1 /*
2 * Copyright (c) 2000-2025 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 #include <net/droptap.h>
86
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet6/in6_var.h>
90 #include <netinet6/in6_ifattach.h>
91 #include <netinet/ip6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet6/nd6.h>
94 #include <netinet6/scope6_var.h>
95 #include <netinet/icmp6.h>
96
97 #if IPSEC
98 #include <netinet6/ipsec.h>
99 #include <netinet6/ipsec6.h>
100 #endif
101
102 #include <net/sockaddr_utils.h>
103
104 struct dadq;
105 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
106 void nd6_dad_stoptimer(struct ifaddr *);
107 static void nd6_dad_timer(struct ifaddr *);
108 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
109 static void nd6_dad_ns_input(struct ifaddr *, char *__sized_by(lladdrlen), int lladdrlen, struct nd_opt_nonce *);
110 static struct mbuf *nd6_dad_na_input(struct mbuf *, struct ifnet *,
111 struct in6_addr *, caddr_t __counted_by(lladdrlen), int lladdrlen);
112 static void dad_addref(struct dadq *, int);
113 static void dad_remref(struct dadq *);
114 static struct dadq *nd6_dad_attach(struct dadq *, struct ifaddr *);
115 static void nd6_dad_detach(struct dadq *, struct ifaddr *);
116 static void nd6_dad_duplicated(struct ifaddr *);
117
118 static int dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */
119
120 #define DAD_LOCK_ASSERT_HELD(_dp) \
121 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_OWNED)
122
123 #define DAD_LOCK_ASSERT_NOTHELD(_dp) \
124 LCK_MTX_ASSERT(&(_dp)->dad_lock, LCK_MTX_ASSERT_NOTOWNED)
125
126 #define DAD_LOCK(_dp) \
127 lck_mtx_lock(&(_dp)->dad_lock)
128
129 #define DAD_LOCK_SPIN(_dp) \
130 lck_mtx_lock_spin(&(_dp)->dad_lock)
131
132 #define DAD_CONVERT_LOCK(_dp) do { \
133 DAD_LOCK_ASSERT_HELD(_dp); \
134 lck_mtx_convert_spin(&(_dp)->dad_lock); \
135 } while (0)
136
137 #define DAD_UNLOCK(_dp) \
138 lck_mtx_unlock(&(_dp)->dad_lock)
139
140 #define DAD_ADDREF(_dp) \
141 dad_addref(_dp, 0)
142
143 #define DAD_ADDREF_LOCKED(_dp) \
144 dad_addref(_dp, 1)
145
146 #define DAD_REMREF(_dp) \
147 dad_remref(_dp)
148
149 static LCK_MTX_DECLARE_ATTR(dad6_mutex, &ip6_mutex_grp, &ip6_mutex_attr);
150
151 static struct sockaddr_in6 hostrtmask;
152
153 static int nd6_llreach_base = 30; /* seconds */
154 SYSCTL_DECL(_net_inet6_icmp6);
155 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_llreach_base,
156 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_llreach_base, 0,
157 "default ND6 link-layer reachability max lifetime (in seconds)");
158
159 int dad_enhanced = ND6_DAD_ENHANCED_DEFAULT;
160 SYSCTL_DECL(_net_inet6_ip6);
161 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_RW | CTLFLAG_LOCKED,
162 &dad_enhanced, 0,
163 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
164
165 static uint32_t nd6_dad_nonce_max_count = 3;
166 SYSCTL_UINT(_net_inet6_ip6, OID_AUTO, nd6_dad_nonce_max_count,
167 CTLFLAG_RW | CTLFLAG_LOCKED, &nd6_dad_nonce_max_count, 0, "Number of times to ignore same nonce for DAD");
168
169 #if DEBUG || DEVELOPMENT
170 static int ip6_p2p_debug = 0;
171 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, ip6_p2p_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
172 &ip6_p2p_debug, 0,
173 "Enable more instrumentation for IPv6 P2P use-case");
174 #endif
175
176 /*
177 * Obtain a link-layer source cache entry for the sender.
178 *
179 * NOTE: This is currently only for ND6/Ethernet.
180 */
181 void
nd6_llreach_alloc(struct rtentry * rt,struct ifnet * ifp,void * addr __sized_by (alen),unsigned int alen,boolean_t solicited)182 nd6_llreach_alloc(struct rtentry *rt, struct ifnet *ifp, void *addr __sized_by(alen),
183 unsigned int alen, boolean_t solicited)
184 {
185 struct llinfo_nd6 *__single ln = rt->rt_llinfo;
186
187 if (nd6_llreach_base != 0 &&
188 (ln->ln_expire != 0 || (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) &&
189 !(rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
190 ifp->if_addrlen == IF_LLREACH_MAXLEN && /* Ethernet */
191 alen == ifp->if_addrlen) {
192 struct if_llreach *__single lr;
193 const char *why = NULL, *type = "";
194
195 /* Become a regular mutex, just in case */
196 RT_CONVERT_LOCK(rt);
197
198 if ((lr = ln->ln_llreach) != NULL) {
199 type = (solicited ? "ND6 advertisement" :
200 "ND6 unsolicited announcement");
201 /*
202 * If target has changed, create a new record;
203 * otherwise keep existing record.
204 */
205 IFLR_LOCK(lr);
206 if (bcmp(addr, lr->lr_key.addr, alen) != 0) {
207 IFLR_UNLOCK(lr);
208 /* Purge any link-layer info caching */
209 VERIFY(rt->rt_llinfo_purge != NULL);
210 rt->rt_llinfo_purge(rt);
211 lr = NULL;
212 why = " for different target HW address; "
213 "using new llreach record";
214 } else {
215 lr->lr_probes = 0; /* reset probe count */
216 IFLR_UNLOCK(lr);
217 if (solicited) {
218 why = " for same target HW address; "
219 "keeping existing llreach record";
220 }
221 }
222 }
223
224 if (lr == NULL) {
225 lr = ln->ln_llreach = ifnet_llreach_alloc(ifp,
226 ETHERTYPE_IPV6, addr, alen, nd6_llreach_base);
227 if (lr != NULL) {
228 lr->lr_probes = 0; /* reset probe count */
229 if (why == NULL) {
230 why = "creating new llreach record";
231 }
232 }
233 }
234
235 if (nd6_debug && lr != NULL && why != NULL) {
236 char tmp[MAX_IPv6_STR_LEN];
237
238 nd6log2(debug, "%s: %s%s for %s iface=%s\n", __func__,
239 type, why, inet_ntop(AF_INET6,
240 &SIN6(rt_key(rt))->sin6_addr, tmp, sizeof(tmp)),
241 if_name(ifp));
242 }
243 }
244 }
245
246 void
nd6_llreach_use(struct llinfo_nd6 * ln)247 nd6_llreach_use(struct llinfo_nd6 *ln)
248 {
249 if (ln->ln_llreach != NULL) {
250 ln->ln_lastused = net_uptime();
251 }
252 }
253
254 /*
255 * Input a Neighbor Solicitation Message.
256 *
257 * Based on RFC 4861
258 * Based on RFC 4862 (duplicate address detection)
259 */
260 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)261 nd6_ns_input(
262 struct mbuf *m,
263 int off,
264 int icmp6len)
265 {
266 ifnet_ref_t ifp = m->m_pkthdr.rcvif;
267 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
268 struct nd_neighbor_solicit *nd_ns = NULL;
269 struct in6_addr saddr6 = ip6->ip6_src;
270 struct in6_addr daddr6 = ip6->ip6_dst;
271 uint32_t saddr_ifscope = IN6_IS_SCOPE_EMBED(&saddr6) ? ip6_input_getsrcifscope(m) : IFSCOPE_NONE;
272 struct in6_addr taddr6 = {};
273 struct in6_addr myaddr6 = {};
274 uint32_t myaddr_ifscope = IFSCOPE_NONE;
275 int lladdrlen = 0;
276 char *lladdr = NULL;
277 struct ifaddr *__single ifa = NULL;
278 int anycast = 0, proxy = 0, dadprogress = 0;
279 int tlladdr = 0;
280 union nd_opts ndopts = {};
281 struct sockaddr_dl proxydl = {};
282 boolean_t advrouter = FALSE;
283 boolean_t is_dad_probe = FALSE;
284 int oflgclr = 0;
285 uint32_t taddr_ifscope;
286
287 /* Expect 32-bit aligned data pointer on strict-align platforms */
288 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
289
290 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
291 ip6 = mtod(m, struct ip6_hdr *);
292 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
293 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
294
295 taddr6 = nd_ns->nd_ns_target;
296 if (in6_setscope(&taddr6, ifp, &taddr_ifscope) != 0) {
297 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
298 goto bad;
299 }
300
301 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
302 nd6log(error,
303 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
304 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
305 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
306 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_HLIM, NULL, 0);
307 goto bad;
308 }
309
310 is_dad_probe = IN6_IS_ADDR_UNSPECIFIED(&saddr6);
311 if (is_dad_probe) {
312 /* dst has to be a solicited node multicast address. */
313 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
314 /* don't check ifindex portion */
315 daddr6.s6_addr32[1] == 0 &&
316 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
317 daddr6.s6_addr8[12] == 0xff) {
318 ; /* good */
319 } else {
320 nd6log(info, "nd6_ns_input: bad DAD packet (wrong ip6 dst)\n");
321 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_DAD, NULL, 0);
322 goto bad;
323 }
324 } else if (!nd6_onlink_ns_rfc4861) {
325 struct sockaddr_in6 src_sa6;
326
327 /*
328 * According to recent IETF discussions, it is not a good idea
329 * to accept a NS from an address which would not be deemed
330 * to be a neighbor otherwise. This point is expected to be
331 * clarified in future revisions of the specification.
332 */
333 SOCKADDR_ZERO(&src_sa6, sizeof(src_sa6));
334 src_sa6.sin6_family = AF_INET6;
335 src_sa6.sin6_len = sizeof(src_sa6);
336 src_sa6.sin6_addr = saddr6;
337 if (!in6_embedded_scope) {
338 src_sa6.sin6_scope_id = saddr_ifscope;
339 }
340 if (!nd6_is_addr_neighbor(&src_sa6, ifp, 0)) {
341 nd6log(info, "nd6_ns_input: NS packet from non-neighbor\n");
342 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
343 goto bad;
344 }
345 }
346
347 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
348 nd6log(info, "nd6_ns_input: bad NS target (multicast)\n");
349 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
350 goto bad;
351 }
352
353 icmp6len -= sizeof(*nd_ns);
354
355 nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
356 if (nd6_options(&ndopts) < 0) {
357 nd6log(info, "nd6_ns_input: invalid ND option, ignored\n");
358 /* nd6_options have incremented stats */
359 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
360 goto bad;
361 }
362
363 if (ndopts.nd_opts_src_lladdr) {
364 ND_OPT_LLADDR(ndopts.nd_opts_src_lladdr, nd_opt_len, lladdr, lladdrlen);
365 }
366
367 if (is_dad_probe && lladdr) {
368 nd6log(info, "nd6_ns_input: bad DAD packet "
369 "(link-layer address option)\n");
370 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_DAD, NULL, 0);
371 goto bad;
372 }
373
374 /*
375 * Attaching target link-layer address to the NA?
376 * (RFC 2461 7.2.4)
377 *
378 * NS IP dst is unicast/anycast MUST NOT add
379 * NS IP dst is solicited-node multicast MUST add
380 *
381 * In implementation, we add target link-layer address by default.
382 * We do not add one in MUST NOT cases.
383 */
384 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) {
385 tlladdr = 0;
386 } else {
387 tlladdr = 1;
388 }
389
390 /*
391 * Target address (taddr6) must be either:
392 * (1) Valid unicast/anycast address for my receiving interface,
393 * (2) Unicast address for which I'm offering proxy service, or
394 * (3) "tentative" or "optimistic" address [DAD is in progress].
