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 "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_ipsec.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/counter.h>
41 #include <sys/eventhandler.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/if_private.h>
63 #include <net/route.h>
64 #include <net/vnet.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet/icmp6.h>
76 #include <netinet/ip_carp.h>
77 #include <netinet6/send.h>
78
79 #include <machine/atomic.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);
89 static void nd6_dad_stoptimer(struct dadq *);
90 static void nd6_dad_timer(void *);
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 struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *,
101 const struct in6_addr *, struct sockaddr_dl *);
102
103 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
104 #define V_dad_enhanced VNET(dad_enhanced)
105
106 SYSCTL_DECL(_net_inet6_ip6);
107 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
108 &VNET_NAME(dad_enhanced), 0,
109 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
110
111 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to
112 transmit DAD packet */
113 #define V_dad_maxtry VNET(dad_maxtry)
114
115 VNET_DEFINE_STATIC(int, nd6_onlink_ns_rfc4861) = 0;
116 #define V_nd6_onlink_ns_rfc4861 VNET(nd6_onlink_ns_rfc4861)
117 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_ONLINKNSRFC4861,
118 nd6_onlink_ns_rfc4861, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(nd6_onlink_ns_rfc4861), 0,
120 "Accept 'on-link' ICMPv6 NS messages in compliance with RFC 4861");
121
122 /*
123 * Input a Neighbor Solicitation Message.
124 *
125 * Based on RFC 2461
126 * Based on RFC 2462 (duplicate address detection)
127 */
128 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)129 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
130 {
131 struct ifnet *ifp;
132 struct ip6_hdr *ip6;
133 struct nd_neighbor_solicit *nd_ns;
134 struct in6_addr daddr6, myaddr6, saddr6, taddr6;
135 struct ifaddr *ifa;
136 struct sockaddr_dl proxydl;
137 union nd_opts ndopts;
138 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
139 char *lladdr;
140 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
141
142 ifa = NULL;
143
144 /* RFC 6980: Nodes MUST silently ignore fragments */
145 if(m->m_flags & M_FRAGMENTED)
146 goto freeit;
147
148 ifp = m->m_pkthdr.rcvif;
149 ip6 = mtod(m, struct ip6_hdr *);
150 if (__predict_false(ip6->ip6_hlim != 255)) {
151 ICMP6STAT_INC(icp6s_invlhlim);
152 nd6log((LOG_ERR,
153 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
154 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
155 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
156 goto bads;
157 }
158
159 if (m->m_len < off + icmp6len) {
160 m = m_pullup(m, off + icmp6len);
161 if (m == NULL) {
162 IP6STAT_INC(ip6s_exthdrtoolong);
163 return;
164 }
165 }
166 ip6 = mtod(m, struct ip6_hdr *);
167 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
168
169 saddr6 = ip6->ip6_src;
170 daddr6 = ip6->ip6_dst;
171 taddr6 = nd_ns->nd_ns_target;
172 if (in6_setscope(&taddr6, ifp, NULL) != 0)
173 goto bad;
174
175 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
176 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
177 rflag = 0;
178
179 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
180 /* dst has to be a solicited node multicast address. */
181 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
182 /* don't check ifindex portion */
183 daddr6.s6_addr32[1] == 0 &&
184 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
185 daddr6.s6_addr8[12] == 0xff) {
186 ; /* good */
187 } else {
188 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
189 "(wrong ip6 dst)\n"));
190 goto bad;
191 }
192 } else if (!V_nd6_onlink_ns_rfc4861) {
193 struct sockaddr_in6 src_sa6;
194
195 /*
196 * According to recent IETF discussions, it is not a good idea
197 * to accept a NS from an address which would not be deemed
198 * to be a neighbor otherwise. This point is expected to be
199 * clarified in future revisions of the specification.
200 */
201 bzero(&src_sa6, sizeof(src_sa6));
202 src_sa6.sin6_family = AF_INET6;
203 src_sa6.sin6_len = sizeof(src_sa6);
204 src_sa6.sin6_addr = saddr6;
205 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
206 nd6log((LOG_INFO, "nd6_ns_input: "
207 "NS packet from non-neighbor\n"));
208 goto bad;
209 }
210 }
211
212 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
213 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
214 goto bad;
215 }
216
217 icmp6len -= sizeof(*nd_ns);
218 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
219 if (nd6_options(&ndopts) < 0) {
220 nd6log((LOG_INFO,
221 "nd6_ns_input: invalid ND option, ignored\n"));
222 /* nd6_options have incremented stats */
223 goto freeit;
224 }
225
226 lladdr = NULL;
227 lladdrlen = 0;
228 if (ndopts.nd_opts_src_lladdr) {
229 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
230 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
231 }
232
233 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
234 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
235 "(link-layer address option)\n"));
236 goto bad;
237 }
238
239 /*
240 * Attaching target link-layer address to the NA?
241 * (RFC 2461 7.2.4)
242 *
243 * NS IP dst is unicast/anycast MUST NOT add
244 * NS IP dst is solicited-node multicast MUST add
245 *
246 * In implementation, we add target link-layer address by default.
247 * We do not add one in MUST NOT cases.
248 */
249 tlladdr = 0;
250 if (IN6_IS_ADDR_MULTICAST(&daddr6))
251 tlladdr |= ND6_NA_OPT_LLA;
252
253 /*
254 * Target address (taddr6) must be either:
255 * (1) Valid unicast/anycast address for my receiving interface,
256 * (2) Unicast address for which I'm offering proxy service, or
257 * (3) "tentative" address on which DAD is being performed.
258 */
259 /* (1) and (3) check. */
260 if (ifp->if_carp) {
261 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
262 if (ifa != NULL)
263 tlladdr |= ND6_NA_CARP_MASTER;
264 } else
265 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
266
267 /* (2) check. */
268 proxy = 0;
269 if (ifa == NULL) {
270 if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL)
271 proxy = 1;
272 }
273 if (ifa == NULL) {
274 /*
275 * We've got an NS packet, and we don't have that address
276 * assigned for us. We MUST silently ignore it.
