xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 8fc257994d0ce2396196d7a06d50d20c8015f4b7)
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_carp.h"
39 #include "opt_mpath.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.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/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/callout.h>
55 
56 #include <net/if.h>
57 #include <net/if_types.h>
58 #include <net/if_dl.h>
59 #include <net/if_var.h>
60 #include <net/route.h>
61 #ifdef RADIX_MPATH
62 #include <net/radix_mpath.h>
63 #endif
64 
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <net/if_llatbl.h>
68 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet/icmp6.h>
76 
77 #ifdef DEV_CARP
78 #include <netinet/ip_carp.h>
79 #endif
80 
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 struct dadq;
84 static struct dadq *nd6_dad_find(struct ifaddr *);
85 static void nd6_dad_starttimer(struct dadq *, int);
86 static void nd6_dad_stoptimer(struct dadq *);
87 static void nd6_dad_timer(struct dadq *);
88 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
89 static void nd6_dad_ns_input(struct ifaddr *);
90 static void nd6_dad_na_input(struct ifaddr *);
91 
92 VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
93 VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
94 #define	V_dad_ignore_ns			VNET(dad_ignore_ns)
95 #define	V_dad_maxtry			VNET(dad_maxtry)
96 
97 /*
98  * Input a Neighbor Solicitation Message.
99  *
100  * Based on RFC 2461
101  * Based on RFC 2462 (duplicate address detection)
102  */
103 void
104 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
105 {
106 	struct ifnet *ifp = m->m_pkthdr.rcvif;
107 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
108 	struct nd_neighbor_solicit *nd_ns;
109 	struct in6_addr saddr6 = ip6->ip6_src;
110 	struct in6_addr daddr6 = ip6->ip6_dst;
111 	struct in6_addr taddr6;
112 	struct in6_addr myaddr6;
113 	char *lladdr = NULL;
114 	struct ifaddr *ifa = NULL;
115 	int lladdrlen = 0;
116 	int anycast = 0, proxy = 0, tentative = 0;
117 	int tlladdr;
118 	union nd_opts ndopts;
119 	struct sockaddr_dl *proxydl = NULL;
120 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
121 
122 #ifndef PULLDOWN_TEST
123 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
124 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
125 #else
126 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
127 	if (nd_ns == NULL) {
128 		ICMP6STAT_INC(icp6s_tooshort);
129 		return;
130 	}
131 #endif
132 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
133 	taddr6 = nd_ns->nd_ns_target;
134 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
135 		goto bad;
136 
137 	if (ip6->ip6_hlim != 255) {
138 		nd6log((LOG_ERR,
139 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
140 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
141 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
142 		goto bad;
143 	}
144 
145 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
146 		/* dst has to be a solicited node multicast address. */
147 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
148 		    /* don't check ifindex portion */
149 		    daddr6.s6_addr32[1] == 0 &&
150 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
151 		    daddr6.s6_addr8[12] == 0xff) {
152 			; /* good */
153 		} else {
154 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
155 			    "(wrong ip6 dst)\n"));
156 			goto bad;
157 		}
158 	} else if (!V_nd6_onlink_ns_rfc4861) {
159 		struct sockaddr_in6 src_sa6;
160 
161 		/*
162 		 * According to recent IETF discussions, it is not a good idea
163 		 * to accept a NS from an address which would not be deemed
164 		 * to be a neighbor otherwise.  This point is expected to be
165 		 * clarified in future revisions of the specification.
166 		 */
167 		bzero(&src_sa6, sizeof(src_sa6));
168 		src_sa6.sin6_family = AF_INET6;
169 		src_sa6.sin6_len = sizeof(src_sa6);
170 		src_sa6.sin6_addr = saddr6;
171 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
172 			nd6log((LOG_INFO, "nd6_ns_input: "
173 				"NS packet from non-neighbor\n"));
174 			goto bad;
175 		}
176 	}
177 
178 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
179 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
180 		goto bad;
181 	}
182 
183 	icmp6len -= sizeof(*nd_ns);
184 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
185 	if (nd6_options(&ndopts) < 0) {
186 		nd6log((LOG_INFO,
187 		    "nd6_ns_input: invalid ND option, ignored\n"));
188 		/* nd6_options have incremented stats */
189 		goto freeit;
190 	}
191 
192 	if (ndopts.nd_opts_src_lladdr) {
193 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
194 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
195 	}
196 
197 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
198 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
199 		    "(link-layer address option)\n"));
200 		goto bad;
201 	}
202 
203 	/*
204 	 * Attaching target link-layer address to the NA?
205 	 * (RFC 2461 7.2.4)
206 	 *
207 	 * NS IP dst is unicast/anycast			MUST NOT add
208 	 * NS IP dst is solicited-node multicast	MUST add
209 	 *
210 	 * In implementation, we add target link-layer address by default.
211 	 * We do not add one in MUST NOT cases.
212 	 */
213 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
214 		tlladdr = 0;
215 	else
216 		tlladdr = 1;
217 
218 	/*
219 	 * Target address (taddr6) must be either:
220 	 * (1) Valid unicast/anycast address for my receiving interface,
221 	 * (2) Unicast address for which I'm offering proxy service, or
222 	 * (3) "tentative" address on which DAD is being performed.
