1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/10.4/sys/netinet6/nd6_nbr.c 303458 2016-07-28 20:08:01Z sbruno $");
34
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le))
72 #include <netinet6/in6_var.h>
73 #include <netinet6/in6_ifattach.h>
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
76 #include <netinet6/scope6_var.h>
77 #include <netinet6/nd6.h>
78 #include <netinet/icmp6.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet6/send.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, int);
90 static void nd6_dad_stoptimer(struct dadq *);
91 static void nd6_dad_timer(struct dadq *);
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 *, struct llentry *, uint8_t *, u_int);
100
101 static VNET_DEFINE(int, dad_enhanced) = 1;
102 #define V_dad_enhanced VNET(dad_enhanced)
103
104 SYSCTL_DECL(_net_inet6_ip6);
105 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
106 &VNET_NAME(dad_enhanced), 0,
107 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
108
109 static VNET_DEFINE(int, dad_maxtry) = 15; /* max # of *tries* to
110 transmit DAD packet */
111 #define V_dad_maxtry VNET(dad_maxtry)
112
113 /*
114 * Input a Neighbor Solicitation Message.
115 *
116 * Based on RFC 2461
117 * Based on RFC 2462 (duplicate address detection)
118 */
119 void
120 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
121 {
122 struct ifnet *ifp = m->m_pkthdr.rcvif;
123 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
124 struct nd_neighbor_solicit *nd_ns;
125 struct in6_addr saddr6 = ip6->ip6_src;
126 struct in6_addr daddr6 = ip6->ip6_dst;
127 struct in6_addr taddr6;
128 struct in6_addr myaddr6;
129 char *lladdr = NULL;
130 struct ifaddr *ifa = NULL;
131 int lladdrlen = 0;
132 int anycast = 0, proxy = 0, tentative = 0;
133 int tlladdr;
134 int rflag;
135 union nd_opts ndopts;
136 struct sockaddr_dl proxydl;
137 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
138
139 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
140 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
141 rflag = 0;
142 #ifndef PULLDOWN_TEST
143 IP6_EXTHDR_CHECK(m, off, icmp6len,);
144 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
145 #else
146 IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
147 if (nd_ns == NULL) {
148 ICMP6STAT_INC(icp6s_tooshort);
149 return;
150 }
151 #endif
152 ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
153 taddr6 = nd_ns->nd_ns_target;
154 if (in6_setscope(&taddr6, ifp, NULL) != 0)
155 goto bad;
156
157 if (ip6->ip6_hlim != 255) {
158 nd6log((LOG_ERR,
159 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
160 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
161 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
162 goto bad;
163 }
164
165 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
166 /* dst has to be a solicited node multicast address. */
167 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
168 /* don't check ifindex portion */
169 daddr6.s6_addr32[1] == 0 &&
170 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
171 daddr6.s6_addr8[12] == 0xff) {
172 ; /* good */
173 } else {
174 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
175 "(wrong ip6 dst)\n"));
176 goto bad;
177 }
178 } else if (!V_nd6_onlink_ns_rfc4861) {
179 struct sockaddr_in6 src_sa6;
180
181 /*
182 * According to recent IETF discussions, it is not a good idea
183 * to accept a NS from an address which would not be deemed
184 * to be a neighbor otherwise. This point is expected to be
185 * clarified in future revisions of the specification.
186 */
187 bzero(&src_sa6, sizeof(src_sa6));
188 src_sa6.sin6_family = AF_INET6;
189 src_sa6.sin6_len = sizeof(src_sa6);
190 src_sa6.sin6_addr = saddr6;
191 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
192 nd6log((LOG_INFO, "nd6_ns_input: "
193 "NS packet from non-neighbor\n"));
194 goto bad;
195 }
196 }
197
198 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
199 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
200 goto bad;
201 }
202
203 icmp6len -= sizeof(*nd_ns);
204 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
205 if (nd6_options(&ndopts) < 0) {
206 nd6log((LOG_INFO,
207 "nd6_ns_input: invalid ND option, ignored\n"));
208 /* nd6_options have incremented stats */
209 goto freeit;
210 }
211
212 if (ndopts.nd_opts_src_lladdr) {
213 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
214 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
215 }
216
217 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
218 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
219 "(link-layer address option)\n"));
220 goto bad;
221 }
222
223 /*
224 * Attaching target link-layer address to the NA?
225 * (RFC 2461 7.2.4)
226 *
227 * NS IP dst is unicast/anycast MUST NOT add
228 * NS IP dst is solicited-node multicast MUST add
229 *
230 * In implementation, we add target link-layer address by default.
231 * We do not add one in MUST NOT cases.
232 */
233 if (!IN6_IS_ADDR_MULTICAST(&daddr6))
234 tlladdr = 0;
235 else
236 tlladdr = 1;
237
238 /*
239 * Target address (taddr6) must be either:
240 * (1) Valid unicast/anycast address for my receiving interface,
241 * (2) Unicast address for which I'm offering proxy service, or
242 * (3) "tentative" address on which DAD is being performed.
