1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/11.2/sys/netinet6/nd6_nbr.c 332819 2018-04-20 14:21:37Z jtl $");
34
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #include <netinet6/nd6.h>
77 #include <netinet/icmp6.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet6/send.h>
80
81 #define SDL(s) ((struct sockaddr_dl *)s)
82
83 struct dadq;
84 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
85 static void nd6_dad_add(struct dadq *dp);
86 static void nd6_dad_del(struct dadq *dp);
87 static void nd6_dad_rele(struct dadq *);
88 static void nd6_dad_starttimer(struct dadq *, int, int);
89 static void nd6_dad_stoptimer(struct dadq *);
90 static void nd6_dad_timer(struct dadq *);
91 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
92 static void nd6_dad_ns_output(struct dadq *);
93 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
94 static void nd6_dad_na_input(struct ifaddr *);
95 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
96 const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
97 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
98 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
99
100 static VNET_DEFINE(int, dad_enhanced) = 1;
101 #define V_dad_enhanced VNET(dad_enhanced)
102
103 SYSCTL_DECL(_net_inet6_ip6);
104 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
105 &VNET_NAME(dad_enhanced), 0,
106 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
107
108 static VNET_DEFINE(int, dad_maxtry) = 15; /* max # of *tries* to
109 transmit DAD packet */
110 #define V_dad_maxtry VNET(dad_maxtry)
111
112 /*
113 * Input a Neighbor Solicitation Message.
114 *
115 * Based on RFC 2461
116 * Based on RFC 2462 (duplicate address detection)
117 */
118 void
119 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
120 {
121 struct ifnet *ifp = m->m_pkthdr.rcvif;
122 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 struct nd_neighbor_solicit *nd_ns;
124 struct in6_addr saddr6 = ip6->ip6_src;
125 struct in6_addr daddr6 = ip6->ip6_dst;
126 struct in6_addr taddr6;
127 struct in6_addr myaddr6;
128 char *lladdr = NULL;
129 struct ifaddr *ifa = NULL;
130 int lladdrlen = 0;
131 int anycast = 0, proxy = 0, tentative = 0;
132 int tlladdr;
133 int rflag;
134 union nd_opts ndopts;
135 struct sockaddr_dl proxydl;
136 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
137
138 /* RFC 6980: Nodes MUST silently ignore fragments */
139 if(m->m_flags & M_FRAGMENTED)
140 goto freeit;
141
142 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
143 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
144 rflag = 0;
145 #ifndef PULLDOWN_TEST
146 IP6_EXTHDR_CHECK(m, off, icmp6len,);
147 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
148 #else
149 IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
150 if (nd_ns == NULL) {
151 ICMP6STAT_INC(icp6s_tooshort);
152 return;
153 }
154 #endif
155 ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
156 taddr6 = nd_ns->nd_ns_target;
157 if (in6_setscope(&taddr6, ifp, NULL) != 0)
158 goto bad;
159
160 if (ip6->ip6_hlim != 255) {
161 nd6log((LOG_ERR,
162 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
163 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
164 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
165 goto bad;
166 }
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 if (ndopts.nd_opts_src_lladdr) {
216 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
217 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
218 }
219
220 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
221 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
222 "(link-layer address option)\n"));
223 goto bad;
224 }
225
226 /*
227 * Attaching target link-layer address to the NA?
228 * (RFC 2461 7.2.4)
229 *
230 * NS IP dst is unicast/anycast MUST NOT add
231 * NS IP dst is solicited-node multicast MUST add
232 *
233 * In implementation, we add target link-layer address by default.
234 * We do not add one in MUST NOT cases.
235 */
236 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
237 tlladdr = 0;
238 else
239 tlladdr = 1;
240
241 /*
242 * Target address (taddr6) must be either:
243 * (1) Valid unicast/anycast address for my receiving interface,
244 * (2) Unicast address for which I'm offering proxy service, or
245 * (3) "tentative" address on which DAD is being performed.
246 */
247 /* (1) and (3) check. */
248 if (ifp->if_carp)
249 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
250 else
251 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
252
253 /* (2) check. */
254 if (ifa == NULL) {
255 struct sockaddr_dl rt_gateway;
256 struct rt_addrinfo info;
257 struct sockaddr_in6 dst6;
258
259 bzero(&dst6, sizeof(dst6));
260 dst6.sin6_len = sizeof(struct sockaddr_in6);
261 dst6.sin6_family = AF_INET6;
262 dst6.sin6_addr = taddr6;
263
264 bzero(&rt_gateway, sizeof(rt_gateway));
265 rt_gateway.sdl_len = sizeof(rt_gateway);
266 bzero(&info, sizeof(info));
267 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
268
269 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
270 0, 0, &info) == 0) {
271 if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
272 rt_gateway.sdl_family == AF_LINK) {
273
274 /*
275 * proxy NDP for single entry
276 */
277 proxydl = *SDL(&rt_gateway);
278 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
279 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
280 if (ifa)
281 proxy = 1;
282 }
283 }
284 }
285 if (ifa == NULL) {
286 /*
287 * We've got an NS packet, and we don't have that adddress
288 * assigned for us. We MUST silently ignore it.
289 * See RFC2461 7.2.3.
