xref: /src/sys/netinet6/nd6_nbr.c (revision 15ebee7c2ed34d2dad8457d6595010ea4ed7e1f2)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32  */
33 
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_ipsec.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/counter.h>
41 #include <sys/eventhandler.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/if_private.h>
63 #include <net/route.h>
64 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet/icmp6.h>
76 #include <netinet/ip_carp.h>
77 #include <netinet6/send.h>
78 
79 #include <machine/atomic.h>
80 
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 MALLOC_DECLARE(M_IP6NDP);
84 
85 struct dadq;
86 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
87 static void nd6_dad_add(struct dadq *dp);
88 static void nd6_dad_del(struct dadq *dp);
89 static void nd6_dad_rele(struct dadq *);
90 static void nd6_dad_starttimer(struct dadq *, int);
91 static void nd6_dad_stoptimer(struct dadq *);
92 static void nd6_dad_timer(void *);
93 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
94 static void nd6_dad_ns_output(struct dadq *);
95 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
96 static void nd6_dad_na_input(struct ifaddr *);
97 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
98     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
99 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
100     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
101 static void nd6_queue_add(struct ifaddr *, struct in6_addr *, int, uint32_t);
102 
103 static struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *,
104     const struct in6_addr *, struct sockaddr_dl *);
105 
106 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
107 #define	V_dad_enhanced			VNET(dad_enhanced)
108 
109 SYSCTL_DECL(_net_inet6_ip6);
110 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
111     &VNET_NAME(dad_enhanced), 0,
112     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
113 
114 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
115 						   transmit DAD packet */
116 #define	V_dad_maxtry			VNET(dad_maxtry)
117 
118 VNET_DEFINE_STATIC(int, nd6_onlink_ns_rfc4861) = 0;
119 #define	V_nd6_onlink_ns_rfc4861		VNET(nd6_onlink_ns_rfc4861)
120 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_ONLINKNSRFC4861,
121     nd6_onlink_ns_rfc4861, CTLFLAG_VNET | CTLFLAG_RW,
122     &VNET_NAME(nd6_onlink_ns_rfc4861), 0,
123     "Accept 'on-link' ICMPv6 NS messages in compliance with RFC 4861");
124 
125 struct nd_queue {
126 	TAILQ_ENTRY(nd_queue) ndq_list;
127 	struct ifaddr	*ndq_ifa;
128 	struct in6_addr	ndq_daddr;
129 	uint32_t	ndq_flags;
130 	struct callout	ndq_callout;
131 };
132 
133 /*
134  * Input a Neighbor Solicitation Message.
135  *
136  * Based on RFC 2461
137  * Based on RFC 2462 (duplicate address detection)
138  */
139 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)140 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
141 {
142 	struct ifnet *ifp;
143 	struct ip6_hdr *ip6;
144 	struct nd_neighbor_solicit *nd_ns;
145 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
146 	struct ifaddr *ifa;
147 	struct sockaddr_dl proxydl;
148 	union nd_opts ndopts;
149 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
150 	char *lladdr;
151 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
152 
153 	ifa = NULL;
154 
155 	/* RFC 6980: Nodes MUST silently ignore fragments */
156 	if(m->m_flags & M_FRAGMENTED)
157 		goto freeit;
158 
159 	ifp = m->m_pkthdr.rcvif;
160 	ip6 = mtod(m, struct ip6_hdr *);
161 	if (__predict_false(ip6->ip6_hlim != 255)) {
162 		ICMP6STAT_INC(icp6s_invlhlim);
163 		nd6log((LOG_ERR,
164 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
165 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
166 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
167 		goto bads;
168 	}
169 
170 	if (m->m_len < off + icmp6len) {
171 		m = m_pullup(m, off + icmp6len);
172 		if (m == NULL) {
173 			IP6STAT_INC(ip6s_exthdrtoolong);
174 			return;
175 		}
176 	}
177 	ip6 = mtod(m, struct ip6_hdr *);
178 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
179 
180 	saddr6 = ip6->ip6_src;
181 	daddr6 = ip6->ip6_dst;
182 	taddr6 = nd_ns->nd_ns_target;
183 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
184 		goto bad;
185 
186 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
187 	if (ifp->if_inet6->nd_flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
188 		rflag = 0;
189 
190 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
191 		/* dst has to be a solicited node multicast address. */
192 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
193 		    /* don't check ifindex portion */
194 		    daddr6.s6_addr32[1] == 0 &&
195 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
196 		    daddr6.s6_addr8[12] == 0xff) {
197 			; /* good */
198 		} else {
199 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
200 			    "(wrong ip6 dst)\n"));
201 			goto bad;
202 		}
203 	} else if (!V_nd6_onlink_ns_rfc4861) {
204 		struct sockaddr_in6 src_sa6;
205 
206 		/*
207 		 * According to recent IETF discussions, it is not a good idea
208 		 * to accept a NS from an address which would not be deemed
209 		 * to be a neighbor otherwise.  This point is expected to be
210 		 * clarified in future revisions of the specification.
211 		 */
212 		bzero(&src_sa6, sizeof(src_sa6));
213 		src_sa6.sin6_family = AF_INET6;
214 		src_sa6.sin6_len = sizeof(src_sa6);
215 		src_sa6.sin6_addr = saddr6;
216 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
217 			nd6log((LOG_INFO, "nd6_ns_input: "
218 				"NS packet from non-neighbor\n"));
219 			goto bad;
220 		}
221 	}
222 
223 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
224 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
225 		goto bad;
226 	}
227 
228 	icmp6len -= sizeof(*nd_ns);
229 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
230 	if (nd6_options(&ndopts) < 0) {
231 		nd6log((LOG_INFO,
232 		    "nd6_ns_input: invalid ND option, ignored\n"));
233 		/* nd6_options have incremented stats */
234 		goto freeit;
235 	}
236 
237 	lladdr = NULL;
238 	lladdrlen = 0;
239 	if (ndopts.nd_opts_src_lladdr) {
240 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
241 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
242 	}
243 
244 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
245 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
246 		    "(link-layer address option)\n"));
247 		goto bad;
248 	}
249 
250 	/*
251 	 * Attaching target link-layer address to the NA?
252 	 * (RFC 2461 7.2.4)
253 	 *
254 	 * NS IP dst is unicast/anycast			MUST NOT add
255 	 * NS IP dst is solicited-node multicast	MUST add
256 	 *
257 	 * In implementation, we add target link-layer address by default.
258 	 * We do not add one in MUST NOT cases.
259 	 */
260 	tlladdr = 0;
261 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
262 		tlladdr |= ND6_NA_OPT_LLA;
263 
264 	/*
265 	 * Target address (taddr6) must be either:
266 	 * (1) Valid unicast/anycast address for my receiving interface,
267 	 * (2) Unicast address for which I'm offering proxy service, or
268 	 * (3) "tentative" address on which DAD is being performed.
269 	 */
270 	/* (1) and (3) check. */
271 	if (ifp->if_carp) {
272 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
273 		if (ifa != NULL)
274 			tlladdr |= ND6_NA_CARP_MASTER;
275 	} else
276 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
277 
278 	/* (2) check. */
279 	proxy = 0;
280 	if (ifa == NULL) {
281 		if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL)
282 			proxy = 1;
283 	}
284 	if (ifa == NULL) {
285 		/*
286 		 * We've got an NS packet, and we don't have that address
287 		 * assigned for us.  We MUST silently ignore it.
288 		 * See RFC2461 7.2.3.
289 		 */
290 		goto freeit;
291 	}
292 	myaddr6 = *IFA_IN6(ifa);
293 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
294 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
295 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
296 		goto freeit;
297 
298 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
299 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
300 		    "(if %d, NS packet %d)\n",
301 		    ip6_sprintf(ip6bufs, &taddr6),
302 		    ifp->if_addrlen, lladdrlen - 2));
303 		goto bad;
304 	}
305 
306 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
307 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
308 		    ip6_sprintf(ip6bufs, &saddr6)));
309 		goto freeit;
310 	}
311 
312 	/*
313 	 * We have neighbor solicitation packet, with target address equals to
314 	 * one of my tentative address.
