1 /*
2 * Copyright (c) 2000-2024 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;
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 *__single hsrc = NULL;
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 hsrc = &hip6->ip6_src;
727 } else {
728 hsrc = NULL;
729 }
730 }
731 /* Update probe count, if applicable */
732 if (ln->ln_llreach != NULL) {
733 IFLR_LOCK_SPIN(ln->ln_llreach);
734 ln->ln_llreach->lr_probes++;
735 IFLR_UNLOCK(ln->ln_llreach);
736 }
737 rtflags = ln->ln_rt->rt_flags;
738 RT_UNLOCK(ln->ln_rt);
739 }
740 if (hsrc != NULL && (ia = in6ifa_ifpwithaddr(ifp, hsrc)) &&
741 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0) {
742 src = hsrc;
743 } else {
744 int error;
745 struct sockaddr_in6 dst_sa;
746
747 SOCKADDR_ZERO(&dst_sa, sizeof(dst_sa));
748 dst_sa.sin6_family = AF_INET6;
749 dst_sa.sin6_len = sizeof(dst_sa);
750 dst_sa.sin6_addr = ip6->ip6_dst;
751
752 src = in6_selectsrc(&dst_sa, NULL,
753 NULL, &ro, NULL, &src_storage, ip6oa.ip6oa_boundif,
754 &error);
755 if (src == NULL) {
756 nd6log(info,
757 "nd6_ns_output: source can't be "
758 "determined: dst=%s, error=%d\n",
759 ip6_sprintf(&dst_sa.sin6_addr),
760 error);
761 drop_reason = DROP_REASON_IP_SRC_ADDR_NO_AVAIL;
762 goto bad;
763 }
764
765 if (ia != NULL) {
766 ifa_remref(&ia->ia_ifa);
767 ia = NULL;
768 }
769 /*
770 * RFC 4429 section 3.2:
771 * When a node has a unicast packet to send
772 * from an Optimistic Address to a neighbor,
773 * but does not know the neighbor's link-layer
774 * address, it MUST NOT perform Address
775 * Resolution.
776 */
777 ia = in6ifa_ifpwithaddr(ifp, src);
778 if (ia == NULL) {
779 nd6log(info,
780 "nd6_ns_output: no preferred source "
781 "available: dst=%s\n",
782 ip6_sprintf(&dst_sa.sin6_addr));
783 drop_reason = DROP_REASON_IP_SRC_ADDR_NO_AVAIL;
784 goto bad;
785 }
786 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
787 is_optimistic = TRUE;
788 nd6log(info,
789 "nd6_ns_output: preferred source "
790 "available is optimistic: dst=%s\n",
791 ip6_sprintf(&dst_sa.sin6_addr));
792 }
793 }
794 } else {
795 /*
796 * Source address for DAD packet must always be IPv6
797 * unspecified address. (0::0)
798 * We actually don't have to 0-clear the address (we did it
799 * above), but we do so here explicitly to make the intention
800 * clearer.
801 */
802 bzero(&src_in, sizeof(src_in));
803 src = &src_in;
804 ip6oa.ip6oa_flags &= ~IP6OAF_BOUND_SRCADDR;
805 }
806 ip6->ip6_src = *src;
807 ip6_output_setsrcifscope(m, ifp->if_index, ia);
808 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
809 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
810 nd_ns->nd_ns_code = 0;
811 nd_ns->nd_ns_reserved = 0;
812 nd_ns->nd_ns_target = *taddr6;
813 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
814
815 /*
816 * Add source link-layer address option.
817 *
818 * spec implementation
819 * --- ---
820 * DAD packet MUST NOT do not add the option
821 * Source is optimistic MUST NOT do not add the option
822 * there's no link layer address:
823 * impossible do not add the option
824 * there's link layer address:
825 * Multicast NS MUST add one add the option
826 * Unicast NS SHOULD add one add the option
827 *
828 * XXX We deviate from RFC 4429 and still use optimistic DAD as source
829 * for address resolution. However to ensure that we do not interfere
830 * with neighbor cache entries of other neighbors, we MUST ensure
831 * that SLLAO is not sent. Also note, sending multicast NS without SLLAO
832 * is also a deviation from RFC 4861.
833 */
834 if (nonce == NULL && (mac = nd6_ifptomac(ifp)) && !is_optimistic) {
835 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
836 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
837 /* 8 byte alignments... */
838 optlen = (optlen + 7) & ~7;
839
840 m->m_pkthdr.len += optlen;
841 m->m_len += optlen;
842 icmp6len += optlen;
843 bzero((caddr_t)nd_opt, optlen);
844 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
845 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
846 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
847 }
848 /*
849 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
850 * This behavior is documented as Enhanced Duplicate Address
851 * Detection in draft-ietf-6man-enhanced-dad-13.
852 * net.inet6.ip6.dad_enhanced=0 disables this.
853 */
854 if (dad_enhanced != 0 && nonce != NULL && !(ifp->if_flags & IFF_POINTOPOINT)) {
855 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
856 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
857 /* 8-byte alignment is required. */
858 optlen = (optlen + 7) & ~7;
859
860 m->m_pkthdr.len += optlen;
861 m->m_len += optlen;
862 icmp6len += optlen;
863 bzero((caddr_t)nd_opt, optlen);
864 nd_opt->nd_opt_type = ND_OPT_NONCE;
865 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
866 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
867 }
868 ip6->ip6_plen = htons((u_short)icmp6len);
869 nd_ns->nd_ns_cksum = 0;
870 nd_ns->nd_ns_cksum
871 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
872
873 flags = nonce ? IPV6_UNSPECSRC : 0;
874 flags |= IPV6_OUTARGS;
875
876 /*
877 * PKTF_{INET,INET6}_RESOLVE_RTR are mutually exclusive, so make
878 * sure only one of them is set (just in case.)
879 */
880 m->m_pkthdr.pkt_flags &= ~(PKTF_INET_RESOLVE | PKTF_RESOLVE_RTR);
881 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
882 /*
883 * If this is a NS for resolving the (default) router, mark
884 * the packet accordingly so that the driver can find out,
885 * in case it needs to perform driver-specific action(s).
886 */
887 if (rtflags & RTF_ROUTER) {
888 m->m_pkthdr.pkt_flags |= PKTF_RESOLVE_RTR;
889 }
890
891 if (ifp->if_eflags & IFEF_TXSTART) {
892 /*
893 * Use control service class if the interface
894 * supports transmit-start model
895 */
896 (void) m_set_service_class(m, MBUF_SC_CTL);
897 }
898
899 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
900 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
901 ip6_output(m, NULL, NULL, flags, im6o, &outif, &ip6oa);
902 if (outif) {
903 icmp6_ifstat_inc(outif, ifs6_out_msg);
904 icmp6_ifstat_inc(outif, ifs6_out_neighborsolicit);
905 ifnet_release(outif);
906 }
907 icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
908
909 exit:
910 if (im6o != NULL) {
911 IM6O_REMREF(im6o);
912 }
913
914 ROUTE_RELEASE(&ro); /* we don't cache this route. */
915
916 if (ia != NULL) {
917 ifa_remref(&ia->ia_ifa);
918 }
919 return;
920
921 bad:
922 m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
923 goto exit;
924 }
925
926 /*
927 * Neighbor advertisement input handling.
928 *
929 * Based on RFC 4861
930 * Based on RFC 4862 (duplicate address detection)
931 *
932 * the following items are not implemented yet:
933 * - anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD)
934 * - proxy advertisement delay rule (RFC 4861 7.2.8, last paragraph, "should")
935 */
936 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)937 nd6_na_input(struct mbuf *m, int off, int icmp6len)
938 {
939 ifnet_ref_t ifp = m->m_pkthdr.rcvif;
940 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
941 struct nd_neighbor_advert *nd_na;
942 struct in6_addr saddr6 = ip6->ip6_src;
943 struct in6_addr daddr6 = ip6->ip6_dst;
944 struct in6_addr taddr6;
945 int flags;
946 int is_router;
947 int is_solicited;
948 int is_override;
949 char *lladdr = NULL;
950 int lladdrlen = 0;
951 struct llinfo_nd6 *__single ln;
952 rtentry_ref_t rt;
953 struct sockaddr_dl *sdl;
954 union nd_opts ndopts;
955 uint64_t timenow;
956 bool send_nc_alive_kev = false;
957 drop_reason_t drop_reason = DROP_REASON_UNSPECIFIED;
958
959 if ((ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
960 nd6log(info, "nd6_na_input: on ND6ALT interface!\n");
961 drop_reason = DROP_REASON_IP6_NO_ND6ALT_IF;
962 goto freeit;
963 }
964
965 /* Expect 32-bit aligned data pointer on strict-align platforms */
966 MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
967
968 if (ip6->ip6_hlim != IPV6_MAXHLIM) {
969 nd6log(error,
970 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
971 ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
972 ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
973 drop_reason = DROP_REASON_IP6_BAD_HLIM;
974 goto bad;
975 }
976
977 IP6_EXTHDR_CHECK(m, off, icmp6len, return );
978 ip6 = mtod(m, struct ip6_hdr *);
979 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
980 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
981
982 flags = nd_na->nd_na_flags_reserved;
983 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
984 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
985 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
986
987 taddr6 = nd_na->nd_na_target;
988 if (in6_setscope(&taddr6, ifp, NULL)) {
989 drop_reason = DROP_REASON_IP6_BAD_SCOPE;
990 goto bad; /* XXX: impossible */
991 }
992 if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
993 nd6log(error,
994 "nd6_na_input: invalid target address %s\n",
995 ip6_sprintf(&taddr6));
996 drop_reason = DROP_REASON_IP_DST_ADDR_NO_AVAIL;
997 goto bad;
998 }
999 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1000 if (is_solicited) {
1001 nd6log(error,
1002 "nd6_na_input: a solicited adv is multicasted\n");
1003 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1004 goto bad;
1005 }
1006 }
1007
1008 icmp6len -= sizeof(*nd_na);
1009 nd6_option_init(nd_na + 1, icmp6len, &ndopts);
1010 if (nd6_options(&ndopts) < 0) {
1011 nd6log(info,
1012 "nd6_na_input: invalid ND option, ignored\n");
1013 /* nd6_options have incremented stats */
1014 drop_reason = DROP_REASON_IP6_TOO_MANY_OPTIONS;
1015 goto freeit;
1016 }
1017
1018 if (ndopts.nd_opts_tgt_lladdr) {
1019 ND_OPT_LLADDR(ndopts.nd_opts_tgt_lladdr, nd_opt_len, lladdr, lladdrlen);
1020
1021 if (((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
1022 nd6log(info,
1023 "nd6_na_input: lladdrlen mismatch for %s "
1024 "(if %d, NA packet %d)\n",
1025 ip6_sprintf(&taddr6), ifp->if_addrlen,
1026 lladdrlen - 2);
1027 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1028 goto bad;
1029 }
1030 }
1031
1032 m = nd6_dad_na_input(m, ifp, &taddr6, lladdr, lladdrlen);
1033 if (m == NULL) {
1034 return;
1035 }
1036
1037 /* Forwarding associated with NDPRF_PRPROXY may apply. */
1038 if (ip6_forwarding && nd6_prproxy) {
1039 nd6_prproxy_na_input(ifp, &saddr6, &daddr6, &taddr6, flags);
1040 }
1041
1042 /*
1043 * If no neighbor cache entry is found, NA SHOULD silently be
1044 * discarded. If we are forwarding (and Scoped Routing is in
1045 * effect), try to see if there is a neighbor cache entry on
1046 * another interface (in case we are doing prefix proxying.)