290 */
291 goto freeit;
292 }
293 myaddr6 = *IFA_IN6(ifa);
294 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
295 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
296 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
297 goto freeit;
298
299 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
300 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
301 "(if %d, NS packet %d)\n",
302 ip6_sprintf(ip6bufs, &taddr6),
303 ifp->if_addrlen, lladdrlen - 2));
304 goto bad;
305 }
306
307 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
308 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
309 ip6_sprintf(ip6bufs, &saddr6)));
310 goto freeit;
311 }
312
313 /*
314 * We have neighbor solicitation packet, with target address equals to
315 * one of my tentative address.
316 *
317 * src addr how to process?
318 * --- ---
319 * multicast of course, invalid (rejected in ip6_input)
320 * unicast somebody is doing address resolution -> ignore
321 * unspec dup address detection
322 *
323 * The processing is defined in RFC 2462.
324 */
325 if (tentative) {
326 /*
327 * If source address is unspecified address, it is for
328 * duplicate address detection.
329 *
330 * If not, the packet is for addess resolution;
331 * silently ignore it.
332 */
333 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
334 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
335
336 goto freeit;
337 }
338
339 /*
340 * If the source address is unspecified address, entries must not
341 * be created or updated.
342 * It looks that sender is performing DAD. Output NA toward
343 * all-node multicast address, to tell the sender that I'm using
344 * the address.
345 * S bit ("solicited") must be zero.
346 */
347 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
348 struct in6_addr in6_all;
349
350 in6_all = in6addr_linklocal_allnodes;
351 if (in6_setscope(&in6_all, ifp, NULL) != 0)
352 goto bad;
353 nd6_na_output_fib(ifp, &in6_all, &taddr6,
354 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
356 M_GETFIB(m));
357 goto freeit;
358 }
359
360 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
361 ND_NEIGHBOR_SOLICIT, 0);
362
363 nd6_na_output_fib(ifp, &saddr6, &taddr6,
364 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
365 rflag | ND_NA_FLAG_SOLICITED, tlladdr,
366 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
367 freeit:
368 if (ifa != NULL)
369 ifa_free(ifa);
370 m_freem(m);
371 return;
372
373 bad:
374 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
375 ip6_sprintf(ip6bufs, &saddr6)));
376 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
377 ip6_sprintf(ip6bufs, &daddr6)));
378 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
379 ip6_sprintf(ip6bufs, &taddr6)));
380 ICMP6STAT_INC(icp6s_badns);
381 if (ifa != NULL)
382 ifa_free(ifa);
383 m_freem(m);
384 }
385
386 /*
387 * Output a Neighbor Solicitation Message. Caller specifies:
388 * - ICMP6 header source IP6 address
389 * - ND6 header target IP6 address
390 * - ND6 header source datalink address
391 *
392 * Based on RFC 2461
393 * Based on RFC 2462 (duplicate address detection)
394 *
395 * ln - for source address determination
396 * nonce - If non-NULL, NS is used for duplicate address detection and
397 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
398 */
399 static void
400 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
401 const struct in6_addr *daddr6, const struct in6_addr *taddr6,
402 uint8_t *nonce, u_int fibnum)
403 {
404 struct mbuf *m;
405 struct m_tag *mtag;
406 struct ip6_hdr *ip6;
407 struct nd_neighbor_solicit *nd_ns;
408 struct ip6_moptions im6o;
409 int icmp6len;
410 int maxlen;
411 caddr_t mac;
412
413 if (IN6_IS_ADDR_MULTICAST(taddr6))
414 return;
415
416 /* estimate the size of message */
417 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
418 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
419 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
420 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
421 __func__, max_linkhdr, maxlen, MCLBYTES));
422
423 if (max_linkhdr + maxlen > MHLEN)
424 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
425 else
426 m = m_gethdr(M_NOWAIT, MT_DATA);
427 if (m == NULL)
428 return;
429 M_SETFIB(m, fibnum);
430
431 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
432 m->m_flags |= M_MCAST;
433 im6o.im6o_multicast_ifp = ifp;
434 im6o.im6o_multicast_hlim = 255;
435 im6o.im6o_multicast_loop = 0;
436 }
437
438 icmp6len = sizeof(*nd_ns);
439 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
440 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
441
442 /* fill neighbor solicitation packet */
443 ip6 = mtod(m, struct ip6_hdr *);
444 ip6->ip6_flow = 0;
445 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
446 ip6->ip6_vfc |= IPV6_VERSION;
447 /* ip6->ip6_plen will be set later */
448 ip6->ip6_nxt = IPPROTO_ICMPV6;
449 ip6->ip6_hlim = 255;
450 if (daddr6)
451 ip6->ip6_dst = *daddr6;
452 else {
453 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
454 ip6->ip6_dst.s6_addr16[1] = 0;
455 ip6->ip6_dst.s6_addr32[1] = 0;
456 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
457 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
458 ip6->ip6_dst.s6_addr8[12] = 0xff;
459 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
460 goto bad;
461 }
462 if (nonce == NULL) {
463 struct ifaddr *ifa = NULL;
464
465 /*
466 * RFC2461 7.2.2:
467 * "If the source address of the packet prompting the
468 * solicitation is the same as one of the addresses assigned
469 * to the outgoing interface, that address SHOULD be placed
470 * in the IP Source Address of the outgoing solicitation.
471 * Otherwise, any one of the addresses assigned to the
472 * interface should be used."
