xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 6132212808e8dccedc9e5d85fea4390c2f38059a)
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 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/eventhandler.h>
44 #include <sys/malloc.h>
45 #include <sys/libkern.h>
46 #include <sys/lock.h>
47 #include <sys/rwlock.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/errno.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
56 #include <sys/queue.h>
57 #include <sys/callout.h>
58 #include <sys/refcount.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_dl.h>
63 #include <net/if_var.h>
64 #include <net/route.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <net/if_llatbl.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip_carp.h>
78 #include <netinet6/send.h>
79 
80 #define SDL(s) ((struct sockaddr_dl *)s)
81 
82 struct dadq;
83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
84 static void nd6_dad_add(struct dadq *dp);
85 static void nd6_dad_del(struct dadq *dp);
86 static void nd6_dad_rele(struct dadq *);
87 static void nd6_dad_starttimer(struct dadq *, int, int);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(struct dadq *);
90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
91 static void nd6_dad_ns_output(struct dadq *);
92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
93 static void nd6_dad_na_input(struct ifaddr *);
94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
95     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
97     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
98 
99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
100 #define	V_dad_enhanced			VNET(dad_enhanced)
101 
102 SYSCTL_DECL(_net_inet6_ip6);
103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
104     &VNET_NAME(dad_enhanced), 0,
105     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
106 
107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
108 						   transmit DAD packet */
109 #define	V_dad_maxtry			VNET(dad_maxtry)
110 
111 /*
112  * Input a Neighbor Solicitation Message.
113  *
114  * Based on RFC 2461
115  * Based on RFC 2462 (duplicate address detection)
116  */
117 void
118 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
119 {
120 	struct ifnet *ifp;
121 	struct ip6_hdr *ip6;
122 	struct nd_neighbor_solicit *nd_ns;
123 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
124 	struct ifaddr *ifa;
125 	struct sockaddr_dl proxydl;
126 	union nd_opts ndopts;
127 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
128 	char *lladdr;
129 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
130 
131 	ifa = NULL;
132 
133 	/* RFC 6980: Nodes MUST silently ignore fragments */
134 	if(m->m_flags & M_FRAGMENTED)
135 		goto freeit;
136 
137 	ifp = m->m_pkthdr.rcvif;
138 	ip6 = mtod(m, struct ip6_hdr *);
139 	if (ip6->ip6_hlim != 255) {
140 		nd6log((LOG_ERR,
141 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
142 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
143 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
144 		goto bads;
145 	}
146 
147 	if (m->m_len < off + icmp6len) {
148 		m = m_pullup(m, off + icmp6len);
149 		if (m == NULL) {
150 			IP6STAT_INC(ip6s_exthdrtoolong);
151 			return;
152 		}
153 	}
154 	ip6 = mtod(m, struct ip6_hdr *);
155 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
156 
157 	saddr6 = ip6->ip6_src;
158 	daddr6 = ip6->ip6_dst;
159 	taddr6 = nd_ns->nd_ns_target;
160 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
161 		goto bad;
162 
163 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
164 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
165 		rflag = 0;
166 
167 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
168 		/* dst has to be a solicited node multicast address. */
169 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
170 		    /* don't check ifindex portion */
171 		    daddr6.s6_addr32[1] == 0 &&
172 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
173 		    daddr6.s6_addr8[12] == 0xff) {
174 			; /* good */
175 		} else {
176 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
177 			    "(wrong ip6 dst)\n"));
178 			goto bad;
179 		}
180 	} else if (!V_nd6_onlink_ns_rfc4861) {
181 		struct sockaddr_in6 src_sa6;
182 
183 		/*
184 		 * According to recent IETF discussions, it is not a good idea
185 		 * to accept a NS from an address which would not be deemed
186 		 * to be a neighbor otherwise.  This point is expected to be
187 		 * clarified in future revisions of the specification.
188 		 */
189 		bzero(&src_sa6, sizeof(src_sa6));
190 		src_sa6.sin6_family = AF_INET6;
191 		src_sa6.sin6_len = sizeof(src_sa6);
192 		src_sa6.sin6_addr = saddr6;
193 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
194 			nd6log((LOG_INFO, "nd6_ns_input: "
195 				"NS packet from non-neighbor\n"));
196 			goto bad;
197 		}
198 	}
199 
200 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
201 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
202 		goto bad;
203 	}
204 
205 	icmp6len -= sizeof(*nd_ns);
206 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
207 	if (nd6_options(&ndopts) < 0) {
208 		nd6log((LOG_INFO,
209 		    "nd6_ns_input: invalid ND option, ignored\n"));
210 		/* nd6_options have incremented stats */
211 		goto freeit;
212 	}
213 
214 	lladdr = NULL;
215 	lladdrlen = 0;
216 	if (ndopts.nd_opts_src_lladdr) {
217 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
218 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
219 	}
220 
221 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
222 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
223 		    "(link-layer address option)\n"));
224 		goto bad;
225 	}
226 
227 	/*
228 	 * Attaching target link-layer address to the NA?
229 	 * (RFC 2461 7.2.4)
230 	 *
231 	 * NS IP dst is unicast/anycast			MUST NOT add
232 	 * NS IP dst is solicited-node multicast	MUST add
233 	 *
234 	 * In implementation, we add target link-layer address by default.
235 	 * We do not add one in MUST NOT cases.
236 	 */
237 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
238 		tlladdr = 0;
239 	else
240 		tlladdr = 1;
241 
242 	/*
243 	 * Target address (taddr6) must be either:
244 	 * (1) Valid unicast/anycast address for my receiving interface,
245 	 * (2) Unicast address for which I'm offering proxy service, or
246 	 * (3) "tentative" address on which DAD is being performed.