243 */
244 /* (1) and (3) check. */
245 if (ifp->if_carp)
246 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
247 else
248 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
249
250 /* (2) check. */
251 if (ifa == NULL) {
252 struct route_in6 ro;
253 int need_proxy;
254
255 bzero(&ro, sizeof(ro));
256 ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
257 ro.ro_dst.sin6_family = AF_INET6;
258 ro.ro_dst.sin6_addr = taddr6;
259
260 /* Always use the default FIB. */
261 #ifdef RADIX_MPATH
262 rtalloc_mpath_fib((struct route *)&ro, RTF_ANNOUNCE,
263 RT_DEFAULT_FIB);
264 #else
265 in6_rtalloc(&ro, RT_DEFAULT_FIB);
266 #endif
267 need_proxy = (ro.ro_rt &&
268 (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
269 ro.ro_rt->rt_gateway->sa_family == AF_LINK);
270 if (ro.ro_rt != NULL) {
271 if (need_proxy)
272 proxydl = *SDL(ro.ro_rt->rt_gateway);
273 RTFREE(ro.ro_rt);
274 }
275 if (need_proxy) {
276 /*
277 * proxy NDP for single entry
278 */
279 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
280 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
281 if (ifa)
282 proxy = 1;
283 }
284 }
285 if (ifa == NULL) {
286 /*
287 * We've got an NS packet, and we don't have that adddress
288 * assigned for us. We MUST silently ignore it.
289 * See RFC2461 7.2.3.
290 */
291 goto freeit;
292 }
293 myaddr6 = *IFA_IN6(ifa);
294 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
295 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
296 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
297 goto freeit;
298
299 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
300 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
301 "(if %d, NS packet %d)\n",
302 ip6_sprintf(ip6bufs, &taddr6),
303 ifp->if_addrlen, lladdrlen - 2));
304 goto bad;
305 }
306
307 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
308 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
309 ip6_sprintf(ip6bufs, &saddr6)));
310 goto freeit;
311 }
312
313 /*
314 * We have neighbor solicitation packet, with target address equals to
315 * one of my tentative address.
316 *
317 * src addr how to process?
318 * --- ---
319 * multicast of course, invalid (rejected in ip6_input)
320 * unicast somebody is doing address resolution -> ignore
321 * unspec dup address detection
322 *
323 * The processing is defined in RFC 2462.
324 */
325 if (tentative) {
326 /*
327 * If source address is unspecified address, it is for
328 * duplicate address detection.
329 *
330 * If not, the packet is for addess resolution;
331 * silently ignore it.
332 */
333 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
334 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
335
336 goto freeit;
337 }
338
339 /*
340 * If the source address is unspecified address, entries must not
341 * be created or updated.
342 * It looks that sender is performing DAD. Output NA toward
343 * all-node multicast address, to tell the sender that I'm using
344 * the address.
345 * S bit ("solicited") must be zero.
346 */
347 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
348 struct in6_addr in6_all;
349
350 in6_all = in6addr_linklocal_allnodes;
351 if (in6_setscope(&in6_all, ifp, NULL) != 0)
352 goto bad;
353 nd6_na_output_fib(ifp, &in6_all, &taddr6,
354 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
356 M_GETFIB(m));
357 goto freeit;
358 }
359
360 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
361 ND_NEIGHBOR_SOLICIT, 0);
362
363 nd6_na_output_fib(ifp, &saddr6, &taddr6,
364 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
365 rflag | ND_NA_FLAG_SOLICITED, tlladdr,
366 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
367 freeit:
368 if (ifa != NULL)
369 ifa_free(ifa);
370 m_freem(m);
371 return;
372
373 bad:
374 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
375 ip6_sprintf(ip6bufs, &saddr6)));
376 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
377 ip6_sprintf(ip6bufs, &daddr6)));
378 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
379 ip6_sprintf(ip6bufs, &taddr6)));
380 ICMP6STAT_INC(icp6s_badns);
381 if (ifa != NULL)
382 ifa_free(ifa);
383 m_freem(m);
384 }
385
386 /*
387 * Output a Neighbor Solicitation Message. Caller specifies:
388 * - ICMP6 header source IP6 address
389 * - ND6 header target IP6 address
390 * - ND6 header source datalink address
391 *
392 * Based on RFC 2461
393 * Based on RFC 2462 (duplicate address detection)
394 *
395 * ln - for source address determination
396 * nonce - If non-NULL, NS is used for duplicate address detection and
397 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
398 */
399 static void
400 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6,
401 const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce,
402 u_int fibnum)
403 {
404 struct mbuf *m;
405 struct m_tag *mtag;
406 struct ip6_hdr *ip6;
407 struct nd_neighbor_solicit *nd_ns;
408 struct ip6_moptions im6o;
409 int icmp6len;
410 int maxlen;
411 caddr_t mac;
412 struct route_in6 ro;
413
414 if (IN6_IS_ADDR_MULTICAST(taddr6))
415 return;
416
417 /* estimate the size of message */
418 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
419 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
420 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
421 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
422 __func__, max_linkhdr, maxlen, MCLBYTES));
423
424
425 if (max_linkhdr + maxlen > MHLEN)
426 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
427 else
428 m = m_gethdr(M_NOWAIT, MT_DATA);
429 if (m == NULL)
430 return;
431 M_SETFIB(m, fibnum);
432
433 bzero(&ro, sizeof(ro));
434
435 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
436 m->m_flags |= M_MCAST;
437 im6o.im6o_multicast_ifp = ifp;
438 im6o.im6o_multicast_hlim = 255;
439 im6o.im6o_multicast_loop = 0;
440 }
441
442 icmp6len = sizeof(*nd_ns);
443 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
444 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
445
446 /* fill neighbor solicitation packet */
447 ip6 = mtod(m, struct ip6_hdr *);
448 ip6->ip6_flow = 0;
449 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
450 ip6->ip6_vfc |= IPV6_VERSION;
451 /* ip6->ip6_plen will be set later */
452 ip6->ip6_nxt = IPPROTO_ICMPV6;
453 ip6->ip6_hlim = 255;
454 if (daddr6)
455 ip6->ip6_dst = *daddr6;
456 else {
457 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
458 ip6->ip6_dst.s6_addr16[1] = 0;
459 ip6->ip6_dst.s6_addr32[1] = 0;
460 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
461 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
462 ip6->ip6_dst.s6_addr8[12] = 0xff;
463 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
464 goto bad;
465 }
466 if (nonce == NULL) {
467 struct ifaddr *ifa;
468
469 /*
470 * RFC2461 7.2.2:
471 * "If the source address of the packet prompting the
472 * solicitation is the same as one of the addresses assigned
473 * to the outgoing interface, that address SHOULD be placed
474 * in the IP Source Address of the outgoing solicitation.