473 *
474 * We use the source address for the prompting packet
475 * (saddr6), if saddr6 belongs to the outgoing interface.
476 * Otherwise, we perform the source address selection as usual.
477 */
478
479 if (saddr6 != NULL)
480 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
481 if (ifa != NULL) {
482 /* ip6_src set already. */
483 ip6->ip6_src = *saddr6;
484 ifa_free(ifa);
485 } else {
486 int error;
487 struct in6_addr dst6, src6;
488 uint32_t scopeid;
489
490 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
491 error = in6_selectsrc_addr(fibnum, &dst6,
492 scopeid, ifp, &src6, NULL);
493 if (error) {
494 char ip6buf[INET6_ADDRSTRLEN];
495 nd6log((LOG_DEBUG, "%s: source can't be "
496 "determined: dst=%s, error=%d\n", __func__,
497 ip6_sprintf(ip6buf, &dst6),
498 error));
499 goto bad;
500 }
501 ip6->ip6_src = src6;
502 }
503 } else {
504 /*
505 * Source address for DAD packet must always be IPv6
506 * unspecified address. (0::0)
507 * We actually don't have to 0-clear the address (we did it
508 * above), but we do so here explicitly to make the intention
509 * clearer.
510 */
511 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
512 }
513 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
514 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
515 nd_ns->nd_ns_code = 0;
516 nd_ns->nd_ns_reserved = 0;
517 nd_ns->nd_ns_target = *taddr6;
518 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
519
520 /*
521 * Add source link-layer address option.
522 *
523 * spec implementation
524 * --- ---
525 * DAD packet MUST NOT do not add the option
526 * there's no link layer address:
527 * impossible do not add the option
528 * there's link layer address:
529 * Multicast NS MUST add one add the option
530 * Unicast NS SHOULD add one add the option
531 */
532 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
533 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
534 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
535 /* 8 byte alignments... */
536 optlen = (optlen + 7) & ~7;
537
538 m->m_pkthdr.len += optlen;
539 m->m_len += optlen;
540 icmp6len += optlen;
541 bzero((caddr_t)nd_opt, optlen);
542 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
543 nd_opt->nd_opt_len = optlen >> 3;
544 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
545 }
546 /*
547 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
548 * This behavior is documented as Enhanced Duplicate Address
549 * Detection in RFC 7527.
550 * net.inet6.ip6.dad_enhanced=0 disables this.
551 */
552 if (V_dad_enhanced != 0 && nonce != NULL) {
553 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
554 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
555 /* 8-byte alignment is required. */
556 optlen = (optlen + 7) & ~7;
557
558 m->m_pkthdr.len += optlen;
559 m->m_len += optlen;
560 icmp6len += optlen;
561 bzero((caddr_t)nd_opt, optlen);
562 nd_opt->nd_opt_type = ND_OPT_NONCE;
563 nd_opt->nd_opt_len = optlen >> 3;
564 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
565 }
566 ip6->ip6_plen = htons((u_short)icmp6len);
567 nd_ns->nd_ns_cksum = 0;
568 nd_ns->nd_ns_cksum =
569 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
570
571 if (send_sendso_input_hook != NULL) {
572 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
573 sizeof(unsigned short), M_NOWAIT);
574 if (mtag == NULL)
575 goto bad;
576 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
577 m_tag_prepend(m, mtag);
578 }
579
580 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
581 &im6o, NULL, NULL);
582 icmp6_ifstat_inc(ifp, ifs6_out_msg);
583 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
584 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
585
586 return;
587
588 bad:
589 m_freem(m);
590 }
591
592 #ifndef BURN_BRIDGES
593 void
594 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
595 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
596 {
597
598 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
599 }
600 #endif
601 /*
602 * Neighbor advertisement input handling.
603 *
604 * Based on RFC 2461
605 * Based on RFC 2462 (duplicate address detection)
606 *
607 * the following items are not implemented yet:
608 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
609 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
610 */
611 void
612 nd6_na_input(struct mbuf *m, int off, int icmp6len)
613 {
614 struct ifnet *ifp = m->m_pkthdr.rcvif;
615 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
616 struct nd_neighbor_advert *nd_na;
617 struct in6_addr daddr6 = ip6->ip6_dst;
618 struct in6_addr taddr6;
619 int flags;
620 int is_router;
621 int is_solicited;
622 int is_override;
623 char *lladdr = NULL;
624 int lladdrlen = 0;
625 int checklink = 0;
626 struct ifaddr *ifa;
627 struct llentry *ln = NULL;
628 union nd_opts ndopts;
629 struct mbuf *chain = NULL;
630 struct sockaddr_in6 sin6;
631 u_char linkhdr[LLE_MAX_LINKHDR];
632 size_t linkhdrsize;
633 int lladdr_off;
634 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
635
636 /* RFC 6980: Nodes MUST silently ignore fragments */
637 if(m->m_flags & M_FRAGMENTED)
638 goto freeit;
639
640 if (ip6->ip6_hlim != 255) {
641 nd6log((LOG_ERR,
642 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
643 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
644 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
645 goto bad;
646 }
647
648 #ifndef PULLDOWN_TEST
649 IP6_EXTHDR_CHECK(m, off, icmp6len,);
650 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
651 #else
652 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
653 if (nd_na == NULL) {
654 ICMP6STAT_INC(icp6s_tooshort);
655 return;
656 }
657 #endif
658
659 flags = nd_na->nd_na_flags_reserved;
660 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
661 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
662 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
663 memset(&sin6, 0, sizeof(sin6));
664
665 taddr6 = nd_na->nd_na_target;
666 if (in6_setscope(&taddr6, ifp, NULL))
667 goto bad; /* XXX: impossible */
668
669 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
670 nd6log((LOG_ERR,
671 "nd6_na_input: invalid target address %s\n",
672 ip6_sprintf(ip6bufs, &taddr6)));
673 goto bad;
674 }
675 if (IN6_IS_ADDR_MULTICAST(&daddr6))
676 if (is_solicited) {
677 nd6log((LOG_ERR,
678 "nd6_na_input: a solicited adv is multicasted\n"));
679 goto bad;
680 }
681
682 icmp6len -= sizeof(*nd_na);
683 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
684 if (nd6_options(&ndopts) < 0) {
685 nd6log((LOG_INFO,
686 "nd6_na_input: invalid ND option, ignored\n"));
687 /* nd6_options have incremented stats */
688 goto freeit;
689 }
690
691 if (ndopts.nd_opts_tgt_lladdr) {
692 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
693 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
694 }
695
696 /*
697 * This effectively disables the DAD check on a non-master CARP
698 * address.