247 	 */
248 	/* (1) and (3) check. */
249 	if (ifp->if_carp)
250 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
251 	else
252 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
253 
254 	/* (2) check. */
255 	proxy = 0;
256 	if (ifa == NULL) {
257 		struct sockaddr_dl rt_gateway;
258 		struct rt_addrinfo info;
259 		struct sockaddr_in6 dst6;
260 
261 		bzero(&dst6, sizeof(dst6));
262 		dst6.sin6_len = sizeof(struct sockaddr_in6);
263 		dst6.sin6_family = AF_INET6;
264 		dst6.sin6_addr = taddr6;
265 
266 		bzero(&rt_gateway, sizeof(rt_gateway));
267 		rt_gateway.sdl_len = sizeof(rt_gateway);
268 		bzero(&info, sizeof(info));
269 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
270 
271 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
272 		    0, 0, &info) == 0) {
273 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
274 			    rt_gateway.sdl_family == AF_LINK) {
275 				/*
276 				 * proxy NDP for single entry
277 				 */
278 				proxydl = *SDL(&rt_gateway);
279 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
280 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
281 				if (ifa)
282 					proxy = 1;
283 			}
284 		}
285 	}
286 	if (ifa == NULL) {
287 		/*
288 		 * We've got an NS packet, and we don't have that adddress
289 		 * assigned for us.  We MUST silently ignore it.
290 		 * See RFC2461 7.2.3.
291 		 */
292 		goto freeit;
293 	}
294 	myaddr6 = *IFA_IN6(ifa);
295 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
296 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
297 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
298 		goto freeit;
299 
300 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
301 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
302 		    "(if %d, NS packet %d)\n",
303 		    ip6_sprintf(ip6bufs, &taddr6),
304 		    ifp->if_addrlen, lladdrlen - 2));
305 		goto bad;
306 	}
307 
308 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
309 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
310 		    ip6_sprintf(ip6bufs, &saddr6)));
311 		goto freeit;
312 	}
313 
314 	/*
315 	 * We have neighbor solicitation packet, with target address equals to
316 	 * one of my tentative address.
317 	 *
318 	 * src addr	how to process?
319 	 * ---		---
320 	 * multicast	of course, invalid (rejected in ip6_input)
321 	 * unicast	somebody is doing address resolution -> ignore
322 	 * unspec	dup address detection
323 	 *
324 	 * The processing is defined in RFC 2462.
325 	 */
326 	if (tentative) {
327 		/*
328 		 * If source address is unspecified address, it is for
329 		 * duplicate address detection.
330 		 *
331 		 * If not, the packet is for addess resolution;
332 		 * silently ignore it.
333 		 */
334 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
335 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
336 
337 		goto freeit;
338 	}
339 
340 	/*
341 	 * If the source address is unspecified address, entries must not
342 	 * be created or updated.
343 	 * It looks that sender is performing DAD.  Output NA toward
344 	 * all-node multicast address, to tell the sender that I'm using
345 	 * the address.
346 	 * S bit ("solicited") must be zero.
347 	 */
348 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
349 		struct in6_addr in6_all;
350 
351 		in6_all = in6addr_linklocal_allnodes;
352 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
353 			goto bad;
354 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
355 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
356 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
357 		    M_GETFIB(m));
358 		goto freeit;
359 	}
360 
361 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
362 	    ND_NEIGHBOR_SOLICIT, 0);
363 
364 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
365 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
366 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
367 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
368  freeit:
369 	if (ifa != NULL)
370 		ifa_free(ifa);
371 	m_freem(m);
372 	return;
373 
374  bad:
375 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
376 		ip6_sprintf(ip6bufs, &saddr6)));
377 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
378 		ip6_sprintf(ip6bufs, &daddr6)));
379 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
380 		ip6_sprintf(ip6bufs, &taddr6)));
381  bads:
382 	ICMP6STAT_INC(icp6s_badns);
383 	if (ifa != NULL)
384 		ifa_free(ifa);
385 	m_freem(m);
386 }
387 
388 /*
389  * Output a Neighbor Solicitation Message. Caller specifies:
390  *	- ICMP6 header source IP6 address
391  *	- ND6 header target IP6 address
392  *	- ND6 header source datalink address
393  *
394  * Based on RFC 2461
395  * Based on RFC 2462 (duplicate address detection)
396  *
397  *    ln - for source address determination
398  * nonce - If non-NULL, NS is used for duplicate address detection and
399  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
400  */
401 static void
402 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
403     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
404     uint8_t *nonce, u_int fibnum)
405 {
406 	struct mbuf *m;
407 	struct m_tag *mtag;
408 	struct ip6_hdr *ip6;
409 	struct nd_neighbor_solicit *nd_ns;
410 	struct ip6_moptions im6o;
411 	int icmp6len;
412 	int maxlen;
413 	caddr_t mac;
414 
415 	if (IN6_IS_ADDR_MULTICAST(taddr6))
416 		return;
417 
418 	/* estimate the size of message */
419 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
420 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
421 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
422 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
423 	    __func__, max_linkhdr, maxlen, MCLBYTES));
424 
425 	if (max_linkhdr + maxlen > MHLEN)
426 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
427 	else
428 		m = m_gethdr(M_NOWAIT, MT_DATA);
429 	if (m == NULL)
430 		return;
431 	M_SETFIB(m, fibnum);
432 
433 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
434 		m->m_flags |= M_MCAST;
435 		im6o.im6o_multicast_ifp = ifp;
436 		im6o.im6o_multicast_hlim = 255;
437 		im6o.im6o_multicast_loop = 0;
438 	}
439 
440 	icmp6len = sizeof(*nd_ns);
441 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
442 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
443 
444 	/* fill neighbor solicitation packet */
445 	ip6 = mtod(m, struct ip6_hdr *);
446 	ip6->ip6_flow = 0;
447 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
448 	ip6->ip6_vfc |= IPV6_VERSION;
449 	/* ip6->ip6_plen will be set later */
450 	ip6->ip6_nxt = IPPROTO_ICMPV6;
451 	ip6->ip6_hlim = 255;
452 	if (daddr6)
453 		ip6->ip6_dst = *daddr6;
454 	else {
455 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
456 		ip6->ip6_dst.s6_addr16[1] = 0;
457 		ip6->ip6_dst.s6_addr32[1] = 0;
458 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
459 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
460 		ip6->ip6_dst.s6_addr8[12] = 0xff;
461 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
462 			goto bad;
463 	}
464 	if (nonce == NULL) {
465 		struct ifaddr *ifa = NULL;
466 
467 		/*
468 		 * RFC2461 7.2.2:
469 		 * "If the source address of the packet prompting the
470 		 * solicitation is the same as one of the addresses assigned
471 		 * to the outgoing interface, that address SHOULD be placed
472 		 * in the IP Source Address of the outgoing solicitation.
