1 /*
2 * Copyright (c) 2000-2022 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 struct sockaddr_in6 hostrtmask;
149
150 static int nd6_llreach_base = 30; /* seconds */
151 SYSCTL_DECL(_net_inet6_icmp6);
152 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_llreach_base,
153 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_llreach_base, 0,
154 "default ND6 link-layer reachability max lifetime (in seconds)");
155
156 int dad_enhanced = ND6_DAD_ENHANCED_DEFAULT;
157 SYSCTL_DECL(_net_inet6_ip6);
158 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_RW | CTLFLAG_LOCKED,
159 &dad_enhanced, 0,
160 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
161
162 static uint32_t nd6_dad_nonce_max_count = 3;
163 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, nd6_dad_nonce_max_count,
164 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_dad_nonce_max_count, 0, "Number of times to ignore same nonce for DAD");
165
166 #if DEBUG || DEVELOPMENT
167 static int ip6_p2p_debug = 0;
168 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, ip6_p2p_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
169 &ip6_p2p_debug, 0,
170 "Enable more instrumentation for IPv6 P2P use-case");
171 #endif
172
173 /*
174 * Obtain a link-layer source cache entry for the sender.
175 *
176 * NOTE: This is currently only for ND6/Ethernet.
177 */
178 void
nd6_llreach_alloc(struct rtentry * rt,struct ifnet * ifp,void * addr,unsigned int alen,boolean_t solicited)179 nd6_llreach_alloc(struct rtentry *rt, struct ifnet *ifp, void *addr,
180 unsigned int alen, boolean_t solicited)
181 {
182 struct llinfo_nd6 *ln = rt->rt_llinfo;
183
184 if (nd6_llreach_base != 0 &&
185 (ln->ln_expire != 0 || (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) &&
186 !(rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
187 ifp->if_addrlen == IF_LLREACH_MAXLEN && /* Ethernet */
188 alen == ifp->if_addrlen) {
189 struct if_llreach *lr;
190 const char *why = NULL, *type = "";
191
192 /* Become a regular mutex, just in case */
193 RT_CONVERT_LOCK(rt);
194
195 if ((lr = ln->ln_llreach) != NULL) {
196 type = (solicited ? "ND6 advertisement" :
197 "ND6 unsolicited announcement");
198 /*
199 * If target has changed, create a new record;
200 * otherwise keep existing record.
201 */
202 IFLR_LOCK(lr);
203 if (bcmp(addr, lr->lr_key.addr, alen) != 0) {
204 IFLR_UNLOCK(lr);
205 /* Purge any link-layer info caching */
206 VERIFY(rt->rt_llinfo_purge != NULL);
207 rt->rt_llinfo_purge(rt);
208 lr = NULL;
209 why = " for different target HW address; "
210 "using new llreach record";
211 } else {
212 lr->lr_probes = 0; /* reset probe count */
213 IFLR_UNLOCK(lr);
214 if (solicited) {
215 why = " for same target HW address; "
216 "keeping existing llreach record";
217 }
218 }
219 }
220
221 if (lr == NULL) {
222 lr = ln->ln_llreach = ifnet_llreach_alloc(ifp,
223 ETHERTYPE_IPV6, addr, alen, nd6_llreach_base);
224 if (lr != NULL) {
225 lr->lr_probes = 0; /* reset probe count */
226 if (why == NULL) {
227 why = "creating new llreach record";
228 }
229 }
230 }
231
232 if (nd6_debug && lr != NULL && why != NULL) {
233 char tmp[MAX_IPv6_STR_LEN];
234
235 nd6log(debug, "%s: %s%s for %s\n", if_name(ifp),
236 type, why, inet_ntop(AF_INET6,
237 &SIN6(rt_key(rt))->sin6_addr, tmp, sizeof(tmp)));
238 }
239 }
240 }
241
242 void
nd6_llreach_use(struct llinfo_nd6 * ln)243 nd6_llreach_use(struct llinfo_nd6 *ln)
244 {
245 if (ln->ln_llreach != NULL) {
246 ln->ln_lastused = net_uptime();
247 }
248 }
249
250 /*
251 * Input a Neighbor Solicitation Message.
252 *
253 * Based on RFC 4861
254 * Based on RFC 4862 (duplicate address detection)
255 */
256 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)257 nd6_ns_input(
258 struct mbuf *m,
259 int off,
260 int icmp6len)
261 {
262 struct ifnet *ifp = m->m_pkthdr.rcvif;
263 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
264 struct nd_neighbor_solicit *nd_ns = NULL;
265 struct in6_addr saddr6 = ip6->ip6_src;
266 struct in6_addr daddr6 = ip6->ip6_dst;
267 uint32_t saddr_ifscope = IN6_IS_SCOPE_EMBED(&saddr6) ? ip6_input_getsrcifscope(m) : IFSCOPE_NONE;
268 struct in6_addr taddr6 = {};
269 struct in6_addr myaddr6 = {};
270 uint32_t myaddr_ifscope = IFSCOPE_NONE;
271 char *lladdr = NULL;
272 struct ifaddr *ifa = NULL;
273 int lladdrlen = 0;
274 int anycast = 0, proxy = 0, dadprogress = 0;
275 int tlladdr = 0;
276 union nd_opts ndopts = {};
277 struct sockaddr_dl proxydl = {};
278 boolean_t advrouter = FALSE;
279 boolean_t is_dad_probe = FALSE;
280 int oflgclr = 0;
281 uint32_t taddr_ifscope;
282
283 /* Expect 32-bit aligned data pointer on strict-align platforms */
284 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
285
286 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
287 ip6 = mtod(m, struct ip6_hdr *);
288 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
289 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
290
291 taddr6 = nd_ns->nd_ns_target;
292 if (in6_setscope(&taddr6, ifp, &taddr_ifscope) != 0) {
293 goto bad;
294 }
295
296 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
297 nd6log(error,
298 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
299 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
300 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
301 goto bad;
302 }
303
304 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(&saddr6);
305 if (is_dad_probe) {
306 /* dst has to be a solicited node multicast address. */
307 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
308 /* don't check ifindex portion */
309 daddr6.s6_addr32[1] == 0 &&
310 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
311 daddr6.s6_addr8[12] == 0xff) {
312 ; /* good */
313 } else {
314 nd6log(info, "nd6_ns_input: bad DAD packet "
315 "(wrong ip6 dst)\n");
316 goto bad;
317 }
318 } else if (!nd6_onlink_ns_rfc4861) {
319 struct sockaddr_in6 src_sa6;
320
321 /*
322 * According to recent IETF discussions, it is not a good idea
323 * to accept a NS from an address which would not be deemed
324 * to be a neighbor otherwise. This point is expected to be
325 * clarified in future revisions of the specification.
326 */
327 bzero(&src_sa6, sizeof(src_sa6));
328 src_sa6.sin6_family = AF_INET6;
329 src_sa6.sin6_len = sizeof(src_sa6);
330 src_sa6.sin6_addr = saddr6;
331 if (!in6_embedded_scope) {
332 src_sa6.sin6_scope_id = saddr_ifscope;
333 }
334 if (!nd6_is_addr_neighbor(&src_sa6, ifp, 0)) {
335 nd6log(info, "nd6_ns_input: "
336 "NS packet from non-neighbor\n");
337 goto bad;
338 }
339 }
340
341 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
342 nd6log(info, "nd6_ns_input: bad NS target (multicast)\n");
343 goto bad;
344 }
345
346 icmp6len -= sizeof(*nd_ns);
347 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
348 if (nd6_options(&ndopts) < 0) {
349 nd6log(info,
350 "nd6_ns_input: invalid ND option, ignored\n");
351 /* nd6_options have incremented stats */
352 goto freeit;
353 }
354
355 if (ndopts.nd_opts_src_lladdr) {
356 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
357 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
358 }
359
360 if (is_dad_probe && lladdr) {
361 nd6log(info, "nd6_ns_input: bad DAD packet "
362 "(link-layer address option)\n");
363 goto bad;
364 }
365
366 /*
367 * Attaching target link-layer address to the NA?
368 * (RFC 2461 7.2.4)
369 *
370 * NS IP dst is unicast/anycast MUST NOT add
371 * NS IP dst is solicited-node multicast MUST add
372 *
373 * In implementation, we add target link-layer address by default.
374 * We do not add one in MUST NOT cases.
375 */
376 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) {
377 tlladdr = 0;
378 } else {
379 tlladdr = 1;
380 }
381
382 /*
383 * Target address (taddr6) must be either:
384 * (1) Valid unicast/anycast address for my receiving interface,
385 * (2) Unicast address for which I'm offering proxy service, or
386 * (3) "tentative" or "optimistic" address [DAD is in progress].
387 */
388 /* (1) and (3) check. */
389 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
390
391 /* (2) check. */
392 if (ifa == NULL) {
393 struct rtentry *rt;
394 struct sockaddr_in6 tsin6;
395
396 bzero(&tsin6, sizeof tsin6);
397 tsin6.sin6_len = sizeof(struct sockaddr_in6);
398 tsin6.sin6_family = AF_INET6;
399 tsin6.sin6_addr = taddr6;
400
401 rt = rtalloc1_scoped((struct sockaddr *)&tsin6, 0, 0,
402 ifp->if_index);
403
404 if (rt != NULL) {
405 RT_LOCK(rt);
406 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
407 rt->rt_gateway->sa_family == AF_LINK) {
408 /*
409 * proxy NDP for single entry
410 */
411 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
412 ifp, IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
413 if (ifa) {
414 proxy = 1;
415 proxydl = *SDL(rt->rt_gateway);
416 }
417 }
418 RT_UNLOCK(rt);
419 rtfree(rt);
420 }
421 }
422 if (ifa == NULL && ip6_forwarding && nd6_prproxy) {
423 /*
424 * Is the target address part of the prefix that is being
425 * proxied and installed on another interface?
426 */
427 ifa = (struct ifaddr *)in6ifa_prproxyaddr(&taddr6, taddr_ifscope);
428 }
429 if (ifa == NULL) {
430 /*
431 * We've got an NS packet, and we don't have that address
432 * assigned for us. We MUST silently ignore it on this
433 * interface, c.f. RFC 4861 7.2.3.
434 *
435 * Forwarding associated with NDPRF_PRPROXY may apply.
436 */
437 if (ip6_forwarding && nd6_prproxy) {
438 nd6_prproxy_ns_input(ifp, &saddr6, lladdr,
439 lladdrlen, &daddr6, &taddr6,
440 (ndopts.nd_opts_nonce == NULL) ? NULL :
441 ndopts.nd_opts_nonce->nd_opt_nonce);
442 }
443 goto freeit;
444 }
445 IFA_LOCK(ifa);
446 myaddr6 = *IFA_IN6(ifa);
447 myaddr_ifscope = IFA_SIN6_SCOPE(ifa);
448 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
449 dadprogress =
450 ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DADPROGRESS;
451 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) {
452 IFA_UNLOCK(ifa);
453 goto freeit;
454 }
455 IFA_UNLOCK(ifa);
456
457 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
458 nd6log(info,
459 "nd6_ns_input: lladdrlen mismatch for %s "
460 "(if %d, NS packet %d)\n",
461 ip6_sprintf(&taddr6), ifp->if_addrlen, lladdrlen - 2);
462 goto bad;
463 }
464
465 if (in6_are_addr_equal_scoped(&myaddr6, &saddr6, myaddr_ifscope, saddr_ifscope)) {
466 nd6log(info,
467 "nd6_ns_input: duplicate IP6 address %s\n",
468 ip6_sprintf(&saddr6));
469 goto freeit;
470 }
471
472 /*
473 * We have neighbor solicitation packet, with target address equals to
474 * one of my DAD in-progress addresses.
475 *
476 * src addr how to process?
