xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 9fe48b8076ae9c6dc486d713c468ff03ec056eda)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
72 #include <netinet6/in6_var.h>
73 #include <netinet6/in6_ifattach.h>
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
76 #include <netinet6/scope6_var.h>
77 #include <netinet6/nd6.h>
78 #include <netinet/icmp6.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet6/send.h>
81 
82 #define SDL(s) ((struct sockaddr_dl *)s)
83 
84 struct dadq;
85 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
86 static void nd6_dad_add(struct dadq *dp);
87 static void nd6_dad_del(struct dadq *dp);
88 static void nd6_dad_rele(struct dadq *);
89 static void nd6_dad_starttimer(struct dadq *, int);
90 static void nd6_dad_stoptimer(struct dadq *);
91 static void nd6_dad_timer(struct dadq *);
92 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
93 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
94 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
95 static void nd6_dad_na_input(struct ifaddr *);
96 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
97     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
98 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
99     const struct in6_addr *, struct llentry *, uint8_t *, u_int);
100 
101 static VNET_DEFINE(int, dad_enhanced) = 1;
102 #define	V_dad_enhanced			VNET(dad_enhanced)
103 
104 SYSCTL_DECL(_net_inet6_ip6);
105 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
106     &VNET_NAME(dad_enhanced), 0,
107     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
108 
109 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
110 						   transmit DAD packet */
111 #define	V_dad_maxtry			VNET(dad_maxtry)
112 
113 /*
114  * Input a Neighbor Solicitation Message.
115  *
116  * Based on RFC 2461
117  * Based on RFC 2462 (duplicate address detection)
118  */
119 void
120 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
121 {
122 	struct ifnet *ifp = m->m_pkthdr.rcvif;
123 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
124 	struct nd_neighbor_solicit *nd_ns;
125 	struct in6_addr saddr6 = ip6->ip6_src;
126 	struct in6_addr daddr6 = ip6->ip6_dst;
127 	struct in6_addr taddr6;
128 	struct in6_addr myaddr6;
129 	char *lladdr = NULL;
130 	struct ifaddr *ifa = NULL;
131 	int lladdrlen = 0;
132 	int anycast = 0, proxy = 0, tentative = 0;
133 	int tlladdr;
134 	int rflag;
135 	union nd_opts ndopts;
136 	struct sockaddr_dl proxydl;
137 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
138 
139 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
140 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
141 		rflag = 0;
142 #ifndef PULLDOWN_TEST
143 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
144 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
145 #else
146 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
147 	if (nd_ns == NULL) {
148 		ICMP6STAT_INC(icp6s_tooshort);
149 		return;
150 	}
151 #endif
152 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
153 	taddr6 = nd_ns->nd_ns_target;
154 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
155 		goto bad;
156 
157 	if (ip6->ip6_hlim != 255) {
158 		nd6log((LOG_ERR,
159 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
160 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
161 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
162 		goto bad;
163 	}
164 
165 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
166 		/* dst has to be a solicited node multicast address. */
167 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
168 		    /* don't check ifindex portion */
169 		    daddr6.s6_addr32[1] == 0 &&
170 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
171 		    daddr6.s6_addr8[12] == 0xff) {
172 			; /* good */
173 		} else {
174 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
175 			    "(wrong ip6 dst)\n"));
176 			goto bad;
177 		}
178 	} else if (!V_nd6_onlink_ns_rfc4861) {
179 		struct sockaddr_in6 src_sa6;
180 
181 		/*
182 		 * According to recent IETF discussions, it is not a good idea
183 		 * to accept a NS from an address which would not be deemed
184 		 * to be a neighbor otherwise.  This point is expected to be
185 		 * clarified in future revisions of the specification.
186 		 */
187 		bzero(&src_sa6, sizeof(src_sa6));
188 		src_sa6.sin6_family = AF_INET6;
189 		src_sa6.sin6_len = sizeof(src_sa6);
190 		src_sa6.sin6_addr = saddr6;
191 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
192 			nd6log((LOG_INFO, "nd6_ns_input: "
193 				"NS packet from non-neighbor\n"));
194 			goto bad;
195 		}
196 	}
197 
198 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
199 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
200 		goto bad;
201 	}
202 
203 	icmp6len -= sizeof(*nd_ns);
204 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
205 	if (nd6_options(&ndopts) < 0) {
206 		nd6log((LOG_INFO,
207 		    "nd6_ns_input: invalid ND option, ignored\n"));
208 		/* nd6_options have incremented stats */
209 		goto freeit;
210 	}
211 
212 	if (ndopts.nd_opts_src_lladdr) {
213 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
214 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
215 	}
216 
217 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
218 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
219 		    "(link-layer address option)\n"));
220 		goto bad;
221 	}
222 
223 	/*
224 	 * Attaching target link-layer address to the NA?
225 	 * (RFC 2461 7.2.4)
226 	 *
227 	 * NS IP dst is unicast/anycast			MUST NOT add
228 	 * NS IP dst is solicited-node multicast	MUST add
229 	 *
230 	 * In implementation, we add target link-layer address by default.
231 	 * We do not add one in MUST NOT cases.
232 	 */
233 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
234 		tlladdr = 0;
235 	else
236 		tlladdr = 1;
237 
238 	/*
239 	 * Target address (taddr6) must be either:
240 	 * (1) Valid unicast/anycast address for my receiving interface,
241 	 * (2) Unicast address for which I'm offering proxy service, or
242 	 * (3) "tentative" address on which DAD is being performed.
