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("$FreeBSD$");
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);
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 *, struct ifaddr *);
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
nd6_ns_input(struct mbuf * m,int off,int icmp6len)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
nd6_ns_output_fib(struct ifnet * ifp,const struct in6_addr * daddr6,const struct in6_addr * taddr6,struct llentry * ln,uint8_t * nonce,u_int fibnum)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 draft-ietf-6man-enhanced-dad-13.
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 return;
627 }
628
629 #ifndef BURN_BRIDGES
630 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * daddr6,const struct in6_addr * taddr6,struct llentry * ln,uint8_t * nonce)631 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
632 const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce)
633 {
634
635 nd6_ns_output_fib(ifp, daddr6, taddr6, ln, 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 = m->m_pkthdr.rcvif;
652 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
653 struct nd_neighbor_advert *nd_na;
654 struct in6_addr daddr6 = ip6->ip6_dst;
655 struct in6_addr taddr6;
656 int flags;
657 int is_router;
658 int is_solicited;
659 int is_override;
660 char *lladdr = NULL;
661 int lladdrlen = 0;
662 int checklink = 0;
663 struct ifaddr *ifa;
664 struct llentry *ln = NULL;
665 union nd_opts ndopts;
666 struct mbuf *chain = NULL;
667 struct sockaddr_in6 sin6;
668 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
669
670 if (ip6->ip6_hlim != 255) {
671 nd6log((LOG_ERR,
672 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
673 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
674 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
675 goto bad;
676 }
677
678 #ifndef PULLDOWN_TEST
679 IP6_EXTHDR_CHECK(m, off, icmp6len,);
680 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
681 #else
682 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
683 if (nd_na == NULL) {
684 ICMP6STAT_INC(icp6s_tooshort);
685 return;
686 }
687 #endif
688
689 flags = nd_na->nd_na_flags_reserved;
690 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
691 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
692 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
693 memset(&sin6, 0, sizeof(sin6));
694
695 taddr6 = nd_na->nd_na_target;
696 if (in6_setscope(&taddr6, ifp, NULL))
697 goto bad; /* XXX: impossible */
698
699 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
700 nd6log((LOG_ERR,
701 "nd6_na_input: invalid target address %s\n",
702 ip6_sprintf(ip6bufs, &taddr6)));
703 goto bad;
704 }
705 if (IN6_IS_ADDR_MULTICAST(&daddr6))
706 if (is_solicited) {
707 nd6log((LOG_ERR,
708 "nd6_na_input: a solicited adv is multicasted\n"));
709 goto bad;
710 }
711
712 icmp6len -= sizeof(*nd_na);
713 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
714 if (nd6_options(&ndopts) < 0) {
715 nd6log((LOG_INFO,
716 "nd6_na_input: invalid ND option, ignored\n"));
717 /* nd6_options have incremented stats */
718 goto freeit;
719 }
720
721 if (ndopts.nd_opts_tgt_lladdr) {
722 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
723 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
724 }
725
726 /*
727 * This effectively disables the DAD check on a non-master CARP
728 * address.
729 */
730 if (ifp->if_carp)
731 ifa = (*carp_iamatch6_p)(ifp, &taddr6);
732 else
733 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
734
735 /*
736 * Target address matches one of my interface address.
737 *
738 * If my address is tentative, this means that there's somebody
739 * already using the same address as mine. This indicates DAD failure.
740 * This is defined in RFC 2462.
741 *
742 * Otherwise, process as defined in RFC 2461.
743 */
744 if (ifa
745 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
746 nd6_dad_na_input(ifa);
747 ifa_free(ifa);
748 goto freeit;
749 }
750
751 /* Just for safety, maybe unnecessary. */
752 if (ifa) {
753 ifa_free(ifa);
754 log(LOG_ERR,
755 "nd6_na_input: duplicate IP6 address %s\n",
756 ip6_sprintf(ip6bufs, &taddr6));
757 goto freeit;
758 }
759
760 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
761 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
762 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
763 ifp->if_addrlen, lladdrlen - 2));
764 goto bad;
765 }
766
767 /*
768 * If no neighbor cache entry is found, NA SHOULD silently be
769 * discarded.
