xref: /freebsd/sys/netinet6/nd6_nbr.c (revision e9b148a3185f41e3a09e91ea75cae7828d908845)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 #include "opt_mpath.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
45 #include <sys/malloc.h>
46 #include <sys/libkern.h>
47 #include <sys/lock.h>
48 #include <sys/rwlock.h>
49 #include <sys/mbuf.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/errno.h>
55 #include <sys/sysctl.h>
56 #include <sys/syslog.h>
57 #include <sys/queue.h>
58 #include <sys/callout.h>
59 #include <sys/refcount.h>
60 
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_dl.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #ifdef RADIX_MPATH
67 #include <net/radix_mpath.h>
68 #endif
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <net/if_llatbl.h>
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #include <netinet6/nd6.h>
80 #include <netinet/icmp6.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet6/send.h>
83 
84 #define SDL(s) ((struct sockaddr_dl *)s)
85 
86 struct dadq;
87 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
88 static void nd6_dad_add(struct dadq *dp);
89 static void nd6_dad_del(struct dadq *dp);
90 static void nd6_dad_rele(struct dadq *);
91 static void nd6_dad_starttimer(struct dadq *, int, int);
92 static void nd6_dad_stoptimer(struct dadq *);
93 static void nd6_dad_timer(struct dadq *);
94 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
95 static void nd6_dad_ns_output(struct dadq *);
96 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
97 static void nd6_dad_na_input(struct ifaddr *);
98 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
99     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
100 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
101     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
102 
103 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
104 #define	V_dad_enhanced			VNET(dad_enhanced)
105 
106 SYSCTL_DECL(_net_inet6_ip6);
107 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
108     &VNET_NAME(dad_enhanced), 0,
109     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
110 
111 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
112 						   transmit DAD packet */
113 #define	V_dad_maxtry			VNET(dad_maxtry)
114 
115 /*
116  * Input a Neighbor Solicitation Message.
117  *
118  * Based on RFC 2461
119  * Based on RFC 2462 (duplicate address detection)
120  */
121 void
122 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
123 {
124 	struct ifnet *ifp = m->m_pkthdr.rcvif;
125 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
126 	struct nd_neighbor_solicit *nd_ns;
127 	struct in6_addr saddr6 = ip6->ip6_src;
128 	struct in6_addr daddr6 = ip6->ip6_dst;
129 	struct in6_addr taddr6;
130 	struct in6_addr myaddr6;
131 	char *lladdr = NULL;
132 	struct ifaddr *ifa = NULL;
133 	int lladdrlen = 0;
134 	int anycast = 0, proxy = 0, tentative = 0;
135 	int tlladdr;
136 	int rflag;
137 	union nd_opts ndopts;
138 	struct sockaddr_dl proxydl;
139 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
140 
141 	/* RFC 6980: Nodes MUST silently ignore fragments */
142 	if(m->m_flags & M_FRAGMENTED)
143 		goto freeit;
144 
145 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
146 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
147 		rflag = 0;
148 #ifndef PULLDOWN_TEST
149 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
150 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
151 #else
152 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
153 	if (nd_ns == NULL) {
154 		ICMP6STAT_INC(icp6s_tooshort);
155 		return;
156 	}
157 #endif
158 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
159 	taddr6 = nd_ns->nd_ns_target;
160 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
161 		goto bad;
162 
163 	if (ip6->ip6_hlim != 255) {
164 		nd6log((LOG_ERR,
165 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
166 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
167 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
168 		goto bad;
169 	}
170 
171 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
172 		/* dst has to be a solicited node multicast address. */
173 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
174 		    /* don't check ifindex portion */
175 		    daddr6.s6_addr32[1] == 0 &&
176 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
177 		    daddr6.s6_addr8[12] == 0xff) {
178 			; /* good */
179 		} else {
180 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
181 			    "(wrong ip6 dst)\n"));
182 			goto bad;
183 		}
184 	} else if (!V_nd6_onlink_ns_rfc4861) {
185 		struct sockaddr_in6 src_sa6;
186 
187 		/*
188 		 * According to recent IETF discussions, it is not a good idea
189 		 * to accept a NS from an address which would not be deemed
190 		 * to be a neighbor otherwise.  This point is expected to be
191 		 * clarified in future revisions of the specification.
192 		 */
193 		bzero(&src_sa6, sizeof(src_sa6));
194 		src_sa6.sin6_family = AF_INET6;
195 		src_sa6.sin6_len = sizeof(src_sa6);
196 		src_sa6.sin6_addr = saddr6;
197 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
198 			nd6log((LOG_INFO, "nd6_ns_input: "
199 				"NS packet from non-neighbor\n"));
200 			goto bad;
201 		}
202 	}
203 
204 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
205 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
206 		goto bad;
207 	}
208 
209 	icmp6len -= sizeof(*nd_ns);
210 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
211 	if (nd6_options(&ndopts) < 0) {
212 		nd6log((LOG_INFO,
213 		    "nd6_ns_input: invalid ND option, ignored\n"));
214 		/* nd6_options have incremented stats */
215 		goto freeit;
216 	}
217 
218 	if (ndopts.nd_opts_src_lladdr) {
219 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
220 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
221 	}
222 
223 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
224 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
225 		    "(link-layer address option)\n"));
226 		goto bad;
227 	}
228 
229 	/*
230 	 * Attaching target link-layer address to the NA?
231 	 * (RFC 2461 7.2.4)
232 	 *
233 	 * NS IP dst is unicast/anycast			MUST NOT add
234 	 * NS IP dst is solicited-node multicast	MUST add
235 	 *
236 	 * In implementation, we add target link-layer address by default.
237 	 * We do not add one in MUST NOT cases.
