xref: /f-stack/freebsd/netinet6/in6.c (revision 22ce4aff)
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/protosw.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/lock.h>
86 #include <sys/rmlock.h>
87 #include <sys/sysctl.h>
88 #include <sys/syslog.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/route/route_ctl.h>
95 #include <net/route/nhop.h>
96 #include <net/if_dl.h>
97 #include <net/vnet.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_var.h>
101 #include <net/if_llatbl.h>
102 #include <netinet/if_ether.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/ip.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/ip_carp.h>
107 
108 #include <netinet/ip6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/ip6_mroute.h>
113 #include <netinet6/in6_ifattach.h>
114 #include <netinet6/scope6_var.h>
115 #include <netinet6/in6_fib.h>
116 #include <netinet6/in6_pcb.h>
117 
118 /*
119  * struct in6_ifreq and struct ifreq must be type punnable for common members
120  * of ifr_ifru to allow accessors to be shared.
121  */
122 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
123     offsetof(struct ifreq, ifr_ifru),
124     "struct in6_ifreq and struct ifreq are not type punnable");
125 
126 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
127 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
128 
129 /*
130  * Definitions of some costant IP6 addresses.
131  */
132 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
133 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
134 const struct in6_addr in6addr_nodelocal_allnodes =
135 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
136 const struct in6_addr in6addr_linklocal_allnodes =
137 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
138 const struct in6_addr in6addr_linklocal_allrouters =
139 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
140 const struct in6_addr in6addr_linklocal_allv2routers =
141 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
142 
143 const struct in6_addr in6mask0 = IN6MASK0;
144 const struct in6_addr in6mask32 = IN6MASK32;
145 const struct in6_addr in6mask64 = IN6MASK64;
146 const struct in6_addr in6mask96 = IN6MASK96;
147 const struct in6_addr in6mask128 = IN6MASK128;
148 
149 const struct sockaddr_in6 sa6_any =
150 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
151 
152 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
153 	struct in6_aliasreq *, int);
154 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
155 
156 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr *, int);
158 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
159     struct in6_aliasreq *, int flags);
160 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
161     struct in6_ifaddr *, int, int);
162 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
163     struct in6_ifaddr *, int);
164 
165 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
166 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
167 
168 void
in6_newaddrmsg(struct in6_ifaddr * ia,int cmd)169 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
170 {
171 	struct rt_addrinfo info;
172 	struct ifaddr *ifa;
173 	struct sockaddr_dl gateway;
174 	int fibnum;
175 
176 	ifa = &ia->ia_ifa;
177 
178 	/*
179 	 * Prepare info data for the host route.
180 	 * This code mimics one from ifa_maintain_loopback_route().
181 	 */
182 	bzero(&info, sizeof(struct rt_addrinfo));
183 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
184 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
185 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
186 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
187 	if (cmd != RTM_DELETE)
188 		info.rti_ifp = V_loif;
189 
190 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
191 
192 	if (cmd == RTM_ADD) {
193 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
194 		rt_routemsg_info(cmd, &info, fibnum);
195 	} else if (cmd == RTM_DELETE) {
196 		rt_routemsg_info(cmd, &info, fibnum);
197 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
198 	}
199 }
200 
201 int
in6_mask2len(struct in6_addr * mask,u_char * lim0)202 in6_mask2len(struct in6_addr *mask, u_char *lim0)
203 {
204 	int x = 0, y;
205 	u_char *lim = lim0, *p;
206 
207 	/* ignore the scope_id part */
208 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
209 		lim = (u_char *)mask + sizeof(*mask);
210 	for (p = (u_char *)mask; p < lim; x++, p++) {
211 		if (*p != 0xff)
212 			break;
213 	}
214 	y = 0;
215 	if (p < lim) {
216 		for (y = 0; y < 8; y++) {
217 			if ((*p & (0x80 >> y)) == 0)
218 				break;
219 		}
220 	}
221 
222 	/*
223 	 * when the limit pointer is given, do a stricter check on the
224 	 * remaining bits.
225 	 */
226 	if (p < lim) {
227 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
228 			return (-1);
229 		for (p = p + 1; p < lim; p++)
230 			if (*p != 0)
231 				return (-1);
232 	}
233 
234 	return x * 8 + y;
235 }
236 
237 #ifdef COMPAT_FREEBSD32
238 struct in6_ndifreq32 {
239 	char ifname[IFNAMSIZ];
240 	uint32_t ifindex;
241 };
242 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
243 #endif
244 
245 int
in6_control(struct socket * so,u_long cmd,caddr_t data,struct ifnet * ifp,struct thread * td)246 in6_control(struct socket *so, u_long cmd, caddr_t data,
247     struct ifnet *ifp, struct thread *td)
248 {
249 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
250 	struct	in6_ifaddr *ia = NULL;
251 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
252 	struct sockaddr_in6 *sa6;
253 	int carp_attached = 0;
254 	int error;
255 	u_long ocmd = cmd;
256 
257 	/*
258 	 * Compat to make pre-10.x ifconfig(8) operable.
259 	 */
260 	if (cmd == OSIOCAIFADDR_IN6)
261 		cmd = SIOCAIFADDR_IN6;
262 
263 	switch (cmd) {
264 	case SIOCGETSGCNT_IN6:
265 	case SIOCGETMIFCNT_IN6:
266 		/*
267 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
268 		 * We cannot see how that would be needed, so do not adjust the
269 		 * KPI blindly; more likely should clean up the IPv4 variant.
270 		 */
271 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
272 	}
273 
274 	switch (cmd) {
275 	case SIOCAADDRCTL_POLICY:
276 	case SIOCDADDRCTL_POLICY:
277 		if (td != NULL) {
278 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
279 			if (error)
280 				return (error);
281 		}
282 		return (in6_src_ioctl(cmd, data));
283 	}
284 
285 	if (ifp == NULL)
286 		return (EOPNOTSUPP);
287 
288 	switch (cmd) {
289 	case SIOCSNDFLUSH_IN6:
290 	case SIOCSPFXFLUSH_IN6:
291 	case SIOCSRTRFLUSH_IN6:
292 	case SIOCSDEFIFACE_IN6:
293 	case SIOCSIFINFO_FLAGS:
294 	case SIOCSIFINFO_IN6:
295 		if (td != NULL) {
296 			error = priv_check(td, PRIV_NETINET_ND6);
297 			if (error)
298 				return (error);
299 		}
300 		/* FALLTHROUGH */
301 	case OSIOCGIFINFO_IN6:
302 	case SIOCGIFINFO_IN6:
303 	case SIOCGNBRINFO_IN6:
304 	case SIOCGDEFIFACE_IN6:
305 		return (nd6_ioctl(cmd, data, ifp));
306 
307 #ifdef COMPAT_FREEBSD32
308 	case SIOCGDEFIFACE32_IN6:
309 		{
310 			struct in6_ndifreq ndif;
311 			struct in6_ndifreq32 *ndif32;
312 
313 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
314 			    ifp);
315 			if (error)
316 				return (error);
317 			ndif32 = (struct in6_ndifreq32 *)data;
318 			ndif32->ifindex = ndif.ifindex;
319 			return (0);
320 		}
321 #endif
322 	}
323 
324 	switch (cmd) {
325 	case SIOCSIFPREFIX_IN6:
326 	case SIOCDIFPREFIX_IN6:
327 	case SIOCAIFPREFIX_IN6:
328 	case SIOCCIFPREFIX_IN6:
329 	case SIOCSGIFPREFIX_IN6:
330 	case SIOCGIFPREFIX_IN6:
331 		log(LOG_NOTICE,
332 		    "prefix ioctls are now invalidated. "
333 		    "please use ifconfig.\n");
334 		return (EOPNOTSUPP);
335 	}
336 
337 	switch (cmd) {
338 	case SIOCSSCOPE6:
339 		if (td != NULL) {
340 			error = priv_check(td, PRIV_NETINET_SCOPE6);
341 			if (error)
342 				return (error);
343 		}
344 		/* FALLTHROUGH */
345 	case SIOCGSCOPE6:
346 	case SIOCGSCOPE6DEF:
347 		return (scope6_ioctl(cmd, data, ifp));
348 	}
349 
350 	/*
351 	 * Find address for this interface, if it exists.
352 	 *
353 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
354 	 * only, and used the first interface address as the target of other
355 	 * operations (without checking ifra_addr).  This was because netinet
356 	 * code/API assumed at most 1 interface address per interface.
357 	 * Since IPv6 allows a node to assign multiple addresses
358 	 * on a single interface, we almost always look and check the
359 	 * presence of ifra_addr, and reject invalid ones here.
360 	 * It also decreases duplicated code among SIOC*_IN6 operations.
361 	 */
362 	switch (cmd) {
363 	case SIOCAIFADDR_IN6:
364 	case SIOCSIFPHYADDR_IN6:
365 		sa6 = &ifra->ifra_addr;
366 		break;
367 	case SIOCSIFADDR_IN6:
368 	case SIOCGIFADDR_IN6:
369 	case SIOCSIFDSTADDR_IN6:
370 	case SIOCSIFNETMASK_IN6:
371 	case SIOCGIFDSTADDR_IN6:
372 	case SIOCGIFNETMASK_IN6:
373 	case SIOCDIFADDR_IN6:
374 	case SIOCGIFPSRCADDR_IN6:
375 	case SIOCGIFPDSTADDR_IN6:
376 	case SIOCGIFAFLAG_IN6:
377 	case SIOCSNDFLUSH_IN6:
378 	case SIOCSPFXFLUSH_IN6:
379 	case SIOCSRTRFLUSH_IN6:
380 	case SIOCGIFALIFETIME_IN6:
381 	case SIOCGIFSTAT_IN6:
382 	case SIOCGIFSTAT_ICMP6:
383 		sa6 = &ifr->ifr_addr;
384 		break;
385 	case SIOCSIFADDR:
386 	case SIOCSIFBRDADDR:
387 	case SIOCSIFDSTADDR:
388 	case SIOCSIFNETMASK:
389 		/*
390 		 * Although we should pass any non-INET6 ioctl requests
391 		 * down to driver, we filter some legacy INET requests.
