nd6_nbr.c revision 1.153 1 /* $NetBSD: nd6_nbr.c,v 1.153 2018/03/19 03:14:08 ozaki-r Exp $ */
2 /* $KAME: nd6_nbr.c,v 1.61 2001/02/10 16:06:14 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 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_nbr.c,v 1.153 2018/03/19 03:14:08 ozaki-r Exp $");
35
36 #ifdef _KERNEL_OPT
37 #include "opt_inet.h"
38 #include "opt_net_mpsafe.h"
39 #endif
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kmem.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sockio.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/errno.h>
51 #include <sys/ioctl.h>
52 #include <sys/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/callout.h>
55 #include <sys/cprng.h>
56
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_dl.h>
60 #include <net/route.h>
61
62 #include <netinet/in.h>
63 #include <netinet/in_var.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet6/in6_ifattach.h>
66 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #include <netinet6/scope6_var.h>
69 #include <netinet6/nd6.h>
70 #include <netinet/icmp6.h>
71 #include <netinet6/icmp6_private.h>
72
73 #include "carp.h"
74 #if NCARP > 0
75 #include <netinet/ip_carp.h>
76 #endif
77
78 #include <net/net_osdep.h>
79
80 struct dadq;
81 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
82 static void nd6_dad_starttimer(struct dadq *, int);
83 static void nd6_dad_destroytimer(struct dadq *);
84 static void nd6_dad_timer(struct dadq *);
85 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
87 static void nd6_dad_na_input(struct ifaddr *);
88 static void nd6_dad_duplicated(struct dadq *);
89
90 static int dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */
91
92 /*
93 * Input a Neighbor Solicitation Message.
94 *
95 * Based on RFC 2461
96 * Based on RFC 2462 (duplicate address detection)
97 */
98 void
99 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
100 {
101 struct ifnet *ifp;
102 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
103 struct nd_neighbor_solicit *nd_ns;
104 struct in6_addr saddr6 = ip6->ip6_src;
105 struct in6_addr daddr6 = ip6->ip6_dst;
106 struct in6_addr taddr6;
107 struct in6_addr myaddr6;
108 char *lladdr = NULL;
109 struct ifaddr *ifa = NULL;
110 int lladdrlen = 0;
111 int anycast = 0, proxy = 0, tentative = 0;
112 int router = ip6_forwarding;
113 int tlladdr;
114 union nd_opts ndopts;
115 const struct sockaddr_dl *proxydl = NULL;
116 struct psref psref;
117 struct psref psref_ia;
118 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
119
120 ifp = m_get_rcvif_psref(m, &psref);
121 if (ifp == NULL)
122 goto freeit;
123
124 IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
125 if (nd_ns == NULL) {
126 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
127 m_put_rcvif_psref(ifp, &psref);
128 return;
129 }
130 ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
131 taddr6 = nd_ns->nd_ns_target;
132 if (in6_setscope(&taddr6, ifp, NULL) != 0)
133 goto bad;
134
135 if (ip6->ip6_hlim != 255) {
136 nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
137 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
138 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
139 goto bad;
140 }
141
142 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
143 /* dst has to be a solicited node multicast address. */
144 /* don't check ifindex portion */
145 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
146 daddr6.s6_addr32[1] == 0 &&
147 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
148 daddr6.s6_addr8[12] == 0xff) {
149 ; /* good */
150 } else {
151 nd6log(LOG_INFO, "bad DAD packet (wrong ip6 dst)\n");
152 goto bad;
153 }
154 } else {
155 struct sockaddr_in6 ssin6;
156
157 /*
158 * Make sure the source address is from a neighbor's address.
159 */
160 sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
161 if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
162 nd6log(LOG_INFO,
163 "NS packet from non-neighbor %s on %s\n",
164 IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
165 goto bad;
166 }
167 }
168
169
170 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
171 nd6log(LOG_INFO, "bad NS target (multicast)\n");
172 goto bad;
173 }
174
175 icmp6len -= sizeof(*nd_ns);
176 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
177 if (nd6_options(&ndopts) < 0) {
178 nd6log(LOG_INFO, "invalid ND option, ignored\n");
179 /* nd6_options have incremented stats */
180 goto freeit;
181 }
182
183 if (ndopts.nd_opts_src_lladdr) {
184 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
185 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
186 }
187
188 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
189 nd6log(LOG_INFO,
190 "bad DAD packet (link-layer address option)\n");
191 goto bad;
192 }
193
194 /*
195 * Attaching target link-layer address to the NA?
196 * (RFC 2461 7.2.4)
197 *
198 * NS IP dst is multicast MUST add
199 * Otherwise MAY be omitted
200 *
201 * In this implementation, we omit the target link-layer address
202 * in the "MAY" case.
203 */
204 #if 0 /* too much! */
205 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &daddr6);
206 if (ifa && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST))
207 tlladdr = 0;
208 else
209 #endif
210 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
211 tlladdr = 0;
212 else
213 tlladdr = 1;
214
215 /*
216 * Target address (taddr6) must be either:
217 * (1) Valid unicast/anycast address for my receiving interface,
218 * (2) Unicast address for which I'm offering proxy service, or
219 * (3) "tentative" address on which DAD is being performed.
220 */
221 /* (1) and (3) check. */
222 #if NCARP > 0
223 if (ifp->if_carp && ifp->if_type != IFT_CARP) {
224 int s = pserialize_read_enter();
225 ifa = carp_iamatch6(ifp->if_carp, &taddr6);
226 if (ifa != NULL)
227 ifa_acquire(ifa, &psref_ia);
228 pserialize_read_exit(s);
229 } else
230 ifa = NULL;
231 if (!ifa)
232 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
233 &psref_ia);
234 #else
235 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
236 &psref_ia);
237 #endif
238
239 /* (2) check. */
240 if (ifa == NULL) {
241 struct rtentry *rt;
242 struct sockaddr_in6 tsin6;
243
244 sockaddr_in6_init(&tsin6, &taddr6, 0, 0, 0);
245
246 rt = rtalloc1(sin6tosa(&tsin6), 0);
247 if (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
248 rt->rt_gateway->sa_family == AF_LINK) {
249 /*
250 * proxy NDP for single entry
251 */
252 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
253 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST, &psref_ia);
254 if (ifa) {
255 proxy = 1;
256 proxydl = satocsdl(rt->rt_gateway);
257 router = 0; /* XXX */
258 }
259 }
260 if (rt)
261 rt_unref(rt);
262 }
263 if (ifa == NULL) {
264 /*
265 * We've got an NS packet, and we don't have that address
266 * assigned for us. We MUST silently ignore it.