277 * See RFC2461 7.2.3.
278 */
279 goto freeit;
280 }
281 myaddr6 = *IFA_IN6(ifa);
282 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
283 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
284 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
285 goto freeit;
286
287 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
288 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
289 "(if %d, NS packet %d)\n",
290 ip6_sprintf(ip6bufs, &taddr6),
291 ifp->if_addrlen, lladdrlen - 2));
292 goto bad;
293 }
294
295 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
296 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
297 ip6_sprintf(ip6bufs, &saddr6)));
298 goto freeit;
299 }
300
301 /*
302 * We have neighbor solicitation packet, with target address equals to
303 * one of my tentative address.
304 *
305 * src addr how to process?
306 * --- ---
307 * multicast of course, invalid (rejected in ip6_input)
308 * unicast somebody is doing address resolution -> ignore
309 * unspec dup address detection
310 *
311 * The processing is defined in RFC 2462.
312 */
313 if (tentative) {
314 /*
315 * If source address is unspecified address, it is for
316 * duplicate address detection.
317 *
318 * If not, the packet is for addess resolution;
319 * silently ignore it.
320 */
321 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
322 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
323
324 goto freeit;
325 }
326
327 /*
328 * If the Target Address is either an anycast address or a unicast
329 * address for which the node is providing proxy service, or the Target
330 * Link-Layer Address option is not included, the Override flag SHOULD
331 * be set to zero. Otherwise, the Override flag SHOULD be set to one.
332 */
333 if (anycast == 0 && proxy == 0 && (tlladdr & ND6_NA_OPT_LLA) != 0)
334 rflag |= ND_NA_FLAG_OVERRIDE;
335 /*
336 * If the source address is unspecified address, entries must not
337 * be created or updated.
338 * It looks that sender is performing DAD. nd6_na_output() will
339 * send NA toward all-node multicast address, to tell the sender
340 * that I'm using the address.
341 * S bit ("solicited") must be zero.
342 */
343 if (!IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
344 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
345 ND_NEIGHBOR_SOLICIT, 0);
346 rflag |= ND_NA_FLAG_SOLICITED;
347 }
348
349 nd6_na_output_fib(ifp, &saddr6, &taddr6, rflag, tlladdr,
350 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
351 freeit:
352 if (ifa != NULL)
353 ifa_free(ifa);
354 m_freem(m);
355 return;
356
357 bad:
358 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
359 ip6_sprintf(ip6bufs, &saddr6)));
360 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
361 ip6_sprintf(ip6bufs, &daddr6)));
362 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
363 ip6_sprintf(ip6bufs, &taddr6)));
364 bads:
365 ICMP6STAT_INC(icp6s_badns);
366 if (ifa != NULL)
367 ifa_free(ifa);
368 m_freem(m);
369 }
370
371 static struct ifaddr *
nd6_proxy_fill_sdl(struct ifnet * ifp,const struct in6_addr * taddr6,struct sockaddr_dl * sdl)372 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6,
373 struct sockaddr_dl *sdl)
374 {
375 struct ifaddr *ifa;
376 struct llentry *ln;
377
378 ifa = NULL;
379 ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp);
380 if (ln == NULL)
381 return (ifa);
382 if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) {
383 link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type);
384 sdl->sdl_alen = ifp->if_addrlen;
385 bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen);
386 LLE_RUNLOCK(ln);
387 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
388 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
389 } else
390 LLE_RUNLOCK(ln);
391
392 return (ifa);
393 }
394
395 /*
396 * Output a Neighbor Solicitation Message. Caller specifies:
397 * - ICMP6 header source IP6 address
398 * - ND6 header target IP6 address
399 * - ND6 header source datalink address
400 *
401 * Based on RFC 2461
402 * Based on RFC 2462 (duplicate address detection)
403 *
404 * ln - for source address determination
405 * nonce - If non-NULL, NS is used for duplicate address detection and
406 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
407 */
408 static void
nd6_ns_output_fib(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce,u_int fibnum)409 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
410 const struct in6_addr *daddr6, const struct in6_addr *taddr6,
411 uint8_t *nonce, u_int fibnum)
412 {
413 struct mbuf *m;
414 struct m_tag *mtag;
415 struct ip6_hdr *ip6;
416 struct nd_neighbor_solicit *nd_ns;
417 struct ip6_moptions im6o;
418 int icmp6len;
419 int maxlen;
420
421 NET_EPOCH_ASSERT();
422
423 if (IN6_IS_ADDR_MULTICAST(taddr6))
424 return;
425
426 /* estimate the size of message */
427 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
428 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
429 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
430 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
431 __func__, max_linkhdr, maxlen, MCLBYTES));
432
433 if (max_linkhdr + maxlen > MHLEN)
434 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
435 else
436 m = m_gethdr(M_NOWAIT, MT_DATA);
437 if (m == NULL)
438 return;
439 M_SETFIB(m, fibnum);
440
441 icmp6len = sizeof(*nd_ns);
442 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
443 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
444
445 /* fill neighbor solicitation packet */
446 ip6 = mtod(m, struct ip6_hdr *);
447 ip6->ip6_flow = 0;
448 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
449 ip6->ip6_vfc |= IPV6_VERSION;
450 /* ip6->ip6_plen will be set later */
451 ip6->ip6_nxt = IPPROTO_ICMPV6;
452 ip6->ip6_hlim = 255;
453 if (daddr6)
454 ip6->ip6_dst = *daddr6;
455 else {
456 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
457 ip6->ip6_dst.s6_addr16[1] = 0;
458 ip6->ip6_dst.s6_addr32[1] = 0;
459 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
460 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
461 ip6->ip6_dst.s6_addr8[12] = 0xff;
462 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
463 goto bad;
464 }
465 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
466 m->m_flags |= M_MCAST;
467 im6o.im6o_multicast_ifp = ifp;
468 im6o.im6o_multicast_hlim = 255;
469 im6o.im6o_multicast_loop = 0;
470 }
471 if (nonce == NULL) {
472 char ip6buf[INET6_ADDRSTRLEN];
473 struct ifaddr *ifa = NULL;
474
475 /*
476 * RFC2461 7.2.2:
477 * "If the source address of the packet prompting the
478 * solicitation is the same as one of the addresses assigned
479 * to the outgoing interface, that address SHOULD be placed
480 * in the IP Source Address of the outgoing solicitation.
