xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 71e8eac4fd89dedbcb4130379add151d3ed942fa)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #include <netinet6/nd6.h>
77 #include <netinet/icmp6.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet6/send.h>
80 
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 struct dadq;
84 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
85 static void nd6_dad_add(struct dadq *dp);
86 static void nd6_dad_del(struct dadq *dp);
87 static void nd6_dad_rele(struct dadq *);
88 static void nd6_dad_starttimer(struct dadq *, int, int);
89 static void nd6_dad_stoptimer(struct dadq *);
90 static void nd6_dad_timer(struct dadq *);
91 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
92 static void nd6_dad_ns_output(struct dadq *);
93 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
94 static void nd6_dad_na_input(struct ifaddr *);
95 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
96     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
97 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
98     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
99 
100 static VNET_DEFINE(int, dad_enhanced) = 1;
101 #define	V_dad_enhanced			VNET(dad_enhanced)
102 
103 SYSCTL_DECL(_net_inet6_ip6);
104 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
105     &VNET_NAME(dad_enhanced), 0,
106     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
107 
108 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
109 						   transmit DAD packet */
110 #define	V_dad_maxtry			VNET(dad_maxtry)
111 
112 /*
113  * Input a Neighbor Solicitation Message.
114  *
115  * Based on RFC 2461
116  * Based on RFC 2462 (duplicate address detection)
117  */
118 void
119 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_neighbor_solicit *nd_ns;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	struct in6_addr daddr6 = ip6->ip6_dst;
126 	struct in6_addr taddr6;
127 	struct in6_addr myaddr6;
128 	char *lladdr = NULL;
129 	struct ifaddr *ifa = NULL;
130 	int lladdrlen = 0;
131 	int anycast = 0, proxy = 0, tentative = 0;
132 	int tlladdr;
133 	int rflag;
134 	union nd_opts ndopts;
135 	struct sockaddr_dl proxydl;
136 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
137 
138 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
139 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
140 		rflag = 0;
141 #ifndef PULLDOWN_TEST
142 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
143 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
144 #else
145 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
146 	if (nd_ns == NULL) {
147 		ICMP6STAT_INC(icp6s_tooshort);
148 		return;
149 	}
150 #endif
151 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
152 	taddr6 = nd_ns->nd_ns_target;
153 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
154 		goto bad;
155 
156 	if (ip6->ip6_hlim != 255) {
157 		nd6log((LOG_ERR,
158 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
159 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
160 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
161 		goto bad;
162 	}
163 
164 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
165 		/* dst has to be a solicited node multicast address. */
166 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
167 		    /* don't check ifindex portion */
168 		    daddr6.s6_addr32[1] == 0 &&
169 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
170 		    daddr6.s6_addr8[12] == 0xff) {
171 			; /* good */
172 		} else {
173 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
174 			    "(wrong ip6 dst)\n"));
175 			goto bad;
176 		}
177 	} else if (!V_nd6_onlink_ns_rfc4861) {
178 		struct sockaddr_in6 src_sa6;
179 
180 		/*
181 		 * According to recent IETF discussions, it is not a good idea
182 		 * to accept a NS from an address which would not be deemed
183 		 * to be a neighbor otherwise.  This point is expected to be
184 		 * clarified in future revisions of the specification.
185 		 */
186 		bzero(&src_sa6, sizeof(src_sa6));
187 		src_sa6.sin6_family = AF_INET6;
188 		src_sa6.sin6_len = sizeof(src_sa6);
189 		src_sa6.sin6_addr = saddr6;
190 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
191 			nd6log((LOG_INFO, "nd6_ns_input: "
192 				"NS packet from non-neighbor\n"));
193 			goto bad;
194 		}
195 	}
196 
197 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
198 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
199 		goto bad;
200 	}
201 
202 	icmp6len -= sizeof(*nd_ns);
203 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
204 	if (nd6_options(&ndopts) < 0) {
205 		nd6log((LOG_INFO,
206 		    "nd6_ns_input: invalid ND option, ignored\n"));
207 		/* nd6_options have incremented stats */
208 		goto freeit;
209 	}
210 
211 	if (ndopts.nd_opts_src_lladdr) {
212 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
213 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
214 	}
215 
216 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
217 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
218 		    "(link-layer address option)\n"));
219 		goto bad;
220 	}
221 
222 	/*
223 	 * Attaching target link-layer address to the NA?
224 	 * (RFC 2461 7.2.4)
225 	 *
226 	 * NS IP dst is unicast/anycast			MUST NOT add
227 	 * NS IP dst is solicited-node multicast	MUST add
228 	 *
229 	 * In implementation, we add target link-layer address by default.
230 	 * We do not add one in MUST NOT cases.
231 	 */
232 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
233 		tlladdr = 0;
234 	else
235 		tlladdr = 1;
236 
237 	/*
238 	 * Target address (taddr6) must be either:
239 	 * (1) Valid unicast/anycast address for my receiving interface,
240 	 * (2) Unicast address for which I'm offering proxy service, or
241 	 * (3) "tentative" address on which DAD is being performed.
