nd6_nbr.c revision 1.174 1 /* $NetBSD: nd6_nbr.c,v 1.174 2019/09/25 09:52:32 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 1.174 2019/09/25 09:52:32 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 struct dadq;
79 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *, bool *);
80 static bool nd6_dad_ownnonce(struct ifaddr *, struct nd_opt_nonce *nonce);
81 static void nd6_dad_starttimer(struct dadq *, int);
82 static void nd6_dad_destroytimer(struct dadq *);
83 static void nd6_dad_timer(struct dadq *);
84 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
85 static void nd6_dad_input(struct ifaddr *, struct nd_opt_nonce *,
86 const struct sockaddr_dl *);
87 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *,
88 const struct sockaddr_dl *);
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 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
170 nd6log(LOG_INFO, "bad NS target (multicast)\n");
171 goto bad;
172 }
173
174 icmp6len -= sizeof(*nd_ns);
175 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
176 if (nd6_options(&ndopts) < 0) {
177 nd6log(LOG_INFO, "invalid ND option, ignored\n");
178 /* nd6_options have incremented stats */
179 goto freeit;
180 }
181
182 if (ndopts.nd_opts_src_lladdr) {
183 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
184 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
185 }
186
187 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
188 nd6log(LOG_INFO,
189 "bad DAD packet (link-layer address option)\n");
190 goto bad;
191 }
192
193 /*
194 * Attaching target link-layer address to the NA?
195 * (RFC 2461 7.2.4)
196 *
197 * NS IP dst is multicast MUST add
198 * Otherwise MAY be omitted
199 *
200 * In this implementation, we omit the target link-layer address
201 * in the "MAY" case.
202 */
203 #if 0 /* too much! */
204 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &daddr6);
205 if (ifa && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST))
206 tlladdr = 0;
207 else
208 #endif
209 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
210 tlladdr = 0;
211 else
212 tlladdr = 1;
213
214 /*
215 * Target address (taddr6) must be either:
216 * (1) Valid unicast/anycast address for my receiving interface,
217 * (2) Unicast address for which I'm offering proxy service, or
218 * (3) "tentative" address on which DAD is being performed.
219 */
220 /* (1) and (3) check. */
221 #if NCARP > 0
222 if (ifp->if_carp && ifp->if_type != IFT_CARP) {
223 int s = pserialize_read_enter();
224 ifa = carp_iamatch6(ifp->if_carp, &taddr6);
225 if (ifa != NULL)
226 ifa_acquire(ifa, &psref_ia);
227 pserialize_read_exit(s);
228 } else
229 ifa = NULL;
230 if (!ifa)
231 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
232 &psref_ia);
233 #else
234 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
235 &psref_ia);
236 #endif
237
238 /* (2) check. */
239 if (ifa == NULL) {
240 struct rtentry *rt;
241 struct sockaddr_in6 tsin6;
242
243 sockaddr_in6_init(&tsin6, &taddr6, 0, 0, 0);
244
245 rt = rtalloc1(sin6tosa(&tsin6), 0);
246 if (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
247 rt->rt_gateway->sa_family == AF_LINK) {
248 /*
249 * proxy NDP for single entry
250 */
251 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
252 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST, &psref_ia);
253 if (ifa) {
254 proxy = 1;
255 proxydl = satocsdl(rt->rt_gateway);
256 router = 0; /* XXX */
257 }
258 }
259 if (rt)
260 rt_unref(rt);
261 }
262 if (ifa == NULL) {
263 /*
264 * We've got an NS packet, and we don't have that address
265 * assigned for us. We MUST silently ignore it.
266 * See RFC2461 7.2.3.
267 */
268 goto freeit;
269 }
270 myaddr6 = *IFA_IN6(ifa);
271 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
272 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
273 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
274 goto freeit;
275
276 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
277 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
278 "(if %d, NS packet %d)\n",
279 IN6_PRINT(ip6buf, &taddr6),
280 ifp->if_addrlen, lladdrlen - 2);
281 goto bad;
282 }
283
284 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
285 nd6log(LOG_INFO, "duplicate IP6 address %s\n",
286 IN6_PRINT(ip6buf, &saddr6));
287 goto freeit;
288 }
289
290 /*
291 * We have neighbor solicitation packet, with target address equals to
292 * one of my tentative address.
293 *
294 * src addr how to process?
295 * --- ---
296 * multicast of course, invalid (rejected in ip6_input)
297 * unicast somebody is doing address resolution -> ignore
298 * unspec dup address detection
299 *
300 * The processing is defined in RFC 2462.
301 */
302 if (tentative) {
303 /*
304 * If source address is unspecified address, it is for
305 * duplicate address detection.
306 *
307 * If not, the packet is for addess resolution;
308 * silently ignore it.
309 */
310 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
311 struct sockaddr_dl sdl, *sdlp;
312
313 if (lladdr != NULL)
314 sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
315 ifp->if_index, ifp->if_type,
316 NULL, 0, lladdr, lladdrlen);
317 else
318 sdlp = NULL;
319 nd6_dad_input(ifa, ndopts.nd_opts_nonce, sdlp);
320 }
321 goto freeit;
322 }
323
324 /*
325 * It looks that sender is performing DAD.
326 * Check that the nonce is not being used by the same address
327 * on another interface.
328 */
329 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6) && ndopts.nd_opts_nonce != NULL) {
330 if (nd6_dad_ownnonce(ifa, ndopts.nd_opts_nonce))
331 goto freeit;
332 }
333
334 ifa_release(ifa, &psref_ia);
335 ifa = NULL;
336
337 /*
338 * If the source address is unspecified address, entries must not
339 * be created or updated.
