1 /*
2 * Copyright (c) 2000-2021 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /*
30 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. Neither the name of the project nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
44 *
45 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55 * SUCH DAMAGE.
56 */
57
58 #include <sys/param.h>
59 #include <sys/systm.h>
60 #include <sys/malloc.h>
61 #include <sys/mbuf.h>
62 #include <sys/socket.h>
63 #include <sys/sockio.h>
64 #include <sys/time.h>
65 #include <sys/kernel.h>
66 #include <sys/errno.h>
67 #include <sys/syslog.h>
68 #include <sys/sysctl.h>
69 #include <sys/mcache.h>
70 #include <sys/protosw.h>
71 #include <kern/queue.h>
72 #include <dev/random/randomdev.h>
73
74 #include <kern/locks.h>
75 #include <kern/zalloc.h>
76
77 #include <net/if.h>
78 #include <net/if_var.h>
79 #include <net/if_types.h>
80 #include <net/if_dl.h>
81 #include <net/if_llreach.h>
82 #include <net/route.h>
83 #include <net/dlil.h>
84 #include <net/nwk_wq.h>
85
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet6/in6_var.h>
89 #include <netinet6/in6_ifattach.h>
90 #include <netinet/ip6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/nd6.h>
93 #include <netinet6/scope6_var.h>
94 #include <netinet/icmp6.h>
95
96 #if IPSEC
97 #include <netinet6/ipsec.h>
98 #include <netinet6/ipsec6.h>
99 #endif
100
101 struct dadq;
102 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
103 void nd6_dad_stoptimer(struct ifaddr *);
104 static void nd6_dad_timer(struct ifaddr *);
105 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
106 static void nd6_dad_ns_input(struct ifaddr *, char *, int, struct nd_opt_nonce *);
107 static struct mbuf *nd6_dad_na_input(struct mbuf *, struct ifnet *,
108 struct in6_addr *, caddr_t, int);
109 static void dad_addref(struct dadq *, int);
110 static void dad_remref(struct dadq *);
111 static struct dadq *nd6_dad_attach(struct dadq *, struct ifaddr *);
112 static void nd6_dad_detach(struct dadq *, struct ifaddr *);
113 static void nd6_dad_duplicated(struct ifaddr *);
114
115 static int dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */
116
117 #define DAD_LOCK_ASSERT_HELD(_dp) \
118 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_OWNED)
119
120 #define DAD_LOCK_ASSERT_NOTHELD(_dp) \
121 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_NOTOWNED)
122
123 #define DAD_LOCK(_dp) \
124 lck_mtx_lock(&(_dp)->dad_lock)
125
126 #define DAD_LOCK_SPIN(_dp) \
127 lck_mtx_lock_spin(&(_dp)->dad_lock)
128
129 #define DAD_CONVERT_LOCK(_dp) do { \
130 DAD_LOCK_ASSERT_HELD(_dp); \
131 lck_mtx_convert_spin(&(_dp)->dad_lock); \
132 } while (0)
133
134 #define DAD_UNLOCK(_dp) \
135 lck_mtx_unlock(&(_dp)->dad_lock)
136
137 #define DAD_ADDREF(_dp) \
138 dad_addref(_dp, 0)
139
140 #define DAD_ADDREF_LOCKED(_dp) \
141 dad_addref(_dp, 1)
142
143 #define DAD_REMREF(_dp) \
144 dad_remref(_dp)
145
146 static LCK_MTX_DECLARE_ATTR(dad6_mutex, &ip6_mutex_grp, &ip6_mutex_attr);
147
148 static int nd6_llreach_base = 30; /* seconds */
149
150 static struct sockaddr_in6 hostrtmask;
151
152 SYSCTL_DECL(_net_inet6_icmp6);
153 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_llreach_base,
154 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_llreach_base, 0,
155 "default ND6 link-layer reachability max lifetime (in seconds)");
156
157 int dad_enhanced = ND6_DAD_ENHANCED_DEFAULT;
158 SYSCTL_DECL(_net_inet6_ip6);
159 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_RW | CTLFLAG_LOCKED,
160 &dad_enhanced, 0,
161 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
162
163 static int ip6_p2p_debug = 0;
164 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, ip6_p2p_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
165 &ip6_p2p_debug, 0,
166 "Enable more instrumentation for IPv6 P2P use-case");
167 /*
168 * Obtain a link-layer source cache entry for the sender.
169 *
170 * NOTE: This is currently only for ND6/Ethernet.
171 */
172 void
nd6_llreach_alloc(struct rtentry * rt,struct ifnet * ifp,void * addr,unsigned int alen,boolean_t solicited)173 nd6_llreach_alloc(struct rtentry *rt, struct ifnet *ifp, void *addr,
174 unsigned int alen, boolean_t solicited)
175 {
176 struct llinfo_nd6 *ln = rt->rt_llinfo;
177
178 if (nd6_llreach_base != 0 &&
179 (ln->ln_expire != 0 || (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) &&
180 !(rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
181 ifp->if_addrlen == IF_LLREACH_MAXLEN && /* Ethernet */
182 alen == ifp->if_addrlen) {
183 struct if_llreach *lr;
184 const char *why = NULL, *type = "";
185
186 /* Become a regular mutex, just in case */
187 RT_CONVERT_LOCK(rt);
188
189 if ((lr = ln->ln_llreach) != NULL) {
190 type = (solicited ? "ND6 advertisement" :
191 "ND6 unsolicited announcement");
192 /*
193 * If target has changed, create a new record;
194 * otherwise keep existing record.
195 */
196 IFLR_LOCK(lr);
197 if (bcmp(addr, lr->lr_key.addr, alen) != 0) {
198 IFLR_UNLOCK(lr);
199 /* Purge any link-layer info caching */
200 VERIFY(rt->rt_llinfo_purge != NULL);
201 rt->rt_llinfo_purge(rt);
202 lr = NULL;
203 why = " for different target HW address; "
204 "using new llreach record";
205 } else {
206 lr->lr_probes = 0; /* reset probe count */
207 IFLR_UNLOCK(lr);
208 if (solicited) {
209 why = " for same target HW address; "
210 "keeping existing llreach record";
211 }
212 }
213 }
214
215 if (lr == NULL) {
216 lr = ln->ln_llreach = ifnet_llreach_alloc(ifp,
217 ETHERTYPE_IPV6, addr, alen, nd6_llreach_base);
218 if (lr != NULL) {
219 lr->lr_probes = 0; /* reset probe count */
220 if (why == NULL) {
221 why = "creating new llreach record";
222 }
223 }
224 }
225
226 if (nd6_debug && lr != NULL && why != NULL) {
227 char tmp[MAX_IPv6_STR_LEN];
228
229 nd6log(debug, "%s: %s%s for %s\n", if_name(ifp),
230 type, why, inet_ntop(AF_INET6,
231 &SIN6(rt_key(rt))->sin6_addr, tmp, sizeof(tmp)));
232 }
233 }
234 }
235
236 void
nd6_llreach_use(struct llinfo_nd6 * ln)237 nd6_llreach_use(struct llinfo_nd6 *ln)
238 {
239 if (ln->ln_llreach != NULL) {
240 ln->ln_lastused = net_uptime();
241 }
242 }
243
244 /*
245 * Input a Neighbor Solicitation Message.
246 *
247 * Based on RFC 4861
248 * Based on RFC 4862 (duplicate address detection)
249 */
250 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)251 nd6_ns_input(
252 struct mbuf *m,
253 int off,
254 int icmp6len)
255 {
256 struct ifnet *ifp = m->m_pkthdr.rcvif;
257 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
258 struct nd_neighbor_solicit *nd_ns = NULL;
259 struct in6_addr saddr6 = ip6->ip6_src;
260 struct in6_addr daddr6 = ip6->ip6_dst;
261 uint32_t saddr_ifscope = IN6_IS_SCOPE_EMBED(&saddr6) ? ip6_input_getsrcifscope(m) : IFSCOPE_NONE;
262 struct in6_addr taddr6 = {};
263 struct in6_addr myaddr6 = {};
264 uint32_t myaddr_ifscope = IFSCOPE_NONE;
265 char *lladdr = NULL;
266 struct ifaddr *ifa = NULL;
267 int lladdrlen = 0;
268 int anycast = 0, proxy = 0, dadprogress = 0;
269 int tlladdr = 0;
270 union nd_opts ndopts = {};
271 struct sockaddr_dl proxydl = {};
272 boolean_t advrouter = FALSE;
273 boolean_t is_dad_probe = FALSE;
274 int oflgclr = 0;
275 uint32_t taddr_ifscope;
276
277 /* Expect 32-bit aligned data pointer on strict-align platforms */
278 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
279
280 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
281 ip6 = mtod(m, struct ip6_hdr *);
282 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
283 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
284
285 taddr6 = nd_ns->nd_ns_target;
286 if (in6_setscope(&taddr6, ifp, &taddr_ifscope) != 0) {
287 goto bad;
288 }
289
290 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
291 nd6log(error,
292 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
293 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
294 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
295 goto bad;
296 }
297
298 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(&saddr6);
299 if (is_dad_probe) {
300 /* dst has to be a solicited node multicast address. */
301 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
302 /* don't check ifindex portion */
303 daddr6.s6_addr32[1] == 0 &&
304 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
305 daddr6.s6_addr8[12] == 0xff) {
306 ; /* good */
307 } else {
308 nd6log(info, "nd6_ns_input: bad DAD packet "
309 "(wrong ip6 dst)\n");
310 goto bad;
311 }
312 } else if (!nd6_onlink_ns_rfc4861) {
313 struct sockaddr_in6 src_sa6;
314
315 /*
316 * According to recent IETF discussions, it is not a good idea
317 * to accept a NS from an address which would not be deemed
318 * to be a neighbor otherwise. This point is expected to be
319 * clarified in future revisions of the specification.
320 */
321 bzero(&src_sa6, sizeof(src_sa6));
322 src_sa6.sin6_family = AF_INET6;
323 src_sa6.sin6_len = sizeof(src_sa6);
324 src_sa6.sin6_addr = saddr6;
325 if (!in6_embedded_scope) {
326 src_sa6.sin6_scope_id = saddr_ifscope;
327 }
328 if (!nd6_is_addr_neighbor(&src_sa6, ifp, 0)) {
329 nd6log(info, "nd6_ns_input: "
330 "NS packet from non-neighbor\n");
331 goto bad;
332 }
333 }
334
335 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
336 nd6log(info, "nd6_ns_input: bad NS target (multicast)\n");
337 goto bad;
338 }
339
340 icmp6len -= sizeof(*nd_ns);
341 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
342 if (nd6_options(&ndopts) < 0) {
343 nd6log(info,
344 "nd6_ns_input: invalid ND option, ignored\n");
345 /* nd6_options have incremented stats */
346 goto freeit;
347 }
348
349 if (ndopts.nd_opts_src_lladdr) {
350 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
351 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
352 }
353
354 if (is_dad_probe && lladdr) {
355 nd6log(info, "nd6_ns_input: bad DAD packet "
356 "(link-layer address option)\n");
357 goto bad;
358 }
359
360 /*
361 * Attaching target link-layer address to the NA?
362 * (RFC 2461 7.2.4)
363 *
364 * NS IP dst is unicast/anycast MUST NOT add
365 * NS IP dst is solicited-node multicast MUST add
366 *
367 * In implementation, we add target link-layer address by default.
368 * We do not add one in MUST NOT cases.
369 */
370 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) {
371 tlladdr = 0;
372 } else {
373 tlladdr = 1;
374 }
375
376 /*
377 * Target address (taddr6) must be either:
378 * (1) Valid unicast/anycast address for my receiving interface,
379 * (2) Unicast address for which I'm offering proxy service, or
380 * (3) "tentative" or "optimistic" address [DAD is in progress].
