nd6_nbr.c revision 1.170 1 /* $NetBSD: nd6_nbr.c,v 1.170 2019/08/29 16:26:43 roy 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 1.170 2019/08/29 16:26:43 roy 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 pserialize_read_exit(s);
498 goto bad;
499 }
500 src = &src_in;
501 }
502 } else {
503 /*
504 * Source address for DAD packet must always be IPv6
505 * unspecified address. (0::0)
506 * We actually don't have to 0-clear the address (we did it
507 * above), but we do so here explicitly to make the intention
508 * clearer.
509 */
510 memset(&src_in, 0, sizeof(src_in));
511 src = &src_in;
512 }
513 ip6->ip6_src = *src;
514 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
515 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
516 nd_ns->nd_ns_code = 0;
517 nd_ns->nd_ns_reserved = 0;
518 nd_ns->nd_ns_target = *taddr6;
519 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
520
521 /*
522 * Add source link-layer address option.
523 *
524 * spec implementation
525 * --- ---
526 * DAD packet MUST NOT do not add the option
527 * there's no link layer address:
528 * impossible do not add the option
529 * there's link layer address:
530 * Multicast NS MUST add one add the option
531 * Unicast NS SHOULD add one add the option
532 */
533 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
534 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
535 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
536 /* 8 byte alignments... */
537 optlen = (optlen + 7) & ~7;
538
539 m->m_pkthdr.len += optlen;
540 m->m_len += optlen;
541 icmp6len += optlen;
542 memset((void *)nd_opt, 0, optlen);
543 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
544 nd_opt->nd_opt_len = optlen >> 3;
545 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
546 }
547
548 /* Add a nonce option (RFC 3971) to detect looped back NS messages.
549 * This behavior is documented in RFC 7527. */
550 if (nonce != NULL) {
551 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
552 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
553
554 /* 8-byte alignment is required. */
555 optlen = (optlen + 7) & ~7;
556 m->m_pkthdr.len += optlen;
557 m->m_len += optlen;
558 icmp6len += optlen;
559 memset(nd_opt, 0, optlen);
560 nd_opt->nd_opt_type = ND_OPT_NONCE;
561 nd_opt->nd_opt_len = optlen >> 3;
562 memcpy(nd_opt + 1, nonce, ND_OPT_NONCE_LEN);
563 }
564
565 ip6->ip6_plen = htons((u_int16_t)icmp6len);
566 nd_ns->nd_ns_cksum = 0;
567 nd_ns->nd_ns_cksum =
568 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
569
570 ip6_output(m, NULL, &ro, nonce != NULL ? IPV6_UNSPECSRC : 0,
571 &im6o, NULL, NULL);
572 icmp6_ifstat_inc(ifp, ifs6_out_msg);
573 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
574 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_SOLICIT);
575
576 rtcache_free(&ro);
577 return;
578
579 bad:
580 rtcache_free(&ro);
581 m_freem(m);
582 return;
583 }
584
585 /*
586 * Neighbor advertisement input handling.
587 *
588 * Based on RFC 2461
589 * Based on RFC 2462 (duplicate address detection)
590 *
591 * the following items are not implemented yet:
592 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
593 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
594 */
595 void
596 nd6_na_input(struct mbuf *m, int off, int icmp6len)
597 {
598 struct ifnet *ifp;
599 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
600 struct nd_neighbor_advert *nd_na;
601 struct in6_addr saddr6 = ip6->ip6_src;
602 struct in6_addr daddr6 = ip6->ip6_dst;
603 struct in6_addr taddr6;
604 int flags;
605 int is_router;
606 int is_solicited;
607 int is_override;
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 rt_announce;
615 bool checklink = false;
616 struct psref psref;
617 struct psref psref_ia;
618 char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
619
620 ifp = m_get_rcvif_psref(m, &psref);
621 if (ifp == NULL)
622 goto freeit;
623
624 if (ip6->ip6_hlim != 255) {
625 nd6log(LOG_ERR,
626 "invalid hlim (%d) from %s to %s on %s\n",
627 ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
628 IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
629 goto bad;
630 }
631
632 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
633 if (nd_na == NULL) {
634 m_put_rcvif_psref(ifp, &psref);
635 ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
636 return;
637 }
638