238 	 */
239 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
240 		tlladdr = 0;
241 	else
242 		tlladdr = 1;
243 
244 	/*
245 	 * Target address (taddr6) must be either:
246 	 * (1) Valid unicast/anycast address for my receiving interface,
247 	 * (2) Unicast address for which I'm offering proxy service, or
248 	 * (3) "tentative" address on which DAD is being performed.
249 	 */
250 	/* (1) and (3) check. */
251 	if (ifp->if_carp)
252 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
253 	else
254 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
255 
256 	/* (2) check. */
257 	if (ifa == NULL) {
258 		struct sockaddr_dl rt_gateway;
259 		struct rt_addrinfo info;
260 		struct sockaddr_in6 dst6;
261 
262 		bzero(&dst6, sizeof(dst6));
263 		dst6.sin6_len = sizeof(struct sockaddr_in6);
264 		dst6.sin6_family = AF_INET6;
265 		dst6.sin6_addr = taddr6;
266 
267 		bzero(&rt_gateway, sizeof(rt_gateway));
268 		rt_gateway.sdl_len = sizeof(rt_gateway);
269 		bzero(&info, sizeof(info));
270 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
271 
272 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
273 		    0, 0, &info) == 0) {
274 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
275 			    rt_gateway.sdl_family == AF_LINK) {
276 
277 				/*
278 				 * proxy NDP for single entry
279 				 */
280 				proxydl = *SDL(&rt_gateway);
281 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
282 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
283 				if (ifa)
284 					proxy = 1;
285 			}
286 		}
287 	}
288 	if (ifa == NULL) {
289 		/*
290 		 * We've got an NS packet, and we don't have that adddress
291 		 * assigned for us.  We MUST silently ignore it.
292 		 * See RFC2461 7.2.3.
293 		 */
294 		goto freeit;
295 	}
296 	myaddr6 = *IFA_IN6(ifa);
297 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
298 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
299 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
300 		goto freeit;
301 
302 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
303 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
304 		    "(if %d, NS packet %d)\n",
305 		    ip6_sprintf(ip6bufs, &taddr6),
306 		    ifp->if_addrlen, lladdrlen - 2));
307 		goto bad;
308 	}
309 
310 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
311 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
312 		    ip6_sprintf(ip6bufs, &saddr6)));
313 		goto freeit;
314 	}
315 
316 	/*
317 	 * We have neighbor solicitation packet, with target address equals to
318 	 * one of my tentative address.
319 	 *
320 	 * src addr	how to process?
321 	 * ---		---
322 	 * multicast	of course, invalid (rejected in ip6_input)
323 	 * unicast	somebody is doing address resolution -> ignore
324 	 * unspec	dup address detection
325 	 *
326 	 * The processing is defined in RFC 2462.
327 	 */
328 	if (tentative) {
329 		/*
330 		 * If source address is unspecified address, it is for
331 		 * duplicate address detection.
332 		 *
333 		 * If not, the packet is for addess resolution;
334 		 * silently ignore it.
335 		 */
336 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
337 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
338 
339 		goto freeit;
340 	}
341 
342 	/*
343 	 * If the source address is unspecified address, entries must not
344 	 * be created or updated.
345 	 * It looks that sender is performing DAD.  Output NA toward
346 	 * all-node multicast address, to tell the sender that I'm using
347 	 * the address.
348 	 * S bit ("solicited") must be zero.
349 	 */
350 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
351 		struct in6_addr in6_all;
352 
353 		in6_all = in6addr_linklocal_allnodes;
354 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
355 			goto bad;
356 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
357 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
358 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
359 		    M_GETFIB(m));
360 		goto freeit;
361 	}
362 
363 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
364 	    ND_NEIGHBOR_SOLICIT, 0);
365 
366 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
367 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
368 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
369 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
370  freeit:
371 	if (ifa != NULL)
372 		ifa_free(ifa);
373 	m_freem(m);
374 	return;
375 
376  bad:
377 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
378 		ip6_sprintf(ip6bufs, &saddr6)));
379 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
380 		ip6_sprintf(ip6bufs, &daddr6)));
381 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
382 		ip6_sprintf(ip6bufs, &taddr6)));
383 	ICMP6STAT_INC(icp6s_badns);
384 	if (ifa != NULL)
385 		ifa_free(ifa);
386 	m_freem(m);
387 }
388 
389 /*
390  * Output a Neighbor Solicitation Message. Caller specifies:
391  *	- ICMP6 header source IP6 address
392  *	- ND6 header target IP6 address
393  *	- ND6 header source datalink address
394  *
395  * Based on RFC 2461
396  * Based on RFC 2462 (duplicate address detection)
397  *
398  *    ln - for source address determination
399  * nonce - If non-NULL, NS is used for duplicate address detection and
400  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
401  */
402 static void
403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
404     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
405     uint8_t *nonce, u_int fibnum)
406 {
407 	struct mbuf *m;
408 	struct m_tag *mtag;
409 	struct ip6_hdr *ip6;
410 	struct nd_neighbor_solicit *nd_ns;
411 	struct ip6_moptions im6o;
412 	int icmp6len;
413 	int maxlen;
414 	caddr_t mac;
415 
416 	if (IN6_IS_ADDR_MULTICAST(taddr6))
417 		return;
418 
419 	/* estimate the size of message */
420 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
421 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
422 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
423 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
424 	    __func__, max_linkhdr, maxlen, MCLBYTES));
425 
426 	if (max_linkhdr + maxlen > MHLEN)
427 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
428 	else
429 		m = m_gethdr(M_NOWAIT, MT_DATA);
430 	if (m == NULL)
431 		return;
432 	M_SETFIB(m, fibnum);
433 
434 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
435 		m->m_flags |= M_MCAST;
436 		im6o.im6o_multicast_ifp = ifp;
437 		im6o.im6o_multicast_hlim = 255;
438 		im6o.im6o_multicast_loop = 0;
439 	}
440 
441 	icmp6len = sizeof(*nd_ns);
442 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
443 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
444 
445 	/* fill neighbor solicitation packet */
446 	ip6 = mtod(m, struct ip6_hdr *);
447 	ip6->ip6_flow = 0;
448 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
449 	ip6->ip6_vfc |= IPV6_VERSION;
450 	/* ip6->ip6_plen will be set later */
451 	ip6->ip6_nxt = IPPROTO_ICMPV6;
452 	ip6->ip6_hlim = 255;
453 	if (daddr6)
454 		ip6->ip6_dst = *daddr6;
455 	else {
456 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
457 		ip6->ip6_dst.s6_addr16[1] = 0;
458 		ip6->ip6_dst.s6_addr32[1] = 0;
459 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
460 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
461 		ip6->ip6_dst.s6_addr8[12] = 0xff;
462 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
463 			goto bad;
464 	}
465 	if (nonce == NULL) {
466 		struct ifaddr *ifa = NULL;
467 
468 		/*
469 		 * RFC2461 7.2.2:
470 		 * "If the source address of the packet prompting the
471 		 * solicitation is the same as one of the addresses assigned
472 		 * to the outgoing interface, that address SHOULD be placed
473 		 * in the IP Source Address of the outgoing solicitation.