473 		 * Otherwise, any one of the addresses assigned to the
474 		 * interface should be used."
475 		 *
476 		 * We use the source address for the prompting packet
477 		 * (saddr6), if saddr6 belongs to the outgoing interface.
478 		 * Otherwise, we perform the source address selection as usual.
479 		 */
480 
481 		if (saddr6 != NULL)
482 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
483 		if (ifa != NULL) {
484 			/* ip6_src set already. */
485 			ip6->ip6_src = *saddr6;
486 			ifa_free(ifa);
487 		} else {
488 			int error;
489 			struct in6_addr dst6, src6;
490 			uint32_t scopeid;
491 
492 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
493 			error = in6_selectsrc_addr(fibnum, &dst6,
494 			    scopeid, ifp, &src6, NULL);
495 			if (error) {
496 				char ip6buf[INET6_ADDRSTRLEN];
497 				nd6log((LOG_DEBUG, "%s: source can't be "
498 				    "determined: dst=%s, error=%d\n", __func__,
499 				    ip6_sprintf(ip6buf, &dst6),
500 				    error));
501 				goto bad;
502 			}
503 			ip6->ip6_src = src6;
504 		}
505 	} else {
506 		/*
507 		 * Source address for DAD packet must always be IPv6
508 		 * unspecified address. (0::0)
509 		 * We actually don't have to 0-clear the address (we did it
510 		 * above), but we do so here explicitly to make the intention
511 		 * clearer.
512 		 */
513 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
514 	}
515 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
516 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
517 	nd_ns->nd_ns_code = 0;
518 	nd_ns->nd_ns_reserved = 0;
519 	nd_ns->nd_ns_target = *taddr6;
520 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
521 
522 	/*
523 	 * Add source link-layer address option.
524 	 *
525 	 *				spec		implementation
526 	 *				---		---
527 	 * DAD packet			MUST NOT	do not add the option
528 	 * there's no link layer address:
529 	 *				impossible	do not add the option
530 	 * there's link layer address:
531 	 *	Multicast NS		MUST add one	add the option
532 	 *	Unicast NS		SHOULD add one	add the option
533 	 */
534 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
535 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
536 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
537 		/* 8 byte alignments... */
538 		optlen = (optlen + 7) & ~7;
539 
540 		m->m_pkthdr.len += optlen;
541 		m->m_len += optlen;
542 		icmp6len += optlen;
543 		bzero((caddr_t)nd_opt, optlen);
544 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
545 		nd_opt->nd_opt_len = optlen >> 3;
546 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
547 	}
548 	/*
549 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
550 	 * This behavior is documented as Enhanced Duplicate Address
551 	 * Detection in RFC 7527.
552 	 * net.inet6.ip6.dad_enhanced=0 disables this.
553 	 */
554 	if (V_dad_enhanced != 0 && nonce != NULL) {
555 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
556 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
557 		/* 8-byte alignment is required. */
558 		optlen = (optlen + 7) & ~7;
559 
560 		m->m_pkthdr.len += optlen;
561 		m->m_len += optlen;
562 		icmp6len += optlen;
563 		bzero((caddr_t)nd_opt, optlen);
564 		nd_opt->nd_opt_type = ND_OPT_NONCE;
565 		nd_opt->nd_opt_len = optlen >> 3;
566 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
567 	}
568 	ip6->ip6_plen = htons((u_short)icmp6len);
569 	nd_ns->nd_ns_cksum = 0;
570 	nd_ns->nd_ns_cksum =
571 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
572 
573 	if (send_sendso_input_hook != NULL) {
574 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
575 			sizeof(unsigned short), M_NOWAIT);
576 		if (mtag == NULL)
577 			goto bad;
578 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
579 		m_tag_prepend(m, mtag);
580 	}
581 
582 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
583 	    &im6o, NULL, NULL);
584 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
585 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
586 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
587 
588 	return;
589 
590   bad:
591 	m_freem(m);
592 }
593 
594 #ifndef BURN_BRIDGES
595 void
596 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
597     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
598 {
599 
600 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
601 }
602 #endif
603 /*
604  * Neighbor advertisement input handling.
