1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32 */
33
34 #include <sys/cdefs.h>
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/counter.h>
42 #include <sys/eventhandler.h>
43 #include <sys/malloc.h>
44 #include <sys/libkern.h>
45 #include <sys/lock.h>
46 #include <sys/rwlock.h>
47 #include <sys/mbuf.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/time.h>
51 #include <sys/kernel.h>
52 #include <sys/errno.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
55 #include <sys/queue.h>
56 #include <sys/callout.h>
57 #include <sys/refcount.h>
58
59 #include <net/if.h>
60 #include <net/if_types.h>
61 #include <net/if_dl.h>
62 #include <net/if_var.h>
63 #include <net/if_private.h>
64 #include <net/route.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <net/if_llatbl.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip_carp.h>
78 #include <netinet6/send.h>
79
80 #include <machine/atomic.h>
81
82 #define SDL(s) ((struct sockaddr_dl *)s)
83
84 struct dadq;
85 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
86 static void nd6_dad_add(struct dadq *dp);
87 static void nd6_dad_del(struct dadq *dp);
88 static void nd6_dad_rele(struct dadq *);
89 static void nd6_dad_starttimer(struct dadq *, int);
90 static void nd6_dad_stoptimer(struct dadq *);
91 static void nd6_dad_timer(void *);
92 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
93 static void nd6_dad_ns_output(struct dadq *);
94 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
95 static void nd6_dad_na_input(struct ifaddr *);
96 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
97 const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
98 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
99 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
100
101 static struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *,
102 const struct in6_addr *, struct sockaddr_dl *);
103
104 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
105 #define V_dad_enhanced VNET(dad_enhanced)
106
107 SYSCTL_DECL(_net_inet6_ip6);
108 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
109 &VNET_NAME(dad_enhanced), 0,
110 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
111
112 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to
113 transmit DAD packet */
114 #define V_dad_maxtry VNET(dad_maxtry)
115
116 VNET_DEFINE_STATIC(int, nd6_onlink_ns_rfc4861) = 0;
117 #define V_nd6_onlink_ns_rfc4861 VNET(nd6_onlink_ns_rfc4861)
118 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_ONLINKNSRFC4861,
119 nd6_onlink_ns_rfc4861, CTLFLAG_VNET | CTLFLAG_RW,
120 &VNET_NAME(nd6_onlink_ns_rfc4861), 0,
121 "Accept 'on-link' ICMPv6 NS messages in compliance with RFC 4861");
122
123 /*
124 * Input a Neighbor Solicitation Message.
125 *
126 * Based on RFC 2461
127 * Based on RFC 2462 (duplicate address detection)
128 */
129 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)130 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
131 {
132 struct ifnet *ifp;
133 struct ip6_hdr *ip6;
134 struct nd_neighbor_solicit *nd_ns;
135 struct in6_addr daddr6, myaddr6, saddr6, taddr6;
136 struct ifaddr *ifa;
137 struct sockaddr_dl proxydl;
138 union nd_opts ndopts;
139 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
140 char *lladdr;
141 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
142
143 ifa = NULL;
144
145 /* RFC 6980: Nodes MUST silently ignore fragments */
146 if(m->m_flags & M_FRAGMENTED)
147 goto freeit;
148
149 ifp = m->m_pkthdr.rcvif;
150 ip6 = mtod(m, struct ip6_hdr *);
151 if (__predict_false(ip6->ip6_hlim != 255)) {
152 ICMP6STAT_INC(icp6s_invlhlim);
153 nd6log((LOG_ERR,
154 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
155 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
156 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
157 goto bads;
158 }
159
160 if (m->m_len < off + icmp6len) {
161 m = m_pullup(m, off + icmp6len);
162 if (m == NULL) {
163 IP6STAT_INC(ip6s_exthdrtoolong);
164 return;
165 }
166 }
167 ip6 = mtod(m, struct ip6_hdr *);
168 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
169
170 saddr6 = ip6->ip6_src;
171 daddr6 = ip6->ip6_dst;
172 taddr6 = nd_ns->nd_ns_target;
173 if (in6_setscope(&taddr6, ifp, NULL) != 0)
174 goto bad;
175
176 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
177 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
178 rflag = 0;
179
180 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
181 /* dst has to be a solicited node multicast address. */
182 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
183 /* don't check ifindex portion */
184 daddr6.s6_addr32[1] == 0 &&
185 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
186 daddr6.s6_addr8[12] == 0xff) {
187 ; /* good */
188 } else {
189 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
190 "(wrong ip6 dst)\n"));
191 goto bad;
192 }
193 } else if (!V_nd6_onlink_ns_rfc4861) {
194 struct sockaddr_in6 src_sa6;
195
196 /*
197 * According to recent IETF discussions, it is not a good idea
198 * to accept a NS from an address which would not be deemed
199 * to be a neighbor otherwise. This point is expected to be
200 * clarified in future revisions of the specification.
201 */
202 bzero(&src_sa6, sizeof(src_sa6));
203 src_sa6.sin6_family = AF_INET6;
204 src_sa6.sin6_len = sizeof(src_sa6);
205 src_sa6.sin6_addr = saddr6;
206 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
207 nd6log((LOG_INFO, "nd6_ns_input: "
208 "NS packet from non-neighbor\n"));
209 goto bad;
210 }
211 }
212
213 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
214 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
215 goto bad;
216 }
217
218 icmp6len -= sizeof(*nd_ns);
219 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
220 if (nd6_options(&ndopts) < 0) {
221 nd6log((LOG_INFO,
222 "nd6_ns_input: invalid ND option, ignored\n"));
223 /* nd6_options have incremented stats */
224 goto freeit;
225 }
226
227 lladdr = NULL;
228 lladdrlen = 0;
229 if (ndopts.nd_opts_src_lladdr) {
230 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
231 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
232 }
233
234 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
235 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
236 "(link-layer address option)\n"));
237 goto bad;
238 }
239
240 /*
241 * Attaching target link-layer address to the NA?
242 * (RFC 2461 7.2.4)
243 *
244 * NS IP dst is unicast/anycast MUST NOT add
245 * NS IP dst is solicited-node multicast MUST add
246 *
247 * In implementation, we add target link-layer address by default.