340 * It looks that sender is performing DAD. Output NA toward
341 * all-node multicast address, to tell the sender that I'm using
342 * the address.
343 * S bit ("solicited") must be zero.
344 */
345 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
346 struct in6_addr in6_all;
347
348 in6_all = in6addr_linklocal_allnodes;
349 if (in6_setscope(&in6_all, ifp, NULL) != 0)
350 goto bad;
351 nd6_na_output(ifp, &in6_all, &taddr6,
352 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
353 (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
354 tlladdr, (const struct sockaddr *)proxydl);
355 goto freeit;
356 }
357
358 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_NEIGHBOR_SOLICIT, 0);
359
360 nd6_na_output(ifp, &saddr6, &taddr6,
361 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
362 (router ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
363 tlladdr, (const struct sockaddr *)proxydl);
364 freeit:
365 ifa_release(ifa, &psref_ia);
366 m_put_rcvif_psref(ifp, &psref);
367 m_freem(m);
368 return;
369
370 bad:
371 nd6log(LOG_ERR, "src=%s\n", IN6_PRINT(ip6buf, &saddr6));
372 nd6log(LOG_ERR, "dst=%s\n", IN6_PRINT(ip6buf, &daddr6));
373 nd6log(LOG_ERR, "tgt=%s\n", IN6_PRINT(ip6buf, &taddr6));
374 ICMP6_STATINC(ICMP6_STAT_BADNS);
375 ifa_release(ifa, &psref_ia);
376 m_put_rcvif_psref(ifp, &psref);
377 m_freem(m);
378 }
379
380 /*
381 * Output a Neighbor Solicitation Message. Caller specifies:
382 * - ICMP6 header source IP6 address
383 * - ND6 header target IP6 address
384 * - ND6 header source datalink address
385 *
386 * Based on RFC 2461
387 * Based on RFC 2462 (duplicate address detection)
388 */
389 void
390 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
391 const struct in6_addr *taddr6,
392 struct in6_addr *hsrc,
393 uint8_t *nonce /* duplicate address detection */)
394 {
395 struct mbuf *m;
396 struct ip6_hdr *ip6;
397 struct nd_neighbor_solicit *nd_ns;
398 struct in6_addr *src, src_in;
399 struct ip6_moptions im6o;
400 int icmp6len;
401 int maxlen;
402 const void *mac;
403 struct route ro;
404
405 if (IN6_IS_ADDR_MULTICAST(taddr6))
406 return;
407
408 memset(&ro, 0, sizeof(ro));
409
410 /* estimate the size of message */
411 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
412 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
413 KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
414 "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
415 max_linkhdr, maxlen, MCLBYTES);
416
417 MGETHDR(m, M_DONTWAIT, MT_DATA);
418 if (m && max_linkhdr + maxlen >= MHLEN) {
419 MCLGET(m, M_DONTWAIT);
420 if ((m->m_flags & M_EXT) == 0) {
421 m_free(m);
422 m = NULL;
423 }
424 }
425 if (m == NULL)
426 return;
427 m_reset_rcvif(m);
428
429 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
430 m->m_flags |= M_MCAST;
431 im6o.im6o_multicast_if_index = if_get_index(ifp);
432 im6o.im6o_multicast_hlim = 255;
433 im6o.im6o_multicast_loop = 0;
434 }
435
436 icmp6len = sizeof(*nd_ns);
437 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
438 m->m_data += max_linkhdr; /* or m_align() equivalent? */
439
440 /* fill neighbor solicitation packet */
441 ip6 = mtod(m, struct ip6_hdr *);
442 ip6->ip6_flow = 0;
443 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
444 ip6->ip6_vfc |= IPV6_VERSION;
445 /* ip6->ip6_plen will be set later */
446 ip6->ip6_nxt = IPPROTO_ICMPV6;
447 ip6->ip6_hlim = 255;
448 if (daddr6)
449 ip6->ip6_dst = *daddr6;
450 else {
451 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
452 ip6->ip6_dst.s6_addr16[1] = 0;
453 ip6->ip6_dst.s6_addr32[1] = 0;
454 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
455 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
456 ip6->ip6_dst.s6_addr8[12] = 0xff;
457 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
458 goto bad;
459 }
460 if (nonce == NULL) {
461 int s;
462 /*
463 * RFC2461 7.2.2:
464 * "If the source address of the packet prompting the
465 * solicitation is the same as one of the addresses assigned
466 * to the outgoing interface, that address SHOULD be placed
467 * in the IP Source Address of the outgoing solicitation.
468 * Otherwise, any one of the addresses assigned to the
469 * interface should be used."
470 *
471 * We use the source address for the prompting packet
472 * (hsrc), if:
473 * - hsrc is given from the caller (by giving "ln"), and
474 * - hsrc belongs to the outgoing interface.
475 * Otherwise, we perform the source address selection as usual.
476 */
477 s = pserialize_read_enter();
478 if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc)) {
479 pserialize_read_exit(s);
480 src = hsrc;
481 } else {
482 int error;
483 struct sockaddr_in6 dst_sa;
484
485 pserialize_read_exit(s);
486
487 sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
488
489 error = in6_selectsrc(&dst_sa, NULL,
490 NULL, &ro, NULL, NULL, NULL, &src_in);
491 if (error != 0) {
492 char ip6buf[INET6_ADDRSTRLEN];
493 nd6log(LOG_DEBUG, "source can't be "
494 "determined: dst=%s, error=%d\n",
495 IN6_PRINT(ip6buf, &dst_sa.sin6_addr),
496 error);
497 goto bad;
498 }
499 src = &src_in;
500 }
501 } else {
502 /*
503 * Source address for DAD packet must always be IPv6
504 * unspecified address. (0::0)
505 * We actually don't have to 0-clear the address (we did it
506 * above), but we do so here explicitly to make the intention
507 * clearer.