639 flags = nd_na->nd_na_flags_reserved;
640 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
641 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
642 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
643
644 taddr6 = nd_na->nd_na_target;
645 if (in6_setscope(&taddr6, ifp, NULL)) {
646 goto bad;
647 }
648
649 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
650 nd6log(LOG_ERR, "invalid target address %s\n",
651 IN6_PRINT(ip6buf, &taddr6));
652 goto bad;
653 }
654 if (is_solicited && IN6_IS_ADDR_MULTICAST(&daddr6)) {
655 nd6log(LOG_ERR, "a solicited adv is multicasted\n");
656 goto bad;
657 }
658
659 icmp6len -= sizeof(*nd_na);
660 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
661 if (nd6_options(&ndopts) < 0) {
662 nd6log(LOG_INFO, "invalid ND option, ignored\n");
663 /* nd6_options have incremented stats */
664 goto freeit;
665 }
666
667 if (ndopts.nd_opts_tgt_lladdr != NULL) {
668 struct ifnet *ifp_ll;
669 struct psref psref_ll;
670
671 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
672 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
673
674 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
675 nd6log(LOG_INFO, "lladdrlen mismatch for %s "
676 "(if %d, NA packet %d)\n", IN6_PRINT(ip6buf, &taddr6),
677 ifp->if_addrlen, lladdrlen - 2);
678 goto bad;
679 }
680
681 ifp_ll = if_get_bylla(lladdr, ifp->if_addrlen, &psref_ll);
682 if (ifp_ll != NULL) {
683 /* it's from me, ignore it. */
684 if_put(ifp_ll, &psref_ll);
685 goto freeit;
686 }
687 }
688
689 ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6, &psref_ia);
690
691 /*
692 * Target address matches one of my interface address.
693 *
694 * If my address is tentative, this means that there's somebody
695 * already using the same address as mine. This indicates DAD failure.
696 * This is defined in RFC 2462.
697 *
698 * Otherwise, process as defined in RFC 2461.
699 */
700 if (ifa) {
701 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE) {
702 struct sockaddr_dl sdl, *sdlp;
703
704 if (lladdr != NULL)
705 sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
706 ifp->if_index, ifp->if_type,
707 NULL, 0, lladdr, lladdrlen);
708 else
709 sdlp = NULL;
710 nd6_dad_input(ifa, NULL, sdlp);
711 } else
712 log(LOG_ERR,
713 "nd6_na_input: duplicate IP6 address %s\n",
714 IN6_PRINT(ip6buf, &taddr6));
715 ifa_release(ifa, &psref_ia);
716 ifa = NULL;
717 goto freeit;
718 }
719
720 /*
721 * Make sure the source address is from a neighbor's address.
722 */
723 sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
724 if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
725 nd6log(LOG_INFO, "ND packet from non-neighbor %s on %s\n",
726 IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
727 goto bad;
728 }
729
730 /*
731 * If no neighbor cache entry is found, NA SHOULD silently be
732 * discarded.
733 */
734 ln = nd6_lookup(&taddr6, ifp, true);
735 if (ln == NULL)
736 goto freeit;
737
738 rt_announce = false;
739 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
740 /*
741 * If the link-layer has address, and no lladdr option came,
742 * discard the packet.
743 */
744 if (ifp->if_addrlen && !lladdr)
745 goto freeit;
746
747 /*
748 * Record link-layer address, and update the state.
749 */
750 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
751 ln->la_flags |= LLE_VALID;
752 rt_announce = true;
753 if (is_solicited) {
754 ln->ln_state = ND6_LLINFO_REACHABLE;
755 ln->ln_byhint = 0;
756 if (!ND6_LLINFO_PERMANENT(ln)) {
757 nd6_llinfo_settimer(ln,
758 ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
759 }
760 } else {
761 ln->ln_state = ND6_LLINFO_STALE;
762 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
763 }
764 if ((ln->ln_router = is_router) != 0) {
765 /*
766 * This means a router's state has changed from
767 * non-reachable to probably reachable, and might
768 * affect the status of associated prefixes..
769 */
770 checklink = true;
771 }
772 } else {
773 bool llchange;
774
775 /*
776 * Check if the link-layer address has changed or not.
777 */
778 if (lladdr == NULL)
779 llchange = false;
780 else {
781 if (ln->la_flags & LLE_VALID) {
782 if (memcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
783 llchange = rt_announce = true;
784 else
785 llchange = false;
786 } else
787 llchange = rt_announce = true;
788 }
789
790 /*
791 * This is VERY complex. Look at it with care.