481 * Otherwise, any one of the addresses assigned to the
482 * interface should be used."
483 *
484 * We use the source address for the prompting packet
485 * (saddr6), if saddr6 belongs to the outgoing interface.
486 * Otherwise, we perform the source address selection as usual.
487 */
488 if (saddr6 != NULL)
489 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
490 if (ifa == NULL) {
491 int error;
492
493 error = in6_selectsrc_nbr(fibnum, &ip6->ip6_dst, &im6o,
494 ifp, &ip6->ip6_src);
495 if (error) {
496 nd6log((LOG_DEBUG, "%s: source can't be "
497 "determined: dst=%s, error=%d\n", __func__,
498 ip6_sprintf(ip6buf, &ip6->ip6_dst),
499 error));
500 goto bad;
501 }
502 } else
503 ip6->ip6_src = *saddr6;
504
505 if (ifp->if_carp != NULL) {
506 /*
507 * Check that selected source address belongs to
508 * CARP addresses.
509 */
510 if (ifa == NULL)
511 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
512 &ip6->ip6_src);
513 /*
514 * Do not send NS for CARP address if we are not
515 * the CARP master.
516 */
517 if (ifa != NULL && ifa->ifa_carp != NULL &&
518 !(*carp_master_p)(ifa)) {
519 nd6log((LOG_DEBUG,
520 "nd6_ns_output: NS from BACKUP CARP address %s\n",
521 ip6_sprintf(ip6buf, &ip6->ip6_src)));
522 ifa_free(ifa);
523 goto bad;
524 }
525 }
526 if (ifa != NULL)
527 ifa_free(ifa);
528 } else {
529 /*
530 * Source address for DAD packet must always be IPv6
531 * unspecified address. (0::0)
532 * We actually don't have to 0-clear the address (we did it
533 * above), but we do so here explicitly to make the intention
534 * clearer.
535 */
536 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
537 }
538 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
539 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
540 nd_ns->nd_ns_code = 0;
541 nd_ns->nd_ns_reserved = 0;
542 nd_ns->nd_ns_target = *taddr6;
543 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
544
545 /*
546 * Add source link-layer address option.
547 *
548 * spec implementation
549 * --- ---
550 * DAD packet MUST NOT do not add the option
551 * there's no link layer address:
552 * impossible do not add the option
553 * there's link layer address:
554 * Multicast NS MUST add one add the option
555 * Unicast NS SHOULD add one add the option
556 */
557 if (nonce == NULL) {
558 struct nd_opt_hdr *nd_opt;
559 char *mac;
560 int optlen;
561
562 mac = NULL;
563 if (ifp->if_carp)
564 mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
565 if (mac == NULL)
566 mac = nd6_ifptomac(ifp);
567
568 if (mac != NULL) {
569 nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
570 optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
571 /* 8 byte alignments... */
572 optlen = (optlen + 7) & ~7;
573 m->m_pkthdr.len += optlen;
574 m->m_len += optlen;
575 icmp6len += optlen;
576 bzero(nd_opt, optlen);
577 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
578 nd_opt->nd_opt_len = optlen >> 3;
579 bcopy(mac, nd_opt + 1, ifp->if_addrlen);
580 }
581 }
582 /*
583 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
584 * This behavior is documented as Enhanced Duplicate Address
585 * Detection in RFC 7527.
586 * net.inet6.ip6.dad_enhanced=0 disables this.
587 */
588 if (V_dad_enhanced != 0 && nonce != NULL) {
589 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
590 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
591 /* 8-byte alignment is required. */
592 optlen = (optlen + 7) & ~7;
593
594 m->m_pkthdr.len += optlen;
595 m->m_len += optlen;
596 icmp6len += optlen;
597 bzero((caddr_t)nd_opt, optlen);
598 nd_opt->nd_opt_type = ND_OPT_NONCE;
599 nd_opt->nd_opt_len = optlen >> 3;
600 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
601 }
602 ip6->ip6_plen = htons((u_short)icmp6len);
603 nd_ns->nd_ns_cksum = 0;
604 nd_ns->nd_ns_cksum =
605 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
606
607 if (send_sendso_input_hook != NULL) {
608 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
609 sizeof(unsigned short), M_NOWAIT);
610 if (mtag == NULL)
611 goto bad;
612 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
613 m_tag_prepend(m, mtag);
614 }
615
616 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
617 &im6o, NULL, NULL);
618 icmp6_ifstat_inc(ifp, ifs6_out_msg);
619 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
620 ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_SOLICIT);
621
622 return;
623
624 bad:
625 m_freem(m);
626 }
627
628 #ifndef BURN_BRIDGES
629 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce)630 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
631 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
632 {
633
634 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
635 }
636 #endif
637 /*
638 * Neighbor advertisement input handling.