508 */
509 memset(&src_in, 0, sizeof(src_in));
510 src = &src_in;
511 }
512 ip6->ip6_src = *src;
513 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
514 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
515 nd_ns->nd_ns_code = 0;
516 nd_ns->nd_ns_reserved = 0;
517 nd_ns->nd_ns_target = *taddr6;
518 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
519
520 /*
521 * Add source link-layer address option.
522 *
523 * spec implementation
524 * --- ---
525 * DAD packet MUST NOT do not add the option
526 * there's no link layer address:
527 * impossible do not add the option
528 * there's link layer address:
529 * Multicast NS MUST add one add the option
530 * Unicast NS SHOULD add one add the option
531 */
532 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
533 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
534 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
535 /* 8 byte alignments... */
536 optlen = (optlen + 7) & ~7;
537
538 m->m_pkthdr.len += optlen;
539 m->m_len += optlen;
540 icmp6len += optlen;
541 memset((void *)nd_opt, 0, optlen);
542 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
543 nd_opt->nd_opt_len = optlen >> 3;
544 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
545 }
546
547 /* Add a nonce option (RFC 3971) to detect looped back NS messages.
548 * This behavior is documented in RFC 7527. */
549 if (nonce != NULL) {
550 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
551 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
552
553 /* 8-byte alignment is required. */
554 optlen = (optlen + 7) & ~7;
555 m->m_pkthdr.len += optlen;
556 m->m_len += optlen;
557 icmp6len += optlen;
558 memset(nd_opt, 0, optlen);
559 nd_opt->nd_opt_type = ND_OPT_NONCE;
560 nd_opt->nd_opt_len = optlen >> 3;
561 memcpy(nd_opt + 1, nonce, ND_OPT_NONCE_LEN);
562 }
563
564 ip6->ip6_plen = htons((u_int16_t)icmp6len);
565 nd_ns->nd_ns_cksum = 0;
566 nd_ns->nd_ns_cksum =
567 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
568
569 ip6_output(m, NULL, &ro, nonce != NULL ? IPV6_UNSPECSRC : 0,
570 &im6o, NULL, NULL);
571 icmp6_ifstat_inc(ifp, ifs6_out_msg);
572 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
573 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_SOLICIT);
574
575 rtcache_free(&ro);
576 return;
577
578 bad:
579 rtcache_free(&ro);
580 m_freem(m);
581 return;
582 }
583
584 /*
585 * Neighbor advertisement input handling.
586 *
587 * Based on RFC 2461
588 * Based on RFC 2462 (duplicate address detection)
589 *
590 * the following items are not implemented yet:
591 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
592 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
593 */
594 void
595 nd6_na_input(struct mbuf *m, int off, int icmp6len)
596 {
597 struct ifnet *ifp;
598 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
599 struct nd_neighbor_advert *nd_na;
600 struct in6_addr saddr6 = ip6->ip6_src;
601 struct in6_addr daddr6 = ip6->ip6_dst;
602 struct in6_addr taddr6;
603 int flags;
604 int is_router;
605 int is_solicited;
606 int is_override;
607 int rt_cmd;
608 char *lladdr = NULL;
609 int lladdrlen = 0;
610 struct ifaddr *ifa;
611 struct llentry *ln = NULL;
612 union nd_opts ndopts;
613 struct sockaddr_in6 ssin6;
614 bool checklink = false;
615 struct psref psref;
616 struct psref psref_ia;
617 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
618
619 ifp = m_get_rcvif_psref(m, &psref);
620 if (ifp == NULL)
621 goto freeit;
622
623 if (ip6->ip6_hlim != 255) {
624 nd6log(LOG_ERR,
625 "invalid hlim (%d) from %s to %s on %s\n",
626 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
627 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
628 goto bad;
629 }
630
631 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
632 if (nd_na == NULL) {
633 m_put_rcvif_psref(ifp, &psref);
634 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
635 return;
636 }
637
638 flags = nd_na->nd_na_flags_reserved;
639 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
640 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
641 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
642
643 taddr6 = nd_na->nd_na_target;
644 if (in6_setscope(&taddr6, ifp, NULL)) {
645 goto bad;
646 }
647
648 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
649 nd6log(LOG_ERR, "invalid target address %s\n",
650 IN6_PRINT(ip6buf, &taddr6));
651 goto bad;
652 }
653 if (is_solicited && IN6_IS_ADDR_MULTICAST(&daddr6)) {
654 nd6log(LOG_ERR, "a solicited adv is multicasted\n");
655 goto bad;
656 }
657
658 icmp6len -= sizeof(*nd_na);
659 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
660 if (nd6_options(&ndopts) < 0) {
661 nd6log(LOG_INFO, "invalid ND option, ignored\n");
662 /* nd6_options have incremented stats */
663 goto freeit;
664 }
665
666 if (ndopts.nd_opts_tgt_lladdr != NULL) {
667 struct ifnet *ifp_ll;
668 struct psref psref_ll;
669
670 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
671 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
672
673 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
674 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
675 "(if %d, NA packet %d)\n", IN6_PRINT(ip6buf, &taddr6),
676 ifp->if_addrlen, lladdrlen - 2);
677 goto bad;
678 }
679
680 ifp_ll = if_get_bylla(lladdr, ifp->if_addrlen, &psref_ll);
681 if (ifp_ll != NULL) {
682 /* it's from me, ignore it. */
683 if_put(ifp_ll, &psref_ll);
684 goto freeit;
685 }
686 }
687
688 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6, &psref_ia);
689
690 /*
691 * Target address matches one of my interface address.
