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 #include "opt_mpath.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.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 #ifdef RADIX_MPATH
66 #include <net/radix_mpath.h>
67 #endif
68 #include <net/vnet.h>
69
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <net/if_llatbl.h>
73 #include <netinet6/in6_var.h>
74 #include <netinet6/in6_ifattach.h>
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 #include <netinet6/scope6_var.h>
78 #include <netinet6/nd6.h>
79 #include <netinet/icmp6.h>
80 #include <netinet/ip_carp.h>
81 #include <netinet6/send.h>
82
83 #define SDL(s) ((struct sockaddr_dl *)s)
84
85 struct dadq;
86 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
87 static void nd6_dad_add(struct dadq *dp);
88 static void nd6_dad_del(struct dadq *dp);
89 static void nd6_dad_rele(struct dadq *);
90 static void nd6_dad_starttimer(struct dadq *, int, int);
91 static void nd6_dad_stoptimer(struct dadq *);
92 static void nd6_dad_timer(struct dadq *);
93 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
94 static void nd6_dad_ns_output(struct dadq *);
95 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
96 static void nd6_dad_na_input(struct ifaddr *);
97 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
98 const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
99 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
100 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
101
102 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
103 #define V_dad_enhanced VNET(dad_enhanced)
104
105 SYSCTL_DECL(_net_inet6_ip6);
106 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
107 &VNET_NAME(dad_enhanced), 0,
108 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
109
110 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to
111 transmit DAD packet */
112 #define V_dad_maxtry VNET(dad_maxtry)
113
114 /*
115 * Input a Neighbor Solicitation Message.
116 *
117 * Based on RFC 2461
118 * Based on RFC 2462 (duplicate address detection)
119 */
120 void
121 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
122 {
123 struct ifnet *ifp;
124 struct ip6_hdr *ip6;
125 struct nd_neighbor_solicit *nd_ns;
126 struct in6_addr daddr6, myaddr6, saddr6, taddr6;
127 struct ifaddr *ifa;
128 struct sockaddr_dl proxydl;
129 union nd_opts ndopts;
130 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
131 char *lladdr;
132 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
133
134 ifa = NULL;
135
136 /* RFC 6980: Nodes MUST silently ignore fragments */
137 if(m->m_flags & M_FRAGMENTED)
138 goto freeit;
139
140 ifp = m->m_pkthdr.rcvif;
141 ip6 = mtod(m, struct ip6_hdr *);
142 if (ip6->ip6_hlim != 255) {
143 nd6log((LOG_ERR,
144 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
145 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
146 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
147 goto bads;
148 }
149
150 if (m->m_len < off + icmp6len) {
151 m = m_pullup(m, off + icmp6len);
152 if (m == NULL) {
153 IP6STAT_INC(ip6s_exthdrtoolong);
154 return;
155 }
156 }
157 ip6 = mtod(m, struct ip6_hdr *);
158 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
159
160 saddr6 = ip6->ip6_src;
161 daddr6 = ip6->ip6_dst;
162 taddr6 = nd_ns->nd_ns_target;
163 if (in6_setscope(&taddr6, ifp, NULL) != 0)
164 goto bad;
165
166 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
167 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
168 rflag = 0;
169
170 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
171 /* dst has to be a solicited node multicast address. */
172 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
173 /* don't check ifindex portion */
174 daddr6.s6_addr32[1] == 0 &&
175 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
176 daddr6.s6_addr8[12] == 0xff) {
177 ; /* good */
178 } else {
179 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
180 "(wrong ip6 dst)\n"));
181 goto bad;
182 }
183 } else if (!V_nd6_onlink_ns_rfc4861) {
184 struct sockaddr_in6 src_sa6;
185
186 /*
187 * According to recent IETF discussions, it is not a good idea
188 * to accept a NS from an address which would not be deemed
189 * to be a neighbor otherwise. This point is expected to be
190 * clarified in future revisions of the specification.
191 */
192 bzero(&src_sa6, sizeof(src_sa6));
193 src_sa6.sin6_family = AF_INET6;
194 src_sa6.sin6_len = sizeof(src_sa6);
195 src_sa6.sin6_addr = saddr6;
196 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
197 nd6log((LOG_INFO, "nd6_ns_input: "
198 "NS packet from non-neighbor\n"));
199 goto bad;
200 }
201 }
202
203 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
204 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
205 goto bad;
206 }
207
208 icmp6len -= sizeof(*nd_ns);
209 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
210 if (nd6_options(&ndopts) < 0) {
211 nd6log((LOG_INFO,
212 "nd6_ns_input: invalid ND option, ignored\n"));
213 /* nd6_options have incremented stats */
214 goto freeit;
215 }
216
217 lladdr = NULL;
218 lladdrlen = 0;
219 if (ndopts.nd_opts_src_lladdr) {
220 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
221 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
222 }
223
224 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
225 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
226 "(link-layer address option)\n"));
227 goto bad;
228 }
229
230 /*
231 * Attaching target link-layer address to the NA?
232 * (RFC 2461 7.2.4)
233 *
234 * NS IP dst is unicast/anycast MUST NOT add
235 * NS IP dst is solicited-node multicast MUST add
236 *
237 * In implementation, we add target link-layer address by default.
238 * We do not add one in MUST NOT cases.
239 */
240 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
241 tlladdr = 0;
242 else
243 tlladdr = 1;
244
245 /*
246 * Target address (taddr6) must be either:
247 * (1) Valid unicast/anycast address for my receiving interface,
248 * (2) Unicast address for which I'm offering proxy service, or
249 * (3) "tentative" address on which DAD is being performed.
