xref: /freebsd-13-stable/sys/netinet6/in6.c (revision ff99bfe3ef739a026e0cf554295774c8bf02d326)
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 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 
69 #include <sys/param.h>
70 #include <sys/eventhandler.h>
71 #include <sys/errno.h>
72 #include <sys/jail.h>
73 #include <sys/malloc.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/systm.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/lock.h>
84 #include <sys/rmlock.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/route/route_ctl.h>
93 #include <net/route/nhop.h>
94 #include <net/if_dl.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <net/if_llatbl.h>
100 #include <netinet/if_ether.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/ip_carp.h>
105 
106 #include <netinet/ip6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/mld6_var.h>
110 #include <netinet6/ip6_mroute.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/in6_fib.h>
114 #include <netinet6/in6_pcb.h>
115 
116 /*
117  * struct in6_ifreq and struct ifreq must be type punnable for common members
118  * of ifr_ifru to allow accessors to be shared.
119  */
120 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
121     offsetof(struct ifreq, ifr_ifru),
122     "struct in6_ifreq and struct ifreq are not type punnable");
123 
124 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
125 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
126 
127 /*
128  * Definitions of some costant IP6 addresses.
129  */
130 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
131 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
132 const struct in6_addr in6addr_nodelocal_allnodes =
133 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
134 const struct in6_addr in6addr_linklocal_allnodes =
135 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
136 const struct in6_addr in6addr_linklocal_allrouters =
137 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
138 const struct in6_addr in6addr_linklocal_allv2routers =
139 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
140 
141 const struct in6_addr in6mask0 = IN6MASK0;
142 const struct in6_addr in6mask32 = IN6MASK32;
143 const struct in6_addr in6mask64 = IN6MASK64;
144 const struct in6_addr in6mask96 = IN6MASK96;
145 const struct in6_addr in6mask128 = IN6MASK128;
146 
147 const struct sockaddr_in6 sa6_any =
148 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
149 
150 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
151 	struct in6_aliasreq *, int);
152 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
153 
154 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
155     struct in6_ifaddr *, int);
156 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
157     struct in6_aliasreq *, int flags);
158 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
159     struct in6_ifaddr *, int, int);
160 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
161     struct in6_ifaddr *, int);
162 
163 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
164 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
165 
166 static struct sx in6_control_sx;
167 SX_SYSINIT(in6_control_sx, &in6_control_sx, "in6_control");
168 
169 void
in6_newaddrmsg(struct in6_ifaddr * ia,int cmd)170 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
171 {
172 	struct rt_addrinfo info;
173 	struct ifaddr *ifa;
174 	struct sockaddr_dl gateway;
175 	int fibnum;
176 
177 	ifa = &ia->ia_ifa;
178 
179 	/*
180 	 * Prepare info data for the host route.
181 	 * This code mimics one from ifa_maintain_loopback_route().
182 	 */
183 	bzero(&info, sizeof(struct rt_addrinfo));
184 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
185 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
186 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
187 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
188 	if (cmd != RTM_DELETE)
189 		info.rti_ifp = V_loif;
190 
191 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
192 
193 	if (cmd == RTM_ADD) {
194 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
195 		rt_routemsg_info(cmd, &info, fibnum);
196 	} else if (cmd == RTM_DELETE) {
197 		rt_routemsg_info(cmd, &info, fibnum);
198 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
199 	}
200 }
201 
202 int
in6_mask2len(struct in6_addr * mask,u_char * lim0)203 in6_mask2len(struct in6_addr *mask, u_char *lim0)
204 {
205 	int x = 0, y;
206 	u_char *lim = lim0, *p;
207 
208 	/* ignore the scope_id part */
209 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
210 		lim = (u_char *)mask + sizeof(*mask);
211 	for (p = (u_char *)mask; p < lim; x++, p++) {
212 		if (*p != 0xff)
213 			break;
214 	}
215 	y = 0;
216 	if (p < lim) {
217 		for (y = 0; y < 8; y++) {
218 			if ((*p & (0x80 >> y)) == 0)
219 				break;
220 		}
221 	}
222 
223 	/*
224 	 * when the limit pointer is given, do a stricter check on the
225 	 * remaining bits.