792 *
793 * override solicit lladdr llchange action
794 * (L: record lladdr)
795 *
796 * 0 0 n -- (2c)
797 * 0 0 y n (2b) L
798 * 0 0 y y (1) REACHABLE->STALE
799 * 0 1 n -- (2c) *->REACHABLE
800 * 0 1 y n (2b) L *->REACHABLE
801 * 0 1 y y (1) REACHABLE->STALE
802 * 1 0 n -- (2a)
803 * 1 0 y n (2a) L
804 * 1 0 y y (2a) L *->STALE
805 * 1 1 n -- (2a) *->REACHABLE
806 * 1 1 y n (2a) L *->REACHABLE
807 * 1 1 y y (2a) L *->REACHABLE
808 */
809 if (!is_override && lladdr != NULL && llchange) { /* (1) */
810 /*
811 * If state is REACHABLE, make it STALE.
812 * no other updates should be done.
813 */
814 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
815 ln->ln_state = ND6_LLINFO_STALE;
816 nd6_llinfo_settimer(ln, nd6_gctimer * hz);
817 }
818 goto freeit;
819 } else if (is_override /* (2a) */
820 || (!is_override && lladdr != NULL && !llchange) /* (2b) */
821 || lladdr == NULL) { /* (2c) */
822 /*
823 * Update link-local address, if any.
824 */
825 if (lladdr != NULL) {
826 memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
827 ln->la_flags |= LLE_VALID;
828 }
829
830 /*
831 * If solicited, make the state REACHABLE.
832 * If not solicited and the link-layer address was
833 * changed, make it STALE.
834 */
835 if (is_solicited) {
836 ln->ln_state = ND6_LLINFO_REACHABLE;
837 ln->ln_byhint = 0;
838 if (!ND6_LLINFO_PERMANENT(ln)) {
839 nd6_llinfo_settimer(ln,
840 ND_IFINFO(ifp)->reachable * hz);
841 }
842 } else {
843 if (lladdr && llchange) {
844 ln->ln_state = ND6_LLINFO_STALE;
845 nd6_llinfo_settimer(ln,
846 nd6_gctimer * hz);
847 }
848 }
849 }
850
851 if (ln->ln_router && !is_router) {
852 /*
853 * The peer dropped the router flag.
854 * Remove the sender from the Default Router List and
855 * update the Destination Cache entries.
856 */
857 const struct in6_addr *in6 = &ln->r_l3addr.addr6;
858 struct nd_defrouter *dr;
859 struct sockaddr_in6 sin6;
860
861 /*
862 * Userland really has no business with NA messages.
863 * However, RFC 4861 6.2.5 only says departing routers
864 * *SHOULD* send RA with lifetime of zero and *MUST*
865 * send all subsequent NA messages if the router flag
866 * unset.
867 *
868 * To help userland avoid the expensive process of
869 * parsing NA messages, send RTM_CHANGE without a
870 * lladdr in the gateway.
871 * This is different from the intial RTM_ADD also
872 * without a lladdr in the gateway and RTM_DELETE.
873 */
874 sockaddr_in6_init(&sin6, in6, 0, 0, 0);
875 rt_clonedmsg(RTM_CHANGE, sin6tosa(&sin6), NULL,
876 ln->lle_tbl->llt_ifp);
877 rt_announce = true;
878
879 ND6_WLOCK();
880 dr = nd6_defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
881 if (dr)
882 nd6_defrtrlist_del(dr, NULL);
883 else if (!ip6_forwarding) {
884 /*
885 * Even if the neighbor is not in the default
886 * router list, the neighbor may be used
887 * as a next hop for some destinations
888 * (e.g. redirect case). So we must
889 * call nd6_rt_flush explicitly.
890 */
891 nd6_rt_flush(&ip6->ip6_src, ln->lle_tbl->llt_ifp);
892 }
893 ND6_UNLOCK();
894 }
895 ln->ln_router = is_router;
896 }
897 /*
898 * XXX: does this matter?