639 *
640 * Based on RFC 2461
641 * Based on RFC 2462 (duplicate address detection)
642 *
643 * the following items are not implemented yet:
644 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
645 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
646 */
647 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)648 nd6_na_input(struct mbuf *m, int off, int icmp6len)
649 {
650 struct ifnet *ifp;
651 struct ip6_hdr *ip6;
652 struct ifaddr *ifa;
653 struct llentry *ln;
654 struct mbuf *chain;
655 struct nd_neighbor_advert *nd_na;
656 struct in6_addr daddr6, taddr6;
657 union nd_opts ndopts;
658 u_char linkhdr[LLE_MAX_LINKHDR];
659 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
660 char *lladdr;
661 size_t linkhdrsize;
662 int flags, is_override, is_router, is_solicited;
663 int lladdr_off, lladdrlen, checklink;
664 bool flush_holdchain = false;
665
666 NET_EPOCH_ASSERT();
667
668 chain = NULL;
669 ln = NULL;
670 checklink = 0;
671
672 /* RFC 6980: Nodes MUST silently ignore fragments */
673 if(m->m_flags & M_FRAGMENTED)
674 goto freeit;
675
676 ifp = m->m_pkthdr.rcvif;
677 ip6 = mtod(m, struct ip6_hdr *);
678 if (__predict_false(ip6->ip6_hlim != 255)) {
679 ICMP6STAT_INC(icp6s_invlhlim);
680 nd6log((LOG_ERR,
681 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
682 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
683 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
684 goto bad;
685 }
686
687 if (m->m_len < off + icmp6len) {
688 m = m_pullup(m, off + icmp6len);
689 if (m == NULL) {
690 IP6STAT_INC(ip6s_exthdrtoolong);
691 return;
692 }
693 }
694 ip6 = mtod(m, struct ip6_hdr *);
695 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
696
697 flags = nd_na->nd_na_flags_reserved;
698 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
699 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
700 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
701
702 taddr6 = nd_na->nd_na_target;
703 if (in6_setscope(&taddr6, ifp, NULL))
704 goto bad; /* XXX: impossible */
705
706 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
707 nd6log((LOG_ERR,
708 "nd6_na_input: invalid target address %s\n",
709 ip6_sprintf(ip6bufs, &taddr6)));
710 goto bad;
711 }
712
713 daddr6 = ip6->ip6_dst;
714 if (IN6_IS_ADDR_MULTICAST(&daddr6))
715 if (is_solicited) {
716 nd6log((LOG_ERR,
717 "nd6_na_input: a solicited adv is multicasted\n"));
718 goto bad;
719 }
720
721 icmp6len -= sizeof(*nd_na);
722 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
723 if (nd6_options(&ndopts) < 0) {
724 nd6log((LOG_INFO,
725 "nd6_na_input: invalid ND option, ignored\n"));
726 /* nd6_options have incremented stats */
727 goto freeit;
728 }
729
730 lladdr = NULL;
731 lladdrlen = 0;
732 if (ndopts.nd_opts_tgt_lladdr) {
733 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
734 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
735 }
736
737 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
738 if (ifa != NULL && ifa->ifa_carp != NULL) {
739 /*
740 * Silently ignore NAs for CARP addresses if we are not
741 * the CARP master.
742 */
743 if (!(*carp_master_p)(ifa)) {
744 nd6log((LOG_DEBUG,
745 "nd6_na_input: NA for BACKUP CARP address %s\n",
746 ip6_sprintf(ip6bufs, &taddr6)));
747 ifa_free(ifa);
748 goto freeit;
749 }
750 }
751 /*
752 * Target address matches one of my interface address.
753 *
754 * If my address is tentative, this means that there's somebody
755 * already using the same address as mine. This indicates DAD failure.
756 * This is defined in RFC 2462.
757 *
758 * Otherwise, process as defined in RFC 2461.
759 */
760 if (ifa
761 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
762 nd6_dad_na_input(ifa);
763 ifa_free(ifa);
764 goto freeit;
765 }
766
767 /* Just for safety, maybe unnecessary. */
768 if (ifa) {
769 ifa_free(ifa);
770 log(LOG_ERR,
771 "nd6_na_input: duplicate IP6 address %s\n",
772 ip6_sprintf(ip6bufs, &taddr6));
773 goto freeit;
774 }
775
776 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
777 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
778 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
779 ifp->if_addrlen, lladdrlen - 2));
780 goto bad;
781 }
782
783 /*
784 * If no neighbor cache entry is found, NA SHOULD silently be
785 * discarded.
786 */
787 ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
788 if (ln == NULL) {
789 goto freeit;
790 }
791
792 /*
793 * Do not try to override static entry.
794 */
795 if (ln->la_flags & LLE_STATIC)
796 goto freeit;
797
798 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
799 /*
800 * If the link-layer has address, and no lladdr option came,
801 * discard the packet.
802 */
803 if (ifp->if_addrlen && lladdr == NULL) {
804 goto freeit;
805 }
806
807 /*
808 * Record link-layer address, and update the state.
809 */
810 if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
811 goto freeit;
812
813 flush_holdchain = true;
814 if (is_solicited)
815 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
816 else
817 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
818 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
819 if ((ln->ln_router = is_router) != 0) {
820 /*
821 * This means a router's state has changed from
822 * non-reachable to probably reachable, and might
823 * affect the status of associated prefixes..
824 */
825 checklink = 1;
826 }
827 } else {
828 int llchange;
829
830 /*
831 * Check if the link-layer address has changed or not.
832 */
833 if (lladdr == NULL)
834 llchange = 0;
835 else {
836 if (ln->la_flags & LLE_VALID) {
837 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
838 llchange = 1;
839 else
840 llchange = 0;
841 } else
842 llchange = 1;
843 }
844
845 /*
846 * This is VERY complex. Look at it with care.
847 *
848 * override solicit lladdr llchange action
849 * (L: record lladdr)
850 *
851 * 0 0 n -- (2c)
852 * 0 0 y n (2b) L
853 * 0 0 y y (1) REACHABLE->STALE
854 * 0 1 n -- (2c) *->REACHABLE
855 * 0 1 y n (2b) L *->REACHABLE
856 * 0 1 y y (1) REACHABLE->STALE
857 * 1 0 n -- (2a)
858 * 1 0 y n (2a) L
859 * 1 0 y y (2a) L *->STALE
860 * 1 1 n -- (2a) *->REACHABLE
861 * 1 1 y n (2a) L *->REACHABLE
862 * 1 1 y y (2a) L *->REACHABLE
863 */
864 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
865 /*
866 * If state is REACHABLE, make it STALE.
867 * no other updates should be done.
868 */
869 if (ln->ln_state == ND6_LLINFO_REACHABLE)
870 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
871 goto freeit;
872 } else if (is_override /* (2a) */
873 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
874 || lladdr == NULL) { /* (2c) */
875 /*
876 * Update link-local address, if any.
