nd6_nbr.c revision 211193
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: head/sys/netinet6/nd6_nbr.c 211193 2010-08-11 20:18:19Z will $");
34
35#include "opt_inet.h"
36#include "opt_inet6.h"
37#include "opt_ipsec.h"
38#include "opt_mpath.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/malloc.h>
43#include <sys/lock.h>
44#include <sys/rwlock.h>
45#include <sys/mbuf.h>
46#include <sys/socket.h>
47#include <sys/sockio.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/errno.h>
51#include <sys/syslog.h>
52#include <sys/queue.h>
53#include <sys/callout.h>
54
55#include <net/if.h>
56#include <net/if_types.h>
57#include <net/if_dl.h>
58#include <net/if_var.h>
59#include <net/route.h>
60#ifdef RADIX_MPATH
61#include <net/radix_mpath.h>
62#endif
63
64#include <netinet/in.h>
65#include <netinet/in_var.h>
66#include <net/if_llatbl.h>
67#define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68#include <netinet6/in6_var.h>
69#include <netinet6/in6_ifattach.h>
70#include <netinet/ip6.h>
71#include <netinet6/ip6_var.h>
72#include <netinet6/scope6_var.h>
73#include <netinet6/nd6.h>
74#include <netinet/icmp6.h>
75#include <netinet/ip_carp.h>
76
77#define SDL(s) ((struct sockaddr_dl *)s)
78
79struct dadq;
80static struct dadq *nd6_dad_find(struct ifaddr *);
81static void nd6_dad_starttimer(struct dadq *, int);
82static void nd6_dad_stoptimer(struct dadq *);
83static void nd6_dad_timer(struct dadq *);
84static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
85static void nd6_dad_ns_input(struct ifaddr *);
86static void nd6_dad_na_input(struct ifaddr *);
87
88VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
89VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
90#define	V_dad_ignore_ns			VNET(dad_ignore_ns)
91#define	V_dad_maxtry			VNET(dad_maxtry)
92
93/*
94 * Input a Neighbor Solicitation Message.
95 *
96 * Based on RFC 2461
97 * Based on RFC 2462 (duplicate address detection)
98 */
99void
100nd6_ns_input(struct mbuf *m, int off, int icmp6len)
101{
102	struct ifnet *ifp = m->m_pkthdr.rcvif;
103	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
104	struct nd_neighbor_solicit *nd_ns;
105	struct in6_addr saddr6 = ip6->ip6_src;
106	struct in6_addr daddr6 = ip6->ip6_dst;
107	struct in6_addr taddr6;
108	struct in6_addr myaddr6;
109	char *lladdr = NULL;
110	struct ifaddr *ifa = NULL;
111	int lladdrlen = 0;
112	int anycast = 0, proxy = 0, tentative = 0;
113	int tlladdr;
114	union nd_opts ndopts;
115	struct sockaddr_dl *proxydl = NULL;
116	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
117
118#ifndef PULLDOWN_TEST
119	IP6_EXTHDR_CHECK(m, off, icmp6len,);
120	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
121#else
122	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
123	if (nd_ns == NULL) {
124		ICMP6STAT_INC(icp6s_tooshort);
125		return;
126	}
127#endif
128	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
129	taddr6 = nd_ns->nd_ns_target;
130	if (in6_setscope(&taddr6, ifp, NULL) != 0)
131		goto bad;
132
133	if (ip6->ip6_hlim != 255) {
134		nd6log((LOG_ERR,
135		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
136		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
137		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
138		goto bad;
139	}
140
141	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
142		/* dst has to be a solicited node multicast address. */
143		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
144		    /* don't check ifindex portion */
145		    daddr6.s6_addr32[1] == 0 &&
146		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
147		    daddr6.s6_addr8[12] == 0xff) {
148			; /* good */
149		} else {
150			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
151			    "(wrong ip6 dst)\n"));
152			goto bad;
153		}
154	} else if (!V_nd6_onlink_ns_rfc4861) {
155		struct sockaddr_in6 src_sa6;
156
157		/*
158		 * According to recent IETF discussions, it is not a good idea
159		 * to accept a NS from an address which would not be deemed
160		 * to be a neighbor otherwise.  This point is expected to be
161		 * clarified in future revisions of the specification.
162		 */
163		bzero(&src_sa6, sizeof(src_sa6));
164		src_sa6.sin6_family = AF_INET6;
165		src_sa6.sin6_len = sizeof(src_sa6);
166		src_sa6.sin6_addr = saddr6;
167		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
168			nd6log((LOG_INFO, "nd6_ns_input: "
169				"NS packet from non-neighbor\n"));
170			goto bad;
171		}
172	}
173
174	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
175		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
176		goto bad;
177	}
178
179	icmp6len -= sizeof(*nd_ns);
180	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
181	if (nd6_options(&ndopts) < 0) {
182		nd6log((LOG_INFO,
183		    "nd6_ns_input: invalid ND option, ignored\n"));
184		/* nd6_options have incremented stats */
185		goto freeit;
186	}
187
188	if (ndopts.nd_opts_src_lladdr) {
189		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
190		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
191	}
192
193	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
194		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
195		    "(link-layer address option)\n"));
196		goto bad;
197	}
198
199	/*
200	 * Attaching target link-layer address to the NA?
201	 * (RFC 2461 7.2.4)
202	 *
203	 * NS IP dst is unicast/anycast			MUST NOT add
204	 * NS IP dst is solicited-node multicast	MUST add
205	 *
206	 * In implementation, we add target link-layer address by default.
