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#include <net/sockaddr_utils.h>
102
103struct dadq;
104static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
105void nd6_dad_stoptimer(struct ifaddr *);
106static void nd6_dad_timer(struct ifaddr *);
107static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
108static void nd6_dad_ns_input(struct ifaddr *, char *, int, struct nd_opt_nonce *);
109static struct mbuf *nd6_dad_na_input(struct mbuf *, struct ifnet *,
110 struct in6_addr *, caddr_t, int);
111static void dad_addref(struct dadq *, int);
112static void dad_remref(struct dadq *);
113static struct dadq *nd6_dad_attach(struct dadq *, struct ifaddr *);
114static void nd6_dad_detach(struct dadq *, struct ifaddr *);
115static void nd6_dad_duplicated(struct ifaddr *);
116
117static int dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */
118
119#define DAD_LOCK_ASSERT_HELD(_dp) \
120 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_OWNED)
121
122#define DAD_LOCK_ASSERT_NOTHELD(_dp) \
123 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_NOTOWNED)
124
125#define DAD_LOCK(_dp) \
126 lck_mtx_lock(&(_dp)->dad_lock)
127
128#define DAD_LOCK_SPIN(_dp) \
129 lck_mtx_lock_spin(&(_dp)->dad_lock)
130
131#define DAD_CONVERT_LOCK(_dp) do { \
132 DAD_LOCK_ASSERT_HELD(_dp); \
133 lck_mtx_convert_spin(&(_dp)->dad_lock); \
134} while (0)
135
136#define DAD_UNLOCK(_dp) \
137 lck_mtx_unlock(&(_dp)->dad_lock)
138
139#define DAD_ADDREF(_dp) \
140 dad_addref(_dp, 0)
141
142#define DAD_ADDREF_LOCKED(_dp) \
143 dad_addref(_dp, 1)
144
145#define DAD_REMREF(_dp) \
146 dad_remref(_dp)
147
148static LCK_MTX_DECLARE_ATTR(dad6_mutex, &ip6_mutex_grp, &ip6_mutex_attr);
149
150static struct sockaddr_in6 hostrtmask;
151
152static int nd6_llreach_base = 30; /* seconds */
153SYSCTL_DECL(_net_inet6_icmp6);
154SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_llreach_base,
155 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_llreach_base, 0,
156 "default ND6 link-layer reachability max lifetime (in seconds)");
157
158int dad_enhanced = ND6_DAD_ENHANCED_DEFAULT;
159SYSCTL_DECL(_net_inet6_ip6);
160SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_RW | CTLFLAG_LOCKED,
161 &dad_enhanced, 0,
162 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
163
164static uint32_t nd6_dad_nonce_max_count = 3;
165SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, nd6_dad_nonce_max_count,
166 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_dad_nonce_max_count, 0, "Number of times to ignore same nonce for DAD");
167
168#if DEBUG || DEVELOPMENT
169static int ip6_p2p_debug = 0;
170SYSCTL_INT(_net_inet6_ip6, OID_AUTO, ip6_p2p_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
171 &ip6_p2p_debug, 0,
172 "Enable more instrumentation for IPv6 P2P use-case");
173#endif
174
175/*
176 * Obtain a link-layer source cache entry for the sender.
177 *
178 * NOTE: This is currently only for ND6/Ethernet.
179 */
180void
181nd6_llreach_alloc(struct rtentry *rt, struct ifnet *ifp, void *addr,
182 unsigned int alen, boolean_t solicited)
183{
184 struct llinfo_nd6 *ln = rt->rt_llinfo;
185
186 if (nd6_llreach_base != 0 &&
187 (ln->ln_expire != 0 || (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) &&
188 !(rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
189 ifp->if_addrlen == IF_LLREACH_MAXLEN && /* Ethernet */
190 alen == ifp->if_addrlen) {
191 struct if_llreach *lr;
192 const char *why = NULL, *type = "";
193
194 /* Become a regular mutex, just in case */
195 RT_CONVERT_LOCK(rt);
196
197 if ((lr = ln->ln_llreach) != NULL) {
198 type = (solicited ? "ND6 advertisement" :
199 "ND6 unsolicited announcement");
200 /*
201 * If target has changed, create a new record;
202 * otherwise keep existing record.
203 */
204 IFLR_LOCK(lr);
205 if (bcmp(s1: addr, s2: lr->lr_key.addr, n: alen) != 0) {
206 IFLR_UNLOCK(lr);
207 /* Purge any link-layer info caching */
208 VERIFY(rt->rt_llinfo_purge != NULL);
209 rt->rt_llinfo_purge(rt);
210 lr = NULL;
211 why = " for different target HW address; "
212 "using new llreach record";
213 } else {
214 lr->lr_probes = 0; /* reset probe count */
215 IFLR_UNLOCK(lr);
216 if (solicited) {
217 why = " for same target HW address; "
218 "keeping existing llreach record";
219 }
220 }
221 }
222
223 if (lr == NULL) {
224 lr = ln->ln_llreach = ifnet_llreach_alloc(ifp,
225 ETHERTYPE_IPV6, addr, alen, nd6_llreach_base);
226 if (lr != NULL) {
227 lr->lr_probes = 0; /* reset probe count */
228 if (why == NULL) {
229 why = "creating new llreach record";
230 }
231 }
232 }
233
234 if (nd6_debug && lr != NULL && why != NULL) {
235 char tmp[MAX_IPv6_STR_LEN];
236
237 nd6log(debug, "%s: %s%s for %s\n", if_name(ifp),
238 type, why, inet_ntop(AF_INET6,
239 &SIN6(rt_key(rt))->sin6_addr, tmp, sizeof(tmp)));
240 }
241 }
242}
243
244void
245nd6_llreach_use(struct llinfo_nd6 *ln)
246{
247 if (ln->ln_llreach != NULL) {
248 ln->ln_lastused = net_uptime();
249 }
250}
251
252/*
253 * Input a Neighbor Solicitation Message.
254 *
255 * Based on RFC 4861
256 * Based on RFC 4862 (duplicate address detection)
257 */
258void
259nd6_ns_input(
260 struct mbuf *m,
261 int off,
262 int icmp6len)
263{
264 struct ifnet *ifp = m->m_pkthdr.rcvif;
265 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
266 struct nd_neighbor_solicit *nd_ns = NULL;
267 struct in6_addr saddr6 = ip6->ip6_src;
268 struct in6_addr daddr6 = ip6->ip6_dst;
269 uint32_t saddr_ifscope = IN6_IS_SCOPE_EMBED(&saddr6) ? ip6_input_getsrcifscope(m) : IFSCOPE_NONE;
270 struct in6_addr taddr6 = {};
271 struct in6_addr myaddr6 = {};
272 uint32_t myaddr_ifscope = IFSCOPE_NONE;
273 char *lladdr = NULL;
274 struct ifaddr *ifa = NULL;
275 int lladdrlen = 0;
276 int anycast = 0, proxy = 0, dadprogress = 0;
277 int tlladdr = 0;
278 union nd_opts ndopts = {};
279 struct sockaddr_dl proxydl = {};
280 boolean_t advrouter = FALSE;
281 boolean_t is_dad_probe = FALSE;
282 int oflgclr = 0;
283 uint32_t taddr_ifscope;
284
285 /* Expect 32-bit aligned data pointer on strict-align platforms */
286 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
287
288 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
289 ip6 = mtod(m, struct ip6_hdr *);
290 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
291 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
292
293 taddr6 = nd_ns->nd_ns_target;
294 if (in6_setscope(&taddr6, ifp, &taddr_ifscope) != 0) {
295 goto bad;
296 }
297
298 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
299 nd6log(error,
300 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
301 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
302 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
303 goto bad;
304 }
305
306 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(&saddr6);
307 if (is_dad_probe) {
308 /* dst has to be a solicited node multicast address. */
309 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
310 /* don't check ifindex portion */
311 daddr6.s6_addr32[1] == 0 &&
312 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
313 daddr6.s6_addr8[12] == 0xff) {
314 ; /* good */
315 } else {
316 nd6log(info, "nd6_ns_input: bad DAD packet "
317 "(wrong ip6 dst)\n");
318 goto bad;
319 }
320 } else if (!nd6_onlink_ns_rfc4861) {
321 struct sockaddr_in6 src_sa6;
322
323 /*
324 * According to recent IETF discussions, it is not a good idea
325 * to accept a NS from an address which would not be deemed
326 * to be a neighbor otherwise. This point is expected to be
327 * clarified in future revisions of the specification.