477 * --- ---
478 * multicast of course, invalid (rejected in ip6_input)
479 * unicast somebody is doing address resolution
480 * unspec dup address detection
481 *
482 * The processing is defined in the "draft standard" RFC 4862 (and by
483 * RFC 4429, which is a "proposed standard" update to its obsolete
484 * predecessor, RFC 2462) The reason optimistic DAD is not included
485 * in RFC 4862 is entirely due to IETF procedural considerations.
486 */
487 if (dadprogress) {
488 /*
489 * If source address is unspecified address, it is for
490 * duplicate address detection.
491 *
492 * If not, the packet is for addess resolution;
493 * silently ignore it when not optimistic
494 *
495 * Per RFC 4429 the reply for an optimistic address must
496 * have the Override flag cleared
497 */
498 if (!is_dad_probe && (dadprogress & IN6_IFF_OPTIMISTIC) != 0) {
499 oflgclr = 1;
500 } else {
501 if (is_dad_probe) {
502 nd6_dad_ns_input(ifa, lladdr, lladdrlen, ndopts.nd_opts_nonce);
503 }
504
505 goto freeit;
506 }
507 }
508
509 /* Are we an advertising router on this interface? */
510 advrouter = (ifp->if_ipv6_router_mode != IPV6_ROUTER_MODE_DISABLED);
511
512 /*
513 * If the source address is unspecified address, entries must not
514 * be created or updated.
515 * It looks that sender is performing DAD. If I'm using the address,
516 * and it's a "preferred" address, i.e. not optimistic, then output NA
517 * toward all-node multicast address, to tell the sender that I'm using
518 * the address.
519 * S bit ("solicited") must be zero.
520 */
521 if (is_dad_probe) {
522 saddr6 = in6addr_linklocal_allnodes;
523 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
524 goto bad;
525 }
526 if ((dadprogress & IN6_IFF_OPTIMISTIC) == 0) {
527 nd6_na_output(ifp, &saddr6, &taddr6,
528 ((anycast || proxy || !tlladdr) ? 0 :
529 ND_NA_FLAG_OVERRIDE) | (advrouter ?
530 ND_NA_FLAG_ROUTER : 0), tlladdr, proxy ?
531 (struct sockaddr *)&proxydl : NULL);
532 }
533 goto freeit;
534 }
535
536 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
537 ND_NEIGHBOR_SOLICIT, 0, NULL);
538
539 nd6_na_output(ifp, &saddr6, &taddr6,
540 ((anycast || proxy || !tlladdr || oflgclr) ? 0 : ND_NA_FLAG_OVERRIDE) |
541 (advrouter ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
542 tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL);
543 freeit:
544 m_freem(m);
545 if (ifa != NULL) {
546 IFA_REMREF(ifa);
547 }
548 return;
549
550 bad:
551 nd6log(error, "nd6_ns_input: src=%s\n", ip6_sprintf(&saddr6));
552 nd6log(error, "nd6_ns_input: dst=%s\n", ip6_sprintf(&daddr6));
553 nd6log(error, "nd6_ns_input: tgt=%s\n", ip6_sprintf(&taddr6));
554 icmp6stat.icp6s_badns++;
555 m_freem(m);
556 if (ifa != NULL) {
557 IFA_REMREF(ifa);
558 }
559 }
560
561 /*
562 * Output a Neighbor Solicitation Message. Caller specifies:
563 * - ICMP6 header source IP6 address
564 * - ND6 header target IP6 address
565 * - ND6 header source datalink address
566 *
567 * Based on RFC 4861
568 * Based on RFC 4862 (duplicate address detection)
569 * Based on RFC 4429 (optimistic duplicate address detection)
570 *
571 * Caller must bump up ln->ln_rt refcnt to make sure 'ln' doesn't go
572 * away if there is a llinfo_nd6 passed in.
573 */
574 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * daddr6,const struct in6_addr * taddr6,struct llinfo_nd6 * ln,uint8_t * nonce)575 nd6_ns_output(
576 struct ifnet *ifp,
577 const struct in6_addr *daddr6,
578 const struct in6_addr *taddr6,
579 struct llinfo_nd6 *ln, /* for source address determination */
580 uint8_t *nonce) /* duplicated address detection */
581 {
582 struct mbuf *m;
583 struct ip6_hdr *ip6;
584 struct nd_neighbor_solicit *nd_ns;
585 struct in6_ifaddr *ia = NULL;
586 struct in6_addr *src, src_in, src_storage;
587 struct ip6_moptions *im6o = NULL;
588 struct ifnet *outif = NULL;
589 int icmp6len;
590 int maxlen;
591 int flags;
592 caddr_t mac;
593 struct route_in6 ro;
594 struct ip6_out_args ip6oa;
595 u_int32_t rtflags = 0;
596 boolean_t is_optimistic = FALSE;
597
598 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) || IN6_IS_ADDR_MULTICAST(taddr6)) {
599 return;
600 }
601
602 bzero(&ro, sizeof(ro));
603 bzero(&ip6oa, sizeof(ip6oa));
604 ip6oa.ip6oa_boundif = ifp->if_index;
605 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
606 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
607 IP6OAF_MANAGEMENT_ALLOWED;
608 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
609 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
610
611 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
612
613 /* estimate the size of message */
614 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
615 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
616 if (max_linkhdr + maxlen >= MCLBYTES) {
617 #if DIAGNOSTIC
618 printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
619 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
620 #endif
621 return;
622 }
623
624 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
625 if (m && max_linkhdr + maxlen >= MHLEN) {
626 MCLGET(m, M_DONTWAIT);
627 if ((m->m_flags & M_EXT) == 0) {
628 m_free(m);
629 m = NULL;
630 }
631 }
632 if (m == NULL) {
633 return;
634 }
635 m->m_pkthdr.rcvif = NULL;
636
637 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
638 m->m_flags |= M_MCAST;
639
640 im6o = ip6_allocmoptions(Z_NOWAIT);
641 if (im6o == NULL) {
642 m_freem(m);
643 return;
644 }
645
646 im6o->im6o_multicast_ifp = ifp;
647 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
648 im6o->im6o_multicast_loop = 0;
649 }
650
651 icmp6len = sizeof(*nd_ns);
652 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
653 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
654
655 /* fill neighbor solicitation packet */
656 ip6 = mtod(m, struct ip6_hdr *);
657 ip6->ip6_flow = 0;
658 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
659 ip6->ip6_vfc |= IPV6_VERSION;
660 /* ip6->ip6_plen will be set later */
661 ip6->ip6_nxt = IPPROTO_ICMPV6;
662 ip6->ip6_hlim = IPV6_MAXHLIM;
663 if (daddr6) {
664 ip6->ip6_dst = *daddr6;
665 ip6_output_setdstifscope(m, ifp->if_index, NULL);
666 } else {
667 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
668 ip6->ip6_dst.s6_addr16[1] = 0;
669 ip6->ip6_dst.s6_addr32[1] = 0;
670 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
671 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
672 ip6->ip6_dst.s6_addr8[12] = 0xff;
673 ip6_output_setdstifscope(m, ifp->if_index, NULL);
674 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) {
675 goto bad;
676 }
677 }
678 if (nonce == NULL) {
679 /*
680 * RFC2461 7.2.2:
681 * "If the source address of the packet prompting the
682 * solicitation is the same as one of the addresses assigned
683 * to the outgoing interface, that address SHOULD be placed
684 * in the IP Source Address of the outgoing solicitation.
685 * Otherwise, any one of the addresses assigned to the
686 * interface should be used."
687 *
688 * We use the source address for the prompting packet
689 * (saddr6), if:
690 * - saddr6 is given from the caller (by giving "ln"), and
691 * - saddr6 belongs to the outgoing interface.
692 * Otherwise, we perform the source address selection as usual.
693 */
694 struct ip6_hdr *hip6; /* hold ip6 */
695 struct in6_addr *hsrc = NULL;
696
697 /* Caller holds ref on this route */
698 if (ln != NULL) {
699 RT_LOCK(ln->ln_rt);
700 /*
701 * assuming every packet in ln_hold has the same IP
702 * header
703 */
704 if (ln->ln_hold != NULL) {
705 hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
706 /* XXX pullup? */
707 if (sizeof(*hip6) < ln->ln_hold->m_len) {
708 hsrc = &hip6->ip6_src;
709 } else {
710 hsrc = NULL;
711 }
712 }
713 /* Update probe count, if applicable */
714 if (ln->ln_llreach != NULL) {
715 IFLR_LOCK_SPIN(ln->ln_llreach);
716 ln->ln_llreach->lr_probes++;
717 IFLR_UNLOCK(ln->ln_llreach);
718 }
719 rtflags = ln->ln_rt->rt_flags;
720 RT_UNLOCK(ln->ln_rt);
721 }
722 if (hsrc != NULL && (ia = in6ifa_ifpwithaddr(ifp, hsrc)) &&
723 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0) {
724 src = hsrc;
725 } else {
726 int error;
727 struct sockaddr_in6 dst_sa;
728
729 bzero(&dst_sa, sizeof(dst_sa));
730 dst_sa.sin6_family = AF_INET6;
731 dst_sa.sin6_len = sizeof(dst_sa);
732 dst_sa.sin6_addr = ip6->ip6_dst;
733
734 src = in6_selectsrc(&dst_sa, NULL,
735 NULL, &ro, NULL, &src_storage, ip6oa.ip6oa_boundif,
736 &error);
737 if (src == NULL) {
738 nd6log(debug,
739 "nd6_ns_output: source can't be "
740 "determined: dst=%s, error=%d\n",
741 ip6_sprintf(&dst_sa.sin6_addr),
742 error);
743 goto bad;
744 }
745
746 if (ia != NULL) {
747 IFA_REMREF(&ia->ia_ifa);
748 ia = NULL;
749 }
750 /*
751 * RFC 4429 section 3.2:
752 * When a node has a unicast packet to send
753 * from an Optimistic Address to a neighbor,
754 * but does not know the neighbor's link-layer
755 * address, it MUST NOT perform Address
756 * Resolution.
757 */
758 ia = in6ifa_ifpwithaddr(ifp, src);
759 if (ia == NULL) {
760 nd6log(debug,
761 "nd6_ns_output: no preferred source "
762 "available: dst=%s\n",
763 ip6_sprintf(&dst_sa.sin6_addr));
764 goto bad;
765 }
766 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
767 is_optimistic = TRUE;
768 nd6log(debug,
769 "nd6_ns_output: preferred source "
770 "available is optimistic: dst=%s\n",
771 ip6_sprintf(&dst_sa.sin6_addr));
772 }
773 }
774 } else {
775 /*
776 * Source address for DAD packet must always be IPv6
777 * unspecified address. (0::0)
778 * We actually don't have to 0-clear the address (we did it
779 * above), but we do so here explicitly to make the intention
780 * clearer.
781 */
782 bzero(&src_in, sizeof(src_in));
783 src = &src_in;
784 ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_SRCADDR;
785 }
786 ip6->ip6_src = *src;
787 ip6_output_setsrcifscope(m, ifp->if_index, ia);
788 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
789 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
790 nd_ns->nd_ns_code = 0;
791 nd_ns->nd_ns_reserved = 0;
792 nd_ns->nd_ns_target = *taddr6;
793 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
794
795 /*
796 * Add source link-layer address option.
797 *
798 * spec implementation
799 * --- ---
800 * DAD packet MUST NOT do not add the option
801 * Source is optimistic MUST NOT do not add the option
802 * there's no link layer address:
803 * impossible do not add the option
804 * there's link layer address:
805 * Multicast NS MUST add one add the option
806 * Unicast NS SHOULD add one add the option
807 *
808 * XXX We deviate from RFC 4429 and still use optimistic DAD as source
809 * for address resolution. However to ensure that we do not interfere
810 * with neighbor cache entries of other neighbors, we MUST ensure
811 * that SLLAO is not sent. Also note, sending multicast NS without SLLAO
812 * is also a deviation from RFC 4861.