242 	 */
243 	/* (1) and (3) check. */
244 	if (ifp->if_carp)
245 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
246 	else
247 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
248 
249 	/* (2) check. */
250 	if (ifa == NULL) {
251 		struct route_in6 ro;
252 		int need_proxy;
253 
254 		bzero(&ro, sizeof(ro));
255 		ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
256 		ro.ro_dst.sin6_family = AF_INET6;
257 		ro.ro_dst.sin6_addr = taddr6;
258 
259 		/* Always use the default FIB. */
260 #ifdef RADIX_MPATH
261 		rtalloc_mpath_fib((struct route *)&ro, ntohl(taddr6.s6_addr32[3]),
262 		    RT_DEFAULT_FIB);
263 #else
264 		in6_rtalloc(&ro, RT_DEFAULT_FIB);
265 #endif
266 		need_proxy = (ro.ro_rt &&
267 		    (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
268 		    ro.ro_rt->rt_gateway->sa_family == AF_LINK);
269 		if (ro.ro_rt != NULL) {
270 			if (need_proxy)
271 				proxydl = *SDL(ro.ro_rt->rt_gateway);
272 			RTFREE(ro.ro_rt);
273 		}
274 		if (need_proxy) {
275 			/*
276 			 * proxy NDP for single entry
277 			 */
278 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
279 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
280 			if (ifa)
281 				proxy = 1;
282 		}
283 	}
284 	if (ifa == NULL) {
285 		/*
286 		 * We've got an NS packet, and we don't have that adddress
287 		 * assigned for us.  We MUST silently ignore it.
288 		 * See RFC2461 7.2.3.
289 		 */
290 		goto freeit;
291 	}
292 	myaddr6 = *IFA_IN6(ifa);
293 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
294 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
295 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
296 		goto freeit;
297 
298 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
299 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
300 		    "(if %d, NS packet %d)\n",
301 		    ip6_sprintf(ip6bufs, &taddr6),
302 		    ifp->if_addrlen, lladdrlen - 2));
303 		goto bad;
304 	}
305 
306 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
307 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
308 		    ip6_sprintf(ip6bufs, &saddr6)));
309 		goto freeit;
310 	}
311 
312 	/*
313 	 * We have neighbor solicitation packet, with target address equals to
314 	 * one of my tentative address.
315 	 *
316 	 * src addr	how to process?
317 	 * ---		---
318 	 * multicast	of course, invalid (rejected in ip6_input)
319 	 * unicast	somebody is doing address resolution -> ignore
320 	 * unspec	dup address detection
321 	 *
322 	 * The processing is defined in RFC 2462.
323 	 */
324 	if (tentative) {
325 		/*
326 		 * If source address is unspecified address, it is for
327 		 * duplicate address detection.
328 		 *
329 		 * If not, the packet is for addess resolution;
330 		 * silently ignore it.
331 		 */
332 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
333 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
334 
335 		goto freeit;
336 	}
337 
338 	/*
339 	 * If the source address is unspecified address, entries must not
340 	 * be created or updated.
341 	 * It looks that sender is performing DAD.  Output NA toward
342 	 * all-node multicast address, to tell the sender that I'm using
343 	 * the address.
344 	 * S bit ("solicited") must be zero.
345 	 */
346 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
347 		struct in6_addr in6_all;
348 
349 		in6_all = in6addr_linklocal_allnodes;
350 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
351 			goto bad;
352 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
353 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
354 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
355 		    M_GETFIB(m));
356 		goto freeit;
357 	}
358 
359 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
360 	    ND_NEIGHBOR_SOLICIT, 0);
361 
362 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
363 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
364 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
365 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
366  freeit:
367 	if (ifa != NULL)
368 		ifa_free(ifa);
369 	m_freem(m);
370 	return;
371 
372  bad:
373 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
374 		ip6_sprintf(ip6bufs, &saddr6)));
375 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
376 		ip6_sprintf(ip6bufs, &daddr6)));
377 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
378 		ip6_sprintf(ip6bufs, &taddr6)));
379 	ICMP6STAT_INC(icp6s_badns);
380 	if (ifa != NULL)
381 		ifa_free(ifa);
382 	m_freem(m);
383 }
384 
385 /*
386  * Output a Neighbor Solicitation Message. Caller specifies:
387  *	- ICMP6 header source IP6 address
388  *	- ND6 header target IP6 address
389  *	- ND6 header source datalink address
390  *
391  * Based on RFC 2461
392  * Based on RFC 2462 (duplicate address detection)
393  *
394  *    ln - for source address determination
395  * nonce - If non-NULL, NS is used for duplicate address detection and
396  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
397  */
398 static void
399 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
400     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
401     uint8_t *nonce, u_int fibnum)
402 {
403 	struct mbuf *m;
404 	struct m_tag *mtag;
405 	struct ip6_hdr *ip6;
406 	struct nd_neighbor_solicit *nd_ns;
407 	struct ip6_moptions im6o;
408 	int icmp6len;
409 	int maxlen;
410 	caddr_t mac;
411 	struct route_in6 ro;
412 
413 	if (IN6_IS_ADDR_MULTICAST(taddr6))
414 		return;
415 
416 	/* estimate the size of message */
417 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
418 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
419 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
420 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
421 	    __func__, max_linkhdr, maxlen, MCLBYTES));
422 
423 	if (max_linkhdr + maxlen > MHLEN)
424 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
425 	else
426 		m = m_gethdr(M_NOWAIT, MT_DATA);
427 	if (m == NULL)
428 		return;
429 	M_SETFIB(m, fibnum);
430 
431 	bzero(&ro, sizeof(ro));
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 sockaddr_in6 dst_sa;
490 			struct in6_addr src_in;
491 			struct ifnet *oifp;
492 
493 			bzero(&dst_sa, sizeof(dst_sa));
494 			dst_sa.sin6_family = AF_INET6;
495 			dst_sa.sin6_len = sizeof(dst_sa);
496 			dst_sa.sin6_addr = ip6->ip6_dst;
497 
498 			oifp = ifp;
499 			error = in6_selectsrc(&dst_sa, NULL,
500 			    NULL, &ro, NULL, &oifp, &src_in);
501 			if (error) {
502 				char ip6buf[INET6_ADDRSTRLEN];
503 				nd6log((LOG_DEBUG, "%s: source can't be "
504 				    "determined: dst=%s, error=%d\n", __func__,
505 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
506 				    error));
507 				goto bad;
508 			}
509 			ip6->ip6_src = src_in;
510 		}
511 	} else {
512 		/*
513 		 * Source address for DAD packet must always be IPv6
514 		 * unspecified address. (0::0)
515 		 * We actually don't have to 0-clear the address (we did it
516 		 * above), but we do so here explicitly to make the intention
517 		 * clearer.
