1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32 */
33
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/eventhandler.h>
44 #include <sys/malloc.h>
45 #include <sys/libkern.h>
46 #include <sys/lock.h>
47 #include <sys/rwlock.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/errno.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
56 #include <sys/queue.h>
57 #include <sys/callout.h>
58 #include <sys/refcount.h>
59
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_dl.h>
63 #include <net/if_var.h>
64 #include <net/route.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <net/if_llatbl.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip_carp.h>
78 #include <netinet6/send.h>
79
80 #define SDL(s) ((struct sockaddr_dl *)s)
81
82 struct dadq;
83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
84 static void nd6_dad_add(struct dadq *dp);
85 static void nd6_dad_del(struct dadq *dp);
86 static void nd6_dad_rele(struct dadq *);
87 static void nd6_dad_starttimer(struct dadq *, int);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(void *);
90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
91 static void nd6_dad_ns_output(struct dadq *);
92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
93 static void nd6_dad_na_input(struct ifaddr *);
94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
95 const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
97 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
98
99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
100 #define V_dad_enhanced VNET(dad_enhanced)
101
102 SYSCTL_DECL(_net_inet6_ip6);
103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
104 &VNET_NAME(dad_enhanced), 0,
105 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
106
107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to
108 transmit DAD packet */
109 #define V_dad_maxtry VNET(dad_maxtry)
110
111 /*
112 * Input a Neighbor Solicitation Message.
113 *
114 * Based on RFC 2461
115 * Based on RFC 2462 (duplicate address detection)
116 */
117 void
118 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
119 {
120 struct ifnet *ifp;
121 struct ip6_hdr *ip6;
122 struct nd_neighbor_solicit *nd_ns;
123 struct in6_addr daddr6, myaddr6, saddr6, taddr6;
124 struct ifaddr *ifa;
125 struct sockaddr_dl proxydl;
126 union nd_opts ndopts;
127 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
128 char *lladdr;
129 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
130
131 ifa = NULL;
132
133 /* RFC 6980: Nodes MUST silently ignore fragments */
134 if(m->m_flags & M_FRAGMENTED)
135 goto freeit;
136
137 ifp = m->m_pkthdr.rcvif;
138 ip6 = mtod(m, struct ip6_hdr *);
139 if (__predict_false(ip6->ip6_hlim != 255)) {
140 ICMP6STAT_INC(icp6s_invlhlim);
141 nd6log((LOG_ERR,
142 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
143 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
144 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
145 goto bads;
146 }
147
148 if (m->m_len < off + icmp6len) {
149 m = m_pullup(m, off + icmp6len);
150 if (m == NULL) {
151 IP6STAT_INC(ip6s_exthdrtoolong);
152 return;
153 }
154 }
155 ip6 = mtod(m, struct ip6_hdr *);
156 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
157
158 saddr6 = ip6->ip6_src;
159 daddr6 = ip6->ip6_dst;
160 taddr6 = nd_ns->nd_ns_target;
161 if (in6_setscope(&taddr6, ifp, NULL) != 0)
162 goto bad;
163
164 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
165 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
166 rflag = 0;
167
168 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
169 /* dst has to be a solicited node multicast address. */
170 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
171 /* don't check ifindex portion */
172 daddr6.s6_addr32[1] == 0 &&
173 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
174 daddr6.s6_addr8[12] == 0xff) {
175 ; /* good */
176 } else {
177 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
178 "(wrong ip6 dst)\n"));
179 goto bad;
180 }
181 } else if (!V_nd6_onlink_ns_rfc4861) {
182 struct sockaddr_in6 src_sa6;
183
184 /*
185 * According to recent IETF discussions, it is not a good idea
186 * to accept a NS from an address which would not be deemed
187 * to be a neighbor otherwise. This point is expected to be
188 * clarified in future revisions of the specification.
189 */
190 bzero(&src_sa6, sizeof(src_sa6));
191 src_sa6.sin6_family = AF_INET6;
192 src_sa6.sin6_len = sizeof(src_sa6);
193 src_sa6.sin6_addr = saddr6;
194 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
195 nd6log((LOG_INFO, "nd6_ns_input: "
196 "NS packet from non-neighbor\n"));
197 goto bad;
198 }
199 }
200
201 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
202 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
203 goto bad;
204 }
205
206 icmp6len -= sizeof(*nd_ns);
207 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
208 if (nd6_options(&ndopts) < 0) {
209 nd6log((LOG_INFO,
210 "nd6_ns_input: invalid ND option, ignored\n"));
211 /* nd6_options have incremented stats */
212 goto freeit;
213 }
214
215 lladdr = NULL;
216 lladdrlen = 0;
217 if (ndopts.nd_opts_src_lladdr) {
218 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
219 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
220 }
221
222 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
223 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
224 "(link-layer address option)\n"));
225 goto bad;
226 }
227
228 /*
229 * Attaching target link-layer address to the NA?
230 * (RFC 2461 7.2.4)
231 *
232 * NS IP dst is unicast/anycast MUST NOT add
233 * NS IP dst is solicited-node multicast MUST add
234 *
235 * In implementation, we add target link-layer address by default.
236 * We do not add one in MUST NOT cases.
237 */
238 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
239 tlladdr = 0;
240 else
241 tlladdr = 1;
242
243 /*
244 * Target address (taddr6) must be either:
245 * (1) Valid unicast/anycast address for my receiving interface,
246 * (2) Unicast address for which I'm offering proxy service, or
247 * (3) "tentative" address on which DAD is being performed.