207	 * We do not add one in MUST NOT cases.
208	 */
209	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
210		tlladdr = 0;
211	else
212		tlladdr = 1;
213
214	/*
215	 * Target address (taddr6) must be either:
216	 * (1) Valid unicast/anycast address for my receiving interface,
217	 * (2) Unicast address for which I'm offering proxy service, or
218	 * (3) "tentative" address on which DAD is being performed.
219	 */
220	/* (1) and (3) check. */
221	if (ifp->if_carp)
222		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
223	if (ifa == NULL)
224		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
225
226	/* (2) check. */
227	if (ifa == NULL) {
228		struct rtentry *rt;
229		struct sockaddr_in6 tsin6;
230		int need_proxy;
231#ifdef RADIX_MPATH
232		struct route_in6 ro;
233#endif
234
235		bzero(&tsin6, sizeof tsin6);
236		tsin6.sin6_len = sizeof(struct sockaddr_in6);
237		tsin6.sin6_family = AF_INET6;
238		tsin6.sin6_addr = taddr6;
239
240#ifdef RADIX_MPATH
241		bzero(&ro, sizeof(ro));
242		ro.ro_dst = tsin6;
243		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
244		rt = ro.ro_rt;
245#else
246		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
247#endif
248		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
249		    rt->rt_gateway->sa_family == AF_LINK);
250		if (rt)
251			RTFREE_LOCKED(rt);
252		if (need_proxy) {
253			/*
254			 * proxy NDP for single entry
255			 */
256			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
257				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
258			if (ifa) {
259				proxy = 1;
260				proxydl = SDL(rt->rt_gateway);
261			}
262		}
263	}
264	if (ifa == NULL) {
265		/*
266		 * We've got an NS packet, and we don't have that adddress
267		 * assigned for us.  We MUST silently ignore it.
268		 * See RFC2461 7.2.3.
269		 */
270		goto freeit;
271	}
272	myaddr6 = *IFA_IN6(ifa);
273	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
274	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
275	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
276		goto freeit;
277
278	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
279		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
280		    "(if %d, NS packet %d)\n",
281		    ip6_sprintf(ip6bufs, &taddr6),
282		    ifp->if_addrlen, lladdrlen - 2));
283		goto bad;
284	}
285
286	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
287		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
288		    ip6_sprintf(ip6bufs, &saddr6)));
289		goto freeit;
290	}
291
292	/*
293	 * We have neighbor solicitation packet, with target address equals to
294	 * one of my tentative address.
295	 *
296	 * src addr	how to process?
297	 * ---		---
298	 * multicast	of course, invalid (rejected in ip6_input)
299	 * unicast	somebody is doing address resolution -> ignore
300	 * unspec	dup address detection
301	 *
302	 * The processing is defined in RFC 2462.
303	 */
304	if (tentative) {
305		/*
306		 * If source address is unspecified address, it is for
307		 * duplicate address detection.
308		 *
309		 * If not, the packet is for addess resolution;
310		 * silently ignore it.
311		 */
312		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
313			nd6_dad_ns_input(ifa);
314
315		goto freeit;
316	}
317
318	/*
319	 * If the source address is unspecified address, entries must not
320	 * be created or updated.
321	 * It looks that sender is performing DAD.  Output NA toward
322	 * all-node multicast address, to tell the sender that I'm using
323	 * the address.
324	 * S bit ("solicited") must be zero.
325	 */
326	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
327		struct in6_addr in6_all;
328
329		in6_all = in6addr_linklocal_allnodes;
330		if (in6_setscope(&in6_all, ifp, NULL) != 0)
331			goto bad;
332		nd6_na_output(ifp, &in6_all, &taddr6,
333		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
334		    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
335		    tlladdr, (struct sockaddr *)proxydl);
336		goto freeit;
337	}
338
339	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
340	    ND_NEIGHBOR_SOLICIT, 0);
341
342	nd6_na_output(ifp, &saddr6, &taddr6,
343	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
344	    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
345	    tlladdr, (struct sockaddr *)proxydl);
346 freeit:
347	if (ifa != NULL)
348		ifa_free(ifa);
349	m_freem(m);
350	return;
351
352 bad:
353	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
354		ip6_sprintf(ip6bufs, &saddr6)));
355	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
356		ip6_sprintf(ip6bufs, &daddr6)));
357	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
358		ip6_sprintf(ip6bufs, &taddr6)));
359	ICMP6STAT_INC(icp6s_badns);
360	if (ifa != NULL)
361		ifa_free(ifa);
362	m_freem(m);
363}
364
365/*
366 * Output a Neighbor Solicitation Message. Caller specifies:
367 *	- ICMP6 header source IP6 address
368 *	- ND6 header target IP6 address
369 *	- ND6 header source datalink address
370 *
371 * Based on RFC 2461
372 * Based on RFC 2462 (duplicate address detection)
373 *
374 *   ln - for source address determination
375 *  dad - duplicate address detection
376 */
377void
378nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
379    const struct in6_addr *taddr6, struct llentry *ln, int dad)
380{
381	struct mbuf *m;
382	struct ip6_hdr *ip6;
383	struct nd_neighbor_solicit *nd_ns;
384	struct in6_addr *src, src_in;
385	struct ip6_moptions im6o;
386	int icmp6len;
387	int maxlen;
388	caddr_t mac;
389	struct route_in6 ro;
390
391	bzero(&ro, sizeof(ro));
392
393	if (IN6_IS_ADDR_MULTICAST(taddr6))
394		return;
395
396	/* estimate the size of message */
