xref: /freebsd/sys/netinet6/in6.c (revision c3a3d1fd6c7007e7088936e14fd9b4944b3a3054)
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: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)in.c	8.2 (Berkeley) 11/15/93
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/eventhandler.h>
73 #include <sys/errno.h>
74 #include <sys/jail.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/priv.h>
81 #include <sys/proc.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/rmlock.h>
86 #include <sys/sysctl.h>
87 #include <sys/syslog.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_types.h>
92 #include <net/route.h>
93 #include <net/route/route_ctl.h>
94 #include <net/route/nhop.h>
95 #include <net/if_dl.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <net/if_llatbl.h>
101 #include <netinet/if_ether.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/ip.h>
104 #include <netinet/in_pcb.h>
105 #include <netinet/ip_carp.h>
106 
107 #include <netinet/ip6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/mld6_var.h>
111 #include <netinet6/ip6_mroute.h>
112 #include <netinet6/in6_ifattach.h>
113 #include <netinet6/scope6_var.h>
114 #include <netinet6/in6_fib.h>
115 #include <netinet6/in6_pcb.h>
116 
117 /*
118  * struct in6_ifreq and struct ifreq must be type punnable for common members
119  * of ifr_ifru to allow accessors to be shared.
120  */
121 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
122     offsetof(struct ifreq, ifr_ifru),
123     "struct in6_ifreq and struct ifreq are not type punnable");
124 
125 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
126 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
127 
128 /*
129  * Definitions of some costant IP6 addresses.
130  */
131 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
132 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
133 const struct in6_addr in6addr_nodelocal_allnodes =
134 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
135 const struct in6_addr in6addr_linklocal_allnodes =
136 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
137 const struct in6_addr in6addr_linklocal_allrouters =
138 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
139 const struct in6_addr in6addr_linklocal_allv2routers =
140 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
141 
142 const struct in6_addr in6mask0 = IN6MASK0;
143 const struct in6_addr in6mask32 = IN6MASK32;
144 const struct in6_addr in6mask64 = IN6MASK64;
145 const struct in6_addr in6mask96 = IN6MASK96;
146 const struct in6_addr in6mask128 = IN6MASK128;
147 
148 const struct sockaddr_in6 sa6_any =
149 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
150 
151 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
152 	struct in6_aliasreq *, int);
153 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
154 
155 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
156     struct in6_ifaddr *, int);
157 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
158     struct in6_aliasreq *, int flags);
159 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
160     struct in6_ifaddr *, int, int);
161 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
162     struct in6_ifaddr *, int);
163 
164 static void in6_join_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
165 static void in6_leave_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
166 
167 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
168 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
169 
170 void
171 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
172 {
173 	struct rt_addrinfo info;
174 	struct ifaddr *ifa;
175 	struct sockaddr_dl gateway;
176 	int fibnum;
177 
178 	ifa = &ia->ia_ifa;
179 
180 	/*
181 	 * Prepare info data for the host route.
182 	 * This code mimics one from ifa_maintain_loopback_route().
183 	 */
184 	bzero(&info, sizeof(struct rt_addrinfo));
185 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
186 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
187 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
188 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
189 	if (cmd != RTM_DELETE)
190 		info.rti_ifp = V_loif;
191 
192 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
193 
194 	if (cmd == RTM_ADD) {
195 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
196 		rt_routemsg_info(cmd, &info, fibnum);
197 	} else if (cmd == RTM_DELETE) {
198 		rt_routemsg_info(cmd, &info, fibnum);
199 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
200 	}
201 }
202 
203 int
204 in6_mask2len(struct in6_addr *mask, u_char *lim0)
205 {
206 	int x = 0, y;
207 	u_char *lim = lim0, *p;
208 
209 	/* ignore the scope_id part */
210 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
211 		lim = (u_char *)mask + sizeof(*mask);
212 	for (p = (u_char *)mask; p < lim; x++, p++) {
213 		if (*p != 0xff)
214 			break;
215 	}
216 	y = 0;
217 	if (p < lim) {
218 		for (y = 0; y < 8; y++) {
219 			if ((*p & (0x80 >> y)) == 0)
220 				break;
221 		}
222 	}
223 
224 	/*
225 	 * when the limit pointer is given, do a stricter check on the
226 	 * remaining bits.
227 	 */
228 	if (p < lim) {
229 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
230 			return (-1);
231 		for (p = p + 1; p < lim; p++)
232 			if (*p != 0)
233 				return (-1);
234 	}
235 
236 	return x * 8 + y;
237 }
238 
239 #ifdef COMPAT_FREEBSD32
240 struct in6_ndifreq32 {
241 	char ifname[IFNAMSIZ];
242 	uint32_t ifindex;
243 };
244 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
245 #endif
246 
247 int
248 in6_control(struct socket *so, u_long cmd, void *data,
249     struct ifnet *ifp, struct thread *td)
250 {
251 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
252 	struct	in6_ifaddr *ia = NULL;
253 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
254 	struct sockaddr_in6 *sa6;
255 	int error;
256 
257 	/*
258 	 * Compat to make pre-10.x ifconfig(8) operable.
259 	 */
260 	if (cmd == OSIOCAIFADDR_IN6) {
261 		cmd = SIOCAIFADDR_IN6;
262 		ifra->ifra_vhid = 0;
263 	}
264 
265 	switch (cmd) {
266 	case SIOCGETSGCNT_IN6:
267 	case SIOCGETMIFCNT_IN6:
268 		/*
269 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
270 		 * We cannot see how that would be needed, so do not adjust the
271 		 * KPI blindly; more likely should clean up the IPv4 variant.
272 		 */
273 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
274 	}
275 
276 	switch (cmd) {
277 	case SIOCAADDRCTL_POLICY:
278 	case SIOCDADDRCTL_POLICY:
279 		if (td != NULL) {
280 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
281 			if (error)
282 				return (error);
283 		}
284 		return (in6_src_ioctl(cmd, data));
285 	}
286 
287 	if (ifp == NULL)
288 		return (EOPNOTSUPP);
289 
290 	switch (cmd) {
291 	case SIOCSNDFLUSH_IN6:
292 	case SIOCSPFXFLUSH_IN6:
293 	case SIOCSRTRFLUSH_IN6:
294 	case SIOCSDEFIFACE_IN6:
295 	case SIOCSIFINFO_FLAGS:
296 	case SIOCSIFINFO_IN6:
297 		if (td != NULL) {
298 			error = priv_check(td, PRIV_NETINET_ND6);
299 			if (error)
300 				return (error);
301 		}
302 		/* FALLTHROUGH */
303 	case OSIOCGIFINFO_IN6:
304 	case SIOCGIFINFO_IN6:
305 	case SIOCGNBRINFO_IN6:
306 	case SIOCGDEFIFACE_IN6:
307 		return (nd6_ioctl(cmd, data, ifp));
308 
309 #ifdef COMPAT_FREEBSD32
310 	case SIOCGDEFIFACE32_IN6:
311 		{
312 			struct in6_ndifreq ndif;
313 			struct in6_ndifreq32 *ndif32;
314 
315 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
316 			    ifp);
317 			if (error)
318 				return (error);
319 			ndif32 = (struct in6_ndifreq32 *)data;
320 			ndif32->ifindex = ndif.ifindex;
321 			return (0);
322 		}
323 #endif
324 	}
325 
326 	switch (cmd) {
327 	case SIOCSIFPREFIX_IN6:
328 	case SIOCDIFPREFIX_IN6:
329 	case SIOCAIFPREFIX_IN6:
330 	case SIOCCIFPREFIX_IN6:
331 	case SIOCSGIFPREFIX_IN6:
332 	case SIOCGIFPREFIX_IN6:
333 		log(LOG_NOTICE,
334 		    "prefix ioctls are now invalidated. "
335 		    "please use ifconfig.\n");
336 		return (EOPNOTSUPP);
337 	}
338 
339 	switch (cmd) {
340 	case SIOCSSCOPE6:
341 		if (td != NULL) {
342 			error = priv_check(td, PRIV_NETINET_SCOPE6);
343 			if (error)
344 				return (error);
345 		}
346 		/* FALLTHROUGH */
347 	case SIOCGSCOPE6:
348 	case SIOCGSCOPE6DEF:
349 		return (scope6_ioctl(cmd, data, ifp));
350 	}
351 
352 	/*
353 	 * Find address for this interface, if it exists.
354 	 *
355 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
356 	 * only, and used the first interface address as the target of other
357 	 * operations (without checking ifra_addr).  This was because netinet
358 	 * code/API assumed at most 1 interface address per interface.
359 	 * Since IPv6 allows a node to assign multiple addresses
360 	 * on a single interface, we almost always look and check the
361 	 * presence of ifra_addr, and reject invalid ones here.
362 	 * It also decreases duplicated code among SIOC*_IN6 operations.
363 	 */
364 	switch (cmd) {
365 	case SIOCAIFADDR_IN6:
366 	case SIOCSIFPHYADDR_IN6:
367 		sa6 = &ifra->ifra_addr;
368 		break;
369 	case SIOCSIFADDR_IN6:
370 	case SIOCGIFADDR_IN6:
371 	case SIOCSIFDSTADDR_IN6:
372 	case SIOCSIFNETMASK_IN6:
373 	case SIOCGIFDSTADDR_IN6:
374 	case SIOCGIFNETMASK_IN6:
375 	case SIOCDIFADDR_IN6:
376 	case SIOCGIFPSRCADDR_IN6:
377 	case SIOCGIFPDSTADDR_IN6:
378 	case SIOCGIFAFLAG_IN6:
379 	case SIOCSNDFLUSH_IN6:
380 	case SIOCSPFXFLUSH_IN6:
381 	case SIOCSRTRFLUSH_IN6:
382 	case SIOCGIFALIFETIME_IN6:
383 	case SIOCGIFSTAT_IN6:
384 	case SIOCGIFSTAT_ICMP6:
385 		sa6 = &ifr->ifr_addr;
386 		break;
387 	case SIOCSIFADDR:
388 	case SIOCSIFBRDADDR:
389 	case SIOCSIFDSTADDR:
390 	case SIOCSIFNETMASK:
391 		/*
392 		 * Although we should pass any non-INET6 ioctl requests
393 		 * down to driver, we filter some legacy INET requests.