248 */
249 /* (1) and (3) check. */
250 if (ifp->if_carp)
251 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
252 else
253 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
254
255 /* (2) check. */
256 proxy = 0;
257 if (ifa == NULL) {
258 struct sockaddr_dl rt_gateway;
259 struct rt_addrinfo info;
260 struct sockaddr_in6 dst6;
261
262 bzero(&dst6, sizeof(dst6));
263 dst6.sin6_len = sizeof(struct sockaddr_in6);
264 dst6.sin6_family = AF_INET6;
265 dst6.sin6_addr = taddr6;
266
267 bzero(&rt_gateway, sizeof(rt_gateway));
268 rt_gateway.sdl_len = sizeof(rt_gateway);
269 bzero(&info, sizeof(info));
270 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
271
272 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
273 0, 0, &info) == 0) {
274 if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
275 rt_gateway.sdl_family == AF_LINK) {
276 /*
277 * proxy NDP for single entry
278 */
279 proxydl = *SDL(&rt_gateway);
280 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
281 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
282 if (ifa)
283 proxy = 1;
284 }
285 }
286 }
287 if (ifa == NULL) {
288 /*
289 * We've got an NS packet, and we don't have that adddress
290 * assigned for us. We MUST silently ignore it.
291 * See RFC2461 7.2.3.
292 */
293 goto freeit;
294 }
295 myaddr6 = *IFA_IN6(ifa);
296 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
297 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
298 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
299 goto freeit;
300
301 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
302 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
303 "(if %d, NS packet %d)\n",
304 ip6_sprintf(ip6bufs, &taddr6),
305 ifp->if_addrlen, lladdrlen - 2));
306 goto bad;
307 }
308
309 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
310 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
311 ip6_sprintf(ip6bufs, &saddr6)));
312 goto freeit;
313 }
314
315 /*
316 * We have neighbor solicitation packet, with target address equals to
317 * one of my tentative address.
318 *
319 * src addr how to process?
320 * --- ---
321 * multicast of course, invalid (rejected in ip6_input)
322 * unicast somebody is doing address resolution -> ignore
323 * unspec dup address detection
324 *
325 * The processing is defined in RFC 2462.
326 */
327 if (tentative) {
328 /*
329 * If source address is unspecified address, it is for
330 * duplicate address detection.
331 *
332 * If not, the packet is for addess resolution;
333 * silently ignore it.
334 */
335 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
336 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
337
338 goto freeit;
339 }
340
341 /*
342 * If the source address is unspecified address, entries must not
343 * be created or updated.
344 * It looks that sender is performing DAD. Output NA toward
345 * all-node multicast address, to tell the sender that I'm using
346 * the address.
347 * S bit ("solicited") must be zero.
348 */
349 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
350 struct in6_addr in6_all;
351
352 in6_all = in6addr_linklocal_allnodes;
353 if (in6_setscope(&in6_all, ifp, NULL) != 0)
354 goto bad;
355 nd6_na_output_fib(ifp, &in6_all, &taddr6,
356 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
357 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
358 M_GETFIB(m));
359 goto freeit;
360 }
361
362 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
363 ND_NEIGHBOR_SOLICIT, 0);
364
365 nd6_na_output_fib(ifp, &saddr6, &taddr6,
366 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
367 rflag | ND_NA_FLAG_SOLICITED, tlladdr,
368 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
369 freeit:
370 if (ifa != NULL)
371 ifa_free(ifa);
372 m_freem(m);
373 return;
374
375 bad:
376 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
377 ip6_sprintf(ip6bufs, &saddr6)));
378 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
379 ip6_sprintf(ip6bufs, &daddr6)));
380 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
381 ip6_sprintf(ip6bufs, &taddr6)));
382 bads:
383 ICMP6STAT_INC(icp6s_badns);
384 if (ifa != NULL)
385 ifa_free(ifa);
386 m_freem(m);
387 }
388
389 /*
390 * Output a Neighbor Solicitation Message. Caller specifies:
391 * - ICMP6 header source IP6 address
392 * - ND6 header target IP6 address
393 * - ND6 header source datalink address
394 *
395 * Based on RFC 2461
396 * Based on RFC 2462 (duplicate address detection)
397 *
398 * ln - for source address determination
399 * nonce - If non-NULL, NS is used for duplicate address detection and
400 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
401 */
402 static void
403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
404 const struct in6_addr *daddr6, const struct in6_addr *taddr6,
405 uint8_t *nonce, u_int fibnum)
406 {
407 struct mbuf *m;
408 struct m_tag *mtag;
409 struct ip6_hdr *ip6;
410 struct nd_neighbor_solicit *nd_ns;
411 struct ip6_moptions im6o;
412 int icmp6len;
413 int maxlen;
414
415 NET_EPOCH_ASSERT();
416
417 if (IN6_IS_ADDR_MULTICAST(taddr6))
418 return;
419
420 /* estimate the size of message */
421 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
422 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
423 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
424 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
425 __func__, max_linkhdr, maxlen, MCLBYTES));
426
427 if (max_linkhdr + maxlen > MHLEN)
428 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
429 else
430 m = m_gethdr(M_NOWAIT, MT_DATA);
431 if (m == NULL)
432 return;
433 M_SETFIB(m, fibnum);
434
435 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
436 m->m_flags |= M_MCAST;
437 im6o.im6o_multicast_ifp = ifp;
438 im6o.im6o_multicast_hlim = 255;
439 im6o.im6o_multicast_loop = 0;
440 }
441
442 icmp6len = sizeof(*nd_ns);
443 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
444 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
445
446 /* fill neighbor solicitation packet */
447 ip6 = mtod(m, struct ip6_hdr *);
448 ip6->ip6_flow = 0;
449 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
450 ip6->ip6_vfc |= IPV6_VERSION;
451 /* ip6->ip6_plen will be set later */
452 ip6->ip6_nxt = IPPROTO_ICMPV6;
453 ip6->ip6_hlim = 255;
454 if (daddr6)
455 ip6->ip6_dst = *daddr6;
456 else {
457 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
458 ip6->ip6_dst.s6_addr16[1] = 0;
459 ip6->ip6_dst.s6_addr32[1] = 0;
460 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
461 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
462 ip6->ip6_dst.s6_addr8[12] = 0xff;
463 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
464 goto bad;
465 }
466 if (nonce == NULL) {
467 struct ifaddr *ifa = NULL;
468
469 /*
470 * RFC2461 7.2.2:
471 * "If the source address of the packet prompting the
472 * solicitation is the same as one of the addresses assigned
473 * to the outgoing interface, that address SHOULD be placed
474 * in the IP Source Address of the outgoing solicitation.