397	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
398	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
399	if (max_linkhdr + maxlen >= MCLBYTES) {
400#ifdef DIAGNOSTIC
401		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
402		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
403#endif
404		return;
405	}
406
407	MGETHDR(m, M_DONTWAIT, MT_DATA);
408	if (m && max_linkhdr + maxlen >= MHLEN) {
409		MCLGET(m, M_DONTWAIT);
410		if ((m->m_flags & M_EXT) == 0) {
411			m_free(m);
412			m = NULL;
413		}
414	}
415	if (m == NULL)
416		return;
417	m->m_pkthdr.rcvif = NULL;
418
419	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
420		m->m_flags |= M_MCAST;
421		im6o.im6o_multicast_ifp = ifp;
422		im6o.im6o_multicast_hlim = 255;
423		im6o.im6o_multicast_loop = 0;
424	}
425
426	icmp6len = sizeof(*nd_ns);
427	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
428	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
429
430	/* fill neighbor solicitation packet */
431	ip6 = mtod(m, struct ip6_hdr *);
432	ip6->ip6_flow = 0;
433	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
434	ip6->ip6_vfc |= IPV6_VERSION;
435	/* ip6->ip6_plen will be set later */
436	ip6->ip6_nxt = IPPROTO_ICMPV6;
437	ip6->ip6_hlim = 255;
438	if (daddr6)
439		ip6->ip6_dst = *daddr6;
440	else {
441		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
442		ip6->ip6_dst.s6_addr16[1] = 0;
443		ip6->ip6_dst.s6_addr32[1] = 0;
444		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
445		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
446		ip6->ip6_dst.s6_addr8[12] = 0xff;
447		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
448			goto bad;
449	}
450	if (!dad) {
451		struct ifaddr *ifa;
452
453		/*
454		 * RFC2461 7.2.2:
455		 * "If the source address of the packet prompting the
456		 * solicitation is the same as one of the addresses assigned
457		 * to the outgoing interface, that address SHOULD be placed
458		 * in the IP Source Address of the outgoing solicitation.
459		 * Otherwise, any one of the addresses assigned to the
460		 * interface should be used."
461		 *
462		 * We use the source address for the prompting packet
463		 * (saddr6), if:
464		 * - saddr6 is given from the caller (by giving "ln"), and
465		 * - saddr6 belongs to the outgoing interface.
466		 * Otherwise, we perform the source address selection as usual.
467		 */
468		struct ip6_hdr *hip6;		/* hold ip6 */
469		struct in6_addr *hsrc = NULL;
470
471		if ((ln != NULL) && ln->la_hold) {
472			/*
473			 * assuming every packet in la_hold has the same IP
474			 * header
475			 */
476			hip6 = mtod(ln->la_hold, struct ip6_hdr *);
477			/* XXX pullup? */
478			if (sizeof(*hip6) < ln->la_hold->m_len)
479				hsrc = &hip6->ip6_src;
480			else
481				hsrc = NULL;
482		}
483		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
484		    hsrc)) != NULL) {
485			src = hsrc;
486			ifa_free(ifa);
487		} else {
488			int error;
489			struct sockaddr_in6 dst_sa;
490
491			bzero(&dst_sa, sizeof(dst_sa));
492			dst_sa.sin6_family = AF_INET6;
493			dst_sa.sin6_len = sizeof(dst_sa);
494			dst_sa.sin6_addr = ip6->ip6_dst;
495
496			error = in6_selectsrc(&dst_sa, NULL,
497			    NULL, &ro, NULL, NULL, &src_in);
498			if (error) {
499				char ip6buf[INET6_ADDRSTRLEN];
500				nd6log((LOG_DEBUG,
501				    "nd6_ns_output: source can't be "
502				    "determined: dst=%s, error=%d\n",
503				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
504				    error));
505				goto bad;
506			}
507			src = &src_in;
508		}
509	} else {
510		/*
511		 * Source address for DAD packet must always be IPv6
512		 * unspecified address. (0::0)
513		 * We actually don't have to 0-clear the address (we did it
514		 * above), but we do so here explicitly to make the intention
515		 * clearer.
516		 */
517		bzero(&src_in, sizeof(src_in));
518		src = &src_in;
519	}
520	ip6->ip6_src = *src;
521	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
522	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
523	nd_ns->nd_ns_code = 0;
524	nd_ns->nd_ns_reserved = 0;
525	nd_ns->nd_ns_target = *taddr6;
526	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
527
528	/*
529	 * Add source link-layer address option.
530	 *
531	 *				spec		implementation
532	 *				---		---
533	 * DAD packet			MUST NOT	do not add the option
534	 * there's no link layer address:
535	 *				impossible	do not add the option
536	 * there's link layer address:
537	 *	Multicast NS		MUST add one	add the option
538	 *	Unicast NS		SHOULD add one	add the option
539	 */
540	if (!dad && (mac = nd6_ifptomac(ifp))) {
541		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
542		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
543		/* 8 byte alignments... */
544		optlen = (optlen + 7) & ~7;
545
546		m->m_pkthdr.len += optlen;
547		m->m_len += optlen;
548		icmp6len += optlen;
549		bzero((caddr_t)nd_opt, optlen);
550		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
551		nd_opt->nd_opt_len = optlen >> 3;
552		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
553	}
554
555	ip6->ip6_plen = htons((u_short)icmp6len);
556	nd_ns->nd_ns_cksum = 0;
557	nd_ns->nd_ns_cksum =
558	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
559
560	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
561	icmp6_ifstat_inc(ifp, ifs6_out_msg);
562	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
563	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
564
565	if (ro.ro_rt) {		/* we don't cache this route. */
566		RTFREE(ro.ro_rt);
567	}
568	return;
569
570  bad:
571	if (ro.ro_rt) {
572		RTFREE(ro.ro_rt);
573	}
574	m_freem(m);
575	return;
576}
577
578/*
579 * Neighbor advertisement input handling.
