nd6_nbr.c revision 226340
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 226340 2011-10-13 13:33:23Z glebius $");
34
35#include "opt_inet.h"
36#include "opt_inet6.h"
37#include "opt_ipsec.h"
38#include "opt_mpath.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/malloc.h>
43#include <sys/lock.h>
44#include <sys/rwlock.h>
45#include <sys/mbuf.h>
46#include <sys/socket.h>
47#include <sys/sockio.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/errno.h>
51#include <sys/syslog.h>
52#include <sys/queue.h>
53#include <sys/callout.h>
54
55#include <net/if.h>
56#include <net/if_types.h>
57#include <net/if_dl.h>
58#include <net/if_var.h>
59#include <net/route.h>
60#ifdef RADIX_MPATH
61#include <net/radix_mpath.h>
62#endif
63
64#include <netinet/in.h>
65#include <netinet/in_var.h>
66#include <net/if_llatbl.h>
67#define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68#include <netinet6/in6_var.h>
69#include <netinet6/in6_ifattach.h>
70#include <netinet/ip6.h>
71#include <netinet6/ip6_var.h>
72#include <netinet6/scope6_var.h>
73#include <netinet6/nd6.h>
74#include <netinet/icmp6.h>
75#include <netinet/ip_carp.h>
76#include <netinet6/send.h>
77
78#define SDL(s) ((struct sockaddr_dl *)s)
79
80struct dadq;
81static struct dadq *nd6_dad_find(struct ifaddr *);
82static void nd6_dad_starttimer(struct dadq *, int);
83static void nd6_dad_stoptimer(struct dadq *);
84static void nd6_dad_timer(struct dadq *);
85static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86static void nd6_dad_ns_input(struct ifaddr *);
87static void nd6_dad_na_input(struct ifaddr *);
88
89VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
90VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
91#define	V_dad_ignore_ns			VNET(dad_ignore_ns)
92#define	V_dad_maxtry			VNET(dad_maxtry)
93
94/*
95 * Input a Neighbor Solicitation Message.
96 *
97 * Based on RFC 2461
98 * Based on RFC 2462 (duplicate address detection)
99 */
100void
101nd6_ns_input(struct mbuf *m, int off, int icmp6len)
102{
103	struct ifnet *ifp = m->m_pkthdr.rcvif;
104	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
105	struct nd_neighbor_solicit *nd_ns;
106	struct in6_addr saddr6 = ip6->ip6_src;
107	struct in6_addr daddr6 = ip6->ip6_dst;
108	struct in6_addr taddr6;
109	struct in6_addr myaddr6;
110	char *lladdr = NULL;
111	struct ifaddr *ifa = NULL;
112	int lladdrlen = 0;
113	int anycast = 0, proxy = 0, tentative = 0;
114	int tlladdr;
115	int rflag;
116	union nd_opts ndopts;
117	struct sockaddr_dl proxydl;
118	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
119
120	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
121	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
122		rflag = 0;
123#ifndef PULLDOWN_TEST
124	IP6_EXTHDR_CHECK(m, off, icmp6len,);
125	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
126#else
127	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
128	if (nd_ns == NULL) {
129		ICMP6STAT_INC(icp6s_tooshort);
130		return;
131	}
132#endif
133	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
134	taddr6 = nd_ns->nd_ns_target;
135	if (in6_setscope(&taddr6, ifp, NULL) != 0)
136		goto bad;
137
138	if (ip6->ip6_hlim != 255) {
139		nd6log((LOG_ERR,
140		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
141		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
142		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
143		goto bad;
144	}
145
146	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
147		/* dst has to be a solicited node multicast address. */
148		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
149		    /* don't check ifindex portion */
150		    daddr6.s6_addr32[1] == 0 &&
151		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
152		    daddr6.s6_addr8[12] == 0xff) {
153			; /* good */
154		} else {
155			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
156			    "(wrong ip6 dst)\n"));
157			goto bad;
158		}
159	} else if (!V_nd6_onlink_ns_rfc4861) {
160		struct sockaddr_in6 src_sa6;
161
162		/*
163		 * According to recent IETF discussions, it is not a good idea
164		 * to accept a NS from an address which would not be deemed
165		 * to be a neighbor otherwise.  This point is expected to be
166		 * clarified in future revisions of the specification.
167		 */
168		bzero(&src_sa6, sizeof(src_sa6));
169		src_sa6.sin6_family = AF_INET6;
170		src_sa6.sin6_len = sizeof(src_sa6);
171		src_sa6.sin6_addr = saddr6;
172		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
173			nd6log((LOG_INFO, "nd6_ns_input: "
174				"NS packet from non-neighbor\n"));
175			goto bad;
176		}
177	}
178
179	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
180		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
181		goto bad;
182	}
183
184	icmp6len -= sizeof(*nd_ns);
185	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
186	if (nd6_options(&ndopts) < 0) {
187		nd6log((LOG_INFO,
188		    "nd6_ns_input: invalid ND option, ignored\n"));
189		/* nd6_options have incremented stats */
190		goto freeit;
191	}
192
193	if (ndopts.nd_opts_src_lladdr) {
194		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
195		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
196	}
197
198	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
199		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
200		    "(link-layer address option)\n"));
201		goto bad;
202	}
203
204	/*
205	 * Attaching target link-layer address to the NA?
206	 * (RFC 2461 7.2.4)
207	 *
208	 * NS IP dst is unicast/anycast			MUST NOT add
209	 * NS IP dst is solicited-node multicast	MUST add
210	 *
211	 * In implementation, we add target link-layer address by default.
212	 * We do not add one in MUST NOT cases.