899 * rt->rt_flags &= ~RTF_REJECT;
900 */
901 ln->ln_asked = 0;
902 nd6_llinfo_release_pkts(ln, ifp);
903
904 if (rt_announce) {
905 struct sockaddr_in6 sin6;
906
907 sockaddr_in6_init(&sin6, &ln->r_l3addr.addr6, 0, 0, 0);
908 rt_clonedmsg(RTM_CHANGE, sin6tosa(&sin6),
909 (char *)&ln->ll_addr, ln->lle_tbl->llt_ifp);
910 }
911
912 freeit:
913 if (ln != NULL)
914 LLE_WUNLOCK(ln);
915
916 if (checklink) {
917 ND6_WLOCK();
918 nd6_pfxlist_onlink_check();
919 ND6_UNLOCK();
920 }
921
922 m_put_rcvif_psref(ifp, &psref);
923 m_freem(m);
924 return;
925
926 bad:
927 if (ln != NULL)
928 LLE_WUNLOCK(ln);
929
930 ICMP6_STATINC(ICMP6_STAT_BADNA);
931 m_put_rcvif_psref(ifp, &psref);
932 m_freem(m);
933 }
934
935 /*
936 * Neighbor advertisement output handling.
937 *
938 * Based on RFC 2461
939 *
940 * the following items are not implemented yet:
941 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
942 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
943 */
944 void
945 nd6_na_output(
946 struct ifnet *ifp,
947 const struct in6_addr *daddr6_0,
948 const struct in6_addr *taddr6,
949 u_long flags,
950 int tlladdr, /* 1 if include target link-layer address */
951 const struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
952 {
953 struct mbuf *m;
954 struct ip6_hdr *ip6;
955 struct nd_neighbor_advert *nd_na;
956 struct ip6_moptions im6o;
957 struct sockaddr *dst;
958 union {
959 struct sockaddr dst;
960 struct sockaddr_in6 dst6;
961 } u;
962 struct in6_addr daddr6;
963 int icmp6len, maxlen, error;
964 const void *mac;
965 struct route ro;
966
967 mac = NULL;
968 memset(&ro, 0, sizeof(ro));
969
970 daddr6 = *daddr6_0; /* make a local copy for modification */
971
972 /* estimate the size of message */
973 maxlen = sizeof(*ip6) + sizeof(*nd_na);
974 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
975 KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
976 "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
977 max_linkhdr, maxlen, MCLBYTES);
978
979 MGETHDR(m, M_DONTWAIT, MT_DATA);
980 if (m && max_linkhdr + maxlen >= MHLEN) {
981 MCLGET(m, M_DONTWAIT);
982 if ((m->m_flags & M_EXT) == 0) {
983 m_free(m);
984 m = NULL;
985 }
986 }
987 if (m == NULL)
988 return;
989 m_reset_rcvif(m);
990
991 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
992 m->m_flags |= M_MCAST;
993 im6o.im6o_multicast_if_index = if_get_index(ifp);
994 im6o.im6o_multicast_hlim = 255;
995 im6o.im6o_multicast_loop = 0;
996 }
997
998 icmp6len = sizeof(*nd_na);
999 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1000 m->m_data += max_linkhdr; /* or m_align() equivalent? */
1001
1002 /* fill neighbor advertisement packet */
1003 ip6 = mtod(m, struct ip6_hdr *);
1004 ip6->ip6_flow = 0;
1005 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1006 ip6->ip6_vfc |= IPV6_VERSION;
1007 ip6->ip6_nxt = IPPROTO_ICMPV6;
1008 ip6->ip6_hlim = 255;
1009 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1010 /* reply to DAD */
1011 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1012 daddr6.s6_addr16[1] = 0;
1013 daddr6.s6_addr32[1] = 0;
1014 daddr6.s6_addr32[2] = 0;
1015 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1016 if (in6_setscope(&daddr6, ifp, NULL))
1017 goto bad;
1018
1019 flags &= ~ND_NA_FLAG_SOLICITED;
1020 }
1021 ip6->ip6_dst = daddr6;
1022 sockaddr_in6_init(&u.dst6, &daddr6, 0, 0, 0);
1023 dst = &u.dst;
1024 if (rtcache_setdst(&ro, dst) != 0)
1025 goto bad;
1026
1027 /*
1028 * Select a source whose scope is the same as that of the dest.