392 		 * Drivers trust SIOCSIFADDR et al to come from an already
393 		 * privileged layer, and do not perform any credentials
394 		 * checks or input validation.
395 		 */
396 		return (EINVAL);
397 	default:
398 		sa6 = NULL;
399 		break;
400 	}
401 	if (sa6 && sa6->sin6_family == AF_INET6) {
402 		if (sa6->sin6_scope_id != 0)
403 			error = sa6_embedscope(sa6, 0);
404 		else
405 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
406 		if (error != 0)
407 			return (error);
408 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
409 		    &sa6->sin6_addr)) != 0)
410 			return (error);
411 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
412 	} else
413 		ia = NULL;
414 
415 	switch (cmd) {
416 	case SIOCSIFADDR_IN6:
417 	case SIOCSIFDSTADDR_IN6:
418 	case SIOCSIFNETMASK_IN6:
419 		/*
420 		 * Since IPv6 allows a node to assign multiple addresses
421 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
422 		 */
423 		/* we decided to obsolete this command (20000704) */
424 		error = EINVAL;
425 		goto out;
426 
427 	case SIOCDIFADDR_IN6:
428 		/*
429 		 * for IPv4, we look for existing in_ifaddr here to allow
430 		 * "ifconfig if0 delete" to remove the first IPv4 address on
431 		 * the interface.  For IPv6, as the spec allows multiple
432 		 * interface address from the day one, we consider "remove the
433 		 * first one" semantics to be not preferable.
434 		 */
435 		if (ia == NULL) {
436 			error = EADDRNOTAVAIL;
437 			goto out;
438 		}
439 		/* FALLTHROUGH */
440 	case SIOCAIFADDR_IN6:
441 		/*
442 		 * We always require users to specify a valid IPv6 address for
443 		 * the corresponding operation.
444 		 */
445 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
446 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
447 			error = EAFNOSUPPORT;
448 			goto out;
449 		}
450 
451 		if (td != NULL) {
452 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
453 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
454 			if (error)
455 				goto out;
456 		}
457 		/* FALLTHROUGH */
458 	case SIOCGIFSTAT_IN6:
459 	case SIOCGIFSTAT_ICMP6:
460 		if (ifp->if_afdata[AF_INET6] == NULL) {
461 			error = EPFNOSUPPORT;
462 			goto out;
463 		}
464 		break;
465 
466 	case SIOCGIFADDR_IN6:
467 		/* This interface is basically deprecated. use SIOCGIFCONF. */
468 		/* FALLTHROUGH */
469 	case SIOCGIFAFLAG_IN6:
470 	case SIOCGIFNETMASK_IN6:
471 	case SIOCGIFDSTADDR_IN6:
472 	case SIOCGIFALIFETIME_IN6:
473 		/* must think again about its semantics */
474 		if (ia == NULL) {
475 			error = EADDRNOTAVAIL;
476 			goto out;
477 		}
478 		break;
479 	}
480 
481 	switch (cmd) {
482 	case SIOCGIFADDR_IN6:
483 		ifr->ifr_addr = ia->ia_addr;
484 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
485 			goto out;
486 		break;
487 
488 	case SIOCGIFDSTADDR_IN6:
489 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
490 			error = EINVAL;
491 			goto out;
492 		}
493 		ifr->ifr_dstaddr = ia->ia_dstaddr;
494 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
495 			goto out;
496 		break;
497 
498 	case SIOCGIFNETMASK_IN6:
499 		ifr->ifr_addr = ia->ia_prefixmask;
500 		break;
501 
502 	case SIOCGIFAFLAG_IN6:
503 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
504 		break;
505 
506 	case SIOCGIFSTAT_IN6:
507 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
508 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
509 		    &ifr->ifr_ifru.ifru_stat,
510 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
511 		break;
512 
513 	case SIOCGIFSTAT_ICMP6:
514 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
515 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
516 		    &ifr->ifr_ifru.ifru_icmp6stat,
517 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
518 		break;
519 
520 	case SIOCGIFALIFETIME_IN6:
521 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
522 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
523 			time_t maxexpire;
524 			struct in6_addrlifetime *retlt =
525 			    &ifr->ifr_ifru.ifru_lifetime;
526 
527 			/*
528 			 * XXX: adjust expiration time assuming time_t is
529 			 * signed.
530 			 */
531 			maxexpire = (-1) &
532 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
533 			if (ia->ia6_lifetime.ia6t_vltime <
534 			    maxexpire - ia->ia6_updatetime) {
535 				retlt->ia6t_expire = ia->ia6_updatetime +
536 				    ia->ia6_lifetime.ia6t_vltime;
537 			} else
538 				retlt->ia6t_expire = maxexpire;
539 		}
540 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
541 			time_t maxexpire;
542 			struct in6_addrlifetime *retlt =
543 			    &ifr->ifr_ifru.ifru_lifetime;
544 
545 			/*
546 			 * XXX: adjust expiration time assuming time_t is
547 			 * signed.
548 			 */
549 			maxexpire = (-1) &
550 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
551 			if (ia->ia6_lifetime.ia6t_pltime <
552 			    maxexpire - ia->ia6_updatetime) {
553 				retlt->ia6t_preferred = ia->ia6_updatetime +
554 				    ia->ia6_lifetime.ia6t_pltime;
555 			} else
556 				retlt->ia6t_preferred = maxexpire;
557 		}
558 		break;
559 
560 	case SIOCAIFADDR_IN6:
561 	{
562 		struct nd_prefixctl pr0;
563 		struct nd_prefix *pr;
564 
565 		/*
566 		 * first, make or update the interface address structure,
567 		 * and link it to the list.
568 		 */
569 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
570 			goto out;
571 		if (ia != NULL) {
572 			if (ia->ia_ifa.ifa_carp)
573 				(*carp_detach_p)(&ia->ia_ifa, true);
574 			ifa_free(&ia->ia_ifa);
575 		}
576 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
577 		    == NULL) {
578 			/*
579 			 * this can happen when the user specify the 0 valid
580 			 * lifetime.
581 			 */
582 			break;
583 		}
584 
585 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
586 			if (carp_attach_p != NULL)
587 				error = (*carp_attach_p)(&ia->ia_ifa,
588 				    ifra->ifra_vhid);
589 			else
590 				error = EPROTONOSUPPORT;
591 			if (error)
592 				goto out;
593 			else
594 				carp_attached = 1;
595 		}
596 
597 		/*
598 		 * then, make the prefix on-link on the interface.
599 		 * XXX: we'd rather create the prefix before the address, but
600 		 * we need at least one address to install the corresponding
601 		 * interface route, so we configure the address first.
602 		 */
603 
604 		/*
605 		 * convert mask to prefix length (prefixmask has already
606 		 * been validated in in6_update_ifa().
607 		 */
608 		bzero(&pr0, sizeof(pr0));
609 		pr0.ndpr_ifp = ifp;
610 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
611 		    NULL);
612 		if (pr0.ndpr_plen == 128) {
613 			/* we don't need to install a host route. */
614 			goto aifaddr_out;
615 		}
616 		pr0.ndpr_prefix = ifra->ifra_addr;
617 		/* apply the mask for safety. */
618 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
619 		    &ifra->ifra_prefixmask.sin6_addr);
620 
621 		/*
622 		 * XXX: since we don't have an API to set prefix (not address)
623 		 * lifetimes, we just use the same lifetimes as addresses.
624 		 * The (temporarily) installed lifetimes can be overridden by
625 		 * later advertised RAs (when accept_rtadv is non 0), which is
626 		 * an intended behavior.
627 		 */
628 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
629 		pr0.ndpr_raf_auto =
630 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
631 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
632 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
633 
634 		/* add the prefix if not yet. */
635 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
636 			/*
637 			 * nd6_prelist_add will install the corresponding
638 			 * interface route.
639 			 */
640 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
641 				if (carp_attached)
642 					(*carp_detach_p)(&ia->ia_ifa, false);
643 				goto out;
644 			}
645 		}
646 
647 		/* relate the address to the prefix */
648 		if (ia->ia6_ndpr == NULL) {
649 			ia->ia6_ndpr = pr;
650 			pr->ndpr_addrcnt++;
651 
652 			/*
653 			 * If this is the first autoconf address from the
654 			 * prefix, create a temporary address as well
655 			 * (when required).
656 			 */
657 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
658 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
659 				int e;
660 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
661 					log(LOG_NOTICE, "in6_control: failed "
662 					    "to create a temporary address, "
663 					    "errno=%d\n", e);
664 				}
665 			}
666 		}
667 		nd6_prefix_rele(pr);
668 
669 		/*
670 		 * this might affect the status of autoconfigured addresses,
671 		 * that is, this address might make other addresses detached.
672 		 */
673 		pfxlist_onlink_check();
674 
675 aifaddr_out:
676 		/*
677 		 * Try to clear the flag when a new IPv6 address is added
678 		 * onto an IFDISABLED interface and it succeeds.
679 		 */
680 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
681 			struct in6_ndireq nd;
682 
683 			memset(&nd, 0, sizeof(nd));
684 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
685 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
686 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
687 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
688 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
689 				    "failed.");
690 			/*
691 			 * Ignore failure of clearing the flag intentionally.
692 			 * The failure means address duplication was detected.