213	 */
214	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
215		tlladdr = 0;
216	else
217		tlladdr = 1;
218
219	/*
220	 * Target address (taddr6) must be either:
221	 * (1) Valid unicast/anycast address for my receiving interface,
222	 * (2) Unicast address for which I'm offering proxy service, or
223	 * (3) "tentative" address on which DAD is being performed.
224	 */
225	/* (1) and (3) check. */
226	if (ifp->if_carp)
227		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
228	if (ifa == NULL)
229		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
230
231	/* (2) check. */
232	if (ifa == NULL) {
233		struct rtentry *rt;
234		struct sockaddr_in6 tsin6;
235		int need_proxy;
236#ifdef RADIX_MPATH
237		struct route_in6 ro;
238#endif
239
240		bzero(&tsin6, sizeof tsin6);
241		tsin6.sin6_len = sizeof(struct sockaddr_in6);
242		tsin6.sin6_family = AF_INET6;
243		tsin6.sin6_addr = taddr6;
244
245#ifdef RADIX_MPATH
246		bzero(&ro, sizeof(ro));
247		ro.ro_dst = tsin6;
248		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
249		rt = ro.ro_rt;
250#else
251		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
252#endif
253		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
254		    rt->rt_gateway->sa_family == AF_LINK);
255		if (rt != NULL) {
256			/*
257			 * Make a copy while we can be sure that rt_gateway
258			 * is still stable before unlocking to avoid lock
259			 * order problems.  proxydl will only be used if
260			 * proxy will be set in the next block.
261			 */
262			if (need_proxy)
263				proxydl = *SDL(rt->rt_gateway);
264			RTFREE_LOCKED(rt);
265		}
266		if (need_proxy) {
267			/*
268			 * proxy NDP for single entry
269			 */
270			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
271				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
272			if (ifa)
273				proxy = 1;
274		}
275	}
276	if (ifa == NULL) {
277		/*
278		 * We've got an NS packet, and we don't have that adddress
279		 * assigned for us.  We MUST silently ignore it.
280		 * See RFC2461 7.2.3.
281		 */
282		goto freeit;
283	}
284	myaddr6 = *IFA_IN6(ifa);
285	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
286	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
287	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
288		goto freeit;
289
290	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
291		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
292		    "(if %d, NS packet %d)\n",
293		    ip6_sprintf(ip6bufs, &taddr6),
294		    ifp->if_addrlen, lladdrlen - 2));
295		goto bad;
296	}
297
298	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
299		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
300		    ip6_sprintf(ip6bufs, &saddr6)));
301		goto freeit;
302	}
303
304	/*
305	 * We have neighbor solicitation packet, with target address equals to
306	 * one of my tentative address.
307	 *
308	 * src addr	how to process?
309	 * ---		---
310	 * multicast	of course, invalid (rejected in ip6_input)
311	 * unicast	somebody is doing address resolution -> ignore
312	 * unspec	dup address detection
313	 *
314	 * The processing is defined in RFC 2462.
315	 */
316	if (tentative) {
317		/*
318		 * If source address is unspecified address, it is for
319		 * duplicate address detection.
320		 *
321		 * If not, the packet is for addess resolution;
322		 * silently ignore it.
323		 */
324		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
325			nd6_dad_ns_input(ifa);
326
327		goto freeit;
328	}
329
330	/*
331	 * If the source address is unspecified address, entries must not
332	 * be created or updated.
333	 * It looks that sender is performing DAD.  Output NA toward
334	 * all-node multicast address, to tell the sender that I'm using
335	 * the address.
336	 * S bit ("solicited") must be zero.
337	 */
338	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
339		struct in6_addr in6_all;
340
341		in6_all = in6addr_linklocal_allnodes;
342		if (in6_setscope(&in6_all, ifp, NULL) != 0)
343			goto bad;
344		nd6_na_output(ifp, &in6_all, &taddr6,
345		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
346		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL);
347		goto freeit;
348	}
349
350	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
351	    ND_NEIGHBOR_SOLICIT, 0);
352
353	nd6_na_output(ifp, &saddr6, &taddr6,
354	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
356	    proxy ? (struct sockaddr *)&proxydl : NULL);
357 freeit:
358	if (ifa != NULL)
359		ifa_free(ifa);
360	m_freem(m);
361	return;
362
363 bad:
364	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
365		ip6_sprintf(ip6bufs, &saddr6)));
366	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
367		ip6_sprintf(ip6bufs, &daddr6)));
368	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
369		ip6_sprintf(ip6bufs, &taddr6)));
370	ICMP6STAT_INC(icp6s_badns);
371	if (ifa != NULL)
372		ifa_free(ifa);
373	m_freem(m);
374}
375
376/*
377 * Output a Neighbor Solicitation Message. Caller specifies:
378 *	- ICMP6 header source IP6 address
379 *	- ND6 header target IP6 address
380 *	- ND6 header source datalink address
381 *
382 * Based on RFC 2461
383 * Based on RFC 2462 (duplicate address detection)
384 *
385 *   ln - for source address determination
386 *  dad - duplicate address detection
387 */
388void
389nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
390    const struct in6_addr *taddr6, struct llentry *ln, int dad)
391{
392	struct mbuf *m;
393	struct m_tag *mtag;
394	struct ip6_hdr *ip6;
395	struct nd_neighbor_solicit *nd_ns;
396	struct ip6_moptions im6o;
397	int icmp6len;
398	int maxlen;
399	caddr_t mac;
400	struct route_in6 ro;
401
402	if (IN6_IS_ADDR_MULTICAST(taddr6))
403		return;
404
405	/* estimate the size of message */
406	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
407	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
408	if (max_linkhdr + maxlen >= MCLBYTES) {
409#ifdef DIAGNOSTIC
410		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