243 	 */
244 	/* (1) and (3) check. */
245 	if (ifp->if_carp)
246 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
247 	else
248 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
249 
250 	/* (2) check. */
251 	if (ifa == NULL) {
252 		struct route_in6 ro;
253 		int need_proxy;
254 
255 		bzero(&ro, sizeof(ro));
256 		ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
257 		ro.ro_dst.sin6_family = AF_INET6;
258 		ro.ro_dst.sin6_addr = taddr6;
259 
260 		/* Always use the default FIB. */
261 #ifdef RADIX_MPATH
262 		rtalloc_mpath_fib((struct route *)&ro, ntohl(taddr6.s6_addr32[3]),
263 		    RT_DEFAULT_FIB);
264 #else
265 		in6_rtalloc(&ro, RT_DEFAULT_FIB);
266 #endif
267 		need_proxy = (ro.ro_rt &&
268 		    (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
269 		    ro.ro_rt->rt_gateway->sa_family == AF_LINK);
270 		if (ro.ro_rt != NULL) {
271 			if (need_proxy)
272 				proxydl = *SDL(ro.ro_rt->rt_gateway);
273 			RTFREE(ro.ro_rt);
274 		}
275 		if (need_proxy) {
276 			/*
277 			 * proxy NDP for single entry
278 			 */
279 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
280 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
281 			if (ifa)
282 				proxy = 1;
283 		}
284 	}
285 	if (ifa == NULL) {
286 		/*
287 		 * We've got an NS packet, and we don't have that adddress
288 		 * assigned for us.  We MUST silently ignore it.
289 		 * See RFC2461 7.2.3.
290 		 */
291 		goto freeit;
292 	}
293 	myaddr6 = *IFA_IN6(ifa);
294 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
295 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
296 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
297 		goto freeit;
298 
299 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
300 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
301 		    "(if %d, NS packet %d)\n",
302 		    ip6_sprintf(ip6bufs, &taddr6),
303 		    ifp->if_addrlen, lladdrlen - 2));
304 		goto bad;
305 	}
306 
307 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
308 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
309 		    ip6_sprintf(ip6bufs, &saddr6)));
310 		goto freeit;
311 	}
312 
313 	/*
314 	 * We have neighbor solicitation packet, with target address equals to
315 	 * one of my tentative address.
316 	 *
317 	 * src addr	how to process?
318 	 * ---		---
319 	 * multicast	of course, invalid (rejected in ip6_input)
320 	 * unicast	somebody is doing address resolution -> ignore
321 	 * unspec	dup address detection
322 	 *
323 	 * The processing is defined in RFC 2462.
324 	 */
325 	if (tentative) {
326 		/*
327 		 * If source address is unspecified address, it is for
328 		 * duplicate address detection.
329 		 *
330 		 * If not, the packet is for addess resolution;
331 		 * silently ignore it.
332 		 */
333 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
334 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
335 
336 		goto freeit;
337 	}
338 
339 	/*
340 	 * If the source address is unspecified address, entries must not
341 	 * be created or updated.
342 	 * It looks that sender is performing DAD.  Output NA toward
343 	 * all-node multicast address, to tell the sender that I'm using
344 	 * the address.
345 	 * S bit ("solicited") must be zero.
346 	 */
347 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
348 		struct in6_addr in6_all;
349 
350 		in6_all = in6addr_linklocal_allnodes;
351 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
352 			goto bad;
353 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
354 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
356 		    M_GETFIB(m));
357 		goto freeit;
358 	}
359 
360 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
361 	    ND_NEIGHBOR_SOLICIT, 0);
362 
363 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
364 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
365 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
366 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
367  freeit:
368 	if (ifa != NULL)
369 		ifa_free(ifa);
370 	m_freem(m);
371 	return;
372 
373  bad:
374 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
375 		ip6_sprintf(ip6bufs, &saddr6)));
376 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
377 		ip6_sprintf(ip6bufs, &daddr6)));
378 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
379 		ip6_sprintf(ip6bufs, &taddr6)));
380 	ICMP6STAT_INC(icp6s_badns);
381 	if (ifa != NULL)
382 		ifa_free(ifa);
383 	m_freem(m);
384 }
385 
386 /*
387  * Output a Neighbor Solicitation Message. Caller specifies:
388  *	- ICMP6 header source IP6 address
389  *	- ND6 header target IP6 address
390  *	- ND6 header source datalink address
391  *
392  * Based on RFC 2461
393  * Based on RFC 2462 (duplicate address detection)
394  *
395  *    ln - for source address determination
396  * nonce - If non-NULL, NS is used for duplicate address detection and
397  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
398  */
399 static void
400 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6,
401     const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce,
402     u_int fibnum)
403 {
404 	struct mbuf *m;
405 	struct m_tag *mtag;
406 	struct ip6_hdr *ip6;
407 	struct nd_neighbor_solicit *nd_ns;
408 	struct ip6_moptions im6o;
409 	int icmp6len;
410 	int maxlen;
411 	caddr_t mac;
412 	struct route_in6 ro;
413 
414 	if (IN6_IS_ADDR_MULTICAST(taddr6))
415 		return;
416 
417 	/* estimate the size of message */
418 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
419 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
420 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
421 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
422 	    __func__, max_linkhdr, maxlen, MCLBYTES));
423 
424 	if (max_linkhdr + maxlen > MHLEN)
425 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
426 	else
427 		m = m_gethdr(M_NOWAIT, MT_DATA);
428 	if (m == NULL)
429 		return;
430 	M_SETFIB(m, fibnum);
431 
432 	bzero(&ro, sizeof(ro));
433 
434 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
435 		m->m_flags |= M_MCAST;
436 		im6o.im6o_multicast_ifp = ifp;
437 		im6o.im6o_multicast_hlim = 255;
438 		im6o.im6o_multicast_loop = 0;
439 	}
440 
441 	icmp6len = sizeof(*nd_ns);
442 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
443 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
444 
445 	/* fill neighbor solicitation packet */
446 	ip6 = mtod(m, struct ip6_hdr *);
447 	ip6->ip6_flow = 0;
448 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
449 	ip6->ip6_vfc |= IPV6_VERSION;
450 	/* ip6->ip6_plen will be set later */
451 	ip6->ip6_nxt = IPPROTO_ICMPV6;
452 	ip6->ip6_hlim = 255;
453 	if (daddr6)
454 		ip6->ip6_dst = *daddr6;
455 	else {
456 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
457 		ip6->ip6_dst.s6_addr16[1] = 0;
458 		ip6->ip6_dst.s6_addr32[1] = 0;
459 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
460 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
461 		ip6->ip6_dst.s6_addr8[12] = 0xff;
462 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
463 			goto bad;
464 	}
465 	if (nonce == NULL) {
466 		struct ifaddr *ifa;
467 
468 		/*
469 		 * RFC2461 7.2.2:
470 		 * "If the source address of the packet prompting the
471 		 * solicitation is the same as one of the addresses assigned
472 		 * to the outgoing interface, that address SHOULD be placed
473 		 * in the IP Source Address of the outgoing solicitation.