315 	 *
316 	 * src addr	how to process?
317 	 * ---		---
318 	 * multicast	of course, invalid (rejected in ip6_input)
319 	 * unicast	somebody is doing address resolution -> ignore
320 	 * unspec	dup address detection
321 	 *
322 	 * The processing is defined in RFC 2462.
323 	 */
324 	if (tentative) {
325 		/*
326 		 * If source address is unspecified address, it is for
327 		 * duplicate address detection.
328 		 *
329 		 * If not, the packet is for addess resolution;
330 		 * silently ignore it.
331 		 */
332 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
333 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
334 
335 		goto freeit;
336 	}
337 
338 	/*
339 	 * If the Target Address is either an anycast address or a unicast
340 	 * address for which the node is providing proxy service, or the Target
341 	 * Link-Layer Address option is not included, the Override flag SHOULD
342 	 * be set to zero.  Otherwise, the Override flag SHOULD be set to one.
343 	 */
344 	if (anycast == 0 && proxy == 0 && (tlladdr & ND6_NA_OPT_LLA) != 0)
345 		rflag |= ND_NA_FLAG_OVERRIDE;
346 	/*
347 	 * If the source address is unspecified address, entries must not
348 	 * be created or updated.
349 	 * It looks that sender is performing DAD. nd6_na_output() will
350 	 * send NA toward all-node multicast address, to tell the sender
351 	 * that I'm using the address.
352 	 * S bit ("solicited") must be zero.
353 	 */
354 	if (!IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
355 		nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
356 		    ND_NEIGHBOR_SOLICIT, 0);
357 		rflag |= ND_NA_FLAG_SOLICITED;
358 	}
359 
360 	/*
361 	 * RFC 4861, anycast or proxy NA sent in response to a NS SHOULD
362 	 * be delayed by a random time between 0 and MAX_ANYCAST_DELAY_TIME
363 	 * to reduce the probability of network congestion.
364 	 */
365 	if (anycast == 0)
366 		nd6_na_output_fib(ifp, &saddr6, &taddr6, rflag, tlladdr,
367 		    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
368 	else
369 		nd6_queue_add(ifa, &saddr6, arc4random() %
370 		    (MAX_ANYCAST_DELAY_TIME * hz), ND6_QUEUE_FLAG_ANYCAST);
371  freeit:
372 	if (ifa != NULL)
373 		ifa_free(ifa);
374 	m_freem(m);
375 	return;
376 
377  bad:
378 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
379 		ip6_sprintf(ip6bufs, &saddr6)));
380 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
381 		ip6_sprintf(ip6bufs, &daddr6)));
382 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
383 		ip6_sprintf(ip6bufs, &taddr6)));
384  bads:
385 	ICMP6STAT_INC(icp6s_badns);
386 	if (ifa != NULL)
387 		ifa_free(ifa);
388 	m_freem(m);
389 }
390 
391 static struct ifaddr *
nd6_proxy_fill_sdl(struct ifnet * ifp,const struct in6_addr * taddr6,struct sockaddr_dl * sdl)392 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6,
393     struct sockaddr_dl *sdl)
394 {
395 	struct ifaddr *ifa;
396 	struct llentry *ln;
397 
398 	ifa = NULL;
399 	ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp);
400 	if (ln == NULL)
401 		return (ifa);
402 	if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) {
403 		link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type);
404 		sdl->sdl_alen = ifp->if_addrlen;
405 		bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen);
406 		LLE_RUNLOCK(ln);
407 		ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
408 		    IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
409 	} else
410 		LLE_RUNLOCK(ln);
411 
412 	return (ifa);
413 }
414 
415 /*
416  * Output a Neighbor Solicitation Message. Caller specifies:
417  *	- ICMP6 header source IP6 address
418  *	- ND6 header target IP6 address
419  *	- ND6 header source datalink address
420  *
421  * Based on RFC 2461
422  * Based on RFC 2462 (duplicate address detection)
423  *
424  *    ln - for source address determination
425  * nonce - If non-NULL, NS is used for duplicate address detection and
426  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
427  */
428 static void
nd6_ns_output_fib(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce,u_int fibnum)429 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
430     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
431     uint8_t *nonce, u_int fibnum)
432 {
433 	struct mbuf *m;
434 	struct m_tag *mtag;
435 	struct ip6_hdr *ip6;
436 	struct nd_neighbor_solicit *nd_ns;
437 	struct ip6_moptions im6o;
438 	int icmp6len;
439 	int maxlen;
440 
441 	NET_EPOCH_ASSERT();
442 
443 	if (IN6_IS_ADDR_MULTICAST(taddr6))
444 		return;
445 
446 	/* estimate the size of message */
447 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
448 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
449 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
450 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
451 	    __func__, max_linkhdr, maxlen, MCLBYTES));
452 
453 	if (max_linkhdr + maxlen > MHLEN)
454 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
455 	else
456 		m = m_gethdr(M_NOWAIT, MT_DATA);
457 	if (m == NULL)
458 		return;
459 	M_SETFIB(m, fibnum);
460 
461 	icmp6len = sizeof(*nd_ns);
462 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
463 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
464 
465 	/* fill neighbor solicitation packet */
466 	ip6 = mtod(m, struct ip6_hdr *);
467 	ip6->ip6_flow = 0;
468 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
469 	ip6->ip6_vfc |= IPV6_VERSION;
470 	/* ip6->ip6_plen will be set later */
471 	ip6->ip6_nxt = IPPROTO_ICMPV6;
472 	ip6->ip6_hlim = 255;
473 	if (daddr6)
474 		ip6->ip6_dst = *daddr6;
475 	else {
476 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
477 		ip6->ip6_dst.s6_addr16[1] = 0;
478 		ip6->ip6_dst.s6_addr32[1] = 0;
479 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
480 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
481 		ip6->ip6_dst.s6_addr8[12] = 0xff;
482 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
483 			goto bad;
484 	}
485 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
486 		m->m_flags |= M_MCAST;
487 		im6o.im6o_multicast_ifp = ifp;
488 		im6o.im6o_multicast_hlim = 255;
489 		im6o.im6o_multicast_loop = 0;
490 	}
491 	if (nonce == NULL) {
492 		char ip6buf[INET6_ADDRSTRLEN];
493 		struct ifaddr *ifa = NULL;
494 
495 		/*
496 		 * RFC2461 7.2.2:
497 		 * "If the source address of the packet prompting the
498 		 * solicitation is the same as one of the addresses assigned
499 		 * to the outgoing interface, that address SHOULD be placed
500 		 * in the IP Source Address of the outgoing solicitation.
501 		 * Otherwise, any one of the addresses assigned to the
502 		 * interface should be used."
503 		 *
504 		 * We use the source address for the prompting packet
505 		 * (saddr6), if saddr6 belongs to the outgoing interface.
506 		 * Otherwise, we perform the source address selection as usual.
507 		 */
508 		if (saddr6 != NULL)
509 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
510 		if (ifa == NULL) {
511 			int error;
512 
513 			error = in6_selectsrc_nbr(fibnum, &ip6->ip6_dst, &im6o,
514 			    ifp, &ip6->ip6_src);
515 			if (error) {
516 				nd6log((LOG_DEBUG, "%s: source can't be "
517 				    "determined: dst=%s, error=%d\n", __func__,
518 				    ip6_sprintf(ip6buf, &ip6->ip6_dst),
519 				    error));
520 				goto bad;
521 			}
522 		} else
523 			ip6->ip6_src = *saddr6;
524 
525 		if (ifp->if_carp != NULL) {
526 			/*
527 			 * Check that selected source address belongs to
528 			 * CARP addresses.
529 			 */
530 			if (ifa == NULL)
531 				ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
532 				    &ip6->ip6_src);
533 			/*
534 			 * Do not send NS for CARP address if we are not
535 			 * the CARP master.