411		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
412#endif
413		return;
414	}
415
416	MGETHDR(m, M_DONTWAIT, MT_DATA);
417	if (m && max_linkhdr + maxlen >= MHLEN) {
418		MCLGET(m, M_DONTWAIT);
419		if ((m->m_flags & M_EXT) == 0) {
420			m_free(m);
421			m = NULL;
422		}
423	}
424	if (m == NULL)
425		return;
426	m->m_pkthdr.rcvif = NULL;
427
428	bzero(&ro, sizeof(ro));
429
430	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
431		m->m_flags |= M_MCAST;
432		im6o.im6o_multicast_ifp = ifp;
433		im6o.im6o_multicast_hlim = 255;
434		im6o.im6o_multicast_loop = 0;
435	}
436
437	icmp6len = sizeof(*nd_ns);
438	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
439	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
440
441	/* fill neighbor solicitation packet */
442	ip6 = mtod(m, struct ip6_hdr *);
443	ip6->ip6_flow = 0;
444	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
445	ip6->ip6_vfc |= IPV6_VERSION;
446	/* ip6->ip6_plen will be set later */
447	ip6->ip6_nxt = IPPROTO_ICMPV6;
448	ip6->ip6_hlim = 255;
449	if (daddr6)
450		ip6->ip6_dst = *daddr6;
451	else {
452		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
453		ip6->ip6_dst.s6_addr16[1] = 0;
454		ip6->ip6_dst.s6_addr32[1] = 0;
455		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
456		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
457		ip6->ip6_dst.s6_addr8[12] = 0xff;
458		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
459			goto bad;
460	}
461	if (!dad) {
462		struct ifaddr *ifa;
463
464		/*
465		 * RFC2461 7.2.2:
466		 * "If the source address of the packet prompting the
467		 * solicitation is the same as one of the addresses assigned
468		 * to the outgoing interface, that address SHOULD be placed
469		 * in the IP Source Address of the outgoing solicitation.
470		 * Otherwise, any one of the addresses assigned to the
471		 * interface should be used."
472		 *
473		 * We use the source address for the prompting packet
474		 * (saddr6), if:
475		 * - saddr6 is given from the caller (by giving "ln"), and
476		 * - saddr6 belongs to the outgoing interface.
477		 * Otherwise, we perform the source address selection as usual.
478		 */
479		struct in6_addr *hsrc;
480
481		hsrc = NULL;
482		if (ln != NULL) {
483			LLE_RLOCK(ln);
484			if (ln->la_hold != NULL) {
485				struct ip6_hdr *hip6;		/* hold ip6 */
486
487				/*
488				 * assuming every packet in la_hold has the same IP
489				 * header
490				 */
491				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
492				/* XXX pullup? */
493				if (sizeof(*hip6) < ln->la_hold->m_len) {
494					ip6->ip6_src = hip6->ip6_src;
495					hsrc = &hip6->ip6_src;
496				}
497			}
498			LLE_RUNLOCK(ln);
499		}
500		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
501		    hsrc)) != NULL) {
502			/* ip6_src set already. */
503			ifa_free(ifa);
504		} else {
505			int error;
506			struct sockaddr_in6 dst_sa;
507			struct in6_addr src_in;
508
509			bzero(&dst_sa, sizeof(dst_sa));
510			dst_sa.sin6_family = AF_INET6;
511			dst_sa.sin6_len = sizeof(dst_sa);
512			dst_sa.sin6_addr = ip6->ip6_dst;
513
514			error = in6_selectsrc(&dst_sa, NULL,
515			    NULL, &ro, NULL, NULL, &src_in);
516			if (error) {
517				char ip6buf[INET6_ADDRSTRLEN];
518				nd6log((LOG_DEBUG,
519				    "nd6_ns_output: source can't be "
520				    "determined: dst=%s, error=%d\n",
521				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
522				    error));
523				goto bad;
524			}
525			ip6->ip6_src = src_in;
526		}
527	} else {
528		/*
529		 * Source address for DAD packet must always be IPv6
530		 * unspecified address. (0::0)
531		 * We actually don't have to 0-clear the address (we did it
532		 * above), but we do so here explicitly to make the intention
533		 * clearer.
534		 */
535		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
536	}
537	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
538	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
539	nd_ns->nd_ns_code = 0;
540	nd_ns->nd_ns_reserved = 0;
541	nd_ns->nd_ns_target = *taddr6;
542	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
543
544	/*
545	 * Add source link-layer address option.
546	 *
547	 *				spec		implementation
548	 *				---		---
549	 * DAD packet			MUST NOT	do not add the option
550	 * there's no link layer address:
551	 *				impossible	do not add the option
552	 * there's link layer address:
553	 *	Multicast NS		MUST add one	add the option
554	 *	Unicast NS		SHOULD add one	add the option
555	 */
556	if (!dad && (mac = nd6_ifptomac(ifp))) {
557		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
558		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
559		/* 8 byte alignments... */
560		optlen = (optlen + 7) & ~7;
561
562		m->m_pkthdr.len += optlen;
563		m->m_len += optlen;
564		icmp6len += optlen;
565		bzero((caddr_t)nd_opt, optlen);
566		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
567		nd_opt->nd_opt_len = optlen >> 3;
568		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
569	}
570
571	ip6->ip6_plen = htons((u_short)icmp6len);
572	nd_ns->nd_ns_cksum = 0;
573	nd_ns->nd_ns_cksum =
574	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
575
576	if (send_sendso_input_hook != NULL) {
577		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
578			sizeof(unsigned short), M_NOWAIT);
579		if (mtag == NULL)
580			goto bad;
581		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
582		m_tag_prepend(m, mtag);
583	}
584
585	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
586	icmp6_ifstat_inc(ifp, ifs6_out_msg);
587	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
588	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
589
590	if (ro.ro_rt) {		/* we don't cache this route. */
591		RTFREE(ro.ro_rt);
592	}
593	return;
594
595  bad:
596	if (ro.ro_rt) {
597		RTFREE(ro.ro_rt);
598	}
599	m_freem(m);
600	return;
601}
602
603/*
604 * Neighbor advertisement input handling.