223 	 */
224 	/* (1) and (3) check. */
225 #ifdef DEV_CARP
226 	if (ifp->if_carp)
227 		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
228 	if (ifa == NULL)
229 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
230 #else
231 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
232 #endif
233 
234 	/* (2) check. */
235 	if (ifa == NULL) {
236 		struct rtentry *rt;
237 		struct sockaddr_in6 tsin6;
238 		int need_proxy;
239 #ifdef RADIX_MPATH
240 		struct route_in6 ro;
241 #endif
242 
243 		bzero(&tsin6, sizeof tsin6);
244 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
245 		tsin6.sin6_family = AF_INET6;
246 		tsin6.sin6_addr = taddr6;
247 
248 #ifdef RADIX_MPATH
249 		bzero(&ro, sizeof(ro));
250 		ro.ro_dst = tsin6;
251 		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
252 		rt = ro.ro_rt;
253 #else
254 		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
255 #endif
256 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
257 		    rt->rt_gateway->sa_family == AF_LINK);
258 		if (rt)
259 			RTFREE_LOCKED(rt);
260 		if (need_proxy) {
261 			/*
262 			 * proxy NDP for single entry
263 			 */
264 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
265 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
266 			if (ifa) {
267 				proxy = 1;
268 				proxydl = SDL(rt->rt_gateway);
269 			}
270 		}
271 	}
272 	if (ifa == NULL) {
273 		/*
274 		 * We've got an NS packet, and we don't have that adddress
275 		 * assigned for us.  We MUST silently ignore it.
276 		 * See RFC2461 7.2.3.
277 		 */
278 		goto freeit;
279 	}
280 	myaddr6 = *IFA_IN6(ifa);
281 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
282 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
283 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
284 		goto freeit;
285 
286 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
287 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
288 		    "(if %d, NS packet %d)\n",
289 		    ip6_sprintf(ip6bufs, &taddr6),
290 		    ifp->if_addrlen, lladdrlen - 2));
291 		goto bad;
292 	}
293 
294 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
295 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
296 		    ip6_sprintf(ip6bufs, &saddr6)));
297 		goto freeit;
298 	}
299 
300 	/*
301 	 * We have neighbor solicitation packet, with target address equals to
302 	 * one of my tentative address.
303 	 *
304 	 * src addr	how to process?
305 	 * ---		---
306 	 * multicast	of course, invalid (rejected in ip6_input)
307 	 * unicast	somebody is doing address resolution -> ignore
308 	 * unspec	dup address detection
309 	 *
310 	 * The processing is defined in RFC 2462.
311 	 */
312 	if (tentative) {
313 		/*
314 		 * If source address is unspecified address, it is for
315 		 * duplicate address detection.
316 		 *
317 		 * If not, the packet is for addess resolution;
318 		 * silently ignore it.
319 		 */
320 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
321 			nd6_dad_ns_input(ifa);
322 
323 		goto freeit;
324 	}
325 
326 	/*
327 	 * If the source address is unspecified address, entries must not
328 	 * be created or updated.
329 	 * It looks that sender is performing DAD.  Output NA toward
330 	 * all-node multicast address, to tell the sender that I'm using
331 	 * the address.
332 	 * S bit ("solicited") must be zero.
333 	 */
334 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
335 		struct in6_addr in6_all;
336 
337 		in6_all = in6addr_linklocal_allnodes;
338 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
339 			goto bad;
340 		nd6_na_output(ifp, &in6_all, &taddr6,
341 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
342 		    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
343 		    tlladdr, (struct sockaddr *)proxydl);
344 		goto freeit;
345 	}
346 
347 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
348 	    ND_NEIGHBOR_SOLICIT, 0);
349 
350 	nd6_na_output(ifp, &saddr6, &taddr6,
351 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
352 	    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
353 	    tlladdr, (struct sockaddr *)proxydl);
354  freeit:
355 	if (ifa != NULL)
356 		ifa_free(ifa);
357 	m_freem(m);
358 	return;
359 
360  bad:
361 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
362 		ip6_sprintf(ip6bufs, &saddr6)));
363 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
364 		ip6_sprintf(ip6bufs, &daddr6)));
365 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
366 		ip6_sprintf(ip6bufs, &taddr6)));
367 	ICMP6STAT_INC(icp6s_badns);
368 	if (ifa != NULL)
369 		ifa_free(ifa);
370 	m_freem(m);
371 }
372 
373 /*
374  * Output a Neighbor Solicitation Message. Caller specifies:
375  *	- ICMP6 header source IP6 address
376  *	- ND6 header target IP6 address
377  *	- ND6 header source datalink address
378  *
379  * Based on RFC 2461
380  * Based on RFC 2462 (duplicate address detection)
381  *
382  *   ln - for source address determination
383  *  dad - duplicate address detection
384  */
385 void
386 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
387     const struct in6_addr *taddr6, struct llentry *ln, int dad)
388 {
389 	struct mbuf *m;
390 	struct ip6_hdr *ip6;
391 	struct nd_neighbor_solicit *nd_ns;
392 	struct in6_addr *src, src_in;
393 	struct ip6_moptions im6o;
394 	int icmp6len;
395 	int maxlen;
396 	caddr_t mac;
397 	struct route_in6 ro;
398 
399 	bzero(&ro, sizeof(ro));
400 
401 	if (IN6_IS_ADDR_MULTICAST(taddr6))
402 		return;
403 
404 	/* estimate the size of message */
405 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
406 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
407 	if (max_linkhdr + maxlen >= MCLBYTES) {
408 #ifdef DIAGNOSTIC
409 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
410 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
411 #endif
412 		return;
413 	}
414 
415 	MGETHDR(m, M_DONTWAIT, MT_DATA);
416 	if (m && max_linkhdr + maxlen >= MHLEN) {
417 		MCLGET(m, M_DONTWAIT);
418 		if ((m->m_flags & M_EXT) == 0) {
419 			m_free(m);
420 			m = NULL;
421 		}
422 	}
423 	if (m == NULL)
424 		return;
425 	m->m_pkthdr.rcvif = NULL;
426 
427 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
428 		m->m_flags |= M_MCAST;
429 		im6o.im6o_multicast_ifp = ifp;
430 		im6o.im6o_multicast_hlim = 255;
431 		im6o.im6o_multicast_loop = 0;
432 	}
433 
434 	icmp6len = sizeof(*nd_ns);
435 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
436 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
437 
438 	/* fill neighbor solicitation packet */
439 	ip6 = mtod(m, struct ip6_hdr *);
440 	ip6->ip6_flow = 0;
441 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
442 	ip6->ip6_vfc |= IPV6_VERSION;
443 	/* ip6->ip6_plen will be set later */
444 	ip6->ip6_nxt = IPPROTO_ICMPV6;
445 	ip6->ip6_hlim = 255;
446 	if (daddr6)
447 		ip6->ip6_dst = *daddr6;
448 	else {
449 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
450 		ip6->ip6_dst.s6_addr16[1] = 0;
451 		ip6->ip6_dst.s6_addr32[1] = 0;
452 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
453 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
454 		ip6->ip6_dst.s6_addr8[12] = 0xff;
455 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
456 			goto bad;
457 	}
458 	if (!dad) {
459 		struct ifaddr *ifa;
460 
461 		/*
462 		 * RFC2461 7.2.2:
463 		 * "If the source address of the packet prompting the
464 		 * solicitation is the same as one of the addresses assigned
465 		 * to the outgoing interface, that address SHOULD be placed
466 		 * in the IP Source Address of the outgoing solicitation.