248 * We do not add one in MUST NOT cases.
249 */
250 tlladdr = 0;
251 if (IN6_IS_ADDR_MULTICAST(&daddr6))
252 tlladdr |= ND6_NA_OPT_LLA;
253
254 /*
255 * Target address (taddr6) must be either:
256 * (1) Valid unicast/anycast address for my receiving interface,
257 * (2) Unicast address for which I'm offering proxy service, or
258 * (3) "tentative" address on which DAD is being performed.
259 */
260 /* (1) and (3) check. */
261 if (ifp->if_carp) {
262 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
263 if (ifa != NULL)
264 tlladdr |= ND6_NA_CARP_MASTER;
265 } else
266 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
267
268 /* (2) check. */
269 proxy = 0;
270 if (ifa == NULL) {
271 if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL)
272 proxy = 1;
273 }
274 if (ifa == NULL) {
275 /*
276 * We've got an NS packet, and we don't have that address
277 * assigned for us. We MUST silently ignore it.
278 * See RFC2461 7.2.3.
279 */
280 goto freeit;
281 }
282 myaddr6 = *IFA_IN6(ifa);
283 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
284 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
285 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
286 goto freeit;
287
288 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
289 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
290 "(if %d, NS packet %d)\n",
291 ip6_sprintf(ip6bufs, &taddr6),
292 ifp->if_addrlen, lladdrlen - 2));
293 goto bad;
294 }
295
296 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
297 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
298 ip6_sprintf(ip6bufs, &saddr6)));
299 goto freeit;
300 }
301
302 /*
303 * We have neighbor solicitation packet, with target address equals to
304 * one of my tentative address.
305 *
306 * src addr how to process?
307 * --- ---
308 * multicast of course, invalid (rejected in ip6_input)
309 * unicast somebody is doing address resolution -> ignore
310 * unspec dup address detection
311 *
312 * The processing is defined in RFC 2462.
313 */
314 if (tentative) {
315 /*
316 * If source address is unspecified address, it is for
317 * duplicate address detection.
318 *
319 * If not, the packet is for addess resolution;
320 * silently ignore it.
321 */
322 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
323 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
324
325 goto freeit;
326 }
327
328 /*
329 * If the Target Address is either an anycast address or a unicast
330 * address for which the node is providing proxy service, or the Target
331 * Link-Layer Address option is not included, the Override flag SHOULD
332 * be set to zero. Otherwise, the Override flag SHOULD be set to one.
333 */
334 if (anycast == 0 && proxy == 0 && (tlladdr & ND6_NA_OPT_LLA) != 0)
335 rflag |= ND_NA_FLAG_OVERRIDE;
336 /*
337 * If the source address is unspecified address, entries must not
338 * be created or updated.
339 * It looks that sender is performing DAD. nd6_na_output() will
340 * send NA toward all-node multicast address, to tell the sender
341 * that I'm using the address.
342 * S bit ("solicited") must be zero.
343 */
344 if (!IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
345 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
346 ND_NEIGHBOR_SOLICIT, 0);
347 rflag |= ND_NA_FLAG_SOLICITED;
348 }
349
350 nd6_na_output_fib(ifp, &saddr6, &taddr6, rflag, tlladdr,
351 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
352 freeit:
353 if (ifa != NULL)
354 ifa_free(ifa);
355 m_freem(m);
356 return;
357
358 bad:
359 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
360 ip6_sprintf(ip6bufs, &saddr6)));
361 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
362 ip6_sprintf(ip6bufs, &daddr6)));
363 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
364 ip6_sprintf(ip6bufs, &taddr6)));
365 bads:
366 ICMP6STAT_INC(icp6s_badns);
367 if (ifa != NULL)
368 ifa_free(ifa);
369 m_freem(m);
370 }
371
372 static struct ifaddr *
nd6_proxy_fill_sdl(struct ifnet * ifp,const struct in6_addr * taddr6,struct sockaddr_dl * sdl)373 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6,
374 struct sockaddr_dl *sdl)
375 {
376 struct ifaddr *ifa;
377 struct llentry *ln;
378
379 ifa = NULL;
380 ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp);
381 if (ln == NULL)
382 return (ifa);
383 if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) {
384 link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type);
385 sdl->sdl_alen = ifp->if_addrlen;
386 bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen);
387 LLE_RUNLOCK(ln);
388 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
389 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
390 } else
391 LLE_RUNLOCK(ln);
392
393 return (ifa);
394 }
395
396 /*
397 * Output a Neighbor Solicitation Message. Caller specifies:
398 * - ICMP6 header source IP6 address
399 * - ND6 header target IP6 address
400 * - ND6 header source datalink address
401 *
402 * Based on RFC 2461
403 * Based on RFC 2462 (duplicate address detection)
404 *
405 * ln - for source address determination
406 * nonce - If non-NULL, NS is used for duplicate address detection and
407 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
408 */
409 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)410 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
411 const struct in6_addr *daddr6, const struct in6_addr *taddr6,
412 uint8_t *nonce, u_int fibnum)
413 {
414 struct mbuf *m;
415 struct m_tag *mtag;
416 struct ip6_hdr *ip6;
417 struct nd_neighbor_solicit *nd_ns;
418 struct ip6_moptions im6o;
419 int icmp6len;
420 int maxlen;
421
422 NET_EPOCH_ASSERT();
423
424 if (IN6_IS_ADDR_MULTICAST(taddr6))
425 return;
426
427 /* estimate the size of message */
428 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
429 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
430 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
431 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
432 __func__, max_linkhdr, maxlen, MCLBYTES));
433
434 if (max_linkhdr + maxlen > MHLEN)
435 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
436 else
437 m = m_gethdr(M_NOWAIT, MT_DATA);
438 if (m == NULL)
439 return;
440 M_SETFIB(m, fibnum);
441
442 icmp6len = sizeof(*nd_ns);
443 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
444 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
445
446 /* fill neighbor solicitation packet */
447 ip6 = mtod(m, struct ip6_hdr *);
448 ip6->ip6_flow = 0;
449 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
450 ip6->ip6_vfc |= IPV6_VERSION;
451 /* ip6->ip6_plen will be set later */
452 ip6->ip6_nxt = IPPROTO_ICMPV6;
453 ip6->ip6_hlim = 255;
454 if (daddr6)
455 ip6->ip6_dst = *daddr6;
456 else {
457 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
458 ip6->ip6_dst.s6_addr16[1] = 0;
459 ip6->ip6_dst.s6_addr32[1] = 0;
460 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
461 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
462 ip6->ip6_dst.s6_addr8[12] = 0xff;
463 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
464 goto bad;
465 }
466 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
467 m->m_flags |= M_MCAST;
468 im6o.im6o_multicast_ifp = ifp;
469 im6o.im6o_multicast_hlim = 255;
470 im6o.im6o_multicast_loop = 0;
471 }
472 if (nonce == NULL) {
473 char ip6buf[INET6_ADDRSTRLEN];
474 struct ifaddr *ifa = NULL;
475
476 /*
477 * RFC2461 7.2.2:
478 * "If the source address of the packet prompting the
479 * solicitation is the same as one of the addresses assigned
480 * to the outgoing interface, that address SHOULD be placed
481 * in the IP Source Address of the outgoing solicitation.