394 		 * Drivers trust SIOCSIFADDR et al to come from an already
395 		 * privileged layer, and do not perform any credentials
396 		 * checks or input validation.
397 		 */
398 		return (EINVAL);
399 	default:
400 		sa6 = NULL;
401 		break;
402 	}
403 	if (sa6 && sa6->sin6_family == AF_INET6) {
404 		if (sa6->sin6_scope_id != 0)
405 			error = sa6_embedscope(sa6, 0);
406 		else
407 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
408 		if (error != 0)
409 			return (error);
410 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
411 		    &sa6->sin6_addr)) != 0)
412 			return (error);
413 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
414 	} else
415 		ia = NULL;
416 
417 	switch (cmd) {
418 	case SIOCSIFADDR_IN6:
419 	case SIOCSIFDSTADDR_IN6:
420 	case SIOCSIFNETMASK_IN6:
421 		/*
422 		 * Since IPv6 allows a node to assign multiple addresses
423 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
424 		 */
425 		/* we decided to obsolete this command (20000704) */
426 		error = EINVAL;
427 		goto out;
428 
429 	case SIOCDIFADDR_IN6:
430 		/*
431 		 * for IPv4, we look for existing in_ifaddr here to allow
432 		 * "ifconfig if0 delete" to remove the first IPv4 address on
433 		 * the interface.  For IPv6, as the spec allows multiple
434 		 * interface address from the day one, we consider "remove the
435 		 * first one" semantics to be not preferable.
436 		 */
437 		if (ia == NULL) {
438 			error = EADDRNOTAVAIL;
439 			goto out;
440 		}
441 		/* FALLTHROUGH */
442 	case SIOCAIFADDR_IN6:
443 		/*
444 		 * We always require users to specify a valid IPv6 address for
445 		 * the corresponding operation.
446 		 */
447 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
448 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
449 			error = EAFNOSUPPORT;
450 			goto out;
451 		}
452 
453 		if (td != NULL) {
454 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
455 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
456 			if (error)
457 				goto out;
458 		}
459 		/* FALLTHROUGH */
460 	case SIOCGIFSTAT_IN6:
461 	case SIOCGIFSTAT_ICMP6:
462 		if (ifp->if_afdata[AF_INET6] == NULL) {
463 			error = EPFNOSUPPORT;
464 			goto out;
465 		}
466 		break;
467 
468 	case SIOCGIFADDR_IN6:
469 		/* This interface is basically deprecated. use SIOCGIFCONF. */
470 		/* FALLTHROUGH */
471 	case SIOCGIFAFLAG_IN6:
472 	case SIOCGIFNETMASK_IN6:
473 	case SIOCGIFDSTADDR_IN6:
474 	case SIOCGIFALIFETIME_IN6:
475 		/* must think again about its semantics */
476 		if (ia == NULL) {
477 			error = EADDRNOTAVAIL;
478 			goto out;
479 		}
480 		break;
481 	}
482 
483 	switch (cmd) {
484 	case SIOCGIFADDR_IN6:
485 		ifr->ifr_addr = ia->ia_addr;
486 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
487 			goto out;
488 		break;
489 
490 	case SIOCGIFDSTADDR_IN6:
491 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
492 			error = EINVAL;
493 			goto out;
494 		}
495 		ifr->ifr_dstaddr = ia->ia_dstaddr;
496 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
497 			goto out;
498 		break;
499 
500 	case SIOCGIFNETMASK_IN6:
501 		ifr->ifr_addr = ia->ia_prefixmask;
502 		break;
503 
504 	case SIOCGIFAFLAG_IN6:
505 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
506 		break;
507 
508 	case SIOCGIFSTAT_IN6:
509 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
510 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
511 		    &ifr->ifr_ifru.ifru_stat,
512 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
513 		break;
514 
515 	case SIOCGIFSTAT_ICMP6:
516 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
517 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
518 		    &ifr->ifr_ifru.ifru_icmp6stat,
519 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
520 		break;
521 
522 	case SIOCGIFALIFETIME_IN6:
523 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
524 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
525 			time_t maxexpire;
526 			struct in6_addrlifetime *retlt =
527 			    &ifr->ifr_ifru.ifru_lifetime;
528 
529 			/*
530 			 * XXX: adjust expiration time assuming time_t is
531 			 * signed.
532 			 */
533 			maxexpire = (-1) &
534 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
535 			if (ia->ia6_lifetime.ia6t_vltime <
536 			    maxexpire - ia->ia6_updatetime) {
537 				retlt->ia6t_expire = ia->ia6_updatetime +
538 				    ia->ia6_lifetime.ia6t_vltime;
539 			} else
540 				retlt->ia6t_expire = maxexpire;
541 		}
542 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
543 			time_t maxexpire;
544 			struct in6_addrlifetime *retlt =
545 			    &ifr->ifr_ifru.ifru_lifetime;
546 
547 			/*
548 			 * XXX: adjust expiration time assuming time_t is
549 			 * signed.
550 			 */
551 			maxexpire = (-1) &
552 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
553 			if (ia->ia6_lifetime.ia6t_pltime <
554 			    maxexpire - ia->ia6_updatetime) {
555 				retlt->ia6t_preferred = ia->ia6_updatetime +
556 				    ia->ia6_lifetime.ia6t_pltime;
557 			} else
558 				retlt->ia6t_preferred = maxexpire;
559 		}
560 		break;
561 
562 	case SIOCAIFADDR_IN6:
563 		error = in6_addifaddr(ifp, ifra, ia);
564 		ia = NULL;
565 		break;
566 
567 	case SIOCDIFADDR_IN6:
568 		in6_purgeifaddr(ia);
569 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
570 		    IFADDR_EVENT_DEL);
571 		break;
572 
573 	default:
574 		if (ifp->if_ioctl == NULL) {
575 			error = EOPNOTSUPP;
576 			goto out;
577 		}
578 		error = (*ifp->if_ioctl)(ifp, cmd, data);
579 		goto out;
580 	}
581 
582 	error = 0;
583 out:
584 	if (ia != NULL)
585 		ifa_free(&ia->ia_ifa);
586 	return (error);
587 }
588 
589 static struct in6_multi_mship *
590 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
591     int *errorp, int delay)
592 {
593 	struct in6_multi_mship *imm;
594 	int error;
595 
596 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
597 	if (imm == NULL) {
598 		*errorp = ENOBUFS;
599 		return (NULL);
600 	}
601 
602 	delay = (delay * MLD_FASTHZ) / hz;
603 
604 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
605 	if (error) {
606 		*errorp = error;
607 		free(imm, M_IP6MADDR);
608 		return (NULL);
609 	}
610 
611 	return (imm);
612 }
613 
614 static int
615 in6_solicited_node_maddr(struct in6_addr *maddr,
616     struct ifnet *ifp, const struct in6_addr *base)
617 {
618 	int error;
619 
620 	bzero(maddr, sizeof(struct in6_addr));
621 	maddr->s6_addr32[0] = IPV6_ADDR_INT32_MLL;
622 	maddr->s6_addr32[2] = htonl(1);
623 	maddr->s6_addr32[3] = base->s6_addr32[3];
624 	maddr->s6_addr8[12] = 0xff;
625 	if ((error = in6_setscope(maddr, ifp, NULL)) != 0) {
626 		/* XXX: should not happen */
627 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
628 	}
629 
630 	return error;
631 }
632 
633 /*
634  * Join necessary multicast groups.  Factored out from in6_update_ifa().
635  * This entire work should only be done once, for the default FIB.
636  */
637 static int
638 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
639     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
640 {
641 	char ip6buf[INET6_ADDRSTRLEN];
642 	struct in6_addr mltaddr;
643 	struct in6_multi_mship *imm;
644 	int delay, error;
645 
646 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
647 
648 	/* Join solicited multicast addr for new host id. */
649 	if ((error = in6_solicited_node_maddr(&mltaddr, ifp,
650 	    &ifra->ifra_addr.sin6_addr)) != 0)
651 		goto cleanup;
652 	delay = error = 0;
653 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
654 		/*
655 		 * We need a random delay for DAD on the address being
656 		 * configured.  It also means delaying transmission of the
657 		 * corresponding MLD report to avoid report collision.
658 		 * [RFC 4861, Section 6.3.7]
659 		 */
660 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
661 	}
662 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
663 	if (imm == NULL) {
664 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
665 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
666 		    if_name(ifp), error));
667 		goto cleanup;
668 	}
669 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
670 	*in6m_sol = imm->i6mm_maddr;
671 
672 	/*
673 	 * Join link-local all-nodes address.
674 	 */
675 	mltaddr = in6addr_linklocal_allnodes;
676 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
677 		goto cleanup; /* XXX: should not fail */
678 
679 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
680 	if (imm == NULL) {
681 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
682 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
683 		    if_name(ifp), error));
684 		goto cleanup;
685 	}
686 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
687 
688 	/*
689 	 * Join node information group address.
690 	 */
691 	delay = 0;
692 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
693 		/*
694 		 * The spec does not say anything about delay for this group,
695 		 * but the same logic should apply.
696 		 */
697 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
698 	}
699 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
700 		/* XXX jinmei */
701 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
702 		if (imm == NULL)
703 			nd6log((LOG_WARNING,
704 			    "%s: in6_joingroup failed for %s on %s "
705 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
706 			    &mltaddr), if_name(ifp), error));
707 			/* XXX not very fatal, go on... */
708 		else
709 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
710 	}
711 	if (V_icmp6_nodeinfo_oldmcprefix &&
712 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
713 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
714 		if (imm == NULL)
715 			nd6log((LOG_WARNING,
716 			    "%s: in6_joingroup failed for %s on %s "
717 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
718 			    &mltaddr), if_name(ifp), error));
719 			/* XXX not very fatal, go on... */
720 		else
721 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
722 	}
723 
724 	/*
725 	 * Join interface-local all-nodes address.