250 */
251 /* (1) and (3) check. */
252 if (ifp->if_carp)
253 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
254 else
255 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
256
257 /* (2) check. */
258 proxy = 0;
259 if (ifa == NULL) {
260 struct sockaddr_dl rt_gateway;
261 struct rt_addrinfo info;
262 struct sockaddr_in6 dst6;
263
264 bzero(&dst6, sizeof(dst6));
265 dst6.sin6_len = sizeof(struct sockaddr_in6);
266 dst6.sin6_family = AF_INET6;
267 dst6.sin6_addr = taddr6;
268
269 bzero(&rt_gateway, sizeof(rt_gateway));
270 rt_gateway.sdl_len = sizeof(rt_gateway);
271 bzero(&info, sizeof(info));
272 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
273
274 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
275 0, 0, &info) == 0) {
276 if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
277 rt_gateway.sdl_family == AF_LINK) {
278
279 /*
280 * proxy NDP for single entry
281 */
282 proxydl = *SDL(&rt_gateway);
283 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
284 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
285 if (ifa)
286 proxy = 1;
287 }
288 }
289 }
290 if (ifa == NULL) {
291 /*
292 * We've got an NS packet, and we don't have that adddress
293 * assigned for us. We MUST silently ignore it.
294 * See RFC2461 7.2.3.
295 */
296 goto freeit;
297 }
298 myaddr6 = *IFA_IN6(ifa);
299 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
300 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
301 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
302 goto freeit;
303
304 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
305 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
306 "(if %d, NS packet %d)\n",
307 ip6_sprintf(ip6bufs, &taddr6),
308 ifp->if_addrlen, lladdrlen - 2));
309 goto bad;
310 }
311
312 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
313 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
314 ip6_sprintf(ip6bufs, &saddr6)));
315 goto freeit;
316 }
317
318 /*
319 * We have neighbor solicitation packet, with target address equals to
320 * one of my tentative address.
321 *
322 * src addr how to process?
323 * --- ---
324 * multicast of course, invalid (rejected in ip6_input)
325 * unicast somebody is doing address resolution -> ignore
326 * unspec dup address detection
327 *
328 * The processing is defined in RFC 2462.
329 */
330 if (tentative) {
331 /*
332 * If source address is unspecified address, it is for
333 * duplicate address detection.
334 *
335 * If not, the packet is for addess resolution;
336 * silently ignore it.
337 */
338 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
339 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
340
341 goto freeit;
342 }
343
344 /*
345 * If the source address is unspecified address, entries must not
346 * be created or updated.
347 * It looks that sender is performing DAD. Output NA toward
348 * all-node multicast address, to tell the sender that I'm using
349 * the address.
350 * S bit ("solicited") must be zero.
351 */
352 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
353 struct in6_addr in6_all;
354
355 in6_all = in6addr_linklocal_allnodes;
356 if (in6_setscope(&in6_all, ifp, NULL) != 0)
357 goto bad;
358 nd6_na_output_fib(ifp, &in6_all, &taddr6,
359 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
360 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
361 M_GETFIB(m));
362 goto freeit;
363 }
364
365 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
366 ND_NEIGHBOR_SOLICIT, 0);
367
368 nd6_na_output_fib(ifp, &saddr6, &taddr6,
369 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
370 rflag | ND_NA_FLAG_SOLICITED, tlladdr,
371 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
372 freeit:
373 if (ifa != NULL)
374 ifa_free(ifa);
375 m_freem(m);
376 return;
377
378 bad:
379 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
380 ip6_sprintf(ip6bufs, &saddr6)));
381 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
382 ip6_sprintf(ip6bufs, &daddr6)));
383 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
384 ip6_sprintf(ip6bufs, &taddr6)));
385 bads:
386 ICMP6STAT_INC(icp6s_badns);
387 if (ifa != NULL)
388 ifa_free(ifa);
389 m_freem(m);
390 }
391
392 /*
393 * Output a Neighbor Solicitation Message. Caller specifies:
394 * - ICMP6 header source IP6 address
395 * - ND6 header target IP6 address
396 * - ND6 header source datalink address
397 *
398 * Based on RFC 2461
399 * Based on RFC 2462 (duplicate address detection)
400 *
401 * ln - for source address determination
402 * nonce - If non-NULL, NS is used for duplicate address detection and
403 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
404 */
405 static void
406 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
407 const struct in6_addr *daddr6, const struct in6_addr *taddr6,
408 uint8_t *nonce, u_int fibnum)
409 {
410 struct mbuf *m;
411 struct m_tag *mtag;
412 struct ip6_hdr *ip6;
413 struct nd_neighbor_solicit *nd_ns;
414 struct ip6_moptions im6o;
415 int icmp6len;
416 int maxlen;
417 caddr_t mac;
418
419 if (IN6_IS_ADDR_MULTICAST(taddr6))
420 return;
421
422 /* estimate the size of message */
423 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
424 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
425 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
426 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
427 __func__, max_linkhdr, maxlen, MCLBYTES));
428
429 if (max_linkhdr + maxlen > MHLEN)
430 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
431 else
432 m = m_gethdr(M_NOWAIT, MT_DATA);
433 if (m == NULL)
434 return;
435 M_SETFIB(m, fibnum);
436
437 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
438 m->m_flags |= M_MCAST;
439 im6o.im6o_multicast_ifp = ifp;
440 im6o.im6o_multicast_hlim = 255;
441 im6o.im6o_multicast_loop = 0;
442 }
443
444 icmp6len = sizeof(*nd_ns);
445 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
446 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
447
448 /* fill neighbor solicitation packet */
449 ip6 = mtod(m, struct ip6_hdr *);
450 ip6->ip6_flow = 0;
451 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
452 ip6->ip6_vfc |= IPV6_VERSION;
453 /* ip6->ip6_plen will be set later */
454 ip6->ip6_nxt = IPPROTO_ICMPV6;
455 ip6->ip6_hlim = 255;
456 if (daddr6)
457 ip6->ip6_dst = *daddr6;
458 else {
459 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
460 ip6->ip6_dst.s6_addr16[1] = 0;
461 ip6->ip6_dst.s6_addr32[1] = 0;
462 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
463 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
464 ip6->ip6_dst.s6_addr8[12] = 0xff;
465 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
466 goto bad;
467 }
468 if (nonce == NULL) {
469 struct ifaddr *ifa = NULL;
470
471 /*
472 * RFC2461 7.2.2:
473 * "If the source address of the packet prompting the
474 * solicitation is the same as one of the addresses assigned
475 * to the outgoing interface, that address SHOULD be placed
476 * in the IP Source Address of the outgoing solicitation.