328 */
329 SOCKADDR_ZERO(&src_sa6, sizeof(src_sa6));
330 src_sa6.sin6_family = AF_INET6;
331 src_sa6.sin6_len = sizeof(src_sa6);
332 src_sa6.sin6_addr = saddr6;
333 if (!in6_embedded_scope) {
334 src_sa6.sin6_scope_id = saddr_ifscope;
335 }
336 if (!nd6_is_addr_neighbor(&src_sa6, ifp, 0)) {
337 nd6log(info, "nd6_ns_input: "
338 "NS packet from non-neighbor\n");
339 goto bad;
340 }
341 }
342
343 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
344 nd6log(info, "nd6_ns_input: bad NS target (multicast)\n");
345 goto bad;
346 }
347
348 icmp6len -= sizeof(*nd_ns);
349 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
350 if (nd6_options(&ndopts) < 0) {
351 nd6log(info,
352 "nd6_ns_input: invalid ND option, ignored\n");
353 /* nd6_options have incremented stats */
354 goto freeit;
355 }
356
357 if (ndopts.nd_opts_src_lladdr) {
358 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
359 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
360 }
361
362 if (is_dad_probe && lladdr) {
363 nd6log(info, "nd6_ns_input: bad DAD packet "
364 "(link-layer address option)\n");
365 goto bad;
366 }
367
368 /*
369 * Attaching target link-layer address to the NA?
370 * (RFC 2461 7.2.4)
371 *
372 * NS IP dst is unicast/anycast MUST NOT add
373 * NS IP dst is solicited-node multicast MUST add
374 *
375 * In implementation, we add target link-layer address by default.
376 * We do not add one in MUST NOT cases.
377 */
378 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) {
379 tlladdr = 0;
380 } else {
381 tlladdr = 1;
382 }
383
384 /*
385 * Target address (taddr6) must be either:
386 * (1) Valid unicast/anycast address for my receiving interface,
387 * (2) Unicast address for which I'm offering proxy service, or
388 * (3) "tentative" or "optimistic" address [DAD is in progress].
389 */
390 /* (1) and (3) check. */
391 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
392
393 /* (2) check. */
394 if (ifa == NULL) {
395 struct rtentry *rt;
396 struct sockaddr_in6 tsin6;
397
398 SOCKADDR_ZERO(&tsin6, sizeof tsin6);
399 tsin6.sin6_len = sizeof(struct sockaddr_in6);
400 tsin6.sin6_family = AF_INET6;
401 tsin6.sin6_addr = taddr6;
402
403 rt = rtalloc1_scoped(SA(&tsin6), 0, 0, ifp->if_index);
404
405 if (rt != NULL) {
406 RT_LOCK(rt);
407 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
408 rt->rt_gateway->sa_family == AF_LINK) {
409 /*
410 * proxy NDP for single entry
411 */
412 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
413 ifp, IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
414 if (ifa) {
415 proxy = 1;
416 proxydl = *SDL(rt->rt_gateway);
417 }
418 }
419 RT_UNLOCK(rt);
420 rtfree(rt);
421 }
422 }
423 if (ifa == NULL && ip6_forwarding && nd6_prproxy) {
424 /*
425 * Is the target address part of the prefix that is being
426 * proxied and installed on another interface?
427 */
428 ifa = (struct ifaddr *)in6ifa_prproxyaddr(&taddr6, taddr_ifscope);
429 }
430 if (ifa == NULL) {
431 /*
432 * We've got an NS packet, and we don't have that address
433 * assigned for us. We MUST silently ignore it on this
434 * interface, c.f. RFC 4861 7.2.3.
435 *
436 * Forwarding associated with NDPRF_PRPROXY may apply.
437 */
438 if (ip6_forwarding && nd6_prproxy) {
439 nd6_prproxy_ns_input(ifp, &saddr6, lladdr,
440 lladdrlen, &daddr6, &taddr6,
441 nonce: (ndopts.nd_opts_nonce == NULL) ? NULL :
442 ndopts.nd_opts_nonce->nd_opt_nonce);
443 }
444 goto freeit;
445 }
446 IFA_LOCK(ifa);
447 myaddr6 = *IFA_IN6(ifa);
448 myaddr_ifscope = IFA_SIN6_SCOPE(ifa);
449 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
450 dadprogress =
451 ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DADPROGRESS;
452 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) {
453 IFA_UNLOCK(ifa);
454 goto freeit;
455 }
456 IFA_UNLOCK(ifa);
457
458 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
459 nd6log(info,
460 "nd6_ns_input: lladdrlen mismatch for %s "
461 "(if %d, NS packet %d)\n",
462 ip6_sprintf(&taddr6), ifp->if_addrlen, lladdrlen - 2);
463 goto bad;
464 }
465
466 if (in6_are_addr_equal_scoped(&myaddr6, &saddr6, myaddr_ifscope, saddr_ifscope)) {
467 nd6log(info,
468 "nd6_ns_input: duplicate IP6 address %s\n",
469 ip6_sprintf(&saddr6));
470 goto freeit;
471 }
472
473 /*
474 * We have neighbor solicitation packet, with target address equals to
475 * one of my DAD in-progress addresses.
476 *
477 * src addr how to process?
478 * --- ---
479 * multicast of course, invalid (rejected in ip6_input)
480 * unicast somebody is doing address resolution
481 * unspec dup address detection
482 *
483 * The processing is defined in the "draft standard" RFC 4862 (and by
484 * RFC 4429, which is a "proposed standard" update to its obsolete
485 * predecessor, RFC 2462) The reason optimistic DAD is not included
486 * in RFC 4862 is entirely due to IETF procedural considerations.
487 */
488 if (dadprogress) {
489 /*
490 * If source address is unspecified address, it is for
491 * duplicate address detection.
492 *
493 * If not, the packet is for addess resolution;
494 * silently ignore it when not optimistic
495 *
496 * Per RFC 4429 the reply for an optimistic address must
497 * have the Override flag cleared
498 */
499 if (!is_dad_probe && (dadprogress & IN6_IFF_OPTIMISTIC) != 0) {
500 oflgclr = 1;
501 } else {
502 if (is_dad_probe) {
503 nd6_dad_ns_input(ifa, lladdr, lladdrlen, ndopts.nd_opts_nonce);
504 }
505
506 goto freeit;
507 }
508 }
509
510 /* Are we an advertising router on this interface? */
511 advrouter = (ifp->if_ipv6_router_mode != IPV6_ROUTER_MODE_DISABLED);
512
513 /*
514 * If the source address is unspecified address, entries must not
515 * be created or updated.
516 * It looks that sender is performing DAD. If I'm using the address,
517 * and it's a "preferred" address, i.e. not optimistic, then output NA
518 * toward all-node multicast address, to tell the sender that I'm using
519 * the address.
520 * S bit ("solicited") must be zero.
521 */
522 if (is_dad_probe) {
523 saddr6 = in6addr_linklocal_allnodes;
524 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
525 goto bad;
526 }
527 if ((dadprogress & IN6_IFF_OPTIMISTIC) == 0) {
528 nd6_na_output(ifp, &saddr6, &taddr6,
529 ((anycast || proxy || !tlladdr) ? 0 :
530 ND_NA_FLAG_OVERRIDE) | (advrouter ?
531 ND_NA_FLAG_ROUTER : 0), tlladdr, proxy ?
532 SA(&proxydl) : NULL);
533 }
534 goto freeit;
535 }
536
537 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
538 ND_NEIGHBOR_SOLICIT, 0, NULL);
539
540 nd6_na_output(ifp, &saddr6, &taddr6,
541 ((anycast || proxy || !tlladdr || oflgclr) ? 0 : ND_NA_FLAG_OVERRIDE) |
542 (advrouter ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
543 tlladdr, proxy ? SA(&proxydl) : NULL);
544freeit:
545 m_freem(m);
546 if (ifa != NULL) {
547 ifa_remref(ifa);
548 }
549 return;
550
551bad:
552 nd6log(error, "nd6_ns_input: src=%s\n", ip6_sprintf(&saddr6));
553 nd6log(error, "nd6_ns_input: dst=%s\n", ip6_sprintf(&daddr6));
554 nd6log(error, "nd6_ns_input: tgt=%s\n", ip6_sprintf(&taddr6));
555 icmp6stat.icp6s_badns++;
556 m_freem(m);
557 if (ifa != NULL) {
558 ifa_remref(ifa);
559 }
560}
561
562/*
563 * Output a Neighbor Solicitation Message. Caller specifies:
564 * - ICMP6 header source IP6 address
565 * - ND6 header target IP6 address
566 * - ND6 header source datalink address
567 *
568 * Based on RFC 4861
569 * Based on RFC 4862 (duplicate address detection)
570 * Based on RFC 4429 (optimistic duplicate address detection)
571 *
572 * Caller must bump up ln->ln_rt refcnt to make sure 'ln' doesn't go
573 * away if there is a llinfo_nd6 passed in.