467 		 * Otherwise, any one of the addresses assigned to the
468 		 * interface should be used."
469 		 *
470 		 * We use the source address for the prompting packet
471 		 * (saddr6), if:
472 		 * - saddr6 is given from the caller (by giving "ln"), and
473 		 * - saddr6 belongs to the outgoing interface.
474 		 * Otherwise, we perform the source address selection as usual.
475 		 */
476 		struct ip6_hdr *hip6;		/* hold ip6 */
477 		struct in6_addr *hsrc = NULL;
478 
479 		if ((ln != NULL) && ln->la_hold) {
480 			/*
481 			 * assuming every packet in la_hold has the same IP
482 			 * header
483 			 */
484 			hip6 = mtod(ln->la_hold, struct ip6_hdr *);
485 			/* XXX pullup? */
486 			if (sizeof(*hip6) < ln->la_hold->m_len)
487 				hsrc = &hip6->ip6_src;
488 			else
489 				hsrc = NULL;
490 		}
491 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
492 		    hsrc)) != NULL) {
493 			src = hsrc;
494 			ifa_free(ifa);
495 		} else {
496 			int error;
497 			struct sockaddr_in6 dst_sa;
498 
499 			bzero(&dst_sa, sizeof(dst_sa));
500 			dst_sa.sin6_family = AF_INET6;
501 			dst_sa.sin6_len = sizeof(dst_sa);
502 			dst_sa.sin6_addr = ip6->ip6_dst;
503 
504 			error = in6_selectsrc(&dst_sa, NULL,
505 			    NULL, &ro, NULL, NULL, &src_in);
506 			if (error) {
507 				char ip6buf[INET6_ADDRSTRLEN];
508 				nd6log((LOG_DEBUG,
509 				    "nd6_ns_output: source can't be "
510 				    "determined: dst=%s, error=%d\n",
511 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
512 				    error));
513 				goto bad;
514 			}
515 			src = &src_in;
516 		}
517 	} else {
518 		/*
519 		 * Source address for DAD packet must always be IPv6
520 		 * unspecified address. (0::0)
521 		 * We actually don't have to 0-clear the address (we did it
522 		 * above), but we do so here explicitly to make the intention
523 		 * clearer.
524 		 */
525 		bzero(&src_in, sizeof(src_in));
526 		src = &src_in;
527 	}
528 	ip6->ip6_src = *src;
529 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
530 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
531 	nd_ns->nd_ns_code = 0;
532 	nd_ns->nd_ns_reserved = 0;
533 	nd_ns->nd_ns_target = *taddr6;
534 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
535 
536 	/*
537 	 * Add source link-layer address option.
538 	 *
539 	 *				spec		implementation
540 	 *				---		---
541 	 * DAD packet			MUST NOT	do not add the option
542 	 * there's no link layer address:
543 	 *				impossible	do not add the option
544 	 * there's link layer address:
545 	 *	Multicast NS		MUST add one	add the option
546 	 *	Unicast NS		SHOULD add one	add the option
547 	 */
548 	if (!dad && (mac = nd6_ifptomac(ifp))) {
549 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
550 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
551 		/* 8 byte alignments... */
552 		optlen = (optlen + 7) & ~7;
553 
554 		m->m_pkthdr.len += optlen;
555 		m->m_len += optlen;
556 		icmp6len += optlen;
557 		bzero((caddr_t)nd_opt, optlen);
558 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
559 		nd_opt->nd_opt_len = optlen >> 3;
560 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
561 	}
562 
563 	ip6->ip6_plen = htons((u_short)icmp6len);
564 	nd_ns->nd_ns_cksum = 0;
565 	nd_ns->nd_ns_cksum =
566 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
567 
568 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
569 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
570 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
571 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
572 
573 	if (ro.ro_rt) {		/* we don't cache this route. */
574 		RTFREE(ro.ro_rt);
575 	}
576 	return;
577 
578   bad:
579 	if (ro.ro_rt) {
580 		RTFREE(ro.ro_rt);
581 	}
582 	m_freem(m);
583 	return;
584 }
585 
586 /*
587  * Neighbor advertisement input handling.