381 */
382 /* (1) and (3) check. */
383 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
384
385 /* (2) check. */
386 if (ifa == NULL) {
387 struct rtentry *rt;
388 struct sockaddr_in6 tsin6;
389
390 bzero(&tsin6, sizeof tsin6);
391 tsin6.sin6_len = sizeof(struct sockaddr_in6);
392 tsin6.sin6_family = AF_INET6;
393 tsin6.sin6_addr = taddr6;
394
395 rt = rtalloc1_scoped((struct sockaddr *)&tsin6, 0, 0,
396 ifp->if_index);
397
398 if (rt != NULL) {
399 RT_LOCK(rt);
400 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
401 rt->rt_gateway->sa_family == AF_LINK) {
402 /*
403 * proxy NDP for single entry
404 */
405 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
406 ifp, IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
407 if (ifa) {
408 proxy = 1;
409 proxydl = *SDL(rt->rt_gateway);
410 }
411 }
412 RT_UNLOCK(rt);
413 rtfree(rt);
414 }
415 }
416 if (ifa == NULL && ip6_forwarding && nd6_prproxy) {
417 /*
418 * Is the target address part of the prefix that is being
419 * proxied and installed on another interface?
420 */
421 ifa = (struct ifaddr *)in6ifa_prproxyaddr(&taddr6, taddr_ifscope);
422 }
423 if (ifa == NULL) {
424 /*
425 * We've got an NS packet, and we don't have that address
426 * assigned for us. We MUST silently ignore it on this
427 * interface, c.f. RFC 4861 7.2.3.
428 *
429 * Forwarding associated with NDPRF_PRPROXY may apply.
430 */
431 if (ip6_forwarding && nd6_prproxy) {
432 nd6_prproxy_ns_input(ifp, &saddr6, lladdr,
433 lladdrlen, &daddr6, &taddr6,
434 (ndopts.nd_opts_nonce == NULL) ? NULL :
435 ndopts.nd_opts_nonce->nd_opt_nonce);
436 }
437 goto freeit;
438 }
439 IFA_LOCK(ifa);
440 myaddr6 = *IFA_IN6(ifa);
441 myaddr_ifscope = IFA_SIN6_SCOPE(ifa);
442 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
443 dadprogress =
444 ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DADPROGRESS;
445 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) {
446 IFA_UNLOCK(ifa);
447 goto freeit;
448 }
449 IFA_UNLOCK(ifa);
450
451 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
452 nd6log(info,
453 "nd6_ns_input: lladdrlen mismatch for %s "
454 "(if %d, NS packet %d)\n",
455 ip6_sprintf(&taddr6), ifp->if_addrlen, lladdrlen - 2);
456 goto bad;
457 }
458
459 if (in6_are_addr_equal_scoped(&myaddr6, &saddr6, myaddr_ifscope, saddr_ifscope)) {
460 nd6log(info,
461 "nd6_ns_input: duplicate IP6 address %s\n",
462 ip6_sprintf(&saddr6));
463 goto freeit;
464 }
465
466 /*
467 * We have neighbor solicitation packet, with target address equals to
468 * one of my DAD in-progress addresses.
469 *
470 * src addr how to process?
471 * --- ---
472 * multicast of course, invalid (rejected in ip6_input)
473 * unicast somebody is doing address resolution
474 * unspec dup address detection
475 *
476 * The processing is defined in the "draft standard" RFC 4862 (and by
477 * RFC 4429, which is a "proposed standard" update to its obsolete
478 * predecessor, RFC 2462) The reason optimistic DAD is not included
479 * in RFC 4862 is entirely due to IETF procedural considerations.
480 */
481 if (dadprogress) {
482 /*
483 * If source address is unspecified address, it is for
484 * duplicate address detection.
485 *
486 * If not, the packet is for addess resolution;
487 * silently ignore it when not optimistic
488 *
489 * Per RFC 4429 the reply for an optimistic address must
490 * have the Override flag cleared
491 */
492 if (!is_dad_probe && (dadprogress & IN6_IFF_OPTIMISTIC) != 0) {
493 oflgclr = 1;
494 } else {
495 if (is_dad_probe) {
496 nd6_dad_ns_input(ifa, lladdr, lladdrlen, ndopts.nd_opts_nonce);
497 }
498
499 goto freeit;
500 }
501 }
502
503 /* Are we an advertising router on this interface? */
504 advrouter = (ifp->if_ipv6_router_mode != IPV6_ROUTER_MODE_DISABLED);
505
506 /*
507 * If the source address is unspecified address, entries must not
508 * be created or updated.
509 * It looks that sender is performing DAD. If I'm using the address,
510 * and it's a "preferred" address, i.e. not optimistic, then output NA
511 * toward all-node multicast address, to tell the sender that I'm using
512 * the address.
513 * S bit ("solicited") must be zero.
514 */
515 if (is_dad_probe) {
516 saddr6 = in6addr_linklocal_allnodes;
517 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
518 goto bad;
519 }
520 if ((dadprogress & IN6_IFF_OPTIMISTIC) == 0) {
521 nd6_na_output(ifp, &saddr6, &taddr6,
522 ((anycast || proxy || !tlladdr) ? 0 :
523 ND_NA_FLAG_OVERRIDE) | (advrouter ?
524 ND_NA_FLAG_ROUTER : 0), tlladdr, proxy ?
525 (struct sockaddr *)&proxydl : NULL);
526 }
527 goto freeit;
528 }
529
530 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
531 ND_NEIGHBOR_SOLICIT, 0);
532
533 nd6_na_output(ifp, &saddr6, &taddr6,
534 ((anycast || proxy || !tlladdr || oflgclr) ? 0 : ND_NA_FLAG_OVERRIDE) |
535 (advrouter ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
536 tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL);
537 freeit:
538 m_freem(m);
539 if (ifa != NULL) {
540 IFA_REMREF(ifa);
541 }
542 return;
543
544 bad:
545 nd6log(error, "nd6_ns_input: src=%s\n", ip6_sprintf(&saddr6));
546 nd6log(error, "nd6_ns_input: dst=%s\n", ip6_sprintf(&daddr6));
547 nd6log(error, "nd6_ns_input: tgt=%s\n", ip6_sprintf(&taddr6));
548 icmp6stat.icp6s_badns++;
549 m_freem(m);
550 if (ifa != NULL) {
551 IFA_REMREF(ifa);
552 }
553 }
554
555 /*
556 * Output a Neighbor Solicitation Message. Caller specifies:
557 * - ICMP6 header source IP6 address
558 * - ND6 header target IP6 address
559 * - ND6 header source datalink address
560 *
561 * Based on RFC 4861
562 * Based on RFC 4862 (duplicate address detection)
563 * Based on RFC 4429 (optimistic duplicate address detection)
564 *
565 * Caller must bump up ln->ln_rt refcnt to make sure 'ln' doesn't go
566 * away if there is a llinfo_nd6 passed in.
567 */
568 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * daddr6,const struct in6_addr * taddr6,struct llinfo_nd6 * ln,uint8_t * nonce)569 nd6_ns_output(
570 struct ifnet *ifp,
571 const struct in6_addr *daddr6,
572 const struct in6_addr *taddr6,
573 struct llinfo_nd6 *ln, /* for source address determination */
574 uint8_t *nonce) /* duplicated address detection */
575 {
576 struct mbuf *m;
577 struct ip6_hdr *ip6;
578 struct nd_neighbor_solicit *nd_ns;
579 struct in6_ifaddr *ia = NULL;
580 struct in6_addr *src, src_in, src_storage;
581 struct ip6_moptions *im6o = NULL;
582 struct ifnet *outif = NULL;
583 int icmp6len;
584 int maxlen;
585 int flags;
586 caddr_t mac;
587 struct route_in6 ro;
588 struct ip6_out_args ip6oa;
589 u_int32_t rtflags = 0;
590 boolean_t is_optimistic = FALSE;
591
592 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) || IN6_IS_ADDR_MULTICAST(taddr6)) {
593 return;
594 }
595
596 bzero(&ro, sizeof(ro));
597 bzero(&ip6oa, sizeof(ip6oa));
598 ip6oa.ip6oa_boundif = ifp->if_index;
599 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
600 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED;
601 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
602 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
603
604 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
605
606 /* estimate the size of message */
607 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
608 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
609 if (max_linkhdr + maxlen >= MCLBYTES) {
610 #if DIAGNOSTIC
611 printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
612 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
613 #endif
614 return;
615 }
616
617 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
618 if (m && max_linkhdr + maxlen >= MHLEN) {
619 MCLGET(m, M_DONTWAIT);
620 if ((m->m_flags & M_EXT) == 0) {
621 m_free(m);
622 m = NULL;
623 }
624 }
625 if (m == NULL) {
626 return;
627 }
628 m->m_pkthdr.rcvif = NULL;
629
630 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
631 m->m_flags |= M_MCAST;
632
633 im6o = ip6_allocmoptions(Z_NOWAIT);
634 if (im6o == NULL) {
635 m_freem(m);
636 return;
637 }
638
639 im6o->im6o_multicast_ifp = ifp;
640 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
641 im6o->im6o_multicast_loop = 0;
642 }
643
644 icmp6len = sizeof(*nd_ns);
645 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
646 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
647
648 /* fill neighbor solicitation packet */
649 ip6 = mtod(m, struct ip6_hdr *);
650 ip6->ip6_flow = 0;
651 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
652 ip6->ip6_vfc |= IPV6_VERSION;
653 /* ip6->ip6_plen will be set later */
654 ip6->ip6_nxt = IPPROTO_ICMPV6;
655 ip6->ip6_hlim = IPV6_MAXHLIM;
656 if (daddr6) {
657 ip6->ip6_dst = *daddr6;
658 ip6_output_setdstifscope(m, ifp->if_index, NULL);
659 } else {
660 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
661 ip6->ip6_dst.s6_addr16[1] = 0;
662 ip6->ip6_dst.s6_addr32[1] = 0;
663 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
664 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
665 ip6->ip6_dst.s6_addr8[12] = 0xff;
666 ip6_output_setdstifscope(m, ifp->if_index, NULL);
667 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) {
668 goto bad;
669 }
670 }
671 if (nonce == NULL) {
672 /*
673 * RFC2461 7.2.2:
674 * "If the source address of the packet prompting the
675 * solicitation is the same as one of the addresses assigned
676 * to the outgoing interface, that address SHOULD be placed
677 * in the IP Source Address of the outgoing solicitation.
678 * Otherwise, any one of the addresses assigned to the
679 * interface should be used."
680 *
681 * We use the source address for the prompting packet
682 * (saddr6), if:
683 * - saddr6 is given from the caller (by giving "ln"), and
684 * - saddr6 belongs to the outgoing interface.
685 * Otherwise, we perform the source address selection as usual.
686 */
687 struct ip6_hdr *hip6; /* hold ip6 */
688 struct in6_addr *hsrc = NULL;
689
690 /* Caller holds ref on this route */
691 if (ln != NULL) {
692 RT_LOCK(ln->ln_rt);
693 /*
694 * assuming every packet in ln_hold has the same IP
695 * header
696 */
697 if (ln->ln_hold != NULL) {
698 hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
699 /* XXX pullup? */
700 if (sizeof(*hip6) < ln->ln_hold->m_len) {
701 hsrc = &hip6->ip6_src;
702 } else {
703 hsrc = NULL;
704 }
705 }
706 /* Update probe count, if applicable */
707 if (ln->ln_llreach != NULL) {
708 IFLR_LOCK_SPIN(ln->ln_llreach);
709 ln->ln_llreach->lr_probes++;
710 IFLR_UNLOCK(ln->ln_llreach);
711 }
712 rtflags = ln->ln_rt->rt_flags;
713 RT_UNLOCK(ln->ln_rt);
714 }
715 if (hsrc != NULL && (ia = in6ifa_ifpwithaddr(ifp, hsrc)) &&
716 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0) {
717 src = hsrc;
718 } else {
719 int error;
720 struct sockaddr_in6 dst_sa;
721
722 bzero(&dst_sa, sizeof(dst_sa));
723 dst_sa.sin6_family = AF_INET6;
724 dst_sa.sin6_len = sizeof(dst_sa);
725 dst_sa.sin6_addr = ip6->ip6_dst;
726
727 src = in6_selectsrc(&dst_sa, NULL,
728 NULL, &ro, NULL, &src_storage, ip6oa.ip6oa_boundif,
729 &error);
730 if (src == NULL) {
731 nd6log(debug,
732 "nd6_ns_output: source can't be "
733 "determined: dst=%s, error=%d\n",
734 ip6_sprintf(&dst_sa.sin6_addr),
735 error);
736 goto bad;
737 }
738
739 if (ia != NULL) {
740 IFA_REMREF(&ia->ia_ifa);
741 ia = NULL;
742 }
743 /*
744 * RFC 4429 section 3.2:
745 * When a node has a unicast packet to send
746 * from an Optimistic Address to a neighbor,
747 * but does not know the neighbor's link-layer
748 * address, it MUST NOT perform Address
749 * Resolution.