267 * See RFC2461 7.2.3.
268 */
269 goto freeit;
270 }
271 myaddr6 = *IFA_IN6(ifa);
272 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
273 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
274 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
275 goto freeit;
276
277 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
278 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
279 "(if %d, NS packet %d)\n",
280 IN6_PRINT(ip6buf, &taddr6),
281 ifp->if_addrlen, lladdrlen - 2);
282 goto bad;
283 }
284
285 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
286 nd6log(LOG_INFO, "duplicate IP6 address %s\n",
287 IN6_PRINT(ip6buf, &saddr6));
288 goto freeit;
289 }
290
291 /*
292 * We have neighbor solicitation packet, with target address equals to
293 * one of my tentative address.
294 *
295 * src addr how to process?
296 * --- ---
297 * multicast of course, invalid (rejected in ip6_input)
298 * unicast somebody is doing address resolution -> ignore
299 * unspec dup address detection
300 *
301 * The processing is defined in RFC 2462.
302 */
303 if (tentative) {
304 /*
305 * If source address is unspecified address, it is for
306 * duplicate address detection.
307 *
308 * If not, the packet is for addess resolution;
309 * silently ignore it.
310 */
311 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
312 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
313 ifa_release(ifa, &psref_ia);
314 ifa = NULL;
315
316 goto freeit;
317 }
318 ifa_release(ifa, &psref_ia);
319 ifa = NULL;
320
321 /*
322 * If the source address is unspecified address, entries must not
323 * be created or updated.
324 * It looks that sender is performing DAD. Output NA toward
325 * all-node multicast address, to tell the sender that I'm using
326 * the address.
327 * S bit ("solicited") must be zero.
328 */
329 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
330 struct in6_addr in6_all;
331
332 in6_all = in6addr_linklocal_allnodes;
333 if (in6_setscope(&in6_all, ifp, NULL) != 0)
334 goto bad;
335 nd6_na_output(ifp, &in6_all, &taddr6,
336 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
337 (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
338 tlladdr, (const struct sockaddr *)proxydl);
339 goto freeit;
340 }
341
342 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_NEIGHBOR_SOLICIT, 0);
343
344 nd6_na_output(ifp, &saddr6, &taddr6,
345 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
346 (router ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
347 tlladdr, (const struct sockaddr *)proxydl);
348 freeit:
349 ifa_release(ifa, &psref_ia);
350 m_put_rcvif_psref(ifp, &psref);
351 m_freem(m);
352 return;
353
354 bad:
355 nd6log(LOG_ERR, "src=%s\n", IN6_PRINT(ip6buf, &saddr6));
356 nd6log(LOG_ERR, "dst=%s\n", IN6_PRINT(ip6buf, &daddr6));
357 nd6log(LOG_ERR, "tgt=%s\n", IN6_PRINT(ip6buf, &taddr6));
358 ICMP6_STATINC(ICMP6_STAT_BADNS);
359 ifa_release(ifa, &psref_ia);
360 m_put_rcvif_psref(ifp, &psref);
361 m_freem(m);
362 }
363
364 /*
365 * Output a Neighbor Solicitation Message. Caller specifies:
366 * - ICMP6 header source IP6 address
367 * - ND6 header target IP6 address
368 * - ND6 header source datalink address
369 *
370 * Based on RFC 2461
371 * Based on RFC 2462 (duplicate address detection)
372 */
373 void
374 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
375 const struct in6_addr *taddr6,
376 struct in6_addr *hsrc,
377 uint8_t *nonce /* duplicate address detection */)
378 {
379 struct mbuf *m;
380 struct ip6_hdr *ip6;
381 struct nd_neighbor_solicit *nd_ns;
382 struct in6_addr *src, src_in;
383 struct ip6_moptions im6o;
384 int icmp6len;
385 int maxlen;
386 const void *mac;
387 struct route ro;
388
389 if (IN6_IS_ADDR_MULTICAST(taddr6))
390 return;
391
392 memset(&ro, 0, sizeof(ro));
393
394 /* estimate the size of message */
395 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
396 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
397 KASSERTMSG(max_linkhdr + maxlen < MCLBYTES,
398 "max_linkhdr + maxlen >= MCLBYTES (%d + %d >= %d)",
399 max_linkhdr, maxlen, MCLBYTES);
400
401 MGETHDR(m, M_DONTWAIT, MT_DATA);
402 if (m && max_linkhdr + maxlen >= MHLEN) {
403 MCLGET(m, M_DONTWAIT);
404 if ((m->m_flags & M_EXT) == 0) {
405 m_free(m);
406 m = NULL;
407 }
408 }
409 if (m == NULL)
410 return;
411 m_reset_rcvif(m);
412
413 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
414 m->m_flags |= M_MCAST;
415 im6o.im6o_multicast_if_index = if_get_index(ifp);
416 im6o.im6o_multicast_hlim = 255;
417 im6o.im6o_multicast_loop = 0;
418 }
419
420 icmp6len = sizeof(*nd_ns);
421 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
422 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
423
424 /* fill neighbor solicitation packet */
425 ip6 = mtod(m, struct ip6_hdr *);
426 ip6->ip6_flow = 0;
427 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
428 ip6->ip6_vfc |= IPV6_VERSION;
429 /* ip6->ip6_plen will be set later */
430 ip6->ip6_nxt = IPPROTO_ICMPV6;
431 ip6->ip6_hlim = 255;
432 if (daddr6)
433 ip6->ip6_dst = *daddr6;
434 else {
435 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
436 ip6->ip6_dst.s6_addr16[1] = 0;
437 ip6->ip6_dst.s6_addr32[1] = 0;
438 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
439 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
440 ip6->ip6_dst.s6_addr8[12] = 0xff;
441 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
442 goto bad;
443 }
444 if (nonce == NULL) {
445 int s;
446 /*
447 * RFC2461 7.2.2:
448 * "If the source address of the packet prompting the
449 * solicitation is the same as one of the addresses assigned
450 * to the outgoing interface, that address SHOULD be placed
451 * in the IP Source Address of the outgoing solicitation.