770 */
771 IF_AFDATA_RLOCK(ifp);
772 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
773 IF_AFDATA_RUNLOCK(ifp);
774 if (ln == NULL) {
775 goto freeit;
776 }
777
778 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
779 /*
780 * If the link-layer has address, and no lladdr option came,
781 * discard the packet.
782 */
783 if (ifp->if_addrlen && lladdr == NULL) {
784 goto freeit;
785 }
786
787 /*
788 * Record link-layer address, and update the state.
789 */
790 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
791 ln->la_flags |= LLE_VALID;
792 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
793 if (is_solicited) {
794 ln->ln_state = ND6_LLINFO_REACHABLE;
795 ln->ln_byhint = 0;
796 if (!ND6_LLINFO_PERMANENT(ln)) {
797 nd6_llinfo_settimer_locked(ln,
798 (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
799 }
800 } else {
801 ln->ln_state = ND6_LLINFO_STALE;
802 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
803 }
804 if ((ln->ln_router = is_router) != 0) {
805 /*
806 * This means a router's state has changed from
807 * non-reachable to probably reachable, and might
808 * affect the status of associated prefixes..
809 */
810 checklink = 1;
811 }
812 } else {
813 int llchange;
814
815 /*
816 * Check if the link-layer address has changed or not.
817 */
818 if (lladdr == NULL)
819 llchange = 0;
820 else {
821 if (ln->la_flags & LLE_VALID) {
822 if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
823 llchange = 1;
824 else
825 llchange = 0;
826 } else
827 llchange = 1;
828 }
829
830 /*
831 * This is VERY complex. Look at it with care.
832 *
833 * override solicit lladdr llchange action
834 * (L: record lladdr)
835 *
836 * 0 0 n -- (2c)
837 * 0 0 y n (2b) L
838 * 0 0 y y (1) REACHABLE->STALE
839 * 0 1 n -- (2c) *->REACHABLE
840 * 0 1 y n (2b) L *->REACHABLE
841 * 0 1 y y (1) REACHABLE->STALE
842 * 1 0 n -- (2a)
843 * 1 0 y n (2a) L
844 * 1 0 y y (2a) L *->STALE
845 * 1 1 n -- (2a) *->REACHABLE
846 * 1 1 y n (2a) L *->REACHABLE
847 * 1 1 y y (2a) L *->REACHABLE
848 */
849 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
850 /*
851 * If state is REACHABLE, make it STALE.
852 * no other updates should be done.
853 */
854 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
855 ln->ln_state = ND6_LLINFO_STALE;
856 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
857 }
858 goto freeit;
859 } else if (is_override /* (2a) */
860 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
861 || lladdr == NULL) { /* (2c) */
862 /*
863 * Update link-local address, if any.
864 */
865 if (lladdr != NULL) {
866 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
867 ln->la_flags |= LLE_VALID;
868 EVENTHANDLER_INVOKE(lle_event, ln,
869 LLENTRY_RESOLVED);
870 }
871
872 /*
873 * If solicited, make the state REACHABLE.
874 * If not solicited and the link-layer address was
875 * changed, make it STALE.
876 */
877 if (is_solicited) {
878 ln->ln_state = ND6_LLINFO_REACHABLE;
879 ln->ln_byhint = 0;
880 if (!ND6_LLINFO_PERMANENT(ln)) {
881 nd6_llinfo_settimer_locked(ln,
882 (long)ND_IFINFO(ifp)->reachable * hz);
883 }
884 } else {
885 if (lladdr != NULL && llchange) {
886 ln->ln_state = ND6_LLINFO_STALE;
887 nd6_llinfo_settimer_locked(ln,
888 (long)V_nd6_gctimer * hz);
889 }
890 }
891 }
892
893 if (ln->ln_router && !is_router) {
894 /*
895 * The peer dropped the router flag.