474 		 * Otherwise, any one of the addresses assigned to the
475 		 * interface should be used."
476 		 *
477 		 * We use the source address for the prompting packet
478 		 * (saddr6), if saddr6 belongs to the outgoing interface.
479 		 * Otherwise, we perform the source address selection as usual.
480 		 */
481 
482 		if (saddr6 != NULL)
483 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
484 		if (ifa != NULL) {
485 			/* ip6_src set already. */
486 			ip6->ip6_src = *saddr6;
487 			ifa_free(ifa);
488 		} else {
489 			int error;
490 			struct in6_addr dst6, src6;
491 			uint32_t scopeid;
492 
493 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
494 			error = in6_selectsrc_addr(fibnum, &dst6,
495 			    scopeid, ifp, &src6, NULL);
496 			if (error) {
497 				char ip6buf[INET6_ADDRSTRLEN];
498 				nd6log((LOG_DEBUG, "%s: source can't be "
499 				    "determined: dst=%s, error=%d\n", __func__,
500 				    ip6_sprintf(ip6buf, &dst6),
501 				    error));
502 				goto bad;
503 			}
504 			ip6->ip6_src = src6;
505 		}
506 	} else {
507 		/*
508 		 * Source address for DAD packet must always be IPv6
509 		 * unspecified address. (0::0)
510 		 * We actually don't have to 0-clear the address (we did it
511 		 * above), but we do so here explicitly to make the intention
512 		 * clearer.
513 		 */
514 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
515 	}
516 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
517 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
518 	nd_ns->nd_ns_code = 0;
519 	nd_ns->nd_ns_reserved = 0;
520 	nd_ns->nd_ns_target = *taddr6;
521 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
522 
523 	/*
524 	 * Add source link-layer address option.
525 	 *
526 	 *				spec		implementation
527 	 *				---		---
528 	 * DAD packet			MUST NOT	do not add the option
529 	 * there's no link layer address:
530 	 *				impossible	do not add the option
531 	 * there's link layer address:
532 	 *	Multicast NS		MUST add one	add the option
533 	 *	Unicast NS		SHOULD add one	add the option
534 	 */
535 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
536 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
537 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
538 		/* 8 byte alignments... */
539 		optlen = (optlen + 7) & ~7;
540 
541 		m->m_pkthdr.len += optlen;
542 		m->m_len += optlen;
543 		icmp6len += optlen;
544 		bzero((caddr_t)nd_opt, optlen);
545 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
546 		nd_opt->nd_opt_len = optlen >> 3;
547 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
548 	}
549 	/*
550 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
551 	 * This behavior is documented as Enhanced Duplicate Address
552 	 * Detection in RFC 7527.
553 	 * net.inet6.ip6.dad_enhanced=0 disables this.
554 	 */
555 	if (V_dad_enhanced != 0 && nonce != NULL) {
556 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
557 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
558 		/* 8-byte alignment is required. */
559 		optlen = (optlen + 7) & ~7;
560 
561 		m->m_pkthdr.len += optlen;
562 		m->m_len += optlen;
563 		icmp6len += optlen;
564 		bzero((caddr_t)nd_opt, optlen);
565 		nd_opt->nd_opt_type = ND_OPT_NONCE;
566 		nd_opt->nd_opt_len = optlen >> 3;
567 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
568 	}
569 	ip6->ip6_plen = htons((u_short)icmp6len);
570 	nd_ns->nd_ns_cksum = 0;
571 	nd_ns->nd_ns_cksum =
572 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
573 
574 	if (send_sendso_input_hook != NULL) {
575 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
576 			sizeof(unsigned short), M_NOWAIT);
577 		if (mtag == NULL)
578 			goto bad;
579 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
580 		m_tag_prepend(m, mtag);
581 	}
582 
583 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
584 	    &im6o, NULL, NULL);
585 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
586 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
587 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
588 
589 	return;
590 
591   bad:
592 	m_freem(m);
593 }
594 
595 #ifndef BURN_BRIDGES
596 void
597 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
598     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
599 {
600 
601 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
602 }
603 #endif
604 /*
605  * Neighbor advertisement input handling.
