xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
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/lock.h>
44 #include <sys/rwlock.h>
45 #include <sys/mbuf.h>
46 #include <sys/socket.h>
47 #include <sys/sockio.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/errno.h>
51 #include <sys/syslog.h>
52 #include <sys/queue.h>
53 #include <sys/callout.h>
54 
55 #include <net/if.h>
56 #include <net/if_types.h>
57 #include <net/if_dl.h>
58 #include <net/if_var.h>
59 #include <net/route.h>
60 #ifdef RADIX_MPATH
61 #include <net/radix_mpath.h>
62 #endif
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <net/if_llatbl.h>
67 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68 #include <netinet6/in6_var.h>
69 #include <netinet6/in6_ifattach.h>
70 #include <netinet/ip6.h>
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/scope6_var.h>
73 #include <netinet6/nd6.h>
74 #include <netinet/icmp6.h>
75 #include <netinet/ip_carp.h>
76 #include <netinet6/send.h>
77 
78 #define SDL(s) ((struct sockaddr_dl *)s)
79 
80 struct dadq;
81 static struct dadq *nd6_dad_find(struct ifaddr *);
82 static void nd6_dad_starttimer(struct dadq *, int);
83 static void nd6_dad_stoptimer(struct dadq *);
84 static void nd6_dad_timer(struct dadq *);
85 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86 static void nd6_dad_ns_input(struct ifaddr *);
87 static void nd6_dad_na_input(struct ifaddr *);
88 
89 VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
90 VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
91 #define	V_dad_ignore_ns			VNET(dad_ignore_ns)
92 #define	V_dad_maxtry			VNET(dad_maxtry)
93 
94 /*
95  * Input a Neighbor Solicitation Message.
96  *
97  * Based on RFC 2461
98  * Based on RFC 2462 (duplicate address detection)
99  */
100 void
101 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
102 {
103 	struct ifnet *ifp = m->m_pkthdr.rcvif;
104 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
105 	struct nd_neighbor_solicit *nd_ns;
106 	struct in6_addr saddr6 = ip6->ip6_src;
107 	struct in6_addr daddr6 = ip6->ip6_dst;
108 	struct in6_addr taddr6;
109 	struct in6_addr myaddr6;
110 	char *lladdr = NULL;
111 	struct ifaddr *ifa = NULL;
112 	int lladdrlen = 0;
113 	int anycast = 0, proxy = 0, tentative = 0;
114 	int tlladdr;
115 	union nd_opts ndopts;
116 	struct sockaddr_dl proxydl;
117 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
118 
119 #ifndef PULLDOWN_TEST
120 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
121 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
122 #else
123 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
124 	if (nd_ns == NULL) {
125 		ICMP6STAT_INC(icp6s_tooshort);
126 		return;
127 	}
128 #endif
129 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
130 	taddr6 = nd_ns->nd_ns_target;
131 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
132 		goto bad;
133 
134 	if (ip6->ip6_hlim != 255) {
135 		nd6log((LOG_ERR,
136 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
137 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
138 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
139 		goto bad;
140 	}
141 
142 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
143 		/* dst has to be a solicited node multicast address. */
144 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
145 		    /* don't check ifindex portion */
146 		    daddr6.s6_addr32[1] == 0 &&
147 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
148 		    daddr6.s6_addr8[12] == 0xff) {
149 			; /* good */
150 		} else {
151 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
152 			    "(wrong ip6 dst)\n"));
153 			goto bad;
154 		}
155 	} else if (!V_nd6_onlink_ns_rfc4861) {
156 		struct sockaddr_in6 src_sa6;
157 
158 		/*
159 		 * According to recent IETF discussions, it is not a good idea
160 		 * to accept a NS from an address which would not be deemed
161 		 * to be a neighbor otherwise.  This point is expected to be
162 		 * clarified in future revisions of the specification.
163 		 */
164 		bzero(&src_sa6, sizeof(src_sa6));
165 		src_sa6.sin6_family = AF_INET6;
166 		src_sa6.sin6_len = sizeof(src_sa6);
167 		src_sa6.sin6_addr = saddr6;
168 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
169 			nd6log((LOG_INFO, "nd6_ns_input: "
170 				"NS packet from non-neighbor\n"));
171 			goto bad;
172 		}
173 	}
174 
175 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
176 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
177 		goto bad;
178 	}
179 
180 	icmp6len -= sizeof(*nd_ns);
181 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
182 	if (nd6_options(&ndopts) < 0) {
183 		nd6log((LOG_INFO,
184 		    "nd6_ns_input: invalid ND option, ignored\n"));
185 		/* nd6_options have incremented stats */
186 		goto freeit;
187 	}
188 
189 	if (ndopts.nd_opts_src_lladdr) {
190 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
191 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
192 	}
193 
194 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
195 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
196 		    "(link-layer address option)\n"));
197 		goto bad;
198 	}
199 
200 	/*
201 	 * Attaching target link-layer address to the NA?
202 	 * (RFC 2461 7.2.4)
203 	 *
204 	 * NS IP dst is unicast/anycast			MUST NOT add
205 	 * NS IP dst is solicited-node multicast	MUST add
206 	 *
207 	 * In implementation, we add target link-layer address by default.