574 */
575void
576nd6_ns_output(
577 struct ifnet *ifp,
578 const struct in6_addr *daddr6,
579 const struct in6_addr *taddr6,
580 struct llinfo_nd6 *ln, /* for source address determination */
581 uint8_t *nonce) /* duplicated address detection */
582{
583 struct mbuf *m;
584 struct ip6_hdr *ip6;
585 struct nd_neighbor_solicit *nd_ns;
586 struct in6_ifaddr *ia = NULL;
587 struct in6_addr *src, src_in, src_storage;
588 struct ip6_moptions *im6o = NULL;
589 struct ifnet *outif = NULL;
590 int icmp6len;
591 int maxlen;
592 int flags;
593 caddr_t mac;
594 struct route_in6 ro;
595 struct ip6_out_args ip6oa;
596 u_int32_t rtflags = 0;
597 boolean_t is_optimistic = FALSE;
598
599 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) || IN6_IS_ADDR_MULTICAST(taddr6)) {
600 return;
601 }
602
603 bzero(s: &ro, n: sizeof(ro));
604 bzero(s: &ip6oa, n: sizeof(ip6oa));
605 ip6oa.ip6oa_boundif = ifp->if_index;
606 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
607 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
608 IP6OAF_MANAGEMENT_ALLOWED;
609 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
610 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
611
612 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
613
614 /* estimate the size of message */
615 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
616 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
617 if (max_linkhdr + maxlen >= MCLBYTES) {
618#if DIAGNOSTIC
619 printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
620 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
621#endif
622 return;
623 }
624
625 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
626 if (m && max_linkhdr + maxlen >= MHLEN) {
627 MCLGET(m, M_DONTWAIT);
628 if ((m->m_flags & M_EXT) == 0) {
629 m_free(m);
630 m = NULL;
631 }
632 }
633 if (m == NULL) {
634 return;
635 }
636 m->m_pkthdr.rcvif = NULL;
637
638 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
639 m->m_flags |= M_MCAST;
640
641 im6o = ip6_allocmoptions(Z_NOWAIT);
642 if (im6o == NULL) {
643 m_freem(m);
644 return;
645 }
646
647 im6o->im6o_multicast_ifp = ifp;
648 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
649 im6o->im6o_multicast_loop = 0;
650 }
651
652 icmp6len = sizeof(*nd_ns);
653 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
654 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
655
656 /* fill neighbor solicitation packet */
657 ip6 = mtod(m, struct ip6_hdr *);
658 ip6->ip6_flow = 0;
659 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
660 ip6->ip6_vfc |= IPV6_VERSION;
661 /* ip6->ip6_plen will be set later */
662 ip6->ip6_nxt = IPPROTO_ICMPV6;
663 ip6->ip6_hlim = IPV6_MAXHLIM;
664 if (daddr6) {
665 ip6->ip6_dst = *daddr6;
666 ip6_output_setdstifscope(m, ifp->if_index, NULL);
667 } else {
668 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
669 ip6->ip6_dst.s6_addr16[1] = 0;
670 ip6->ip6_dst.s6_addr32[1] = 0;
671 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
672 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
673 ip6->ip6_dst.s6_addr8[12] = 0xff;
674 ip6_output_setdstifscope(m, ifp->if_index, NULL);
675 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) {
676 goto bad;
677 }
678 }
679 if (nonce == NULL) {
680 /*
681 * RFC2461 7.2.2:
682 * "If the source address of the packet prompting the
683 * solicitation is the same as one of the addresses assigned
684 * to the outgoing interface, that address SHOULD be placed
685 * in the IP Source Address of the outgoing solicitation.
686 * Otherwise, any one of the addresses assigned to the
687 * interface should be used."
688 *
689 * We use the source address for the prompting packet
690 * (saddr6), if:
691 * - saddr6 is given from the caller (by giving "ln"), and
692 * - saddr6 belongs to the outgoing interface.
693 * Otherwise, we perform the source address selection as usual.
694 */
695 struct ip6_hdr *hip6; /* hold ip6 */
696 struct in6_addr *hsrc = NULL;
697
698 /* Caller holds ref on this route */
699 if (ln != NULL) {
700 RT_LOCK(ln->ln_rt);
701 /*
702 * assuming every packet in ln_hold has the same IP
703 * header
704 */
705 if (ln->ln_hold != NULL) {
706 hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
707 /* XXX pullup? */
708 if (sizeof(*hip6) < ln->ln_hold->m_len) {
709 hsrc = &hip6->ip6_src;
710 } else {
711 hsrc = NULL;
712 }
713 }
714 /* Update probe count, if applicable */
715 if (ln->ln_llreach != NULL) {
716 IFLR_LOCK_SPIN(ln->ln_llreach);
717 ln->ln_llreach->lr_probes++;
718 IFLR_UNLOCK(ln->ln_llreach);
719 }
720 rtflags = ln->ln_rt->rt_flags;
721 RT_UNLOCK(ln->ln_rt);
722 }
723 if (hsrc != NULL && (ia = in6ifa_ifpwithaddr(ifp, hsrc)) &&
724 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0) {
725 src = hsrc;
726 } else {
727 int error;
728 struct sockaddr_in6 dst_sa;
729
730 SOCKADDR_ZERO(&dst_sa, sizeof(dst_sa));
731 dst_sa.sin6_family = AF_INET6;
732 dst_sa.sin6_len = sizeof(dst_sa);
733 dst_sa.sin6_addr = ip6->ip6_dst;
734
735 src = in6_selectsrc(&dst_sa, NULL,
736 NULL, &ro, NULL, &src_storage, ip6oa.ip6oa_boundif,
737 &error);
738 if (src == NULL) {
739 nd6log(debug,
740 "nd6_ns_output: source can't be "
741 "determined: dst=%s, error=%d\n",
742 ip6_sprintf(&dst_sa.sin6_addr),
743 error);
744 goto bad;
745 }
746
747 if (ia != NULL) {
748 ifa_remref(ifa: &ia->ia_ifa);
749 ia = NULL;
750 }
751 /*
752 * RFC 4429 section 3.2:
753 * When a node has a unicast packet to send
754 * from an Optimistic Address to a neighbor,
755 * but does not know the neighbor's link-layer
756 * address, it MUST NOT perform Address
757 * Resolution.
758 */
759 ia = in6ifa_ifpwithaddr(ifp, src);
760 if (ia == NULL) {
761 nd6log(debug,
762 "nd6_ns_output: no preferred source "
763 "available: dst=%s\n",
764 ip6_sprintf(&dst_sa.sin6_addr));
765 goto bad;
766 }
767 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
768 is_optimistic = TRUE;
769 nd6log(debug,
770 "nd6_ns_output: preferred source "
771 "available is optimistic: dst=%s\n",
772 ip6_sprintf(&dst_sa.sin6_addr));
773 }
774 }
775 } else {
776 /*
777 * Source address for DAD packet must always be IPv6
778 * unspecified address. (0::0)
779 * We actually don't have to 0-clear the address (we did it
780 * above), but we do so here explicitly to make the intention
781 * clearer.
782 */
783 bzero(s: &src_in, n: sizeof(src_in));
784 src = &src_in;
785 ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_SRCADDR;
786 }
787 ip6->ip6_src = *src;
788 ip6_output_setsrcifscope(m, ifp->if_index, ia);
789 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
790 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
791 nd_ns->nd_ns_code = 0;
792 nd_ns->nd_ns_reserved = 0;
793 nd_ns->nd_ns_target = *taddr6;
794 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
795
796 /*
797 * Add source link-layer address option.
798 *
799 * spec implementation
800 * --- ---
801 * DAD packet MUST NOT do not add the option
802 * Source is optimistic MUST NOT do not add the option
803 * there's no link layer address:
804 * impossible do not add the option
805 * there's link layer address:
806 * Multicast NS MUST add one add the option
807 * Unicast NS SHOULD add one add the option
808 *
809 * XXX We deviate from RFC 4429 and still use optimistic DAD as source
810 * for address resolution. However to ensure that we do not interfere
811 * with neighbor cache entries of other neighbors, we MUST ensure
812 * that SLLAO is not sent. Also note, sending multicast NS without SLLAO
813 * is also a deviation from RFC 4861.