693 			 */
694 		}
695 		break;
696 	}
697 
698 	case SIOCDIFADDR_IN6:
699 		in6_purgeifaddr(ia);
700 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
701 		    IFADDR_EVENT_DEL);
702 		break;
703 
704 	default:
705 		if (ifp->if_ioctl == NULL) {
706 			error = EOPNOTSUPP;
707 			goto out;
708 		}
709 		error = (*ifp->if_ioctl)(ifp, cmd, data);
710 		goto out;
711 	}
712 
713 	error = 0;
714 out:
715 	if (ia != NULL)
716 		ifa_free(&ia->ia_ifa);
717 	return (error);
718 }
719 
720 static struct in6_multi_mship *
in6_joingroup_legacy(struct ifnet * ifp,const struct in6_addr * mcaddr,int * errorp,int delay)721 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
722     int *errorp, int delay)
723 {
724 	struct in6_multi_mship *imm;
725 	int error;
726 
727 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
728 	if (imm == NULL) {
729 		*errorp = ENOBUFS;
730 		return (NULL);
731 	}
732 
733 	delay = (delay * PR_FASTHZ) / hz;
734 
735 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
736 	if (error) {
737 		*errorp = error;
738 		free(imm, M_IP6MADDR);
739 		return (NULL);
740 	}
741 
742 	return (imm);
743 }
744 /*
745  * Join necessary multicast groups.  Factored out from in6_update_ifa().
746  * This entire work should only be done once, for the default FIB.
747  */
748 static int
in6_update_ifa_join_mc(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags,struct in6_multi ** in6m_sol)749 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
750     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
751 {
752 	char ip6buf[INET6_ADDRSTRLEN];
753 	struct in6_addr mltaddr;
754 	struct in6_multi_mship *imm;
755 	int delay, error;
756 
757 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
758 
759 	/* Join solicited multicast addr for new host id. */
760 	bzero(&mltaddr, sizeof(struct in6_addr));
761 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
762 	mltaddr.s6_addr32[2] = htonl(1);
763 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
764 	mltaddr.s6_addr8[12] = 0xff;
765 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
766 		/* XXX: should not happen */
767 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
768 		goto cleanup;
769 	}
770 	delay = error = 0;
771 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
772 		/*
773 		 * We need a random delay for DAD on the address being
774 		 * configured.  It also means delaying transmission of the
775 		 * corresponding MLD report to avoid report collision.
776 		 * [RFC 4861, Section 6.3.7]
777 		 */
778 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
779 	}
780 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
781 	if (imm == NULL) {
782 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
783 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
784 		    if_name(ifp), error));
785 		goto cleanup;
786 	}
787 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
788 	*in6m_sol = imm->i6mm_maddr;
789 
790 	/*
791 	 * Join link-local all-nodes address.
792 	 */
793 	mltaddr = in6addr_linklocal_allnodes;
794 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
795 		goto cleanup; /* XXX: should not fail */
796 
797 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
798 	if (imm == NULL) {
799 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
800 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
801 		    if_name(ifp), error));
802 		goto cleanup;
803 	}
804 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
805 
806 	/*
807 	 * Join node information group address.
808 	 */
809 	delay = 0;
810 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
811 		/*
812 		 * The spec does not say anything about delay for this group,
813 		 * but the same logic should apply.
814 		 */
815 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
816 	}
817 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
818 		/* XXX jinmei */
819 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
820 		if (imm == NULL)
821 			nd6log((LOG_WARNING,
822 			    "%s: in6_joingroup failed for %s on %s "
823 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
824 			    &mltaddr), if_name(ifp), error));
825 			/* XXX not very fatal, go on... */
826 		else
827 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
828 	}
829 	if (V_icmp6_nodeinfo_oldmcprefix &&
830 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
831 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
832 		if (imm == NULL)
833 			nd6log((LOG_WARNING,
834 			    "%s: in6_joingroup failed for %s on %s "
835 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
836 			    &mltaddr), if_name(ifp), error));
837 			/* XXX not very fatal, go on... */
838 		else
839 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
840 	}
841 
842 	/*
843 	 * Join interface-local all-nodes address.
844 	 * (ff01::1%ifN, and ff01::%ifN/32)
845 	 */
846 	mltaddr = in6addr_nodelocal_allnodes;
847 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
848 		goto cleanup; /* XXX: should not fail */
849 
850 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
851 	if (imm == NULL) {
852 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
853 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
854 		    &mltaddr), if_name(ifp), error));
855 		goto cleanup;
856 	}
857 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
858 
859 cleanup:
860 	return (error);
861 }
862 
863 /*
864  * Update parameters of an IPv6 interface address.
865  * If necessary, a new entry is created and linked into address chains.
866  * This function is separated from in6_control().
867  */
868 int
in6_update_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)869 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
870     struct in6_ifaddr *ia, int flags)
871 {
872 	int error, hostIsNew = 0;
873 
874 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
875 		return (error);
876 
877 	if (ia == NULL) {
878 		hostIsNew = 1;
879 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
880 			return (ENOBUFS);
881 	}
882 
883 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
884 	if (error != 0) {
885 		if (hostIsNew != 0) {
886 			in6_unlink_ifa(ia, ifp);
887 			ifa_free(&ia->ia_ifa);
888 		}
889 		return (error);
890 	}
891 
892 	if (hostIsNew)
893 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
894 
895 	return (error);
896 }
897 
898 /*
899  * Fill in basic IPv6 address request info.
900  */
901 void
in6_prepare_ifra(struct in6_aliasreq * ifra,const struct in6_addr * addr,const struct in6_addr * mask)902 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
903     const struct in6_addr *mask)
904 {
905 
906 	memset(ifra, 0, sizeof(struct in6_aliasreq));
907 
908 	ifra->ifra_addr.sin6_family = AF_INET6;
909 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
910 	if (addr != NULL)
911 		ifra->ifra_addr.sin6_addr = *addr;
912 
913 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
914 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
915 	if (mask != NULL)
916 		ifra->ifra_prefixmask.sin6_addr = *mask;
917 }
918 
919 static int
in6_validate_ifra(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)920 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
921     struct in6_ifaddr *ia, int flags)
922 {
923 	int plen = -1;
924 	struct sockaddr_in6 dst6;
925 	struct in6_addrlifetime *lt;
926 	char ip6buf[INET6_ADDRSTRLEN];
927 
928 	/* Validate parameters */
929 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
930 		return (EINVAL);
931 
932 	/*
933 	 * The destination address for a p2p link must have a family
934 	 * of AF_UNSPEC or AF_INET6.
935 	 */
936 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
937 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
938 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
939 		return (EAFNOSUPPORT);
940 
941 	/*
942 	 * Validate address
943 	 */
944 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
945 	    ifra->ifra_addr.sin6_family != AF_INET6)
946 		return (EINVAL);
947 
948 	/*
949 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
950 	 * does not carry fields other than sin6_len.
951 	 */
952 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
953 		return (EINVAL);
954 	/*
955 	 * Because the IPv6 address architecture is classless, we require
956 	 * users to specify a (non 0) prefix length (mask) for a new address.
957 	 * We also require the prefix (when specified) mask is valid, and thus
958 	 * reject a non-consecutive mask.
959 	 */
960 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
961 		return (EINVAL);
962 	if (ifra->ifra_prefixmask.sin6_len != 0) {
963 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
964 		    (u_char *)&ifra->ifra_prefixmask +
965 		    ifra->ifra_prefixmask.sin6_len);
966 		if (plen <= 0)
967 			return (EINVAL);
968 	} else {
969 		/*
970 		 * In this case, ia must not be NULL.  We just use its prefix
971 		 * length.
972 		 */
973 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
974 	}
975 	/*
976 	 * If the destination address on a p2p interface is specified,
977 	 * and the address is a scoped one, validate/set the scope
978 	 * zone identifier.
979 	 */
980 	dst6 = ifra->ifra_dstaddr;
981 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
982 	    (dst6.sin6_family == AF_INET6)) {
983 		struct in6_addr in6_tmp;
984 		u_int32_t zoneid;
985 
986 		in6_tmp = dst6.sin6_addr;
987 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
988 			return (EINVAL); /* XXX: should be impossible */
989 
990 		if (dst6.sin6_scope_id != 0) {
991 			if (dst6.sin6_scope_id != zoneid)
992 				return (EINVAL);
993 		} else		/* user omit to specify the ID. */
994 			dst6.sin6_scope_id = zoneid;
995 
996 		/* convert into the internal form */
997 		if (sa6_embedscope(&dst6, 0))
998 			return (EINVAL); /* XXX: should be impossible */
999 	}
1000 	/* Modify original ifra_dstaddr to reflect changes */
1001 	ifra->ifra_dstaddr = dst6;
1002 
1003 	/*
1004 	 * The destination address can be specified only for a p2p or a
1005 	 * loopback interface.  If specified, the corresponding prefix length
1006 	 * must be 128.
1007 	 */
1008 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1009 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1010 			/* XXX: noisy message */
1011 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1012 			    "be specified for a p2p or a loopback IF only\n"));
1013 			return (EINVAL);
1014 		}
1015 		if (plen != 128) {
1016 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1017 			    "be 128 when dstaddr is specified\n"));
1018 			return (EINVAL);
1019 		}
1020 	}
1021 	/* lifetime consistency check */
1022 	lt = &ifra->ifra_lifetime;
1023 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1024 		return (EINVAL);
1025 	if (lt->ia6t_vltime == 0) {
1026 		/*
1027 		 * the following log might be noisy, but this is a typical
1028 		 * configuration mistake or a tool's bug.
1029 		 */
1030 		nd6log((LOG_INFO,
1031 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1032 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1033 
1034 		if (ia == NULL)
1035 			return (0); /* there's nothing to do */
1036 	}
1037 
1038 	/* Check prefix mask */
1039 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1040 		/*
1041 		 * We prohibit changing the prefix length of an existing
1042 		 * address, because
1043 		 * + such an operation should be rare in IPv6, and
1044 		 * + the operation would confuse prefix management.