226 	 */
227 	if (p < lim) {
228 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
229 			return (-1);
230 		for (p = p + 1; p < lim; p++)
231 			if (*p != 0)
232 				return (-1);
233 	}
234 
235 	return x * 8 + y;
236 }
237 
238 #ifdef COMPAT_FREEBSD32
239 struct in6_ndifreq32 {
240 	char ifname[IFNAMSIZ];
241 	uint32_t ifindex;
242 };
243 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
244 #endif
245 
246 int
in6_control(struct socket * so,u_long cmd,caddr_t data,struct ifnet * ifp,struct thread * td)247 in6_control(struct socket *so, u_long cmd, caddr_t data,
248     struct ifnet *ifp, struct thread *td)
249 {
250 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
251 	struct	in6_ifaddr *ia = NULL;
252 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
253 	struct sockaddr_in6 *sa6;
254 	int carp_attached = 0;
255 	int error;
256 	u_long ocmd = cmd;
257 	bool control_locked = false;
258 
259 	/*
260 	 * Compat to make pre-10.x ifconfig(8) operable.
261 	 */
262 	if (cmd == OSIOCAIFADDR_IN6)
263 		cmd = SIOCAIFADDR_IN6;
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 		sx_xlock(&in6_control_sx);
414 		control_locked = true;
415 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
416 	} else
417 		ia = NULL;
418 
419 	switch (cmd) {
420 	case SIOCSIFADDR_IN6:
421 	case SIOCSIFDSTADDR_IN6:
422 	case SIOCSIFNETMASK_IN6:
423 		/*
424 		 * Since IPv6 allows a node to assign multiple addresses
425 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
426 		 */
427 		/* we decided to obsolete this command (20000704) */
428 		error = EINVAL;
429 		goto out;
430 
431 	case SIOCDIFADDR_IN6:
432 		/*
433 		 * for IPv4, we look for existing in_ifaddr here to allow
434 		 * "ifconfig if0 delete" to remove the first IPv4 address on
435 		 * the interface.  For IPv6, as the spec allows multiple
436 		 * interface address from the day one, we consider "remove the
437 		 * first one" semantics to be not preferable.
438 		 */
439 		if (ia == NULL) {
440 			error = EADDRNOTAVAIL;
441 			goto out;
442 		}
443 		/* FALLTHROUGH */
444 	case SIOCAIFADDR_IN6:
445 		/*
446 		 * We always require users to specify a valid IPv6 address for
447 		 * the corresponding operation.
448 		 */
449 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
450 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
451 			error = EAFNOSUPPORT;
452 			goto out;
453 		}
454 
455 		if (td != NULL) {
456 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
457 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
458 			if (error)
459 				goto out;
460 		}
461 		/* FALLTHROUGH */
462 	case SIOCGIFSTAT_IN6:
463 	case SIOCGIFSTAT_ICMP6:
464 		if (ifp->if_afdata[AF_INET6] == NULL) {
465 			error = EPFNOSUPPORT;
466 			goto out;
467 		}
468 		break;
469 
470 	case SIOCGIFADDR_IN6:
471 		/* This interface is basically deprecated. use SIOCGIFCONF. */
472 		/* FALLTHROUGH */
473 	case SIOCGIFAFLAG_IN6:
474 	case SIOCGIFNETMASK_IN6:
475 	case SIOCGIFDSTADDR_IN6:
476 	case SIOCGIFALIFETIME_IN6:
477 		/* must think again about its semantics */
478 		if (ia == NULL) {
479 			error = EADDRNOTAVAIL;
480 			goto out;
481 		}
482 		break;
483 	}
484 
485 	switch (cmd) {
486 	case SIOCGIFADDR_IN6:
487 		ifr->ifr_addr = ia->ia_addr;
488 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
489 			goto out;
490 		break;
491 
492 	case SIOCGIFDSTADDR_IN6:
493 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
494 			error = EINVAL;
495 			goto out;