1029 */
1030 error = in6_selectsrc(satosin6(dst), NULL, NULL, &ro, NULL, NULL, NULL,
1031 &ip6->ip6_src);
1032 if (error != 0) {
1033 char ip6buf[INET6_ADDRSTRLEN];
1034 nd6log(LOG_DEBUG, "source can't be "
1035 "determined: dst=%s, error=%d\n",
1036 IN6_PRINT(ip6buf, &satocsin6(dst)->sin6_addr), error);
1037 goto bad;
1038 }
1039 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1040 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1041 nd_na->nd_na_code = 0;
1042 nd_na->nd_na_target = *taddr6;
1043 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1044
1045 /*
1046 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1047 * see nd6_ns_input() for details.
1048 * Basically, if NS packet is sent to unicast/anycast addr,
1049 * target lladdr option SHOULD NOT be included.
1050 */
1051 if (tlladdr) {
1052 /*
1053 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1054 * lladdr in sdl0. If we are not proxying (sending NA for
1055 * my address) use lladdr configured for the interface.
1056 */
1057 if (sdl0 == NULL)
1058 mac = nd6_ifptomac(ifp);
1059 else if (sdl0->sa_family == AF_LINK) {
1060 const struct sockaddr_dl *sdl;
1061 sdl = satocsdl(sdl0);
1062 if (sdl->sdl_alen == ifp->if_addrlen)
1063 mac = CLLADDR(sdl);
1064 }
1065 }
1066 if (tlladdr && mac) {
1067 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1068 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1069
1070 /* roundup to 8 bytes alignment! */
1071 optlen = (optlen + 7) & ~7;
1072
1073 m->m_pkthdr.len += optlen;
1074 m->m_len += optlen;
1075 icmp6len += optlen;
1076 memset((void *)nd_opt, 0, optlen);
1077 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1078 nd_opt->nd_opt_len = optlen >> 3;
1079 memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
1080 } else
1081 flags &= ~ND_NA_FLAG_OVERRIDE;
1082
1083 ip6->ip6_plen = htons((u_int16_t)icmp6len);
1084 nd_na->nd_na_flags_reserved = flags;
1085 nd_na->nd_na_cksum = 0;
1086 nd_na->nd_na_cksum =
1087 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1088
1089 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1090
1091 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1092 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1093 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_ADVERT);
1094
1095 rtcache_free(&ro);
1096 return;
1097
1098 bad:
1099 rtcache_free(&ro);
1100 m_freem(m);
1101 return;
1102 }
1103
1104 const void *
1105 nd6_ifptomac(const struct ifnet *ifp)
1106 {
1107 switch (ifp->if_type) {
1108 case IFT_ARCNET:
1109 case IFT_ETHER:
1110 case IFT_FDDI:
1111 case IFT_IEEE1394:
1112 case IFT_PROPVIRTUAL:
1113 case IFT_CARP:
1114 case IFT_L2VLAN:
1115 case IFT_IEEE80211:
1116 return CLLADDR(ifp->if_sadl);
1117 default:
1118 return NULL;
1119 }
1120 }
1121
1122 TAILQ_HEAD(dadq_head, dadq);
1123 struct dadq {
1124 TAILQ_ENTRY(dadq) dad_list;
1125 struct ifaddr *dad_ifa;
1126 int dad_count; /* max NS to send */
1127 int dad_ns_tcount; /* # of trials to send NS */
1128 int dad_ns_ocount; /* NS sent so far */
1129 int dad_ns_lcount; /* looped back NS */
1130 struct callout dad_timer_ch;
1131 #define ND_OPT_NONCE_STORE 3 /* dad_count should not exceed this */
1132 /*
1133 * The default ip6_dad_count is 1 as specified by RFC 4862 and
1134 * practically must users won't exceed this.
1135 * A storage of 3 is defaulted to here, in-case the administrator wants
1136 * to match the equivalent behaviour in our ARP implementation.
1137 * This constraint could be removed by sending the on wire nonce as
1138 * hmac(key, dad_ns_ocount), but that would increase the nonce size
1139 * sent on the wire.
