xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 4ce386ff25d77954b8cfa11534f632172e848244)
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 	if (max_linkhdr + maxlen >= MCLBYTES) {
421 #ifdef DIAGNOSTIC
422 		printf("%s: max_linkhdr + maxlen >= MCLBYTES "
423 		    "(%d + %d > %d)\n", __func__, max_linkhdr, maxlen,
424 		    MCLBYTES);
425 #endif
426 		return;
427 	}
428 
429 	if (max_linkhdr + maxlen > MHLEN)
430 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
431 	else
432 		m = m_gethdr(M_NOWAIT, MT_DATA);
433 	if (m == NULL)
434 		return;
435 	M_SETFIB(m, fibnum);
436 
437 	bzero(&ro, sizeof(ro));
438 
439 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
440 		m->m_flags |= M_MCAST;
441 		im6o.im6o_multicast_ifp = ifp;
442 		im6o.im6o_multicast_hlim = 255;
443 		im6o.im6o_multicast_loop = 0;
444 	}
445 
446 	icmp6len = sizeof(*nd_ns);
447 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
448 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
449 
450 	/* fill neighbor solicitation packet */
451 	ip6 = mtod(m, struct ip6_hdr *);
452 	ip6->ip6_flow = 0;
453 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
454 	ip6->ip6_vfc |= IPV6_VERSION;
455 	/* ip6->ip6_plen will be set later */
456 	ip6->ip6_nxt = IPPROTO_ICMPV6;
457 	ip6->ip6_hlim = 255;
458 	if (daddr6)
459 		ip6->ip6_dst = *daddr6;
460 	else {
461 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
462 		ip6->ip6_dst.s6_addr16[1] = 0;
463 		ip6->ip6_dst.s6_addr32[1] = 0;
464 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
465 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
466 		ip6->ip6_dst.s6_addr8[12] = 0xff;
467 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
468 			goto bad;
469 	}
470 	if (nonce == NULL) {
471 		struct ifaddr *ifa;
472 
473 		/*
474 		 * RFC2461 7.2.2:
475 		 * "If the source address of the packet prompting the
476 		 * solicitation is the same as one of the addresses assigned
477 		 * to the outgoing interface, that address SHOULD be placed
478 		 * in the IP Source Address of the outgoing solicitation.
479 		 * Otherwise, any one of the addresses assigned to the
480 		 * interface should be used."
481 		 *
482 		 * We use the source address for the prompting packet
483 		 * (saddr6), if:
484 		 * - saddr6 is given from the caller (by giving "ln"), and
485 		 * - saddr6 belongs to the outgoing interface.
486 		 * Otherwise, we perform the source address selection as usual.
487 		 */
488 		struct in6_addr *hsrc;
489 
490 		hsrc = NULL;
491 		if (ln != NULL) {
492 			LLE_RLOCK(ln);
493 			if (ln->la_hold != NULL) {
494 				struct ip6_hdr *hip6;		/* hold ip6 */
495 
496 				/*
497 				 * assuming every packet in la_hold has the same IP
498 				 * header
499 				 */
500 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
501 				/* XXX pullup? */
502 				if (sizeof(*hip6) < ln->la_hold->m_len) {
503 					ip6->ip6_src = hip6->ip6_src;
504 					hsrc = &hip6->ip6_src;
505 				}
506 			}
507 			LLE_RUNLOCK(ln);
508 		}
509 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
510 		    hsrc)) != NULL) {
511 			/* ip6_src set already. */
512 			ifa_free(ifa);
513 		} else {
514 			int error;
515 			struct sockaddr_in6 dst_sa;
516 			struct in6_addr src_in;
517 			struct ifnet *oifp;
518 
519 			bzero(&dst_sa, sizeof(dst_sa));
520 			dst_sa.sin6_family = AF_INET6;
521 			dst_sa.sin6_len = sizeof(dst_sa);
522 			dst_sa.sin6_addr = ip6->ip6_dst;
523 
524 			oifp = ifp;
525 			error = in6_selectsrc(&dst_sa, NULL,
526 			    NULL, &ro, NULL, &oifp, &src_in);
527 			if (error) {
528 				char ip6buf[INET6_ADDRSTRLEN];
529 				nd6log((LOG_DEBUG, "%s: source can't be "
530 				    "determined: dst=%s, error=%d\n", __func__,
531 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
532 				    error));
533 				goto bad;
534 			}
535 			ip6->ip6_src = src_in;
536 		}
537 	} else {
538 		/*
539 		 * Source address for DAD packet must always be IPv6
540 		 * unspecified address. (0::0)
541 		 * We actually don't have to 0-clear the address (we did it
542 		 * above), but we do so here explicitly to make the intention
543 		 * clearer.
544 		 */
545 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
546 	}
547 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
548 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
549 	nd_ns->nd_ns_code = 0;
550 	nd_ns->nd_ns_reserved = 0;
551 	nd_ns->nd_ns_target = *taddr6;
552 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
553 
554 	/*
555 	 * Add source link-layer address option.