814 */
815 if (nonce == NULL && (mac = nd6_ifptomac(ifp)) && !is_optimistic) {
816 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
817 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
818 /* 8 byte alignments... */
819 optlen = (optlen + 7) & ~7;
820
821 m->m_pkthdr.len += optlen;
822 m->m_len += optlen;
823 icmp6len += optlen;
824 bzero(s: (caddr_t)nd_opt, n: optlen);
825 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
826 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
827 bcopy(src: mac, dst: (caddr_t)(nd_opt + 1), n: ifp->if_addrlen);
828 }
829 /*
830 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
831 * This behavior is documented as Enhanced Duplicate Address
832 * Detection in draft-ietf-6man-enhanced-dad-13.
833 * net.inet6.ip6.dad_enhanced=0 disables this.
834 */
835 if (dad_enhanced != 0 && nonce != NULL && !(ifp->if_flags & IFF_POINTOPOINT)) {
836 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
837 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
838 /* 8-byte alignment is required. */
839 optlen = (optlen + 7) & ~7;
840
841 m->m_pkthdr.len += optlen;
842 m->m_len += optlen;
843 icmp6len += optlen;
844 bzero(s: (caddr_t)nd_opt, n: optlen);
845 nd_opt->nd_opt_type = ND_OPT_NONCE;
846 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
847 bcopy(src: nonce, dst: (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
848 }
849 ip6->ip6_plen = htons((u_short)icmp6len);
850 nd_ns->nd_ns_cksum = 0;
851 nd_ns->nd_ns_cksum
852 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
853
854 flags = nonce ? IPV6_UNSPECSRC : 0;
855 flags |= IPV6_OUTARGS;
856
857 /*
858 * PKTF_{INET,INET6}_RESOLVE_RTR are mutually exclusive, so make
859 * sure only one of them is set (just in case.)
860 */
861 m->m_pkthdr.pkt_flags &= ~(PKTF_INET_RESOLVE | PKTF_RESOLVE_RTR);
862 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
863 /*
864 * If this is a NS for resolving the (default) router, mark
865 * the packet accordingly so that the driver can find out,
866 * in case it needs to perform driver-specific action(s).
867 */
868 if (rtflags & RTF_ROUTER) {
869 m->m_pkthdr.pkt_flags |= PKTF_RESOLVE_RTR;
870 }
871
872 if (ifp->if_eflags & IFEF_TXSTART) {
873 /*
874 * Use control service class if the interface
875 * supports transmit-start model
876 */
877 (void) m_set_service_class(m, MBUF_SC_CTL);
878 }
879
880 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
881 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
882 ip6_output(m, NULL, NULL, flags, im6o, &outif, &ip6oa);
883 if (outif) {
884 icmp6_ifstat_inc(outif, ifs6_out_msg);
885 icmp6_ifstat_inc(outif, ifs6_out_neighborsolicit);
886 ifnet_release(interface: outif);
887 }
888 icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
889
890exit:
891 if (im6o != NULL) {
892 IM6O_REMREF(im6o);
893 }
894
895 ROUTE_RELEASE(&ro); /* we don't cache this route. */
896
897 if (ia != NULL) {
898 ifa_remref(ifa: &ia->ia_ifa);
899 }
900 return;
901
902bad:
903 m_freem(m);
904 goto exit;
905}
906
907/*
908 * Neighbor advertisement input handling.
909 *
910 * Based on RFC 4861
911 * Based on RFC 4862 (duplicate address detection)
912 *
913 * the following items are not implemented yet:
914 * - anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD)
915 * - proxy advertisement delay rule (RFC 4861 7.2.8, last paragraph, "should")
916 */
917void
918nd6_na_input(struct mbuf *m, int off, int icmp6len)
919{
920 struct ifnet *ifp = m->m_pkthdr.rcvif;
921 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
922 struct nd_neighbor_advert *nd_na;
923 struct in6_addr saddr6 = ip6->ip6_src;
924 struct in6_addr daddr6 = ip6->ip6_dst;
925 struct in6_addr taddr6;
926 int flags;
927 int is_router;
928 int is_solicited;
929 int is_override;
930 char *lladdr = NULL;
931 int lladdrlen = 0;
932 struct llinfo_nd6 *ln;
933 struct rtentry *rt;
934 struct sockaddr_dl *sdl;
935 union nd_opts ndopts;
936 uint64_t timenow;
937 bool send_nc_alive_kev = false;
938
939 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
940 nd6log(info, "nd6_na_input: on ND6ALT interface!\n");
941 goto freeit;
942 }
943
944 /* Expect 32-bit aligned data pointer on strict-align platforms */
945 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
946
947 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
948 nd6log(error,
949 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
950 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
951 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
952 goto bad;
953 }
954
955 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
956 ip6 = mtod(m, struct ip6_hdr *);
957 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
958 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
959
960 flags = nd_na->nd_na_flags_reserved;
961 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
962 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
963 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
964
965 taddr6 = nd_na->nd_na_target;
966 if (in6_setscope(&taddr6, ifp, NULL)) {
967 goto bad; /* XXX: impossible */
968 }
969 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
970 nd6log(error,
971 "nd6_na_input: invalid target address %s\n",
972 ip6_sprintf(&taddr6));
973 goto bad;
974 }
975 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
976 if (is_solicited) {
977 nd6log(error,
978 "nd6_na_input: a solicited adv is multicasted\n");
979 goto bad;
980 }
981 }
982
983 icmp6len -= sizeof(*nd_na);
984 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
985 if (nd6_options(&ndopts) < 0) {
986 nd6log(info,
987 "nd6_na_input: invalid ND option, ignored\n");
988 /* nd6_options have incremented stats */
989 goto freeit;
990 }
991
992 if (ndopts.nd_opts_tgt_lladdr) {
993 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
994 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
995
996 if (((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
997 nd6log(info,
998 "nd6_na_input: lladdrlen mismatch for %s "
999 "(if %d, NA packet %d)\n",
1000 ip6_sprintf(&taddr6), ifp->if_addrlen,
1001 lladdrlen - 2);
1002 goto bad;
1003 }
1004 }
1005
1006 m = nd6_dad_na_input(m, ifp, &taddr6, lladdr, lladdrlen);
1007 if (m == NULL) {
1008 return;
1009 }
1010
1011 /* Forwarding associated with NDPRF_PRPROXY may apply. */
1012 if (ip6_forwarding && nd6_prproxy) {
1013 nd6_prproxy_na_input(ifp, &saddr6, &daddr6, &taddr6, flags);
1014 }
1015
1016 /*
1017 * If no neighbor cache entry is found, NA SHOULD silently be
1018 * discarded. If we are forwarding (and Scoped Routing is in
1019 * effect), try to see if there is a neighbor cache entry on
1020 * another interface (in case we are doing prefix proxying.)
1021 */
1022 if ((rt = nd6_lookup(&taddr6, 0, ifp, 0)) == NULL) {
1023 if (!ip6_forwarding || !nd6_prproxy) {
1024 goto freeit;
1025 }
1026
1027 if ((rt = nd6_lookup(&taddr6, 0, NULL, 0)) == NULL) {
1028 goto freeit;
1029 }
1030
1031 RT_LOCK_ASSERT_HELD(rt);
1032 if (rt->rt_ifp != ifp) {
1033 /*
1034 * Purge any link-layer info caching.
1035 */
1036 if (rt->rt_llinfo_purge != NULL) {
1037 rt->rt_llinfo_purge(rt);
1038 }
1039
1040 /* Adjust route ref count for the interfaces */
1041 if (rt->rt_if_ref_fn != NULL) {
1042 rt->rt_if_ref_fn(ifp, 1);
1043 rt->rt_if_ref_fn(rt->rt_ifp, -1);
1044 }
1045
1046 /* Change the interface when the existing route is on */
1047 rt->rt_ifp = ifp;
1048
1049 /*
1050 * If rmx_mtu is not locked, update it
1051 * to the MTU used by the new interface.
1052 */
1053 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1054 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1055 }
1056 }
1057 }
1058
1059 RT_LOCK_ASSERT_HELD(rt);
1060 if ((ln = rt->rt_llinfo) == NULL ||
1061 (sdl = SDL(rt->rt_gateway)) == NULL) {
1062 RT_REMREF_LOCKED(rt);
1063 RT_UNLOCK(rt);
1064 goto freeit;
1065 }
1066
1067 timenow = net_uptime();
1068
1069 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1070 /*
1071 * If the link-layer has address, and no lladdr option came,
1072 * discard the packet.
1073 */
1074 if (ifp->if_addrlen && !lladdr) {
1075 RT_REMREF_LOCKED(rt);
1076 RT_UNLOCK(rt);
1077 goto freeit;
1078 }
1079
1080 /*
1081 * Record link-layer address, and update the state.