896 * Remove the sender from the Default Router List and
897 * update the Destination Cache entries.
898 */
899 struct nd_defrouter *dr;
900 struct in6_addr *in6;
901
902 in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
903
904 /*
905 * Lock to protect the default router list.
906 * XXX: this might be unnecessary, since this function
907 * is only called under the network software interrupt
908 * context. However, we keep it just for safety.
909 */
910 dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
911 if (dr)
912 defrtrlist_del(dr);
913 else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
914 ND6_IFF_ACCEPT_RTADV) {
915 /*
916 * Even if the neighbor is not in the default
917 * router list, the neighbor may be used
918 * as a next hop for some destinations
919 * (e.g. redirect case). So we must
920 * call rt6_flush explicitly.
921 */
922 rt6_flush(&ip6->ip6_src, ifp);
923 }
924 }
925 ln->ln_router = is_router;
926 }
927 /* XXX - QL
928 * Does this matter?
929 * rt->rt_flags &= ~RTF_REJECT;
930 */
931 ln->la_asked = 0;
932 if (ln->la_hold != NULL)
933 nd6_grab_holdchain(ln, &chain, &sin6);
934 freeit:
935 if (ln != NULL)
936 LLE_WUNLOCK(ln);
937
938 if (chain != NULL)
939 nd6_flush_holdchain(ifp, ifp, chain, &sin6);
940
941 if (checklink)
942 pfxlist_onlink_check();
943
944 m_freem(m);
945 return;
946
947 bad:
948 if (ln != NULL)
949 LLE_WUNLOCK(ln);
950
951 ICMP6STAT_INC(icp6s_badna);
952 m_freem(m);
953 }
954
955 /*
956 * Neighbor advertisement output handling.
957 *
958 * Based on RFC 2461
959 *
960 * the following items are not implemented yet:
961 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
962 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
963 *
964 * tlladdr - 1 if include target link-layer address
965 * sdl0 - sockaddr_dl (= proxy NA) or NULL
966 */
967 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)968 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
969 const struct in6_addr *taddr6, u_long flags, int tlladdr,
970 struct sockaddr *sdl0, u_int fibnum)
971 {
972 struct mbuf *m;
973 struct m_tag *mtag;
974 struct ifnet *oifp;
975 struct ip6_hdr *ip6;
976 struct nd_neighbor_advert *nd_na;
977 struct ip6_moptions im6o;
978 struct in6_addr src, daddr6;
979 struct sockaddr_in6 dst_sa;
980 int icmp6len, maxlen, error;
981 caddr_t mac = NULL;
982 struct route_in6 ro;
983
984 bzero(&ro, sizeof(ro));
985
986 daddr6 = *daddr6_0; /* make a local copy for modification */
987
988 /* estimate the size of message */
989 maxlen = sizeof(*ip6) + sizeof(*nd_na);
990 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
991 KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
992 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
993 __func__, max_linkhdr, maxlen, MCLBYTES));
994
995 if (max_linkhdr + maxlen > MHLEN)
996 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
997 else
998 m = m_gethdr(M_NOWAIT, MT_DATA);
999 if (m == NULL)
1000 return;
1001 M_SETFIB(m, fibnum);
1002
1003 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1004 m->m_flags |= M_MCAST;
1005 im6o.im6o_multicast_ifp = ifp;
1006 im6o.im6o_multicast_hlim = 255;
1007 im6o.im6o_multicast_loop = 0;
1008 }
1009
1010 icmp6len = sizeof(*nd_na);
1011 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1012 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
1013
1014 /* fill neighbor advertisement packet */
1015 ip6 = mtod(m, struct ip6_hdr *);
1016 ip6->ip6_flow = 0;
1017 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1018 ip6->ip6_vfc |= IPV6_VERSION;
1019 ip6->ip6_nxt = IPPROTO_ICMPV6;
1020 ip6->ip6_hlim = 255;
1021 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1022 /* reply to DAD */
1023 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1024 daddr6.s6_addr16[1] = 0;
1025 daddr6.s6_addr32[1] = 0;
1026 daddr6.s6_addr32[2] = 0;
1027 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1028 if (in6_setscope(&daddr6, ifp, NULL))
1029 goto bad;
1030
1031 flags &= ~ND_NA_FLAG_SOLICITED;
1032 }
1033 ip6->ip6_dst = daddr6;
1034 bzero(&dst_sa, sizeof(struct sockaddr_in6));
1035 dst_sa.sin6_family = AF_INET6;
1036 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1037 dst_sa.sin6_addr = daddr6;
1038
1039 /*
1040 * Select a source whose scope is the same as that of the dest.