518 		 */
519 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
520 	}
521 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
522 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
523 	nd_ns->nd_ns_code = 0;
524 	nd_ns->nd_ns_reserved = 0;
525 	nd_ns->nd_ns_target = *taddr6;
526 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
527 
528 	/*
529 	 * Add source link-layer address option.
530 	 *
531 	 *				spec		implementation
532 	 *				---		---
533 	 * DAD packet			MUST NOT	do not add the option
534 	 * there's no link layer address:
535 	 *				impossible	do not add the option
536 	 * there's link layer address:
537 	 *	Multicast NS		MUST add one	add the option
538 	 *	Unicast NS		SHOULD add one	add the option
539 	 */
540 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
541 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
542 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
543 		/* 8 byte alignments... */
544 		optlen = (optlen + 7) & ~7;
545 
546 		m->m_pkthdr.len += optlen;
547 		m->m_len += optlen;
548 		icmp6len += optlen;
549 		bzero((caddr_t)nd_opt, optlen);
550 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
551 		nd_opt->nd_opt_len = optlen >> 3;
552 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
553 	}
554 	/*
555 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
556 	 * This behavior is documented as Enhanced Duplicate Address
557 	 * Detection in RFC 7527.
558 	 * net.inet6.ip6.dad_enhanced=0 disables this.
559 	 */
560 	if (V_dad_enhanced != 0 && nonce != NULL) {
561 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
562 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
563 		/* 8-byte alignment is required. */
564 		optlen = (optlen + 7) & ~7;
565 
566 		m->m_pkthdr.len += optlen;
567 		m->m_len += optlen;
568 		icmp6len += optlen;
569 		bzero((caddr_t)nd_opt, optlen);
570 		nd_opt->nd_opt_type = ND_OPT_NONCE;
571 		nd_opt->nd_opt_len = optlen >> 3;
572 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
573 	}
574 	ip6->ip6_plen = htons((u_short)icmp6len);
575 	nd_ns->nd_ns_cksum = 0;
576 	nd_ns->nd_ns_cksum =
577 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
578 
579 	if (send_sendso_input_hook != NULL) {
580 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
581 			sizeof(unsigned short), M_NOWAIT);
582 		if (mtag == NULL)
583 			goto bad;
584 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
585 		m_tag_prepend(m, mtag);
586 	}
587 
588 	ip6_output(m, NULL, &ro, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
589 	    &im6o, NULL, NULL);
590 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
591 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
592 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
593 
594 	/* We don't cache this route. */
595 	RO_RTFREE(&ro);
596 
597 	return;
598 
599   bad:
600 	if (ro.ro_rt) {
601 		RTFREE(ro.ro_rt);
602 	}
603 	m_freem(m);
604 	return;
605 }
606 
607 #ifndef BURN_BRIDGES
608 void
609 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
610     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
611 {
612 
613 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
614 }
615 #endif
616 /*
617  * Neighbor advertisement input handling.
618  *
619  * Based on RFC 2461
620  * Based on RFC 2462 (duplicate address detection)
621  *
622  * the following items are not implemented yet:
623  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
624  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
625  */
626 void
627 nd6_na_input(struct mbuf *m, int off, int icmp6len)
628 {
629 	struct ifnet *ifp = m->m_pkthdr.rcvif;
630 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
631 	struct nd_neighbor_advert *nd_na;
632 	struct in6_addr daddr6 = ip6->ip6_dst;
633 	struct in6_addr taddr6;
634 	int flags;
635 	int is_router;
636 	int is_solicited;
637 	int is_override;
638 	char *lladdr = NULL;
639 	int lladdrlen = 0;
640 	int checklink = 0;
641 	struct ifaddr *ifa;
642 	struct llentry *ln = NULL;
643 	union nd_opts ndopts;
644 	struct mbuf *chain = NULL;
645 	struct sockaddr_in6 sin6;
646 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
647 
648 	if (ip6->ip6_hlim != 255) {
649 		nd6log((LOG_ERR,
650 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
651 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
652 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
653 		goto bad;
654 	}
655 
656 #ifndef PULLDOWN_TEST
657 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
658 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
659 #else
660 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
661 	if (nd_na == NULL) {
662 		ICMP6STAT_INC(icp6s_tooshort);
663 		return;
664 	}
665 #endif
666 
667 	flags = nd_na->nd_na_flags_reserved;
668 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
669 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
670 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
671 	memset(&sin6, 0, sizeof(sin6));
672 
673 	taddr6 = nd_na->nd_na_target;
674 	if (in6_setscope(&taddr6, ifp, NULL))
675 		goto bad;	/* XXX: impossible */
676 
677 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
678 		nd6log((LOG_ERR,
679 		    "nd6_na_input: invalid target address %s\n",
680 		    ip6_sprintf(ip6bufs, &taddr6)));
681 		goto bad;
682 	}
683 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
684 		if (is_solicited) {
685 			nd6log((LOG_ERR,
686 			    "nd6_na_input: a solicited adv is multicasted\n"));
687 			goto bad;
688 		}
689 
690 	icmp6len -= sizeof(*nd_na);
691 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
692 	if (nd6_options(&ndopts) < 0) {
693 		nd6log((LOG_INFO,
694 		    "nd6_na_input: invalid ND option, ignored\n"));
695 		/* nd6_options have incremented stats */
696 		goto freeit;
697 	}
698 
699 	if (ndopts.nd_opts_tgt_lladdr) {
700 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
701 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
702 	}
703 
704 	/*
705 	 * This effectively disables the DAD check on a non-master CARP
706 	 * address.