482 * Otherwise, any one of the addresses assigned to the
483 * interface should be used."
484 *
485 * We use the source address for the prompting packet
486 * (saddr6), if saddr6 belongs to the outgoing interface.
487 * Otherwise, we perform the source address selection as usual.
488 */
489 if (saddr6 != NULL)
490 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
491 if (ifa == NULL) {
492 int error;
493
494 error = in6_selectsrc_nbr(fibnum, &ip6->ip6_dst, &im6o,
495 ifp, &ip6->ip6_src);
496 if (error) {
497 nd6log((LOG_DEBUG, "%s: source can't be "
498 "determined: dst=%s, error=%d\n", __func__,
499 ip6_sprintf(ip6buf, &ip6->ip6_dst),
500 error));
501 goto bad;
502 }
503 } else
504 ip6->ip6_src = *saddr6;
505
506 if (ifp->if_carp != NULL) {
507 /*
508 * Check that selected source address belongs to
509 * CARP addresses.
510 */
511 if (ifa == NULL)
512 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
513 &ip6->ip6_src);
514 /*
515 * Do not send NS for CARP address if we are not
516 * the CARP master.
517 */
518 if (ifa != NULL && ifa->ifa_carp != NULL &&
519 !(*carp_master_p)(ifa)) {
520 nd6log((LOG_DEBUG,
521 "nd6_ns_output: NS from BACKUP CARP address %s\n",
522 ip6_sprintf(ip6buf, &ip6->ip6_src)));
523 ifa_free(ifa);
524 goto bad;
525 }
526 }
527 if (ifa != NULL)
528 ifa_free(ifa);
529 } else {
530 /*
531 * Source address for DAD packet must always be IPv6
532 * unspecified address. (0::0)
533 * We actually don't have to 0-clear the address (we did it
534 * above), but we do so here explicitly to make the intention
535 * clearer.
536 */
537 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
538 }
539 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
540 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
541 nd_ns->nd_ns_code = 0;
542 nd_ns->nd_ns_reserved = 0;
543 nd_ns->nd_ns_target = *taddr6;
544 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
545
546 /*
547 * Add source link-layer address option.
548 *
549 * spec implementation
550 * --- ---
551 * DAD packet MUST NOT do not add the option
552 * there's no link layer address:
553 * impossible do not add the option
554 * there's link layer address:
555 * Multicast NS MUST add one add the option
556 * Unicast NS SHOULD add one add the option
557 */
558 if (nonce == NULL) {
559 struct nd_opt_hdr *nd_opt;
560 char *mac;
561 int optlen;
562
563 mac = NULL;
564 if (ifp->if_carp)
565 mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
566 if (mac == NULL)
567 mac = nd6_ifptomac(ifp);
568
569 if (mac != NULL) {
570 nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
571 optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
572 /* 8 byte alignments... */
573 optlen = (optlen + 7) & ~7;
574 m->m_pkthdr.len += optlen;
575 m->m_len += optlen;
576 icmp6len += optlen;
577 bzero(nd_opt, optlen);
578 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
579 nd_opt->nd_opt_len = optlen >> 3;
580 bcopy(mac, nd_opt + 1, ifp->if_addrlen);
581 }
582 }
583 /*
584 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
585 * This behavior is documented as Enhanced Duplicate Address
586 * Detection in RFC 7527.
587 * net.inet6.ip6.dad_enhanced=0 disables this.
588 */
589 if (V_dad_enhanced != 0 && nonce != NULL) {
590 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
591 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
592 /* 8-byte alignment is required. */
593 optlen = (optlen + 7) & ~7;
594
595 m->m_pkthdr.len += optlen;
596 m->m_len += optlen;
597 icmp6len += optlen;
598 bzero((caddr_t)nd_opt, optlen);
599 nd_opt->nd_opt_type = ND_OPT_NONCE;
600 nd_opt->nd_opt_len = optlen >> 3;
601 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
602 }
603 ip6->ip6_plen = htons((u_short)icmp6len);
604 nd_ns->nd_ns_cksum = 0;
605 nd_ns->nd_ns_cksum =
606 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
607
608 if (send_sendso_input_hook != NULL) {
609 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
610 sizeof(unsigned short), M_NOWAIT);
611 if (mtag == NULL)
612 goto bad;
613 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
614 m_tag_prepend(m, mtag);
615 }
616
617 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
618 &im6o, NULL, NULL);
619 icmp6_ifstat_inc(ifp, ifs6_out_msg);
620 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
621 ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_SOLICIT);
622
623 return;
624
625 bad:
626 m_freem(m);
627 }
628
629 #ifndef BURN_BRIDGES
630 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)631 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
632 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
633 {
634
635 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
636 }
637 #endif
638 /*
639 * Neighbor advertisement input handling.