726 	 * (ff01::1%ifN, and ff01::%ifN/32)
727 	 */
728 	mltaddr = in6addr_nodelocal_allnodes;
729 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
730 		goto cleanup; /* XXX: should not fail */
731 
732 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
733 	if (imm == NULL) {
734 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
735 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
736 		    &mltaddr), if_name(ifp), error));
737 		goto cleanup;
738 	}
739 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
740 
741 cleanup:
742 	return (error);
743 }
744 
745 /*
746  * Update parameters of an IPv6 interface address.
747  * If necessary, a new entry is created and linked into address chains.
748  * This function is separated from in6_control().
749  */
750 int
751 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
752     struct in6_ifaddr *ia, int flags)
753 {
754 	int error, hostIsNew = 0;
755 
756 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
757 		return (error);
758 
759 	if (ia == NULL) {
760 		hostIsNew = 1;
761 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
762 			return (ENOBUFS);
763 	}
764 
765 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
766 	if (error != 0) {
767 		if (hostIsNew != 0) {
768 			in6_unlink_ifa(ia, ifp);
769 			ifa_free(&ia->ia_ifa);
770 		}
771 		return (error);
772 	}
773 
774 	if (hostIsNew)
775 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
776 
777 	return (error);
778 }
779 
780 /*
781  * Fill in basic IPv6 address request info.
782  */
783 void
784 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
785     const struct in6_addr *mask)
786 {
787 
788 	memset(ifra, 0, sizeof(struct in6_aliasreq));
789 
790 	ifra->ifra_addr.sin6_family = AF_INET6;
791 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
792 	if (addr != NULL)
793 		ifra->ifra_addr.sin6_addr = *addr;
794 
795 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
796 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
797 	if (mask != NULL)
798 		ifra->ifra_prefixmask.sin6_addr = *mask;
799 }
800 
801 static int
802 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
803     struct in6_ifaddr *ia, int flags)
804 {
805 	int plen = -1;
806 	struct sockaddr_in6 dst6;
807 	struct in6_addrlifetime *lt;
808 	char ip6buf[INET6_ADDRSTRLEN];
809 
810 	/* Validate parameters */
811 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
812 		return (EINVAL);
813 
814 	/*
815 	 * The destination address for a p2p link must have a family
816 	 * of AF_UNSPEC or AF_INET6.
817 	 */
818 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
819 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
820 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
821 		return (EAFNOSUPPORT);
822 
823 	/*
824 	 * Validate address
825 	 */
826 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
827 	    ifra->ifra_addr.sin6_family != AF_INET6)
828 		return (EINVAL);
829 
830 	/*
831 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
832 	 * does not carry fields other than sin6_len.
833 	 */
834 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
835 		return (EINVAL);
836 	/*
837 	 * Because the IPv6 address architecture is classless, we require
838 	 * users to specify a (non 0) prefix length (mask) for a new address.
839 	 * We also require the prefix (when specified) mask is valid, and thus
840 	 * reject a non-consecutive mask.
841 	 */
842 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
843 		return (EINVAL);
844 	if (ifra->ifra_prefixmask.sin6_len != 0) {
845 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
846 		    (u_char *)&ifra->ifra_prefixmask +
847 		    ifra->ifra_prefixmask.sin6_len);
848 		if (plen <= 0)
849 			return (EINVAL);
850 	} else {
851 		/*
852 		 * In this case, ia must not be NULL.  We just use its prefix
853 		 * length.
854 		 */
855 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
856 	}
857 	/*
858 	 * If the destination address on a p2p interface is specified,
859 	 * and the address is a scoped one, validate/set the scope
860 	 * zone identifier.
861 	 */
862 	dst6 = ifra->ifra_dstaddr;
863 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
864 	    (dst6.sin6_family == AF_INET6)) {
865 		struct in6_addr in6_tmp;
866 		u_int32_t zoneid;
867 
868 		in6_tmp = dst6.sin6_addr;
869 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
870 			return (EINVAL); /* XXX: should be impossible */
871 
872 		if (dst6.sin6_scope_id != 0) {
873 			if (dst6.sin6_scope_id != zoneid)
874 				return (EINVAL);
875 		} else		/* user omit to specify the ID. */
876 			dst6.sin6_scope_id = zoneid;
877 
878 		/* convert into the internal form */
879 		if (sa6_embedscope(&dst6, 0))
880 			return (EINVAL); /* XXX: should be impossible */
881 	}
882 	/* Modify original ifra_dstaddr to reflect changes */
883 	ifra->ifra_dstaddr = dst6;
884 
885 	/*
886 	 * The destination address can be specified only for a p2p or a
887 	 * loopback interface.  If specified, the corresponding prefix length
888 	 * must be 128.
889 	 */
890 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
891 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
892 			/* XXX: noisy message */
893 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
894 			    "be specified for a p2p or a loopback IF only\n"));
895 			return (EINVAL);
896 		}
897 		if (plen != 128) {
898 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
899 			    "be 128 when dstaddr is specified\n"));
900 			return (EINVAL);
901 		}
902 	}
903 	/* lifetime consistency check */
904 	lt = &ifra->ifra_lifetime;
905 	if (lt->ia6t_pltime > lt->ia6t_vltime)
906 		return (EINVAL);
907 	if (lt->ia6t_vltime == 0) {
908 		/*
909 		 * the following log might be noisy, but this is a typical
910 		 * configuration mistake or a tool's bug.
911 		 */
912 		nd6log((LOG_INFO,
913 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
914 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
915 
916 		if (ia == NULL)
917 			return (0); /* there's nothing to do */
918 	}
919 
920 	/* Check prefix mask */
921 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
922 		/*
923 		 * We prohibit changing the prefix length of an existing
924 		 * address, because
925 		 * + such an operation should be rare in IPv6, and
926 		 * + the operation would confuse prefix management.
927 		 */
928 		if (ia->ia_prefixmask.sin6_len != 0 &&
929 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
930 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
931 			    "of an existing %s address should not be changed\n",
932 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
933 
934 			return (EINVAL);
935 		}
936 	}
937 
938 	return (0);
939 }
940 
941 /*
942  * Allocate a new ifaddr and link it into chains.
943  */
944 static struct in6_ifaddr *
945 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
946 {
947 	struct in6_ifaddr *ia;
948 
949 	/*
950 	 * When in6_alloc_ifa() is called in a process of a received
951 	 * RA, it is called under an interrupt context.  So, we should
952 	 * call malloc with M_NOWAIT.
953 	 */
954 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
955 	if (ia == NULL)
956 		return (NULL);
957 	LIST_INIT(&ia->ia6_memberships);
958 	/* Initialize the address and masks, and put time stamp */
959 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
960 	ia->ia_addr.sin6_family = AF_INET6;
961 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
962 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
963 	ia->ia_addr = ifra->ifra_addr;
964 	ia->ia6_createtime = time_uptime;
965 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
966 		/*
967 		 * Some functions expect that ifa_dstaddr is not
968 		 * NULL for p2p interfaces.
969 		 */
970 		ia->ia_ifa.ifa_dstaddr =
971 		    (struct sockaddr *)&ia->ia_dstaddr;
972 	} else {
973 		ia->ia_ifa.ifa_dstaddr = NULL;
974 	}
975 
976 	/* set prefix mask if any */
977 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
978 	if (ifra->ifra_prefixmask.sin6_len != 0) {
979 		ia->ia_prefixmask.sin6_family = AF_INET6;
980 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
981 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
982 	}
983 
984 	ia->ia_ifp = ifp;
985 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
986 	IF_ADDR_WLOCK(ifp);
987 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
988 	IF_ADDR_WUNLOCK(ifp);
989 
990 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
991 	IN6_IFADDR_WLOCK();
992 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
993 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
994 	IN6_IFADDR_WUNLOCK();
995 
996 	return (ia);
997 }
998 
999 /*
1000  * Update/configure interface address parameters:
1001  *
1002  * 1) Update lifetime
1003  * 2) Update interface metric ad flags
1004  * 3) Notify other subsystems
1005  */
1006 static int
1007 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1008     struct in6_ifaddr *ia, int hostIsNew, int flags)
1009 {
1010 	int error;
1011 
1012 	/* update timestamp */
1013 	ia->ia6_updatetime = time_uptime;
1014 
1015 	/*
1016 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1017 	 * to see if the address is deprecated or invalidated, but initialize
1018 	 * these members for applications.
1019 	 */
1020 	ia->ia6_lifetime = ifra->ifra_lifetime;
1021 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1022 		ia->ia6_lifetime.ia6t_expire =
1023 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1024 	} else
1025 		ia->ia6_lifetime.ia6t_expire = 0;
1026 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1027 		ia->ia6_lifetime.ia6t_preferred =
1028 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1029 	} else
1030 		ia->ia6_lifetime.ia6t_preferred = 0;
1031 
1032 	/*
1033 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1034 	 * userland, make it deprecated.
1035 	 */
1036 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1037 		ia->ia6_lifetime.ia6t_pltime = 0;
1038 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1039 	}
1040 
1041 	/*
1042 	 * configure address flags.
1043 	 */
1044 	ia->ia6_flags = ifra->ifra_flags;
1045 
1046 	/*
1047 	 * Make the address tentative before joining multicast addresses,
1048 	 * so that corresponding MLD responses would not have a tentative
1049 	 * source address.
1050 	 */
1051 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1052 
1053 	/*
1054 	 * DAD should be performed for an new address or addresses on
1055 	 * an interface with ND6_IFF_IFDISABLED.