707 	 */
708 	if (ifp->if_carp)
709 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
710 	else
711 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
712 
713 	/*
714 	 * Target address matches one of my interface address.
715 	 *
716 	 * If my address is tentative, this means that there's somebody
717 	 * already using the same address as mine.  This indicates DAD failure.
718 	 * This is defined in RFC 2462.
719 	 *
720 	 * Otherwise, process as defined in RFC 2461.
721 	 */
722 	if (ifa
723 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
724 		nd6_dad_na_input(ifa);
725 		ifa_free(ifa);
726 		goto freeit;
727 	}
728 
729 	/* Just for safety, maybe unnecessary. */
730 	if (ifa) {
731 		ifa_free(ifa);
732 		log(LOG_ERR,
733 		    "nd6_na_input: duplicate IP6 address %s\n",
734 		    ip6_sprintf(ip6bufs, &taddr6));
735 		goto freeit;
736 	}
737 
738 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
739 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
740 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
741 		    ifp->if_addrlen, lladdrlen - 2));
742 		goto bad;
743 	}
744 
745 	/*
746 	 * If no neighbor cache entry is found, NA SHOULD silently be
747 	 * discarded.
748 	 */
749 	IF_AFDATA_RLOCK(ifp);
750 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
751 	IF_AFDATA_RUNLOCK(ifp);
752 	if (ln == NULL) {
753 		goto freeit;
754 	}
755 
756 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
757 		/*
758 		 * If the link-layer has address, and no lladdr option came,
759 		 * discard the packet.
760 		 */
761 		if (ifp->if_addrlen && lladdr == NULL) {
762 			goto freeit;
763 		}
764 
765 		/*
766 		 * Record link-layer address, and update the state.
767 		 */
768 		if (lltable_try_set_entry_addr(ifp, ln, lladdr) == 0) {
769 			ln = NULL;
770 			goto freeit;
771 		}
772 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
773 		if (is_solicited)
774 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
775 		else
776 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
777 		if ((ln->ln_router = is_router) != 0) {
778 			/*
779 			 * This means a router's state has changed from
780 			 * non-reachable to probably reachable, and might
781 			 * affect the status of associated prefixes..
782 			 */
783 			checklink = 1;
784 		}
785 	} else {
786 		int llchange;
787 
788 		/*
789 		 * Check if the link-layer address has changed or not.
790 		 */
791 		if (lladdr == NULL)
792 			llchange = 0;
793 		else {
794 			if (ln->la_flags & LLE_VALID) {
795 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
796 					llchange = 1;
797 				else
798 					llchange = 0;
799 			} else
800 				llchange = 1;
801 		}
802 
803 		/*
804 		 * This is VERY complex.  Look at it with care.
805 		 *
806 		 * override solicit lladdr llchange	action
807 		 *					(L: record lladdr)
808 		 *
809 		 *	0	0	n	--	(2c)
810 		 *	0	0	y	n	(2b) L
811 		 *	0	0	y	y	(1)    REACHABLE->STALE
812 		 *	0	1	n	--	(2c)   *->REACHABLE
813 		 *	0	1	y	n	(2b) L *->REACHABLE
814 		 *	0	1	y	y	(1)    REACHABLE->STALE
815 		 *	1	0	n	--	(2a)
816 		 *	1	0	y	n	(2a) L
817 		 *	1	0	y	y	(2a) L *->STALE
818 		 *	1	1	n	--	(2a)   *->REACHABLE
819 		 *	1	1	y	n	(2a) L *->REACHABLE
820 		 *	1	1	y	y	(2a) L *->REACHABLE
821 		 */
822 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
823 			/*
824 			 * If state is REACHABLE, make it STALE.
825 			 * no other updates should be done.
826 			 */
827 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
828 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
829 			goto freeit;
830 		} else if (is_override				   /* (2a) */
831 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
832 			|| lladdr == NULL) {			   /* (2c) */
833 			/*
834 			 * Update link-local address, if any.
835 			 */
836 			if (lladdr != NULL) {
837 				int ret;
838 				ret = lltable_try_set_entry_addr(ifp, ln,lladdr);
839 				if (ret == 0) {
840 					ln = NULL;
841 					goto freeit;
842 				}
843 				EVENTHANDLER_INVOKE(lle_event, ln,
844 				    LLENTRY_RESOLVED);
845 			}
846 
847 			/*
848 			 * If solicited, make the state REACHABLE.
849 			 * If not solicited and the link-layer address was
850 			 * changed, make it STALE.
851 			 */
852 			if (is_solicited)
853 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
854 			else {
855 				if (lladdr != NULL && llchange)
856 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
857 			}
858 		}
859 
860 		if (ln->ln_router && !is_router) {
861 			/*
862 			 * The peer dropped the router flag.
863 			 * Remove the sender from the Default Router List and
864 			 * update the Destination Cache entries.