556 	 *
557 	 *				spec		implementation
558 	 *				---		---
559 	 * DAD packet			MUST NOT	do not add the option
560 	 * there's no link layer address:
561 	 *				impossible	do not add the option
562 	 * there's link layer address:
563 	 *	Multicast NS		MUST add one	add the option
564 	 *	Unicast NS		SHOULD add one	add the option
565 	 */
566 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
567 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
568 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
569 		/* 8 byte alignments... */
570 		optlen = (optlen + 7) & ~7;
571 
572 		m->m_pkthdr.len += optlen;
573 		m->m_len += optlen;
574 		icmp6len += optlen;
575 		bzero((caddr_t)nd_opt, optlen);
576 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
577 		nd_opt->nd_opt_len = optlen >> 3;
578 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
579 	}
580 	/*
581 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
582 	 * This behavior is documented as Enhanced Duplicate Address
583 	 * Detection in draft-ietf-6man-enhanced-dad-13.
584 	 * net.inet6.ip6.dad_enhanced=0 disables this.
585 	 */
586 	if (V_dad_enhanced != 0 && nonce != NULL) {
587 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
588 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
589 		/* 8-byte alignment is required. */
590 		optlen = (optlen + 7) & ~7;
591 
592 		m->m_pkthdr.len += optlen;
593 		m->m_len += optlen;
594 		icmp6len += optlen;
595 		bzero((caddr_t)nd_opt, optlen);
596 		nd_opt->nd_opt_type = ND_OPT_NONCE;
597 		nd_opt->nd_opt_len = optlen >> 3;
598 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
599 	}
600 	ip6->ip6_plen = htons((u_short)icmp6len);
601 	nd_ns->nd_ns_cksum = 0;
602 	nd_ns->nd_ns_cksum =
603 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
604 
605 	if (send_sendso_input_hook != NULL) {
606 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
607 			sizeof(unsigned short), M_NOWAIT);
608 		if (mtag == NULL)
609 			goto bad;
610 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
611 		m_tag_prepend(m, mtag);
612 	}
613 
614 	ip6_output(m, NULL, &ro, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
615 	    &im6o, NULL, NULL);
616 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
617 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
618 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
619 
620 	/* We don't cache this route. */
621 	RO_RTFREE(&ro);
622 
623 	return;
624 
625   bad:
626 	if (ro.ro_rt) {
627 		RTFREE(ro.ro_rt);
628 	}
629 	m_freem(m);
630 	return;
631 }
632 
633 #ifndef BURN_BRIDGES
634 void
635 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
636     const struct in6_addr *taddr6, struct llentry *ln, uint8_t *nonce)
637 {
638 
639 	nd6_ns_output_fib(ifp, daddr6, taddr6, ln, nonce, RT_DEFAULT_FIB);
640 }
641 #endif
642 /*
643  * Neighbor advertisement input handling.
644  *
645  * Based on RFC 2461
646  * Based on RFC 2462 (duplicate address detection)
647  *
648  * the following items are not implemented yet:
649  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
650  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
651  */
652 void
653 nd6_na_input(struct mbuf *m, int off, int icmp6len)
654 {
655 	struct ifnet *ifp = m->m_pkthdr.rcvif;
656 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
657 	struct nd_neighbor_advert *nd_na;
658 	struct in6_addr daddr6 = ip6->ip6_dst;
659 	struct in6_addr taddr6;
660 	int flags;
661 	int is_router;
662 	int is_solicited;
663 	int is_override;
664 	char *lladdr = NULL;
665 	int lladdrlen = 0;
666 	int checklink = 0;
667 	struct ifaddr *ifa;
668 	struct llentry *ln = NULL;
669 	union nd_opts ndopts;
670 	struct mbuf *chain = NULL;
671 	struct sockaddr_in6 sin6;
672 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
673 
674 	if (ip6->ip6_hlim != 255) {
675 		nd6log((LOG_ERR,
676 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
677 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
678 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
679 		goto bad;
680 	}
681 
682 #ifndef PULLDOWN_TEST
683 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
684 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
685 #else
686 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
687 	if (nd_na == NULL) {
688 		ICMP6STAT_INC(icp6s_tooshort);
689 		return;
690 	}
691 #endif
692 
693 	flags = nd_na->nd_na_flags_reserved;
694 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
695 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
696 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
697 	memset(&sin6, 0, sizeof(sin6));
698 
699 	taddr6 = nd_na->nd_na_target;
700 	if (in6_setscope(&taddr6, ifp, NULL))
701 		goto bad;	/* XXX: impossible */
702 
703 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
704 		nd6log((LOG_ERR,
705 		    "nd6_na_input: invalid target address %s\n",
706 		    ip6_sprintf(ip6bufs, &taddr6)));
707 		goto bad;
708 	}
709 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
710 		if (is_solicited) {
711 			nd6log((LOG_ERR,
712 			    "nd6_na_input: a solicited adv is multicasted\n"));
713 			goto bad;
714 		}
715 
716 	icmp6len -= sizeof(*nd_na);
717 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
718 	if (nd6_options(&ndopts) < 0) {
719 		nd6log((LOG_INFO,
720 		    "nd6_na_input: invalid ND option, ignored\n"));
721 		/* nd6_options have incremented stats */
722 		goto freeit;
723 	}
724 
725 	if (ndopts.nd_opts_tgt_lladdr) {
726 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
727 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
728 	}
729 
730 	/*
731 	 * This effectively disables the DAD check on a non-master CARP
732 	 * address.