474 		 * Otherwise, any one of the addresses assigned to the
475 		 * interface should be used."
476 		 *
477 		 * We use the source address for the prompting packet
478 		 * (saddr6), if:
479 		 * - saddr6 is given from the caller (by giving "ln"), and
480 		 * - saddr6 belongs to the outgoing interface.
481 		 * Otherwise, we perform the source address selection as usual.
482 		 */
483 		struct in6_addr *hsrc;
484 
485 		hsrc = NULL;
486 		if (ln != NULL) {
487 			LLE_RLOCK(ln);
488 			if (ln->la_hold != NULL) {
489 				struct ip6_hdr *hip6;		/* hold ip6 */
490 
491 				/*
492 				 * assuming every packet in la_hold has the same IP
493 				 * header
494 				 */
495 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
496 				/* XXX pullup? */
497 				if (sizeof(*hip6) < ln->la_hold->m_len) {
498 					ip6->ip6_src = hip6->ip6_src;
499 					hsrc = &hip6->ip6_src;
500 				}
501 			}
502 			LLE_RUNLOCK(ln);
503 		}
504 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
505 		    hsrc)) != NULL) {
506 			/* ip6_src set already. */
507 			ifa_free(ifa);
508 		} else {
509 			int error;
510 			struct sockaddr_in6 dst_sa;
511 			struct in6_addr src_in;
512 			struct ifnet *oifp;
513 
514 			bzero(&dst_sa, sizeof(dst_sa));
515 			dst_sa.sin6_family = AF_INET6;
516 			dst_sa.sin6_len = sizeof(dst_sa);
517 			dst_sa.sin6_addr = ip6->ip6_dst;
518 
519 			oifp = ifp;
520 			error = in6_selectsrc(&dst_sa, NULL,
521 			    NULL, &ro, NULL, &oifp, &src_in);
522 			if (error) {
523 				char ip6buf[INET6_ADDRSTRLEN];
524 				nd6log((LOG_DEBUG, "%s: source can't be "
525 				    "determined: dst=%s, error=%d\n", __func__,
526 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
527 				    error));
528 				goto bad;
529 			}
530 			ip6->ip6_src = src_in;
531 		}
532 	} else {
533 		/*
534 		 * Source address for DAD packet must always be IPv6
535 		 * unspecified address. (0::0)
536 		 * We actually don't have to 0-clear the address (we did it
537 		 * above), but we do so here explicitly to make the intention
538 		 * clearer.
539 		 */
540 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
541 	}
542 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
543 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
544 	nd_ns->nd_ns_code = 0;
545 	nd_ns->nd_ns_reserved = 0;
546 	nd_ns->nd_ns_target = *taddr6;
547 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
548 
549 	/*
550 	 * Add source link-layer address option.
551 	 *
552 	 *				spec		implementation
553 	 *				---		---
554 	 * DAD packet			MUST NOT	do not add the option
555 	 * there's no link layer address:
556 	 *				impossible	do not add the option
557 	 * there's link layer address:
558 	 *	Multicast NS		MUST add one	add the option
559 	 *	Unicast NS		SHOULD add one	add the option
560 	 */
561 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
562 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
563 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
564 		/* 8 byte alignments... */
565 		optlen = (optlen + 7) & ~7;
566 
567 		m->m_pkthdr.len += optlen;
568 		m->m_len += optlen;
569 		icmp6len += optlen;
570 		bzero((caddr_t)nd_opt, optlen);
571 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
572 		nd_opt->nd_opt_len = optlen >> 3;
573 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
574 	}
575 	/*
576 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
577 	 * This behavior is documented as Enhanced Duplicate Address
578 	 * Detection in draft-ietf-6man-enhanced-dad-13.
579 	 * net.inet6.ip6.dad_enhanced=0 disables this.
580 	 */
581 	if (V_dad_enhanced != 0 && nonce != NULL) {
582 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
583 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
584 		/* 8-byte alignment is required. */
585 		optlen = (optlen + 7) & ~7;
586 
587 		m->m_pkthdr.len += optlen;
588 		m->m_len += optlen;
589 		icmp6len += optlen;
590 		bzero((caddr_t)nd_opt, optlen);
591 		nd_opt->nd_opt_type = ND_OPT_NONCE;
592 		nd_opt->nd_opt_len = optlen >> 3;
593 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
594 	}
595 	ip6->ip6_plen = htons((u_short)icmp6len);
596 	nd_ns->nd_ns_cksum = 0;
597 	nd_ns->nd_ns_cksum =
598 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
599 
600 	if (send_sendso_input_hook != NULL) {
601 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
602 			sizeof(unsigned short), M_NOWAIT);
603 		if (mtag == NULL)
604 			goto bad;
605 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
606 		m_tag_prepend(m, mtag);
607 	}
608 
609 	ip6_output(m, NULL, &ro, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
610 	    &im6o, NULL, NULL);
611 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
612 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
613 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
614 
615 	/* We don't cache this route. */
616 	RO_RTFREE(&ro);
617 
618 	return;
619 
620   bad:
621 	if (ro.ro_rt) {
622 		RTFREE(ro.ro_rt);
623 	}
624 	m_freem(m);
625 	return;
626 }
627 
628 #ifndef BURN_BRIDGES
629 void
630 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
631     const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce)
632 {
633 
634 	nd6_ns_output_fib(ifp, daddr6, taddr6, ln, nonce, RT_DEFAULT_FIB);
635 }
636 #endif
637 /*
638  * Neighbor advertisement input handling.