1047 */
1048 if ((rt = nd6_lookup(&taddr6, 0, ifp, 0)) == NULL) {
1049 if (!ip6_forwarding || !nd6_prproxy) {
1050 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1051 goto freeit;
1052 }
1053
1054 if ((rt = nd6_lookup(&taddr6, 0, NULL, 0)) == NULL) {
1055 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1056 goto freeit;
1057 }
1058
1059 RT_LOCK_ASSERT_HELD(rt);
1060 if (rt->rt_ifp != ifp) {
1061 /*
1062 * Purge any link-layer info caching.
1063 */
1064 if (rt->rt_llinfo_purge != NULL) {
1065 rt->rt_llinfo_purge(rt);
1066 }
1067
1068 /* Adjust route ref count for the interfaces */
1069 if (rt->rt_if_ref_fn != NULL) {
1070 rt->rt_if_ref_fn(ifp, 1);
1071 rt->rt_if_ref_fn(rt->rt_ifp, -1);
1072 }
1073
1074 /* Change the interface when the existing route is on */
1075 rt->rt_ifp = ifp;
1076
1077 /*
1078 * If rmx_mtu is not locked, update it
1079 * to the MTU used by the new interface.
1080 */
1081 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1082 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
1083 }
1084 }
1085 }
1086
1087 RT_LOCK_ASSERT_HELD(rt);
1088 if ((ln = rt->rt_llinfo) == NULL ||
1089 (sdl = SDL(rt->rt_gateway)) == NULL) {
1090 RT_REMREF_LOCKED(rt);
1091 RT_UNLOCK(rt);
1092 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1093 goto freeit;
1094 }
1095
1096 timenow = net_uptime();
1097
1098 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
1099 /*
1100 * If the link-layer has address, and no lladdr option came,
1101 * discard the packet.
1102 */
1103 if (ifp->if_addrlen && !lladdr) {
1104 RT_REMREF_LOCKED(rt);
1105 RT_UNLOCK(rt);
1106 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1107 goto freeit;
1108 }
1109
1110 /*
1111 * Record link-layer address, and update the state.
1112 */
1113 sdl->sdl_alen = ifp->if_addrlen;
1114 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1115 if (is_solicited) {
1116 send_nc_alive_kev = (rt->rt_flags & RTF_ROUTER) ? true : false;
1117 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1118 if (ln->ln_expire != 0) {
1119 struct nd_ifinfo *__single ndi = NULL;
1120
1121 ndi = ND_IFINFO(rt->rt_ifp);
1122 VERIFY(ndi != NULL && ndi->initialized);
1123 lck_mtx_lock(&ndi->lock);
1124 ln_setexpire(ln, timenow + ndi->reachable);
1125 lck_mtx_unlock(&ndi->lock);
1126 RT_UNLOCK(rt);
1127 lck_mtx_lock(rnh_lock);
1128 nd6_sched_timeout(NULL, NULL);
1129 lck_mtx_unlock(rnh_lock);
1130 RT_LOCK(rt);
1131 }
1132 } else {
1133 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1134 ln_setexpire(ln, timenow + nd6_gctimer);
1135 }
1136
1137 /*
1138 * Enqueue work item to invoke callback for this
1139 * route entry
1140 */
1141 route_event_enqueue_nwk_wq_entry(rt, NULL,
1142 ROUTE_LLENTRY_RESOLVED, NULL, TRUE);
1143
1144 if ((ln->ln_router = (short)is_router) != 0) {
1145 struct radix_node_head *__single rnh = NULL;
1146 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1147 ifnet_ref_t rt_ifp = rt->rt_ifp;
1148
1149 struct route_event rt_ev;
1150 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_RESOLVED);
1151 /*
1152 * This means a router's state has changed from
1153 * non-reachable to probably reachable, and might
1154 * affect the status of associated prefixes..
1155 * We already have a reference on rt. Don't need to
1156 * take one for the unlock/lock.
1157 */
1158 RT_UNLOCK(rt);
1159 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1160 lck_mtx_lock(rnh_lock);
1161 rnh = rt_tables[AF_INET6];
1162
1163 if (rnh != NULL) {
1164 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1165 (void *)&rt_ev);
1166 }
1167 lck_mtx_unlock(rnh_lock);
1168 lck_mtx_lock(nd6_mutex);
1169 pfxlist_onlink_check();
1170 lck_mtx_unlock(nd6_mutex);
1171 RT_LOCK(rt);
1172 }
1173 } else {
1174 int llchange = 0;
1175
1176 /*
1177 * Check if the link-layer address has changed or not.
1178 */
1179 if (lladdr == NULL) {
1180 llchange = 0;
1181 } else {
1182 if (sdl->sdl_alen) {
1183 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) {
1184 llchange = 1;
1185 } else {
1186 llchange = 0;
1187 }
1188 } else {
1189 llchange = 1;
1190 }
1191 }
1192
1193 /*
1194 * This is VERY complex. Look at it with care.
1195 *
1196 * override solicit lladdr llchange action
1197 * (L: record lladdr)
1198 *
1199 * 0 0 n -- (2c)
1200 * 0 0 y n (2b) L
1201 * 0 0 y y (1) REACHABLE->STALE
1202 * 0 1 n -- (2c) *->REACHABLE
1203 * 0 1 y n (2b) L *->REACHABLE
1204 * 0 1 y y (1) REACHABLE->STALE
1205 * 1 0 n -- (2a)
1206 * 1 0 y n (2a) L
1207 * 1 0 y y (2a) L *->STALE
1208 * 1 1 n -- (2a) *->REACHABLE
1209 * 1 1 y n (2a) L *->REACHABLE
1210 * 1 1 y y (2a) L *->REACHABLE
1211 */
1212 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */
1213 /*
1214 * If state is REACHABLE, make it STALE.
1215 * no other updates should be done.
1216 */
1217 if (ln->ln_state == ND6_LLINFO_REACHABLE) {
1218 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1219 ln_setexpire(ln, timenow + nd6_gctimer);
1220 }
1221 RT_REMREF_LOCKED(rt);
1222 RT_UNLOCK(rt);
1223 drop_reason = DROP_REASON_IP6_BAD_ND_STATE;
1224 goto freeit;
1225 } else if (is_override /* (2a) */
1226 || (!is_override && (lladdr && !llchange)) /* (2b) */
1227 || !lladdr) { /* (2c) */
1228 /*
1229 * Update link-local address, if any.
1230 */
1231 if (lladdr) {
1232 sdl->sdl_alen = ifp->if_addrlen;
1233 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
1234 }
1235
1236 /*
1237 * If solicited, make the state REACHABLE.
1238 * If not solicited and the link-layer address was
1239 * changed, make it STALE.
1240 */
1241 if (is_solicited) {
1242 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
1243 if (ln->ln_expire != 0) {
1244 struct nd_ifinfo *__single ndi = NULL;
1245
1246 ndi = ND_IFINFO(ifp);
1247 VERIFY(ndi != NULL && ndi->initialized);
1248 lck_mtx_lock(&ndi->lock);
1249 ln_setexpire(ln,
1250 timenow + ndi->reachable);
1251 lck_mtx_unlock(&ndi->lock);
1252 RT_UNLOCK(rt);
1253 lck_mtx_lock(rnh_lock);
1254 nd6_sched_timeout(NULL, NULL);
1255 lck_mtx_unlock(rnh_lock);
1256 RT_LOCK(rt);
1257 }
1258 } else {
1259 if (lladdr && llchange) {
1260 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_STALE);
1261 ln_setexpire(ln, timenow + nd6_gctimer);
1262 }
1263 }
1264
1265 /*
1266 * XXX
1267 * The above is somewhat convoluted, for now just
1268 * issue a callback for LLENTRY changed.
1269 */
1270 /* Enqueue work item to invoke callback for this route entry */
1271 if (llchange) {
1272 route_event_enqueue_nwk_wq_entry(rt, NULL,
1273 ROUTE_LLENTRY_CHANGED, NULL, TRUE);
1274 }
1275
1276 /*
1277 * If the router's link-layer address has changed,
1278 * notify routes using this as gateway so they can
1279 * update any cached information.