1082 */
1083 sdl->sdl_alen = ifp->if_addrlen;
1084 bcopy(src: lladdr, LLADDR(sdl), n: ifp->if_addrlen);
1085 if (is_solicited) {
1086 send_nc_alive_kev = (rt->rt_flags & RTF_ROUTER) ? true : false;
1087 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1088 if (ln->ln_expire != 0) {
1089 struct nd_ifinfo *ndi = NULL;
1090
1091 ndi = ND_IFINFO(rt->rt_ifp);
1092 VERIFY(ndi != NULL && ndi->initialized);
1093 lck_mtx_lock(lck: &ndi->lock);
1094 ln_setexpire(ln, timenow + ndi->reachable);
1095 lck_mtx_unlock(lck: &ndi->lock);
1096 RT_UNLOCK(rt);
1097 lck_mtx_lock(rnh_lock);
1098 nd6_sched_timeout(NULL, NULL);
1099 lck_mtx_unlock(rnh_lock);
1100 RT_LOCK(rt);
1101 }
1102 } else {
1103 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1104 ln_setexpire(ln, timenow + nd6_gctimer);
1105 }
1106
1107 /*
1108 * Enqueue work item to invoke callback for this
1109 * route entry
1110 */
1111 route_event_enqueue_nwk_wq_entry(rt, NULL,
1112 ROUTE_LLENTRY_RESOLVED, NULL, TRUE);
1113
1114 if ((ln->ln_router = (short)is_router) != 0) {
1115 struct radix_node_head *rnh = NULL;
1116 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1117 struct ifnet *rt_ifp = rt->rt_ifp;
1118
1119 struct route_event rt_ev;
1120 route_event_init(p_route_ev: &rt_ev, rt, NULL, route_ev_code: ROUTE_LLENTRY_RESOLVED);
1121 /*
1122 * This means a router's state has changed from
1123 * non-reachable to probably reachable, and might
1124 * affect the status of associated prefixes..
1125 * We already have a reference on rt. Don't need to
1126 * take one for the unlock/lock.
1127 */
1128 RT_UNLOCK(rt);
1129 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1130 lck_mtx_lock(rnh_lock);
1131 rnh = rt_tables[AF_INET6];
1132
1133 if (rnh != NULL) {
1134 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1135 (void *)&rt_ev);
1136 }
1137 lck_mtx_unlock(rnh_lock);
1138 lck_mtx_lock(nd6_mutex);
1139 pfxlist_onlink_check();
1140 lck_mtx_unlock(nd6_mutex);
1141 RT_LOCK(rt);
1142 }
1143 } else {
1144 int llchange = 0;
1145
1146 /*
1147 * Check if the link-layer address has changed or not.
1148 */
1149 if (lladdr == NULL) {
1150 llchange = 0;
1151 } else {
1152 if (sdl->sdl_alen) {
1153 if (bcmp(s1: lladdr, LLADDR(sdl), n: ifp->if_addrlen)) {
1154 llchange = 1;
1155 } else {
1156 llchange = 0;
1157 }
1158 } else {
1159 llchange = 1;
1160 }
1161 }
1162
1163 /*
1164 * This is VERY complex. Look at it with care.
1165 *
1166 * override solicit lladdr llchange action
1167 * (L: record lladdr)
1168 *
1169 * 0 0 n -- (2c)
1170 * 0 0 y n (2b) L
1171 * 0 0 y y (1) REACHABLE->STALE
1172 * 0 1 n -- (2c) *->REACHABLE
1173 * 0 1 y n (2b) L *->REACHABLE
1174 * 0 1 y y (1) REACHABLE->STALE
1175 * 1 0 n -- (2a)
1176 * 1 0 y n (2a) L
1177 * 1 0 y y (2a) L *->STALE
1178 * 1 1 n -- (2a) *->REACHABLE
1179 * 1 1 y n (2a) L *->REACHABLE
1180 * 1 1 y y (2a) L *->REACHABLE
1181 */
1182 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
1183 /*
1184 * If state is REACHABLE, make it STALE.
1185 * no other updates should be done.
1186 */
1187 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
1188 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1189 ln_setexpire(ln, timenow + nd6_gctimer);
1190 }
1191 RT_REMREF_LOCKED(rt);
1192 RT_UNLOCK(rt);
1193 goto freeit;
1194 } else if (is_override /* (2a) */
1195 || (!is_override && (lladdr && !llchange)) /* (2b) */
1196 || !lladdr) { /* (2c) */
1197 /*
1198 * Update link-local address, if any.
1199 */
1200 if (lladdr) {
1201 sdl->sdl_alen = ifp->if_addrlen;
1202 bcopy(src: lladdr, LLADDR(sdl), n: ifp->if_addrlen);
1203 }
1204
1205 /*
1206 * If solicited, make the state REACHABLE.
1207 * If not solicited and the link-layer address was
1208 * changed, make it STALE.
1209 */
1210 if (is_solicited) {
1211 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1212 if (ln->ln_expire != 0) {
1213 struct nd_ifinfo *ndi = NULL;
1214
1215 ndi = ND_IFINFO(ifp);
1216 VERIFY(ndi != NULL && ndi->initialized);
1217 lck_mtx_lock(lck: &ndi->lock);
1218 ln_setexpire(ln,
1219 timenow + ndi->reachable);
1220 lck_mtx_unlock(lck: &ndi->lock);
1221 RT_UNLOCK(rt);
1222 lck_mtx_lock(rnh_lock);
1223 nd6_sched_timeout(NULL, NULL);
1224 lck_mtx_unlock(rnh_lock);
1225 RT_LOCK(rt);
1226 }
1227 } else {
1228 if (lladdr && llchange) {
1229 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1230 ln_setexpire(ln, timenow + nd6_gctimer);
1231 }
1232 }
1233
1234 /*
1235 * XXX
1236 * The above is somewhat convoluted, for now just
1237 * issue a callback for LLENTRY changed.
1238 */
1239 /* Enqueue work item to invoke callback for this route entry */
1240 if (llchange) {
1241 route_event_enqueue_nwk_wq_entry(rt, NULL,
1242 ROUTE_LLENTRY_CHANGED, NULL, TRUE);
1243 }
1244
1245 /*
1246 * If the router's link-layer address has changed,
1247 * notify routes using this as gateway so they can
1248 * update any cached information.
1249 */
1250 if (ln->ln_router && is_router && llchange) {
1251 struct radix_node_head *rnh = NULL;
1252 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1253 struct ifnet *rt_ifp = rt->rt_ifp;
1254 struct route_event rt_ev;
1255 route_event_init(p_route_ev: &rt_ev, rt, NULL, route_ev_code: ROUTE_LLENTRY_CHANGED);
1256
1257 /*
1258 * This means a router's state has changed from
1259 * non-reachable to probably reachable, and might
1260 * affect the status of associated prefixes..
1261 *
1262 * We already have a valid rt reference here.
1263 * We don't need to take another one for unlock/lock.
1264 */
1265 RT_UNLOCK(rt);
1266 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1267 lck_mtx_lock(rnh_lock);
1268 rnh = rt_tables[AF_INET6];
1269
1270 if (rnh != NULL) {
1271 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1272 (void *)&rt_ev);
1273 }
1274 lck_mtx_unlock(rnh_lock);
1275 RT_LOCK(rt);
1276 }
1277 }
1278
1279 if (ln->ln_router && !is_router) {
1280 /*
1281 * The peer dropped the router flag.
1282 * Remove the sender from the Default Router List and
1283 * update the Destination Cache entries.