1140 */
1141 uint8_t dad_nonce[ND_OPT_NONCE_STORE][ND_OPT_NONCE_LEN];
1142 };
1143
1144 static struct dadq_head dadq;
1145 static int dad_init = 0;
1146 static kmutex_t nd6_dad_lock;
1147
1148 static struct dadq *
1149 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce, bool *found_nonce)
1150 {
1151 struct in6_addr *myaddr6, *dadaddr6;
1152 bool match_ifa;
1153 struct dadq *dp;
1154 int i, nonce_max;
1155
1156 KASSERT(mutex_owned(&nd6_dad_lock));
1157 KASSERT(ifa != NULL);
1158
1159 myaddr6 = IFA_IN6(ifa);
1160 if (nonce != NULL &&
1161 nonce->nd_opt_nonce_len != (ND_OPT_NONCE_LEN + 2) / 8)
1162 nonce = NULL;
1163 match_ifa = nonce == NULL || found_nonce == NULL || *found_nonce == false;
1164 if (found_nonce != NULL)
1165 *found_nonce = false;
1166
1167 TAILQ_FOREACH(dp, &dadq, dad_list) {
1168 if (match_ifa) {
1169 if (dp->dad_ifa != ifa)
1170 continue;
1171 } else {
1172 dadaddr6 = IFA_IN6(dp->dad_ifa);
1173 if (!IN6_ARE_ADDR_EQUAL(myaddr6, dadaddr6))
1174 continue;
1175 }
1176
1177 if (nonce == NULL)
1178 break;
1179
1180 nonce_max = MIN(dp->dad_ns_ocount, ND_OPT_NONCE_STORE);
1181 for (i = 0; i < nonce_max; i++) {
1182 if (memcmp(nonce->nd_opt_nonce,
1183 dp->dad_nonce[i],
1184 ND_OPT_NONCE_LEN) == 0)
1185 break;
1186 }
1187 if (i < nonce_max) {
1188 char ip6buf[INET6_ADDRSTRLEN];
1189
1190 *found_nonce = true;
1191 log(LOG_DEBUG,
1192 "%s: detected a looped back NS message for %s\n",
1193 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???",
1194 IN6_PRINT(ip6buf, myaddr6));
1195 dp->dad_ns_lcount++;
1196 continue;
1197 }
1198
1199 break;
1200 }
1201 return dp;
1202 }
1203
1204 static bool
1205 nd6_dad_ownnonce(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1206 {
1207 bool found_nonce = true;
1208
1209 mutex_enter(&nd6_dad_lock);
1210 nd6_dad_find(ifa, nonce, &found_nonce);
1211 mutex_exit(&nd6_dad_lock);
1212
1213 return found_nonce;
1214 }
1215
1216 static void
1217 nd6_dad_starttimer(struct dadq *dp, int ticks)
1218 {
1219
1220 callout_reset(&dp->dad_timer_ch, ticks,
1221 (void (*)(void *))nd6_dad_timer, dp);
1222 }
1223
1224 static void
1225 nd6_dad_stoptimer(struct dadq *dp)
1226 {
1227
1228 KASSERT(mutex_owned(&nd6_dad_lock));
1229
1230 TAILQ_REMOVE(&dadq, dp, dad_list);
1231 /* Tell the timer that dp is being destroyed. */
1232 dp->dad_ifa = NULL;
1233 callout_halt(&dp->dad_timer_ch, &nd6_dad_lock);
1234 }
1235
1236 static void
1237 nd6_dad_destroytimer(struct dadq *dp)
1238 {
1239
1240 KASSERT(dp->dad_ifa == NULL);
1241 callout_destroy(&dp->dad_timer_ch);
1242 kmem_intr_free(dp, sizeof(*dp));
1243 }
1244
1245 /*
1246 * Start Duplicate Address Detection (DAD) for specified interface address.
1247 *
1248 * Note that callout is used when xtick > 0 and not when xtick == 0.
1249 *
1250 * xtick: minimum delay ticks for IFF_UP event
1251 */
1252 void
1253 nd6_dad_start(struct ifaddr *ifa, int xtick)
1254 {
1255 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1256 struct dadq *dp;
1257 char ip6buf[INET6_ADDRSTRLEN];
1258
1259 if (!dad_init) {
1260 TAILQ_INIT(&dadq);
1261 mutex_init(&nd6_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1262 dad_init++;
1263 }
1264
1265 /*
1266 * If we don't need DAD, don't do it.