395 */
396 /* (1) and (3) check. */
397 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
398
399 /* (2) check. */
400 if (ifa == NULL) {
401 rtentry_ref_t rt;
402 struct sockaddr_in6 tsin6;
403
404 SOCKADDR_ZERO(&tsin6, sizeof tsin6);
405 tsin6.sin6_len = sizeof(struct sockaddr_in6);
406 tsin6.sin6_family = AF_INET6;
407 tsin6.sin6_addr = taddr6;
408
409 rt = rtalloc1_scoped(SA(&tsin6), 0, 0, ifp->if_index);
410
411 if (rt != NULL) {
412 RT_LOCK(rt);
413 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
414 rt->rt_gateway->sa_family == AF_LINK) {
415 /*
416 * proxy NDP for single entry
417 */
418 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
419 ifp, IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
420 if (ifa) {
421 proxy = 1;
422 proxydl = *SDL(rt->rt_gateway);
423 }
424 }
425 RT_UNLOCK(rt);
426 rtfree(rt);
427 }
428 }
429 if (ifa == NULL && ip6_forwarding && nd6_prproxy) {
430 /*
431 * Is the target address part of the prefix that is being
432 * proxied and installed on another interface?
433 */
434 ifa = (struct ifaddr *)in6ifa_prproxyaddr(&taddr6, taddr_ifscope);
435 }
436 if (ifa == NULL) {
437 /*
438 * We've got an NS packet, and we don't have that address
439 * assigned for us. We MUST silently ignore it on this
440 * interface, c.f. RFC 4861 7.2.3.
441 *
442 * Forwarding associated with NDPRF_PRPROXY may apply.
443 */
444 if (ip6_forwarding && nd6_prproxy) {
445 size_t noncelen = (ndopts.nd_opts_nonce == NULL) ? 0 :
446 ndopts.nd_opts_nonce->nd_opt_nonce_len << 3;
447 uint8_t *nonce = (ndopts.nd_opts_nonce == NULL) ? NULL :
448 __unsafe_forge_bidi_indexable(uint8_t *, ndopts.nd_opts_nonce->nd_opt_nonce, noncelen);
449
450 nd6_prproxy_ns_input(ifp, &saddr6, lladdr,
451 lladdrlen, &daddr6, &taddr6,
452 nonce, noncelen);
453 }
454 goto freeit;
455 }
456 IFA_LOCK(ifa);
457 myaddr6 = *IFA_IN6(ifa);
458 myaddr_ifscope = IFA_SIN6_SCOPE(ifa);
459 anycast = (ifatoia6(ifa))->ia6_flags & IN6_IFF_ANYCAST;
460 dadprogress =
461 (ifatoia6(ifa))->ia6_flags & IN6_IFF_DADPROGRESS;
462 if ((ifatoia6(ifa))->ia6_flags & IN6_IFF_DUPLICATED) {
463 IFA_UNLOCK(ifa);
464 goto freeit;
465 }
466 IFA_UNLOCK(ifa);
467
468 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
469 nd6log(info,
470 "nd6_ns_input: lladdrlen mismatch for %s "
471 "(if %d, NS packet %d)\n",
472 ip6_sprintf(&taddr6), ifp->if_addrlen, lladdrlen - 2);
473 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
474 goto bad;
475 }
476
477 if (in6_are_addr_equal_scoped(&myaddr6, &saddr6, myaddr_ifscope, saddr_ifscope)) {
478 nd6log(info,
479 "nd6_ns_input: duplicate IP6 address %s\n",
480 ip6_sprintf(&saddr6));
481 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
482 goto bad;
483 }
484
485 /*
486 * We have neighbor solicitation packet, with target address equals to
487 * one of my DAD in-progress addresses.
488 *
489 * src addr how to process?
490 * --- ---
491 * multicast of course, invalid (rejected in ip6_input)
492 * unicast somebody is doing address resolution
493 * unspec dup address detection
494 *
495 * The processing is defined in the "draft standard" RFC 4862 (and by
496 * RFC 4429, which is a "proposed standard" update to its obsolete
497 * predecessor, RFC 2462) The reason optimistic DAD is not included
498 * in RFC 4862 is entirely due to IETF procedural considerations.
499 */
500 if (dadprogress) {
501 /*
502 * If source address is unspecified address, it is for
503 * duplicate address detection.
504 *
505 * If not, the packet is for addess resolution;
506 * silently ignore it when not optimistic
507 *
508 * Per RFC 4429 the reply for an optimistic address must
509 * have the Override flag cleared
510 */
511 if (!is_dad_probe && (dadprogress & IN6_IFF_OPTIMISTIC) != 0) {
512 oflgclr = 1;
513 } else {
514 if (is_dad_probe) {
515 nd6_dad_ns_input(ifa, lladdr, lladdrlen, ndopts.nd_opts_nonce);
516 }
517
518 goto freeit;
519 }
520 }
521
522 /* Are we an advertising router on this interface? */
523 advrouter = (ifp->if_ipv6_router_mode != IPV6_ROUTER_MODE_DISABLED);
524
525 /*
526 * If the source address is unspecified address, entries must not
527 * be created or updated.
528 * It looks that sender is performing DAD. If I'm using the address,
529 * and it's a "preferred" address, i.e. not optimistic, then output NA
530 * toward all-node multicast address, to tell the sender that I'm using
531 * the address.
532 * S bit ("solicited") must be zero.
533 */
534 if (is_dad_probe) {
535 saddr6 = in6addr_linklocal_allnodes;
536 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
537 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
538 goto bad;
539 }
540 if ((dadprogress & IN6_IFF_OPTIMISTIC) == 0) {
541 nd6_na_output(ifp, &saddr6, &taddr6,
542 ((anycast || proxy || !tlladdr) ? 0 :
543 ND_NA_FLAG_OVERRIDE) | (advrouter ?
544 ND_NA_FLAG_ROUTER : 0), tlladdr, proxy ?
545 SA(&proxydl) : NULL);
546 }
547 goto freeit;
548 }
549
550 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
551 ND_NEIGHBOR_SOLICIT, 0, NULL);
552
553 nd6_na_output(ifp, &saddr6, &taddr6,
554 ((anycast || proxy || !tlladdr || oflgclr) ? 0 : ND_NA_FLAG_OVERRIDE) |
555 (advrouter ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
556 tlladdr, proxy ? SA(&proxydl) : NULL);
557 freeit:
558 m_freem(m);
559 if (ifa != NULL) {
560 ifa_remref(ifa);
561 }
562 return;
563
564 bad:
565 nd6log(error, "nd6_ns_input: src=%s\n", ip6_sprintf(&saddr6));
566 nd6log(error, "nd6_ns_input: dst=%s\n", ip6_sprintf(&daddr6));
567 nd6log(error, "nd6_ns_input: tgt=%s\n", ip6_sprintf(&taddr6));
568 icmp6stat.icp6s_badns++;
569 if (ifa != NULL) {
570 ifa_remref(ifa);
571 }
572 }
573
574 /*
575 * Output a Neighbor Solicitation Message. Caller specifies:
576 * - ICMP6 header source IP6 address
577 * - ND6 header target IP6 address
578 * - ND6 header source datalink address
579 *
580 * Based on RFC 4861
581 * Based on RFC 4862 (duplicate address detection)
582 * Based on RFC 4429 (optimistic duplicate address detection)
583 *
584 * Caller must bump up ln->ln_rt refcnt to make sure 'ln' doesn't go
585 * away if there is a llinfo_nd6 passed in.
586 */
587 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * daddr6,const struct in6_addr * taddr6,struct llinfo_nd6 * ln,uint8_t * __counted_by (noncelen)nonce,size_t noncelen)588 nd6_ns_output(
589 struct ifnet *ifp,
590 const struct in6_addr *daddr6,
591 const struct in6_addr *taddr6,
592 struct llinfo_nd6 *ln, /* for source address determination */
593 uint8_t *__counted_by(noncelen) nonce,
594 size_t noncelen)
595 {
596 #pragma unused(noncelen)
597
598 mbuf_ref_t m;
599 struct ip6_hdr *ip6;
600 struct nd_neighbor_solicit *nd_ns;
601 struct in6_ifaddr *ia = NULL;
602 struct in6_addr *__single src, src_in, src_storage;
603 struct ip6_moptions *__single im6o = NULL;
604 ifnet_ref_t outif = NULL;
605 int icmp6len;
606 int maxlen;
607 int flags;
608 caddr_t mac;
609 struct route_in6 ro;
610 struct ip6_out_args ip6oa;
611 u_int32_t rtflags = 0;
612 boolean_t is_optimistic = FALSE;
613 drop_reason_t drop_reason = DROP_REASON_UNSPECIFIED;
614
615 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) || IN6_IS_ADDR_MULTICAST(taddr6)) {
616 return;
617 }
618
619 bzero(&ro, sizeof(ro));
620 bzero(&ip6oa, sizeof(ip6oa));
621 ip6oa.ip6oa_boundif = ifp->if_index;
622 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
623 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
624 IP6OAF_MANAGEMENT_ALLOWED | IP6OAF_ULTRA_CONSTRAINED_ALLOWED;
625 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
626 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
627
628 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
629
630 /* estimate the size of message */
631 maxlen = sizeof(*ip6) + sizeof(*nd_ns);
632 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
633 if (max_linkhdr + maxlen >= MCLBYTES) {
634 #if DIAGNOSTIC
635 printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
636 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
637 #endif
638 return;
639 }
640
641 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
642 if (m && max_linkhdr + maxlen >= MHLEN) {
643 MCLGET(m, M_DONTWAIT);
644 if ((m->m_flags & M_EXT) == 0) {
645 m_free(m);
646 m = NULL;
647 }
648 }
649 if (m == NULL) {
650 return;
651 }
652 m->m_pkthdr.rcvif = NULL;
653
654 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
655 m->m_flags |= M_MCAST;
656
657 im6o = ip6_allocmoptions(Z_NOWAIT);
658 if (im6o == NULL) {
659 m_freem(m);
660 return;
661 }
662
663 im6o->im6o_multicast_ifp = ifp;
664 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
665 im6o->im6o_multicast_loop = 0;
666 }
667
668 icmp6len = sizeof(*nd_ns);
669 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
670 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
671
672 /* fill neighbor solicitation packet */
673 ip6 = mtod(m, struct ip6_hdr *);
674 ip6->ip6_flow = 0;
675 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
676 ip6->ip6_vfc |= IPV6_VERSION;
677 /* ip6->ip6_plen will be set later */
678 ip6->ip6_nxt = IPPROTO_ICMPV6;
679 ip6->ip6_hlim = IPV6_MAXHLIM;
680 if (daddr6) {
681 ip6->ip6_dst = *daddr6;
682 ip6_output_setdstifscope(m, ifp->if_index, NULL);
683 } else {
684 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
685 ip6->ip6_dst.s6_addr16[1] = 0;
686 ip6->ip6_dst.s6_addr32[1] = 0;
687 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
688 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
689 ip6->ip6_dst.s6_addr8[12] = 0xff;
690 ip6_output_setdstifscope(m, ifp->if_index, NULL);
691 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) {
692 drop_reason = DROP_REASON_IP6_BAD_SCOPE;
693 goto bad;
694 }
695 }
696 if (nonce == NULL) {
697 /*
698 * RFC2461 7.2.2:
699 * "If the source address of the packet prompting the
700 * solicitation is the same as one of the addresses assigned
701 * to the outgoing interface, that address SHOULD be placed
702 * in the IP Source Address of the outgoing solicitation.
703 * Otherwise, any one of the addresses assigned to the
704 * interface should be used."
705 *
706 * We use the source address for the prompting packet
707 * (saddr6), if:
708 * - saddr6 is given from the caller (by giving "ln"), and
709 * - saddr6 belongs to the outgoing interface.
710 * Otherwise, we perform the source address selection as usual.