588  *
589  * Based on RFC 2461
590  * Based on RFC 2462 (duplicate address detection)
591  *
592  * the following items are not implemented yet:
593  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
594  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
595  */
596 void
597 nd6_na_input(struct mbuf *m, int off, int icmp6len)
598 {
599 	struct ifnet *ifp = m->m_pkthdr.rcvif;
600 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
601 	struct nd_neighbor_advert *nd_na;
602 	struct in6_addr daddr6 = ip6->ip6_dst;
603 	struct in6_addr taddr6;
604 	int flags;
605 	int is_router;
606 	int is_solicited;
607 	int is_override;
608 	char *lladdr = NULL;
609 	int lladdrlen = 0;
610 	int checklink = 0;
611 	struct ifaddr *ifa;
612 	struct llentry *ln = NULL;
613 	union nd_opts ndopts;
614 	struct mbuf *chain = NULL;
615 	struct sockaddr_in6 sin6;
616 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
617 
618 	if (ip6->ip6_hlim != 255) {
619 		nd6log((LOG_ERR,
620 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
621 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
622 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
623 		goto bad;
624 	}
625 
626 #ifndef PULLDOWN_TEST
627 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
628 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
629 #else
630 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
631 	if (nd_na == NULL) {
632 		ICMP6STAT_INC(icp6s_tooshort);
633 		return;
634 	}
635 #endif
636 
637 	flags = nd_na->nd_na_flags_reserved;
638 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
639 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
640 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
641 
642 	taddr6 = nd_na->nd_na_target;
643 	if (in6_setscope(&taddr6, ifp, NULL))
644 		goto bad;	/* XXX: impossible */
645 
646 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
647 		nd6log((LOG_ERR,
648 		    "nd6_na_input: invalid target address %s\n",
649 		    ip6_sprintf(ip6bufs, &taddr6)));
650 		goto bad;
651 	}
652 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
653 		if (is_solicited) {
654 			nd6log((LOG_ERR,
655 			    "nd6_na_input: a solicited adv is multicasted\n"));
656 			goto bad;
657 		}
658 
659 	icmp6len -= sizeof(*nd_na);
660 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
661 	if (nd6_options(&ndopts) < 0) {
662 		nd6log((LOG_INFO,
663 		    "nd6_na_input: invalid ND option, ignored\n"));
664 		/* nd6_options have incremented stats */
665 		goto freeit;
666 	}
667 
668 	if (ndopts.nd_opts_tgt_lladdr) {
669 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
670 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
671 	}
672 
673 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
674 
675 	/*
676 	 * Target address matches one of my interface address.
677 	 *
678 	 * If my address is tentative, this means that there's somebody
679 	 * already using the same address as mine.  This indicates DAD failure.
680 	 * This is defined in RFC 2462.
681 	 *
682 	 * Otherwise, process as defined in RFC 2461.
683 	 */
684 	if (ifa
685 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
686 		ifa_free(ifa);
687 		nd6_dad_na_input(ifa);
688 		goto freeit;
689 	}
690 
691 	/* Just for safety, maybe unnecessary. */
692 	if (ifa) {
693 		ifa_free(ifa);
694 		log(LOG_ERR,
695 		    "nd6_na_input: duplicate IP6 address %s\n",
696 		    ip6_sprintf(ip6bufs, &taddr6));
697 		goto freeit;
698 	}
699 
700 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
701 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
702 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
703 		    ifp->if_addrlen, lladdrlen - 2));
704 		goto bad;
705 	}
706 
707 	/*
708 	 * If no neighbor cache entry is found, NA SHOULD silently be
709 	 * discarded.
710 	 */
711 	IF_AFDATA_LOCK(ifp);
712 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
713 	IF_AFDATA_UNLOCK(ifp);
714 	if (ln == NULL) {
715 		goto freeit;
716 	}
717 
718 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
719 		/*
720 		 * If the link-layer has address, and no lladdr option came,
721 		 * discard the packet.
722 		 */
723 		if (ifp->if_addrlen && lladdr == NULL) {
724 			goto freeit;
725 		}
726 
727 		/*
728 		 * Record link-layer address, and update the state.
729 		 */
730 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
731 		ln->la_flags |= LLE_VALID;
732 		if (is_solicited) {
733 			ln->ln_state = ND6_LLINFO_REACHABLE;
734 			ln->ln_byhint = 0;
735 			if (!ND6_LLINFO_PERMANENT(ln)) {
736 				nd6_llinfo_settimer_locked(ln,
737 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
738 			}
739 		} else {
740 			ln->ln_state = ND6_LLINFO_STALE;
741 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
742 		}
743 		if ((ln->ln_router = is_router) != 0) {
744 			/*
745 			 * This means a router's state has changed from
746 			 * non-reachable to probably reachable, and might
747 			 * affect the status of associated prefixes..
748 			 */
749 			checklink = 1;
750 		}
751 	} else {
752 		int llchange;
753 
754 		/*
755 		 * Check if the link-layer address has changed or not.
756 		 */
757 		if (lladdr == NULL)
758 			llchange = 0;
759 		else {
760 			if (ln->la_flags & LLE_VALID) {
761 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
762 					llchange = 1;
763 				else
764 					llchange = 0;
765 			} else
766 				llchange = 1;
767 		}
768 
769 		/*
770 		 * This is VERY complex.  Look at it with care.
771 		 *
772 		 * override solicit lladdr llchange	action
773 		 *					(L: record lladdr)
774 		 *
775 		 *	0	0	n	--	(2c)
776 		 *	0	0	y	n	(2b) L
777 		 *	0	0	y	y	(1)    REACHABLE->STALE
778 		 *	0	1	n	--	(2c)   *->REACHABLE
779 		 *	0	1	y	n	(2b) L *->REACHABLE
780 		 *	0	1	y	y	(1)    REACHABLE->STALE
781 		 *	1	0	n	--	(2a)
782 		 *	1	0	y	n	(2a) L
783 		 *	1	0	y	y	(2a) L *->STALE
784 		 *	1	1	n	--	(2a)   *->REACHABLE
785 		 *	1	1	y	n	(2a) L *->REACHABLE
786 		 *	1	1	y	y	(2a) L *->REACHABLE
787 		 */
788 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
789 			/*
790 			 * If state is REACHABLE, make it STALE.