1280 */
1281 if (ln->ln_router && is_router && llchange) {
1282 struct radix_node_head *__single rnh = NULL;
1283 struct in6_addr rt_addr = SIN6(rt_key(rt))->sin6_addr;
1284 ifnet_ref_t rt_ifp = rt->rt_ifp;
1285 struct route_event rt_ev;
1286 route_event_init(&rt_ev, rt, NULL, ROUTE_LLENTRY_CHANGED);
1287
1288 /*
1289 * This means a router's state has changed from
1290 * non-reachable to probably reachable, and might
1291 * affect the status of associated prefixes..
1292 *
1293 * We already have a valid rt reference here.
1294 * We don't need to take another one for unlock/lock.
1295 */
1296 RT_UNLOCK(rt);
1297 defrouter_set_reachability(&rt_addr, rt_ifp, TRUE);
1298 lck_mtx_lock(rnh_lock);
1299 rnh = rt_tables[AF_INET6];
1300
1301 if (rnh != NULL) {
1302 (void) rnh->rnh_walktree(rnh, route_event_walktree,
1303 (void *)&rt_ev);
1304 }
1305 lck_mtx_unlock(rnh_lock);
1306 RT_LOCK(rt);
1307 }
1308 }
1309
1310 if (ln->ln_router && !is_router) {
1311 /*
1312 * The peer dropped the router flag.
1313 * Remove the sender from the Default Router List and
1314 * update the Destination Cache entries.
1315 */
1316 struct nd_defrouter *__single dr;
1317 struct in6_addr *__single in6;
1318 ifnet_ref_t rt_ifp = rt->rt_ifp;
1319
1320 in6 = &SIN6(rt_key(rt))->sin6_addr;
1321
1322 RT_UNLOCK(rt);
1323 lck_mtx_lock(nd6_mutex);
1324 /*
1325 * XXX Handle router lists for route information option
1326 * as well.
1327 */
1328 dr = defrouter_lookup(NULL, in6, rt_ifp);
1329 if (dr) {
1330 TAILQ_REMOVE(&nd_defrouter_list, dr, dr_entry);
1331 defrtrlist_del(dr, NULL);
1332 NDDR_REMREF(dr); /* remove list reference */
1333 NDDR_REMREF(dr);
1334 lck_mtx_unlock(nd6_mutex);
1335 } else {
1336 lck_mtx_unlock(nd6_mutex);
1337 /*
1338 * Even if the neighbor is not in the
1339 * default router list, the neighbor
1340 * may be used as a next hop for some
1341 * destinations (e.g. redirect case).
1342 * So we must call rt6_flush explicitly.
1343 */
1344 rt6_flush(&ip6->ip6_src, rt_ifp);
1345 }
1346 RT_LOCK(rt);
1347 }
1348 ln->ln_router = (short)is_router;
1349 }
1350
1351 if (send_nc_alive_kev && (ifp->if_addrlen == IF_LLREACH_MAXLEN)) {
1352 struct kev_msg ev_msg;
1353 struct kev_nd6_ndalive nd6_ndalive;
1354 bzero(&ev_msg, sizeof(ev_msg));
1355 bzero(&nd6_ndalive, sizeof(nd6_ndalive));
1356 ev_msg.vendor_code = KEV_VENDOR_APPLE;
1357 ev_msg.kev_class = KEV_NETWORK_CLASS;
1358 ev_msg.kev_subclass = KEV_ND6_SUBCLASS;
1359 ev_msg.event_code = KEV_ND6_NDALIVE;
1360
1361 nd6_ndalive.link_data.if_family = ifp->if_family;
1362 nd6_ndalive.link_data.if_unit = ifp->if_unit;
1363 strlcpy(nd6_ndalive.link_data.if_name,
1364 ifp->if_name,
1365 sizeof(nd6_ndalive.link_data.if_name));
1366 ev_msg.dv[0].data_ptr = &nd6_ndalive;
1367 ev_msg.dv[0].data_length =
1368 sizeof(nd6_ndalive);
1369 dlil_post_complete_msg(NULL, &ev_msg);
1370 }
1371
1372 RT_LOCK_ASSERT_HELD(rt);
1373 rt->rt_flags &= ~RTF_REJECT;
1374
1375 /* cache the gateway (sender HW) address */
1376 nd6_llreach_alloc(rt, ifp, LLADDR(sdl), sdl->sdl_alen, TRUE);
1377
1378 /* update the llinfo, send a queued packet if there is one */
1379 ln->ln_asked = 0;
1380 if (ln->ln_hold != NULL) {
1381 mbuf_ref_t m_hold, m_hold_next;
1382 struct sockaddr_in6 sin6;
1383
1384 rtkey_to_sa6(rt, &sin6);
1385 /*
1386 * reset the ln_hold in advance, to explicitly
1387 * prevent a ln_hold lookup in nd6_output()
1388 * (wouldn't happen, though...)
1389 */
1390 m_hold = ln->ln_hold;
1391 ln->ln_hold = NULL;
1392 for (; m_hold; m_hold = m_hold_next) {
1393 m_hold_next = m_hold->m_nextpkt;
1394 m_hold->m_nextpkt = NULL;
1395 /*
1396 * we assume ifp is not a loopback here, so just set
1397 * the 2nd argument as the 1st one.
1398 */
1399 RT_UNLOCK(rt);
1400 nd6_output(ifp, ifp, m_hold, &sin6, rt, NULL);
1401 RT_LOCK_SPIN(rt);
1402 }
1403 }
1404 RT_REMREF_LOCKED(rt);
1405 RT_UNLOCK(rt);
1406 m_freem(m);
1407 return;
1408
1409 bad:
1410 icmp6stat.icp6s_badna++;
1411 /* fall through */
1412
1413 freeit:
1414 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, drop_reason, NULL, 0);
1415 return;
1416 }
1417
1418 /*
1419 * Neighbor advertisement output handling.
1420 *
1421 * Based on RFC 2461
1422 *
1423 * the following items are not implemented yet:
1424 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
1425 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
1426 *
1427 * tlladdr - 1 if include target link-layer address
1428 * sdl0 - sockaddr_dl (= proxy NA) or NULL
1429 */
1430 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)1431 nd6_na_output(
1432 struct ifnet *ifp,
1433 const struct in6_addr *daddr6_0,
1434 const struct in6_addr *taddr6,
1435 uint32_t flags,
1436 int tlladdr, /* 1 if include target link-layer address */
1437 struct sockaddr *sdl0) /* sockaddr_dl (= proxy NA) or NULL */
1438 {
1439 mbuf_ref_t m;
1440 struct ip6_hdr *ip6;
1441 struct nd_neighbor_advert *nd_na;
1442 struct ip6_moptions *__single im6o = NULL;
1443 caddr_t mac = NULL;
1444 struct route_in6 ro;
1445 struct in6_addr *__single src;
1446 struct in6_addr src_storage, daddr6;
1447 struct in6_ifaddr *__single ia;
1448 struct sockaddr_in6 dst_sa;
1449 int icmp6len, maxlen, error;
1450 ifnet_ref_t outif = NULL;
1451
1452 struct ip6_out_args ip6oa;
1453 bzero(&ro, sizeof(ro));
1454
1455 daddr6 = *daddr6_0; /* make a local copy for modification */
1456
1457 bzero(&ip6oa, sizeof(ip6oa));
1458 ip6oa.ip6oa_boundif = ifp->if_index;
1459 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR |
1460 IP6OAF_AWDL_UNRESTRICTED | IP6OAF_INTCOPROC_ALLOWED |
1461 IP6OAF_MANAGEMENT_ALLOWED;
1462 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
1463 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
1464
1465 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
1466
1467 /* estimate the size of message */
1468 maxlen = sizeof(*ip6) + sizeof(*nd_na);
1469 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
1470 if (max_linkhdr + maxlen >= MCLBYTES) {
1471 #if DIAGNOSTIC
1472 printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
1473 "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
1474 #endif
1475 return;
1476 }
1477
1478 MGETHDR(m, M_DONTWAIT, MT_DATA); /* XXXMAC: mac_create_mbuf_linklayer() probably */
1479 if (m && max_linkhdr + maxlen >= MHLEN) {
1480 MCLGET(m, M_DONTWAIT);
1481 if ((m->m_flags & M_EXT) == 0) {
1482 m_free(m);
1483 m = NULL;
1484 }
1485 }
1486 if (m == NULL) {
1487 return;
1488 }
1489 m->m_pkthdr.rcvif = NULL;
1490
1491 if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1492 m->m_flags |= M_MCAST;
1493
1494 im6o = ip6_allocmoptions(Z_NOWAIT);
1495 if (im6o == NULL) {
1496 m_freem(m);
1497 return;
1498 }
1499
1500 im6o->im6o_multicast_ifp = ifp;
1501 im6o->im6o_multicast_hlim = IPV6_MAXHLIM;
1502 im6o->im6o_multicast_loop = 0;
1503 }
1504
1505 icmp6len = sizeof(*nd_na);
1506 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1507 m->m_data += max_linkhdr; /* or MH_ALIGN() equivalent? */
1508
1509 /* fill neighbor advertisement packet */
1510 ip6 = mtod(m, struct ip6_hdr *);
1511 ip6->ip6_flow = 0;
1512 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1513 ip6->ip6_vfc |= IPV6_VERSION;
1514 ip6->ip6_nxt = IPPROTO_ICMPV6;
1515 ip6->ip6_hlim = IPV6_MAXHLIM;
1516 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1517 /* reply to DAD */
1518 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1519 daddr6.s6_addr16[1] = 0;
1520 daddr6.s6_addr32[1] = 0;
1521 daddr6.s6_addr32[2] = 0;
1522 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1523 if (in6_setscope(&daddr6, ifp, NULL)) {
1524 m_drop(m, DROPTAP_FLAG_DIR_OUT | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
1525 goto exit;
1526 }
1527
1528 flags &= ~ND_NA_FLAG_SOLICITED;
1529 } else {
1530 ip6->ip6_dst = daddr6;
1531 ip6_output_setdstifscope(m, ifp->if_index, NULL);
1532 }
1533
1534 SOCKADDR_ZERO(&dst_sa, sizeof(struct sockaddr_in6));
1535 dst_sa.sin6_family = AF_INET6;
1536 dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1537 dst_sa.sin6_addr = daddr6;
1538
1539 /*
1540 * Select a source whose scope is the same as that of the dest.