208 	 * We do not add one in MUST NOT cases.
209 	 */
210 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
211 		tlladdr = 0;
212 	else
213 		tlladdr = 1;
214 
215 	/*
216 	 * Target address (taddr6) must be either:
217 	 * (1) Valid unicast/anycast address for my receiving interface,
218 	 * (2) Unicast address for which I'm offering proxy service, or
219 	 * (3) "tentative" address on which DAD is being performed.
220 	 */
221 	/* (1) and (3) check. */
222 	if (ifp->if_carp)
223 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
224 	if (ifa == NULL)
225 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
226 
227 	/* (2) check. */
228 	if (ifa == NULL) {
229 		struct rtentry *rt;
230 		struct sockaddr_in6 tsin6;
231 		int need_proxy;
232 #ifdef RADIX_MPATH
233 		struct route_in6 ro;
234 #endif
235 
236 		bzero(&tsin6, sizeof tsin6);
237 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
238 		tsin6.sin6_family = AF_INET6;
239 		tsin6.sin6_addr = taddr6;
240 
241 #ifdef RADIX_MPATH
242 		bzero(&ro, sizeof(ro));
243 		ro.ro_dst = tsin6;
244 		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
245 		rt = ro.ro_rt;
246 #else
247 		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
248 #endif
249 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
250 		    rt->rt_gateway->sa_family == AF_LINK);
251 		if (rt != NULL) {
252 			/*
253 			 * Make a copy while we can be sure that rt_gateway
254 			 * is still stable before unlocking to avoid lock
255 			 * order problems.  proxydl will only be used if
256 			 * proxy will be set in the next block.
257 			 */
258 			if (need_proxy)
259 				proxydl = *SDL(rt->rt_gateway);
260 			RTFREE_LOCKED(rt);
261 		}
262 		if (need_proxy) {
263 			/*
264 			 * proxy NDP for single entry
265 			 */
266 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
267 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
268 			if (ifa)
269 				proxy = 1;
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, proxy ? (struct sockaddr *)&proxydl : NULL);
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, proxy ? (struct sockaddr *)&proxydl : NULL);
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 m_tag *mtag;
391 	struct ip6_hdr *ip6;
392 	struct nd_neighbor_solicit *nd_ns;
393 	struct ip6_moptions im6o;
394 	int icmp6len;
395 	int maxlen;
396 	caddr_t mac;
397 	struct route_in6 ro;
398 
399 	if (IN6_IS_ADDR_MULTICAST(taddr6))
400 		return;
401 
402 	/* estimate the size of message */
403 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
404 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
405 	if (max_linkhdr + maxlen >= MCLBYTES) {
406 #ifdef DIAGNOSTIC
407 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
408 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
409 #endif
410 		return;
411 	}
412 
413 	MGETHDR(m, M_DONTWAIT, MT_DATA);
414 	if (m && max_linkhdr + maxlen >= MHLEN) {
415 		MCLGET(m, M_DONTWAIT);
416 		if ((m->m_flags & M_EXT) == 0) {
417 			m_free(m);
418 			m = NULL;
419 		}
420 	}
421 	if (m == NULL)
422 		return;
423 	m->m_pkthdr.rcvif = NULL;
424 
425 	bzero(&ro, sizeof(ro));
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 in6_addr *hsrc;
477 
478 		hsrc = NULL;
479 		if (ln != NULL) {
480 			LLE_RLOCK(ln);
481 			if (ln->la_hold != NULL) {
482 				struct ip6_hdr *hip6;		/* hold ip6 */
483 
484 				/*
485 				 * assuming every packet in la_hold has the same IP
486 				 * header
487 				 */
488 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
489 				/* XXX pullup? */
490 				if (sizeof(*hip6) < ln->la_hold->m_len) {
491 					ip6->ip6_src = hip6->ip6_src;
492 					hsrc = &hip6->ip6_src;
493 				}
494 			}
495 			LLE_RUNLOCK(ln);
496 		}
497 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
498 		    hsrc)) != NULL) {
499 			/* ip6_src set already. */
500 			ifa_free(ifa);
501 		} else {
502 			int error;
503 			struct sockaddr_in6 dst_sa;
504 			struct in6_addr src_in;
505 
506 			bzero(&dst_sa, sizeof(dst_sa));
507 			dst_sa.sin6_family = AF_INET6;
508 			dst_sa.sin6_len = sizeof(dst_sa);
509 			dst_sa.sin6_addr = ip6->ip6_dst;
510 
511 			error = in6_selectsrc(&dst_sa, NULL,
512 			    NULL, &ro, NULL, NULL, &src_in);
513 			if (error) {
514 				char ip6buf[INET6_ADDRSTRLEN];
515 				nd6log((LOG_DEBUG,
516 				    "nd6_ns_output: source can't be "
517 				    "determined: dst=%s, error=%d\n",
518 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
519 				    error));
520 				goto bad;
521 			}
522 			ip6->ip6_src = src_in;
523 		}
524 	} else {
525 		/*
526 		 * Source address for DAD packet must always be IPv6
527 		 * unspecified address. (0::0)
528 		 * We actually don't have to 0-clear the address (we did it
529 		 * above), but we do so here explicitly to make the intention
530 		 * clearer.