639  *
640  * Based on RFC 2461
641  * Based on RFC 2462 (duplicate address detection)
642  *
643  * the following items are not implemented yet:
644  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
645  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
646  */
647 void
648 nd6_na_input(struct mbuf *m, int off, int icmp6len)
649 {
650 	struct ifnet *ifp = m->m_pkthdr.rcvif;
651 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
652 	struct nd_neighbor_advert *nd_na;
653 	struct in6_addr daddr6 = ip6->ip6_dst;
654 	struct in6_addr taddr6;
655 	int flags;
656 	int is_router;
657 	int is_solicited;
658 	int is_override;
659 	char *lladdr = NULL;
660 	int lladdrlen = 0;
661 	int checklink = 0;
662 	struct ifaddr *ifa;
663 	struct llentry *ln = NULL;
664 	union nd_opts ndopts;
665 	struct mbuf *chain = NULL;
666 	struct sockaddr_in6 sin6;
667 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
668 
669 	if (ip6->ip6_hlim != 255) {
670 		nd6log((LOG_ERR,
671 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
672 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
673 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
674 		goto bad;
675 	}
676 
677 #ifndef PULLDOWN_TEST
678 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
679 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
680 #else
681 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
682 	if (nd_na == NULL) {
683 		ICMP6STAT_INC(icp6s_tooshort);
684 		return;
685 	}
686 #endif
687 
688 	flags = nd_na->nd_na_flags_reserved;
689 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
690 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
691 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
692 	memset(&sin6, 0, sizeof(sin6));
693 
694 	taddr6 = nd_na->nd_na_target;
695 	if (in6_setscope(&taddr6, ifp, NULL))
696 		goto bad;	/* XXX: impossible */
697 
698 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
699 		nd6log((LOG_ERR,
700 		    "nd6_na_input: invalid target address %s\n",
701 		    ip6_sprintf(ip6bufs, &taddr6)));
702 		goto bad;
703 	}
704 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
705 		if (is_solicited) {
706 			nd6log((LOG_ERR,
707 			    "nd6_na_input: a solicited adv is multicasted\n"));
708 			goto bad;
709 		}
710 
711 	icmp6len -= sizeof(*nd_na);
712 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
713 	if (nd6_options(&ndopts) < 0) {
714 		nd6log((LOG_INFO,
715 		    "nd6_na_input: invalid ND option, ignored\n"));
716 		/* nd6_options have incremented stats */
717 		goto freeit;
718 	}
719 
720 	if (ndopts.nd_opts_tgt_lladdr) {
721 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
722 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
723 	}
724 
725 	/*
726 	 * This effectively disables the DAD check on a non-master CARP
727 	 * address.
728 	 */
729 	if (ifp->if_carp)
730 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
731 	else
732 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
733 
734 	/*
735 	 * Target address matches one of my interface address.
736 	 *
737 	 * If my address is tentative, this means that there's somebody
738 	 * already using the same address as mine.  This indicates DAD failure.
739 	 * This is defined in RFC 2462.
740 	 *
741 	 * Otherwise, process as defined in RFC 2461.
742 	 */
743 	if (ifa
744 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
745 		nd6_dad_na_input(ifa);
746 		ifa_free(ifa);
747 		goto freeit;
748 	}
749 
750 	/* Just for safety, maybe unnecessary. */
751 	if (ifa) {
752 		ifa_free(ifa);
753 		log(LOG_ERR,
754 		    "nd6_na_input: duplicate IP6 address %s\n",
755 		    ip6_sprintf(ip6bufs, &taddr6));
756 		goto freeit;
757 	}
758 
759 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
760 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
761 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
762 		    ifp->if_addrlen, lladdrlen - 2));
763 		goto bad;
764 	}
765 
766 	/*
767 	 * If no neighbor cache entry is found, NA SHOULD silently be
768 	 * discarded.
769 	 */
770 	IF_AFDATA_RLOCK(ifp);
771 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
772 	IF_AFDATA_RUNLOCK(ifp);
773 	if (ln == NULL) {
774 		goto freeit;
775 	}
776 
777 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
778 		/*
779 		 * If the link-layer has address, and no lladdr option came,
780 		 * discard the packet.
781 		 */
782 		if (ifp->if_addrlen && lladdr == NULL) {
783 			goto freeit;
784 		}
785 
786 		/*
787 		 * Record link-layer address, and update the state.
788 		 */
789 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
790 		ln->la_flags |= LLE_VALID;
791 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
792 		if (is_solicited) {
793 			ln->ln_state = ND6_LLINFO_REACHABLE;
794 			ln->ln_byhint = 0;
795 			if (!ND6_LLINFO_PERMANENT(ln)) {
796 				nd6_llinfo_settimer_locked(ln,
797 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
798 			}
799 		} else {
800 			ln->ln_state = ND6_LLINFO_STALE;
801 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
802 		}
803 		if ((ln->ln_router = is_router) != 0) {
804 			/*
805 			 * This means a router's state has changed from
806 			 * non-reachable to probably reachable, and might
807 			 * affect the status of associated prefixes..
808 			 */
809 			checklink = 1;
810 		}
811 	} else {
812 		int llchange;
813 
814 		/*
815 		 * Check if the link-layer address has changed or not.
816 		 */
817 		if (lladdr == NULL)
818 			llchange = 0;
819 		else {
820 			if (ln->la_flags & LLE_VALID) {
821 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
822 					llchange = 1;
823 				else
824 					llchange = 0;
825 			} else
826 				llchange = 1;
827 		}
828 
829 		/*
830 		 * This is VERY complex.  Look at it with care.
831 		 *
832 		 * override solicit lladdr llchange	action
833 		 *					(L: record lladdr)
834 		 *
835 		 *	0	0	n	--	(2c)
836 		 *	0	0	y	n	(2b) L
837 		 *	0	0	y	y	(1)    REACHABLE->STALE
838 		 *	0	1	n	--	(2c)   *->REACHABLE
839 		 *	0	1	y	n	(2b) L *->REACHABLE
840 		 *	0	1	y	y	(1)    REACHABLE->STALE
841 		 *	1	0	n	--	(2a)
842 		 *	1	0	y	n	(2a) L
843 		 *	1	0	y	y	(2a) L *->STALE
844 		 *	1	1	n	--	(2a)   *->REACHABLE
845 		 *	1	1	y	n	(2a) L *->REACHABLE
846 		 *	1	1	y	y	(2a) L *->REACHABLE
847 		 */
848 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
849 			/*
850 			 * If state is REACHABLE, make it STALE.