640 *
641 * Based on RFC 2461
642 * Based on RFC 2462 (duplicate address detection)
643 *
644 * the following items are not implemented yet:
645 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
646 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
647 */
648 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)649 nd6_na_input(struct mbuf *m, int off, int icmp6len)
650 {
651 struct ifnet *ifp;
652 struct ip6_hdr *ip6;
653 struct ifaddr *ifa;
654 struct llentry *ln;
655 struct mbuf *chain;
656 struct nd_neighbor_advert *nd_na;
657 struct in6_addr daddr6, taddr6;
658 union nd_opts ndopts;
659 u_char linkhdr[LLE_MAX_LINKHDR];
660 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
661 char *lladdr;
662 size_t linkhdrsize;
663 int flags, is_override, is_router, is_solicited;
664 int lladdr_off, lladdrlen, checklink;
665 bool flush_holdchain = false;
666
667 NET_EPOCH_ASSERT();
668
669 chain = NULL;
670 ln = NULL;
671 checklink = 0;
672
673 /* RFC 6980: Nodes MUST silently ignore fragments */
674 if(m->m_flags & M_FRAGMENTED)
675 goto freeit;
676
677 ifp = m->m_pkthdr.rcvif;
678 ip6 = mtod(m, struct ip6_hdr *);
679 if (__predict_false(ip6->ip6_hlim != 255)) {
680 ICMP6STAT_INC(icp6s_invlhlim);
681 nd6log((LOG_ERR,
682 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
683 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
684 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
685 goto bad;
686 }
687
688 if (m->m_len < off + icmp6len) {
689 m = m_pullup(m, off + icmp6len);
690 if (m == NULL) {
691 IP6STAT_INC(ip6s_exthdrtoolong);
692 return;
693 }
694 }
695 ip6 = mtod(m, struct ip6_hdr *);
696 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
697
698 flags = nd_na->nd_na_flags_reserved;
699 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
700 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
701 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
702
703 taddr6 = nd_na->nd_na_target;
704 if (in6_setscope(&taddr6, ifp, NULL))
705 goto bad; /* XXX: impossible */
706
707 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
708 nd6log((LOG_ERR,
709 "nd6_na_input: invalid target address %s\n",
710 ip6_sprintf(ip6bufs, &taddr6)));
711 goto bad;
712 }
713
714 daddr6 = ip6->ip6_dst;
715 if (IN6_IS_ADDR_MULTICAST(&daddr6))
716 if (is_solicited) {
717 nd6log((LOG_ERR,
718 "nd6_na_input: a solicited adv is multicasted\n"));
719 goto bad;
720 }
721
722 icmp6len -= sizeof(*nd_na);
723 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
724 if (nd6_options(&ndopts) < 0) {
725 nd6log((LOG_INFO,
726 "nd6_na_input: invalid ND option, ignored\n"));
727 /* nd6_options have incremented stats */
728 goto freeit;
729 }
730
731 lladdr = NULL;
732 lladdrlen = 0;
733 if (ndopts.nd_opts_tgt_lladdr) {
734 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
735 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
736 }
737
738 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
739 if (ifa != NULL && ifa->ifa_carp != NULL) {
740 /*
741 * Silently ignore NAs for CARP addresses if we are not
742 * the CARP master.
743 */
744 if (!(*carp_master_p)(ifa)) {
745 nd6log((LOG_DEBUG,
746 "nd6_na_input: NA for BACKUP CARP address %s\n",
747 ip6_sprintf(ip6bufs, &taddr6)));
748 ifa_free(ifa);
749 goto freeit;
750 }
751 }
752 /*
753 * Target address matches one of my interface address.
754 *
755 * If my address is tentative, this means that there's somebody
756 * already using the same address as mine. This indicates DAD failure.
757 * This is defined in RFC 2462.
758 *
759 * Otherwise, process as defined in RFC 2461.
760 */
761 if (ifa
762 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
763 nd6_dad_na_input(ifa);
764 ifa_free(ifa);
765 goto freeit;
766 }
767
768 /* Just for safety, maybe unnecessary. */
769 if (ifa) {
770 ifa_free(ifa);
771 log(LOG_ERR,
772 "nd6_na_input: duplicate IP6 address %s\n",
773 ip6_sprintf(ip6bufs, &taddr6));
774 goto freeit;
775 }
776
777 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
778 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
779 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
780 ifp->if_addrlen, lladdrlen - 2));
781 goto bad;
782 }
783
784 /*
785 * If no neighbor cache entry is found, NA SHOULD silently be
786 * discarded.
787 */
788 ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
789 if (ln == NULL) {
790 goto freeit;
791 }
792
793 /*
794 * Do not try to override static entry.
795 */
796 if (ln->la_flags & LLE_STATIC)
797 goto freeit;
798
799 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
800 /*
801 * If the link-layer has address, and no lladdr option came,
802 * discard the packet.
803 */
804 if (ifp->if_addrlen && lladdr == NULL) {
805 goto freeit;
806 }
807
808 /*
809 * Record link-layer address, and update the state.
810 */
811 if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
812 goto freeit;
813
814 flush_holdchain = true;
815 if (is_solicited)
816 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
817 else
818 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
819 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
820 if ((ln->ln_router = is_router) != 0) {
821 /*
822 * This means a router's state has changed from
823 * non-reachable to probably reachable, and might
824 * affect the status of associated prefixes..
825 */
826 checklink = 1;
827 }
828 } else {
829 int llchange;
830
831 /*
832 * Check if the link-layer address has changed or not.
833 */
834 if (lladdr == NULL)
835 llchange = 0;
836 else {
837 if (ln->la_flags & LLE_VALID) {
838 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
839 llchange = 1;
840 else
841 llchange = 0;
842 } else
843 llchange = 1;
844 }
845
846 /*
847 * This is VERY complex. Look at it with care.
848 *
849 * override solicit lladdr llchange action
850 * (L: record lladdr)
851 *
852 * 0 0 n -- (2c)
853 * 0 0 y n (2b) L
854 * 0 0 y y (1) REACHABLE->STALE
855 * 0 1 n -- (2c) *->REACHABLE
856 * 0 1 y n (2b) L *->REACHABLE
857 * 0 1 y y (1) REACHABLE->STALE
858 * 1 0 n -- (2a)
859 * 1 0 y n (2a) L
860 * 1 0 y y (2a) L *->STALE
861 * 1 1 n -- (2a) *->REACHABLE
862 * 1 1 y n (2a) L *->REACHABLE
863 * 1 1 y y (2a) L *->REACHABLE
864 */
865 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
866 /*
867 * If state is REACHABLE, make it STALE.