531 		 */
532 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
533 	}
534 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
535 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
536 	nd_ns->nd_ns_code = 0;
537 	nd_ns->nd_ns_reserved = 0;
538 	nd_ns->nd_ns_target = *taddr6;
539 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
540 
541 	/*
542 	 * Add source link-layer address option.
543 	 *
544 	 *				spec		implementation
545 	 *				---		---
546 	 * DAD packet			MUST NOT	do not add the option
547 	 * there's no link layer address:
548 	 *				impossible	do not add the option
549 	 * there's link layer address:
550 	 *	Multicast NS		MUST add one	add the option
551 	 *	Unicast NS		SHOULD add one	add the option
552 	 */
553 	if (!dad && (mac = nd6_ifptomac(ifp))) {
554 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
555 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
556 		/* 8 byte alignments... */
557 		optlen = (optlen + 7) & ~7;
558 
559 		m->m_pkthdr.len += optlen;
560 		m->m_len += optlen;
561 		icmp6len += optlen;
562 		bzero((caddr_t)nd_opt, optlen);
563 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
564 		nd_opt->nd_opt_len = optlen >> 3;
565 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
566 	}
567 
568 	ip6->ip6_plen = htons((u_short)icmp6len);
569 	nd_ns->nd_ns_cksum = 0;
570 	nd_ns->nd_ns_cksum =
571 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
572 
573 	if (send_sendso_input_hook != NULL) {
574 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
575 			sizeof(unsigned short), M_NOWAIT);
576 		if (mtag == NULL)
577 			goto bad;
578 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
579 		m_tag_prepend(m, mtag);
580 	}
581 
582 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
583 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
584 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
585 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
586 
587 	if (ro.ro_rt) {		/* we don't cache this route. */
588 		RTFREE(ro.ro_rt);
589 	}
590 	return;
591 
592   bad:
593 	if (ro.ro_rt) {
594 		RTFREE(ro.ro_rt);
595 	}
596 	m_freem(m);
597 	return;
598 }
599 
600 /*
601  * Neighbor advertisement input handling.
602  *
603  * Based on RFC 2461
604  * Based on RFC 2462 (duplicate address detection)
605  *
606  * the following items are not implemented yet:
607  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
608  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
609  */
610 void
611 nd6_na_input(struct mbuf *m, int off, int icmp6len)
612 {
613 	struct ifnet *ifp = m->m_pkthdr.rcvif;
614 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
615 	struct nd_neighbor_advert *nd_na;
616 	struct in6_addr daddr6 = ip6->ip6_dst;
617 	struct in6_addr taddr6;
618 	int flags;
619 	int is_router;
620 	int is_solicited;
621 	int is_override;
622 	char *lladdr = NULL;
623 	int lladdrlen = 0;
624 	int checklink = 0;
625 	struct ifaddr *ifa;
626 	struct llentry *ln = NULL;
627 	union nd_opts ndopts;
628 	struct mbuf *chain = NULL;
629 	struct m_tag *mtag;
630 	struct sockaddr_in6 sin6;
631 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
632 
633 	if (ip6->ip6_hlim != 255) {
634 		nd6log((LOG_ERR,
635 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
636 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
637 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
638 		goto bad;
639 	}
640 
641 #ifndef PULLDOWN_TEST
642 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
643 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
644 #else
645 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
646 	if (nd_na == NULL) {
647 		ICMP6STAT_INC(icp6s_tooshort);
648 		return;
649 	}
650 #endif
651 
652 	flags = nd_na->nd_na_flags_reserved;
653 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
654 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
655 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
656 
657 	taddr6 = nd_na->nd_na_target;
658 	if (in6_setscope(&taddr6, ifp, NULL))
659 		goto bad;	/* XXX: impossible */
660 
661 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
662 		nd6log((LOG_ERR,
663 		    "nd6_na_input: invalid target address %s\n",
664 		    ip6_sprintf(ip6bufs, &taddr6)));
665 		goto bad;
666 	}
667 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
668 		if (is_solicited) {
669 			nd6log((LOG_ERR,
670 			    "nd6_na_input: a solicited adv is multicasted\n"));
671 			goto bad;
672 		}
673 
674 	icmp6len -= sizeof(*nd_na);
675 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
676 	if (nd6_options(&ndopts) < 0) {
677 		nd6log((LOG_INFO,
678 		    "nd6_na_input: invalid ND option, ignored\n"));
679 		/* nd6_options have incremented stats */
680 		goto freeit;
681 	}
682 
683 	if (ndopts.nd_opts_tgt_lladdr) {
684 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
685 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
686 	}
687 
688 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
689 
690 	/*
691 	 * Target address matches one of my interface address.
692 	 *
693 	 * If my address is tentative, this means that there's somebody
694 	 * already using the same address as mine.  This indicates DAD failure.
695 	 * This is defined in RFC 2462.
696 	 *
697 	 * Otherwise, process as defined in RFC 2461.