452 * Otherwise, any one of the addresses assigned to the
453 * interface should be used."
454 *
455 * We use the source address for the prompting packet
456 * (hsrc), if:
457 * - hsrc is given from the caller (by giving "ln"), and
458 * - hsrc belongs to the outgoing interface.
459 * Otherwise, we perform the source address selection as usual.
460 */
461 s = pserialize_read_enter();
462 if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc)) {
463 pserialize_read_exit(s);
464 src = hsrc;
465 } else {
466 int error;
467 struct sockaddr_in6 dst_sa;
468
469 pserialize_read_exit(s);
470
471 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
472
473 error = in6_selectsrc(&dst_sa, NULL,
474 NULL, &ro, NULL, NULL, NULL, &src_in);
475 if (error != 0) {
476 char ip6buf[INET6_ADDRSTRLEN];
477 nd6log(LOG_DEBUG, "source can't be "
478 "determined: dst=%s, error=%d\n",
479 IN6_PRINT(ip6buf, &dst_sa.sin6_addr),
480 error);
481 pserialize_read_exit(s);
482 goto bad;
483 }
484 src = &src_in;
485 }
486 } else {
487 /*
488 * Source address for DAD packet must always be IPv6
489 * unspecified address. (0::0)
490 * We actually don't have to 0-clear the address (we did it
491 * above), but we do so here explicitly to make the intention
492 * clearer.
493 */
494 memset(&src_in, 0, sizeof(src_in));
495 src = &src_in;
496 }
497 ip6->ip6_src = *src;
498 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
499 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
500 nd_ns->nd_ns_code = 0;
501 nd_ns->nd_ns_reserved = 0;
502 nd_ns->nd_ns_target = *taddr6;
503 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
504
505 /*
506 * Add source link-layer address option.
507 *
508 * spec implementation
509 * --- ---
510 * DAD packet MUST NOT do not add the option
511 * there's no link layer address:
512 * impossible do not add the option
513 * there's link layer address:
514 * Multicast NS MUST add one add the option
515 * Unicast NS SHOULD add one add the option
516 */
517 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
518 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
519 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
520 /* 8 byte alignments... */
521 optlen = (optlen + 7) & ~7;
522
523 m->m_pkthdr.len += optlen;
524 m->m_len += optlen;
525 icmp6len += optlen;
526 memset((void *)nd_opt, 0, optlen);
527 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
528 nd_opt->nd_opt_len = optlen >> 3;
529 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
530 }
531
532 /* Add a nonce option (RFC 3971) to detect looped back NS messages.
533 * This behavior is documented in RFC 7527. */
534 if (nonce != NULL) {
535 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
536 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
537
538 /* 8-byte alignment is required. */
539 optlen = (optlen + 7) & ~7;
540 m->m_pkthdr.len += optlen;
541 m->m_len += optlen;
542 icmp6len += optlen;
543 memset(nd_opt, 0, optlen);
544 nd_opt->nd_opt_type = ND_OPT_NONCE;
545 nd_opt->nd_opt_len = optlen >> 3;
546 memcpy(nd_opt + 1, nonce, ND_OPT_NONCE_LEN);
547 }
548
549 ip6->ip6_plen = htons((u_int16_t)icmp6len);
550 nd_ns->nd_ns_cksum = 0;
551 nd_ns->nd_ns_cksum =
552 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
553
554 ip6_output(m, NULL, &ro, nonce != NULL ? IPV6_UNSPECSRC : 0,
555 &im6o, NULL, NULL);
556 icmp6_ifstat_inc(ifp, ifs6_out_msg);
557 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
558 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_SOLICIT);
559
560 rtcache_free(&ro);
561 return;
562
563 bad:
564 rtcache_free(&ro);
565 m_freem(m);
566 return;
567 }
568
569 /*
570 * Neighbor advertisement input handling.
571 *
572 * Based on RFC 2461
573 * Based on RFC 2462 (duplicate address detection)
574 *
575 * the following items are not implemented yet:
576 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
577 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
578 */
579 void
580 nd6_na_input(struct mbuf *m, int off, int icmp6len)
581 {
582 struct ifnet *ifp;
583 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
584 struct nd_neighbor_advert *nd_na;
585 struct in6_addr saddr6 = ip6->ip6_src;
586 struct in6_addr daddr6 = ip6->ip6_dst;
587 struct in6_addr taddr6;
588 int flags;
589 int is_router;
590 int is_solicited;
591 int is_override;
592 char *lladdr = NULL;
593 int lladdrlen = 0;
594 struct ifaddr *ifa;
595 struct llentry *ln = NULL;
596 union nd_opts ndopts;
597 struct sockaddr_in6 ssin6;
598 int rt_announce;
599 bool checklink = false;
600 struct psref psref;
601 struct psref psref_ia;
602 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
603
604 ifp = m_get_rcvif_psref(m, &psref);
605 if (ifp == NULL)
606 goto freeit;
607
608 if (ip6->ip6_hlim != 255) {
609 nd6log(LOG_ERR,
610 "invalid hlim (%d) from %s to %s on %s\n",
611 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
612 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
613 goto bad;
614 }
615
616 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
617 if (nd_na == NULL) {
618 m_put_rcvif_psref(ifp, &psref);
619 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
620 return;
621 }
622
623 flags = nd_na->nd_na_flags_reserved;
624 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
625 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
626 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
627
628 taddr6 = nd_na->nd_na_target;
629 if (in6_setscope(&taddr6, ifp, NULL)) {
630 goto bad;
631 }
632
633 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
634 nd6log(LOG_ERR, "invalid target address %s\n",
635 IN6_PRINT(ip6buf, &taddr6));
636 goto bad;
637 }
638 if (is_solicited && IN6_IS_ADDR_MULTICAST(&daddr6)) {
639 nd6log(LOG_ERR, "a solicited adv is multicasted\n");
640 goto bad;
641 }
642
643 icmp6len -= sizeof(*nd_na);
644 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
645 if (nd6_options(&ndopts) < 0) {
646 nd6log(LOG_INFO, "invalid ND option, ignored\n");
647 /* nd6_options have incremented stats */
648 goto freeit;
649 }
650
651 if (ndopts.nd_opts_tgt_lladdr) {
652 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
653 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
654 }
655
656 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6, &psref_ia);
657
658 /*
659 * Target address matches one of my interface address.