605  *
606  * Based on RFC 2461
607  * Based on RFC 2462 (duplicate address detection)
608  *
609  * the following items are not implemented yet:
610  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
611  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
612  */
613 void
614 nd6_na_input(struct mbuf *m, int off, int icmp6len)
615 {
616 	struct ifnet *ifp;
617 	struct ip6_hdr *ip6;
618 	struct ifaddr *ifa;
619 	struct llentry *ln;
620 	struct mbuf *chain;
621 	struct nd_neighbor_advert *nd_na;
622 	struct in6_addr daddr6, taddr6;
623 	struct sockaddr_in6 sin6;
624 	union nd_opts ndopts;
625 	u_char linkhdr[LLE_MAX_LINKHDR];
626 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
627 	char *lladdr;
628 	size_t linkhdrsize;
629 	int flags, is_override, is_router, is_solicited;
630 	int lladdr_off, lladdrlen, checklink;
631 
632 	NET_EPOCH_ASSERT();
633 
634 	chain = NULL;
635 	ln = NULL;
636 	checklink = 0;
637 
638 	/* RFC 6980: Nodes MUST silently ignore fragments */
639 	if(m->m_flags & M_FRAGMENTED)
640 		goto freeit;
641 
642 	ifp = m->m_pkthdr.rcvif;
643 	ip6 = mtod(m, struct ip6_hdr *);
644 	if (ip6->ip6_hlim != 255) {
645 		nd6log((LOG_ERR,
646 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
647 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
648 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
649 		goto bad;
650 	}
651 
652 	if (m->m_len < off + icmp6len) {
653 		m = m_pullup(m, off + icmp6len);
654 		if (m == NULL) {
655 			IP6STAT_INC(ip6s_exthdrtoolong);
656 			return;
657 		}
658 	}
659 	ip6 = mtod(m, struct ip6_hdr *);
660 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
661 
662 	flags = nd_na->nd_na_flags_reserved;
663 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
664 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
665 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
666 
667 	taddr6 = nd_na->nd_na_target;
668 	if (in6_setscope(&taddr6, ifp, NULL))
669 		goto bad;	/* XXX: impossible */
670 
671 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
672 		nd6log((LOG_ERR,
673 		    "nd6_na_input: invalid target address %s\n",
674 		    ip6_sprintf(ip6bufs, &taddr6)));
675 		goto bad;
676 	}
677 
678 	daddr6 = ip6->ip6_dst;
679 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
680 		if (is_solicited) {
681 			nd6log((LOG_ERR,
682 			    "nd6_na_input: a solicited adv is multicasted\n"));
683 			goto bad;
684 		}
685 
686 	icmp6len -= sizeof(*nd_na);
687 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
688 	if (nd6_options(&ndopts) < 0) {
689 		nd6log((LOG_INFO,
690 		    "nd6_na_input: invalid ND option, ignored\n"));
691 		/* nd6_options have incremented stats */
692 		goto freeit;
693 	}
694 
695 	lladdr = NULL;
696 	lladdrlen = 0;
697 	if (ndopts.nd_opts_tgt_lladdr) {
698 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
699 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
700 	}
701 
702 	/*
703 	 * This effectively disables the DAD check on a non-master CARP
704 	 * address.
705 	 */
706 	if (ifp->if_carp)
707 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
708 	else
709 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
710 
711 	/*
712 	 * Target address matches one of my interface address.
713 	 *
714 	 * If my address is tentative, this means that there's somebody
715 	 * already using the same address as mine.  This indicates DAD failure.
716 	 * This is defined in RFC 2462.
717 	 *
718 	 * Otherwise, process as defined in RFC 2461.
719 	 */
720 	if (ifa
721 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
722 		nd6_dad_na_input(ifa);
723 		ifa_free(ifa);
724 		goto freeit;
725 	}
726 
727 	/* Just for safety, maybe unnecessary. */
728 	if (ifa) {
729 		ifa_free(ifa);
730 		log(LOG_ERR,
731 		    "nd6_na_input: duplicate IP6 address %s\n",
732 		    ip6_sprintf(ip6bufs, &taddr6));
733 		goto freeit;
734 	}
735 
736 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
737 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
738 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
739 		    ifp->if_addrlen, lladdrlen - 2));
740 		goto bad;
741 	}
742 
743 	/*
744 	 * If no neighbor cache entry is found, NA SHOULD silently be
745 	 * discarded.
746 	 */
747 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
748 	if (ln == NULL) {
749 		goto freeit;
750 	}
751 
752 	/*
753 	 * Do not try to override static entry.
754 	 */
755 	if (ln->la_flags & LLE_STATIC)
756 		goto freeit;
757 
758 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
759 		/*
760 		 * If the link-layer has address, and no lladdr option came,
761 		 * discard the packet.
762 		 */
763 		if (ifp->if_addrlen && lladdr == NULL) {
764 			goto freeit;
765 		}
766 
767 		/*
768 		 * Record link-layer address, and update the state.
769 		 */
770 		linkhdrsize = sizeof(linkhdr);
771 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
772 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
773 			return;
774 
775 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
776 		    lladdr_off) == 0) {
777 			ln = NULL;
778 			goto freeit;
779 		}
780 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
781 		if (is_solicited)
782 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
783 		else
784 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
785 		if ((ln->ln_router = is_router) != 0) {
786 			/*
787 			 * This means a router's state has changed from
788 			 * non-reachable to probably reachable, and might
789 			 * affect the status of associated prefixes..
790 			 */
791 			checklink = 1;
792 		}
793 	} else {
794 		int llchange;
795 
796 		/*
797 		 * Check if the link-layer address has changed or not.
798 		 */
799 		if (lladdr == NULL)
800 			llchange = 0;
801 		else {
802 			if (ln->la_flags & LLE_VALID) {
803 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
804 					llchange = 1;
805 				else
806 					llchange = 0;
807 			} else
808 				llchange = 1;
809 		}
810 
811 		/*
812 		 * This is VERY complex.  Look at it with care.