1056 	 */
1057 	if (in6if_do_dad(ifp) &&
1058 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1059 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1060 
1061 	/* notify other subsystems */
1062 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1063 
1064 	return (error);
1065 }
1066 
1067 /*
1068  * Do link-level ifa job:
1069  * 1) Add lle entry for added address
1070  * 2) Notifies routing socket users about new address
1071  * 3) join appropriate multicast group
1072  * 4) start DAD if enabled
1073  */
1074 static int
1075 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1076     struct in6_ifaddr *ia, int flags)
1077 {
1078 	struct in6_multi *in6m_sol;
1079 	int error = 0;
1080 
1081 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1082 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1083 		in6_purgeaddr(&ia->ia_ifa);
1084 		ifa_free(&ia->ia_ifa);
1085 		return (error);
1086 	}
1087 
1088 	/* Join necessary multicast groups. */
1089 	in6m_sol = NULL;
1090 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1091 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1092 		if (error != 0) {
1093 			in6_purgeaddr(&ia->ia_ifa);
1094 			ifa_free(&ia->ia_ifa);
1095 			return (error);
1096 		}
1097 	}
1098 
1099 	/* Perform DAD, if the address is TENTATIVE. */
1100 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1101 		int delay, mindelay, maxdelay;
1102 
1103 		delay = 0;
1104 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1105 			/*
1106 			 * We need to impose a delay before sending an NS
1107 			 * for DAD.  Check if we also needed a delay for the
1108 			 * corresponding MLD message.  If we did, the delay
1109 			 * should be larger than the MLD delay (this could be
1110 			 * relaxed a bit, but this simple logic is at least
1111 			 * safe).
1112 			 * XXX: Break data hiding guidelines and look at
1113 			 * state for the solicited multicast group.
1114 			 */
1115 			mindelay = 0;
1116 			if (in6m_sol != NULL &&
1117 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1118 				mindelay = in6m_sol->in6m_timer;
1119 			}
1120 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1121 			if (maxdelay - mindelay == 0)
1122 				delay = 0;
1123 			else {
1124 				delay =
1125 				    (arc4random() % (maxdelay - mindelay)) +
1126 				    mindelay;
1127 			}
1128 		}
1129 		nd6_dad_start((struct ifaddr *)ia, delay);
1130 	}
1131 
1132 	in6_newaddrmsg(ia, RTM_ADD);
1133 	ifa_free(&ia->ia_ifa);
1134 	return (error);
1135 }
1136 
1137 /*
1138  * Adds or deletes interface route for p2p ifa.
1139  * Returns 0 on success or errno.
1140  */
1141 static int
1142 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1143 {
1144 	struct epoch_tracker et;
1145 	struct ifaddr *ifa = &ia->ia_ifa;
1146 	int error;
1147 
1148 	/* Prepare gateway */
1149 	struct sockaddr_dl_short sdl = {
1150 		.sdl_family = AF_LINK,
1151 		.sdl_len = sizeof(struct sockaddr_dl_short),
1152 		.sdl_type = ifa->ifa_ifp->if_type,
1153 		.sdl_index = ifa->ifa_ifp->if_index,
1154 	};
1155 
1156 	struct sockaddr_in6 dst = {
1157 		.sin6_family = AF_INET6,
1158 		.sin6_len = sizeof(struct sockaddr_in6),
1159 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1160 	};
1161 
1162 	struct rt_addrinfo info = {
1163 		.rti_ifa = ifa,
1164 		.rti_ifp = ifa->ifa_ifp,
1165 		.rti_flags = RTF_PINNED | RTF_HOST,
1166 		.rti_info = {
1167 			[RTAX_DST] = (struct sockaddr *)&dst,
1168 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1169 		},
1170 	};
1171 	/* Don't set additional per-gw filters on removal */
1172 
1173 	NET_EPOCH_ENTER(et);
1174 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1175 	NET_EPOCH_EXIT(et);
1176 
1177 	return (error);
1178 }
1179 
1180 static bool
1181 ifa_is_p2p(struct in6_ifaddr *ia)
1182 {
1183 	int plen;
1184 
1185 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1186 
1187 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1188 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1189 		return (true);
1190 
1191 	return (false);
1192 }
1193 
1194 int
1195 in6_addifaddr(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia)
1196 {
1197 	struct nd_prefixctl pr0;
1198 	struct nd_prefix *pr;
1199 	int carp_attached = 0;
1200 	int error;
1201 
1202 	/*
1203 	 * first, make or update the interface address structure,
1204 	 * and link it to the list.
1205 	 */
1206 	if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
1207 		goto out;
1208 	if (ia != NULL) {
1209 		if (ia->ia_ifa.ifa_carp)
1210 			(*carp_detach_p)(&ia->ia_ifa, true);
1211 		ifa_free(&ia->ia_ifa);
1212 	}
1213 	if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) {
1214 		/*
1215 		 * this can happen when the user specify the 0 valid
1216 		 * lifetime.
1217 		 */
1218 		return (0);
1219 	}
1220 
1221 	if (ifra->ifra_vhid > 0) {
1222 		if (carp_attach_p != NULL)
1223 			error = (*carp_attach_p)(&ia->ia_ifa,
1224 			    ifra->ifra_vhid);
1225 		else
1226 			error = EPROTONOSUPPORT;
1227 		if (error)
1228 			goto out;
1229 		else
1230 			carp_attached = 1;
1231 	}
1232 
1233 	/*
1234 	 * then, make the prefix on-link on the interface.
1235 	 * XXX: we'd rather create the prefix before the address, but
1236 	 * we need at least one address to install the corresponding
1237 	 * interface route, so we configure the address first.
1238 	 */
1239 
1240 	/*
1241 	 * convert mask to prefix length (prefixmask has already
1242 	 * been validated in in6_update_ifa().
1243 	 */
1244 	bzero(&pr0, sizeof(pr0));
1245 	pr0.ndpr_ifp = ifp;
1246 	pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1247 	    NULL);
1248 	if (pr0.ndpr_plen == 128) {
1249 		/* we don't need to install a host route. */
1250 		goto aifaddr_out;
1251 	}
1252 	pr0.ndpr_prefix = ifra->ifra_addr;
1253 	/* apply the mask for safety. */
1254 	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
1255 	    &ifra->ifra_prefixmask.sin6_addr);
1256 
1257 	/*
1258 	 * XXX: since we don't have an API to set prefix (not address)
1259 	 * lifetimes, we just use the same lifetimes as addresses.
1260 	 * The (temporarily) installed lifetimes can be overridden by
1261 	 * later advertised RAs (when accept_rtadv is non 0), which is
1262 	 * an intended behavior.
1263 	 */
1264 	pr0.ndpr_raf_onlink = 1; /* should be configurable? */
1265 	pr0.ndpr_raf_auto =
1266 	    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
1267 	pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
1268 	pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
1269 
1270 	/* add the prefix if not yet. */
1271 	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
1272 		/*
1273 		 * nd6_prelist_add will install the corresponding
1274 		 * interface route.
1275 		 */
1276 		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
1277 			if (carp_attached)
1278 				(*carp_detach_p)(&ia->ia_ifa, false);
1279 			goto out;
1280 		}
1281 	}
1282 
1283 	/* relate the address to the prefix */
1284 	if (ia->ia6_ndpr == NULL) {
1285 		ia->ia6_ndpr = pr;
1286 		pr->ndpr_addrcnt++;
1287 
1288 		/*
1289 		 * If this is the first autoconf address from the
1290 		 * prefix, create a temporary address as well
1291 		 * (when required).
1292 		 */
1293 		if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
1294 		    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
1295 			int e;
1296 			if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
1297 				log(LOG_NOTICE, "in6_control: failed "
1298 				    "to create a temporary address, "
1299 				    "errno=%d\n", e);
1300 			}
1301 		}
1302 	}
1303 	nd6_prefix_rele(pr);
1304 
1305 	/*
1306 	 * this might affect the status of autoconfigured addresses,
1307 	 * that is, this address might make other addresses detached.
1308 	 */
1309 	pfxlist_onlink_check();
1310 
1311 aifaddr_out:
1312 	/*
1313 	 * Try to clear the flag when a new IPv6 address is added
1314 	 * onto an IFDISABLED interface and it succeeds.
1315 	 */
1316 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1317 		struct in6_ndireq nd;
1318 
1319 		memset(&nd, 0, sizeof(nd));
1320 		nd.ndi.flags = ND_IFINFO(ifp)->flags;
1321 		nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
1322 		if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
1323 			log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
1324 			    "SIOCSIFINFO_FLAGS for -ifdisabled "
1325 			    "failed.");
1326 		/*
1327 		 * Ignore failure of clearing the flag intentionally.
1328 		 * The failure means address duplication was detected.
1329 		 */
1330 	}
1331 	error = 0;
1332 
1333 out:
1334 	if (ia != NULL)
1335 		ifa_free(&ia->ia_ifa);
1336 	return (error);
1337 }
1338 
1339 void
1340 in6_purgeaddr(struct ifaddr *ifa)
1341 {
1342 	struct ifnet *ifp = ifa->ifa_ifp;
1343 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1344 	struct in6_multi_mship *imm;
1345 	int error;
1346 
1347 	if (ifa->ifa_carp)
1348 		(*carp_detach_p)(ifa, false);
1349 
1350 	/*
1351 	 * Remove the loopback route to the interface address.
1352 	 * The check for the current setting of "nd6_useloopback"
1353 	 * is not needed.
1354 	 */
1355 	if (ia->ia_flags & IFA_RTSELF) {
1356 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1357 		    (struct sockaddr *)&ia->ia_addr);
1358 		if (error == 0)
1359 			ia->ia_flags &= ~IFA_RTSELF;
1360 	}
1361 
1362 	/* stop DAD processing */
1363 	nd6_dad_stop(ifa);
1364 
1365 	/* Leave multicast groups. */
1366 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1367 		LIST_REMOVE(imm, i6mm_chain);
1368 		if (imm->i6mm_maddr != NULL)
1369 			in6_leavegroup(imm->i6mm_maddr, NULL);
1370 		free(imm, M_IP6MADDR);
1371 	}
1372 	/* Check if we need to remove p2p route */
1373 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1374 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1375 		if (error != 0)
1376 			log(LOG_INFO, "%s: err=%d, destination address delete "
1377 			    "failed\n", __func__, error);
1378 		ia->ia_flags &= ~IFA_ROUTE;
1379 	}
1380 
1381 	in6_newaddrmsg(ia, RTM_DELETE);
1382 	in6_unlink_ifa(ia, ifp);
1383 }
1384 
1385 /*
1386  * Removes @ia from the corresponding interfaces and unlinks corresponding
1387  *  prefix if no addresses are using it anymore.