865 			 */
866 			struct nd_defrouter *dr;
867 			struct in6_addr *in6;
868 
869 			in6 = &ln->r_l3addr.addr6;
870 
871 			/*
872 			 * Lock to protect the default router list.
873 			 * XXX: this might be unnecessary, since this function
874 			 * is only called under the network software interrupt
875 			 * context.  However, we keep it just for safety.
876 			 */
877 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
878 			if (dr)
879 				defrtrlist_del(dr);
880 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
881 			    ND6_IFF_ACCEPT_RTADV) {
882 				/*
883 				 * Even if the neighbor is not in the default
884 				 * router list, the neighbor may be used
885 				 * as a next hop for some destinations
886 				 * (e.g. redirect case). So we must
887 				 * call rt6_flush explicitly.
888 				 */
889 				rt6_flush(&ip6->ip6_src, ifp);
890 			}
891 		}
892 		ln->ln_router = is_router;
893 	}
894         /* XXX - QL
895 	 *  Does this matter?
896 	 *  rt->rt_flags &= ~RTF_REJECT;
897 	 */
898 	ln->la_asked = 0;
899 	if (ln->la_hold != NULL)
900 		nd6_grab_holdchain(ln, &chain, &sin6);
901  freeit:
902 	if (ln != NULL)
903 		LLE_WUNLOCK(ln);
904 
905 	if (chain != NULL)
906 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
907 
908 	if (checklink)
909 		pfxlist_onlink_check();
910 
911 	m_freem(m);
912 	return;
913 
914  bad:
915 	if (ln != NULL)
916 		LLE_WUNLOCK(ln);
917 
918 	ICMP6STAT_INC(icp6s_badna);
919 	m_freem(m);
920 }
921 
922 /*
923  * Neighbor advertisement output handling.
924  *
925  * Based on RFC 2461
926  *
927  * the following items are not implemented yet:
928  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
929  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
930  *
931  * tlladdr - 1 if include target link-layer address
932  * sdl0 - sockaddr_dl (= proxy NA) or NULL
933  */
934 static void
935 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
936     const struct in6_addr *taddr6, u_long flags, int tlladdr,
937     struct sockaddr *sdl0, u_int fibnum)
938 {
939 	struct mbuf *m;
940 	struct m_tag *mtag;
941 	struct ifnet *oifp;
942 	struct ip6_hdr *ip6;
943 	struct nd_neighbor_advert *nd_na;
944 	struct ip6_moptions im6o;
945 	struct in6_addr src, daddr6;
946 	struct sockaddr_in6 dst_sa;
947 	int icmp6len, maxlen, error;
948 	caddr_t mac = NULL;
949 	struct route_in6 ro;
950 
951 	bzero(&ro, sizeof(ro));
952 
953 	daddr6 = *daddr6_0;	/* make a local copy for modification */
954 
955 	/* estimate the size of message */
956 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
957 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
958 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
959 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
960 	    __func__, max_linkhdr, maxlen, MCLBYTES));
961 
962 	if (max_linkhdr + maxlen > MHLEN)
963 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
964 	else
965 		m = m_gethdr(M_NOWAIT, MT_DATA);
966 	if (m == NULL)
967 		return;
968 	M_SETFIB(m, fibnum);
969 
970 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
971 		m->m_flags |= M_MCAST;
972 		im6o.im6o_multicast_ifp = ifp;
973 		im6o.im6o_multicast_hlim = 255;
974 		im6o.im6o_multicast_loop = 0;
975 	}
976 
977 	icmp6len = sizeof(*nd_na);
978 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
979 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
980 
981 	/* fill neighbor advertisement packet */
982 	ip6 = mtod(m, struct ip6_hdr *);
983 	ip6->ip6_flow = 0;
984 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
985 	ip6->ip6_vfc |= IPV6_VERSION;
986 	ip6->ip6_nxt = IPPROTO_ICMPV6;
987 	ip6->ip6_hlim = 255;
988 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
989 		/* reply to DAD */
990 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
991 		daddr6.s6_addr16[1] = 0;
992 		daddr6.s6_addr32[1] = 0;
993 		daddr6.s6_addr32[2] = 0;
994 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
995 		if (in6_setscope(&daddr6, ifp, NULL))
996 			goto bad;
997 
998 		flags &= ~ND_NA_FLAG_SOLICITED;
999 	}
1000 	ip6->ip6_dst = daddr6;
1001 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1002 	dst_sa.sin6_family = AF_INET6;
1003 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1004 	dst_sa.sin6_addr = daddr6;
1005 
1006 	/*
1007 	 * Select a source whose scope is the same as that of the dest.
1008 	 */
1009 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1010 	oifp = ifp;
1011 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1012 	if (error) {
1013 		char ip6buf[INET6_ADDRSTRLEN];
1014 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1015 		    "determined: dst=%s, error=%d\n",
1016 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1017 		goto bad;
1018 	}
1019 	ip6->ip6_src = src;
1020 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1021 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1022 	nd_na->nd_na_code = 0;
1023 	nd_na->nd_na_target = *taddr6;
1024 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1025 
1026 	/*
1027 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1028 	 * see nd6_ns_input() for details.
1029 	 * Basically, if NS packet is sent to unicast/anycast addr,
1030 	 * target lladdr option SHOULD NOT be included.
1031 	 */
1032 	if (tlladdr) {
1033 		/*
1034 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1035 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1036 		 * my address) use lladdr configured for the interface.