851 			 * no other updates should be done.
852 			 */
853 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
854 				ln->ln_state = ND6_LLINFO_STALE;
855 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
856 			}
857 			goto freeit;
858 		} else if (is_override				   /* (2a) */
859 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
860 			|| lladdr == NULL) {			   /* (2c) */
861 			/*
862 			 * Update link-local address, if any.
863 			 */
864 			if (lladdr != NULL) {
865 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
866 				ln->la_flags |= LLE_VALID;
867 				EVENTHANDLER_INVOKE(lle_event, ln,
868 				    LLENTRY_RESOLVED);
869 			}
870 
871 			/*
872 			 * If solicited, make the state REACHABLE.
873 			 * If not solicited and the link-layer address was
874 			 * changed, make it STALE.
875 			 */
876 			if (is_solicited) {
877 				ln->ln_state = ND6_LLINFO_REACHABLE;
878 				ln->ln_byhint = 0;
879 				if (!ND6_LLINFO_PERMANENT(ln)) {
880 					nd6_llinfo_settimer_locked(ln,
881 					    (long)ND_IFINFO(ifp)->reachable * hz);
882 				}
883 			} else {
884 				if (lladdr != NULL && llchange) {
885 					ln->ln_state = ND6_LLINFO_STALE;
886 					nd6_llinfo_settimer_locked(ln,
887 					    (long)V_nd6_gctimer * hz);
888 				}
889 			}
890 		}
891 
892 		if (ln->ln_router && !is_router) {
893 			/*
894 			 * The peer dropped the router flag.
895 			 * Remove the sender from the Default Router List and
896 			 * update the Destination Cache entries.
897 			 */
898 			struct nd_defrouter *dr;
899 			struct in6_addr *in6;
900 
901 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
902 
903 			/*
904 			 * Lock to protect the default router list.
905 			 * XXX: this might be unnecessary, since this function
906 			 * is only called under the network software interrupt
907 			 * context.  However, we keep it just for safety.
908 			 */
909 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
910 			if (dr)
911 				defrtrlist_del(dr);
912 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
913 			    ND6_IFF_ACCEPT_RTADV) {
914 				/*
915 				 * Even if the neighbor is not in the default
916 				 * router list, the neighbor may be used
917 				 * as a next hop for some destinations
918 				 * (e.g. redirect case). So we must
919 				 * call rt6_flush explicitly.
920 				 */
921 				rt6_flush(&ip6->ip6_src, ifp);
922 			}
923 		}
924 		ln->ln_router = is_router;
925 	}
926         /* XXX - QL
927 	 *  Does this matter?
928 	 *  rt->rt_flags &= ~RTF_REJECT;
929 	 */
930 	ln->la_asked = 0;
931 	if (ln->la_hold != NULL)
932 		nd6_grab_holdchain(ln, &chain, &sin6);
933  freeit:
934 	if (ln != NULL)
935 		LLE_WUNLOCK(ln);
936 
937 	if (chain != NULL)
938 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
939 
940 	if (checklink)
941 		pfxlist_onlink_check();
942 
943 	m_freem(m);
944 	return;
945 
946  bad:
947 	if (ln != NULL)
948 		LLE_WUNLOCK(ln);
949 
950 	ICMP6STAT_INC(icp6s_badna);
951 	m_freem(m);
952 }
953 
954 /*
955  * Neighbor advertisement output handling.
956  *
957  * Based on RFC 2461
958  *
959  * the following items are not implemented yet:
960  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
961  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
962  *
963  * tlladdr - 1 if include target link-layer address
964  * sdl0 - sockaddr_dl (= proxy NA) or NULL
965  */
966 static void
967 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
968     const struct in6_addr *taddr6, u_long flags, int tlladdr,
969     struct sockaddr *sdl0, u_int fibnum)
970 {
971 	struct mbuf *m;
972 	struct m_tag *mtag;
973 	struct ifnet *oifp;
974 	struct ip6_hdr *ip6;
975 	struct nd_neighbor_advert *nd_na;
976 	struct ip6_moptions im6o;
977 	struct in6_addr src, daddr6;
978 	struct sockaddr_in6 dst_sa;
979 	int icmp6len, maxlen, error;
980 	caddr_t mac = NULL;
981 	struct route_in6 ro;
982 
983 	bzero(&ro, sizeof(ro));
984 
985 	daddr6 = *daddr6_0;	/* make a local copy for modification */
986 
987 	/* estimate the size of message */
988 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
989 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
990 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
991 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
992 	    __func__, max_linkhdr, maxlen, MCLBYTES));
993 
994 	if (max_linkhdr + maxlen > MHLEN)
995 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
996 	else
997 		m = m_gethdr(M_NOWAIT, MT_DATA);
998 	if (m == NULL)
999 		return;
1000 	M_SETFIB(m, fibnum);
1001 
1002 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1003 		m->m_flags |= M_MCAST;
1004 		im6o.im6o_multicast_ifp = ifp;
1005 		im6o.im6o_multicast_hlim = 255;
1006 		im6o.im6o_multicast_loop = 0;
1007 	}
1008 
1009 	icmp6len = sizeof(*nd_na);
1010 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1011 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
1012 
1013 	/* fill neighbor advertisement packet */
1014 	ip6 = mtod(m, struct ip6_hdr *);
1015 	ip6->ip6_flow = 0;
1016 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1017 	ip6->ip6_vfc |= IPV6_VERSION;
1018 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1019 	ip6->ip6_hlim = 255;
1020 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1021 		/* reply to DAD */
1022 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1023 		daddr6.s6_addr16[1] = 0;
1024 		daddr6.s6_addr32[1] = 0;
1025 		daddr6.s6_addr32[2] = 0;
1026 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1027 		if (in6_setscope(&daddr6, ifp, NULL))
1028 			goto bad;
1029 
1030 		flags &= ~ND_NA_FLAG_SOLICITED;
1031 	}
1032 	ip6->ip6_dst = daddr6;
1033 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1034 	dst_sa.sin6_family = AF_INET6;
1035 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1036 	dst_sa.sin6_addr = daddr6;
1037 
1038 	/*
1039 	 * Select a source whose scope is the same as that of the dest.