711 */
712 struct ip6_hdr *__single hip6; /* hold ip6 */
713 struct in6_addr hsrc = {};
714
715 /* Caller holds ref on this route */
716 if (ln != NULL) {
717 RT_LOCK(ln->ln_rt);
718 /*
719 * assuming every packet in ln_hold has the same IP
720 * header
721 */
722 if (ln->ln_hold != NULL) {
723 hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
724 /* XXX pullup? */
725 if (sizeof(*hip6) < ln->ln_hold->m_len) {
726 memcpy(&hsrc, &hip6->ip6_src, sizeof(struct in6_addr));
727 }
728 }
729 /* Update probe count, if applicable */
730 if (ln->ln_llreach != NULL) {
731 IFLR_LOCK_SPIN(ln->ln_llreach);
732 ln->ln_llreach->lr_probes++;
733 IFLR_UNLOCK(ln->ln_llreach);
734 }
735 rtflags = ln->ln_rt->rt_flags;
736 RT_UNLOCK(ln->ln_rt);
737 }
738 if (!IN6_IS_ADDR_UNSPECIFIED(&hsrc) && (ia = in6ifa_ifpwithaddr(ifp, &hsrc)) &&
739 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0) {
740 src = &hsrc;
741 } else {
742 int error;
743 struct sockaddr_in6 dst_sa;
744
745 SOCKADDR_ZERO(&dst_sa, sizeof(dst_sa));
746 dst_sa.sin6_family = AF_INET6;
747 dst_sa.sin6_len = sizeof(dst_sa);
748 dst_sa.sin6_addr = ip6->ip6_dst;
749
750 src = in6_selectsrc(&dst_sa, NULL,
751 NULL, &ro, NULL, &src_storage, ip6oa.ip6oa_boundif,
752 &error);
753 if (src == NULL) {
754 nd6log(info,
755 "nd6_ns_output: source can't be "
756 "determined: dst=%s, error=%d\n",
757 ip6_sprintf(&dst_sa.sin6_addr),
758 error);
759 drop_reason = DROP_REASON_IP_SRC_ADDR_NO_AVAIL;
760 goto bad;
761 }
762
763 if (ia != NULL) {
764 ifa_remref(&ia->ia_ifa);
765 ia = NULL;
766 }
767 /*
768 * RFC 4429 section 3.2:
769 * When a node has a unicast packet to send
770 * from an Optimistic Address to a neighbor,
771 * but does not know the neighbor's link-layer
772 * address, it MUST NOT perform Address
773 * Resolution.
774 */
775 ia = in6ifa_ifpwithaddr(ifp, src);
776 if (ia == NULL) {
777 nd6log(info,
778 "nd6_ns_output: no preferred source "
779 "available: dst=%s\n",
780 ip6_sprintf(&dst_sa.sin6_addr));
781 drop_reason = DROP_REASON_IP_SRC_ADDR_NO_AVAIL;
782 goto bad;
783 }
784 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
785 is_optimistic = TRUE;
786 nd6log(info,
787 "nd6_ns_output: preferred source "
788 "available is optimistic: dst=%s\n",
789 ip6_sprintf(&dst_sa.sin6_addr));
790 }
791 }
792 } else {
793 /*
794 * Source address for DAD packet must always be IPv6
795 * unspecified address. (0::0)
796 * We actually don't have to 0-clear the address (we did it
797 * above), but we do so here explicitly to make the intention
798 * clearer.
799 */
800 bzero(&src_in, sizeof(src_in));
801 src = &src_in;
802 ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_SRCADDR;
803 }
804
805 ip6->ip6_src = *src;
806 ip6_output_setsrcifscope(m, ifp->if_index, ia);
807 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
808 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
809 nd_ns->nd_ns_code = 0;
810 nd_ns->nd_ns_reserved = 0;
811 nd_ns->nd_ns_target = *taddr6;
812 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
813
814 /*
815 * Add source link-layer address option.
816 *
817 * spec implementation
818 * --- ---
819 * DAD packet MUST NOT do not add the option
820 * Source is optimistic MUST NOT do not add the option
821 * there's no link layer address:
822 * impossible do not add the option
823 * there's link layer address:
824 * Multicast NS MUST add one add the option
825 * Unicast NS SHOULD add one add the option
826 *
827 * XXX We deviate from RFC 4429 and still use optimistic DAD as source
828 * for address resolution. However to ensure that we do not interfere
829 * with neighbor cache entries of other neighbors, we MUST ensure
830 * that SLLAO is not sent. Also note, sending multicast NS without SLLAO
831 * is also a deviation from RFC 4861.
832 */
833 if (nonce == NULL && (mac = nd6_ifptomac(ifp)) && !is_optimistic) {
834 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
835 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
836 /* 8 byte alignments... */
837 optlen = (optlen + 7) & ~7;
838
839 m->m_pkthdr.len += optlen;
840 m->m_len += optlen;
841 icmp6len += optlen;
842 bzero((caddr_t)nd_opt, optlen);
843 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
844 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
845 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
846 }
847 /*
848 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
849 * This behavior is documented as Enhanced Duplicate Address
850 * Detection in draft-ietf-6man-enhanced-dad-13.
851 * net.inet6.ip6.dad_enhanced=0 disables this.
852 */
853 if (dad_enhanced != 0 && nonce != NULL && !(ifp->if_flags & IFF_POINTOPOINT)) {
854 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
855 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
856 /* 8-byte alignment is required. */
857 optlen = (optlen + 7) & ~7;
858
859 m->m_pkthdr.len += optlen;
860 m->m_len += optlen;
861 icmp6len += optlen;
862 bzero((caddr_t)nd_opt, optlen);
863 nd_opt->nd_opt_type = ND_OPT_NONCE;
864 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
865 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
866 }
867 ip6->ip6_plen = htons((u_short)icmp6len);
868 nd_ns->nd_ns_cksum = 0;
869 nd_ns->nd_ns_cksum
870 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
871
872 flags = nonce ? IPV6_UNSPECSRC : 0;
873 flags |= IPV6_OUTARGS;
874
875 /*
876 * PKTF_{INET,INET6}_RESOLVE_RTR are mutually exclusive, so make
877 * sure only one of them is set (just in case.)
878 */
879 m->m_pkthdr.pkt_flags &= ~(PKTF_INET_RESOLVE | PKTF_RESOLVE_RTR);
880 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
881 /*
882 * If this is a NS for resolving the (default) router, mark
883 * the packet accordingly so that the driver can find out,
884 * in case it needs to perform driver-specific action(s).
885 */
886 if (rtflags & RTF_ROUTER) {
887 m->m_pkthdr.pkt_flags |= PKTF_RESOLVE_RTR;
888 }
889
890 if (ifp->if_eflags & IFEF_TXSTART) {
891 /*
892 * Use control service class if the interface
893 * supports transmit-start model
894 */
895 (void) m_set_service_class(m, MBUF_SC_CTL);
896 }
897
898 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
899 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
900 ip6_output(m, NULL, NULL, flags, im6o, &outif, &ip6oa);
901 if (outif) {
902 icmp6_ifstat_inc(outif, ifs6_out_msg);
903 icmp6_ifstat_inc(outif, ifs6_out_neighborsolicit);
904 ifnet_release(outif);
905 }
906 icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
907
908 exit:
909 if (im6o != NULL) {
910 IM6O_REMREF(im6o);
911 }
912
913 ROUTE_RELEASE(&ro); /* we don't cache this route. */
914
915 if (ia != NULL) {
916 ifa_remref(&ia->ia_ifa);
917 }
918 return;
919
920 bad:
921 m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
922 goto exit;
923 }
924
925 /*
926 * Neighbor advertisement input handling.
927 *
928 * Based on RFC 4861
929 * Based on RFC 4862 (duplicate address detection)
930 *
931 * the following items are not implemented yet:
932 * - anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD)
933 * - proxy advertisement delay rule (RFC 4861 7.2.8, last paragraph, "should")
934 */
935 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)936 nd6_na_input(struct mbuf *m, int off, int icmp6len)
937 {
938 ifnet_ref_t ifp = m->m_pkthdr.rcvif;
939 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
940 struct nd_neighbor_advert *nd_na;
941 struct in6_addr saddr6 = ip6->ip6_src;
942 struct in6_addr daddr6 = ip6->ip6_dst;
943 struct in6_addr taddr6;
944 int flags;
945 int is_router;
946 int is_solicited;
947 int is_override;
948 char *lladdr = NULL;
949 int lladdrlen = 0;
950 struct llinfo_nd6 *__single ln;
951 rtentry_ref_t rt;
952 struct sockaddr_dl *sdl;
953 union nd_opts ndopts;
954 uint64_t timenow;
955 bool send_nc_alive_kev = false;
956 drop_reason_t drop_reason = DROP_REASON_UNSPECIFIED;
957
958 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
959 nd6log(info, "nd6_na_input: on ND6ALT interface!\n");
960 drop_reason = DROP_REASON_IP6_NO_ND6ALT_IF;
961 goto freeit;
962 }
963
964 /* Expect 32-bit aligned data pointer on strict-align platforms */
965 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
966
967 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
968 nd6log(error,
969 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
970 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
971 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
972 drop_reason = DROP_REASON_IP6_BAD_HLIM;
973 goto bad;
974 }
975
976 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
977 ip6 = mtod(m, struct ip6_hdr *);
978 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
979 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
980
981 flags = nd_na->nd_na_flags_reserved;
982 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
983 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
984 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
985
986 taddr6 = nd_na->nd_na_target;
987 if (in6_setscope(&taddr6, ifp, NULL)) {
988 drop_reason = DROP_REASON_IP6_BAD_SCOPE;
989 goto bad; /* XXX: impossible */
990 }
991 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
992 nd6log(error,
993 "nd6_na_input: invalid target address %s\n",
994 ip6_sprintf(&taddr6));
995 drop_reason = DROP_REASON_IP_DST_ADDR_NO_AVAIL;
996 goto bad;
997 }
998 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
999 if (is_solicited) {
1000 nd6log(error,
1001 "nd6_na_input: a solicited adv is multicasted\n");
1002 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1003 goto bad;
1004 }
1005 }
1006
1007 icmp6len -= sizeof(*nd_na);
1008 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
1009 if (nd6_options(&ndopts) < 0) {
1010 nd6log(info,
1011 "nd6_na_input: invalid ND option, ignored\n");
1012 /* nd6_options have incremented stats */
1013 drop_reason = DROP_REASON_IP6_TOO_MANY_OPTIONS;
1014 goto freeit;
1015 }
1016
1017 if (ndopts.nd_opts_tgt_lladdr) {
1018 ND_OPT_LLADDR(ndopts.nd_opts_tgt_lladdr, nd_opt_len, lladdr, lladdrlen);
1019
1020 if (((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
1021 nd6log(info,
1022 "nd6_na_input: lladdrlen mismatch for %s "
1023 "(if %d, NA packet %d)\n",
1024 ip6_sprintf(&taddr6), ifp->if_addrlen,
1025 lladdrlen - 2);
1026 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1027 goto bad;
1028 }
1029 }
1030
1031 m = nd6_dad_na_input(m, ifp, &taddr6, lladdr, lladdrlen);
1032 if (m == NULL) {
1033 return;
1034 }
1035
1036 /* Forwarding associated with NDPRF_PRPROXY may apply. */
1037 if (ip6_forwarding && nd6_prproxy) {
1038 nd6_prproxy_na_input(ifp, &saddr6, &daddr6, &taddr6, flags);
1039 }
1040
1041 /*
1042 * If no neighbor cache entry is found, NA SHOULD silently be
1043 * discarded. If we are forwarding (and Scoped Routing is in
1044 * effect), try to see if there is a neighbor cache entry on
1045 * another interface (in case we are doing prefix proxying.)
1046 */
1047 if ((rt = nd6_lookup(&taddr6, 0, ifp, 0)) == NULL) {
1048 if (!ip6_forwarding || !nd6_prproxy) {
1049 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1050 goto freeit;
1051 }
1052
1053 if ((rt = nd6_lookup(&taddr6, 0, NULL, 0)) == NULL) {
1054 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1055 goto freeit;
1056 }
1057
1058 RT_LOCK_ASSERT_HELD(rt);
1059 if (rt->rt_ifp != ifp) {
1060 /*
1061 * Purge any link-layer info caching.
1062 */
1063 if (rt->rt_llinfo_purge != NULL) {
1064 rt->rt_llinfo_purge(rt);
1065 }
1066
1067 /* Adjust route ref count for the interfaces */
1068 if (rt->rt_if_ref_fn != NULL) {
1069 rt->rt_if_ref_fn(ifp, 1);
1070 rt->rt_if_ref_fn(rt->rt_ifp, -1);
1071 }
1072
1073 /* Change the interface when the existing route is on */
1074 rt->rt_ifp = ifp;
1075
1076 /*
1077 * If rmx_mtu is not locked, update it
1078 * to the MTU used by the new interface.
