH_BEYONDSCOPE:
512 /* I mean "source address was incorrect." */
513 code = PRC_PARAMPROB;
514 break;
515 #endif
516 case ICMP6_DST_UNREACH_NOPORT:
517 code = PRC_UNREACH_PORT;
518 break;
519 default:
520 goto badcode;
521 }
522 goto deliver;
523
524 case ICMP6_PACKET_TOO_BIG:
525 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
526
527 code = PRC_MSGSIZE;
528
529 /*
530 * Updating the path MTU will be done after examining
531 * intermediate extension headers.
532 */
533 goto deliver;
534
535 case ICMP6_TIME_EXCEEDED:
536 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
537 switch (code) {
538 case ICMP6_TIME_EXCEED_TRANSIT:
539 code = PRC_TIMXCEED_INTRANS;
540 break;
541 case ICMP6_TIME_EXCEED_REASSEMBLY:
542 code = PRC_TIMXCEED_REASS;
543 break;
544 default:
545 goto badcode;
546 }
547 goto deliver;
548
549 case ICMP6_PARAM_PROB:
550 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
551 switch (code) {
552 case ICMP6_PARAMPROB_NEXTHEADER:
553 code = PRC_UNREACH_PROTOCOL;
554 break;
555 case ICMP6_PARAMPROB_HEADER:
556 case ICMP6_PARAMPROB_OPTION:
557 code = PRC_PARAMPROB;
558 break;
559 default:
560 goto badcode;
561 }
562 goto deliver;
563
564 case ICMP6_ECHO_REQUEST:
565 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
566 if (code != 0)
567 goto badcode;
568 /*
569 * Copy mbuf to send to two data paths: userland socket(s),
570 * and to the querier (echo reply).
571 * m: a copy for socket, n: a copy for querier
572 */
573 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
574 /* Give up local */
575 n = m;
576 m = NULL;
577 goto deliverecho;
578 }
579 /*
580 * If the first mbuf is shared, or the first mbuf is too short,
581 * copy the first part of the data into a fresh mbuf.
582 * Otherwise, we will wrongly overwrite both copies.
583 */
584 if ((n->m_flags & M_EXT) != 0 ||
585 n->m_len < off + sizeof(struct icmp6_hdr)) {
586 struct mbuf *n0 = n;
587 const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
588
589 /*
590 * Prepare an internal mbuf. m_pullup() doesn't
591 * always copy the length we specified.
592 */
593 if (maxlen >= MCLBYTES) {
594 /* Give up remote */
595 m_freem(n0);
596 break;
597 }
598 MGETHDR(n, M_DONTWAIT, n0->m_type);
599 if (n && maxlen >= MHLEN) {
600 MCLGET(n, M_DONTWAIT);
601 if ((n->m_flags & M_EXT) == 0) {
602 m_free(n);
603 n = NULL;
604 }
605 }
606 if (n == NULL) {
607 /* Give up local */
608 m_freem(n0);
609 n = m;
610 m = NULL;
611 goto deliverecho;
612 }
613 M_MOVE_PKTHDR(n, n0);
614 /*
615 * Copy IPv6 and ICMPv6 only.
616 */
617 nip6 = mtod(n, struct ip6_hdr *);
618 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
619 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
620 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
621 noff = sizeof(struct ip6_hdr);
622 n->m_len = noff + sizeof(struct icmp6_hdr);
623 /*
624 * Adjust mbuf. ip6_plen will be adjusted in
625 * ip6_output().
626 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
627 */
628 n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
629 n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
630 m_adj(n0, off + sizeof(struct icmp6_hdr));
631 n->m_next = n0;
632 } else {
633 deliverecho:
634 nip6 = mtod(n, struct ip6_hdr *);
635 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
636 sizeof(*nicmp6));
637 noff = off;
638 }
639 nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
640 nicmp6->icmp6_code = 0;
641 if (n) {
642 icmp6stat.icp6s_reflect++;
643 icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
644 icmp6_reflect(n, noff);
645 }
646 if (!m)
647 goto freeit;
648 break;
649
650 case ICMP6_ECHO_REPLY:
651 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
652 if (code != 0)
653 goto badcode;
654 break;
655
656 case MLD_LISTENER_QUERY:
657 case MLD_LISTENER_REPORT:
658 if (icmp6len < sizeof(struct mld_hdr))
659 goto badlen;
660 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
661 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
662 else
663 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
664 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
665 /* give up local */
666 mld6_input(m, off);
667 m = NULL;
668 goto freeit;
669 }
670 mld6_input(n, off);
671 /* m stays. */
672 break;
673
674 case MLD_LISTENER_DONE:
675 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
676 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */
677 goto badlen;
678 break; /* nothing to be done in kernel */
679
680 case MLD_MTRACE_RESP:
681 case MLD_MTRACE:
682 /* XXX: these two are experimental. not officially defined. */
683 /* XXX: per-interface statistics? */
684 break; /* just pass it to applications */
685
686 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */
687 {
688 enum { WRU, FQDN } mode;
689
690 if (!icmp6_nodeinfo)
691 break;
692
693 if (icmp6len == sizeof(struct icmp6_hdr) + 4)
694 mode = WRU;
695 else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
696 mode = FQDN;
697 else
698 goto badlen;
699
700 if (mode == FQDN) {
701 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
702 if (n)
703 n = ni6_input(n, off);
704 /* XXX meaningless if n == NULL */
705 noff = sizeof(struct ip6_hdr);
706 } else {
707 u_char *p;
708 int maxlen, maxhlen;
709
710 if ((icmp6_nodeinfo & 1) == 0)
711 break;
712
713 if (code != 0)
714 goto badcode;
715 maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
716 if (maxlen >= MCLBYTES) {
717 /* Give up remote */
718 break;
719 }
720 MGETHDR(n, M_DONTWAIT, m->m_type);
721 if (n && maxlen > MHLEN) {
722 MCLGET(n, M_DONTWAIT);
723 if ((n->m_flags & M_EXT) == 0) {
724 m_free(n);
725 n = NULL;
726 }
727 }
728 if (n == NULL) {
729 /* Give up remote */
730 break;
731 }
732 n->m_pkthdr.rcvif = NULL;
733 n->m_len = 0;
734 maxhlen = M_TRAILINGSPACE(n) - maxlen;
735 if (maxhlen > hostnamelen)
736 maxhlen = hostnamelen;
737 /*
738 * Copy IPv6 and ICMPv6 only.