813 */
814 if (nonce == NULL && (mac = nd6_ifptomac(ifp)) && !is_optimistic) {
815 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
816 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
817 /* 8 byte alignments... */
818 optlen = (optlen + 7) & ~7;
819
820 m->m_pkthdr.len += optlen;
821 m->m_len += optlen;
822 icmp6len += optlen;
823 bzero((caddr_t)nd_opt, optlen);
824 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
825 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
826 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
827 }
828 /*
829 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
830 * This behavior is documented as Enhanced Duplicate Address
831 * Detection in draft-ietf-6man-enhanced-dad-13.
832 * net.inet6.ip6.dad_enhanced=0 disables this.
833 */
834 if (dad_enhanced != 0 && nonce != NULL && !(ifp->if_flags & IFF_POINTOPOINT)) {
835 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
836 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
837 /* 8-byte alignment is required. */
838 optlen = (optlen + 7) & ~7;
839
840 m->m_pkthdr.len += optlen;
841 m->m_len += optlen;
842 icmp6len += optlen;
843 bzero((caddr_t)nd_opt, optlen);
844 nd_opt->nd_opt_type = ND_OPT_NONCE;
845 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
846 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
847 }
848 ip6->ip6_plen = htons((u_short)icmp6len);
849 nd_ns->nd_ns_cksum = 0;
850 nd_ns->nd_ns_cksum
851 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
852
853 flags = nonce ? IPV6_UNSPECSRC : 0;
854 flags |= IPV6_OUTARGS;
855
856 /*
857 * PKTF_{INET,INET6}_RESOLVE_RTR are mutually exclusive, so make
858 * sure only one of them is set (just in case.)
859 */
860 m->m_pkthdr.pkt_flags &= ~(PKTF_INET_RESOLVE | PKTF_RESOLVE_RTR);
861 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
862 /*
863 * If this is a NS for resolving the (default) router, mark
864 * the packet accordingly so that the driver can find out,
865 * in case it needs to perform driver-specific action(s).
866 */
867 if (rtflags & RTF_ROUTER) {
868 m->m_pkthdr.pkt_flags |= PKTF_RESOLVE_RTR;
869 }
870
871 if (ifp->if_eflags & IFEF_TXSTART) {
872 /*
873 * Use control service class if the interface
874 * supports transmit-start model
875 */
876 (void) m_set_service_class(m, MBUF_SC_CTL);
877 }
878
879 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
880 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
881 ip6_output(m, NULL, NULL, flags, im6o, &outif, &ip6oa);
882 if (outif) {
883 icmp6_ifstat_inc(outif, ifs6_out_msg);
884 icmp6_ifstat_inc(outif, ifs6_out_neighborsolicit);
885 ifnet_release(outif);
886 }
887 icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
888
889 exit:
890 if (im6o != NULL) {
891 IM6O_REMREF(im6o);
892 }
893
894 ROUTE_RELEASE(&ro); /* we don't cache this route. */
895
896 if (ia != NULL) {
897 IFA_REMREF(&ia->ia_ifa);
898 }
899 return;
900
901 bad:
902 m_freem(m);
903 goto exit;
904 }
905
906 /*
907 * Neighbor advertisement input handling.
908 *
909 * Based on RFC 4861
910 * Based on RFC 4862 (duplicate address detection)
911 *
912 * the following items are not implemented yet:
913 * - anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD)
914 * - proxy advertisement delay rule (RFC 4861 7.2.8, last paragraph, "should")
915 */
916 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)917 nd6_na_input(struct mbuf *m, int off, int icmp6len)
918 {
919 struct ifnet *ifp = m->m_pkthdr.rcvif;
920 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
921 struct nd_neighbor_advert *nd_na;
922 struct in6_addr saddr6 = ip6->ip6_src;
923 struct in6_addr daddr6 = ip6->ip6_dst;
924 struct in6_addr taddr6;
925 int flags;
926 int is_router;
927 int is_solicited;
928 int is_override;
929 char *lladdr = NULL;
930 int lladdrlen = 0;
931 struct llinfo_nd6 *ln;
932 struct rtentry *rt;
933 struct sockaddr_dl *sdl;
934 union nd_opts ndopts;
935 uint64_t timenow;
936 bool send_nc_alive_kev = false;
937
938 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
939 nd6log(info, "nd6_na_input: on ND6ALT interface!\n");
940 goto freeit;
941 }
942
943 /* Expect 32-bit aligned data pointer on strict-align platforms */
944 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
945
946 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
947 nd6log(error,
948 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
949 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
950 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
951 goto bad;
952 }
953
954 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
955 ip6 = mtod(m, struct ip6_hdr *);
956 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
957 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
958
959 flags = nd_na->nd_na_flags_reserved;
960 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
961 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
962 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
963
964 taddr6 = nd_na->nd_na_target;
965 if (in6_setscope(&taddr6, ifp, NULL)) {
966 goto bad; /* XXX: impossible */
967 }
968 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
969 nd6log(error,
970 "nd6_na_input: invalid target address %s\n",
971 ip6_sprintf(&taddr6));
972 goto bad;
973 }
974 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
975 if (is_solicited) {
976 nd6log(error,
977 "nd6_na_input: a solicited adv is multicasted\n");
978 goto bad;
979 }
980 }
981
982 icmp6len -= sizeof(*nd_na);
983 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
984 if (nd6_options(&ndopts) < 0) {
985 nd6log(info,
986 "nd6_na_input: invalid ND option, ignored\n");
987 /* nd6_options have incremented stats */
988 goto freeit;
989 }
990
991 if (ndopts.nd_opts_tgt_lladdr) {
992 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
993 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
994
995 if (((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
996 nd6log(info,
997 "nd6_na_input: lladdrlen mismatch for %s "
998 "(if %d, NA packet %d)\n",
999 ip6_sprintf(&taddr6), ifp->if_addrlen,
1000 lladdrlen - 2);
1001 goto bad;
1002 }
1003 }
1004
1005 m = nd6_dad_na_input(m, ifp, &taddr6, lladdr, lladdrlen);
1006 if (m == NULL) {
1007 return;
1008 }
1009
1010 /* Forwarding associated with NDPRF_PRPROXY may apply. */
1011 if (ip6_forwarding && nd6_prproxy) {
1012 nd6_prproxy_na_input(ifp, &saddr6, &daddr6, &taddr6, flags);
1013 }
1014
1015 /*
1016 * If no neighbor cache entry is found, NA SHOULD silently be
1017 * discarded. If we are forwarding (and Scoped Routing is in
1018 * effect), try to see if there is a neighbor cache entry on
1019 * another interface (in case we are doing prefix proxying.)
1020 */
1021 if ((rt = nd6_lookup(&taddr6, 0, ifp, 0)) == NULL) {
1022 if (!ip6_forwarding || !nd6_prproxy) {
1023 goto freeit;
1024 }
1025
1026 if ((rt = nd6_lookup(&taddr6, 0, NULL, 0)) == NULL) {
1027 goto freeit;
1028 }
1029
1030 RT_LOCK_ASSERT_HELD(rt);
1031 if (rt->rt_ifp != ifp) {
1032 /*
1033 * Purge any link-layer info caching.
1034 */
1035 if (rt->rt_llinfo_purge != NULL) {
1036 rt->rt_llinfo_purge(rt);
1037 }
1038
1039 /* Adjust route ref count for the interfaces */
1040 if (rt->rt_if_ref_fn != NULL) {
1041 rt->rt_if_ref_fn(ifp, 1);
1042 rt->rt_if_ref_fn(rt->rt_ifp, -1);
1043 }
1044
1045 /* Change the interface when the existing route is on */
1046 rt->rt_ifp = ifp;
1047
1048 /*
1049 * If rmx_mtu is not locked, update it
1050 * to the MTU used by the new interface.
1051 */
1052 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1053 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1054 }
1055 }
1056 }
1057
1058 RT_LOCK_ASSERT_HELD(rt);
1059 if ((ln = rt->rt_llinfo) == NULL ||
1060 (sdl = SDL(rt->rt_gateway)) == NULL) {
1061 RT_REMREF_LOCKED(rt);
1062 RT_UNLOCK(rt);
1063 goto freeit;
1064 }
1065
1066 timenow = net_uptime();
1067
1068 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1069 /*
1070 * If the link-layer has address, and no lladdr option came,
1071 * discard the packet.
1072 */
1073 if (ifp->if_addrlen && !lladdr) {
1074 RT_REMREF_LOCKED(rt);
1075 RT_UNLOCK(rt);
1076 goto freeit;
1077 }
1078
1079 /*
1080 * Record link-layer address, and update the state.
1081 */
1082 sdl->sdl_alen = ifp->if_addrlen;
1083 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1084 if (is_solicited) {
1085 send_nc_alive_kev = (rt->rt_flags & RTF_ROUTER) ? true : false;
1086 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1087 if (ln->ln_expire != 0) {
1088 struct nd_ifinfo *ndi = NULL;
1089
1090 ndi = ND_IFINFO(rt->rt_ifp);
1091 VERIFY(ndi != NULL && ndi->initialized);
1092 lck_mtx_lock(&ndi->lock);
1093 ln_setexpire(ln, timenow + ndi->reachable);
1094 lck_mtx_unlock(&ndi->lock);
1095 RT_UNLOCK(rt);
1096 lck_mtx_lock(rnh_lock);
1097 nd6_sched_timeout(NULL, NULL);
1098 lck_mtx_unlock(rnh_lock);
1099 RT_LOCK(rt);
1100 }
1101 } else {
1102 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1103 ln_setexpire(ln, timenow + nd6_gctimer);
1104 }
1105
1106 /*
1107 * Enqueue work item to invoke callback for this
1108 * route entry
1109 */
1110 route_event_enqueue_nwk_wq_entry(rt, NULL,
1111 ROUTE_LLENTRY_RESOLVED, NULL, TRUE);
1112
1113 if ((ln->ln_router = (short)is_router) != 0) {
1114 struct radix_node_head *rnh = NULL;
1115 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1116 struct ifnet *rt_ifp = rt->rt_ifp;
1117
1118 struct route_event rt_ev;
1119 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_RESOLVED);
1120 /*
1121 * This means a router's state has changed from
1122 * non-reachable to probably reachable, and might
1123 * affect the status of associated prefixes..
1124 * We already have a reference on rt. Don't need to
1125 * take one for the unlock/lock.
1126 */
1127 RT_UNLOCK(rt);
1128 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1129 lck_mtx_lock(rnh_lock);
1130 rnh = rt_tables[AF_INET6];
1131
1132 if (rnh != NULL) {
1133 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1134 (void *)&rt_ev);
1135 }
1136 lck_mtx_unlock(rnh_lock);
1137 lck_mtx_lock(nd6_mutex);
1138 pfxlist_onlink_check();
1139 lck_mtx_unlock(nd6_mutex);
1140 RT_LOCK(rt);
1141 }
1142 } else {
1143 int llchange = 0;
1144
1145 /*
1146 * Check if the link-layer address has changed or not.
1147 */
1148 if (lladdr == NULL) {
1149 llchange = 0;
1150 } else {
1151 if (sdl->sdl_alen) {
1152 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) {
1153 llchange = 1;
1154 } else {
1155 llchange = 0;
1156 }
1157 } else {
1158 llchange = 1;
1159 }
1160 }
1161
1162 /*
1163 * This is VERY complex. Look at it with care.