1388  */
1389 void
1390 in6_purgeifaddr(struct in6_ifaddr *ia)
1391 {
1392 	struct nd_prefix *pr;
1393 
1394 	/*
1395 	 * If the address being deleted is the only one that owns
1396 	 * the corresponding prefix, expire the prefix as well.
1397 	 * XXX: theoretically, we don't have to worry about such
1398 	 * relationship, since we separate the address management
1399 	 * and the prefix management.  We do this, however, to provide
1400 	 * as much backward compatibility as possible in terms of
1401 	 * the ioctl operation.
1402 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1403 	 */
1404 	pr = ia->ia6_ndpr;
1405 	in6_purgeaddr(&ia->ia_ifa);
1406 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1407 		ND6_WLOCK();
1408 		nd6_prefix_unlink(pr, NULL);
1409 		ND6_WUNLOCK();
1410 		nd6_prefix_del(pr);
1411 	}
1412 }
1413 
1414 
1415 static void
1416 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1417 {
1418 	char ip6buf[INET6_ADDRSTRLEN];
1419 	int remove_lle;
1420 
1421 	IF_ADDR_WLOCK(ifp);
1422 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1423 	IF_ADDR_WUNLOCK(ifp);
1424 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1425 
1426 	/*
1427 	 * Defer the release of what might be the last reference to the
1428 	 * in6_ifaddr so that it can't be freed before the remainder of the
1429 	 * cleanup.
1430 	 */
1431 	IN6_IFADDR_WLOCK();
1432 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1433 	CK_LIST_REMOVE(ia, ia6_hash);
1434 	IN6_IFADDR_WUNLOCK();
1435 
1436 	/*
1437 	 * Release the reference to the base prefix.  There should be a
1438 	 * positive reference.
1439 	 */
1440 	remove_lle = 0;
1441 	if (ia->ia6_ndpr == NULL) {
1442 		nd6log((LOG_NOTICE,
1443 		    "in6_unlink_ifa: autoconf'ed address "
1444 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1445 	} else {
1446 		ia->ia6_ndpr->ndpr_addrcnt--;
1447 		/* Do not delete lles within prefix if refcont != 0 */
1448 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1449 			remove_lle = 1;
1450 		ia->ia6_ndpr = NULL;
1451 	}
1452 
1453 	nd6_rem_ifa_lle(ia, remove_lle);
1454 
1455 	/*
1456 	 * Also, if the address being removed is autoconf'ed, call
1457 	 * pfxlist_onlink_check() since the release might affect the status of
1458 	 * other (detached) addresses.
1459 	 */
1460 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1461 		pfxlist_onlink_check();
1462 	}
1463 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1464 }
1465 
1466 /*
1467  * Notifies other subsystems about address change/arrival:
1468  * 1) Notifies device handler on the first IPv6 address assignment
1469  * 2) Handle routing table changes for P2P links and route
1470  * 3) Handle routing table changes for address host route
1471  */
1472 static int
1473 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1474     struct in6_aliasreq *ifra, int hostIsNew)
1475 {
1476 	int	error = 0, ifacount = 0;
1477 	struct ifaddr *ifa;
1478 	struct sockaddr_in6 *pdst;
1479 	char ip6buf[INET6_ADDRSTRLEN];
1480 
1481 	/*
1482 	 * Give the interface a chance to initialize
1483 	 * if this is its first address,
1484 	 */
1485 	if (hostIsNew != 0) {
1486 		struct epoch_tracker et;
1487 
1488 		NET_EPOCH_ENTER(et);
1489 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1490 			if (ifa->ifa_addr->sa_family != AF_INET6)
1491 				continue;
1492 			ifacount++;
1493 		}
1494 		NET_EPOCH_EXIT(et);
1495 	}
1496 
1497 	if (ifacount <= 1 && ifp->if_ioctl) {
1498 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1499 		if (error)
1500 			goto done;
1501 	}
1502 
1503 	/*
1504 	 * If a new destination address is specified, scrub the old one and
1505 	 * install the new destination.  Note that the interface must be
1506 	 * p2p or loopback.
1507 	 */
1508 	pdst = &ifra->ifra_dstaddr;
1509 	if (pdst->sin6_family == AF_INET6 &&
1510 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1511 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1512 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1513 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1514 			    "remove a route to the old destination: %s\n",
1515 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1516 			/* proceed anyway... */
1517 		} else
1518 			ia->ia_flags &= ~IFA_ROUTE;
1519 		ia->ia_dstaddr = *pdst;
1520 	}
1521 
1522 	/*
1523 	 * If a new destination address is specified for a point-to-point
1524 	 * interface, install a route to the destination as an interface
1525 	 * direct route.
1526 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1527 	 * interface that share the same destination.
1528 	 */
1529 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1530 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1531 		if (error)
1532 			goto done;
1533 		ia->ia_flags |= IFA_ROUTE;
1534 	}
1535 
1536 	/*
1537 	 * add a loopback route to self if not exists
1538 	 */
1539 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1540 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1541 		    (struct sockaddr *)&ia->ia_addr);
1542 		if (error == 0)
1543 			ia->ia_flags |= IFA_RTSELF;
1544 	}
1545 done:
1546 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1547 	    "Invoking IPv6 network device address event may sleep");
1548 
1549 	ifa_ref(&ia->ia_ifa);
1550 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1551 	    IFADDR_EVENT_ADD);
1552 	ifa_free(&ia->ia_ifa);
1553 
1554 	return (error);
1555 }
1556 
1557 /*
1558  * Find an IPv6 interface link-local address specific to an interface.
1559  * ifaddr is returned referenced.
1560  */
1561 struct in6_ifaddr *
1562 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1563 {
1564 	struct ifaddr *ifa;
1565 
1566 	NET_EPOCH_ASSERT();
1567 
1568 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1569 		if (ifa->ifa_addr->sa_family != AF_INET6)
1570 			continue;
1571 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1572 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1573 			    ignoreflags) != 0)
1574 				continue;
1575 			ifa_ref(ifa);
1576 			break;
1577 		}
1578 	}
1579 
1580 	return ((struct in6_ifaddr *)ifa);
1581 }
1582 
1583 /*
1584  * find the interface address corresponding to a given IPv6 address.
1585  * ifaddr is returned referenced if @referenced flag is set.
1586  */
1587 struct in6_ifaddr *
1588 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1589 {
1590 	struct rm_priotracker in6_ifa_tracker;
1591 	struct in6_ifaddr *ia;
1592 
1593 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1594 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1595 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1596 			if (zoneid != 0 &&
1597 			    zoneid != ia->ia_addr.sin6_scope_id)
1598 				continue;
1599 			if (referenced)
1600 				ifa_ref(&ia->ia_ifa);
1601 			break;
1602 		}
1603 	}
1604 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1605 	return (ia);
1606 }
1607 
1608 /*
1609  * find the internet address corresponding to a given interface and address.
1610  * ifaddr is returned referenced.
1611  */
1612 struct in6_ifaddr *
1613 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1614 {
1615 	struct epoch_tracker et;
1616 	struct ifaddr *ifa;
1617 
1618 	NET_EPOCH_ENTER(et);
1619 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1620 		if (ifa->ifa_addr->sa_family != AF_INET6)
1621 			continue;
1622 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1623 			ifa_ref(ifa);
1624 			break;
1625 		}
1626 	}
1627 	NET_EPOCH_EXIT(et);
1628 
1629 	return ((struct in6_ifaddr *)ifa);
1630 }
1631 
1632 /*
1633  * Find a link-local scoped address on ifp and return it if any.
1634  */
1635 struct in6_ifaddr *
1636 in6ifa_llaonifp(struct ifnet *ifp)
1637 {
1638 	struct epoch_tracker et;
1639 	struct sockaddr_in6 *sin6;
1640 	struct ifaddr *ifa;
1641 
1642 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1643 		return (NULL);
1644 	NET_EPOCH_ENTER(et);
1645 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1646 		if (ifa->ifa_addr->sa_family != AF_INET6)
1647 			continue;
1648 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1649 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1650 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1651 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1652 			break;
1653 	}
1654 	NET_EPOCH_EXIT(et);
1655 
1656 	return ((struct in6_ifaddr *)ifa);
1657 }
1658 
1659 /*
1660  * Convert IP6 address to printable (loggable) representation. Caller
1661  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1662  */
1663 static char digits[] = "0123456789abcdef";
1664 char *
1665 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1666 {
1667 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1668 	char *cp;
1669 	const u_int16_t *a = (const u_int16_t *)addr;
1670 	const u_int8_t *d;
1671 	int dcolon = 0, zero = 0;
1672 
1673 	cp = ip6buf;
1674 
1675 	for (i = 0; i < 8; i++) {
1676 		if (*(a + i) == 0) {
1677 			cnt++;
1678 			if (cnt == 1)
1679 				idx = i;
1680 		}
1681 		else if (maxcnt < cnt) {
1682 			maxcnt = cnt;
1683 			index = idx;
1684 			cnt = 0;
1685 		}
1686 	}
1687 	if (maxcnt < cnt) {
1688 		maxcnt = cnt;
1689 		index = idx;
1690 	}
1691 
1692 	for (i = 0; i < 8; i++) {
1693 		if (dcolon == 1) {
1694 			if (*a == 0) {
1695 				if (i == 7)
1696 					*cp++ = ':';
1697 				a++;
1698 				continue;
1699 			} else
1700 				dcolon = 2;
1701 		}
1702 		if (*a == 0) {
1703 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1704 				if (i == 0)
1705 					*cp++ = ':';
1706 				*cp++ = ':';
1707 				dcolon = 1;
1708 			} else {
1709 				*cp++ = '0';
1710 				*cp++ = ':';
1711 			}
1712 			a++;
1713 			continue;
1714 		}
1715 		d = (const u_char *)a;
1716 		/* Try to eliminate leading zeros in printout like in :0001. */
1717 		zero = 1;
1718 		*cp = digits[*d >> 4];
1719 		if (*cp != '0') {
1720 			zero = 0;
1721 			cp++;
1722 		}
1723 		*cp = digits[*d++ & 0xf];
1724 		if (zero == 0 || (*cp != '0')) {
1725 			zero = 0;
1726 			cp++;
1727 		}
1728 		*cp = digits[*d >> 4];
1729 		if (zero == 0 || (*cp != '0')) {
1730 			zero = 0;
1731 			cp++;
1732 		}
1733 		*cp++ = digits[*d & 0xf];
1734 		*cp++ = ':';
1735 		a++;
1736 	}
1737 	*--cp = '\0';
1738 	return (ip6buf);
1739 }
1740 
1741 int
1742 in6_localaddr(struct in6_addr *in6)
1743 {
1744 	struct rm_priotracker in6_ifa_tracker;
1745 	struct in6_ifaddr *ia;
1746 
1747 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1748 		return 1;
1749 
1750 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1751 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1752 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1753 		    &ia->ia_prefixmask.sin6_addr)) {
1754 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1755 			return 1;
1756 		}
1757 	}
1758 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1759 
1760 	return (0);
1761 }
1762 
1763 /*
1764  * Return 1 if an internet address is for the local host and configured
1765  * on one of its interfaces.