606  *
607  * Based on RFC 2461
608  * Based on RFC 2462 (duplicate address detection)
609  *
610  * the following items are not implemented yet:
611  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
612  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
613  */
614 void
615 nd6_na_input(struct mbuf *m, int off, int icmp6len)
616 {
617 	struct ifnet *ifp = m->m_pkthdr.rcvif;
618 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
619 	struct nd_neighbor_advert *nd_na;
620 	struct in6_addr daddr6 = ip6->ip6_dst;
621 	struct in6_addr taddr6;
622 	int flags;
623 	int is_router;
624 	int is_solicited;
625 	int is_override;
626 	char *lladdr = NULL;
627 	int lladdrlen = 0;
628 	int checklink = 0;
629 	struct ifaddr *ifa;
630 	struct llentry *ln = NULL;
631 	union nd_opts ndopts;
632 	struct mbuf *chain = NULL;
633 	struct sockaddr_in6 sin6;
634 	u_char linkhdr[LLE_MAX_LINKHDR];
635 	size_t linkhdrsize;
636 	int lladdr_off;
637 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
638 
639 	NET_EPOCH_ASSERT();
640 
641 	/* RFC 6980: Nodes MUST silently ignore fragments */
642 	if(m->m_flags & M_FRAGMENTED)
643 		goto freeit;
644 
645 	if (ip6->ip6_hlim != 255) {
646 		nd6log((LOG_ERR,
647 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
648 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
649 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
650 		goto bad;
651 	}
652 
653 #ifndef PULLDOWN_TEST
654 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
655 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
656 #else
657 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
658 	if (nd_na == NULL) {
659 		ICMP6STAT_INC(icp6s_tooshort);
660 		return;
661 	}
662 #endif
663 
664 	flags = nd_na->nd_na_flags_reserved;
665 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
666 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
667 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
668 	memset(&sin6, 0, sizeof(sin6));
669 
670 	taddr6 = nd_na->nd_na_target;
671 	if (in6_setscope(&taddr6, ifp, NULL))
672 		goto bad;	/* XXX: impossible */
673 
674 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
675 		nd6log((LOG_ERR,
676 		    "nd6_na_input: invalid target address %s\n",
677 		    ip6_sprintf(ip6bufs, &taddr6)));
678 		goto bad;
679 	}
680 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
681 		if (is_solicited) {
682 			nd6log((LOG_ERR,
683 			    "nd6_na_input: a solicited adv is multicasted\n"));
684 			goto bad;
685 		}
686 
687 	icmp6len -= sizeof(*nd_na);
688 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
689 	if (nd6_options(&ndopts) < 0) {
690 		nd6log((LOG_INFO,
691 		    "nd6_na_input: invalid ND option, ignored\n"));
692 		/* nd6_options have incremented stats */
693 		goto freeit;
694 	}
695 
696 	if (ndopts.nd_opts_tgt_lladdr) {
697 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
698 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
699 	}
700 
701 	/*
702 	 * This effectively disables the DAD check on a non-master CARP
703 	 * address.
704 	 */
705 	if (ifp->if_carp)
706 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
707 	else
708 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
709 
710 	/*
711 	 * Target address matches one of my interface address.
712 	 *
713 	 * If my address is tentative, this means that there's somebody
714 	 * already using the same address as mine.  This indicates DAD failure.
715 	 * This is defined in RFC 2462.
716 	 *
717 	 * Otherwise, process as defined in RFC 2461.
718 	 */
719 	if (ifa
720 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
721 		nd6_dad_na_input(ifa);
722 		ifa_free(ifa);
723 		goto freeit;
724 	}
725 
726 	/* Just for safety, maybe unnecessary. */
727 	if (ifa) {
728 		ifa_free(ifa);
729 		log(LOG_ERR,
730 		    "nd6_na_input: duplicate IP6 address %s\n",
731 		    ip6_sprintf(ip6bufs, &taddr6));
732 		goto freeit;
733 	}
734 
735 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
736 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
737 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
738 		    ifp->if_addrlen, lladdrlen - 2));
739 		goto bad;
740 	}
741 
742 	/*
743 	 * If no neighbor cache entry is found, NA SHOULD silently be
744 	 * discarded.
745 	 */
746 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
747 	if (ln == NULL) {
748 		goto freeit;
749 	}
750 
751 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
752 		/*
753 		 * If the link-layer has address, and no lladdr option came,
754 		 * discard the packet.
755 		 */
756 		if (ifp->if_addrlen && lladdr == NULL) {
757 			goto freeit;
758 		}
759 
760 		/*
761 		 * Record link-layer address, and update the state.
762 		 */
763 		linkhdrsize = sizeof(linkhdr);
764 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
765 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
766 			return;
767 
768 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
769 		    lladdr_off) == 0) {
770 			ln = NULL;
771 			goto freeit;
772 		}
773 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
774 		if (is_solicited)
775 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
776 		else
777 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
778 		if ((ln->ln_router = is_router) != 0) {
779 			/*
780 			 * This means a router's state has changed from
781 			 * non-reachable to probably reachable, and might
782 			 * affect the status of associated prefixes..
783 			 */
784 			checklink = 1;
785 		}
786 	} else {
787 		int llchange;
788 
789 		/*
790 		 * Check if the link-layer address has changed or not.
791 		 */
792 		if (lladdr == NULL)
793 			llchange = 0;
794 		else {
795 			if (ln->la_flags & LLE_VALID) {
796 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
797 					llchange = 1;
798 				else
799 					llchange = 0;
800 			} else
801 				llchange = 1;
802 		}
803 
804 		/*
805 		 * This is VERY complex.  Look at it with care.