698 	 */
699 	if (ifa
700 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
701 		ifa_free(ifa);
702 		nd6_dad_na_input(ifa);
703 		goto freeit;
704 	}
705 
706 	/* Just for safety, maybe unnecessary. */
707 	if (ifa) {
708 		ifa_free(ifa);
709 		log(LOG_ERR,
710 		    "nd6_na_input: duplicate IP6 address %s\n",
711 		    ip6_sprintf(ip6bufs, &taddr6));
712 		goto freeit;
713 	}
714 
715 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
716 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
717 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
718 		    ifp->if_addrlen, lladdrlen - 2));
719 		goto bad;
720 	}
721 
722 	/*
723 	 * If no neighbor cache entry is found, NA SHOULD silently be
724 	 * discarded.
725 	 */
726 	IF_AFDATA_LOCK(ifp);
727 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
728 	IF_AFDATA_UNLOCK(ifp);
729 	if (ln == NULL) {
730 		goto freeit;
731 	}
732 
733 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
734 		/*
735 		 * If the link-layer has address, and no lladdr option came,
736 		 * discard the packet.
737 		 */
738 		if (ifp->if_addrlen && lladdr == NULL) {
739 			goto freeit;
740 		}
741 
742 		/*
743 		 * Record link-layer address, and update the state.
744 		 */
745 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
746 		ln->la_flags |= LLE_VALID;
747 		if (is_solicited) {
748 			ln->ln_state = ND6_LLINFO_REACHABLE;
749 			ln->ln_byhint = 0;
750 			if (!ND6_LLINFO_PERMANENT(ln)) {
751 				nd6_llinfo_settimer_locked(ln,
752 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
753 			}
754 		} else {
755 			ln->ln_state = ND6_LLINFO_STALE;
756 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
757 		}
758 		if ((ln->ln_router = is_router) != 0) {
759 			/*
760 			 * This means a router's state has changed from
761 			 * non-reachable to probably reachable, and might
762 			 * affect the status of associated prefixes..
763 			 */
764 			checklink = 1;
765 		}
766 	} else {
767 		int llchange;
768 
769 		/*
770 		 * Check if the link-layer address has changed or not.
771 		 */
772 		if (lladdr == NULL)
773 			llchange = 0;
774 		else {
775 			if (ln->la_flags & LLE_VALID) {
776 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
777 					llchange = 1;
778 				else
779 					llchange = 0;
780 			} else
781 				llchange = 1;
782 		}
783 
784 		/*
785 		 * This is VERY complex.  Look at it with care.
786 		 *
787 		 * override solicit lladdr llchange	action
788 		 *					(L: record lladdr)
789 		 *
790 		 *	0	0	n	--	(2c)
791 		 *	0	0	y	n	(2b) L
792 		 *	0	0	y	y	(1)    REACHABLE->STALE
793 		 *	0	1	n	--	(2c)   *->REACHABLE
794 		 *	0	1	y	n	(2b) L *->REACHABLE
795 		 *	0	1	y	y	(1)    REACHABLE->STALE
796 		 *	1	0	n	--	(2a)
797 		 *	1	0	y	n	(2a) L
798 		 *	1	0	y	y	(2a) L *->STALE
799 		 *	1	1	n	--	(2a)   *->REACHABLE
800 		 *	1	1	y	n	(2a) L *->REACHABLE
801 		 *	1	1	y	y	(2a) L *->REACHABLE
802 		 */
803 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
804 			/*
805 			 * If state is REACHABLE, make it STALE.
806 			 * no other updates should be done.
807 			 */
808 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
809 				ln->ln_state = ND6_LLINFO_STALE;
810 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
811 			}
812 			goto freeit;
813 		} else if (is_override				   /* (2a) */
814 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
815 			|| lladdr == NULL) {			   /* (2c) */
816 			/*
817 			 * Update link-local address, if any.
818 			 */
819 			if (lladdr != NULL) {
820 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
821 				ln->la_flags |= LLE_VALID;
822 			}
823 
824 			/*
825 			 * If solicited, make the state REACHABLE.
826 			 * If not solicited and the link-layer address was
827 			 * changed, make it STALE.
828 			 */
829 			if (is_solicited) {
830 				ln->ln_state = ND6_LLINFO_REACHABLE;
831 				ln->ln_byhint = 0;
832 				if (!ND6_LLINFO_PERMANENT(ln)) {
833 					nd6_llinfo_settimer_locked(ln,
834 					    (long)ND_IFINFO(ifp)->reachable * hz);
835 				}
836 			} else {
837 				if (lladdr != NULL && llchange) {
838 					ln->ln_state = ND6_LLINFO_STALE;
839 					nd6_llinfo_settimer_locked(ln,
840 					    (long)V_nd6_gctimer * hz);
841 				}
842 			}
843 		}
844 
845 		if (ln->ln_router && !is_router) {
846 			/*
847 			 * The peer dropped the router flag.
848 			 * Remove the sender from the Default Router List and
849 			 * update the Destination Cache entries.
850 			 */
851 			struct nd_defrouter *dr;
852 			struct in6_addr *in6;
853 
854 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
855 
856 			/*
857 			 * Lock to protect the default router list.
858 			 * XXX: this might be unnecessary, since this function
859 			 * is only called under the network software interrupt
860 			 * context.  However, we keep it just for safety.
861 			 */
862 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
863 			if (dr)
864 				defrtrlist_del(dr);
865 			else if (!V_ip6_forwarding) {
866 				/*
867 				 * Even if the neighbor is not in the default
868 				 * router list, the neighbor may be used
869 				 * as a next hop for some destinations
870 				 * (e.g. redirect case). So we must
871 				 * call rt6_flush explicitly.