1766  */
1767 int
1768 in6_localip(struct in6_addr *in6)
1769 {
1770 	struct rm_priotracker in6_ifa_tracker;
1771 	struct in6_ifaddr *ia;
1772 
1773 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1774 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1775 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1776 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1777 			return (1);
1778 		}
1779 	}
1780 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1781 	return (0);
1782 }
1783 
1784 /*
1785  * Like in6_localip(), but FIB-aware.
1786  */
1787 bool
1788 in6_localip_fib(struct in6_addr *in6, uint16_t fib)
1789 {
1790 	struct rm_priotracker in6_ifa_tracker;
1791 	struct in6_ifaddr *ia;
1792 
1793 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1794 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1795 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr) &&
1796 		    ia->ia_ifa.ifa_ifp->if_fib == fib) {
1797 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1798 			return (true);
1799 		}
1800 	}
1801 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1802 	return (false);
1803 }
1804 
1805 /*
1806  * Return 1 if an internet address is configured on an interface.
1807  */
1808 int
1809 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1810 {
1811 	struct in6_addr in6;
1812 	struct ifaddr *ifa;
1813 	struct in6_ifaddr *ia6;
1814 
1815 	NET_EPOCH_ASSERT();
1816 
1817 	in6 = *addr;
1818 	if (in6_clearscope(&in6))
1819 		return (0);
1820 	in6_setscope(&in6, ifp, NULL);
1821 
1822 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1823 		if (ifa->ifa_addr->sa_family != AF_INET6)
1824 			continue;
1825 		ia6 = (struct in6_ifaddr *)ifa;
1826 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1827 			return (1);
1828 	}
1829 
1830 	return (0);
1831 }
1832 
1833 int
1834 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1835 {
1836 	struct rm_priotracker in6_ifa_tracker;
1837 	struct in6_ifaddr *ia;
1838 
1839 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1840 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1841 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1842 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1843 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1844 				return (1); /* true */
1845 			}
1846 			break;
1847 		}
1848 	}
1849 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1850 
1851 	return (0);		/* false */
1852 }
1853 
1854 /*
1855  * return length of part which dst and src are equal
1856  * hard coding...
1857  */
1858 int
1859 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1860 {
1861 	int match = 0;
1862 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1863 	u_char *lim = s + 16, r;
1864 
1865 	while (s < lim)
1866 		if ((r = (*d++ ^ *s++)) != 0) {
1867 			while (r < 128) {
1868 				match++;
1869 				r <<= 1;
1870 			}
1871 			break;
1872 		} else
1873 			match += 8;
1874 	return match;
1875 }
1876 
1877 /* XXX: to be scope conscious */
1878 int
1879 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1880 {
1881 	int bytelen, bitlen;
1882 
1883 	/* sanity check */
1884 	if (0 > len || len > 128) {
1885 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1886 		    len);
1887 		return (0);
1888 	}
1889 
1890 	bytelen = len / 8;
1891 	bitlen = len % 8;
1892 
1893 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1894 		return (0);
1895 	if (bitlen != 0 &&
1896 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1897 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1898 		return (0);
1899 
1900 	return (1);
1901 }
1902 
1903 void
1904 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1905 {
1906 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1907 	int bytelen, bitlen, i;
1908 
1909 	/* sanity check */
1910 	if (0 > len || len > 128) {
1911 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1912 		    len);
1913 		return;
1914 	}
1915 
1916 	bzero(maskp, sizeof(*maskp));
1917 	bytelen = len / 8;
1918 	bitlen = len % 8;
1919 	for (i = 0; i < bytelen; i++)
1920 		maskp->s6_addr[i] = 0xff;
1921 	if (bitlen)
1922 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1923 }
1924 
1925 /*
1926  * return the best address out of the same scope. if no address was
1927  * found, return the first valid address from designated IF.
1928  */
1929 struct in6_ifaddr *
1930 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1931 {
1932 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1933 	struct ifaddr *ifa;
1934 	struct in6_ifaddr *besta = NULL;
1935 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1936 
1937 	NET_EPOCH_ASSERT();
1938 
1939 	dep[0] = dep[1] = NULL;
1940 
1941 	/*
1942 	 * We first look for addresses in the same scope.
1943 	 * If there is one, return it.
1944 	 * If two or more, return one which matches the dst longest.
1945 	 * If none, return one of global addresses assigned other ifs.
1946 	 */
1947 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1948 		if (ifa->ifa_addr->sa_family != AF_INET6)
1949 			continue;
1950 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1951 			continue; /* XXX: is there any case to allow anycast? */
1952 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1953 			continue; /* don't use this interface */
1954 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1955 			continue;
1956 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1957 			if (V_ip6_use_deprecated)
1958 				dep[0] = (struct in6_ifaddr *)ifa;
1959 			continue;
1960 		}
1961 
1962 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1963 			/*
1964 			 * call in6_matchlen() as few as possible
1965 			 */
1966 			if (besta) {
1967 				if (blen == -1)
1968 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1969 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1970 				if (tlen > blen) {
1971 					blen = tlen;
1972 					besta = (struct in6_ifaddr *)ifa;
1973 				}
1974 			} else
1975 				besta = (struct in6_ifaddr *)ifa;
1976 		}
1977 	}
1978 	if (besta)
1979 		return (besta);
1980 
1981 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1982 		if (ifa->ifa_addr->sa_family != AF_INET6)
1983 			continue;
1984 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1985 			continue; /* XXX: is there any case to allow anycast? */
1986 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1987 			continue; /* don't use this interface */
1988 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1989 			continue;
1990 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1991 			if (V_ip6_use_deprecated)
1992 				dep[1] = (struct in6_ifaddr *)ifa;
1993 			continue;
1994 		}
1995 
1996 		return (struct in6_ifaddr *)ifa;
1997 	}
1998 
1999 	/* use the last-resort values, that are, deprecated addresses */
2000 	if (dep[0])
2001 		return dep[0];
2002 	if (dep[1])
2003 		return dep[1];
2004 
2005 	return NULL;
2006 }
2007 
2008 /*
2009  * perform DAD when interface becomes IFF_UP.
2010  */
2011 void
2012 in6_if_up(struct ifnet *ifp)
2013 {
2014 	struct epoch_tracker et;
2015 	struct ifaddr *ifa;
2016 	struct in6_ifaddr *ia;
2017 
2018 	NET_EPOCH_ENTER(et);
2019 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2020 		if (ifa->ifa_addr->sa_family != AF_INET6)
2021 			continue;
2022 		ia = (struct in6_ifaddr *)ifa;
2023 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2024 			/*
2025 			 * The TENTATIVE flag was likely set by hand
2026 			 * beforehand, implicitly indicating the need for DAD.
2027 			 * We may be able to skip the random delay in this
2028 			 * case, but we impose delays just in case.
2029 			 */
2030 			nd6_dad_start(ifa,
2031 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2032 		}
2033 	}
2034 	NET_EPOCH_EXIT(et);
2035 
2036 	/*
2037 	 * special cases, like 6to4, are handled in in6_ifattach
2038 	 */
2039 	in6_ifattach(ifp, NULL);
2040 }
2041 
2042 int
2043 in6if_do_dad(struct ifnet *ifp)
2044 {
2045 
2046 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2047 		return (0);
2048 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2049 		return (0);
2050 	if ((ND_IFINFO(ifp)->flags &
2051 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2052 		return (0);
2053 	return (1);
2054 }
2055 
2056 /*
2057  * Calculate max IPv6 MTU through all the interfaces and store it
2058  * to in6_maxmtu.
2059  */
2060 void
2061 in6_setmaxmtu(void)
2062 {
2063 	struct epoch_tracker et;
2064 	unsigned long maxmtu = 0;
2065 	struct ifnet *ifp;
2066 
2067 	NET_EPOCH_ENTER(et);
2068 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2069 		/* this function can be called during ifnet initialization */
2070 		if (!ifp->if_afdata[AF_INET6])
2071 			continue;
2072 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2073 		    IN6_LINKMTU(ifp) > maxmtu)
2074 			maxmtu = IN6_LINKMTU(ifp);
2075 	}
2076 	NET_EPOCH_EXIT(et);
2077 	if (maxmtu)	/* update only when maxmtu is positive */
2078 		V_in6_maxmtu = maxmtu;
2079 }
2080 
2081 /*
2082  * Provide the length of interface identifiers to be used for the link attached
2083  * to the given interface.  The length should be defined in "IPv6 over
2084  * xxx-link" document.  Note that address architecture might also define
2085  * the length for a particular set of address prefixes, regardless of the
2086  * link type.  As clarified in rfc2462bis, those two definitions should be
2087  * consistent, and those really are as of August 2004.