605 *
606 * Based on RFC 2461
607 * Based on RFC 2462 (duplicate address detection)
608 *
609 * the following items are not implemented yet:
610 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
611 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
612 */
613void
614nd6_na_input(struct mbuf *m, int off, int icmp6len)
615{
616	struct ifnet *ifp = m->m_pkthdr.rcvif;
617	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
618	struct nd_neighbor_advert *nd_na;
619	struct in6_addr daddr6 = ip6->ip6_dst;
620	struct in6_addr taddr6;
621	int flags;
622	int is_router;
623	int is_solicited;
624	int is_override;
625	char *lladdr = NULL;
626	int lladdrlen = 0;
627	int checklink = 0;
628	struct ifaddr *ifa;
629	struct llentry *ln = NULL;
630	union nd_opts ndopts;
631	struct mbuf *chain = NULL;
632	struct m_tag *mtag;
633	struct sockaddr_in6 sin6;
634	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
635
636	if (ip6->ip6_hlim != 255) {
637		nd6log((LOG_ERR,
638		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
639		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
640		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
641		goto bad;
642	}
643
644#ifndef PULLDOWN_TEST
645	IP6_EXTHDR_CHECK(m, off, icmp6len,);
646	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
647#else
648	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
649	if (nd_na == NULL) {
650		ICMP6STAT_INC(icp6s_tooshort);
651		return;
652	}
653#endif
654
655	flags = nd_na->nd_na_flags_reserved;
656	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
657	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
658	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
659
660	taddr6 = nd_na->nd_na_target;
661	if (in6_setscope(&taddr6, ifp, NULL))
662		goto bad;	/* XXX: impossible */
663
664	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
665		nd6log((LOG_ERR,
666		    "nd6_na_input: invalid target address %s\n",
667		    ip6_sprintf(ip6bufs, &taddr6)));
668		goto bad;
669	}
670	if (IN6_IS_ADDR_MULTICAST(&daddr6))
671		if (is_solicited) {
672			nd6log((LOG_ERR,
673			    "nd6_na_input: a solicited adv is multicasted\n"));
674			goto bad;
675		}
676
677	icmp6len -= sizeof(*nd_na);
678	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
679	if (nd6_options(&ndopts) < 0) {
680		nd6log((LOG_INFO,
681		    "nd6_na_input: invalid ND option, ignored\n"));
682		/* nd6_options have incremented stats */
683		goto freeit;
684	}
685
686	if (ndopts.nd_opts_tgt_lladdr) {
687		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
688		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
689	}
690
691	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
692
693	/*
694	 * Target address matches one of my interface address.
695	 *
696	 * If my address is tentative, this means that there's somebody
697	 * already using the same address as mine.  This indicates DAD failure.
698	 * This is defined in RFC 2462.
699	 *
700	 * Otherwise, process as defined in RFC 2461.
701	 */
702	if (ifa
703	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
704		ifa_free(ifa);
705		nd6_dad_na_input(ifa);
706		goto freeit;
707	}
708
709	/* Just for safety, maybe unnecessary. */
710	if (ifa) {
711		ifa_free(ifa);
712		log(LOG_ERR,
713		    "nd6_na_input: duplicate IP6 address %s\n",
714		    ip6_sprintf(ip6bufs, &taddr6));
715		goto freeit;
716	}
717
718	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
719		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
720		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
721		    ifp->if_addrlen, lladdrlen - 2));
722		goto bad;
723	}
724
725	/*
726	 * If no neighbor cache entry is found, NA SHOULD silently be
727	 * discarded.
728	 */
729	IF_AFDATA_LOCK(ifp);
730	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
731	IF_AFDATA_UNLOCK(ifp);
732	if (ln == NULL) {
733		goto freeit;
734	}
735
736	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
737		/*
738		 * If the link-layer has address, and no lladdr option came,
739		 * discard the packet.
740		 */
741		if (ifp->if_addrlen && lladdr == NULL) {
742			goto freeit;
743		}
744
745		/*
746		 * Record link-layer address, and update the state.
747		 */
748		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
749		ln->la_flags |= LLE_VALID;
750		if (is_solicited) {
751			ln->ln_state = ND6_LLINFO_REACHABLE;
752			ln->ln_byhint = 0;
753			if (!ND6_LLINFO_PERMANENT(ln)) {
754				nd6_llinfo_settimer_locked(ln,
755				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
756			}
757		} else {
758			ln->ln_state = ND6_LLINFO_STALE;
759			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
760		}
761		if ((ln->ln_router = is_router) != 0) {
762			/*
763			 * This means a router's state has changed from
764			 * non-reachable to probably reachable, and might
765			 * affect the status of associated prefixes..
766			 */
767			checklink = 1;
768		}
769	} else {
770		int llchange;
771
772		/*
773		 * Check if the link-layer address has changed or not.
774		 */
775		if (lladdr == NULL)
776			llchange = 0;
777		else {
778			if (ln->la_flags & LLE_VALID) {
779				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
780					llchange = 1;
781				else
782					llchange = 0;
783			} else
784				llchange = 1;
785		}
786
787		/*
788		 * This is VERY complex.  Look at it with care.