877 */
878 if (lladdr != NULL) {
879 linkhdrsize = sizeof(linkhdr);
880 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
881 linkhdr, &linkhdrsize, &lladdr_off) != 0)
882 goto freeit;
883 if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
884 linkhdrsize, lladdr_off) == 0)
885 goto freeit;
886 EVENTHANDLER_INVOKE(lle_event, ln,
887 LLENTRY_RESOLVED);
888 }
889
890 /*
891 * If solicited, make the state REACHABLE.
892 * If not solicited and the link-layer address was
893 * changed, make it STALE.
894 */
895 if (is_solicited)
896 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
897 else {
898 if (lladdr != NULL && llchange)
899 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
900 }
901 }
902
903 if (ln->ln_router && !is_router) {
904 /*
905 * The peer dropped the router flag.
906 * Remove the sender from the Default Router List and
907 * update the Destination Cache entries.
908 */
909 struct ifnet *nd6_ifp;
910
911 nd6_ifp = lltable_get_ifp(ln->lle_tbl);
912 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
913 (ND_IFINFO(nd6_ifp)->flags &
914 ND6_IFF_ACCEPT_RTADV) != 0)
915 /*
916 * Even if the neighbor is not in the default
917 * router list, the neighbor may be used as a
918 * next hop for some destinations (e.g. redirect
919 * case). So we must call rt6_flush explicitly.
920 */
921 rt6_flush(&ip6->ip6_src, ifp);
922 }
923 ln->ln_router = is_router;
924 }
925 /* XXX - QL
926 * Does this matter?
927 * rt->rt_flags &= ~RTF_REJECT;
928 */
929 ln->la_asked = 0;
930 if (ln->la_hold != NULL)
931 chain = nd6_grab_holdchain(ln);
932 freeit:
933 if (ln != NULL)
934 LLE_WUNLOCK(ln);
935
936 if (chain != NULL)
937 nd6_flush_holdchain(ifp, ln, chain);
938 if (flush_holdchain)
939 nd6_flush_children_holdchain(ifp, ln);
940
941 if (checklink)
942 pfxlist_onlink_check();
943
944 m_freem(m);
945 return;
946
947 bad:
948 if (ln != NULL)
949 LLE_WUNLOCK(ln);
950
951 ICMP6STAT_INC(icp6s_badna);
952 m_freem(m);
953 }
954
955 /*
956 * Neighbor advertisement output handling.
957 *
958 * Based on RFC 2461
959 *
960 * the following items are not implemented yet:
961 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
962 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
963 *
964 * tlladdr:
965 * - 0x01 if include target link-layer address
966 * - 0x02 if target address is CARP MASTER
967 * sdl0 - sockaddr_dl (= proxy NA) or NULL
968 */
969 static void
nd6_na_output_fib(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0,u_int fibnum)970 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
971 const struct in6_addr *taddr6, u_long flags, int tlladdr,
972 struct sockaddr *sdl0, u_int fibnum)
973 {
974 struct mbuf *m;
975 struct m_tag *mtag;
976 struct ip6_hdr *ip6;
977 struct nd_neighbor_advert *nd_na;
978 struct ip6_moptions im6o;
979 struct in6_addr daddr6;
980
981 NET_EPOCH_ASSERT();
982
983 int icmp6len, maxlen, error;
984 caddr_t mac = NULL;
985
986 daddr6 = *daddr6_0; /* make a local copy for modification */
987
988 /* estimate the size of message */
989 maxlen = sizeof(*ip6) + sizeof(*nd_na);
990 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
991 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
992 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
993 __func__, max_linkhdr, maxlen, MCLBYTES));
994
995 if (max_linkhdr + maxlen > MHLEN)
996 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
997 else
998 m = m_gethdr(M_NOWAIT, MT_DATA);
999 if (m == NULL)
1000 return;
1001 M_SETFIB(m, fibnum);
1002
1003 icmp6len = sizeof(*nd_na);
1004 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1005 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
1006
1007 /* fill neighbor advertisement packet */
1008 ip6 = mtod(m, struct ip6_hdr *);
1009 ip6->ip6_flow = 0;
1010 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1011 ip6->ip6_vfc |= IPV6_VERSION;
1012 ip6->ip6_nxt = IPPROTO_ICMPV6;
1013 ip6->ip6_hlim = 255;
1014
1015 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1016 /* reply to DAD */
1017 daddr6 = in6addr_linklocal_allnodes;
1018 if (in6_setscope(&daddr6, ifp, NULL))
1019 goto bad;
1020
1021 flags &= ~ND_NA_FLAG_SOLICITED;
1022 }
1023 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1024 m->m_flags |= M_MCAST;
1025 im6o.im6o_multicast_ifp = ifp;
1026 im6o.im6o_multicast_hlim = 255;
1027 im6o.im6o_multicast_loop = 0;
1028 }
1029
1030 ip6->ip6_dst = daddr6;
1031 error = in6_selectsrc_nbr(fibnum, &daddr6, &im6o, ifp, &ip6->ip6_src);
1032 if (error) {
1033 char ip6buf[INET6_ADDRSTRLEN];
1034 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1035 "determined: dst=%s, error=%d\n",
1036 ip6_sprintf(ip6buf, &daddr6), error));
1037 goto bad;
1038 }
1039 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1040 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1041 nd_na->nd_na_code = 0;
1042 nd_na->nd_na_target = *taddr6;
1043 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1044
1045 /*
1046 * If we respond from CARP address, we need to prepare mac address
1047 * for carp_output().
1048 */
1049 if (ifp->if_carp && (tlladdr & ND6_NA_CARP_MASTER))
1050 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1051 /*
1052 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1053 * see nd6_ns_input() for details.
1054 * Basically, if NS packet is sent to unicast/anycast addr,
1055 * target lladdr option SHOULD NOT be included.
1056 */
1057 if (tlladdr & ND6_NA_OPT_LLA) {
1058 /*
1059 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1060 * lladdr in sdl0. If we are not proxying (sending NA for
1061 * my address) use lladdr configured for the interface.