475 * Otherwise, any one of the addresses assigned to the
476 * interface should be used."
477 *
478 * We use the source address for the prompting packet
479 * (saddr6), if saddr6 belongs to the outgoing interface.
480 * Otherwise, we perform the source address selection as usual.
481 */
482
483 if (saddr6 != NULL)
484 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
485 if (ifa != NULL) {
486 /* ip6_src set already. */
487 ip6->ip6_src = *saddr6;
488 ifa_free(ifa);
489 } else {
490 int error;
491 struct in6_addr dst6, src6;
492 uint32_t scopeid;
493
494 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
495 error = in6_selectsrc_addr(fibnum, &dst6,
496 scopeid, ifp, &src6, NULL);
497 if (error) {
498 char ip6buf[INET6_ADDRSTRLEN];
499 nd6log((LOG_DEBUG, "%s: source can't be "
500 "determined: dst=%s, error=%d\n", __func__,
501 ip6_sprintf(ip6buf, &dst6),
502 error));
503 goto bad;
504 }
505 ip6->ip6_src = src6;
506 }
507 } else {
508 /*
509 * Source address for DAD packet must always be IPv6
510 * unspecified address. (0::0)
511 * We actually don't have to 0-clear the address (we did it
512 * above), but we do so here explicitly to make the intention
513 * clearer.
514 */
515 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
516 }
517 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
518 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
519 nd_ns->nd_ns_code = 0;
520 nd_ns->nd_ns_reserved = 0;
521 nd_ns->nd_ns_target = *taddr6;
522 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
523
524 /*
525 * Add source link-layer address option.
526 *
527 * spec implementation
528 * --- ---
529 * DAD packet MUST NOT do not add the option
530 * there's no link layer address:
531 * impossible do not add the option
532 * there's link layer address:
533 * Multicast NS MUST add one add the option
534 * Unicast NS SHOULD add one add the option
535 */
536 if (nonce == NULL) {
537 struct nd_opt_hdr *nd_opt;
538 char *mac;
539 int optlen;
540
541 mac = NULL;
542 if (ifp->if_carp)
543 mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
544 if (mac == NULL)
545 mac = nd6_ifptomac(ifp);
546
547 if (mac != NULL) {
548 nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
549 optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
550 /* 8 byte alignments... */
551 optlen = (optlen + 7) & ~7;
552 m->m_pkthdr.len += optlen;
553 m->m_len += optlen;
554 icmp6len += optlen;
555 bzero(nd_opt, optlen);
556 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
557 nd_opt->nd_opt_len = optlen >> 3;
558 bcopy(mac, nd_opt + 1, ifp->if_addrlen);
559 }
560 }
561 /*
562 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
563 * This behavior is documented as Enhanced Duplicate Address
564 * Detection in RFC 7527.
565 * net.inet6.ip6.dad_enhanced=0 disables this.
566 */
567 if (V_dad_enhanced != 0 && nonce != NULL) {
568 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
569 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
570 /* 8-byte alignment is required. */
571 optlen = (optlen + 7) & ~7;
572
573 m->m_pkthdr.len += optlen;
574 m->m_len += optlen;
575 icmp6len += optlen;
576 bzero((caddr_t)nd_opt, optlen);
577 nd_opt->nd_opt_type = ND_OPT_NONCE;
578 nd_opt->nd_opt_len = optlen >> 3;
579 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
580 }
581 ip6->ip6_plen = htons((u_short)icmp6len);
582 nd_ns->nd_ns_cksum = 0;
583 nd_ns->nd_ns_cksum =
584 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
585
586 if (send_sendso_input_hook != NULL) {
587 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
588 sizeof(unsigned short), M_NOWAIT);
589 if (mtag == NULL)
590 goto bad;
591 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
592 m_tag_prepend(m, mtag);
593 }
594
595 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
596 &im6o, NULL, NULL);
597 icmp6_ifstat_inc(ifp, ifs6_out_msg);
598 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
599 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
600
601 return;
602
603 bad:
604 m_freem(m);
605 }
606
607 #ifndef BURN_BRIDGES
608 void
609 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
610 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
611 {
612
613 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
614 }
615 #endif
616 /*
617 * Neighbor advertisement input handling.