536 			 */
537 			if (ifa != NULL && ifa->ifa_carp != NULL &&
538 			    !(*carp_master_p)(ifa)) {
539 				nd6log((LOG_DEBUG,
540 				    "nd6_ns_output: NS from BACKUP CARP address %s\n",
541 				    ip6_sprintf(ip6buf, &ip6->ip6_src)));
542 				ifa_free(ifa);
543 				goto bad;
544 			}
545 		}
546 		if (ifa != NULL)
547 			ifa_free(ifa);
548 	} else {
549 		/*
550 		 * Source address for DAD packet must always be IPv6
551 		 * unspecified address. (0::0)
552 		 * We actually don't have to 0-clear the address (we did it
553 		 * above), but we do so here explicitly to make the intention
554 		 * clearer.
555 		 */
556 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
557 	}
558 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
559 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
560 	nd_ns->nd_ns_code = 0;
561 	nd_ns->nd_ns_reserved = 0;
562 	nd_ns->nd_ns_target = *taddr6;
563 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
564 
565 	/*
566 	 * Add source link-layer address option.
567 	 *
568 	 *				spec		implementation
569 	 *				---		---
570 	 * DAD packet			MUST NOT	do not add the option
571 	 * there's no link layer address:
572 	 *				impossible	do not add the option
573 	 * there's link layer address:
574 	 *	Multicast NS		MUST add one	add the option
575 	 *	Unicast NS		SHOULD add one	add the option
576 	 */
577 	if (nonce == NULL) {
578 		struct nd_opt_hdr *nd_opt;
579 		char *mac;
580 		int optlen;
581 
582 		mac = NULL;
583 		if (ifp->if_carp)
584 			mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
585 		if (mac == NULL)
586 			mac = nd6_ifptomac(ifp);
587 
588 		if (mac != NULL) {
589 			nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
590 			optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
591 			/* 8 byte alignments... */
592 			optlen = (optlen + 7) & ~7;
593 			m->m_pkthdr.len += optlen;
594 			m->m_len += optlen;
595 			icmp6len += optlen;
596 			bzero(nd_opt, optlen);
597 			nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
598 			nd_opt->nd_opt_len = optlen >> 3;
599 			bcopy(mac, nd_opt + 1, ifp->if_addrlen);
600 		}
601 	}
602 	/*
603 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
604 	 * This behavior is documented as Enhanced Duplicate Address
605 	 * Detection in RFC 7527.
606 	 * net.inet6.ip6.dad_enhanced=0 disables this.
607 	 */
608 	if (V_dad_enhanced != 0 && nonce != NULL) {
609 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
610 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
611 		/* 8-byte alignment is required. */
612 		optlen = (optlen + 7) & ~7;
613 
614 		m->m_pkthdr.len += optlen;
615 		m->m_len += optlen;
616 		icmp6len += optlen;
617 		bzero((caddr_t)nd_opt, optlen);
618 		nd_opt->nd_opt_type = ND_OPT_NONCE;
619 		nd_opt->nd_opt_len = optlen >> 3;
620 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
621 	}
622 	ip6->ip6_plen = htons((u_short)icmp6len);
623 	nd_ns->nd_ns_cksum = 0;
624 	nd_ns->nd_ns_cksum =
625 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
626 
627 	if (send_sendso_input_hook != NULL) {
628 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
629 			sizeof(unsigned short), M_NOWAIT);
630 		if (mtag == NULL)
631 			goto bad;
632 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
633 		m_tag_prepend(m, mtag);
634 	}
635 
636 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
637 	    &im6o, NULL, NULL);
638 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
639 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
640 	ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_SOLICIT);
641 
642 	return;
643 
644   bad:
645 	m_freem(m);
646 }
647 
648 #ifndef BURN_BRIDGES
649 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce)650 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
651     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
652 {
653 
654 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
655 }
656 #endif
657 /*
658  * Neighbor advertisement input handling.
659  *
660  * Based on RFC 2461
661  * Based on RFC 2462 (duplicate address detection)
662  */
663 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)664 nd6_na_input(struct mbuf *m, int off, int icmp6len)
665 {
666 	struct ifnet *ifp;
667 	struct ip6_hdr *ip6;
668 	struct ifaddr *ifa;
669 	struct llentry *ln;
670 	struct mbuf *chain;
671 	struct nd_neighbor_advert *nd_na;
672 	struct in6_addr daddr6, taddr6;
673 	union nd_opts ndopts;
674 	u_char linkhdr[LLE_MAX_LINKHDR];
675 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
676 	char *lladdr;
677 	size_t linkhdrsize;
678 	int flags, is_override, is_router, is_solicited;
679 	int lladdr_off, lladdrlen, checklink;
680 	bool flush_holdchain = false;
681 
682 	NET_EPOCH_ASSERT();
683 
684 	chain = NULL;
685 	ln = NULL;
686 	checklink = 0;
687 
688 	/* RFC 6980: Nodes MUST silently ignore fragments */
689 	if(m->m_flags & M_FRAGMENTED)
690 		goto freeit;
691 
692 	ifp = m->m_pkthdr.rcvif;
693 	ip6 = mtod(m, struct ip6_hdr *);
694 	if (__predict_false(ip6->ip6_hlim != 255)) {
695 		ICMP6STAT_INC(icp6s_invlhlim);
696 		nd6log((LOG_ERR,
697 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
698 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
699 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
700 		goto bad;
701 	}
702 
703 	if (m->m_len < off + icmp6len) {
704 		m = m_pullup(m, off + icmp6len);
705 		if (m == NULL) {
706 			IP6STAT_INC(ip6s_exthdrtoolong);
707 			return;
708 		}
709 	}
710 	ip6 = mtod(m, struct ip6_hdr *);
711 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
712 
713 	flags = nd_na->nd_na_flags_reserved;
714 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
715 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
716 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
717 
718 	taddr6 = nd_na->nd_na_target;
719 	if (in6_setscope(&taddr6, ifp, NULL))
720 		goto bad;	/* XXX: impossible */
721 
722 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
723 		nd6log((LOG_ERR,
724 		    "nd6_na_input: invalid target address %s\n",
725 		    ip6_sprintf(ip6bufs, &taddr6)));
726 		goto bad;
727 	}
728 
729 	daddr6 = ip6->ip6_dst;
730 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
731 		if (is_solicited) {
732 			nd6log((LOG_ERR,
733 			    "nd6_na_input: a solicited adv is multicasted\n"));
734 			goto bad;
735 		}
736 
737 	icmp6len -= sizeof(*nd_na);
738 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
739 	if (nd6_options(&ndopts) < 0) {
740 		nd6log((LOG_INFO,
741 		    "nd6_na_input: invalid ND option, ignored\n"));
742 		/* nd6_options have incremented stats */
743 		goto freeit;
744 	}
745 
746 	lladdr = NULL;
747 	lladdrlen = 0;
748 	if (ndopts.nd_opts_tgt_lladdr) {
749 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
750 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
751 	}
752 
753 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
754 	if (ifa != NULL && ifa->ifa_carp != NULL) {
755 		/*
756 		 * Silently ignore NAs for CARP addresses if we are not
757 		 * the CARP master.
758 		 */
759 		if (!(*carp_master_p)(ifa)) {
760 			nd6log((LOG_DEBUG,
761 			    "nd6_na_input: NA for BACKUP CARP address %s\n",
762 			    ip6_sprintf(ip6bufs, &taddr6)));
763 			ifa_free(ifa);
764 			goto freeit;
765 		}
766 	}
767 	/*
768 	 * Target address matches one of my interface address.
769 	 *
770 	 * If my address is tentative, this means that there's somebody
771 	 * already using the same address as mine.  This indicates DAD failure.
772 	 * This is defined in RFC 2462.
773 	 *
774 	 * Otherwise, process as defined in RFC 2461.