475 * Otherwise, any one of the addresses assigned to the
476 * interface should be used."
477 *
478 * We use the source address for the prompting packet
479 * (saddr6), if:
480 * - saddr6 is given from the caller (by giving "ln"), and
481 * - saddr6 belongs to the outgoing interface.
482 * Otherwise, we perform the source address selection as usual.
483 */
484 struct in6_addr *hsrc;
485
486 hsrc = NULL;
487 if (ln != NULL) {
488 LLE_RLOCK(ln);
489 if (ln->la_hold != NULL) {
490 struct ip6_hdr *hip6; /* hold ip6 */
491
492 /*
493 * assuming every packet in la_hold has the same IP
494 * header
495 */
496 hip6 = mtod(ln->la_hold, struct ip6_hdr *);
497 /* XXX pullup? */
498 if (sizeof(*hip6) < ln->la_hold->m_len) {
499 ip6->ip6_src = hip6->ip6_src;
500 hsrc = &hip6->ip6_src;
501 }
502 }
503 LLE_RUNLOCK(ln);
504 }
505 if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
506 hsrc)) != NULL) {
507 /* ip6_src set already. */
508 ifa_free(ifa);
509 } else {
510 int error;
511 struct sockaddr_in6 dst_sa;
512 struct in6_addr src_in;
513 struct ifnet *oifp;
514
515 bzero(&dst_sa, sizeof(dst_sa));
516 dst_sa.sin6_family = AF_INET6;
517 dst_sa.sin6_len = sizeof(dst_sa);
518 dst_sa.sin6_addr = ip6->ip6_dst;
519
520 oifp = ifp;
521 error = in6_selectsrc(&dst_sa, NULL,
522 NULL, &ro, NULL, &oifp, &src_in);
523 if (error) {
524 char ip6buf[INET6_ADDRSTRLEN];
525 nd6log((LOG_DEBUG, "%s: source can't be "
526 "determined: dst=%s, error=%d\n", __func__,
527 ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
528 error));
529 goto bad;
530 }
531 ip6->ip6_src = src_in;
532 }
533 } else {
534 /*
535 * Source address for DAD packet must always be IPv6
536 * unspecified address. (0::0)
537 * We actually don't have to 0-clear the address (we did it
538 * above), but we do so here explicitly to make the intention
539 * clearer.
540 */
541 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
542 }
543 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
544 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
545 nd_ns->nd_ns_code = 0;
546 nd_ns->nd_ns_reserved = 0;
547 nd_ns->nd_ns_target = *taddr6;
548 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
549
550 /*
551 * Add source link-layer address option.
552 *
553 * spec implementation
554 * --- ---
555 * DAD packet MUST NOT do not add the option
556 * there's no link layer address:
557 * impossible do not add the option
558 * there's link layer address:
559 * Multicast NS MUST add one add the option
560 * Unicast NS SHOULD add one add the option
561 */
562 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
563 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
564 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
565 /* 8 byte alignments... */
566 optlen = (optlen + 7) & ~7;
567
568 m->m_pkthdr.len += optlen;
569 m->m_len += optlen;
570 icmp6len += optlen;
571 bzero((caddr_t)nd_opt, optlen);
572 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
573 nd_opt->nd_opt_len = optlen >> 3;
574 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
575 }
576 /*
577 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
578 * This behavior is documented as Enhanced Duplicate Address
579 * Detection in RFC 7527.
580 * net.inet6.ip6.dad_enhanced=0 disables this.
581 */
582 if (V_dad_enhanced != 0 && nonce != NULL) {
583 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
584 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
585 /* 8-byte alignment is required. */
586 optlen = (optlen + 7) & ~7;
587
588 m->m_pkthdr.len += optlen;
589 m->m_len += optlen;
590 icmp6len += optlen;
591 bzero((caddr_t)nd_opt, optlen);
592 nd_opt->nd_opt_type = ND_OPT_NONCE;
593 nd_opt->nd_opt_len = optlen >> 3;
594 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
595 }
596 ip6->ip6_plen = htons((u_short)icmp6len);
597 nd_ns->nd_ns_cksum = 0;
598 nd_ns->nd_ns_cksum =
599 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
600
601 if (send_sendso_input_hook != NULL) {
602 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
603 sizeof(unsigned short), M_NOWAIT);
604 if (mtag == NULL)
605 goto bad;
606 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
607 m_tag_prepend(m, mtag);
608 }
609
610 ip6_output(m, NULL, &ro, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
611 &im6o, NULL, NULL);
612 icmp6_ifstat_inc(ifp, ifs6_out_msg);
613 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
614 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
615
616 /* We don't cache this route. */
617 RO_RTFREE(&ro);
618
619 return;
620
621 bad:
622 if (ro.ro_rt) {
623 RTFREE(ro.ro_rt);
624 }
625 m_freem(m);
626 }
627
628 #ifndef BURN_BRIDGES
629 void
630 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
631 const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce)
632 {
633
634 nd6_ns_output_fib(ifp, daddr6, taddr6, ln, 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
648 nd6_na_input(struct mbuf *m, int off, int icmp6len)
649 {
650 struct ifnet *ifp = m->m_pkthdr.rcvif;
651 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
652 struct nd_neighbor_advert *nd_na;
653 struct in6_addr daddr6 = ip6->ip6_dst;
654 struct in6_addr taddr6;
655 int flags;
656 int is_router;
657 int is_solicited;
658 int is_override;
659 char *lladdr = NULL;
660 int lladdrlen = 0;