699 */
700 if (ifp->if_carp)
701 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
702 else
703 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
704
705 /*
706 * Target address matches one of my interface address.
707 *
708 * If my address is tentative, this means that there's somebody
709 * already using the same address as mine. This indicates DAD failure.
710 * This is defined in RFC 2462.
711 *
712 * Otherwise, process as defined in RFC 2461.
713 */
714 if (ifa
715 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
716 nd6_dad_na_input(ifa);
717 ifa_free(ifa);
718 goto freeit;
719 }
720
721 /* Just for safety, maybe unnecessary. */
722 if (ifa) {
723 ifa_free(ifa);
724 log(LOG_ERR,
725 "nd6_na_input: duplicate IP6 address %s\n",
726 ip6_sprintf(ip6bufs, &taddr6));
727 goto freeit;
728 }
729
730 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
731 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
732 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
733 ifp->if_addrlen, lladdrlen - 2));
734 goto bad;
735 }
736
737 /*
738 * If no neighbor cache entry is found, NA SHOULD silently be
739 * discarded.
740 */
741 IF_AFDATA_RLOCK(ifp);
742 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
743 IF_AFDATA_RUNLOCK(ifp);
744 if (ln == NULL) {
745 goto freeit;
746 }
747
748 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
749 /*
750 * If the link-layer has address, and no lladdr option came,
751 * discard the packet.
752 */
753 if (ifp->if_addrlen && lladdr == NULL) {
754 goto freeit;
755 }
756
757 /*
758 * Record link-layer address, and update the state.
759 */
760 linkhdrsize = sizeof(linkhdr);
761 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
762 linkhdr, &linkhdrsize, &lladdr_off) != 0)
763 return;
764
765 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
766 lladdr_off) == 0) {
767 ln = NULL;
768 goto freeit;
769 }
770 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
771 if (is_solicited)
772 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
773 else
774 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
775 if ((ln->ln_router = is_router) != 0) {
776 /*
777 * This means a router's state has changed from
778 * non-reachable to probably reachable, and might
779 * affect the status of associated prefixes..
780 */
781 checklink = 1;
782 }
783 } else {
784 int llchange;
785
786 /*
787 * Check if the link-layer address has changed or not.
788 */
789 if (lladdr == NULL)
790 llchange = 0;
791 else {
792 if (ln->la_flags & LLE_VALID) {
793 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
794 llchange = 1;
795 else
796 llchange = 0;
797 } else
798 llchange = 1;
799 }
800
801 /*
802 * This is VERY complex. Look at it with care.
803 *
804 * override solicit lladdr llchange action
805 * (L: record lladdr)
806 *
807 * 0 0 n -- (2c)
808 * 0 0 y n (2b) L
809 * 0 0 y y (1) REACHABLE->STALE
810 * 0 1 n -- (2c) *->REACHABLE
811 * 0 1 y n (2b) L *->REACHABLE
812 * 0 1 y y (1) REACHABLE->STALE
813 * 1 0 n -- (2a)
814 * 1 0 y n (2a) L
815 * 1 0 y y (2a) L *->STALE
816 * 1 1 n -- (2a) *->REACHABLE
817 * 1 1 y n (2a) L *->REACHABLE
818 * 1 1 y y (2a) L *->REACHABLE
819 */
820 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
821 /*
822 * If state is REACHABLE, make it STALE.
823 * no other updates should be done.
824 */
825 if (ln->ln_state == ND6_LLINFO_REACHABLE)
826 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
827 goto freeit;
828 } else if (is_override /* (2a) */
829 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
830 || lladdr == NULL) { /* (2c) */
831 /*
832 * Update link-local address, if any.
833 */
834 if (lladdr != NULL) {
835 linkhdrsize = sizeof(linkhdr);
836 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
837 linkhdr, &linkhdrsize, &lladdr_off) != 0)
838 goto freeit;
839 if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
840 linkhdrsize, lladdr_off) == 0) {
841 ln = NULL;
842 goto freeit;
843 }
844 EVENTHANDLER_INVOKE(lle_event, ln,
845 LLENTRY_RESOLVED);
846 }
847
848 /*
849 * If solicited, make the state REACHABLE.