868 * no other updates should be done.
869 */
870 if (ln->ln_state == ND6_LLINFO_REACHABLE)
871 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
872 goto freeit;
873 } else if (is_override /* (2a) */
874 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
875 || lladdr == NULL) { /* (2c) */
876 /*
877 * Update link-local address, if any.
878 */
879 if (lladdr != NULL) {
880 linkhdrsize = sizeof(linkhdr);
881 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
882 linkhdr, &linkhdrsize, &lladdr_off) != 0)
883 goto freeit;
884 if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
885 linkhdrsize, lladdr_off) == 0)
886 goto freeit;
887 EVENTHANDLER_INVOKE(lle_event, ln,
888 LLENTRY_RESOLVED);
889 }
890
891 /*
892 * If solicited, make the state REACHABLE.
893 * If not solicited and the link-layer address was
894 * changed, make it STALE.
895 */
896 if (is_solicited)
897 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
898 else {
899 if (lladdr != NULL && llchange)
900 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
901 }
902 }
903
904 if (ln->ln_router && !is_router) {
905 /*
906 * The peer dropped the router flag.
907 * Remove the sender from the Default Router List and
908 * update the Destination Cache entries.
909 */
910 struct ifnet *nd6_ifp;
911
912 nd6_ifp = lltable_get_ifp(ln->lle_tbl);
913 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
914 (ND_IFINFO(nd6_ifp)->flags &
915 ND6_IFF_ACCEPT_RTADV) != 0)
916 /*
917 * Even if the neighbor is not in the default
918 * router list, the neighbor may be used as a
919 * next hop for some destinations (e.g. redirect
920 * case). So we must call rt6_flush explicitly.
921 */
922 rt6_flush(&ip6->ip6_src, ifp);
923 }
924 ln->ln_router = is_router;
925 }
926 /* XXX - QL
927 * Does this matter?
928 * rt->rt_flags &= ~RTF_REJECT;
929 */
930 ln->la_asked = 0;
931 if (ln->la_hold != NULL)
932 chain = nd6_grab_holdchain(ln);
933 freeit:
934 if (ln != NULL)
935 LLE_WUNLOCK(ln);
936
937 if (chain != NULL)
938 nd6_flush_holdchain(ifp, ln, chain);
939 if (flush_holdchain)
940 nd6_flush_children_holdchain(ifp, ln);
941
942 if (checklink)
943 pfxlist_onlink_check();
944
945 m_freem(m);
946 return;
947
948 bad:
949 if (ln != NULL)
950 LLE_WUNLOCK(ln);
951
952 ICMP6STAT_INC(icp6s_badna);
953 m_freem(m);
954 }
955
956 /*
957 * Neighbor advertisement output handling.
958 *
959 * Based on RFC 2461
960 *
961 * the following items are not implemented yet:
962 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
963 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
964 *
965 * tlladdr:
966 * - 0x01 if include target link-layer address
967 * - 0x02 if target address is CARP MASTER
968 * sdl0 - sockaddr_dl (= proxy NA) or NULL
969 */
970 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)971 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
972 const struct in6_addr *taddr6, u_long flags, int tlladdr,
973 struct sockaddr *sdl0, u_int fibnum)
974 {
975 struct mbuf *m;
976 struct m_tag *mtag;
977 struct ip6_hdr *ip6;
978 struct nd_neighbor_advert *nd_na;
979 struct ip6_moptions im6o;
980 struct in6_addr daddr6;
981
982 NET_EPOCH_ASSERT();
983
984 int icmp6len, maxlen, error;
985 caddr_t mac = NULL;
986
987 daddr6 = *daddr6_0; /* make a local copy for modification */
988
989 /* estimate the size of message */
990 maxlen = sizeof(*ip6) + sizeof(*nd_na);
991 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
992 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
993 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
994 __func__, max_linkhdr, maxlen, MCLBYTES));
995
996 if (max_linkhdr + maxlen > MHLEN)
997 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
998 else
999 m = m_gethdr(M_NOWAIT, MT_DATA);
1000 if (m == NULL)
1001 return;
1002 M_SETFIB(m, fibnum);
1003
1004 icmp6len = sizeof(*nd_na);
1005 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1006 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */
1007
1008 /* fill neighbor advertisement packet */
1009 ip6 = mtod(m, struct ip6_hdr *);
1010 ip6->ip6_flow = 0;
1011 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1012 ip6->ip6_vfc |= IPV6_VERSION;
1013 ip6->ip6_nxt = IPPROTO_ICMPV6;
1014 ip6->ip6_hlim = 255;
1015
1016 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1017 /* reply to DAD */
1018 daddr6 = in6addr_linklocal_allnodes;
1019 if (in6_setscope(&daddr6, ifp, NULL))
1020 goto bad;
1021
1022 flags &= ~ND_NA_FLAG_SOLICITED;
1023 }
1024 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1025 m->m_flags |= M_MCAST;
1026 im6o.im6o_multicast_ifp = ifp;
1027 im6o.im6o_multicast_hlim = 255;
1028 im6o.im6o_multicast_loop = 0;
1029 }
1030
1031 ip6->ip6_dst = daddr6;
1032 error = in6_selectsrc_nbr(fibnum, &daddr6, &im6o, ifp, &ip6->ip6_src);
1033 if (error) {
1034 char ip6buf[INET6_ADDRSTRLEN];
1035 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1036 "determined: dst=%s, error=%d\n",
1037 ip6_sprintf(ip6buf, &daddr6), error));
1038 goto bad;
1039 }
1040 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1041 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1042 nd_na->nd_na_code = 0;
1043 nd_na->nd_na_target = *taddr6;
1044 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1045
1046 /*
1047 * If we respond from CARP address, we need to prepare mac address
1048 * for carp_output().