750 */
751 ia = in6ifa_ifpwithaddr(ifp, src);
752 if (ia == NULL) {
753 nd6log(debug,
754 "nd6_ns_output: no preferred source "
755 "available: dst=%s\n",
756 ip6_sprintf(&dst_sa.sin6_addr));
757 goto bad;
758 }
759 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
760 is_optimistic = TRUE;
761 nd6log(debug,
762 "nd6_ns_output: preferred source "
763 "available is optimistic: dst=%s\n",
764 ip6_sprintf(&dst_sa.sin6_addr));
765 }
766 }
767 } else {
768 /*
769 * Source address for DAD packet must always be IPv6
770 * unspecified address. (0::0)
771 * We actually don't have to 0-clear the address (we did it
772 * above), but we do so here explicitly to make the intention
773 * clearer.
774 */
775 bzero(&src_in, sizeof(src_in));
776 src = &src_in;
777 ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_SRCADDR;
778 }
779 ip6->ip6_src = *src;
780 ip6_output_setsrcifscope(m, ifp->if_index, ia);
781 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
782 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
783 nd_ns->nd_ns_code = 0;
784 nd_ns->nd_ns_reserved = 0;
785 nd_ns->nd_ns_target = *taddr6;
786 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
787
788 /*
789 * Add source link-layer address option.
790 *
791 * spec implementation
792 * --- ---
793 * DAD packet MUST NOT do not add the option
794 * Source is optimistic MUST NOT do not add the option
795 * there's no link layer address:
796 * impossible do not add the option
797 * there's link layer address:
798 * Multicast NS MUST add one add the option
799 * Unicast NS SHOULD add one add the option
800 *
801 * XXX We deviate from RFC 4429 and still use optimistic DAD as source
802 * for address resolution. However to ensure that we do not interfere
803 * with neighbor cache entries of other neighbors, we MUST ensure
804 * that SLLAO is not sent. Also note, sending multicast NS without SLLAO
805 * is also a deviation from RFC 4861.
806 */
807 if (nonce == NULL && (mac = nd6_ifptomac(ifp)) && !is_optimistic) {
808 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
809 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
810 /* 8 byte alignments... */
811 optlen = (optlen + 7) & ~7;
812
813 m->m_pkthdr.len += optlen;
814 m->m_len += optlen;
815 icmp6len += optlen;
816 bzero((caddr_t)nd_opt, optlen);
817 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
818 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
819 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
820 }
821 /*
822 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
823 * This behavior is documented as Enhanced Duplicate Address
824 * Detection in draft-ietf-6man-enhanced-dad-13.
825 * net.inet6.ip6.dad_enhanced=0 disables this.
826 */
827 if (dad_enhanced != 0 && nonce != NULL && !(ifp->if_flags & IFF_POINTOPOINT)) {
828 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
829 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
830 /* 8-byte alignment is required. */
831 optlen = (optlen + 7) & ~7;
832
833 m->m_pkthdr.len += optlen;
834 m->m_len += optlen;
835 icmp6len += optlen;
836 bzero((caddr_t)nd_opt, optlen);
837 nd_opt->nd_opt_type = ND_OPT_NONCE;
838 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
839 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
840 }
841 ip6->ip6_plen = htons((u_short)icmp6len);
842 nd_ns->nd_ns_cksum = 0;
843 nd_ns->nd_ns_cksum
844 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
845
846 flags = nonce ? IPV6_UNSPECSRC : 0;
847 flags |= IPV6_OUTARGS;
848
849 /*
850 * PKTF_{INET,INET6}_RESOLVE_RTR are mutually exclusive, so make
851 * sure only one of them is set (just in case.)
852 */
853 m->m_pkthdr.pkt_flags &= ~(PKTF_INET_RESOLVE | PKTF_RESOLVE_RTR);
854 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
855 /*
856 * If this is a NS for resolving the (default) router, mark
857 * the packet accordingly so that the driver can find out,
858 * in case it needs to perform driver-specific action(s).
859 */
860 if (rtflags & RTF_ROUTER) {
861 m->m_pkthdr.pkt_flags |= PKTF_RESOLVE_RTR;
862 }
863
864 if (ifp->if_eflags & IFEF_TXSTART) {
865 /*
866 * Use control service class if the interface
867 * supports transmit-start model
868 */
869 (void) m_set_service_class(m, MBUF_SC_CTL);
870 }
871
872 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
873 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
874 ip6_output(m, NULL, NULL, flags, im6o, &outif, &ip6oa);
875 if (outif) {
876 icmp6_ifstat_inc(outif, ifs6_out_msg);
877 icmp6_ifstat_inc(outif, ifs6_out_neighborsolicit);
878 ifnet_release(outif);
879 }
880 icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
881
882 exit:
883 if (im6o != NULL) {
884 IM6O_REMREF(im6o);
885 }
886
887 ROUTE_RELEASE(&ro); /* we don't cache this route. */
888
889 if (ia != NULL) {
890 IFA_REMREF(&ia->ia_ifa);
891 }
892 return;
893
894 bad:
895 m_freem(m);
896 goto exit;
897 }
898
899 /*
900 * Neighbor advertisement input handling.
901 *
902 * Based on RFC 4861
903 * Based on RFC 4862 (duplicate address detection)
904 *
905 * the following items are not implemented yet:
906 * - anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD)
907 * - proxy advertisement delay rule (RFC 4861 7.2.8, last paragraph, "should")
908 */
909 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)910 nd6_na_input(struct mbuf *m, int off, int icmp6len)
911 {
912 struct ifnet *ifp = m->m_pkthdr.rcvif;
913 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
914 struct nd_neighbor_advert *nd_na;
915 struct in6_addr saddr6 = ip6->ip6_src;
916 struct in6_addr daddr6 = ip6->ip6_dst;
917 struct in6_addr taddr6;
918 int flags;
919 int is_router;
920 int is_solicited;
921 int is_override;
922 char *lladdr = NULL;
923 int lladdrlen = 0;
924 struct llinfo_nd6 *ln;
925 struct rtentry *rt;
926 struct sockaddr_dl *sdl;
927 union nd_opts ndopts;
928 uint64_t timenow;
929 bool send_nc_alive_kev = false;
930
931 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
932 nd6log(info, "nd6_na_input: on ND6ALT interface!\n");
933 goto freeit;
934 }
935
936 /* Expect 32-bit aligned data pointer on strict-align platforms */
937 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
938
939 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
940 nd6log(error,
941 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
942 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
943 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
944 goto bad;
945 }
946
947 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
948 ip6 = mtod(m, struct ip6_hdr *);
949 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
950 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
951
952 flags = nd_na->nd_na_flags_reserved;
953 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
954 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
955 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
956
957 taddr6 = nd_na->nd_na_target;
958 if (in6_setscope(&taddr6, ifp, NULL)) {
959 goto bad; /* XXX: impossible */
960 }
961 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
962 nd6log(error,
963 "nd6_na_input: invalid target address %s\n",
964 ip6_sprintf(&taddr6));
965 goto bad;
966 }
967 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
968 if (is_solicited) {
969 nd6log(error,
970 "nd6_na_input: a solicited adv is multicasted\n");
971 goto bad;
972 }
973 }
974
975 icmp6len -= sizeof(*nd_na);
976 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
977 if (nd6_options(&ndopts) < 0) {
978 nd6log(info,
979 "nd6_na_input: invalid ND option, ignored\n");
980 /* nd6_options have incremented stats */
981 goto freeit;
982 }
983
984 if (ndopts.nd_opts_tgt_lladdr) {
985 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
986 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
987
988 if (((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
989 nd6log(info,
990 "nd6_na_input: lladdrlen mismatch for %s "
991 "(if %d, NA packet %d)\n",
992 ip6_sprintf(&taddr6), ifp->if_addrlen,
993 lladdrlen - 2);
994 goto bad;
995 }
996 }
997
998 m = nd6_dad_na_input(m, ifp, &taddr6, lladdr, lladdrlen);
999 if (m == NULL) {
1000 return;
1001 }
1002
1003 /* Forwarding associated with NDPRF_PRPROXY may apply. */
1004 if (ip6_forwarding && nd6_prproxy) {
1005 nd6_prproxy_na_input(ifp, &saddr6, &daddr6, &taddr6, flags);
1006 }
1007
1008 /*
1009 * If no neighbor cache entry is found, NA SHOULD silently be
1010 * discarded. If we are forwarding (and Scoped Routing is in
1011 * effect), try to see if there is a neighbor cache entry on
1012 * another interface (in case we are doing prefix proxying.)
1013 */
1014 if ((rt = nd6_lookup(&taddr6, 0, ifp, 0)) == NULL) {
1015 if (!ip6_forwarding || !nd6_prproxy) {
1016 goto freeit;
1017 }
1018
1019 if ((rt = nd6_lookup(&taddr6, 0, NULL, 0)) == NULL) {
1020 goto freeit;
1021 }
1022
1023 RT_LOCK_ASSERT_HELD(rt);
1024 if (rt->rt_ifp != ifp) {
1025 /*
1026 * Purge any link-layer info caching.
1027 */
1028 if (rt->rt_llinfo_purge != NULL) {
1029 rt->rt_llinfo_purge(rt);
1030 }
1031
1032 /* Adjust route ref count for the interfaces */
1033 if (rt->rt_if_ref_fn != NULL) {
1034 rt->rt_if_ref_fn(ifp, 1);
1035 rt->rt_if_ref_fn(rt->rt_ifp, -1);
1036 }
1037
1038 /* Change the interface when the existing route is on */
1039 rt->rt_ifp = ifp;
1040
1041 /*
1042 * If rmx_mtu is not locked, update it
1043 * to the MTU used by the new interface.
1044 */
1045 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1046 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1047 }
1048 }
1049 }
1050
1051 RT_LOCK_ASSERT_HELD(rt);
1052 if ((ln = rt->rt_llinfo) == NULL ||
1053 (sdl = SDL(rt->rt_gateway)) == NULL) {
1054 RT_REMREF_LOCKED(rt);
1055 RT_UNLOCK(rt);
1056 goto freeit;
1057 }
1058
1059 timenow = net_uptime();
1060
1061 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1062 /*
1063 * If the link-layer has address, and no lladdr option came,
1064 * discard the packet.
1065 */
1066 if (ifp->if_addrlen && !lladdr) {
1067 RT_REMREF_LOCKED(rt);
1068 RT_UNLOCK(rt);
1069 goto freeit;
1070 }
1071
1072 /*
1073 * Record link-layer address, and update the state.