1079 */
1080 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1081 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1082 }
1083 }
1084 }
1085
1086 RT_LOCK_ASSERT_HELD(rt);
1087 if ((ln = rt->rt_llinfo) == NULL ||
1088 (sdl = SDL(rt->rt_gateway)) == NULL) {
1089 RT_REMREF_LOCKED(rt);
1090 RT_UNLOCK(rt);
1091 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1092 goto freeit;
1093 }
1094
1095 timenow = net_uptime();
1096
1097 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1098 /*
1099 * If the link-layer has address, and no lladdr option came,
1100 * discard the packet.
1101 */
1102 if (ifp->if_addrlen && !lladdr) {
1103 RT_REMREF_LOCKED(rt);
1104 RT_UNLOCK(rt);
1105 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1106 goto freeit;
1107 }
1108
1109 /*
1110 * Record link-layer address, and update the state.
1111 */
1112 sdl->sdl_alen = ifp->if_addrlen;
1113 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1114 if (is_solicited) {
1115 send_nc_alive_kev = (rt->rt_flags & RTF_ROUTER) ? true : false;
1116 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1117 if (ln->ln_expire != 0) {
1118 struct nd_ifinfo *__single ndi = NULL;
1119
1120 ndi = ND_IFINFO(rt->rt_ifp);
1121 VERIFY(ndi != NULL && ndi->initialized);
1122 lck_mtx_lock(&ndi->lock);
1123 ln_setexpire(ln, timenow + ndi->reachable);
1124 lck_mtx_unlock(&ndi->lock);
1125 RT_UNLOCK(rt);
1126 lck_mtx_lock(rnh_lock);
1127 nd6_sched_timeout(NULL, NULL);
1128 lck_mtx_unlock(rnh_lock);
1129 RT_LOCK(rt);
1130 }
1131 } else {
1132 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1133 ln_setexpire(ln, timenow + nd6_gctimer);
1134 }
1135
1136 /*
1137 * Enqueue work item to invoke callback for this
1138 * route entry
1139 */
1140 route_event_enqueue_nwk_wq_entry(rt, NULL,
1141 ROUTE_LLENTRY_RESOLVED, NULL, TRUE);
1142
1143 if ((ln->ln_router = (short)is_router) != 0) {
1144 struct radix_node_head *__single rnh = NULL;
1145 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1146 ifnet_ref_t rt_ifp = rt->rt_ifp;
1147
1148 struct route_event rt_ev;
1149 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_RESOLVED);
1150 /*
1151 * This means a router's state has changed from
1152 * non-reachable to probably reachable, and might
1153 * affect the status of associated prefixes..
1154 * We already have a reference on rt. Don't need to
1155 * take one for the unlock/lock.
1156 */
1157 RT_UNLOCK(rt);
1158 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1159 lck_mtx_lock(rnh_lock);
1160 rnh = rt_tables[AF_INET6];
1161
1162 if (rnh != NULL) {
1163 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1164 (void *)&rt_ev);
1165 }
1166 lck_mtx_unlock(rnh_lock);
1167 lck_mtx_lock(nd6_mutex);
1168 pfxlist_onlink_check();
1169 lck_mtx_unlock(nd6_mutex);
1170 RT_LOCK(rt);
1171 }
1172 } else {
1173 int llchange = 0;
1174
1175 /*
1176 * Check if the link-layer address has changed or not.
1177 */
1178 if (lladdr == NULL) {
1179 llchange = 0;
1180 } else {
1181 if (sdl->sdl_alen) {
1182 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) {
1183 llchange = 1;
1184 } else {
1185 llchange = 0;
1186 }
1187 } else {
1188 llchange = 1;
1189 }
1190 }
1191
1192 /*
1193 * This is VERY complex. Look at it with care.
1194 *
1195 * override solicit lladdr llchange action
1196 * (L: record lladdr)
1197 *
1198 * 0 0 n -- (2c)
1199 * 0 0 y n (2b) L
1200 * 0 0 y y (1) REACHABLE->STALE
1201 * 0 1 n -- (2c) *->REACHABLE
1202 * 0 1 y n (2b) L *->REACHABLE
1203 * 0 1 y y (1) REACHABLE->STALE
1204 * 1 0 n -- (2a)
1205 * 1 0 y n (2a) L
1206 * 1 0 y y (2a) L *->STALE
1207 * 1 1 n -- (2a) *->REACHABLE
1208 * 1 1 y n (2a) L *->REACHABLE
1209 * 1 1 y y (2a) L *->REACHABLE
1210 */
1211 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
1212 /*
1213 * If state is REACHABLE, make it STALE.
1214 * no other updates should be done.
1215 */
1216 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
1217 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1218 ln_setexpire(ln, timenow + nd6_gctimer);
1219 }
1220 RT_REMREF_LOCKED(rt);
1221 RT_UNLOCK(rt);
1222 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1223 goto freeit;
1224 } else if (is_override /* (2a) */
1225 || (!is_override && (lladdr && !llchange)) /* (2b) */
1226 || !lladdr) { /* (2c) */
1227 /*
1228 * Update link-local address, if any.
1229 */
1230 if (lladdr) {
1231 sdl->sdl_alen = ifp->if_addrlen;
1232 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1233 }
1234
1235 /*
1236 * If solicited, make the state REACHABLE.
1237 * If not solicited and the link-layer address was
1238 * changed, make it STALE.
1239 */
1240 if (is_solicited) {
1241 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1242 if (ln->ln_expire != 0) {
1243 struct nd_ifinfo *__single ndi = NULL;
1244
1245 ndi = ND_IFINFO(ifp);
1246 VERIFY(ndi != NULL && ndi->initialized);
1247 lck_mtx_lock(&ndi->lock);
1248 ln_setexpire(ln,
1249 timenow + ndi->reachable);
1250 lck_mtx_unlock(&ndi->lock);
1251 RT_UNLOCK(rt);
1252 lck_mtx_lock(rnh_lock);
1253 nd6_sched_timeout(NULL, NULL);
1254 lck_mtx_unlock(rnh_lock);
1255 RT_LOCK(rt);
1256 }
1257 } else {
1258 if (lladdr && llchange) {
1259 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1260 ln_setexpire(ln, timenow + nd6_gctimer);
1261 }
1262 }
1263
1264 /*
1265 * XXX
1266 * The above is somewhat convoluted, for now just
1267 * issue a callback for LLENTRY changed.
1268 */
1269 /* Enqueue work item to invoke callback for this route entry */
1270 if (llchange) {
1271 route_event_enqueue_nwk_wq_entry(rt, NULL,
1272 ROUTE_LLENTRY_CHANGED, NULL, TRUE);
1273 }
1274
1275 /*
1276 * If the router's link-layer address has changed,
1277 * notify routes using this as gateway so they can
1278 * update any cached information.
1279 */
1280 if (ln->ln_router && is_router && llchange) {
1281 struct radix_node_head *__single rnh = NULL;
1282 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1283 ifnet_ref_t rt_ifp = rt->rt_ifp;
1284 struct route_event rt_ev;
1285 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_CHANGED);
1286
1287 /*
1288 * This means a router's state has changed from
1289 * non-reachable to probably reachable, and might
1290 * affect the status of associated prefixes..
1291 *
1292 * We already have a valid rt reference here.
1293 * We don't need to take another one for unlock/lock.
1294 */
1295 RT_UNLOCK(rt);
1296 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1297 lck_mtx_lock(rnh_lock);
1298 rnh = rt_tables[AF_INET6];
1299
1300 if (rnh != NULL) {
1301 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1302 (void *)&rt_ev);
1303 }
1304 lck_mtx_unlock(rnh_lock);
1305 RT_LOCK(rt);
1306 }
1307 }
1308
1309 if (ln->ln_router && !is_router) {
1310 /*
1311 * The peer dropped the router flag.
1312 * Remove the sender from the Default Router List and
1313 * update the Destination Cache entries.
1314 */
1315 struct nd_defrouter *__single dr;
1316 struct in6_addr *__single in6;
1317 ifnet_ref_t rt_ifp = rt->rt_ifp;
1318
1319 in6 = &SIN6(rt_key(rt))->sin6_addr;
1320
1321 RT_UNLOCK(rt);
1322 lck_mtx_lock(nd6_mutex);
1323 /*
1324 * XXX Handle router lists for route information option
1325 * as well.
1326 */
1327 dr = defrouter_lookup(NULL, in6, rt_ifp);
1328 if (dr) {
1329 TAILQ_REMOVE(&nd_defrouter_list, dr, dr_entry);
1330 defrtrlist_del(dr, NULL);
1331 NDDR_REMREF(dr); /* remove list reference */
1332 NDDR_REMREF(dr);
1333 lck_mtx_unlock(nd6_mutex);
1334 } else {
1335 lck_mtx_unlock(nd6_mutex);
1336 /*
1337 * Even if the neighbor is not in the
1338 * default router list, the neighbor
1339 * may be used as a next hop for some
1340 * destinations (e.g. redirect case).
1341 * So we must call rt6_flush explicitly.
1342 */
1343 rt6_flush(&ip6->ip6_src, rt_ifp);
1344 }
1345 RT_LOCK(rt);
1346 }
1347 ln->ln_router = (short)is_router;
1348 }
1349
1350 if (send_nc_alive_kev && (ifp->if_addrlen == IF_LLREACH_MAXLEN)) {
1351 struct kev_msg ev_msg;
1352 struct kev_nd6_ndalive nd6_ndalive;
1353 bzero(&ev_msg, sizeof(ev_msg));
1354 bzero(&nd6_ndalive, sizeof(nd6_ndalive));
1355 ev_msg.vendor_code = KEV_VENDOR_APPLE;
1356 ev_msg.kev_class = KEV_NETWORK_CLASS;
1357 ev_msg.kev_subclass = KEV_ND6_SUBCLASS;
1358 ev_msg.event_code = KEV_ND6_NDALIVE;
1359
1360 nd6_ndalive.link_data.if_family = ifp->if_family;
1361 nd6_ndalive.link_data.if_unit = ifp->if_unit;
1362 strlcpy(nd6_ndalive.link_data.if_name,
1363 ifp->if_name,
1364 sizeof(nd6_ndalive.link_data.if_name));
1365 ev_msg.dv[0].data_ptr = &nd6_ndalive;
1366 ev_msg.dv[0].data_length =
1367 sizeof(nd6_ndalive);
1368 dlil_post_complete_msg(NULL, &ev_msg);
1369 }
1370
1371 RT_LOCK_ASSERT_HELD(rt);
1372 rt->rt_flags &= ~RTF_REJECT;
1373
1374 /* cache the gateway (sender HW) address */
1375 nd6_llreach_alloc(rt, ifp, LLADDR(sdl), sdl->sdl_alen, TRUE);
1376
1377 /* update the llinfo, send a queued packet if there is one */
1378 ln->ln_asked = 0;
1379 if (ln->ln_hold != NULL) {
1380 mbuf_ref_t m_hold, m_hold_next;
1381 struct sockaddr_in6 sin6;
1382
1383 rtkey_to_sa6(rt, &sin6);
1384 /*
1385 * reset the ln_hold in advance, to explicitly
1386 * prevent a ln_hold lookup in nd6_output()
1387 * (wouldn't happen, though...)
1388 */
1389 m_hold = ln->ln_hold;
1390 ln->ln_hold = NULL;
1391 for (; m_hold; m_hold = m_hold_next) {
1392 m_hold_next = m_hold->m_nextpkt;
1393 m_hold->m_nextpkt = NULL;
1394 /*
1395 * we assume ifp is not a loopback here, so just set
1396 * the 2nd argument as the 1st one.
1397 */
1398 RT_UNLOCK(rt);
1399 nd6_output(ifp, ifp, m_hold, &sin6, rt, NULL);
1400 RT_LOCK_SPIN(rt);
1401 }
1402 }
1403 RT_REMREF_LOCKED(rt);
1404 RT_UNLOCK(rt);
1405 m_freem(m);
1406 return;
1407
1408 bad:
1409 icmp6stat.icp6s_badna++;
1410 /* fall through */
1411
1412 freeit:
1413 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
1414 return;
1415 }
1416
1417 /*
1418 * Neighbor advertisement output handling.