496 		}
497 		ifr->ifr_dstaddr = ia->ia_dstaddr;
498 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
499 			goto out;
500 		break;
501 
502 	case SIOCGIFNETMASK_IN6:
503 		ifr->ifr_addr = ia->ia_prefixmask;
504 		break;
505 
506 	case SIOCGIFAFLAG_IN6:
507 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
508 		break;
509 
510 	case SIOCGIFSTAT_IN6:
511 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
512 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
513 		    &ifr->ifr_ifru.ifru_stat,
514 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
515 		break;
516 
517 	case SIOCGIFSTAT_ICMP6:
518 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
519 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
520 		    &ifr->ifr_ifru.ifru_icmp6stat,
521 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
522 		break;
523 
524 	case SIOCGIFALIFETIME_IN6:
525 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
526 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
527 			time_t maxexpire;
528 			struct in6_addrlifetime *retlt =
529 			    &ifr->ifr_ifru.ifru_lifetime;
530 
531 			/*
532 			 * XXX: adjust expiration time assuming time_t is
533 			 * signed.
534 			 */
535 			maxexpire = (-1) &
536 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
537 			if (ia->ia6_lifetime.ia6t_vltime <
538 			    maxexpire - ia->ia6_updatetime) {
539 				retlt->ia6t_expire = ia->ia6_updatetime +
540 				    ia->ia6_lifetime.ia6t_vltime;
541 			} else
542 				retlt->ia6t_expire = maxexpire;
543 		}
544 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
545 			time_t maxexpire;
546 			struct in6_addrlifetime *retlt =
547 			    &ifr->ifr_ifru.ifru_lifetime;
548 
549 			/*
550 			 * XXX: adjust expiration time assuming time_t is
551 			 * signed.
552 			 */
553 			maxexpire = (-1) &
554 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
555 			if (ia->ia6_lifetime.ia6t_pltime <
556 			    maxexpire - ia->ia6_updatetime) {
557 				retlt->ia6t_preferred = ia->ia6_updatetime +
558 				    ia->ia6_lifetime.ia6t_pltime;
559 			} else
560 				retlt->ia6t_preferred = maxexpire;
561 		}
562 		break;
563 
564 	case SIOCAIFADDR_IN6:
565 	{
566 		struct nd_prefixctl pr0;
567 		struct nd_prefix *pr;
568 
569 		/*
570 		 * first, make or update the interface address structure,
571 		 * and link it to the list.
572 		 */
573 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
574 			goto out;
575 		if (ia != NULL) {
576 			if (ia->ia_ifa.ifa_carp)
577 				(*carp_detach_p)(&ia->ia_ifa, true);
578 			ifa_free(&ia->ia_ifa);
579 		}
580 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
581 		    == NULL) {
582 			/*
583 			 * this can happen when the user specify the 0 valid
584 			 * lifetime.
585 			 */
586 			break;
587 		}
588 
589 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
590 			if (carp_attach_p != NULL)
591 				error = (*carp_attach_p)(&ia->ia_ifa,
592 				    ifra->ifra_vhid);
593 			else
594 				error = EPROTONOSUPPORT;
595 			if (error)
596 				goto out;
597 			else
598 				carp_attached = 1;
599 		}
600 
601 		/*
602 		 * then, make the prefix on-link on the interface.
603 		 * XXX: we'd rather create the prefix before the address, but
604 		 * we need at least one address to install the corresponding
605 		 * interface route, so we configure the address first.
606 		 */
607 
608 		/*
609 		 * convert mask to prefix length (prefixmask has already
610 		 * been validated in in6_update_ifa().