1037 		 */
1038 		if (sdl0 == NULL) {
1039 			if (ifp->if_carp)
1040 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1041 			if (mac == NULL)
1042 				mac = nd6_ifptomac(ifp);
1043 		} else if (sdl0->sa_family == AF_LINK) {
1044 			struct sockaddr_dl *sdl;
1045 			sdl = (struct sockaddr_dl *)sdl0;
1046 			if (sdl->sdl_alen == ifp->if_addrlen)
1047 				mac = LLADDR(sdl);
1048 		}
1049 	}
1050 	if (tlladdr && mac) {
1051 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1052 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1053 
1054 		/* roundup to 8 bytes alignment! */
1055 		optlen = (optlen + 7) & ~7;
1056 
1057 		m->m_pkthdr.len += optlen;
1058 		m->m_len += optlen;
1059 		icmp6len += optlen;
1060 		bzero((caddr_t)nd_opt, optlen);
1061 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1062 		nd_opt->nd_opt_len = optlen >> 3;
1063 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1064 	} else
1065 		flags &= ~ND_NA_FLAG_OVERRIDE;
1066 
1067 	ip6->ip6_plen = htons((u_short)icmp6len);
1068 	nd_na->nd_na_flags_reserved = flags;
1069 	nd_na->nd_na_cksum = 0;
1070 	nd_na->nd_na_cksum =
1071 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1072 
1073 	if (send_sendso_input_hook != NULL) {
1074 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1075 		    sizeof(unsigned short), M_NOWAIT);
1076 		if (mtag == NULL)
1077 			goto bad;
1078 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1079 		m_tag_prepend(m, mtag);
1080 	}
1081 
1082 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1083 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1084 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1085 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1086 
1087 	/* We don't cache this route. */
1088 	RO_RTFREE(&ro);
1089 
1090 	return;
1091 
1092   bad:
1093 	if (ro.ro_rt) {
1094 		RTFREE(ro.ro_rt);
1095 	}
1096 	m_freem(m);
1097 	return;
1098 }
1099 
1100 #ifndef BURN_BRIDGES
1101 void
1102 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1103     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1104     struct sockaddr *sdl0)
1105 {
1106 
1107 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1108 	    RT_DEFAULT_FIB);
1109 }
1110 #endif
1111 
1112 caddr_t
1113 nd6_ifptomac(struct ifnet *ifp)
1114 {
1115 	switch (ifp->if_type) {
1116 	case IFT_ARCNET:
1117 	case IFT_ETHER:
1118 	case IFT_FDDI:
1119 	case IFT_IEEE1394:
1120 	case IFT_L2VLAN:
1121 	case IFT_IEEE80211:
1122 	case IFT_INFINIBAND:
1123 	case IFT_BRIDGE:
1124 	case IFT_ISO88025:
1125 		return IF_LLADDR(ifp);
1126 	default:
1127 		return NULL;
1128 	}
1129 }
1130 
1131 struct dadq {
1132 	TAILQ_ENTRY(dadq) dad_list;
1133 	struct ifaddr *dad_ifa;
1134 	int dad_count;		/* max NS to send */
1135 	int dad_ns_tcount;	/* # of trials to send NS */
1136 	int dad_ns_ocount;	/* NS sent so far */
1137 	int dad_ns_icount;
1138 	int dad_na_icount;
1139 	int dad_ns_lcount;	/* looped back NS */
1140 	int dad_loopbackprobe;	/* probing state for loopback detection */
1141 	struct callout dad_timer_ch;
1142 	struct vnet *dad_vnet;
1143 	u_int dad_refcnt;
1144 #define	ND_OPT_NONCE_LEN32 \
1145 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1146 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1147 };
1148 
1149 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1150 static VNET_DEFINE(struct rwlock, dad_rwlock);
1151 #define	V_dadq			VNET(dadq)
1152 #define	V_dad_rwlock		VNET(dad_rwlock)
1153 
1154 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1155 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1156 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1157 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1158 
1159 static void
1160 nd6_dad_add(struct dadq *dp)
1161 {
1162 
1163 	DADQ_WLOCK();
1164 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1165 	DADQ_WUNLOCK();
1166 }
1167 
1168 static void
1169 nd6_dad_del(struct dadq *dp)
1170 {
1171 
1172 	DADQ_WLOCK();
1173 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1174 	DADQ_WUNLOCK();
1175 	nd6_dad_rele(dp);
1176 }
1177 
1178 static struct dadq *
1179 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1180 {
1181 	struct dadq *dp;
1182 
1183 	DADQ_RLOCK();
1184 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1185 		if (dp->dad_ifa != ifa)
1186 			continue;
1187 		/*
1188 		 * Skip if the nonce matches the received one.
1189 		 * +2 in the length is required because of type and
1190 		 * length fields are included in a header.
1191 		 */
1192 		if (n != NULL &&
1193 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1194 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1195 		        ND_OPT_NONCE_LEN) == 0) {
1196 			dp->dad_ns_lcount++;
1197 			continue;
1198 		}
1199 		refcount_acquire(&dp->dad_refcnt);
1200 		break;
1201 	}
1202 	DADQ_RUNLOCK();
1203 
1204 	return (dp);
1205 }
1206 
1207 static void
1208 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1209 {
1210 
1211 	if (send_ns != 0)
1212 		nd6_dad_ns_output(dp);
1213 	callout_reset(&dp->dad_timer_ch, ticks,
1214 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1215 }
1216 
1217 static void
1218 nd6_dad_stoptimer(struct dadq *dp)
1219 {
1220 
1221 	callout_drain(&dp->dad_timer_ch);
1222 }
1223 
1224 static void
1225 nd6_dad_rele(struct dadq *dp)
1226 {
1227 
1228 	if (refcount_release(&dp->dad_refcnt)) {
1229 		ifa_free(dp->dad_ifa);
1230 		free(dp, M_IP6NDP);
1231 	}
1232 }
1233 
1234 void
1235 nd6_dad_init(void)
1236 {
1237 
1238 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1239 	TAILQ_INIT(&V_dadq);
1240 }
1241 
1242 /*
1243  * Start Duplicate Address Detection (DAD) for specified interface address.