1040 	 */
1041 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1042 	oifp = ifp;
1043 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1044 	if (error) {
1045 		char ip6buf[INET6_ADDRSTRLEN];
1046 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1047 		    "determined: dst=%s, error=%d\n",
1048 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1049 		goto bad;
1050 	}
1051 	ip6->ip6_src = src;
1052 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1053 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1054 	nd_na->nd_na_code = 0;
1055 	nd_na->nd_na_target = *taddr6;
1056 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1057 
1058 	/*
1059 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1060 	 * see nd6_ns_input() for details.
1061 	 * Basically, if NS packet is sent to unicast/anycast addr,
1062 	 * target lladdr option SHOULD NOT be included.
1063 	 */
1064 	if (tlladdr) {
1065 		/*
1066 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1067 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1068 		 * my address) use lladdr configured for the interface.
1069 		 */
1070 		if (sdl0 == NULL) {
1071 			if (ifp->if_carp)
1072 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1073 			if (mac == NULL)
1074 				mac = nd6_ifptomac(ifp);
1075 		} else if (sdl0->sa_family == AF_LINK) {
1076 			struct sockaddr_dl *sdl;
1077 			sdl = (struct sockaddr_dl *)sdl0;
1078 			if (sdl->sdl_alen == ifp->if_addrlen)
1079 				mac = LLADDR(sdl);
1080 		}
1081 	}
1082 	if (tlladdr && mac) {
1083 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1084 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1085 
1086 		/* roundup to 8 bytes alignment! */
1087 		optlen = (optlen + 7) & ~7;
1088 
1089 		m->m_pkthdr.len += optlen;
1090 		m->m_len += optlen;
1091 		icmp6len += optlen;
1092 		bzero((caddr_t)nd_opt, optlen);
1093 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1094 		nd_opt->nd_opt_len = optlen >> 3;
1095 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1096 	} else
1097 		flags &= ~ND_NA_FLAG_OVERRIDE;
1098 
1099 	ip6->ip6_plen = htons((u_short)icmp6len);
1100 	nd_na->nd_na_flags_reserved = flags;
1101 	nd_na->nd_na_cksum = 0;
1102 	nd_na->nd_na_cksum =
1103 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1104 
1105 	if (send_sendso_input_hook != NULL) {
1106 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1107 		    sizeof(unsigned short), M_NOWAIT);
1108 		if (mtag == NULL)
1109 			goto bad;
1110 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1111 		m_tag_prepend(m, mtag);
1112 	}
1113 
1114 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1115 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1116 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1117 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1118 
1119 	/* We don't cache this route. */
1120 	RO_RTFREE(&ro);
1121 
1122 	return;
1123 
1124   bad:
1125 	if (ro.ro_rt) {
1126 		RTFREE(ro.ro_rt);
1127 	}
1128 	m_freem(m);
1129 	return;
1130 }
1131 
1132 #ifndef BURN_BRIDGES
1133 void
1134 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1135     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1136     struct sockaddr *sdl0)
1137 {
1138 
1139 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1140 	    RT_DEFAULT_FIB);
1141 }
1142 #endif
1143 
1144 caddr_t
1145 nd6_ifptomac(struct ifnet *ifp)
1146 {
1147 	switch (ifp->if_type) {
1148 	case IFT_ARCNET:
1149 	case IFT_ETHER:
1150 	case IFT_FDDI:
1151 	case IFT_IEEE1394:
1152 #ifdef IFT_L2VLAN
1153 	case IFT_L2VLAN:
1154 #endif
1155 #ifdef IFT_IEEE80211
1156 	case IFT_IEEE80211:
1157 #endif
1158 	case IFT_INFINIBAND:
1159 	case IFT_BRIDGE:
1160 	case IFT_ISO88025:
1161 		return IF_LLADDR(ifp);
1162 	default:
1163 		return NULL;
1164 	}
1165 }
1166 
1167 struct dadq {
1168 	TAILQ_ENTRY(dadq) dad_list;
1169 	struct ifaddr *dad_ifa;
1170 	int dad_count;		/* max NS to send */
1171 	int dad_ns_tcount;	/* # of trials to send NS */
1172 	int dad_ns_ocount;	/* NS sent so far */
1173 	int dad_ns_icount;
1174 	int dad_na_icount;
1175 	int dad_ns_lcount;	/* looped back NS */
1176 	int dad_loopbackprobe;	/* probing state for loopback detection */
1177 	struct callout dad_timer_ch;
1178 	struct vnet *dad_vnet;
1179 	u_int dad_refcnt;
1180 #define	ND_OPT_NONCE_LEN32 \
1181 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1182 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1183 };
1184 
1185 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1186 static VNET_DEFINE(struct rwlock, dad_rwlock);
1187 #define	V_dadq			VNET(dadq)
1188 #define	V_dad_rwlock		VNET(dad_rwlock)
1189 
1190 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1191 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1192 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1193 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1194 
1195 static void
1196 nd6_dad_add(struct dadq *dp)
1197 {
1198 
1199 	DADQ_WLOCK();
1200 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1201 	DADQ_WUNLOCK();
1202 }
1203 
1204 static void
1205 nd6_dad_del(struct dadq *dp)
1206 {
1207 
1208 	DADQ_WLOCK();
1209 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1210 	DADQ_WUNLOCK();
1211 	nd6_dad_rele(dp);
1212 }
1213 
1214 static struct dadq *
1215 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1216 {
1217 	struct dadq *dp;
1218 
1219 	DADQ_RLOCK();
1220 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1221 		if (dp->dad_ifa != ifa)
1222 			continue;
1223 		/*
1224 		 * Skip if the nonce matches the received one.
1225 		 * +2 in the length is required because of type and
1226 		 * length fields are included in a header.