872 				 */
873 				rt6_flush(&ip6->ip6_src, ifp);
874 			}
875 		}
876 		ln->ln_router = is_router;
877 	}
878         /* XXX - QL
879 	 *  Does this matter?
880 	 *  rt->rt_flags &= ~RTF_REJECT;
881 	 */
882 	ln->la_asked = 0;
883 	if (ln->la_hold) {
884 		struct mbuf *m_hold, *m_hold_next;
885 
886 		/*
887 		 * reset the la_hold in advance, to explicitly
888 		 * prevent a la_hold lookup in nd6_output()
889 		 * (wouldn't happen, though...)
890 		 */
891 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
892 		    m_hold; m_hold = m_hold_next) {
893 			m_hold_next = m_hold->m_nextpkt;
894 			m_hold->m_nextpkt = NULL;
895 			/*
896 			 * we assume ifp is not a loopback here, so just set
897 			 * the 2nd argument as the 1st one.
898 			 */
899 
900 			if (send_sendso_input_hook != NULL) {
901 				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
902 				    sizeof(unsigned short), M_NOWAIT);
903 				if (mtag == NULL)
904 					goto bad;
905 				m_tag_prepend(m, mtag);
906 			}
907 
908 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
909 		}
910 	}
911  freeit:
912 	if (ln != NULL) {
913 		if (chain)
914 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
915 		LLE_WUNLOCK(ln);
916 
917 		if (chain)
918 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
919 	}
920 	if (checklink)
921 		pfxlist_onlink_check();
922 
923 	m_freem(m);
924 	return;
925 
926  bad:
927 	if (ln != NULL)
928 		LLE_WUNLOCK(ln);
929 
930 	ICMP6STAT_INC(icp6s_badna);
931 	m_freem(m);
932 }
933 
934 /*
935  * Neighbor advertisement output handling.
936  *
937  * Based on RFC 2461
938  *
939  * the following items are not implemented yet:
940  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
941  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
942  *
943  * tlladdr - 1 if include target link-layer address
944  * sdl0 - sockaddr_dl (= proxy NA) or NULL
945  */
946 void
947 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
948     const struct in6_addr *taddr6, u_long flags, int tlladdr,
949     struct sockaddr *sdl0)
950 {
951 	struct mbuf *m;
952 	struct m_tag *mtag;
953 	struct ip6_hdr *ip6;
954 	struct nd_neighbor_advert *nd_na;
955 	struct ip6_moptions im6o;
956 	struct in6_addr src, daddr6;
957 	struct sockaddr_in6 dst_sa;
958 	int icmp6len, maxlen, error;
959 	caddr_t mac = NULL;
960 	struct route_in6 ro;
961 
962 	bzero(&ro, sizeof(ro));
963 
964 	daddr6 = *daddr6_0;	/* make a local copy for modification */
965 
966 	/* estimate the size of message */
967 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
968 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
969 	if (max_linkhdr + maxlen >= MCLBYTES) {
970 #ifdef DIAGNOSTIC
971 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
972 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
973 #endif
974 		return;
975 	}
976 
977 	MGETHDR(m, M_DONTWAIT, MT_DATA);
978 	if (m && max_linkhdr + maxlen >= MHLEN) {
979 		MCLGET(m, M_DONTWAIT);
980 		if ((m->m_flags & M_EXT) == 0) {
981 			m_free(m);
982 			m = NULL;
983 		}
984 	}
985 	if (m == NULL)
986 		return;
987 	m->m_pkthdr.rcvif = NULL;
988 
989 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
990 		m->m_flags |= M_MCAST;
991 		im6o.im6o_multicast_ifp = ifp;
992 		im6o.im6o_multicast_hlim = 255;
993 		im6o.im6o_multicast_loop = 0;
994 	}
995 
996 	icmp6len = sizeof(*nd_na);
997 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
998 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
999 
1000 	/* fill neighbor advertisement packet */
1001 	ip6 = mtod(m, struct ip6_hdr *);
1002 	ip6->ip6_flow = 0;
1003 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1004 	ip6->ip6_vfc |= IPV6_VERSION;
1005 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1006 	ip6->ip6_hlim = 255;
1007 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1008 		/* reply to DAD */
1009 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1010 		daddr6.s6_addr16[1] = 0;
1011 		daddr6.s6_addr32[1] = 0;
1012 		daddr6.s6_addr32[2] = 0;
1013 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1014 		if (in6_setscope(&daddr6, ifp, NULL))
1015 			goto bad;
1016 
1017 		flags &= ~ND_NA_FLAG_SOLICITED;
1018 	}
1019 	ip6->ip6_dst = daddr6;
1020 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1021 	dst_sa.sin6_family = AF_INET6;
1022 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1023 	dst_sa.sin6_addr = daddr6;
1024 
1025 	/*
1026 	 * Select a source whose scope is the same as that of the dest.