1045 		 */
1046 		if (ia->ia_prefixmask.sin6_len != 0 &&
1047 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1048 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1049 			    "of an existing %s address should not be changed\n",
1050 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1051 
1052 			return (EINVAL);
1053 		}
1054 	}
1055 
1056 	return (0);
1057 }
1058 
1059 /*
1060  * Allocate a new ifaddr and link it into chains.
1061  */
1062 static struct in6_ifaddr *
in6_alloc_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,int flags)1063 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1064 {
1065 	struct in6_ifaddr *ia;
1066 
1067 	/*
1068 	 * When in6_alloc_ifa() is called in a process of a received
1069 	 * RA, it is called under an interrupt context.  So, we should
1070 	 * call malloc with M_NOWAIT.
1071 	 */
1072 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1073 	if (ia == NULL)
1074 		return (NULL);
1075 	LIST_INIT(&ia->ia6_memberships);
1076 	/* Initialize the address and masks, and put time stamp */
1077 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1078 	ia->ia_addr.sin6_family = AF_INET6;
1079 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1080 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1081 	ia->ia_addr = ifra->ifra_addr;
1082 	ia->ia6_createtime = time_uptime;
1083 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1084 		/*
1085 		 * Some functions expect that ifa_dstaddr is not
1086 		 * NULL for p2p interfaces.
1087 		 */
1088 		ia->ia_ifa.ifa_dstaddr =
1089 		    (struct sockaddr *)&ia->ia_dstaddr;
1090 	} else {
1091 		ia->ia_ifa.ifa_dstaddr = NULL;
1092 	}
1093 
1094 	/* set prefix mask if any */
1095 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1096 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1097 		ia->ia_prefixmask.sin6_family = AF_INET6;
1098 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1099 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1100 	}
1101 
1102 	ia->ia_ifp = ifp;
1103 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1104 	IF_ADDR_WLOCK(ifp);
1105 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1106 	IF_ADDR_WUNLOCK(ifp);
1107 
1108 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1109 	IN6_IFADDR_WLOCK();
1110 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1111 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1112 	IN6_IFADDR_WUNLOCK();
1113 
1114 	return (ia);
1115 }
1116 
1117 /*
1118  * Update/configure interface address parameters:
1119  *
1120  * 1) Update lifetime
1121  * 2) Update interface metric ad flags
1122  * 3) Notify other subsystems
1123  */
1124 static int
in6_update_ifa_internal(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int hostIsNew,int flags)1125 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1126     struct in6_ifaddr *ia, int hostIsNew, int flags)
1127 {
1128 	int error;
1129 
1130 	/* update timestamp */
1131 	ia->ia6_updatetime = time_uptime;
1132 
1133 	/*
1134 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1135 	 * to see if the address is deprecated or invalidated, but initialize
1136 	 * these members for applications.
1137 	 */
1138 	ia->ia6_lifetime = ifra->ifra_lifetime;
1139 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1140 		ia->ia6_lifetime.ia6t_expire =
1141 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1142 	} else
1143 		ia->ia6_lifetime.ia6t_expire = 0;
1144 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1145 		ia->ia6_lifetime.ia6t_preferred =
1146 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1147 	} else
1148 		ia->ia6_lifetime.ia6t_preferred = 0;
1149 
1150 	/*
1151 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1152 	 * userland, make it deprecated.
1153 	 */
1154 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1155 		ia->ia6_lifetime.ia6t_pltime = 0;
1156 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1157 	}
1158 
1159 	/*
1160 	 * configure address flags.
1161 	 */
1162 	ia->ia6_flags = ifra->ifra_flags;
1163 
1164 	/*
1165 	 * Make the address tentative before joining multicast addresses,
1166 	 * so that corresponding MLD responses would not have a tentative
1167 	 * source address.
1168 	 */
1169 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1170 
1171 	/*
1172 	 * DAD should be performed for an new address or addresses on
1173 	 * an interface with ND6_IFF_IFDISABLED.
1174 	 */
1175 	if (in6if_do_dad(ifp) &&
1176 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1177 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1178 
1179 	/* notify other subsystems */
1180 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1181 
1182 	return (error);
1183 }
1184 
1185 /*
1186  * Do link-level ifa job:
1187  * 1) Add lle entry for added address
1188  * 2) Notifies routing socket users about new address
1189  * 3) join appropriate multicast group
1190  * 4) start DAD if enabled
1191  */
1192 static int
in6_broadcast_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)1193 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1194     struct in6_ifaddr *ia, int flags)
1195 {
1196 	struct in6_multi *in6m_sol;
1197 	int error = 0;
1198 
1199 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1200 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1201 		in6_purgeaddr(&ia->ia_ifa);
1202 		ifa_free(&ia->ia_ifa);
1203 		return (error);
1204 	}
1205 
1206 	/* Join necessary multicast groups. */
1207 	in6m_sol = NULL;
1208 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1209 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1210 		if (error != 0) {
1211 			in6_purgeaddr(&ia->ia_ifa);
1212 			ifa_free(&ia->ia_ifa);
1213 			return (error);
1214 		}
1215 	}
1216 
1217 	/* Perform DAD, if the address is TENTATIVE. */
1218 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1219 		int delay, mindelay, maxdelay;
1220 
1221 		delay = 0;
1222 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1223 			/*
1224 			 * We need to impose a delay before sending an NS
1225 			 * for DAD.  Check if we also needed a delay for the
1226 			 * corresponding MLD message.  If we did, the delay
1227 			 * should be larger than the MLD delay (this could be
1228 			 * relaxed a bit, but this simple logic is at least
1229 			 * safe).
1230 			 * XXX: Break data hiding guidelines and look at
1231 			 * state for the solicited multicast group.
1232 			 */
1233 			mindelay = 0;
1234 			if (in6m_sol != NULL &&
1235 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1236 				mindelay = in6m_sol->in6m_timer;
1237 			}
1238 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1239 			if (maxdelay - mindelay == 0)
1240 				delay = 0;
1241 			else {
1242 				delay =
1243 				    (arc4random() % (maxdelay - mindelay)) +
1244 				    mindelay;
1245 			}
1246 		}
1247 		nd6_dad_start((struct ifaddr *)ia, delay);
1248 	}
1249 
1250 	in6_newaddrmsg(ia, RTM_ADD);
1251 	ifa_free(&ia->ia_ifa);
1252 	return (error);
1253 }
1254 
1255 /*
1256  * Adds or deletes interface route for p2p ifa.
1257  * Returns 0 on success or errno.
1258  */
1259 static int
in6_handle_dstaddr_rtrequest(int cmd,struct in6_ifaddr * ia)1260 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1261 {
1262 	struct epoch_tracker et;
1263 	struct ifaddr *ifa = &ia->ia_ifa;
1264 	int error;
1265 
1266 	/* Prepare gateway */
1267 	struct sockaddr_dl_short sdl = {
1268 		.sdl_family = AF_LINK,
1269 		.sdl_len = sizeof(struct sockaddr_dl_short),
1270 		.sdl_type = ifa->ifa_ifp->if_type,
1271 		.sdl_index = ifa->ifa_ifp->if_index,
1272 	};
1273 
1274 	struct sockaddr_in6 dst = {
1275 		.sin6_family = AF_INET6,
1276 		.sin6_len = sizeof(struct sockaddr_in6),
1277 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1278 	};
1279 
1280 	struct rt_addrinfo info = {
1281 		.rti_ifa = ifa,
1282 		.rti_flags = RTF_PINNED | RTF_HOST,
1283 		.rti_info = {
1284 			[RTAX_DST] = (struct sockaddr *)&dst,
1285 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1286 		},
1287 	};
1288 	/* Don't set additional per-gw filters on removal */
1289 
1290 	NET_EPOCH_ENTER(et);
1291 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1292 	NET_EPOCH_EXIT(et);
1293 
1294 	return (error);
1295 }
1296 
1297 static bool
ifa_is_p2p(struct in6_ifaddr * ia)1298 ifa_is_p2p(struct in6_ifaddr *ia)
1299 {
1300 	int plen;
1301 
1302 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1303 
1304 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1305 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1306 		return (true);
1307 
1308 	return (false);
1309 }
1310 
1311 void
in6_purgeaddr(struct ifaddr * ifa)1312 in6_purgeaddr(struct ifaddr *ifa)
1313 {
1314 	struct ifnet *ifp = ifa->ifa_ifp;
1315 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1316 	struct in6_multi_mship *imm;
1317 	int error;
1318 
1319 	if (ifa->ifa_carp)
1320 		(*carp_detach_p)(ifa, false);
1321 
1322 	/*
1323 	 * Remove the loopback route to the interface address.
1324 	 * The check for the current setting of "nd6_useloopback"
1325 	 * is not needed.
1326 	 */
1327 	if (ia->ia_flags & IFA_RTSELF) {
1328 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1329 		    (struct sockaddr *)&ia->ia_addr);
1330 		if (error == 0)
1331 			ia->ia_flags &= ~IFA_RTSELF;
1332 	}
1333 
1334 	/* stop DAD processing */
1335 	nd6_dad_stop(ifa);
1336 
1337 	/* Leave multicast groups. */
1338 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1339 		LIST_REMOVE(imm, i6mm_chain);
1340 		if (imm->i6mm_maddr != NULL)
1341 			in6_leavegroup(imm->i6mm_maddr, NULL);
1342 		free(imm, M_IP6MADDR);
1343 	}
1344 	/* Check if we need to remove p2p route */
1345 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1346 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1347 		if (error != 0)
1348 			log(LOG_INFO, "%s: err=%d, destination address delete "
1349 			    "failed\n", __func__, error);
1350 		ia->ia_flags &= ~IFA_ROUTE;
1351 	}
1352 
1353 	in6_newaddrmsg(ia, RTM_DELETE);
1354 	in6_unlink_ifa(ia, ifp);
1355 }
1356 
1357 /*
1358  * Removes @ia from the corresponding interfaces and unlinks corresponding
1359  *  prefix if no addresses are using it anymore.