806 		 *
807 		 * override solicit lladdr llchange	action
808 		 *					(L: record lladdr)
809 		 *
810 		 *	0	0	n	--	(2c)
811 		 *	0	0	y	n	(2b) L
812 		 *	0	0	y	y	(1)    REACHABLE->STALE
813 		 *	0	1	n	--	(2c)   *->REACHABLE
814 		 *	0	1	y	n	(2b) L *->REACHABLE
815 		 *	0	1	y	y	(1)    REACHABLE->STALE
816 		 *	1	0	n	--	(2a)
817 		 *	1	0	y	n	(2a) L
818 		 *	1	0	y	y	(2a) L *->STALE
819 		 *	1	1	n	--	(2a)   *->REACHABLE
820 		 *	1	1	y	n	(2a) L *->REACHABLE
821 		 *	1	1	y	y	(2a) L *->REACHABLE
822 		 */
823 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
824 			/*
825 			 * If state is REACHABLE, make it STALE.
826 			 * no other updates should be done.
827 			 */
828 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
829 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
830 			goto freeit;
831 		} else if (is_override				   /* (2a) */
832 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
833 			|| lladdr == NULL) {			   /* (2c) */
834 			/*
835 			 * Update link-local address, if any.
836 			 */
837 			if (lladdr != NULL) {
838 				linkhdrsize = sizeof(linkhdr);
839 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
840 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
841 					goto freeit;
842 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
843 				    linkhdrsize, lladdr_off) == 0) {
844 					ln = NULL;
845 					goto freeit;
846 				}
847 				EVENTHANDLER_INVOKE(lle_event, ln,
848 				    LLENTRY_RESOLVED);
849 			}
850 
851 			/*
852 			 * If solicited, make the state REACHABLE.
853 			 * If not solicited and the link-layer address was
854 			 * changed, make it STALE.
855 			 */
856 			if (is_solicited)
857 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
858 			else {
859 				if (lladdr != NULL && llchange)
860 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
861 			}
862 		}
863 
864 		if (ln->ln_router && !is_router) {
865 			/*
866 			 * The peer dropped the router flag.
867 			 * Remove the sender from the Default Router List and
868 			 * update the Destination Cache entries.
869 			 */
870 			struct ifnet *nd6_ifp;
871 
872 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
873 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
874 			    (ND_IFINFO(nd6_ifp)->flags &
875 			     ND6_IFF_ACCEPT_RTADV) != 0)
876 				/*
877 				 * Even if the neighbor is not in the default
878 				 * router list, the neighbor may be used as a
879 				 * next hop for some destinations (e.g. redirect
880 				 * case). So we must call rt6_flush explicitly.
881 				 */
882 				rt6_flush(&ip6->ip6_src, ifp);
883 		}
884 		ln->ln_router = is_router;
885 	}
886         /* XXX - QL
887 	 *  Does this matter?
888 	 *  rt->rt_flags &= ~RTF_REJECT;
889 	 */
890 	ln->la_asked = 0;
891 	if (ln->la_hold != NULL)
892 		nd6_grab_holdchain(ln, &chain, &sin6);
893  freeit:
894 	if (ln != NULL)
895 		LLE_WUNLOCK(ln);
896 
897 	if (chain != NULL)
898 		nd6_flush_holdchain(ifp, chain, &sin6);
899 
900 	if (checklink)
901 		pfxlist_onlink_check();
902 
903 	m_freem(m);
904 	return;
905 
906  bad:
907 	if (ln != NULL)
908 		LLE_WUNLOCK(ln);
909 
910 	ICMP6STAT_INC(icp6s_badna);
911 	m_freem(m);
912 }
913 
914 /*
915  * Neighbor advertisement output handling.
916  *
917  * Based on RFC 2461
918  *
919  * the following items are not implemented yet:
920  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
921  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
922  *
923  * tlladdr - 1 if include target link-layer address
924  * sdl0 - sockaddr_dl (= proxy NA) or NULL
925  */
926 static void
927 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
928     const struct in6_addr *taddr6, u_long flags, int tlladdr,
929     struct sockaddr *sdl0, u_int fibnum)
930 {
931 	struct mbuf *m;
932 	struct m_tag *mtag;
933 	struct ip6_hdr *ip6;
934 	struct nd_neighbor_advert *nd_na;
935 	struct ip6_moptions im6o;
936 	struct in6_addr daddr6, dst6, src6;
937 	uint32_t scopeid;
938 
939 	int icmp6len, maxlen, error;
940 	caddr_t mac = NULL;
941 
942 	daddr6 = *daddr6_0;	/* make a local copy for modification */
943 
944 	/* estimate the size of message */
945 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
946 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
947 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
948 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
949 	    __func__, max_linkhdr, maxlen, MCLBYTES));
950 
951 	if (max_linkhdr + maxlen > MHLEN)
952 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
953 	else
954 		m = m_gethdr(M_NOWAIT, MT_DATA);
955 	if (m == NULL)
956 		return;
957 	M_SETFIB(m, fibnum);
958 
959 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
960 		m->m_flags |= M_MCAST;
961 		im6o.im6o_multicast_ifp = ifp;
962 		im6o.im6o_multicast_hlim = 255;
963 		im6o.im6o_multicast_loop = 0;
964 	}
965 
966 	icmp6len = sizeof(*nd_na);
967 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
968 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
969 
970 	/* fill neighbor advertisement packet */
971 	ip6 = mtod(m, struct ip6_hdr *);
972 	ip6->ip6_flow = 0;
973 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
974 	ip6->ip6_vfc |= IPV6_VERSION;
975 	ip6->ip6_nxt = IPPROTO_ICMPV6;
976 	ip6->ip6_hlim = 255;
977 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
978 		/* reply to DAD */
979 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
980 		daddr6.s6_addr16[1] = 0;
981 		daddr6.s6_addr32[1] = 0;
982 		daddr6.s6_addr32[2] = 0;
983 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
984 		if (in6_setscope(&daddr6, ifp, NULL))
985 			goto bad;
986 
987 		flags &= ~ND_NA_FLAG_SOLICITED;
988 	}
989 	ip6->ip6_dst = daddr6;
990 
991 	/*
992 	 * Select a source whose scope is the same as that of the dest.