850 * If not solicited and the link-layer address was
851 * changed, make it STALE.
852 */
853 if (is_solicited)
854 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
855 else {
856 if (lladdr != NULL && llchange)
857 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
858 }
859 }
860
861 if (ln->ln_router && !is_router) {
862 /*
863 * The peer dropped the router flag.
864 * Remove the sender from the Default Router List and
865 * update the Destination Cache entries.
866 */
867 struct ifnet *nd6_ifp;
868
869 nd6_ifp = lltable_get_ifp(ln->lle_tbl);
870 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
871 (ND_IFINFO(nd6_ifp)->flags &
872 ND6_IFF_ACCEPT_RTADV) != 0)
873 /*
874 * Even if the neighbor is not in the default
875 * router list, the neighbor may be used as a
876 * next hop for some destinations (e.g. redirect
877 * case). So we must call rt6_flush explicitly.
878 */
879 rt6_flush(&ip6->ip6_src, ifp);
880 }
881 ln->ln_router = is_router;
882 }
883 /* XXX - QL
884 * Does this matter?
885 * rt->rt_flags &= ~RTF_REJECT;
886 */
887 ln->la_asked = 0;
888 if (ln->la_hold != NULL)
889 nd6_grab_holdchain(ln, &chain, &sin6);
890 freeit:
891 if (ln != NULL)
892 LLE_WUNLOCK(ln);
893
894 if (chain != NULL)
895 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
896
897 if (checklink)
898 pfxlist_onlink_check();
899
900 m_freem(m);
901 return;
902
903 bad:
904 if (ln != NULL)
905 LLE_WUNLOCK(ln);
906
907 ICMP6STAT_INC(icp6s_badna);
908 m_freem(m);
909 }
910
911 /*
912 * Neighbor advertisement output handling.
913 *
914 * Based on RFC 2461
915 *
916 * the following items are not implemented yet:
917 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
918 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
919 *
920 * tlladdr - 1 if include target link-layer address
921 * sdl0 - sockaddr_dl (= proxy NA) or NULL
922 */
923 static void
924 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
925 const struct in6_addr *taddr6, u_long flags, int tlladdr,
926 struct sockaddr *sdl0, u_int fibnum)
927 {
928 struct mbuf *m;
929 struct m_tag *mtag;
930 struct ip6_hdr *ip6;
931 struct nd_neighbor_advert *nd_na;
932 struct ip6_moptions im6o;
933 struct in6_addr daddr6, dst6, src6;
934 uint32_t scopeid;
935
936 int icmp6len, maxlen, error;
937 caddr_t mac = NULL;
938
939 daddr6 = *daddr6_0; /* make a local copy for modification */
940
941 /* estimate the size of message */
942 maxlen = sizeof(*ip6) + sizeof(*nd_na);
943 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
944 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
945 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
946 __func__, max_linkhdr, maxlen, MCLBYTES));
947
948 if (max_linkhdr + maxlen > MHLEN)
949 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
950 else
951 m = m_gethdr(M_NOWAIT, MT_DATA);
952 if (m == NULL)
953 return;
954 M_SETFIB(m, fibnum);
955
956 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
957 m->m_flags |= M_MCAST;
958 im6o.im6o_multicast_ifp = ifp;
959 im6o.im6o_multicast_hlim = 255;
960 im6o.im6o_multicast_loop = 0;
961 }
962
963 icmp6len = sizeof(*nd_na);
964 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
965 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
966
967 /* fill neighbor advertisement packet */
968 ip6 = mtod(m, struct ip6_hdr *);
969 ip6->ip6_flow = 0;
970 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
971 ip6->ip6_vfc |= IPV6_VERSION;
972 ip6->ip6_nxt = IPPROTO_ICMPV6;
973 ip6->ip6_hlim = 255;
974 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
975 /* reply to DAD */
976 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
977 daddr6.s6_addr16[1] = 0;
978 daddr6.s6_addr32[1] = 0;
979 daddr6.s6_addr32[2] = 0;
980 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
981 if (in6_setscope(&daddr6, ifp, NULL))
982 goto bad;
983
984 flags &= ~ND_NA_FLAG_SOLICITED;
985 }
986 ip6->ip6_dst = daddr6;
987
988 /*
989 * Select a source whose scope is the same as that of the dest.
990 */
991 in6_splitscope(&daddr6, &dst6, &scopeid);
992 error = in6_selectsrc_addr(fibnum, &dst6,
993 scopeid, ifp, &src6, NULL);
994 if (error) {
995 char ip6buf[INET6_ADDRSTRLEN];
996 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
997 "determined: dst=%s, error=%d\n",
998 ip6_sprintf(ip6buf, &daddr6), error));
999 goto bad;
1000 }
1001 ip6->ip6_src = src6;
1002 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1003 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1004 nd_na->nd_na_code = 0;
1005 nd_na->nd_na_target = *taddr6;
1006 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1007
1008 /*
1009 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1010 * see nd6_ns_input() for details.
1011 * Basically, if NS packet is sent to unicast/anycast addr,
1012 * target lladdr option SHOULD NOT be included.
1013 */
1014 if (tlladdr) {
1015 /*
1016 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1017 * lladdr in sdl0. If we are not proxying (sending NA for
1018 * my address) use lladdr configured for the interface.