813 		 *
814 		 * override solicit lladdr llchange	action
815 		 *					(L: record lladdr)
816 		 *
817 		 *	0	0	n	--	(2c)
818 		 *	0	0	y	n	(2b) L
819 		 *	0	0	y	y	(1)    REACHABLE->STALE
820 		 *	0	1	n	--	(2c)   *->REACHABLE
821 		 *	0	1	y	n	(2b) L *->REACHABLE
822 		 *	0	1	y	y	(1)    REACHABLE->STALE
823 		 *	1	0	n	--	(2a)
824 		 *	1	0	y	n	(2a) L
825 		 *	1	0	y	y	(2a) L *->STALE
826 		 *	1	1	n	--	(2a)   *->REACHABLE
827 		 *	1	1	y	n	(2a) L *->REACHABLE
828 		 *	1	1	y	y	(2a) L *->REACHABLE
829 		 */
830 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
831 			/*
832 			 * If state is REACHABLE, make it STALE.
833 			 * no other updates should be done.
834 			 */
835 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
836 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
837 			goto freeit;
838 		} else if (is_override				   /* (2a) */
839 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
840 			|| lladdr == NULL) {			   /* (2c) */
841 			/*
842 			 * Update link-local address, if any.
843 			 */
844 			if (lladdr != NULL) {
845 				linkhdrsize = sizeof(linkhdr);
846 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
847 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
848 					goto freeit;
849 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
850 				    linkhdrsize, lladdr_off) == 0) {
851 					ln = NULL;
852 					goto freeit;
853 				}
854 				EVENTHANDLER_INVOKE(lle_event, ln,
855 				    LLENTRY_RESOLVED);
856 			}
857 
858 			/*
859 			 * If solicited, make the state REACHABLE.
860 			 * If not solicited and the link-layer address was
861 			 * changed, make it STALE.
862 			 */
863 			if (is_solicited)
864 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
865 			else {
866 				if (lladdr != NULL && llchange)
867 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
868 			}
869 		}
870 
871 		if (ln->ln_router && !is_router) {
872 			/*
873 			 * The peer dropped the router flag.
874 			 * Remove the sender from the Default Router List and
875 			 * update the Destination Cache entries.
876 			 */
877 			struct ifnet *nd6_ifp;
878 
879 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
880 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
881 			    (ND_IFINFO(nd6_ifp)->flags &
882 			     ND6_IFF_ACCEPT_RTADV) != 0)
883 				/*
884 				 * Even if the neighbor is not in the default
885 				 * router list, the neighbor may be used as a
886 				 * next hop for some destinations (e.g. redirect
887 				 * case). So we must call rt6_flush explicitly.
888 				 */
889 				rt6_flush(&ip6->ip6_src, ifp);
890 		}
891 		ln->ln_router = is_router;
892 	}
893         /* XXX - QL
894 	 *  Does this matter?
895 	 *  rt->rt_flags &= ~RTF_REJECT;
896 	 */
897 	ln->la_asked = 0;
898 	if (ln->la_hold != NULL) {
899 		memset(&sin6, 0, sizeof(sin6));
900 		nd6_grab_holdchain(ln, &chain, &sin6);
901 	}
902  freeit:
903 	if (ln != NULL)
904 		LLE_WUNLOCK(ln);
905 
906 	if (chain != NULL)
907 		nd6_flush_holdchain(ifp, chain, &sin6);
908 
909 	if (checklink)
910 		pfxlist_onlink_check();
911 
912 	m_freem(m);
913 	return;
914 
915  bad:
916 	if (ln != NULL)
917 		LLE_WUNLOCK(ln);
918 
919 	ICMP6STAT_INC(icp6s_badna);
920 	m_freem(m);
921 }
922 
923 /*
924  * Neighbor advertisement output handling.
925  *
926  * Based on RFC 2461
927  *
928  * the following items are not implemented yet:
929  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
930  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
931  *
932  * tlladdr - 1 if include target link-layer address
933  * sdl0 - sockaddr_dl (= proxy NA) or NULL
934  */
935 static void
936 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
937     const struct in6_addr *taddr6, u_long flags, int tlladdr,
938     struct sockaddr *sdl0, u_int fibnum)
939 {
940 	struct mbuf *m;
941 	struct m_tag *mtag;
942 	struct ip6_hdr *ip6;
943 	struct nd_neighbor_advert *nd_na;
944 	struct ip6_moptions im6o;
945 	struct in6_addr daddr6, dst6, src6;
946 	uint32_t scopeid;
947 
948 	int icmp6len, maxlen, error;
949 	caddr_t mac = NULL;
950 
951 	daddr6 = *daddr6_0;	/* make a local copy for modification */
952 
953 	/* estimate the size of message */
954 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
955 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
956 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
957 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
958 	    __func__, max_linkhdr, maxlen, MCLBYTES));
959 
960 	if (max_linkhdr + maxlen > MHLEN)
961 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
962 	else
963 		m = m_gethdr(M_NOWAIT, MT_DATA);
964 	if (m == NULL)
965 		return;
966 	M_SETFIB(m, fibnum);
967 
968 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
969 		m->m_flags |= M_MCAST;
970 		im6o.im6o_multicast_ifp = ifp;
971 		im6o.im6o_multicast_hlim = 255;
972 		im6o.im6o_multicast_loop = 0;
973 	}
974 
975 	icmp6len = sizeof(*nd_na);
976 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
977 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
978 
979 	/* fill neighbor advertisement packet */
980 	ip6 = mtod(m, struct ip6_hdr *);
981 	ip6->ip6_flow = 0;
982 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
983 	ip6->ip6_vfc |= IPV6_VERSION;
984 	ip6->ip6_nxt = IPPROTO_ICMPV6;
985 	ip6->ip6_hlim = 255;
986 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
987 		/* reply to DAD */
988 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
989 		daddr6.s6_addr16[1] = 0;
990 		daddr6.s6_addr32[1] = 0;
991 		daddr6.s6_addr32[2] = 0;
992 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
993 		if (in6_setscope(&daddr6, ifp, NULL))
994 			goto bad;
995 
996 		flags &= ~ND_NA_FLAG_SOLICITED;
997 	}
998 	ip6->ip6_dst = daddr6;
999 
1000 	/*
1001 	 * Select a source whose scope is the same as that of the dest.