660 *
661 * If my address is tentative, this means that there's somebody
662 * already using the same address as mine. This indicates DAD failure.
663 * This is defined in RFC 2462.
664 *
665 * Otherwise, process as defined in RFC 2461.
666 */
667 if (ifa
668 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
669 nd6_dad_na_input(ifa);
670 ifa_release(ifa, &psref_ia);
671 ifa = NULL;
672 goto freeit;
673 }
674
675 /* Just for safety, maybe unnecessary. */
676 if (ifa) {
677 log(LOG_ERR,
678 "nd6_na_input: duplicate IP6 address %s\n",
679 IN6_PRINT(ip6buf, &taddr6));
680 ifa_release(ifa, &psref_ia);
681 ifa = NULL;
682 goto freeit;
683 }
684
685 /*
686 * Make sure the source address is from a neighbor's address.
687 */
688 sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
689 if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
690 nd6log(LOG_INFO, "ND packet from non-neighbor %s on %s\n",
691 IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
692 goto bad;
693 }
694
695 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
696 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
697 "(if %d, NA packet %d)\n", IN6_PRINT(ip6buf, &taddr6),
698 ifp->if_addrlen, lladdrlen - 2);
699 goto bad;
700 }
701
702 /*
703 * If no neighbor cache entry is found, NA SHOULD silently be
704 * discarded.
705 */
706 ln = nd6_lookup(&taddr6, ifp, true);
707 if (ln == NULL)
708 goto freeit;
709
710 rt_announce = 0;
711 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
712 /*
713 * If the link-layer has address, and no lladdr option came,
714 * discard the packet.
715 */
716 if (ifp->if_addrlen && !lladdr)
717 goto freeit;
718
719 /*
720 * Record link-layer address, and update the state.
721 */
722 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
723 ln->la_flags |= LLE_VALID;
724 rt_announce = 1;
725 if (is_solicited) {
726 ln->ln_state = ND6_LLINFO_REACHABLE;
727 ln->ln_byhint = 0;
728 if (!ND6_LLINFO_PERMANENT(ln)) {
729 nd6_llinfo_settimer(ln,
730 ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
731 }
732 } else {
733 ln->ln_state = ND6_LLINFO_STALE;
734 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
735 }
736 if ((ln->ln_router = is_router) != 0) {
737 /*
738 * This means a router's state has changed from
739 * non-reachable to probably reachable, and might
740 * affect the status of associated prefixes..
741 */
742 checklink = true;
743 }
744 } else {
745 int llchange;
746
747 /*
748 * Check if the link-layer address has changed or not.
749 */
750 if (lladdr == NULL)
751 llchange = 0;
752 else {
753 if (ln->la_flags & LLE_VALID) {
754 if (memcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
755 llchange = rt_announce = 1;
756 else
757 llchange = 0;
758 } else
759 llchange = rt_announce = 1;
760 }
761
762 /*
763 * This is VERY complex. Look at it with care.
764 *
765 * override solicit lladdr llchange action
766 * (L: record lladdr)
767 *
768 * 0 0 n -- (2c)
769 * 0 0 y n (2b) L
770 * 0 0 y y (1) REACHABLE->STALE
771 * 0 1 n -- (2c) *->REACHABLE
772 * 0 1 y n (2b) L *->REACHABLE
773 * 0 1 y y (1) REACHABLE->STALE
774 * 1 0 n -- (2a)
775 * 1 0 y n (2a) L
776 * 1 0 y y (2a) L *->STALE
777 * 1 1 n -- (2a) *->REACHABLE
778 * 1 1 y n (2a) L *->REACHABLE
779 * 1 1 y y (2a) L *->REACHABLE
780 */
781 if (!is_override && lladdr != NULL && llchange) { /* (1) */
782 /*
783 * If state is REACHABLE, make it STALE.
784 * no other updates should be done.
785 */
786 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
787 ln->ln_state = ND6_LLINFO_STALE;
788 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
789 }
790 goto freeit;
791 } else if (is_override /* (2a) */
792 || (!is_override && lladdr != NULL && !llchange) /* (2b) */
793 || lladdr == NULL) { /* (2c) */
794 /*
795 * Update link-local address, if any.
796 */
797 if (lladdr != NULL) {
798 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
799 ln->la_flags |= LLE_VALID;
800 }
801
802 /*
803 * If solicited, make the state REACHABLE.
804 * If not solicited and the link-layer address was
805 * changed, make it STALE.