993 	 */
994 	in6_splitscope(&daddr6, &dst6, &scopeid);
995 	error = in6_selectsrc_addr(fibnum, &dst6,
996 	    scopeid, ifp, &src6, NULL);
997 	if (error) {
998 		char ip6buf[INET6_ADDRSTRLEN];
999 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1000 		    "determined: dst=%s, error=%d\n",
1001 		    ip6_sprintf(ip6buf, &daddr6), error));
1002 		goto bad;
1003 	}
1004 	ip6->ip6_src = src6;
1005 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1006 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1007 	nd_na->nd_na_code = 0;
1008 	nd_na->nd_na_target = *taddr6;
1009 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1010 
1011 	/*
1012 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1013 	 * see nd6_ns_input() for details.
1014 	 * Basically, if NS packet is sent to unicast/anycast addr,
1015 	 * target lladdr option SHOULD NOT be included.
1016 	 */
1017 	if (tlladdr) {
1018 		/*
1019 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1020 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1021 		 * my address) use lladdr configured for the interface.
1022 		 */
1023 		if (sdl0 == NULL) {
1024 			if (ifp->if_carp)
1025 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1026 			if (mac == NULL)
1027 				mac = nd6_ifptomac(ifp);
1028 		} else if (sdl0->sa_family == AF_LINK) {
1029 			struct sockaddr_dl *sdl;
1030 			sdl = (struct sockaddr_dl *)sdl0;
1031 			if (sdl->sdl_alen == ifp->if_addrlen)
1032 				mac = LLADDR(sdl);
1033 		}
1034 	}
1035 	if (tlladdr && mac) {
1036 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1037 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1038 
1039 		/* roundup to 8 bytes alignment! */
1040 		optlen = (optlen + 7) & ~7;
1041 
1042 		m->m_pkthdr.len += optlen;
1043 		m->m_len += optlen;
1044 		icmp6len += optlen;
1045 		bzero((caddr_t)nd_opt, optlen);
1046 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1047 		nd_opt->nd_opt_len = optlen >> 3;
1048 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1049 	} else
1050 		flags &= ~ND_NA_FLAG_OVERRIDE;
1051 
1052 	ip6->ip6_plen = htons((u_short)icmp6len);
1053 	nd_na->nd_na_flags_reserved = flags;
1054 	nd_na->nd_na_cksum = 0;
1055 	nd_na->nd_na_cksum =
1056 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1057 
1058 	if (send_sendso_input_hook != NULL) {
1059 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1060 		    sizeof(unsigned short), M_NOWAIT);
1061 		if (mtag == NULL)
1062 			goto bad;
1063 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1064 		m_tag_prepend(m, mtag);
1065 	}
1066 
1067 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1068 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1069 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1070 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1071 
1072 	return;
1073 
1074   bad:
1075 	m_freem(m);
1076 }
1077 
1078 #ifndef BURN_BRIDGES
1079 void
1080 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1081     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1082     struct sockaddr *sdl0)
1083 {
1084 
1085 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1086 	    RT_DEFAULT_FIB);
1087 }
1088 #endif
1089 
1090 caddr_t
1091 nd6_ifptomac(struct ifnet *ifp)
1092 {
1093 	switch (ifp->if_type) {
1094 	case IFT_ETHER:
1095 	case IFT_IEEE1394:
1096 	case IFT_L2VLAN:
1097 	case IFT_INFINIBAND:
1098 	case IFT_BRIDGE:
1099 		return IF_LLADDR(ifp);
1100 	default:
1101 		return NULL;
1102 	}
1103 }
1104 
1105 struct dadq {
1106 	TAILQ_ENTRY(dadq) dad_list;
1107 	struct ifaddr *dad_ifa;
1108 	int dad_count;		/* max NS to send */
1109 	int dad_ns_tcount;	/* # of trials to send NS */
1110 	int dad_ns_ocount;	/* NS sent so far */
1111 	int dad_ns_icount;
1112 	int dad_na_icount;
1113 	int dad_ns_lcount;	/* looped back NS */
1114 	int dad_loopbackprobe;	/* probing state for loopback detection */
1115 	struct callout dad_timer_ch;
1116 	struct vnet *dad_vnet;
1117 	u_int dad_refcnt;
1118 #define	ND_OPT_NONCE_LEN32 \
1119 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1120 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1121 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1122 };
1123 
1124 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1125 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1126 #define	V_dadq			VNET(dadq)
1127 #define	V_dad_rwlock		VNET(dad_rwlock)
1128 
1129 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1130 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1131 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1132 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1133 
1134 static void
1135 nd6_dad_add(struct dadq *dp)
1136 {
1137 
1138 	DADQ_WLOCK();
1139 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1140 	dp->dad_ondadq = true;
1141 	DADQ_WUNLOCK();
1142 }
1143 
1144 static void
1145 nd6_dad_del(struct dadq *dp)
1146 {
1147 
1148 	DADQ_WLOCK();
1149 	if (dp->dad_ondadq) {
1150 		/*
1151 		 * Remove dp from the dadq and release the dadq's
1152 		 * reference.
1153 		 */
1154 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1155 		dp->dad_ondadq = false;
1156 		DADQ_WUNLOCK();
1157 		nd6_dad_rele(dp);
1158 	} else
1159 		DADQ_WUNLOCK();
1160 }
1161 
1162 static struct dadq *
1163 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1164 {
1165 	struct dadq *dp;
1166 
1167 	DADQ_RLOCK();
1168 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1169 		if (dp->dad_ifa != ifa)
1170 			continue;
1171 		/*
1172 		 * Skip if the nonce matches the received one.