1419 *
1420 * Based on RFC 2461
1421 *
1422 * the following items are not implemented yet:
1423 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
1424 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
1425 *
1426 * tlladdr - 1 if include target link-layer address
1427 * sdl0 - sockaddr_dl (= proxy NA) or NULL
1428 */
1429 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)1430 nd6_na_output(
1431 struct ifnet *ifp,
1432 const struct in6_addr *daddr6_0,
1433 const struct in6_addr *taddr6,
1434 uint32_t flags,
1435 int tlladdr, /* 1 if include target link-layer address */
1436 struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
1437 {
1438 mbuf_ref_t m;
1439 struct ip6_hdr *ip6;
1440 struct nd_neighbor_advert *nd_na;
1441 struct ip6_moptions *__single im6o = NULL;
1442 caddr_t mac = NULL;
1443 struct route_in6 ro;
1444 struct in6_addr *__single src;
1445 struct in6_addr src_storage, daddr6;
1446 struct in6_ifaddr *__single ia;
1447 struct sockaddr_in6 dst_sa;
1448 int icmp6len, maxlen, error;
1449 ifnet_ref_t outif = NULL;
1450
1451 struct ip6_out_args ip6oa;
1452 bzero(&ro, sizeof(ro));
1453
1454 daddr6 = *daddr6_0; /* make a local copy for modification */
1455
1456 bzero(&ip6oa, sizeof(ip6oa));
1457 ip6oa.ip6oa_boundif = ifp->if_index;
1458 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
1459 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
1460 IP6OAF_MANAGEMENT_ALLOWED | IP6OAF_ULTRA_CONSTRAINED_ALLOWED;
1461 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
1462 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1463
1464 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
1465
1466 /* estimate the size of message */
1467 maxlen = sizeof(*ip6) + sizeof(*nd_na);
1468 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
1469 if (max_linkhdr + maxlen >= MCLBYTES) {
1470 #if DIAGNOSTIC
1471 printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
1472 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
1473 #endif
1474 return;
1475 }
1476
1477 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
1478 if (m && max_linkhdr + maxlen >= MHLEN) {
1479 MCLGET(m, M_DONTWAIT);
1480 if ((m->m_flags & M_EXT) == 0) {
1481 m_free(m);
1482 m = NULL;
1483 }
1484 }
1485 if (m == NULL) {
1486 return;
1487 }
1488 m->m_pkthdr.rcvif = NULL;
1489
1490 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1491 m->m_flags |= M_MCAST;
1492
1493 im6o = ip6_allocmoptions(Z_NOWAIT);
1494 if (im6o == NULL) {
1495 m_freem(m);
1496 return;
1497 }
1498
1499 im6o->im6o_multicast_ifp = ifp;
1500 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
1501 im6o->im6o_multicast_loop = 0;
1502 }
1503
1504 icmp6len = sizeof(*nd_na);
1505 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1506 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
1507
1508 /* fill neighbor advertisement packet */
1509 ip6 = mtod(m, struct ip6_hdr *);
1510 ip6->ip6_flow = 0;
1511 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1512 ip6->ip6_vfc |= IPV6_VERSION;
1513 ip6->ip6_nxt = IPPROTO_ICMPV6;
1514 ip6->ip6_hlim = IPV6_MAXHLIM;
1515 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1516 /* reply to DAD */
1517 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1518 daddr6.s6_addr16[1] = 0;
1519 daddr6.s6_addr32[1] = 0;
1520 daddr6.s6_addr32[2] = 0;
1521 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1522 if (in6_setscope(&daddr6, ifp, NULL)) {
1523 m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1524 goto exit;
1525 }
1526
1527 flags &= ~ND_NA_FLAG_SOLICITED;
1528 } else {
1529 ip6->ip6_dst = daddr6;
1530 ip6_output_setdstifscope(m, ifp->if_index, NULL);
1531 }
1532
1533 SOCKADDR_ZERO(&dst_sa, sizeof(struct sockaddr_in6));
1534 dst_sa.sin6_family = AF_INET6;
1535 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1536 dst_sa.sin6_addr = daddr6;
1537
1538 /*
1539 * Select a source whose scope is the same as that of the dest.
1540 */
1541 SOCKADDR_COPY(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1542 src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &src_storage,
1543 ip6oa.ip6oa_boundif, &error);
1544 if (src == NULL) {
1545 nd6log(info, "nd6_na_output: source can't be "
1546 "determined: dst=%s, error=%d\n",
1547 ip6_sprintf(&dst_sa.sin6_addr), error);
1548 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
1549 goto exit;
1550 }
1551 ip6->ip6_src = *src;
1552
1553 /*
1554 * RFC 4429 requires not setting "override" flag on NA packets sent
1555 * from optimistic addresses.
1556 */
1557 ia = in6ifa_ifpwithaddr(ifp, src);
1558 ip6_output_setsrcifscope(m, ifp->if_index, ia);
1559 if (ia != NULL) {
1560 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
1561 flags &= ~ND_NA_FLAG_OVERRIDE;
1562 }
1563 ifa_remref(&ia->ia_ifa);
1564 }
1565
1566 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1567 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1568 nd_na->nd_na_code = 0;
1569 nd_na->nd_na_target = *taddr6;
1570 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1571
1572 /*
1573 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1574 * see nd6_ns_input() for details.
1575 * Basically, if NS packet is sent to unicast/anycast addr,
1576 * target lladdr option SHOULD NOT be included.
1577 */
1578 if (tlladdr) {
1579 /*
1580 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1581 * lladdr in sdl0. If we are not proxying (sending NA for
1582 * my address) use lladdr configured for the interface.
1583 */
1584 if (sdl0 == NULL) {
1585 mac = nd6_ifptomac(ifp);
1586 } else if (sdl0->sa_family == AF_LINK) {
1587 struct sockaddr_dl *sdl;
1588 sdl = SDL(sdl0);
1589 if (sdl->sdl_alen == ifp->if_addrlen) {
1590 mac = LLADDR(sdl);
1591 }
1592 }
1593 }
1594 if (tlladdr && mac) {
1595 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1596 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1597
1598 /* roundup to 8 bytes alignment! */
1599 optlen = (optlen + 7) & ~7;
1600
1601 m->m_pkthdr.len += optlen;
1602 m->m_len += optlen;
1603 icmp6len += optlen;
1604 bzero((caddr_t)nd_opt, optlen);
1605 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1606 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
1607 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1608 } else {
1609 flags &= ~ND_NA_FLAG_OVERRIDE;
1610 }
1611
1612 ip6->ip6_plen = htons((u_short)icmp6len);
1613 nd_na->nd_na_flags_reserved = flags;
1614 nd_na->nd_na_cksum = 0;
1615 nd_na->nd_na_cksum =
1616 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1617
1618 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
1619
1620 if (ifp->if_eflags & IFEF_TXSTART) {
1621 /* Use control service class if the interface supports
1622 * transmit-start model.
1623 */
1624 (void) m_set_service_class(m, MBUF_SC_CTL);
1625 }
1626
1627 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
1628 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
1629 ip6_output(m, NULL, NULL, IPV6_OUTARGS, im6o, &outif, &ip6oa);
1630 if (outif) {
1631 icmp6_ifstat_inc(outif, ifs6_out_msg);
1632 icmp6_ifstat_inc(outif, ifs6_out_neighboradvert);
1633 ifnet_release(outif);
1634 }
1635 icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1636
1637 exit:
1638 if (im6o != NULL) {
1639 IM6O_REMREF(im6o);
1640 }
1641
1642 ROUTE_RELEASE(&ro);
1643 }
1644
1645 TAILQ_HEAD(dadq_head, dadq);
1646 struct dadq {
1647 decl_lck_mtx_data(, dad_lock);
1648 u_int32_t dad_refcount; /* reference count */
1649 int dad_attached;
1650 TAILQ_ENTRY(dadq) dad_list;
1651 struct ifaddr *dad_ifa;
1652 int dad_count; /* max NS to send */
1653 int dad_ns_tcount; /* # of trials to send NS */
1654 int dad_ns_ocount; /* NS sent so far */
1655 int dad_ns_icount;
1656 int dad_na_icount;
1657 int dad_ns_lcount; /* looped back NS */
1658 int dad_loopbackprobe; /* probing state for loopback detection */
1659 uint8_t dad_lladdr[ETHER_ADDR_LEN];
1660 uint8_t dad_lladdrlen;
1661 #define ND_OPT_NONCE_LEN32 \
1662 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1663 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1664 uint32_t dad_same_nonce_count; /* # of consecutive times we've ignored DAD failure because of optimistic DAD */
1665 };
1666
1667 static KALLOC_TYPE_DEFINE(dad_zone, struct dadq, NET_KT_DEFAULT);
1668 static struct dadq_head dadq;
1669
1670 void
nd6_nbr_init(void)1671 nd6_nbr_init(void)
1672 {
1673 int i;
1674
1675 TAILQ_INIT(&dadq);
1676
1677 SOCKADDR_ZERO(&hostrtmask, sizeof hostrtmask);
1678 hostrtmask.sin6_family = AF_INET6;
1679 hostrtmask.sin6_len = sizeof hostrtmask;
1680 for (i = 0; i < sizeof hostrtmask.sin6_addr; ++i) {
1681 hostrtmask.sin6_addr.s6_addr[i] = 0xff;
1682 }
1683 }
1684
1685 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * nonce)1686 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1687 {
1688 struct dadq *__single dp;
1689 boolean_t same_nonce = false;
1690
1691 lck_mtx_lock(&dad6_mutex);
1692 for (dp = dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1693 DAD_LOCK_SPIN(dp);
1694 if (dp->dad_ifa != ifa) {
1695 DAD_UNLOCK(dp);
1696 continue;
1697 }
1698
1699 /*
1700 * Skip if the nonce matches the received one.
1701 * +2 in the length is required because of type and
1702 * length fields are included in a header.
1703 */
1704 same_nonce = nonce != NULL &&
1705 nonce->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1706 memcmp(&nonce->nd_opt_nonce[0], &dp->dad_nonce[0],
1707 ND_OPT_NONCE_LEN) == 0;
1708
1709 if (same_nonce &&
1710 dp->dad_same_nonce_count <= nd6_dad_nonce_max_count) {
1711 nd6log(error, "%s: a looped back NS message is "
1712 "detected during DAD for if=%s %s. Ignoring.\n",
1713 __func__,
1714 if_name(ifa->ifa_ifp),
1715 ip6_sprintf(IFA_IN6(ifa)));
1716 dp->dad_same_nonce_count++;
1717 dp->dad_ns_lcount++;
1718 ++ip6stat.ip6s_dad_loopcount;
1719 DAD_UNLOCK(dp);
1720 continue;
1721 } else if (!same_nonce) {
1722 // Not the same nonce, reset counter
1723 dp->dad_same_nonce_count = 1;
1724 }
1725
1726 DAD_ADDREF_LOCKED(dp);
1727 DAD_UNLOCK(dp);
1728 break;
1729 }
1730 lck_mtx_unlock(&dad6_mutex);
1731 return dp;
1732 }
1733
1734 void
nd6_dad_stoptimer(struct ifaddr * ifa)1735 nd6_dad_stoptimer(
1736 struct ifaddr *ifa)
1737 {
1738 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1739 }
1740
1741 /*
1742 * Start Duplicate Address Detection (DAD) for specified interface address.
1743 */
1744 void
nd6_dad_start(struct ifaddr * ifa,int * tick_delay)1745 nd6_dad_start(
1746 struct ifaddr *ifa,
1747 int *tick_delay) /* minimum delay ticks for IFF_UP event */
1748 {
1749 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1750 struct dadq *__single dp;
1751
1752 if (ifa->ifa_ifp == NULL) {
1753 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1754 }
1755
1756 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1757 __func__,
1758 ip6_sprintf(&ia->ia_addr.sin6_addr),
1759 if_name(ia->ia_ifp),
1760 ia->ia6_flags);
1761
1762 /*
1763 * If we don't need DAD, don't do it.