1267 * There are several cases:
1268 * - DAD is disabled
1269 * - the interface address is anycast
1270 */
1271 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1272 log(LOG_DEBUG,
1273 "nd6_dad_start: called with non-tentative address "
1274 "%s(%s)\n",
1275 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1276 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1277 return;
1278 }
1279 if (ia->ia6_flags & IN6_IFF_ANYCAST || !ip6_dad_enabled()) {
1280 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1281 rt_addrmsg(RTM_NEWADDR, ifa);
1282 return;
1283 }
1284 KASSERT(ifa->ifa_ifp != NULL);
1285 if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1286 return;
1287
1288 dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1289
1290 mutex_enter(&nd6_dad_lock);
1291 if (nd6_dad_find(ifa, NULL, NULL) != NULL) {
1292 mutex_exit(&nd6_dad_lock);
1293 /* DAD already in progress */
1294 if (dp != NULL)
1295 kmem_intr_free(dp, sizeof(*dp));
1296 return;
1297 }
1298
1299 if (dp == NULL) {
1300 mutex_exit(&nd6_dad_lock);
1301 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1302 "%s(%s)\n",
1303 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1304 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1305 return;
1306 }
1307
1308 /*
1309 * Send NS packet for DAD, ip6_dad_count times.
1310 * Note that we must delay the first transmission, if this is the
1311 * first packet to be sent from the interface after interface
1312 * (re)initialization.
1313 */
1314 callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1315 dp->dad_ifa = ifa;
1316 ifaref(ifa); /* just for safety */
1317 dp->dad_count = ip6_dad_count;
1318 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1319 dp->dad_ns_lcount = 0;
1320 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1321
1322 nd6log(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1323 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1324
1325 if (xtick == 0) {
1326 nd6_dad_ns_output(dp, ifa);
1327 nd6_dad_starttimer(dp,
1328 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1329 } else
1330 nd6_dad_starttimer(dp, xtick);
1331 mutex_exit(&nd6_dad_lock);
1332 }
1333
1334 /*
1335 * terminate DAD unconditionally. used for address removals.
1336 */
1337 void
1338 nd6_dad_stop(struct ifaddr *ifa)
1339 {
1340 struct dadq *dp;
1341
1342 if (!dad_init)
1343 return;
1344
1345 mutex_enter(&nd6_dad_lock);
1346 dp = nd6_dad_find(ifa, NULL, NULL);
1347 if (dp == NULL) {
1348 mutex_exit(&nd6_dad_lock);
1349 /* DAD wasn't started yet */
1350 return;
1351 }
1352
1353 /* Prevent the timer from running anymore. */
1354 nd6_dad_stoptimer(dp);
1355
1356 mutex_exit(&nd6_dad_lock);
1357
1358 nd6_dad_destroytimer(dp);
1359 ifafree(ifa);
1360 }
1361
1362 static void
1363 nd6_dad_timer(struct dadq *dp)
1364 {
1365 struct ifaddr *ifa;
1366 struct in6_ifaddr *ia;
1367 char ip6buf[INET6_ADDRSTRLEN];
1368 bool need_free = false;
1369
1370 KERNEL_LOCK_UNLESS_NET_MPSAFE();
1371 mutex_enter(&nd6_dad_lock);
1372
1373 ifa = dp->dad_ifa;
1374 if (ifa == NULL) {
1375 /* dp is being destroyed by someone. Do nothing. */
1376 goto done;
1377 }
1378
1379 ia = (struct in6_ifaddr *)ifa;
1380 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1381 log(LOG_ERR, "nd6_dad_timer: called with duplicate address "
1382 "%s(%s)\n",
1383 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1384 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1385 goto done;
1386 }
1387 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1388 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1389 "%s(%s)\n",
1390 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1391 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1392 goto done;
1393 }
1394
1395 /* timeouted with IFF_{RUNNING,UP} check */
1396 if (dp->dad_ns_tcount > dad_maxtry) {
1397 nd6log(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1398 if_name(ifa->ifa_ifp));
1399
1400 nd6_dad_stoptimer(dp);
1401 need_free = true;
1402 goto done;
1403 }
1404
1405 /* Need more checks? */
1406 if (dp->dad_ns_ocount < dp->dad_count) {
1407 /*
1408 * We have more NS to go. Send NS packet for DAD.
1409 */
1410 nd6_dad_ns_output(dp, ifa);
1411 nd6_dad_starttimer(dp,
1412 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1413 } else {
1414 /*
1415 * We are done with DAD. No NA came, no NS came.
1416 * No duplicate address found.