1173 		 * +2 in the length is required because of type and
1174 		 * length fields are included in a header.
1175 		 */
1176 		if (n != NULL &&
1177 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1178 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1179 		        ND_OPT_NONCE_LEN) == 0) {
1180 			dp->dad_ns_lcount++;
1181 			continue;
1182 		}
1183 		refcount_acquire(&dp->dad_refcnt);
1184 		break;
1185 	}
1186 	DADQ_RUNLOCK();
1187 
1188 	return (dp);
1189 }
1190 
1191 static void
1192 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1193 {
1194 
1195 	NET_EPOCH_ASSERT();
1196 
1197 	if (send_ns != 0)
1198 		nd6_dad_ns_output(dp);
1199 	callout_reset(&dp->dad_timer_ch, ticks,
1200 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1201 }
1202 
1203 static void
1204 nd6_dad_stoptimer(struct dadq *dp)
1205 {
1206 
1207 	callout_drain(&dp->dad_timer_ch);
1208 }
1209 
1210 static void
1211 nd6_dad_rele(struct dadq *dp)
1212 {
1213 
1214 	if (refcount_release(&dp->dad_refcnt)) {
1215 		ifa_free(dp->dad_ifa);
1216 		free(dp, M_IP6NDP);
1217 	}
1218 }
1219 
1220 void
1221 nd6_dad_init(void)
1222 {
1223 
1224 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1225 	TAILQ_INIT(&V_dadq);
1226 }
1227 
1228 /*
1229  * Start Duplicate Address Detection (DAD) for specified interface address.
1230  */
1231 void
1232 nd6_dad_start(struct ifaddr *ifa, int delay)
1233 {
1234 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1235 	struct dadq *dp;
1236 	char ip6buf[INET6_ADDRSTRLEN];
1237 	struct epoch_tracker et;
1238 
1239 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1240 	    ("starting DAD on non-tentative address %p", ifa));
1241 
1242 	/*
1243 	 * If we don't need DAD, don't do it.
1244 	 * There are several cases:
1245 	 * - DAD is disabled globally or on the interface
1246 	 * - the interface address is anycast
1247 	 */
1248 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1249 	    V_ip6_dad_count == 0 ||
1250 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1251 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1252 		return;
1253 	}
1254 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1255 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1256 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1257 		return;
1258 
1259 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1260 		/*
1261 		 * DAD is already in progress.  Let the existing entry
1262 		 * finish it.
1263 		 */
1264 		nd6_dad_rele(dp);
1265 		return;
1266 	}
1267 
1268 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1269 	if (dp == NULL) {
1270 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1271 			"%s(%s)\n",
1272 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1273 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1274 		return;
1275 	}
1276 	callout_init(&dp->dad_timer_ch, 0);
1277 #ifdef VIMAGE
1278 	dp->dad_vnet = curvnet;
1279 #endif
1280 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1281 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1282 
1283 	/*
1284 	 * Send NS packet for DAD, ip6_dad_count times.
1285 	 * Note that we must delay the first transmission, if this is the
1286 	 * first packet to be sent from the interface after interface
1287 	 * (re)initialization.
1288 	 */
1289 	dp->dad_ifa = ifa;
1290 	ifa_ref(dp->dad_ifa);
1291 	dp->dad_count = V_ip6_dad_count;
1292 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1293 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1294 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1295 
1296 	/* Add this to the dadq and add a reference for the dadq. */
1297 	refcount_init(&dp->dad_refcnt, 1);
1298 	nd6_dad_add(dp);
1299 	NET_EPOCH_ENTER(et);
1300 	nd6_dad_starttimer(dp, delay, 0);
1301 	NET_EPOCH_EXIT(et);
1302 }
1303 
1304 /*
1305  * terminate DAD unconditionally.  used for address removals.
1306  */
1307 void
1308 nd6_dad_stop(struct ifaddr *ifa)
1309 {
1310 	struct dadq *dp;
1311 
1312 	dp = nd6_dad_find(ifa, NULL);
1313 	if (!dp) {
1314 		/* DAD wasn't started yet */
1315 		return;
1316 	}
1317 
1318 	nd6_dad_stoptimer(dp);
1319 	nd6_dad_del(dp);
1320 
1321 	/* Release this function's reference, acquired by nd6_dad_find(). */
1322 	nd6_dad_rele(dp);
1323 }
1324 
1325 static void
1326 nd6_dad_timer(struct dadq *dp)
1327 {
1328 	CURVNET_SET(dp->dad_vnet);
1329 	struct ifaddr *ifa = dp->dad_ifa;
1330 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1331 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1332 	char ip6buf[INET6_ADDRSTRLEN];
1333 	struct epoch_tracker et;
1334 
1335 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1336 
1337 	NET_EPOCH_ENTER(et);
1338 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1339 		/* Do not need DAD for ifdisabled interface. */
1340 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1341 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1342 		goto err;
1343 	}
1344 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1345 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1346 			"%s(%s)\n",
1347 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1348 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1349 		goto err;
1350 	}
1351 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1352 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1353 			"%s(%s)\n",
1354 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1355 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1356 		goto err;
1357 	}
1358 
1359 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1360 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1361 	    (((ifp->if_flags & IFF_UP) == 0) ||
1362 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1363 		nd6log((LOG_INFO, "%s: could not run DAD "
1364 		    "because the interface was down or not running.\n",
1365 		    if_name(ifa->ifa_ifp)));
1366 		goto err;
1367 	}
1368 
1369 	/* Need more checks? */
1370 	if (dp->dad_ns_ocount < dp->dad_count) {
1371 		/*
1372 		 * We have more NS to go.  Send NS packet for DAD.