1360  */
1361 void
in6_purgeifaddr(struct in6_ifaddr * ia)1362 in6_purgeifaddr(struct in6_ifaddr *ia)
1363 {
1364 	struct nd_prefix *pr;
1365 
1366 	/*
1367 	 * If the address being deleted is the only one that owns
1368 	 * the corresponding prefix, expire the prefix as well.
1369 	 * XXX: theoretically, we don't have to worry about such
1370 	 * relationship, since we separate the address management
1371 	 * and the prefix management.  We do this, however, to provide
1372 	 * as much backward compatibility as possible in terms of
1373 	 * the ioctl operation.
1374 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1375 	 */
1376 	pr = ia->ia6_ndpr;
1377 	in6_purgeaddr(&ia->ia_ifa);
1378 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1379 		ND6_WLOCK();
1380 		nd6_prefix_unlink(pr, NULL);
1381 		ND6_WUNLOCK();
1382 		nd6_prefix_del(pr);
1383 	}
1384 }
1385 
1386 
1387 static void
in6_unlink_ifa(struct in6_ifaddr * ia,struct ifnet * ifp)1388 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1389 {
1390 	char ip6buf[INET6_ADDRSTRLEN];
1391 	int remove_lle;
1392 
1393 	IF_ADDR_WLOCK(ifp);
1394 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1395 	IF_ADDR_WUNLOCK(ifp);
1396 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1397 
1398 	/*
1399 	 * Defer the release of what might be the last reference to the
1400 	 * in6_ifaddr so that it can't be freed before the remainder of the
1401 	 * cleanup.
1402 	 */
1403 	IN6_IFADDR_WLOCK();
1404 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1405 	CK_LIST_REMOVE(ia, ia6_hash);
1406 	IN6_IFADDR_WUNLOCK();
1407 
1408 	/*
1409 	 * Release the reference to the base prefix.  There should be a
1410 	 * positive reference.
1411 	 */
1412 	remove_lle = 0;
1413 	if (ia->ia6_ndpr == NULL) {
1414 		nd6log((LOG_NOTICE,
1415 		    "in6_unlink_ifa: autoconf'ed address "
1416 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1417 	} else {
1418 		ia->ia6_ndpr->ndpr_addrcnt--;
1419 		/* Do not delete lles within prefix if refcont != 0 */
1420 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1421 			remove_lle = 1;
1422 		ia->ia6_ndpr = NULL;
1423 	}
1424 
1425 	nd6_rem_ifa_lle(ia, remove_lle);
1426 
1427 	/*
1428 	 * Also, if the address being removed is autoconf'ed, call
1429 	 * pfxlist_onlink_check() since the release might affect the status of
1430 	 * other (detached) addresses.
1431 	 */
1432 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1433 		pfxlist_onlink_check();
1434 	}
1435 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1436 }
1437 
1438 /*
1439  * Notifies other subsystems about address change/arrival:
1440  * 1) Notifies device handler on the first IPv6 address assignment
1441  * 2) Handle routing table changes for P2P links and route
1442  * 3) Handle routing table changes for address host route
1443  */
1444 static int
in6_notify_ifa(struct ifnet * ifp,struct in6_ifaddr * ia,struct in6_aliasreq * ifra,int hostIsNew)1445 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1446     struct in6_aliasreq *ifra, int hostIsNew)
1447 {
1448 	int	error = 0, ifacount = 0;
1449 	struct ifaddr *ifa;
1450 	struct sockaddr_in6 *pdst;
1451 	char ip6buf[INET6_ADDRSTRLEN];
1452 
1453 	/*
1454 	 * Give the interface a chance to initialize
1455 	 * if this is its first address,
1456 	 */
1457 	if (hostIsNew != 0) {
1458 		struct epoch_tracker et;
1459 
1460 		NET_EPOCH_ENTER(et);
1461 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1462 			if (ifa->ifa_addr->sa_family != AF_INET6)
1463 				continue;
1464 			ifacount++;
1465 		}
1466 		NET_EPOCH_EXIT(et);
1467 	}
1468 
1469 	if (ifacount <= 1 && ifp->if_ioctl) {
1470 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1471 		if (error)
1472 			goto done;
1473 	}
1474 
1475 	/*
1476 	 * If a new destination address is specified, scrub the old one and
1477 	 * install the new destination.  Note that the interface must be
1478 	 * p2p or loopback.
1479 	 */
1480 	pdst = &ifra->ifra_dstaddr;
1481 	if (pdst->sin6_family == AF_INET6 &&
1482 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1483 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1484 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1485 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1486 			    "remove a route to the old destination: %s\n",
1487 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1488 			/* proceed anyway... */
1489 		} else
1490 			ia->ia_flags &= ~IFA_ROUTE;
1491 		ia->ia_dstaddr = *pdst;
1492 	}
1493 
1494 	/*
1495 	 * If a new destination address is specified for a point-to-point
1496 	 * interface, install a route to the destination as an interface
1497 	 * direct route.
1498 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1499 	 * interface that share the same destination.
1500 	 */
1501 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1502 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1503 		if (error)
1504 			goto done;
1505 		ia->ia_flags |= IFA_ROUTE;
1506 	}
1507 
1508 	/*
1509 	 * add a loopback route to self if not exists
1510 	 */
1511 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1512 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1513 		    (struct sockaddr *)&ia->ia_addr);
1514 		if (error == 0)
1515 			ia->ia_flags |= IFA_RTSELF;
1516 	}
1517 done:
1518 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1519 	    "Invoking IPv6 network device address event may sleep");
1520 
1521 	ifa_ref(&ia->ia_ifa);
1522 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1523 	    IFADDR_EVENT_ADD);
1524 	ifa_free(&ia->ia_ifa);
1525 
1526 	return (error);
1527 }
1528 
1529 /*
1530  * Find an IPv6 interface link-local address specific to an interface.
1531  * ifaddr is returned referenced.
1532  */
1533 struct in6_ifaddr *
in6ifa_ifpforlinklocal(struct ifnet * ifp,int ignoreflags)1534 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1535 {
1536 	struct ifaddr *ifa;
1537 
1538 	NET_EPOCH_ASSERT();
1539 
1540 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1541 		if (ifa->ifa_addr->sa_family != AF_INET6)
1542 			continue;
1543 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1544 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1545 			    ignoreflags) != 0)
1546 				continue;
1547 			ifa_ref(ifa);
1548 			break;
1549 		}
1550 	}
1551 
1552 	return ((struct in6_ifaddr *)ifa);
1553 }
1554 
1555 /*
1556  * find the interface address corresponding to a given IPv6 address.
1557  * ifaddr is returned referenced.
1558  */
1559 struct in6_ifaddr *
in6ifa_ifwithaddr(const struct in6_addr * addr,uint32_t zoneid)1560 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1561 {
1562 	struct rm_priotracker in6_ifa_tracker;
1563 	struct in6_ifaddr *ia;
1564 
1565 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1566 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1567 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1568 			if (zoneid != 0 &&
1569 			    zoneid != ia->ia_addr.sin6_scope_id)
1570 				continue;
1571 			ifa_ref(&ia->ia_ifa);
1572 			break;
1573 		}
1574 	}
1575 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1576 	return (ia);
1577 }
1578 
1579 /*
1580  * find the internet address corresponding to a given interface and address.
1581  * ifaddr is returned referenced.
1582  */
1583 struct in6_ifaddr *
in6ifa_ifpwithaddr(struct ifnet * ifp,const struct in6_addr * addr)1584 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1585 {
1586 	struct epoch_tracker et;
1587 	struct ifaddr *ifa;
1588 
1589 	NET_EPOCH_ENTER(et);
1590 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1591 		if (ifa->ifa_addr->sa_family != AF_INET6)
1592 			continue;
1593 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1594 			ifa_ref(ifa);
1595 			break;
1596 		}
1597 	}
1598 	NET_EPOCH_EXIT(et);
1599 
1600 	return ((struct in6_ifaddr *)ifa);
1601 }
1602 
1603 /*
1604  * Find a link-local scoped address on ifp and return it if any.