1027 	 */
1028 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1029 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &src);
1030 	if (error) {
1031 		char ip6buf[INET6_ADDRSTRLEN];
1032 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1033 		    "determined: dst=%s, error=%d\n",
1034 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1035 		goto bad;
1036 	}
1037 	ip6->ip6_src = src;
1038 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1039 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1040 	nd_na->nd_na_code = 0;
1041 	nd_na->nd_na_target = *taddr6;
1042 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1043 
1044 	/*
1045 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1046 	 * see nd6_ns_input() for details.
1047 	 * Basically, if NS packet is sent to unicast/anycast addr,
1048 	 * target lladdr option SHOULD NOT be included.
1049 	 */
1050 	if (tlladdr) {
1051 		/*
1052 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1053 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1054 		 * my address) use lladdr configured for the interface.
1055 		 */
1056 		if (sdl0 == NULL) {
1057 			if (ifp->if_carp)
1058 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1059 			if (mac == NULL)
1060 				mac = nd6_ifptomac(ifp);
1061 		} else if (sdl0->sa_family == AF_LINK) {
1062 			struct sockaddr_dl *sdl;
1063 			sdl = (struct sockaddr_dl *)sdl0;
1064 			if (sdl->sdl_alen == ifp->if_addrlen)
1065 				mac = LLADDR(sdl);
1066 		}
1067 	}
1068 	if (tlladdr && mac) {
1069 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1070 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1071 
1072 		/* roundup to 8 bytes alignment! */
1073 		optlen = (optlen + 7) & ~7;
1074 
1075 		m->m_pkthdr.len += optlen;
1076 		m->m_len += optlen;
1077 		icmp6len += optlen;
1078 		bzero((caddr_t)nd_opt, optlen);
1079 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1080 		nd_opt->nd_opt_len = optlen >> 3;
1081 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1082 	} else
1083 		flags &= ~ND_NA_FLAG_OVERRIDE;
1084 
1085 	ip6->ip6_plen = htons((u_short)icmp6len);
1086 	nd_na->nd_na_flags_reserved = flags;
1087 	nd_na->nd_na_cksum = 0;
1088 	nd_na->nd_na_cksum =
1089 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1090 
1091 	if (send_sendso_input_hook != NULL) {
1092 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1093 		    sizeof(unsigned short), M_NOWAIT);
1094 		if (mtag == NULL)
1095 			goto bad;
1096 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1097 		m_tag_prepend(m, mtag);
1098 	}
1099 
1100 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1101 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1102 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1103 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1104 
1105 	if (ro.ro_rt) {		/* we don't cache this route. */
1106 		RTFREE(ro.ro_rt);
1107 	}
1108 	return;
1109 
1110   bad:
1111 	if (ro.ro_rt) {
1112 		RTFREE(ro.ro_rt);
1113 	}
1114 	m_freem(m);
1115 	return;
1116 }
1117 
1118 caddr_t
1119 nd6_ifptomac(struct ifnet *ifp)
1120 {
1121 	switch (ifp->if_type) {
1122 	case IFT_ARCNET:
1123 	case IFT_ETHER:
1124 	case IFT_FDDI:
1125 	case IFT_IEEE1394:
1126 #ifdef IFT_L2VLAN
1127 	case IFT_L2VLAN:
1128 #endif
1129 #ifdef IFT_IEEE80211
1130 	case IFT_IEEE80211:
1131 #endif
1132 #ifdef IFT_CARP
1133 	case IFT_CARP:
1134 #endif
1135 	case IFT_INFINIBAND:
1136 	case IFT_BRIDGE:
1137 	case IFT_ISO88025:
1138 		return IF_LLADDR(ifp);
1139 	default:
1140 		return NULL;
1141 	}
1142 }
1143 
1144 struct dadq {
1145 	TAILQ_ENTRY(dadq) dad_list;
1146 	struct ifaddr *dad_ifa;
1147 	int dad_count;		/* max NS to send */
1148 	int dad_ns_tcount;	/* # of trials to send NS */
1149 	int dad_ns_ocount;	/* NS sent so far */
1150 	int dad_ns_icount;
1151 	int dad_na_icount;
1152 	struct callout dad_timer_ch;
1153 	struct vnet *dad_vnet;
1154 };
1155 
1156 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1157 VNET_DEFINE(int, dad_init) = 0;
1158 #define	V_dadq				VNET(dadq)
1159 #define	V_dad_init			VNET(dad_init)
1160 
1161 static struct dadq *
1162 nd6_dad_find(struct ifaddr *ifa)
1163 {
1164 	struct dadq *dp;
1165 
1166 	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1167 		if (dp->dad_ifa == ifa)
1168 			return dp;
1169 	}
1170 	return NULL;
1171 }
1172 
1173 static void
1174 nd6_dad_starttimer(struct dadq *dp, int ticks)
1175 {
1176 
1177 	callout_reset(&dp->dad_timer_ch, ticks,
1178 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1179 }
1180 
1181 static void
1182 nd6_dad_stoptimer(struct dadq *dp)
1183 {
1184 
1185 	callout_stop(&dp->dad_timer_ch);
1186 }
1187 
1188 /*
1189  * Start Duplicate Address Detection (DAD) for specified interface address.