1284 */
1285 struct nd_defrouter *dr;
1286 struct in6_addr *in6;
1287 struct ifnet *rt_ifp = rt->rt_ifp;
1288
1289 in6 = &SIN6(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(s: &ev_msg, n: sizeof(ev_msg));
1324 bzero(s: &nd6_ndalive, n: 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(dst: nd6_ndalive.link_data.if_name,
1333 src: ifp->if_name,
1334 n: 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), alen: 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
1378bad:
1379 icmp6stat.icp6s_badna++;
1380 /* fall through */
1381freeit:
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 */
1398void
1399nd6_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(s: &ro, n: sizeof(ro));
1421
1422 daddr6 = *daddr6_0; /* make a local copy for modification */
1423
1424 bzero(s: &ip6oa, n: 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 SOCKADDR_ZERO(&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 SOCKADDR_COPY(&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(ifa: &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 = SDL(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(s: (caddr_t)nd_opt, n: optlen);
1571 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1572 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
1573 bcopy(src: mac, dst: (caddr_t)(nd_opt + 1), n: 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(interface: outif);
1600 }
1601 icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1602
1603exit:
1604 if (im6o != NULL) {
1605 IM6O_REMREF(im6o);
1606 }
1607
1608 ROUTE_RELEASE(&ro);
1609 return;
1610
1611bad:
1612 m_freem(m);
1613 goto exit;
1614}
1615
1616caddr_t
1617nd6_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
1643TAILQ_HEAD(dadq_head, dadq);
1644struct 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
1665static KALLOC_TYPE_DEFINE(dad_zone, struct dadq, NET_KT_DEFAULT);
1666static struct dadq_head dadq;
1667
1668void
1669nd6_nbr_init(void)
1670{
1671 int i;
1672
1673 TAILQ_INIT(&dadq);
1674
1675 SOCKADDR_ZERO(&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
1683static struct dadq *
1684nd6_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(lck: &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(s1: &nonce->nd_opt_nonce[0], s2: &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(lck: &dad6_mutex);
1729 return dp;
1730}
1731
1732void
1733nd6_dad_stoptimer(
1734 struct ifaddr *ifa)
1735{
1736 untimeout((void (*)(void *))nd6_dad_timer, arg: (void *)ifa);
1737}
1738
1739/*
1740 * Start Duplicate Address Detection (DAD) for specified interface address.
1741 */
1742void
1743nd6_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(lck: &dp->dad_lock, grp: &ifa_mtx_grp, attr: &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(lck: &ndi->lock);
1819 retrans = ndi->retrans * hz / 1000;
1820 lck_mtx_unlock(lck: &ndi->lock);
1821 timeout((void (*)(void *))nd6_dad_timer, arg: (void *)ifa, ticks: 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, arg: (void *)ifa,
1832 ticks: ntick);
1833 }
1834
1835 DAD_REMREF(dp); /* drop our reference */
1836}
1837
1838static struct dadq *
1839nd6_dad_attach(struct dadq *dp, struct ifaddr *ifa)
1840{
1841 lck_mtx_lock(lck: &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(lck: &dad6_mutex);
1858
1859 return dp;
1860}
1861
1862static void
1863nd6_dad_detach(struct dadq *dp, struct ifaddr *ifa)
1864{
1865 int detached;
1866
1867 lck_mtx_lock(lck: &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(lck: &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 */
1886void
1887nd6_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, arg: (void *)ifa);
1898
1899 nd6_dad_detach(dp, ifa);
1900 DAD_REMREF(dp); /* drop our reference */
1901}
1902
1903static void
1904nd6_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, daddr6_0: &saddr6, taddr6: &taddr6, ND_NA_FLAG_OVERRIDE, tlladdr: 1, NULL);
1931}
1932
1933static void
1934nd6_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(lck: &ndi->lock);
1998 retrans = ndi->retrans * hz / 1000;
1999 lck_mtx_unlock(lck: &ndi->lock);
2000 timeout((void (*)(void *))nd6_dad_timer, arg: (void *)ifa, ticks: 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(lck: &ndi->lock);
2030 retrans = ndi->retrans * hz / 1000;
2031 lck_mtx_unlock(lck: &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, arg: (void *)ifa, ticks: 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(lck: &ndi->lock);
2063 txunsolna = (ndi->flags & ND6_IFF_REPLICATED) != 0;
2064 lck_mtx_unlock(lck: &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 mac: 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
2095done:
2096 if (dp != NULL) {
2097 DAD_REMREF(dp); /* drop our reference */
2098 }
2099}
2100
2101static void
2102nd6_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 = SDL(llifa->ifa_addr);
2150 if (lladdrlen == sdl->sdl_alen &&
2151 bcmp(s1: lladdr, LLADDR(sdl), n: lladdrlen) == 0) {
2152 candisable = TRUE;
2153 }
2154 IFA_UNLOCK(llifa);
2155
2156 in6 = ia->ia_addr.sin6_addr;
2157 if (in6_iid_from_hw(ifp, &in6) != 0) {
2158 break;
2159 }
2160
2161 /* Refine decision about whether IPv6 can be disabled */
2162 if (candisable &&
2163 !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2164 /*
2165 * Apply this logic only to the embedded MAC
2166 * address form of link-local IPv6 address.
2167 */
2168 candisable = FALSE;
2169 } else if (lladdr == NULL &&
2170 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2171 /*
2172 * We received a NA with no target link-layer
2173 * address option. This means that someone else
2174 * has our address. Mark it as a hardware
2175 * duplicate so we disable IPv6 later on.
2176 */
2177 candisable = TRUE;
2178 }
2179 break;
2180 default:
2181 break;
2182 }
2183 }
2184 DAD_UNLOCK(dp);
2185
2186 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2187 ia->ia6_flags |= IN6_IFF_DUPLICATED;
2188 in6_event_enqueue_nwk_wq_entry(IN6_ADDR_MARKED_DUPLICATED,
2189 ia->ia_ifa.ifa_ifp, &ia->ia_addr.sin6_addr,
2190 0);
2191 IFA_UNLOCK(&ia->ia_ifa);
2192
2193 /* increment DAD collision counter */
2194 ++ip6stat.ip6s_dad_collide;
2195
2196 /* We are done with DAD, with duplicated address found. (failure) */
2197 untimeout((void (*)(void *))nd6_dad_timer, arg: (void *)ifa);
2198
2199 IFA_LOCK(&ia->ia_ifa);
2200 log(LOG_ERR, "%s: DAD complete for %s - duplicate found.\n",
2201 if_name(ifp), ip6_sprintf(&ia->ia_addr.sin6_addr));
2202 IFA_UNLOCK(&ia->ia_ifa);
2203
2204 if (candisable) {
2205 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
2206 log(LOG_ERR, "%s: possible hardware address duplication "
2207 "detected, disabling IPv6 for interface.\n", if_name(ifp));
2208
2209 VERIFY((NULL != ndi) && (TRUE == ndi->initialized));
2210 ndi->flags |= ND6_IFF_IFDISABLED;
2211 /* Make sure to set IFEF_IPV6_DISABLED too */
2212 nd6_if_disable(ifp, TRUE);
2213 }
2214
2215 log(LOG_ERR,
2216 "%s: manual intervention may be required.\n",
2217 if_name(ifp));
2218
2219 /* Send an event to the configuration agent so that the
2220 * duplicate address will be notified to the user and will
2221 * be removed.
2222 */
2223 in6_post_msg(ifp, KEV_INET6_NEW_USER_ADDR, ia, mac: dp->dad_lladdr);
2224 nd6_dad_detach(dp, ifa);
2225 DAD_REMREF(dp); /* drop our reference */
2226}
2227
2228static void
2229nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
2230{
2231 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
2232 struct ifnet *ifp = ifa->ifa_ifp;
2233 int i = 0;
2234 struct in6_addr taddr6;
2235
2236 DAD_LOCK(dp);
2237 dp->dad_ns_tcount++;
2238 if ((ifp->if_flags & IFF_UP) == 0) {
2239 DAD_UNLOCK(dp);
2240 return;
2241 }
2242 if ((ifp->if_flags & IFF_RUNNING) == 0) {
2243 DAD_UNLOCK(dp);
2244 return;
2245 }
2246
2247 dp->dad_ns_ocount++;
2248 DAD_UNLOCK(dp);
2249 IFA_LOCK_SPIN(&ia->ia_ifa);
2250 taddr6 = ia->ia_addr.sin6_addr;
2251 IFA_UNLOCK(&ia->ia_ifa);
2252 if (dad_enhanced != 0 && !(ifp->if_flags & IFF_POINTOPOINT)) {
2253 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) {
2254 dp->dad_nonce[i] = RandomULong();
2255 }
2256
2257 /*
2258 * XXXHRS: Note that in the case that
2259 * DupAddrDetectTransmits > 1, multiple NS messages with
2260 * different nonces can be looped back in an unexpected
2261 * order. The current implementation recognizes only
2262 * the latest nonce on the sender side. Practically it
2263 * should work well in almost all cases.