806 */
807 if (is_solicited) {
808 ln->ln_state = ND6_LLINFO_REACHABLE;
809 ln->ln_byhint = 0;
810 if (!ND6_LLINFO_PERMANENT(ln)) {
811 nd6_llinfo_settimer(ln,
812 ND_IFINFO(ifp)->reachable * hz);
813 }
814 } else {
815 if (lladdr && llchange) {
816 ln->ln_state = ND6_LLINFO_STALE;
817 nd6_llinfo_settimer(ln,
818 nd6_gctimer * hz);
819 }
820 }
821 }
822
823 if (ln->ln_router && !is_router) {
824 /*
825 * The peer dropped the router flag.
826 * Remove the sender from the Default Router List and
827 * update the Destination Cache entries.
828 */
829 struct nd_defrouter *dr;
830 const struct in6_addr *in6;
831
832 in6 = &ln->r_l3addr.addr6;
833
834 ND6_WLOCK();
835 dr = nd6_defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
836 if (dr)
837 nd6_defrtrlist_del(dr, NULL);
838 else if (!ip6_forwarding) {
839 /*
840 * Even if the neighbor is not in the default
841 * router list, the neighbor may be used
842 * as a next hop for some destinations
843 * (e.g. redirect case). So we must
844 * call nd6_rt_flush explicitly.
845 */
846 nd6_rt_flush(&ip6->ip6_src, ln->lle_tbl->llt_ifp);
847 }
848 ND6_UNLOCK();
849 }
850 ln->ln_router = is_router;
851 }
852 /*
853 * XXX: does this matter?
854 * rt->rt_flags &= ~RTF_REJECT;
855 */
856 ln->ln_asked = 0;
857 nd6_llinfo_release_pkts(ln, ifp);
858 /* FIXME */
859 #if 0
860 if (rt_announce) /* tell user process about any new lladdr */
861 rt_newmsg(RTM_CHANGE, rt);
862 #endif
863
864 freeit:
865 if (ln != NULL)
866 LLE_WUNLOCK(ln);
867
868 if (checklink) {
869 ND6_WLOCK();
870 nd6_pfxlist_onlink_check();
871 ND6_UNLOCK();
872 }
873
874 m_put_rcvif_psref(ifp, &psref);
875 m_freem(m);
876 return;
877
878 bad:
879 if (ln != NULL)
880 LLE_WUNLOCK(ln);
881
882 ICMP6_STATINC(ICMP6_STAT_BADNA);
883 m_put_rcvif_psref(ifp, &psref);
884 m_freem(m);
885 }
886
887 /*
888 * Neighbor advertisement output handling.
889 *
890 * Based on RFC 2461
891 *
892 * the following items are not implemented yet:
893 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
894 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
895 */
896 void
897 nd6_na_output(
898 struct ifnet *ifp,
899 const struct in6_addr *daddr6_0,
900 const struct in6_addr *taddr6,
901 u_long flags,
902 int tlladdr, /* 1 if include target link-layer address */
903 const struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
904 {
905 struct mbuf *m;
906 struct ip6_hdr *ip6;
907 struct nd_neighbor_advert *nd_na;
908 struct ip6_moptions im6o;
909 struct sockaddr *dst;
910 union {
911 struct sockaddr dst;
912 struct sockaddr_in6 dst6;
913 } u;
914 struct in6_addr daddr6;
915 int icmp6len, maxlen, error;
916 const void *mac;
917 struct route ro;
918
919 mac = NULL;
920 memset(&ro, 0, sizeof(ro));
921
922 daddr6 = *daddr6_0; /* make a local copy for modification */
923
924 /* estimate the size of message */
925 maxlen = sizeof(*ip6) + sizeof(*nd_na);
926 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
927 KASSERTMSG(max_linkhdr + maxlen < MCLBYTES,
928 "max_linkhdr + maxlen >= MCLBYTES (%d + %d >= %d)",
929 max_linkhdr, maxlen, MCLBYTES);
930
931 MGETHDR(m, M_DONTWAIT, MT_DATA);
932 if (m && max_linkhdr + maxlen >= MHLEN) {
933 MCLGET(m, M_DONTWAIT);
934 if ((m->m_flags & M_EXT) == 0) {
935 m_free(m);
936 m = NULL;
937 }
938 }
939 if (m == NULL)
940 return;
941 m_reset_rcvif(m);
942
943 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
944 m->m_flags |= M_MCAST;
945 im6o.im6o_multicast_if_index = if_get_index(ifp);
946 im6o.im6o_multicast_hlim = 255;
947 im6o.im6o_multicast_loop = 0;
948 }
949
950 icmp6len = sizeof(*nd_na);
951 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
952 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
953
954 /* fill neighbor advertisement packet */
955 ip6 = mtod(m, struct ip6_hdr *);
956 ip6->ip6_flow = 0;
957 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
958 ip6->ip6_vfc |= IPV6_VERSION;
959 ip6->ip6_nxt = IPPROTO_ICMPV6;
960 ip6->ip6_hlim = 255;
961 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
962 /* reply to DAD */
963 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
964 daddr6.s6_addr16[1] = 0;
965 daddr6.s6_addr32[1] = 0;
966 daddr6.s6_addr32[2] = 0;
967 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
968 if (in6_setscope(&daddr6, ifp, NULL))
969 goto bad;
970
971 flags &= ~ND_NA_FLAG_SOLICITED;
972 }
973 ip6->ip6_dst = daddr6;
974 sockaddr_in6_init(&u.dst6, &daddr6, 0, 0, 0);
975 dst = &u.dst;
976 if (rtcache_setdst(&ro, dst) != 0)
977 goto bad;
978
979 /*
980 * Select a source whose scope is the same as that of the dest.
981 */
982 error = in6_selectsrc(satosin6(dst), NULL, NULL, &ro, NULL, NULL, NULL,
983 &ip6->ip6_src);
984 if (error != 0) {
985 char ip6buf[INET6_ADDRSTRLEN];
986 nd6log(LOG_DEBUG, "source can't be "
987 "determined: dst=%s, error=%d\n",
988 IN6_PRINT(ip6buf, &satocsin6(dst)->sin6_addr), error);
989 goto bad;
990 }
991 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
992 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
993 nd_na->nd_na_code = 0;
994 nd_na->nd_na_target = *taddr6;
995 in6_clearscope(&nd_na->nd_na_target); /* XXX */
996
997 /*
998 * "tlladdr" indicates NS's condition for adding tlladdr or not.