1049 */
1050 if (ifp->if_carp && (tlladdr & ND6_NA_CARP_MASTER))
1051 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1052 /*
1053 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1054 * see nd6_ns_input() for details.
1055 * Basically, if NS packet is sent to unicast/anycast addr,
1056 * target lladdr option SHOULD NOT be included.
1057 */
1058 if (tlladdr & ND6_NA_OPT_LLA) {
1059 /*
1060 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1061 * lladdr in sdl0. If we are not proxying (sending NA for
1062 * my address) use lladdr configured for the interface.
1063 */
1064 if (sdl0 == NULL) {
1065 if (mac == NULL)
1066 mac = nd6_ifptomac(ifp);
1067 } else if (sdl0->sa_family == AF_LINK) {
1068 struct sockaddr_dl *sdl;
1069 sdl = (struct sockaddr_dl *)sdl0;
1070 if (sdl->sdl_alen == ifp->if_addrlen)
1071 mac = LLADDR(sdl);
1072 }
1073 }
1074 if ((tlladdr & ND6_NA_OPT_LLA) && mac != NULL) {
1075 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1076 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1077
1078 /* roundup to 8 bytes alignment! */
1079 optlen = (optlen + 7) & ~7;
1080
1081 m->m_pkthdr.len += optlen;
1082 m->m_len += optlen;
1083 icmp6len += optlen;
1084 bzero((caddr_t)nd_opt, optlen);
1085 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1086 nd_opt->nd_opt_len = optlen >> 3;
1087 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1088 } else
1089 flags &= ~ND_NA_FLAG_OVERRIDE;
1090
1091 ip6->ip6_plen = htons((u_short)icmp6len);
1092 nd_na->nd_na_flags_reserved = flags;
1093 nd_na->nd_na_cksum = 0;
1094 nd_na->nd_na_cksum =
1095 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1096
1097 if (send_sendso_input_hook != NULL) {
1098 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1099 sizeof(unsigned short), M_NOWAIT);
1100 if (mtag == NULL)
1101 goto bad;
1102 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1103 m_tag_prepend(m, mtag);
1104 }
1105
1106 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1107 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1108 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1109 ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_ADVERT);
1110
1111 return;
1112
1113 bad:
1114 m_freem(m);
1115 }
1116
1117 #ifndef BURN_BRIDGES
1118 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)1119 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1120 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1121 struct sockaddr *sdl0)
1122 {
1123
1124 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1125 RT_DEFAULT_FIB);
1126 }
1127 #endif
1128
1129 caddr_t
nd6_ifptomac(struct ifnet * ifp)1130 nd6_ifptomac(struct ifnet *ifp)
1131 {
1132 switch (ifp->if_type) {
1133 case IFT_ETHER:
1134 case IFT_IEEE1394:
1135 case IFT_L2VLAN:
1136 case IFT_INFINIBAND:
1137 case IFT_BRIDGE:
1138 return IF_LLADDR(ifp);
1139 default:
1140 return NULL;
1141 }
1142 }
1143
1144 struct dadq {
1145 TAILQ_ENTRY(dadq) dad_list;
1146 struct ifaddr *dad_ifa;
1147 int dad_count; /* max NS to send */
1148 int dad_ns_tcount; /* # of trials to send NS */
1149 int dad_ns_ocount; /* NS sent so far */
1150 int dad_ns_icount;
1151 int dad_na_icount;
1152 int dad_ns_lcount; /* looped back NS */
1153 int dad_loopbackprobe; /* probing state for loopback detection */
1154 struct callout dad_timer_ch;
1155 struct vnet *dad_vnet;
1156 u_int dad_refcnt;
1157 #define ND_OPT_NONCE_LEN32 \
1158 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1159 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1160 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */
1161 };
1162
1163 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1164 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1165 #define V_dadq VNET(dadq)
1166 #define V_dad_rwlock VNET(dad_rwlock)
1167
1168 #define DADQ_LOCKPTR() (&V_dad_rwlock)
1169 #define DADQ_LOCK_INIT() rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1170 #define DADQ_RLOCK() rw_rlock(DADQ_LOCKPTR())
1171 #define DADQ_RUNLOCK() rw_runlock(DADQ_LOCKPTR())
1172 #define DADQ_WLOCK() rw_wlock(DADQ_LOCKPTR())
1173 #define DADQ_WUNLOCK() rw_wunlock(DADQ_LOCKPTR())
1174
1175 #define DADQ_LOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1176 #define DADQ_RLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1177 #define DADQ_WLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1178
1179 static void
nd6_dad_add(struct dadq * dp)1180 nd6_dad_add(struct dadq *dp)
1181 {
1182 DADQ_WLOCK_ASSERT();
1183
1184 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1185 dp->dad_ondadq = true;
1186 }
1187
1188 static void
nd6_dad_del(struct dadq * dp)1189 nd6_dad_del(struct dadq *dp)
1190 {
1191 DADQ_WLOCK_ASSERT();
1192
1193 if (dp->dad_ondadq) {
1194 /*
1195 * Remove dp from the dadq and release the dadq's
1196 * reference.
1197 */
1198 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1199 dp->dad_ondadq = false;
1200 nd6_dad_rele(dp);
1201 }
1202 }
1203
1204 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * n)1205 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1206 {
1207 struct dadq *dp;
1208
1209 DADQ_LOCK_ASSERT();
1210
1211 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1212 if (dp->dad_ifa != ifa)
1213 continue;
1214
1215 /*
1216 * Skip if the nonce matches the received one.
1217 * +2 in the length is required because of type and
1218 * length fields are included in a header.