733 	 */
734 	if (ifp->if_carp)
735 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
736 	else
737 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
738 
739 	/*
740 	 * Target address matches one of my interface address.
741 	 *
742 	 * If my address is tentative, this means that there's somebody
743 	 * already using the same address as mine.  This indicates DAD failure.
744 	 * This is defined in RFC 2462.
745 	 *
746 	 * Otherwise, process as defined in RFC 2461.
747 	 */
748 	if (ifa
749 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
750 		ifa_free(ifa);
751 		nd6_dad_na_input(ifa);
752 		goto freeit;
753 	}
754 
755 	/* Just for safety, maybe unnecessary. */
756 	if (ifa) {
757 		ifa_free(ifa);
758 		log(LOG_ERR,
759 		    "nd6_na_input: duplicate IP6 address %s\n",
760 		    ip6_sprintf(ip6bufs, &taddr6));
761 		goto freeit;
762 	}
763 
764 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
765 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
766 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
767 		    ifp->if_addrlen, lladdrlen - 2));
768 		goto bad;
769 	}
770 
771 	/*
772 	 * If no neighbor cache entry is found, NA SHOULD silently be
773 	 * discarded.
774 	 */
775 	IF_AFDATA_RLOCK(ifp);
776 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
777 	IF_AFDATA_RUNLOCK(ifp);
778 	if (ln == NULL) {
779 		goto freeit;
780 	}
781 
782 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
783 		/*
784 		 * If the link-layer has address, and no lladdr option came,
785 		 * discard the packet.
786 		 */
787 		if (ifp->if_addrlen && lladdr == NULL) {
788 			goto freeit;
789 		}
790 
791 		/*
792 		 * Record link-layer address, and update the state.
793 		 */
794 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
795 		ln->la_flags |= LLE_VALID;
796 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
797 		if (is_solicited) {
798 			ln->ln_state = ND6_LLINFO_REACHABLE;
799 			ln->ln_byhint = 0;
800 			if (!ND6_LLINFO_PERMANENT(ln)) {
801 				nd6_llinfo_settimer_locked(ln,
802 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
803 			}
804 		} else {
805 			ln->ln_state = ND6_LLINFO_STALE;
806 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
807 		}
808 		if ((ln->ln_router = is_router) != 0) {
809 			/*
810 			 * This means a router's state has changed from
811 			 * non-reachable to probably reachable, and might
812 			 * affect the status of associated prefixes..
813 			 */
814 			checklink = 1;
815 		}
816 	} else {
817 		int llchange;
818 
819 		/*
820 		 * Check if the link-layer address has changed or not.
821 		 */
822 		if (lladdr == NULL)
823 			llchange = 0;
824 		else {
825 			if (ln->la_flags & LLE_VALID) {
826 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
827 					llchange = 1;
828 				else
829 					llchange = 0;
830 			} else
831 				llchange = 1;
832 		}
833 
834 		/*
835 		 * This is VERY complex.  Look at it with care.
836 		 *
837 		 * override solicit lladdr llchange	action
838 		 *					(L: record lladdr)
839 		 *
840 		 *	0	0	n	--	(2c)
841 		 *	0	0	y	n	(2b) L
842 		 *	0	0	y	y	(1)    REACHABLE->STALE
843 		 *	0	1	n	--	(2c)   *->REACHABLE
844 		 *	0	1	y	n	(2b) L *->REACHABLE
845 		 *	0	1	y	y	(1)    REACHABLE->STALE
846 		 *	1	0	n	--	(2a)
847 		 *	1	0	y	n	(2a) L
848 		 *	1	0	y	y	(2a) L *->STALE
849 		 *	1	1	n	--	(2a)   *->REACHABLE
850 		 *	1	1	y	n	(2a) L *->REACHABLE
851 		 *	1	1	y	y	(2a) L *->REACHABLE
852 		 */
853 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
854 			/*
855 			 * If state is REACHABLE, make it STALE.
856 			 * no other updates should be done.
857 			 */
858 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
859 				ln->ln_state = ND6_LLINFO_STALE;
860 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
861 			}
862 			goto freeit;
863 		} else if (is_override				   /* (2a) */
864 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
865 			|| lladdr == NULL) {			   /* (2c) */
866 			/*
867 			 * Update link-local address, if any.