1074 */
1075 sdl->sdl_alen = ifp->if_addrlen;
1076 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1077 if (is_solicited) {
1078 send_nc_alive_kev = (rt->rt_flags & RTF_ROUTER) ? true : false;
1079 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1080 if (ln->ln_expire != 0) {
1081 struct nd_ifinfo *ndi = NULL;
1082
1083 ndi = ND_IFINFO(rt->rt_ifp);
1084 VERIFY(ndi != NULL && ndi->initialized);
1085 lck_mtx_lock(&ndi->lock);
1086 ln_setexpire(ln, timenow + ndi->reachable);
1087 lck_mtx_unlock(&ndi->lock);
1088 RT_UNLOCK(rt);
1089 lck_mtx_lock(rnh_lock);
1090 nd6_sched_timeout(NULL, NULL);
1091 lck_mtx_unlock(rnh_lock);
1092 RT_LOCK(rt);
1093 }
1094 } else {
1095 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1096 ln_setexpire(ln, timenow + nd6_gctimer);
1097 }
1098
1099
1100 /*
1101 * Enqueue work item to invoke callback for this
1102 * route entry
1103 */
1104 route_event_enqueue_nwk_wq_entry(rt, NULL,
1105 ROUTE_LLENTRY_RESOLVED, NULL, TRUE);
1106
1107 if ((ln->ln_router = (short)is_router) != 0) {
1108 struct radix_node_head *rnh = NULL;
1109 struct route_event rt_ev;
1110 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_RESOLVED);
1111 /*
1112 * This means a router's state has changed from
1113 * non-reachable to probably reachable, and might
1114 * affect the status of associated prefixes..
1115 * We already have a reference on rt. Don't need to
1116 * take one for the unlock/lock.
1117 */
1118 RT_UNLOCK(rt);
1119 lck_mtx_lock(rnh_lock);
1120 rnh = rt_tables[AF_INET6];
1121
1122 if (rnh != NULL) {
1123 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1124 (void *)&rt_ev);
1125 }
1126 lck_mtx_unlock(rnh_lock);
1127 lck_mtx_lock(nd6_mutex);
1128 pfxlist_onlink_check();
1129 lck_mtx_unlock(nd6_mutex);
1130 RT_LOCK(rt);
1131 }
1132 } else {
1133 int llchange = 0;
1134
1135 /*
1136 * Check if the link-layer address has changed or not.
1137 */
1138 if (lladdr == NULL) {
1139 llchange = 0;
1140 } else {
1141 if (sdl->sdl_alen) {
1142 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) {
1143 llchange = 1;
1144 } else {
1145 llchange = 0;
1146 }
1147 } else {
1148 llchange = 1;
1149 }
1150 }
1151
1152 /*
1153 * This is VERY complex. Look at it with care.
1154 *
1155 * override solicit lladdr llchange action
1156 * (L: record lladdr)
1157 *
1158 * 0 0 n -- (2c)
1159 * 0 0 y n (2b) L
1160 * 0 0 y y (1) REACHABLE->STALE
1161 * 0 1 n -- (2c) *->REACHABLE
1162 * 0 1 y n (2b) L *->REACHABLE
1163 * 0 1 y y (1) REACHABLE->STALE
1164 * 1 0 n -- (2a)
1165 * 1 0 y n (2a) L
1166 * 1 0 y y (2a) L *->STALE
1167 * 1 1 n -- (2a) *->REACHABLE
1168 * 1 1 y n (2a) L *->REACHABLE
1169 * 1 1 y y (2a) L *->REACHABLE
1170 */
1171 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
1172 /*
1173 * If state is REACHABLE, make it STALE.
1174 * no other updates should be done.
1175 */
1176 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
1177 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1178 ln_setexpire(ln, timenow + nd6_gctimer);
1179 }
1180 RT_REMREF_LOCKED(rt);
1181 RT_UNLOCK(rt);
1182 goto freeit;
1183 } else if (is_override /* (2a) */
1184 || (!is_override && (lladdr && !llchange)) /* (2b) */
1185 || !lladdr) { /* (2c) */
1186 /*
1187 * Update link-local address, if any.
1188 */
1189 if (lladdr) {
1190 sdl->sdl_alen = ifp->if_addrlen;
1191 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1192 }
1193
1194 /*
1195 * If solicited, make the state REACHABLE.
1196 * If not solicited and the link-layer address was
1197 * changed, make it STALE.
1198 */
1199 if (is_solicited) {
1200 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1201 if (ln->ln_expire != 0) {
1202 struct nd_ifinfo *ndi = NULL;
1203
1204 ndi = ND_IFINFO(ifp);
1205 VERIFY(ndi != NULL && ndi->initialized);
1206 lck_mtx_lock(&ndi->lock);
1207 ln_setexpire(ln,
1208 timenow + ndi->reachable);
1209 lck_mtx_unlock(&ndi->lock);
1210 RT_UNLOCK(rt);
1211 lck_mtx_lock(rnh_lock);
1212 nd6_sched_timeout(NULL, NULL);
1213 lck_mtx_unlock(rnh_lock);
1214 RT_LOCK(rt);
1215 }
1216 } else {
1217 if (lladdr && llchange) {
1218 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1219 ln_setexpire(ln, timenow + nd6_gctimer);
1220 }
1221 }
1222
1223 /*
1224 * XXX
1225 * The above is somewhat convoluted, for now just
1226 * issue a callback for LLENTRY changed.
1227 */
1228 /* Enqueue work item to invoke callback for this route entry */
1229 if (llchange) {
1230 route_event_enqueue_nwk_wq_entry(rt, NULL,
1231 ROUTE_LLENTRY_CHANGED, NULL, TRUE);
1232 }
1233
1234 /*
1235 * If the router's link-layer address has changed,
1236 * notify routes using this as gateway so they can
1237 * update any cached information.
1238 */
1239 if (ln->ln_router && is_router && llchange) {
1240 struct radix_node_head *rnh = NULL;
1241 struct route_event rt_ev;
1242 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_CHANGED);
1243 /*
1244 * This means a router's state has changed from
1245 * non-reachable to probably reachable, and might
1246 * affect the status of associated prefixes..
1247 *
1248 * We already have a valid rt reference here.
1249 * We don't need to take another one for unlock/lock.
1250 */
1251 RT_UNLOCK(rt);
1252 lck_mtx_lock(rnh_lock);
1253 rnh = rt_tables[AF_INET6];
1254
1255 if (rnh != NULL) {
1256 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1257 (void *)&rt_ev);
1258 }
1259 lck_mtx_unlock(rnh_lock);
1260 RT_LOCK(rt);
1261 }
1262 }
1263
1264 if (ln->ln_router && !is_router) {
1265 /*
1266 * The peer dropped the router flag.
1267 * Remove the sender from the Default Router List and
1268 * update the Destination Cache entries.
1269 */
1270 struct nd_defrouter *dr;
1271 struct in6_addr *in6;
1272 struct ifnet *rt_ifp = rt->rt_ifp;
1273
1274 in6 = &((struct sockaddr_in6 *)
1275 (void *)rt_key(rt))->sin6_addr;
1276
1277 RT_UNLOCK(rt);
1278 lck_mtx_lock(nd6_mutex);
1279 /*
1280 * XXX Handle router lists for route information option
1281 * as well.
1282 */
1283 dr = defrouter_lookup(NULL, in6, rt_ifp);
1284 if (dr) {
1285 TAILQ_REMOVE(&nd_defrouter_list, dr, dr_entry);
1286 defrtrlist_del(dr, NULL);
1287 NDDR_REMREF(dr); /* remove list reference */
1288 NDDR_REMREF(dr);
1289 lck_mtx_unlock(nd6_mutex);
1290 } else {
1291 lck_mtx_unlock(nd6_mutex);
1292 /*
1293 * Even if the neighbor is not in the
1294 * default router list, the neighbor
1295 * may be used as a next hop for some
1296 * destinations (e.g. redirect case).
1297 * So we must call rt6_flush explicitly.
1298 */
1299 rt6_flush(&ip6->ip6_src, rt_ifp);
1300 }
1301 RT_LOCK(rt);
1302 }
1303 ln->ln_router = (short)is_router;
1304 }
1305
1306 if (send_nc_alive_kev && (ifp->if_addrlen == IF_LLREACH_MAXLEN)) {
1307 struct kev_msg ev_msg;
1308 struct kev_nd6_ndalive nd6_ndalive;
1309 bzero(&ev_msg, sizeof(ev_msg));
1310 bzero(&nd6_ndalive, sizeof(nd6_ndalive));
1311 ev_msg.vendor_code = KEV_VENDOR_APPLE;
1312 ev_msg.kev_class = KEV_NETWORK_CLASS;
1313 ev_msg.kev_subclass = KEV_ND6_SUBCLASS;
1314 ev_msg.event_code = KEV_ND6_NDALIVE;
1315
1316 nd6_ndalive.link_data.if_family = ifp->if_family;
1317 nd6_ndalive.link_data.if_unit = ifp->if_unit;
1318 strlcpy(nd6_ndalive.link_data.if_name,
1319 ifp->if_name,
1320 sizeof(nd6_ndalive.link_data.if_name));
1321 ev_msg.dv[0].data_ptr = &nd6_ndalive;
1322 ev_msg.dv[0].data_length =
1323 sizeof(nd6_ndalive);
1324 dlil_post_complete_msg(NULL, &ev_msg);
1325 }
1326
1327 RT_LOCK_ASSERT_HELD(rt);
1328 rt->rt_flags &= ~RTF_REJECT;
1329
1330 /* cache the gateway (sender HW) address */
1331 nd6_llreach_alloc(rt, ifp, LLADDR(sdl), sdl->sdl_alen, TRUE);
1332
1333 /* update the llinfo, send a queued packet if there is one */
1334 ln->ln_asked = 0;
1335 if (ln->ln_hold != NULL) {
1336 struct mbuf *m_hold, *m_hold_next;
1337 struct sockaddr_in6 sin6;
1338
1339 rtkey_to_sa6(rt, &sin6);
1340 /*
1341 * reset the ln_hold in advance, to explicitly
1342 * prevent a ln_hold lookup in nd6_output()
1343 * (wouldn't happen, though...)
1344 */
1345 m_hold = ln->ln_hold;
1346 ln->ln_hold = NULL;
1347 for (; m_hold; m_hold = m_hold_next) {
1348 m_hold_next = m_hold->m_nextpkt;
1349 m_hold->m_nextpkt = NULL;
1350 /*
1351 * we assume ifp is not a loopback here, so just set
1352 * the 2nd argument as the 1st one.
1353 */
1354 RT_UNLOCK(rt);
1355 nd6_output(ifp, ifp, m_hold, &sin6, rt, NULL);
1356 RT_LOCK_SPIN(rt);
1357 }
1358 }
1359 RT_REMREF_LOCKED(rt);
1360 RT_UNLOCK(rt);
1361 m_freem(m);
1362 return;
1363
1364 bad:
1365 icmp6stat.icp6s_badna++;
1366 /* fall through */
1367 freeit:
1368 m_freem(m);
1369 return;
1370 }
1371
1372 /*
1373 * Neighbor advertisement output handling.