1605  */
1606 struct in6_ifaddr *
in6ifa_llaonifp(struct ifnet * ifp)1607 in6ifa_llaonifp(struct ifnet *ifp)
1608 {
1609 	struct epoch_tracker et;
1610 	struct sockaddr_in6 *sin6;
1611 	struct ifaddr *ifa;
1612 
1613 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1614 		return (NULL);
1615 	NET_EPOCH_ENTER(et);
1616 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1617 		if (ifa->ifa_addr->sa_family != AF_INET6)
1618 			continue;
1619 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1620 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1621 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1622 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1623 			break;
1624 	}
1625 	NET_EPOCH_EXIT(et);
1626 
1627 	return ((struct in6_ifaddr *)ifa);
1628 }
1629 
1630 /*
1631  * Convert IP6 address to printable (loggable) representation. Caller
1632  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1633  */
1634 static char digits[] = "0123456789abcdef";
1635 char *
ip6_sprintf(char * ip6buf,const struct in6_addr * addr)1636 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1637 {
1638 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1639 	char *cp;
1640 	const u_int16_t *a = (const u_int16_t *)addr;
1641 	const u_int8_t *d;
1642 	int dcolon = 0, zero = 0;
1643 
1644 	cp = ip6buf;
1645 
1646 	for (i = 0; i < 8; i++) {
1647 		if (*(a + i) == 0) {
1648 			cnt++;
1649 			if (cnt == 1)
1650 				idx = i;
1651 		}
1652 		else if (maxcnt < cnt) {
1653 			maxcnt = cnt;
1654 			index = idx;
1655 			cnt = 0;
1656 		}
1657 	}
1658 	if (maxcnt < cnt) {
1659 		maxcnt = cnt;
1660 		index = idx;
1661 	}
1662 
1663 	for (i = 0; i < 8; i++) {
1664 		if (dcolon == 1) {
1665 			if (*a == 0) {
1666 				if (i == 7)
1667 					*cp++ = ':';
1668 				a++;
1669 				continue;
1670 			} else
1671 				dcolon = 2;
1672 		}
1673 		if (*a == 0) {
1674 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1675 				if (i == 0)
1676 					*cp++ = ':';
1677 				*cp++ = ':';
1678 				dcolon = 1;
1679 			} else {
1680 				*cp++ = '0';
1681 				*cp++ = ':';
1682 			}
1683 			a++;
1684 			continue;
1685 		}
1686 		d = (const u_char *)a;
1687 		/* Try to eliminate leading zeros in printout like in :0001. */
1688 		zero = 1;
1689 		*cp = digits[*d >> 4];
1690 		if (*cp != '0') {
1691 			zero = 0;
1692 			cp++;
1693 		}
1694 		*cp = digits[*d++ & 0xf];
1695 		if (zero == 0 || (*cp != '0')) {
1696 			zero = 0;
1697 			cp++;
1698 		}
1699 		*cp = digits[*d >> 4];
1700 		if (zero == 0 || (*cp != '0')) {
1701 			zero = 0;
1702 			cp++;
1703 		}
1704 		*cp++ = digits[*d & 0xf];
1705 		*cp++ = ':';
1706 		a++;
1707 	}
1708 	*--cp = '\0';
1709 	return (ip6buf);
1710 }
1711 
1712 int
in6_localaddr(struct in6_addr * in6)1713 in6_localaddr(struct in6_addr *in6)
1714 {
1715 	struct rm_priotracker in6_ifa_tracker;
1716 	struct in6_ifaddr *ia;
1717 
1718 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1719 		return 1;
1720 
1721 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1722 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1723 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1724 		    &ia->ia_prefixmask.sin6_addr)) {
1725 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1726 			return 1;
1727 		}
1728 	}
1729 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1730 
1731 	return (0);
1732 }
1733 
1734 /*
1735  * Return 1 if an internet address is for the local host and configured
1736  * on one of its interfaces.
1737  */
1738 int
in6_localip(struct in6_addr * in6)1739 in6_localip(struct in6_addr *in6)
1740 {
1741 	struct rm_priotracker in6_ifa_tracker;
1742 	struct in6_ifaddr *ia;
1743 
1744 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1745 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1746 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1747 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1748 			return (1);
1749 		}
1750 	}
1751 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1752 	return (0);
1753 }
1754 
1755 /*
1756  * Return 1 if an internet address is configured on an interface.
1757  */
1758 int
in6_ifhasaddr(struct ifnet * ifp,struct in6_addr * addr)1759 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1760 {
1761 	struct in6_addr in6;
1762 	struct ifaddr *ifa;
1763 	struct in6_ifaddr *ia6;
1764 
1765 	NET_EPOCH_ASSERT();
1766 
1767 	in6 = *addr;
1768 	if (in6_clearscope(&in6))
1769 		return (0);
1770 	in6_setscope(&in6, ifp, NULL);
1771 
1772 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1773 		if (ifa->ifa_addr->sa_family != AF_INET6)
1774 			continue;
1775 		ia6 = (struct in6_ifaddr *)ifa;
1776 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1777 			return (1);
1778 	}
1779 
1780 	return (0);
1781 }
1782 
1783 int
in6_is_addr_deprecated(struct sockaddr_in6 * sa6)1784 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1785 {
1786 	struct rm_priotracker in6_ifa_tracker;
1787 	struct in6_ifaddr *ia;
1788 
1789 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1790 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1791 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1792 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1793 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1794 				return (1); /* true */
1795 			}
1796 			break;
1797 		}
1798 	}
1799 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1800 
1801 	return (0);		/* false */
1802 }
1803 
1804 /*
1805  * return length of part which dst and src are equal
1806  * hard coding...
1807  */
1808 int
in6_matchlen(struct in6_addr * src,struct in6_addr * dst)1809 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1810 {
1811 	int match = 0;
1812 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1813 	u_char *lim = s + 16, r;
1814 
1815 	while (s < lim)
1816 		if ((r = (*d++ ^ *s++)) != 0) {
1817 			while (r < 128) {
1818 				match++;
1819 				r <<= 1;
1820 			}
1821 			break;
1822 		} else
1823 			match += 8;
1824 	return match;
1825 }
1826 
1827 /* XXX: to be scope conscious */
1828 int
in6_are_prefix_equal(struct in6_addr * p1,struct in6_addr * p2,int len)1829 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1830 {
1831 	int bytelen, bitlen;
1832 
1833 	/* sanity check */
1834 	if (0 > len || len > 128) {
1835 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1836 		    len);
1837 		return (0);
1838 	}
1839 
1840 	bytelen = len / 8;
1841 	bitlen = len % 8;
1842 
1843 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1844 		return (0);
1845 	if (bitlen != 0 &&
1846 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1847 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1848 		return (0);
1849 
1850 	return (1);
1851 }
1852 
1853 void
in6_prefixlen2mask(struct in6_addr * maskp,int len)1854 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1855 {
1856 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1857 	int bytelen, bitlen, i;
1858 
1859 	/* sanity check */
1860 	if (0 > len || len > 128) {
1861 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1862 		    len);
1863 		return;
1864 	}
1865 
1866 	bzero(maskp, sizeof(*maskp));
1867 	bytelen = len / 8;
1868 	bitlen = len % 8;
1869 	for (i = 0; i < bytelen; i++)
1870 		maskp->s6_addr[i] = 0xff;
1871 	if (bitlen)
1872 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1873 }
1874 
1875 /*
1876  * return the best address out of the same scope. if no address was
1877  * found, return the first valid address from designated IF.
1878  */
1879 struct in6_ifaddr *
in6_ifawithifp(struct ifnet * ifp,struct in6_addr * dst)1880 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1881 {
1882 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1883 	struct ifaddr *ifa;
1884 	struct in6_ifaddr *besta = NULL;
1885 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1886 
1887 	NET_EPOCH_ASSERT();
1888 
1889 	dep[0] = dep[1] = NULL;
1890 
1891 	/*
1892 	 * We first look for addresses in the same scope.
1893 	 * If there is one, return it.
1894 	 * If two or more, return one which matches the dst longest.
1895 	 * If none, return one of global addresses assigned other ifs.
1896 	 */
1897 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1898 		if (ifa->ifa_addr->sa_family != AF_INET6)
1899 			continue;
1900 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1901 			continue; /* XXX: is there any case to allow anycast? */
1902 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1903 			continue; /* don't use this interface */
1904 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1905 			continue;
1906 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1907 			if (V_ip6_use_deprecated)
1908 				dep[0] = (struct in6_ifaddr *)ifa;
1909 			continue;
1910 		}
1911 
1912 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1913 			/*
1914 			 * call in6_matchlen() as few as possible
1915 			 */
1916 			if (besta) {
1917 				if (blen == -1)
1918 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1919 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1920 				if (tlen > blen) {
1921 					blen = tlen;
1922 					besta = (struct in6_ifaddr *)ifa;
1923 				}
1924 			} else
1925 				besta = (struct in6_ifaddr *)ifa;
1926 		}
1927 	}
1928 	if (besta) {
1929 		ifa_ref(&besta->ia_ifa);
1930 		return (besta);
1931 	}
1932 
1933 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1934 		if (ifa->ifa_addr->sa_family != AF_INET6)
1935 			continue;
1936 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1937 			continue; /* XXX: is there any case to allow anycast? */
1938 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1939 			continue; /* don't use this interface */
1940 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1941 			continue;
1942 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1943 			if (V_ip6_use_deprecated)
1944 				dep[1] = (struct in6_ifaddr *)ifa;
1945 			continue;
1946 		}
1947 
1948 		if (ifa != NULL)
1949 			ifa_ref(ifa);
1950 		return (struct in6_ifaddr *)ifa;
1951 	}
1952 
1953 	/* use the last-resort values, that are, deprecated addresses */
1954 	if (dep[0]) {
1955 		ifa_ref((struct ifaddr *)dep[0]);
1956 		return dep[0];
1957 	}
1958 	if (dep[1]) {
1959 		ifa_ref((struct ifaddr *)dep[1]);
1960 		return dep[1];
1961 	}
1962 
1963 	return NULL;
1964 }
1965 
1966 /*
1967  * perform DAD when interface becomes IFF_UP.
1968  */
1969 void
in6_if_up(struct ifnet * ifp)1970 in6_if_up(struct ifnet *ifp)
1971 {
1972 	struct epoch_tracker et;
1973 	struct ifaddr *ifa;
1974 	struct in6_ifaddr *ia;
1975 
1976 	NET_EPOCH_ENTER(et);
1977 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1978 		if (ifa->ifa_addr->sa_family != AF_INET6)
1979 			continue;
1980 		ia = (struct in6_ifaddr *)ifa;
1981 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1982 			/*
1983 			 * The TENTATIVE flag was likely set by hand
1984 			 * beforehand, implicitly indicating the need for DAD.
1985 			 * We may be able to skip the random delay in this
1986 			 * case, but we impose delays just in case.
1987 			 */
1988 			nd6_dad_start(ifa,
1989 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1990 		}
1991 	}
1992 	NET_EPOCH_EXIT(et);
1993 
1994 	/*
1995 	 * special cases, like 6to4, are handled in in6_ifattach
1996 	 */
1997 	in6_ifattach(ifp, NULL);
1998 }
1999 
2000 int
in6if_do_dad(struct ifnet * ifp)2001 in6if_do_dad(struct ifnet *ifp)
2002 {
2003 
2004 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2005 		return (0);
2006 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2007 		return (0);
2008 	if ((ND_IFINFO(ifp)->flags &
2009 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2010 		return (0);
2011 	return (1);
2012 }
2013 
2014 /*
2015  * Calculate max IPv6 MTU through all the interfaces and store it
2016  * to in6_maxmtu.