618 *
619 * Based on RFC 2461
620 * Based on RFC 2462 (duplicate address detection)
621 *
622 * the following items are not implemented yet:
623 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
624 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
625 */
626 void
627 nd6_na_input(struct mbuf *m, int off, int icmp6len)
628 {
629 struct ifnet *ifp;
630 struct ip6_hdr *ip6;
631 struct ifaddr *ifa;
632 struct llentry *ln;
633 struct mbuf *chain;
634 struct nd_neighbor_advert *nd_na;
635 struct in6_addr daddr6, taddr6;
636 union nd_opts ndopts;
637 u_char linkhdr[LLE_MAX_LINKHDR];
638 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
639 char *lladdr;
640 size_t linkhdrsize;
641 int flags, is_override, is_router, is_solicited;
642 int lladdr_off, lladdrlen, checklink;
643 bool flush_holdchain = false;
644
645 NET_EPOCH_ASSERT();
646
647 chain = NULL;
648 ln = NULL;
649 checklink = 0;
650
651 /* RFC 6980: Nodes MUST silently ignore fragments */
652 if(m->m_flags & M_FRAGMENTED)
653 goto freeit;
654
655 ifp = m->m_pkthdr.rcvif;
656 ip6 = mtod(m, struct ip6_hdr *);
657 if (__predict_false(ip6->ip6_hlim != 255)) {
658 ICMP6STAT_INC(icp6s_invlhlim);
659 nd6log((LOG_ERR,
660 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
661 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
662 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
663 goto bad;
664 }
665
666 if (m->m_len < off + icmp6len) {
667 m = m_pullup(m, off + icmp6len);
668 if (m == NULL) {
669 IP6STAT_INC(ip6s_exthdrtoolong);
670 return;
671 }
672 }
673 ip6 = mtod(m, struct ip6_hdr *);
674 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
675
676 flags = nd_na->nd_na_flags_reserved;
677 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
678 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
679 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
680
681 taddr6 = nd_na->nd_na_target;
682 if (in6_setscope(&taddr6, ifp, NULL))
683 goto bad; /* XXX: impossible */
684
685 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
686 nd6log((LOG_ERR,
687 "nd6_na_input: invalid target address %s\n",
688 ip6_sprintf(ip6bufs, &taddr6)));
689 goto bad;
690 }
691
692 daddr6 = ip6->ip6_dst;
693 if (IN6_IS_ADDR_MULTICAST(&daddr6))
694 if (is_solicited) {
695 nd6log((LOG_ERR,
696 "nd6_na_input: a solicited adv is multicasted\n"));
697 goto bad;
698 }
699
700 icmp6len -= sizeof(*nd_na);
701 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
702 if (nd6_options(&ndopts) < 0) {
703 nd6log((LOG_INFO,
704 "nd6_na_input: invalid ND option, ignored\n"));
705 /* nd6_options have incremented stats */
706 goto freeit;
707 }
708
709 lladdr = NULL;
710 lladdrlen = 0;
711 if (ndopts.nd_opts_tgt_lladdr) {
712 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
713 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
714 }
715
716 /*
717 * This effectively disables the DAD check on a non-master CARP
718 * address.
719 */
720 if (ifp->if_carp)
721 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
722 else
723 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
724
725 /*
726 * Target address matches one of my interface address.
727 *
728 * If my address is tentative, this means that there's somebody
729 * already using the same address as mine. This indicates DAD failure.
730 * This is defined in RFC 2462.
731 *
732 * Otherwise, process as defined in RFC 2461.
733 */
734 if (ifa
735 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
736 nd6_dad_na_input(ifa);
737 ifa_free(ifa);
738 goto freeit;
739 }
740
741 /* Just for safety, maybe unnecessary. */
742 if (ifa) {
743 ifa_free(ifa);
744 log(LOG_ERR,
745 "nd6_na_input: duplicate IP6 address %s\n",
746 ip6_sprintf(ip6bufs, &taddr6));
747 goto freeit;
748 }
749
750 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
751 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
752 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
753 ifp->if_addrlen, lladdrlen - 2));
754 goto bad;
755 }
756
757 /*
758 * If no neighbor cache entry is found, NA SHOULD silently be
759 * discarded.
760 */
761 ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
762 if (ln == NULL) {
763 goto freeit;
764 }
765
766 /*
767 * Do not try to override static entry.
768 */
769 if (ln->la_flags & LLE_STATIC)
770 goto freeit;
771
772 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
773 /*
774 * If the link-layer has address, and no lladdr option came,
775 * discard the packet.
776 */
777 if (ifp->if_addrlen && lladdr == NULL) {
778 goto freeit;
779 }
780
781 /*
782 * Record link-layer address, and update the state.
783 */
784 if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
785 goto freeit;
786
787 flush_holdchain = true;
788 if (is_solicited)
789 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
790 else
791 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
792 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
793 if ((ln->ln_router = is_router) != 0) {
794 /*
795 * This means a router's state has changed from
796 * non-reachable to probably reachable, and might
797 * affect the status of associated prefixes..
798 */
799 checklink = 1;
800 }
801 } else {
802 int llchange;
803
804 /*
805 * Check if the link-layer address has changed or not.
806 */
807 if (lladdr == NULL)
808 llchange = 0;
809 else {
810 if (ln->la_flags & LLE_VALID) {
811 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
812 llchange = 1;
813 else
814 llchange = 0;
815 } else
816 llchange = 1;
817 }
818
819 /*
820 * This is VERY complex. Look at it with care.
821 *
822 * override solicit lladdr llchange action
823 * (L: record lladdr)
824 *
825 * 0 0 n -- (2c)
826 * 0 0 y n (2b) L
827 * 0 0 y y (1) REACHABLE->STALE
828 * 0 1 n -- (2c) *->REACHABLE
829 * 0 1 y n (2b) L *->REACHABLE
830 * 0 1 y y (1) REACHABLE->STALE
831 * 1 0 n -- (2a)
832 * 1 0 y n (2a) L
833 * 1 0 y y (2a) L *->STALE
834 * 1 1 n -- (2a) *->REACHABLE
835 * 1 1 y n (2a) L *->REACHABLE
836 * 1 1 y y (2a) L *->REACHABLE
837 */
838 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
839 /*
840 * If state is REACHABLE, make it STALE.