580 *
581 * Based on RFC 2461
582 * Based on RFC 2462 (duplicate address detection)
583 *
584 * the following items are not implemented yet:
585 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
586 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
587 */
588void
589nd6_na_input(struct mbuf *m, int off, int icmp6len)
590{
591	struct ifnet *ifp = m->m_pkthdr.rcvif;
592	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
593	struct nd_neighbor_advert *nd_na;
594	struct in6_addr daddr6 = ip6->ip6_dst;
595	struct in6_addr taddr6;
596	int flags;
597	int is_router;
598	int is_solicited;
599	int is_override;
600	char *lladdr = NULL;
601	int lladdrlen = 0;
602	int checklink = 0;
603	struct ifaddr *ifa;
604	struct llentry *ln = NULL;
605	union nd_opts ndopts;
606	struct mbuf *chain = NULL;
607	struct sockaddr_in6 sin6;
608	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
609
610	if (ip6->ip6_hlim != 255) {
611		nd6log((LOG_ERR,
612		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
613		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
614		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
615		goto bad;
616	}
617
618#ifndef PULLDOWN_TEST
619	IP6_EXTHDR_CHECK(m, off, icmp6len,);
620	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
621#else
622	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
623	if (nd_na == NULL) {
624		ICMP6STAT_INC(icp6s_tooshort);
625		return;
626	}
627#endif
628
629	flags = nd_na->nd_na_flags_reserved;
630	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
631	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
632	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
633
634	taddr6 = nd_na->nd_na_target;
635	if (in6_setscope(&taddr6, ifp, NULL))
636		goto bad;	/* XXX: impossible */
637
638	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
639		nd6log((LOG_ERR,
640		    "nd6_na_input: invalid target address %s\n",
641		    ip6_sprintf(ip6bufs, &taddr6)));
642		goto bad;
643	}
644	if (IN6_IS_ADDR_MULTICAST(&daddr6))
645		if (is_solicited) {
646			nd6log((LOG_ERR,
647			    "nd6_na_input: a solicited adv is multicasted\n"));
648			goto bad;
649		}
650
651	icmp6len -= sizeof(*nd_na);
652	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
653	if (nd6_options(&ndopts) < 0) {
654		nd6log((LOG_INFO,
655		    "nd6_na_input: invalid ND option, ignored\n"));
656		/* nd6_options have incremented stats */
657		goto freeit;
658	}
659
660	if (ndopts.nd_opts_tgt_lladdr) {
661		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
662		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
663	}
664
665	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
666
667	/*
668	 * Target address matches one of my interface address.
669	 *
670	 * If my address is tentative, this means that there's somebody
671	 * already using the same address as mine.  This indicates DAD failure.
672	 * This is defined in RFC 2462.
673	 *
674	 * Otherwise, process as defined in RFC 2461.
675	 */
676	if (ifa
677	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
678		ifa_free(ifa);
679		nd6_dad_na_input(ifa);
680		goto freeit;
681	}
682
683	/* Just for safety, maybe unnecessary. */
684	if (ifa) {
685		ifa_free(ifa);
686		log(LOG_ERR,
687		    "nd6_na_input: duplicate IP6 address %s\n",
688		    ip6_sprintf(ip6bufs, &taddr6));
689		goto freeit;
690	}
691
692	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
693		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
694		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
695		    ifp->if_addrlen, lladdrlen - 2));
696		goto bad;
697	}
698
699	/*
700	 * If no neighbor cache entry is found, NA SHOULD silently be
701	 * discarded.
702	 */
703	IF_AFDATA_LOCK(ifp);
704	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
705	IF_AFDATA_UNLOCK(ifp);
706	if (ln == NULL) {
707		goto freeit;
708	}
709
710	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
711		/*
712		 * If the link-layer has address, and no lladdr option came,
713		 * discard the packet.
714		 */
715		if (ifp->if_addrlen && lladdr == NULL) {
716			goto freeit;
717		}
718
719		/*
720		 * Record link-layer address, and update the state.
721		 */
722		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
723		ln->la_flags |= LLE_VALID;
724		if (is_solicited) {
725			ln->ln_state = ND6_LLINFO_REACHABLE;
726			ln->ln_byhint = 0;
727			if (!ND6_LLINFO_PERMANENT(ln)) {
728				nd6_llinfo_settimer_locked(ln,
729				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
730			}
731		} else {
732			ln->ln_state = ND6_LLINFO_STALE;
733			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
734		}
735		if ((ln->ln_router = is_router) != 0) {
736			/*
737			 * This means a router's state has changed from
738			 * non-reachable to probably reachable, and might
739			 * affect the status of associated prefixes..
740			 */
741			checklink = 1;
742		}
743	} else {
744		int llchange;
745
746		/*
747		 * Check if the link-layer address has changed or not.
748		 */
749		if (lladdr == NULL)
750			llchange = 0;
751		else {
752			if (ln->la_flags & LLE_VALID) {
753				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
754					llchange = 1;
755				else
756					llchange = 0;
757			} else
758				llchange = 1;
759		}
760
761		/*
762		 * This is VERY complex.  Look at it with care.