1041 */
1042 bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1043 oifp = ifp;
1044 error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1045 if (error) {
1046 char ip6buf[INET6_ADDRSTRLEN];
1047 nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1048 "determined: dst=%s, error=%d\n",
1049 ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1050 goto bad;
1051 }
1052 ip6->ip6_src = src;
1053 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1054 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1055 nd_na->nd_na_code = 0;
1056 nd_na->nd_na_target = *taddr6;
1057 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1058
1059 /*
1060 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1061 * see nd6_ns_input() for details.
1062 * Basically, if NS packet is sent to unicast/anycast addr,
1063 * target lladdr option SHOULD NOT be included.
1064 */
1065 if (tlladdr) {
1066 /*
1067 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1068 * lladdr in sdl0. If we are not proxying (sending NA for
1069 * my address) use lladdr configured for the interface.
1070 */
1071 if (sdl0 == NULL) {
1072 if (ifp->if_carp)
1073 mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1074 if (mac == NULL)
1075 mac = nd6_ifptomac(ifp);
1076 } else if (sdl0->sa_family == AF_LINK) {
1077 struct sockaddr_dl *sdl;
1078 sdl = (struct sockaddr_dl *)sdl0;
1079 if (sdl->sdl_alen == ifp->if_addrlen)
1080 mac = LLADDR(sdl);
1081 }
1082 }
1083 if (tlladdr && mac) {
1084 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1085 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1086
1087 /* roundup to 8 bytes alignment! */
1088 optlen = (optlen + 7) & ~7;
1089
1090 m->m_pkthdr.len += optlen;
1091 m->m_len += optlen;
1092 icmp6len += optlen;
1093 bzero((caddr_t)nd_opt, optlen);
1094 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1095 nd_opt->nd_opt_len = optlen >> 3;
1096 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1097 } else
1098 flags &= ~ND_NA_FLAG_OVERRIDE;
1099
1100 ip6->ip6_plen = htons((u_short)icmp6len);
1101 nd_na->nd_na_flags_reserved = flags;
1102 nd_na->nd_na_cksum = 0;
1103 nd_na->nd_na_cksum =
1104 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1105
1106 if (send_sendso_input_hook != NULL) {
1107 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1108 sizeof(unsigned short), M_NOWAIT);
1109 if (mtag == NULL)
1110 goto bad;
1111 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1112 m_tag_prepend(m, mtag);
1113 }
1114
1115 ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1116 icmp6_ifstat_inc(ifp, ifs6_out_msg);
1117 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1118 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1119
1120 /* We don't cache this route. */
1121 RO_RTFREE(&ro);
1122
1123 return;
1124
1125 bad:
1126 if (ro.ro_rt) {
1127 RTFREE(ro.ro_rt);
1128 }
1129 m_freem(m);
1130 return;
1131 }
1132
1133 #ifndef BURN_BRIDGES
1134 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)1135 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1136 const struct in6_addr *taddr6, u_long flags, int tlladdr,
1137 struct sockaddr *sdl0)
1138 {
1139
1140 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1141 RT_DEFAULT_FIB);
1142 }
1143 #endif
1144
1145 caddr_t
nd6_ifptomac(struct ifnet * ifp)1146 nd6_ifptomac(struct ifnet *ifp)
1147 {
1148 switch (ifp->if_type) {
1149 case IFT_ARCNET:
1150 case IFT_ETHER:
1151 case IFT_FDDI:
1152 case IFT_IEEE1394:
1153 #ifdef IFT_L2VLAN
1154 case IFT_L2VLAN:
1155 #endif
1156 #ifdef IFT_IEEE80211
1157 case IFT_IEEE80211:
1158 #endif
1159 case IFT_INFINIBAND:
1160 case IFT_BRIDGE:
1161 case IFT_ISO88025:
1162 return IF_LLADDR(ifp);
1163 default:
1164 return NULL;
1165 }
1166 }
1167
1168 struct dadq {
1169 TAILQ_ENTRY(dadq) dad_list;
1170 struct ifaddr *dad_ifa;
1171 int dad_count; /* max NS to send */
1172 int dad_ns_tcount; /* # of trials to send NS */
1173 int dad_ns_ocount; /* NS sent so far */
1174 int dad_ns_icount;
1175 int dad_na_icount;
1176 int dad_ns_lcount; /* looped back NS */
1177 int dad_loopbackprobe; /* probing state for loopback detection */
1178 struct callout dad_timer_ch;
1179 struct vnet *dad_vnet;
1180 u_int dad_refcnt;
1181 #define ND_OPT_NONCE_LEN32 \
1182 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1183 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1184 };
1185
1186 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1187 static VNET_DEFINE(struct rwlock, dad_rwlock);
1188 #define V_dadq VNET(dadq)
1189 #define V_dad_rwlock VNET(dad_rwlock)
1190
1191 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock)
1192 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock)
1193 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock)
1194 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock)
1195
1196 static void
nd6_dad_add(struct dadq * dp)1197 nd6_dad_add(struct dadq *dp)
1198 {
1199
1200 DADQ_WLOCK();
1201 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1202 DADQ_WUNLOCK();
1203 }
1204
1205 static void
nd6_dad_del(struct dadq * dp)1206 nd6_dad_del(struct dadq *dp)
1207 {
1208
1209 DADQ_WLOCK();
1210 TAILQ_REMOVE(&V_dadq, dp, dad_list);
1211 DADQ_WUNLOCK();
1212 nd6_dad_rele(dp);
1213 }
1214
1215 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * n)1216 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1217 {
1218 struct dadq *dp;
1219
1220 DADQ_RLOCK();
1221 TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1222 if (dp->dad_ifa != ifa)
1223 continue;
1224 /*
1225 * Skip if the nonce matches the received one.
1226 * +2 in the length is required because of type and
1227 * length fields are included in a header.
1228 */
1229 if (n != NULL &&
1230 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1231 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1232 ND_OPT_NONCE_LEN) == 0) {
1233 dp->dad_ns_lcount++;
1234 continue;
1235 }
1236 refcount_acquire(&dp->dad_refcnt);
1237 break;
1238 }
1239 DADQ_RUNLOCK();
1240
1241 return (dp);
1242 }
1243
1244 static void
nd6_dad_starttimer(struct dadq * dp,int ticks)1245 nd6_dad_starttimer(struct dadq *dp, int ticks)
1246 {
1247
1248 callout_reset(&dp->dad_timer_ch, ticks,
1249 (void (*)(void *))nd6_dad_timer, (void *)dp);
1250 }
1251
1252 static void
nd6_dad_stoptimer(struct dadq * dp)1253 nd6_dad_stoptimer(struct dadq *dp)
1254 {
1255
1256 callout_drain(&dp->dad_timer_ch);
1257 }
1258
1259 static void
nd6_dad_rele(struct dadq * dp)1260 nd6_dad_rele(struct dadq *dp)
1261 {
1262
1263 if (refcount_release(&dp->dad_refcnt)) {
1264 ifa_free(dp->dad_ifa);
1265 free(dp, M_IP6NDP);
1266 }
1267 }
1268
1269 void
nd6_dad_init(void)1270 nd6_dad_init(void)
1271 {
1272
1273 rw_init(&V_dad_rwlock, "nd6 DAD queue");
1274 TAILQ_INIT(&V_dadq);
1275 }
1276
1277 /*
1278 * Start Duplicate Address Detection (DAD) for specified interface address.