868 			 */
869 			if (lladdr != NULL) {
870 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
871 				ln->la_flags |= LLE_VALID;
872 				EVENTHANDLER_INVOKE(lle_event, ln,
873 				    LLENTRY_RESOLVED);
874 			}
875 
876 			/*
877 			 * If solicited, make the state REACHABLE.
878 			 * If not solicited and the link-layer address was
879 			 * changed, make it STALE.
880 			 */
881 			if (is_solicited) {
882 				ln->ln_state = ND6_LLINFO_REACHABLE;
883 				ln->ln_byhint = 0;
884 				if (!ND6_LLINFO_PERMANENT(ln)) {
885 					nd6_llinfo_settimer_locked(ln,
886 					    (long)ND_IFINFO(ifp)->reachable * hz);
887 				}
888 			} else {
889 				if (lladdr != NULL && llchange) {
890 					ln->ln_state = ND6_LLINFO_STALE;
891 					nd6_llinfo_settimer_locked(ln,
892 					    (long)V_nd6_gctimer * hz);
893 				}
894 			}
895 		}
896 
897 		if (ln->ln_router && !is_router) {
898 			/*
899 			 * The peer dropped the router flag.
900 			 * Remove the sender from the Default Router List and
901 			 * update the Destination Cache entries.
902 			 */
903 			struct nd_defrouter *dr;
904 			struct in6_addr *in6;
905 
906 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
907 
908 			/*
909 			 * Lock to protect the default router list.
910 			 * XXX: this might be unnecessary, since this function
911 			 * is only called under the network software interrupt
912 			 * context.  However, we keep it just for safety.
913 			 */
914 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
915 			if (dr)
916 				defrtrlist_del(dr);
917 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
918 			    ND6_IFF_ACCEPT_RTADV) {
919 				/*
920 				 * Even if the neighbor is not in the default
921 				 * router list, the neighbor may be used
922 				 * as a next hop for some destinations
923 				 * (e.g. redirect case). So we must
924 				 * call rt6_flush explicitly.
925 				 */
926 				rt6_flush(&ip6->ip6_src, ifp);
927 			}
928 		}
929 		ln->ln_router = is_router;
930 	}
931         /* XXX - QL
932 	 *  Does this matter?
933 	 *  rt->rt_flags &= ~RTF_REJECT;
934 	 */
935 	ln->la_asked = 0;
936 	if (ln->la_hold != NULL)
937 		nd6_grab_holdchain(ln, &chain, &sin6);
938  freeit:
939 	if (ln != NULL)
940 		LLE_WUNLOCK(ln);
941 
942 	if (chain != NULL)
943 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
944 
945 	if (checklink)
946 		pfxlist_onlink_check();
947 
948 	m_freem(m);
949 	return;
950 
951  bad:
952 	if (ln != NULL)
953 		LLE_WUNLOCK(ln);
954 
955 	ICMP6STAT_INC(icp6s_badna);
956 	m_freem(m);
957 }
958 
959 /*
960  * Neighbor advertisement output handling.
961  *
962  * Based on RFC 2461
963  *
964  * the following items are not implemented yet:
965  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
966  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
967  *
968  * tlladdr - 1 if include target link-layer address
969  * sdl0 - sockaddr_dl (= proxy NA) or NULL
970  */
971 static void
972 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
973     const struct in6_addr *taddr6, u_long flags, int tlladdr,
974     struct sockaddr *sdl0, u_int fibnum)
975 {
976 	struct mbuf *m;
977 	struct m_tag *mtag;
978 	struct ifnet *oifp;
979 	struct ip6_hdr *ip6;
980 	struct nd_neighbor_advert *nd_na;
981 	struct ip6_moptions im6o;
982 	struct in6_addr src, daddr6;
983 	struct sockaddr_in6 dst_sa;
984 	int icmp6len, maxlen, error;
985 	caddr_t mac = NULL;
986 	struct route_in6 ro;
987 
988 	bzero(&ro, sizeof(ro));
989 
990 	daddr6 = *daddr6_0;	/* make a local copy for modification */
991 
992 	/* estimate the size of message */
993 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
994 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
995 	if (max_linkhdr + maxlen >= MCLBYTES) {
996 #ifdef DIAGNOSTIC
997 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
998 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
999 #endif
1000 		return;
1001 	}
1002 
1003 	if (max_linkhdr + maxlen > MHLEN)
1004 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1005 	else
1006 		m = m_gethdr(M_NOWAIT, MT_DATA);
1007 	if (m == NULL)
1008 		return;
1009 	M_SETFIB(m, fibnum);
1010 
1011 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1012 		m->m_flags |= M_MCAST;
1013 		im6o.im6o_multicast_ifp = ifp;
1014 		im6o.im6o_multicast_hlim = 255;
1015 		im6o.im6o_multicast_loop = 0;
1016 	}
1017 
1018 	icmp6len = sizeof(*nd_na);
1019 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1020 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
1021 
1022 	/* fill neighbor advertisement packet */
1023 	ip6 = mtod(m, struct ip6_hdr *);
1024 	ip6->ip6_flow = 0;
1025 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1026 	ip6->ip6_vfc |= IPV6_VERSION;
1027 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1028 	ip6->ip6_hlim = 255;
1029 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1030 		/* reply to DAD */
1031 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1032 		daddr6.s6_addr16[1] = 0;
1033 		daddr6.s6_addr32[1] = 0;
1034 		daddr6.s6_addr32[2] = 0;
1035 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1036 		if (in6_setscope(&daddr6, ifp, NULL))
1037 			goto bad;
1038 
1039 		flags &= ~ND_NA_FLAG_SOLICITED;
1040 	}
1041 	ip6->ip6_dst = daddr6;
1042 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1043 	dst_sa.sin6_family = AF_INET6;
1044 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1045 	dst_sa.sin6_addr = daddr6;
1046 
1047 	/*
1048 	 * Select a source whose scope is the same as that of the dest.