1164 *
1165 * override solicit lladdr llchange action
1166 * (L: record lladdr)
1167 *
1168 * 0 0 n -- (2c)
1169 * 0 0 y n (2b) L
1170 * 0 0 y y (1) REACHABLE->STALE
1171 * 0 1 n -- (2c) *->REACHABLE
1172 * 0 1 y n (2b) L *->REACHABLE
1173 * 0 1 y y (1) REACHABLE->STALE
1174 * 1 0 n -- (2a)
1175 * 1 0 y n (2a) L
1176 * 1 0 y y (2a) L *->STALE
1177 * 1 1 n -- (2a) *->REACHABLE
1178 * 1 1 y n (2a) L *->REACHABLE
1179 * 1 1 y y (2a) L *->REACHABLE
1180 */
1181 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
1182 /*
1183 * If state is REACHABLE, make it STALE.
1184 * no other updates should be done.
1185 */
1186 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
1187 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1188 ln_setexpire(ln, timenow + nd6_gctimer);
1189 }
1190 RT_REMREF_LOCKED(rt);
1191 RT_UNLOCK(rt);
1192 goto freeit;
1193 } else if (is_override /* (2a) */
1194 || (!is_override && (lladdr && !llchange)) /* (2b) */
1195 || !lladdr) { /* (2c) */
1196 /*
1197 * Update link-local address, if any.
1198 */
1199 if (lladdr) {
1200 sdl->sdl_alen = ifp->if_addrlen;
1201 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1202 }
1203
1204 /*
1205 * If solicited, make the state REACHABLE.
1206 * If not solicited and the link-layer address was
1207 * changed, make it STALE.
1208 */
1209 if (is_solicited) {
1210 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1211 if (ln->ln_expire != 0) {
1212 struct nd_ifinfo *ndi = NULL;
1213
1214 ndi = ND_IFINFO(ifp);
1215 VERIFY(ndi != NULL && ndi->initialized);
1216 lck_mtx_lock(&ndi->lock);
1217 ln_setexpire(ln,
1218 timenow + ndi->reachable);
1219 lck_mtx_unlock(&ndi->lock);
1220 RT_UNLOCK(rt);
1221 lck_mtx_lock(rnh_lock);
1222 nd6_sched_timeout(NULL, NULL);
1223 lck_mtx_unlock(rnh_lock);
1224 RT_LOCK(rt);
1225 }
1226 } else {
1227 if (lladdr && llchange) {
1228 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1229 ln_setexpire(ln, timenow + nd6_gctimer);
1230 }
1231 }
1232
1233 /*
1234 * XXX
1235 * The above is somewhat convoluted, for now just
1236 * issue a callback for LLENTRY changed.
1237 */
1238 /* Enqueue work item to invoke callback for this route entry */
1239 if (llchange) {
1240 route_event_enqueue_nwk_wq_entry(rt, NULL,
1241 ROUTE_LLENTRY_CHANGED, NULL, TRUE);
1242 }
1243
1244 /*
1245 * If the router's link-layer address has changed,
1246 * notify routes using this as gateway so they can
1247 * update any cached information.
1248 */
1249 if (ln->ln_router && is_router && llchange) {
1250 struct radix_node_head *rnh = NULL;
1251 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1252 struct ifnet *rt_ifp = rt->rt_ifp;
1253 struct route_event rt_ev;
1254 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_CHANGED);
1255
1256 /*
1257 * This means a router's state has changed from
1258 * non-reachable to probably reachable, and might
1259 * affect the status of associated prefixes..
1260 *
1261 * We already have a valid rt reference here.
1262 * We don't need to take another one for unlock/lock.
1263 */
1264 RT_UNLOCK(rt);
1265 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1266 lck_mtx_lock(rnh_lock);
1267 rnh = rt_tables[AF_INET6];
1268
1269 if (rnh != NULL) {
1270 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1271 (void *)&rt_ev);
1272 }
1273 lck_mtx_unlock(rnh_lock);
1274 RT_LOCK(rt);
1275 }
1276 }
1277
1278 if (ln->ln_router && !is_router) {
1279 /*
1280 * The peer dropped the router flag.
1281 * Remove the sender from the Default Router List and
1282 * update the Destination Cache entries.
1283 */
1284 struct nd_defrouter *dr;
1285 struct in6_addr *in6;
1286 struct ifnet *rt_ifp = rt->rt_ifp;
1287
1288 in6 = &((struct sockaddr_in6 *)
1289 (void *)rt_key(rt))->sin6_addr;
1290
1291 RT_UNLOCK(rt);
1292 lck_mtx_lock(nd6_mutex);
1293 /*
1294 * XXX Handle router lists for route information option
1295 * as well.
1296 */
1297 dr = defrouter_lookup(NULL, in6, rt_ifp);
1298 if (dr) {
1299 TAILQ_REMOVE(&nd_defrouter_list, dr, dr_entry);
1300 defrtrlist_del(dr, NULL);
1301 NDDR_REMREF(dr); /* remove list reference */
1302 NDDR_REMREF(dr);
1303 lck_mtx_unlock(nd6_mutex);
1304 } else {
1305 lck_mtx_unlock(nd6_mutex);
1306 /*
1307 * Even if the neighbor is not in the
1308 * default router list, the neighbor
1309 * may be used as a next hop for some
1310 * destinations (e.g. redirect case).
1311 * So we must call rt6_flush explicitly.
1312 */
1313 rt6_flush(&ip6->ip6_src, rt_ifp);
1314 }
1315 RT_LOCK(rt);
1316 }
1317 ln->ln_router = (short)is_router;
1318 }
1319
1320 if (send_nc_alive_kev && (ifp->if_addrlen == IF_LLREACH_MAXLEN)) {
1321 struct kev_msg ev_msg;
1322 struct kev_nd6_ndalive nd6_ndalive;
1323 bzero(&ev_msg, sizeof(ev_msg));
1324 bzero(&nd6_ndalive, sizeof(nd6_ndalive));
1325 ev_msg.vendor_code = KEV_VENDOR_APPLE;
1326 ev_msg.kev_class = KEV_NETWORK_CLASS;
1327 ev_msg.kev_subclass = KEV_ND6_SUBCLASS;
1328 ev_msg.event_code = KEV_ND6_NDALIVE;
1329
1330 nd6_ndalive.link_data.if_family = ifp->if_family;
1331 nd6_ndalive.link_data.if_unit = ifp->if_unit;
1332 strlcpy(nd6_ndalive.link_data.if_name,
1333 ifp->if_name,
1334 sizeof(nd6_ndalive.link_data.if_name));
1335 ev_msg.dv[0].data_ptr = &nd6_ndalive;
1336 ev_msg.dv[0].data_length =
1337 sizeof(nd6_ndalive);
1338 dlil_post_complete_msg(NULL, &ev_msg);
1339 }
1340
1341 RT_LOCK_ASSERT_HELD(rt);
1342 rt->rt_flags &= ~RTF_REJECT;
1343
1344 /* cache the gateway (sender HW) address */
1345 nd6_llreach_alloc(rt, ifp, LLADDR(sdl), sdl->sdl_alen, TRUE);
1346
1347 /* update the llinfo, send a queued packet if there is one */
1348 ln->ln_asked = 0;
1349 if (ln->ln_hold != NULL) {
1350 struct mbuf *m_hold, *m_hold_next;
1351 struct sockaddr_in6 sin6;
1352
1353 rtkey_to_sa6(rt, &sin6);
1354 /*
1355 * reset the ln_hold in advance, to explicitly
1356 * prevent a ln_hold lookup in nd6_output()
1357 * (wouldn't happen, though...)
1358 */
1359 m_hold = ln->ln_hold;
1360 ln->ln_hold = NULL;
1361 for (; m_hold; m_hold = m_hold_next) {
1362 m_hold_next = m_hold->m_nextpkt;
1363 m_hold->m_nextpkt = NULL;
1364 /*
1365 * we assume ifp is not a loopback here, so just set
1366 * the 2nd argument as the 1st one.
1367 */
1368 RT_UNLOCK(rt);
1369 nd6_output(ifp, ifp, m_hold, &sin6, rt, NULL);
1370 RT_LOCK_SPIN(rt);
1371 }
1372 }
1373 RT_REMREF_LOCKED(rt);
1374 RT_UNLOCK(rt);
1375 m_freem(m);
1376 return;
1377
1378 bad:
1379 icmp6stat.icp6s_badna++;
1380 /* fall through */
1381 freeit:
1382 m_freem(m);
1383 return;
1384 }
1385
1386 /*
1387 * Neighbor advertisement output handling.
1388 *
1389 * Based on RFC 2461
1390 *
1391 * the following items are not implemented yet:
1392 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
1393 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
1394 *
1395 * tlladdr - 1 if include target link-layer address
1396 * sdl0 - sockaddr_dl (= proxy NA) or NULL
1397 */
1398 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)1399 nd6_na_output(
1400 struct ifnet *ifp,
1401 const struct in6_addr *daddr6_0,
1402 const struct in6_addr *taddr6,
1403 uint32_t flags,
1404 int tlladdr, /* 1 if include target link-layer address */
1405 struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
1406 {
1407 struct mbuf *m;
1408 struct ip6_hdr *ip6;
1409 struct nd_neighbor_advert *nd_na;
1410 struct ip6_moptions *im6o = NULL;
1411 caddr_t mac = NULL;
1412 struct route_in6 ro;
1413 struct in6_addr *src, src_storage, daddr6;
1414 struct in6_ifaddr *ia;
1415 struct sockaddr_in6 dst_sa;
1416 int icmp6len, maxlen, error;
1417 struct ifnet *outif = NULL;
1418
1419 struct ip6_out_args ip6oa;
1420 bzero(&ro, sizeof(ro));
1421
1422 daddr6 = *daddr6_0; /* make a local copy for modification */
1423
1424 bzero(&ip6oa, sizeof(ip6oa));
1425 ip6oa.ip6oa_boundif = ifp->if_index;
1426 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
1427 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
1428 IP6OAF_MANAGEMENT_ALLOWED;
1429 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
1430 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1431
1432 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
1433
1434 /* estimate the size of message */
1435 maxlen = sizeof(*ip6) + sizeof(*nd_na);
1436 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
1437 if (max_linkhdr + maxlen >= MCLBYTES) {
1438 #if DIAGNOSTIC
1439 printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
1440 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
1441 #endif
1442 return;
1443 }
1444
1445 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
1446 if (m && max_linkhdr + maxlen >= MHLEN) {
1447 MCLGET(m, M_DONTWAIT);
1448 if ((m->m_flags & M_EXT) == 0) {
1449 m_free(m);
1450 m = NULL;
1451 }
1452 }
1453 if (m == NULL) {
1454 return;
1455 }
1456 m->m_pkthdr.rcvif = NULL;
1457
1458 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1459 m->m_flags |= M_MCAST;
1460
1461 im6o = ip6_allocmoptions(Z_NOWAIT);
1462 if (im6o == NULL) {
1463 m_freem(m);
1464 return;
1465 }
1466
1467 im6o->im6o_multicast_ifp = ifp;
1468 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
1469 im6o->im6o_multicast_loop = 0;
1470 }
1471
1472 icmp6len = sizeof(*nd_na);
1473 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1474 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
1475
1476 /* fill neighbor advertisement packet */
1477 ip6 = mtod(m, struct ip6_hdr *);
1478 ip6->ip6_flow = 0;
1479 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1480 ip6->ip6_vfc |= IPV6_VERSION;
1481 ip6->ip6_nxt = IPPROTO_ICMPV6;
1482 ip6->ip6_hlim = IPV6_MAXHLIM;
1483 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1484 /* reply to DAD */
1485 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1486 daddr6.s6_addr16[1] = 0;
1487 daddr6.s6_addr32[1] = 0;
1488 daddr6.s6_addr32[2] = 0;
1489 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1490 if (in6_setscope(&daddr6, ifp, NULL)) {
1491 goto bad;
1492 }
1493
1494 flags &= ~ND_NA_FLAG_SOLICITED;
1495 } else {
1496 ip6->ip6_dst = daddr6;
1497 ip6_output_setdstifscope(m, ifp->if_index, NULL);
1498 }
1499
1500 bzero(&dst_sa, sizeof(struct sockaddr_in6));
1501 dst_sa.sin6_family = AF_INET6;
1502 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1503 dst_sa.sin6_addr = daddr6;
1504
1505 /*
1506 * Select a source whose scope is the same as that of the dest.