2017  */
2018 void
in6_setmaxmtu(void)2019 in6_setmaxmtu(void)
2020 {
2021 	struct epoch_tracker et;
2022 	unsigned long maxmtu = 0;
2023 	struct ifnet *ifp;
2024 
2025 	NET_EPOCH_ENTER(et);
2026 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2027 		/* this function can be called during ifnet initialization */
2028 		if (!ifp->if_afdata[AF_INET6])
2029 			continue;
2030 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2031 		    IN6_LINKMTU(ifp) > maxmtu)
2032 			maxmtu = IN6_LINKMTU(ifp);
2033 	}
2034 	NET_EPOCH_EXIT(et);
2035 	if (maxmtu)	/* update only when maxmtu is positive */
2036 		V_in6_maxmtu = maxmtu;
2037 }
2038 
2039 /*
2040  * Provide the length of interface identifiers to be used for the link attached
2041  * to the given interface.  The length should be defined in "IPv6 over
2042  * xxx-link" document.  Note that address architecture might also define
2043  * the length for a particular set of address prefixes, regardless of the
2044  * link type.  As clarified in rfc2462bis, those two definitions should be
2045  * consistent, and those really are as of August 2004.
2046  */
2047 int
in6_if2idlen(struct ifnet * ifp)2048 in6_if2idlen(struct ifnet *ifp)
2049 {
2050 	switch (ifp->if_type) {
2051 	case IFT_ETHER:		/* RFC2464 */
2052 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2053 	case IFT_L2VLAN:	/* ditto */
2054 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2055 	case IFT_INFINIBAND:
2056 		return (64);
2057 	case IFT_PPP:		/* RFC2472 */
2058 		return (64);
2059 	case IFT_FRELAY:	/* RFC2590 */
2060 		return (64);
2061 	case IFT_IEEE1394:	/* RFC3146 */
2062 		return (64);
2063 	case IFT_GIF:
2064 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2065 	case IFT_LOOP:
2066 		return (64);	/* XXX: is this really correct? */
2067 	default:
2068 		/*
2069 		 * Unknown link type:
2070 		 * It might be controversial to use the today's common constant
2071 		 * of 64 for these cases unconditionally.  For full compliance,
2072 		 * we should return an error in this case.  On the other hand,
2073 		 * if we simply miss the standard for the link type or a new
2074 		 * standard is defined for a new link type, the IFID length
2075 		 * is very likely to be the common constant.  As a compromise,
2076 		 * we always use the constant, but make an explicit notice
2077 		 * indicating the "unknown" case.
2078 		 */
2079 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2080 		return (64);
2081 	}
2082 }
2083 
2084 struct in6_llentry {
2085 	struct llentry		base;
2086 };
2087 
2088 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2089 #define	IN6_LLTBL_HASH(k, h) \
2090 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2091 
2092 /*
2093  * Do actual deallocation of @lle.
2094  */
2095 static void
in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)2096 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2097 {
2098 	struct llentry *lle;
2099 
2100 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2101 	LLE_LOCK_DESTROY(lle);
2102 	LLE_REQ_DESTROY(lle);
2103 	free(lle, M_LLTABLE);
2104 }
2105 
2106 /*
2107  * Called by LLE_FREE_LOCKED when number of references
2108  * drops to zero.
2109  */
2110 static void
in6_lltable_destroy_lle(struct llentry * lle)2111 in6_lltable_destroy_lle(struct llentry *lle)
2112 {
2113 
2114 	LLE_WUNLOCK(lle);
2115 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2116 }
2117 
2118 static struct llentry *
in6_lltable_new(const struct in6_addr * addr6,u_int flags)2119 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2120 {
2121 	struct in6_llentry *lle;
2122 
2123 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2124 	if (lle == NULL)		/* NB: caller generates msg */
2125 		return NULL;
2126 
2127 	lle->base.r_l3addr.addr6 = *addr6;
2128 	lle->base.lle_refcnt = 1;
2129 	lle->base.lle_free = in6_lltable_destroy_lle;
2130 	LLE_LOCK_INIT(&lle->base);
2131 	LLE_REQ_INIT(&lle->base);
2132 	callout_init(&lle->base.lle_timer, 1);
2133 
2134 	return (&lle->base);
2135 }
2136 
2137 static int
in6_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)2138 in6_lltable_match_prefix(const struct sockaddr *saddr,
2139     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2140 {
2141 	const struct in6_addr *addr, *mask, *lle_addr;
2142 
2143 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2144 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2145 	lle_addr = &lle->r_l3addr.addr6;
2146 
2147 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2148 		return (0);
2149 
2150 	if (lle->la_flags & LLE_IFADDR) {
2151 		/*
2152 		 * Delete LLE_IFADDR records IFF address & flag matches.
2153 		 * Note that addr is the interface address within prefix
2154 		 * being matched.
2155 		 */
2156 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2157 		    (flags & LLE_STATIC) != 0)
2158 			return (1);
2159 		return (0);
2160 	}
2161 
2162 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2163 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2164 		return (1);
2165 
2166 	return (0);
2167 }
2168 
2169 static void
in6_lltable_free_entry(struct lltable * llt,struct llentry * lle)2170 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2171 {
2172 	struct ifnet *ifp;
2173 
2174 	LLE_WLOCK_ASSERT(lle);
2175 	KASSERT(llt != NULL, ("lltable is NULL"));
2176 
2177 	/* Unlink entry from table */
2178 	if ((lle->la_flags & LLE_LINKED) != 0) {
2179 		ifp = llt->llt_ifp;
2180 		IF_AFDATA_WLOCK_ASSERT(ifp);
2181 		lltable_unlink_entry(llt, lle);
2182 	}
2183 
2184 	llentry_free(lle);
2185 }
2186 
2187 static int
in6_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)2188 in6_lltable_rtcheck(struct ifnet *ifp,
2189 		    u_int flags,
2190 		    const struct sockaddr *l3addr)
2191 {
2192 	const struct sockaddr_in6 *sin6;
2193 	struct nhop_object *nh;
2194 	struct in6_addr dst;
2195 	uint32_t scopeid;
2196 	char ip6buf[INET6_ADDRSTRLEN];
2197 	int fibnum;
2198 
2199 	NET_EPOCH_ASSERT();
2200 	KASSERT(l3addr->sa_family == AF_INET6,
2201 	    ("sin_family %d", l3addr->sa_family));
2202 
2203 	sin6 = (const struct sockaddr_in6 *)l3addr;
2204 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2205 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2206 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2207 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2208 		struct ifaddr *ifa;
2209 		/*
2210 		 * Create an ND6 cache for an IPv6 neighbor
2211 		 * that is not covered by our own prefix.
2212 		 */
2213 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2214 		if (ifa != NULL) {
2215 			return 0;
2216 		}
2217 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2218 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2219 		return EINVAL;
2220 	}
2221 	return 0;
2222 }
2223 
2224 /*
2225  * Called by the datapath to indicate that the entry was used.
2226  */
2227 static void
in6_lltable_mark_used(struct llentry * lle)2228 in6_lltable_mark_used(struct llentry *lle)
2229 {
2230 
2231 	LLE_REQ_LOCK(lle);
2232 	lle->r_skip_req = 0;
2233 
2234 	/*
2235 	 * Set the hit time so the callback function
2236 	 * can determine the remaining time before
2237 	 * transiting to the DELAY state.
2238 	 */
2239 	lle->lle_hittime = time_uptime;
2240 	LLE_REQ_UNLOCK(lle);
2241 }
2242 
2243 static inline uint32_t
in6_lltable_hash_dst(const struct in6_addr * dst,uint32_t hsize)2244 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2245 {
2246 
2247 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2248 }
2249 
2250 static uint32_t
in6_lltable_hash(const struct llentry * lle,uint32_t hsize)2251 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2252 {
2253 
2254 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2255 }
2256 
2257 static void
in6_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)2258 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2259 {
2260 	struct sockaddr_in6 *sin6;
2261 
2262 	sin6 = (struct sockaddr_in6 *)sa;
2263 	bzero(sin6, sizeof(*sin6));
2264 	sin6->sin6_family = AF_INET6;
2265 	sin6->sin6_len = sizeof(*sin6);
2266 	sin6->sin6_addr = lle->r_l3addr.addr6;
2267 }
2268 
2269 static inline struct llentry *
in6_lltable_find_dst(struct lltable * llt,const struct in6_addr * dst)2270 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2271 {
2272 	struct llentry *lle;
2273 	struct llentries *lleh;
2274 	u_int hashidx;
2275 
2276 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2277 	lleh = &llt->lle_head[hashidx];
2278 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2279 		if (lle->la_flags & LLE_DELETED)
2280 			continue;
2281 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2282 			break;
2283 	}
2284 
2285 	return (lle);
2286 }
2287 
2288 static void
in6_lltable_delete_entry(struct lltable * llt,struct llentry * lle)2289 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2290 {
2291 
2292 	lle->la_flags |= LLE_DELETED;
2293 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2294 #ifdef DIAGNOSTIC
2295 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2296 #endif
2297 	llentry_free(lle);
2298 }
2299 
2300 static struct llentry *
in6_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)2301 in6_lltable_alloc(struct lltable *llt, u_int flags,
2302 	const struct sockaddr *l3addr)
2303 {
2304 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2305 	struct ifnet *ifp = llt->llt_ifp;
2306 	struct llentry *lle;
2307 	char linkhdr[LLE_MAX_LINKHDR];
2308 	size_t linkhdrsize;
2309 	int lladdr_off;
2310 
2311 	KASSERT(l3addr->sa_family == AF_INET6,
2312 	    ("sin_family %d", l3addr->sa_family));
2313 
2314 	/*
2315 	 * A route that covers the given address must have
2316 	 * been installed 1st because we are doing a resolution,
2317 	 * verify this.