1227 		 */
1228 		if (n != NULL &&
1229 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1230 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1231 		        ND_OPT_NONCE_LEN) == 0) {
1232 			dp->dad_ns_lcount++;
1233 			continue;
1234 		}
1235 		refcount_acquire(&dp->dad_refcnt);
1236 		break;
1237 	}
1238 	DADQ_RUNLOCK();
1239 
1240 	return (dp);
1241 }
1242 
1243 static void
1244 nd6_dad_starttimer(struct dadq *dp, int ticks)
1245 {
1246 
1247 	callout_reset(&dp->dad_timer_ch, ticks,
1248 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1249 }
1250 
1251 static void
1252 nd6_dad_stoptimer(struct dadq *dp)
1253 {
1254 
1255 	callout_drain(&dp->dad_timer_ch);
1256 }
1257 
1258 static void
1259 nd6_dad_rele(struct dadq *dp)
1260 {
1261 
1262 	if (refcount_release(&dp->dad_refcnt)) {
1263 		ifa_free(dp->dad_ifa);
1264 		free(dp, M_IP6NDP);
1265 	}
1266 }
1267 
1268 void
1269 nd6_dad_init(void)
1270 {
1271 
1272 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1273 	TAILQ_INIT(&V_dadq);
1274 }
1275 
1276 /*
1277  * Start Duplicate Address Detection (DAD) for specified interface address.
1278  */
1279 void
1280 nd6_dad_start(struct ifaddr *ifa, int delay)
1281 {
1282 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1283 	struct dadq *dp;
1284 	char ip6buf[INET6_ADDRSTRLEN];
1285 
1286 	/*
1287 	 * If we don't need DAD, don't do it.
1288 	 * There are several cases:
1289 	 * - DAD is disabled (ip6_dad_count == 0)
1290 	 * - the interface address is anycast
1291 	 */
1292 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1293 		log(LOG_DEBUG,
1294 			"nd6_dad_start: called with non-tentative address "
1295 			"%s(%s)\n",
1296 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1297 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1298 		return;
1299 	}
1300 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1301 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1302 		return;
1303 	}
1304 	if (!V_ip6_dad_count) {
1305 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1306 		return;
1307 	}
1308 	if (ifa->ifa_ifp == NULL)
1309 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1310 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1311 		return;
1312 	}
1313 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1314 		return;
1315 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1316 		/* DAD already in progress */
1317 		nd6_dad_rele(dp);
1318 		return;
1319 	}
1320 
1321 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1322 	if (dp == NULL) {
1323 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1324 			"%s(%s)\n",
1325 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1326 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1327 		return;
1328 	}
1329 	callout_init(&dp->dad_timer_ch, 0);
1330 #ifdef VIMAGE
1331 	dp->dad_vnet = curvnet;
1332 #endif
1333 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1334 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1335 
1336 	/*
1337 	 * Send NS packet for DAD, ip6_dad_count times.
1338 	 * Note that we must delay the first transmission, if this is the
1339 	 * first packet to be sent from the interface after interface
1340 	 * (re)initialization.
1341 	 */
1342 	dp->dad_ifa = ifa;
1343 	ifa_ref(dp->dad_ifa);
1344 	dp->dad_count = V_ip6_dad_count;
1345 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1346 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1347 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1348 	refcount_init(&dp->dad_refcnt, 1);
1349 	nd6_dad_add(dp);
1350 	if (delay == 0) {
1351 		nd6_dad_ns_output(dp, ifa);
1352 		nd6_dad_starttimer(dp,
1353 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1354 	} else {
1355 		nd6_dad_starttimer(dp, delay);
1356 	}
1357 }
1358 
1359 /*
1360  * terminate DAD unconditionally.  used for address removals.
1361  */
1362 void
1363 nd6_dad_stop(struct ifaddr *ifa)
1364 {
1365 	struct dadq *dp;
1366 
1367 	dp = nd6_dad_find(ifa, NULL);
1368 	if (!dp) {
1369 		/* DAD wasn't started yet */
1370 		return;
1371 	}
1372 
1373 	nd6_dad_stoptimer(dp);
1374 
1375 	/*
1376 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1377 	 * we were waiting for it to stop, so re-do the lookup.
1378 	 */
1379 	nd6_dad_rele(dp);
1380 	if (nd6_dad_find(ifa, NULL) == NULL)
1381 		return;
1382 
1383 	nd6_dad_del(dp);
1384 	nd6_dad_rele(dp);
1385 }
1386 
1387 static void
1388 nd6_dad_timer(struct dadq *dp)
1389 {
1390 	CURVNET_SET(dp->dad_vnet);
1391 	struct ifaddr *ifa = dp->dad_ifa;
1392 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1393 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1394 	char ip6buf[INET6_ADDRSTRLEN];
1395 
1396 	/* Sanity check */
1397 	if (ia == NULL) {
1398 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1399 		goto err;
1400 	}
1401 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1402 		/* Do not need DAD for ifdisabled interface. */
1403 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1404 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1405 		goto err;
1406 	}
1407 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1408 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1409 			"%s(%s)\n",
1410 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1411 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1412 		goto err;
1413 	}
1414 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1415 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1416 			"%s(%s)\n",
1417 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1418 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1419 		goto err;
1420 	}
1421 
1422 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1423 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1424 	    (((ifp->if_flags & IFF_UP) == 0) ||
1425 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1426 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1427 		    if_name(ifa->ifa_ifp)));
1428 		goto err;
1429 	}
1430 
1431 	/* Need more checks? */
1432 	if (dp->dad_ns_ocount < dp->dad_count) {
1433 		/*
1434 		 * We have more NS to go.  Send NS packet for DAD.
1435 		 */
1436 		nd6_dad_ns_output(dp, ifa);
1437 		nd6_dad_starttimer(dp,
1438 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1439 		goto done;
1440 	} else {
1441 		/*
1442 		 * We have transmitted sufficient number of DAD packets.
1443 		 * See what we've got.
1444 		 */
1445 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1446 			/* We've seen NS or NA, means DAD has failed. */
1447 			nd6_dad_duplicated(ifa, dp);
1448 		else if (V_dad_enhanced != 0 &&
1449 		    dp->dad_ns_lcount > 0 &&
1450 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1451 			/*
1452 			 * A looped back probe is detected,
1453 			 * Sec. 4.1 in draft-ietf-6man-enhanced-dad-13
1454 			 * requires transmission of additional probes until
1455 			 * the loopback condition becomes clear.