661 int checklink = 0;
662 struct ifaddr *ifa;
663 struct llentry *ln = NULL;
664 union nd_opts ndopts;
665 struct mbuf *chain = NULL;
666 struct sockaddr_in6 sin6;
667 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
668
669 if (ip6->ip6_hlim != 255) {
670 nd6log((LOG_ERR,
671 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
672 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
673 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
674 goto bad;
675 }
676
677 #ifndef PULLDOWN_TEST
678 IP6_EXTHDR_CHECK(m, off, icmp6len,);
679 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
680 #else
681 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
682 if (nd_na == NULL) {
683 ICMP6STAT_INC(icp6s_tooshort);
684 return;
685 }
686 #endif
687
688 flags = nd_na->nd_na_flags_reserved;
689 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
690 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
691 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
692 memset(&sin6, 0, sizeof(sin6));
693
694 taddr6 = nd_na->nd_na_target;
695 if (in6_setscope(&taddr6, ifp, NULL))
696 goto bad; /* XXX: impossible */
697
698 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
699 nd6log((LOG_ERR,
700 "nd6_na_input: invalid target address %s\n",
701 ip6_sprintf(ip6bufs, &taddr6)));
702 goto bad;
703 }
704 if (IN6_IS_ADDR_MULTICAST(&daddr6))
705 if (is_solicited) {
706 nd6log((LOG_ERR,
707 "nd6_na_input: a solicited adv is multicasted\n"));
708 goto bad;
709 }
710
711 icmp6len -= sizeof(*nd_na);
712 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
713 if (nd6_options(&ndopts) < 0) {
714 nd6log((LOG_INFO,
715 "nd6_na_input: invalid ND option, ignored\n"));
716 /* nd6_options have incremented stats */
717 goto freeit;
718 }
719
720 if (ndopts.nd_opts_tgt_lladdr) {
721 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
722 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
723 }
724
725 /*
726 * This effectively disables the DAD check on a non-master CARP
727 * address.
728 */
729 if (ifp->if_carp)
730 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
731 else
732 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
733
734 /*
735 * Target address matches one of my interface address.
736 *
737 * If my address is tentative, this means that there's somebody
738 * already using the same address as mine. This indicates DAD failure.
739 * This is defined in RFC 2462.
740 *
741 * Otherwise, process as defined in RFC 2461.
742 */
743 if (ifa
744 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
745 nd6_dad_na_input(ifa);
746 ifa_free(ifa);
747 goto freeit;
748 }
749
750 /* Just for safety, maybe unnecessary. */
751 if (ifa) {
752 ifa_free(ifa);
753 log(LOG_ERR,
754 "nd6_na_input: duplicate IP6 address %s\n",
755 ip6_sprintf(ip6bufs, &taddr6));
756 goto freeit;
757 }
758
759 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
760 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
761 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
762 ifp->if_addrlen, lladdrlen - 2));
763 goto bad;
764 }
765
766 /*
767 * If no neighbor cache entry is found, NA SHOULD silently be
768 * discarded.
769 */
770 IF_AFDATA_RLOCK(ifp);
771 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
772 IF_AFDATA_RUNLOCK(ifp);
773 if (ln == NULL) {
774 goto freeit;
775 }
776
777 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
778 /*
779 * If the link-layer has address, and no lladdr option came,
780 * discard the packet.
781 */
782 if (ifp->if_addrlen && lladdr == NULL) {
783 goto freeit;
784 }
785
786 /*
787 * Record link-layer address, and update the state.
788 */
789 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
790 ln->la_flags |= LLE_VALID;
791 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
792 if (is_solicited) {
793 ln->ln_state = ND6_LLINFO_REACHABLE;
794 ln->ln_byhint = 0;
795 if (!ND6_LLINFO_PERMANENT(ln)) {
796 nd6_llinfo_settimer_locked(ln,
797 (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
798 }
799 } else {
800 ln->ln_state = ND6_LLINFO_STALE;
801 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
802 }
803 if ((ln->ln_router = is_router) != 0) {
804 /*
805 * This means a router's state has changed from
806 * non-reachable to probably reachable, and might
807 * affect the status of associated prefixes..
808 */
809 checklink = 1;
810 }
811 } else {
812 int llchange;
813
814 /*
815 * Check if the link-layer address has changed or not.
816 */
817 if (lladdr == NULL)
818 llchange = 0;
819 else {
820 if (ln->la_flags & LLE_VALID) {
821 if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
822 llchange = 1;
823 else
824 llchange = 0;
825 } else
826 llchange = 1;
827 }
828
829 /*
830 * This is VERY complex. Look at it with care.
831 *
832 * override solicit lladdr llchange action
833 * (L: record lladdr)
834 *
835 * 0 0 n -- (2c)
836 * 0 0 y n (2b) L
837 * 0 0 y y (1) REACHABLE->STALE
838 * 0 1 n -- (2c) *->REACHABLE
839 * 0 1 y n (2b) L *->REACHABLE
840 * 0 1 y y (1) REACHABLE->STALE
841 * 1 0 n -- (2a)
842 * 1 0 y n (2a) L
843 * 1 0 y y (2a) L *->STALE
844 * 1 1 n -- (2a) *->REACHABLE
845 * 1 1 y n (2a) L *->REACHABLE
846 * 1 1 y y (2a) L *->REACHABLE
847 */
848 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
849 /*
850 * If state is REACHABLE, make it STALE.