1374 *
1375 * Based on RFC 2461
1376 *
1377 * the following items are not implemented yet:
1378 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
1379 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
1380 *
1381 * tlladdr - 1 if include target link-layer address
1382 * sdl0 - sockaddr_dl (= proxy NA) or NULL
1383 */
1384 void
nd6_na_output(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,uint32_t flags,int tlladdr,struct sockaddr * sdl0)1385 nd6_na_output(
1386 struct ifnet *ifp,
1387 const struct in6_addr *daddr6_0,
1388 const struct in6_addr *taddr6,
1389 uint32_t flags,
1390 int tlladdr, /* 1 if include target link-layer address */
1391 struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
1392 {
1393 struct mbuf *m;
1394 struct ip6_hdr *ip6;
1395 struct nd_neighbor_advert *nd_na;
1396 struct ip6_moptions *im6o = NULL;
1397 caddr_t mac = NULL;
1398 struct route_in6 ro;
1399 struct in6_addr *src, src_storage, daddr6;
1400 struct in6_ifaddr *ia;
1401 struct sockaddr_in6 dst_sa;
1402 int icmp6len, maxlen, error;
1403 struct ifnet *outif = NULL;
1404
1405 struct ip6_out_args ip6oa;
1406 bzero(&ro, sizeof(ro));
1407
1408 daddr6 = *daddr6_0; /* make a local copy for modification */
1409
1410 bzero(&ip6oa, sizeof(ip6oa));
1411 ip6oa.ip6oa_boundif = ifp->if_index;
1412 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
1413 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED;
1414 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
1415 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1416
1417 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
1418
1419 /* estimate the size of message */
1420 maxlen = sizeof(*ip6) + sizeof(*nd_na);
1421 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
1422 if (max_linkhdr + maxlen >= MCLBYTES) {
1423 #if DIAGNOSTIC
1424 printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
1425 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
1426 #endif
1427 return;
1428 }
1429
1430 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
1431 if (m && max_linkhdr + maxlen >= MHLEN) {
1432 MCLGET(m, M_DONTWAIT);
1433 if ((m->m_flags & M_EXT) == 0) {
1434 m_free(m);
1435 m = NULL;
1436 }
1437 }
1438 if (m == NULL) {
1439 return;
1440 }
1441 m->m_pkthdr.rcvif = NULL;
1442
1443 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1444 m->m_flags |= M_MCAST;
1445
1446 im6o = ip6_allocmoptions(Z_NOWAIT);
1447 if (im6o == NULL) {
1448 m_freem(m);
1449 return;
1450 }
1451
1452 im6o->im6o_multicast_ifp = ifp;
1453 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
1454 im6o->im6o_multicast_loop = 0;
1455 }
1456
1457 icmp6len = sizeof(*nd_na);
1458 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1459 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
1460
1461 /* fill neighbor advertisement packet */
1462 ip6 = mtod(m, struct ip6_hdr *);
1463 ip6->ip6_flow = 0;
1464 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1465 ip6->ip6_vfc |= IPV6_VERSION;
1466 ip6->ip6_nxt = IPPROTO_ICMPV6;
1467 ip6->ip6_hlim = IPV6_MAXHLIM;
1468 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1469 /* reply to DAD */
1470 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1471 daddr6.s6_addr16[1] = 0;
1472 daddr6.s6_addr32[1] = 0;
1473 daddr6.s6_addr32[2] = 0;
1474 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1475 if (in6_setscope(&daddr6, ifp, NULL)) {
1476 goto bad;
1477 }
1478
1479 flags &= ~ND_NA_FLAG_SOLICITED;
1480 } else {
1481 ip6->ip6_dst = daddr6;
1482 ip6_output_setdstifscope(m, ifp->if_index, NULL);
1483 }
1484
1485 bzero(&dst_sa, sizeof(struct sockaddr_in6));
1486 dst_sa.sin6_family = AF_INET6;
1487 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1488 dst_sa.sin6_addr = daddr6;
1489
1490 /*
1491 * Select a source whose scope is the same as that of the dest.
1492 */
1493 bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1494 src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &src_storage,
1495 ip6oa.ip6oa_boundif, &error);
1496 if (src == NULL) {
1497 nd6log(debug, "nd6_na_output: source can't be "
1498 "determined: dst=%s, error=%d\n",
1499 ip6_sprintf(&dst_sa.sin6_addr), error);
1500 goto bad;
1501 }
1502 ip6->ip6_src = *src;
1503
1504 /*
1505 * RFC 4429 requires not setting "override" flag on NA packets sent
1506 * from optimistic addresses.
1507 */
1508 ia = in6ifa_ifpwithaddr(ifp, src);
1509 ip6_output_setsrcifscope(m, ifp->if_index, ia);
1510 if (ia != NULL) {
1511 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
1512 flags &= ~ND_NA_FLAG_OVERRIDE;
1513 }
1514 IFA_REMREF(&ia->ia_ifa);
1515 }
1516
1517 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1518 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1519 nd_na->nd_na_code = 0;
1520 nd_na->nd_na_target = *taddr6;
1521 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1522
1523 /*
1524 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1525 * see nd6_ns_input() for details.
1526 * Basically, if NS packet is sent to unicast/anycast addr,
1527 * target lladdr option SHOULD NOT be included.
1528 */
1529 if (tlladdr) {
1530 /*
1531 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1532 * lladdr in sdl0. If we are not proxying (sending NA for
1533 * my address) use lladdr configured for the interface.
1534 */
1535 if (sdl0 == NULL) {
1536 mac = nd6_ifptomac(ifp);
1537 } else if (sdl0->sa_family == AF_LINK) {
1538 struct sockaddr_dl *sdl;
1539 sdl = (struct sockaddr_dl *)(void *)sdl0;
1540 if (sdl->sdl_alen == ifp->if_addrlen) {
1541 mac = LLADDR(sdl);
1542 }
1543 }
1544 }
1545 if (tlladdr && mac) {
1546 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1547 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1548
1549 /* roundup to 8 bytes alignment! */
1550 optlen = (optlen + 7) & ~7;
1551
1552 m->m_pkthdr.len += optlen;
1553 m->m_len += optlen;
1554 icmp6len += optlen;
1555 bzero((caddr_t)nd_opt, optlen);
1556 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1557 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
1558 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1559 } else {
1560 flags &= ~ND_NA_FLAG_OVERRIDE;
1561 }
1562
1563 ip6->ip6_plen = htons((u_short)icmp6len);
1564 nd_na->nd_na_flags_reserved = flags;
1565 nd_na->nd_na_cksum = 0;
1566 nd_na->nd_na_cksum =
1567 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1568
1569 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
1570
1571 if (ifp->if_eflags & IFEF_TXSTART) {
1572 /* Use control service class if the interface supports
1573 * transmit-start model.
1574 */
1575 (void) m_set_service_class(m, MBUF_SC_CTL);
1576 }
1577
1578 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
1579 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
1580 ip6_output(m, NULL, NULL, IPV6_OUTARGS, im6o, &outif, &ip6oa);
1581 if (outif) {
1582 icmp6_ifstat_inc(outif, ifs6_out_msg);
1583 icmp6_ifstat_inc(outif, ifs6_out_neighboradvert);
1584 ifnet_release(outif);
1585 }
1586 icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1587
1588 exit:
1589 if (im6o != NULL) {
1590 IM6O_REMREF(im6o);
1591 }
1592
1593 ROUTE_RELEASE(&ro);
1594 return;
1595
1596 bad:
1597 m_freem(m);
1598 goto exit;
1599 }
1600
1601 caddr_t
nd6_ifptomac(struct ifnet * ifp)1602 nd6_ifptomac(
1603 struct ifnet *ifp)
1604 {
1605 switch (ifp->if_type) {
1606 case IFT_ARCNET:
1607 case IFT_ETHER:
1608 case IFT_IEEE8023ADLAG:
1609 case IFT_FDDI:
1610 case IFT_IEEE1394:
1611 #ifdef IFT_L2VLAN
1612 case IFT_L2VLAN:
1613 #endif
1614 #ifdef IFT_IEEE80211
1615 case IFT_IEEE80211:
1616 #endif
1617 #ifdef IFT_CARP
1618 case IFT_CARP:
1619 #endif
1620 case IFT_BRIDGE:
1621 case IFT_ISO88025:
1622 case IFT_6LOWPAN:
1623 return (caddr_t)IF_LLADDR(ifp);
1624 default:
1625 return NULL;
1626 }
1627 }
1628
1629 TAILQ_HEAD(dadq_head, dadq);
1630 struct dadq {
1631 decl_lck_mtx_data(, dad_lock);
1632 u_int32_t dad_refcount; /* reference count */
1633 int dad_attached;
1634 TAILQ_ENTRY(dadq) dad_list;
1635 struct ifaddr *dad_ifa;
1636 int dad_count; /* max NS to send */
1637 int dad_ns_tcount; /* # of trials to send NS */
1638 int dad_ns_ocount; /* NS sent so far */
1639 int dad_ns_icount;
1640 int dad_na_icount;
1641 int dad_ns_lcount; /* looped back NS */
1642 int dad_loopbackprobe; /* probing state for loopback detection */
1643 uint8_t dad_lladdr[ETHER_ADDR_LEN];
1644 uint8_t dad_lladdrlen;
1645 #define ND_OPT_NONCE_LEN32 \
1646 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1647 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1648 };
1649
1650 static ZONE_DECLARE(dad_zone, "nd6_dad", sizeof(struct dadq), ZC_ZFREE_CLEARMEM);
1651 static struct dadq_head dadq;
1652
1653 void
nd6_nbr_init(void)1654 nd6_nbr_init(void)
1655 {
1656 int i;
1657
1658 TAILQ_INIT(&dadq);
1659
1660 bzero(&hostrtmask, sizeof hostrtmask);
1661 hostrtmask.sin6_family = AF_INET6;
1662 hostrtmask.sin6_len = sizeof hostrtmask;
1663 for (i = 0; i < sizeof hostrtmask.sin6_addr; ++i) {
1664 hostrtmask.sin6_addr.s6_addr[i] = 0xff;
1665 }
1666 }
1667
1668 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * nonce)1669 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1670 {
1671 struct dadq *dp;
1672
1673 lck_mtx_lock(&dad6_mutex);
1674 for (dp = dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1675 DAD_LOCK_SPIN(dp);
1676 if (dp->dad_ifa != ifa) {
1677 DAD_UNLOCK(dp);
1678 continue;
1679 }
1680
1681 /*
1682 * Skip if the nonce matches the received one.
1683 * +2 in the length is required because of type and
1684 * length fields are included in a header.
1685 */
1686 if (nonce != NULL &&
1687 nonce->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1688 memcmp(&nonce->nd_opt_nonce[0], &dp->dad_nonce[0],
1689 ND_OPT_NONCE_LEN) == 0) {
1690 nd6log(error, "%s: a looped back NS message is "
1691 "detected during DAD for %s. Ignoring.\n",
1692 if_name(ifa->ifa_ifp),
1693 ip6_sprintf(IFA_IN6(ifa)));
1694 dp->dad_ns_lcount++;
1695 ++ip6stat.ip6s_dad_loopcount;
1696 DAD_UNLOCK(dp);
1697 continue;
1698 }
1699
1700 DAD_ADDREF_LOCKED(dp);
1701 DAD_UNLOCK(dp);
1702 break;
1703 }
1704 lck_mtx_unlock(&dad6_mutex);
1705 return dp;
1706 }
1707
1708 void
nd6_dad_stoptimer(struct ifaddr * ifa)1709 nd6_dad_stoptimer(
1710 struct ifaddr *ifa)
1711 {
1712 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1713 }
1714
1715 /*
1716 * Start Duplicate Address Detection (DAD) for specified interface address.
1717 */
1718 void
nd6_dad_start(struct ifaddr * ifa,int * tick_delay)1719 nd6_dad_start(
1720 struct ifaddr *ifa,
1721 int *tick_delay) /* minimum delay ticks for IFF_UP event */
1722 {
1723 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1724 struct dadq *dp;
1725
1726 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1727 __func__,
1728 ip6_sprintf(&ia->ia_addr.sin6_addr),
1729 if_name(ia->ia_ifp),
1730 ia->ia6_flags);
1731
1732 /*
1733 * If we don't need DAD, don't do it.
1734 * There are several cases:
1735 * - DAD is disabled (ip6_dad_count == 0)
1736 * - the interface address is anycast
1737 */
1738 IFA_LOCK(&ia->ia_ifa);
1739 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
1740 nd6log0(debug,
1741 "nd6_dad_start: not a tentative or optimistic address "
1742 "%s(%s)\n",
1743 ip6_sprintf(&ia->ia_addr.sin6_addr),
1744 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1745 IFA_UNLOCK(&ia->ia_ifa);
1746 return;
1747 }
1748 if (!ip6_dad_count || (ia->ia6_flags & IN6_IFF_ANYCAST) != 0) {
1749 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
1750 IFA_UNLOCK(&ia->ia_ifa);
1751 return;
1752 }
1753 IFA_UNLOCK(&ia->ia_ifa);
1754 if (ifa->ifa_ifp == NULL) {
1755 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1756 }
1757 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1758 (ifa->ifa_ifp->if_eflags & IFEF_IPV6_ND6ALT)) {
1759 return;
1760 }
1761 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1762 DAD_REMREF(dp);
1763 /* DAD already in progress */
1764 return;
1765 }
1766
1767 dp = zalloc_flags(dad_zone, Z_WAITOK | Z_ZERO);
1768 lck_mtx_init(&dp->dad_lock, &ifa_mtx_grp, &ifa_mtx_attr);
1769
1770 /* Callee adds one reference for us */
1771 dp = nd6_dad_attach(dp, ifa);
1772
1773 nd6log0(debug, "%s: starting %sDAD %sfor %s\n",
1774 if_name(ifa->ifa_ifp),
1775 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) ? "optimistic " : "",
1776 (tick_delay == NULL) ? "immediately " : "",
1777 ip6_sprintf(&ia->ia_addr.sin6_addr));
1778
1779 /*
1780 * Send NS packet for DAD, ip6_dad_count times.