1019 */
1020 if (sdl0 == NULL) {
1021 if (ifp->if_carp)
1022 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1023 if (mac == NULL)
1024 mac = nd6_ifptomac(ifp);
1025 } else if (sdl0->sa_family == AF_LINK) {
1026 struct sockaddr_dl *sdl;
1027 sdl = (struct sockaddr_dl *)sdl0;
1028 if (sdl->sdl_alen == ifp->if_addrlen)
1029 mac = LLADDR(sdl);
1030 }
1031 }
1032 if (tlladdr && mac) {
1033 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1034 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1035
1036 /* roundup to 8 bytes alignment! */
1037 optlen = (optlen + 7) & ~7;
1038
1039 m->m_pkthdr.len += optlen;
1040 m->m_len += optlen;
1041 icmp6len += optlen;
1042 bzero((caddr_t)nd_opt, optlen);
1043 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1044 nd_opt->nd_opt_len = optlen >> 3;
1045 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1046 } else
1047 flags &= ~ND_NA_FLAG_OVERRIDE;
1048
1049 ip6->ip6_plen = htons((u_short)icmp6len);
1050 nd_na->nd_na_flags_reserved = flags;
1051 nd_na->nd_na_cksum = 0;
1052 nd_na->nd_na_cksum =
1053 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1054
1055 if (send_sendso_input_hook != NULL) {
1056 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1057 sizeof(unsigned short), M_NOWAIT);
1058 if (mtag == NULL)
1059 goto bad;
1060 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1061 m_tag_prepend(m, mtag);
1062 }
1063
1064 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1065 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1066 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1067 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1068
1069 return;
1070
1071 bad:
1072 m_freem(m);
1073 }
1074
1075 #ifndef BURN_BRIDGES
1076 void
1077 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1078 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1079 struct sockaddr *sdl0)
1080 {
1081
1082 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1083 RT_DEFAULT_FIB);
1084 }
1085 #endif
1086
1087 caddr_t
1088 nd6_ifptomac(struct ifnet *ifp)
1089 {
1090 switch (ifp->if_type) {
1091 case IFT_ARCNET:
1092 case IFT_ETHER:
1093 case IFT_FDDI:
1094 case IFT_IEEE1394:
1095 case IFT_L2VLAN:
1096 case IFT_IEEE80211:
1097 case IFT_INFINIBAND:
1098 case IFT_BRIDGE:
1099 case IFT_ISO88025:
1100 return IF_LLADDR(ifp);
1101 default:
1102 return NULL;
1103 }
1104 }
1105
1106 struct dadq {
1107 TAILQ_ENTRY(dadq) dad_list;
1108 struct ifaddr *dad_ifa;
1109 int dad_count; /* max NS to send */
1110 int dad_ns_tcount; /* # of trials to send NS */
1111 int dad_ns_ocount; /* NS sent so far */
1112 int dad_ns_icount;
1113 int dad_na_icount;
1114 int dad_ns_lcount; /* looped back NS */
1115 int dad_loopbackprobe; /* probing state for loopback detection */
1116 struct callout dad_timer_ch;
1117 struct vnet *dad_vnet;
1118 u_int dad_refcnt;
1119 #define ND_OPT_NONCE_LEN32 \
1120 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1121 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1122 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */
1123 };
1124
1125 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1126 static VNET_DEFINE(struct rwlock, dad_rwlock);
1127 #define V_dadq VNET(dadq)
1128 #define V_dad_rwlock VNET(dad_rwlock)
1129
1130 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock)
1131 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock)
1132 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock)
1133 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock)
1134
1135 static void
1136 nd6_dad_add(struct dadq *dp)
1137 {
1138
1139 DADQ_WLOCK();
1140 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1141 dp->dad_ondadq = true;
1142 DADQ_WUNLOCK();
1143 }
1144
1145 static void
1146 nd6_dad_del(struct dadq *dp)
1147 {
1148
1149 DADQ_WLOCK();
1150 if (dp->dad_ondadq) {
1151 /*
1152 * Remove dp from the dadq and release the dadq's
1153 * reference.
1154 */
1155 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1156 dp->dad_ondadq = false;
1157 DADQ_WUNLOCK();
1158 nd6_dad_rele(dp);
1159 } else
1160 DADQ_WUNLOCK();
1161 }
1162
1163 static struct dadq *
1164 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1165 {
1166 struct dadq *dp;
1167
1168 DADQ_RLOCK();
1169 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1170 if (dp->dad_ifa != ifa)
1171 continue;
1172 /*
1173 * Skip if the nonce matches the received one.
1174 * +2 in the length is required because of type and
1175 * length fields are included in a header.