851 * no other updates should be done.
852 */
853 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
854 ln->ln_state = ND6_LLINFO_STALE;
855 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
856 }
857 goto freeit;
858 } else if (is_override /* (2a) */
859 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
860 || lladdr == NULL) { /* (2c) */
861 /*
862 * Update link-local address, if any.
863 */
864 if (lladdr != NULL) {
865 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
866 ln->la_flags |= LLE_VALID;
867 EVENTHANDLER_INVOKE(lle_event, ln,
868 LLENTRY_RESOLVED);
869 }
870
871 /*
872 * If solicited, make the state REACHABLE.
873 * If not solicited and the link-layer address was
874 * changed, make it STALE.
875 */
876 if (is_solicited) {
877 ln->ln_state = ND6_LLINFO_REACHABLE;
878 ln->ln_byhint = 0;
879 if (!ND6_LLINFO_PERMANENT(ln)) {
880 nd6_llinfo_settimer_locked(ln,
881 (long)ND_IFINFO(ifp)->reachable * hz);
882 }
883 } else {
884 if (lladdr != NULL && llchange) {
885 ln->ln_state = ND6_LLINFO_STALE;
886 nd6_llinfo_settimer_locked(ln,
887 (long)V_nd6_gctimer * hz);
888 }
889 }
890 }
891
892 if (ln->ln_router && !is_router) {
893 /*
894 * The peer dropped the router flag.
895 * Remove the sender from the Default Router List and
896 * update the Destination Cache entries.
897 */
898 if (!defrouter_remove(&L3_ADDR_SIN6(ln)->sin6_addr,
899 ln->lle_tbl->llt_ifp) &&
900 (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
901 ND6_IFF_ACCEPT_RTADV) != 0)
902 /*
903 * Even if the neighbor is not in the default
904 * router list, the neighbor may be used as a
905 * next hop for some destinations (e.g. redirect
906 * case). So we must call rt6_flush explicitly.
907 */
908 rt6_flush(&ip6->ip6_src, ifp);
909 }
910 ln->ln_router = is_router;
911 }
912 /* XXX - QL
913 * Does this matter?
914 * rt->rt_flags &= ~RTF_REJECT;
915 */
916 ln->la_asked = 0;
917 if (ln->la_hold != NULL)
918 nd6_grab_holdchain(ln, &chain, &sin6);
919 freeit:
920 if (ln != NULL)
921 LLE_WUNLOCK(ln);
922
923 if (chain != NULL)
924 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
925
926 if (checklink)
927 pfxlist_onlink_check();
928
929 m_freem(m);
930 return;
931
932 bad:
933 if (ln != NULL)
934 LLE_WUNLOCK(ln);
935
936 ICMP6STAT_INC(icp6s_badna);
937 m_freem(m);
938 }
939
940 /*
941 * Neighbor advertisement output handling.
942 *
943 * Based on RFC 2461
944 *
945 * the following items are not implemented yet:
946 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
947 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
948 *
949 * tlladdr - 1 if include target link-layer address
950 * sdl0 - sockaddr_dl (= proxy NA) or NULL
951 */
952 static void
953 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
954 const struct in6_addr *taddr6, u_long flags, int tlladdr,
955 struct sockaddr *sdl0, u_int fibnum)
956 {
957 struct mbuf *m;
958 struct m_tag *mtag;
959 struct ifnet *oifp;
960 struct ip6_hdr *ip6;
961 struct nd_neighbor_advert *nd_na;
962 struct ip6_moptions im6o;
963 struct in6_addr src, daddr6;
964 struct sockaddr_in6 dst_sa;
965 int icmp6len, maxlen, error;
966 caddr_t mac = NULL;
967 struct route_in6 ro;
968
969 bzero(&ro, sizeof(ro));
970
971 daddr6 = *daddr6_0; /* make a local copy for modification */
972
973 /* estimate the size of message */
974 maxlen = sizeof(*ip6) + sizeof(*nd_na);
975 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
976 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
977 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
978 __func__, max_linkhdr, maxlen, MCLBYTES));
979
980 if (max_linkhdr + maxlen > MHLEN)
981 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
982 else
983 m = m_gethdr(M_NOWAIT, MT_DATA);
984 if (m == NULL)
985 return;
986 M_SETFIB(m, fibnum);
987
988 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
989 m->m_flags |= M_MCAST;
990 im6o.im6o_multicast_ifp = ifp;
991 im6o.im6o_multicast_hlim = 255;
992 im6o.im6o_multicast_loop = 0;
993 }
994
995 icmp6len = sizeof(*nd_na);
996 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
997 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
998
999 /* fill neighbor advertisement packet */
1000 ip6 = mtod(m, struct ip6_hdr *);
1001 ip6->ip6_flow = 0;
1002 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1003 ip6->ip6_vfc |= IPV6_VERSION;
1004 ip6->ip6_nxt = IPPROTO_ICMPV6;
1005 ip6->ip6_hlim = 255;
1006 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1007 /* reply to DAD */
1008 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1009 daddr6.s6_addr16[1] = 0;
1010 daddr6.s6_addr32[1] = 0;
1011 daddr6.s6_addr32[2] = 0;
1012 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1013 if (in6_setscope(&daddr6, ifp, NULL))
1014 goto bad;
1015
1016 flags &= ~ND_NA_FLAG_SOLICITED;
1017 }
1018 ip6->ip6_dst = daddr6;
1019 bzero(&dst_sa, sizeof(struct sockaddr_in6));
1020 dst_sa.sin6_family = AF_INET6;
1021 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1022 dst_sa.sin6_addr = daddr6;
1023
1024 /*
1025 * Select a source whose scope is the same as that of the dest.