1049 	 */
1050 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1051 	oifp = ifp;
1052 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1053 	if (error) {
1054 		char ip6buf[INET6_ADDRSTRLEN];
1055 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1056 		    "determined: dst=%s, error=%d\n",
1057 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1058 		goto bad;
1059 	}
1060 	ip6->ip6_src = src;
1061 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1062 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1063 	nd_na->nd_na_code = 0;
1064 	nd_na->nd_na_target = *taddr6;
1065 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1066 
1067 	/*
1068 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1069 	 * see nd6_ns_input() for details.
1070 	 * Basically, if NS packet is sent to unicast/anycast addr,
1071 	 * target lladdr option SHOULD NOT be included.
1072 	 */
1073 	if (tlladdr) {
1074 		/*
1075 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1076 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1077 		 * my address) use lladdr configured for the interface.
1078 		 */
1079 		if (sdl0 == NULL) {
1080 			if (ifp->if_carp)
1081 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1082 			if (mac == NULL)
1083 				mac = nd6_ifptomac(ifp);
1084 		} else if (sdl0->sa_family == AF_LINK) {
1085 			struct sockaddr_dl *sdl;
1086 			sdl = (struct sockaddr_dl *)sdl0;
1087 			if (sdl->sdl_alen == ifp->if_addrlen)
1088 				mac = LLADDR(sdl);
1089 		}
1090 	}
1091 	if (tlladdr && mac) {
1092 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1093 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1094 
1095 		/* roundup to 8 bytes alignment! */
1096 		optlen = (optlen + 7) & ~7;
1097 
1098 		m->m_pkthdr.len += optlen;
1099 		m->m_len += optlen;
1100 		icmp6len += optlen;
1101 		bzero((caddr_t)nd_opt, optlen);
1102 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1103 		nd_opt->nd_opt_len = optlen >> 3;
1104 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1105 	} else
1106 		flags &= ~ND_NA_FLAG_OVERRIDE;
1107 
1108 	ip6->ip6_plen = htons((u_short)icmp6len);
1109 	nd_na->nd_na_flags_reserved = flags;
1110 	nd_na->nd_na_cksum = 0;
1111 	nd_na->nd_na_cksum =
1112 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1113 
1114 	if (send_sendso_input_hook != NULL) {
1115 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1116 		    sizeof(unsigned short), M_NOWAIT);
1117 		if (mtag == NULL)
1118 			goto bad;
1119 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1120 		m_tag_prepend(m, mtag);
1121 	}
1122 
1123 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1124 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1125 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1126 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1127 
1128 	/* We don't cache this route. */
1129 	RO_RTFREE(&ro);
1130 
1131 	return;
1132 
1133   bad:
1134 	if (ro.ro_rt) {
1135 		RTFREE(ro.ro_rt);
1136 	}
1137 	m_freem(m);
1138 	return;
1139 }
1140 
1141 #ifndef BURN_BRIDGES
1142 void
1143 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1144     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1145     struct sockaddr *sdl0)
1146 {
1147 
1148 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1149 	    RT_DEFAULT_FIB);
1150 }
1151 #endif
1152 
1153 caddr_t
1154 nd6_ifptomac(struct ifnet *ifp)
1155 {
1156 	switch (ifp->if_type) {
1157 	case IFT_ARCNET:
1158 	case IFT_ETHER:
1159 	case IFT_FDDI:
1160 	case IFT_IEEE1394:
1161 #ifdef IFT_L2VLAN
1162 	case IFT_L2VLAN:
1163 #endif
1164 #ifdef IFT_IEEE80211
1165 	case IFT_IEEE80211:
1166 #endif
1167 	case IFT_INFINIBAND:
1168 	case IFT_BRIDGE:
1169 	case IFT_ISO88025:
1170 		return IF_LLADDR(ifp);
1171 	default:
1172 		return NULL;
1173 	}
1174 }
1175 
1176 struct dadq {
1177 	TAILQ_ENTRY(dadq) dad_list;
1178 	struct ifaddr *dad_ifa;
1179 	int dad_count;		/* max NS to send */
1180 	int dad_ns_tcount;	/* # of trials to send NS */
1181 	int dad_ns_ocount;	/* NS sent so far */
1182 	int dad_ns_icount;
1183 	int dad_na_icount;
1184 	int dad_ns_lcount;	/* looped back NS */
1185 	int dad_loopbackprobe;	/* probing state for loopback detection */
1186 	struct callout dad_timer_ch;
1187 	struct vnet *dad_vnet;
1188 	u_int dad_refcnt;
1189 #define	ND_OPT_NONCE_LEN32 \