477 * Otherwise, any one of the addresses assigned to the
478 * interface should be used."
479 *
480 * We use the source address for the prompting packet
481 * (saddr6), if saddr6 belongs to the outgoing interface.
482 * Otherwise, we perform the source address selection as usual.
483 */
484
485 if (saddr6 != NULL)
486 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
487 if (ifa != NULL) {
488 /* ip6_src set already. */
489 ip6->ip6_src = *saddr6;
490 ifa_free(ifa);
491 } else {
492 int error;
493 struct in6_addr dst6, src6;
494 uint32_t scopeid;
495
496 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
497 error = in6_selectsrc_addr(fibnum, &dst6,
498 scopeid, ifp, &src6, NULL);
499 if (error) {
500 char ip6buf[INET6_ADDRSTRLEN];
501 nd6log((LOG_DEBUG, "%s: source can't be "
502 "determined: dst=%s, error=%d\n", __func__,
503 ip6_sprintf(ip6buf, &dst6),
504 error));
505 goto bad;
506 }
507 ip6->ip6_src = src6;
508 }
509 } else {
510 /*
511 * Source address for DAD packet must always be IPv6
512 * unspecified address. (0::0)
513 * We actually don't have to 0-clear the address (we did it
514 * above), but we do so here explicitly to make the intention
515 * clearer.
516 */
517 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
518 }
519 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
520 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
521 nd_ns->nd_ns_code = 0;
522 nd_ns->nd_ns_reserved = 0;
523 nd_ns->nd_ns_target = *taddr6;
524 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
525
526 /*
527 * Add source link-layer address option.
528 *
529 * spec implementation
530 * --- ---
531 * DAD packet MUST NOT do not add the option
532 * there's no link layer address:
533 * impossible do not add the option
534 * there's link layer address:
535 * Multicast NS MUST add one add the option
536 * Unicast NS SHOULD add one add the option
537 */
538 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
539 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
540 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
541 /* 8 byte alignments... */
542 optlen = (optlen + 7) & ~7;
543
544 m->m_pkthdr.len += optlen;
545 m->m_len += optlen;
546 icmp6len += optlen;
547 bzero((caddr_t)nd_opt, optlen);
548 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
549 nd_opt->nd_opt_len = optlen >> 3;
550 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
551 }
552 /*
553 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
554 * This behavior is documented as Enhanced Duplicate Address
555 * Detection in RFC 7527.
556 * net.inet6.ip6.dad_enhanced=0 disables this.
557 */
558 if (V_dad_enhanced != 0 && nonce != NULL) {
559 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
560 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
561 /* 8-byte alignment is required. */
562 optlen = (optlen + 7) & ~7;
563
564 m->m_pkthdr.len += optlen;
565 m->m_len += optlen;
566 icmp6len += optlen;
567 bzero((caddr_t)nd_opt, optlen);
568 nd_opt->nd_opt_type = ND_OPT_NONCE;
569 nd_opt->nd_opt_len = optlen >> 3;
570 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
571 }
572 ip6->ip6_plen = htons((u_short)icmp6len);
573 nd_ns->nd_ns_cksum = 0;
574 nd_ns->nd_ns_cksum =
575 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
576
577 if (send_sendso_input_hook != NULL) {
578 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
579 sizeof(unsigned short), M_NOWAIT);
580 if (mtag == NULL)
581 goto bad;
582 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
583 m_tag_prepend(m, mtag);
584 }
585
586 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
587 &im6o, NULL, NULL);
588 icmp6_ifstat_inc(ifp, ifs6_out_msg);
589 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
590 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
591
592 return;
593
594 bad:
595 m_freem(m);
596 }
597
598 #ifndef BURN_BRIDGES
599 void
600 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
601 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
602 {
603
604 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
605 }
606 #endif
607 /*
608 * Neighbor advertisement input handling.