1541 */
1542 SOCKADDR_COPY(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1543 src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &src_storage,
1544 ip6oa.ip6oa_boundif, &error);
1545 if (src == NULL) {
1546 nd6log(info, "nd6_na_output: source can't be "
1547 "determined: dst=%s, error=%d\n",
1548 ip6_sprintf(&dst_sa.sin6_addr), error);
1549 m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_ND_STATE, NULL, 0);
1550 goto exit;
1551 }
1552 ip6->ip6_src = *src;
1553
1554 /*
1555 * RFC 4429 requires not setting "override" flag on NA packets sent
1556 * from optimistic addresses.
1557 */
1558 ia = in6ifa_ifpwithaddr(ifp, src);
1559 ip6_output_setsrcifscope(m, ifp->if_index, ia);
1560 if (ia != NULL) {
1561 if (ia->ia6_flags & IN6_IFF_OPTIMISTIC) {
1562 flags &= ~ND_NA_FLAG_OVERRIDE;
1563 }
1564 ifa_remref(&ia->ia_ifa);
1565 }
1566
1567 nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1568 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1569 nd_na->nd_na_code = 0;
1570 nd_na->nd_na_target = *taddr6;
1571 in6_clearscope(&nd_na->nd_na_target); /* XXX */
1572
1573 /*
1574 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1575 * see nd6_ns_input() for details.
1576 * Basically, if NS packet is sent to unicast/anycast addr,
1577 * target lladdr option SHOULD NOT be included.
1578 */
1579 if (tlladdr) {
1580 /*
1581 * sdl0 != NULL indicates proxy NA. If we do proxy, use
1582 * lladdr in sdl0. If we are not proxying (sending NA for
1583 * my address) use lladdr configured for the interface.
1584 */
1585 if (sdl0 == NULL) {
1586 mac = nd6_ifptomac(ifp);
1587 } else if (sdl0->sa_family == AF_LINK) {
1588 struct sockaddr_dl *sdl;
1589 sdl = SDL(sdl0);
1590 if (sdl->sdl_alen == ifp->if_addrlen) {
1591 mac = LLADDR(sdl);
1592 }
1593 }
1594 }
1595 if (tlladdr && mac) {
1596 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1597 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1598
1599 /* roundup to 8 bytes alignment! */
1600 optlen = (optlen + 7) & ~7;
1601
1602 m->m_pkthdr.len += optlen;
1603 m->m_len += optlen;
1604 icmp6len += optlen;
1605 bzero((caddr_t)nd_opt, optlen);
1606 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1607 nd_opt->nd_opt_len = (uint8_t)(optlen >> 3);
1608 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1609 } else {
1610 flags &= ~ND_NA_FLAG_OVERRIDE;
1611 }
1612
1613 ip6->ip6_plen = htons((u_short)icmp6len);
1614 nd_na->nd_na_flags_reserved = flags;
1615 nd_na->nd_na_cksum = 0;
1616 nd_na->nd_na_cksum =
1617 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1618
1619 m->m_pkthdr.pkt_flags |= PKTF_INET6_RESOLVE;
1620
1621 if (ifp->if_eflags & IFEF_TXSTART) {
1622 /* Use control service class if the interface supports
1623 * transmit-start model.
1624 */
1625 (void) m_set_service_class(m, MBUF_SC_CTL);
1626 }
1627
1628 ip6oa.ip6oa_flags |= IP6OAF_SKIP_PF;
1629 ip6oa.ip6oa_flags |= IP6OAF_DONT_FRAG;
1630 ip6_output(m, NULL, NULL, IPV6_OUTARGS, im6o, &outif, &ip6oa);
1631 if (outif) {
1632 icmp6_ifstat_inc(outif, ifs6_out_msg);
1633 icmp6_ifstat_inc(outif, ifs6_out_neighboradvert);
1634 ifnet_release(outif);
1635 }
1636 icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1637
1638 exit:
1639 if (im6o != NULL) {
1640 IM6O_REMREF(im6o);
1641 }
1642
1643 ROUTE_RELEASE(&ro);
1644 }
1645
1646 TAILQ_HEAD(dadq_head, dadq);
1647 struct dadq {
1648 decl_lck_mtx_data(, dad_lock);
1649 u_int32_t dad_refcount; /* reference count */
1650 int dad_attached;
1651 TAILQ_ENTRY(dadq) dad_list;
1652 struct ifaddr *dad_ifa;
1653 int dad_count; /* max NS to send */
1654 int dad_ns_tcount; /* # of trials to send NS */
1655 int dad_ns_ocount; /* NS sent so far */
1656 int dad_ns_icount;
1657 int dad_na_icount;
1658 int dad_ns_lcount; /* looped back NS */
1659 int dad_loopbackprobe; /* probing state for loopback detection */
1660 uint8_t dad_lladdr[ETHER_ADDR_LEN];
1661 uint8_t dad_lladdrlen;
1662 #define ND_OPT_NONCE_LEN32 \
1663 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1664 uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1665 uint32_t dad_same_nonce_count; /* # of consecutive times we've ignored DAD failure because of optimistic DAD */
1666 };
1667
1668 static KALLOC_TYPE_DEFINE(dad_zone, struct dadq, NET_KT_DEFAULT);
1669 static struct dadq_head dadq;
1670
1671 void
nd6_nbr_init(void)1672 nd6_nbr_init(void)
1673 {
1674 int i;
1675
1676 TAILQ_INIT(&dadq);
1677
1678 SOCKADDR_ZERO(&hostrtmask, sizeof hostrtmask);
1679 hostrtmask.sin6_family = AF_INET6;
1680 hostrtmask.sin6_len = sizeof hostrtmask;
1681 for (i = 0; i < sizeof hostrtmask.sin6_addr; ++i) {
1682 hostrtmask.sin6_addr.s6_addr[i] = 0xff;
1683 }
1684 }
1685
1686 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * nonce)1687 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1688 {
1689 struct dadq *__single dp;
1690 boolean_t same_nonce = false;
1691
1692 lck_mtx_lock(&dad6_mutex);
1693 for (dp = dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1694 DAD_LOCK_SPIN(dp);
1695 if (dp->dad_ifa != ifa) {
1696 DAD_UNLOCK(dp);
1697 continue;
1698 }
1699
1700 /*
1701 * Skip if the nonce matches the received one.
1702 * +2 in the length is required because of type and
1703 * length fields are included in a header.
1704 */
1705 same_nonce = nonce != NULL &&
1706 nonce->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1707 memcmp(&nonce->nd_opt_nonce[0], &dp->dad_nonce[0],
1708 ND_OPT_NONCE_LEN) == 0;
1709
1710 if (same_nonce &&
1711 dp->dad_same_nonce_count <= nd6_dad_nonce_max_count) {
1712 nd6log(error, "%s: a looped back NS message is "
1713 "detected during DAD for if=%s %s. Ignoring.\n",
1714 __func__,
1715 if_name(ifa->ifa_ifp),
1716 ip6_sprintf(IFA_IN6(ifa)));
1717 dp->dad_same_nonce_count++;
1718 dp->dad_ns_lcount++;
1719 ++ip6stat.ip6s_dad_loopcount;
1720 DAD_UNLOCK(dp);
1721 continue;
1722 } else if (!same_nonce) {
1723 // Not the same nonce, reset counter
1724 dp->dad_same_nonce_count = 1;
1725 }
1726
1727 DAD_ADDREF_LOCKED(dp);
1728 DAD_UNLOCK(dp);
1729 break;
1730 }
1731 lck_mtx_unlock(&dad6_mutex);
1732 return dp;
1733 }
1734
1735 void
nd6_dad_stoptimer(struct ifaddr * ifa)1736 nd6_dad_stoptimer(
1737 struct ifaddr *ifa)
1738 {
1739 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1740 }
1741
1742 /*
1743 * Start Duplicate Address Detection (DAD) for specified interface address.
1744 */
1745 void
nd6_dad_start(struct ifaddr * ifa,int * tick_delay)1746 nd6_dad_start(
1747 struct ifaddr *ifa,
1748 int *tick_delay) /* minimum delay ticks for IFF_UP event */
1749 {
1750 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1751 struct dadq *__single dp;
1752
1753 if (ifa->ifa_ifp == NULL) {
1754 panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1755 }
1756
1757 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1758 __func__,
1759 ip6_sprintf(&ia->ia_addr.sin6_addr),
1760 if_name(ia->ia_ifp),
1761 ia->ia6_flags);
1762
1763 /*
1764 * If we don't need DAD, don't do it.