1279 */
1280 void
nd6_dad_start(struct ifaddr * ifa,int delay)1281 nd6_dad_start(struct ifaddr *ifa, int delay)
1282 {
1283 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1284 struct dadq *dp;
1285 char ip6buf[INET6_ADDRSTRLEN];
1286
1287 /*
1288 * If we don't need DAD, don't do it.
1289 * There are several cases:
1290 * - DAD is disabled (ip6_dad_count == 0)
1291 * - the interface address is anycast
1292 */
1293 if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1294 log(LOG_DEBUG,
1295 "nd6_dad_start: called with non-tentative address "
1296 "%s(%s)\n",
1297 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1298 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1299 return;
1300 }
1301 if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1302 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1303 return;
1304 }
1305 if (!V_ip6_dad_count) {
1306 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1307 return;
1308 }
1309 if (ifa->ifa_ifp == NULL)
1310 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1311 if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1312 return;
1313 }
1314 if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1315 return;
1316 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1317 /* DAD already in progress */
1318 nd6_dad_rele(dp);
1319 return;
1320 }
1321
1322 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1323 if (dp == NULL) {
1324 log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1325 "%s(%s)\n",
1326 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1327 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1328 return;
1329 }
1330 callout_init(&dp->dad_timer_ch, 0);
1331 #ifdef VIMAGE
1332 dp->dad_vnet = curvnet;
1333 #endif
1334 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1335 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1336
1337 /*
1338 * Send NS packet for DAD, ip6_dad_count times.
1339 * Note that we must delay the first transmission, if this is the
1340 * first packet to be sent from the interface after interface
1341 * (re)initialization.
1342 */
1343 dp->dad_ifa = ifa;
1344 ifa_ref(dp->dad_ifa);
1345 dp->dad_count = V_ip6_dad_count;
1346 dp->dad_ns_icount = dp->dad_na_icount = 0;
1347 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1348 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1349 refcount_init(&dp->dad_refcnt, 1);
1350 nd6_dad_add(dp);
1351 if (delay == 0) {
1352 nd6_dad_ns_output(dp, ifa);
1353 nd6_dad_starttimer(dp,
1354 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1355 } else {
1356 nd6_dad_starttimer(dp, delay);
1357 }
1358 }
1359
1360 /*
1361 * terminate DAD unconditionally. used for address removals.
1362 */
1363 void
nd6_dad_stop(struct ifaddr * ifa)1364 nd6_dad_stop(struct ifaddr *ifa)
1365 {
1366 struct dadq *dp;
1367
1368 dp = nd6_dad_find(ifa, NULL);
1369 if (!dp) {
1370 /* DAD wasn't started yet */
1371 return;
1372 }
1373
1374 nd6_dad_stoptimer(dp);
1375
1376 /*
1377 * The DAD queue entry may have been removed by nd6_dad_timer() while
1378 * we were waiting for it to stop, so re-do the lookup.