1764 * There are several cases:
1765 * - DAD is disabled (ip6_dad_count == 0)
1766 * - the interface address is anycast
1767 */
1768 IFA_LOCK(&ia->ia_ifa);
1769 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
1770 nd6log0(info,
1771 "nd6_dad_start: not a tentative or optimistic address "
1772 "%s(%s)\n",
1773 ip6_sprintf(&ia->ia_addr.sin6_addr),
1774 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1775 IFA_UNLOCK(&ia->ia_ifa);
1776 return;
1777 }
1778 if (!ip6_dad_count || (ia->ia6_flags & IN6_IFF_ANYCAST) != 0) {
1779 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
1780 IFA_UNLOCK(&ia->ia_ifa);
1781 return;
1782 }
1783 IFA_UNLOCK(&ia->ia_ifa);
1784
1785 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1786 (ifa->ifa_ifp->if_eflags & IFEF_IPV6_ND6ALT)) {
1787 return;
1788 }
1789 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1790 DAD_REMREF(dp);
1791 /* DAD already in progress */
1792 return;
1793 }
1794
1795 dp = zalloc_flags(dad_zone, Z_WAITOK | Z_ZERO);
1796 lck_mtx_init(&dp->dad_lock, &ifa_mtx_grp, &ifa_mtx_attr);
1797
1798 /* Callee adds one reference for us */
1799 dp = nd6_dad_attach(dp, ifa);
1800
1801 nd6log0(info, "%s: starting %sDAD %sfor %s\n",
1802 if_name(ifa->ifa_ifp),
1803 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) ? "optimistic " : "",
1804 (tick_delay == NULL) ? "immediately " : "",
1805 ip6_sprintf(&ia->ia_addr.sin6_addr));
1806
1807 /*
1808 * Send NS packet for DAD, ip6_dad_count times.
1809 * Note that we must delay the first transmission, if this is the
1810 * first packet to be sent from the interface after interface
1811 * (re)initialization.
1812 */
1813 if (tick_delay == NULL) {
1814 u_int32_t retrans;
1815 struct nd_ifinfo *__single ndi = NULL;
1816
1817 nd6_dad_ns_output(dp, ifa);
1818 ndi = ND_IFINFO(ifa->ifa_ifp);
1819 VERIFY(ndi != NULL && ndi->initialized);
1820 lck_mtx_lock(&ndi->lock);
1821 retrans = ndi->retrans * hz / 1000;
1822 lck_mtx_unlock(&ndi->lock);
1823 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
1824 } else {
1825 int ntick;
1826
1827 if (*tick_delay == 0) {
1828 ntick = random() % (MAX_RTR_SOLICITATION_DELAY * hz);
1829 } else {
1830 ntick = *tick_delay + random() % (hz / 2);
1831 }
1832 *tick_delay = ntick;
1833 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa,
1834 ntick);
1835 }
1836
1837 DAD_REMREF(dp); /* drop our reference */
1838 }
1839
1840 static struct dadq *
nd6_dad_attach(struct dadq * dp,struct ifaddr * ifa)1841 nd6_dad_attach(struct dadq *dp, struct ifaddr *ifa)
1842 {
1843 lck_mtx_lock(&dad6_mutex);
1844 DAD_LOCK(dp);
1845 dp->dad_ifa = ifa;
1846 ifa_addref(ifa); /* for dad_ifa */
1847 dp->dad_count = ip6_dad_count;
1848 dp->dad_ns_icount = dp->dad_na_icount = 0;
1849 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1850 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1851 VERIFY(!dp->dad_attached);
1852 dp->dad_same_nonce_count = 1;
1853 dp->dad_attached = 1;
1854 dp->dad_lladdrlen = 0;
1855 DAD_ADDREF_LOCKED(dp); /* for caller */
1856 DAD_ADDREF_LOCKED(dp); /* for dadq_head list */
1857 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1858 DAD_UNLOCK(dp);
1859 lck_mtx_unlock(&dad6_mutex);
1860
1861 return dp;
1862 }
1863
1864 static void
nd6_dad_detach(struct dadq * dp,struct ifaddr * ifa)1865 nd6_dad_detach(struct dadq *dp, struct ifaddr *ifa)
1866 {
1867 int detached;
1868
1869 lck_mtx_lock(&dad6_mutex);
1870 DAD_LOCK(dp);
1871 if ((detached = dp->dad_attached)) {
1872 VERIFY(dp->dad_ifa == ifa);
1873 TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1874 dp->dad_list.tqe_next = NULL;
1875 dp->dad_list.tqe_prev = NULL;
1876 dp->dad_attached = 0;
1877 }
1878 DAD_UNLOCK(dp);
1879 lck_mtx_unlock(&dad6_mutex);
1880 if (detached) {
1881 DAD_REMREF(dp); /* drop dadq_head reference */
1882 }
1883 }
1884
1885 /*
1886 * terminate DAD unconditionally. used for address removals.
1887 */
1888 void
nd6_dad_stop(struct ifaddr * ifa)1889 nd6_dad_stop(struct ifaddr *ifa)
1890 {
1891 struct dadq *__single dp;
1892
1893 dp = nd6_dad_find(ifa, NULL);
1894 if (!dp) {
1895 /* DAD wasn't started yet */
1896 return;
1897 }
1898
1899 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1900
1901 nd6_dad_detach(dp, ifa);
1902 DAD_REMREF(dp); /* drop our reference */
1903 }
1904
1905 static void
nd6_unsol_na_output(struct ifaddr * ifa)1906 nd6_unsol_na_output(struct ifaddr *ifa)
1907 {
1908 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1909 ifnet_ref_t ifp = ifa->ifa_ifp;
1910 struct in6_addr saddr6, taddr6;
1911
1912 if ((ifp->if_flags & IFF_UP) == 0 ||
1913 (ifp->if_flags & IFF_RUNNING) == 0 ||
1914 (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
1915 return;
1916 }
1917
1918 IFA_LOCK_SPIN(&ia->ia_ifa);
1919 taddr6 = ia->ia_addr.sin6_addr;
1920 IFA_UNLOCK(&ia->ia_ifa);
1921 if (in6_setscope(&taddr6, ifp, NULL) != 0) {
1922 return;
1923 }
1924 saddr6 = in6addr_linklocal_allnodes;
1925 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
1926 return;
1927 }
1928
1929 nd6log(info, "%s: sending unsolicited NA\n",
1930 if_name(ifa->ifa_ifp));
1931
1932 nd6_na_output(ifp, &saddr6, &taddr6, ND_NA_FLAG_OVERRIDE, 1, NULL);
1933 }
1934
1935 static void
nd6_dad_timer(struct ifaddr * ifa)1936 nd6_dad_timer(struct ifaddr *ifa)
1937 {
1938 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1939 struct dadq *__single dp = NULL;
1940 struct nd_ifinfo *ndi = NULL;
1941 u_int32_t retrans;
1942
1943 /* Sanity check */
1944 if (ia == NULL) {
1945 nd6log0(error, "nd6_dad_timer: called with null parameter\n");
1946 goto done;
1947 }
1948
1949 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1950 __func__,
1951 ip6_sprintf(&ia->ia_addr.sin6_addr),
1952 if_name(ia->ia_ifp),
1953 ia->ia6_flags);
1954
1955 dp = nd6_dad_find(ifa, NULL);
1956 if (dp == NULL) {
1957 nd6log0(error, "nd6_dad_timer: DAD structure not found\n");
1958 goto done;
1959 }
1960 IFA_LOCK(&ia->ia_ifa);
1961 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1962 nd6log0(error, "nd6_dad_timer: called with duplicated address "
1963 "%s(%s)\n",
1964 ip6_sprintf(&ia->ia_addr.sin6_addr),
1965 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1966 IFA_UNLOCK(&ia->ia_ifa);
1967 goto done;
1968 }
1969 if ((ia->ia6_flags & IN6_IFF_DADPROGRESS) == 0) {
1970 nd6log0(error, "nd6_dad_timer: not a tentative or optimistic "
1971 "address %s(%s)\n",
1972 ip6_sprintf(&ia->ia_addr.sin6_addr),
1973 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1974 IFA_UNLOCK(&ia->ia_ifa);
1975 goto done;
1976 }
1977 IFA_UNLOCK(&ia->ia_ifa);
1978
1979 /* timeouted with IFF_{RUNNING,UP} check */
1980 DAD_LOCK(dp);
1981 if (dp->dad_ns_tcount > dad_maxtry) {
1982 DAD_UNLOCK(dp);
1983 nd6log0(info, "%s: could not run DAD, driver problem?\n",
1984 if_name(ifa->ifa_ifp));
1985
1986 nd6_dad_detach(dp, ifa);
1987 goto done;
1988 }
1989
1990 /* Need more checks? */
1991 if (dp->dad_ns_ocount < dp->dad_count) {
1992 DAD_UNLOCK(dp);
1993 /*
1994 * We have more NS to go. Send NS packet for DAD.
1995 */
1996 nd6_dad_ns_output(dp, ifa);
1997 ndi = ND_IFINFO(ifa->ifa_ifp);
1998 VERIFY(ndi != NULL && ndi->initialized);
1999 lck_mtx_lock(&ndi->lock);
2000 retrans = ndi->retrans * hz / 1000;
2001 lck_mtx_unlock(&ndi->lock);
2002 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2003 } else {
2004 /*
2005 * We have transmitted sufficient number of DAD packets.
2006 * See what we've got.
2007 */
2008 if (dp->dad_na_icount > 0 || dp->dad_ns_icount) {
2009 /* We've seen NS or NA, means DAD has failed. */
2010 DAD_UNLOCK(dp);
2011 nd6log0(info,
2012 "%s: duplicate IPv6 address %s if:%s [timer]\n",
2013 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2014 if_name(ia->ia_ifp));
2015 nd6_dad_duplicated(ifa);
2016 /* (*dp) will be freed in nd6_dad_duplicated() */
2017 #if SKYWALK
2018 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2019 #endif /* SKYWALK */
2020 } else if (dad_enhanced != 0 &&
2021 dp->dad_ns_lcount > 0 &&
2022 dp->dad_ns_lcount > dp->dad_loopbackprobe &&
2023 dp->dad_same_nonce_count > 0 &&
2024 dp->dad_same_nonce_count > nd6_dad_nonce_max_count) {
2025 dp->dad_loopbackprobe = dp->dad_ns_lcount;
2026 dp->dad_count =
2027 dp->dad_ns_ocount + dad_maxtry - 1;
2028 DAD_UNLOCK(dp);
2029 ndi = ND_IFINFO(ifa->ifa_ifp);
2030 VERIFY(ndi != NULL && ndi->initialized);
2031 lck_mtx_lock(&ndi->lock);
2032 retrans = ndi->retrans * hz / 1000;
2033 lck_mtx_unlock(&ndi->lock);
2034
2035 /*
2036 * Sec. 4.1 in RFC 7527 requires transmission of
2037 * additional probes until the loopback condition
2038 * becomes clear when a looped back probe is detected.
2039 */
2040 nd6log0(info,
2041 "%s: a looped back NS message is detected during DAD for %s. Another DAD probe is being sent on interface %s.\n",
2042 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2043 if_name(ia->ia_ifp));
2044 /*
2045 * Send an NS immediately and increase dad_count by
2046 * nd6_mmaxtries - 1.
2047 */
2048 nd6_dad_ns_output(dp, ifa);
2049 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2050 goto done;
2051 } else {
2052 boolean_t txunsolna;
2053 DAD_UNLOCK(dp);
2054 /*
2055 * We are done with DAD. No NA came, no NS came.
2056 * No duplicate address found.