692 *
693 * If my address is tentative, this means that there's somebody
694 * already using the same address as mine. This indicates DAD failure.
695 * This is defined in RFC 2462.
696 *
697 * Otherwise, process as defined in RFC 2461.
698 */
699 if (ifa) {
700 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE) {
701 struct sockaddr_dl sdl, *sdlp;
702
703 if (lladdr != NULL)
704 sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
705 ifp->if_index, ifp->if_type,
706 NULL, 0, lladdr, lladdrlen);
707 else
708 sdlp = NULL;
709 nd6_dad_input(ifa, NULL, sdlp);
710 } else
711 log(LOG_ERR,
712 "nd6_na_input: duplicate IP6 address %s\n",
713 IN6_PRINT(ip6buf, &taddr6));
714 ifa_release(ifa, &psref_ia);
715 ifa = NULL;
716 goto freeit;
717 }
718
719 /*
720 * Make sure the source address is from a neighbor's address.
721 */
722 sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
723 if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
724 nd6log(LOG_INFO, "ND packet from non-neighbor %s on %s\n",
725 IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
726 goto bad;
727 }
728
729 /*
730 * If no neighbor cache entry is found, NA SHOULD silently be
731 * discarded.
732 */
733 ln = nd6_lookup(&taddr6, ifp, true);
734 if (ln == NULL)
735 goto freeit;
736
737 rt_cmd = 0;
738 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
739 /*
740 * If the link-layer has address, and no lladdr option came,
741 * discard the packet.
742 */
743 if (ifp->if_addrlen && !lladdr)
744 goto freeit;
745
746 /*
747 * Record link-layer address, and update the state.
748 */
749 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
750 ln->la_flags |= LLE_VALID;
751 rt_cmd = RTM_ADD;
752 if (is_solicited) {
753 ln->ln_state = ND6_LLINFO_REACHABLE;
754 ln->ln_byhint = 0;
755 if (!ND6_LLINFO_PERMANENT(ln)) {
756 nd6_llinfo_settimer(ln,
757 ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
758 }
759 } else {
760 ln->ln_state = ND6_LLINFO_STALE;
761 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
762 }
763 if ((ln->ln_router = is_router) != 0) {
764 /*
765 * This means a router's state has changed from
766 * non-reachable to probably reachable, and might
767 * affect the status of associated prefixes..
768 */
769 checklink = true;
770 }
771 } else {
772 bool llchange;
773
774 /*
775 * Check if the link-layer address has changed or not.
776 */
777 if (lladdr == NULL)
778 llchange = false;
779 else {
780 if (ln->la_flags & LLE_VALID) {
781 if (memcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
782 llchange = true;
783 else
784 llchange = false;
785 } else
786 llchange = true;
787 }
788 if (llchange)
789 rt_cmd = RTM_CHANGE;
790
791 /*
792 * This is VERY complex. Look at it with care.
793 *
794 * override solicit lladdr llchange action
795 * (L: record lladdr)
796 *
797 * 0 0 n -- (2c)
798 * 0 0 y n (2b) L
799 * 0 0 y y (1) REACHABLE->STALE
800 * 0 1 n -- (2c) *->REACHABLE
801 * 0 1 y n (2b) L *->REACHABLE
802 * 0 1 y y (1) REACHABLE->STALE
803 * 1 0 n -- (2a)
804 * 1 0 y n (2a) L
805 * 1 0 y y (2a) L *->STALE
806 * 1 1 n -- (2a) *->REACHABLE
807 * 1 1 y n (2a) L *->REACHABLE
808 * 1 1 y y (2a) L *->REACHABLE
809 */
810 if (!is_override && lladdr != NULL && llchange) { /* (1) */
811 /*
812 * If state is REACHABLE, make it STALE.
813 * no other updates should be done.
814 */
815 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
816 ln->ln_state = ND6_LLINFO_STALE;
817 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
818 }
819 goto freeit;
820 } else if (is_override /* (2a) */
821 || (!is_override && lladdr != NULL && !llchange) /* (2b) */
822 || lladdr == NULL) { /* (2c) */
823 /*
824 * Update link-local address, if any.
825 */
826 if (lladdr != NULL) {
827 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
828 ln->la_flags |= LLE_VALID;
829 }
830
831 /*
832 * If solicited, make the state REACHABLE.
833 * If not solicited and the link-layer address was
834 * changed, make it STALE.
835 */
836 if (is_solicited) {
837 ln->ln_state = ND6_LLINFO_REACHABLE;
838 ln->ln_byhint = 0;
839 if (!ND6_LLINFO_PERMANENT(ln)) {
840 nd6_llinfo_settimer(ln,
841 ND_IFINFO(ifp)->reachable * hz);
842 }
843 } else {
844 if (lladdr && llchange) {
845 ln->ln_state = ND6_LLINFO_STALE;
846 nd6_llinfo_settimer(ln,
847 nd6_gctimer * hz);
848 }
849 }
850 }
851
852 if (ln->ln_router && !is_router) {
853 /*
854 * The peer dropped the router flag.
855 * Remove the sender from the Default Router List and
856 * update the Destination Cache entries.