609 *
610 * Based on RFC 2461
611 * Based on RFC 2462 (duplicate address detection)
612 *
613 * the following items are not implemented yet:
614 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
615 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
616 */
617 void
618 nd6_na_input(struct mbuf *m, int off, int icmp6len)
619 {
620 struct ifnet *ifp;
621 struct ip6_hdr *ip6;
622 struct ifaddr *ifa;
623 struct llentry *ln;
624 struct mbuf *chain;
625 struct nd_neighbor_advert *nd_na;
626 struct in6_addr daddr6, taddr6;
627 struct sockaddr_in6 sin6;
628 union nd_opts ndopts;
629 u_char linkhdr[LLE_MAX_LINKHDR];
630 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
631 char *lladdr;
632 size_t linkhdrsize;
633 int flags, is_override, is_router, is_solicited;
634 int lladdr_off, lladdrlen, checklink;
635
636 chain = NULL;
637 ln = NULL;
638 checklink = 0;
639
640 /* RFC 6980: Nodes MUST silently ignore fragments */
641 if(m->m_flags & M_FRAGMENTED)
642 goto freeit;
643
644 ifp = m->m_pkthdr.rcvif;
645 ip6 = mtod(m, struct ip6_hdr *);
646 if (ip6->ip6_hlim != 255) {
647 nd6log((LOG_ERR,
648 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
649 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
650 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
651 goto bad;
652 }
653
654 if (m->m_len < off + icmp6len) {
655 m = m_pullup(m, off + icmp6len);
656 if (m == NULL) {
657 IP6STAT_INC(ip6s_exthdrtoolong);
658 return;
659 }
660 }
661 ip6 = mtod(m, struct ip6_hdr *);
662 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
663
664 flags = nd_na->nd_na_flags_reserved;
665 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
666 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
667 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
668
669 taddr6 = nd_na->nd_na_target;
670 if (in6_setscope(&taddr6, ifp, NULL))
671 goto bad; /* XXX: impossible */
672
673 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
674 nd6log((LOG_ERR,
675 "nd6_na_input: invalid target address %s\n",
676 ip6_sprintf(ip6bufs, &taddr6)));
677 goto bad;
678 }
679
680 daddr6 = ip6->ip6_dst;
681 if (IN6_IS_ADDR_MULTICAST(&daddr6))
682 if (is_solicited) {
683 nd6log((LOG_ERR,
684 "nd6_na_input: a solicited adv is multicasted\n"));
685 goto bad;
686 }
687
688 icmp6len -= sizeof(*nd_na);
689 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
690 if (nd6_options(&ndopts) < 0) {
691 nd6log((LOG_INFO,
692 "nd6_na_input: invalid ND option, ignored\n"));
693 /* nd6_options have incremented stats */
694 goto freeit;
695 }
696
697 lladdr = NULL;
698 lladdrlen = 0;
699 if (ndopts.nd_opts_tgt_lladdr) {
700 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
701 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
702 }
703
704 /*
705 * This effectively disables the DAD check on a non-master CARP
706 * address.
707 */
708 if (ifp->if_carp)
709 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
710 else
711 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
712
713 /*
714 * Target address matches one of my interface address.
715 *
716 * If my address is tentative, this means that there's somebody
717 * already using the same address as mine. This indicates DAD failure.
718 * This is defined in RFC 2462.
719 *
720 * Otherwise, process as defined in RFC 2461.
721 */
722 if (ifa
723 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
724 nd6_dad_na_input(ifa);
725 ifa_free(ifa);
726 goto freeit;
727 }
728
729 /* Just for safety, maybe unnecessary. */
730 if (ifa) {
731 ifa_free(ifa);
732 log(LOG_ERR,
733 "nd6_na_input: duplicate IP6 address %s\n",
734 ip6_sprintf(ip6bufs, &taddr6));
735 goto freeit;
736 }
737
738 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
739 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
740 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
741 ifp->if_addrlen, lladdrlen - 2));
742 goto bad;
743 }
744
745 /*
746 * If no neighbor cache entry is found, NA SHOULD silently be
747 * discarded.
748 */
749 IF_AFDATA_RLOCK(ifp);
750 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
751 IF_AFDATA_RUNLOCK(ifp);
752 if (ln == NULL) {
753 goto freeit;
754 }
755
756 /*
757 * Do not try to override static entry.
758 */
759 if (ln->la_flags & LLE_STATIC)
760 goto freeit;
761
762 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
763 /*
764 * If the link-layer has address, and no lladdr option came,
765 * discard the packet.
766 */
767 if (ifp->if_addrlen && lladdr == NULL) {
768 goto freeit;
769 }
770
771 /*
772 * Record link-layer address, and update the state.
773 */
774 linkhdrsize = sizeof(linkhdr);
775 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
776 linkhdr, &linkhdrsize, &lladdr_off) != 0)
777 return;
778
779 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
780 lladdr_off) == 0) {
781 ln = NULL;
782 goto freeit;
783 }
784 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
785 if (is_solicited)
786 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
787 else
788 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
789 if ((ln->ln_router = is_router) != 0) {
790 /*
791 * This means a router's state has changed from
792 * non-reachable to probably reachable, and might
793 * affect the status of associated prefixes..
794 */
795 checklink = 1;
796 }
797 } else {
798 int llchange;
799
800 /*
801 * Check if the link-layer address has changed or not.
802 */
803 if (lladdr == NULL)
804 llchange = 0;
805 else {
806 if (ln->la_flags & LLE_VALID) {
807 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
808 llchange = 1;
809 else
810 llchange = 0;
811 } else
812 llchange = 1;
813 }
814
815 /*
816 * This is VERY complex. Look at it with care.