1244  */
1245 void
1246 nd6_dad_start(struct ifaddr *ifa, int delay)
1247 {
1248 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1249 	struct dadq *dp;
1250 	char ip6buf[INET6_ADDRSTRLEN];
1251 	int send_ns;
1252 
1253 	/*
1254 	 * If we don't need DAD, don't do it.
1255 	 * There are several cases:
1256 	 * - DAD is disabled (ip6_dad_count == 0)
1257 	 * - the interface address is anycast
1258 	 */
1259 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1260 		log(LOG_DEBUG,
1261 			"nd6_dad_start: called with non-tentative address "
1262 			"%s(%s)\n",
1263 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1264 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1265 		return;
1266 	}
1267 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1268 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1269 		return;
1270 	}
1271 	if (!V_ip6_dad_count) {
1272 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1273 		return;
1274 	}
1275 	if (ifa->ifa_ifp == NULL)
1276 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1277 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) {
1278 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1279 		return;
1280 	}
1281 	if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1282 	    !(ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) ||
1283 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)) {
1284 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1285 		return;
1286 	}
1287 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1288 		/*
1289 		 * DAD already in progress.  Let the existing entry
1290 		 * to finish it.
1291 		 */
1292 		return;
1293 	}
1294 
1295 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1296 	if (dp == NULL) {
1297 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1298 			"%s(%s)\n",
1299 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1300 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1301 		return;
1302 	}
1303 	callout_init(&dp->dad_timer_ch, 0);
1304 #ifdef VIMAGE
1305 	dp->dad_vnet = curvnet;
1306 #endif
1307 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1308 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1309 
1310 	/*
1311 	 * Send NS packet for DAD, ip6_dad_count times.
1312 	 * Note that we must delay the first transmission, if this is the
1313 	 * first packet to be sent from the interface after interface
1314 	 * (re)initialization.
1315 	 */
1316 	dp->dad_ifa = ifa;
1317 	ifa_ref(dp->dad_ifa);
1318 	dp->dad_count = V_ip6_dad_count;
1319 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1320 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1321 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1322 	refcount_init(&dp->dad_refcnt, 1);
1323 	nd6_dad_add(dp);
1324 	send_ns = 0;
1325 	if (delay == 0) {
1326 		send_ns = 1;
1327 		delay = (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000;
1328 	}
1329 	nd6_dad_starttimer(dp, delay, send_ns);
1330 }
1331 
1332 /*
1333  * terminate DAD unconditionally.  used for address removals.
1334  */
1335 void
1336 nd6_dad_stop(struct ifaddr *ifa)
1337 {
1338 	struct dadq *dp;
1339 
1340 	dp = nd6_dad_find(ifa, NULL);
1341 	if (!dp) {
1342 		/* DAD wasn't started yet */
1343 		return;
1344 	}
1345 
1346 	nd6_dad_stoptimer(dp);
1347 
1348 	/*
1349 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1350 	 * we were waiting for it to stop, so re-do the lookup.
1351 	 */
1352 	nd6_dad_rele(dp);
1353 	if (nd6_dad_find(ifa, NULL) == NULL)
1354 		return;
1355 
1356 	nd6_dad_del(dp);
1357 	nd6_dad_rele(dp);
1358 }
1359 
1360 static void
1361 nd6_dad_timer(struct dadq *dp)
1362 {
1363 	CURVNET_SET(dp->dad_vnet);
1364 	struct ifaddr *ifa = dp->dad_ifa;
1365 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1366 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1367 	char ip6buf[INET6_ADDRSTRLEN];
1368 
1369 	/* Sanity check */
1370 	if (ia == NULL) {
1371 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1372 		goto err;
1373 	}
1374 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1375 		/* Do not need DAD for ifdisabled interface. */
1376 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1377 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1378 		goto err;
1379 	}
1380 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1381 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1382 			"%s(%s)\n",
1383 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1384 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1385 		goto err;
1386 	}
1387 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1388 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1389 			"%s(%s)\n",
1390 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1391 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1392 		goto err;
1393 	}
1394 
1395 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1396 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1397 	    (((ifp->if_flags & IFF_UP) == 0) ||
1398 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1399 		nd6log((LOG_INFO, "%s: could not run DAD "
1400 		    "because the interface was down or not running.\n",
1401 		    if_name(ifa->ifa_ifp)));
1402 		goto err;
1403 	}
1404 
1405 	/* Need more checks? */
1406 	if (dp->dad_ns_ocount < dp->dad_count) {
1407 		/*
1408 		 * We have more NS to go.  Send NS packet for DAD.
1409 		 */
1410 		nd6_dad_starttimer(dp,
1411 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1412 		goto done;
1413 	} else {
1414 		/*
1415 		 * We have transmitted sufficient number of DAD packets.
1416 		 * See what we've got.
1417 		 */
1418 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1419 			/* We've seen NS or NA, means DAD has failed. */
1420 			nd6_dad_duplicated(ifa, dp);
1421 		else if (V_dad_enhanced != 0 &&
1422 		    dp->dad_ns_lcount > 0 &&
1423 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1424 			/*
1425 			 * Sec. 4.1 in RFC 7527 requires transmission of
1426 			 * additional probes until the loopback condition
1427 			 * becomes clear when a looped back probe is detected.