1765 * There are several cases:
1766 * - DAD is disabled (ip6_dad_count == 0)
1767 * - the interface address is anycast
1768 */
1769 IFA_LOCK(&ia->ia_ifa);
1770 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
1771 nd6log0(info,
1772 "nd6_dad_start: not a tentative or optimistic address "
1773 "%s(%s)\n",
1774 ip6_sprintf(&ia->ia_addr.sin6_addr),
1775 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1776 IFA_UNLOCK(&ia->ia_ifa);
1777 return;
1778 }
1779 if (!ip6_dad_count || (ia->ia6_flags & IN6_IFF_ANYCAST) != 0) {
1780 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
1781 IFA_UNLOCK(&ia->ia_ifa);
1782 return;
1783 }
1784 IFA_UNLOCK(&ia->ia_ifa);
1785
1786 if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1787 (ifa->ifa_ifp->if_eflags & IFEF_IPV6_ND6ALT)) {
1788 return;
1789 }
1790 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1791 DAD_REMREF(dp);
1792 /* DAD already in progress */
1793 return;
1794 }
1795
1796 dp = zalloc_flags(dad_zone, Z_WAITOK | Z_ZERO);
1797 lck_mtx_init(&dp->dad_lock, &ifa_mtx_grp, &ifa_mtx_attr);
1798
1799 /* Callee adds one reference for us */
1800 dp = nd6_dad_attach(dp, ifa);
1801
1802 nd6log0(info, "%s: starting %sDAD %sfor %s\n",
1803 if_name(ifa->ifa_ifp),
1804 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) ? "optimistic " : "",
1805 (tick_delay == NULL) ? "immediately " : "",
1806 ip6_sprintf(&ia->ia_addr.sin6_addr));
1807
1808 /*
1809 * Send NS packet for DAD, ip6_dad_count times.
1810 * Note that we must delay the first transmission, if this is the
1811 * first packet to be sent from the interface after interface
1812 * (re)initialization.
1813 */
1814 if (tick_delay == NULL) {
1815 u_int32_t retrans;
1816 struct nd_ifinfo *__single ndi = NULL;
1817
1818 nd6_dad_ns_output(dp, ifa);
1819 ndi = ND_IFINFO(ifa->ifa_ifp);
1820 VERIFY(ndi != NULL && ndi->initialized);
1821 lck_mtx_lock(&ndi->lock);
1822 retrans = ndi->retrans * hz / 1000;
1823 lck_mtx_unlock(&ndi->lock);
1824 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
1825 } else {
1826 int ntick;
1827
1828 if (*tick_delay == 0) {
1829 ntick = random() % (MAX_RTR_SOLICITATION_DELAY * hz);
1830 } else {
1831 ntick = *tick_delay + random() % (hz / 2);
1832 }
1833 *tick_delay = ntick;
1834 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa,
1835 ntick);
1836 }
1837
1838 DAD_REMREF(dp); /* drop our reference */
1839 }
1840
1841 static struct dadq *
nd6_dad_attach(struct dadq * dp,struct ifaddr * ifa)1842 nd6_dad_attach(struct dadq *dp, struct ifaddr *ifa)
1843 {
1844 lck_mtx_lock(&dad6_mutex);
1845 DAD_LOCK(dp);
1846 dp->dad_ifa = ifa;
1847 ifa_addref(ifa); /* for dad_ifa */
1848 dp->dad_count = ip6_dad_count;
1849 dp->dad_ns_icount = dp->dad_na_icount = 0;
1850 dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1851 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1852 VERIFY(!dp->dad_attached);
1853 dp->dad_same_nonce_count = 1;
1854 dp->dad_attached = 1;
1855 dp->dad_lladdrlen = 0;
1856 DAD_ADDREF_LOCKED(dp); /* for caller */
1857 DAD_ADDREF_LOCKED(dp); /* for dadq_head list */
1858 TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1859 DAD_UNLOCK(dp);
1860 lck_mtx_unlock(&dad6_mutex);
1861
1862 return dp;
1863 }
1864
1865 static void
nd6_dad_detach(struct dadq * dp,struct ifaddr * ifa)1866 nd6_dad_detach(struct dadq *dp, struct ifaddr *ifa)
1867 {
1868 int detached;
1869
1870 lck_mtx_lock(&dad6_mutex);
1871 DAD_LOCK(dp);
1872 if ((detached = dp->dad_attached)) {
1873 VERIFY(dp->dad_ifa == ifa);
1874 TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1875 dp->dad_list.tqe_next = NULL;
1876 dp->dad_list.tqe_prev = NULL;
1877 dp->dad_attached = 0;
1878 }
1879 DAD_UNLOCK(dp);
1880 lck_mtx_unlock(&dad6_mutex);
1881 if (detached) {
1882 DAD_REMREF(dp); /* drop dadq_head reference */
1883 }
1884 }
1885
1886 /*
1887 * terminate DAD unconditionally. used for address removals.
1888 */
1889 void
nd6_dad_stop(struct ifaddr * ifa)1890 nd6_dad_stop(struct ifaddr *ifa)
1891 {
1892 struct dadq *__single dp;
1893
1894 dp = nd6_dad_find(ifa, NULL);
1895 if (!dp) {
1896 /* DAD wasn't started yet */
1897 return;
1898 }
1899
1900 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
1901
1902 nd6_dad_detach(dp, ifa);
1903 DAD_REMREF(dp); /* drop our reference */
1904 }
1905
1906 static void
nd6_unsol_na_output(struct ifaddr * ifa)1907 nd6_unsol_na_output(struct ifaddr *ifa)
1908 {
1909 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1910 ifnet_ref_t ifp = ifa->ifa_ifp;
1911 struct in6_addr saddr6, taddr6;
1912
1913 if ((ifp->if_flags & IFF_UP) == 0 ||
1914 (ifp->if_flags & IFF_RUNNING) == 0 ||
1915 (ifp->if_eflags & IFEF_IPV6_ND6ALT) != 0) {
1916 return;
1917 }
1918
1919 IFA_LOCK_SPIN(&ia->ia_ifa);
1920 taddr6 = ia->ia_addr.sin6_addr;
1921 IFA_UNLOCK(&ia->ia_ifa);
1922 if (in6_setscope(&taddr6, ifp, NULL) != 0) {
1923 return;
1924 }
1925 saddr6 = in6addr_linklocal_allnodes;
1926 if (in6_setscope(&saddr6, ifp, NULL) != 0) {
1927 return;
1928 }
1929
1930 nd6log(info, "%s: sending unsolicited NA\n",
1931 if_name(ifa->ifa_ifp));
1932
1933 nd6_na_output(ifp, &saddr6, &taddr6, ND_NA_FLAG_OVERRIDE, 1, NULL);
1934 }
1935
1936 static void
nd6_dad_timer(struct ifaddr * ifa)1937 nd6_dad_timer(struct ifaddr *ifa)
1938 {
1939 struct in6_ifaddr *__single ia = ifatoia6(ifa);
1940 struct dadq *__single dp = NULL;
1941 struct nd_ifinfo *ndi = NULL;
1942 u_int32_t retrans;
1943
1944 /* Sanity check */
1945 if (ia == NULL) {
1946 nd6log0(error, "nd6_dad_timer: called with null parameter\n");
1947 goto done;
1948 }
1949
1950 nd6log2(debug, "%s - %s ifp %s ia6_flags 0x%x\n",
1951 __func__,
1952 ip6_sprintf(&ia->ia_addr.sin6_addr),
1953 if_name(ia->ia_ifp),
1954 ia->ia6_flags);
1955
1956 dp = nd6_dad_find(ifa, NULL);
1957 if (dp == NULL) {
1958 nd6log0(error, "nd6_dad_timer: DAD structure not found\n");
1959 goto done;
1960 }
1961 IFA_LOCK(&ia->ia_ifa);
1962 if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1963 nd6log0(error, "nd6_dad_timer: called with duplicated address "
1964 "%s(%s)\n",
1965 ip6_sprintf(&ia->ia_addr.sin6_addr),
1966 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1967 IFA_UNLOCK(&ia->ia_ifa);
1968 goto done;
1969 }
1970 if ((ia->ia6_flags & IN6_IFF_DADPROGRESS) == 0) {
1971 nd6log0(error, "nd6_dad_timer: not a tentative or optimistic "
1972 "address %s(%s)\n",
1973 ip6_sprintf(&ia->ia_addr.sin6_addr),
1974 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1975 IFA_UNLOCK(&ia->ia_ifa);
1976 goto done;
1977 }
1978 IFA_UNLOCK(&ia->ia_ifa);
1979
1980 /* timeouted with IFF_{RUNNING,UP} check */
1981 DAD_LOCK(dp);
1982 if (dp->dad_ns_tcount > dad_maxtry) {
1983 DAD_UNLOCK(dp);
1984 nd6log0(info, "%s: could not run DAD, driver problem?\n",
1985 if_name(ifa->ifa_ifp));
1986
1987 nd6_dad_detach(dp, ifa);
1988 goto done;
1989 }
1990
1991 /* Need more checks? */
1992 if (dp->dad_ns_ocount < dp->dad_count) {
1993 DAD_UNLOCK(dp);
1994 /*
1995 * We have more NS to go. Send NS packet for DAD.
1996 */
1997 nd6_dad_ns_output(dp, ifa);
1998 ndi = ND_IFINFO(ifa->ifa_ifp);
1999 VERIFY(ndi != NULL && ndi->initialized);
2000 lck_mtx_lock(&ndi->lock);
2001 retrans = ndi->retrans * hz / 1000;
2002 lck_mtx_unlock(&ndi->lock);
2003 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2004 } else {
2005 /*
2006 * We have transmitted sufficient number of DAD packets.
2007 * See what we've got.
2008 */
2009 if (dp->dad_na_icount > 0 || dp->dad_ns_icount) {
2010 /* We've seen NS or NA, means DAD has failed. */
2011 DAD_UNLOCK(dp);
2012 nd6log0(info,
2013 "%s: duplicate IPv6 address %s if:%s [timer]\n",
2014 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2015 if_name(ia->ia_ifp));
2016 nd6_dad_duplicated(ifa);
2017 /* (*dp) will be freed in nd6_dad_duplicated() */
2018 #if SKYWALK
2019 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2020 #endif /* SKYWALK */
2021 } else if (dad_enhanced != 0 &&
2022 dp->dad_ns_lcount > 0 &&
2023 dp->dad_ns_lcount > dp->dad_loopbackprobe &&
2024 dp->dad_same_nonce_count > 0 &&
2025 dp->dad_same_nonce_count > nd6_dad_nonce_max_count) {
2026 dp->dad_loopbackprobe = dp->dad_ns_lcount;
2027 dp->dad_count =
2028 dp->dad_ns_ocount + dad_maxtry - 1;
2029 DAD_UNLOCK(dp);
2030 ndi = ND_IFINFO(ifa->ifa_ifp);
2031 VERIFY(ndi != NULL && ndi->initialized);
2032 lck_mtx_lock(&ndi->lock);
2033 retrans = ndi->retrans * hz / 1000;
2034 lck_mtx_unlock(&ndi->lock);
2035
2036 /*
2037 * Sec. 4.1 in RFC 7527 requires transmission of
2038 * additional probes until the loopback condition
2039 * becomes clear when a looped back probe is detected.