1219 */
1220 if (n != NULL &&
1221 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1222 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1223 ND_OPT_NONCE_LEN) == 0) {
1224 dp->dad_ns_lcount++;
1225 continue;
1226 }
1227 break;
1228 }
1229
1230 return (dp);
1231 }
1232
1233 static void
nd6_dad_starttimer(struct dadq * dp,int ticks)1234 nd6_dad_starttimer(struct dadq *dp, int ticks)
1235 {
1236 DADQ_WLOCK_ASSERT();
1237
1238 callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1239 }
1240
1241 static void
nd6_dad_stoptimer(struct dadq * dp)1242 nd6_dad_stoptimer(struct dadq *dp)
1243 {
1244 callout_drain(&dp->dad_timer_ch);
1245 }
1246
1247 static void
nd6_dad_rele(struct dadq * dp)1248 nd6_dad_rele(struct dadq *dp)
1249 {
1250 if (refcount_release(&dp->dad_refcnt)) {
1251 KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1252 ifa_free(dp->dad_ifa);
1253 free(dp, M_IP6NDP);
1254 }
1255 }
1256
1257 void
nd6_dad_init(void)1258 nd6_dad_init(void)
1259 {
1260 DADQ_LOCK_INIT();
1261 TAILQ_INIT(&V_dadq);
1262 }
1263
1264 /*
1265 * Start Duplicate Address Detection (DAD) for specified interface address.
1266 */
1267 void
nd6_dad_start(struct ifaddr * ifa,int delay)1268 nd6_dad_start(struct ifaddr *ifa, int delay)
1269 {
1270 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1271 struct dadq *dp;
1272 char ip6buf[INET6_ADDRSTRLEN];
1273
1274 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1275 ("starting DAD on non-tentative address %p", ifa));
1276
1277 /*
1278 * If we don't need DAD, don't do it.
1279 * There are several cases:
1280 * - DAD is disabled globally or on the interface
1281 * - the interface address is anycast
1282 */
1283 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1284 V_ip6_dad_count == 0 ||
1285 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1286 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1287 return;
1288 }
1289 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1290 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1291 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1292 return;
1293
1294 DADQ_WLOCK();
1295 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1296 /*
1297 * DAD is already in progress. Let the existing entry
1298 * finish it.
1299 */
1300 DADQ_WUNLOCK();
1301 return;
1302 }
1303
1304 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1305 if (dp == NULL) {
1306 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1307 "%s(%s)\n",
1308 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1309 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1310 return;
1311 }
1312 callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1313 CALLOUT_RETURNUNLOCKED);
1314 #ifdef VIMAGE
1315 dp->dad_vnet = curvnet;
1316 #endif
1317 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1318 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1319
1320 /*
1321 * Send NS packet for DAD, ip6_dad_count times.
1322 * Note that we must delay the first transmission, if this is the
1323 * first packet to be sent from the interface after interface
1324 * (re)initialization.
1325 */
1326 dp->dad_ifa = ifa;
1327 ifa_ref(dp->dad_ifa);
1328 dp->dad_count = V_ip6_dad_count;
1329 dp->dad_ns_icount = dp->dad_na_icount = 0;
1330 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1331 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1332
1333 /* Add this to the dadq and add a reference for the dadq. */
1334 refcount_init(&dp->dad_refcnt, 1);
1335 nd6_dad_add(dp);
1336 nd6_dad_starttimer(dp, delay);
1337 DADQ_WUNLOCK();
1338 }
1339
1340 /*
1341 * terminate DAD unconditionally. used for address removals.
1342 */
1343 void
nd6_dad_stop(struct ifaddr * ifa)1344 nd6_dad_stop(struct ifaddr *ifa)
1345 {
1346 struct dadq *dp;
1347
1348 DADQ_WLOCK();
1349 dp = nd6_dad_find(ifa, NULL);
1350 if (dp == NULL) {
1351 DADQ_WUNLOCK();
1352 /* DAD wasn't started yet */
1353 return;
1354 }
1355
1356 /*
1357 * Acquire a temporary reference so that we can safely stop the callout.
1358 */
1359 (void)refcount_acquire(&dp->dad_refcnt);
1360 nd6_dad_del(dp);
1361 DADQ_WUNLOCK();
1362
1363 nd6_dad_stoptimer(dp);
1364 nd6_dad_rele(dp);
1365 }
1366
1367 static void
nd6_dad_timer(void * arg)1368 nd6_dad_timer(void *arg)
1369 {
1370 struct dadq *dp = arg;
1371 struct ifaddr *ifa = dp->dad_ifa;
1372 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1373 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1374 char ip6buf[INET6_ADDRSTRLEN];
1375 struct epoch_tracker et;
1376
1377 CURVNET_SET(dp->dad_vnet);
1378 KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1379
1380 NET_EPOCH_ENTER(et);
1381 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1382 /* Do not need DAD for ifdisabled interface. */
1383 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1384 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1385 goto err;
1386 }
1387 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1388 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1389 "%s(%s)\n",
1390 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1391 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1392 goto err;
1393 }
1394 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1395 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1396 "%s(%s)\n",
1397 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1398 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1399 goto err;
1400 }
1401
1402 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1403 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1404 (((ifp->if_flags & IFF_UP) == 0) ||
1405 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1406 nd6log((LOG_INFO, "%s: could not run DAD "
1407 "because the interface was down or not running.\n",
1408 if_name(ifa->ifa_ifp)));
1409 goto err;
1410 }
1411
1412 /* Need more checks? */
1413 if (dp->dad_ns_ocount < dp->dad_count) {
1414 /*
1415 * We have more NS to go. Send NS packet for DAD.
1416 */
1417 nd6_dad_starttimer(dp,
1418 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1419 nd6_dad_ns_output(dp);
1420 goto done;
1421 } else {
1422 /*
1423 * We have transmitted sufficient number of DAD packets.
1424 * See what we've got.
1425 */
1426 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1427 /* We've seen NS or NA, means DAD has failed. */
1428 nd6_dad_duplicated(ifa, dp);
1429 } else if (V_dad_enhanced != 0 &&
1430 dp->dad_ns_lcount > 0 &&
1431 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1432 /*
1433 * Sec. 4.1 in RFC 7527 requires transmission of
1434 * additional probes until the loopback condition
1435 * becomes clear when a looped back probe is detected.
1436 */
1437 log(LOG_ERR, "%s: a looped back NS message is "
1438 "detected during DAD for %s. "
1439 "Another DAD probes are being sent.\n",
1440 if_name(ifa->ifa_ifp),
1441 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1442 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1443 /*
1444 * Send an NS immediately and increase dad_count by
1445 * V_nd6_mmaxtries - 1.