857 */
858 const struct in6_addr *in6 = &ln->r_l3addr.addr6;
859 struct nd_defrouter *dr;
860
861 ND6_WLOCK();
862 dr = nd6_defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
863 if (dr)
864 nd6_defrtrlist_del(dr, NULL);
865 else if (!ip6_forwarding) {
866 /*
867 * Even if the neighbor is not in the default
868 * router list, the neighbor may be used
869 * as a next hop for some destinations
870 * (e.g. redirect case). So we must
871 * call nd6_rt_flush explicitly.
872 */
873 nd6_rt_flush(&ip6->ip6_src, ln->lle_tbl->llt_ifp);
874 }
875 ND6_UNLOCK();
876 }
877 ln->ln_router = is_router;
878 }
879 /*
880 * XXX: does this matter?
881 * rt->rt_flags &= ~RTF_REJECT;
882 */
883 ln->ln_asked = 0;
884 nd6_llinfo_release_pkts(ln, ifp);
885
886 if (rt_cmd != 0) {
887 struct sockaddr_in6 sin6;
888
889 sockaddr_in6_init(&sin6, &ln->r_l3addr.addr6, 0, 0, 0);
890 rt_clonedmsg(rt_cmd, sin6tosa(&sin6),
891 (char *)&ln->ll_addr, ln->lle_tbl->llt_ifp);
892 }
893
894 freeit:
895 if (ln != NULL)
896 LLE_WUNLOCK(ln);
897
898 if (checklink) {
899 ND6_WLOCK();
900 nd6_pfxlist_onlink_check();
901 ND6_UNLOCK();
902 }
903
904 m_put_rcvif_psref(ifp, &psref);
905 m_freem(m);
906 return;
907
908 bad:
909 if (ln != NULL)
910 LLE_WUNLOCK(ln);
911
912 ICMP6_STATINC(ICMP6_STAT_BADNA);
913 m_put_rcvif_psref(ifp, &psref);
914 m_freem(m);
915 }
916
917 /*
918 * Neighbor advertisement output handling.
919 *
920 * Based on RFC 2461
921 *
922 * the following items are not implemented yet:
923 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
924 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
925 */
926 void
927 nd6_na_output(
928 struct ifnet *ifp,
929 const struct in6_addr *daddr6_0,
930 const struct in6_addr *taddr6,
931 u_long flags,
932 int tlladdr, /* 1 if include target link-layer address */
933 const struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
934 {
935 struct mbuf *m;
936 struct ip6_hdr *ip6;
937 struct nd_neighbor_advert *nd_na;
938 struct ip6_moptions im6o;
939 struct sockaddr *dst;
940 union {
941 struct sockaddr dst;
942 struct sockaddr_in6 dst6;
943 } u;
944 struct in6_addr daddr6;
945 int icmp6len, maxlen, error;
946 const void *mac;
947 struct route ro;
948
949 mac = NULL;
950 memset(&ro, 0, sizeof(ro));
951
952 daddr6 = *daddr6_0; /* make a local copy for modification */
953
954 /* estimate the size of message */
955 maxlen = sizeof(*ip6) + sizeof(*nd_na);
956 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
957 KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
958 "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
959 max_linkhdr, maxlen, MCLBYTES);
960
961 MGETHDR(m, M_DONTWAIT, MT_DATA);
962 if (m && max_linkhdr + maxlen >= MHLEN) {
963 MCLGET(m, M_DONTWAIT);
964 if ((m->m_flags & M_EXT) == 0) {
965 m_free(m);
966 m = NULL;
967 }
968 }
969 if (m == NULL)
970 return;
971 m_reset_rcvif(m);
972
973 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
974 m->m_flags |= M_MCAST;
975 im6o.im6o_multicast_if_index = if_get_index(ifp);
976 im6o.im6o_multicast_hlim = 255;
977 im6o.im6o_multicast_loop = 0;
978 }
979
980 icmp6len = sizeof(*nd_na);
981 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
982 m->m_data += max_linkhdr; /* or m_align() equivalent? */
983
984 /* fill neighbor advertisement packet */
985 ip6 = mtod(m, struct ip6_hdr *);
986 ip6->ip6_flow = 0;
987 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
988 ip6->ip6_vfc |= IPV6_VERSION;
989 ip6->ip6_nxt = IPPROTO_ICMPV6;
990 ip6->ip6_hlim = 255;
991 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
992 /* reply to DAD */
993 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
994 daddr6.s6_addr16[1] = 0;
995 daddr6.s6_addr32[1] = 0;
996 daddr6.s6_addr32[2] = 0;
997 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
998 if (in6_setscope(&daddr6, ifp, NULL))
999 goto bad;
1000
1001 flags &= ~ND_NA_FLAG_SOLICITED;
1002 }
1003 ip6->ip6_dst = daddr6;
1004 sockaddr_in6_init(&u.dst6, &daddr6, 0, 0, 0);
1005 dst = &u.dst;
1006 if (rtcache_setdst(&ro, dst) != 0)
1007 goto bad;
1008
1009 /*
1010 * Select a source whose scope is the same as that of the dest.
1011 */
1012 error = in6_selectsrc(satosin6(dst), NULL, NULL, &ro, NULL, NULL, NULL,
1013 &ip6->ip6_src);
1014 if (error != 0) {
1015 char ip6buf[INET6_ADDRSTRLEN];
1016 nd6log(LOG_DEBUG, "source can't be "
1017 "determined: dst=%s, error=%d\n",
1018 IN6_PRINT(ip6buf, &satocsin6(dst)->sin6_addr), error);
1019 goto bad;
1020 }
1021 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1022 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1023 nd_na->nd_na_code = 0;
1024 nd_na->nd_na_target = *taddr6;
1025 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1026
1027 /*
1028 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1029 * see nd6_ns_input() for details.