841 * no other updates should be done.
842 */
843 if (ln->ln_state == ND6_LLINFO_REACHABLE)
844 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
845 goto freeit;
846 } else if (is_override /* (2a) */
847 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
848 || lladdr == NULL) { /* (2c) */
849 /*
850 * Update link-local address, if any.
851 */
852 if (lladdr != NULL) {
853 linkhdrsize = sizeof(linkhdr);
854 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
855 linkhdr, &linkhdrsize, &lladdr_off) != 0)
856 goto freeit;
857 if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
858 linkhdrsize, lladdr_off) == 0)
859 goto freeit;
860 EVENTHANDLER_INVOKE(lle_event, ln,
861 LLENTRY_RESOLVED);
862 }
863
864 /*
865 * If solicited, make the state REACHABLE.
866 * If not solicited and the link-layer address was
867 * changed, make it STALE.
868 */
869 if (is_solicited)
870 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
871 else {
872 if (lladdr != NULL && llchange)
873 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
874 }
875 }
876
877 if (ln->ln_router && !is_router) {
878 /*
879 * The peer dropped the router flag.
880 * Remove the sender from the Default Router List and
881 * update the Destination Cache entries.
882 */
883 struct ifnet *nd6_ifp;
884
885 nd6_ifp = lltable_get_ifp(ln->lle_tbl);
886 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
887 (ND_IFINFO(nd6_ifp)->flags &
888 ND6_IFF_ACCEPT_RTADV) != 0)
889 /*
890 * Even if the neighbor is not in the default
891 * router list, the neighbor may be used as a
892 * next hop for some destinations (e.g. redirect
893 * case). So we must call rt6_flush explicitly.
894 */
895 rt6_flush(&ip6->ip6_src, ifp);
896 }
897 ln->ln_router = is_router;
898 }
899 /* XXX - QL
900 * Does this matter?
901 * rt->rt_flags &= ~RTF_REJECT;
902 */
903 ln->la_asked = 0;
904 if (ln->la_hold != NULL)
905 chain = nd6_grab_holdchain(ln);
906 freeit:
907 if (ln != NULL)
908 LLE_WUNLOCK(ln);
909
910 if (chain != NULL)
911 nd6_flush_holdchain(ifp, ln, chain);
912 if (flush_holdchain)
913 nd6_flush_children_holdchain(ifp, ln);
914
915 if (checklink)
916 pfxlist_onlink_check();
917
918 m_freem(m);
919 return;
920
921 bad:
922 if (ln != NULL)
923 LLE_WUNLOCK(ln);
924
925 ICMP6STAT_INC(icp6s_badna);
926 m_freem(m);
927 }
928
929 /*
930 * Neighbor advertisement output handling.
931 *
932 * Based on RFC 2461
933 *
934 * the following items are not implemented yet:
935 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
936 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
937 *
938 * tlladdr - 1 if include target link-layer address
939 * sdl0 - sockaddr_dl (= proxy NA) or NULL
940 */
941 static void
942 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
943 const struct in6_addr *taddr6, u_long flags, int tlladdr,
944 struct sockaddr *sdl0, u_int fibnum)
945 {
946 struct mbuf *m;
947 struct m_tag *mtag;
948 struct ip6_hdr *ip6;
949 struct nd_neighbor_advert *nd_na;
950 struct ip6_moptions im6o;
951 struct in6_addr daddr6, dst6, src6;
952 uint32_t scopeid;
953
954 NET_EPOCH_ASSERT();
955
956 int icmp6len, maxlen, error;
957 caddr_t mac = NULL;
958
959 daddr6 = *daddr6_0; /* make a local copy for modification */
960
961 /* estimate the size of message */
962 maxlen = sizeof(*ip6) + sizeof(*nd_na);
963 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
964 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
965 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
966 __func__, max_linkhdr, maxlen, MCLBYTES));
967
968 if (max_linkhdr + maxlen > MHLEN)
969 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
970 else
971 m = m_gethdr(M_NOWAIT, MT_DATA);
972 if (m == NULL)
973 return;
974 M_SETFIB(m, fibnum);
975
976 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
977 m->m_flags |= M_MCAST;
978 im6o.im6o_multicast_ifp = ifp;
979 im6o.im6o_multicast_hlim = 255;
980 im6o.im6o_multicast_loop = 0;
981 }
982
983 icmp6len = sizeof(*nd_na);
984 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
985 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
986
987 /* fill neighbor advertisement packet */
988 ip6 = mtod(m, struct ip6_hdr *);
989 ip6->ip6_flow = 0;
990 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
991 ip6->ip6_vfc |= IPV6_VERSION;
992 ip6->ip6_nxt = IPPROTO_ICMPV6;
993 ip6->ip6_hlim = 255;
994 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
995 /* reply to DAD */
996 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
997 daddr6.s6_addr16[1] = 0;
998 daddr6.s6_addr32[1] = 0;
999 daddr6.s6_addr32[2] = 0;
1000 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1001 if (in6_setscope(&daddr6, ifp, NULL))
1002 goto bad;
1003
1004 flags &= ~ND_NA_FLAG_SOLICITED;
1005 }
1006 ip6->ip6_dst = daddr6;
1007
1008 /*
1009 * Select a source whose scope is the same as that of the dest.