739 */
740 nip6 = mtod(n, struct ip6_hdr *);
741 bcopy(ip6, nip6, sizeof(struct ip6_hdr));
742 nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
743 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
744 p = (u_char *)(nicmp6 + 1);
745 bzero(p, 4);
746 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
747 noff = sizeof(struct ip6_hdr);
748 M_DUP_PKTHDR(n, m); /* just for rcvif */
749 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
750 sizeof(struct icmp6_hdr) + 4 + maxhlen;
751 nicmp6->icmp6_type = ICMP6_WRUREPLY;
752 nicmp6->icmp6_code = 0;
753 }
754 #undef hostnamelen
755 if (n) {
756 icmp6stat.icp6s_reflect++;
757 icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
758 icmp6_reflect(n, noff);
759 }
760 break;
761 }
762
763 case ICMP6_WRUREPLY:
764 if (code != 0)
765 goto badcode;
766 break;
767
768 case ND_ROUTER_SOLICIT:
769 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
770 if (code != 0)
771 goto badcode;
772 if (icmp6len < sizeof(struct nd_router_solicit))
773 goto badlen;
774 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
775 /* give up local */
776 nd6_rs_input(m, off, icmp6len);
777 m = NULL;
778 goto freeit;
779 }
780 nd6_rs_input(n, off, icmp6len);
781 /* m stays. */
782 break;
783
784 case ND_ROUTER_ADVERT:
785 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
786 if (code != 0)
787 goto badcode;
788 if (icmp6len < sizeof(struct nd_router_advert))
789 goto badlen;
790 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
791 /* give up local */
792 nd6_ra_input(m, off, icmp6len);
793 m = NULL;
794 goto freeit;
795 }
796 nd6_ra_input(n, off, icmp6len);
797 /* m stays. */
798 break;
799
800 case ND_NEIGHBOR_SOLICIT:
801 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
802 if (code != 0)
803 goto badcode;
804 if (icmp6len < sizeof(struct nd_neighbor_solicit))
805 goto badlen;
806 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
807 /* give up local */
808 nd6_ns_input(m, off, icmp6len);
809 m = NULL;
810 goto freeit;
811 }
812 nd6_ns_input(n, off, icmp6len);
813 /* m stays. */
814 break;
815
816 case ND_NEIGHBOR_ADVERT:
817 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
818 if (code != 0)
819 goto badcode;
820 if (icmp6len < sizeof(struct nd_neighbor_advert))
821 goto badlen;
822 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
823 /* give up local */
824 nd6_na_input(m, off, icmp6len);
825 m = NULL;
826 goto freeit;
827 }
828 nd6_na_input(n, off, icmp6len);
829 /* m stays. */
830 break;
831
832 case ND_REDIRECT:
833 icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
834 if (code != 0)
835 goto badcode;
836 if (icmp6len < sizeof(struct nd_redirect))
837 goto badlen;
838 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
839 /* give up local */
840 icmp6_redirect_input(m, off);
841 m = NULL;
842 goto freeit;
843 }
844 icmp6_redirect_input(n, off);
845 /* m stays. */
846 break;
847
848 case ICMP6_ROUTER_RENUMBERING:
849 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
850 code != ICMP6_ROUTER_RENUMBERING_RESULT)
851 goto badcode;
852 if (icmp6len < sizeof(struct icmp6_router_renum))
853 goto badlen;
854 break;
855
856 default:
857 nd6log((LOG_DEBUG,
858 "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
859 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
860 ip6_sprintf(&ip6->ip6_dst),
861 m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
862 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
863 /* ICMPv6 error: MUST deliver it by spec... */
864 code = PRC_NCMDS;
865 /* deliver */
866 } else {
867 /* ICMPv6 informational: MUST not deliver */
868 break;
869 }
870 deliver:
871 if (icmp6_notify_error(m, off, icmp6len, code)) {
872 /* In this case, m should've been freed. */
873 return (IPPROTO_DONE);
874 }
875 break;
876
877 badcode:
878 icmp6stat.icp6s_badcode++;
879 break;
880
881 badlen:
882 icmp6stat.icp6s_badlen++;
883 break;
884 }
885
886 /* deliver the packet to appropriate sockets */
887 icmp6_rip6_input(&m, *offp);
888
889 return IPPROTO_DONE;
890
891 freeit:
892 m_freem(m);
893 return IPPROTO_DONE;
894 }
895
896 static int
897 icmp6_notify_error(m, off, icmp6len, code)
898 struct mbuf *m;
899 int off, icmp6len, code;
900 {
901 struct icmp6_hdr *icmp6;
902 struct ip6_hdr *eip6;
903 u_int32_t notifymtu;
904 struct sockaddr_in6 icmp6src, icmp6dst;
905
906 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
907 icmp6stat.icp6s_tooshort++;
908 goto freeit;
909 }
910 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
911 sizeof(*icmp6) + sizeof(struct ip6_hdr));
912 if (icmp6 == NULL) {
913 icmp6stat.icp6s_tooshort++;
914 return (-1);
915 }
916 eip6 = (struct ip6_hdr *)(icmp6 + 1);
917
918 /* Detect the upper level protocol */
919 {
920 void (*ctlfunc)(int, struct sockaddr *, void *);
921 u_int8_t nxt = eip6->ip6_nxt;
922 int eoff = off + sizeof(struct icmp6_hdr) +
923 sizeof(struct ip6_hdr);
924 struct ip6ctlparam ip6cp;
925 struct in6_addr *finaldst = NULL;
926 int icmp6type = icmp6->icmp6_type;
927 struct ip6_frag *fh;
928 struct ip6_rthdr *rth;
929 struct ip6_rthdr0 *rth0;
930 int rthlen;
931
932 while (1) { /* XXX: should avoid infinite loop explicitly? */
933 struct ip6_ext *eh;
934
935 switch (nxt) {
936 case IPPROTO_HOPOPTS:
937 case IPPROTO_DSTOPTS:
938 case IPPROTO_AH:
939 IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
940 eoff, sizeof(*eh));
941 if (eh == NULL) {
942 icmp6stat.icp6s_tooshort++;
943 return (-1);
944 }
945
946 if (nxt == IPPROTO_AH)
947 eoff += (eh->ip6e_len + 2) << 2;
948 else
949 eoff += (eh->ip6e_len + 1) << 3;
950 nxt = eh->ip6e_nxt;
951 break;
952 case IPPROTO_ROUTING:
953 /*
954 * When the erroneous packet contains a
955 * routing header, we should examine the
956 * header to determine the final destination.
957 * Otherwise, we can't properly update
958 * information that depends on the final
959 * destination (e.g. path MTU).
960 */
961 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
962 eoff, sizeof(*rth));
963 if (rth == NULL) {
964 icmp6stat.icp6s_tooshort++;
965 return (-1);
966 }
967 rthlen = (rth->ip6r_len + 1) << 3;
968 /*
969 * XXX: currently there is no
970 * officially defined type other
971 * than type-0.
972 * Note that if the segment left field
973 * is 0, all intermediate hops must
974 * have been passed.
975 */
976 if (rth->ip6r_segleft &&
977 rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
978 int hops;
979
980 IP6_EXTHDR_GET(rth0,
981 struct ip6_rthdr0 *, m,
982 eoff, rthlen);
983 if (rth0 == NULL) {
984 icmp6stat.icp6s_tooshort++;
985 return (-1);
986 }
987 /* just ignore a bogus header */
988 if ((rth0->ip6r0_len % 2) == 0 &&
989 (hops = rth0->ip6r0_len/2))
990 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
991 }
992 eoff += rthlen;
993 nxt = rth->ip6r_nxt;
994 break;
995 case IPPROTO_FRAGMENT:
996 IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
997 eoff, sizeof(*fh));
998 if (fh == NULL) {
999 icmp6stat.icp6s_tooshort++;
1000 return (-1);
1001 }
1002 /*
1003 * Data after a fragment header is meaningless
1004 * unless it is the first fragment, but
1005 * we'll go to the notify label for path MTU
1006 * discovery.
1007 */
1008 if (fh->ip6f_offlg & IP6F_OFF_MASK)
1009 goto notify;
1010
1011 eoff += sizeof(struct ip6_frag);
1012 nxt = fh->ip6f_nxt;
1013 break;
1014 default:
1015 /*
1016 * This case includes ESP and the No Next
1017 * Header. In such cases going to the notify
1018 * label does not have any meaning
1019 * (i.e. ctlfunc will be NULL), but we go
1020 * anyway since we might have to update
1021 * path MTU information.
1022 */
1023 goto notify;
1024 }
1025 }
1026 notify:
1027 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1028 sizeof(*icmp6) + sizeof(struct ip6_hdr));
1029 if (icmp6 == NULL) {
1030 icmp6stat.icp6s_tooshort++;
1031 return (-1);
1032 }
1033
1034 eip6 = (struct ip6_hdr *)(icmp6 + 1);
1035 bzero(&icmp6dst, sizeof(icmp6dst));
1036 icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1037 icmp6dst.sin6_family = AF_INET6;
1038 if (finaldst == NULL)
1039 icmp6dst.sin6_addr = eip6->ip6_dst;
1040 else
1041 icmp6dst.sin6_addr = *finaldst;
1042 icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1043 &icmp6dst.sin6_addr);
1044 if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1045 NULL, NULL)) {
1046 /* should be impossbile */
1047 nd6log((LOG_DEBUG,
1048 "icmp6_notify_error: in6_embedscope failed\n"));
1049 goto freeit;
1050 }
1051
1052 /*
1053 * retrieve parameters from the inner IPv6 header, and convert
1054 * them into sockaddr structures.
1055 */
1056 bzero(&icmp6src, sizeof(icmp6src));
1057 icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1058 icmp6src.sin6_family = AF_INET6;
1059 icmp6src.sin6_addr = eip6->ip6_src;
1060 icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1061 &icmp6src.sin6_addr);
1062 if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1063 NULL, NULL)) {
1064 /* should be impossbile */
1065 nd6log((LOG_DEBUG,
1066 "icmp6_notify_error: in6_embedscope failed\n"));
1067 goto freeit;
1068 }
1069 icmp6src.sin6_flowinfo =
1070 (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 }
1086
1087 ctlfunc = (void (*)(int, struct sockaddr *, void *))
1088 (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1089 if (ctlfunc) {
1090 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1091 &ip6cp);
1092 }
1093 }
1094 return (0);
1095
1096 freeit:
1097 m_freem(m);
1098 return (-1);
1099 }
1100
1101 void
1102 icmp6_mtudisc_update(ip6cp, validated)
1103 struct ip6ctlparam *ip6cp;
1104 int validated;
1105 {
1106 unsigned long rtcount;
1107 struct icmp6_mtudisc_callback *mc;
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 rtentry *rt = NULL;
1113 struct sockaddr_in6 sin6;
1114
1115 /*
1116 * allow non-validated cases if memory is plenty, to make traffic
1117 * from non-connected pcb happy.