2057 */
2058 IFA_LOCK_SPIN(&ia->ia_ifa);
2059 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2060 IFA_UNLOCK(&ia->ia_ifa);
2061
2062 ndi = ND_IFINFO(ifa->ifa_ifp);
2063 VERIFY(ndi != NULL && ndi->initialized);
2064 lck_mtx_lock(&ndi->lock);
2065 txunsolna = (ndi->flags & ND6_IFF_REPLICATED) != 0;
2066 lck_mtx_unlock(&ndi->lock);
2067
2068 if (txunsolna) {
2069 nd6_unsol_na_output(ifa);
2070 }
2071
2072 nd6log0(info,
2073 "%s: DAD complete for %s - no duplicates found %s\n",
2074 if_name(ifa->ifa_ifp),
2075 ip6_sprintf(&ia->ia_addr.sin6_addr),
2076 txunsolna ? ", tx unsolicited NA with O=1" : ".");
2077
2078 if (dp->dad_ns_lcount > 0) {
2079 nd6log0(info,
2080 "%s: DAD completed while "
2081 "a looped back NS message is detected "
2082 "during DAD for %s om interface %s\n",
2083 __func__,
2084 ip6_sprintf(&ia->ia_addr.sin6_addr),
2085 if_name(ia->ia_ifp));
2086 }
2087
2088 in6_post_msg(ia->ia_ifp, KEV_INET6_NEW_USER_ADDR, ia,
2089 dp->dad_lladdr, ETHER_ADDR_LEN);
2090 nd6_dad_detach(dp, ifa);
2091 #if SKYWALK
2092 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2093 #endif /* SKYWALK */
2094 }
2095 }
2096
2097 done:
2098 if (dp != NULL) {
2099 DAD_REMREF(dp); /* drop our reference */
2100 }
2101 }
2102
2103 static void
nd6_dad_duplicated(struct ifaddr * ifa)2104 nd6_dad_duplicated(struct ifaddr *ifa)
2105 {
2106 struct in6_ifaddr *__single ia = ifatoia6(ifa);
2107 struct dadq *__single dp;
2108 ifnet_ref_t ifp = ifa->ifa_ifp;
2109 boolean_t candisable;
2110
2111 dp = nd6_dad_find(ifa, NULL);
2112 if (dp == NULL) {
2113 log(LOG_ERR, "%s: DAD structure not found.\n", __func__);
2114 return;
2115 }
2116 IFA_LOCK(&ia->ia_ifa);
2117 DAD_LOCK(dp);
2118 nd6log(error, "%s: NS in/out/loopback=%d/%d/%d, NA in=%d\n",
2119 __func__, dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
2120 dp->dad_na_icount);
2121 candisable = FALSE;
2122
2123 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr) &&
2124 !(ia->ia6_flags & IN6_IFF_SECURED)) {
2125 struct in6_addr in6;
2126 struct ifaddr *__single llifa = NULL;
2127 struct sockaddr_dl *sdl = NULL;
2128 uint8_t *lladdr = dp->dad_lladdr;
2129 uint8_t lladdrlen = dp->dad_lladdrlen;
2130
2131 /*
2132 * To avoid over-reaction, we only apply this logic when we are
2133 * very sure that hardware addresses are supposed to be unique.
2134 */
2135 switch (ifp->if_type) {
2136 case IFT_BRIDGE:
2137 case IFT_ETHER:
2138 case IFT_FDDI:
2139 case IFT_ATM:
2140 case IFT_IEEE1394:
2141 #ifdef IFT_IEEE80211
2142 case IFT_IEEE80211:
2143 #endif
2144 /*
2145 * Check if our hardware address matches the
2146 * link layer information received in the
2147 * NS/NA
2148 */
2149 llifa = ifp->if_lladdr;
2150 IFA_LOCK(llifa);
2151 sdl = SDL(llifa->ifa_addr);
2152 if (lladdrlen == sdl->sdl_alen &&
2153 bcmp(lladdr, LLADDR(sdl), lladdrlen) == 0) {
2154 candisable = TRUE;
2155 }
2156 IFA_UNLOCK(llifa);
2157
2158 in6 = ia->ia_addr.sin6_addr;
2159 if (in6_iid_from_hw(ifp, &in6) != 0) {
2160 break;
2161 }
2162
2163 /* Refine decision about whether IPv6 can be disabled */
2164 if (candisable &&
2165 !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2166 /*
2167 * Apply this logic only to the embedded MAC
2168 * address form of link-local IPv6 address.
2169 */
2170 candisable = FALSE;
2171 } else if (lladdr == NULL &&
2172 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2173 /*
2174 * We received a NA with no target link-layer
2175 * address option. This means that someone else
2176 * has our address. Mark it as a hardware
2177 * duplicate so we disable IPv6 later on.
2178 */
2179 candisable = TRUE;
2180 }
2181 break;
2182 default:
2183 break;
2184 }
2185 }
2186 DAD_UNLOCK(dp);
2187
2188 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2189 ia->ia6_flags |= IN6_IFF_DUPLICATED;
2190 in6_event_enqueue_nwk_wq_entry(IN6_ADDR_MARKED_DUPLICATED,
2191 ia->ia_ifa.ifa_ifp, &ia->ia_addr.sin6_addr,
2192 0);
2193 IFA_UNLOCK(&ia->ia_ifa);
2194
2195 /* increment DAD collision counter */
2196 ++ip6stat.ip6s_dad_collide;
2197
2198 /* We are done with DAD, with duplicated address found. (failure) */
2199 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
2200
2201 IFA_LOCK(&ia->ia_ifa);
2202 log(LOG_ERR, "%s: DAD complete for %s - duplicate found.\n",
2203 if_name(ifp), ip6_sprintf(&ia->ia_addr.sin6_addr));
2204 IFA_UNLOCK(&ia->ia_ifa);
2205
2206 if (candisable) {
2207 struct nd_ifinfo *__single ndi = ND_IFINFO(ifp);
2208 log(LOG_ERR, "%s: possible hardware address duplication "
2209 "detected, disabling IPv6 for interface.\n", if_name(ifp));
2210
2211 VERIFY((NULL != ndi) && (TRUE == ndi->initialized));
2212 ndi->flags |= ND6_IFF_IFDISABLED;
2213 /* Make sure to set IFEF_IPV6_DISABLED too */
2214 nd6_if_disable(ifp, TRUE);
2215 }
2216
2217 log(LOG_ERR,
2218 "%s: manual intervention may be required.\n",
2219 if_name(ifp));
2220
2221 /* Send an event to the configuration agent so that the
2222 * duplicate address will be notified to the user and will
2223 * be removed.
2224 */
2225 in6_post_msg(ifp, KEV_INET6_NEW_USER_ADDR, ia, dp->dad_lladdr, ETHER_ADDR_LEN);
2226 nd6_dad_detach(dp, ifa);
2227 DAD_REMREF(dp); /* drop our reference */
2228 }
2229
2230 static void
nd6_dad_ns_output(struct dadq * dp,struct ifaddr * ifa)2231 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
2232 {
2233 struct in6_ifaddr *__single ia = ifatoia6(ifa);
2234 ifnet_ref_t ifp = ifa->ifa_ifp;
2235 int i = 0;
2236 struct in6_addr taddr6;
2237
2238 DAD_LOCK(dp);
2239 dp->dad_ns_tcount++;
2240 if ((ifp->if_flags & IFF_UP) == 0) {
2241 DAD_UNLOCK(dp);
2242 return;
2243 }
2244 if ((ifp->if_flags & IFF_RUNNING) == 0) {
2245 DAD_UNLOCK(dp);
2246 return;
2247 }
2248
2249 dp->dad_ns_ocount++;
2250 DAD_UNLOCK(dp);
2251 IFA_LOCK_SPIN(&ia->ia_ifa);
2252 taddr6 = ia->ia_addr.sin6_addr;
2253 IFA_UNLOCK(&ia->ia_ifa);
2254 if (dad_enhanced != 0 && !(ifp->if_flags & IFF_POINTOPOINT)) {
2255 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) {
2256 dp->dad_nonce[i] = RandomULong();
2257 }
2258
2259 /*
2260 * XXXHRS: Note that in the case that
2261 * DupAddrDetectTransmits > 1, multiple NS messages with
2262 * different nonces can be looped back in an unexpected
2263 * order. The current implementation recognizes only
2264 * the latest nonce on the sender side. Practically it
2265 * should work well in almost all cases.
2266 */
2267 }
2268 nd6_ns_output(ifp, NULL, &taddr6, NULL,
2269 (uint8_t *)&dp->dad_nonce[0], ND_OPT_NONCE_LEN);
2270 }
2271
2272 /*
2273 * @brief Called to process DAD NS
2274 *
2275 * @param ifa is the pointer to the interface's address
2276 * @param lladdr is source link layer information
2277 * @param lladdrlen is source's linklayer length
2278 *
2279 * @return void
2280 */
2281 static void
nd6_dad_ns_input(struct ifaddr * ifa,char * lladdr __sized_by (lladdrlen),int lladdrlen,struct nd_opt_nonce * ndopt_nonce)2282 nd6_dad_ns_input(struct ifaddr *ifa, char *lladdr __sized_by(lladdrlen),
2283 int lladdrlen, struct nd_opt_nonce *ndopt_nonce)
2284 {
2285 struct dadq *__single dp;
2286 VERIFY(ifa != NULL);
2287
2288 /* Ignore Nonce option when Enhanced DAD is disabled. */
2289 if (dad_enhanced == 0) {
2290 ndopt_nonce = NULL;
2291 }
2292
2293 dp = nd6_dad_find(ifa, ndopt_nonce);
2294 if (dp == NULL) {
2295 return;
2296 }
2297
2298 DAD_LOCK(dp);
2299 ++dp->dad_ns_icount;
2300 if (lladdr && lladdrlen >= ETHER_ADDR_LEN) {
2301 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2302 /* fine to truncate as it is compared against sdl_alen */
2303 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2304 }
2305 DAD_UNLOCK(dp);
2306 DAD_REMREF(dp);
2307 }
2308
2309 /*
2310 * @brief Called to process received NA for DAD
2311 *
2312 * @param m is the pointer to the packet's mbuf
2313 * @param ifp is the pointer to the interface on which packet
2314 * was receicved.
2315 * @param taddr is pointer to target's IPv6 address
2316 * @param lladdr is target's link layer information
2317 * @param lladdrlen is target's linklayer length
2318 *
2319 * @return NULL if the packet is consumed by DAD processing, else
2320 * pointer to the mbuf.
2321 */
2322 static struct mbuf *
nd6_dad_na_input(struct mbuf * m,struct ifnet * ifp,struct in6_addr * taddr,caddr_t lladdr __counted_by (lladdrlen),int lladdrlen)2323 nd6_dad_na_input(struct mbuf *m, struct ifnet *ifp, struct in6_addr *taddr,
2324 caddr_t lladdr __counted_by(lladdrlen), int lladdrlen)
2325 {
2326 struct ifaddr *__single ifa = NULL;
2327 struct in6_ifaddr *__single ia = NULL;
2328 struct dadq *__single dp = NULL;
2329 struct nd_ifinfo *__single ndi = NULL;
2330 boolean_t replicated;
2331
2332 ifa = (struct ifaddr *) in6ifa_ifpwithaddr(ifp, taddr);
2333 if (ifa == NULL) {
2334 return m;
2335 }
2336
2337 replicated = FALSE;
2338
2339 /* Get the ND6_IFF_REPLICATED flag. */
2340 ndi = ND_IFINFO(ifp);
2341 if (ndi != NULL && ndi->initialized) {
2342 lck_mtx_lock(&ndi->lock);
2343 replicated = !!(ndi->flags & ND6_IFF_REPLICATED);
2344 lck_mtx_unlock(&ndi->lock);
2345 }
2346
2347 if (replicated) {
2348 nd6log(info, "%s: ignoring duplicate NA on "
2349 "replicated interface %s\n", __func__, if_name(ifp));
2350 goto done;
2351 }
2352
2353 /* Lock the interface address until done (see label below). */
2354 IFA_LOCK(ifa);
2355 ia = ifatoia6(ifa);
2356
2357 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
2358 IFA_UNLOCK(ifa);
2359 nd6log(info, "%s: ignoring duplicate NA on "
2360 "%s [DAD not in progress]\n", __func__,
2361 if_name(ifp));
2362 goto done;
2363 }
2364
2365 /* Some sleep proxies improperly send the client's Ethernet address in
2366 * the target link-layer address option, so detect this by comparing
2367 * the L2-header source address, if we have seen it, with the target
2368 * address, and ignoring the NA if they don't match.