1026 */
1027 bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1028 oifp = ifp;
1029 error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1030 if (error) {
1031 char ip6buf[INET6_ADDRSTRLEN];
1032 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1033 "determined: dst=%s, error=%d\n",
1034 ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1035 goto bad;
1036 }
1037 ip6->ip6_src = src;
1038 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1039 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1040 nd_na->nd_na_code = 0;
1041 nd_na->nd_na_target = *taddr6;
1042 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1043
1044 /*
1045 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1046 * see nd6_ns_input() for details.
1047 * Basically, if NS packet is sent to unicast/anycast addr,
1048 * target lladdr option SHOULD NOT be included.
1049 */
1050 if (tlladdr) {
1051 /*
1052 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1053 * lladdr in sdl0. If we are not proxying (sending NA for
1054 * my address) use lladdr configured for the interface.
1055 */
1056 if (sdl0 == NULL) {
1057 if (ifp->if_carp)
1058 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1059 if (mac == NULL)
1060 mac = nd6_ifptomac(ifp);
1061 } else if (sdl0->sa_family == AF_LINK) {
1062 struct sockaddr_dl *sdl;
1063 sdl = (struct sockaddr_dl *)sdl0;
1064 if (sdl->sdl_alen == ifp->if_addrlen)
1065 mac = LLADDR(sdl);
1066 }
1067 }
1068 if (tlladdr && mac) {
1069 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1070 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1071
1072 /* roundup to 8 bytes alignment! */
1073 optlen = (optlen + 7) & ~7;
1074
1075 m->m_pkthdr.len += optlen;
1076 m->m_len += optlen;
1077 icmp6len += optlen;
1078 bzero((caddr_t)nd_opt, optlen);
1079 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1080 nd_opt->nd_opt_len = optlen >> 3;
1081 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1082 } else
1083 flags &= ~ND_NA_FLAG_OVERRIDE;
1084
1085 ip6->ip6_plen = htons((u_short)icmp6len);
1086 nd_na->nd_na_flags_reserved = flags;
1087 nd_na->nd_na_cksum = 0;
1088 nd_na->nd_na_cksum =
1089 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1090
1091 if (send_sendso_input_hook != NULL) {
1092 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1093 sizeof(unsigned short), M_NOWAIT);
1094 if (mtag == NULL)
1095 goto bad;
1096 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1097 m_tag_prepend(m, mtag);
1098 }
1099
1100 ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1101 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1102 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1103 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1104
1105 /* We don't cache this route. */
1106 RO_RTFREE(&ro);
1107
1108 return;
1109
1110 bad:
1111 if (ro.ro_rt) {
1112 RTFREE(ro.ro_rt);
1113 }
1114 m_freem(m);
1115 }
1116
1117 #ifndef BURN_BRIDGES
1118 void
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
1130 nd6_ifptomac(struct ifnet *ifp)
1131 {
1132 switch (ifp->if_type) {
1133 case IFT_ARCNET:
1134 case IFT_ETHER:
1135 case IFT_FDDI:
1136 case IFT_IEEE1394:
1137 #ifdef IFT_L2VLAN
1138 case IFT_L2VLAN:
1139 #endif
1140 #ifdef IFT_IEEE80211
1141 case IFT_IEEE80211:
1142 #endif
1143 case IFT_INFINIBAND:
1144 case IFT_BRIDGE:
1145 case IFT_ISO88025:
1146 return IF_LLADDR(ifp);
1147 default:
1148 return NULL;
1149 }
1150 }
1151
1152 struct dadq {
1153 TAILQ_ENTRY(dadq) dad_list;
1154 struct ifaddr *dad_ifa;
1155 int dad_count; /* max NS to send */
1156 int dad_ns_tcount; /* # of trials to send NS */
1157 int dad_ns_ocount; /* NS sent so far */
1158 int dad_ns_icount;
1159 int dad_na_icount;
1160 int dad_ns_lcount; /* looped back NS */
1161 int dad_loopbackprobe; /* probing state for loopback detection */
1162 struct callout dad_timer_ch;
1163 struct vnet *dad_vnet;
1164 u_int dad_refcnt;
1165 #define ND_OPT_NONCE_LEN32 \
1166 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1167 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1168 };
1169
1170 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1171 static VNET_DEFINE(struct rwlock, dad_rwlock);
1172 #define V_dadq VNET(dadq)
1173 #define V_dad_rwlock VNET(dad_rwlock)
1174
1175 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock)
1176 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock)
1177 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock)
1178 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock)
1179
1180 static void
1181 nd6_dad_add(struct dadq *dp)
1182 {
1183
1184 DADQ_WLOCK();
1185 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1186 DADQ_WUNLOCK();
1187 }
1188
1189 static void
1190 nd6_dad_del(struct dadq *dp)
1191 {
1192
1193 DADQ_WLOCK();
1194 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1195 DADQ_WUNLOCK();
1196 nd6_dad_rele(dp);
1197 }
1198
1199 static struct dadq *
1200 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1201 {
1202 struct dadq *dp;
1203
1204 DADQ_RLOCK();
1205 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1206 if (dp->dad_ifa != ifa)
1207 continue;
1208 /*
1209 * Skip if the nonce matches the received one.
1210 * +2 in the length is required because of type and
1211 * length fields are included in a header.