1062 */
1063 if (sdl0 == NULL) {
1064 if (mac == NULL)
1065 mac = nd6_ifptomac(ifp);
1066 } else if (sdl0->sa_family == AF_LINK) {
1067 struct sockaddr_dl *sdl;
1068 sdl = (struct sockaddr_dl *)sdl0;
1069 if (sdl->sdl_alen == ifp->if_addrlen)
1070 mac = LLADDR(sdl);
1071 }
1072 }
1073 if ((tlladdr & ND6_NA_OPT_LLA) && mac != NULL) {
1074 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1075 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1076
1077 /* roundup to 8 bytes alignment! */
1078 optlen = (optlen + 7) & ~7;
1079
1080 m->m_pkthdr.len += optlen;
1081 m->m_len += optlen;
1082 icmp6len += optlen;
1083 bzero((caddr_t)nd_opt, optlen);
1084 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1085 nd_opt->nd_opt_len = optlen >> 3;
1086 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1087 } else
1088 flags &= ~ND_NA_FLAG_OVERRIDE;
1089
1090 ip6->ip6_plen = htons((u_short)icmp6len);
1091 nd_na->nd_na_flags_reserved = flags;
1092 nd_na->nd_na_cksum = 0;
1093 nd_na->nd_na_cksum =
1094 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1095
1096 if (send_sendso_input_hook != NULL) {
1097 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1098 sizeof(unsigned short), M_NOWAIT);
1099 if (mtag == NULL)
1100 goto bad;
1101 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1102 m_tag_prepend(m, mtag);
1103 }
1104
1105 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1106 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1107 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1108 ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_ADVERT);
1109
1110 return;
1111
1112 bad:
1113 m_freem(m);
1114 }
1115
1116 #ifndef BURN_BRIDGES
1117 void
nd6_na_output(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0)1118 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1119 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1120 struct sockaddr *sdl0)
1121 {
1122
1123 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1124 RT_DEFAULT_FIB);
1125 }
1126 #endif
1127
1128 caddr_t
nd6_ifptomac(struct ifnet * ifp)1129 nd6_ifptomac(struct ifnet *ifp)
1130 {
1131 switch (ifp->if_type) {
1132 case IFT_ETHER:
1133 case IFT_IEEE1394:
1134 case IFT_L2VLAN:
1135 case IFT_INFINIBAND:
1136 case IFT_BRIDGE:
1137 return IF_LLADDR(ifp);
1138 default:
1139 return NULL;
1140 }
1141 }
1142
1143 struct dadq {
1144 TAILQ_ENTRY(dadq) dad_list;
1145 struct ifaddr *dad_ifa;
1146 int dad_count; /* max NS to send */
1147 int dad_ns_tcount; /* # of trials to send NS */
1148 int dad_ns_ocount; /* NS sent so far */
1149 int dad_ns_icount;
1150 int dad_na_icount;
1151 int dad_ns_lcount; /* looped back NS */
1152 int dad_loopbackprobe; /* probing state for loopback detection */
1153 struct callout dad_timer_ch;
1154 struct vnet *dad_vnet;
1155 u_int dad_refcnt;
1156 #define ND_OPT_NONCE_LEN32 \
1157 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1158 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1159 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */
1160 };
1161
1162 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1163 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1164 #define V_dadq VNET(dadq)
1165 #define V_dad_rwlock VNET(dad_rwlock)
1166
1167 #define DADQ_LOCKPTR() (&V_dad_rwlock)
1168 #define DADQ_LOCK_INIT() rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1169 #define DADQ_RLOCK() rw_rlock(DADQ_LOCKPTR())
1170 #define DADQ_RUNLOCK() rw_runlock(DADQ_LOCKPTR())
1171 #define DADQ_WLOCK() rw_wlock(DADQ_LOCKPTR())
1172 #define DADQ_WUNLOCK() rw_wunlock(DADQ_LOCKPTR())
1173
1174 #define DADQ_LOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1175 #define DADQ_RLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1176 #define DADQ_WLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1177
1178 static void
nd6_dad_add(struct dadq * dp)1179 nd6_dad_add(struct dadq *dp)
1180 {
1181 DADQ_WLOCK_ASSERT();
1182
1183 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1184 dp->dad_ondadq = true;
1185 }
1186
1187 static void
nd6_dad_del(struct dadq * dp)1188 nd6_dad_del(struct dadq *dp)
1189 {
1190 DADQ_WLOCK_ASSERT();
1191
1192 if (dp->dad_ondadq) {
1193 /*
1194 * Remove dp from the dadq and release the dadq's
1195 * reference.
1196 */
1197 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1198 dp->dad_ondadq = false;
1199 nd6_dad_rele(dp);
1200 }
1201 }
1202
1203 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * n)1204 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1205 {
1206 struct dadq *dp;
1207
1208 DADQ_LOCK_ASSERT();
1209
1210 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1211 if (dp->dad_ifa != ifa)
1212 continue;
1213
1214 /*
1215 * Skip if the nonce matches the received one.
1216 * +2 in the length is required because of type and
1217 * length fields are included in a header.
1218 */
1219 if (n != NULL &&
1220 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1221 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1222 ND_OPT_NONCE_LEN) == 0) {
1223 dp->dad_ns_lcount++;
1224 continue;
1225 }
1226 break;
1227 }
1228
1229 return (dp);
1230 }
1231
1232 static void
nd6_dad_starttimer(struct dadq * dp,int ticks)1233 nd6_dad_starttimer(struct dadq *dp, int ticks)
1234 {
1235 DADQ_WLOCK_ASSERT();
1236
1237 callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1238 }
1239
1240 static void
nd6_dad_stoptimer(struct dadq * dp)1241 nd6_dad_stoptimer(struct dadq *dp)
1242 {
1243 callout_drain(&dp->dad_timer_ch);
1244 }
1245
1246 static void
nd6_dad_rele(struct dadq * dp)1247 nd6_dad_rele(struct dadq *dp)
1248 {
1249 if (refcount_release(&dp->dad_refcnt)) {
1250 KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1251 ifa_free(dp->dad_ifa);
1252 free(dp, M_IP6NDP);
1253 }
1254 }
1255
1256 void
nd6_dad_init(void)1257 nd6_dad_init(void)
1258 {
1259 DADQ_LOCK_INIT();
1260 TAILQ_INIT(&V_dadq);
1261 }
1262
1263 /*
1264 * Start Duplicate Address Detection (DAD) for specified interface address.