1507 */
1508 bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1509 src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &src_storage,
1510 ip6oa.ip6oa_boundif, &error);
1511 if (src == NULL) {
1512 nd6log(debug, "nd6_na_output: source can't be "
1513 "determined: dst=%s, error=%d\n",
1514 ip6_sprintf(&dst_sa.sin6_addr), error);
1515 goto bad;
1516 }
1517 ip6->ip6_src = *src;
1518
1519 /*
1520 * RFC 4429 requires not setting "override" flag on NA packets sent
1521 * from optimistic addresses.
1522 */
1523 ia = in6ifa_ifpwithaddr(ifp, src);
1524 ip6_output_setsrcifscope(m, ifp->if_index, ia);
1525 if (ia != NULL) {
1526 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
1527 flags &= ~ND_NA_FLAG_OVERRIDE;
1528 }
1529 IFA_REMREF(&ia->ia_ifa);
1530 }
1531
1532 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1533 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1534 nd_na->nd_na_code = 0;
1535 nd_na->nd_na_target = *taddr6;
1536 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1537
1538 /*
1539 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1540 * see nd6_ns_input() for details.
1541 * Basically, if NS packet is sent to unicast/anycast addr,
1542 * target lladdr option SHOULD NOT be included.
1543 */
1544 if (tlladdr) {
1545 /*
1546 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1547 * lladdr in sdl0. If we are not proxying (sending NA for
1548 * my address) use lladdr configured for the interface.
1549 */
1550 if (sdl0 == NULL) {
1551 mac = nd6_ifptomac(ifp);
1552 } else if (sdl0->sa_family == AF_LINK) {
1553 struct sockaddr_dl *sdl;
1554 sdl = (struct sockaddr_dl *)(void *)sdl0;
1555 if (sdl->sdl_alen == ifp->if_addrlen) {
1556 mac = LLADDR(sdl);
1557 }
1558 }
1559 }
1560 if (tlladdr && mac) {
1561 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1562 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1563
1564 /* roundup to 8 bytes alignment! */
1565 optlen = (optlen + 7) & ~7;
1566
1567 m->m_pkthdr.len += optlen;
1568 m->m_len += optlen;
1569 icmp6len += optlen;
1570 bzero((caddr_t)nd_opt, optlen);
1571 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1572 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
1573 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1574 } else {
1575 flags &= ~ND_NA_FLAG_OVERRIDE;
1576 }
1577
1578 ip6->ip6_plen = htons((u_short)icmp6len);
1579 nd_na->nd_na_flags_reserved = flags;
1580 nd_na->nd_na_cksum = 0;
1581 nd_na->nd_na_cksum =
1582 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1583
1584 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
1585
1586 if (ifp->if_eflags & IFEF_TXSTART) {
1587 /* Use control service class if the interface supports
1588 * transmit-start model.
1589 */
1590 (void) m_set_service_class(m, MBUF_SC_CTL);
1591 }
1592
1593 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
1594 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
1595 ip6_output(m, NULL, NULL, IPV6_OUTARGS, im6o, &outif, &ip6oa);
1596 if (outif) {
1597 icmp6_ifstat_inc(outif, ifs6_out_msg);
1598 icmp6_ifstat_inc(outif, ifs6_out_neighboradvert);
1599 ifnet_release(outif);
1600 }
1601 icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1602
1603 exit:
1604 if (im6o != NULL) {
1605 IM6O_REMREF(im6o);
1606 }
1607
1608 ROUTE_RELEASE(&ro);
1609 return;
1610
1611 bad:
1612 m_freem(m);
1613 goto exit;
1614 }
1615
1616 caddr_t
nd6_ifptomac(struct ifnet * ifp)1617 nd6_ifptomac(
1618 struct ifnet *ifp)
1619 {
1620 switch (ifp->if_type) {
1621 case IFT_ARCNET:
1622 case IFT_ETHER:
1623 case IFT_IEEE8023ADLAG:
1624 case IFT_FDDI:
1625 case IFT_IEEE1394:
1626 #ifdef IFT_L2VLAN
1627 case IFT_L2VLAN:
1628 #endif
1629 #ifdef IFT_IEEE80211
1630 case IFT_IEEE80211:
1631 #endif
1632 #ifdef IFT_CARP
1633 case IFT_CARP:
1634 #endif
1635 case IFT_BRIDGE:
1636 case IFT_ISO88025:
1637 return (caddr_t)IF_LLADDR(ifp);
1638 default:
1639 return NULL;
1640 }
1641 }
1642
1643 TAILQ_HEAD(dadq_head, dadq);
1644 struct dadq {
1645 decl_lck_mtx_data(, dad_lock);
1646 u_int32_t dad_refcount; /* reference count */
1647 int dad_attached;
1648 TAILQ_ENTRY(dadq) dad_list;
1649 struct ifaddr *dad_ifa;
1650 int dad_count; /* max NS to send */
1651 int dad_ns_tcount; /* # of trials to send NS */
1652 int dad_ns_ocount; /* NS sent so far */
1653 int dad_ns_icount;
1654 int dad_na_icount;
1655 int dad_ns_lcount; /* looped back NS */
1656 int dad_loopbackprobe; /* probing state for loopback detection */
1657 uint8_t dad_lladdr[ETHER_ADDR_LEN];
1658 uint8_t dad_lladdrlen;
1659 #define ND_OPT_NONCE_LEN32 \
1660 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1661 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1662 uint32_t dad_same_nonce_count; /* # of consecutive times we've ignored DAD failure because of optimistic DAD */
1663 };
1664
1665 static KALLOC_TYPE_DEFINE(dad_zone, struct dadq, NET_KT_DEFAULT);
1666 static struct dadq_head dadq;
1667
1668 void
nd6_nbr_init(void)1669 nd6_nbr_init(void)
1670 {
1671 int i;
1672
1673 TAILQ_INIT(&dadq);
1674
1675 bzero(&hostrtmask, sizeof hostrtmask);
1676 hostrtmask.sin6_family = AF_INET6;
1677 hostrtmask.sin6_len = sizeof hostrtmask;
1678 for (i = 0; i < sizeof hostrtmask.sin6_addr; ++i) {
1679 hostrtmask.sin6_addr.s6_addr[i] = 0xff;
1680 }
1681 }
1682
1683 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * nonce)1684 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1685 {
1686 struct dadq *dp;
1687 boolean_t same_nonce = false;
1688
1689 lck_mtx_lock(&dad6_mutex);
1690 for (dp = dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1691 DAD_LOCK_SPIN(dp);
1692 if (dp->dad_ifa != ifa) {
1693 DAD_UNLOCK(dp);
1694 continue;
1695 }
1696
1697 /*
1698 * Skip if the nonce matches the received one.
1699 * +2 in the length is required because of type and
1700 * length fields are included in a header.
1701 */
1702 same_nonce = nonce != NULL &&
1703 nonce->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1704 memcmp(&nonce->nd_opt_nonce[0], &dp->dad_nonce[0],
1705 ND_OPT_NONCE_LEN) == 0;
1706
1707 if (same_nonce &&
1708 dp->dad_same_nonce_count <= nd6_dad_nonce_max_count) {
1709 nd6log(error, "%s: a looped back NS message is "
1710 "detected during DAD for if=%s %s. Ignoring.\n",
1711 __func__,
1712 if_name(ifa->ifa_ifp),
1713 ip6_sprintf(IFA_IN6(ifa)));
1714 dp->dad_same_nonce_count++;
1715 dp->dad_ns_lcount++;
1716 ++ip6stat.ip6s_dad_loopcount;
1717 DAD_UNLOCK(dp);
1718 continue;
1719 } else if (!same_nonce) {
1720 // Not the same nonce, reset counter
1721 dp->dad_same_nonce_count = 1;
1722 }
1723
1724 DAD_ADDREF_LOCKED(dp);
1725 DAD_UNLOCK(dp);
1726 break;
1727 }
1728 lck_mtx_unlock(&dad6_mutex);
1729 return dp;
1730 }
1731
1732 void
nd6_dad_stoptimer(struct ifaddr * ifa)1733 nd6_dad_stoptimer(
1734 struct ifaddr *ifa)
1735 {
1736 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1737 }
1738
1739 /*
1740 * Start Duplicate Address Detection (DAD) for specified interface address.
1741 */
1742 void
nd6_dad_start(struct ifaddr * ifa,int * tick_delay)1743 nd6_dad_start(
1744 struct ifaddr *ifa,
1745 int *tick_delay) /* minimum delay ticks for IFF_UP event */
1746 {
1747 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1748 struct dadq *dp;
1749
1750 if (ifa->ifa_ifp == NULL) {
1751 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1752 }
1753
1754 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1755 __func__,
1756 ip6_sprintf(&ia->ia_addr.sin6_addr),
1757 if_name(ia->ia_ifp),
1758 ia->ia6_flags);
1759
1760 /*
1761 * If we don't need DAD, don't do it.
1762 * There are several cases:
1763 * - DAD is disabled (ip6_dad_count == 0)
1764 * - the interface address is anycast
1765 */
1766 IFA_LOCK(&ia->ia_ifa);
1767 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
1768 nd6log0(debug,
1769 "nd6_dad_start: not a tentative or optimistic address "
1770 "%s(%s)\n",
1771 ip6_sprintf(&ia->ia_addr.sin6_addr),
1772 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1773 IFA_UNLOCK(&ia->ia_ifa);
1774 return;
1775 }
1776 if (!ip6_dad_count || (ia->ia6_flags & IN6_IFF_ANYCAST) != 0) {
1777 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
1778 IFA_UNLOCK(&ia->ia_ifa);
1779 return;
1780 }
1781 IFA_UNLOCK(&ia->ia_ifa);
1782
1783 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1784 (ifa->ifa_ifp->if_eflags & IFEF_IPV6_ND6ALT)) {
1785 return;
1786 }
1787 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1788 DAD_REMREF(dp);
1789 /* DAD already in progress */
1790 return;
1791 }
1792
1793 dp = zalloc_flags(dad_zone, Z_WAITOK | Z_ZERO);
1794 lck_mtx_init(&dp->dad_lock, &ifa_mtx_grp, &ifa_mtx_attr);
1795
1796 /* Callee adds one reference for us */
1797 dp = nd6_dad_attach(dp, ifa);
1798
1799 nd6log0(debug, "%s: starting %sDAD %sfor %s\n",
1800 if_name(ifa->ifa_ifp),
1801 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) ? "optimistic " : "",
1802 (tick_delay == NULL) ? "immediately " : "",
1803 ip6_sprintf(&ia->ia_addr.sin6_addr));
1804
1805 /*
1806 * Send NS packet for DAD, ip6_dad_count times.
1807 * Note that we must delay the first transmission, if this is the
1808 * first packet to be sent from the interface after interface
1809 * (re)initialization.