1010 */
1011 in6_splitscope(&daddr6, &dst6, &scopeid);
1012 error = in6_selectsrc_addr(fibnum, &dst6,
1013 scopeid, ifp, &src6, NULL);
1014 if (error) {
1015 char ip6buf[INET6_ADDRSTRLEN];
1016 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1017 "determined: dst=%s, error=%d\n",
1018 ip6_sprintf(ip6buf, &daddr6), error));
1019 goto bad;
1020 }
1021 ip6->ip6_src = src6;
1022 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1023 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1024 nd_na->nd_na_code = 0;
1025 nd_na->nd_na_target = *taddr6;
1026 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1027
1028 /*
1029 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1030 * see nd6_ns_input() for details.
1031 * Basically, if NS packet is sent to unicast/anycast addr,
1032 * target lladdr option SHOULD NOT be included.
1033 */
1034 if (tlladdr) {
1035 /*
1036 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1037 * lladdr in sdl0. If we are not proxying (sending NA for
1038 * my address) use lladdr configured for the interface.
1039 */
1040 if (sdl0 == NULL) {
1041 if (ifp->if_carp)
1042 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1043 if (mac == NULL)
1044 mac = nd6_ifptomac(ifp);
1045 } else if (sdl0->sa_family == AF_LINK) {
1046 struct sockaddr_dl *sdl;
1047 sdl = (struct sockaddr_dl *)sdl0;
1048 if (sdl->sdl_alen == ifp->if_addrlen)
1049 mac = LLADDR(sdl);
1050 }
1051 }
1052 if (tlladdr && mac) {
1053 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1054 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1055
1056 /* roundup to 8 bytes alignment! */
1057 optlen = (optlen + 7) & ~7;
1058
1059 m->m_pkthdr.len += optlen;
1060 m->m_len += optlen;
1061 icmp6len += optlen;
1062 bzero((caddr_t)nd_opt, optlen);
1063 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1064 nd_opt->nd_opt_len = optlen >> 3;
1065 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1066 } else
1067 flags &= ~ND_NA_FLAG_OVERRIDE;
1068
1069 ip6->ip6_plen = htons((u_short)icmp6len);
1070 nd_na->nd_na_flags_reserved = flags;
1071 nd_na->nd_na_cksum = 0;
1072 nd_na->nd_na_cksum =
1073 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1074
1075 if (send_sendso_input_hook != NULL) {
1076 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1077 sizeof(unsigned short), M_NOWAIT);
1078 if (mtag == NULL)
1079 goto bad;
1080 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1081 m_tag_prepend(m, mtag);
1082 }
1083
1084 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1085 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1086 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1087 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1088
1089 return;
1090
1091 bad:
1092 m_freem(m);
1093 }
1094
1095 #ifndef BURN_BRIDGES
1096 void
1097 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1098 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1099 struct sockaddr *sdl0)
1100 {
1101
1102 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1103 RT_DEFAULT_FIB);
1104 }
1105 #endif
1106
1107 caddr_t
1108 nd6_ifptomac(struct ifnet *ifp)
1109 {
1110 switch (ifp->if_type) {
1111 case IFT_ETHER:
1112 case IFT_IEEE1394:
1113 case IFT_L2VLAN:
1114 case IFT_INFINIBAND:
1115 case IFT_BRIDGE:
1116 return IF_LLADDR(ifp);
1117 default:
1118 return NULL;
1119 }
1120 }
1121
1122 struct dadq {
1123 TAILQ_ENTRY(dadq) dad_list;
1124 struct ifaddr *dad_ifa;
1125 int dad_count; /* max NS to send */
1126 int dad_ns_tcount; /* # of trials to send NS */
1127 int dad_ns_ocount; /* NS sent so far */
1128 int dad_ns_icount;
1129 int dad_na_icount;
1130 int dad_ns_lcount; /* looped back NS */
1131 int dad_loopbackprobe; /* probing state for loopback detection */
1132 struct callout dad_timer_ch;
1133 struct vnet *dad_vnet;
1134 u_int dad_refcnt;
1135 #define ND_OPT_NONCE_LEN32 \
1136 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1137 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1138 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */
1139 };
1140
1141 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1142 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1143 #define V_dadq VNET(dadq)
1144 #define V_dad_rwlock VNET(dad_rwlock)
1145
1146 #define DADQ_LOCKPTR() (&V_dad_rwlock)
1147 #define DADQ_LOCK_INIT() rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1148 #define DADQ_RLOCK() rw_rlock(DADQ_LOCKPTR())
1149 #define DADQ_RUNLOCK() rw_runlock(DADQ_LOCKPTR())
1150 #define DADQ_WLOCK() rw_wlock(DADQ_LOCKPTR())
1151 #define DADQ_WUNLOCK() rw_wunlock(DADQ_LOCKPTR())
1152
1153 #define DADQ_LOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1154 #define DADQ_RLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1155 #define DADQ_WLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1156
1157 static void
1158 nd6_dad_add(struct dadq *dp)
1159 {
1160 DADQ_WLOCK_ASSERT();
1161
1162 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1163 dp->dad_ondadq = true;
1164 }
1165
1166 static void
1167 nd6_dad_del(struct dadq *dp)
1168 {
1169 DADQ_WLOCK_ASSERT();
1170
1171 if (dp->dad_ondadq) {
1172 /*
1173 * Remove dp from the dadq and release the dadq's
1174 * reference.
1175 */
1176 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1177 dp->dad_ondadq = false;
1178 nd6_dad_rele(dp);
1179 }
1180 }
1181
1182 static struct dadq *
1183 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1184 {
1185 struct dadq *dp;
1186
1187 DADQ_LOCK_ASSERT();
1188
1189 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1190 if (dp->dad_ifa != ifa)
1191 continue;
1192
1193 /*
1194 * Skip if the nonce matches the received one.
1195 * +2 in the length is required because of type and
1196 * length fields are included in a header.