789		 *
790		 * override solicit lladdr llchange	action
791		 *					(L: record lladdr)
792		 *
793		 *	0	0	n	--	(2c)
794		 *	0	0	y	n	(2b) L
795		 *	0	0	y	y	(1)    REACHABLE->STALE
796		 *	0	1	n	--	(2c)   *->REACHABLE
797		 *	0	1	y	n	(2b) L *->REACHABLE
798		 *	0	1	y	y	(1)    REACHABLE->STALE
799		 *	1	0	n	--	(2a)
800		 *	1	0	y	n	(2a) L
801		 *	1	0	y	y	(2a) L *->STALE
802		 *	1	1	n	--	(2a)   *->REACHABLE
803		 *	1	1	y	n	(2a) L *->REACHABLE
804		 *	1	1	y	y	(2a) L *->REACHABLE
805		 */
806		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
807			/*
808			 * If state is REACHABLE, make it STALE.
809			 * no other updates should be done.
810			 */
811			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
812				ln->ln_state = ND6_LLINFO_STALE;
813				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
814			}
815			goto freeit;
816		} else if (is_override				   /* (2a) */
817			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
818			|| lladdr == NULL) {			   /* (2c) */
819			/*
820			 * Update link-local address, if any.
821			 */
822			if (lladdr != NULL) {
823				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
824				ln->la_flags |= LLE_VALID;
825			}
826
827			/*
828			 * If solicited, make the state REACHABLE.
829			 * If not solicited and the link-layer address was
830			 * changed, make it STALE.
831			 */
832			if (is_solicited) {
833				ln->ln_state = ND6_LLINFO_REACHABLE;
834				ln->ln_byhint = 0;
835				if (!ND6_LLINFO_PERMANENT(ln)) {
836					nd6_llinfo_settimer_locked(ln,
837					    (long)ND_IFINFO(ifp)->reachable * hz);
838				}
839			} else {
840				if (lladdr != NULL && llchange) {
841					ln->ln_state = ND6_LLINFO_STALE;
842					nd6_llinfo_settimer_locked(ln,
843					    (long)V_nd6_gctimer * hz);
844				}
845			}
846		}
847
848		if (ln->ln_router && !is_router) {
849			/*
850			 * The peer dropped the router flag.
851			 * Remove the sender from the Default Router List and
852			 * update the Destination Cache entries.
853			 */
854			struct nd_defrouter *dr;
855			struct in6_addr *in6;
856
857			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
858
859			/*
860			 * Lock to protect the default router list.
861			 * XXX: this might be unnecessary, since this function
862			 * is only called under the network software interrupt
863			 * context.  However, we keep it just for safety.
864			 */
865			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
866			if (dr)
867				defrtrlist_del(dr);
868			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
869			    ND6_IFF_ACCEPT_RTADV) {
870				/*
871				 * Even if the neighbor is not in the default
872				 * router list, the neighbor may be used
873				 * as a next hop for some destinations
874				 * (e.g. redirect case). So we must
875				 * call rt6_flush explicitly.
876				 */
877				rt6_flush(&ip6->ip6_src, ifp);
878			}
879		}
880		ln->ln_router = is_router;
881	}
882        /* XXX - QL
883	 *  Does this matter?
884	 *  rt->rt_flags &= ~RTF_REJECT;
885	 */
886	ln->la_asked = 0;
887	if (ln->la_hold) {
888		struct mbuf *m_hold, *m_hold_next;
889
890		/*
891		 * reset the la_hold in advance, to explicitly
892		 * prevent a la_hold lookup in nd6_output()
893		 * (wouldn't happen, though...)
894		 */
895		for (m_hold = ln->la_hold, ln->la_hold = NULL;
896		    m_hold; m_hold = m_hold_next) {
897			m_hold_next = m_hold->m_nextpkt;
898			m_hold->m_nextpkt = NULL;
899			/*
900			 * we assume ifp is not a loopback here, so just set
901			 * the 2nd argument as the 1st one.
902			 */
903
904			if (send_sendso_input_hook != NULL) {
905				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
906				    sizeof(unsigned short), M_NOWAIT);
907				if (mtag == NULL)
908					goto bad;
909				m_tag_prepend(m, mtag);
910			}
911
912			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
913		}
914	}
915 freeit:
916	if (ln != NULL) {
917		if (chain)
918			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
919		LLE_WUNLOCK(ln);
920
921		if (chain)
922			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
923	}
924	if (checklink)
925		pfxlist_onlink_check();
926
927	m_freem(m);
928	return;
929
930 bad:
931	if (ln != NULL)
932		LLE_WUNLOCK(ln);
933
934	ICMP6STAT_INC(icp6s_badna);
935	m_freem(m);
936}
937
938/*
939 * Neighbor advertisement output handling.