1265 */
1266 void
nd6_dad_start(struct ifaddr * ifa,int delay)1267 nd6_dad_start(struct ifaddr *ifa, int delay)
1268 {
1269 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1270 struct dadq *dp;
1271 char ip6buf[INET6_ADDRSTRLEN];
1272
1273 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1274 ("starting DAD on non-tentative address %p", ifa));
1275
1276 /*
1277 * If we don't need DAD, don't do it.
1278 * There are several cases:
1279 * - DAD is disabled globally or on the interface
1280 * - the interface address is anycast
1281 */
1282 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1283 V_ip6_dad_count == 0 ||
1284 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1285 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1286 return;
1287 }
1288 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1289 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1290 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1291 return;
1292
1293 DADQ_WLOCK();
1294 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1295 /*
1296 * DAD is already in progress. Let the existing entry
1297 * finish it.
1298 */
1299 DADQ_WUNLOCK();
1300 return;
1301 }
1302
1303 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1304 if (dp == NULL) {
1305 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1306 "%s(%s)\n",
1307 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1308 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1309 return;
1310 }
1311 callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1312 CALLOUT_RETURNUNLOCKED);
1313 #ifdef VIMAGE
1314 dp->dad_vnet = curvnet;
1315 #endif
1316 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1317 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1318
1319 /*
1320 * Send NS packet for DAD, ip6_dad_count times.
1321 * Note that we must delay the first transmission, if this is the
1322 * first packet to be sent from the interface after interface
1323 * (re)initialization.
1324 */
1325 dp->dad_ifa = ifa;
1326 ifa_ref(dp->dad_ifa);
1327 dp->dad_count = V_ip6_dad_count;
1328 dp->dad_ns_icount = dp->dad_na_icount = 0;
1329 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1330 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1331
1332 /* Add this to the dadq and add a reference for the dadq. */
1333 refcount_init(&dp->dad_refcnt, 1);
1334 nd6_dad_add(dp);
1335 nd6_dad_starttimer(dp, delay);
1336 DADQ_WUNLOCK();
1337 }
1338
1339 /*
1340 * terminate DAD unconditionally. used for address removals.
1341 */
1342 void
nd6_dad_stop(struct ifaddr * ifa)1343 nd6_dad_stop(struct ifaddr *ifa)
1344 {
1345 struct dadq *dp;
1346
1347 DADQ_WLOCK();
1348 dp = nd6_dad_find(ifa, NULL);
1349 if (dp == NULL) {
1350 DADQ_WUNLOCK();
1351 /* DAD wasn't started yet */
1352 return;
1353 }
1354
1355 /*
1356 * Acquire a temporary reference so that we can safely stop the callout.
1357 */
1358 (void)refcount_acquire(&dp->dad_refcnt);
1359 nd6_dad_del(dp);
1360 DADQ_WUNLOCK();
1361
1362 nd6_dad_stoptimer(dp);
1363 nd6_dad_rele(dp);
1364 }
1365
1366 static void
nd6_dad_timer(void * arg)1367 nd6_dad_timer(void *arg)
1368 {
1369 struct dadq *dp = arg;
1370 struct ifaddr *ifa = dp->dad_ifa;
1371 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1372 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1373 char ip6buf[INET6_ADDRSTRLEN];
1374 struct epoch_tracker et;
1375
1376 CURVNET_SET(dp->dad_vnet);
1377 KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1378
1379 NET_EPOCH_ENTER(et);
1380 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1381 /* Do not need DAD for ifdisabled interface. */
1382 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1383 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1384 goto err;
1385 }
1386 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1387 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1388 "%s(%s)\n",
1389 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1390 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1391 goto err;
1392 }
1393 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1394 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1395 "%s(%s)\n",
1396 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1397 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1398 goto err;
1399 }
1400
1401 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1402 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1403 (((ifp->if_flags & IFF_UP) == 0) ||
1404 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1405 nd6log((LOG_INFO, "%s: could not run DAD "
1406 "because the interface was down or not running.\n",
1407 if_name(ifa->ifa_ifp)));
1408 goto err;
1409 }
1410
1411 /* Need more checks? */
1412 if (dp->dad_ns_ocount < dp->dad_count) {
1413 /*
1414 * We have more NS to go. Send NS packet for DAD.
1415 */
1416 nd6_dad_starttimer(dp,
1417 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1418 nd6_dad_ns_output(dp);
1419 goto done;
1420 } else {
1421 /*
1422 * We have transmitted sufficient number of DAD packets.
1423 * See what we've got.
1424 */
1425 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1426 /* We've seen NS or NA, means DAD has failed. */
1427 nd6_dad_duplicated(ifa, dp);
1428 } else if (V_dad_enhanced != 0 &&
1429 dp->dad_ns_lcount > 0 &&
1430 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1431 /*
1432 * Sec. 4.1 in RFC 7527 requires transmission of
1433 * additional probes until the loopback condition
1434 * becomes clear when a looped back probe is detected.
1435 */
1436 log(LOG_ERR, "%s: a looped back NS message is "
1437 "detected during DAD for %s. "
1438 "Another DAD probes are being sent.\n",
1439 if_name(ifa->ifa_ifp),
1440 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1441 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1442 /*
1443 * Send an NS immediately and increase dad_count by
1444 * V_nd6_mmaxtries - 1.
1445 */
1446 dp->dad_count =
1447 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1448 nd6_dad_starttimer(dp,
1449 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1450 nd6_dad_ns_output(dp);
1451 goto done;
1452 } else {
1453 /*
1454 * We are done with DAD. No NA came, no NS came.