2040 */
2041 nd6log0(info,
2042 "%s: a looped back NS message is detected during DAD for %s. Another DAD probe is being sent on interface %s.\n",
2043 __func__, ip6_sprintf(&ia->ia_addr.sin6_addr),
2044 if_name(ia->ia_ifp));
2045 /*
2046 * Send an NS immediately and increase dad_count by
2047 * nd6_mmaxtries - 1.
2048 */
2049 nd6_dad_ns_output(dp, ifa);
2050 timeout((void (*)(void *))nd6_dad_timer, (void *)ifa, retrans);
2051 goto done;
2052 } else {
2053 boolean_t txunsolna;
2054 DAD_UNLOCK(dp);
2055 /*
2056 * We are done with DAD. No NA came, no NS came.
2057 * No duplicate address found.
2058 */
2059 IFA_LOCK_SPIN(&ia->ia_ifa);
2060 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2061 IFA_UNLOCK(&ia->ia_ifa);
2062
2063 ndi = ND_IFINFO(ifa->ifa_ifp);
2064 VERIFY(ndi != NULL && ndi->initialized);
2065 lck_mtx_lock(&ndi->lock);
2066 txunsolna = (ndi->flags & ND6_IFF_REPLICATED) != 0;
2067 lck_mtx_unlock(&ndi->lock);
2068
2069 if (txunsolna) {
2070 nd6_unsol_na_output(ifa);
2071 }
2072
2073 nd6log0(info,
2074 "%s: DAD complete for %s - no duplicates found %s\n",
2075 if_name(ifa->ifa_ifp),
2076 ip6_sprintf(&ia->ia_addr.sin6_addr),
2077 txunsolna ? ", tx unsolicited NA with O=1" : ".");
2078
2079 if (dp->dad_ns_lcount > 0) {
2080 nd6log0(info,
2081 "%s: DAD completed while "
2082 "a looped back NS message is detected "
2083 "during DAD for %s om interface %s\n",
2084 __func__,
2085 ip6_sprintf(&ia->ia_addr.sin6_addr),
2086 if_name(ia->ia_ifp));
2087 }
2088
2089 in6_post_msg(ia->ia_ifp, KEV_INET6_NEW_USER_ADDR, ia,
2090 dp->dad_lladdr, ETHER_ADDR_LEN);
2091 nd6_dad_detach(dp, ifa);
2092 #if SKYWALK
2093 SK_NXS_MS_IF_ADDR_GENCNT_INC(ia->ia_ifp);
2094 #endif /* SKYWALK */
2095 }
2096 }
2097
2098 done:
2099 if (dp != NULL) {
2100 DAD_REMREF(dp); /* drop our reference */
2101 }
2102 }
2103
2104 static void
nd6_dad_duplicated(struct ifaddr * ifa)2105 nd6_dad_duplicated(struct ifaddr *ifa)
2106 {
2107 struct in6_ifaddr *__single ia = ifatoia6(ifa);
2108 struct dadq *__single dp;
2109 ifnet_ref_t ifp = ifa->ifa_ifp;
2110 boolean_t candisable;
2111
2112 dp = nd6_dad_find(ifa, NULL);
2113 if (dp == NULL) {
2114 log(LOG_ERR, "%s: DAD structure not found.\n", __func__);
2115 return;
2116 }
2117 IFA_LOCK(&ia->ia_ifa);
2118 DAD_LOCK(dp);
2119 nd6log(error, "%s: NS in/out/loopback=%d/%d/%d, NA in=%d\n",
2120 __func__, dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
2121 dp->dad_na_icount);
2122 candisable = FALSE;
2123
2124 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr) &&
2125 !(ia->ia6_flags & IN6_IFF_SECURED)) {
2126 struct in6_addr in6;
2127 struct ifaddr *__single llifa = NULL;
2128 struct sockaddr_dl *sdl = NULL;
2129 uint8_t *lladdr = dp->dad_lladdr;
2130 uint8_t lladdrlen = dp->dad_lladdrlen;
2131
2132 /*
2133 * To avoid over-reaction, we only apply this logic when we are
2134 * very sure that hardware addresses are supposed to be unique.
2135 */
2136 switch (ifp->if_type) {
2137 case IFT_BRIDGE:
2138 case IFT_ETHER:
2139 case IFT_FDDI:
2140 case IFT_ATM:
2141 case IFT_IEEE1394:
2142 #ifdef IFT_IEEE80211
2143 case IFT_IEEE80211:
2144 #endif
2145 /*
2146 * Check if our hardware address matches the
2147 * link layer information received in the
2148 * NS/NA
2149 */
2150 llifa = ifp->if_lladdr;
2151 IFA_LOCK(llifa);
2152 sdl = SDL(llifa->ifa_addr);
2153 if (lladdrlen == sdl->sdl_alen &&
2154 bcmp(lladdr, LLADDR(sdl), lladdrlen) == 0) {
2155 candisable = TRUE;
2156 }
2157 IFA_UNLOCK(llifa);
2158
2159 in6 = ia->ia_addr.sin6_addr;
2160 if (in6_iid_from_hw(ifp, &in6) != 0) {
2161 break;
2162 }
2163
2164 /* Refine decision about whether IPv6 can be disabled */
2165 if (candisable &&
2166 !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2167 /*
2168 * Apply this logic only to the embedded MAC
2169 * address form of link-local IPv6 address.
2170 */
2171 candisable = FALSE;
2172 } else if (lladdr == NULL &&
2173 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
2174 /*
2175 * We received a NA with no target link-layer
2176 * address option. This means that someone else
2177 * has our address. Mark it as a hardware
2178 * duplicate so we disable IPv6 later on.
2179 */
2180 candisable = TRUE;
2181 }
2182 break;
2183 default:
2184 break;
2185 }
2186 }
2187 DAD_UNLOCK(dp);
2188
2189 ia->ia6_flags &= ~IN6_IFF_DADPROGRESS;
2190 ia->ia6_flags |= IN6_IFF_DUPLICATED;
2191 in6_event_enqueue_nwk_wq_entry(IN6_ADDR_MARKED_DUPLICATED,
2192 ia->ia_ifa.ifa_ifp, &ia->ia_addr.sin6_addr,
2193 0);
2194 IFA_UNLOCK(&ia->ia_ifa);
2195
2196 /* increment DAD collision counter */
2197 ++ip6stat.ip6s_dad_collide;
2198
2199 /* We are done with DAD, with duplicated address found. (failure) */
2200 untimeout((void (*)(void *))nd6_dad_timer, (void *)ifa);
2201
2202 IFA_LOCK(&ia->ia_ifa);
2203 log(LOG_ERR, "%s: DAD complete for %s - duplicate found.\n",
2204 if_name(ifp), ip6_sprintf(&ia->ia_addr.sin6_addr));
2205 IFA_UNLOCK(&ia->ia_ifa);
2206
2207 if (candisable) {
2208 struct nd_ifinfo *__single ndi = ND_IFINFO(ifp);
2209 log(LOG_ERR, "%s: possible hardware address duplication "
2210 "detected, disabling IPv6 for interface.\n", if_name(ifp));
2211
2212 VERIFY((NULL != ndi) && (TRUE == ndi->initialized));
2213 ndi->flags |= ND6_IFF_IFDISABLED;
2214 /* Make sure to set IFEF_IPV6_DISABLED too */
2215 nd6_if_disable(ifp, TRUE);
2216 }
2217
2218 log(LOG_ERR,
2219 "%s: manual intervention may be required.\n",
2220 if_name(ifp));
2221
2222 /* Send an event to the configuration agent so that the
2223 * duplicate address will be notified to the user and will
2224 * be removed.
2225 */
2226 in6_post_msg(ifp, KEV_INET6_NEW_USER_ADDR, ia, dp->dad_lladdr, ETHER_ADDR_LEN);
2227 nd6_dad_detach(dp, ifa);
2228 DAD_REMREF(dp); /* drop our reference */
2229 }
2230
2231 static void
nd6_dad_ns_output(struct dadq * dp,struct ifaddr * ifa)2232 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
2233 {
2234 struct in6_ifaddr *__single ia = ifatoia6(ifa);
2235 ifnet_ref_t ifp = ifa->ifa_ifp;
2236 int i = 0;
2237 struct in6_addr taddr6;
2238
2239 DAD_LOCK(dp);
2240 dp->dad_ns_tcount++;
2241 if ((ifp->if_flags & IFF_UP) == 0) {
2242 DAD_UNLOCK(dp);
2243 return;
2244 }
2245 if ((ifp->if_flags & IFF_RUNNING) == 0) {
2246 DAD_UNLOCK(dp);
2247 return;
2248 }
2249
2250 dp->dad_ns_ocount++;
2251 DAD_UNLOCK(dp);
2252 IFA_LOCK_SPIN(&ia->ia_ifa);
2253 taddr6 = ia->ia_addr.sin6_addr;
2254 IFA_UNLOCK(&ia->ia_ifa);
2255 if (dad_enhanced != 0 && !(ifp->if_flags & IFF_POINTOPOINT)) {
2256 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) {
2257 dp->dad_nonce[i] = RandomULong();
2258 }
2259
2260 /*
2261 * XXXHRS: Note that in the case that
2262 * DupAddrDetectTransmits > 1, multiple NS messages with
2263 * different nonces can be looped back in an unexpected
2264 * order. The current implementation recognizes only
2265 * the latest nonce on the sender side. Practically it
2266 * should work well in almost all cases.