2264 */
2265 }
2266 nd6_ns_output(ifp, NULL, taddr6: &taddr6, NULL,
2267 nonce: (uint8_t *)&dp->dad_nonce[0]);
2268}
2269
2270/*
2271 * @brief Called to process DAD NS
2272 *
2273 * @param ifa is the pointer to the interface's address
2274 * @param lladdr is source link layer information
2275 * @param lladdrlen is source's linklayer length
2276 *
2277 * @return void
2278 */
2279static void
2280nd6_dad_ns_input(struct ifaddr *ifa, char *lladdr,
2281 int lladdrlen, struct nd_opt_nonce *ndopt_nonce)
2282{
2283 struct dadq *dp;
2284 VERIFY(ifa != NULL);
2285
2286 /* Ignore Nonce option when Enhanced DAD is disabled. */
2287 if (dad_enhanced == 0) {
2288 ndopt_nonce = NULL;
2289 }
2290
2291 dp = nd6_dad_find(ifa, nonce: ndopt_nonce);
2292 if (dp == NULL) {
2293 return;
2294 }
2295
2296 DAD_LOCK(dp);
2297 ++dp->dad_ns_icount;
2298 if (lladdr && lladdrlen >= ETHER_ADDR_LEN) {
2299 memcpy(dst: dp->dad_lladdr, src: lladdr, ETHER_ADDR_LEN);
2300 /* fine to truncate as it is compared against sdl_alen */
2301 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2302 }
2303 DAD_UNLOCK(dp);
2304 DAD_REMREF(dp);
2305}
2306
2307/*
2308 * @brief Called to process received NA for DAD
2309 *
2310 * @param m is the pointer to the packet's mbuf
2311 * @param ifp is the pointer to the interface on which packet
2312 * was receicved.
2313 * @param taddr is pointer to target's IPv6 address
2314 * @param lladdr is target's link layer information
2315 * @param lladdrlen is target's linklayer length
2316 *
2317 * @return NULL if the packet is consumed by DAD processing, else
2318 * pointer to the mbuf.
2319 */
2320static struct mbuf *
2321nd6_dad_na_input(struct mbuf *m, struct ifnet *ifp, struct in6_addr *taddr,
2322 caddr_t lladdr, int lladdrlen)
2323{
2324 struct ifaddr *ifa = NULL;
2325 struct in6_ifaddr *ia = NULL;
2326 struct dadq *dp = NULL;
2327 struct nd_ifinfo *ndi = NULL;
2328 boolean_t replicated;
2329
2330 ifa = (struct ifaddr *) in6ifa_ifpwithaddr(ifp, taddr);
2331 if (ifa == NULL) {
2332 return m;
2333 }
2334
2335 replicated = FALSE;
2336
2337 /* Get the ND6_IFF_REPLICATED flag. */
2338 ndi = ND_IFINFO(ifp);
2339 if (ndi != NULL && ndi->initialized) {
2340 lck_mtx_lock(lck: &ndi->lock);
2341 replicated = !!(ndi->flags & ND6_IFF_REPLICATED);
2342 lck_mtx_unlock(lck: &ndi->lock);
2343 }
2344
2345 if (replicated) {
2346 nd6log(info, "%s: ignoring duplicate NA on "
2347 "replicated interface %s\n", __func__, if_name(ifp));
2348 goto done;
2349 }
2350
2351 /* Lock the interface address until done (see label below). */
2352 IFA_LOCK(ifa);
2353 ia = (struct in6_ifaddr *) ifa;
2354
2355 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
2356 IFA_UNLOCK(ifa);
2357 nd6log(info, "%s: ignoring duplicate NA on "
2358 "%s [DAD not in progress]\n", __func__,
2359 if_name(ifp));
2360 goto done;
2361 }
2362
2363 /* Some sleep proxies improperly send the client's Ethernet address in
2364 * the target link-layer address option, so detect this by comparing
2365 * the L2-header source address, if we have seen it, with the target
2366 * address, and ignoring the NA if they don't match.
2367 */
2368 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2369 struct ip6aux *ip6a = ip6_findaux(m);
2370 if (ip6a && (ip6a->ip6a_flags & IP6A_HASEEN) != 0 &&
2371 bcmp(s1: ip6a->ip6a_ehsrc, s2: lladdr, ETHER_ADDR_LEN) != 0) {
2372 IFA_UNLOCK(ifa);
2373 nd6log(error, "%s: ignoring duplicate NA on %s "
2374 "[eh_src != tgtlladdr]\n", __func__, if_name(ifp));
2375 goto done;
2376 }
2377 }
2378
2379 IFA_UNLOCK(ifa);
2380
2381 dp = nd6_dad_find(ifa, NULL);
2382 if (dp == NULL) {
2383 nd6log(info, "%s: no DAD structure for %s on %s.\n",
2384 __func__, ip6_sprintf(taddr), if_name(ifp));
2385 goto done;
2386 }
2387
2388 DAD_LOCK_SPIN(dp);
2389 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2390 memcpy(dst: dp->dad_lladdr, src: lladdr, ETHER_ADDR_LEN);
2391 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2392 }
2393 dp->dad_na_icount++;
2394 DAD_UNLOCK(dp);
2395 DAD_REMREF(dp);
2396
2397 /* remove the address. */
2398 nd6log(info,
2399 "%s: duplicate IPv6 address %s [processing NA on %s]\n", __func__,
2400 ip6_sprintf(taddr), if_name(ifp));
2401done:
2402 IFA_LOCK_ASSERT_NOTHELD(ifa);
2403 ifa_remref(ifa);
2404 m_freem(m);
2405 return NULL;
2406}
2407
2408static void
2409dad_addref(struct dadq *dp, int locked)
2410{
2411 if (!locked) {
2412 DAD_LOCK_SPIN(dp);
2413 } else {
2414 DAD_LOCK_ASSERT_HELD(dp);
2415 }
2416
2417 if (++dp->dad_refcount == 0) {
2418 panic("%s: dad %p wraparound refcnt", __func__, dp);
2419 /* NOTREACHED */
2420 }
2421 if (!locked) {
2422 DAD_UNLOCK(dp);
2423 }
2424}
2425
2426static void
2427dad_remref(struct dadq *dp)
2428{
2429 struct ifaddr *ifa;
2430
2431 DAD_LOCK_SPIN(dp);
2432 if (dp->dad_refcount == 0) {
2433 panic("%s: dad %p negative refcnt", __func__, dp);
2434 }
2435 --dp->dad_refcount;
2436 if (dp->dad_refcount > 0) {
2437 DAD_UNLOCK(dp);
2438 return;
2439 }
2440 DAD_UNLOCK(dp);
2441
2442 if (dp->dad_attached ||
2443 dp->dad_list.tqe_next != NULL || dp->dad_list.tqe_prev != NULL) {
2444 panic("%s: attached dad=%p is being freed", __func__, dp);
2445 /* NOTREACHED */
2446 }
2447
2448 if ((ifa = dp->dad_ifa) != NULL) {
2449 ifa_remref(ifa); /* drop dad_ifa reference */
2450 dp->dad_ifa = NULL;
2451 }
2452
2453 lck_mtx_destroy(lck: &dp->dad_lock, grp: &ifa_mtx_grp);
2454 zfree(dad_zone, dp);
2455}
2456
2457void
2458nd6_llreach_set_reachable(struct ifnet *ifp, void *addr, unsigned int alen)
2459{
2460 /* Nothing more to do if it's disabled */
2461 if (nd6_llreach_base == 0) {
2462 return;
2463 }
2464
2465 ifnet_llreach_set_reachable(ifp, ETHERTYPE_IPV6, addr, alen);
2466}
2467
2468void
2469nd6_alt_node_addr_decompose(struct ifnet *ifp, struct sockaddr *sa,
2470 struct sockaddr_dl* sdl, struct sockaddr_in6 *sin6)
2471{
2472 static const size_t EUI64_LENGTH = 8;
2473
2474 VERIFY(nd6_need_cache(ifp));
2475 VERIFY(sa);
2476 VERIFY(sdl && (void *)sa != (void *)sdl);
2477 VERIFY(sin6 && (void *)sa != (void *)sin6);
2478
2479 SOCKADDR_ZERO(sin6, sizeof(*sin6));
2480 sin6->sin6_len = sizeof *sin6;
2481 sin6->sin6_family = AF_INET6;
2482
2483 SOCKADDR_ZERO(sdl, sizeof(*sdl));
2484 sdl->sdl_len = sizeof *sdl;
2485 sdl->sdl_family = AF_LINK;
2486 sdl->sdl_type = ifp->if_type;
2487 sdl->sdl_index = ifp->if_index;
2488 sdl->sdl_nlen = 0;
2489
2490 switch (sa->sa_family) {
2491 case AF_INET6: {
2492 struct sockaddr_in6 *sin6a = SIN6(sa);