2088  */
2089 int
2090 in6_if2idlen(struct ifnet *ifp)
2091 {
2092 	switch (ifp->if_type) {
2093 	case IFT_ETHER:		/* RFC2464 */
2094 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2095 	case IFT_L2VLAN:	/* ditto */
2096 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2097 	case IFT_INFINIBAND:
2098 		return (64);
2099 	case IFT_PPP:		/* RFC2472 */
2100 		return (64);
2101 	case IFT_FRELAY:	/* RFC2590 */
2102 		return (64);
2103 	case IFT_IEEE1394:	/* RFC3146 */
2104 		return (64);
2105 	case IFT_GIF:
2106 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2107 	case IFT_LOOP:
2108 		return (64);	/* XXX: is this really correct? */
2109 	default:
2110 		/*
2111 		 * Unknown link type:
2112 		 * It might be controversial to use the today's common constant
2113 		 * of 64 for these cases unconditionally.  For full compliance,
2114 		 * we should return an error in this case.  On the other hand,
2115 		 * if we simply miss the standard for the link type or a new
2116 		 * standard is defined for a new link type, the IFID length
2117 		 * is very likely to be the common constant.  As a compromise,
2118 		 * we always use the constant, but make an explicit notice
2119 		 * indicating the "unknown" case.
2120 		 */
2121 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2122 		return (64);
2123 	}
2124 }
2125 
2126 struct in6_llentry {
2127 	struct llentry		base;
2128 };
2129 
2130 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2131 #define	IN6_LLTBL_HASH(k, h) \
2132 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2133 
2134 /*
2135  * Do actual deallocation of @lle.
2136  */
2137 static void
2138 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2139 {
2140 	struct llentry *lle;
2141 
2142 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2143 	LLE_LOCK_DESTROY(lle);
2144 	LLE_REQ_DESTROY(lle);
2145 	free(lle, M_LLTABLE);
2146 }
2147 
2148 /*
2149  * Called by LLE_FREE_LOCKED when number of references
2150  * drops to zero.
2151  */
2152 static void
2153 in6_lltable_destroy_lle(struct llentry *lle)
2154 {
2155 
2156 	LLE_WUNLOCK(lle);
2157 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2158 }
2159 
2160 static struct llentry *
2161 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2162 {
2163 	struct in6_llentry *lle;
2164 
2165 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2166 	if (lle == NULL)		/* NB: caller generates msg */
2167 		return NULL;
2168 
2169 	lle->base.r_l3addr.addr6 = *addr6;
2170 	lle->base.lle_refcnt = 1;
2171 	lle->base.lle_free = in6_lltable_destroy_lle;
2172 	LLE_LOCK_INIT(&lle->base);
2173 	LLE_REQ_INIT(&lle->base);
2174 	callout_init(&lle->base.lle_timer, 1);
2175 
2176 	return (&lle->base);
2177 }
2178 
2179 static int
2180 in6_lltable_match_prefix(const struct sockaddr *saddr,
2181     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2182 {
2183 	const struct in6_addr *addr, *mask, *lle_addr;
2184 
2185 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2186 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2187 	lle_addr = &lle->r_l3addr.addr6;
2188 
2189 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2190 		return (0);
2191 
2192 	if (lle->la_flags & LLE_IFADDR) {
2193 		/*
2194 		 * Delete LLE_IFADDR records IFF address & flag matches.
2195 		 * Note that addr is the interface address within prefix
2196 		 * being matched.
2197 		 */
2198 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2199 		    (flags & LLE_STATIC) != 0)
2200 			return (1);
2201 		return (0);
2202 	}
2203 
2204 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2205 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2206 		return (1);
2207 
2208 	return (0);
2209 }
2210 
2211 static void
2212 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2213 {
2214 	struct ifnet *ifp __diagused;
2215 
2216 	LLE_WLOCK_ASSERT(lle);
2217 	KASSERT(llt != NULL, ("lltable is NULL"));
2218 
2219 	/* Unlink entry from table */
2220 	if ((lle->la_flags & LLE_LINKED) != 0) {
2221 		ifp = llt->llt_ifp;
2222 		IF_AFDATA_WLOCK_ASSERT(ifp);
2223 		lltable_unlink_entry(llt, lle);
2224 	}
2225 
2226 	llentry_free(lle);
2227 }
2228 
2229 static int
2230 in6_lltable_rtcheck(struct ifnet *ifp,
2231 		    u_int flags,
2232 		    const struct sockaddr *l3addr)
2233 {
2234 	const struct sockaddr_in6 *sin6;
2235 	struct nhop_object *nh;
2236 	struct in6_addr dst;
2237 	uint32_t scopeid;
2238 	char ip6buf[INET6_ADDRSTRLEN];
2239 	int fibnum;
2240 
2241 	NET_EPOCH_ASSERT();
2242 	KASSERT(l3addr->sa_family == AF_INET6,
2243 	    ("sin_family %d", l3addr->sa_family));
2244 
2245 	sin6 = (const struct sockaddr_in6 *)l3addr;
2246 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2247 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2248 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2249 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2250 		struct ifaddr *ifa;
2251 		/*
2252 		 * Create an ND6 cache for an IPv6 neighbor
2253 		 * that is not covered by our own prefix.
2254 		 */
2255 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2256 		if (ifa != NULL) {
2257 			return 0;
2258 		}
2259 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2260 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2261 		return EINVAL;
2262 	}
2263 	return 0;
2264 }
2265 
2266 static inline uint32_t
2267 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2268 {
2269 
2270 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2271 }
2272 
2273 static uint32_t
2274 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2275 {
2276 
2277 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2278 }
2279 
2280 static void
2281 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2282 {
2283 	struct sockaddr_in6 *sin6;
2284 
2285 	sin6 = (struct sockaddr_in6 *)sa;
2286 	bzero(sin6, sizeof(*sin6));
2287 	sin6->sin6_family = AF_INET6;
2288 	sin6->sin6_len = sizeof(*sin6);
2289 	sin6->sin6_addr = lle->r_l3addr.addr6;
2290 }
2291 
2292 static inline struct llentry *
2293 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2294 {
2295 	struct llentry *lle;
2296 	struct llentries *lleh;
2297 	u_int hashidx;
2298 
2299 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2300 	lleh = &llt->lle_head[hashidx];
2301 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2302 		if (lle->la_flags & LLE_DELETED)
2303 			continue;
2304 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2305 			break;
2306 	}
2307 
2308 	return (lle);
2309 }
2310 
2311 static void
2312 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2313 {
2314 
2315 	lle->la_flags |= LLE_DELETED;
2316 
2317 	/* Leave the solicited multicast group. */
2318 	if ((lle->la_flags & LLE_PUB) != 0)
2319 		in6_leave_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2320 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2321 #ifdef DIAGNOSTIC
2322 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2323 #endif
2324 	llentry_free(lle);
2325 }
2326 
2327 static struct llentry *
2328 in6_lltable_alloc(struct lltable *llt, u_int flags,
2329 	const struct sockaddr *l3addr)
2330 {
2331 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2332 	struct ifnet *ifp = llt->llt_ifp;
2333 	struct llentry *lle;
2334 	char linkhdr[LLE_MAX_LINKHDR];
2335 	size_t linkhdrsize;
2336 	int lladdr_off;
2337 
2338 	KASSERT(l3addr->sa_family == AF_INET6,
2339 	    ("sin_family %d", l3addr->sa_family));
2340 
2341 	/*
2342 	 * A route that covers the given address must have
2343 	 * been installed 1st because we are doing a resolution,
2344 	 * verify this.
2345 	 */
2346 	if (!(flags & LLE_IFADDR) &&
2347 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2348 		return (NULL);
2349 
2350 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2351 	if (lle == NULL) {
2352 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2353 		return (NULL);
2354 	}
2355 	lle->la_flags = flags;
2356 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2357 		linkhdrsize = LLE_MAX_LINKHDR;
2358 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2359 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2360 			in6_lltable_free_entry(llt, lle);
2361 			return (NULL);
2362 		}
2363 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2364 		    lladdr_off);
2365 		lle->la_flags |= LLE_STATIC;
2366 	}
2367 
2368 	if ((lle->la_flags & LLE_STATIC) != 0)
2369 		lle->ln_state = ND6_LLINFO_REACHABLE;
2370 
2371 	return (lle);
2372 }
2373 
2374 static struct llentry *
2375 in6_lltable_lookup(struct lltable *llt, u_int flags,
2376 	const struct sockaddr *l3addr)
2377 {
2378 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2379 	int family = flags >> 16;
2380 	struct llentry *lle;
2381 
2382 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2383 	KASSERT(l3addr->sa_family == AF_INET6,
2384 	    ("sin_family %d", l3addr->sa_family));
2385 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2386 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2387 	    ("wrong lle request flags: %#x", flags));
2388 
2389 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2390 
2391 	if (__predict_false(family != AF_INET6))
2392 		lle = llentry_lookup_family(lle, family);
2393 
2394 	if (lle == NULL)
2395 		return (NULL);
2396 
2397 	if (flags & LLE_UNLOCKED)
2398 		return (lle);
2399 
2400 	if (flags & LLE_EXCLUSIVE)
2401 		LLE_WLOCK(lle);
2402 	else
2403 		LLE_RLOCK(lle);
2404 
2405 	/*
2406 	 * If the afdata lock is not held, the LLE may have been unlinked while
2407 	 * we were blocked on the LLE lock.  Check for this case.