2267 */
2268 }
2269 nd6_ns_output(ifp, NULL, &taddr6, NULL,
2270 (uint8_t *)&dp->dad_nonce[0], ND_OPT_NONCE_LEN);
2271 }
2272
2273 /*
2274 * @brief Called to process DAD NS
2275 *
2276 * @param ifa is the pointer to the interface's address
2277 * @param lladdr is source link layer information
2278 * @param lladdrlen is source's linklayer length
2279 *
2280 * @return void
2281 */
2282 static void
nd6_dad_ns_input(struct ifaddr * ifa,char * lladdr __sized_by (lladdrlen),int lladdrlen,struct nd_opt_nonce * ndopt_nonce)2283 nd6_dad_ns_input(struct ifaddr *ifa, char *lladdr __sized_by(lladdrlen),
2284 int lladdrlen, struct nd_opt_nonce *ndopt_nonce)
2285 {
2286 struct dadq *__single dp;
2287 VERIFY(ifa != NULL);
2288
2289 /* Ignore Nonce option when Enhanced DAD is disabled. */
2290 if (dad_enhanced == 0) {
2291 ndopt_nonce = NULL;
2292 }
2293
2294 dp = nd6_dad_find(ifa, ndopt_nonce);
2295 if (dp == NULL) {
2296 return;
2297 }
2298
2299 DAD_LOCK(dp);
2300 ++dp->dad_ns_icount;
2301 if (lladdr && lladdrlen >= ETHER_ADDR_LEN) {
2302 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2303 /* fine to truncate as it is compared against sdl_alen */
2304 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2305 }
2306 DAD_UNLOCK(dp);
2307 DAD_REMREF(dp);
2308 }
2309
2310 /*
2311 * @brief Called to process received NA for DAD
2312 *
2313 * @param m is the pointer to the packet's mbuf
2314 * @param ifp is the pointer to the interface on which packet
2315 * was receicved.
2316 * @param taddr is pointer to target's IPv6 address
2317 * @param lladdr is target's link layer information
2318 * @param lladdrlen is target's linklayer length
2319 *
2320 * @return NULL if the packet is consumed by DAD processing, else
2321 * pointer to the mbuf.
2322 */
2323 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)2324 nd6_dad_na_input(struct mbuf *m, struct ifnet *ifp, struct in6_addr *taddr,
2325 caddr_t lladdr __counted_by(lladdrlen), int lladdrlen)
2326 {
2327 struct ifaddr *__single ifa = NULL;
2328 struct in6_ifaddr *__single ia = NULL;
2329 struct dadq *__single dp = NULL;
2330 struct nd_ifinfo *__single ndi = NULL;
2331 boolean_t replicated;
2332
2333 ifa = (struct ifaddr *) in6ifa_ifpwithaddr(ifp, taddr);
2334 if (ifa == NULL) {
2335 return m;
2336 }
2337
2338 replicated = FALSE;
2339
2340 /* Get the ND6_IFF_REPLICATED flag. */
2341 ndi = ND_IFINFO(ifp);
2342 if (ndi != NULL && ndi->initialized) {
2343 lck_mtx_lock(&ndi->lock);
2344 replicated = !!(ndi->flags & ND6_IFF_REPLICATED);
2345 lck_mtx_unlock(&ndi->lock);
2346 }
2347
2348 if (replicated) {
2349 nd6log(info, "%s: ignoring duplicate NA on "
2350 "replicated interface %s\n", __func__, if_name(ifp));
2351 goto done;
2352 }
2353
2354 /* Lock the interface address until done (see label below). */
2355 IFA_LOCK(ifa);
2356 ia = ifatoia6(ifa);
2357
2358 if (!(ia->ia6_flags & IN6_IFF_DADPROGRESS)) {
2359 IFA_UNLOCK(ifa);
2360 nd6log(info, "%s: ignoring duplicate NA on "
2361 "%s [DAD not in progress]\n", __func__,
2362 if_name(ifp));
2363 goto done;
2364 }
2365
2366 /* Some sleep proxies improperly send the client's Ethernet address in
2367 * the target link-layer address option, so detect this by comparing
2368 * the L2-header source address, if we have seen it, with the target
2369 * address, and ignoring the NA if they don't match.
2370 */
2371 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2372 struct ip6aux *__single ip6a = ip6_findaux(m);
2373 if (ip6a && (ip6a->ip6a_flags & IP6A_HASEEN) != 0 &&
2374 bcmp(ip6a->ip6a_ehsrc, lladdr, ETHER_ADDR_LEN) != 0) {
2375 IFA_UNLOCK(ifa);
2376 nd6log(error, "%s: ignoring duplicate NA on %s "
2377 "[eh_src != tgtlladdr]\n", __func__, if_name(ifp));
2378 goto done;
2379 }
2380 }
2381
2382 IFA_UNLOCK(ifa);
2383
2384 dp = nd6_dad_find(ifa, NULL);
2385 if (dp == NULL) {
2386 nd6log(info, "%s: no DAD structure for %s on %s.\n",
2387 __func__, ip6_sprintf(taddr), if_name(ifp));
2388 goto done;
2389 }
2390
2391 DAD_LOCK_SPIN(dp);
2392 if (lladdr != NULL && lladdrlen >= ETHER_ADDR_LEN) {
2393 memcpy(dp->dad_lladdr, lladdr, ETHER_ADDR_LEN);
2394 dp->dad_lladdrlen = (uint8_t)lladdrlen;
2395 }
2396 dp->dad_na_icount++;
2397 DAD_UNLOCK(dp);
2398 DAD_REMREF(dp);
2399
2400 /* remove the address. */
2401 nd6log(info,
2402 "%s: duplicate IPv6 address %s [processing NA on %s]\n", __func__,
2403 ip6_sprintf(taddr), if_name(ifp));
2404 done:
2405 IFA_LOCK_ASSERT_NOTHELD(ifa);
2406 ifa_remref(ifa);
2407 m_freem(m);
2408 return NULL;
2409 }
2410
2411 static void
dad_addref(struct dadq * dp,int locked)2412 dad_addref(struct dadq *dp, int locked)
2413 {
2414 if (!locked) {
2415 DAD_LOCK_SPIN(dp);
2416 } else {
2417 DAD_LOCK_ASSERT_HELD(dp);
2418 }
2419
2420 if (++dp->dad_refcount == 0) {
2421 panic("%s: dad %p wraparound refcnt", __func__, dp);
2422 /* NOTREACHED */
2423 }
2424 if (!locked) {
2425 DAD_UNLOCK(dp);
2426 }
2427 }
2428
2429 static void
dad_remref(struct dadq * dp)2430 dad_remref(struct dadq *dp)
2431 {
2432 struct ifaddr *__single ifa;
2433
2434 DAD_LOCK_SPIN(dp);
2435 if (dp->dad_refcount == 0) {
2436 panic("%s: dad %p negative refcnt", __func__, dp);
2437 }
2438 --dp->dad_refcount;
2439 if (dp->dad_refcount > 0) {
2440 DAD_UNLOCK(dp);
2441 return;
2442 }
2443 DAD_UNLOCK(dp);
2444
2445 if (dp->dad_attached ||
2446 dp->dad_list.tqe_next != NULL || dp->dad_list.tqe_prev != NULL) {
2447 panic("%s: attached dad=%p is being freed", __func__, dp);
2448 /* NOTREACHED */
2449 }
2450
2451 if ((ifa = dp->dad_ifa) != NULL) {
2452 ifa_remref(ifa); /* drop dad_ifa reference */
2453 dp->dad_ifa = NULL;
2454 }
2455
2456 lck_mtx_destroy(&dp->dad_lock, &ifa_mtx_grp);
2457 zfree(dad_zone, dp);
2458 }
2459
2460 void
nd6_llreach_set_reachable(struct ifnet * ifp,void * __sized_by (alen)addr,unsigned int alen)2461 nd6_llreach_set_reachable(struct ifnet *ifp, void *__sized_by(alen) addr, unsigned int alen)
2462 {
2463 /* Nothing more to do if it's disabled */
2464 if (nd6_llreach_base == 0) {
2465 return;
2466 }
2467
2468 ifnet_llreach_set_reachable(ifp, ETHERTYPE_IPV6, addr, alen);
2469 }
2470
2471 void
nd6_alt_node_addr_decompose(struct ifnet * ifp,struct sockaddr * sa,struct sockaddr_dl * isdl,struct sockaddr_in6 * sin6)2472 nd6_alt_node_addr_decompose(struct ifnet *ifp, struct sockaddr *sa,
2473 struct sockaddr_dl* isdl, struct sockaddr_in6 *sin6)
2474 {
2475 static const size_t EUI64_LENGTH = 8;
2476 struct sockaddr_dl *sdl = SDL(isdl);
2477
2478 VERIFY(nd6_need_cache(ifp));
2479 VERIFY(sa);
2480 VERIFY(sdl && (void *)sa != (void *)sdl);
2481 VERIFY(sin6 && (void *)sa != (void *)sin6);
2482
2483 SOCKADDR_ZERO(sin6, sizeof(*sin6));
2484 sin6->sin6_len = sizeof *sin6;
2485 sin6->sin6_family = AF_INET6;
2486
2487 SOCKADDR_ZERO(sdl, sizeof(*sdl));
2488 sdl->sdl_len = sizeof *sdl;
2489 sdl->sdl_family = AF_LINK;
2490 sdl->sdl_type = ifp->if_type;
2491 sdl->sdl_index = ifp->if_index;
2492 sdl->sdl_nlen = 0;