1446 */
1447 dp->dad_count =
1448 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1449 nd6_dad_starttimer(dp,
1450 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1451 nd6_dad_ns_output(dp);
1452 goto done;
1453 } else {
1454 /*
1455 * We are done with DAD. No NA came, no NS came.
1456 * No duplicate address found. Check IFDISABLED flag
1457 * again in case that it is changed between the
1458 * beginning of this function and here.
1459 *
1460 * Reset DAD failures counter if using stable addresses.
1461 */
1462 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) {
1463 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1464 if ((ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY))
1465 atomic_store_int(&DAD_FAILURES(ifp), 0);
1466 }
1467
1468 nd6log((LOG_DEBUG,
1469 "%s: DAD complete for %s - no duplicates found\n",
1470 if_name(ifa->ifa_ifp),
1471 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1472 if (dp->dad_ns_lcount > 0)
1473 log(LOG_ERR, "%s: DAD completed while "
1474 "a looped back NS message is detected "
1475 "during DAD for %s.\n",
1476 if_name(ifa->ifa_ifp),
1477 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1478 }
1479 }
1480 err:
1481 nd6_dad_del(dp);
1482 DADQ_WUNLOCK();
1483 done:
1484 NET_EPOCH_EXIT(et);
1485 CURVNET_RESTORE();
1486 }
1487
1488 static void
nd6_dad_duplicated(struct ifaddr * ifa,struct dadq * dp)1489 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1490 {
1491 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1492 struct ifnet *ifp;
1493 char ip6buf[INET6_ADDRSTRLEN];
1494
1495 ifp = ifa->ifa_ifp;
1496
1497 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1498 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1499 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1500 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1501 dp->dad_na_icount);
1502
1503 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1504 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1505
1506 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1507 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1508
1509 /*
1510 * For RFC 7217 stable addresses, increment failure counter here if we still have retries.
1511 * More addresses will be generated as long as retries are not exhausted.
1512 */
1513 if ((ND_IFINFO(ifp)->flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
1514 u_int dad_failures = atomic_load_int(&DAD_FAILURES(ifp));
1515
1516 if (dad_failures <= V_ip6_stableaddr_maxretries) {
1517 atomic_add_int(&DAD_FAILURES(ifp), 1);
1518 /* if retries exhausted, output an informative error message */
1519 if (dad_failures == V_ip6_stableaddr_maxretries)
1520 log(LOG_ERR, "%s: manual intervention required, consider disabling \"stableaddr\" on the interface"
1521 " or checking hostuuid for uniqueness\n",
1522 if_name(ifp));
1523 }
1524 } else {
1525 log(LOG_ERR, "%s: manual intervention required\n",
1526 if_name(ifp));
1527 }
1528
1529 /*
1530 * If the address is a link-local address formed from an interface
1531 * identifier based on the hardware address which is supposed to be
1532 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1533 * operation on the interface SHOULD be disabled.
1534 * [RFC 4862, Section 5.4.5]
1535 */
1536 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1537 struct in6_addr in6;
1538
1539 /*
1540 * To avoid over-reaction, we only apply this logic when we are
1541 * very sure that hardware addresses are supposed to be unique.
1542 */
1543 switch (ifp->if_type) {
1544 case IFT_ETHER:
1545 case IFT_ATM:
1546 case IFT_IEEE1394:
1547 case IFT_INFINIBAND:
1548 in6 = ia->ia_addr.sin6_addr;
1549 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1550 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1551 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1552 log(LOG_ERR, "%s: possible hardware address "
1553 "duplication detected, disable IPv6\n",
1554 if_name(ifp));
1555 }
1556 break;
1557 }
1558 }
1559 }
1560
1561 /*
1562 * Transmit a neighbour solicitation for the purpose of DAD. Returns with the
1563 * DAD queue unlocked.
1564 */
1565 static void
nd6_dad_ns_output(struct dadq * dp)1566 nd6_dad_ns_output(struct dadq *dp)
1567 {
1568 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1569 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1570 int i;
1571
1572 DADQ_WLOCK_ASSERT();
1573
1574 dp->dad_ns_tcount++;
1575 if ((ifp->if_flags & IFF_UP) == 0) {
1576 DADQ_WUNLOCK();
1577 return;
1578 }
1579 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1580 DADQ_WUNLOCK();
1581 return;
1582 }
1583
1584 dp->dad_ns_ocount++;
1585 if (V_dad_enhanced != 0) {
1586 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1587 dp->dad_nonce[i] = arc4random();
1588 /*
1589 * XXXHRS: Note that in the case that
1590 * DupAddrDetectTransmits > 1, multiple NS messages with
1591 * different nonces can be looped back in an unexpected
1592 * order. The current implementation recognizes only
1593 * the latest nonce on the sender side. Practically it
1594 * should work well in almost all cases.
1595 */
1596 }
1597 DADQ_WUNLOCK();
1598 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1599 (uint8_t *)&dp->dad_nonce[0]);
1600 }
1601
1602 static void
nd6_dad_ns_input(struct ifaddr * ifa,struct nd_opt_nonce * ndopt_nonce)1603 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1604 {
1605 struct dadq *dp;
1606
1607 if (ifa == NULL)
1608 panic("ifa == NULL in nd6_dad_ns_input");
1609
1610 /* Ignore Nonce option when Enhanced DAD is disabled. */
1611 if (V_dad_enhanced == 0)
1612 ndopt_nonce = NULL;
1613 DADQ_RLOCK();
1614 dp = nd6_dad_find(ifa, ndopt_nonce);
1615 if (dp != NULL)
1616 dp->dad_ns_icount++;
1617 DADQ_RUNLOCK();
1618 }
1619
1620 static void
nd6_dad_na_input(struct ifaddr * ifa)1621 nd6_dad_na_input(struct ifaddr *ifa)
1622 {
1623 struct dadq *dp;
1624
1625 if (ifa == NULL)
1626 panic("ifa == NULL in nd6_dad_na_input");
1627
1628 DADQ_RLOCK();
1629 dp = nd6_dad_find(ifa, NULL);
1630 if (dp != NULL)
1631 dp->dad_na_icount++;
1632 DADQ_RUNLOCK();
1633 }
1634