1781 * Note that we must delay the first transmission, if this is the
1782 * first packet to be sent from the interface after interface
1783 * (re)initialization.
1784 */
1785 if (tick_delay == NULL) {
1786 u_int32_t retrans;
1787 struct nd_ifinfo *ndi = NULL;
1788
1789 nd6_dad_ns_output(dp, ifa);
1790 ndi = ND_IFINFO(ifa->ifa_ifp);
1791 VERIFY(ndi != NULL && ndi->initialized);
1792 lck_mtx_lock(&ndi->lock);
1793 retrans = ndi->retrans * hz / 1000;
1794 lck_mtx_unlock(&ndi->lock);
1795 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
1796 } else {
1797 int ntick;
1798
1799 if (*tick_delay == 0) {
1800 ntick = random() % (MAX_RTR_SOLICITATION_DELAY * hz);
1801 } else {
1802 ntick = *tick_delay + random() % (hz / 2);
1803 }
1804 *tick_delay = ntick;
1805 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa,
1806 ntick);
1807 }
1808
1809 DAD_REMREF(dp); /* drop our reference */
1810 }
1811
1812 static struct dadq *
nd6_dad_attach(struct dadq * dp,struct ifaddr * ifa)1813 nd6_dad_attach(struct dadq *dp, struct ifaddr *ifa)
1814 {
1815 lck_mtx_lock(&dad6_mutex);
1816 DAD_LOCK(dp);
1817 dp->dad_ifa = ifa;
1818 IFA_ADDREF(ifa); /* for dad_ifa */
1819 dp->dad_count = ip6_dad_count;
1820 dp->dad_ns_icount = dp->dad_na_icount = 0;
1821 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1822 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1823 VERIFY(!dp->dad_attached);
1824 dp->dad_attached = 1;
1825 dp->dad_lladdrlen = 0;
1826 DAD_ADDREF_LOCKED(dp); /* for caller */
1827 DAD_ADDREF_LOCKED(dp); /* for dadq_head list */
1828 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1829 DAD_UNLOCK(dp);
1830 lck_mtx_unlock(&dad6_mutex);
1831
1832 return dp;
1833 }
1834
1835 static void
nd6_dad_detach(struct dadq * dp,struct ifaddr * ifa)1836 nd6_dad_detach(struct dadq *dp, struct ifaddr *ifa)
1837 {
1838 int detached;
1839
1840 lck_mtx_lock(&dad6_mutex);
1841 DAD_LOCK(dp);
1842 if ((detached = dp->dad_attached)) {
1843 VERIFY(dp->dad_ifa == ifa);
1844 TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1845 dp->dad_list.tqe_next = NULL;
1846 dp->dad_list.tqe_prev = NULL;
1847 dp->dad_attached = 0;
1848 }
1849 DAD_UNLOCK(dp);
1850 lck_mtx_unlock(&dad6_mutex);
1851 if (detached) {
1852 DAD_REMREF(dp); /* drop dadq_head reference */
1853 }
1854 }
1855
1856 /*
1857 * terminate DAD unconditionally. used for address removals.
1858 */
1859 void
nd6_dad_stop(struct ifaddr * ifa)1860 nd6_dad_stop(struct ifaddr *ifa)
1861 {
1862 struct dadq *dp;
1863
1864 dp = nd6_dad_find(ifa, NULL);
1865 if (!dp) {
1866 /* DAD wasn't started yet */
1867 return;
1868 }
1869
1870 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1871
1872 nd6_dad_detach(dp, ifa);
1873 DAD_REMREF(dp); /* drop our reference */
1874 }
1875
1876 static void
nd6_unsol_na_output(struct ifaddr * ifa)1877 nd6_unsol_na_output(struct ifaddr *ifa)
1878 {
1879 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1880 struct ifnet *ifp = ifa->ifa_ifp;
1881 struct in6_addr saddr6, taddr6;
1882
1883 if ((ifp->if_flags & IFF_UP) == 0 ||
1884 (ifp->if_flags & IFF_RUNNING) == 0 ||
1885 (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
1886 return;
1887 }
1888
1889 IFA_LOCK_SPIN(&ia->ia_ifa);
1890 taddr6 = ia->ia_addr.sin6_addr;
1891 IFA_UNLOCK(&ia->ia_ifa);
1892 if (in6_setscope(&taddr6, ifp, NULL) != 0) {
1893 return;
1894 }
1895 saddr6 = in6addr_linklocal_allnodes;
1896 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
1897 return;
1898 }
1899
1900 nd6log(info, "%s: sending unsolicited NA\n",
1901 if_name(ifa->ifa_ifp));
1902
1903 nd6_na_output(ifp, &saddr6, &taddr6, ND_NA_FLAG_OVERRIDE, 1, NULL);
1904 }
1905
1906 static void
nd6_dad_timer(struct ifaddr * ifa)1907 nd6_dad_timer(struct ifaddr *ifa)
1908 {
1909 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1910 struct dadq *dp = NULL;
1911 struct nd_ifinfo *ndi = NULL;
1912 u_int32_t retrans;
1913
1914 /* Sanity check */
1915 if (ia == NULL) {
1916 nd6log0(error, "nd6_dad_timer: called with null parameter\n");
1917 goto done;
1918 }
1919
1920 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1921 __func__,
1922 ip6_sprintf(&ia->ia_addr.sin6_addr),
1923 if_name(ia->ia_ifp),
1924 ia->ia6_flags);
1925
1926 dp = nd6_dad_find(ifa, NULL);
1927 if (dp == NULL) {
1928 nd6log0(error, "nd6_dad_timer: DAD structure not found\n");
1929 goto done;
1930 }
1931 IFA_LOCK(&ia->ia_ifa);
1932 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1933 nd6log0(error, "nd6_dad_timer: called with duplicated address "
1934 "%s(%s)\n",
1935 ip6_sprintf(&ia->ia_addr.sin6_addr),
1936 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1937 IFA_UNLOCK(&ia->ia_ifa);
1938 goto done;
1939 }
1940 if ((ia->ia6_flags & IN6_IFF_DADPROGRESS) == 0) {
1941 nd6log0(error, "nd6_dad_timer: not a tentative or optimistic "
1942 "address %s(%s)\n",
1943 ip6_sprintf(&ia->ia_addr.sin6_addr),
1944 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1945 IFA_UNLOCK(&ia->ia_ifa);
1946 goto done;
1947 }
1948 IFA_UNLOCK(&ia->ia_ifa);
1949
1950 /* timeouted with IFF_{RUNNING,UP} check */
1951 DAD_LOCK(dp);
1952 if (dp->dad_ns_tcount > dad_maxtry) {
1953 DAD_UNLOCK(dp);
1954 nd6log0(info, "%s: could not run DAD, driver problem?\n",
1955 if_name(ifa->ifa_ifp));
1956
1957 nd6_dad_detach(dp, ifa);
1958 goto done;
1959 }
1960
1961 /* Need more checks? */
1962 if (dp->dad_ns_ocount < dp->dad_count) {
1963 DAD_UNLOCK(dp);
1964 /*
1965 * We have more NS to go. Send NS packet for DAD.
1966 */
1967 nd6_dad_ns_output(dp, ifa);
1968 ndi = ND_IFINFO(ifa->ifa_ifp);
1969 VERIFY(ndi != NULL && ndi->initialized);
1970 lck_mtx_lock(&ndi->lock);
1971 retrans = ndi->retrans * hz / 1000;
1972 lck_mtx_unlock(&ndi->lock);
1973 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
1974 } else {
1975 /*
1976 * We have transmitted sufficient number of DAD packets.
1977 * See what we've got.
1978 */
1979 if (dp->dad_na_icount > 0 || dp->dad_ns_icount) {
1980 /* We've seen NS or NA, means DAD has failed. */
1981 DAD_UNLOCK(dp);
1982 nd6log0(info,
1983 "%s: duplicate IPv6 address %s if:%s [timer]\n",
1984 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
1985 if_name(ia->ia_ifp));
1986 nd6_dad_duplicated(ifa);
1987 /* (*dp) will be freed in nd6_dad_duplicated() */
1988 #if SKYWALK
1989 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
1990 #endif /* SKYWALK */
1991 } else if (dad_enhanced != 0 &&
1992 dp->dad_ns_lcount > 0 &&
1993 dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1994 dp->dad_loopbackprobe = dp->dad_ns_lcount;
1995 dp->dad_count =
1996 dp->dad_ns_ocount + dad_maxtry - 1;
1997 DAD_UNLOCK(dp);
1998 ndi = ND_IFINFO(ifa->ifa_ifp);
1999 VERIFY(ndi != NULL && ndi->initialized);
2000 lck_mtx_lock(&ndi->lock);
2001 retrans = ndi->retrans * hz / 1000;
2002 lck_mtx_unlock(&ndi->lock);
2003
2004 /*
2005 * Sec. 4.1 in RFC 7527 requires transmission of
2006 * additional probes until the loopback condition
2007 * becomes clear when a looped back probe is detected.
2008 */
2009 nd6log0(info,
2010 "%s: a looped back NS message is detected during DAD for %s. Another DAD probe is being sent on interface %s.\n",
2011 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2012 if_name(ia->ia_ifp));
2013 /*
2014 * Send an NS immediately and increase dad_count by
2015 * nd6_mmaxtries - 1.
2016 */
2017 nd6_dad_ns_output(dp, ifa);
2018 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2019 goto done;
2020 } else {
2021 boolean_t txunsolna;
2022 DAD_UNLOCK(dp);
2023 /*
2024 * We are done with DAD. No NA came, no NS came.
2025 * No duplicate address found.
2026 */
2027 IFA_LOCK_SPIN(&ia->ia_ifa);
2028 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2029 IFA_UNLOCK(&ia->ia_ifa);
2030
2031 ndi = ND_IFINFO(ifa->ifa_ifp);
2032 VERIFY(ndi != NULL && ndi->initialized);
2033 lck_mtx_lock(&ndi->lock);
2034 txunsolna = (ndi->flags & ND6_IFF_REPLICATED) != 0;
2035 lck_mtx_unlock(&ndi->lock);
2036
2037 if (txunsolna) {
2038 nd6_unsol_na_output(ifa);
2039 }
2040
2041 nd6log0(debug,
2042 "%s: DAD complete for %s - no duplicates found %s\n",
2043 if_name(ifa->ifa_ifp),
2044 ip6_sprintf(&ia->ia_addr.sin6_addr),
2045 txunsolna ? ", tx unsolicited NA with O=1" : ".");
2046
2047 if (dp->dad_ns_lcount > 0) {
2048 nd6log0(debug,
2049 "%s: DAD completed while "
2050 "a looped back NS message is detected "
2051 "during DAD for %s om interface %s\n",
2052 __func__,
2053 ip6_sprintf(&ia->ia_addr.sin6_addr),
2054 if_name(ia->ia_ifp));
2055 }
2056
2057 in6_post_msg(ia->ia_ifp, KEV_INET6_NEW_USER_ADDR, ia,
2058 dp->dad_lladdr);
2059 nd6_dad_detach(dp, ifa);
2060 #if SKYWALK
2061 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2062 #endif /* SKYWALK */
2063 }
2064 }
2065
2066 done:
2067 if (dp != NULL) {
2068 DAD_REMREF(dp); /* drop our reference */
2069 }
2070 }
2071
2072 static void
nd6_dad_duplicated(struct ifaddr * ifa)2073 nd6_dad_duplicated(struct ifaddr *ifa)
2074 {
2075 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
2076 struct dadq *dp;
2077 struct ifnet *ifp = ifa->ifa_ifp;
2078 boolean_t candisable;
2079
2080 dp = nd6_dad_find(ifa, NULL);
2081 if (dp == NULL) {
2082 log(LOG_ERR, "%s: DAD structure not found.\n", __func__);
2083 return;
2084 }
2085 IFA_LOCK(&ia->ia_ifa);
2086 DAD_LOCK(dp);
2087 nd6log(error, "%s: NS in/out/loopback=%d/%d/%d, NA in=%d\n",
2088 __func__, dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
2089 dp->dad_na_icount);
2090 candisable = FALSE;
2091
2092 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr) &&
2093 !(ia->ia6_flags & IN6_IFF_SECURED)) {
2094 struct in6_addr in6;
2095 struct ifaddr *llifa = NULL;
2096 struct sockaddr_dl *sdl = NULL;
2097 uint8_t *lladdr = dp->dad_lladdr;
2098 uint8_t lladdrlen = dp->dad_lladdrlen;
2099
2100 /*
2101 * To avoid over-reaction, we only apply this logic when we are
2102 * very sure that hardware addresses are supposed to be unique.