775 	 */
776 	if (ifa
777 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
778 		nd6_dad_na_input(ifa);
779 		ifa_free(ifa);
780 		goto freeit;
781 	}
782 
783 	/* Just for safety, maybe unnecessary. */
784 	if (ifa) {
785 		ifa_free(ifa);
786 		log(LOG_ERR,
787 		    "nd6_na_input: duplicate IP6 address %s\n",
788 		    ip6_sprintf(ip6bufs, &taddr6));
789 		goto freeit;
790 	}
791 
792 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
793 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
794 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
795 		    ifp->if_addrlen, lladdrlen - 2));
796 		goto bad;
797 	}
798 
799 	/*
800 	 * If no neighbor cache entry is found, NA SHOULD silently be
801 	 * discarded.
802 	 */
803 	ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
804 	if (ln == NULL) {
805 		goto freeit;
806 	}
807 
808 	/*
809 	 * Do not try to override static entry.
810 	 */
811 	if (ln->la_flags & LLE_STATIC)
812 		goto freeit;
813 
814 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
815 		/*
816 		 * If the link-layer has address, and no lladdr option came,
817 		 * discard the packet.
818 		 */
819 		if (ifp->if_addrlen && lladdr == NULL) {
820 			goto freeit;
821 		}
822 
823 		/*
824 		 * Record link-layer address, and update the state.
825 		 */
826 		if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
827 			goto freeit;
828 
829 		flush_holdchain = true;
830 		if (is_solicited)
831 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
832 		else
833 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
834 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
835 		if ((ln->ln_router = is_router) != 0) {
836 			/*
837 			 * This means a router's state has changed from
838 			 * non-reachable to probably reachable, and might
839 			 * affect the status of associated prefixes..
840 			 */
841 			checklink = 1;
842 		}
843 	} else {
844 		int llchange;
845 
846 		/*
847 		 * Check if the link-layer address has changed or not.
848 		 */
849 		if (lladdr == NULL)
850 			llchange = 0;
851 		else {
852 			if (ln->la_flags & LLE_VALID) {
853 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
854 					llchange = 1;
855 				else
856 					llchange = 0;
857 			} else
858 				llchange = 1;
859 		}
860 
861 		/*
862 		 * This is VERY complex.  Look at it with care.
863 		 *
864 		 * override solicit lladdr llchange	action
865 		 *					(L: record lladdr)
866 		 *
867 		 *	0	0	n	--	(2c)
868 		 *	0	0	y	n	(2b) L
869 		 *	0	0	y	y	(1)    REACHABLE->STALE
870 		 *	0	1	n	--	(2c)   *->REACHABLE
871 		 *	0	1	y	n	(2b) L *->REACHABLE
872 		 *	0	1	y	y	(1)    REACHABLE->STALE
873 		 *	1	0	n	--	(2a)
874 		 *	1	0	y	n	(2a) L
875 		 *	1	0	y	y	(2a) L *->STALE
876 		 *	1	1	n	--	(2a)   *->REACHABLE
877 		 *	1	1	y	n	(2a) L *->REACHABLE
878 		 *	1	1	y	y	(2a) L *->REACHABLE
879 		 */
880 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
881 			/*
882 			 * If state is REACHABLE, make it STALE.
883 			 * no other updates should be done.
884 			 */
885 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
886 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
887 			goto freeit;
888 		} else if (is_override				   /* (2a) */
889 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
890 			|| lladdr == NULL) {			   /* (2c) */
891 			/*
892 			 * Update link-local address, if any.
893 			 */
894 			if (lladdr != NULL) {
895 				linkhdrsize = sizeof(linkhdr);
896 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
897 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
898 					goto freeit;
899 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
900 				    linkhdrsize, lladdr_off) == 0)
901 					goto freeit;
902 				EVENTHANDLER_INVOKE(lle_event, ln,
903 				    LLENTRY_RESOLVED);
904 			}
905 
906 			/*
907 			 * If solicited, make the state REACHABLE.
908 			 * If not solicited and the link-layer address was
909 			 * changed, make it STALE.
910 			 */
911 			if (is_solicited)
912 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
913 			else {
914 				if (lladdr != NULL && llchange)
915 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
916 			}
917 		}
918 
919 		if (ln->ln_router && !is_router) {
920 			/*
921 			 * The peer dropped the router flag.
922 			 * Remove the sender from the Default Router List and
923 			 * update the Destination Cache entries.
924 			 */
925 			struct ifnet *nd6_ifp;
926 
927 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
928 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
929 			    (nd6_ifp->if_inet6->nd_flags &
930 			     ND6_IFF_ACCEPT_RTADV) != 0)
931 				/*
932 				 * Even if the neighbor is not in the default
933 				 * router list, the neighbor may be used as a
934 				 * next hop for some destinations (e.g. redirect
935 				 * case). So we must call rt6_flush explicitly.
936 				 */
937 				rt6_flush(&ip6->ip6_src, ifp);
938 		}
939 		ln->ln_router = is_router;
940 	}
941         /* XXX - QL
942 	 *  Does this matter?
943 	 *  rt->rt_flags &= ~RTF_REJECT;
944 	 */
945 	ln->la_asked = 0;
946 	if (ln->la_hold != NULL)
947 		chain = nd6_grab_holdchain(ln);
948  freeit:
949 	if (ln != NULL)
950 		LLE_WUNLOCK(ln);
951 
952 	if (chain != NULL)
953 		nd6_flush_holdchain(ifp, ln, chain);
954 	if (flush_holdchain)
955 		nd6_flush_children_holdchain(ifp, ln);
956 
957 	if (checklink)
958 		pfxlist_onlink_check();
959 
960 	m_freem(m);
961 	return;
962 
963  bad:
964 	if (ln != NULL)
965 		LLE_WUNLOCK(ln);
966 
967 	ICMP6STAT_INC(icp6s_badna);
968 	m_freem(m);
969 }
970 
971 /*
972  * Neighbor advertisement output handling.
973  *
974  * Based on RFC 2461
975  *
976  * tlladdr:
977  * - 0x01 if include target link-layer address
978  * - 0x02 if target address is CARP MASTER
979  * sdl0 - sockaddr_dl (= proxy NA) or NULL
980  */
981 static void
nd6_na_output_fib(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0,u_int fibnum)982 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
983     const struct in6_addr *taddr6, u_long flags, int tlladdr,
984     struct sockaddr *sdl0, u_int fibnum)
985 {
986 	struct mbuf *m;
987 	struct m_tag *mtag;
988 	struct ip6_hdr *ip6;
989 	struct nd_neighbor_advert *nd_na;
990 	struct ip6_moptions im6o;
991 	struct in6_addr daddr6;
992 
993 	NET_EPOCH_ASSERT();
994 
995 	int icmp6len, maxlen, error;
996 	caddr_t mac = NULL;
997 
998 	daddr6 = *daddr6_0;	/* make a local copy for modification */
999 
1000 	/* estimate the size of message */
1001 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
1002 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
1003 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
1004 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
1005 	    __func__, max_linkhdr, maxlen, MCLBYTES));
1006 
1007 	if (max_linkhdr + maxlen > MHLEN)
1008 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1009 	else
1010 		m = m_gethdr(M_NOWAIT, MT_DATA);
1011 	if (m == NULL)
1012 		return;
1013 	M_SETFIB(m, fibnum);
1014 
1015 	icmp6len = sizeof(*nd_na);
1016 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1017 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
1018 
1019 	/* fill neighbor advertisement packet */
1020 	ip6 = mtod(m, struct ip6_hdr *);
1021 	ip6->ip6_flow = 0;
1022 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1023 	ip6->ip6_vfc |= IPV6_VERSION;
1024 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1025 	ip6->ip6_hlim = 255;
1026 
1027 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1028 		/* reply to DAD */
1029 		daddr6 = in6addr_linklocal_allnodes;
1030 		if (in6_setscope(&daddr6, ifp, NULL))
1031 			goto bad;
1032 
1033 		flags &= ~ND_NA_FLAG_SOLICITED;
1034 	}
1035 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1036 		m->m_flags |= M_MCAST;
1037 		im6o.im6o_multicast_ifp = ifp;
1038 		im6o.im6o_multicast_hlim = 255;
1039 		im6o.im6o_multicast_loop = 0;
1040 	}
1041 
1042 	ip6->ip6_dst = daddr6;
1043 	error = in6_selectsrc_nbr(fibnum, &daddr6, &im6o, ifp, &ip6->ip6_src);
1044 	if (error) {
1045 		char ip6buf[INET6_ADDRSTRLEN];
1046 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1047 		    "determined: dst=%s, error=%d\n",
1048 		    ip6_sprintf(ip6buf, &daddr6), error));
1049 		goto bad;
1050 	}
1051 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1052 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1053 	nd_na->nd_na_code = 0;
1054 	nd_na->nd_na_target = *taddr6;
1055 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1056 
1057 	/*
1058 	 * If we respond from CARP address, we need to prepare mac address
1059 	 * for carp_output().