817 *
818 * override solicit lladdr llchange action
819 * (L: record lladdr)
820 *
821 * 0 0 n -- (2c)
822 * 0 0 y n (2b) L
823 * 0 0 y y (1) REACHABLE->STALE
824 * 0 1 n -- (2c) *->REACHABLE
825 * 0 1 y n (2b) L *->REACHABLE
826 * 0 1 y y (1) REACHABLE->STALE
827 * 1 0 n -- (2a)
828 * 1 0 y n (2a) L
829 * 1 0 y y (2a) L *->STALE
830 * 1 1 n -- (2a) *->REACHABLE
831 * 1 1 y n (2a) L *->REACHABLE
832 * 1 1 y y (2a) L *->REACHABLE
833 */
834 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
835 /*
836 * If state is REACHABLE, make it STALE.
837 * no other updates should be done.
838 */
839 if (ln->ln_state == ND6_LLINFO_REACHABLE)
840 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
841 goto freeit;
842 } else if (is_override /* (2a) */
843 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
844 || lladdr == NULL) { /* (2c) */
845 /*
846 * Update link-local address, if any.
847 */
848 if (lladdr != NULL) {
849 linkhdrsize = sizeof(linkhdr);
850 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
851 linkhdr, &linkhdrsize, &lladdr_off) != 0)
852 goto freeit;
853 if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
854 linkhdrsize, lladdr_off) == 0) {
855 ln = NULL;
856 goto freeit;
857 }
858 EVENTHANDLER_INVOKE(lle_event, ln,
859 LLENTRY_RESOLVED);
860 }
861
862 /*
863 * If solicited, make the state REACHABLE.
864 * If not solicited and the link-layer address was
865 * changed, make it STALE.
866 */
867 if (is_solicited)
868 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
869 else {
870 if (lladdr != NULL && llchange)
871 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
872 }
873 }
874
875 if (ln->ln_router && !is_router) {
876 /*
877 * The peer dropped the router flag.
878 * Remove the sender from the Default Router List and
879 * update the Destination Cache entries.
880 */
881 struct ifnet *nd6_ifp;
882
883 nd6_ifp = lltable_get_ifp(ln->lle_tbl);
884 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
885 (ND_IFINFO(nd6_ifp)->flags &
886 ND6_IFF_ACCEPT_RTADV) != 0)
887 /*
888 * Even if the neighbor is not in the default
889 * router list, the neighbor may be used as a
890 * next hop for some destinations (e.g. redirect
891 * case). So we must call rt6_flush explicitly.
892 */
893 rt6_flush(&ip6->ip6_src, ifp);
894 }
895 ln->ln_router = is_router;
896 }
897 /* XXX - QL
898 * Does this matter?
899 * rt->rt_flags &= ~RTF_REJECT;
900 */
901 ln->la_asked = 0;
902 if (ln->la_hold != NULL) {
903 memset(&sin6, 0, sizeof(sin6));
904 nd6_grab_holdchain(ln, &chain, &sin6);
905 }
906 freeit:
907 if (ln != NULL)
908 LLE_WUNLOCK(ln);
909
910 if (chain != NULL)
911 nd6_flush_holdchain(ifp, chain, &sin6);
912
913 if (checklink)
914 pfxlist_onlink_check();
915
916 m_freem(m);
917 return;
918
919 bad:
920 if (ln != NULL)
921 LLE_WUNLOCK(ln);
922
923 ICMP6STAT_INC(icp6s_badna);
924 m_freem(m);
925 }
926
927 /*
928 * Neighbor advertisement output handling.
929 *
930 * Based on RFC 2461
931 *
932 * the following items are not implemented yet:
933 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
934 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
935 *
936 * tlladdr - 1 if include target link-layer address
937 * sdl0 - sockaddr_dl (= proxy NA) or NULL
938 */
939 static void
940 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
941 const struct in6_addr *taddr6, u_long flags, int tlladdr,
942 struct sockaddr *sdl0, u_int fibnum)
943 {
944 struct mbuf *m;
945 struct m_tag *mtag;
946 struct ip6_hdr *ip6;
947 struct nd_neighbor_advert *nd_na;
948 struct ip6_moptions im6o;
949 struct in6_addr daddr6, dst6, src6;
950 uint32_t scopeid;
951
952 int icmp6len, maxlen, error;
953 caddr_t mac = NULL;
954
955 daddr6 = *daddr6_0; /* make a local copy for modification */
956
957 /* estimate the size of message */
958 maxlen = sizeof(*ip6) + sizeof(*nd_na);
959 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
960 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
961 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
962 __func__, max_linkhdr, maxlen, MCLBYTES));
963
964 if (max_linkhdr + maxlen > MHLEN)
965 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
966 else
967 m = m_gethdr(M_NOWAIT, MT_DATA);
968 if (m == NULL)
969 return;
970 M_SETFIB(m, fibnum);
971
972 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
973 m->m_flags |= M_MCAST;
974 im6o.im6o_multicast_ifp = ifp;
975 im6o.im6o_multicast_hlim = 255;
976 im6o.im6o_multicast_loop = 0;
977 }
978
979 icmp6len = sizeof(*nd_na);
980 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
981 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
982
983 /* fill neighbor advertisement packet */
984 ip6 = mtod(m, struct ip6_hdr *);
985 ip6->ip6_flow = 0;
986 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
987 ip6->ip6_vfc |= IPV6_VERSION;
988 ip6->ip6_nxt = IPPROTO_ICMPV6;
989 ip6->ip6_hlim = 255;
990 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
991 /* reply to DAD */
992 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
993 daddr6.s6_addr16[1] = 0;
994 daddr6.s6_addr32[1] = 0;
995 daddr6.s6_addr32[2] = 0;
996 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
997 if (in6_setscope(&daddr6, ifp, NULL))
998 goto bad;
999
1000 flags &= ~ND_NA_FLAG_SOLICITED;
1001 }
1002 ip6->ip6_dst = daddr6;
1003
1004 /*
1005 * Select a source whose scope is the same as that of the dest.