1118 */
1119 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1120 if (validated) {
1121 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1122 return;
1123 else if (0 <= icmp6_mtudisc_lowat &&
1124 rtcount > icmp6_mtudisc_lowat) {
1125 /*
1126 * XXX nuke a victim, install the new one.
1127 */
1128 }
1129 } else {
1130 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1131 return;
1132 }
1133
1134 bzero(&sin6, sizeof(sin6));
1135 sin6.sin6_family = PF_INET6;
1136 sin6.sin6_len = sizeof(struct sockaddr_in6);
1137 sin6.sin6_addr = *dst;
1138 /* XXX normally, this won't happen */
1139 if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1140 sin6.sin6_addr.s6_addr16[1] =
1141 htons(m->m_pkthdr.rcvif->if_index);
1142 }
1143 /* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1144 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1145
1146 if (rt && (rt->rt_flags & RTF_HOST) &&
1147 !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1148 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1149 if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1150 icmp6stat.icp6s_pmtuchg++;
1151 rt->rt_rmx.rmx_mtu = mtu;
1152 }
1153 }
1154 if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1155 RTFREE(rt);
1156 }
1157
1158 /*
1159 * Notify protocols that the MTU for this destination
1160 * has changed.
1161 */
1162 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1163 mc = LIST_NEXT(mc, mc_list))
1164 (*mc->mc_func)(&sin6.sin6_addr);
1165 }
1166
1167 /*
1168 * Process a Node Information Query packet, based on
1169 * draft-ietf-ipngwg-icmp-name-lookups-07.
1170 *
1171 * Spec incompatibilities:
1172 * - IPv6 Subject address handling
1173 * - IPv4 Subject address handling support missing
1174 * - Proxy reply (answer even if it's not for me)
1175 * - joins NI group address at in6_ifattach() time only, does not cope
1176 * with hostname changes by sethostname(3)
1177 */
1178 #ifndef offsetof /* XXX */
1179 #define offsetof(type, member) ((size_t)(&((type *)0)->member))
1180 #endif
1181 static struct mbuf *
1182 ni6_input(m, off)
1183 struct mbuf *m;
1184 int off;
1185 {
1186 struct icmp6_nodeinfo *ni6, *nni6;
1187 struct mbuf *n = NULL;
1188 u_int16_t qtype;
1189 int subjlen;
1190 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1191 struct ni_reply_fqdn *fqdn;
1192 int addrs; /* for NI_QTYPE_NODEADDR */
1193 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1194 struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1195 struct ip6_hdr *ip6;
1196 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */
1197 char *subj = NULL;
1198
1199 ip6 = mtod(m, struct ip6_hdr *);
1200 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1201 if (ni6 == NULL) {
1202 /* m is already reclaimed */
1203 return NULL;
1204 }
1205
1206 /*
1207 * Validate IPv6 destination address.
1208 *
1209 * The Responder must discard the Query without further processing
1210 * unless it is one of the Responder's unicast or anycast addresses, or
1211 * a link-local scope multicast address which the Responder has joined.
1212 * [icmp-name-lookups-07, Section 4.]
1213 */
1214 bzero(&sin6, sizeof(sin6));
1215 sin6.sin6_family = AF_INET6;
1216 sin6.sin6_len = sizeof(struct sockaddr_in6);
1217 bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1218 /* XXX scopeid */
1219 if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1220 ; /* unicast/anycast, fine */
1221 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1222 ; /* link-local multicast, fine */
1223 else
1224 goto bad;
1225
1226 /* validate query Subject field. */
1227 qtype = ntohs(ni6->ni_qtype);
1228 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1229 switch (qtype) {
1230 case NI_QTYPE_NOOP:
1231 case NI_QTYPE_SUPTYPES:
1232 /* 07 draft */
1233 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1234 break;
1235 /* FALLTHROUGH */
1236 case NI_QTYPE_FQDN:
1237 case NI_QTYPE_NODEADDR:
1238 switch (ni6->ni_code) {
1239 case ICMP6_NI_SUBJ_IPV6:
1240 #if ICMP6_NI_SUBJ_IPV6 != 0
1241 case 0:
1242 #endif
1243 /*
1244 * backward compatibility - try to accept 03 draft
1245 * format, where no Subject is present.
1246 */
1247 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1248 subjlen == 0) {
1249 oldfqdn++;
1250 break;
1251 }
1252 #if ICMP6_NI_SUBJ_IPV6 != 0
1253 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1254 goto bad;
1255 #endif
1256
1257 if (subjlen != sizeof(sin6.sin6_addr))
1258 goto bad;
1259
1260 /*
1261 * Validate Subject address.
1262 *
1263 * Not sure what exactly "address belongs to the node"
1264 * means in the spec, is it just unicast, or what?
1265 *
1266 * At this moment we consider Subject address as
1267 * "belong to the node" if the Subject address equals
1268 * to the IPv6 destination address; validation for
1269 * IPv6 destination address should have done enough
1270 * check for us.
1271 *
1272 * We do not do proxy at this moment.
1273 */
1274 /* m_pulldown instead of copy? */
1275 m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1276 subjlen, (caddr_t)&sin6.sin6_addr);
1277 /* XXX kame scope hack */
1278 if (IN6_IS_SCOPE_EMBED(&sin6.sin6_addr)) {
1279 if ((m->m_flags & M_PKTHDR) != 0 &&
1280 m->m_pkthdr.rcvif) {
1281 sin6.sin6_addr.s6_addr16[1] =
1282 htons(m->m_pkthdr.rcvif->if_index);
1283 }
1284 }
1285 subj = (char *)&sin6;
1286 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1287 break;
1288
1289 /*
1290 * XXX if we are to allow other cases, we should really
1291 * be careful about scope here.
1292 * basically, we should disallow queries toward IPv6
1293 * destination X with subject Y, if scope(X) > scope(Y).
1294 * if we allow scope(X) > scope(Y), it will result in
1295 * information leakage across scope boundary.
1296 */
1297 goto bad;
1298
1299 case ICMP6_NI_SUBJ_FQDN:
1300 /*
1301 * Validate Subject name with gethostname(3).
1302 *
1303 * The behavior may need some debate, since:
1304 * - we are not sure if the node has FQDN as
1305 * hostname (returned by gethostname(3)).
1306 * - the code does wildcard match for truncated names.
1307 * however, we are not sure if we want to perform
1308 * wildcard match, if gethostname(3) side has
1309 * truncated hostname.
1310 */
1311 n = ni6_nametodns(hostname, hostnamelen, 0);
1312 if (!n || n->m_next || n->m_len == 0)
1313 goto bad;
1314 IP6_EXTHDR_GET(subj, char *, m,
1315 off + sizeof(struct icmp6_nodeinfo), subjlen);
1316 if (subj == NULL)
1317 goto bad;
1318 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1319 n->m_len)) {
1320 goto bad;
1321 }
1322 m_freem(n);
1323 n = NULL;
1324 break;
1325
1326 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */
1327 default:
1328 goto bad;
1329 }
1330 break;
1331 }
1332
1333 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */
1334 switch (qtype) {
1335 case NI_QTYPE_FQDN:
1336 if ((icmp6_nodeinfo & 1) == 0)
1337 goto bad;
1338 break;
1339 case NI_QTYPE_NODEADDR:
1340 if ((icmp6_nodeinfo & 2) == 0)
1341 goto bad;
1342 break;
1343 }
1344
1345 /* guess reply length */
1346 switch (qtype) {
1347 case NI_QTYPE_NOOP:
1348 break; /* no reply data */
1349 case NI_QTYPE_SUPTYPES:
1350 replylen += sizeof(u_int32_t);
1351 break;
1352 case NI_QTYPE_FQDN:
1353 /* XXX will append an mbuf */
1354 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1355 break;
1356 case NI_QTYPE_NODEADDR:
1357 addrs = ni6_addrs(ni6, m, &ifp, subj);
1358 if ((replylen += addrs * (sizeof(struct in6_addr) +
1359 sizeof(u_int32_t))) > MCLBYTES)
1360 replylen = MCLBYTES; /* XXX: will truncate pkt later */
1361 break;
1362 default:
1363 /*
1364 * XXX: We must return a reply with the ICMP6 code
1365 * `unknown Qtype' in this case. However we regard the case
1366 * as an FQDN query for backward compatibility.
1367 * Older versions set a random value to this field,
1368 * so it rarely varies in the defined qtypes.
1369 * But the mechanism is not reliable...
1370 * maybe we should obsolete older versions.