2369 */
2370 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2371 struct ip6aux *__single ip6a = ip6_findaux(m);
2372 if (ip6a && (ip6a->ip6a_flags & IP6A_HASEEN) != 0 &&
2373 bcmp(ip6a->ip6a_ehsrc, lladdr, ETHER_ADDR_LEN) != 0) {
2374 IFA_UNLOCK(ifa);
2375 nd6log(error, "%s: ignoring duplicate NA on %s "
2376 "[eh_src != tgtlladdr]\n", __func__, if_name(ifp));
2377 goto done;
2378 }
2379 }
2380
2381 IFA_UNLOCK(ifa);
2382
2383 dp = nd6_dad_find(ifa, NULL);
2384 if (dp == NULL) {
2385 nd6log(info, "%s: no DAD structure for %s on %s.\n",
2386 __func__, ip6_sprintf(taddr), if_name(ifp));
2387 goto done;
2388 }
2389
2390 DAD_LOCK_SPIN(dp);
2391 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2392 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2393 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2394 }
2395 dp->dad_na_icount++;
2396 DAD_UNLOCK(dp);
2397 DAD_REMREF(dp);
2398
2399 /* remove the address. */
2400 nd6log(info,
2401 "%s: duplicate IPv6 address %s [processing NA on %s]\n", __func__,
2402 ip6_sprintf(taddr), if_name(ifp));
2403 done:
2404 IFA_LOCK_ASSERT_NOTHELD(ifa);
2405 ifa_remref(ifa);
2406 m_freem(m);
2407 return NULL;
2408 }
2409
2410 static void
dad_addref(struct dadq * dp,int locked)2411 dad_addref(struct dadq *dp, int locked)
2412 {
2413 if (!locked) {
2414 DAD_LOCK_SPIN(dp);
2415 } else {
2416 DAD_LOCK_ASSERT_HELD(dp);
2417 }
2418
2419 if (++dp->dad_refcount == 0) {
2420 panic("%s: dad %p wraparound refcnt", __func__, dp);
2421 /* NOTREACHED */
2422 }
2423 if (!locked) {
2424 DAD_UNLOCK(dp);
2425 }
2426 }
2427
2428 static void
dad_remref(struct dadq * dp)2429 dad_remref(struct dadq *dp)
2430 {
2431 struct ifaddr *__single ifa;
2432
2433 DAD_LOCK_SPIN(dp);
2434 if (dp->dad_refcount == 0) {
2435 panic("%s: dad %p negative refcnt", __func__, dp);
2436 }
2437 --dp->dad_refcount;
2438 if (dp->dad_refcount > 0) {
2439 DAD_UNLOCK(dp);
2440 return;
2441 }
2442 DAD_UNLOCK(dp);
2443
2444 if (dp->dad_attached ||
2445 dp->dad_list.tqe_next != NULL || dp->dad_list.tqe_prev != NULL) {
2446 panic("%s: attached dad=%p is being freed", __func__, dp);
2447 /* NOTREACHED */
2448 }
2449
2450 if ((ifa = dp->dad_ifa) != NULL) {
2451 ifa_remref(ifa); /* drop dad_ifa reference */
2452 dp->dad_ifa = NULL;
2453 }
2454
2455 lck_mtx_destroy(&dp->dad_lock, &ifa_mtx_grp);
2456 zfree(dad_zone, dp);
2457 }
2458
2459 void
nd6_llreach_set_reachable(struct ifnet * ifp,void * __sized_by (alen)addr,unsigned int alen)2460 nd6_llreach_set_reachable(struct ifnet *ifp, void *__sized_by(alen) addr, unsigned int alen)
2461 {
2462 /* Nothing more to do if it's disabled */
2463 if (nd6_llreach_base == 0) {
2464 return;
2465 }
2466
2467 ifnet_llreach_set_reachable(ifp, ETHERTYPE_IPV6, addr, alen);
2468 }
2469
2470 void
nd6_alt_node_addr_decompose(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr_dl * isdl,struct sockaddr_in6 * sin6)2471 nd6_alt_node_addr_decompose(struct ifnet *ifp, struct sockaddr *sa,
2472 struct sockaddr_dl* isdl, struct sockaddr_in6 *sin6)
2473 {
2474 static const size_t EUI64_LENGTH = 8;
2475 struct sockaddr_dl *sdl = SDL(isdl);
2476
2477 VERIFY(nd6_need_cache(ifp));
2478 VERIFY(sa);
2479 VERIFY(sdl && (void *)sa != (void *)sdl);
2480 VERIFY(sin6 && (void *)sa != (void *)sin6);
2481
2482 SOCKADDR_ZERO(sin6, sizeof(*sin6));
2483 sin6->sin6_len = sizeof *sin6;
2484 sin6->sin6_family = AF_INET6;
2485
2486 SOCKADDR_ZERO(sdl, sizeof(*sdl));
2487 sdl->sdl_len = sizeof *sdl;
2488 sdl->sdl_family = AF_LINK;
2489 sdl->sdl_type = ifp->if_type;
2490 sdl->sdl_index = ifp->if_index;
2491 sdl->sdl_nlen = 0;
2492
2493 switch (sa->sa_family) {
2494 case AF_INET6: {
2495 struct sockaddr_in6 *__single sin6a = SIN6(sa);
2496 struct in6_addr *__single in6 = &sin6a->sin6_addr;
2497
2498 VERIFY(sa->sa_len == sizeof *sin6);
2499 if (in6->s6_addr[11] == 0xff && in6->s6_addr[12] == 0xfe) {
2500 sdl->sdl_alen = ETHER_ADDR_LEN;
2501 LLADDR(sdl)[0] = (in6->s6_addr[8] ^ ND6_EUI64_UBIT);
2502 LLADDR(sdl)[1] = in6->s6_addr[9];
2503 LLADDR(sdl)[2] = in6->s6_addr[10];
2504 LLADDR(sdl)[3] = in6->s6_addr[13];
2505 LLADDR(sdl)[4] = in6->s6_addr[14];
2506 LLADDR(sdl)[5] = in6->s6_addr[15];
2507 } else {
2508 sdl->sdl_alen = EUI64_LENGTH;
2509 bcopy(&in6->s6_addr[8], LLADDR(sdl), EUI64_LENGTH);
2510 }
2511
2512 sdl->sdl_slen = 0;
2513 break;
2514 }
2515 case AF_LINK: {
2516 struct sockaddr_dl *sdla = SDL(sa);
2517 struct in6_addr *__single in6 = &sin6->sin6_addr;
2518 caddr_t lla = LLADDR(sdla);
2519 VERIFY(sa->sa_len <= sizeof(*sdl));
2520 SOCKADDR_COPY(sa, sdl, sa->sa_len);
2521
2522 sin6->sin6_scope_id = sdla->sdl_index;
2523 if (sin6->sin6_scope_id == 0) {
2524 sin6->sin6_scope_id = ifp->if_index;
2525 }
2526 in6->s6_addr[0] = 0xfe;
2527 in6->s6_addr[1] = 0x80;
2528 if (sdla->sdl_alen == EUI64_LENGTH) {
2529 bcopy(lla, &in6->s6_addr[8], EUI64_LENGTH);
2530 } else {
2531 VERIFY(sdla->sdl_alen == ETHER_ADDR_LEN);
2532
2533 in6->s6_addr[8] = ((uint8_t) lla[0] ^ ND6_EUI64_UBIT);
2534 in6->s6_addr[9] = (uint8_t) lla[1];
2535 in6->s6_addr[10] = (uint8_t) lla[2];
2536 in6->s6_addr[11] = 0xff;
2537 in6->s6_addr[12] = 0xfe;
2538 in6->s6_addr[13] = (uint8_t) lla[3];
2539 in6->s6_addr[14] = (uint8_t) lla[4];
2540 in6->s6_addr[15] = (uint8_t) lla[5];
2541 }
2542
2543 break;
2544 }
2545 default:
2546 VERIFY(false);
2547 break;
2548 }
2549 }
2550
2551 int
nd6_alt_node_present(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * isdl,int32_t rssi,int lqm,int npm)2552 nd6_alt_node_present(struct ifnet *ifp, struct sockaddr_in6 *sin6,
2553 struct sockaddr_dl *isdl, int32_t rssi, int lqm, int npm)
2554 {
2555 struct sockaddr_dl *sdl = SDL(isdl);
2556 rtentry_ref_t rt = NULL;
2557 struct llinfo_nd6 *__single ln = NULL;
2558 struct if_llreach *__single lr = NULL;
2559 int nd6_nc_updated = 0;
2560 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2561 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2562
2563 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2564 if (in6_embedded_scope) {
2565 if (temp_embedded_id == 0) {
2566 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2567 }
2568 } else if (temp_ifscope_id == 0) {
2569 sin6->sin6_scope_id = ifp->if_index;
2570 }
2571 }
2572
2573 nd6_cache_lladdr(ifp, &sin6->sin6_addr, LLADDR(sdl), sdl->sdl_alen,
2574 ND_NEIGHBOR_ADVERT, 0, &nd6_nc_updated);
2575
2576 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2577 lck_mtx_lock(rnh_lock);
2578
2579 rt = rtalloc1_scoped_locked(SA(sin6), 1, 0, ifp->if_index);
2580
2581 /* Restore the address that was passed to us */
2582 if (in6_embedded_scope) {
2583 if (temp_embedded_id == 0) {
2584 sin6->sin6_addr.s6_addr16[1] = 0;
2585 }
2586 } else if (temp_ifscope_id == 0) {
2587 sin6->sin6_scope_id = 0;
2588 }
2589
2590 if (rt != NULL) {
2591 RT_LOCK(rt);
2592 VERIFY(rt->rt_flags & RTF_LLINFO);
2593 VERIFY(rt->rt_llinfo);
2594
2595 ln = rt->rt_llinfo;
2596 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
2597 ln_setexpire(ln, 0);
2598
2599 lr = ln->ln_llreach;
2600 if (lr) {
2601 IFLR_LOCK(lr);
2602 lr->lr_rssi = rssi;
2603 lr->lr_lqm = (int32_t) lqm;
2604 lr->lr_npm = (int32_t) npm;
2605 IFLR_UNLOCK(lr);
2606 }
2607
2608 RT_UNLOCK(rt);
2609 RT_REMREF(rt);
2610 }
2611
2612 lck_mtx_unlock(rnh_lock);
2613
2614 if (rt == NULL) {
2615 log(LOG_ERR, "%s: failed to add/update host route to %s.\n",
2616 __func__, ip6_sprintf(&sin6->sin6_addr));
2617 #if DEBUG || DEVELOPMENT
2618 if (ip6_p2p_debug) {
2619 panic("%s: failed to add/update host route to %s.\n",
2620 __func__, ip6_sprintf(&sin6->sin6_addr));
2621 }
2622 #endif
2623 return EHOSTUNREACH;
2624 }
2625
2626 nd6log(info, "%s: Successfully added/updated host route to %s [lr=0x%llx]\n",
2627 __func__, ip6_sprintf(&sin6->sin6_addr),
2628 (uint64_t)VM_KERNEL_ADDRPERM(lr));
2629 /*
2630 * nd6_nc_updated not set implies that nothing was updated
2631 * in the neighbor cache. Convey that as EEXIST to callers.
2632 */
2633 if (nd6_nc_updated == 0) {
2634 return EEXIST;
2635 }
2636 return 0;
2637 }
2638
2639 int
nd6_alt_node_absent(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl)2640 nd6_alt_node_absent(struct ifnet *ifp, struct sockaddr_in6 *sin6, struct sockaddr_dl *sdl)
2641 {
2642 rtentry_ref_t rt = NULL;
2643 int error = 0;
2644 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2645 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2646
2647 nd6log2(debug, "%s: host route to %s\n", __func__,
2648 ip6_sprintf(&sin6->sin6_addr));
2649
2650 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2651 if (in6_embedded_scope) {
2652 if (temp_embedded_id == 0) {
2653 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2654 }
2655 } else if (temp_ifscope_id == 0) {
2656 sin6->sin6_scope_id = ifp->if_index;
2657 }
2658 }
2659
2660 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2661 lck_mtx_lock(rnh_lock);
2662
2663 rt = rtalloc1_scoped_locked(SA(sin6), 0, 0, 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 SOCKADDR_COPY(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), NULL, rt_mask(rt),
2692 0, (struct rtentry **)NULL);
2693
2694 rtfree_locked(rt);
2695 } else {
2696 error = EHOSTUNREACH;
2697 RT_REMREF_LOCKED(rt);
2698 RT_UNLOCK(rt);
2699 }
2700 } else {
2701 error = EHOSTUNREACH;
2702 }
2703
2704 if (error == 0) {
2705 nd6log(info, "%s: Successfully deleted host route to %s "
2706 "for interface %s.\n", __func__, ip6_sprintf(&sin6->sin6_addr),
2707 ifp->if_xname);
2708 } else {
2709 nd6log(error, "%s: Failed to delete host route to %s "
2710 "for interface %s with error :%d.\n", __func__,
2711 ip6_sprintf(&sin6->sin6_addr),
2712 ifp->if_xname, error);
2713 }
2714
2715 lck_mtx_unlock(rnh_lock);
2716 return error;
2717 }
2718