1428 			 */
1429 			log(LOG_ERR, "%s: a looped back NS message is "
1430 			    "detected during DAD for %s.  "
1431 			    "Another DAD probes are being sent.\n",
1432 			    if_name(ifa->ifa_ifp),
1433 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1434 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1435 			/*
1436 			 * Send an NS immediately and increase dad_count by
1437 			 * V_nd6_mmaxtries - 1.
1438 			 */
1439 			dp->dad_count =
1440 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1441 			nd6_dad_starttimer(dp,
1442 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1443 			    1);
1444 			goto done;
1445 		} else {
1446 			/*
1447 			 * We are done with DAD.  No NA came, no NS came.
1448 			 * No duplicate address found.  Check IFDISABLED flag
1449 			 * again in case that it is changed between the
1450 			 * beginning of this function and here.
1451 			 */
1452 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1453 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1454 
1455 			nd6log((LOG_DEBUG,
1456 			    "%s: DAD complete for %s - no duplicates found\n",
1457 			    if_name(ifa->ifa_ifp),
1458 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1459 			if (dp->dad_ns_lcount > 0)
1460 				log(LOG_ERR, "%s: DAD completed while "
1461 				    "a looped back NS message is detected "
1462 				    "during DAD for %s.\n",
1463 				    if_name(ifa->ifa_ifp),
1464 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1465 		}
1466 	}
1467 err:
1468 	nd6_dad_del(dp);
1469 done:
1470 	CURVNET_RESTORE();
1471 }
1472 
1473 static void
1474 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1475 {
1476 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1477 	struct ifnet *ifp;
1478 	char ip6buf[INET6_ADDRSTRLEN];
1479 
1480 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1481 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1482 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1483 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1484 	    dp->dad_na_icount);
1485 
1486 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1487 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1488 
1489 	ifp = ifa->ifa_ifp;
1490 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1491 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1492 	log(LOG_ERR, "%s: manual intervention required\n",
1493 	    if_name(ifp));
1494 
1495 	/*
1496 	 * If the address is a link-local address formed from an interface
1497 	 * identifier based on the hardware address which is supposed to be
1498 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1499 	 * operation on the interface SHOULD be disabled.
1500 	 * [RFC 4862, Section 5.4.5]
1501 	 */
1502 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1503 		struct in6_addr in6;
1504 
1505 		/*
1506 		 * To avoid over-reaction, we only apply this logic when we are
1507 		 * very sure that hardware addresses are supposed to be unique.
1508 		 */
1509 		switch (ifp->if_type) {
1510 		case IFT_ETHER:
1511 		case IFT_FDDI:
1512 		case IFT_ATM:
1513 		case IFT_IEEE1394:
1514 		case IFT_IEEE80211:
1515 		case IFT_INFINIBAND:
1516 			in6 = ia->ia_addr.sin6_addr;
1517 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1518 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1519 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1520 				log(LOG_ERR, "%s: possible hardware address "
1521 				    "duplication detected, disable IPv6\n",
1522 				    if_name(ifp));
1523 			}
1524 			break;
1525 		}
1526 	}
1527 }
1528 
1529 static void
1530 nd6_dad_ns_output(struct dadq *dp)
1531 {
1532 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1533 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1534 	int i;
1535 
1536 	dp->dad_ns_tcount++;
1537 	if ((ifp->if_flags & IFF_UP) == 0) {
1538 		return;
1539 	}
1540 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1541 		return;
1542 	}
1543 
1544 	dp->dad_ns_ocount++;
1545 	if (V_dad_enhanced != 0) {
1546 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1547 			dp->dad_nonce[i] = arc4random();
1548 		/*
1549 		 * XXXHRS: Note that in the case that
1550 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1551 		 * different nonces can be looped back in an unexpected
1552 		 * order.  The current implementation recognizes only
1553 		 * the latest nonce on the sender side.  Practically it
1554 		 * should work well in almost all cases.
1555 		 */
1556 	}
1557 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1558 	    (uint8_t *)&dp->dad_nonce[0]);
1559 }
1560 
1561 static void
1562 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1563 {
1564 	struct in6_ifaddr *ia;
1565 	struct ifnet *ifp;
1566 	const struct in6_addr *taddr6;
1567 	struct dadq *dp;
1568 
1569 	if (ifa == NULL)
1570 		panic("ifa == NULL in nd6_dad_ns_input");
1571 
1572 	ia = (struct in6_ifaddr *)ifa;
1573 	ifp = ifa->ifa_ifp;
1574 	taddr6 = &ia->ia_addr.sin6_addr;
1575 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1576 	if (V_dad_enhanced == 0)
1577 		ndopt_nonce = NULL;
1578 	dp = nd6_dad_find(ifa, ndopt_nonce);
1579 	if (dp == NULL)
1580 		return;
1581 
1582 	dp->dad_ns_icount++;
1583 	nd6_dad_rele(dp);
1584 }
1585 
1586 static void
1587 nd6_dad_na_input(struct ifaddr *ifa)
1588 {
1589 	struct dadq *dp;
1590 
1591 	if (ifa == NULL)
1592 		panic("ifa == NULL in nd6_dad_na_input");
1593 
1594 	dp = nd6_dad_find(ifa, NULL);
1595 	if (dp != NULL) {
1596 		dp->dad_na_icount++;
1597 		nd6_dad_rele(dp);
1598 	}
1599 }
1600