791 			 * no other updates should be done.
792 			 */
793 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
794 				ln->ln_state = ND6_LLINFO_STALE;
795 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
796 			}
797 			goto freeit;
798 		} else if (is_override				   /* (2a) */
799 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
800 			|| lladdr == NULL) {			   /* (2c) */
801 			/*
802 			 * Update link-local address, if any.
803 			 */
804 			if (lladdr != NULL) {
805 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
806 				ln->la_flags |= LLE_VALID;
807 			}
808 
809 			/*
810 			 * If solicited, make the state REACHABLE.
811 			 * If not solicited and the link-layer address was
812 			 * changed, make it STALE.
813 			 */
814 			if (is_solicited) {
815 				ln->ln_state = ND6_LLINFO_REACHABLE;
816 				ln->ln_byhint = 0;
817 				if (!ND6_LLINFO_PERMANENT(ln)) {
818 					nd6_llinfo_settimer_locked(ln,
819 					    (long)ND_IFINFO(ifp)->reachable * hz);
820 				}
821 			} else {
822 				if (lladdr != NULL && llchange) {
823 					ln->ln_state = ND6_LLINFO_STALE;
824 					nd6_llinfo_settimer_locked(ln,
825 					    (long)V_nd6_gctimer * hz);
826 				}
827 			}
828 		}
829 
830 		if (ln->ln_router && !is_router) {
831 			/*
832 			 * The peer dropped the router flag.
833 			 * Remove the sender from the Default Router List and
834 			 * update the Destination Cache entries.
835 			 */
836 			struct nd_defrouter *dr;
837 			struct in6_addr *in6;
838 
839 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
840 
841 			/*
842 			 * Lock to protect the default router list.
843 			 * XXX: this might be unnecessary, since this function
844 			 * is only called under the network software interrupt
845 			 * context.  However, we keep it just for safety.
846 			 */
847 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
848 			if (dr)
849 				defrtrlist_del(dr);
850 			else if (!V_ip6_forwarding) {
851 				/*
852 				 * Even if the neighbor is not in the default
853 				 * router list, the neighbor may be used
854 				 * as a next hop for some destinations
855 				 * (e.g. redirect case). So we must
856 				 * call rt6_flush explicitly.
857 				 */
858 				rt6_flush(&ip6->ip6_src, ifp);
859 			}
860 		}
861 		ln->ln_router = is_router;
862 	}
863         /* XXX - QL
864 	 *  Does this matter?
865 	 *  rt->rt_flags &= ~RTF_REJECT;
866 	 */
867 	ln->la_asked = 0;
868 	if (ln->la_hold) {
869 		struct mbuf *m_hold, *m_hold_next;
870 
871 		/*
872 		 * reset the la_hold in advance, to explicitly
873 		 * prevent a la_hold lookup in nd6_output()
874 		 * (wouldn't happen, though...)
875 		 */
876 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
877 		    m_hold; m_hold = m_hold_next) {
878 			m_hold_next = m_hold->m_nextpkt;
879 			m_hold->m_nextpkt = NULL;
880 			/*
881 			 * we assume ifp is not a loopback here, so just set
882 			 * the 2nd argument as the 1st one.
883 			 */
884 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
885 		}
886 	}
887  freeit:
888 	if (ln != NULL) {
889 		if (chain)
890 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
891 		LLE_WUNLOCK(ln);
892 
893 		if (chain)
894 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
895 	}
896 	if (checklink)
897 		pfxlist_onlink_check();
898 
899 	m_freem(m);
900 	return;
901 
902  bad:
903 	if (ln != NULL)
904 		LLE_WUNLOCK(ln);
905 
906 	ICMP6STAT_INC(icp6s_badna);
907 	m_freem(m);
908 }
909 
910 /*
911  * Neighbor advertisement output handling.
912  *
913  * Based on RFC 2461
914  *
915  * the following items are not implemented yet:
916  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
917  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
918  *
919  * tlladdr - 1 if include target link-layer address
920  * sdl0 - sockaddr_dl (= proxy NA) or NULL
921  */
922 void
923 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
924     const struct in6_addr *taddr6, u_long flags, int tlladdr,
925     struct sockaddr *sdl0)
926 {
927 	struct mbuf *m;
928 	struct ip6_hdr *ip6;
929 	struct nd_neighbor_advert *nd_na;
930 	struct ip6_moptions im6o;
931 	struct in6_addr src, daddr6;
932 	struct sockaddr_in6 dst_sa;
933 	int icmp6len, maxlen, error;
934 	caddr_t mac = NULL;
935 	struct route_in6 ro;
936 
937 	bzero(&ro, sizeof(ro));
938 
939 	daddr6 = *daddr6_0;	/* make a local copy for modification */
940 
941 	/* estimate the size of message */
942 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
943 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
944 	if (max_linkhdr + maxlen >= MCLBYTES) {
945 #ifdef DIAGNOSTIC
946 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
947 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
948 #endif
949 		return;
950 	}
951 
952 	MGETHDR(m, M_DONTWAIT, MT_DATA);
953 	if (m && max_linkhdr + maxlen >= MHLEN) {
954 		MCLGET(m, M_DONTWAIT);
955 		if ((m->m_flags & M_EXT) == 0) {
956 			m_free(m);
957 			m = NULL;
958 		}
959 	}
960 	if (m == NULL)
961 		return;
962 	m->m_pkthdr.rcvif = NULL;
963 
964 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
965 		m->m_flags |= M_MCAST;
966 		im6o.im6o_multicast_ifp = ifp;
967 		im6o.im6o_multicast_hlim = 255;
968 		im6o.im6o_multicast_loop = 0;
969 	}
970 
971 	icmp6len = sizeof(*nd_na);
972 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
973 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
974 
975 	/* fill neighbor advertisement packet */
976 	ip6 = mtod(m, struct ip6_hdr *);
977 	ip6->ip6_flow = 0;
978 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
979 	ip6->ip6_vfc |= IPV6_VERSION;
980 	ip6->ip6_nxt = IPPROTO_ICMPV6;
981 	ip6->ip6_hlim = 255;
982 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
983 		/* reply to DAD */
984 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
985 		daddr6.s6_addr16[1] = 0;
986 		daddr6.s6_addr32[1] = 0;
987 		daddr6.s6_addr32[2] = 0;
988 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
989 		if (in6_setscope(&daddr6, ifp, NULL))
990 			goto bad;
991 
992 		flags &= ~ND_NA_FLAG_SOLICITED;
993 	}
994 	ip6->ip6_dst = daddr6;
995 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
996 	dst_sa.sin6_family = AF_INET6;
997 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
998 	dst_sa.sin6_addr = daddr6;
999 
1000 	/*
1001 	 * Select a source whose scope is the same as that of the dest.