1190  */
1191 void
1192 nd6_dad_start(struct ifaddr *ifa, int delay)
1193 {
1194 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1195 	struct dadq *dp;
1196 	char ip6buf[INET6_ADDRSTRLEN];
1197 
1198 	if (!V_dad_init) {
1199 		TAILQ_INIT(&V_dadq);
1200 		V_dad_init++;
1201 	}
1202 
1203 	/*
1204 	 * If we don't need DAD, don't do it.
1205 	 * There are several cases:
1206 	 * - DAD is disabled (ip6_dad_count == 0)
1207 	 * - the interface address is anycast
1208 	 */
1209 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1210 		log(LOG_DEBUG,
1211 			"nd6_dad_start: called with non-tentative address "
1212 			"%s(%s)\n",
1213 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1214 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1215 		return;
1216 	}
1217 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1218 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1219 		return;
1220 	}
1221 	if (!V_ip6_dad_count) {
1222 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1223 		return;
1224 	}
1225 	if (ifa->ifa_ifp == NULL)
1226 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1227 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1228 		return;
1229 	}
1230 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1231 		return;
1232 	if (nd6_dad_find(ifa) != NULL) {
1233 		/* DAD already in progress */
1234 		return;
1235 	}
1236 
1237 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1238 	if (dp == NULL) {
1239 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1240 			"%s(%s)\n",
1241 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1242 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1243 		return;
1244 	}
1245 	bzero(dp, sizeof(*dp));
1246 	callout_init(&dp->dad_timer_ch, 0);
1247 #ifdef VIMAGE
1248 	dp->dad_vnet = curvnet;
1249 #endif
1250 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1251 
1252 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1253 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1254 
1255 	/*
1256 	 * Send NS packet for DAD, ip6_dad_count times.
1257 	 * Note that we must delay the first transmission, if this is the
1258 	 * first packet to be sent from the interface after interface
1259 	 * (re)initialization.
1260 	 */
1261 	dp->dad_ifa = ifa;
1262 	ifa_ref(ifa);	/* just for safety */
1263 	dp->dad_count = V_ip6_dad_count;
1264 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1265 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1266 	if (delay == 0) {
1267 		nd6_dad_ns_output(dp, ifa);
1268 		nd6_dad_starttimer(dp,
1269 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1270 	} else {
1271 		nd6_dad_starttimer(dp, delay);
1272 	}
1273 }
1274 
1275 /*
1276  * terminate DAD unconditionally.  used for address removals.
1277  */
1278 void
1279 nd6_dad_stop(struct ifaddr *ifa)
1280 {
1281 	struct dadq *dp;
1282 
1283 	if (!V_dad_init)
1284 		return;
1285 	dp = nd6_dad_find(ifa);
1286 	if (!dp) {
1287 		/* DAD wasn't started yet */
1288 		return;
1289 	}
1290 
1291 	nd6_dad_stoptimer(dp);
1292 
1293 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1294 	free(dp, M_IP6NDP);
1295 	dp = NULL;
1296 	ifa_free(ifa);
1297 }
1298 
1299 static void
1300 nd6_dad_timer(struct dadq *dp)
1301 {
1302 	CURVNET_SET(dp->dad_vnet);
1303 	int s;
1304 	struct ifaddr *ifa = dp->dad_ifa;
1305 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1306 	char ip6buf[INET6_ADDRSTRLEN];
1307 
1308 	s = splnet();		/* XXX */
1309 
1310 	/* Sanity check */
1311 	if (ia == NULL) {
1312 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1313 		goto done;
1314 	}
1315 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1316 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1317 			"%s(%s)\n",
1318 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1319 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1320 		goto done;
1321 	}
1322 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1323 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1324 			"%s(%s)\n",
1325 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1326 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1327 		goto done;
1328 	}
1329 
1330 	/* timeouted with IFF_{RUNNING,UP} check */
1331 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1332 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1333 		    if_name(ifa->ifa_ifp)));
1334 
1335 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1336 		free(dp, M_IP6NDP);
1337 		dp = NULL;
1338 		ifa_free(ifa);
1339 		goto done;
1340 	}
1341 
1342 	/* Need more checks? */
1343 	if (dp->dad_ns_ocount < dp->dad_count) {
1344 		/*
1345 		 * We have more NS to go.  Send NS packet for DAD.
1346 		 */
1347 		nd6_dad_ns_output(dp, ifa);
1348 		nd6_dad_starttimer(dp,
1349 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1350 	} else {
1351 		/*
1352 		 * We have transmitted sufficient number of DAD packets.
1353 		 * See what we've got.
1354 		 */
1355 		int duplicate;
1356 
1357 		duplicate = 0;
1358 
1359 		if (dp->dad_na_icount) {
1360 			/*
1361 			 * the check is in nd6_dad_na_input(),
1362 			 * but just in case
1363 			 */
1364 			duplicate++;
1365 		}
1366 
1367 		if (dp->dad_ns_icount) {
1368 			/* We've seen NS, means DAD has failed. */
1369 			duplicate++;
1370 		}
1371 
1372 		if (duplicate) {
1373 			/* (*dp) will be freed in nd6_dad_duplicated() */
1374 			dp = NULL;
1375 			nd6_dad_duplicated(ifa);
1376 		} else {
1377 			/*
1378 			 * We are done with DAD.  No NA came, no NS came.
1379 			 * No duplicate address found.