2103 */
2104 switch (ifp->if_type) {
2105 case IFT_BRIDGE:
2106 case IFT_ETHER:
2107 case IFT_FDDI:
2108 case IFT_ATM:
2109 case IFT_IEEE1394:
2110 #ifdef IFT_IEEE80211
2111 case IFT_IEEE80211:
2112 #endif
2113 /*
2114 * Check if our hardware address matches the
2115 * link layer information received in the
2116 * NS/NA
2117 */
2118 llifa = ifp->if_lladdr;
2119 IFA_LOCK(llifa);
2120 sdl = (struct sockaddr_dl *)(void *)
2121 llifa->ifa_addr;
2122 if (lladdrlen == sdl->sdl_alen &&
2123 bcmp(lladdr, LLADDR(sdl), lladdrlen) == 0) {
2124 candisable = TRUE;
2125 }
2126 IFA_UNLOCK(llifa);
2127
2128 in6 = ia->ia_addr.sin6_addr;
2129 if (in6_iid_from_hw(ifp, &in6) != 0) {
2130 break;
2131 }
2132
2133 /* Refine decision about whether IPv6 can be disabled */
2134 if (candisable &&
2135 !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2136 /*
2137 * Apply this logic only to the embedded MAC
2138 * address form of link-local IPv6 address.
2139 */
2140 candisable = FALSE;
2141 } else if (lladdr == NULL &&
2142 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2143 /*
2144 * We received a NA with no target link-layer
2145 * address option. This means that someone else
2146 * has our address. Mark it as a hardware
2147 * duplicate so we disable IPv6 later on.
2148 */
2149 candisable = TRUE;
2150 }
2151 break;
2152 default:
2153 break;
2154 }
2155 }
2156 DAD_UNLOCK(dp);
2157
2158 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2159 ia->ia6_flags |= IN6_IFF_DUPLICATED;
2160 in6_event_enqueue_nwk_wq_entry(IN6_ADDR_MARKED_DUPLICATED,
2161 ia->ia_ifa.ifa_ifp, &ia->ia_addr.sin6_addr,
2162 0);
2163 IFA_UNLOCK(&ia->ia_ifa);
2164
2165 /* increment DAD collision counter */
2166 ++ip6stat.ip6s_dad_collide;
2167
2168 /* We are done with DAD, with duplicated address found. (failure) */
2169 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
2170
2171 IFA_LOCK(&ia->ia_ifa);
2172 log(LOG_ERR, "%s: DAD complete for %s - duplicate found.\n",
2173 if_name(ifp), ip6_sprintf(&ia->ia_addr.sin6_addr));
2174 IFA_UNLOCK(&ia->ia_ifa);
2175
2176 if (candisable) {
2177 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
2178 log(LOG_ERR, "%s: possible hardware address duplication "
2179 "detected, disabling IPv6 for interface.\n", if_name(ifp));
2180
2181 VERIFY((NULL != ndi) && (TRUE == ndi->initialized));
2182 ndi->flags |= ND6_IFF_IFDISABLED;
2183 /* Make sure to set IFEF_IPV6_DISABLED too */
2184 nd6_if_disable(ifp, TRUE);
2185 }
2186
2187 log(LOG_ERR,
2188 "%s: manual intervention may be required.\n",
2189 if_name(ifp));
2190
2191 /* Send an event to the configuration agent so that the
2192 * duplicate address will be notified to the user and will
2193 * be removed.
2194 */
2195 in6_post_msg(ifp, KEV_INET6_NEW_USER_ADDR, ia, dp->dad_lladdr);
2196 nd6_dad_detach(dp, ifa);
2197 DAD_REMREF(dp); /* drop our reference */
2198 }
2199
2200 static void
nd6_dad_ns_output(struct dadq * dp,struct ifaddr * ifa)2201 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
2202 {
2203 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
2204 struct ifnet *ifp = ifa->ifa_ifp;
2205 int i = 0;
2206 struct in6_addr taddr6;
2207
2208 DAD_LOCK(dp);
2209 dp->dad_ns_tcount++;
2210 if ((ifp->if_flags & IFF_UP) == 0) {
2211 DAD_UNLOCK(dp);
2212 return;
2213 }
2214 if ((ifp->if_flags & IFF_RUNNING) == 0) {
2215 DAD_UNLOCK(dp);
2216 return;
2217 }
2218
2219 dp->dad_ns_ocount++;
2220 DAD_UNLOCK(dp);
2221 IFA_LOCK_SPIN(&ia->ia_ifa);
2222 taddr6 = ia->ia_addr.sin6_addr;
2223 IFA_UNLOCK(&ia->ia_ifa);
2224 if (dad_enhanced != 0 && !(ifp->if_flags & IFF_POINTOPOINT)) {
2225 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) {
2226 dp->dad_nonce[i] = RandomULong();
2227 }
2228 /*
2229 * XXXHRS: Note that in the case that
2230 * DupAddrDetectTransmits > 1, multiple NS messages with
2231 * different nonces can be looped back in an unexpected
2232 * order. The current implementation recognizes only
2233 * the latest nonce on the sender side. Practically it
2234 * should work well in almost all cases.
2235 */
2236 }
2237 nd6_ns_output(ifp, NULL, &taddr6, NULL,
2238 (uint8_t *)&dp->dad_nonce[0]);
2239 }
2240
2241 /*
2242 * @brief Called to process DAD NS
2243 *
2244 * @param ifa is the pointer to the interface's address
2245 * @param lladdr is source link layer information
2246 * @param lladdrlen is source's linklayer length
2247 *
2248 * @return void
2249 */
2250 static void
nd6_dad_ns_input(struct ifaddr * ifa,char * lladdr,int lladdrlen,struct nd_opt_nonce * ndopt_nonce)2251 nd6_dad_ns_input(struct ifaddr *ifa, char *lladdr,
2252 int lladdrlen, struct nd_opt_nonce *ndopt_nonce)
2253 {
2254 struct dadq *dp;
2255 VERIFY(ifa != NULL);
2256
2257 /* Ignore Nonce option when Enhanced DAD is disabled. */
2258 if (dad_enhanced == 0) {
2259 ndopt_nonce = NULL;
2260 }
2261
2262 dp = nd6_dad_find(ifa, ndopt_nonce);
2263 if (dp == NULL) {
2264 return;
2265 }
2266
2267 DAD_LOCK(dp);
2268 ++dp->dad_ns_icount;
2269 if (lladdr && lladdrlen >= ETHER_ADDR_LEN) {
2270 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2271 /* fine to truncate as it is compared against sdl_alen */
2272 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2273 }
2274 DAD_UNLOCK(dp);
2275 DAD_REMREF(dp);
2276 }
2277
2278 /*
2279 * @brief Called to process received NA for DAD
2280 *
2281 * @param m is the pointer to the packet's mbuf
2282 * @param ifp is the pointer to the interface on which packet
2283 * was receicved.
2284 * @param taddr is pointer to target's IPv6 address
2285 * @param lladdr is target's link layer information
2286 * @param lladdrlen is target's linklayer length
2287 *
2288 * @return NULL if the packet is consumed by DAD processing, else
2289 * pointer to the mbuf.
2290 */
2291 static struct mbuf *
nd6_dad_na_input(struct mbuf * m,struct ifnet * ifp,struct in6_addr * taddr,caddr_t lladdr,int lladdrlen)2292 nd6_dad_na_input(struct mbuf *m, struct ifnet *ifp, struct in6_addr *taddr,
2293 caddr_t lladdr, int lladdrlen)
2294 {
2295 struct ifaddr *ifa = NULL;
2296 struct in6_ifaddr *ia = NULL;
2297 struct dadq *dp = NULL;
2298 struct nd_ifinfo *ndi = NULL;
2299 boolean_t replicated;
2300
2301 ifa = (struct ifaddr *) in6ifa_ifpwithaddr(ifp, taddr);
2302 if (ifa == NULL) {
2303 return m;
2304 }
2305
2306 replicated = FALSE;
2307
2308 /* Get the ND6_IFF_REPLICATED flag. */
2309 ndi = ND_IFINFO(ifp);
2310 if (ndi != NULL && ndi->initialized) {
2311 lck_mtx_lock(&ndi->lock);
2312 replicated = !!(ndi->flags & ND6_IFF_REPLICATED);
2313 lck_mtx_unlock(&ndi->lock);
2314 }
2315
2316 if (replicated) {
2317 nd6log(info, "%s: ignoring duplicate NA on "
2318 "replicated interface %s\n", __func__, if_name(ifp));
2319 goto done;
2320 }
2321
2322 /* Lock the interface address until done (see label below). */
2323 IFA_LOCK(ifa);
2324 ia = (struct in6_ifaddr *) ifa;
2325
2326 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
2327 IFA_UNLOCK(ifa);
2328 nd6log(info, "%s: ignoring duplicate NA on "
2329 "%s [DAD not in progress]\n", __func__,
2330 if_name(ifp));
2331 goto done;
2332 }
2333
2334 /* Some sleep proxies improperly send the client's Ethernet address in
2335 * the target link-layer address option, so detect this by comparing
2336 * the L2-header source address, if we have seen it, with the target
2337 * address, and ignoring the NA if they don't match.