1456 			 */
1457 			log(LOG_ERR, "%s: a looped back NS message is "
1458 			    "detected during DAD for %s.  "
1459 			    "Another DAD probes are being sent.\n",
1460 			    if_name(ifa->ifa_ifp),
1461 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1462 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1463 			/*
1464 			 * An interface with IGNORELOOP is one which a
1465 			 * loopback is permanently expected while regular
1466 			 * traffic works.  In that case, stop DAD after
1467 			 * MAX_MULTICAST_SOLICIT number of NS messages
1468 			 * regardless of the number of received loopback NS
1469 			 * by increasing dad_loopbackprobe in advance.
1470 			 */
1471 			if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IGNORELOOP)
1472 				dp->dad_loopbackprobe += V_nd6_mmaxtries;
1473 			/*
1474 			 * Send an NS immediately and increase dad_count by
1475 			 * V_nd6_mmaxtries - 1.
1476 			 */
1477 			nd6_dad_ns_output(dp, ifa);
1478 			dp->dad_count =
1479 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1480 			nd6_dad_starttimer(dp,
1481 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1482 			goto done;
1483 		} else {
1484 			/*
1485 			 * We are done with DAD.  No NA came, no NS came.
1486 			 * No duplicate address found.  Check IFDISABLED flag
1487 			 * again in case that it is changed between the
1488 			 * beginning of this function and here.
1489 			 */
1490 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1491 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1492 
1493 			nd6log((LOG_DEBUG,
1494 			    "%s: DAD complete for %s - no duplicates found\n",
1495 			    if_name(ifa->ifa_ifp),
1496 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1497 			if (dp->dad_ns_lcount > 0)
1498 				log(LOG_ERR, "%s: DAD completed while "
1499 				    "a looped back NS message is detected "
1500 				    "during DAD for %s.\n",
1501 				    if_name(ifa->ifa_ifp),
1502 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1503 		}
1504 	}
1505 err:
1506 	nd6_dad_del(dp);
1507 done:
1508 	CURVNET_RESTORE();
1509 }
1510 
1511 static void
1512 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1513 {
1514 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1515 	struct ifnet *ifp;
1516 	char ip6buf[INET6_ADDRSTRLEN];
1517 
1518 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1519 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1520 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1521 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1522 	    dp->dad_na_icount);
1523 
1524 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1525 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1526 
1527 	ifp = ifa->ifa_ifp;
1528 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1529 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1530 	log(LOG_ERR, "%s: manual intervention required\n",
1531 	    if_name(ifp));
1532 
1533 	/*
1534 	 * If the address is a link-local address formed from an interface
1535 	 * identifier based on the hardware address which is supposed to be
1536 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1537 	 * operation on the interface SHOULD be disabled.
1538 	 * [RFC 4862, Section 5.4.5]
1539 	 */
1540 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1541 		struct in6_addr in6;
1542 
1543 		/*
1544 		 * To avoid over-reaction, we only apply this logic when we are
1545 		 * very sure that hardware addresses are supposed to be unique.
1546 		 */
1547 		switch (ifp->if_type) {
1548 		case IFT_ETHER:
1549 		case IFT_FDDI:
1550 		case IFT_ATM:
1551 		case IFT_IEEE1394:
1552 #ifdef IFT_IEEE80211
1553 		case IFT_IEEE80211:
1554 #endif
1555 		case IFT_INFINIBAND:
1556 			in6 = ia->ia_addr.sin6_addr;
1557 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1558 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1559 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1560 				log(LOG_ERR, "%s: possible hardware address "
1561 				    "duplication detected, disable IPv6\n",
1562 				    if_name(ifp));
1563 			}
1564 			break;
1565 		}
1566 	}
1567 }
1568 
1569 static void
1570 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1571 {
1572 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1573 	struct ifnet *ifp = ifa->ifa_ifp;
1574 	int i;
1575 
1576 	dp->dad_ns_tcount++;
1577 	if ((ifp->if_flags & IFF_UP) == 0) {
1578 		return;
1579 	}
1580 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1581 		return;
1582 	}
1583 
1584 	dp->dad_ns_ocount++;
1585 	if (V_dad_enhanced != 0) {
1586 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1587 			dp->dad_nonce[i] = arc4random();
1588 		/*
1589 		 * XXXHRS: Note that in the case that
1590 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1591 		 * different nonces can be looped back in an unexpected
1592 		 * order.  The current implementation recognizes only
1593 		 * the latest nonce on the sender side.  Practically it
1594 		 * should work well in almost all cases.
1595 		 */
1596 	}
1597 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL,
1598 	    (uint8_t *)&dp->dad_nonce[0]);
1599 }
1600 
1601 static void
1602 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1603 {
1604 	struct in6_ifaddr *ia;
1605 	struct ifnet *ifp;
1606 	const struct in6_addr *taddr6;
1607 	struct dadq *dp;
1608 
1609 	if (ifa == NULL)
1610 		panic("ifa == NULL in nd6_dad_ns_input");
1611 
1612 	ia = (struct in6_ifaddr *)ifa;
1613 	ifp = ifa->ifa_ifp;
1614 	taddr6 = &ia->ia_addr.sin6_addr;
1615 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1616 	if (V_dad_enhanced == 0)
1617 		ndopt_nonce = NULL;
1618 	dp = nd6_dad_find(ifa, ndopt_nonce);
1619 	if (dp == NULL)
1620 		return;
1621 
1622 	dp->dad_ns_icount++;
1623 	nd6_dad_rele(dp);
1624 }
1625 
1626 static void
1627 nd6_dad_na_input(struct ifaddr *ifa)
1628 {
1629 	struct dadq *dp;
1630 
1631 	if (ifa == NULL)
1632 		panic("ifa == NULL in nd6_dad_na_input");
1633 
1634 	dp = nd6_dad_find(ifa, NULL);
1635 	if (dp != NULL) {
1636 		dp->dad_na_icount++;
1637 		nd6_dad_rele(dp);
1638 	}
1639 }
1640