1002 	 */
1003 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1004 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &src);
1005 	if (error) {
1006 		char ip6buf[INET6_ADDRSTRLEN];
1007 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1008 		    "determined: dst=%s, error=%d\n",
1009 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1010 		goto bad;
1011 	}
1012 	ip6->ip6_src = src;
1013 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1014 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1015 	nd_na->nd_na_code = 0;
1016 	nd_na->nd_na_target = *taddr6;
1017 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1018 
1019 	/*
1020 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1021 	 * see nd6_ns_input() for details.
1022 	 * Basically, if NS packet is sent to unicast/anycast addr,
1023 	 * target lladdr option SHOULD NOT be included.
1024 	 */
1025 	if (tlladdr) {
1026 		/*
1027 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1028 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1029 		 * my address) use lladdr configured for the interface.
1030 		 */
1031 		if (sdl0 == NULL) {
1032 #ifdef DEV_CARP
1033 			if (ifp->if_carp)
1034 				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
1035 			if (mac == NULL)
1036 				mac = nd6_ifptomac(ifp);
1037 #else
1038 			mac = nd6_ifptomac(ifp);
1039 #endif
1040 		} else if (sdl0->sa_family == AF_LINK) {
1041 			struct sockaddr_dl *sdl;
1042 			sdl = (struct sockaddr_dl *)sdl0;
1043 			if (sdl->sdl_alen == ifp->if_addrlen)
1044 				mac = LLADDR(sdl);
1045 		}
1046 	}
1047 	if (tlladdr && mac) {
1048 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1049 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1050 
1051 		/* roundup to 8 bytes alignment! */
1052 		optlen = (optlen + 7) & ~7;
1053 
1054 		m->m_pkthdr.len += optlen;
1055 		m->m_len += optlen;
1056 		icmp6len += optlen;
1057 		bzero((caddr_t)nd_opt, optlen);
1058 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1059 		nd_opt->nd_opt_len = optlen >> 3;
1060 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1061 	} else
1062 		flags &= ~ND_NA_FLAG_OVERRIDE;
1063 
1064 	ip6->ip6_plen = htons((u_short)icmp6len);
1065 	nd_na->nd_na_flags_reserved = flags;
1066 	nd_na->nd_na_cksum = 0;
1067 	nd_na->nd_na_cksum =
1068 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1069 
1070 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1071 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1072 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1073 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1074 
1075 	if (ro.ro_rt) {		/* we don't cache this route. */
1076 		RTFREE(ro.ro_rt);
1077 	}
1078 	return;
1079 
1080   bad:
1081 	if (ro.ro_rt) {
1082 		RTFREE(ro.ro_rt);
1083 	}
1084 	m_freem(m);
1085 	return;
1086 }
1087 
1088 caddr_t
1089 nd6_ifptomac(struct ifnet *ifp)
1090 {
1091 	switch (ifp->if_type) {
1092 	case IFT_ARCNET:
1093 	case IFT_ETHER:
1094 	case IFT_FDDI:
1095 	case IFT_IEEE1394:
1096 #ifdef IFT_L2VLAN
1097 	case IFT_L2VLAN:
1098 #endif
1099 #ifdef IFT_IEEE80211
1100 	case IFT_IEEE80211:
1101 #endif
1102 #ifdef IFT_CARP
1103 	case IFT_CARP:
1104 #endif
1105 	case IFT_BRIDGE:
1106 	case IFT_ISO88025:
1107 		return IF_LLADDR(ifp);
1108 	default:
1109 		return NULL;
1110 	}
1111 }
1112 
1113 struct dadq {
1114 	TAILQ_ENTRY(dadq) dad_list;
1115 	struct ifaddr *dad_ifa;
1116 	int dad_count;		/* max NS to send */
1117 	int dad_ns_tcount;	/* # of trials to send NS */
1118 	int dad_ns_ocount;	/* NS sent so far */
1119 	int dad_ns_icount;
1120 	int dad_na_icount;
1121 	struct callout dad_timer_ch;
1122 	struct vnet *dad_vnet;
1123 };
1124 
1125 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1126 VNET_DEFINE(int, dad_init) = 0;
1127 #define	V_dadq				VNET(dadq)
1128 #define	V_dad_init			VNET(dad_init)
1129 
1130 static struct dadq *
1131 nd6_dad_find(struct ifaddr *ifa)
1132 {
1133 	struct dadq *dp;
1134 
1135 	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1136 		if (dp->dad_ifa == ifa)
1137 			return dp;
1138 	}
1139 	return NULL;
1140 }
1141 
1142 static void
1143 nd6_dad_starttimer(struct dadq *dp, int ticks)
1144 {
1145 
1146 	callout_reset(&dp->dad_timer_ch, ticks,
1147 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1148 }
1149 
1150 static void
1151 nd6_dad_stoptimer(struct dadq *dp)
1152 {
1153 
1154 	callout_stop(&dp->dad_timer_ch);
1155 }
1156 
1157 /*
1158  * Start Duplicate Address Detection (DAD) for specified interface address.