763		 *
764		 * override solicit lladdr llchange	action
765		 *					(L: record lladdr)
766		 *
767		 *	0	0	n	--	(2c)
768		 *	0	0	y	n	(2b) L
769		 *	0	0	y	y	(1)    REACHABLE->STALE
770		 *	0	1	n	--	(2c)   *->REACHABLE
771		 *	0	1	y	n	(2b) L *->REACHABLE
772		 *	0	1	y	y	(1)    REACHABLE->STALE
773		 *	1	0	n	--	(2a)
774		 *	1	0	y	n	(2a) L
775		 *	1	0	y	y	(2a) L *->STALE
776		 *	1	1	n	--	(2a)   *->REACHABLE
777		 *	1	1	y	n	(2a) L *->REACHABLE
778		 *	1	1	y	y	(2a) L *->REACHABLE
779		 */
780		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
781			/*
782			 * If state is REACHABLE, make it STALE.
783			 * no other updates should be done.
784			 */
785			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
786				ln->ln_state = ND6_LLINFO_STALE;
787				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
788			}
789			goto freeit;
790		} else if (is_override				   /* (2a) */
791			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
792			|| lladdr == NULL) {			   /* (2c) */
793			/*
794			 * Update link-local address, if any.
795			 */
796			if (lladdr != NULL) {
797				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
798				ln->la_flags |= LLE_VALID;
799			}
800
801			/*
802			 * If solicited, make the state REACHABLE.
803			 * If not solicited and the link-layer address was
804			 * changed, make it STALE.
805			 */
806			if (is_solicited) {
807				ln->ln_state = ND6_LLINFO_REACHABLE;
808				ln->ln_byhint = 0;
809				if (!ND6_LLINFO_PERMANENT(ln)) {
810					nd6_llinfo_settimer_locked(ln,
811					    (long)ND_IFINFO(ifp)->reachable * hz);
812				}
813			} else {
814				if (lladdr != NULL && llchange) {
815					ln->ln_state = ND6_LLINFO_STALE;
816					nd6_llinfo_settimer_locked(ln,
817					    (long)V_nd6_gctimer * hz);
818				}
819			}
820		}
821
822		if (ln->ln_router && !is_router) {
823			/*
824			 * The peer dropped the router flag.
825			 * Remove the sender from the Default Router List and
826			 * update the Destination Cache entries.
827			 */
828			struct nd_defrouter *dr;
829			struct in6_addr *in6;
830
831			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
832
833			/*
834			 * Lock to protect the default router list.
835			 * XXX: this might be unnecessary, since this function
836			 * is only called under the network software interrupt
837			 * context.  However, we keep it just for safety.
838			 */
839			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
840			if (dr)
841				defrtrlist_del(dr);
842			else if (!V_ip6_forwarding) {
843				/*
844				 * Even if the neighbor is not in the default
845				 * router list, the neighbor may be used
846				 * as a next hop for some destinations
847				 * (e.g. redirect case). So we must
848				 * call rt6_flush explicitly.
849				 */
850				rt6_flush(&ip6->ip6_src, ifp);
851			}
852		}
853		ln->ln_router = is_router;
854	}
855        /* XXX - QL
856	 *  Does this matter?
857	 *  rt->rt_flags &= ~RTF_REJECT;
858	 */
859	ln->la_asked = 0;
860	if (ln->la_hold) {
861		struct mbuf *m_hold, *m_hold_next;
862
863		/*
864		 * reset the la_hold in advance, to explicitly
865		 * prevent a la_hold lookup in nd6_output()
866		 * (wouldn't happen, though...)
867		 */
868		for (m_hold = ln->la_hold, ln->la_hold = NULL;
869		    m_hold; m_hold = m_hold_next) {
870			m_hold_next = m_hold->m_nextpkt;
871			m_hold->m_nextpkt = NULL;
872			/*
873			 * we assume ifp is not a loopback here, so just set
874			 * the 2nd argument as the 1st one.
875			 */
876			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
877		}
878	}
879 freeit:
880	if (ln != NULL) {
881		if (chain)
882			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
883		LLE_WUNLOCK(ln);
884
885		if (chain)
886			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
887	}
888	if (checklink)
889		pfxlist_onlink_check();
890
891	m_freem(m);
892	return;
893
894 bad:
895	if (ln != NULL)
896		LLE_WUNLOCK(ln);
897
898	ICMP6STAT_INC(icp6s_badna);
899	m_freem(m);
900}
901
902/*
903 * Neighbor advertisement output handling.