1176 */
1177 if (n != NULL &&
1178 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1179 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1180 ND_OPT_NONCE_LEN) == 0) {
1181 dp->dad_ns_lcount++;
1182 continue;
1183 }
1184 refcount_acquire(&dp->dad_refcnt);
1185 break;
1186 }
1187 DADQ_RUNLOCK();
1188
1189 return (dp);
1190 }
1191
1192 static void
1193 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1194 {
1195
1196 if (send_ns != 0)
1197 nd6_dad_ns_output(dp);
1198 callout_reset(&dp->dad_timer_ch, ticks,
1199 (void (*)(void *))nd6_dad_timer, (void *)dp);
1200 }
1201
1202 static void
1203 nd6_dad_stoptimer(struct dadq *dp)
1204 {
1205
1206 callout_drain(&dp->dad_timer_ch);
1207 }
1208
1209 static void
1210 nd6_dad_rele(struct dadq *dp)
1211 {
1212
1213 if (refcount_release(&dp->dad_refcnt)) {
1214 ifa_free(dp->dad_ifa);
1215 free(dp, M_IP6NDP);
1216 }
1217 }
1218
1219 void
1220 nd6_dad_init(void)
1221 {
1222
1223 rw_init(&V_dad_rwlock, "nd6 DAD queue");
1224 TAILQ_INIT(&V_dadq);
1225 }
1226
1227 /*
1228 * Start Duplicate Address Detection (DAD) for specified interface address.
1229 */
1230 void
1231 nd6_dad_start(struct ifaddr *ifa, int delay)
1232 {
1233 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1234 struct dadq *dp;
1235 char ip6buf[INET6_ADDRSTRLEN];
1236
1237 /*
1238 * If we don't need DAD, don't do it.
1239 * There are several cases:
1240 * - DAD is disabled (ip6_dad_count == 0)
1241 * - the interface address is anycast
1242 */
1243 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1244 log(LOG_DEBUG,
1245 "nd6_dad_start: called with non-tentative address "
1246 "%s(%s)\n",
1247 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1248 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1249 return;
1250 }
1251 if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1252 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1253 return;
1254 }
1255 if (!V_ip6_dad_count) {
1256 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1257 return;
1258 }
1259 if (ifa->ifa_ifp == NULL)
1260 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1261 if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) {
1262 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1263 return;
1264 }
1265 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1266 !(ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) ||
1267 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)) {
1268 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1269 return;
1270 }
1271 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1272 /*
1273 * DAD is already in progress. Let the existing entry
1274 * finish it.
1275 */
1276 nd6_dad_rele(dp);
1277 return;
1278 }
1279
1280 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1281 if (dp == NULL) {
1282 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1283 "%s(%s)\n",
1284 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1285 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1286 return;
1287 }
1288 callout_init(&dp->dad_timer_ch, 0);
1289 #ifdef VIMAGE
1290 dp->dad_vnet = curvnet;
1291 #endif
1292 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1293 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1294
1295 /*
1296 * Send NS packet for DAD, ip6_dad_count times.
1297 * Note that we must delay the first transmission, if this is the
1298 * first packet to be sent from the interface after interface
1299 * (re)initialization.
1300 */
1301 dp->dad_ifa = ifa;
1302 ifa_ref(dp->dad_ifa);
1303 dp->dad_count = V_ip6_dad_count;
1304 dp->dad_ns_icount = dp->dad_na_icount = 0;
1305 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1306 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1307
1308 /* Add this to the dadq and add a reference for the dadq. */
1309 refcount_init(&dp->dad_refcnt, 1);
1310 nd6_dad_add(dp);
1311 nd6_dad_starttimer(dp, delay, 0);
1312 }
1313
1314 /*
1315 * terminate DAD unconditionally. used for address removals.
1316 */
1317 void
1318 nd6_dad_stop(struct ifaddr *ifa)
1319 {
1320 struct dadq *dp;
1321
1322 dp = nd6_dad_find(ifa, NULL);
1323 if (!dp) {
1324 /* DAD wasn't started yet */
1325 return;
1326 }
1327
1328 nd6_dad_stoptimer(dp);
1329 nd6_dad_del(dp);
1330
1331 /* Release this function's reference, acquired by nd6_dad_find(). */
1332 nd6_dad_rele(dp);
1333 }
1334
1335 static void
1336 nd6_dad_timer(struct dadq *dp)
1337 {
1338 CURVNET_SET(dp->dad_vnet);
1339 struct ifaddr *ifa = dp->dad_ifa;
1340 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1341 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1342 char ip6buf[INET6_ADDRSTRLEN];
1343
1344 /* Sanity check */
1345 if (ia == NULL) {
1346 log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1347 goto err;
1348 }
1349 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1350 /* Do not need DAD for ifdisabled interface. */
1351 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1352 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1353 goto err;
1354 }
1355 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1356 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1357 "%s(%s)\n",
1358 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1359 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1360 goto err;
1361 }
1362 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1363 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1364 "%s(%s)\n",
1365 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1366 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1367 goto err;
1368 }
1369
1370 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1371 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1372 (((ifp->if_flags & IFF_UP) == 0) ||
1373 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1374 nd6log((LOG_INFO, "%s: could not run DAD "
1375 "because the interface was down or not running.\n",
1376 if_name(ifa->ifa_ifp)));
1377 goto err;
1378 }
1379
1380 /* Need more checks? */
1381 if (dp->dad_ns_ocount < dp->dad_count) {
1382 /*
1383 * We have more NS to go. Send NS packet for DAD.
1384 */
1385 nd6_dad_starttimer(dp,
1386 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1387 goto done;
1388 } else {
1389 /*
1390 * We have transmitted sufficient number of DAD packets.
1391 * See what we've got.
1392 */
1393 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1394 /* We've seen NS or NA, means DAD has failed. */
1395 nd6_dad_duplicated(ifa, dp);
1396 else if (V_dad_enhanced != 0 &&
1397 dp->dad_ns_lcount > 0 &&
1398 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1399 /*
1400 * Sec. 4.1 in RFC 7527 requires transmission of
1401 * additional probes until the loopback condition
1402 * becomes clear when a looped back probe is detected.