611 		 */
612 		bzero(&pr0, sizeof(pr0));
613 		pr0.ndpr_ifp = ifp;
614 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
615 		    NULL);
616 		if (pr0.ndpr_plen == 128) {
617 			/* we don't need to install a host route. */
618 			goto aifaddr_out;
619 		}
620 		pr0.ndpr_prefix = ifra->ifra_addr;
621 		/* apply the mask for safety. */
622 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
623 		    &ifra->ifra_prefixmask.sin6_addr);
624 
625 		/*
626 		 * XXX: since we don't have an API to set prefix (not address)
627 		 * lifetimes, we just use the same lifetimes as addresses.
628 		 * The (temporarily) installed lifetimes can be overridden by
629 		 * later advertised RAs (when accept_rtadv is non 0), which is
630 		 * an intended behavior.
631 		 */
632 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
633 		pr0.ndpr_raf_auto =
634 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
635 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
636 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
637 
638 		/* add the prefix if not yet. */
639 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
640 			/*
641 			 * nd6_prelist_add will install the corresponding
642 			 * interface route.
643 			 */
644 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
645 				if (carp_attached)
646 					(*carp_detach_p)(&ia->ia_ifa, false);
647 				goto out;
648 			}
649 		}
650 
651 		/* relate the address to the prefix */
652 		if (ia->ia6_ndpr == NULL) {
653 			ia->ia6_ndpr = pr;
654 			pr->ndpr_addrcnt++;
655 
656 			/*
657 			 * If this is the first autoconf address from the
658 			 * prefix, create a temporary address as well
659 			 * (when required).
660 			 */
661 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
662 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
663 				int e;
664 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
665 					log(LOG_NOTICE, "in6_control: failed "
666 					    "to create a temporary address, "
667 					    "errno=%d\n", e);
668 				}
669 			}
670 		}
671 		nd6_prefix_rele(pr);
672 
673 		/*
674 		 * this might affect the status of autoconfigured addresses,
675 		 * that is, this address might make other addresses detached.
676 		 */
677 		pfxlist_onlink_check();
678 
679 aifaddr_out:
680 		/*
681 		 * Try to clear the flag when a new IPv6 address is added
682 		 * onto an IFDISABLED interface and it succeeds.
683 		 */
684 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
685 			struct in6_ndireq nd;
686 
687 			memset(&nd, 0, sizeof(nd));
688 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
689 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
690 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
691 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
692 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
693 				    "failed.");
694 			/*
695 			 * Ignore failure of clearing the flag intentionally.
696 			 * The failure means address duplication was detected.
697 			 */
698 		}
699 		break;
700 	}
701 
702 	case SIOCDIFADDR_IN6:
703 		in6_purgeifaddr(ia);
704 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
705 		    IFADDR_EVENT_DEL);
706 		break;
707 
708 	default:
709 		if (ifp->if_ioctl == NULL) {
710 			error = EOPNOTSUPP;
711 			goto out;
712 		}
713 		error = (*ifp->if_ioctl)(ifp, cmd, data);
714 		goto out;
715 	}
716 
717 	error = 0;
718 out:
719 	if (control_locked)
720 		sx_xunlock(&in6_control_sx);
721 
722 	if (ia != NULL)
723 		ifa_free(&ia->ia_ifa);
724 	return (error);
725 }
726 
727 static struct in6_multi_mship *
in6_joingroup_legacy(struct ifnet * ifp,const struct in6_addr * mcaddr,int * errorp,int delay)728 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
729     int *errorp, int delay)
730 {
731 	struct in6_multi_mship *imm;
732 	int error;
733 
734 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
735 	if (imm == NULL) {
736 		*errorp = ENOBUFS;
737 		return (NULL);
738 	}
739 
740 	delay = (delay * PR_FASTHZ) / hz;
741 
742 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
743 	if (error) {
744 		*errorp = error;
745 		free(imm, M_IP6MADDR);
746 		return (NULL);
747 	}
748 
749 	return (imm);
750 }
751 /*
752  * Join necessary multicast groups.  Factored out from in6_update_ifa().
753  * This entire work should only be done once, for the default FIB.