1006 */
1007 in6_splitscope(&daddr6, &dst6, &scopeid);
1008 error = in6_selectsrc_addr(fibnum, &dst6,
1009 scopeid, ifp, &src6, NULL);
1010 if (error) {
1011 char ip6buf[INET6_ADDRSTRLEN];
1012 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1013 "determined: dst=%s, error=%d\n",
1014 ip6_sprintf(ip6buf, &daddr6), error));
1015 goto bad;
1016 }
1017 ip6->ip6_src = src6;
1018 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1019 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1020 nd_na->nd_na_code = 0;
1021 nd_na->nd_na_target = *taddr6;
1022 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1023
1024 /*
1025 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1026 * see nd6_ns_input() for details.
1027 * Basically, if NS packet is sent to unicast/anycast addr,
1028 * target lladdr option SHOULD NOT be included.
1029 */
1030 if (tlladdr) {
1031 /*
1032 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1033 * lladdr in sdl0. If we are not proxying (sending NA for
1034 * my address) use lladdr configured for the interface.
1035 */
1036 if (sdl0 == NULL) {
1037 if (ifp->if_carp)
1038 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1039 if (mac == NULL)
1040 mac = nd6_ifptomac(ifp);
1041 } else if (sdl0->sa_family == AF_LINK) {
1042 struct sockaddr_dl *sdl;
1043 sdl = (struct sockaddr_dl *)sdl0;
1044 if (sdl->sdl_alen == ifp->if_addrlen)
1045 mac = LLADDR(sdl);
1046 }
1047 }
1048 if (tlladdr && mac) {
1049 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1050 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1051
1052 /* roundup to 8 bytes alignment! */
1053 optlen = (optlen + 7) & ~7;
1054
1055 m->m_pkthdr.len += optlen;
1056 m->m_len += optlen;
1057 icmp6len += optlen;
1058 bzero((caddr_t)nd_opt, optlen);
1059 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1060 nd_opt->nd_opt_len = optlen >> 3;
1061 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1062 } else
1063 flags &= ~ND_NA_FLAG_OVERRIDE;
1064
1065 ip6->ip6_plen = htons((u_short)icmp6len);
1066 nd_na->nd_na_flags_reserved = flags;
1067 nd_na->nd_na_cksum = 0;
1068 nd_na->nd_na_cksum =
1069 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1070
1071 if (send_sendso_input_hook != NULL) {
1072 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1073 sizeof(unsigned short), M_NOWAIT);
1074 if (mtag == NULL)
1075 goto bad;
1076 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1077 m_tag_prepend(m, mtag);
1078 }
1079
1080 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1081 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1082 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1083 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1084
1085 return;
1086
1087 bad:
1088 m_freem(m);
1089 }
1090
1091 #ifndef BURN_BRIDGES
1092 void
1093 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1094 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1095 struct sockaddr *sdl0)
1096 {
1097
1098 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1099 RT_DEFAULT_FIB);
1100 }
1101 #endif
1102
1103 caddr_t
1104 nd6_ifptomac(struct ifnet *ifp)
1105 {
1106 switch (ifp->if_type) {
1107 case IFT_ETHER:
1108 case IFT_IEEE1394:
1109 case IFT_L2VLAN:
1110 case IFT_INFINIBAND:
1111 case IFT_BRIDGE:
1112 return IF_LLADDR(ifp);
1113 default:
1114 return NULL;
1115 }
1116 }
1117
1118 struct dadq {
1119 TAILQ_ENTRY(dadq) dad_list;
1120 struct ifaddr *dad_ifa;
1121 int dad_count; /* max NS to send */
1122 int dad_ns_tcount; /* # of trials to send NS */
1123 int dad_ns_ocount; /* NS sent so far */
1124 int dad_ns_icount;
1125 int dad_na_icount;
1126 int dad_ns_lcount; /* looped back NS */
1127 int dad_loopbackprobe; /* probing state for loopback detection */
1128 struct callout dad_timer_ch;
1129 struct vnet *dad_vnet;
1130 u_int dad_refcnt;
1131 #define ND_OPT_NONCE_LEN32 \
1132 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1133 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1134 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */
1135 };
1136
1137 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1138 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1139 #define V_dadq VNET(dadq)
1140 #define V_dad_rwlock VNET(dad_rwlock)
1141
1142 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock)
1143 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock)
1144 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock)
1145 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock)
1146
1147 static void
1148 nd6_dad_add(struct dadq *dp)
1149 {
1150
1151 DADQ_WLOCK();
1152 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1153 dp->dad_ondadq = true;
1154 DADQ_WUNLOCK();
1155 }
1156
1157 static void
1158 nd6_dad_del(struct dadq *dp)
1159 {
1160
1161 DADQ_WLOCK();
1162 if (dp->dad_ondadq) {
1163 /*
1164 * Remove dp from the dadq and release the dadq's
1165 * reference.