1810 */
1811 if (tick_delay == NULL) {
1812 u_int32_t retrans;
1813 struct nd_ifinfo *ndi = NULL;
1814
1815 nd6_dad_ns_output(dp, ifa);
1816 ndi = ND_IFINFO(ifa->ifa_ifp);
1817 VERIFY(ndi != NULL && ndi->initialized);
1818 lck_mtx_lock(&ndi->lock);
1819 retrans = ndi->retrans * hz / 1000;
1820 lck_mtx_unlock(&ndi->lock);
1821 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
1822 } else {
1823 int ntick;
1824
1825 if (*tick_delay == 0) {
1826 ntick = random() % (MAX_RTR_SOLICITATION_DELAY * hz);
1827 } else {
1828 ntick = *tick_delay + random() % (hz / 2);
1829 }
1830 *tick_delay = ntick;
1831 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa,
1832 ntick);
1833 }
1834
1835 DAD_REMREF(dp); /* drop our reference */
1836 }
1837
1838 static struct dadq *
nd6_dad_attach(struct dadq * dp,struct ifaddr * ifa)1839 nd6_dad_attach(struct dadq *dp, struct ifaddr *ifa)
1840 {
1841 lck_mtx_lock(&dad6_mutex);
1842 DAD_LOCK(dp);
1843 dp->dad_ifa = ifa;
1844 IFA_ADDREF(ifa); /* for dad_ifa */
1845 dp->dad_count = ip6_dad_count;
1846 dp->dad_ns_icount = dp->dad_na_icount = 0;
1847 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1848 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1849 VERIFY(!dp->dad_attached);
1850 dp->dad_same_nonce_count = 1;
1851 dp->dad_attached = 1;
1852 dp->dad_lladdrlen = 0;
1853 DAD_ADDREF_LOCKED(dp); /* for caller */
1854 DAD_ADDREF_LOCKED(dp); /* for dadq_head list */
1855 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1856 DAD_UNLOCK(dp);
1857 lck_mtx_unlock(&dad6_mutex);
1858
1859 return dp;
1860 }
1861
1862 static void
nd6_dad_detach(struct dadq * dp,struct ifaddr * ifa)1863 nd6_dad_detach(struct dadq *dp, struct ifaddr *ifa)
1864 {
1865 int detached;
1866
1867 lck_mtx_lock(&dad6_mutex);
1868 DAD_LOCK(dp);
1869 if ((detached = dp->dad_attached)) {
1870 VERIFY(dp->dad_ifa == ifa);
1871 TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1872 dp->dad_list.tqe_next = NULL;
1873 dp->dad_list.tqe_prev = NULL;
1874 dp->dad_attached = 0;
1875 }
1876 DAD_UNLOCK(dp);
1877 lck_mtx_unlock(&dad6_mutex);
1878 if (detached) {
1879 DAD_REMREF(dp); /* drop dadq_head reference */
1880 }
1881 }
1882
1883 /*
1884 * terminate DAD unconditionally. used for address removals.
1885 */
1886 void
nd6_dad_stop(struct ifaddr * ifa)1887 nd6_dad_stop(struct ifaddr *ifa)
1888 {
1889 struct dadq *dp;
1890
1891 dp = nd6_dad_find(ifa, NULL);
1892 if (!dp) {
1893 /* DAD wasn't started yet */
1894 return;
1895 }
1896
1897 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1898
1899 nd6_dad_detach(dp, ifa);
1900 DAD_REMREF(dp); /* drop our reference */
1901 }
1902
1903 static void
nd6_unsol_na_output(struct ifaddr * ifa)1904 nd6_unsol_na_output(struct ifaddr *ifa)
1905 {
1906 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1907 struct ifnet *ifp = ifa->ifa_ifp;
1908 struct in6_addr saddr6, taddr6;
1909
1910 if ((ifp->if_flags & IFF_UP) == 0 ||
1911 (ifp->if_flags & IFF_RUNNING) == 0 ||
1912 (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
1913 return;
1914 }
1915
1916 IFA_LOCK_SPIN(&ia->ia_ifa);
1917 taddr6 = ia->ia_addr.sin6_addr;
1918 IFA_UNLOCK(&ia->ia_ifa);
1919 if (in6_setscope(&taddr6, ifp, NULL) != 0) {
1920 return;
1921 }
1922 saddr6 = in6addr_linklocal_allnodes;
1923 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
1924 return;
1925 }
1926
1927 nd6log(info, "%s: sending unsolicited NA\n",
1928 if_name(ifa->ifa_ifp));
1929
1930 nd6_na_output(ifp, &saddr6, &taddr6, ND_NA_FLAG_OVERRIDE, 1, NULL);
1931 }
1932
1933 static void
nd6_dad_timer(struct ifaddr * ifa)1934 nd6_dad_timer(struct ifaddr *ifa)
1935 {
1936 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1937 struct dadq *dp = NULL;
1938 struct nd_ifinfo *ndi = NULL;
1939 u_int32_t retrans;
1940
1941 /* Sanity check */
1942 if (ia == NULL) {
1943 nd6log0(error, "nd6_dad_timer: called with null parameter\n");
1944 goto done;
1945 }
1946
1947 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1948 __func__,
1949 ip6_sprintf(&ia->ia_addr.sin6_addr),
1950 if_name(ia->ia_ifp),
1951 ia->ia6_flags);
1952
1953 dp = nd6_dad_find(ifa, NULL);
1954 if (dp == NULL) {
1955 nd6log0(error, "nd6_dad_timer: DAD structure not found\n");
1956 goto done;
1957 }
1958 IFA_LOCK(&ia->ia_ifa);
1959 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1960 nd6log0(error, "nd6_dad_timer: called with duplicated address "
1961 "%s(%s)\n",
1962 ip6_sprintf(&ia->ia_addr.sin6_addr),
1963 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1964 IFA_UNLOCK(&ia->ia_ifa);
1965 goto done;
1966 }
1967 if ((ia->ia6_flags & IN6_IFF_DADPROGRESS) == 0) {
1968 nd6log0(error, "nd6_dad_timer: not a tentative or optimistic "
1969 "address %s(%s)\n",
1970 ip6_sprintf(&ia->ia_addr.sin6_addr),
1971 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1972 IFA_UNLOCK(&ia->ia_ifa);
1973 goto done;
1974 }
1975 IFA_UNLOCK(&ia->ia_ifa);
1976
1977 /* timeouted with IFF_{RUNNING,UP} check */
1978 DAD_LOCK(dp);
1979 if (dp->dad_ns_tcount > dad_maxtry) {
1980 DAD_UNLOCK(dp);
1981 nd6log0(info, "%s: could not run DAD, driver problem?\n",
1982 if_name(ifa->ifa_ifp));
1983
1984 nd6_dad_detach(dp, ifa);
1985 goto done;
1986 }
1987
1988 /* Need more checks? */
1989 if (dp->dad_ns_ocount < dp->dad_count) {
1990 DAD_UNLOCK(dp);
1991 /*
1992 * We have more NS to go. Send NS packet for DAD.
1993 */
1994 nd6_dad_ns_output(dp, ifa);
1995 ndi = ND_IFINFO(ifa->ifa_ifp);
1996 VERIFY(ndi != NULL && ndi->initialized);
1997 lck_mtx_lock(&ndi->lock);
1998 retrans = ndi->retrans * hz / 1000;
1999 lck_mtx_unlock(&ndi->lock);
2000 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2001 } else {
2002 /*
2003 * We have transmitted sufficient number of DAD packets.
2004 * See what we've got.
2005 */
2006 if (dp->dad_na_icount > 0 || dp->dad_ns_icount) {
2007 /* We've seen NS or NA, means DAD has failed. */
2008 DAD_UNLOCK(dp);
2009 nd6log0(info,
2010 "%s: duplicate IPv6 address %s if:%s [timer]\n",
2011 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2012 if_name(ia->ia_ifp));
2013 nd6_dad_duplicated(ifa);
2014 /* (*dp) will be freed in nd6_dad_duplicated() */
2015 #if SKYWALK
2016 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2017 #endif /* SKYWALK */
2018 } else if (dad_enhanced != 0 &&
2019 dp->dad_ns_lcount > 0 &&
2020 dp->dad_ns_lcount > dp->dad_loopbackprobe &&
2021 dp->dad_same_nonce_count > 0 &&
2022 dp->dad_same_nonce_count > nd6_dad_nonce_max_count) {
2023 dp->dad_loopbackprobe = dp->dad_ns_lcount;
2024 dp->dad_count =
2025 dp->dad_ns_ocount + dad_maxtry - 1;
2026 DAD_UNLOCK(dp);
2027 ndi = ND_IFINFO(ifa->ifa_ifp);
2028 VERIFY(ndi != NULL && ndi->initialized);
2029 lck_mtx_lock(&ndi->lock);
2030 retrans = ndi->retrans * hz / 1000;
2031 lck_mtx_unlock(&ndi->lock);
2032
2033 /*
2034 * Sec. 4.1 in RFC 7527 requires transmission of
2035 * additional probes until the loopback condition
2036 * becomes clear when a looped back probe is detected.
2037 */
2038 nd6log0(info,
2039 "%s: a looped back NS message is detected during DAD for %s. Another DAD probe is being sent on interface %s.\n",
2040 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2041 if_name(ia->ia_ifp));
2042 /*
2043 * Send an NS immediately and increase dad_count by
2044 * nd6_mmaxtries - 1.
2045 */
2046 nd6_dad_ns_output(dp, ifa);
2047 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2048 goto done;
2049 } else {
2050 boolean_t txunsolna;
2051 DAD_UNLOCK(dp);
2052 /*
2053 * We are done with DAD. No NA came, no NS came.
2054 * No duplicate address found.
2055 */
2056 IFA_LOCK_SPIN(&ia->ia_ifa);
2057 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2058 IFA_UNLOCK(&ia->ia_ifa);
2059
2060 ndi = ND_IFINFO(ifa->ifa_ifp);
2061 VERIFY(ndi != NULL && ndi->initialized);
2062 lck_mtx_lock(&ndi->lock);
2063 txunsolna = (ndi->flags & ND6_IFF_REPLICATED) != 0;
2064 lck_mtx_unlock(&ndi->lock);
2065
2066 if (txunsolna) {
2067 nd6_unsol_na_output(ifa);
2068 }
2069
2070 nd6log0(debug,
2071 "%s: DAD complete for %s - no duplicates found %s\n",
2072 if_name(ifa->ifa_ifp),
2073 ip6_sprintf(&ia->ia_addr.sin6_addr),
2074 txunsolna ? ", tx unsolicited NA with O=1" : ".");
2075
2076 if (dp->dad_ns_lcount > 0) {
2077 nd6log0(debug,
2078 "%s: DAD completed while "
2079 "a looped back NS message is detected "
2080 "during DAD for %s om interface %s\n",
2081 __func__,
2082 ip6_sprintf(&ia->ia_addr.sin6_addr),
2083 if_name(ia->ia_ifp));
2084 }
2085
2086 in6_post_msg(ia->ia_ifp, KEV_INET6_NEW_USER_ADDR, ia,
2087 dp->dad_lladdr);
2088 nd6_dad_detach(dp, ifa);
2089 #if SKYWALK
2090 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2091 #endif /* SKYWALK */
2092 }
2093 }
2094
2095 done:
2096 if (dp != NULL) {
2097 DAD_REMREF(dp); /* drop our reference */
2098 }
2099 }
2100
2101 static void
nd6_dad_duplicated(struct ifaddr * ifa)2102 nd6_dad_duplicated(struct ifaddr *ifa)
2103 {
2104 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
2105 struct dadq *dp;
2106 struct ifnet *ifp = ifa->ifa_ifp;
2107 boolean_t candisable;
2108
2109 dp = nd6_dad_find(ifa, NULL);
2110 if (dp == NULL) {
2111 log(LOG_ERR, "%s: DAD structure not found.\n", __func__);
2112 return;
2113 }
2114 IFA_LOCK(&ia->ia_ifa);
2115 DAD_LOCK(dp);
2116 nd6log(error, "%s: NS in/out/loopback=%d/%d/%d, NA in=%d\n",
2117 __func__, dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
2118 dp->dad_na_icount);
2119 candisable = FALSE;
2120
2121 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr) &&
2122 !(ia->ia6_flags & IN6_IFF_SECURED)) {
2123 struct in6_addr in6;
2124 struct ifaddr *llifa = NULL;
2125 struct sockaddr_dl *sdl = NULL;
2126 uint8_t *lladdr = dp->dad_lladdr;
2127 uint8_t lladdrlen = dp->dad_lladdrlen;
2128
2129 /*
2130 * To avoid over-reaction, we only apply this logic when we are
2131 * very sure that hardware addresses are supposed to be unique.