999 * see nd6_ns_input() for details.
1000 * Basically, if NS packet is sent to unicast/anycast addr,
1001 * target lladdr option SHOULD NOT be included.
1002 */
1003 if (tlladdr) {
1004 /*
1005 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1006 * lladdr in sdl0. If we are not proxying (sending NA for
1007 * my address) use lladdr configured for the interface.
1008 */
1009 if (sdl0 == NULL)
1010 mac = nd6_ifptomac(ifp);
1011 else if (sdl0->sa_family == AF_LINK) {
1012 const struct sockaddr_dl *sdl;
1013 sdl = satocsdl(sdl0);
1014 if (sdl->sdl_alen == ifp->if_addrlen)
1015 mac = CLLADDR(sdl);
1016 }
1017 }
1018 if (tlladdr && mac) {
1019 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1020 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1021
1022 /* roundup to 8 bytes alignment! */
1023 optlen = (optlen + 7) & ~7;
1024
1025 m->m_pkthdr.len += optlen;
1026 m->m_len += optlen;
1027 icmp6len += optlen;
1028 memset((void *)nd_opt, 0, optlen);
1029 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1030 nd_opt->nd_opt_len = optlen >> 3;
1031 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
1032 } else
1033 flags &= ~ND_NA_FLAG_OVERRIDE;
1034
1035 ip6->ip6_plen = htons((u_int16_t)icmp6len);
1036 nd_na->nd_na_flags_reserved = flags;
1037 nd_na->nd_na_cksum = 0;
1038 nd_na->nd_na_cksum =
1039 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1040
1041 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1042
1043 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1044 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1045 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_ADVERT);
1046
1047 rtcache_free(&ro);
1048 return;
1049
1050 bad:
1051 rtcache_free(&ro);
1052 m_freem(m);
1053 return;
1054 }
1055
1056 const void *
1057 nd6_ifptomac(const struct ifnet *ifp)
1058 {
1059 switch (ifp->if_type) {
1060 case IFT_ARCNET:
1061 case IFT_ETHER:
1062 case IFT_FDDI:
1063 case IFT_IEEE1394:
1064 case IFT_PROPVIRTUAL:
1065 case IFT_CARP:
1066 case IFT_L2VLAN:
1067 case IFT_IEEE80211:
1068 return CLLADDR(ifp->if_sadl);
1069 default:
1070 return NULL;
1071 }
1072 }
1073
1074 TAILQ_HEAD(dadq_head, dadq);
1075 struct dadq {
1076 TAILQ_ENTRY(dadq) dad_list;
1077 struct ifaddr *dad_ifa;
1078 int dad_count; /* max NS to send */
1079 int dad_ns_tcount; /* # of trials to send NS */
1080 int dad_ns_ocount; /* NS sent so far */
1081 int dad_ns_icount;
1082 int dad_na_icount;
1083 int dad_ns_lcount; /* looped back NS */
1084 struct callout dad_timer_ch;
1085 #define ND_OPT_NONCE_STORE 3 /* dad_count should not exceed this */
1086 /*
1087 * The default ip6_dad_count is 1 as specified by RFC 4862 and
1088 * practically must users won't exceed this.
1089 * A storage of 3 is defaulted to here, in-case the administrator wants
1090 * to match the equivalent behaviour in our ARP implementation.
1091 * This constraint could be removed by sending the on wire nonce as
1092 * hmac(key, dad_ns_ocount), but that would increase the nonce size
1093 * sent on the wire.
1094 */
1095 uint8_t dad_nonce[ND_OPT_NONCE_STORE][ND_OPT_NONCE_LEN];
1096 };
1097
1098 static struct dadq_head dadq;
1099 static int dad_init = 0;
1100 static kmutex_t nd6_dad_lock;
1101
1102 static struct dadq *
1103 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1104 {
1105 struct dadq *dp;
1106 int i, nonce_max;
1107
1108 KASSERT(mutex_owned(&nd6_dad_lock));
1109
1110 TAILQ_FOREACH(dp, &dadq, dad_list) {
1111 if (dp->dad_ifa != ifa)
1112 continue;
1113
1114 if (nonce == NULL ||
1115 nonce->nd_opt_nonce_len != (ND_OPT_NONCE_LEN + 2) / 8)
1116 break;
1117
1118 nonce_max = MIN(dp->dad_ns_ocount, ND_OPT_NONCE_STORE);
1119 for (i = 0; i < nonce_max; i++) {
1120 if (memcmp(nonce->nd_opt_nonce,
1121 dp->dad_nonce[i],
1122 ND_OPT_NONCE_LEN) == 0)
1123 break;
1124 }
1125 if (i < nonce_max) {
1126 char ip6buf[INET6_ADDRSTRLEN];
1127
1128 log(LOG_DEBUG,
1129 "%s: detected a looped back NS message for %s\n",
1130 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???",
1131 IN6_PRINT(ip6buf, IFA_IN6(ifa)));
1132 dp->dad_ns_lcount++;
1133 continue;
1134 }
1135
1136 break;
1137 }
1138 return dp;
1139 }
1140
1141 static void
1142 nd6_dad_starttimer(struct dadq *dp, int ticks)
1143 {
1144
1145 callout_reset(&dp->dad_timer_ch, ticks,
1146 (void (*)(void *))nd6_dad_timer, dp);
1147 }
1148
1149 static void
1150 nd6_dad_stoptimer(struct dadq *dp)
1151 {
1152
1153 KASSERT(mutex_owned(&nd6_dad_lock));
1154
1155 TAILQ_REMOVE(&dadq, dp, dad_list);
1156 /* Tell the timer that dp is being destroyed. */
1157 dp->dad_ifa = NULL;
1158 callout_halt(&dp->dad_timer_ch, &nd6_dad_lock);
1159 }
1160
1161 static void
1162 nd6_dad_destroytimer(struct dadq *dp)
1163 {
1164
1165 KASSERT(dp->dad_ifa == NULL);
1166 callout_destroy(&dp->dad_timer_ch);
1167 kmem_intr_free(dp, sizeof(*dp));
1168 }
1169
1170 /*
1171 * Start Duplicate Address Detection (DAD) for specified interface address.