2493 struct in6_addr *in6 = &sin6a->sin6_addr;
2494
2495 VERIFY(sa->sa_len == sizeof *sin6);
2496 if (in6->s6_addr[11] == 0xff && in6->s6_addr[12] == 0xfe) {
2497 sdl->sdl_alen = ETHER_ADDR_LEN;
2498 LLADDR(sdl)[0] = (in6->s6_addr[8] ^ ND6_EUI64_UBIT);
2499 LLADDR(sdl)[1] = in6->s6_addr[9];
2500 LLADDR(sdl)[2] = in6->s6_addr[10];
2501 LLADDR(sdl)[3] = in6->s6_addr[13];
2502 LLADDR(sdl)[4] = in6->s6_addr[14];
2503 LLADDR(sdl)[5] = in6->s6_addr[15];
2504 } else {
2505 sdl->sdl_alen = EUI64_LENGTH;
2506 bcopy(src: &in6->s6_addr[8], LLADDR(sdl), n: EUI64_LENGTH);
2507 }
2508
2509 sdl->sdl_slen = 0;
2510 break;
2511 }
2512 case AF_LINK: {
2513 struct sockaddr_dl *sdla = SDL(sa);
2514 struct in6_addr *in6 = &sin6->sin6_addr;
2515 caddr_t lla = LLADDR(sdla);
2516
2517 VERIFY(sa->sa_len <= sizeof(*sdl));
2518 SOCKADDR_COPY(sa, sdl, sa->sa_len);
2519
2520 sin6->sin6_scope_id = sdla->sdl_index;
2521 if (sin6->sin6_scope_id == 0) {
2522 sin6->sin6_scope_id = ifp->if_index;
2523 }
2524 in6->s6_addr[0] = 0xfe;
2525 in6->s6_addr[1] = 0x80;
2526 if (sdla->sdl_alen == EUI64_LENGTH) {
2527 bcopy(src: lla, dst: &in6->s6_addr[8], n: EUI64_LENGTH);
2528 } else {
2529 VERIFY(sdla->sdl_alen == ETHER_ADDR_LEN);
2530
2531 in6->s6_addr[8] = ((uint8_t) lla[0] ^ ND6_EUI64_UBIT);
2532 in6->s6_addr[9] = (uint8_t) lla[1];
2533 in6->s6_addr[10] = (uint8_t) lla[2];
2534 in6->s6_addr[11] = 0xff;
2535 in6->s6_addr[12] = 0xfe;
2536 in6->s6_addr[13] = (uint8_t) lla[3];
2537 in6->s6_addr[14] = (uint8_t) lla[4];
2538 in6->s6_addr[15] = (uint8_t) lla[5];
2539 }
2540
2541 break;
2542 }
2543 default:
2544 VERIFY(false);
2545 break;
2546 }
2547}
2548
2549int
2550nd6_alt_node_present(struct ifnet *ifp, struct sockaddr_in6 *sin6,
2551 struct sockaddr_dl *sdl, int32_t rssi, int lqm, int npm)
2552{
2553 struct rtentry *rt = NULL;
2554 struct llinfo_nd6 *ln = NULL;
2555 struct if_llreach *lr = NULL;
2556 int nd6_nc_updated = 0;
2557 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2558 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2559
2560 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2561 if (in6_embedded_scope) {
2562 if (temp_embedded_id == 0) {
2563 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2564 }
2565 } else if (temp_ifscope_id == 0) {
2566 sin6->sin6_scope_id = ifp->if_index;
2567 }
2568 }
2569
2570 nd6_cache_lladdr(ifp, &sin6->sin6_addr, LLADDR(sdl), sdl->sdl_alen,
2571 ND_NEIGHBOR_ADVERT, 0, &nd6_nc_updated);
2572
2573 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2574 lck_mtx_lock(rnh_lock);
2575
2576 rt = rtalloc1_scoped_locked(SA(sin6), 1, 0, ifp->if_index);
2577
2578 /* Restore the address that was passed to us */
2579 if (in6_embedded_scope) {
2580 if (temp_embedded_id == 0) {
2581 sin6->sin6_addr.s6_addr16[1] = 0;
2582 }
2583 } else if (temp_ifscope_id == 0) {
2584 sin6->sin6_scope_id = 0;
2585 }
2586
2587 if (rt != NULL) {
2588 RT_LOCK(rt);
2589 VERIFY(rt->rt_flags & RTF_LLINFO);
2590 VERIFY(rt->rt_llinfo);
2591
2592 ln = rt->rt_llinfo;
2593 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
2594 ln_setexpire(ln, 0);
2595
2596 lr = ln->ln_llreach;
2597 if (lr) {
2598 IFLR_LOCK(lr);
2599 lr->lr_rssi = rssi;
2600 lr->lr_lqm = (int32_t) lqm;
2601 lr->lr_npm = (int32_t) npm;
2602 IFLR_UNLOCK(lr);
2603 }
2604
2605 RT_UNLOCK(rt);
2606 RT_REMREF(rt);
2607 }
2608
2609 lck_mtx_unlock(rnh_lock);
2610
2611 if (rt == NULL) {
2612 log(LOG_ERR, "%s: failed to add/update host route to %s.\n",
2613 __func__, ip6_sprintf(&sin6->sin6_addr));
2614#if DEBUG || DEVELOPMENT
2615 if (ip6_p2p_debug) {
2616 panic("%s: failed to add/update host route to %s.\n",
2617 __func__, ip6_sprintf(&sin6->sin6_addr));
2618 }
2619#endif
2620 return EHOSTUNREACH;
2621 }
2622
2623 nd6log(debug, "%s: Successfully added/updated host route to %s [lr=0x%llx]\n",
2624 __func__, ip6_sprintf(&sin6->sin6_addr),
2625 (uint64_t)VM_KERNEL_ADDRPERM(lr));
2626 /*
2627 * nd6_nc_updated not set implies that nothing was updated
2628 * in the neighbor cache. Convey that as EEXIST to callers.
2629 */
2630 if (nd6_nc_updated == 0) {
2631 return EEXIST;
2632 }
2633 return 0;
2634}
2635
2636int
2637nd6_alt_node_absent(struct ifnet *ifp, struct sockaddr_in6 *sin6, struct sockaddr_dl *sdl)
2638{
2639 struct rtentry *rt = NULL;
2640 int error = 0;
2641 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2642 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2643
2644 nd6log(debug, "%s: host route to %s\n", __func__,
2645 ip6_sprintf(&sin6->sin6_addr));
2646
2647 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2648 if (in6_embedded_scope) {
2649 if (temp_embedded_id == 0) {
2650 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2651 }
2652 } else if (temp_ifscope_id == 0) {
2653 sin6->sin6_scope_id = ifp->if_index;
2654 }
2655 }
2656
2657 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2658 lck_mtx_lock(rnh_lock);
2659
2660 rt = rtalloc1_scoped_locked(SA(sin6), 0, 0, ifp->if_index);
2661
2662 /* Restore the address that was passed to us */
2663 if (in6_embedded_scope) {
2664 if (temp_embedded_id == 0) {
2665 sin6->sin6_addr.s6_addr16[1] = 0;
2666 }
2667 } else if (temp_ifscope_id == 0) {
2668 sin6->sin6_scope_id = 0;
2669 }
2670
2671 if (rt != NULL) {
2672 RT_LOCK(rt);
2673 if (IS_DYNAMIC_DIRECT_HOSTROUTE(rt)) {
2674 /*
2675 * Copy the link layer information in SDL when present
2676 * as it later gets used to issue the kernel event for
2677 * node absence.
2678 */
2679 if (sdl != NULL && rt->rt_gateway != NULL &&
2680 rt->rt_gateway->sa_family == AF_LINK &&
2681 SDL(rt->rt_gateway)->sdl_len <= sizeof(*sdl)) {
2682 SOCKADDR_COPY(rt->rt_gateway, sdl, SDL(rt->rt_gateway)->sdl_len);
2683 }
2684
2685 rt->rt_flags |= RTF_CONDEMNED;
2686 RT_UNLOCK(rt);
2687
2688 error = rtrequest_locked(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
2689 0, (struct rtentry **)NULL);
2690
2691 rtfree_locked(rt);
2692 } else {
2693 error = EHOSTUNREACH;
2694 RT_REMREF_LOCKED(rt);
2695 RT_UNLOCK(rt);
2696 }
2697 } else {
2698 error = EHOSTUNREACH;
2699 }
2700
2701 if (error == 0) {
2702 nd6log(debug, "%s: Successfully deleted host route to %s "
2703 "for interface %s.\n", __func__, ip6_sprintf(&sin6->sin6_addr),
2704 ifp->if_xname);
2705 } else {
2706 nd6log(error, "%s: Failed to delete host route to %s "
2707 "for interface %s with error :%d.\n", __func__,
2708 ip6_sprintf(&sin6->sin6_addr),
2709 ifp->if_xname, error);
2710 }
2711
2712 lck_mtx_unlock(rnh_lock);
2713 return error;
2714}
2715