1371 */
1372 qtype = NI_QTYPE_FQDN;
1373 /* XXX will append an mbuf */
1374 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1375 oldfqdn++;
1376 break;
1377 }
1378
1379 /* allocate an mbuf to reply. */
1380 MGETHDR(n, M_DONTWAIT, m->m_type);
1381 if (n == NULL) {
1382 m_freem(m);
1383 return (NULL);
1384 }
1385 M_DUP_PKTHDR(n, m); /* just for rcvif */
1386 if (replylen > MHLEN) {
1387 if (replylen > MCLBYTES) {
1388 /*
1389 * XXX: should we try to allocate more? But MCLBYTES
1390 * is probably much larger than IPV6_MMTU...
1391 */
1392 goto bad;
1393 }
1394 MCLGET(n, M_DONTWAIT);
1395 if ((n->m_flags & M_EXT) == 0) {
1396 goto bad;
1397 }
1398 }
1399 n->m_pkthdr.len = n->m_len = replylen;
1400
1401 /* copy mbuf header and IPv6 + Node Information base headers */
1402 bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1403 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1404 bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1405
1406 /* qtype dependent procedure */
1407 switch (qtype) {
1408 case NI_QTYPE_NOOP:
1409 nni6->ni_code = ICMP6_NI_SUCCESS;
1410 nni6->ni_flags = 0;
1411 break;
1412 case NI_QTYPE_SUPTYPES:
1413 {
1414 u_int32_t v;
1415 nni6->ni_code = ICMP6_NI_SUCCESS;
1416 nni6->ni_flags = htons(0x0000); /* raw bitmap */
1417 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1418 v = (u_int32_t)htonl(0x0000000f);
1419 bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1420 break;
1421 }
1422 case NI_QTYPE_FQDN:
1423 nni6->ni_code = ICMP6_NI_SUCCESS;
1424 fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1425 sizeof(struct ip6_hdr) +
1426 sizeof(struct icmp6_nodeinfo));
1427 nni6->ni_flags = 0; /* XXX: meaningless TTL */
1428 fqdn->ni_fqdn_ttl = 0; /* ditto. */
1429 /*
1430 * XXX do we really have FQDN in variable "hostname"?
1431 */
1432 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1433 if (n->m_next == NULL)
1434 goto bad;
1435 /* XXX we assume that n->m_next is not a chain */
1436 if (n->m_next->m_next != NULL)
1437 goto bad;
1438 n->m_pkthdr.len += n->m_next->m_len;
1439 break;
1440 case NI_QTYPE_NODEADDR:
1441 {
1442 int lenlim, copied;
1443
1444 nni6->ni_code = ICMP6_NI_SUCCESS;
1445 n->m_pkthdr.len = n->m_len =
1446 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1447 lenlim = M_TRAILINGSPACE(n);
1448 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1449 /* XXX: reset mbuf length */
1450 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1451 sizeof(struct icmp6_nodeinfo) + copied;
1452 break;
1453 }
1454 default:
1455 break; /* XXX impossible! */
1456 }
1457
1458 nni6->ni_type = ICMP6_NI_REPLY;
1459 m_freem(m);
1460 return (n);
1461
1462 bad:
1463 m_freem(m);
1464 if (n)
1465 m_freem(n);
1466 return (NULL);
1467 }
1468 #undef hostnamelen
1469
1470 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1471 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1472 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1473 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1474
1475 /*
1476 * make a mbuf with DNS-encoded string. no compression support.
1477 *
1478 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1479 * treated as truncated name (two \0 at the end). this is a wild guess.
1480 */
1481 static struct mbuf *
1482 ni6_nametodns(name, namelen, old)
1483 const char *name;
1484 int namelen;
1485 int old; /* return pascal string if non-zero */
1486 {
1487 struct mbuf *m;
1488 char *cp, *ep;
1489 const char *p, *q;
1490 int i, len, nterm;
1491
1492 if (old)
1493 len = namelen + 1;
1494 else
1495 len = MCLBYTES;
1496
1497 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1498 MGET(m, M_DONTWAIT, MT_DATA);
1499 if (m && len > MLEN) {
1500 MCLGET(m, M_DONTWAIT);
1501 if ((m->m_flags & M_EXT) == 0)
1502 goto fail;
1503 }
1504 if (!m)
1505 goto fail;
1506 m->m_next = NULL;
1507
1508 if (old) {
1509 m->m_len = len;
1510 *mtod(m, char *) = namelen;
1511 bcopy(name, mtod(m, char *) + 1, namelen);
1512 return m;
1513 } else {
1514 m->m_len = 0;
1515 cp = mtod(m, char *);
1516 ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1517
1518 /* if not certain about my name, return empty buffer */
1519 if (namelen == 0)
1520 return m;
1521
1522 /*
1523 * guess if it looks like shortened hostname, or FQDN.
1524 * shortened hostname needs two trailing "\0".
1525 */
1526 i = 0;
1527 for (p = name; p < name + namelen; p++) {
1528 if (*p && *p == '.')
1529 i++;
1530 }
1531 if (i < 2)
1532 nterm = 2;
1533 else
1534 nterm = 1;
1535
1536 p = name;
1537 while (cp < ep && p < name + namelen) {
1538 i = 0;
1539 for (q = p; q < name + namelen && *q && *q != '.'; q++)
1540 i++;
1541 /* result does not fit into mbuf */
1542 if (cp + i + 1 >= ep)
1543 goto fail;
1544 /*
1545 * DNS label length restriction, RFC1035 page 8.
1546 * "i == 0" case is included here to avoid returning
1547 * 0-length label on "foo..bar".
1548 */
1549 if (i <= 0 || i >= 64)
1550 goto fail;
1551 *cp++ = i;
1552 if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1553 goto fail;
1554 while (i > 0) {
1555 if (!isalnum(*p) && *p != '-')
1556 goto fail;
1557 if (isupper(*p)) {
1558 *cp++ = tolower(*p);
1559 p++;
1560 } else
1561 *cp++ = *p++;
1562 i--;
1563 }
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 TAILQ_FOREACH(ifp, &ifnet, if_list) {
1686 addrsofif = 0;
1687 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1688 if (ifa->ifa_addr->sa_family != AF_INET6)
1689 continue;
1690 ifa6 = (struct in6_ifaddr *)ifa;
1691
1692 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1693 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1694 &ifa6->ia_addr.sin6_addr))
1695 iffound = 1;
1696
1697 /*
1698 * IPv4-mapped addresses can only be returned by a
1699 * Node Information proxy, since they represent
1700 * addresses of IPv4-only nodes, which perforce do
1701 * not implement this protocol.
1702 * [icmp-name-lookups-07, Section 5.4]
1703 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1704 * this function at this moment.
1705 */
1706
1707 /* What do we have to do about ::1? */
1708 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1709 case IPV6_ADDR_SCOPE_LINKLOCAL:
1710 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1711 continue;
1712 break;
1713 case IPV6_ADDR_SCOPE_SITELOCAL:
1714 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1715 continue;
1716 break;
1717 case IPV6_ADDR_SCOPE_GLOBAL:
1718 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1719 continue;
1720 break;
1721 default:
1722 continue;
1723 }
1724
1725 /*
1726 * check if anycast is okay.
1727 * XXX: just experimental. not in the spec.
1728 */
1729 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1730 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1731 continue; /* we need only unicast addresses */
1732
1733 addrsofif++; /* count the address */
1734 }
1735 if (iffound) {
1736 *ifpp = ifp;
1737 return (addrsofif);
1738 }
1739
1740 addrs += addrsofif;
1741 }
1742
1743 return (addrs);
1744 }
1745
1746 static int
1747 ni6_store_addrs(ni6, nni6, ifp0, resid)
1748 struct icmp6_nodeinfo *ni6, *nni6;
1749 struct ifnet *ifp0;
1750 int resid;
1751 {
1752 struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1753 struct in6_ifaddr *ifa6;
1754 struct ifaddr *ifa;
1755 struct ifnet *ifp_dep = NULL;
1756 int copied = 0, allow_deprecated = 0;
1757 u_char *cp = (u_char *)(nni6 + 1);
1758 int niflags = ni6->ni_flags;
1759 u_int32_t ltime;
1760
1761 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1762 return (0); /* needless to copy */
1763
1764 again:
1765
1766 for (; ifp != TAILQ_END(&ifnet); ifp = TAILQ_NEXT(ifp, if_list)) {
1767 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1768 if (ifa->ifa_addr->sa_family != AF_INET6)
1769 continue;
1770 ifa6 = (struct in6_ifaddr *)ifa;
1771
1772 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1773 allow_deprecated == 0) {
1774 /*
1775 * prefererred address should be put before
1776 * deprecated addresses.