940 *
941 * Based on RFC 2461
942 *
943 * the following items are not implemented yet:
944 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
945 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
946 *
947 * tlladdr - 1 if include target link-layer address
948 * sdl0 - sockaddr_dl (= proxy NA) or NULL
949 */
950void
951nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
952    const struct in6_addr *taddr6, u_long flags, int tlladdr,
953    struct sockaddr *sdl0)
954{
955	struct mbuf *m;
956	struct m_tag *mtag;
957	struct ip6_hdr *ip6;
958	struct nd_neighbor_advert *nd_na;
959	struct ip6_moptions im6o;
960	struct in6_addr src, daddr6;
961	struct sockaddr_in6 dst_sa;
962	int icmp6len, maxlen, error;
963	caddr_t mac = NULL;
964	struct route_in6 ro;
965
966	bzero(&ro, sizeof(ro));
967
968	daddr6 = *daddr6_0;	/* make a local copy for modification */
969
970	/* estimate the size of message */
971	maxlen = sizeof(*ip6) + sizeof(*nd_na);
972	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
973	if (max_linkhdr + maxlen >= MCLBYTES) {
974#ifdef DIAGNOSTIC
975		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
976		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
977#endif
978		return;
979	}
980
981	MGETHDR(m, M_DONTWAIT, MT_DATA);
982	if (m && max_linkhdr + maxlen >= MHLEN) {
983		MCLGET(m, M_DONTWAIT);
984		if ((m->m_flags & M_EXT) == 0) {
985			m_free(m);
986			m = NULL;
987		}
988	}
989	if (m == NULL)
990		return;
991	m->m_pkthdr.rcvif = NULL;
992
993	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
994		m->m_flags |= M_MCAST;
995		im6o.im6o_multicast_ifp = ifp;
996		im6o.im6o_multicast_hlim = 255;
997		im6o.im6o_multicast_loop = 0;
998	}
999
1000	icmp6len = sizeof(*nd_na);
1001	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1002	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1003
1004	/* fill neighbor advertisement packet */
1005	ip6 = mtod(m, struct ip6_hdr *);
1006	ip6->ip6_flow = 0;
1007	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1008	ip6->ip6_vfc |= IPV6_VERSION;
1009	ip6->ip6_nxt = IPPROTO_ICMPV6;
1010	ip6->ip6_hlim = 255;
1011	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1012		/* reply to DAD */
1013		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1014		daddr6.s6_addr16[1] = 0;
1015		daddr6.s6_addr32[1] = 0;
1016		daddr6.s6_addr32[2] = 0;
1017		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1018		if (in6_setscope(&daddr6, ifp, NULL))
1019			goto bad;
1020
1021		flags &= ~ND_NA_FLAG_SOLICITED;
1022	}
1023	ip6->ip6_dst = daddr6;
1024	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1025	dst_sa.sin6_family = AF_INET6;
1026	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1027	dst_sa.sin6_addr = daddr6;
1028
1029	/*
1030	 * Select a source whose scope is the same as that of the dest.
1031	 */
1032	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1033	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &src);
1034	if (error) {
1035		char ip6buf[INET6_ADDRSTRLEN];
1036		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1037		    "determined: dst=%s, error=%d\n",
1038		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1039		goto bad;
1040	}
1041	ip6->ip6_src = src;
1042	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1043	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1044	nd_na->nd_na_code = 0;
1045	nd_na->nd_na_target = *taddr6;
1046	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1047
1048	/*
1049	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1050	 * see nd6_ns_input() for details.
1051	 * Basically, if NS packet is sent to unicast/anycast addr,
1052	 * target lladdr option SHOULD NOT be included.
1053	 */
1054	if (tlladdr) {
1055		/*
1056		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1057		 * lladdr in sdl0.  If we are not proxying (sending NA for
1058		 * my address) use lladdr configured for the interface.
1059		 */
1060		if (sdl0 == NULL) {
1061			if (ifp->if_carp)
1062				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1063			if (mac == NULL)
1064				mac = nd6_ifptomac(ifp);
1065		} else if (sdl0->sa_family == AF_LINK) {
1066			struct sockaddr_dl *sdl;
1067			sdl = (struct sockaddr_dl *)sdl0;
1068			if (sdl->sdl_alen == ifp->if_addrlen)
1069				mac = LLADDR(sdl);
1070		}
1071	}
1072	if (tlladdr && mac) {
1073		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1074		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1075
1076		/* roundup to 8 bytes alignment! */
1077		optlen = (optlen + 7) & ~7;
1078
1079		m->m_pkthdr.len += optlen;
1080		m->m_len += optlen;
1081		icmp6len += optlen;
1082		bzero((caddr_t)nd_opt, optlen);
1083		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1084		nd_opt->nd_opt_len = optlen >> 3;
1085		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1086	} else
1087		flags &= ~ND_NA_FLAG_OVERRIDE;
1088
1089	ip6->ip6_plen = htons((u_short)icmp6len);
1090	nd_na->nd_na_flags_reserved = flags;
1091	nd_na->nd_na_cksum = 0;
1092	nd_na->nd_na_cksum =
1093	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1094
1095	if (send_sendso_input_hook != NULL) {
1096		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1097		    sizeof(unsigned short), M_NOWAIT);
1098		if (mtag == NULL)
1099			goto bad;
1100		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1101		m_tag_prepend(m, mtag);
1102	}
1103
1104	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1105	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1106	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1107	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1108
1109	if (ro.ro_rt) {		/* we don't cache this route. */
1110		RTFREE(ro.ro_rt);
1111	}
1112	return;
1113
1114  bad:
1115	if (ro.ro_rt) {
1116		RTFREE(ro.ro_rt);
1117	}
1118	m_freem(m);
1119	return;
1120}
1121
1122caddr_t
1123nd6_ifptomac(struct ifnet *ifp)
1124{
1125	switch (ifp->if_type) {
1126	case IFT_ARCNET:
1127	case IFT_ETHER:
1128	case IFT_FDDI:
1129	case IFT_IEEE1394:
1130#ifdef IFT_L2VLAN
1131	case IFT_L2VLAN:
1132#endif
1133#ifdef IFT_IEEE80211
1134	case IFT_IEEE80211:
1135#endif
1136#ifdef IFT_CARP
1137	case IFT_CARP:
1138#endif
1139	case IFT_INFINIBAND:
1140	case IFT_BRIDGE:
1141	case IFT_ISO88025:
1142		return IF_LLADDR(ifp);
1143	default:
1144		return NULL;
1145	}
1146}
1147
1148struct dadq {
1149	TAILQ_ENTRY(dadq) dad_list;
1150	struct ifaddr *dad_ifa;
1151	int dad_count;		/* max NS to send */
1152	int dad_ns_tcount;	/* # of trials to send NS */
1153	int dad_ns_ocount;	/* NS sent so far */
1154	int dad_ns_icount;
1155	int dad_na_icount;
1156	struct callout dad_timer_ch;
1157	struct vnet *dad_vnet;
1158};
1159
1160static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1161VNET_DEFINE(int, dad_init) = 0;
1162#define	V_dadq				VNET(dadq)
1163#define	V_dad_init			VNET(dad_init)
1164
1165static struct dadq *
1166nd6_dad_find(struct ifaddr *ifa)
1167{
1168	struct dadq *dp;
1169
1170	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1171		if (dp->dad_ifa == ifa)
1172			return (dp);
1173
1174	return (NULL);
1175}
1176
1177static void
1178nd6_dad_starttimer(struct dadq *dp, int ticks)
1179{
1180
1181	callout_reset(&dp->dad_timer_ch, ticks,
1182	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1183}
1184
1185static void
1186nd6_dad_stoptimer(struct dadq *dp)
1187{
1188
1189	callout_stop(&dp->dad_timer_ch);
1190}
1191
1192/*
1193 * Start Duplicate Address Detection (DAD) for specified interface address.