2132 */
2133 switch (ifp->if_type) {
2134 case IFT_BRIDGE:
2135 case IFT_ETHER:
2136 case IFT_FDDI:
2137 case IFT_ATM:
2138 case IFT_IEEE1394:
2139 #ifdef IFT_IEEE80211
2140 case IFT_IEEE80211:
2141 #endif
2142 /*
2143 * Check if our hardware address matches the
2144 * link layer information received in the
2145 * NS/NA
2146 */
2147 llifa = ifp->if_lladdr;
2148 IFA_LOCK(llifa);
2149 sdl = (struct sockaddr_dl *)(void *)
2150 llifa->ifa_addr;
2151 if (lladdrlen == sdl->sdl_alen &&
2152 bcmp(lladdr, LLADDR(sdl), lladdrlen) == 0) {
2153 candisable = TRUE;
2154 }
2155 IFA_UNLOCK(llifa);
2156
2157 in6 = ia->ia_addr.sin6_addr;
2158 if (in6_iid_from_hw(ifp, &in6) != 0) {
2159 break;
2160 }
2161
2162 /* Refine decision about whether IPv6 can be disabled */
2163 if (candisable &&
2164 !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2165 /*
2166 * Apply this logic only to the embedded MAC
2167 * address form of link-local IPv6 address.
2168 */
2169 candisable = FALSE;
2170 } else if (lladdr == NULL &&
2171 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2172 /*
2173 * We received a NA with no target link-layer
2174 * address option. This means that someone else
2175 * has our address. Mark it as a hardware
2176 * duplicate so we disable IPv6 later on.
2177 */
2178 candisable = TRUE;
2179 }
2180 break;
2181 default:
2182 break;
2183 }
2184 }
2185 DAD_UNLOCK(dp);
2186
2187 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2188 ia->ia6_flags |= IN6_IFF_DUPLICATED;
2189 in6_event_enqueue_nwk_wq_entry(IN6_ADDR_MARKED_DUPLICATED,
2190 ia->ia_ifa.ifa_ifp, &ia->ia_addr.sin6_addr,
2191 0);
2192 IFA_UNLOCK(&ia->ia_ifa);
2193
2194 /* increment DAD collision counter */
2195 ++ip6stat.ip6s_dad_collide;
2196
2197 /* We are done with DAD, with duplicated address found. (failure) */
2198 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
2199
2200 IFA_LOCK(&ia->ia_ifa);
2201 log(LOG_ERR, "%s: DAD complete for %s - duplicate found.\n",
2202 if_name(ifp), ip6_sprintf(&ia->ia_addr.sin6_addr));
2203 IFA_UNLOCK(&ia->ia_ifa);
2204
2205 if (candisable) {
2206 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
2207 log(LOG_ERR, "%s: possible hardware address duplication "
2208 "detected, disabling IPv6 for interface.\n", if_name(ifp));
2209
2210 VERIFY((NULL != ndi) && (TRUE == ndi->initialized));
2211 ndi->flags |= ND6_IFF_IFDISABLED;
2212 /* Make sure to set IFEF_IPV6_DISABLED too */
2213 nd6_if_disable(ifp, TRUE);
2214 }
2215
2216 log(LOG_ERR,
2217 "%s: manual intervention may be required.\n",
2218 if_name(ifp));
2219
2220 /* Send an event to the configuration agent so that the
2221 * duplicate address will be notified to the user and will
2222 * be removed.
2223 */
2224 in6_post_msg(ifp, KEV_INET6_NEW_USER_ADDR, ia, dp->dad_lladdr);
2225 nd6_dad_detach(dp, ifa);
2226 DAD_REMREF(dp); /* drop our reference */
2227 }
2228
2229 static void
nd6_dad_ns_output(struct dadq * dp,struct ifaddr * ifa)2230 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
2231 {
2232 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
2233 struct ifnet *ifp = ifa->ifa_ifp;
2234 int i = 0;
2235 struct in6_addr taddr6;
2236
2237 DAD_LOCK(dp);
2238 dp->dad_ns_tcount++;
2239 if ((ifp->if_flags & IFF_UP) == 0) {
2240 DAD_UNLOCK(dp);
2241 return;
2242 }
2243 if ((ifp->if_flags & IFF_RUNNING) == 0) {
2244 DAD_UNLOCK(dp);
2245 return;
2246 }
2247
2248 dp->dad_ns_ocount++;
2249 DAD_UNLOCK(dp);
2250 IFA_LOCK_SPIN(&ia->ia_ifa);
2251 taddr6 = ia->ia_addr.sin6_addr;
2252 IFA_UNLOCK(&ia->ia_ifa);
2253 if (dad_enhanced != 0 && !(ifp->if_flags & IFF_POINTOPOINT)) {
2254 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) {
2255 dp->dad_nonce[i] = RandomULong();
2256 }
2257
2258 /*
2259 * XXXHRS: Note that in the case that
2260 * DupAddrDetectTransmits > 1, multiple NS messages with
2261 * different nonces can be looped back in an unexpected
2262 * order. The current implementation recognizes only
2263 * the latest nonce on the sender side. Practically it
2264 * should work well in almost all cases.
2265 */
2266 }
2267 nd6_ns_output(ifp, NULL, &taddr6, NULL,
2268 (uint8_t *)&dp->dad_nonce[0]);
2269 }
2270
2271 /*
2272 * @brief Called to process DAD NS
2273 *
2274 * @param ifa is the pointer to the interface's address
2275 * @param lladdr is source link layer information
2276 * @param lladdrlen is source's linklayer length
2277 *
2278 * @return void
2279 */
2280 static void
nd6_dad_ns_input(struct ifaddr * ifa,char * lladdr,int lladdrlen,struct nd_opt_nonce * ndopt_nonce)2281 nd6_dad_ns_input(struct ifaddr *ifa, char *lladdr,
2282 int lladdrlen, struct nd_opt_nonce *ndopt_nonce)
2283 {
2284 struct dadq *dp;
2285 VERIFY(ifa != NULL);
2286
2287 /* Ignore Nonce option when Enhanced DAD is disabled. */
2288 if (dad_enhanced == 0) {
2289 ndopt_nonce = NULL;
2290 }
2291
2292 dp = nd6_dad_find(ifa, ndopt_nonce);
2293 if (dp == NULL) {
2294 return;
2295 }
2296
2297 DAD_LOCK(dp);
2298 ++dp->dad_ns_icount;
2299 if (lladdr && lladdrlen >= ETHER_ADDR_LEN) {
2300 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2301 /* fine to truncate as it is compared against sdl_alen */
2302 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2303 }
2304 DAD_UNLOCK(dp);
2305 DAD_REMREF(dp);
2306 }
2307
2308 /*
2309 * @brief Called to process received NA for DAD
2310 *
2311 * @param m is the pointer to the packet's mbuf
2312 * @param ifp is the pointer to the interface on which packet
2313 * was receicved.
2314 * @param taddr is pointer to target's IPv6 address
2315 * @param lladdr is target's link layer information
2316 * @param lladdrlen is target's linklayer length
2317 *
2318 * @return NULL if the packet is consumed by DAD processing, else
2319 * pointer to the mbuf.
2320 */
2321 static struct mbuf *
nd6_dad_na_input(struct mbuf * m,struct ifnet * ifp,struct in6_addr * taddr,caddr_t lladdr,int lladdrlen)2322 nd6_dad_na_input(struct mbuf *m, struct ifnet *ifp, struct in6_addr *taddr,
2323 caddr_t lladdr, int lladdrlen)
2324 {
2325 struct ifaddr *ifa = NULL;
2326 struct in6_ifaddr *ia = NULL;
2327 struct dadq *dp = NULL;
2328 struct nd_ifinfo *ndi = NULL;
2329 boolean_t replicated;
2330
2331 ifa = (struct ifaddr *) in6ifa_ifpwithaddr(ifp, taddr);
2332 if (ifa == NULL) {
2333 return m;
2334 }
2335
2336 replicated = FALSE;
2337
2338 /* Get the ND6_IFF_REPLICATED flag. */
2339 ndi = ND_IFINFO(ifp);
2340 if (ndi != NULL && ndi->initialized) {
2341 lck_mtx_lock(&ndi->lock);
2342 replicated = !!(ndi->flags & ND6_IFF_REPLICATED);
2343 lck_mtx_unlock(&ndi->lock);
2344 }
2345
2346 if (replicated) {
2347 nd6log(info, "%s: ignoring duplicate NA on "
2348 "replicated interface %s\n", __func__, if_name(ifp));
2349 goto done;
2350 }
2351
2352 /* Lock the interface address until done (see label below). */
2353 IFA_LOCK(ifa);
2354 ia = (struct in6_ifaddr *) ifa;
2355
2356 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
2357 IFA_UNLOCK(ifa);
2358 nd6log(info, "%s: ignoring duplicate NA on "
2359 "%s [DAD not in progress]\n", __func__,
2360 if_name(ifp));
2361 goto done;
2362 }
2363
2364 /* Some sleep proxies improperly send the client's Ethernet address in
2365 * the target link-layer address option, so detect this by comparing
2366 * the L2-header source address, if we have seen it, with the target
2367 * address, and ignoring the NA if they don't match.