1777 */
1778
1779 /* record the interface for later search */
1780 if (ifp_dep == NULL)
1781 ifp_dep = ifp;
1782
1783 continue;
1784 }
1785 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1786 allow_deprecated != 0)
1787 continue; /* we now collect deprecated addrs */
1788
1789 /* What do we have to do about ::1? */
1790 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1791 case IPV6_ADDR_SCOPE_LINKLOCAL:
1792 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1793 continue;
1794 break;
1795 case IPV6_ADDR_SCOPE_SITELOCAL:
1796 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1797 continue;
1798 break;
1799 case IPV6_ADDR_SCOPE_GLOBAL:
1800 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1801 continue;
1802 break;
1803 default:
1804 continue;
1805 }
1806
1807 /*
1808 * check if anycast is okay.
1809 * XXX: just experimental. not in the spec.
1810 */
1811 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1812 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1813 continue;
1814
1815 /* now we can copy the address */
1816 if (resid < sizeof(struct in6_addr) +
1817 sizeof(u_int32_t)) {
1818 /*
1819 * We give up much more copy.
1820 * Set the truncate flag and return.
1821 */
1822 nni6->ni_flags |=
1823 NI_NODEADDR_FLAG_TRUNCATE;
1824 return (copied);
1825 }
1826
1827 /*
1828 * Set the TTL of the address.
1829 * The TTL value should be one of the following
1830 * according to the specification:
1831 *
1832 * 1. The remaining lifetime of a DHCP lease on the
1833 * address, or
1834 * 2. The remaining Valid Lifetime of a prefix from
1835 * which the address was derived through Stateless
1836 * Autoconfiguration.
1837 *
1838 * Note that we currently do not support stateful
1839 * address configuration by DHCPv6, so the former
1840 * case can't happen.
1841 *
1842 * TTL must be 2^31 > TTL >= 0.
1843 */
1844 if (ifa6->ia6_lifetime.ia6t_expire == 0)
1845 ltime = ND6_INFINITE_LIFETIME;
1846 else {
1847 if (ifa6->ia6_lifetime.ia6t_expire >
1848 time_second)
1849 ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1850 else
1851 ltime = 0;
1852 }
1853
1854 bcopy(<ime, cp, sizeof(u_int32_t));
1855 cp += sizeof(u_int32_t);
1856
1857 /* copy the address itself */
1858 bcopy(&ifa6->ia_addr.sin6_addr, cp,
1859 sizeof(struct in6_addr));
1860 /* XXX: KAME link-local hack; remove ifindex */
1861 if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1862 ((struct in6_addr *)cp)->s6_addr16[1] = 0;
1863 cp += sizeof(struct in6_addr);
1864
1865 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1866 copied += (sizeof(struct in6_addr) +
1867 sizeof(u_int32_t));
1868 }
1869 if (ifp0) /* we need search only on the specified IF */
1870 break;
1871 }
1872
1873 if (allow_deprecated == 0 && ifp_dep != NULL) {
1874 ifp = ifp_dep;
1875 allow_deprecated = 1;
1876
1877 goto again;
1878 }
1879
1880 return (copied);
1881 }
1882
1883 /*
1884 * XXX almost dup'ed code with rip6_input.
1885 */
1886 static int
1887 icmp6_rip6_input(mp, off)
1888 struct mbuf **mp;
1889 int off;
1890 {
1891 struct mbuf *m = *mp;
1892 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1893 struct in6pcb *in6p;
1894 struct in6pcb *last = NULL;
1895 struct sockaddr_in6 rip6src;
1896 struct icmp6_hdr *icmp6;
1897 struct mbuf *opts = NULL;
1898
1899 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1900 if (icmp6 == NULL) {
1901 /* m is already reclaimed */
1902 return IPPROTO_DONE;
1903 }
1904
1905 bzero(&rip6src, sizeof(rip6src));
1906 rip6src.sin6_len = sizeof(struct sockaddr_in6);
1907 rip6src.sin6_family = AF_INET6;
1908 /* KAME hack: recover scopeid */
1909 (void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1910
1911 CIRCLEQ_FOREACH(in6p, &rawin6pcbtable.inpt_queue, inp_queue) {
1912 if (!(in6p->in6p_flags & INP_IPV6))
1913 continue;
1914 if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1915 continue;
1916 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1917 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1918 continue;
1919 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1920 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1921 continue;
1922 if (in6p->in6p_icmp6filt
1923 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1924 in6p->in6p_icmp6filt))
1925 continue;
1926 if (last) {
1927 struct mbuf *n;
1928 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1929 if (last->in6p_flags & IN6P_CONTROLOPTS)
1930 ip6_savecontrol(last, n, &opts);
1931 /* strip intermediate headers */
1932 m_adj(n, off);
1933 if (sbappendaddr(&last->in6p_socket->so_rcv,
1934 (struct sockaddr *)&rip6src,
1935 n, opts) == 0) {
1936 /* should notify about lost packet */
1937 m_freem(n);
1938 if (opts)
1939 m_freem(opts);
1940 } else
1941 sorwakeup(last->in6p_socket);
1942 opts = NULL;
1943 }
1944 }
1945 last = in6p;
1946 }
1947 if (last) {
1948 if (last->in6p_flags & IN6P_CONTROLOPTS)
1949 ip6_savecontrol(last, m, &opts);
1950 /* strip intermediate headers */
1951 m_adj(m, off);
1952 if (sbappendaddr(&last->in6p_socket->so_rcv,
1953 (struct sockaddr *)&rip6src,
1954 m, opts) == 0) {
1955 m_freem(m);
1956 if (opts)
1957 m_freem(opts);
1958 } else
1959 sorwakeup(last->in6p_socket);
1960 } else {
1961 m_freem(m);
1962 ip6stat.ip6s_delivered--;
1963 }
1964 return IPPROTO_DONE;
1965 }
1966
1967 /*
1968 * Reflect the ip6 packet back to the source.
1969 * OFF points to the icmp6 header, counted from the top of the mbuf.
1970 *
1971 * Note: RFC 1885 required that an echo reply should be truncated if it
1972 * did not fit in with (return) path MTU, and KAME code supported the
1973 * behavior. However, as a clarification after the RFC, this limitation
1974 * was removed in a revised version of the spec, RFC 2463. We had kept the
1975 * old behavior, with a (non-default) ifdef block, while the new version of
1976 * the spec was an internet-draft status, and even after the new RFC was
1977 * published. But it would rather make sense to clean the obsoleted part
1978 * up, and to make the code simpler at this stage.
1979 */
1980 void
1981 icmp6_reflect(m, off)
1982 struct mbuf *m;
1983 size_t off;
1984 {
1985 struct ip6_hdr *ip6;
1986 struct icmp6_hdr *icmp6;
1987 struct in6_ifaddr *ia;
1988 struct in6_addr t, *src = 0;
1989 int plen;
1990 int type, code;
1991 struct ifnet *outif = NULL;
1992 struct sockaddr_in6 sa6_src, sa6_dst;
1993
1994 /* too short to reflect */
1995 if (off < sizeof(struct ip6_hdr)) {
1996 nd6log((LOG_DEBUG,
1997 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1998 (u_long)off, (u_long)sizeof(struct ip6_hdr),
1999 __FILE__, __LINE__));
2000 goto bad;
2001 }
2002
2003 /*
2004 * If there are extra headers between IPv6 and ICMPv6, strip
2005 * off that header first.
2006 */
2007 #ifdef DIAGNOSTIC
2008 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2009 panic("assumption failed in icmp6_reflect");
2010 #endif
2011 if (off > sizeof(struct ip6_hdr)) {
2012 size_t l;
2013 struct ip6_hdr nip6;
2014
2015 l = off - sizeof(struct ip6_hdr);
2016 m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2017 m_adj(m, l);
2018 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2019 if (m->m_len < l) {
2020 if ((m = m_pullup(m, l)) == NULL)
2021 return;
2022 }
2023 bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2024 } else /* off == sizeof(struct ip6_hdr) */ {
2025 size_t l;
2026 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2027 if (m->m_len < l) {
2028 if ((m = m_pullup(m, l)) == NULL)
2029 return;
2030 }
2031 }
2032 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2033 ip6 = mtod(m, struct ip6_hdr *);
2034 ip6->ip6_nxt = IPPROTO_ICMPV6;
2035 icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2036 type = icmp6->icmp6_type; /* keep type for statistics */
2037 code = icmp6->icmp6_code; /* ditto. */
2038
2039 t = ip6->ip6_dst;
2040 /*
2041 * ip6_input() drops a packet if its src is multicast.