1060 	 */
1061 	if (ifp->if_carp && (tlladdr & ND6_NA_CARP_MASTER))
1062 		mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1063 	/*
1064 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1065 	 * see nd6_ns_input() for details.
1066 	 * Basically, if NS packet is sent to unicast/anycast addr,
1067 	 * target lladdr option SHOULD NOT be included.
1068 	 */
1069 	if (tlladdr & ND6_NA_OPT_LLA) {
1070 		/*
1071 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1072 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1073 		 * my address) use lladdr configured for the interface.
1074 		 */
1075 		if (sdl0 == NULL) {
1076 			if (mac == NULL)
1077 				mac = nd6_ifptomac(ifp);
1078 		} else if (sdl0->sa_family == AF_LINK) {
1079 			struct sockaddr_dl *sdl;
1080 			sdl = (struct sockaddr_dl *)sdl0;
1081 			if (sdl->sdl_alen == ifp->if_addrlen)
1082 				mac = LLADDR(sdl);
1083 		}
1084 	}
1085 	if ((tlladdr & ND6_NA_OPT_LLA) && mac != NULL) {
1086 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1087 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1088 
1089 		/* roundup to 8 bytes alignment! */
1090 		optlen = (optlen + 7) & ~7;
1091 
1092 		m->m_pkthdr.len += optlen;
1093 		m->m_len += optlen;
1094 		icmp6len += optlen;
1095 		bzero((caddr_t)nd_opt, optlen);
1096 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1097 		nd_opt->nd_opt_len = optlen >> 3;
1098 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1099 	} else
1100 		flags &= ~ND_NA_FLAG_OVERRIDE;
1101 
1102 	ip6->ip6_plen = htons((u_short)icmp6len);
1103 	nd_na->nd_na_flags_reserved = flags;
1104 	nd_na->nd_na_cksum = 0;
1105 	nd_na->nd_na_cksum =
1106 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1107 
1108 	if (send_sendso_input_hook != NULL) {
1109 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1110 		    sizeof(unsigned short), M_NOWAIT);
1111 		if (mtag == NULL)
1112 			goto bad;
1113 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1114 		m_tag_prepend(m, mtag);
1115 	}
1116 
1117 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1118 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1119 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1120 	ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_ADVERT);
1121 
1122 	return;
1123 
1124   bad:
1125 	m_freem(m);
1126 }
1127 
1128 #ifndef BURN_BRIDGES
1129 void
nd6_na_output(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0)1130 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1131     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1132     struct sockaddr *sdl0)
1133 {
1134 
1135 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1136 	    RT_DEFAULT_FIB);
1137 }
1138 #endif
1139 
1140 caddr_t
nd6_ifptomac(struct ifnet * ifp)1141 nd6_ifptomac(struct ifnet *ifp)
1142 {
1143 	switch (ifp->if_type) {
1144 	case IFT_ETHER:
1145 	case IFT_IEEE1394:
1146 	case IFT_L2VLAN:
1147 	case IFT_INFINIBAND:
1148 	case IFT_BRIDGE:
1149 		return IF_LLADDR(ifp);
1150 	default:
1151 		return NULL;
1152 	}
1153 }
1154 
1155 struct dadq {
1156 	TAILQ_ENTRY(dadq) dad_list;
1157 	struct ifaddr *dad_ifa;
1158 	int dad_count;		/* max NS to send */
1159 	int dad_ns_tcount;	/* # of trials to send NS */
1160 	int dad_ns_ocount;	/* NS sent so far */
1161 	int dad_ns_icount;
1162 	int dad_na_icount;
1163 	int dad_ns_lcount;	/* looped back NS */
1164 	int dad_loopbackprobe;	/* probing state for loopback detection */
1165 	struct callout dad_timer_ch;
1166 	struct vnet *dad_vnet;
1167 	u_int dad_refcnt;
1168 #define	ND_OPT_NONCE_LEN32 \
1169 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1170 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1171 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1172 };
1173 
1174 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1175 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1176 #define	V_dadq			VNET(dadq)
1177 #define	V_dad_rwlock		VNET(dad_rwlock)
1178 
1179 #define	DADQ_LOCKPTR()		(&V_dad_rwlock)
1180 #define	DADQ_LOCK_INIT()	rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1181 #define	DADQ_RLOCK()		rw_rlock(DADQ_LOCKPTR())
1182 #define	DADQ_RUNLOCK()		rw_runlock(DADQ_LOCKPTR())
1183 #define	DADQ_WLOCK()		rw_wlock(DADQ_LOCKPTR())
1184 #define	DADQ_WUNLOCK()		rw_wunlock(DADQ_LOCKPTR())
1185 
1186 #define	DADQ_LOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1187 #define	DADQ_RLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1188 #define	DADQ_WLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1189 
1190 static void
nd6_dad_add(struct dadq * dp)1191 nd6_dad_add(struct dadq *dp)
1192 {
1193 	DADQ_WLOCK_ASSERT();
1194 
1195 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1196 	dp->dad_ondadq = true;
1197 }
1198 
1199 static void
nd6_dad_del(struct dadq * dp)1200 nd6_dad_del(struct dadq *dp)
1201 {
1202 	DADQ_WLOCK_ASSERT();
1203 
1204 	if (dp->dad_ondadq) {
1205 		/*
1206 		 * Remove dp from the dadq and release the dadq's
1207 		 * reference.
1208 		 */
1209 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1210 		dp->dad_ondadq = false;
1211 		nd6_dad_rele(dp);
1212 	}
1213 }
1214 
1215 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * n)1216 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1217 {
1218 	struct dadq *dp;
1219 
1220 	DADQ_LOCK_ASSERT();
1221 
1222 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1223 		if (dp->dad_ifa != ifa)
1224 			continue;
1225 
1226 		/*
1227 		 * Skip if the nonce matches the received one.
1228 		 * +2 in the length is required because of type and
1229 		 * length fields are included in a header.
1230 		 */
1231 		if (n != NULL &&
1232 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1233 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1234 		    ND_OPT_NONCE_LEN) == 0) {
1235 			dp->dad_ns_lcount++;
1236 			continue;
1237 		}
1238 		break;
1239 	}
1240 
1241 	return (dp);
1242 }
1243 
1244 static void
nd6_dad_starttimer(struct dadq * dp,int ticks)1245 nd6_dad_starttimer(struct dadq *dp, int ticks)
1246 {
1247 	DADQ_WLOCK_ASSERT();
1248 
1249 	callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1250 }
1251 
1252 static void
nd6_dad_stoptimer(struct dadq * dp)1253 nd6_dad_stoptimer(struct dadq *dp)
1254 {
1255 	callout_drain(&dp->dad_timer_ch);
1256 }
1257 
1258 static void
nd6_dad_rele(struct dadq * dp)1259 nd6_dad_rele(struct dadq *dp)
1260 {
1261 	if (refcount_release(&dp->dad_refcnt)) {
1262 		KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1263 		ifa_free(dp->dad_ifa);
1264 		free(dp, M_IP6NDP);
1265 	}
1266 }
1267 
1268 void
nd6_dad_init(void)1269 nd6_dad_init(void)
1270 {
1271 	DADQ_LOCK_INIT();
1272 	TAILQ_INIT(&V_dadq);
1273 }
1274 
1275 /*
1276  * Start Duplicate Address Detection (DAD) for specified interface address.