904 *
905 * Based on RFC 2461
906 *
907 * the following items are not implemented yet:
908 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
909 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
910 *
911 * tlladdr - 1 if include target link-layer address
912 * sdl0 - sockaddr_dl (= proxy NA) or NULL
913 */
914void
915nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
916    const struct in6_addr *taddr6, u_long flags, int tlladdr,
917    struct sockaddr *sdl0)
918{
919	struct mbuf *m;
920	struct ip6_hdr *ip6;
921	struct nd_neighbor_advert *nd_na;
922	struct ip6_moptions im6o;
923	struct in6_addr src, daddr6;
924	struct sockaddr_in6 dst_sa;
925	int icmp6len, maxlen, error;
926	caddr_t mac = NULL;
927	struct route_in6 ro;
928
929	bzero(&ro, sizeof(ro));
930
931	daddr6 = *daddr6_0;	/* make a local copy for modification */
932
933	/* estimate the size of message */
934	maxlen = sizeof(*ip6) + sizeof(*nd_na);
935	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
936	if (max_linkhdr + maxlen >= MCLBYTES) {
937#ifdef DIAGNOSTIC
938		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
939		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
940#endif
941		return;
942	}
943
944	MGETHDR(m, M_DONTWAIT, MT_DATA);
945	if (m && max_linkhdr + maxlen >= MHLEN) {
946		MCLGET(m, M_DONTWAIT);
947		if ((m->m_flags & M_EXT) == 0) {
948			m_free(m);
949			m = NULL;
950		}
951	}
952	if (m == NULL)
953		return;
954	m->m_pkthdr.rcvif = NULL;
955
956	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
957		m->m_flags |= M_MCAST;
958		im6o.im6o_multicast_ifp = ifp;
959		im6o.im6o_multicast_hlim = 255;
960		im6o.im6o_multicast_loop = 0;
961	}
962
963	icmp6len = sizeof(*nd_na);
964	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
965	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
966
967	/* fill neighbor advertisement packet */
968	ip6 = mtod(m, struct ip6_hdr *);
969	ip6->ip6_flow = 0;
970	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
971	ip6->ip6_vfc |= IPV6_VERSION;
972	ip6->ip6_nxt = IPPROTO_ICMPV6;
973	ip6->ip6_hlim = 255;
974	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
975		/* reply to DAD */
976		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
977		daddr6.s6_addr16[1] = 0;
978		daddr6.s6_addr32[1] = 0;
979		daddr6.s6_addr32[2] = 0;
980		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
981		if (in6_setscope(&daddr6, ifp, NULL))
982			goto bad;
983
984		flags &= ~ND_NA_FLAG_SOLICITED;
985	}
986	ip6->ip6_dst = daddr6;
987	bzero(&dst_sa, sizeof(struct sockaddr_in6));
988	dst_sa.sin6_family = AF_INET6;
989	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
990	dst_sa.sin6_addr = daddr6;
991
992	/*
993	 * Select a source whose scope is the same as that of the dest.
994	 */
995	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
996	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &src);
997	if (error) {
998		char ip6buf[INET6_ADDRSTRLEN];
999		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1000		    "determined: dst=%s, error=%d\n",
1001		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1002		goto bad;
1003	}
1004	ip6->ip6_src = src;
1005	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1006	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1007	nd_na->nd_na_code = 0;
1008	nd_na->nd_na_target = *taddr6;
1009	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1010
1011	/*
1012	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1013	 * see nd6_ns_input() for details.
1014	 * Basically, if NS packet is sent to unicast/anycast addr,
1015	 * target lladdr option SHOULD NOT be included.
1016	 */
1017	if (tlladdr) {
1018		/*
1019		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1020		 * lladdr in sdl0.  If we are not proxying (sending NA for
1021		 * my address) use lladdr configured for the interface.
1022		 */
1023		if (sdl0 == NULL) {
1024			if (ifp->if_carp)
1025				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1026			if (mac == NULL)
1027				mac = nd6_ifptomac(ifp);
1028		} else if (sdl0->sa_family == AF_LINK) {
1029			struct sockaddr_dl *sdl;
1030			sdl = (struct sockaddr_dl *)sdl0;
1031			if (sdl->sdl_alen == ifp->if_addrlen)
1032				mac = LLADDR(sdl);
1033		}
1034	}
1035	if (tlladdr && mac) {
1036		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1037		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1038
1039		/* roundup to 8 bytes alignment! */
1040		optlen = (optlen + 7) & ~7;
1041
1042		m->m_pkthdr.len += optlen;
1043		m->m_len += optlen;
1044		icmp6len += optlen;
1045		bzero((caddr_t)nd_opt, optlen);
1046		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1047		nd_opt->nd_opt_len = optlen >> 3;
1048		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1049	} else
1050		flags &= ~ND_NA_FLAG_OVERRIDE;
1051
1052	ip6->ip6_plen = htons((u_short)icmp6len);
1053	nd_na->nd_na_flags_reserved = flags;
1054	nd_na->nd_na_cksum = 0;
1055	nd_na->nd_na_cksum =
1056	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1057
1058	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1059	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1060	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1061	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1062
1063	if (ro.ro_rt) {		/* we don't cache this route. */
1064		RTFREE(ro.ro_rt);
1065	}
1066	return;
1067
1068  bad:
1069	if (ro.ro_rt) {
1070		RTFREE(ro.ro_rt);
1071	}
1072	m_freem(m);
1073	return;
1074}
1075
1076caddr_t
1077nd6_ifptomac(struct ifnet *ifp)
1078{
1079	switch (ifp->if_type) {
1080	case IFT_ARCNET:
1081	case IFT_ETHER:
1082	case IFT_FDDI:
1083	case IFT_IEEE1394:
1084#ifdef IFT_L2VLAN
1085	case IFT_L2VLAN:
1086#endif
1087#ifdef IFT_IEEE80211
1088	case IFT_IEEE80211:
1089#endif
1090#ifdef IFT_CARP
1091	case IFT_CARP:
1092#endif
1093	case IFT_BRIDGE:
1094	case IFT_ISO88025:
1095		return IF_LLADDR(ifp);
1096	default:
1097		return NULL;
1098	}
1099}
1100
1101struct dadq {
1102	TAILQ_ENTRY(dadq) dad_list;
1103	struct ifaddr *dad_ifa;
1104	int dad_count;		/* max NS to send */
1105	int dad_ns_tcount;	/* # of trials to send NS */
1106	int dad_ns_ocount;	/* NS sent so far */
1107	int dad_ns_icount;
1108	int dad_na_icount;
1109	struct callout dad_timer_ch;
1110	struct vnet *dad_vnet;
1111};
1112
1113static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1114VNET_DEFINE(int, dad_init) = 0;
1115#define	V_dadq				VNET(dadq)
1116#define	V_dad_init			VNET(dad_init)
1117
1118static struct dadq *
1119nd6_dad_find(struct ifaddr *ifa)
1120{
1121	struct dadq *dp;
1122
1123	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1124		if (dp->dad_ifa == ifa)
1125			return dp;
1126	}
1127	return NULL;
1128}
1129
1130static void
1131nd6_dad_starttimer(struct dadq *dp, int ticks)
1132{
1133
1134	callout_reset(&dp->dad_timer_ch, ticks,
1135	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1136}
1137
1138static void
1139nd6_dad_stoptimer(struct dadq *dp)
1140{
1141
1142	callout_stop(&dp->dad_timer_ch);
1143}
1144
1145/*
1146 * Start Duplicate Address Detection (DAD) for specified interface address.