2042 * So, the src is never multicast.
2043 */
2044 ip6->ip6_dst = ip6->ip6_src;
2045
2046 /*
2047 * XXX: make sure to embed scope zone information, using
2048 * already embedded IDs or the received interface (if any).
2049 * Note that rcvif may be NULL.
2050 * TODO: scoped routing case (XXX).
2051 */
2052 bzero(&sa6_src, sizeof(sa6_src));
2053 sa6_src.sin6_family = AF_INET6;
2054 sa6_src.sin6_len = sizeof(sa6_src);
2055 sa6_src.sin6_addr = ip6->ip6_dst;
2056 in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2057 in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2058 bzero(&sa6_dst, sizeof(sa6_dst));
2059 sa6_dst.sin6_family = AF_INET6;
2060 sa6_dst.sin6_len = sizeof(sa6_dst);
2061 sa6_dst.sin6_addr = t;
2062 in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2063 in6_embedscope(&t, &sa6_dst, NULL, NULL);
2064
2065 /*
2066 * If the incoming packet was addressed directly to us (i.e. unicast),
2067 * use dst as the src for the reply.
2068 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2069 * (for example) when we encounter an error while forwarding procedure
2070 * destined to a duplicated address of ours.
2071 */
2072 for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2073 if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2074 (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2075 src = &t;
2076 break;
2077 }
2078 if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2079 /*
2080 * This is the case if the dst is our link-local address
2081 * and the sender is also ourselves.
2082 */
2083 src = &t;
2084 }
2085
2086 if (src == 0) {
2087 int e;
2088 struct route_in6 ro;
2089
2090 /*
2091 * This case matches to multicasts, our anycast, or unicasts
2092 * that we do not own. Select a source address based on the
2093 * source address of the erroneous packet.
2094 */
2095 bzero(&ro, sizeof(ro));
2096 src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2097 if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2098 RTFREE(ro.ro_rt); /* XXX: we could use this */
2099 }
2100 if (src == NULL) {
2101 nd6log((LOG_DEBUG,
2102 "icmp6_reflect: source can't be determined: "
2103 "dst=%s, error=%d\n",
2104 ip6_sprintf(&sa6_src.sin6_addr), e));
2105 goto bad;
2106 }
2107 }
2108
2109 ip6->ip6_src = *src;
2110
2111 ip6->ip6_flow = 0;
2112 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2113 ip6->ip6_vfc |= IPV6_VERSION;
2114 ip6->ip6_nxt = IPPROTO_ICMPV6;
2115 if (m->m_pkthdr.rcvif) {
2116 /* XXX: This may not be the outgoing interface */
2117 ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2118 } else
2119 ip6->ip6_hlim = ip6_defhlim;
2120
2121 icmp6->icmp6_cksum = 0;
2122 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2123 sizeof(struct ip6_hdr), plen);
2124
2125 /*
2126 * XXX option handling
2127 */
2128
2129 m->m_flags &= ~(M_BCAST|M_MCAST);
2130
2131 /*
2132 * To avoid a "too big" situation at an intermediate router
2133 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2134 * Note that only echo and node information replies are affected,
2135 * since the length of ICMP6 errors is limited to the minimum MTU.
2136 */
2137 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif, NULL) != 0 &&
2138 outif)
2139 icmp6_ifstat_inc(outif, ifs6_out_error);
2140
2141 if (outif)
2142 icmp6_ifoutstat_inc(outif, type, code);
2143
2144 return;
2145
2146 bad:
2147 m_freem(m);
2148 return;
2149 }
2150
2151 void
2152 icmp6_fasttimo()
2153 {
2154
2155 mld6_fasttimeo();
2156 }
2157
2158 static const char *
2159 icmp6_redirect_diag(src6, dst6, tgt6)
2160 struct in6_addr *src6;
2161 struct in6_addr *dst6;
2162 struct in6_addr *tgt6;
2163 {
2164 static char buf[1024];
2165 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2166 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2167 return buf;
2168 }
2169
2170 void
2171 icmp6_redirect_input(m, off)
2172 struct mbuf *m;
2173 int off;
2174 {
2175 struct ifnet *ifp = m->m_pkthdr.rcvif;
2176 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2177 struct nd_redirect *nd_rd;
2178 int icmp6len = ntohs(ip6->ip6_plen);
2179 char *lladdr = NULL;
2180 int lladdrlen = 0;
2181 u_char *redirhdr = NULL;
2182 int redirhdrlen = 0;
2183 struct rtentry *rt = NULL;
2184 int is_router;
2185 int is_onlink;
2186 struct in6_addr src6 = ip6->ip6_src;
2187 struct in6_addr redtgt6;
2188 struct in6_addr reddst6;
2189 union nd_opts ndopts;
2190
2191 if (!ifp)
2192 return;
2193
2194 /* XXX if we are router, we don't update route by icmp6 redirect */
2195 if (ip6_forwarding)
2196 goto freeit;
2197 if (!icmp6_rediraccept)
2198 goto freeit;
2199
2200 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2201 if (nd_rd == NULL) {
2202 icmp6stat.icp6s_tooshort++;
2203 return;
2204 }
2205 redtgt6 = nd_rd->nd_rd_target;
2206 reddst6 = nd_rd->nd_rd_dst;
2207
2208 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2209 redtgt6.s6_addr16[1] = htons(ifp->if_index);
2210 if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2211 reddst6.s6_addr16[1] = htons(ifp->if_index);
2212
2213 /* validation */
2214 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2215 nd6log((LOG_ERR,
2216 "ICMP6 redirect sent from %s rejected; "
2217 "must be from linklocal\n", ip6_sprintf(&src6)));
2218 goto bad;
2219 }
2220 if (ip6->ip6_hlim != 255) {
2221 nd6log((LOG_ERR,
2222 "ICMP6 redirect sent from %s rejected; "
2223 "hlim=%d (must be 255)\n",
2224 ip6_sprintf(&src6), ip6->ip6_hlim));
2225 goto bad;
2226 }
2227 {
2228 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2229 struct sockaddr_in6 sin6;
2230 struct in6_addr *gw6;
2231
2232 bzero(&sin6, sizeof(sin6));
2233 sin6.sin6_family = AF_INET6;
2234 sin6.sin6_len = sizeof(struct sockaddr_in6);
2235 bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2236 rt = rtalloc1((struct sockaddr *)&sin6, 0, 0);
2237 if (rt) {
2238 if (rt->rt_gateway == NULL ||
2239 rt->rt_gateway->sa_family != AF_INET6) {
2240 nd6log((LOG_ERR,
2241 "ICMP6 redirect rejected; no route "
2242 "with inet6 gateway found for redirect dst: %s\n",
2243 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2244 RTFREE(rt);
2245 goto bad;
2246 }
2247
2248 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2249 if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2250 nd6log((LOG_ERR,
2251 "ICMP6 redirect rejected; "
2252 "not equal to gw-for-src=%s (must be same): "
2253 "%s\n",
2254 ip6_sprintf(gw6),
2255 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2256 RTFREE(rt);
2257 goto bad;
2258 }
2259 } else {
2260 nd6log((LOG_ERR,
2261 "ICMP6 redirect rejected; "
2262 "no route found for redirect dst: %s\n",
2263 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2264 goto bad;
2265 }
2266 RTFREE(rt);
2267 rt = NULL;
2268 }
2269 if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2270 nd6log((LOG_ERR,
2271 "ICMP6 redirect rejected; "
2272 "redirect dst must be unicast: %s\n",
2273 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2274 goto bad;
2275 }
2276
2277 is_router = is_onlink = 0;
2278 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2279 is_router = 1; /* router case */
2280 if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2281 is_onlink = 1; /* on-link destination case */
2282 if (!is_router && !is_onlink) {
2283 nd6log((LOG_ERR,
2284 "ICMP6 redirect rejected; "
2285 "neither router case nor onlink case: %s\n",
2286 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2287 goto bad;
2288 }
2289 /* validation passed */
2290
2291 icmp6len -= sizeof(*nd_rd);
2292 nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2293 if (nd6_options(&ndopts) < 0) {
2294 nd6log((LOG_INFO, "icmp6_redirect_input: "
2295 "invalid ND option, rejected: %s\n",
2296 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2297 /* nd6_options have incremented stats */
2298 goto freeit;
2299 }
2300
2301 if (ndopts.nd_opts_tgt_lladdr) {
2302 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2303 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2304 }
2305
2306 if (ndopts.nd_opts_rh) {
2307 redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2308 redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2309 }
2310
2311 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2312 nd6log((LOG_INFO,
2313 "icmp6_redirect_input: lladdrlen mismatch for %s "
2314 "(if %d, icmp6 packet %d): %s\n",
2315 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2316 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2317 goto bad;
2318 }
2319
2320 /* RFC 2461 8.3 */
2321 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2322 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2323
2324 if (!is_onlink) { /* better router case. perform rtredirect. */
2325 /* perform rtredirect */
2326 struct sockaddr_in6 sdst;
2327 struct sockaddr_in6 sgw;
2328 struct sockaddr_in6 ssrc;
2329 unsigned long rtcount;
2330 struct rtentry *newrt = NULL;
2331
2332 /*
2333 * do not install redirect route, if the number of entries
2334 * is too much (> hiwat). note that, the node (= host) will
2335 * work just fine even if we do not install redirect route
2336 * (there will be additional hops, though).