1166 */
1167 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1168 dp->dad_ondadq = false;
1169 DADQ_WUNLOCK();
1170 nd6_dad_rele(dp);
1171 } else
1172 DADQ_WUNLOCK();
1173 }
1174
1175 static struct dadq *
1176 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1177 {
1178 struct dadq *dp;
1179
1180 DADQ_RLOCK();
1181 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1182 if (dp->dad_ifa != ifa)
1183 continue;
1184 /*
1185 * Skip if the nonce matches the received one.
1186 * +2 in the length is required because of type and
1187 * length fields are included in a header.
1188 */
1189 if (n != NULL &&
1190 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1191 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1192 ND_OPT_NONCE_LEN) == 0) {
1193 dp->dad_ns_lcount++;
1194 continue;
1195 }
1196 refcount_acquire(&dp->dad_refcnt);
1197 break;
1198 }
1199 DADQ_RUNLOCK();
1200
1201 return (dp);
1202 }
1203
1204 static void
1205 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1206 {
1207
1208 if (send_ns != 0)
1209 nd6_dad_ns_output(dp);
1210 callout_reset(&dp->dad_timer_ch, ticks,
1211 (void (*)(void *))nd6_dad_timer, (void *)dp);
1212 }
1213
1214 static void
1215 nd6_dad_stoptimer(struct dadq *dp)
1216 {
1217
1218 callout_drain(&dp->dad_timer_ch);
1219 }
1220
1221 static void
1222 nd6_dad_rele(struct dadq *dp)
1223 {
1224
1225 if (refcount_release(&dp->dad_refcnt)) {
1226 ifa_free(dp->dad_ifa);
1227 free(dp, M_IP6NDP);
1228 }
1229 }
1230
1231 void
1232 nd6_dad_init(void)
1233 {
1234
1235 rw_init(&V_dad_rwlock, "nd6 DAD queue");
1236 TAILQ_INIT(&V_dadq);
1237 }
1238
1239 /*
1240 * Start Duplicate Address Detection (DAD) for specified interface address.
1241 */
1242 void
1243 nd6_dad_start(struct ifaddr *ifa, int delay)
1244 {
1245 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1246 struct dadq *dp;
1247 char ip6buf[INET6_ADDRSTRLEN];
1248
1249 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1250 ("starting DAD on non-tentative address %p", ifa));
1251
1252 /*
1253 * If we don't need DAD, don't do it.
1254 * There are several cases:
1255 * - DAD is disabled globally or on the interface
1256 * - the interface address is anycast
1257 */
1258 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1259 V_ip6_dad_count == 0 ||
1260 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1261 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1262 return;
1263 }
1264 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1265 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1266 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1267 return;
1268
1269 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1270 /*
1271 * DAD is already in progress. Let the existing entry
1272 * finish it.
1273 */
1274 nd6_dad_rele(dp);
1275 return;
1276 }
1277
1278 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1279 if (dp == NULL) {
1280 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1281 "%s(%s)\n",
1282 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1283 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1284 return;
1285 }
1286 callout_init(&dp->dad_timer_ch, 0);
1287 #ifdef VIMAGE
1288 dp->dad_vnet = curvnet;
1289 #endif
1290 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1291 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1292
1293 /*
1294 * Send NS packet for DAD, ip6_dad_count times.
1295 * Note that we must delay the first transmission, if this is the
1296 * first packet to be sent from the interface after interface
1297 * (re)initialization.
1298 */
1299 dp->dad_ifa = ifa;
1300 ifa_ref(dp->dad_ifa);
1301 dp->dad_count = V_ip6_dad_count;
1302 dp->dad_ns_icount = dp->dad_na_icount = 0;
1303 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1304 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1305
1306 /* Add this to the dadq and add a reference for the dadq. */
1307 refcount_init(&dp->dad_refcnt, 1);
1308 nd6_dad_add(dp);
1309 nd6_dad_starttimer(dp, delay, 0);
1310 }
1311
1312 /*
1313 * terminate DAD unconditionally. used for address removals.
1314 */
1315 void
1316 nd6_dad_stop(struct ifaddr *ifa)
1317 {
1318 struct dadq *dp;
1319
1320 dp = nd6_dad_find(ifa, NULL);
1321 if (!dp) {
1322 /* DAD wasn't started yet */
1323 return;
1324 }
1325
1326 nd6_dad_stoptimer(dp);
1327 nd6_dad_del(dp);
1328
1329 /* Release this function's reference, acquired by nd6_dad_find(). */
1330 nd6_dad_rele(dp);
1331 }
1332
1333 static void
1334 nd6_dad_timer(struct dadq *dp)
1335 {
1336 CURVNET_SET(dp->dad_vnet);
1337 struct ifaddr *ifa = dp->dad_ifa;
1338 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1339 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1340 char ip6buf[INET6_ADDRSTRLEN];
1341
1342 KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1343
1344 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1345 /* Do not need DAD for ifdisabled interface. */
1346 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1347 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1348 goto err;
1349 }
1350 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1351 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1352 "%s(%s)\n",
1353 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1354 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1355 goto err;
1356 }
1357 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1358 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1359 "%s(%s)\n",
1360 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1361 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1362 goto err;
1363 }
1364
1365 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1366 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1367 (((ifp->if_flags & IFF_UP) == 0) ||
1368 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1369 nd6log((LOG_INFO, "%s: could not run DAD "
1370 "because the interface was down or not running.\n",
1371 if_name(ifa->ifa_ifp)));
1372 goto err;
1373 }
1374
1375 /* Need more checks? */
1376 if (dp->dad_ns_ocount < dp->dad_count) {
1377 /*
1378 * We have more NS to go. Send NS packet for DAD.