2493
2494 switch (sa->sa_family) {
2495 case AF_INET6: {
2496 struct sockaddr_in6 *__single sin6a = SIN6(sa);
2497 struct in6_addr *__single in6 = &sin6a->sin6_addr;
2498
2499 VERIFY(sa->sa_len == sizeof *sin6);
2500 if (in6->s6_addr[11] == 0xff && in6->s6_addr[12] == 0xfe) {
2501 sdl->sdl_alen = ETHER_ADDR_LEN;
2502 LLADDR(sdl)[0] = (in6->s6_addr[8] ^ ND6_EUI64_UBIT);
2503 LLADDR(sdl)[1] = in6->s6_addr[9];
2504 LLADDR(sdl)[2] = in6->s6_addr[10];
2505 LLADDR(sdl)[3] = in6->s6_addr[13];
2506 LLADDR(sdl)[4] = in6->s6_addr[14];
2507 LLADDR(sdl)[5] = in6->s6_addr[15];
2508 } else {
2509 sdl->sdl_alen = EUI64_LENGTH;
2510 bcopy(&in6->s6_addr[8], LLADDR(sdl), EUI64_LENGTH);
2511 }
2512
2513 sdl->sdl_slen = 0;
2514 break;
2515 }
2516 case AF_LINK: {
2517 struct sockaddr_dl *sdla = SDL(sa);
2518 struct in6_addr *__single in6 = &sin6->sin6_addr;
2519 caddr_t lla = LLADDR(sdla);
2520 VERIFY(sa->sa_len <= sizeof(*sdl));
2521 SOCKADDR_COPY(sa, sdl, sa->sa_len);
2522
2523 sin6->sin6_scope_id = sdla->sdl_index;
2524 if (sin6->sin6_scope_id == 0) {
2525 sin6->sin6_scope_id = ifp->if_index;
2526 }
2527 in6->s6_addr[0] = 0xfe;
2528 in6->s6_addr[1] = 0x80;
2529 if (sdla->sdl_alen == EUI64_LENGTH) {
2530 bcopy(lla, &in6->s6_addr[8], EUI64_LENGTH);
2531 } else {
2532 VERIFY(sdla->sdl_alen == ETHER_ADDR_LEN);
2533
2534 in6->s6_addr[8] = ((uint8_t) lla[0] ^ ND6_EUI64_UBIT);
2535 in6->s6_addr[9] = (uint8_t) lla[1];
2536 in6->s6_addr[10] = (uint8_t) lla[2];
2537 in6->s6_addr[11] = 0xff;
2538 in6->s6_addr[12] = 0xfe;
2539 in6->s6_addr[13] = (uint8_t) lla[3];
2540 in6->s6_addr[14] = (uint8_t) lla[4];
2541 in6->s6_addr[15] = (uint8_t) lla[5];
2542 }
2543
2544 break;
2545 }
2546 default:
2547 VERIFY(false);
2548 break;
2549 }
2550 }
2551
2552 int
nd6_alt_node_present(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * isdl,int32_t rssi,int lqm,int npm)2553 nd6_alt_node_present(struct ifnet *ifp, struct sockaddr_in6 *sin6,
2554 struct sockaddr_dl *isdl, int32_t rssi, int lqm, int npm)
2555 {
2556 struct sockaddr_dl *sdl = SDL(isdl);
2557 rtentry_ref_t rt = NULL;
2558 struct llinfo_nd6 *__single ln = NULL;
2559 struct if_llreach *__single lr = NULL;
2560 int nd6_nc_updated = 0;
2561 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2562 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2563
2564 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2565 if (in6_embedded_scope) {
2566 if (temp_embedded_id == 0) {
2567 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2568 }
2569 } else if (temp_ifscope_id == 0) {
2570 sin6->sin6_scope_id = ifp->if_index;
2571 }
2572 }
2573
2574 nd6_cache_lladdr(ifp, &sin6->sin6_addr, LLADDR(sdl), sdl->sdl_alen,
2575 ND_NEIGHBOR_ADVERT, 0, &nd6_nc_updated);
2576
2577 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2578 lck_mtx_lock(rnh_lock);
2579
2580 rt = rtalloc1_scoped_locked(SA(sin6), 1, 0, ifp->if_index);
2581
2582 /* Restore the address that was passed to us */
2583 if (in6_embedded_scope) {
2584 if (temp_embedded_id == 0) {
2585 sin6->sin6_addr.s6_addr16[1] = 0;
2586 }
2587 } else if (temp_ifscope_id == 0) {
2588 sin6->sin6_scope_id = 0;
2589 }
2590
2591 if (rt != NULL) {
2592 RT_LOCK(rt);
2593 VERIFY(rt->rt_flags & RTF_LLINFO);
2594 VERIFY(rt->rt_llinfo);
2595
2596 ln = rt->rt_llinfo;
2597 ND6_CACHE_STATE_TRANSITION(ln, ND6_LLINFO_REACHABLE);
2598 ln_setexpire(ln, 0);
2599
2600 lr = ln->ln_llreach;
2601 if (lr) {
2602 IFLR_LOCK(lr);
2603 lr->lr_rssi = rssi;
2604 lr->lr_lqm = (int32_t) lqm;
2605 lr->lr_npm = (int32_t) npm;
2606 IFLR_UNLOCK(lr);
2607 }
2608
2609 RT_UNLOCK(rt);
2610 RT_REMREF(rt);
2611 }
2612
2613 lck_mtx_unlock(rnh_lock);
2614
2615 if (rt == NULL) {
2616 log(LOG_ERR, "%s: failed to add/update host route to %s.\n",
2617 __func__, ip6_sprintf(&sin6->sin6_addr));
2618 #if DEBUG || DEVELOPMENT
2619 if (ip6_p2p_debug) {
2620 panic("%s: failed to add/update host route to %s.\n",
2621 __func__, ip6_sprintf(&sin6->sin6_addr));
2622 }
2623 #endif
2624 return EHOSTUNREACH;
2625 }
2626
2627 nd6log(info, "%s: Successfully added/updated host route to %s [lr=0x%llx]\n",
2628 __func__, ip6_sprintf(&sin6->sin6_addr),
2629 (uint64_t)VM_KERNEL_ADDRPERM(lr));
2630 /*
2631 * nd6_nc_updated not set implies that nothing was updated
2632 * in the neighbor cache. Convey that as EEXIST to callers.
2633 */
2634 if (nd6_nc_updated == 0) {
2635 return EEXIST;
2636 }
2637 return 0;
2638 }
2639
2640 int
nd6_alt_node_absent(struct ifnet * ifp,struct sockaddr_in6 * sin6,struct sockaddr_dl * sdl)2641 nd6_alt_node_absent(struct ifnet *ifp, struct sockaddr_in6 *sin6, struct sockaddr_dl *sdl)
2642 {
2643 rtentry_ref_t rt = NULL;
2644 int error = 0;
2645 const uint32_t temp_embedded_id = sin6->sin6_addr.s6_addr16[1];
2646 const uint32_t temp_ifscope_id = sin6->sin6_scope_id;
2647
2648 nd6log2(debug, "%s: host route to %s\n", __func__,
2649 ip6_sprintf(&sin6->sin6_addr));
2650
2651 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
2652 if (in6_embedded_scope) {
2653 if (temp_embedded_id == 0) {
2654 sin6->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
2655 }
2656 } else if (temp_ifscope_id == 0) {
2657 sin6->sin6_scope_id = ifp->if_index;
2658 }
2659 }
2660
2661 LCK_MTX_ASSERT(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
2662 lck_mtx_lock(rnh_lock);
2663
2664 rt = rtalloc1_scoped_locked(SA(sin6), 0, 0, ifp->if_index);
2665
2666 /* Restore the address that was passed to us */
2667 if (in6_embedded_scope) {
2668 if (temp_embedded_id == 0) {
2669 sin6->sin6_addr.s6_addr16[1] = 0;
2670 }
2671 } else if (temp_ifscope_id == 0) {
2672 sin6->sin6_scope_id = 0;
2673 }
2674
2675 if (rt != NULL) {
2676 RT_LOCK(rt);
2677 if (IS_DYNAMIC_DIRECT_HOSTROUTE(rt)) {
2678 /*
2679 * Copy the link layer information in SDL when present
2680 * as it later gets used to issue the kernel event for
2681 * node absence.
2682 */
2683 if (sdl != NULL && rt->rt_gateway != NULL &&
2684 rt->rt_gateway->sa_family == AF_LINK &&
2685 SDL(rt->rt_gateway)->sdl_len <= sizeof(*sdl)) {
2686 SOCKADDR_COPY(rt->rt_gateway, sdl, SDL(rt->rt_gateway)->sdl_len);
2687 }
2688
2689 rt->rt_flags |= RTF_CONDEMNED;
2690 RT_UNLOCK(rt);
2691
2692 error = rtrequest_locked(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt),
2693 0, (struct rtentry **)NULL);
2694
2695 rtfree_locked(rt);
2696 } else {
2697 error = EHOSTUNREACH;
2698 RT_REMREF_LOCKED(rt);
2699 RT_UNLOCK(rt);
2700 }
2701 } else {
2702 error = EHOSTUNREACH;
2703 }
2704
2705 if (error == 0) {
2706 nd6log(info, "%s: Successfully deleted host route to %s "
2707 "for interface %s.\n", __func__, ip6_sprintf(&sin6->sin6_addr),
2708 ifp->if_xname);
2709 } else {
2710 nd6log(error, "%s: Failed to delete host route to %s "
2711 "for interface %s with error :%d.\n", __func__,
2712 ip6_sprintf(&sin6->sin6_addr),
2713 ifp->if_xname, error);
2714 }
2715
2716 lck_mtx_unlock(rnh_lock);
2717 return error;
2718 }
2719