1197 */
1198 if (n != NULL &&
1199 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1200 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1201 ND_OPT_NONCE_LEN) == 0) {
1202 dp->dad_ns_lcount++;
1203 continue;
1204 }
1205 break;
1206 }
1207
1208 return (dp);
1209 }
1210
1211 static void
1212 nd6_dad_starttimer(struct dadq *dp, int ticks)
1213 {
1214 DADQ_WLOCK_ASSERT();
1215
1216 callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1217 }
1218
1219 static void
1220 nd6_dad_stoptimer(struct dadq *dp)
1221 {
1222 callout_drain(&dp->dad_timer_ch);
1223 }
1224
1225 static void
1226 nd6_dad_rele(struct dadq *dp)
1227 {
1228 if (refcount_release(&dp->dad_refcnt)) {
1229 KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1230 ifa_free(dp->dad_ifa);
1231 free(dp, M_IP6NDP);
1232 }
1233 }
1234
1235 void
1236 nd6_dad_init(void)
1237 {
1238 DADQ_LOCK_INIT();
1239 TAILQ_INIT(&V_dadq);
1240 }
1241
1242 /*
1243 * Start Duplicate Address Detection (DAD) for specified interface address.
1244 */
1245 void
1246 nd6_dad_start(struct ifaddr *ifa, int delay)
1247 {
1248 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1249 struct dadq *dp;
1250 char ip6buf[INET6_ADDRSTRLEN];
1251
1252 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1253 ("starting DAD on non-tentative address %p", ifa));
1254
1255 /*
1256 * If we don't need DAD, don't do it.
1257 * There are several cases:
1258 * - DAD is disabled globally or on the interface
1259 * - the interface address is anycast
1260 */
1261 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1262 V_ip6_dad_count == 0 ||
1263 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1264 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1265 return;
1266 }
1267 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1268 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1269 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1270 return;
1271
1272 DADQ_WLOCK();
1273 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1274 /*
1275 * DAD is already in progress. Let the existing entry
1276 * finish it.
1277 */
1278 DADQ_WUNLOCK();
1279 return;
1280 }
1281
1282 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1283 if (dp == NULL) {
1284 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1285 "%s(%s)\n",
1286 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1287 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1288 return;
1289 }
1290 callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1291 CALLOUT_RETURNUNLOCKED);
1292 #ifdef VIMAGE
1293 dp->dad_vnet = curvnet;
1294 #endif
1295 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1296 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1297
1298 /*
1299 * Send NS packet for DAD, ip6_dad_count times.
1300 * Note that we must delay the first transmission, if this is the
1301 * first packet to be sent from the interface after interface
1302 * (re)initialization.
1303 */
1304 dp->dad_ifa = ifa;
1305 ifa_ref(dp->dad_ifa);
1306 dp->dad_count = V_ip6_dad_count;
1307 dp->dad_ns_icount = dp->dad_na_icount = 0;
1308 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1309 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1310
1311 /* Add this to the dadq and add a reference for the dadq. */
1312 refcount_init(&dp->dad_refcnt, 1);
1313 nd6_dad_add(dp);
1314 nd6_dad_starttimer(dp, delay);
1315 DADQ_WUNLOCK();
1316 }
1317
1318 /*
1319 * terminate DAD unconditionally. used for address removals.
1320 */
1321 void
1322 nd6_dad_stop(struct ifaddr *ifa)
1323 {
1324 struct dadq *dp;
1325
1326 DADQ_WLOCK();
1327 dp = nd6_dad_find(ifa, NULL);
1328 if (dp == NULL) {
1329 DADQ_WUNLOCK();
1330 /* DAD wasn't started yet */
1331 return;
1332 }
1333
1334 /*
1335 * Acquire a temporary reference so that we can safely stop the callout.
1336 */
1337 (void)refcount_acquire(&dp->dad_refcnt);
1338 nd6_dad_del(dp);
1339 DADQ_WUNLOCK();
1340
1341 nd6_dad_stoptimer(dp);
1342 nd6_dad_rele(dp);
1343 }
1344
1345 static void
1346 nd6_dad_timer(void *arg)
1347 {
1348 struct dadq *dp = arg;
1349 struct ifaddr *ifa = dp->dad_ifa;
1350 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1351 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1352 char ip6buf[INET6_ADDRSTRLEN];
1353 struct epoch_tracker et;
1354
1355 CURVNET_SET(dp->dad_vnet);
1356 KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1357
1358 NET_EPOCH_ENTER(et);
1359 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1360 /* Do not need DAD for ifdisabled interface. */
1361 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1362 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1363 goto err;
1364 }
1365 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1366 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1367 "%s(%s)\n",
1368 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1369 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1370 goto err;
1371 }
1372 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1373 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1374 "%s(%s)\n",
1375 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1376 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1377 goto err;
1378 }
1379
1380 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1381 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1382 (((ifp->if_flags & IFF_UP) == 0) ||
1383 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1384 nd6log((LOG_INFO, "%s: could not run DAD "
1385 "because the interface was down or not running.\n",
1386 if_name(ifa->ifa_ifp)));
1387 goto err;
1388 }
1389
1390 /* Need more checks? */
1391 if (dp->dad_ns_ocount < dp->dad_count) {
1392 /*
1393 * We have more NS to go. Send NS packet for DAD.
1394 */
1395 nd6_dad_starttimer(dp,
1396 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1397 nd6_dad_ns_output(dp);
1398 goto done;
1399 } else {
1400 /*
1401 * We have transmitted sufficient number of DAD packets.
1402 * See what we've got.