1379 */
1380 nd6_dad_rele(dp);
1381 if (nd6_dad_find(ifa, NULL) == NULL)
1382 return;
1383
1384 nd6_dad_del(dp);
1385 nd6_dad_rele(dp);
1386 }
1387
1388 static void
nd6_dad_timer(struct dadq * dp)1389 nd6_dad_timer(struct dadq *dp)
1390 {
1391 CURVNET_SET(dp->dad_vnet);
1392 struct ifaddr *ifa = dp->dad_ifa;
1393 struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1394 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1395 char ip6buf[INET6_ADDRSTRLEN];
1396
1397 /* Sanity check */
1398 if (ia == NULL) {
1399 log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1400 goto err;
1401 }
1402 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1403 /* Do not need DAD for ifdisabled interface. */
1404 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1405 "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1406 goto err;
1407 }
1408 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1409 log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1410 "%s(%s)\n",
1411 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1412 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1413 goto err;
1414 }
1415 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1416 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1417 "%s(%s)\n",
1418 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1419 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1420 goto err;
1421 }
1422
1423 /* Stop DAD if the interface is down even after dad_maxtry attempts. */
1424 if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1425 (((ifp->if_flags & IFF_UP) == 0) ||
1426 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1427 nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1428 if_name(ifa->ifa_ifp)));
1429 goto err;
1430 }
1431
1432 /* Need more checks? */
1433 if (dp->dad_ns_ocount < dp->dad_count) {
1434 /*
1435 * We have more NS to go. Send NS packet for DAD.
1436 */
1437 nd6_dad_ns_output(dp, ifa);
1438 nd6_dad_starttimer(dp,
1439 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1440 goto done;
1441 } else {
1442 /*
1443 * We have transmitted sufficient number of DAD packets.
1444 * See what we've got.
1445 */
1446 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1447 /* We've seen NS or NA, means DAD has failed. */
1448 nd6_dad_duplicated(ifa, dp);
1449 else if (V_dad_enhanced != 0 &&
1450 dp->dad_ns_lcount > 0 &&
1451 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1452 /*
1453 * A looped back probe is detected,
1454 * Sec. 4.1 in draft-ietf-6man-enhanced-dad-13
1455 * requires transmission of additional probes until
1456 * the loopback condition becomes clear.
1457 */
1458 log(LOG_ERR, "%s: a looped back NS message is "
1459 "detected during DAD for %s. "
1460 "Another DAD probes are being sent.\n",
1461 if_name(ifa->ifa_ifp),
1462 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1463 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1464 /*
1465 * An interface with IGNORELOOP is one which a
1466 * loopback is permanently expected while regular
1467 * traffic works. In that case, stop DAD after
1468 * MAX_MULTICAST_SOLICIT number of NS messages
1469 * regardless of the number of received loopback NS
1470 * by increasing dad_loopbackprobe in advance.
1471 */
1472 if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IGNORELOOP)
1473 dp->dad_loopbackprobe += V_nd6_mmaxtries;
1474 /*
1475 * Send an NS immediately and increase dad_count by
1476 * V_nd6_mmaxtries - 1.
1477 */
1478 nd6_dad_ns_output(dp, ifa);
1479 dp->dad_count =
1480 dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1481 nd6_dad_starttimer(dp,
1482 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1483 goto done;
1484 } else {
1485 /*
1486 * We are done with DAD. No NA came, no NS came.
1487 * No duplicate address found. Check IFDISABLED flag
1488 * again in case that it is changed between the
1489 * beginning of this function and here.
1490 */
1491 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1492 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1493
1494 nd6log((LOG_DEBUG,
1495 "%s: DAD complete for %s - no duplicates found\n",
1496 if_name(ifa->ifa_ifp),
1497 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1498 if (dp->dad_ns_lcount > 0)
1499 log(LOG_ERR, "%s: DAD completed while "
1500 "a looped back NS message is detected "
1501 "during DAD for %s.\n",
1502 if_name(ifa->ifa_ifp),
1503 ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1504 }
1505 }
1506 err:
1507 nd6_dad_del(dp);
1508 done:
1509 CURVNET_RESTORE();
1510 }
1511
1512 static void
nd6_dad_duplicated(struct ifaddr * ifa,struct dadq * dp)1513 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1514 {
1515 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1516 struct ifnet *ifp;
1517 char ip6buf[INET6_ADDRSTRLEN];
1518
1519 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1520 "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1521 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1522 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1523 dp->dad_na_icount);
1524
1525 ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1526 ia->ia6_flags |= IN6_IFF_DUPLICATED;
1527
1528 ifp = ifa->ifa_ifp;
1529 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1530 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1531 log(LOG_ERR, "%s: manual intervention required\n",
1532 if_name(ifp));
1533
1534 /*
1535 * If the address is a link-local address formed from an interface
1536 * identifier based on the hardware address which is supposed to be
1537 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1538 * operation on the interface SHOULD be disabled.