1172 *
1173 * Note that callout is used when xtick > 0 and not when xtick == 0.
1174 *
1175 * xtick: minimum delay ticks for IFF_UP event
1176 */
1177 void
1178 nd6_dad_start(struct ifaddr *ifa, int xtick)
1179 {
1180 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1181 struct dadq *dp;
1182 char ip6buf[INET6_ADDRSTRLEN];
1183
1184 if (!dad_init) {
1185 TAILQ_INIT(&dadq);
1186 mutex_init(&nd6_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1187 dad_init++;
1188 }
1189
1190 /*
1191 * If we don't need DAD, don't do it.
1192 * There are several cases:
1193 * - DAD is disabled (ip6_dad_count == 0)
1194 * - the interface address is anycast
1195 */
1196 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1197 log(LOG_DEBUG,
1198 "nd6_dad_start: called with non-tentative address "
1199 "%s(%s)\n",
1200 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1201 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1202 return;
1203 }
1204 if (ia->ia6_flags & IN6_IFF_ANYCAST || !ip6_dad_count) {
1205 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1206 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1207 return;
1208 }
1209 KASSERT(ifa->ifa_ifp != NULL);
1210 if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1211 return;
1212
1213 dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1214
1215 mutex_enter(&nd6_dad_lock);
1216 if (nd6_dad_find(ifa, NULL) != NULL) {
1217 mutex_exit(&nd6_dad_lock);
1218 /* DAD already in progress */
1219 if (dp != NULL)
1220 kmem_intr_free(dp, sizeof(*dp));
1221 return;
1222 }
1223
1224 if (dp == NULL) {
1225 mutex_exit(&nd6_dad_lock);
1226 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1227 "%s(%s)\n",
1228 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1229 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1230 return;
1231 }
1232
1233 /*
1234 * Send NS packet for DAD, ip6_dad_count times.
1235 * Note that we must delay the first transmission, if this is the
1236 * first packet to be sent from the interface after interface
1237 * (re)initialization.
1238 */
1239 callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1240 dp->dad_ifa = ifa;
1241 ifaref(ifa); /* just for safety */
1242 dp->dad_count = ip6_dad_count;
1243 dp->dad_ns_icount = dp->dad_na_icount = 0;
1244 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1245 dp->dad_ns_lcount = 0;
1246 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1247
1248 nd6log(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1249 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1250
1251 if (xtick == 0) {
1252 nd6_dad_ns_output(dp, ifa);
1253 nd6_dad_starttimer(dp,
1254 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1255 } else
1256 nd6_dad_starttimer(dp, xtick);
1257 mutex_exit(&nd6_dad_lock);
1258 }
1259
1260 /*
1261 * terminate DAD unconditionally. used for address removals.
1262 */
1263 void
1264 nd6_dad_stop(struct ifaddr *ifa)
1265 {
1266 struct dadq *dp;
1267
1268 if (!dad_init)
1269 return;
1270
1271 mutex_enter(&nd6_dad_lock);
1272 dp = nd6_dad_find(ifa, NULL);
1273 if (dp == NULL) {
1274 mutex_exit(&nd6_dad_lock);
1275 /* DAD wasn't started yet */
1276 return;
1277 }
1278
1279 /* Prevent the timer from running anymore. */
1280 nd6_dad_stoptimer(dp);
1281
1282 mutex_exit(&nd6_dad_lock);
1283
1284 nd6_dad_destroytimer(dp);
1285 ifafree(ifa);
1286 }
1287
1288 static void
1289 nd6_dad_timer(struct dadq *dp)
1290 {
1291 struct ifaddr *ifa;
1292 struct in6_ifaddr *ia;
1293 int duplicate = 0;
1294 char ip6buf[INET6_ADDRSTRLEN];
1295 bool need_free = false;
1296
1297 KERNEL_LOCK_UNLESS_NET_MPSAFE();
1298 mutex_enter(&nd6_dad_lock);
1299
1300 ifa = dp->dad_ifa;
1301 if (ifa == NULL) {
1302 /* dp is being destroyed by someone. Do nothing. */
1303 goto done;
1304 }
1305
1306 ia = (struct in6_ifaddr *)ifa;
1307 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1308 log(LOG_ERR, "nd6_dad_timer: called with duplicate address "
1309 "%s(%s)\n",
1310 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1311 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1312 goto done;
1313 }
1314 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1315 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1316 "%s(%s)\n",
1317 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1318 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1319 goto done;
1320 }
1321
1322 /* timeouted with IFF_{RUNNING,UP} check */
1323 if (dp->dad_ns_tcount > dad_maxtry) {
1324 nd6log(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1325 if_name(ifa->ifa_ifp));
1326
1327 nd6_dad_stoptimer(dp);
1328 need_free = true;
1329 goto done;
1330 }
1331
1332 /* Need more checks? */
1333 if (dp->dad_ns_ocount < dp->dad_count) {
1334 /*
1335 * We have more NS to go. Send NS packet for DAD.
1336 */
1337 nd6_dad_ns_output(dp, ifa);
1338 nd6_dad_starttimer(dp,
1339 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1340 } else {
1341 /*
1342 * We have transmitted sufficient number of DAD packets.
1343 * See what we've got.