1277  */
1278 void
nd6_dad_start(struct ifaddr * ifa,int delay)1279 nd6_dad_start(struct ifaddr *ifa, int delay)
1280 {
1281 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1282 	struct dadq *dp;
1283 	char ip6buf[INET6_ADDRSTRLEN];
1284 
1285 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1286 	    ("starting DAD on non-tentative address %p", ifa));
1287 
1288 	/*
1289 	 * If we don't need DAD, don't do it.
1290 	 * There are several cases:
1291 	 * - DAD is disabled globally or on the interface
1292 	 * - the interface address is anycast
1293 	 */
1294 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1295 	    V_ip6_dad_count == 0 ||
1296 	    (ifa->ifa_ifp->if_inet6->nd_flags & ND6_IFF_NO_DAD) != 0) {
1297 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1298 		return;
1299 	}
1300 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1301 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1302 	    (ifa->ifa_ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) != 0)
1303 		return;
1304 
1305 	DADQ_WLOCK();
1306 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1307 		/*
1308 		 * DAD is already in progress.  Let the existing entry
1309 		 * finish it.
1310 		 */
1311 		DADQ_WUNLOCK();
1312 		return;
1313 	}
1314 
1315 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1316 	if (dp == NULL) {
1317 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1318 			"%s(%s)\n",
1319 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1320 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1321 		return;
1322 	}
1323 	callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1324 	    CALLOUT_RETURNUNLOCKED);
1325 #ifdef VIMAGE
1326 	dp->dad_vnet = curvnet;
1327 #endif
1328 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1329 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1330 
1331 	/*
1332 	 * Send NS packet for DAD, ip6_dad_count times.
1333 	 * Note that we must delay the first transmission, if this is the
1334 	 * first packet to be sent from the interface after interface
1335 	 * (re)initialization.
1336 	 */
1337 	dp->dad_ifa = ifa;
1338 	ifa_ref(dp->dad_ifa);
1339 	dp->dad_count = V_ip6_dad_count;
1340 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1341 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1342 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1343 
1344 	/* Add this to the dadq and add a reference for the dadq. */
1345 	refcount_init(&dp->dad_refcnt, 1);
1346 	nd6_dad_add(dp);
1347 	nd6_dad_starttimer(dp, delay);
1348 	DADQ_WUNLOCK();
1349 }
1350 
1351 /*
1352  * terminate DAD unconditionally.  used for address removals.
1353  */
1354 void
nd6_dad_stop(struct ifaddr * ifa)1355 nd6_dad_stop(struct ifaddr *ifa)
1356 {
1357 	struct dadq *dp;
1358 
1359 	DADQ_WLOCK();
1360 	dp = nd6_dad_find(ifa, NULL);
1361 	if (dp == NULL) {
1362 		DADQ_WUNLOCK();
1363 		/* DAD wasn't started yet */
1364 		return;
1365 	}
1366 
1367 	/*
1368 	 * Acquire a temporary reference so that we can safely stop the callout.
1369 	 */
1370 	(void)refcount_acquire(&dp->dad_refcnt);
1371 	nd6_dad_del(dp);
1372 	DADQ_WUNLOCK();
1373 
1374 	nd6_dad_stoptimer(dp);
1375 	nd6_dad_rele(dp);
1376 }
1377 
1378 static void
nd6_dad_timer(void * arg)1379 nd6_dad_timer(void *arg)
1380 {
1381 	struct dadq *dp = arg;
1382 	struct ifaddr *ifa = dp->dad_ifa;
1383 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1384 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1385 	char ip6buf[INET6_ADDRSTRLEN];
1386 	struct epoch_tracker et;
1387 
1388 	CURVNET_SET(dp->dad_vnet);
1389 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1390 
1391 	NET_EPOCH_ENTER(et);
1392 	if (ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) {
1393 		/* Do not need DAD for ifdisabled interface. */
1394 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1395 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1396 		goto err;
1397 	}
1398 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1399 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1400 			"%s(%s)\n",
1401 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1402 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1403 		goto err;
1404 	}
1405 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1406 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1407 			"%s(%s)\n",
1408 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1409 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1410 		goto err;
1411 	}
1412 
1413 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1414 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1415 	    (((ifp->if_flags & IFF_UP) == 0) ||
1416 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1417 		nd6log((LOG_INFO, "%s: could not run DAD "
1418 		    "because the interface was down or not running.\n",
1419 		    if_name(ifa->ifa_ifp)));
1420 		goto err;
1421 	}
1422 
1423 	/* Need more checks? */
1424 	if (dp->dad_ns_ocount < dp->dad_count) {
1425 		/*
1426 		 * We have more NS to go.  Send NS packet for DAD.
1427 		 */
1428 		nd6_dad_starttimer(dp,
1429 		    (long)ifa->ifa_ifp->if_inet6->nd_retrans * hz / 1000);
1430 		nd6_dad_ns_output(dp);
1431 		goto done;
1432 	} else {
1433 		/*
1434 		 * We have transmitted sufficient number of DAD packets.
1435 		 * See what we've got.
1436 		 */
1437 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1438 			/* We've seen NS or NA, means DAD has failed. */
1439 			nd6_dad_duplicated(ifa, dp);
1440 		} else if (V_dad_enhanced != 0 &&
1441 		    dp->dad_ns_lcount > 0 &&
1442 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1443 			/*
1444 			 * Sec. 4.1 in RFC 7527 requires transmission of
1445 			 * additional probes until the loopback condition
1446 			 * becomes clear when a looped back probe is detected.
1447 			 */
1448 			log(LOG_ERR, "%s: a looped back NS message is "
1449 			    "detected during DAD for %s.  "
1450 			    "Another DAD probes are being sent.\n",
1451 			    if_name(ifa->ifa_ifp),
1452 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1453 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1454 			/*
1455 			 * Send an NS immediately and increase dad_count by
1456 			 * V_nd6_mmaxtries - 1.
1457 			 */
1458 			dp->dad_count =
1459 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1460 			nd6_dad_starttimer(dp,
1461 			    (long)ifa->ifa_ifp->if_inet6->nd_retrans * hz / 1000);
1462 			nd6_dad_ns_output(dp);
1463 			goto done;
1464 		} else {
1465 			/*
1466 			 * We are done with DAD.  No NA came, no NS came.
1467 			 * No duplicate address found.  Check IFDISABLED flag
1468 			 * again in case that it is changed between the
1469 			 * beginning of this function and here.
1470 			 *
1471 			 * Reset DAD failures counter if using stable addresses.
1472 			 */
1473 			if ((ifp->if_inet6->nd_flags & ND6_IFF_IFDISABLED) == 0) {
1474 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1475 				if ((ifp->if_inet6->nd_flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY))
1476 					atomic_store_int(&DAD_FAILURES(ifp), 0);
1477 				/*
1478 				 * RFC 9131 Section 6.1.2: The first advertisement
1479 				 * SHOULD be sent as soon as an address changes the
1480 				 * state from tentative to preferred.