1190 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1191 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1192 };
1193 
1194 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1195 static VNET_DEFINE(struct rwlock, dad_rwlock);
1196 #define	V_dadq			VNET(dadq)
1197 #define	V_dad_rwlock		VNET(dad_rwlock)
1198 
1199 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1200 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1201 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1202 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1203 
1204 static void
1205 nd6_dad_add(struct dadq *dp)
1206 {
1207 
1208 	DADQ_WLOCK();
1209 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1210 	DADQ_WUNLOCK();
1211 }
1212 
1213 static void
1214 nd6_dad_del(struct dadq *dp)
1215 {
1216 
1217 	DADQ_WLOCK();
1218 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1219 	DADQ_WUNLOCK();
1220 	nd6_dad_rele(dp);
1221 }
1222 
1223 static struct dadq *
1224 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1225 {
1226 	struct dadq *dp;
1227 
1228 	DADQ_RLOCK();
1229 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1230 		if (dp->dad_ifa != ifa)
1231 			continue;
1232 		/*
1233 		 * Skip if the nonce matches the received one.
1234 		 * +2 in the length is required because of type and
1235 		 * length fields are included in a header.
1236 		 */
1237 		if (n != NULL &&
1238 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1239 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1240 		        ND_OPT_NONCE_LEN) == 0) {
1241 			dp->dad_ns_lcount++;
1242 			continue;
1243 		}
1244 		refcount_acquire(&dp->dad_refcnt);
1245 		break;
1246 	}
1247 	DADQ_RUNLOCK();
1248 
1249 	return (dp);
1250 }
1251 
1252 static void
1253 nd6_dad_starttimer(struct dadq *dp, int ticks)
1254 {
1255 
1256 	callout_reset(&dp->dad_timer_ch, ticks,
1257 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1258 }
1259 
1260 static void
1261 nd6_dad_stoptimer(struct dadq *dp)
1262 {
1263 
1264 	callout_drain(&dp->dad_timer_ch);
1265 }
1266 
1267 static void
1268 nd6_dad_rele(struct dadq *dp)
1269 {
1270 
1271 	if (refcount_release(&dp->dad_refcnt)) {
1272 		ifa_free(dp->dad_ifa);
1273 		free(dp, M_IP6NDP);
1274 	}
1275 }
1276 
1277 void
1278 nd6_dad_init(void)
1279 {
1280 
1281 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1282 	TAILQ_INIT(&V_dadq);
1283 }
1284 
1285 /*
1286  * Start Duplicate Address Detection (DAD) for specified interface address.
1287  */
1288 void
1289 nd6_dad_start(struct ifaddr *ifa, int delay)
1290 {
1291 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1292 	struct dadq *dp;
1293 	char ip6buf[INET6_ADDRSTRLEN];
1294 
1295 	/*
1296 	 * If we don't need DAD, don't do it.
1297 	 * There are several cases:
1298 	 * - DAD is disabled (ip6_dad_count == 0)
1299 	 * - the interface address is anycast
1300 	 */
1301 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1302 		log(LOG_DEBUG,
1303 			"nd6_dad_start: called with non-tentative address "
1304 			"%s(%s)\n",
1305 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1306 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1307 		return;
1308 	}
1309 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1310 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1311 		return;
1312 	}
1313 	if (!V_ip6_dad_count) {
1314 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1315 		return;
1316 	}
1317 	if (ifa->ifa_ifp == NULL)
1318 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1319 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1320 		return;
1321 	}
1322 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1323 		return;
1324 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1325 		/* DAD already in progress */
1326 		nd6_dad_rele(dp);
1327 		return;
1328 	}
1329 
1330 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1331 	if (dp == NULL) {
1332 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1333 			"%s(%s)\n",
1334 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1335 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1336 		return;
1337 	}
1338 	callout_init(&dp->dad_timer_ch, 0);
1339 #ifdef VIMAGE
1340 	dp->dad_vnet = curvnet;
1341 #endif
1342 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1343 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1344 
1345 	/*
1346 	 * Send NS packet for DAD, ip6_dad_count times.