1344 */
1345 if (dp->dad_na_icount) {
1346 /*
1347 * the check is in nd6_dad_na_input(),
1348 * but just in case
1349 */
1350 duplicate++;
1351 }
1352
1353 if (dp->dad_ns_icount) {
1354 /* We've seen NS, means DAD has failed. */
1355 duplicate++;
1356 }
1357
1358 if (duplicate) {
1359 nd6_dad_duplicated(dp);
1360 nd6_dad_stoptimer(dp);
1361 need_free = true;
1362 } else {
1363 /*
1364 * We are done with DAD. No NA came, no NS came.
1365 * No duplicate address found.
1366 */
1367 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1368 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1369
1370 nd6log(LOG_DEBUG,
1371 "%s: DAD complete for %s - no duplicates found\n",
1372 if_name(ifa->ifa_ifp),
1373 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1374
1375 nd6_dad_stoptimer(dp);
1376 need_free = true;
1377 }
1378 }
1379 done:
1380 mutex_exit(&nd6_dad_lock);
1381
1382 if (need_free) {
1383 nd6_dad_destroytimer(dp);
1384 KASSERT(ifa != NULL);
1385 ifafree(ifa);
1386 }
1387
1388 KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1389 }
1390
1391 static void
1392 nd6_dad_duplicated(struct dadq *dp)
1393 {
1394 struct ifaddr *ifa = dp->dad_ifa;
1395 struct in6_ifaddr *ia;
1396 struct ifnet *ifp;
1397 char ip6buf[INET6_ADDRSTRLEN];
1398
1399 KASSERT(mutex_owned(&nd6_dad_lock));
1400 KASSERT(ifa != NULL);
1401
1402 ifp = ifa->ifa_ifp;
1403 ia = (struct in6_ifaddr *)ifa;
1404 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1405 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1406 if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1407 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1408 dp->dad_na_icount);
1409
1410 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1411 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1412
1413 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1414 if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1415 log(LOG_ERR, "%s: manual intervention required\n",
1416 if_name(ifp));
1417
1418 /* Inform the routing socket that DAD has completed */
1419 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1420
1421 /*
1422 * If the address is a link-local address formed from an interface
1423 * identifier based on the hardware address which is supposed to be
1424 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1425 * operation on the interface SHOULD be disabled.
1426 * [rfc2462bis-03 Section 5.4.5]
1427 */
1428 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1429 struct in6_addr in6;
1430
1431 /*
1432 * To avoid over-reaction, we only apply this logic when we are
1433 * very sure that hardware addresses are supposed to be unique.
1434 */
1435 switch (ifp->if_type) {
1436 case IFT_ETHER:
1437 case IFT_FDDI:
1438 case IFT_ATM:
1439 case IFT_IEEE1394:
1440 case IFT_IEEE80211:
1441 in6 = ia->ia_addr.sin6_addr;
1442 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1443 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1444 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1445 log(LOG_ERR, "%s: possible hardware address "
1446 "duplication detected, disable IPv6\n",
1447 if_name(ifp));
1448 }
1449 break;
1450 }
1451 }
1452 }
1453
1454 static void
1455 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1456 {
1457 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1458 struct ifnet *ifp = ifa->ifa_ifp;
1459 uint8_t *nonce;
1460
1461 dp->dad_ns_tcount++;
1462 if ((ifp->if_flags & IFF_UP) == 0) {
1463 #if 0
1464 printf("%s: interface down?\n", if_name(ifp));
1465 #endif
1466 return;
1467 }
1468 if ((ifp->if_flags & IFF_RUNNING) == 0) {
1469 #if 0
1470 printf("%s: interface not running?\n", if_name(ifp));
1471 #endif
1472 return;
1473 }
1474
1475 dp->dad_ns_tcount = 0;
1476 nonce = dp->dad_nonce[dp->dad_ns_ocount % ND_OPT_NONCE_STORE];
1477 cprng_fast(nonce, ND_OPT_NONCE_LEN);
1478 dp->dad_ns_ocount++;
1479
1480 nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, nonce);
1481 }
1482
1483 static void
1484 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1485 {
1486 struct dadq *dp;
1487 int duplicate;
1488
1489 if (ifa == NULL)
1490 panic("ifa == NULL in nd6_dad_ns_input");
1491
1492 duplicate = 0;
1493
1494 mutex_enter(&nd6_dad_lock);
1495 dp = nd6_dad_find(ifa, nonce);
1496
1497 /*
1498 * if I'm yet to start DAD, someone else started using this address
1499 * first. I have a duplicate and you win.
1500 */
1501 if (dp == NULL || dp->dad_ns_ocount == 0)
1502 duplicate++;
1503
1504 /* XXX more checks for loopback situation - see nd6_dad_timer too */
1505
1506 if (duplicate) {
1507 if (dp) {
1508 nd6_dad_duplicated(dp);
1509 nd6_dad_stoptimer(dp);
1510 }
1511 } else {
1512 /*
1513 * not sure if I got a duplicate.
1514 * increment ns count and see what happens.
1515 */
1516 if (dp)
1517 dp->dad_ns_icount++;
1518 }
1519 mutex_exit(&nd6_dad_lock);
1520
1521 if (duplicate && dp) {
1522 nd6_dad_destroytimer(dp);
1523 ifafree(ifa);
1524 }
1525 }
1526
1527 static void
1528 nd6_dad_na_input(struct ifaddr *ifa)
1529 {
1530 struct dadq *dp;
1531
1532 KASSERT(ifa != NULL);
1533
1534 mutex_enter(&nd6_dad_lock);
1535
1536 dp = nd6_dad_find(ifa, NULL);
1537 if (dp == NULL) {
1538 mutex_exit(&nd6_dad_lock);
1539 return;
1540 }
1541
1542 dp->dad_na_icount++;
1543
1544 /* remove the address. */
1545 nd6_dad_duplicated(dp);
1546 nd6_dad_stoptimer(dp);
1547
1548 mutex_exit(&nd6_dad_lock);
1549
1550 nd6_dad_destroytimer(dp);
1551 ifafree(ifa);
1552 }
1553