1030 * Basically, if NS packet is sent to unicast/anycast addr,
1031 * target lladdr option SHOULD NOT be included.
1032 */
1033 if (tlladdr) {
1034 /*
1035 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1036 * lladdr in sdl0. If we are not proxying (sending NA for
1037 * my address) use lladdr configured for the interface.
1038 */
1039 if (sdl0 == NULL)
1040 mac = nd6_ifptomac(ifp);
1041 else if (sdl0->sa_family == AF_LINK) {
1042 const struct sockaddr_dl *sdl;
1043 sdl = satocsdl(sdl0);
1044 if (sdl->sdl_alen == ifp->if_addrlen)
1045 mac = CLLADDR(sdl);
1046 }
1047 }
1048 if (tlladdr && mac) {
1049 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1050 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1051
1052 /* roundup to 8 bytes alignment! */
1053 optlen = (optlen + 7) & ~7;
1054
1055 m->m_pkthdr.len += optlen;
1056 m->m_len += optlen;
1057 icmp6len += optlen;
1058 memset((void *)nd_opt, 0, optlen);
1059 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1060 nd_opt->nd_opt_len = optlen >> 3;
1061 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
1062 } else
1063 flags &= ~ND_NA_FLAG_OVERRIDE;
1064
1065 ip6->ip6_plen = htons((u_int16_t)icmp6len);
1066 nd_na->nd_na_flags_reserved = flags;
1067 nd_na->nd_na_cksum = 0;
1068 nd_na->nd_na_cksum =
1069 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1070
1071 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1072
1073 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1074 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1075 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_ADVERT);
1076
1077 rtcache_free(&ro);
1078 return;
1079
1080 bad:
1081 rtcache_free(&ro);
1082 m_freem(m);
1083 return;
1084 }
1085
1086 const void *
1087 nd6_ifptomac(const struct ifnet *ifp)
1088 {
1089 switch (ifp->if_type) {
1090 case IFT_ARCNET:
1091 case IFT_ETHER:
1092 case IFT_FDDI:
1093 case IFT_IEEE1394:
1094 case IFT_PROPVIRTUAL:
1095 case IFT_CARP:
1096 case IFT_L2VLAN:
1097 case IFT_IEEE80211:
1098 return CLLADDR(ifp->if_sadl);
1099 default:
1100 return NULL;
1101 }
1102 }
1103
1104 TAILQ_HEAD(dadq_head, dadq);
1105 struct dadq {
1106 TAILQ_ENTRY(dadq) dad_list;
1107 struct ifaddr *dad_ifa;
1108 int dad_count; /* max NS to send */
1109 int dad_ns_tcount; /* # of trials to send NS */
1110 int dad_ns_ocount; /* NS sent so far */
1111 int dad_ns_lcount; /* looped back NS */
1112 struct callout dad_timer_ch;
1113 #define ND_OPT_NONCE_STORE 3 /* dad_count should not exceed this */
1114 /*
1115 * The default ip6_dad_count is 1 as specified by RFC 4862 and
1116 * practically must users won't exceed this.
1117 * A storage of 3 is defaulted to here, in-case the administrator wants
1118 * to match the equivalent behaviour in our ARP implementation.
1119 * This constraint could be removed by sending the on wire nonce as
1120 * hmac(key, dad_ns_ocount), but that would increase the nonce size
1121 * sent on the wire.
1122 */
1123 uint8_t dad_nonce[ND_OPT_NONCE_STORE][ND_OPT_NONCE_LEN];
1124 };
1125
1126 static struct dadq_head dadq;
1127 static kmutex_t nd6_dad_lock;
1128
1129 void
1130 nd6_nbr_init(void)
1131 {
1132
1133 TAILQ_INIT(&dadq);
1134 mutex_init(&nd6_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1135 }
1136
1137 static struct dadq *
1138 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce, bool *found_nonce)
1139 {
1140 struct in6_addr *myaddr6, *dadaddr6;
1141 bool match_ifa;
1142 struct dadq *dp;
1143 int i, nonce_max;
1144
1145 KASSERT(mutex_owned(&nd6_dad_lock));
1146 KASSERT(ifa != NULL);
1147
1148 myaddr6 = IFA_IN6(ifa);
1149 if (nonce != NULL &&
1150 nonce->nd_opt_nonce_len != (ND_OPT_NONCE_LEN + 2) / 8)
1151 nonce = NULL;
1152 match_ifa = nonce == NULL || found_nonce == NULL || *found_nonce == false;
1153 if (found_nonce != NULL)
1154 *found_nonce = false;
1155
1156 TAILQ_FOREACH(dp, &dadq, dad_list) {
1157 if (match_ifa) {
1158 if (dp->dad_ifa != ifa)
1159 continue;
1160 } else {
1161 dadaddr6 = IFA_IN6(dp->dad_ifa);
1162 if (!IN6_ARE_ADDR_EQUAL(myaddr6, dadaddr6))
1163 continue;
1164 }
1165
1166 if (nonce == NULL)
1167 break;
1168
1169 nonce_max = MIN(dp->dad_ns_ocount, ND_OPT_NONCE_STORE);
1170 for (i = 0; i < nonce_max; i++) {
1171 if (memcmp(nonce->nd_opt_nonce,
1172 dp->dad_nonce[i],
1173 ND_OPT_NONCE_LEN) == 0)
1174 break;
1175 }
1176 if (i < nonce_max) {
1177 char ip6buf[INET6_ADDRSTRLEN];
1178
1179 *found_nonce = true;
1180 log(LOG_DEBUG,
1181 "%s: detected a looped back NS message for %s\n",
1182 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???",
1183 IN6_PRINT(ip6buf, myaddr6));
1184 dp->dad_ns_lcount++;
1185 continue;
1186 }
1187
1188 break;
1189 }
1190 return dp;
1191 }
1192
1193 static bool
1194 nd6_dad_ownnonce(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1195 {
1196 bool found_nonce = true;
1197
1198 mutex_enter(&nd6_dad_lock);
1199 nd6_dad_find(ifa, nonce, &found_nonce);
1200 mutex_exit(&nd6_dad_lock);
1201
1202 return found_nonce;
1203 }
1204
1205 static void
1206 nd6_dad_starttimer(struct dadq *dp, int ticks)
1207 {
1208
1209 callout_reset(&dp->dad_timer_ch, ticks,