1147 */
1148void
1149nd6_dad_start(struct ifaddr *ifa, int delay)
1150{
1151	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1152	struct dadq *dp;
1153	char ip6buf[INET6_ADDRSTRLEN];
1154
1155	if (!V_dad_init) {
1156		TAILQ_INIT(&V_dadq);
1157		V_dad_init++;
1158	}
1159
1160	/*
1161	 * If we don't need DAD, don't do it.
1162	 * There are several cases:
1163	 * - DAD is disabled (ip6_dad_count == 0)
1164	 * - the interface address is anycast
1165	 */
1166	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1167		log(LOG_DEBUG,
1168			"nd6_dad_start: called with non-tentative address "
1169			"%s(%s)\n",
1170			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1171			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1172		return;
1173	}
1174	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1175		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1176		return;
1177	}
1178	if (!V_ip6_dad_count) {
1179		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1180		return;
1181	}
1182	if (ifa->ifa_ifp == NULL)
1183		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1184	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1185		return;
1186	}
1187	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1188		return;
1189	if (nd6_dad_find(ifa) != NULL) {
1190		/* DAD already in progress */
1191		return;
1192	}
1193
1194	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1195	if (dp == NULL) {
1196		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1197			"%s(%s)\n",
1198			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1199			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1200		return;
1201	}
1202	bzero(dp, sizeof(*dp));
1203	callout_init(&dp->dad_timer_ch, 0);
1204#ifdef VIMAGE
1205	dp->dad_vnet = curvnet;
1206#endif
1207	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1208
1209	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1210	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1211
1212	/*
1213	 * Send NS packet for DAD, ip6_dad_count times.
1214	 * Note that we must delay the first transmission, if this is the
1215	 * first packet to be sent from the interface after interface
1216	 * (re)initialization.
1217	 */
1218	dp->dad_ifa = ifa;
1219	ifa_ref(ifa);	/* just for safety */
1220	dp->dad_count = V_ip6_dad_count;
1221	dp->dad_ns_icount = dp->dad_na_icount = 0;
1222	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1223	if (delay == 0) {
1224		nd6_dad_ns_output(dp, ifa);
1225		nd6_dad_starttimer(dp,
1226		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1227	} else {
1228		nd6_dad_starttimer(dp, delay);
1229	}
1230}
1231
1232/*
1233 * terminate DAD unconditionally.  used for address removals.
1234 */
1235void
1236nd6_dad_stop(struct ifaddr *ifa)
1237{
1238	struct dadq *dp;
1239
1240	if (!V_dad_init)
1241		return;
1242	dp = nd6_dad_find(ifa);
1243	if (!dp) {
1244		/* DAD wasn't started yet */
1245		return;
1246	}
1247
1248	nd6_dad_stoptimer(dp);
1249
1250	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1251	free(dp, M_IP6NDP);
1252	dp = NULL;
1253	ifa_free(ifa);
1254}
1255
1256static void
1257nd6_dad_timer(struct dadq *dp)
1258{
1259	CURVNET_SET(dp->dad_vnet);
1260	int s;
1261	struct ifaddr *ifa = dp->dad_ifa;
1262	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1263	char ip6buf[INET6_ADDRSTRLEN];
1264
1265	s = splnet();		/* XXX */
1266
1267	/* Sanity check */
1268	if (ia == NULL) {
1269		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1270		goto done;
1271	}
1272	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1273		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1274			"%s(%s)\n",
1275			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1276			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1277		goto done;
1278	}
1279	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1280		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1281			"%s(%s)\n",
1282			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1283			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1284		goto done;
1285	}
1286
1287	/* timeouted with IFF_{RUNNING,UP} check */
1288	if (dp->dad_ns_tcount > V_dad_maxtry) {
1289		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1290		    if_name(ifa->ifa_ifp)));
1291
1292		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1293		free(dp, M_IP6NDP);
1294		dp = NULL;
1295		ifa_free(ifa);
1296		goto done;
1297	}
1298
1299	/* Need more checks? */
1300	if (dp->dad_ns_ocount < dp->dad_count) {
1301		/*
1302		 * We have more NS to go.  Send NS packet for DAD.
1303		 */
1304		nd6_dad_ns_output(dp, ifa);
1305		nd6_dad_starttimer(dp,
1306		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1307	} else {
1308		/*
1309		 * We have transmitted sufficient number of DAD packets.
1310		 * See what we've got.