2408 	 */
2409 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2410 		if (flags & LLE_EXCLUSIVE)
2411 			LLE_WUNLOCK(lle);
2412 		else
2413 			LLE_RUNLOCK(lle);
2414 		return (NULL);
2415 	}
2416 	return (lle);
2417 }
2418 
2419 static int
2420 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2421     struct sysctl_req *wr)
2422 {
2423 	struct ifnet *ifp = llt->llt_ifp;
2424 	/* XXX stack use */
2425 	struct {
2426 		struct rt_msghdr	rtm;
2427 		struct sockaddr_in6	sin6;
2428 		/*
2429 		 * ndp.c assumes that sdl is word aligned
2430 		 */
2431 #ifdef __LP64__
2432 		uint32_t		pad;
2433 #endif
2434 		struct sockaddr_dl	sdl;
2435 	} ndpc;
2436 	struct sockaddr_dl *sdl;
2437 	int error;
2438 
2439 	bzero(&ndpc, sizeof(ndpc));
2440 	/* skip deleted entries */
2441 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2442 		return (0);
2443 	/* Skip if jailed and not a valid IP of the prison. */
2444 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2445 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2446 		return (0);
2447 	/*
2448 	 * produce a msg made of:
2449 	 *  struct rt_msghdr;
2450 	 *  struct sockaddr_in6 (IPv6)
2451 	 *  struct sockaddr_dl;
2452 	 */
2453 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2454 	ndpc.rtm.rtm_version = RTM_VERSION;
2455 	ndpc.rtm.rtm_type = RTM_GET;
2456 	ndpc.rtm.rtm_flags = RTF_UP;
2457 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2458 	sa6_recoverscope(&ndpc.sin6);
2459 
2460 	/* publish */
2461 	if (lle->la_flags & LLE_PUB)
2462 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2463 
2464 	sdl = &ndpc.sdl;
2465 	sdl->sdl_family = AF_LINK;
2466 	sdl->sdl_len = sizeof(*sdl);
2467 	sdl->sdl_index = ifp->if_index;
2468 	sdl->sdl_type = ifp->if_type;
2469 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2470 		sdl->sdl_alen = ifp->if_addrlen;
2471 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2472 	} else {
2473 		sdl->sdl_alen = 0;
2474 		bzero(LLADDR(sdl), ifp->if_addrlen);
2475 	}
2476 	if (lle->la_expire != 0)
2477 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2478 		    lle->lle_remtime / hz + time_second - time_uptime;
2479 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2480 	if (lle->la_flags & LLE_STATIC)
2481 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2482 	if (lle->la_flags & LLE_IFADDR)
2483 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2484 	if (lle->ln_router != 0)
2485 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2486 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2487 	/* Store state in rmx_weight value */
2488 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2489 	ndpc.rtm.rtm_index = ifp->if_index;
2490 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2491 
2492 	return (error);
2493 }
2494 
2495 static void
2496 in6_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
2497 {
2498 	/* Join the solicited multicast group for dst. */
2499 	if ((lle->la_flags & LLE_PUB) == LLE_PUB)
2500 		in6_join_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2501 }
2502 
2503 static struct lltable *
2504 in6_lltattach(struct ifnet *ifp)
2505 {
2506 	struct lltable *llt;
2507 
2508 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2509 	llt->llt_af = AF_INET6;
2510 	llt->llt_ifp = ifp;
2511 
2512 	llt->llt_lookup = in6_lltable_lookup;
2513 	llt->llt_alloc_entry = in6_lltable_alloc;
2514 	llt->llt_delete_entry = in6_lltable_delete_entry;
2515 	llt->llt_dump_entry = in6_lltable_dump_entry;
2516 	llt->llt_hash = in6_lltable_hash;
2517 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2518 	llt->llt_free_entry = in6_lltable_free_entry;
2519 	llt->llt_match_prefix = in6_lltable_match_prefix;
2520 	llt->llt_mark_used = llentry_mark_used;
2521 	llt->llt_post_resolved = in6_lltable_post_resolved;
2522  	lltable_link(llt);
2523 
2524 	return (llt);
2525 }
2526 
2527 struct lltable *
2528 in6_lltable_get(struct ifnet *ifp)
2529 {
2530 	struct lltable *llt = NULL;
2531 
2532 	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2533 	if (afdata_ptr != NULL)
2534 		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2535 	return (llt);
2536 }
2537 
2538 void *
2539 in6_domifattach(struct ifnet *ifp)
2540 {
2541 	struct in6_ifextra *ext;
2542 
2543 	/* There are not IPv6-capable interfaces. */
2544 	switch (ifp->if_type) {
2545 	case IFT_PFLOG:
2546 	case IFT_PFSYNC:
2547 	case IFT_USB:
2548 		return (NULL);
2549 	}
2550 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2551 	bzero(ext, sizeof(*ext));
2552 
2553 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2554 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2555 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2556 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2557 
2558 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2559 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2560 	    M_WAITOK);
2561 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2562 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2563 
2564 	ext->nd_ifinfo = nd6_ifattach(ifp);
2565 	ext->scope6_id = scope6_ifattach(ifp);
2566 	ext->lltable = in6_lltattach(ifp);
2567 
2568 	ext->mld_ifinfo = mld_domifattach(ifp);
2569 
2570 	return ext;
2571 }
2572 
2573 int
2574 in6_domifmtu(struct ifnet *ifp)
2575 {
2576 	if (ifp->if_afdata[AF_INET6] == NULL)
2577 		return ifp->if_mtu;
2578 
2579 	return (IN6_LINKMTU(ifp));
2580 }
2581 
2582 void
2583 in6_domifdetach(struct ifnet *ifp, void *aux)
2584 {
2585 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2586 
2587 	mld_domifdetach(ifp);
2588 	scope6_ifdetach(ext->scope6_id);
2589 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2590 	lltable_free(ext->lltable);
2591 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2592 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2593 	free(ext->in6_ifstat, M_IFADDR);
2594 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2595 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2596 	free(ext->icmp6_ifstat, M_IFADDR);
2597 	free(ext, M_IFADDR);
2598 }
2599 
2600 /*
2601  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2602  * v4 mapped addr or v4 compat addr
2603  */
2604 void
2605 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2606 {
2607 
2608 	bzero(sin, sizeof(*sin));
2609 	sin->sin_len = sizeof(struct sockaddr_in);
2610 	sin->sin_family = AF_INET;
2611 	sin->sin_port = sin6->sin6_port;
2612 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2613 }
2614 
2615 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2616 void
2617 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2618 {
2619 	bzero(sin6, sizeof(*sin6));
2620 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2621 	sin6->sin6_family = AF_INET6;
2622 	sin6->sin6_port = sin->sin_port;
2623 	sin6->sin6_addr.s6_addr32[0] = 0;
2624 	sin6->sin6_addr.s6_addr32[1] = 0;
2625 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2626 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2627 }
2628 
2629 /* Convert sockaddr_in6 into sockaddr_in. */
2630 void
2631 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2632 {
2633 	struct sockaddr_in *sin_p;
2634 	struct sockaddr_in6 sin6;
2635 
2636 	/*
2637 	 * Save original sockaddr_in6 addr and convert it
2638 	 * to sockaddr_in.
2639 	 */
2640 	sin6 = *(struct sockaddr_in6 *)nam;
2641 	sin_p = (struct sockaddr_in *)nam;
2642 	in6_sin6_2_sin(sin_p, &sin6);
2643 }
2644 
2645 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2646 void
2647 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2648 {
2649 	struct sockaddr_in *sin_p;
2650 	struct sockaddr_in6 *sin6_p;
2651 
2652 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2653 	sin_p = (struct sockaddr_in *)*nam;
2654 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2655 	free(*nam, M_SONAME);
2656 	*nam = (struct sockaddr *)sin6_p;
2657 }
2658 
2659 /*
2660  * Join/leave the solicited multicast groups for proxy NDP entries.
2661  */
2662 static void
2663 in6_join_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2664 {
2665 	struct in6_multi *inm;
2666 	struct in6_addr mltaddr;
2667 	char ip6buf[INET6_ADDRSTRLEN];
2668 	int error;
2669 
2670 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2671 		return;	/* error logged in in6_solicited_node_maddr. */
2672 
2673 	error = in6_joingroup(ifp, &mltaddr, NULL, &inm, 0);
2674 	if (error != 0) {
2675 		nd6log((LOG_WARNING,
2676 		    "%s: in6_joingroup failed for %s on %s (errno=%d)\n",
2677 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp),
2678 		    error));
2679 	}
2680 }
2681 
2682 static void
2683 in6_leave_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2684 {
2685 	struct epoch_tracker et;
2686 	struct in6_multi *inm;
2687 	struct in6_addr mltaddr;
2688 	char ip6buf[INET6_ADDRSTRLEN];
2689 
2690 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2691 		return;	/* error logged in in6_solicited_node_maddr. */
2692 
2693 	NET_EPOCH_ENTER(et);
2694 	inm = in6m_lookup(ifp, &mltaddr);
2695 	NET_EPOCH_EXIT(et);
2696 	if (inm != NULL)
2697 		in6_leavegroup(inm, NULL);
2698 	else
2699 		nd6log((LOG_WARNING, "%s: in6m_lookup failed for %s on %s\n",
2700 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp)));
2701 }
2702 
2703 static bool
2704 in6_lle_match_pub(struct lltable *llt, struct llentry *lle, void *farg)
2705 {
2706 	return ((lle->la_flags & LLE_PUB) != 0);
2707 }
2708 
2709 void
2710 in6_purge_proxy_ndp(struct ifnet *ifp)
2711 {
2712 	struct lltable *llt;
2713 	bool need_purge;
2714 
2715 	if (ifp->if_afdata[AF_INET6] == NULL)
2716 		return;
2717 
2718 	llt = LLTABLE6(ifp);
2719 	IF_AFDATA_WLOCK(ifp);
2720 	need_purge = ((llt->llt_flags & LLT_ADDEDPROXY) != 0);
2721 	IF_AFDATA_WUNLOCK(ifp);
2722 
2723 	/*
2724 	 * Ever added proxy ndp entries, leave solicited node multicast
2725 	 * before deleting the llentry.
2726 	 */
2727 	if (need_purge)
2728 		lltable_delete_conditional(llt, in6_lle_match_pub, NULL);
2729 }
2730