2337 */
2338 rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2339 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2340 return;
2341 else if (0 <= icmp6_redirect_lowat &&
2342 rtcount > icmp6_redirect_lowat) {
2343 /*
2344 * XXX nuke a victim, install the new one.
2345 */
2346 }
2347
2348 bzero(&sdst, sizeof(sdst));
2349 bzero(&sgw, sizeof(sgw));
2350 bzero(&ssrc, sizeof(ssrc));
2351 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2352 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2353 sizeof(struct sockaddr_in6);
2354 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2355 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2356 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2357 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2358 (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2359 (struct sockaddr *)&ssrc,
2360 &newrt);
2361
2362 if (newrt) {
2363 (void)rt_timer_add(newrt, icmp6_redirect_timeout,
2364 icmp6_redirect_timeout_q);
2365 rtfree(newrt);
2366 }
2367 }
2368 /* finally update cached route in each socket via pfctlinput */
2369 {
2370 struct sockaddr_in6 sdst;
2371
2372 bzero(&sdst, sizeof(sdst));
2373 sdst.sin6_family = AF_INET6;
2374 sdst.sin6_len = sizeof(struct sockaddr_in6);
2375 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2376 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2377 }
2378
2379 freeit:
2380 m_freem(m);
2381 return;
2382
2383 bad:
2384 icmp6stat.icp6s_badredirect++;
2385 m_freem(m);
2386 }
2387
2388 void
2389 icmp6_redirect_output(m0, rt)
2390 struct mbuf *m0;
2391 struct rtentry *rt;
2392 {
2393 struct ifnet *ifp; /* my outgoing interface */
2394 struct in6_addr *ifp_ll6;
2395 struct in6_addr *nexthop;
2396 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */
2397 struct mbuf *m = NULL; /* newly allocated one */
2398 struct ip6_hdr *ip6; /* m as struct ip6_hdr */
2399 struct nd_redirect *nd_rd;
2400 size_t maxlen;
2401 u_char *p;
2402 struct sockaddr_in6 src_sa;
2403
2404 icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2405
2406 /* if we are not router, we don't send icmp6 redirect */
2407 if (!ip6_forwarding)
2408 goto fail;
2409
2410 /* sanity check */
2411 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2412 goto fail;
2413
2414 /*
2415 * Address check:
2416 * the source address must identify a neighbor, and
2417 * the destination address must not be a multicast address
2418 * [RFC 2461, sec 8.2]
2419 */
2420 sip6 = mtod(m0, struct ip6_hdr *);
2421 bzero(&src_sa, sizeof(src_sa));
2422 src_sa.sin6_family = AF_INET6;
2423 src_sa.sin6_len = sizeof(src_sa);
2424 src_sa.sin6_addr = sip6->ip6_src;
2425 /* we don't currently use sin6_scope_id, but eventually use it */
2426 src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2427 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2428 goto fail;
2429 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2430 goto fail; /* what should we do here? */
2431
2432 /* rate limit */
2433 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2434 goto fail;
2435
2436 /*
2437 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2438 * we almost always ask for an mbuf cluster for simplicity.
2439 * (MHLEN < IPV6_MMTU is almost always true)
2440 */
2441 #if IPV6_MMTU >= MCLBYTES
2442 # error assumption failed about IPV6_MMTU and MCLBYTES
2443 #endif
2444 MGETHDR(m, M_DONTWAIT, MT_HEADER);
2445 if (m && IPV6_MMTU >= MHLEN)
2446 MCLGET(m, M_DONTWAIT);
2447 if (!m)
2448 goto fail;
2449 m->m_pkthdr.rcvif = NULL;
2450 m->m_len = 0;
2451 maxlen = M_TRAILINGSPACE(m);
2452 maxlen = min(IPV6_MMTU, maxlen);
2453 /* just for safety */
2454 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2455 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2456 goto fail;
2457 }
2458
2459 {
2460 /* get ip6 linklocal address for ifp(my outgoing interface). */
2461 struct in6_ifaddr *ia;
2462 if ((ia = in6ifa_ifpforlinklocal(ifp,
2463 IN6_IFF_NOTREADY|
2464 IN6_IFF_ANYCAST)) == NULL)
2465 goto fail;
2466 ifp_ll6 = &ia->ia_addr.sin6_addr;
2467 }
2468
2469 /* get ip6 linklocal address for the router. */
2470 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2471 struct sockaddr_in6 *sin6;
2472 sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2473 nexthop = &sin6->sin6_addr;
2474 if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2475 nexthop = NULL;
2476 } else
2477 nexthop = NULL;
2478
2479 /* ip6 */
2480 ip6 = mtod(m, struct ip6_hdr *);
2481 ip6->ip6_flow = 0;
2482 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2483 ip6->ip6_vfc |= IPV6_VERSION;
2484 /* ip6->ip6_plen will be set later */
2485 ip6->ip6_nxt = IPPROTO_ICMPV6;
2486 ip6->ip6_hlim = 255;
2487 /* ip6->ip6_src must be linklocal addr for my outgoing if. */
2488 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2489 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2490
2491 /* ND Redirect */
2492 nd_rd = (struct nd_redirect *)(ip6 + 1);
2493 nd_rd->nd_rd_type = ND_REDIRECT;
2494 nd_rd->nd_rd_code = 0;
2495 nd_rd->nd_rd_reserved = 0;
2496 if (rt->rt_flags & RTF_GATEWAY) {
2497 /*
2498 * nd_rd->nd_rd_target must be a link-local address in
2499 * better router cases.
2500 */
2501 if (!nexthop)
2502 goto fail;
2503 bcopy(nexthop, &nd_rd->nd_rd_target,
2504 sizeof(nd_rd->nd_rd_target));
2505 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2506 sizeof(nd_rd->nd_rd_dst));
2507 } else {
2508 /* make sure redtgt == reddst */
2509 nexthop = &sip6->ip6_dst;
2510 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2511 sizeof(nd_rd->nd_rd_target));
2512 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2513 sizeof(nd_rd->nd_rd_dst));
2514 }
2515
2516 p = (u_char *)(nd_rd + 1);
2517
2518 {
2519 /* target lladdr option */
2520 struct rtentry *rt_nexthop = NULL;
2521 int len;
2522 struct sockaddr_dl *sdl;
2523 struct nd_opt_hdr *nd_opt;
2524 char *lladdr;
2525
2526 rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2527 if (!rt_nexthop)
2528 goto nolladdropt;
2529 len = sizeof(*nd_opt) + ifp->if_addrlen;
2530 len = (len + 7) & ~7; /* round by 8 */
2531 /* safety check */
2532 if (len + (p - (u_char *)ip6) > maxlen)
2533 goto nolladdropt;
2534 if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2535 (rt_nexthop->rt_flags & RTF_LLINFO) &&
2536 (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2537 (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2538 sdl->sdl_alen) {
2539 nd_opt = (struct nd_opt_hdr *)p;
2540 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2541 nd_opt->nd_opt_len = len >> 3;
2542 lladdr = (char *)(nd_opt + 1);
2543 bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2544 p += len;
2545 }
2546 }
2547 nolladdropt:;
2548
2549 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2550
2551 /* just to be safe */
2552 if (p - (u_char *)ip6 > maxlen)
2553 goto noredhdropt;
2554
2555 {
2556 /* redirected header option */
2557 int len;
2558 struct nd_opt_rd_hdr *nd_opt_rh;
2559
2560 /*
2561 * compute the maximum size for icmp6 redirect header option.
2562 * XXX room for auth header?
2563 */
2564 len = maxlen - (p - (u_char *)ip6);
2565 len &= ~7;
2566
2567 /*
2568 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2569 * about padding/truncate rule for the original IP packet.
2570 * From the discussion on IPv6imp in Feb 1999,
2571 * the consensus was:
2572 * - "attach as much as possible" is the goal
2573 * - pad if not aligned (original size can be guessed by
2574 * original ip6 header)
2575 * Following code adds the padding if it is simple enough,
2576 * and truncates if not.