1212 */
1213 if (n != NULL &&
1214 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1215 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1216 ND_OPT_NONCE_LEN) == 0) {
1217 dp->dad_ns_lcount++;
1218 continue;
1219 }
1220 refcount_acquire(&dp->dad_refcnt);
1221 break;
1222 }
1223 DADQ_RUNLOCK();
1224
1225 return (dp);
1226 }
1227
1228 static void
1229 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1230 {
1231
1232 if (send_ns != 0)
1233 nd6_dad_ns_output(dp);
1234 callout_reset(&dp->dad_timer_ch, ticks,
1235 (void (*)(void *))nd6_dad_timer, (void *)dp);
1236 }
1237
1238 static void
1239 nd6_dad_stoptimer(struct dadq *dp)
1240 {
1241
1242 callout_drain(&dp->dad_timer_ch);
1243 }
1244
1245 static void
1246 nd6_dad_rele(struct dadq *dp)
1247 {
1248
1249 if (refcount_release(&dp->dad_refcnt)) {
1250 ifa_free(dp->dad_ifa);
1251 free(dp, M_IP6NDP);
1252 }
1253 }
1254
1255 void
1256 nd6_dad_init(void)
1257 {
1258
1259 rw_init(&V_dad_rwlock, "nd6 DAD queue");
1260 TAILQ_INIT(&V_dadq);
1261 }
1262
1263 /*
1264 * Start Duplicate Address Detection (DAD) for specified interface address.
1265 */
1266 void
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 int send_ns;
1273
1274 /*
1275 * If we don't need DAD, don't do it.
1276 * There are several cases:
1277 * - DAD is disabled (ip6_dad_count == 0)
1278 * - the interface address is anycast
1279 */
1280 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1281 log(LOG_DEBUG,
1282 "nd6_dad_start: called with non-tentative address "
1283 "%s(%s)\n",
1284 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1285 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1286 return;
1287 }
1288 if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1289 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1290 return;
1291 }
1292 if (!V_ip6_dad_count) {
1293 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1294 return;
1295 }
1296 if (ifa->ifa_ifp == NULL)
1297 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1298 if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) {
1299 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1300 return;
1301 }
1302 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1303 !(ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) ||
1304 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)) {
1305 ia->ia6_flags |= IN6_IFF_TENTATIVE;
1306 return;
1307 }
1308 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1309 /*
1310 * DAD is already in progress. Let the existing entry
1311 * finish it.
1312 */
1313 nd6_dad_rele(dp);
1314 return;
1315 }
1316
1317 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1318 if (dp == NULL) {
1319 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1320 "%s(%s)\n",
1321 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1322 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1323 return;
1324 }
1325 callout_init(&dp->dad_timer_ch, 0);
1326 #ifdef VIMAGE
1327 dp->dad_vnet = curvnet;
1328 #endif
1329 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1330 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1331
1332 /*
1333 * Send NS packet for DAD, ip6_dad_count times.
1334 * Note that we must delay the first transmission, if this is the
1335 * first packet to be sent from the interface after interface
1336 * (re)initialization.
1337 */
1338 dp->dad_ifa = ifa;
1339 ifa_ref(dp->dad_ifa);
1340 dp->dad_count = V_ip6_dad_count;
1341 dp->dad_ns_icount = dp->dad_na_icount = 0;
1342 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1343 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1344 refcount_init(&dp->dad_refcnt, 1);
1345 nd6_dad_add(dp);
1346 send_ns = 0;
1347 if (delay == 0) {
1348 send_ns = 1;
1349 delay = (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000;
1350 }
1351 nd6_dad_starttimer(dp, delay, send_ns);
1352 }
1353
1354 /*
1355 * terminate DAD unconditionally. used for address removals.
1356 */
1357 void
1358 nd6_dad_stop(struct ifaddr *ifa)
1359 {
1360 struct dadq *dp;
1361
1362 dp = nd6_dad_find(ifa, NULL);
1363 if (!dp) {
1364 /* DAD wasn't started yet */
1365 return;
1366 }
1367
1368 nd6_dad_stoptimer(dp);
1369
1370 /*
1371 * The DAD queue entry may have been removed by nd6_dad_timer() while
1372 * we were waiting for it to stop, so re-do the lookup.
1373 */
1374 nd6_dad_rele(dp);
1375 if (nd6_dad_find(ifa, NULL) == NULL)
1376 return;
1377
1378 nd6_dad_del(dp);
1379 nd6_dad_rele(dp);
1380 }
1381
1382 static void
1383 nd6_dad_timer(struct dadq *dp)
1384 {
1385 CURVNET_SET(dp->dad_vnet);
1386 struct ifaddr *ifa = dp->dad_ifa;
1387 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1388 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1389 char ip6buf[INET6_ADDRSTRLEN];
1390
1391 /* Sanity check */
1392 if (ia == NULL) {
1393 log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1394 goto err;
1395 }
1396 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1397 /* Do not need DAD for ifdisabled interface. */
1398 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1399 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1400 goto err;
1401 }
1402 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1403 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1404 "%s(%s)\n",
1405 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1406 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1407 goto err;
1408 }
1409 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1410 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1411 "%s(%s)\n",
1412 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1413 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1414 goto err;
1415 }
1416
1417 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1418 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1419 (((ifp->if_flags & IFF_UP) == 0) ||
1420 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1421 nd6log((LOG_INFO, "%s: could not run DAD "
1422 "because the interface was down or not running.\n",
1423 if_name(ifa->ifa_ifp)));
1424 goto err;
1425 }
1426
1427 /* Need more checks? */
1428 if (dp->dad_ns_ocount < dp->dad_count) {
1429 /*
1430 * We have more NS to go. Send NS packet for DAD.
1431 */
1432 nd6_dad_starttimer(dp,
1433 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1434 goto done;
1435 } else {
1436 /*
1437 * We have transmitted sufficient number of DAD packets.
1438 * See what we've got.