1373 		 */
1374 		nd6_dad_starttimer(dp,
1375 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1376 		goto done;
1377 	} else {
1378 		/*
1379 		 * We have transmitted sufficient number of DAD packets.
1380 		 * See what we've got.
1381 		 */
1382 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1383 			/* We've seen NS or NA, means DAD has failed. */
1384 			nd6_dad_duplicated(ifa, dp);
1385 		else if (V_dad_enhanced != 0 &&
1386 		    dp->dad_ns_lcount > 0 &&
1387 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1388 			/*
1389 			 * Sec. 4.1 in RFC 7527 requires transmission of
1390 			 * additional probes until the loopback condition
1391 			 * becomes clear when a looped back probe is detected.
1392 			 */
1393 			log(LOG_ERR, "%s: a looped back NS message is "
1394 			    "detected during DAD for %s.  "
1395 			    "Another DAD probes are being sent.\n",
1396 			    if_name(ifa->ifa_ifp),
1397 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1398 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1399 			/*
1400 			 * Send an NS immediately and increase dad_count by
1401 			 * V_nd6_mmaxtries - 1.
1402 			 */
1403 			dp->dad_count =
1404 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1405 			nd6_dad_starttimer(dp,
1406 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1407 			    1);
1408 			goto done;
1409 		} else {
1410 			/*
1411 			 * We are done with DAD.  No NA came, no NS came.
1412 			 * No duplicate address found.  Check IFDISABLED flag
1413 			 * again in case that it is changed between the
1414 			 * beginning of this function and here.
1415 			 */
1416 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1417 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1418 
1419 			nd6log((LOG_DEBUG,
1420 			    "%s: DAD complete for %s - no duplicates found\n",
1421 			    if_name(ifa->ifa_ifp),
1422 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1423 			if (dp->dad_ns_lcount > 0)
1424 				log(LOG_ERR, "%s: DAD completed while "
1425 				    "a looped back NS message is detected "
1426 				    "during DAD for %s.\n",
1427 				    if_name(ifa->ifa_ifp),
1428 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1429 		}
1430 	}
1431 err:
1432 	nd6_dad_del(dp);
1433 done:
1434 	NET_EPOCH_EXIT(et);
1435 	CURVNET_RESTORE();
1436 }
1437 
1438 static void
1439 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1440 {
1441 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1442 	struct ifnet *ifp;
1443 	char ip6buf[INET6_ADDRSTRLEN];
1444 
1445 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1446 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1447 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1448 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1449 	    dp->dad_na_icount);
1450 
1451 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1452 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1453 
1454 	ifp = ifa->ifa_ifp;
1455 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1456 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1457 	log(LOG_ERR, "%s: manual intervention required\n",
1458 	    if_name(ifp));
1459 
1460 	/*
1461 	 * If the address is a link-local address formed from an interface
1462 	 * identifier based on the hardware address which is supposed to be
1463 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1464 	 * operation on the interface SHOULD be disabled.
1465 	 * [RFC 4862, Section 5.4.5]
1466 	 */
1467 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1468 		struct in6_addr in6;
1469 
1470 		/*
1471 		 * To avoid over-reaction, we only apply this logic when we are
1472 		 * very sure that hardware addresses are supposed to be unique.
1473 		 */
1474 		switch (ifp->if_type) {
1475 		case IFT_ETHER:
1476 		case IFT_ATM:
1477 		case IFT_IEEE1394:
1478 		case IFT_INFINIBAND:
1479 			in6 = ia->ia_addr.sin6_addr;
1480 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1481 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1482 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1483 				log(LOG_ERR, "%s: possible hardware address "
1484 				    "duplication detected, disable IPv6\n",
1485 				    if_name(ifp));
1486 			}
1487 			break;
1488 		}
1489 	}
1490 }
1491 
1492 static void
1493 nd6_dad_ns_output(struct dadq *dp)
1494 {
1495 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1496 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1497 	int i;
1498 
1499 	dp->dad_ns_tcount++;
1500 	if ((ifp->if_flags & IFF_UP) == 0) {
1501 		return;
1502 	}
1503 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1504 		return;
1505 	}
1506 
1507 	dp->dad_ns_ocount++;
1508 	if (V_dad_enhanced != 0) {
1509 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1510 			dp->dad_nonce[i] = arc4random();
1511 		/*
1512 		 * XXXHRS: Note that in the case that
1513 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1514 		 * different nonces can be looped back in an unexpected
1515 		 * order.  The current implementation recognizes only
1516 		 * the latest nonce on the sender side.  Practically it
1517 		 * should work well in almost all cases.
1518 		 */
1519 	}
1520 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1521 	    (uint8_t *)&dp->dad_nonce[0]);
1522 }
1523 
1524 static void
1525 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1526 {
1527 	struct dadq *dp;
1528 
1529 	if (ifa == NULL)
1530 		panic("ifa == NULL in nd6_dad_ns_input");
1531 
1532 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1533 	if (V_dad_enhanced == 0)
1534 		ndopt_nonce = NULL;
1535 	dp = nd6_dad_find(ifa, ndopt_nonce);
1536 	if (dp == NULL)
1537 		return;
1538 
1539 	dp->dad_ns_icount++;
1540 	nd6_dad_rele(dp);
1541 }
1542 
1543 static void
1544 nd6_dad_na_input(struct ifaddr *ifa)
1545 {
1546 	struct dadq *dp;
1547 
1548 	if (ifa == NULL)
1549 		panic("ifa == NULL in nd6_dad_na_input");
1550 
1551 	dp = nd6_dad_find(ifa, NULL);
1552 	if (dp != NULL) {
1553 		dp->dad_na_icount++;
1554 		nd6_dad_rele(dp);
1555 	}
1556 }
1557