1403 */
1404 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1405 /* We've seen NS or NA, means DAD has failed. */
1406 nd6_dad_duplicated(ifa, dp);
1407 } else if (V_dad_enhanced != 0 &&
1408 dp->dad_ns_lcount > 0 &&
1409 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1410 /*
1411 * Sec. 4.1 in RFC 7527 requires transmission of
1412 * additional probes until the loopback condition
1413 * becomes clear when a looped back probe is detected.
1414 */
1415 log(LOG_ERR, "%s: a looped back NS message is "
1416 "detected during DAD for %s. "
1417 "Another DAD probes are being sent.\n",
1418 if_name(ifa->ifa_ifp),
1419 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1420 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1421 /*
1422 * Send an NS immediately and increase dad_count by
1423 * V_nd6_mmaxtries - 1.
1424 */
1425 dp->dad_count =
1426 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1427 nd6_dad_starttimer(dp,
1428 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1429 nd6_dad_ns_output(dp);
1430 goto done;
1431 } else {
1432 /*
1433 * We are done with DAD. No NA came, no NS came.
1434 * No duplicate address found. Check IFDISABLED flag
1435 * again in case that it is changed between the
1436 * beginning of this function and here.
1437 */
1438 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1439 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1440
1441 nd6log((LOG_DEBUG,
1442 "%s: DAD complete for %s - no duplicates found\n",
1443 if_name(ifa->ifa_ifp),
1444 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1445 if (dp->dad_ns_lcount > 0)
1446 log(LOG_ERR, "%s: DAD completed while "
1447 "a looped back NS message is detected "
1448 "during DAD for %s.\n",
1449 if_name(ifa->ifa_ifp),
1450 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1451 }
1452 }
1453 err:
1454 nd6_dad_del(dp);
1455 DADQ_WUNLOCK();
1456 done:
1457 NET_EPOCH_EXIT(et);
1458 CURVNET_RESTORE();
1459 }
1460
1461 static void
1462 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1463 {
1464 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1465 struct ifnet *ifp;
1466 char ip6buf[INET6_ADDRSTRLEN];
1467
1468 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1469 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1470 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1471 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1472 dp->dad_na_icount);
1473
1474 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1475 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1476
1477 ifp = ifa->ifa_ifp;
1478 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1479 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1480 log(LOG_ERR, "%s: manual intervention required\n",
1481 if_name(ifp));
1482
1483 /*
1484 * If the address is a link-local address formed from an interface
1485 * identifier based on the hardware address which is supposed to be
1486 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1487 * operation on the interface SHOULD be disabled.
1488 * [RFC 4862, Section 5.4.5]
1489 */
1490 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1491 struct in6_addr in6;
1492
1493 /*
1494 * To avoid over-reaction, we only apply this logic when we are
1495 * very sure that hardware addresses are supposed to be unique.
1496 */
1497 switch (ifp->if_type) {
1498 case IFT_ETHER:
1499 case IFT_ATM:
1500 case IFT_IEEE1394:
1501 case IFT_INFINIBAND:
1502 in6 = ia->ia_addr.sin6_addr;
1503 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1504 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1505 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1506 log(LOG_ERR, "%s: possible hardware address "
1507 "duplication detected, disable IPv6\n",
1508 if_name(ifp));
1509 }
1510 break;
1511 }
1512 }
1513 }
1514
1515 /*
1516 * Transmit a neighbour solicitation for the purpose of DAD. Returns with the
1517 * DAD queue unlocked.
1518 */
1519 static void
1520 nd6_dad_ns_output(struct dadq *dp)
1521 {
1522 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1523 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1524 int i;
1525
1526 DADQ_WLOCK_ASSERT();
1527
1528 dp->dad_ns_tcount++;
1529 if ((ifp->if_flags & IFF_UP) == 0) {
1530 DADQ_WUNLOCK();
1531 return;
1532 }
1533 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1534 DADQ_WUNLOCK();
1535 return;
1536 }
1537
1538 dp->dad_ns_ocount++;
1539 if (V_dad_enhanced != 0) {
1540 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1541 dp->dad_nonce[i] = arc4random();
1542 /*
1543 * XXXHRS: Note that in the case that
1544 * DupAddrDetectTransmits > 1, multiple NS messages with
1545 * different nonces can be looped back in an unexpected
1546 * order. The current implementation recognizes only
1547 * the latest nonce on the sender side. Practically it
1548 * should work well in almost all cases.
1549 */
1550 }
1551 DADQ_WUNLOCK();
1552 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1553 (uint8_t *)&dp->dad_nonce[0]);
1554 }
1555
1556 static void
1557 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1558 {
1559 struct dadq *dp;
1560
1561 if (ifa == NULL)
1562 panic("ifa == NULL in nd6_dad_ns_input");
1563
1564 /* Ignore Nonce option when Enhanced DAD is disabled. */
1565 if (V_dad_enhanced == 0)
1566 ndopt_nonce = NULL;
1567 DADQ_RLOCK();
1568 dp = nd6_dad_find(ifa, ndopt_nonce);
1569 if (dp != NULL)
1570 dp->dad_ns_icount++;
1571 DADQ_RUNLOCK();
1572 }
1573
1574 static void
1575 nd6_dad_na_input(struct ifaddr *ifa)
1576 {
1577 struct dadq *dp;
1578
1579 if (ifa == NULL)
1580 panic("ifa == NULL in nd6_dad_na_input");
1581
1582 DADQ_RLOCK();
1583 dp = nd6_dad_find(ifa, NULL);
1584 if (dp != NULL)
1585 dp->dad_na_icount++;
1586 DADQ_RUNLOCK();
1587 }
Cache object: 27448a96858edfa1991954ea176d9732
|