1439 */
1440 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1441 /* We've seen NS or NA, means DAD has failed. */
1442 nd6_dad_duplicated(ifa, dp);
1443 else if (V_dad_enhanced != 0 &&
1444 dp->dad_ns_lcount > 0 &&
1445 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1446 /*
1447 * Sec. 4.1 in RFC 7527 requires transmission of
1448 * additional probes until the loopback condition
1449 * becomes clear when a looped back probe is detected.
1450 */
1451 log(LOG_ERR, "%s: a looped back NS message is "
1452 "detected during DAD for %s. "
1453 "Another DAD probes are being sent.\n",
1454 if_name(ifa->ifa_ifp),
1455 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1456 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1457 /*
1458 * Send an NS immediately and increase dad_count by
1459 * V_nd6_mmaxtries - 1.
1460 */
1461 dp->dad_count =
1462 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1463 nd6_dad_starttimer(dp,
1464 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1465 1);
1466 goto done;
1467 } else {
1468 /*
1469 * We are done with DAD. No NA came, no NS came.
1470 * No duplicate address found. Check IFDISABLED flag
1471 * again in case that it is changed between the
1472 * beginning of this function and here.
1473 */
1474 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1475 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1476
1477 nd6log((LOG_DEBUG,
1478 "%s: DAD complete for %s - no duplicates found\n",
1479 if_name(ifa->ifa_ifp),
1480 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1481 if (dp->dad_ns_lcount > 0)
1482 log(LOG_ERR, "%s: DAD completed while "
1483 "a looped back NS message is detected "
1484 "during DAD for %s.\n",
1485 if_name(ifa->ifa_ifp),
1486 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1487 }
1488 }
1489 err:
1490 nd6_dad_del(dp);
1491 done:
1492 CURVNET_RESTORE();
1493 }
1494
1495 static void
1496 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1497 {
1498 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1499 struct ifnet *ifp;
1500 char ip6buf[INET6_ADDRSTRLEN];
1501
1502 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1503 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1504 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1505 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1506 dp->dad_na_icount);
1507
1508 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1509 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1510
1511 ifp = ifa->ifa_ifp;
1512 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1513 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1514 log(LOG_ERR, "%s: manual intervention required\n",
1515 if_name(ifp));
1516
1517 /*
1518 * If the address is a link-local address formed from an interface
1519 * identifier based on the hardware address which is supposed to be
1520 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1521 * operation on the interface SHOULD be disabled.
1522 * [RFC 4862, Section 5.4.5]
1523 */
1524 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1525 struct in6_addr in6;
1526
1527 /*
1528 * To avoid over-reaction, we only apply this logic when we are
1529 * very sure that hardware addresses are supposed to be unique.
1530 */
1531 switch (ifp->if_type) {
1532 case IFT_ETHER:
1533 case IFT_FDDI:
1534 case IFT_ATM:
1535 case IFT_IEEE1394:
1536 #ifdef IFT_IEEE80211
1537 case IFT_IEEE80211:
1538 #endif
1539 case IFT_INFINIBAND:
1540 in6 = ia->ia_addr.sin6_addr;
1541 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1542 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1543 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1544 log(LOG_ERR, "%s: possible hardware address "
1545 "duplication detected, disable IPv6\n",
1546 if_name(ifp));
1547 }
1548 break;
1549 }
1550 }
1551 }
1552
1553 static void
1554 nd6_dad_ns_output(struct dadq *dp)
1555 {
1556 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1557 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1558 int i;
1559
1560 dp->dad_ns_tcount++;
1561 if ((ifp->if_flags & IFF_UP) == 0) {
1562 return;
1563 }
1564 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1565 return;
1566 }
1567
1568 dp->dad_ns_ocount++;
1569 if (V_dad_enhanced != 0) {
1570 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1571 dp->dad_nonce[i] = arc4random();
1572 /*
1573 * XXXHRS: Note that in the case that
1574 * DupAddrDetectTransmits > 1, multiple NS messages with
1575 * different nonces can be looped back in an unexpected
1576 * order. The current implementation recognizes only
1577 * the latest nonce on the sender side. Practically it
1578 * should work well in almost all cases.
1579 */
1580 }
1581 nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL,
1582 (uint8_t *)&dp->dad_nonce[0]);
1583 }
1584
1585 static void
1586 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1587 {
1588 struct in6_ifaddr *ia;
1589 struct ifnet *ifp;
1590 const struct in6_addr *taddr6;
1591 struct dadq *dp;
1592
1593 if (ifa == NULL)
1594 panic("ifa == NULL in nd6_dad_ns_input");
1595
1596 ia = (struct in6_ifaddr *)ifa;
1597 ifp = ifa->ifa_ifp;
1598 taddr6 = &ia->ia_addr.sin6_addr;
1599 /* Ignore Nonce option when Enhanced DAD is disabled. */
1600 if (V_dad_enhanced == 0)
1601 ndopt_nonce = NULL;
1602 dp = nd6_dad_find(ifa, ndopt_nonce);
1603 if (dp == NULL)
1604 return;
1605
1606 dp->dad_ns_icount++;
1607 nd6_dad_rele(dp);
1608 }
1609
1610 static void
1611 nd6_dad_na_input(struct ifaddr *ifa)
1612 {
1613 struct dadq *dp;
1614
1615 if (ifa == NULL)
1616 panic("ifa == NULL in nd6_dad_na_input");
1617
1618 dp = nd6_dad_find(ifa, NULL);
1619 if (dp != NULL) {
1620 dp->dad_na_icount++;
1621 nd6_dad_rele(dp);
1622 }
1623 }
Cache object: 1eb9147400e99d40982a1df1d3da62fb
|