1481 				 */
1482 				nd6_grand_start(ifa, ND6_QUEUE_FLAG_NEWGUA);
1483 			}
1484 
1485 			nd6log((LOG_DEBUG,
1486 			    "%s: DAD complete for %s - no duplicates found\n",
1487 			    if_name(ifa->ifa_ifp),
1488 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1489 			if (dp->dad_ns_lcount > 0)
1490 				log(LOG_ERR, "%s: DAD completed while "
1491 				    "a looped back NS message is detected "
1492 				    "during DAD for %s.\n",
1493 				    if_name(ifa->ifa_ifp),
1494 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1495 		}
1496 	}
1497 err:
1498 	nd6_dad_del(dp);
1499 	DADQ_WUNLOCK();
1500 done:
1501 	NET_EPOCH_EXIT(et);
1502 	CURVNET_RESTORE();
1503 }
1504 
1505 static void
nd6_dad_duplicated(struct ifaddr * ifa,struct dadq * dp)1506 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1507 {
1508 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1509 	struct ifnet *ifp;
1510 	char ip6buf[INET6_ADDRSTRLEN];
1511 
1512 	ifp = ifa->ifa_ifp;
1513 
1514 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1515 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1516 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1517 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1518 	    dp->dad_na_icount);
1519 
1520 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1521 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1522 
1523 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1524 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1525 
1526 	/*
1527 	 * For RFC 7217 stable addresses, increment failure counter here if we still have retries.
1528 	 * More addresses will be generated as long as retries are not exhausted.
1529 	 */
1530 	if ((ifp->if_inet6->nd_flags & ND6_IFF_STABLEADDR) && !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
1531 		u_int dad_failures = atomic_load_int(&DAD_FAILURES(ifp));
1532 
1533 		if (dad_failures <= V_ip6_stableaddr_maxretries) {
1534 			atomic_add_int(&DAD_FAILURES(ifp), 1);
1535 			/* if retries exhausted, output an informative error message */
1536 			if (dad_failures == V_ip6_stableaddr_maxretries)
1537 				log(LOG_ERR, "%s: manual intervention required, consider disabling \"stableaddr\" on the interface"
1538 				    " or checking hostuuid for uniqueness\n",
1539 				    if_name(ifp));
1540 		}
1541 	} else {
1542 		log(LOG_ERR, "%s: manual intervention required\n",
1543 		    if_name(ifp));
1544 	}
1545 
1546 	/*
1547 	 * If the address is a link-local address formed from an interface
1548 	 * identifier based on the hardware address which is supposed to be
1549 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1550 	 * operation on the interface SHOULD be disabled.
1551 	 * [RFC 4862, Section 5.4.5]
1552 	 */
1553 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1554 		struct in6_addr in6;
1555 
1556 		/*
1557 		 * To avoid over-reaction, we only apply this logic when we are
1558 		 * very sure that hardware addresses are supposed to be unique.
1559 		 */
1560 		switch (ifp->if_type) {
1561 		case IFT_ETHER:
1562 		case IFT_ATM:
1563 		case IFT_IEEE1394:
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 				ifp->if_inet6->nd_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 /*
1579  * Transmit a neighbour solicitation for the purpose of DAD.  Returns with the
1580  * DAD queue unlocked.
1581  */
1582 static void
nd6_dad_ns_output(struct dadq * dp)1583 nd6_dad_ns_output(struct dadq *dp)
1584 {
1585 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1586 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1587 	int i;
1588 
1589 	DADQ_WLOCK_ASSERT();
1590 
1591 	dp->dad_ns_tcount++;
1592 	if ((ifp->if_flags & IFF_UP) == 0) {
1593 		DADQ_WUNLOCK();
1594 		return;
1595 	}
1596 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1597 		DADQ_WUNLOCK();
1598 		return;
1599 	}
1600 
1601 	dp->dad_ns_ocount++;
1602 	if (V_dad_enhanced != 0) {
1603 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1604 			dp->dad_nonce[i] = arc4random();
1605 		/*
1606 		 * XXXHRS: Note that in the case that
1607 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1608 		 * different nonces can be looped back in an unexpected
1609 		 * order.  The current implementation recognizes only
1610 		 * the latest nonce on the sender side.  Practically it
1611 		 * should work well in almost all cases.
1612 		 */
1613 	}
1614 	DADQ_WUNLOCK();
1615 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1616 	    (uint8_t *)&dp->dad_nonce[0]);
1617 }
1618 
1619 static void
nd6_dad_ns_input(struct ifaddr * ifa,struct nd_opt_nonce * ndopt_nonce)1620 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1621 {
1622 	struct dadq *dp;
1623 
1624 	if (ifa == NULL)
1625 		panic("ifa == NULL in nd6_dad_ns_input");
1626 
1627 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1628 	if (V_dad_enhanced == 0)
1629 		ndopt_nonce = NULL;
1630 	DADQ_RLOCK();
1631 	dp = nd6_dad_find(ifa, ndopt_nonce);
1632 	if (dp != NULL)
1633 		dp->dad_ns_icount++;
1634 	DADQ_RUNLOCK();
1635 }
1636 
1637 static void
nd6_dad_na_input(struct ifaddr * ifa)1638 nd6_dad_na_input(struct ifaddr *ifa)
1639 {
1640 	struct dadq *dp;
1641 
1642 	if (ifa == NULL)
1643 		panic("ifa == NULL in nd6_dad_na_input");
1644 
1645 	DADQ_RLOCK();
1646 	dp = nd6_dad_find(ifa, NULL);
1647 	if (dp != NULL)
1648 		dp->dad_na_icount++;
1649 	DADQ_RUNLOCK();
1650 }
1651 
1652 static void
nd6_queue_rel(void * arg)1653 nd6_queue_rel(void *arg)
1654 {
1655 	struct nd_queue *ndq = arg;
1656 	struct ifaddr *ifa;
1657 
1658 	ifa = ndq->ndq_ifa;
1659 	IF_ADDR_WLOCK_ASSERT(ifa->ifa_ifp);
1660 
1661 	/* Remove ndq from the nd_queue list and free it */
1662 	TAILQ_REMOVE(&ifa->ifa_ifp->if_inet6->nd_queue, ndq, ndq_list);
1663 	IF_ADDR_WUNLOCK(ifa->ifa_ifp);
1664 
1665 	free(ndq, M_IP6NDP);
1666 	ifa_free(ifa);
1667 }
1668 
1669 static void
nd6_queue_timer(void * arg)1670 nd6_queue_timer(void *arg)
1671 {
1672 	struct nd_queue *ndq = arg;
1673 	struct ifaddr *ifa = ndq->ndq_ifa;
1674 	struct ifnet *ifp;
1675 	struct in6_addr daddr;
1676 	struct epoch_tracker et;
1677 	int delay, tlladdr;
1678 	u_long flags;
1679 
1680 	KASSERT(ifa != NULL, ("ND6 queue entry %p with no address", ndq));
1681 
1682 	ifp = ifa->ifa_ifp;
1683 	CURVNET_SET(ifp->if_vnet);
1684 	NET_EPOCH_ENTER(et);
1685 
1686 	daddr = ndq->ndq_daddr;
1687 	tlladdr = ND6_NA_OPT_LLA;
1688 	flags = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
1689 	if ((ifp->if_inet6->nd_flags & ND6_IFF_ACCEPT_RTADV) != 0 && V_ip6_norbit_raif)
1690 		flags &= ~ND_NA_FLAG_ROUTER;
1691 
1692 	/*
1693 	 * RFC 9131 Section 6.1.2: If the address is preferred,
1694 	 * then the Override flag SHOULD NOT be set.
1695 	 */
1696 	if ((ndq->ndq_flags & ND6_QUEUE_FLAG_NEWGUA) != 0) {
1697 		/*
1698 		 * XXX: If the address is in the Optimistic state,
1699 		 * then the Override flag MUST NOT be set.
1700 		 * We don't support RFC 4429 yet.
1701 		 */
1702 		if ((ifp->if_inet6->nd_flags & ND6_IFF_PREFER_SOURCE) == 0)
1703 			flags |= ND_NA_FLAG_OVERRIDE;
1704 	}
1705 	/*
1706 	 * RFC 4861 Section 7.2.6: if link-layer address changed,
1707 	 * The Override flag MAY be set to either zero or one.