1379 */
1380 nd6_dad_starttimer(dp,
1381 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1382 goto done;
1383 } else {
1384 /*
1385 * We have transmitted sufficient number of DAD packets.
1386 * See what we've got.
1387 */
1388 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1389 /* We've seen NS or NA, means DAD has failed. */
1390 nd6_dad_duplicated(ifa, dp);
1391 else if (V_dad_enhanced != 0 &&
1392 dp->dad_ns_lcount > 0 &&
1393 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1394 /*
1395 * Sec. 4.1 in RFC 7527 requires transmission of
1396 * additional probes until the loopback condition
1397 * becomes clear when a looped back probe is detected.
1398 */
1399 log(LOG_ERR, "%s: a looped back NS message is "
1400 "detected during DAD for %s. "
1401 "Another DAD probes are being sent.\n",
1402 if_name(ifa->ifa_ifp),
1403 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1404 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1405 /*
1406 * Send an NS immediately and increase dad_count by
1407 * V_nd6_mmaxtries - 1.
1408 */
1409 dp->dad_count =
1410 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1411 nd6_dad_starttimer(dp,
1412 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1413 1);
1414 goto done;
1415 } else {
1416 /*
1417 * We are done with DAD. No NA came, no NS came.
1418 * No duplicate address found. Check IFDISABLED flag
1419 * again in case that it is changed between the
1420 * beginning of this function and here.
1421 */
1422 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1423 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1424
1425 nd6log((LOG_DEBUG,
1426 "%s: DAD complete for %s - no duplicates found\n",
1427 if_name(ifa->ifa_ifp),
1428 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1429 if (dp->dad_ns_lcount > 0)
1430 log(LOG_ERR, "%s: DAD completed while "
1431 "a looped back NS message is detected "
1432 "during DAD for %s.\n",
1433 if_name(ifa->ifa_ifp),
1434 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1435 }
1436 }
1437 err:
1438 nd6_dad_del(dp);
1439 done:
1440 CURVNET_RESTORE();
1441 }
1442
1443 static void
1444 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1445 {
1446 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1447 struct ifnet *ifp;
1448 char ip6buf[INET6_ADDRSTRLEN];
1449
1450 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1451 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1452 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1453 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1454 dp->dad_na_icount);
1455
1456 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1457 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1458
1459 ifp = ifa->ifa_ifp;
1460 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1461 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1462 log(LOG_ERR, "%s: manual intervention required\n",
1463 if_name(ifp));
1464
1465 /*
1466 * If the address is a link-local address formed from an interface
1467 * identifier based on the hardware address which is supposed to be
1468 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1469 * operation on the interface SHOULD be disabled.
1470 * [RFC 4862, Section 5.4.5]
1471 */
1472 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1473 struct in6_addr in6;
1474
1475 /*
1476 * To avoid over-reaction, we only apply this logic when we are
1477 * very sure that hardware addresses are supposed to be unique.
1478 */
1479 switch (ifp->if_type) {
1480 case IFT_ETHER:
1481 case IFT_ATM:
1482 case IFT_IEEE1394:
1483 case IFT_INFINIBAND:
1484 in6 = ia->ia_addr.sin6_addr;
1485 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1486 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1487 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1488 log(LOG_ERR, "%s: possible hardware address "
1489 "duplication detected, disable IPv6\n",
1490 if_name(ifp));
1491 }
1492 break;
1493 }
1494 }
1495 }
1496
1497 static void
1498 nd6_dad_ns_output(struct dadq *dp)
1499 {
1500 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1501 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1502 int i;
1503
1504 dp->dad_ns_tcount++;
1505 if ((ifp->if_flags & IFF_UP) == 0) {
1506 return;
1507 }
1508 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1509 return;
1510 }
1511
1512 dp->dad_ns_ocount++;
1513 if (V_dad_enhanced != 0) {
1514 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1515 dp->dad_nonce[i] = arc4random();
1516 /*
1517 * XXXHRS: Note that in the case that
1518 * DupAddrDetectTransmits > 1, multiple NS messages with
1519 * different nonces can be looped back in an unexpected
1520 * order. The current implementation recognizes only
1521 * the latest nonce on the sender side. Practically it
1522 * should work well in almost all cases.
1523 */
1524 }
1525 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1526 (uint8_t *)&dp->dad_nonce[0]);
1527 }
1528
1529 static void
1530 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1531 {
1532 struct dadq *dp;
1533
1534 if (ifa == NULL)
1535 panic("ifa == NULL in nd6_dad_ns_input");
1536
1537 /* Ignore Nonce option when Enhanced DAD is disabled. */
1538 if (V_dad_enhanced == 0)
1539 ndopt_nonce = NULL;
1540 dp = nd6_dad_find(ifa, ndopt_nonce);
1541 if (dp == NULL)
1542 return;
1543
1544 dp->dad_ns_icount++;
1545 nd6_dad_rele(dp);
1546 }
1547
1548 static void
1549 nd6_dad_na_input(struct ifaddr *ifa)
1550 {
1551 struct dadq *dp;
1552
1553 if (ifa == NULL)
1554 panic("ifa == NULL in nd6_dad_na_input");
1555
1556 dp = nd6_dad_find(ifa, NULL);
1557 if (dp != NULL) {
1558 dp->dad_na_icount++;
1559 nd6_dad_rele(dp);
1560 }
1561 }
Cache object: 7eea84aae5050fdedb8403b35f52430d
|