2368 */
2369 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2370 struct ip6aux *ip6a = ip6_findaux(m);
2371 if (ip6a && (ip6a->ip6a_flags & IP6A_HASEEN) != 0 &&
2372 bcmp(ip6a->ip6a_ehsrc, lladdr, ETHER_ADDR_LEN) != 0) {
2373 IFA_UNLOCK(ifa);
2374 nd6log(error, "%s: ignoring duplicate NA on %s "
2375 "[eh_src != tgtlladdr]\n", __func__, if_name(ifp));
2376 goto done;
2377 }
2378 }
2379
2380 IFA_UNLOCK(ifa);
2381
2382 dp = nd6_dad_find(ifa, NULL);
2383 if (dp == NULL) {
2384 nd6log(info, "%s: no DAD structure for %s on %s.\n",
2385 __func__, ip6_sprintf(taddr), if_name(ifp));
2386 goto done;
2387 }
2388
2389 DAD_LOCK_SPIN(dp);
2390 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2391 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2392 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2393 }
2394 dp->dad_na_icount++;
2395 DAD_UNLOCK(dp);
2396 DAD_REMREF(dp);
2397
2398 /* remove the address. */
2399 nd6log(info,
2400 "%s: duplicate IPv6 address %s [processing NA on %s]\n", __func__,
2401 ip6_sprintf(taddr), if_name(ifp));
2402 done:
2403 IFA_LOCK_ASSERT_NOTHELD(ifa);
2404 IFA_REMREF(ifa);
2405 m_freem(m);
2406 return NULL;
2407 }
2408
2409 static void
dad_addref(struct dadq * dp,int locked)2410 dad_addref(struct dadq *dp, int locked)
2411 {
2412 if (!locked) {
2413 DAD_LOCK_SPIN(dp);
2414 } else {
2415 DAD_LOCK_ASSERT_HELD(dp);
2416 }
2417
2418 if (++dp->dad_refcount == 0) {
2419 panic("%s: dad %p wraparound refcnt", __func__, dp);
2420 /* NOTREACHED */
2421 }
2422 if (!locked) {
2423 DAD_UNLOCK(dp);
2424 }
2425 }
2426
2427 static void
dad_remref(struct dadq * dp)2428 dad_remref(struct dadq *dp)
2429 {
2430 struct ifaddr *ifa;
2431
2432 DAD_LOCK_SPIN(dp);
2433 if (dp->dad_refcount == 0) {
2434 panic("%s: dad %p negative refcnt", __func__, dp);
2435 }
2436 --dp->dad_refcount;
2437 if (dp->dad_refcount > 0) {
2438 DAD_UNLOCK(dp);
2439 return;
2440 }
2441 DAD_UNLOCK(dp);
2442
2443 if (dp->dad_attached ||
2444 dp->dad_list.tqe_next != NULL || dp->dad_list.tqe_prev != NULL) {
2445 panic("%s: attached dad=%p is being freed", __func__, dp);
2446 /* NOTREACHED */
2447 }
2448
2449 if ((ifa = dp->dad_ifa) != NULL) {
2450 IFA_REMREF(ifa); /* drop dad_ifa reference */
2451 dp->dad_ifa = NULL;
2452 }
2453
2454 lck_mtx_destroy(&dp->dad_lock, &ifa_mtx_grp);
2455 zfree(dad_zone, dp);
2456 }
2457
2458 void
nd6_llreach_set_reachable(struct ifnet * ifp,void * addr,unsigned int alen)2459 nd6_llreach_set_reachable(struct ifnet *ifp, void *addr, unsigned int alen)
2460 {
2461 /* Nothing more to do if it's disabled */
2462 if (nd6_llreach_base == 0) {
2463 return;
2464 }
2465
2466 ifnet_llreach_set_reachable(ifp, ETHERTYPE_IPV6, addr, alen);
2467 }
2468
2469 void
nd6_alt_node_addr_decompose(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr_dl * sdl,struct sockaddr_in6 * sin6)2470 nd6_alt_node_addr_decompose(struct ifnet *ifp, struct sockaddr *sa,
2471 struct sockaddr_dl* sdl, struct sockaddr_in6 *sin6)
2472 {
2473 static const size_t EUI64_LENGTH = 8;
2474
2475 VERIFY(nd6_need_cache(ifp));
2476 VERIFY(sa);
2477 VERIFY(sdl && (void *)sa != (void *)sdl);
2478 VERIFY(sin6 && (void *)sa != (void *)sin6);
2479
2480 bzero(sin6, sizeof(*sin6));
2481 sin6->sin6_len = sizeof *sin6;
2482 sin6->sin6_family = AF_INET6;
2483
2484 bzero(sdl, sizeof(*sdl));
2485 sdl->sdl_len = sizeof *sdl;
2486 sdl->sdl_family = AF_LINK;
2487 sdl->sdl_type = ifp->if_type;
2488 sdl->sdl_index = ifp->if_index;
2489 sdl->sdl_nlen = 0;
2490
2491 switch (sa->sa_family) {
2492 case AF_INET6: {
2493 struct sockaddr_in6 *sin6a = (struct sockaddr_in6 *)(void *)sa;
2494 struct in6_addr *in6 = &sin6a->sin6_addr;
2495
2496 VERIFY(sa->sa_len == sizeof *sin6);
2497 if (in6->s6_addr[11] == 0xff && in6->s6_addr[12] == 0xfe) {
2498 sdl->sdl_alen = ETHER_ADDR_LEN;
2499 LLADDR(sdl)[0] = (in6->s6_addr[8] ^ ND6_EUI64_UBIT);
2500 LLADDR(sdl)[1] = in6->s6_addr[9];
2501 LLADDR(sdl)[2] = in6->s6_addr[10];
2502 LLADDR(sdl)[3] = in6->s6_addr[13];
2503 LLADDR(sdl)[4] = in6->s6_addr[14];
2504 LLADDR(sdl)[5] = in6->s6_addr[15];
2505 } else {
2506 sdl->sdl_alen = EUI64_LENGTH;
2507 bcopy(&in6->s6_addr[8], LLADDR(sdl), EUI64_LENGTH);
2508 }
2509
2510 sdl->sdl_slen = 0;
2511 break;
2512 }
2513 case AF_LINK: {
2514 struct sockaddr_dl *sdla = (struct sockaddr_dl *)(void *)sa;
2515 struct in6_addr *in6 = &sin6->sin6_addr;
2516 caddr_t lla = LLADDR(sdla);
2517
2518 VERIFY(sa->sa_len <= sizeof(*sdl));
2519 bcopy(sa, sdl, sa->sa_len);
2520
2521 sin6->sin6_scope_id = sdla->sdl_index;
2522 if (sin6->sin6_scope_id == 0) {
2523 sin6->sin6_scope_id = ifp->if_index;
2524 }
2525 in6->s6_addr[0] = 0xfe;
2526 in6->s6_addr[1] = 0x80;
2527 if (sdla->sdl_alen == EUI64_LENGTH) {
2528 bcopy(lla, &in6->s6_addr[8], EUI64_LENGTH);
2529 } else {
2530 VERIFY(sdla->sdl_alen == ETHER_ADDR_LEN);
2531
2532 in6->s6_addr[8] = ((uint8_t) lla[0] ^ ND6_EUI64_UBIT);
2533 in6->s6_addr[9] = (uint8_t) lla[1];
2534 in6->s6_addr[10] = (uint8_t) lla[2];
2535 in6->s6_addr[11] = 0xff;
2536 in6->s6_addr[12] = 0xfe;
2537 in6->s6_addr[13] = (uint8_t) lla[3];
2538 in6->s6_addr[14] = (uint8_t) lla[4];
2539 in6->s6_addr[15] = (uint8_t) lla[5];
2540 }
2541
2542 break;
2543 }
2544 default:
2545 VERIFY(false);
2546 break;
2547 }
2548 }
2549
2550 int
nd6_alt_node_present(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl,int32_t rssi,int lqm,int npm)2551 nd6_alt_node_present(struct ifnet *ifp, struct sockaddr_in6 *sin6,
2552 struct sockaddr_dl *sdl, int32_t rssi, int lqm, int npm)
2553 {
2554 struct rtentry *rt = NULL;
2555 struct llinfo_nd6 *ln = NULL;
2556 struct if_llreach *lr = NULL;
2557 int nd6_nc_updated = 0;
2558 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2559 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2560
2561 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2562 if (in6_embedded_scope) {
2563 if (temp_embedded_id == 0) {
2564 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2565 }
2566 } else if (temp_ifscope_id == 0) {
2567 sin6->sin6_scope_id = ifp->if_index;
2568 }
2569 }
2570
2571 nd6_cache_lladdr(ifp, &sin6->sin6_addr, LLADDR(sdl), sdl->sdl_alen,
2572 ND_NEIGHBOR_ADVERT, 0, &nd6_nc_updated);
2573
2574 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2575 lck_mtx_lock(rnh_lock);
2576
2577 rt = rtalloc1_scoped_locked((struct sockaddr *)sin6, 1, 0,
2578 ifp->if_index);
2579
2580 /* Restore the address that was passed to us */
2581 if (in6_embedded_scope) {
2582 if (temp_embedded_id == 0) {
2583 sin6->sin6_addr.s6_addr16[1] = 0;
2584 }
2585 } else if (temp_ifscope_id == 0) {
2586 sin6->sin6_scope_id = 0;
2587 }
2588
2589 if (rt != NULL) {
2590 RT_LOCK(rt);
2591 VERIFY(rt->rt_flags & RTF_LLINFO);
2592 VERIFY(rt->rt_llinfo);
2593
2594 ln = rt->rt_llinfo;
2595 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
2596 ln_setexpire(ln, 0);
2597
2598 lr = ln->ln_llreach;
2599 if (lr) {
2600 IFLR_LOCK(lr);
2601 lr->lr_rssi = rssi;
2602 lr->lr_lqm = (int32_t) lqm;
2603 lr->lr_npm = (int32_t) npm;
2604 IFLR_UNLOCK(lr);
2605 }
2606
2607 RT_UNLOCK(rt);
2608 RT_REMREF(rt);
2609 }
2610
2611 lck_mtx_unlock(rnh_lock);
2612
2613 if (rt == NULL) {
2614 log(LOG_ERR, "%s: failed to add/update host route to %s.\n",
2615 __func__, ip6_sprintf(&sin6->sin6_addr));
2616 #if DEBUG || DEVELOPMENT
2617 if (ip6_p2p_debug) {
2618 panic("%s: failed to add/update host route to %s.\n",
2619 __func__, ip6_sprintf(&sin6->sin6_addr));
2620 }
2621 #endif
2622 return EHOSTUNREACH;
2623 }
2624
2625 nd6log(debug, "%s: Successfully added/updated host route to %s [lr=0x%llx]\n",
2626 __func__, ip6_sprintf(&sin6->sin6_addr),
2627 (uint64_t)VM_KERNEL_ADDRPERM(lr));
2628 /*
2629 * nd6_nc_updated not set implies that nothing was updated
2630 * in the neighbor cache. Convey that as EEXIST to callers.
2631 */
2632 if (nd6_nc_updated == 0) {
2633 return EEXIST;
2634 }
2635 return 0;
2636 }
2637
2638 int
nd6_alt_node_absent(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl)2639 nd6_alt_node_absent(struct ifnet *ifp, struct sockaddr_in6 *sin6, struct sockaddr_dl *sdl)
2640 {
2641 struct rtentry *rt = NULL;
2642 int error = 0;
2643 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2644 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2645
2646 nd6log(debug, "%s: host route to %s\n", __func__,
2647 ip6_sprintf(&sin6->sin6_addr));
2648
2649 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2650 if (in6_embedded_scope) {
2651 if (temp_embedded_id == 0) {
2652 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2653 }
2654 } else if (temp_ifscope_id == 0) {
2655 sin6->sin6_scope_id = ifp->if_index;
2656 }
2657 }
2658
2659 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2660 lck_mtx_lock(rnh_lock);
2661
2662 rt = rtalloc1_scoped_locked((struct sockaddr *)sin6, 0, 0,
2663 ifp->if_index);
2664
2665 /* Restore the address that was passed to us */
2666 if (in6_embedded_scope) {
2667 if (temp_embedded_id == 0) {
2668 sin6->sin6_addr.s6_addr16[1] = 0;
2669 }
2670 } else if (temp_ifscope_id == 0) {
2671 sin6->sin6_scope_id = 0;
2672 }
2673
2674 if (rt != NULL) {
2675 RT_LOCK(rt);
2676 if (IS_DYNAMIC_DIRECT_HOSTROUTE(rt)) {
2677 /*
2678 * Copy the link layer information in SDL when present
2679 * as it later gets used to issue the kernel event for
2680 * node absence.
2681 */
2682 if (sdl != NULL && rt->rt_gateway != NULL &&
2683 rt->rt_gateway->sa_family == AF_LINK &&
2684 SDL(rt->rt_gateway)->sdl_len <= sizeof(*sdl)) {
2685 bcopy(rt->rt_gateway, sdl, SDL(rt->rt_gateway)->sdl_len);
2686 }
2687
2688 rt->rt_flags |= RTF_CONDEMNED;
2689 RT_UNLOCK(rt);
2690
2691 error = rtrequest_locked(RTM_DELETE, rt_key(rt),
2692 (struct sockaddr *)NULL, rt_mask(rt), 0,
2693 (struct rtentry **)NULL);
2694
2695 rtfree_locked(rt);
2696 } else {
2697 error = EHOSTUNREACH;
2698 RT_REMREF_LOCKED(rt);
2699 RT_UNLOCK(rt);
2700 }
2701 } else {
2702 error = EHOSTUNREACH;
2703 }
2704
2705 if (error == 0) {
2706 nd6log(debug, "%s: Successfully deleted host route to %s "
2707 "for interface %s.\n", __func__, ip6_sprintf(&sin6->sin6_addr),
2708 ifp->if_xname);
2709 } else {
2710 nd6log(error, "%s: Failed to delete host route to %s "
2711 "for interface %s with error :%d.\n", __func__,
2712 ip6_sprintf(&sin6->sin6_addr),
2713 ifp->if_xname, error);
2714 }
2715
2716 lck_mtx_unlock(rnh_lock);
2717 return error;
2718 }
2719