1708 	 */
1709 	if ((ndq->ndq_flags & ND6_QUEUE_FLAG_LLADDR) != 0)
1710 		flags |= ND_NA_FLAG_OVERRIDE;
1711 	/* anycast advertisement delay rule (RFC 4861 7.2.7, SHOULD) */
1712 	if ((ndq->ndq_flags & ND6_QUEUE_FLAG_ANYCAST) != 0)
1713 		flags |= ND_NA_FLAG_SOLICITED;
1714 
1715 	/*
1716 	 * if it was GRAND, wait at least a RetransTimer
1717 	 * before removing from queue.
1718 	 */
1719 	if ((ndq->ndq_flags & ND6_QUEUE_GRAND_MASK) != 0) {
1720 		delay = ifp->if_inet6->nd_retrans * hz / 1000;
1721 		callout_reset(&ndq->ndq_callout, delay, nd6_queue_rel, ndq);
1722 		IF_ADDR_WUNLOCK(ifp);
1723 	} else
1724 		nd6_queue_rel(ndq);
1725 
1726 	if (__predict_true(in6_setscope(&daddr, ifp, NULL) == 0))
1727 		nd6_na_output_fib(ifp, &daddr, IFA_IN6(ifa), flags, tlladdr,
1728 		    NULL, ifp->if_fib);
1729 
1730 	NET_EPOCH_EXIT(et);
1731 	CURVNET_RESTORE();
1732 }
1733 
1734 static void
nd6_queue_add(struct ifaddr * ifa,struct in6_addr * daddr,int delay,uint32_t flags)1735 nd6_queue_add(struct ifaddr *ifa, struct in6_addr *daddr,
1736     int delay, uint32_t flags)
1737 {
1738 	struct nd_queue *ndq = NULL;
1739 	struct ifnet *ifp;
1740 	struct in6_ifextra *ext;
1741 	char ip6buf[INET6_ADDRSTRLEN];
1742 
1743 	NET_EPOCH_ASSERT();
1744 
1745 	ifp = ifa->ifa_ifp;
1746 	ext = ifp->if_inet6;
1747 	IF_ADDR_WLOCK(ifp);
1748 	/*
1749 	 * if request comes from GRAND, check whether another delayed
1750 	 * GRAND NA exists in the queue.
1751 	 * If it exists, cancel previous one and reuse its ndq.
1752 	 */
1753 	if ((flags & ND6_QUEUE_GRAND_MASK) != 0) {
1754 		TAILQ_FOREACH(ndq, &ext->nd_queue, ndq_list) {
1755 			if (ndq->ndq_ifa == ifa &&
1756 			    (flags & ND6_QUEUE_GRAND_MASK) != 0)
1757 				break;
1758 		}
1759 	}
1760 	if (ndq == NULL) {
1761 		ndq = malloc(sizeof(*ndq), M_IP6NDP, M_NOWAIT | M_ZERO);
1762 		if (ndq == NULL) {
1763 			log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
1764 			    __func__, ip6_sprintf(ip6buf, IFA_IN6(ifa)),
1765 			    ifp ? if_name(ifp) : "???");
1766 			IF_ADDR_WUNLOCK(ifp);
1767 			return;
1768 		}
1769 
1770 		callout_init_mtx(&ndq->ndq_callout, &ifp->if_addr_lock,
1771 		    CALLOUT_TRYLOCK | CALLOUT_RETURNUNLOCKED);
1772 		ifa_ref(ifa);
1773 		ndq->ndq_ifa = ifa;
1774 		TAILQ_INSERT_TAIL(&ext->nd_queue, ndq, ndq_list);
1775 	}
1776 
1777 	memcpy(&ndq->ndq_daddr, daddr, sizeof(struct in6_addr));
1778 	ndq->ndq_flags = flags;
1779 
1780 	nd6log((LOG_DEBUG, "%s: delay IPv6 NA for %s\n", if_name(ifp),
1781 	    ip6_sprintf(ip6buf, IFA_IN6(ifa))));
1782 	callout_reset(&ndq->ndq_callout, delay, nd6_queue_timer, ndq);
1783 	IF_ADDR_WUNLOCK(ifp);
1784 }
1785 
1786 /*
1787  * Start Gratuitous Neighbor Discovery (GRAND) for specified address.
1788  * Called after DAD completes and by interface link layer change event.
1789  */
1790 void
nd6_grand_start(struct ifaddr * ifa,uint32_t flags)1791 nd6_grand_start(struct ifaddr *ifa, uint32_t flags)
1792 {
1793 	struct nd_queue *ndq;
1794 	struct in6_ifextra *ext = ifa->ifa_ifp->if_inet6;
1795 	struct in6_addr daddr = IN6ADDR_ANY_INIT;
1796 	int delay, count = 0;
1797 
1798 	NET_EPOCH_ASSERT();
1799 	/* If we don't need GRAND, don't do it. */
1800 	if (V_ip6_grand_count == 0 ||
1801 	    ifa->ifa_carp != NULL)
1802 		return;
1803 
1804 	/* Check if new address is global */
1805 	if ((flags & ND6_QUEUE_FLAG_NEWGUA) != 0 &&
1806 	    in6_addrscope(IFA_IN6(ifa)) != IPV6_ADDR_SCOPE_GLOBAL)
1807 		return;
1808 
1809 	/*
1810 	 * RFC 9131 Section 6.1.2: These advertisements MUST be
1811 	 * separated by at least RetransTimer seconds.
1812 	 */
1813 	TAILQ_FOREACH(ndq, &ext->nd_queue, ndq_list) {
1814 		/*
1815 		 * RFC 9131 Section 6.1.2: a node SHOULD send
1816 		 * up to MAX_NEIGHBOR_ADVERTISEMENT Neighbor Advertisement messages.
1817 		 * Make sure we don't queue GRAND more than V_ip6_grand_count
1818 		 * per interface.
1819 		 * Since this limitation only applies to GRAND, don't
1820 		 * count non-GRAND ndq.
1821 		 */
1822 		if ((ndq->ndq_flags & ND6_QUEUE_GRAND_MASK) == 0)
1823 			continue;
1824 
1825 		count++;
1826 		if (count >= V_ip6_grand_count)
1827 			return;
1828 	}
1829 
1830 	/*
1831 	 * RFC 9131 Section 6.1.2: if new global address added,
1832 	 * use the all-routers multicast address.
1833 	 */
1834 	if ((flags & ND6_QUEUE_FLAG_NEWGUA) != 0)
1835 		daddr = in6addr_linklocal_allrouters;
1836 
1837 	/*
1838 	 * RFC 4861 Section 7.2.6: if link-layer address changed,
1839 	 * use the all-nodes multicast address.
1840 	 */
1841 	if ((flags & ND6_QUEUE_FLAG_LLADDR) != 0)
1842 		daddr = in6addr_linklocal_allnodes;
1843 
1844 	delay = ext->nd_retrans * hz / 1000;
1845 	nd6_queue_add(ifa, &daddr, count * delay, flags);
1846 }
1847 
1848 /*
1849  * drain nd6 queue. used for address removals.
1850  */
1851 void
nd6_queue_stop(struct ifaddr * ifa)1852 nd6_queue_stop(struct ifaddr *ifa)
1853 {
1854 	struct nd_queue *ndq, *dndq;
1855 	struct ifnet *ifp;
1856 
1857 	ifp = ifa->ifa_ifp;
1858 	IF_ADDR_WLOCK(ifp);
1859 	TAILQ_FOREACH_SAFE(ndq, &ifp->if_inet6->nd_queue, ndq_list, dndq) {
1860 		if (ndq->ndq_ifa != ifa)
1861 			continue;
1862 
1863 		callout_stop(&ndq->ndq_callout);
1864 
1865 		/* Remove ndq from the nd_queue list and free it */
1866 		TAILQ_REMOVE(&ifa->ifa_ifp->if_inet6->nd_queue, ndq, ndq_list);
1867 		free(ndq, M_IP6NDP);
1868 		ifa_free(ifa);
1869 	}
1870 	IF_ADDR_WUNLOCK(ifp);
1871 }
1872