1539 * [RFC 4862, Section 5.4.5]
1540 */
1541 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1542 struct in6_addr in6;
1543
1544 /*
1545 * To avoid over-reaction, we only apply this logic when we are
1546 * very sure that hardware addresses are supposed to be unique.
1547 */
1548 switch (ifp->if_type) {
1549 case IFT_ETHER:
1550 case IFT_FDDI:
1551 case IFT_ATM:
1552 case IFT_IEEE1394:
1553 #ifdef IFT_IEEE80211
1554 case IFT_IEEE80211:
1555 #endif
1556 case IFT_INFINIBAND:
1557 in6 = ia->ia_addr.sin6_addr;
1558 if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1559 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1560 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1561 log(LOG_ERR, "%s: possible hardware address "
1562 "duplication detected, disable IPv6\n",
1563 if_name(ifp));
1564 }
1565 break;
1566 }
1567 }
1568 }
1569
1570 static void
nd6_dad_ns_output(struct dadq * dp,struct ifaddr * ifa)1571 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1572 {
1573 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1574 struct ifnet *ifp = ifa->ifa_ifp;
1575 int i;
1576
1577 dp->dad_ns_tcount++;
1578 if ((ifp->if_flags & IFF_UP) == 0) {
1579 return;
1580 }
1581 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1582 return;
1583 }
1584
1585 dp->dad_ns_ocount++;
1586 if (V_dad_enhanced != 0) {
1587 for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1588 dp->dad_nonce[i] = arc4random();
1589 /*
1590 * XXXHRS: Note that in the case that
1591 * DupAddrDetectTransmits > 1, multiple NS messages with
1592 * different nonces can be looped back in an unexpected
1593 * order. The current implementation recognizes only
1594 * the latest nonce on the sender side. Practically it
1595 * should work well in almost all cases.
1596 */
1597 }
1598 nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL,
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 in6_ifaddr *ia;
1606 struct ifnet *ifp;
1607 const struct in6_addr *taddr6;
1608 struct dadq *dp;
1609
1610 if (ifa == NULL)
1611 panic("ifa == NULL in nd6_dad_ns_input");
1612
1613 ia = (struct in6_ifaddr *)ifa;
1614 ifp = ifa->ifa_ifp;
1615 taddr6 = &ia->ia_addr.sin6_addr;
1616 /* Ignore Nonce option when Enhanced DAD is disabled. */
1617 if (V_dad_enhanced == 0)
1618 ndopt_nonce = NULL;
1619 dp = nd6_dad_find(ifa, ndopt_nonce);
1620 if (dp == NULL)
1621 return;
1622
1623 dp->dad_ns_icount++;
1624 nd6_dad_rele(dp);
1625 }
1626
1627 static void
nd6_dad_na_input(struct ifaddr * ifa)1628 nd6_dad_na_input(struct ifaddr *ifa)
1629 {
1630 struct dadq *dp;
1631
1632 if (ifa == NULL)
1633 panic("ifa == NULL in nd6_dad_na_input");
1634
1635 dp = nd6_dad_find(ifa, NULL);
1636 if (dp != NULL) {
1637 dp->dad_na_icount++;
1638 nd6_dad_rele(dp);
1639 }
1640 }
1641