1403 */
1404 log(LOG_ERR, "%s: a looped back NS message is "
1405 "detected during DAD for %s. "
1406 "Another DAD probes are being sent.\n",
1407 if_name(ifa->ifa_ifp),
1408 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1409 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1410 /*
1411 * Send an NS immediately and increase dad_count by
1412 * V_nd6_mmaxtries - 1.
1413 */
1414 dp->dad_count =
1415 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1416 nd6_dad_starttimer(dp,
1417 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1418 1);
1419 goto done;
1420 } else {
1421 /*
1422 * We are done with DAD. No NA came, no NS came.
1423 * No duplicate address found. Check IFDISABLED flag
1424 * again in case that it is changed between the
1425 * beginning of this function and here.
1426 */
1427 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1428 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1429
1430 nd6log((LOG_DEBUG,
1431 "%s: DAD complete for %s - no duplicates found\n",
1432 if_name(ifa->ifa_ifp),
1433 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1434 if (dp->dad_ns_lcount > 0)
1435 log(LOG_ERR, "%s: DAD completed while "
1436 "a looped back NS message is detected "
1437 "during DAD for %s.\n",
1438 if_name(ifa->ifa_ifp),
1439 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1440 }
1441 }
1442 err:
1443 nd6_dad_del(dp);
1444 done:
1445 CURVNET_RESTORE();
1446 }
1447
1448 static void
1449 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1450 {
1451 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1452 struct ifnet *ifp;
1453 char ip6buf[INET6_ADDRSTRLEN];
1454
1455 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1456 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1457 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1458 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1459 dp->dad_na_icount);
1460
1461 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1462 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1463
1464 ifp = ifa->ifa_ifp;
1465 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1466 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1467 log(LOG_ERR, "%s: manual intervention required\n",
1468 if_name(ifp));
1469
1470 /*
1471 * If the address is a link-local address formed from an interface
1472 * identifier based on the hardware address which is supposed to be
1473 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1474 * operation on the interface SHOULD be disabled.
1475 * [RFC 4862, Section 5.4.5]
1476 */
1477 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1478 struct in6_addr in6;
1479
1480 /*
1481 * To avoid over-reaction, we only apply this logic when we are
1482 * very sure that hardware addresses are supposed to be unique.
1483 */
1484 switch (ifp->if_type) {
1485 case IFT_ETHER:
1486 case IFT_FDDI:
1487 case IFT_ATM:
1488 case IFT_IEEE1394:
1489 case IFT_IEEE80211:
1490 case IFT_INFINIBAND:
1491 in6 = ia->ia_addr.sin6_addr;
1492 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1493 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1494 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1495 log(LOG_ERR, "%s: possible hardware address "
1496 "duplication detected, disable IPv6\n",
1497 if_name(ifp));
1498 }
1499 break;
1500 }
1501 }
1502 }
1503
1504 static void
1505 nd6_dad_ns_output(struct dadq *dp)
1506 {
1507 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1508 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1509 int i;
1510
1511 dp->dad_ns_tcount++;
1512 if ((ifp->if_flags & IFF_UP) == 0) {
1513 return;
1514 }
1515 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1516 return;
1517 }
1518
1519 dp->dad_ns_ocount++;
1520 if (V_dad_enhanced != 0) {
1521 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1522 dp->dad_nonce[i] = arc4random();
1523 /*
1524 * XXXHRS: Note that in the case that
1525 * DupAddrDetectTransmits > 1, multiple NS messages with
1526 * different nonces can be looped back in an unexpected
1527 * order. The current implementation recognizes only
1528 * the latest nonce on the sender side. Practically it
1529 * should work well in almost all cases.
1530 */
1531 }
1532 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1533 (uint8_t *)&dp->dad_nonce[0]);
1534 }
1535
1536 static void
1537 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1538 {
1539 struct in6_ifaddr *ia;
1540 struct ifnet *ifp;
1541 const struct in6_addr *taddr6;
1542 struct dadq *dp;
1543
1544 if (ifa == NULL)
1545 panic("ifa == NULL in nd6_dad_ns_input");
1546
1547 ia = (struct in6_ifaddr *)ifa;
1548 ifp = ifa->ifa_ifp;
1549 taddr6 = &ia->ia_addr.sin6_addr;
1550 /* Ignore Nonce option when Enhanced DAD is disabled. */
1551 if (V_dad_enhanced == 0)
1552 ndopt_nonce = NULL;
1553 dp = nd6_dad_find(ifa, ndopt_nonce);
1554 if (dp == NULL)
1555 return;
1556
1557 dp->dad_ns_icount++;
1558 nd6_dad_rele(dp);
1559 }
1560
1561 static void
1562 nd6_dad_na_input(struct ifaddr *ifa)
1563 {
1564 struct dadq *dp;
1565
1566 if (ifa == NULL)
1567 panic("ifa == NULL in nd6_dad_na_input");
1568
1569 dp = nd6_dad_find(ifa, NULL);
1570 if (dp != NULL) {
1571 dp->dad_na_icount++;
1572 nd6_dad_rele(dp);
1573 }
1574 }
Cache object: e3dbb9ef423b0a0b5e10404d55c6acd9
|