2577 */
2578 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2579 /* not enough room, truncate */
2580 m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2581 m0->m_pkthdr.len);
2582 } else {
2583 /*
2584 * enough room, truncate if not aligned.
2585 * we don't pad here for simplicity.
2586 */
2587 size_t extra;
2588
2589 extra = m0->m_pkthdr.len % 8;
2590 if (extra) {
2591 /* truncate */
2592 m_adj(m0, -extra);
2593 }
2594 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2595 }
2596
2597 nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2598 bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2599 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2600 nd_opt_rh->nd_opt_rh_len = len >> 3;
2601 p += sizeof(*nd_opt_rh);
2602 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2603
2604 /* connect m0 to m */
2605 m->m_pkthdr.len += m0->m_pkthdr.len;
2606 m_cat(m, m0);
2607 m0 = NULL;
2608 }
2609 noredhdropt:
2610 if (m0) {
2611 m_freem(m0);
2612 m0 = NULL;
2613 }
2614
2615 sip6 = mtod(m, struct ip6_hdr *);
2616 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2617 sip6->ip6_src.s6_addr16[1] = 0;
2618 if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2619 sip6->ip6_dst.s6_addr16[1] = 0;
2620 #if 0
2621 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2622 ip6->ip6_src.s6_addr16[1] = 0;
2623 if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2624 ip6->ip6_dst.s6_addr16[1] = 0;
2625 #endif
2626 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2627 nd_rd->nd_rd_target.s6_addr16[1] = 0;
2628 if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2629 nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2630
2631 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2632
2633 nd_rd->nd_rd_cksum = 0;
2634 nd_rd->nd_rd_cksum
2635 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2636
2637 /* send the packet to outside... */
2638 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2639 icmp6_ifstat_inc(ifp, ifs6_out_error);
2640
2641 icmp6_ifstat_inc(ifp, ifs6_out_msg);
2642 icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2643 icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2644
2645 return;
2646
2647 fail:
2648 if (m)
2649 m_freem(m);
2650 if (m0)
2651 m_freem(m0);
2652 }
2653
2654 /* NRL PCB */
2655 #define sotoin6pcb sotoinpcb
2656 #define in6pcb inpcb
2657 #define in6p_icmp6filt inp_icmp6filt
2658
2659 /*
2660 * ICMPv6 socket option processing.
2661 */
2662 int
2663 icmp6_ctloutput(op, so, level, optname, mp)
2664 int op;
2665 struct socket *so;
2666 int level, optname;
2667 struct mbuf **mp;
2668 {
2669 int error = 0;
2670 int optlen;
2671 struct in6pcb *in6p = sotoin6pcb(so);
2672 struct mbuf *m = *mp;
2673
2674 optlen = m ? m->m_len : 0;
2675
2676 if (level != IPPROTO_ICMPV6) {
2677 if (op == PRCO_SETOPT && m)
2678 (void)m_free(m);
2679 return EINVAL;
2680 }
2681
2682 switch (op) {
2683 case PRCO_SETOPT:
2684 switch (optname) {
2685 case ICMP6_FILTER:
2686 {
2687 struct icmp6_filter *p;
2688
2689 if (optlen != sizeof(*p)) {
2690 error = EMSGSIZE;
2691 break;
2692 }
2693 p = mtod(m, struct icmp6_filter *);
2694 if (!p || !in6p->in6p_icmp6filt) {
2695 error = EINVAL;
2696 break;
2697 }
2698 bcopy(p, in6p->in6p_icmp6filt,
2699 sizeof(struct icmp6_filter));
2700 error = 0;
2701 break;
2702 }
2703
2704 default:
2705 error = ENOPROTOOPT;
2706 break;
2707 }
2708 if (m)
2709 (void)m_freem(m);
2710 break;
2711
2712 case PRCO_GETOPT:
2713 switch (optname) {
2714 case ICMP6_FILTER:
2715 {
2716 struct icmp6_filter *p;
2717
2718 if (!in6p->in6p_icmp6filt) {
2719 error = EINVAL;
2720 break;
2721 }
2722 *mp = m = m_get(M_WAIT, MT_SOOPTS);
2723 m->m_len = sizeof(struct icmp6_filter);
2724 p = mtod(m, struct icmp6_filter *);
2725 bcopy(in6p->in6p_icmp6filt, p,
2726 sizeof(struct icmp6_filter));
2727 error = 0;
2728 break;
2729 }
2730
2731 default:
2732 error = ENOPROTOOPT;
2733 break;
2734 }
2735 break;
2736 }
2737
2738 return (error);
2739 }
2740
2741 /* NRL PCB */
2742 #undef sotoin6pcb
2743 #undef in6pcb
2744 #undef in6p_icmp6filt
2745
2746 /*
2747 * Perform rate limit check.
2748 * Returns 0 if it is okay to send the icmp6 packet.
2749 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2750 * limitation.
2751 *
2752 * XXX per-destination/type check necessary?
2753 */
2754 static int
2755 icmp6_ratelimit(dst, type, code)
2756 const struct in6_addr *dst; /* not used at this moment */
2757 const int type; /* not used at this moment */
2758 const int code; /* not used at this moment */
2759 {
2760 int ret;
2761
2762 ret = 0; /* okay to send */
2763
2764 /* PPS limit */
2765 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2766 icmp6errppslim)) {
2767 /* The packet is subject to rate limit */
2768 ret++;
2769 }
2770
2771 return ret;
2772 }
2773
2774 static struct rtentry *
2775 icmp6_mtudisc_clone(dst)
2776 struct sockaddr *dst;
2777 {
2778 struct rtentry *rt;
2779 int error;
2780
2781 rt = rtalloc1(dst, 1, 0);
2782 if (rt == 0)
2783 return NULL;
2784
2785 /* If we didn't get a host route, allocate one */
2786 if ((rt->rt_flags & RTF_HOST) == 0) {
2787 struct rtentry *nrt;
2788
2789 error = rtrequest((int) RTM_ADD, dst,
2790 (struct sockaddr *) rt->rt_gateway,
2791 (struct sockaddr *) 0,
2792 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt, 0);
2793 if (error) {
2794 rtfree(rt);
2795 return NULL;
2796 }
2797 nrt->rt_rmx = rt->rt_rmx;
2798 rtfree(rt);
2799 rt = nrt;
2800 }
2801 error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2802 icmp6_mtudisc_timeout_q);
2803 if (error) {
2804 rtfree(rt);
2805 return NULL;
2806 }
2807
2808 return rt; /* caller need to call rtfree() */
2809 }
2810
2811 static void
2812 icmp6_mtudisc_timeout(rt, r)
2813 struct rtentry *rt;
2814 struct rttimer *r;
2815 {
2816 if (rt == NULL)
2817 panic("icmp6_mtudisc_timeout: bad route to timeout");
2818 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2819 (RTF_DYNAMIC | RTF_HOST)) {
2820 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2821 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0, 0);
2822 } else {
2823 if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2824 rt->rt_rmx.rmx_mtu = 0;
2825 }
2826 }
2827
2828 static void
2829 icmp6_redirect_timeout(rt, r)
2830 struct rtentry *rt;
2831 struct rttimer *r;
2832 {
2833 if (rt == NULL)
2834 panic("icmp6_redirect_timeout: bad route to timeout");
2835 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2836 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2837 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2838 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0, 0);
2839 }
2840 }
2841
2842 #include <uvm/uvm_extern.h>
2843 #include <sys/sysctl.h>
2844
2845 int *icmpv6ctl_vars[ICMPV6CTL_MAXID] = ICMPV6CTL_VARS;
2846
2847 int
2848 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
2849 int *name;
2850 u_int namelen;
2851 void *oldp;
2852 size_t *oldlenp;
2853 void *newp;
2854 size_t newlen;
2855 {
2856
2857 /* All sysctl names at this level are terminal. */
2858 if (namelen != 1)
2859 return ENOTDIR;
2860
2861 switch (name[0]) {
2862
2863 case ICMPV6CTL_STATS:
2864 return sysctl_rdstruct(oldp, oldlenp, newp,
2865 &icmp6stat, sizeof(icmp6stat));
2866 case ICMPV6CTL_ND6_DRLIST:
2867 case ICMPV6CTL_ND6_PRLIST:
2868 return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen);
2869 default:
2870 if (name[0] < ICMPV6CTL_MAXID)
2871 return (sysctl_int_arr(icmpv6ctl_vars, name, namelen,
2872 oldp, oldlenp, newp, newlen));
2873 return ENOPROTOOPT;
2874 }
2875 /* NOTREACHED */
2876 }