1347 	 * Note that we must delay the first transmission, if this is the
1348 	 * first packet to be sent from the interface after interface
1349 	 * (re)initialization.
1350 	 */
1351 	dp->dad_ifa = ifa;
1352 	ifa_ref(dp->dad_ifa);
1353 	dp->dad_count = V_ip6_dad_count;
1354 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1355 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1356 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1357 	refcount_init(&dp->dad_refcnt, 1);
1358 	nd6_dad_add(dp);
1359 	if (delay == 0) {
1360 		nd6_dad_ns_output(dp, ifa);
1361 		nd6_dad_starttimer(dp,
1362 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1363 	} else {
1364 		nd6_dad_starttimer(dp, delay);
1365 	}
1366 }
1367 
1368 /*
1369  * terminate DAD unconditionally.  used for address removals.
1370  */
1371 void
1372 nd6_dad_stop(struct ifaddr *ifa)
1373 {
1374 	struct dadq *dp;
1375 
1376 	dp = nd6_dad_find(ifa, NULL);
1377 	if (!dp) {
1378 		/* DAD wasn't started yet */
1379 		return;
1380 	}
1381 
1382 	nd6_dad_stoptimer(dp);
1383 
1384 	/*
1385 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1386 	 * we were waiting for it to stop, so re-do the lookup.
1387 	 */
1388 	nd6_dad_rele(dp);
1389 	if (nd6_dad_find(ifa, NULL) == NULL)
1390 		return;
1391 
1392 	nd6_dad_del(dp);
1393 	nd6_dad_rele(dp);
1394 }
1395 
1396 static void
1397 nd6_dad_timer(struct dadq *dp)
1398 {
1399 	CURVNET_SET(dp->dad_vnet);
1400 	struct ifaddr *ifa = dp->dad_ifa;
1401 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1402 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1403 	char ip6buf[INET6_ADDRSTRLEN];
1404 
1405 	/* Sanity check */
1406 	if (ia == NULL) {
1407 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1408 		goto err;
1409 	}
1410 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1411 		/* Do not need DAD for ifdisabled interface. */
1412 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1413 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1414 		goto err;
1415 	}
1416 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1417 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1418 			"%s(%s)\n",
1419 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1420 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1421 		goto err;
1422 	}
1423 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1424 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1425 			"%s(%s)\n",
1426 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1427 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1428 		goto err;
1429 	}
1430 
1431 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1432 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1433 	    (((ifp->if_flags & IFF_UP) == 0) ||
1434 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1435 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1436 		    if_name(ifa->ifa_ifp)));
1437 		goto err;
1438 	}
1439 
1440 	/* Need more checks? */
1441 	if (dp->dad_ns_ocount < dp->dad_count) {
1442 		/*
1443 		 * We have more NS to go.  Send NS packet for DAD.
1444 		 */
1445 		nd6_dad_ns_output(dp, ifa);
1446 		nd6_dad_starttimer(dp,
1447 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1448 		goto done;
1449 	} else {
1450 		/*
1451 		 * We have transmitted sufficient number of DAD packets.
1452 		 * See what we've got.
1453 		 */
1454 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1455 			/* We've seen NS or NA, means DAD has failed. */
1456 			nd6_dad_duplicated(ifa, dp);
1457 		else if (V_dad_enhanced != 0 &&
1458 		    dp->dad_ns_lcount > 0 &&
1459 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1460 			/*
1461 			 * A looped back probe is detected,
1462 			 * Sec. 4.1 in draft-ietf-6man-enhanced-dad-13
1463 			 * requires transmission of additional probes until
1464 			 * the loopback condition becomes clear.
1465 			 */
1466 			log(LOG_ERR, "%s: a looped back NS message is "
1467 			    "detected during DAD for %s.  "
1468 			    "Another DAD probes are being sent.\n",
1469 			    if_name(ifa->ifa_ifp),
1470 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1471 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1472 			/*
1473 			 * An interface with IGNORELOOP is one which a
1474 			 * loopback is permanently expected while regular
1475 			 * traffic works.  In that case, stop DAD after
1476 			 * MAX_MULTICAST_SOLICIT number of NS messages
1477 			 * regardless of the number of received loopback NS
1478 			 * by increasing dad_loopbackprobe in advance.
1479 			 */
1480 			if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IGNORELOOP)
1481 				dp->dad_loopbackprobe += V_nd6_mmaxtries;
1482 			/*
1483 			 * Send an NS immediately and increase dad_count by
1484 			 * V_nd6_mmaxtries - 1.
1485 			 */
1486 			nd6_dad_ns_output(dp, ifa);
1487 			dp->dad_count =
1488 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1489 			nd6_dad_starttimer(dp,
1490 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1491 			goto done;
1492 		} else {
1493 			/*
1494 			 * We are done with DAD.  No NA came, no NS came.