1311		 */
1312		int duplicate;
1313
1314		duplicate = 0;
1315
1316		if (dp->dad_na_icount) {
1317			/*
1318			 * the check is in nd6_dad_na_input(),
1319			 * but just in case
1320			 */
1321			duplicate++;
1322		}
1323
1324		if (dp->dad_ns_icount) {
1325			/* We've seen NS, means DAD has failed. */
1326			duplicate++;
1327		}
1328
1329		if (duplicate) {
1330			/* (*dp) will be freed in nd6_dad_duplicated() */
1331			dp = NULL;
1332			nd6_dad_duplicated(ifa);
1333		} else {
1334			/*
1335			 * We are done with DAD.  No NA came, no NS came.
1336			 * No duplicate address found.
1337			 */
1338			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1339
1340			nd6log((LOG_DEBUG,
1341			    "%s: DAD complete for %s - no duplicates found\n",
1342			    if_name(ifa->ifa_ifp),
1343			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1344
1345			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1346			free(dp, M_IP6NDP);
1347			dp = NULL;
1348			ifa_free(ifa);
1349		}
1350	}
1351
1352done:
1353	splx(s);
1354	CURVNET_RESTORE();
1355}
1356
1357void
1358nd6_dad_duplicated(struct ifaddr *ifa)
1359{
1360	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1361	struct ifnet *ifp;
1362	struct dadq *dp;
1363	char ip6buf[INET6_ADDRSTRLEN];
1364
1365	dp = nd6_dad_find(ifa);
1366	if (dp == NULL) {
1367		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1368		return;
1369	}
1370
1371	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1372	    "NS in/out=%d/%d, NA in=%d\n",
1373	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1374	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1375
1376	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1377	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1378
1379	/* We are done with DAD, with duplicate address found. (failure) */
1380	nd6_dad_stoptimer(dp);
1381
1382	ifp = ifa->ifa_ifp;
1383	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1384	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1385	log(LOG_ERR, "%s: manual intervention required\n",
1386	    if_name(ifp));
1387
1388	/*
1389	 * If the address is a link-local address formed from an interface
1390	 * identifier based on the hardware address which is supposed to be
1391	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1392	 * operation on the interface SHOULD be disabled.
1393	 * [RFC 4862, Section 5.4.5]
1394	 */
1395	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1396		struct in6_addr in6;
1397
1398		/*
1399		 * To avoid over-reaction, we only apply this logic when we are
1400		 * very sure that hardware addresses are supposed to be unique.
1401		 */
1402		switch (ifp->if_type) {
1403		case IFT_ETHER:
1404		case IFT_FDDI:
1405		case IFT_ATM:
1406		case IFT_IEEE1394:
1407#ifdef IFT_IEEE80211
1408		case IFT_IEEE80211:
1409#endif
1410			in6 = ia->ia_addr.sin6_addr;
1411			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1412			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1413				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1414				log(LOG_ERR, "%s: possible hardware address "
1415				    "duplication detected, disable IPv6\n",
1416				    if_name(ifp));
1417			}
1418			break;
1419		}
1420	}
1421
1422	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1423	free(dp, M_IP6NDP);
1424	dp = NULL;
1425	ifa_free(ifa);
1426}
1427
1428static void
1429nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1430{
1431	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1432	struct ifnet *ifp = ifa->ifa_ifp;
1433
1434	dp->dad_ns_tcount++;
1435	if ((ifp->if_flags & IFF_UP) == 0) {
1436		return;
1437	}
1438	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1439		return;
1440	}
1441
1442	dp->dad_ns_ocount++;
1443	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1444}
1445
1446static void
1447nd6_dad_ns_input(struct ifaddr *ifa)
1448{
1449	struct in6_ifaddr *ia;
1450	struct ifnet *ifp;
1451	const struct in6_addr *taddr6;
1452	struct dadq *dp;
1453	int duplicate;
1454
1455	if (ifa == NULL)
1456		panic("ifa == NULL in nd6_dad_ns_input");
1457
1458	ia = (struct in6_ifaddr *)ifa;
1459	ifp = ifa->ifa_ifp;
1460	taddr6 = &ia->ia_addr.sin6_addr;
1461	duplicate = 0;
1462	dp = nd6_dad_find(ifa);
1463
1464	/* Quickhack - completely ignore DAD NS packets */
1465	if (V_dad_ignore_ns) {
1466		char ip6buf[INET6_ADDRSTRLEN];
1467		nd6log((LOG_INFO,
1468		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1469		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1470		    if_name(ifa->ifa_ifp)));
1471		return;
1472	}
1473
1474	/*
1475	 * if I'm yet to start DAD, someone else started using this address
1476	 * first.  I have a duplicate and you win.
1477	 */
1478	if (dp == NULL || dp->dad_ns_ocount == 0)
1479		duplicate++;
1480
1481	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1482
1483	if (duplicate) {
1484		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1485		nd6_dad_duplicated(ifa);
1486	} else {
1487		/*
1488		 * not sure if I got a duplicate.
1489		 * increment ns count and see what happens.
1490		 */
1491		if (dp)
1492			dp->dad_ns_icount++;
1493	}
1494}
1495
1496static void
1497nd6_dad_na_input(struct ifaddr *ifa)
1498{
1499	struct dadq *dp;
1500
1501	if (ifa == NULL)
1502		panic("ifa == NULL in nd6_dad_na_input");
1503
1504	dp = nd6_dad_find(ifa);
1505	if (dp)
1506		dp->dad_na_icount++;
1507
1508	/* remove the address. */
1509	nd6_dad_duplicated(ifa);
1510}
1511