1159  */
1160 void
1161 nd6_dad_start(struct ifaddr *ifa, int delay)
1162 {
1163 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1164 	struct dadq *dp;
1165 	char ip6buf[INET6_ADDRSTRLEN];
1166 
1167 	if (!V_dad_init) {
1168 		TAILQ_INIT(&V_dadq);
1169 		V_dad_init++;
1170 	}
1171 
1172 	/*
1173 	 * If we don't need DAD, don't do it.
1174 	 * There are several cases:
1175 	 * - DAD is disabled (ip6_dad_count == 0)
1176 	 * - the interface address is anycast
1177 	 */
1178 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1179 		log(LOG_DEBUG,
1180 			"nd6_dad_start: called with non-tentative address "
1181 			"%s(%s)\n",
1182 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1183 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1184 		return;
1185 	}
1186 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1187 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1188 		return;
1189 	}
1190 	if (!V_ip6_dad_count) {
1191 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1192 		return;
1193 	}
1194 	if (ifa->ifa_ifp == NULL)
1195 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1196 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1197 		return;
1198 	}
1199 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1200 		return;
1201 	if (nd6_dad_find(ifa) != NULL) {
1202 		/* DAD already in progress */
1203 		return;
1204 	}
1205 
1206 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1207 	if (dp == NULL) {
1208 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1209 			"%s(%s)\n",
1210 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1211 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1212 		return;
1213 	}
1214 	bzero(dp, sizeof(*dp));
1215 	callout_init(&dp->dad_timer_ch, 0);
1216 #ifdef VIMAGE
1217 	dp->dad_vnet = curvnet;
1218 #endif
1219 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1220 
1221 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1222 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1223 
1224 	/*
1225 	 * Send NS packet for DAD, ip6_dad_count times.
1226 	 * Note that we must delay the first transmission, if this is the
1227 	 * first packet to be sent from the interface after interface
1228 	 * (re)initialization.
1229 	 */
1230 	dp->dad_ifa = ifa;
1231 	ifa_ref(ifa);	/* just for safety */
1232 	dp->dad_count = V_ip6_dad_count;
1233 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1234 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1235 	if (delay == 0) {
1236 		nd6_dad_ns_output(dp, ifa);
1237 		nd6_dad_starttimer(dp,
1238 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1239 	} else {
1240 		nd6_dad_starttimer(dp, delay);
1241 	}
1242 }
1243 
1244 /*
1245  * terminate DAD unconditionally.  used for address removals.
1246  */
1247 void
1248 nd6_dad_stop(struct ifaddr *ifa)
1249 {
1250 	struct dadq *dp;
1251 
1252 	if (!V_dad_init)
1253 		return;
1254 	dp = nd6_dad_find(ifa);
1255 	if (!dp) {
1256 		/* DAD wasn't started yet */
1257 		return;
1258 	}
1259 
1260 	nd6_dad_stoptimer(dp);
1261 
1262 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1263 	free(dp, M_IP6NDP);
1264 	dp = NULL;
1265 	ifa_free(ifa);
1266 }
1267 
1268 static void
1269 nd6_dad_timer(struct dadq *dp)
1270 {
1271 	CURVNET_SET(dp->dad_vnet);
1272 	int s;
1273 	struct ifaddr *ifa = dp->dad_ifa;
1274 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1275 	char ip6buf[INET6_ADDRSTRLEN];
1276 
1277 	s = splnet();		/* XXX */
1278 
1279 	/* Sanity check */
1280 	if (ia == NULL) {
1281 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1282 		goto done;
1283 	}
1284 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1285 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1286 			"%s(%s)\n",
1287 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1288 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1289 		goto done;
1290 	}
1291 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1292 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1293 			"%s(%s)\n",
1294 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1295 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1296 		goto done;
1297 	}
1298 
1299 	/* timeouted with IFF_{RUNNING,UP} check */
1300 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1301 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1302 		    if_name(ifa->ifa_ifp)));
1303 
1304 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1305 		free(dp, M_IP6NDP);
1306 		dp = NULL;
1307 		ifa_free(ifa);
1308 		goto done;
1309 	}
1310 
1311 	/* Need more checks? */
1312 	if (dp->dad_ns_ocount < dp->dad_count) {
1313 		/*
1314 		 * We have more NS to go.  Send NS packet for DAD.
1315 		 */
1316 		nd6_dad_ns_output(dp, ifa);
1317 		nd6_dad_starttimer(dp,
1318 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1319 	} else {
1320 		/*
1321 		 * We have transmitted sufficient number of DAD packets.
1322 		 * See what we've got.