1002 	 */
1003 	in6_splitscope(&daddr6, &dst6, &scopeid);
1004 	error = in6_selectsrc_addr(fibnum, &dst6,
1005 	    scopeid, ifp, &src6, NULL);
1006 	if (error) {
1007 		char ip6buf[INET6_ADDRSTRLEN];
1008 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1009 		    "determined: dst=%s, error=%d\n",
1010 		    ip6_sprintf(ip6buf, &daddr6), error));
1011 		goto bad;
1012 	}
1013 	ip6->ip6_src = src6;
1014 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1015 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1016 	nd_na->nd_na_code = 0;
1017 	nd_na->nd_na_target = *taddr6;
1018 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1019 
1020 	/*
1021 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1022 	 * see nd6_ns_input() for details.
1023 	 * Basically, if NS packet is sent to unicast/anycast addr,
1024 	 * target lladdr option SHOULD NOT be included.
1025 	 */
1026 	if (tlladdr) {
1027 		/*
1028 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1029 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1030 		 * my address) use lladdr configured for the interface.
1031 		 */
1032 		if (sdl0 == NULL) {
1033 			if (ifp->if_carp)
1034 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1035 			if (mac == NULL)
1036 				mac = nd6_ifptomac(ifp);
1037 		} else if (sdl0->sa_family == AF_LINK) {
1038 			struct sockaddr_dl *sdl;
1039 			sdl = (struct sockaddr_dl *)sdl0;
1040 			if (sdl->sdl_alen == ifp->if_addrlen)
1041 				mac = LLADDR(sdl);
1042 		}
1043 	}
1044 	if (tlladdr && mac) {
1045 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1046 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1047 
1048 		/* roundup to 8 bytes alignment! */
1049 		optlen = (optlen + 7) & ~7;
1050 
1051 		m->m_pkthdr.len += optlen;
1052 		m->m_len += optlen;
1053 		icmp6len += optlen;
1054 		bzero((caddr_t)nd_opt, optlen);
1055 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1056 		nd_opt->nd_opt_len = optlen >> 3;
1057 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1058 	} else
1059 		flags &= ~ND_NA_FLAG_OVERRIDE;
1060 
1061 	ip6->ip6_plen = htons((u_short)icmp6len);
1062 	nd_na->nd_na_flags_reserved = flags;
1063 	nd_na->nd_na_cksum = 0;
1064 	nd_na->nd_na_cksum =
1065 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1066 
1067 	if (send_sendso_input_hook != NULL) {
1068 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1069 		    sizeof(unsigned short), M_NOWAIT);
1070 		if (mtag == NULL)
1071 			goto bad;
1072 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1073 		m_tag_prepend(m, mtag);
1074 	}
1075 
1076 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1077 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1078 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1079 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1080 
1081 	return;
1082 
1083   bad:
1084 	m_freem(m);
1085 }
1086 
1087 #ifndef BURN_BRIDGES
1088 void
1089 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1090     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1091     struct sockaddr *sdl0)
1092 {
1093 
1094 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1095 	    RT_DEFAULT_FIB);
1096 }
1097 #endif
1098 
1099 caddr_t
1100 nd6_ifptomac(struct ifnet *ifp)
1101 {
1102 	switch (ifp->if_type) {
1103 	case IFT_ETHER:
1104 	case IFT_IEEE1394:
1105 	case IFT_L2VLAN:
1106 	case IFT_INFINIBAND:
1107 	case IFT_BRIDGE:
1108 		return IF_LLADDR(ifp);
1109 	default:
1110 		return NULL;
1111 	}
1112 }
1113 
1114 struct dadq {
1115 	TAILQ_ENTRY(dadq) dad_list;
1116 	struct ifaddr *dad_ifa;
1117 	int dad_count;		/* max NS to send */
1118 	int dad_ns_tcount;	/* # of trials to send NS */
1119 	int dad_ns_ocount;	/* NS sent so far */
1120 	int dad_ns_icount;
1121 	int dad_na_icount;
1122 	int dad_ns_lcount;	/* looped back NS */
1123 	int dad_loopbackprobe;	/* probing state for loopback detection */
1124 	struct callout dad_timer_ch;
1125 	struct vnet *dad_vnet;
1126 	u_int dad_refcnt;
1127 #define	ND_OPT_NONCE_LEN32 \
1128 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1129 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1130 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1131 };
1132 
1133 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1134 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1135 #define	V_dadq			VNET(dadq)
1136 #define	V_dad_rwlock		VNET(dad_rwlock)
1137 
1138 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1139 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1140 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1141 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1142 
1143 static void
1144 nd6_dad_add(struct dadq *dp)
1145 {
1146 
1147 	DADQ_WLOCK();
1148 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1149 	dp->dad_ondadq = true;
1150 	DADQ_WUNLOCK();
1151 }
1152 
1153 static void
1154 nd6_dad_del(struct dadq *dp)
1155 {
1156 
1157 	DADQ_WLOCK();
1158 	if (dp->dad_ondadq) {
1159 		/*
1160 		 * Remove dp from the dadq and release the dadq's
1161 		 * reference.
1162 		 */
1163 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1164 		dp->dad_ondadq = false;
1165 		DADQ_WUNLOCK();
1166 		nd6_dad_rele(dp);
1167 	} else
1168 		DADQ_WUNLOCK();
1169 }
1170 
1171 static struct dadq *
1172 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1173 {
1174 	struct dadq *dp;
1175 
1176 	DADQ_RLOCK();
1177 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1178 		if (dp->dad_ifa != ifa)
1179 			continue;
1180 		/*
1181 		 * Skip if the nonce matches the received one.