1455 * No duplicate address found. Check IFDISABLED flag
1456 * again in case that it is changed between the
1457 * beginning of this function and here.
1458 *
1459 * Reset DAD failures counter if using stable addresses.
1460 */
1461 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) {
1462 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1463 if ((ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY))
1464 atomic_store_int(&DAD_FAILURES(ifp), 0);
1465 }
1466
1467 nd6log((LOG_DEBUG,
1468 "%s: DAD complete for %s - no duplicates found\n",
1469 if_name(ifa->ifa_ifp),
1470 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1471 if (dp->dad_ns_lcount > 0)
1472 log(LOG_ERR, "%s: DAD completed while "
1473 "a looped back NS message is detected "
1474 "during DAD for %s.\n",
1475 if_name(ifa->ifa_ifp),
1476 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1477 }
1478 }
1479 err:
1480 nd6_dad_del(dp);
1481 DADQ_WUNLOCK();
1482 done:
1483 NET_EPOCH_EXIT(et);
1484 CURVNET_RESTORE();
1485 }
1486
1487 static void
nd6_dad_duplicated(struct ifaddr * ifa,struct dadq * dp)1488 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1489 {
1490 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1491 struct ifnet *ifp;
1492 char ip6buf[INET6_ADDRSTRLEN];
1493
1494 ifp = ifa->ifa_ifp;
1495
1496 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1497 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1498 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1499 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1500 dp->dad_na_icount);
1501
1502 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1503 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1504
1505 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1506 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1507
1508 /*
1509 * For RFC 7217 stable addresses, increment failure counter here if we still have retries.
1510 * More addresses will be generated as long as retries are not exhausted.
1511 */
1512 if ((ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
1513 u_int dad_failures = atomic_load_int(&DAD_FAILURES(ifp));
1514
1515 if (dad_failures <= V_ip6_stableaddr_maxretries) {
1516 atomic_add_int(&DAD_FAILURES(ifp), 1);
1517 /* if retries exhausted, output an informative error message */
1518 if (dad_failures == V_ip6_stableaddr_maxretries)
1519 log(LOG_ERR, "%s: manual intervention required, consider disabling \"stableaddr\" on the interface"
1520 " or checking hostuuid for uniqueness\n",
1521 if_name(ifp));
1522 }
1523 } else {
1524 log(LOG_ERR, "%s: manual intervention required\n",
1525 if_name(ifp));
1526 }
1527
1528 /*
1529 * If the address is a link-local address formed from an interface
1530 * identifier based on the hardware address which is supposed to be
1531 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1532 * operation on the interface SHOULD be disabled.
1533 * [RFC 4862, Section 5.4.5]
1534 */
1535 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1536 struct in6_addr in6;
1537
1538 /*
1539 * To avoid over-reaction, we only apply this logic when we are
1540 * very sure that hardware addresses are supposed to be unique.
1541 */
1542 switch (ifp->if_type) {
1543 case IFT_ETHER:
1544 case IFT_ATM:
1545 case IFT_IEEE1394:
1546 case IFT_INFINIBAND:
1547 in6 = ia->ia_addr.sin6_addr;
1548 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1549 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1550 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1551 log(LOG_ERR, "%s: possible hardware address "
1552 "duplication detected, disable IPv6\n",
1553 if_name(ifp));
1554 }
1555 break;
1556 }
1557 }
1558 }
1559
1560 /*
1561 * Transmit a neighbour solicitation for the purpose of DAD. Returns with the
1562 * DAD queue unlocked.
1563 */
1564 static void
nd6_dad_ns_output(struct dadq * dp)1565 nd6_dad_ns_output(struct dadq *dp)
1566 {
1567 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1568 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1569 int i;
1570
1571 DADQ_WLOCK_ASSERT();
1572
1573 dp->dad_ns_tcount++;
1574 if ((ifp->if_flags & IFF_UP) == 0) {
1575 DADQ_WUNLOCK();
1576 return;
1577 }
1578 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1579 DADQ_WUNLOCK();
1580 return;
1581 }
1582
1583 dp->dad_ns_ocount++;
1584 if (V_dad_enhanced != 0) {
1585 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1586 dp->dad_nonce[i] = arc4random();
1587 /*
1588 * XXXHRS: Note that in the case that
1589 * DupAddrDetectTransmits > 1, multiple NS messages with
1590 * different nonces can be looped back in an unexpected
1591 * order. The current implementation recognizes only
1592 * the latest nonce on the sender side. Practically it
1593 * should work well in almost all cases.
1594 */
1595 }
1596 DADQ_WUNLOCK();
1597 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1598 (uint8_t *)&dp->dad_nonce[0]);
1599 }
1600
1601 static void
nd6_dad_ns_input(struct ifaddr * ifa,struct nd_opt_nonce * ndopt_nonce)1602 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1603 {
1604 struct dadq *dp;
1605
1606 if (ifa == NULL)
1607 panic("ifa == NULL in nd6_dad_ns_input");
1608
1609 /* Ignore Nonce option when Enhanced DAD is disabled. */
1610 if (V_dad_enhanced == 0)
1611 ndopt_nonce = NULL;
1612 DADQ_RLOCK();
1613 dp = nd6_dad_find(ifa, ndopt_nonce);
1614 if (dp != NULL)
1615 dp->dad_ns_icount++;
1616 DADQ_RUNLOCK();
1617 }
1618
1619 static void
nd6_dad_na_input(struct ifaddr * ifa)1620 nd6_dad_na_input(struct ifaddr *ifa)
1621 {
1622 struct dadq *dp;
1623
1624 if (ifa == NULL)
1625 panic("ifa == NULL in nd6_dad_na_input");
1626
1627 DADQ_RLOCK();
1628 dp = nd6_dad_find(ifa, NULL);
1629 if (dp != NULL)
1630 dp->dad_na_icount++;
1631 DADQ_RUNLOCK();
1632 }
1633