1380 			 */
1381 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1382 
1383 			nd6log((LOG_DEBUG,
1384 			    "%s: DAD complete for %s - no duplicates found\n",
1385 			    if_name(ifa->ifa_ifp),
1386 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1387 
1388 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1389 			free(dp, M_IP6NDP);
1390 			dp = NULL;
1391 			ifa_free(ifa);
1392 		}
1393 	}
1394 
1395 done:
1396 	splx(s);
1397 	CURVNET_RESTORE();
1398 }
1399 
1400 void
1401 nd6_dad_duplicated(struct ifaddr *ifa)
1402 {
1403 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1404 	struct ifnet *ifp;
1405 	struct dadq *dp;
1406 	char ip6buf[INET6_ADDRSTRLEN];
1407 
1408 	dp = nd6_dad_find(ifa);
1409 	if (dp == NULL) {
1410 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1411 		return;
1412 	}
1413 
1414 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1415 	    "NS in/out=%d/%d, NA in=%d\n",
1416 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1417 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1418 
1419 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1420 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1421 
1422 	/* We are done with DAD, with duplicate address found. (failure) */
1423 	nd6_dad_stoptimer(dp);
1424 
1425 	ifp = ifa->ifa_ifp;
1426 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1427 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1428 	log(LOG_ERR, "%s: manual intervention required\n",
1429 	    if_name(ifp));
1430 
1431 	/*
1432 	 * If the address is a link-local address formed from an interface
1433 	 * identifier based on the hardware address which is supposed to be
1434 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1435 	 * operation on the interface SHOULD be disabled.
1436 	 * [RFC 4862, Section 5.4.5]
1437 	 */
1438 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1439 		struct in6_addr in6;
1440 
1441 		/*
1442 		 * To avoid over-reaction, we only apply this logic when we are
1443 		 * very sure that hardware addresses are supposed to be unique.
1444 		 */
1445 		switch (ifp->if_type) {
1446 		case IFT_ETHER:
1447 		case IFT_FDDI:
1448 		case IFT_ATM:
1449 		case IFT_IEEE1394:
1450 #ifdef IFT_IEEE80211
1451 		case IFT_IEEE80211:
1452 #endif
1453 		case IFT_INFINIBAND:
1454 			in6 = ia->ia_addr.sin6_addr;
1455 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1456 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1457 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1458 				log(LOG_ERR, "%s: possible hardware address "
1459 				    "duplication detected, disable IPv6\n",
1460 				    if_name(ifp));
1461 			}
1462 			break;
1463 		}
1464 	}
1465 
1466 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1467 	free(dp, M_IP6NDP);
1468 	dp = NULL;
1469 	ifa_free(ifa);
1470 }
1471 
1472 static void
1473 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1474 {
1475 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1476 	struct ifnet *ifp = ifa->ifa_ifp;
1477 
1478 	dp->dad_ns_tcount++;
1479 	if ((ifp->if_flags & IFF_UP) == 0) {
1480 		return;
1481 	}
1482 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1483 		return;
1484 	}
1485 
1486 	dp->dad_ns_ocount++;
1487 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1488 }
1489 
1490 static void
1491 nd6_dad_ns_input(struct ifaddr *ifa)
1492 {
1493 	struct in6_ifaddr *ia;
1494 	struct ifnet *ifp;
1495 	const struct in6_addr *taddr6;
1496 	struct dadq *dp;
1497 	int duplicate;
1498 
1499 	if (ifa == NULL)
1500 		panic("ifa == NULL in nd6_dad_ns_input");
1501 
1502 	ia = (struct in6_ifaddr *)ifa;
1503 	ifp = ifa->ifa_ifp;
1504 	taddr6 = &ia->ia_addr.sin6_addr;
1505 	duplicate = 0;
1506 	dp = nd6_dad_find(ifa);
1507 
1508 	/* Quickhack - completely ignore DAD NS packets */
1509 	if (V_dad_ignore_ns) {
1510 		char ip6buf[INET6_ADDRSTRLEN];
1511 		nd6log((LOG_INFO,
1512 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1513 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1514 		    if_name(ifa->ifa_ifp)));
1515 		return;
1516 	}
1517 
1518 	/*
1519 	 * if I'm yet to start DAD, someone else started using this address
1520 	 * first.  I have a duplicate and you win.
1521 	 */
1522 	if (dp == NULL || dp->dad_ns_ocount == 0)
1523 		duplicate++;
1524 
1525 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1526 
1527 	if (duplicate) {
1528 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1529 		nd6_dad_duplicated(ifa);
1530 	} else {
1531 		/*
1532 		 * not sure if I got a duplicate.
1533 		 * increment ns count and see what happens.
1534 		 */
1535 		if (dp)
1536 			dp->dad_ns_icount++;
1537 	}
1538 }
1539 
1540 static void
1541 nd6_dad_na_input(struct ifaddr *ifa)
1542 {
1543 	struct dadq *dp;
1544 
1545 	if (ifa == NULL)
1546 		panic("ifa == NULL in nd6_dad_na_input");
1547 
1548 	dp = nd6_dad_find(ifa);
1549 	if (dp)
1550 		dp->dad_na_icount++;
1551 
1552 	/* remove the address. */
1553 	nd6_dad_duplicated(ifa);
1554 }
1555