1182 		 * +2 in the length is required because of type and
1183 		 * length fields are included in a header.
1184 		 */
1185 		if (n != NULL &&
1186 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1187 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1188 		        ND_OPT_NONCE_LEN) == 0) {
1189 			dp->dad_ns_lcount++;
1190 			continue;
1191 		}
1192 		refcount_acquire(&dp->dad_refcnt);
1193 		break;
1194 	}
1195 	DADQ_RUNLOCK();
1196 
1197 	return (dp);
1198 }
1199 
1200 static void
1201 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1202 {
1203 
1204 	NET_EPOCH_ASSERT();
1205 
1206 	if (send_ns != 0)
1207 		nd6_dad_ns_output(dp);
1208 	callout_reset(&dp->dad_timer_ch, ticks,
1209 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1210 }
1211 
1212 static void
1213 nd6_dad_stoptimer(struct dadq *dp)
1214 {
1215 
1216 	callout_drain(&dp->dad_timer_ch);
1217 }
1218 
1219 static void
1220 nd6_dad_rele(struct dadq *dp)
1221 {
1222 
1223 	if (refcount_release(&dp->dad_refcnt)) {
1224 		ifa_free(dp->dad_ifa);
1225 		free(dp, M_IP6NDP);
1226 	}
1227 }
1228 
1229 void
1230 nd6_dad_init(void)
1231 {
1232 
1233 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1234 	TAILQ_INIT(&V_dadq);
1235 }
1236 
1237 /*
1238  * Start Duplicate Address Detection (DAD) for specified interface address.
1239  */
1240 void
1241 nd6_dad_start(struct ifaddr *ifa, int delay)
1242 {
1243 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1244 	struct dadq *dp;
1245 	char ip6buf[INET6_ADDRSTRLEN];
1246 	struct epoch_tracker et;
1247 
1248 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1249 	    ("starting DAD on non-tentative address %p", ifa));
1250 
1251 	/*
1252 	 * If we don't need DAD, don't do it.
1253 	 * There are several cases:
1254 	 * - DAD is disabled globally or on the interface
1255 	 * - the interface address is anycast
1256 	 */
1257 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1258 	    V_ip6_dad_count == 0 ||
1259 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1260 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1261 		return;
1262 	}
1263 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1264 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1265 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1266 		return;
1267 
1268 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1269 		/*
1270 		 * DAD is already in progress.  Let the existing entry
1271 		 * finish it.
1272 		 */
1273 		nd6_dad_rele(dp);
1274 		return;
1275 	}
1276 
1277 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1278 	if (dp == NULL) {
1279 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1280 			"%s(%s)\n",
1281 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1282 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1283 		return;
1284 	}
1285 	callout_init(&dp->dad_timer_ch, 0);
1286 #ifdef VIMAGE
1287 	dp->dad_vnet = curvnet;
1288 #endif
1289 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1290 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1291 
1292 	/*
1293 	 * Send NS packet for DAD, ip6_dad_count times.
1294 	 * Note that we must delay the first transmission, if this is the
1295 	 * first packet to be sent from the interface after interface
1296 	 * (re)initialization.
1297 	 */
1298 	dp->dad_ifa = ifa;
1299 	ifa_ref(dp->dad_ifa);
1300 	dp->dad_count = V_ip6_dad_count;
1301 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1302 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1303 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1304 
1305 	/* Add this to the dadq and add a reference for the dadq. */
1306 	refcount_init(&dp->dad_refcnt, 1);
1307 	nd6_dad_add(dp);
1308 	NET_EPOCH_ENTER(et);
1309 	nd6_dad_starttimer(dp, delay, 0);
1310 	NET_EPOCH_EXIT(et);
1311 }
1312 
1313 /*
1314  * terminate DAD unconditionally.  used for address removals.
1315  */
1316 void
1317 nd6_dad_stop(struct ifaddr *ifa)
1318 {
1319 	struct dadq *dp;
1320 
1321 	dp = nd6_dad_find(ifa, NULL);
1322 	if (!dp) {
1323 		/* DAD wasn't started yet */
1324 		return;
1325 	}
1326 
1327 	nd6_dad_stoptimer(dp);
1328 	nd6_dad_del(dp);
1329 
1330 	/* Release this function's reference, acquired by nd6_dad_find(). */
1331 	nd6_dad_rele(dp);
1332 }
1333 
1334 static void
1335 nd6_dad_timer(struct dadq *dp)
1336 {
1337 	CURVNET_SET(dp->dad_vnet);
1338 	struct ifaddr *ifa = dp->dad_ifa;
1339 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1340 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1341 	char ip6buf[INET6_ADDRSTRLEN];
1342 	struct epoch_tracker et;
1343 
1344 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1345 
1346 	NET_EPOCH_ENTER(et);
1347 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1348 		/* Do not need DAD for ifdisabled interface. */
1349 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1350 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1351 		goto err;
1352 	}
1353 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1354 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1355 			"%s(%s)\n",
1356 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1357 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1358 		goto err;
1359 	}
1360 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1361 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1362 			"%s(%s)\n",
1363 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1364 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1365 		goto err;
1366 	}
1367 
1368 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1369 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1370 	    (((ifp->if_flags & IFF_UP) == 0) ||
1371 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1372 		nd6log((LOG_INFO, "%s: could not run DAD "
1373 		    "because the interface was down or not running.\n",
1374 		    if_name(ifa->ifa_ifp)));
1375 		goto err;
1376 	}
1377 
1378 	/* Need more checks? */
1379 	if (dp->dad_ns_ocount < dp->dad_count) {
1380 		/*
1381 		 * We have more NS to go.  Send NS packet for DAD.