1323 		 */
1324 		int duplicate;
1325 
1326 		duplicate = 0;
1327 
1328 		if (dp->dad_na_icount) {
1329 			/*
1330 			 * the check is in nd6_dad_na_input(),
1331 			 * but just in case
1332 			 */
1333 			duplicate++;
1334 		}
1335 
1336 		if (dp->dad_ns_icount) {
1337 			/* We've seen NS, means DAD has failed. */
1338 			duplicate++;
1339 		}
1340 
1341 		if (duplicate) {
1342 			/* (*dp) will be freed in nd6_dad_duplicated() */
1343 			dp = NULL;
1344 			nd6_dad_duplicated(ifa);
1345 		} else {
1346 			/*
1347 			 * We are done with DAD.  No NA came, no NS came.
1348 			 * No duplicate address found.
1349 			 */
1350 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1351 
1352 			nd6log((LOG_DEBUG,
1353 			    "%s: DAD complete for %s - no duplicates found\n",
1354 			    if_name(ifa->ifa_ifp),
1355 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1356 
1357 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1358 			free(dp, M_IP6NDP);
1359 			dp = NULL;
1360 			ifa_free(ifa);
1361 		}
1362 	}
1363 
1364 done:
1365 	splx(s);
1366 	CURVNET_RESTORE();
1367 }
1368 
1369 void
1370 nd6_dad_duplicated(struct ifaddr *ifa)
1371 {
1372 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1373 	struct ifnet *ifp;
1374 	struct dadq *dp;
1375 	char ip6buf[INET6_ADDRSTRLEN];
1376 
1377 	dp = nd6_dad_find(ifa);
1378 	if (dp == NULL) {
1379 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1380 		return;
1381 	}
1382 
1383 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1384 	    "NS in/out=%d/%d, NA in=%d\n",
1385 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1386 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1387 
1388 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1389 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1390 
1391 	/* We are done with DAD, with duplicate address found. (failure) */
1392 	nd6_dad_stoptimer(dp);
1393 
1394 	ifp = ifa->ifa_ifp;
1395 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1396 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1397 	log(LOG_ERR, "%s: manual intervention required\n",
1398 	    if_name(ifp));
1399 
1400 	/*
1401 	 * If the address is a link-local address formed from an interface
1402 	 * identifier based on the hardware address which is supposed to be
1403 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1404 	 * operation on the interface SHOULD be disabled.
1405 	 * [RFC 4862, Section 5.4.5]
1406 	 */
1407 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1408 		struct in6_addr in6;
1409 
1410 		/*
1411 		 * To avoid over-reaction, we only apply this logic when we are
1412 		 * very sure that hardware addresses are supposed to be unique.
1413 		 */
1414 		switch (ifp->if_type) {
1415 		case IFT_ETHER:
1416 		case IFT_FDDI:
1417 		case IFT_ATM:
1418 		case IFT_IEEE1394:
1419 #ifdef IFT_IEEE80211
1420 		case IFT_IEEE80211:
1421 #endif
1422 			in6 = ia->ia_addr.sin6_addr;
1423 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1424 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1425 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1426 				log(LOG_ERR, "%s: possible hardware address "
1427 				    "duplication detected, disable IPv6\n",
1428 				    if_name(ifp));
1429 			}
1430 			break;
1431 		}
1432 	}
1433 
1434 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1435 	free(dp, M_IP6NDP);
1436 	dp = NULL;
1437 	ifa_free(ifa);
1438 }
1439 
1440 static void
1441 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1442 {
1443 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1444 	struct ifnet *ifp = ifa->ifa_ifp;
1445 
1446 	dp->dad_ns_tcount++;
1447 	if ((ifp->if_flags & IFF_UP) == 0) {
1448 		return;
1449 	}
1450 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1451 		return;
1452 	}
1453 
1454 	dp->dad_ns_ocount++;
1455 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1456 }
1457 
1458 static void
1459 nd6_dad_ns_input(struct ifaddr *ifa)
1460 {
1461 	struct in6_ifaddr *ia;
1462 	struct ifnet *ifp;
1463 	const struct in6_addr *taddr6;
1464 	struct dadq *dp;
1465 	int duplicate;
1466 
1467 	if (ifa == NULL)
1468 		panic("ifa == NULL in nd6_dad_ns_input");
1469 
1470 	ia = (struct in6_ifaddr *)ifa;
1471 	ifp = ifa->ifa_ifp;
1472 	taddr6 = &ia->ia_addr.sin6_addr;
1473 	duplicate = 0;
1474 	dp = nd6_dad_find(ifa);
1475 
1476 	/* Quickhack - completely ignore DAD NS packets */
1477 	if (V_dad_ignore_ns) {
1478 		char ip6buf[INET6_ADDRSTRLEN];
1479 		nd6log((LOG_INFO,
1480 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1481 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1482 		    if_name(ifa->ifa_ifp)));
1483 		return;
1484 	}
1485 
1486 	/*
1487 	 * if I'm yet to start DAD, someone else started using this address
1488 	 * first.  I have a duplicate and you win.
1489 	 */
1490 	if (dp == NULL || dp->dad_ns_ocount == 0)
1491 		duplicate++;
1492 
1493 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1494 
1495 	if (duplicate) {
1496 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1497 		nd6_dad_duplicated(ifa);
1498 	} else {
1499 		/*
1500 		 * not sure if I got a duplicate.
1501 		 * increment ns count and see what happens.
1502 		 */
1503 		if (dp)
1504 			dp->dad_ns_icount++;
1505 	}
1506 }
1507 
1508 static void
1509 nd6_dad_na_input(struct ifaddr *ifa)
1510 {
1511 	struct dadq *dp;
1512 
1513 	if (ifa == NULL)
1514 		panic("ifa == NULL in nd6_dad_na_input");
1515 
1516 	dp = nd6_dad_find(ifa);
1517 	if (dp)
1518 		dp->dad_na_icount++;
1519 
1520 	/* remove the address. */
1521 	nd6_dad_duplicated(ifa);
1522 }
1523