1210 (void (*)(void *))nd6_dad_timer, dp);
1211 }
1212
1213 static void
1214 nd6_dad_stoptimer(struct dadq *dp)
1215 {
1216
1217 KASSERT(mutex_owned(&nd6_dad_lock));
1218
1219 TAILQ_REMOVE(&dadq, dp, dad_list);
1220 /* Tell the timer that dp is being destroyed. */
1221 dp->dad_ifa = NULL;
1222 callout_halt(&dp->dad_timer_ch, &nd6_dad_lock);
1223 }
1224
1225 static void
1226 nd6_dad_destroytimer(struct dadq *dp)
1227 {
1228
1229 KASSERT(dp->dad_ifa == NULL);
1230 callout_destroy(&dp->dad_timer_ch);
1231 kmem_intr_free(dp, sizeof(*dp));
1232 }
1233
1234 /*
1235 * Start Duplicate Address Detection (DAD) for specified interface address.
1236 *
1237 * Note that callout is used when xtick > 0 and not when xtick == 0.
1238 *
1239 * xtick: minimum delay ticks for IFF_UP event
1240 */
1241 void
1242 nd6_dad_start(struct ifaddr *ifa, int xtick)
1243 {
1244 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1245 struct dadq *dp;
1246 char ip6buf[INET6_ADDRSTRLEN];
1247
1248 /*
1249 * If we don't need DAD, don't do it.
1250 * There are several cases:
1251 * - DAD is disabled
1252 * - the interface address is anycast
1253 */
1254 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1255 log(LOG_DEBUG,
1256 "nd6_dad_start: called with non-tentative address "
1257 "%s(%s)\n",
1258 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1259 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1260 return;
1261 }
1262 if (ia->ia6_flags & IN6_IFF_ANYCAST || !ip6_dad_enabled()) {
1263 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1264 rt_addrmsg(RTM_NEWADDR, ifa);
1265 return;
1266 }
1267 KASSERT(ifa->ifa_ifp != NULL);
1268 if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1269 return;
1270
1271 dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1272
1273 mutex_enter(&nd6_dad_lock);
1274 if (nd6_dad_find(ifa, NULL, NULL) != NULL) {
1275 mutex_exit(&nd6_dad_lock);
1276 /* DAD already in progress */
1277 if (dp != NULL)
1278 kmem_intr_free(dp, sizeof(*dp));
1279 return;
1280 }
1281
1282 if (dp == NULL) {
1283 mutex_exit(&nd6_dad_lock);
1284 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1285 "%s(%s)\n",
1286 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1287 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1288 return;
1289 }
1290
1291 /*
1292 * Send NS packet for DAD, ip6_dad_count times.
1293 * Note that we must delay the first transmission, if this is the
1294 * first packet to be sent from the interface after interface
1295 * (re)initialization.
1296 */
1297 callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1298 dp->dad_ifa = ifa;
1299 ifaref(ifa); /* just for safety */
1300 dp->dad_count = ip6_dad_count;
1301 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1302 dp->dad_ns_lcount = 0;
1303 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1304
1305 nd6log(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1306 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1307
1308 if (xtick == 0) {
1309 nd6_dad_ns_output(dp, ifa);
1310 nd6_dad_starttimer(dp,
1311 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1312 } else
1313 nd6_dad_starttimer(dp, xtick);
1314 mutex_exit(&nd6_dad_lock);
1315 }
1316
1317 /*
1318 * terminate DAD unconditionally. used for address removals.
1319 */
1320 void
1321 nd6_dad_stop(struct ifaddr *ifa)
1322 {
1323 struct dadq *dp;
1324
1325 mutex_enter(&nd6_dad_lock);
1326 dp = nd6_dad_find(ifa, NULL, NULL);
1327 if (dp == NULL) {
1328 mutex_exit(&nd6_dad_lock);
1329 /* DAD wasn't started yet */
1330 return;
1331 }
1332
1333 /* Prevent the timer from running anymore. */
1334 nd6_dad_stoptimer(dp);
1335
1336 mutex_exit(&nd6_dad_lock);
1337
1338 nd6_dad_destroytimer(dp);
1339 ifafree(ifa);
1340 }
1341
1342 static void
1343 nd6_dad_timer(struct dadq *dp)
1344 {
1345 struct ifaddr *ifa;
1346 struct in6_ifaddr *ia;
1347 char ip6buf[INET6_ADDRSTRLEN];
1348 bool need_free = false;
1349
1350 KERNEL_LOCK_UNLESS_NET_MPSAFE();
1351 mutex_enter(&nd6_dad_lock);
1352
1353 ifa = dp->dad_ifa;
1354 if (ifa == NULL) {
1355 /* dp is being destroyed by someone. Do nothing. */
1356 goto done;
1357 }
1358
1359 ia = (struct in6_ifaddr *)ifa;
1360 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1361 log(LOG_ERR, "nd6_dad_timer: called with duplicate address "
1362 "%s(%s)\n",
1363 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1364 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1365 goto done;
1366 }
1367 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1368 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1369 "%s(%s)\n",
1370 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1371 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1372 goto done;
1373 }
1374
1375 /* timeouted with IFF_{RUNNING,UP} check */
1376 if (dp->dad_ns_tcount > dad_maxtry) {
1377 nd6log(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1378 if_name(ifa->ifa_ifp));
1379
1380 nd6_dad_stoptimer(dp);
1381 need_free = true;
1382 goto done;
1383 }
1384
1385 /* Need more checks? */
1386 if (dp->dad_ns_ocount < dp->dad_count) {
1387 /*
1388 * We have more NS to go. Send NS packet for DAD.