1194 */
1195void
1196nd6_dad_start(struct ifaddr *ifa, int delay)
1197{
1198	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1199	struct dadq *dp;
1200	char ip6buf[INET6_ADDRSTRLEN];
1201
1202	if (!V_dad_init) {
1203		TAILQ_INIT(&V_dadq);
1204		V_dad_init++;
1205	}
1206
1207	/*
1208	 * If we don't need DAD, don't do it.
1209	 * There are several cases:
1210	 * - DAD is disabled (ip6_dad_count == 0)
1211	 * - the interface address is anycast
1212	 */
1213	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1214		log(LOG_DEBUG,
1215			"nd6_dad_start: called with non-tentative address "
1216			"%s(%s)\n",
1217			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1218			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1219		return;
1220	}
1221	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1222		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1223		return;
1224	}
1225	if (!V_ip6_dad_count) {
1226		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1227		return;
1228	}
1229	if (ifa->ifa_ifp == NULL)
1230		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1231	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1232		return;
1233	}
1234	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1235		return;
1236	if (nd6_dad_find(ifa) != NULL) {
1237		/* DAD already in progress */
1238		return;
1239	}
1240
1241	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1242	if (dp == NULL) {
1243		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1244			"%s(%s)\n",
1245			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1246			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1247		return;
1248	}
1249	bzero(dp, sizeof(*dp));
1250	callout_init(&dp->dad_timer_ch, 0);
1251#ifdef VIMAGE
1252	dp->dad_vnet = curvnet;
1253#endif
1254	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1255
1256	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1257	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1258
1259	/*
1260	 * Send NS packet for DAD, ip6_dad_count times.
1261	 * Note that we must delay the first transmission, if this is the
1262	 * first packet to be sent from the interface after interface
1263	 * (re)initialization.
1264	 */
1265	dp->dad_ifa = ifa;
1266	ifa_ref(ifa);	/* just for safety */
1267	dp->dad_count = V_ip6_dad_count;
1268	dp->dad_ns_icount = dp->dad_na_icount = 0;
1269	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1270	if (delay == 0) {
1271		nd6_dad_ns_output(dp, ifa);
1272		nd6_dad_starttimer(dp,
1273		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1274	} else {
1275		nd6_dad_starttimer(dp, delay);
1276	}
1277}
1278
1279/*
1280 * terminate DAD unconditionally.  used for address removals.
1281 */
1282void
1283nd6_dad_stop(struct ifaddr *ifa)
1284{
1285	struct dadq *dp;
1286
1287	if (!V_dad_init)
1288		return;
1289	dp = nd6_dad_find(ifa);
1290	if (!dp) {
1291		/* DAD wasn't started yet */
1292		return;
1293	}
1294
1295	nd6_dad_stoptimer(dp);
1296
1297	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1298	free(dp, M_IP6NDP);
1299	dp = NULL;
1300	ifa_free(ifa);
1301}
1302
1303static void
1304nd6_dad_timer(struct dadq *dp)
1305{
1306	CURVNET_SET(dp->dad_vnet);
1307	int s;
1308	struct ifaddr *ifa = dp->dad_ifa;
1309	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1310	char ip6buf[INET6_ADDRSTRLEN];
1311
1312	s = splnet();		/* XXX */
1313
1314	/* Sanity check */
1315	if (ia == NULL) {
1316		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1317		goto done;
1318	}
1319	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1320		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1321			"%s(%s)\n",
1322			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1323			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1324		goto done;
1325	}
1326	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1327		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1328			"%s(%s)\n",
1329			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1330			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1331		goto done;
1332	}
1333
1334	/* timeouted with IFF_{RUNNING,UP} check */
1335	if (dp->dad_ns_tcount > V_dad_maxtry) {
1336		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1337		    if_name(ifa->ifa_ifp)));
1338
1339		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1340		free(dp, M_IP6NDP);
1341		dp = NULL;
1342		ifa_free(ifa);
1343		goto done;
1344	}
1345
1346	/* Need more checks? */
1347	if (dp->dad_ns_ocount < dp->dad_count) {
1348		/*
1349		 * We have more NS to go.  Send NS packet for DAD.
1350		 */
1351		nd6_dad_ns_output(dp, ifa);
1352		nd6_dad_starttimer(dp,
1353		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1354	} else {
1355		/*
1356		 * We have transmitted sufficient number of DAD packets.