2338 */
2339 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2340 struct ip6aux *ip6a = ip6_findaux(m);
2341 if (ip6a && (ip6a->ip6a_flags & IP6A_HASEEN) != 0 &&
2342 bcmp(ip6a->ip6a_ehsrc, lladdr, ETHER_ADDR_LEN) != 0) {
2343 IFA_UNLOCK(ifa);
2344 nd6log(error, "%s: ignoring duplicate NA on %s "
2345 "[eh_src != tgtlladdr]\n", __func__, if_name(ifp));
2346 goto done;
2347 }
2348 }
2349
2350 IFA_UNLOCK(ifa);
2351
2352 dp = nd6_dad_find(ifa, NULL);
2353 if (dp == NULL) {
2354 nd6log(info, "%s: no DAD structure for %s on %s.\n",
2355 __func__, ip6_sprintf(taddr), if_name(ifp));
2356 goto done;
2357 }
2358
2359 DAD_LOCK_SPIN(dp);
2360 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2361 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2362 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2363 }
2364 dp->dad_na_icount++;
2365 DAD_UNLOCK(dp);
2366 DAD_REMREF(dp);
2367
2368 /* remove the address. */
2369 nd6log(info,
2370 "%s: duplicate IPv6 address %s [processing NA on %s]\n", __func__,
2371 ip6_sprintf(taddr), if_name(ifp));
2372 done:
2373 IFA_LOCK_ASSERT_NOTHELD(ifa);
2374 IFA_REMREF(ifa);
2375 m_freem(m);
2376 return NULL;
2377 }
2378
2379 static void
dad_addref(struct dadq * dp,int locked)2380 dad_addref(struct dadq *dp, int locked)
2381 {
2382 if (!locked) {
2383 DAD_LOCK_SPIN(dp);
2384 } else {
2385 DAD_LOCK_ASSERT_HELD(dp);
2386 }
2387
2388 if (++dp->dad_refcount == 0) {
2389 panic("%s: dad %p wraparound refcnt", __func__, dp);
2390 /* NOTREACHED */
2391 }
2392 if (!locked) {
2393 DAD_UNLOCK(dp);
2394 }
2395 }
2396
2397 static void
dad_remref(struct dadq * dp)2398 dad_remref(struct dadq *dp)
2399 {
2400 struct ifaddr *ifa;
2401
2402 DAD_LOCK_SPIN(dp);
2403 if (dp->dad_refcount == 0) {
2404 panic("%s: dad %p negative refcnt", __func__, dp);
2405 }
2406 --dp->dad_refcount;
2407 if (dp->dad_refcount > 0) {
2408 DAD_UNLOCK(dp);
2409 return;
2410 }
2411 DAD_UNLOCK(dp);
2412
2413 if (dp->dad_attached ||
2414 dp->dad_list.tqe_next != NULL || dp->dad_list.tqe_prev != NULL) {
2415 panic("%s: attached dad=%p is being freed", __func__, dp);
2416 /* NOTREACHED */
2417 }
2418
2419 if ((ifa = dp->dad_ifa) != NULL) {
2420 IFA_REMREF(ifa); /* drop dad_ifa reference */
2421 dp->dad_ifa = NULL;
2422 }
2423
2424 lck_mtx_destroy(&dp->dad_lock, &ifa_mtx_grp);
2425 zfree(dad_zone, dp);
2426 }
2427
2428 void
nd6_llreach_set_reachable(struct ifnet * ifp,void * addr,unsigned int alen)2429 nd6_llreach_set_reachable(struct ifnet *ifp, void *addr, unsigned int alen)
2430 {
2431 /* Nothing more to do if it's disabled */
2432 if (nd6_llreach_base == 0) {
2433 return;
2434 }
2435
2436 ifnet_llreach_set_reachable(ifp, ETHERTYPE_IPV6, addr, alen);
2437 }
2438
2439 void
nd6_alt_node_addr_decompose(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr_dl * sdl,struct sockaddr_in6 * sin6)2440 nd6_alt_node_addr_decompose(struct ifnet *ifp, struct sockaddr *sa,
2441 struct sockaddr_dl* sdl, struct sockaddr_in6 *sin6)
2442 {
2443 static const size_t EUI64_LENGTH = 8;
2444
2445 VERIFY(nd6_need_cache(ifp));
2446 VERIFY(sa);
2447 VERIFY(sdl && (void *)sa != (void *)sdl);
2448 VERIFY(sin6 && (void *)sa != (void *)sin6);
2449
2450 bzero(sin6, sizeof(*sin6));
2451 sin6->sin6_len = sizeof *sin6;
2452 sin6->sin6_family = AF_INET6;
2453
2454 bzero(sdl, sizeof(*sdl));
2455 sdl->sdl_len = sizeof *sdl;
2456 sdl->sdl_family = AF_LINK;
2457 sdl->sdl_type = ifp->if_type;
2458 sdl->sdl_index = ifp->if_index;
2459
2460 switch (sa->sa_family) {
2461 case AF_INET6: {
2462 struct sockaddr_in6 *sin6a = (struct sockaddr_in6 *)(void *)sa;
2463 struct in6_addr *in6 = &sin6a->sin6_addr;
2464
2465 VERIFY(sa->sa_len == sizeof *sin6);
2466 VERIFY(strlen(ifp->if_name) <= IFNAMSIZ);
2467
2468 sdl->sdl_nlen = (u_char)strlen(ifp->if_name);
2469 bcopy(ifp->if_name, sdl->sdl_data, sdl->sdl_nlen);
2470 if (in6->s6_addr[11] == 0xff && in6->s6_addr[12] == 0xfe) {
2471 sdl->sdl_alen = ETHER_ADDR_LEN;
2472 LLADDR(sdl)[0] = (in6->s6_addr[8] ^ ND6_EUI64_UBIT);
2473 LLADDR(sdl)[1] = in6->s6_addr[9];
2474 LLADDR(sdl)[2] = in6->s6_addr[10];
2475 LLADDR(sdl)[3] = in6->s6_addr[13];
2476 LLADDR(sdl)[4] = in6->s6_addr[14];
2477 LLADDR(sdl)[5] = in6->s6_addr[15];
2478 } else {
2479 sdl->sdl_alen = EUI64_LENGTH;
2480 bcopy(&in6->s6_addr[8], LLADDR(sdl), EUI64_LENGTH);
2481 }
2482
2483 sdl->sdl_slen = 0;
2484 break;
2485 }
2486 case AF_LINK: {
2487 struct sockaddr_dl *sdla = (struct sockaddr_dl *)(void *)sa;
2488 struct in6_addr *in6 = &sin6->sin6_addr;
2489 caddr_t lla = LLADDR(sdla);
2490
2491 VERIFY(sa->sa_len <= sizeof(*sdl));
2492 bcopy(sa, sdl, sa->sa_len);
2493
2494 sin6->sin6_scope_id = sdla->sdl_index;
2495 if (sin6->sin6_scope_id == 0) {
2496 sin6->sin6_scope_id = ifp->if_index;
2497 }
2498 in6->s6_addr[0] = 0xfe;
2499 in6->s6_addr[1] = 0x80;
2500 if (sdla->sdl_alen == EUI64_LENGTH) {
2501 bcopy(lla, &in6->s6_addr[8], EUI64_LENGTH);
2502 } else {
2503 VERIFY(sdla->sdl_alen == ETHER_ADDR_LEN);
2504
2505 in6->s6_addr[8] = ((uint8_t) lla[0] ^ ND6_EUI64_UBIT);
2506 in6->s6_addr[9] = (uint8_t) lla[1];
2507 in6->s6_addr[10] = (uint8_t) lla[2];
2508 in6->s6_addr[11] = 0xff;
2509 in6->s6_addr[12] = 0xfe;
2510 in6->s6_addr[13] = (uint8_t) lla[3];
2511 in6->s6_addr[14] = (uint8_t) lla[4];
2512 in6->s6_addr[15] = (uint8_t) lla[5];
2513 }
2514
2515 break;
2516 }
2517 default:
2518 VERIFY(false);
2519 break;
2520 }
2521 }
2522
2523 int
nd6_alt_node_present(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl,int32_t rssi,int lqm,int npm)2524 nd6_alt_node_present(struct ifnet *ifp, struct sockaddr_in6 *sin6,
2525 struct sockaddr_dl *sdl, int32_t rssi, int lqm, int npm)
2526 {
2527 struct rtentry *rt;
2528 struct llinfo_nd6 *ln;
2529 struct if_llreach *lr = NULL;
2530 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2531 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2532
2533 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) &&
2534 (temp_embedded_id == 0)) {
2535 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2536 }
2537 if (!in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) && (temp_ifscope_id == 0)) {
2538 sin6->sin6_scope_id = ifp->if_index;
2539 }
2540
2541 nd6_cache_lladdr(ifp, &sin6->sin6_addr, LLADDR(sdl), sdl->sdl_alen,
2542 ND_NEIGHBOR_ADVERT, 0);
2543
2544 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2545 lck_mtx_lock(rnh_lock);
2546
2547 rt = rtalloc1_scoped_locked((struct sockaddr *)sin6, 1, 0,
2548 ifp->if_index);
2549
2550 /* Restore the address that was passed to us */
2551 if (in6_embedded_scope && temp_embedded_id == 0) {
2552 sin6->sin6_addr.s6_addr16[1] = 0;
2553 }
2554 if (!in6_embedded_scope && temp_ifscope_id == 0) {
2555 sin6->sin6_scope_id = 0;
2556 }
2557
2558 if (rt != NULL) {
2559 RT_LOCK(rt);
2560 VERIFY(rt->rt_flags & RTF_LLINFO);
2561 VERIFY(rt->rt_llinfo);
2562
2563 ln = rt->rt_llinfo;
2564 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
2565 ln_setexpire(ln, 0);
2566
2567 lr = ln->ln_llreach;
2568 if (lr) {
2569 IFLR_LOCK(lr);
2570 lr->lr_rssi = rssi;
2571 lr->lr_lqm = (int32_t) lqm;
2572 lr->lr_npm = (int32_t) npm;
2573 IFLR_UNLOCK(lr);
2574 }
2575
2576 RT_UNLOCK(rt);
2577 RT_REMREF(rt);
2578 }
2579
2580 lck_mtx_unlock(rnh_lock);
2581
2582 if (rt == NULL) {
2583 log(LOG_ERR, "%s: failed to add/update host route to %s.\n",
2584 __func__, ip6_sprintf(&sin6->sin6_addr));
2585 if (ip6_p2p_debug) {
2586 panic("%s: failed to add/update host route to %s.\n",
2587 __func__, ip6_sprintf(&sin6->sin6_addr));
2588 }
2589 return EHOSTUNREACH;
2590 } else {
2591 nd6log(debug, "%s: host route to %s [lr=0x%llx]\n",
2592 __func__, ip6_sprintf(&sin6->sin6_addr),
2593 (uint64_t)VM_KERNEL_ADDRPERM(lr));
2594 return 0;
2595 }
2596 }
2597
2598 void
nd6_alt_node_absent(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl)2599 nd6_alt_node_absent(struct ifnet *ifp, struct sockaddr_in6 *sin6, struct sockaddr_dl *sdl)
2600 {
2601 struct rtentry *rt;
2602 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2603 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2604
2605 nd6log(debug, "%s: host route to %s\n", __func__,
2606 ip6_sprintf(&sin6->sin6_addr));
2607
2608 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) &&
2609 (temp_embedded_id == 0)) {
2610 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2611 }
2612 if (!in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) && (temp_ifscope_id == 0)) {
2613 sin6->sin6_scope_id = ifp->if_index;
2614 }
2615
2616 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2617 lck_mtx_lock(rnh_lock);
2618
2619 rt = rtalloc1_scoped_locked((struct sockaddr *)sin6, 0, 0,
2620 ifp->if_index);
2621
2622 /* Restore the address that was passed to us */
2623 if (in6_embedded_scope && temp_embedded_id == 0) {
2624 sin6->sin6_addr.s6_addr16[1] = 0;
2625 }
2626 if (!in6_embedded_scope && temp_ifscope_id == 0) {
2627 sin6->sin6_scope_id = 0;
2628 }
2629
2630 if (rt != NULL) {
2631 RT_LOCK(rt);
2632
2633 if (!(rt->rt_flags & (RTF_CLONING | RTF_PRCLONING)) &&
2634 (rt->rt_flags & (RTF_HOST | RTF_LLINFO | RTF_WASCLONED)) ==
2635 (RTF_HOST | RTF_LLINFO | RTF_WASCLONED)) {
2636 /*
2637 * Copy the link layer information in SDL when present
2638 * as it later gets used to issue the kernel event for
2639 * node absence.
2640 */
2641 if (sdl != NULL && rt->rt_gateway != NULL &&
2642 rt->rt_gateway->sa_family == AF_LINK &&
2643 SDL(rt->rt_gateway)->sdl_len <= sizeof(*sdl)) {
2644 bcopy(rt->rt_gateway, sdl, SDL(rt->rt_gateway)->sdl_len);
2645 }
2646
2647 rt->rt_flags |= RTF_CONDEMNED;
2648 RT_UNLOCK(rt);
2649
2650 (void) rtrequest_locked(RTM_DELETE, rt_key(rt),
2651 (struct sockaddr *)NULL, rt_mask(rt), 0,
2652 (struct rtentry **)NULL);
2653
2654 rtfree_locked(rt);
2655 } else {
2656 RT_REMREF_LOCKED(rt);
2657 RT_UNLOCK(rt);
2658 }
2659 }
2660
2661 lck_mtx_unlock(rnh_lock);
2662 }
2663