1389 */
1390 nd6_dad_ns_output(dp, ifa);
1391 nd6_dad_starttimer(dp,
1392 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1393 } else {
1394 /*
1395 * We are done with DAD. No NA came, no NS came.
1396 * No duplicate address found.
1397 */
1398 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1399 rt_addrmsg(RTM_NEWADDR, ifa);
1400
1401 nd6log(LOG_DEBUG,
1402 "%s: DAD complete for %s - no duplicates found\n",
1403 if_name(ifa->ifa_ifp),
1404 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1405
1406 nd6_dad_stoptimer(dp);
1407 need_free = true;
1408 }
1409 done:
1410 mutex_exit(&nd6_dad_lock);
1411
1412 if (need_free) {
1413 nd6_dad_destroytimer(dp);
1414 KASSERT(ifa != NULL);
1415 ifafree(ifa);
1416 }
1417
1418 KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1419 }
1420
1421 static void
1422 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp,
1423 const struct sockaddr_dl *from)
1424 {
1425 struct in6_ifaddr *ia;
1426 struct ifnet *ifp;
1427 char ip6buf[INET6_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN], *llastr;
1428
1429 KASSERT(mutex_owned(&nd6_dad_lock));
1430 KASSERT(ifa != NULL);
1431
1432 ifp = ifa->ifa_ifp;
1433 ia = (struct in6_ifaddr *)ifa;
1434
1435 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1436 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1437
1438 if (__predict_false(from == NULL))
1439 llastr = NULL;
1440 else
1441 llastr = lla_snprintf(llabuf, sizeof(llabuf),
1442 CLLADDR(from), from->sdl_alen);
1443
1444 log(LOG_ERR, "%s: DAD duplicate address %s from %s\n",
1445 if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr), llastr);
1446
1447 /* Inform the routing socket that DAD has completed */
1448 rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1449
1450 /*
1451 * If the address is a link-local address formed from an interface
1452 * identifier based on the hardware address which is supposed to be
1453 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1454 * operation on the interface SHOULD be disabled.
1455 * [rfc2462bis-03 Section 5.4.5]
1456 */
1457 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1458 struct in6_addr in6;
1459
1460 /*
1461 * To avoid over-reaction, we only apply this logic when we are
1462 * very sure that hardware addresses are supposed to be unique.
1463 */
1464 switch (ifp->if_type) {
1465 case IFT_ETHER:
1466 case IFT_FDDI:
1467 case IFT_ATM:
1468 case IFT_IEEE1394:
1469 case IFT_IEEE80211:
1470 in6 = ia->ia_addr.sin6_addr;
1471 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1472 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1473 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1474 log(LOG_ERR, "%s: possible hardware address "
1475 "duplication detected, disable IPv6\n",
1476 if_name(ifp));
1477 }
1478 break;
1479 }
1480 }
1481 }
1482
1483 static void
1484 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1485 {
1486 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1487 struct ifnet *ifp = ifa->ifa_ifp;
1488 uint8_t *nonce;
1489
1490 dp->dad_ns_tcount++;
1491 if ((ifp->if_flags & IFF_UP) == 0) {
1492 #if 0
1493 printf("%s: interface down?\n", if_name(ifp));
1494 #endif
1495 return;
1496 }
1497 if ((ifp->if_flags & IFF_RUNNING) == 0) {
1498 #if 0
1499 printf("%s: interface not running?\n", if_name(ifp));
1500 #endif
1501 return;
1502 }
1503
1504 dp->dad_ns_tcount = 0;
1505 nonce = dp->dad_nonce[dp->dad_ns_ocount % ND_OPT_NONCE_STORE];
1506 cprng_fast(nonce, ND_OPT_NONCE_LEN);
1507 dp->dad_ns_ocount++;
1508
1509 nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, nonce);
1510 }
1511
1512 static void
1513 nd6_dad_input(struct ifaddr *ifa, struct nd_opt_nonce *nonce,
1514 const struct sockaddr_dl *from)
1515 {
1516 struct dadq *dp;
1517 bool found_nonce = false;
1518
1519 KASSERT(ifa != NULL);
1520
1521 mutex_enter(&nd6_dad_lock);
1522 dp = nd6_dad_find(ifa, nonce, &found_nonce);
1523 if (!found_nonce) {
1524 nd6_dad_duplicated(ifa, dp, from);
1525 if (dp != NULL)
1526 nd6_dad_stoptimer(dp);
1527 }
1528 mutex_exit(&nd6_dad_lock);
1529 if (dp != NULL) {
1530 nd6_dad_destroytimer(dp);
1531 ifafree(ifa);
1532 }
1533 }
1534