1357		 * See what we've got.
1358		 */
1359		int duplicate;
1360
1361		duplicate = 0;
1362
1363		if (dp->dad_na_icount) {
1364			/*
1365			 * the check is in nd6_dad_na_input(),
1366			 * but just in case
1367			 */
1368			duplicate++;
1369		}
1370
1371		if (dp->dad_ns_icount) {
1372			/* We've seen NS, means DAD has failed. */
1373			duplicate++;
1374		}
1375
1376		if (duplicate) {
1377			/* (*dp) will be freed in nd6_dad_duplicated() */
1378			dp = NULL;
1379			nd6_dad_duplicated(ifa);
1380		} else {
1381			/*
1382			 * We are done with DAD.  No NA came, no NS came.
1383			 * No duplicate address found.
1384			 */
1385			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1386
1387			nd6log((LOG_DEBUG,
1388			    "%s: DAD complete for %s - no duplicates found\n",
1389			    if_name(ifa->ifa_ifp),
1390			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1391
1392			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1393			free(dp, M_IP6NDP);
1394			dp = NULL;
1395			ifa_free(ifa);
1396		}
1397	}
1398
1399done:
1400	splx(s);
1401	CURVNET_RESTORE();
1402}
1403
1404void
1405nd6_dad_duplicated(struct ifaddr *ifa)
1406{
1407	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1408	struct ifnet *ifp;
1409	struct dadq *dp;
1410	char ip6buf[INET6_ADDRSTRLEN];
1411
1412	dp = nd6_dad_find(ifa);
1413	if (dp == NULL) {
1414		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1415		return;
1416	}
1417
1418	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1419	    "NS in/out=%d/%d, NA in=%d\n",
1420	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1421	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1422
1423	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1424	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1425
1426	/* We are done with DAD, with duplicate address found. (failure) */
1427	nd6_dad_stoptimer(dp);
1428
1429	ifp = ifa->ifa_ifp;
1430	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1431	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1432	log(LOG_ERR, "%s: manual intervention required\n",
1433	    if_name(ifp));
1434
1435	/*
1436	 * If the address is a link-local address formed from an interface
1437	 * identifier based on the hardware address which is supposed to be
1438	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1439	 * operation on the interface SHOULD be disabled.
1440	 * [RFC 4862, Section 5.4.5]
1441	 */
1442	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1443		struct in6_addr in6;
1444
1445		/*
1446		 * To avoid over-reaction, we only apply this logic when we are
1447		 * very sure that hardware addresses are supposed to be unique.
1448		 */
1449		switch (ifp->if_type) {
1450		case IFT_ETHER:
1451		case IFT_FDDI:
1452		case IFT_ATM:
1453		case IFT_IEEE1394:
1454#ifdef IFT_IEEE80211
1455		case IFT_IEEE80211:
1456#endif
1457		case IFT_INFINIBAND:
1458			in6 = ia->ia_addr.sin6_addr;
1459			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1460			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1461				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1462				log(LOG_ERR, "%s: possible hardware address "
1463				    "duplication detected, disable IPv6\n",
1464				    if_name(ifp));
1465			}
1466			break;
1467		}
1468	}
1469
1470	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1471	free(dp, M_IP6NDP);
1472	dp = NULL;
1473	ifa_free(ifa);
1474}
1475
1476static void
1477nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1478{
1479	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1480	struct ifnet *ifp = ifa->ifa_ifp;
1481
1482	dp->dad_ns_tcount++;
1483	if ((ifp->if_flags & IFF_UP) == 0) {
1484		return;
1485	}
1486	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1487		return;
1488	}
1489
1490	dp->dad_ns_ocount++;
1491	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1492}
1493
1494static void
1495nd6_dad_ns_input(struct ifaddr *ifa)
1496{
1497	struct in6_ifaddr *ia;
1498	struct ifnet *ifp;
1499	const struct in6_addr *taddr6;
1500	struct dadq *dp;
1501	int duplicate;
1502
1503	if (ifa == NULL)
1504		panic("ifa == NULL in nd6_dad_ns_input");
1505
1506	ia = (struct in6_ifaddr *)ifa;
1507	ifp = ifa->ifa_ifp;
1508	taddr6 = &ia->ia_addr.sin6_addr;
1509	duplicate = 0;
1510	dp = nd6_dad_find(ifa);
1511
1512	/* Quickhack - completely ignore DAD NS packets */
1513	if (V_dad_ignore_ns) {
1514		char ip6buf[INET6_ADDRSTRLEN];
1515		nd6log((LOG_INFO,
1516		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1517		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1518		    if_name(ifa->ifa_ifp)));
1519		return;
1520	}
1521
1522	/*
1523	 * if I'm yet to start DAD, someone else started using this address
1524	 * first.  I have a duplicate and you win.
1525	 */
1526	if (dp == NULL || dp->dad_ns_ocount == 0)
1527		duplicate++;
1528
1529	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1530
1531	if (duplicate) {
1532		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1533		nd6_dad_duplicated(ifa);
1534	} else {
1535		/*
1536		 * not sure if I got a duplicate.
1537		 * increment ns count and see what happens.
1538		 */
1539		if (dp)
1540			dp->dad_ns_icount++;
1541	}
1542}
1543
1544static void
1545nd6_dad_na_input(struct ifaddr *ifa)
1546{
1547	struct dadq *dp;
1548
1549	if (ifa == NULL)
1550		panic("ifa == NULL in nd6_dad_na_input");
1551
1552	dp = nd6_dad_find(ifa);
1553	if (dp)
1554		dp->dad_na_icount++;
1555
1556	/* remove the address. */
1557	nd6_dad_duplicated(ifa);
1558}
1559