1417 */
1418 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1419 rt_addrmsg(RTM_NEWADDR, ifa);
1420
1421 nd6log(LOG_DEBUG,
1422 "%s: DAD complete for %s - no duplicates found\n",
1423 if_name(ifa->ifa_ifp),
1424 IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1425
1426 nd6_dad_stoptimer(dp);
1427 need_free = true;
1428 }
1429 done:
1430 mutex_exit(&nd6_dad_lock);
1431
1432 if (need_free) {
1433 nd6_dad_destroytimer(dp);
1434 KASSERT(ifa != NULL);
1435 ifafree(ifa);
1436 }
1437
1438 KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1439 }
1440
1441 static void
1442 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp,
1443 const struct sockaddr_dl *from)
1444 {
1445 struct in6_ifaddr *ia;
1446 struct ifnet *ifp;
1447 char ip6buf[INET6_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN], *llastr;
1448
1449 KASSERT(mutex_owned(&nd6_dad_lock));
1450 KASSERT(ifa != NULL);
1451
1452 ifp = ifa->ifa_ifp;
1453 ia = (struct in6_ifaddr *)ifa;
1454
1455 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1456 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1457
1458 if (__predict_false(from == NULL))
1459 llastr = NULL;
1460 else
1461 llastr = lla_snprintf(llabuf, sizeof(llabuf),
1462 CLLADDR(from), from->sdl_alen);
1463
1464 log(LOG_ERR, "%s: DAD duplicate address %s from %s\n",
1465 if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr), llastr);
1466
1467 /* Inform the routing socket that DAD has completed */
1468 rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1469
1470 /*
1471 * If the address is a link-local address formed from an interface
1472 * identifier based on the hardware address which is supposed to be
1473 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1474 * operation on the interface SHOULD be disabled.
1475 * [rfc2462bis-03 Section 5.4.5]
1476 */
1477 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1478 struct in6_addr in6;
1479
1480 /*
1481 * To avoid over-reaction, we only apply this logic when we are
1482 * very sure that hardware addresses are supposed to be unique.
1483 */
1484 switch (ifp->if_type) {
1485 case IFT_ETHER:
1486 case IFT_FDDI:
1487 case IFT_ATM:
1488 case IFT_IEEE1394:
1489 case IFT_IEEE80211:
1490 in6 = ia->ia_addr.sin6_addr;
1491 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1492 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1493 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1494 log(LOG_ERR, "%s: possible hardware address "
1495 "duplication detected, disable IPv6\n",
1496 if_name(ifp));
1497 }
1498 break;
1499 }
1500 }
1501 }
1502
1503 static void
1504 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1505 {
1506 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1507 struct ifnet *ifp = ifa->ifa_ifp;
1508 uint8_t *nonce;
1509
1510 dp->dad_ns_tcount++;
1511 if ((ifp->if_flags & IFF_UP) == 0) {
1512 #if 0
1513 printf("%s: interface down?\n", if_name(ifp));
1514 #endif
1515 return;
1516 }
1517 if ((ifp->if_flags & IFF_RUNNING) == 0) {
1518 #if 0
1519 printf("%s: interface not running?\n", if_name(ifp));
1520 #endif
1521 return;
1522 }
1523
1524 dp->dad_ns_tcount = 0;
1525 nonce = dp->dad_nonce[dp->dad_ns_ocount % ND_OPT_NONCE_STORE];
1526 cprng_fast(nonce, ND_OPT_NONCE_LEN);
1527 dp->dad_ns_ocount++;
1528
1529 nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, nonce);
1530 }
1531
1532 static void
1533 nd6_dad_input(struct ifaddr *ifa, struct nd_opt_nonce *nonce,
1534 const struct sockaddr_dl *from)
1535 {
1536 struct dadq *dp;
1537 bool found_nonce = false;
1538
1539 KASSERT(ifa != NULL);
1540
1541 mutex_enter(&nd6_dad_lock);
1542 dp = nd6_dad_find(ifa, nonce, &found_nonce);
1543 if (!found_nonce) {
1544 nd6_dad_duplicated(ifa, dp, from);
1545 if (dp != NULL)
1546 nd6_dad_stoptimer(dp);
1547 }
1548 mutex_exit(&nd6_dad_lock);
1549 if (dp != NULL) {
1550 nd6_dad_destroytimer(dp);
1551 ifafree(ifa);
1552 }
1553 }
1554