2318 	 */
2319 	if (!(flags & LLE_IFADDR) &&
2320 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2321 		return (NULL);
2322 
2323 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2324 	if (lle == NULL) {
2325 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2326 		return (NULL);
2327 	}
2328 	lle->la_flags = flags;
2329 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2330 		linkhdrsize = LLE_MAX_LINKHDR;
2331 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2332 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2333 			NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2334 			return (NULL);
2335 		}
2336 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2337 		    lladdr_off);
2338 		lle->la_flags |= LLE_STATIC;
2339 	}
2340 
2341 	if ((lle->la_flags & LLE_STATIC) != 0)
2342 		lle->ln_state = ND6_LLINFO_REACHABLE;
2343 
2344 	return (lle);
2345 }
2346 
2347 static struct llentry *
in6_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)2348 in6_lltable_lookup(struct lltable *llt, u_int flags,
2349 	const struct sockaddr *l3addr)
2350 {
2351 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2352 	struct llentry *lle;
2353 
2354 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2355 	KASSERT(l3addr->sa_family == AF_INET6,
2356 	    ("sin_family %d", l3addr->sa_family));
2357 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2358 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2359 	    ("wrong lle request flags: %#x", flags));
2360 
2361 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2362 	if (lle == NULL)
2363 		return (NULL);
2364 	if (flags & LLE_UNLOCKED)
2365 		return (lle);
2366 
2367 	if (flags & LLE_EXCLUSIVE)
2368 		LLE_WLOCK(lle);
2369 	else
2370 		LLE_RLOCK(lle);
2371 
2372 	/*
2373 	 * If the afdata lock is not held, the LLE may have been unlinked while
2374 	 * we were blocked on the LLE lock.  Check for this case.
2375 	 */
2376 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2377 		if (flags & LLE_EXCLUSIVE)
2378 			LLE_WUNLOCK(lle);
2379 		else
2380 			LLE_RUNLOCK(lle);
2381 		return (NULL);
2382 	}
2383 	return (lle);
2384 }
2385 
2386 static int
in6_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)2387 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2388     struct sysctl_req *wr)
2389 {
2390 	struct ifnet *ifp = llt->llt_ifp;
2391 	/* XXX stack use */
2392 	struct {
2393 		struct rt_msghdr	rtm;
2394 		struct sockaddr_in6	sin6;
2395 		/*
2396 		 * ndp.c assumes that sdl is word aligned
2397 		 */
2398 #ifdef __LP64__
2399 		uint32_t		pad;
2400 #endif
2401 		struct sockaddr_dl	sdl;
2402 	} ndpc;
2403 	struct sockaddr_dl *sdl;
2404 	int error;
2405 
2406 	bzero(&ndpc, sizeof(ndpc));
2407 	/* skip deleted entries */
2408 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2409 		return (0);
2410 	/* Skip if jailed and not a valid IP of the prison. */
2411 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2412 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2413 		return (0);
2414 	/*
2415 	 * produce a msg made of:
2416 	 *  struct rt_msghdr;
2417 	 *  struct sockaddr_in6 (IPv6)
2418 	 *  struct sockaddr_dl;
2419 	 */
2420 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2421 	ndpc.rtm.rtm_version = RTM_VERSION;
2422 	ndpc.rtm.rtm_type = RTM_GET;
2423 	ndpc.rtm.rtm_flags = RTF_UP;
2424 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2425 	sa6_recoverscope(&ndpc.sin6);
2426 
2427 	/* publish */
2428 	if (lle->la_flags & LLE_PUB)
2429 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2430 
2431 	sdl = &ndpc.sdl;
2432 	sdl->sdl_family = AF_LINK;
2433 	sdl->sdl_len = sizeof(*sdl);
2434 	sdl->sdl_index = ifp->if_index;
2435 	sdl->sdl_type = ifp->if_type;
2436 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2437 		sdl->sdl_alen = ifp->if_addrlen;
2438 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2439 	} else {
2440 		sdl->sdl_alen = 0;
2441 		bzero(LLADDR(sdl), ifp->if_addrlen);
2442 	}
2443 	if (lle->la_expire != 0)
2444 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2445 		    lle->lle_remtime / hz + time_second - time_uptime;
2446 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2447 	if (lle->la_flags & LLE_STATIC)
2448 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2449 	if (lle->la_flags & LLE_IFADDR)
2450 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2451 	if (lle->ln_router != 0)
2452 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2453 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2454 	/* Store state in rmx_weight value */
2455 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2456 	ndpc.rtm.rtm_index = ifp->if_index;
2457 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2458 
2459 	return (error);
2460 }
2461 
2462 static struct lltable *
in6_lltattach(struct ifnet * ifp)2463 in6_lltattach(struct ifnet *ifp)
2464 {
2465 	struct lltable *llt;
2466 
2467 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2468 	llt->llt_af = AF_INET6;
2469 	llt->llt_ifp = ifp;
2470 
2471 	llt->llt_lookup = in6_lltable_lookup;
2472 	llt->llt_alloc_entry = in6_lltable_alloc;
2473 	llt->llt_delete_entry = in6_lltable_delete_entry;
2474 	llt->llt_dump_entry = in6_lltable_dump_entry;
2475 	llt->llt_hash = in6_lltable_hash;
2476 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2477 	llt->llt_free_entry = in6_lltable_free_entry;
2478 	llt->llt_match_prefix = in6_lltable_match_prefix;
2479 	llt->llt_mark_used = in6_lltable_mark_used;
2480  	lltable_link(llt);
2481 
2482 	return (llt);
2483 }
2484 
2485 void *
in6_domifattach(struct ifnet * ifp)2486 in6_domifattach(struct ifnet *ifp)
2487 {
2488 	struct in6_ifextra *ext;
2489 
2490 	/* There are not IPv6-capable interfaces. */
2491 	switch (ifp->if_type) {
2492 	case IFT_PFLOG:
2493 	case IFT_PFSYNC:
2494 	case IFT_USB:
2495 		return (NULL);
2496 	}
2497 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2498 	bzero(ext, sizeof(*ext));
2499 
2500 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2501 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2502 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2503 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2504 
2505 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2506 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2507 	    M_WAITOK);
2508 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2509 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2510 
2511 	ext->nd_ifinfo = nd6_ifattach(ifp);
2512 	ext->scope6_id = scope6_ifattach(ifp);
2513 	ext->lltable = in6_lltattach(ifp);
2514 
2515 	ext->mld_ifinfo = mld_domifattach(ifp);
2516 
2517 	return ext;
2518 }
2519 
2520 int
in6_domifmtu(struct ifnet * ifp)2521 in6_domifmtu(struct ifnet *ifp)
2522 {
2523 	if (ifp->if_afdata[AF_INET6] == NULL)
2524 		return ifp->if_mtu;
2525 
2526 	return (IN6_LINKMTU(ifp));
2527 }
2528 
2529 void
in6_domifdetach(struct ifnet * ifp,void * aux)2530 in6_domifdetach(struct ifnet *ifp, void *aux)
2531 {
2532 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2533 
2534 	mld_domifdetach(ifp);
2535 	scope6_ifdetach(ext->scope6_id);
2536 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2537 	lltable_free(ext->lltable);
2538 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2539 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2540 	free(ext->in6_ifstat, M_IFADDR);
2541 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2542 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2543 	free(ext->icmp6_ifstat, M_IFADDR);
2544 	free(ext, M_IFADDR);
2545 }
2546 
2547 /*
2548  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2549  * v4 mapped addr or v4 compat addr
2550  */
2551 void
in6_sin6_2_sin(struct sockaddr_in * sin,struct sockaddr_in6 * sin6)2552 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2553 {
2554 
2555 	bzero(sin, sizeof(*sin));
2556 	sin->sin_len = sizeof(struct sockaddr_in);
2557 	sin->sin_family = AF_INET;
2558 	sin->sin_port = sin6->sin6_port;
2559 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2560 }
2561 
2562 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2563 void
in6_sin_2_v4mapsin6(struct sockaddr_in * sin,struct sockaddr_in6 * sin6)2564 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2565 {
2566 	bzero(sin6, sizeof(*sin6));
2567 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2568 	sin6->sin6_family = AF_INET6;
2569 	sin6->sin6_port = sin->sin_port;
2570 	sin6->sin6_addr.s6_addr32[0] = 0;
2571 	sin6->sin6_addr.s6_addr32[1] = 0;
2572 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2573 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2574 }
2575 
2576 /* Convert sockaddr_in6 into sockaddr_in. */
2577 void
in6_sin6_2_sin_in_sock(struct sockaddr * nam)2578 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2579 {
2580 	struct sockaddr_in *sin_p;
2581 	struct sockaddr_in6 sin6;
2582 
2583 	/*
2584 	 * Save original sockaddr_in6 addr and convert it
2585 	 * to sockaddr_in.
2586 	 */
2587 	sin6 = *(struct sockaddr_in6 *)nam;
2588 	sin_p = (struct sockaddr_in *)nam;
2589 	in6_sin6_2_sin(sin_p, &sin6);
2590 }
2591 
2592 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2593 void
in6_sin_2_v4mapsin6_in_sock(struct sockaddr ** nam)2594 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2595 {
2596 	struct sockaddr_in *sin_p;
2597 	struct sockaddr_in6 *sin6_p;
2598 
2599 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2600 	sin_p = (struct sockaddr_in *)*nam;
2601 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2602 	free(*nam, M_SONAME);
2603 	*nam = (struct sockaddr *)sin6_p;
2604 }
2605