1495 			 * No duplicate address found.  Check IFDISABLED flag
1496 			 * again in case that it is changed between the
1497 			 * beginning of this function and here.
1498 			 */
1499 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1500 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1501 
1502 			nd6log((LOG_DEBUG,
1503 			    "%s: DAD complete for %s - no duplicates found\n",
1504 			    if_name(ifa->ifa_ifp),
1505 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1506 			if (dp->dad_ns_lcount > 0)
1507 				log(LOG_ERR, "%s: DAD completed while "
1508 				    "a looped back NS message is detected "
1509 				    "during DAD for %s.\n",
1510 				    if_name(ifa->ifa_ifp),
1511 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1512 		}
1513 	}
1514 err:
1515 	nd6_dad_del(dp);
1516 done:
1517 	CURVNET_RESTORE();
1518 }
1519 
1520 static void
1521 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1522 {
1523 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1524 	struct ifnet *ifp;
1525 	char ip6buf[INET6_ADDRSTRLEN];
1526 
1527 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1528 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1529 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1530 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1531 	    dp->dad_na_icount);
1532 
1533 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1534 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1535 
1536 	ifp = ifa->ifa_ifp;
1537 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1538 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1539 	log(LOG_ERR, "%s: manual intervention required\n",
1540 	    if_name(ifp));
1541 
1542 	/*
1543 	 * If the address is a link-local address formed from an interface
1544 	 * identifier based on the hardware address which is supposed to be
1545 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1546 	 * operation on the interface SHOULD be disabled.
1547 	 * [RFC 4862, Section 5.4.5]
1548 	 */
1549 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1550 		struct in6_addr in6;
1551 
1552 		/*
1553 		 * To avoid over-reaction, we only apply this logic when we are
1554 		 * very sure that hardware addresses are supposed to be unique.
1555 		 */
1556 		switch (ifp->if_type) {
1557 		case IFT_ETHER:
1558 		case IFT_FDDI:
1559 		case IFT_ATM:
1560 		case IFT_IEEE1394:
1561 #ifdef IFT_IEEE80211
1562 		case IFT_IEEE80211:
1563 #endif
1564 		case IFT_INFINIBAND:
1565 			in6 = ia->ia_addr.sin6_addr;
1566 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1567 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1568 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1569 				log(LOG_ERR, "%s: possible hardware address "
1570 				    "duplication detected, disable IPv6\n",
1571 				    if_name(ifp));
1572 			}
1573 			break;
1574 		}
1575 	}
1576 }
1577 
1578 static void
1579 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1580 {
1581 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1582 	struct ifnet *ifp = ifa->ifa_ifp;
1583 	int i;
1584 
1585 	dp->dad_ns_tcount++;
1586 	if ((ifp->if_flags & IFF_UP) == 0) {
1587 		return;
1588 	}
1589 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1590 		return;
1591 	}
1592 
1593 	dp->dad_ns_ocount++;
1594 	if (V_dad_enhanced != 0) {
1595 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1596 			dp->dad_nonce[i] = arc4random();
1597 		/*
1598 		 * XXXHRS: Note that in the case that
1599 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1600 		 * different nonces can be looped back in an unexpected
1601 		 * order.  The current implementation recognizes only
1602 		 * the latest nonce on the sender side.  Practically it
1603 		 * should work well in almost all cases.
1604 		 */
1605 	}
1606 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL,
1607 	    (uint8_t *)&dp->dad_nonce[0]);
1608 }
1609 
1610 static void
1611 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1612 {
1613 	struct in6_ifaddr *ia;
1614 	struct ifnet *ifp;
1615 	const struct in6_addr *taddr6;
1616 	struct dadq *dp;
1617 
1618 	if (ifa == NULL)
1619 		panic("ifa == NULL in nd6_dad_ns_input");
1620 
1621 	ia = (struct in6_ifaddr *)ifa;
1622 	ifp = ifa->ifa_ifp;
1623 	taddr6 = &ia->ia_addr.sin6_addr;
1624 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1625 	if (V_dad_enhanced == 0)
1626 		ndopt_nonce = NULL;
1627 	dp = nd6_dad_find(ifa, ndopt_nonce);
1628 	if (dp == NULL)
1629 		return;
1630 
1631 	dp->dad_ns_icount++;
1632 	nd6_dad_rele(dp);
1633 }
1634 
1635 static void
1636 nd6_dad_na_input(struct ifaddr *ifa)
1637 {
1638 	struct dadq *dp;
1639 
1640 	if (ifa == NULL)
1641 		panic("ifa == NULL in nd6_dad_na_input");
1642 
1643 	dp = nd6_dad_find(ifa, NULL);
1644 	if (dp != NULL) {
1645 		dp->dad_na_icount++;
1646 		nd6_dad_rele(dp);
1647 	}
1648 }
1649