1382 		 */
1383 		nd6_dad_starttimer(dp,
1384 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1385 		goto done;
1386 	} else {
1387 		/*
1388 		 * We have transmitted sufficient number of DAD packets.
1389 		 * See what we've got.
1390 		 */
1391 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1392 			/* We've seen NS or NA, means DAD has failed. */
1393 			nd6_dad_duplicated(ifa, dp);
1394 		else if (V_dad_enhanced != 0 &&
1395 		    dp->dad_ns_lcount > 0 &&
1396 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1397 			/*
1398 			 * Sec. 4.1 in RFC 7527 requires transmission of
1399 			 * additional probes until the loopback condition
1400 			 * becomes clear when a looped back probe is detected.
1401 			 */
1402 			log(LOG_ERR, "%s: a looped back NS message is "
1403 			    "detected during DAD for %s.  "
1404 			    "Another DAD probes are being sent.\n",
1405 			    if_name(ifa->ifa_ifp),
1406 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1407 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1408 			/*
1409 			 * Send an NS immediately and increase dad_count by
1410 			 * V_nd6_mmaxtries - 1.
1411 			 */
1412 			dp->dad_count =
1413 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1414 			nd6_dad_starttimer(dp,
1415 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1416 			    1);
1417 			goto done;
1418 		} else {
1419 			/*
1420 			 * We are done with DAD.  No NA came, no NS came.
1421 			 * No duplicate address found.  Check IFDISABLED flag
1422 			 * again in case that it is changed between the
1423 			 * beginning of this function and here.
1424 			 */
1425 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1426 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1427 
1428 			nd6log((LOG_DEBUG,
1429 			    "%s: DAD complete for %s - no duplicates found\n",
1430 			    if_name(ifa->ifa_ifp),
1431 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1432 			if (dp->dad_ns_lcount > 0)
1433 				log(LOG_ERR, "%s: DAD completed while "
1434 				    "a looped back NS message is detected "
1435 				    "during DAD for %s.\n",
1436 				    if_name(ifa->ifa_ifp),
1437 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1438 		}
1439 	}
1440 err:
1441 	nd6_dad_del(dp);
1442 done:
1443 	NET_EPOCH_EXIT(et);
1444 	CURVNET_RESTORE();
1445 }
1446 
1447 static void
1448 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1449 {
1450 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1451 	struct ifnet *ifp;
1452 	char ip6buf[INET6_ADDRSTRLEN];
1453 
1454 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1455 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1456 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1457 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1458 	    dp->dad_na_icount);
1459 
1460 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1461 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1462 
1463 	ifp = ifa->ifa_ifp;
1464 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1465 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1466 	log(LOG_ERR, "%s: manual intervention required\n",
1467 	    if_name(ifp));
1468 
1469 	/*
1470 	 * If the address is a link-local address formed from an interface
1471 	 * identifier based on the hardware address which is supposed to be
1472 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1473 	 * operation on the interface SHOULD be disabled.
1474 	 * [RFC 4862, Section 5.4.5]
1475 	 */
1476 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1477 		struct in6_addr in6;
1478 
1479 		/*
1480 		 * To avoid over-reaction, we only apply this logic when we are
1481 		 * very sure that hardware addresses are supposed to be unique.
1482 		 */
1483 		switch (ifp->if_type) {
1484 		case IFT_ETHER:
1485 		case IFT_ATM:
1486 		case IFT_IEEE1394:
1487 		case IFT_INFINIBAND:
1488 			in6 = ia->ia_addr.sin6_addr;
1489 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1490 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1491 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1492 				log(LOG_ERR, "%s: possible hardware address "
1493 				    "duplication detected, disable IPv6\n",
1494 				    if_name(ifp));
1495 			}
1496 			break;
1497 		}
1498 	}
1499 }
1500 
1501 static void
1502 nd6_dad_ns_output(struct dadq *dp)
1503 {
1504 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1505 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1506 	int i;
1507 
1508 	dp->dad_ns_tcount++;
1509 	if ((ifp->if_flags & IFF_UP) == 0) {
1510 		return;
1511 	}
1512 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1513 		return;
1514 	}
1515 
1516 	dp->dad_ns_ocount++;
1517 	if (V_dad_enhanced != 0) {
1518 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1519 			dp->dad_nonce[i] = arc4random();
1520 		/*
1521 		 * XXXHRS: Note that in the case that
1522 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1523 		 * different nonces can be looped back in an unexpected
1524 		 * order.  The current implementation recognizes only
1525 		 * the latest nonce on the sender side.  Practically it
1526 		 * should work well in almost all cases.
1527 		 */
1528 	}
1529 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1530 	    (uint8_t *)&dp->dad_nonce[0]);
1531 }
1532 
1533 static void
1534 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1535 {
1536 	struct dadq *dp;
1537 
1538 	if (ifa == NULL)
1539 		panic("ifa == NULL in nd6_dad_ns_input");
1540 
1541 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1542 	if (V_dad_enhanced == 0)
1543 		ndopt_nonce = NULL;
1544 	dp = nd6_dad_find(ifa, ndopt_nonce);
1545 	if (dp == NULL)
1546 		return;
1547 
1548 	dp->dad_ns_icount++;
1549 	nd6_dad_rele(dp);
1550 }
1551 
1552 static void
1553 nd6_dad_na_input(struct ifaddr *ifa)
1554 {
1555 	struct dadq *dp;
1556 
1557 	if (ifa == NULL)
1558 		panic("ifa == NULL in nd6_dad_na_input");
1559 
1560 	dp = nd6_dad_find(ifa, NULL);
1561 	if (dp != NULL) {
1562 		dp->dad_na_icount++;
1563 		nd6_dad_rele(dp);
1564 	}
1565 }
1566