754  */
755 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)756 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
757     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
758 {
759 	char ip6buf[INET6_ADDRSTRLEN];
760 	struct in6_addr mltaddr;
761 	struct in6_multi_mship *imm;
762 	int delay, error;
763 
764 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
765 
766 	/* Join solicited multicast addr for new host id. */
767 	bzero(&mltaddr, sizeof(struct in6_addr));
768 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
769 	mltaddr.s6_addr32[2] = htonl(1);
770 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
771 	mltaddr.s6_addr8[12] = 0xff;
772 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
773 		/* XXX: should not happen */
774 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
775 		goto cleanup;
776 	}
777 	delay = error = 0;
778 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
779 		/*
780 		 * We need a random delay for DAD on the address being
781 		 * configured.  It also means delaying transmission of the
782 		 * corresponding MLD report to avoid report collision.
783 		 * [RFC 4861, Section 6.3.7]
784 		 */
785 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
786 	}
787 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
788 	if (imm == NULL) {
789 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
790 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
791 		    if_name(ifp), error));
792 		goto cleanup;
793 	}
794 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
795 	*in6m_sol = imm->i6mm_maddr;
796 
797 	/*
798 	 * Join link-local all-nodes address.
799 	 */
800 	mltaddr = in6addr_linklocal_allnodes;
801 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
802 		goto cleanup; /* XXX: should not fail */
803 
804 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
805 	if (imm == NULL) {
806 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
807 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
808 		    if_name(ifp), error));
809 		goto cleanup;
810 	}
811 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
812 
813 	/*
814 	 * Join node information group address.
815 	 */
816 	delay = 0;
817 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
818 		/*
819 		 * The spec does not say anything about delay for this group,
820 		 * but the same logic should apply.
821 		 */
822 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
823 	}
824 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
825 		/* XXX jinmei */
826 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
827 		if (imm == NULL)
828 			nd6log((LOG_WARNING,
829 			    "%s: in6_joingroup failed for %s on %s "
830 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
831 			    &mltaddr), if_name(ifp), error));
832 			/* XXX not very fatal, go on... */
833 		else
834 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
835 	}
836 	if (V_icmp6_nodeinfo_oldmcprefix &&
837 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
838 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
839 		if (imm == NULL)
840 			nd6log((LOG_WARNING,
841 			    "%s: in6_joingroup failed for %s on %s "
842 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
843 			    &mltaddr), if_name(ifp), error));
844 			/* XXX not very fatal, go on... */
845 		else
846 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
847 	}
848 
849 	/*
850 	 * Join interface-local all-nodes address.
851 	 * (ff01::1%ifN, and ff01::%ifN/32)
852 	 */
853 	mltaddr = in6addr_nodelocal_allnodes;
854 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
855 		goto cleanup; /* XXX: should not fail */
856 
857 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
858 	if (imm == NULL) {
859 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
860 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
861 		    &mltaddr), if_name(ifp), error));
862 		goto cleanup;
863 	}
864 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
865 
866 cleanup:
867 	return (error);
868 }
869 
870 /*
871  * Update parameters of an IPv6 interface address.
872  * If necessary, a new entry is created and linked into address chains.
873  * This function is separated from in6_control().
874  */
875 int
in6_update_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)876 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
877     struct in6_ifaddr *ia, int flags)
878 {
879 	int error, hostIsNew = 0;
880 
881 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
882 		return (error);
883 
884 	if (ia == NULL) {
885 		hostIsNew = 1;
886 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
887 			return (ENOBUFS);
888 	}
889 
890 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
891 	if (error != 0) {
892 		if (hostIsNew != 0) {
893 			in6_unlink_ifa(ia, ifp);
894 			ifa_free(&ia->ia_ifa);
895 		}
896 		return (error);
897 	}
898 
899 	if (hostIsNew)
900 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
901 
902 	return (error);
903 }
904 
905 /*
906  * Fill in basic IPv6 address request info.
907  */
908 void
in6_prepare_ifra(struct in6_aliasreq * ifra,const struct in6_addr * addr,const struct in6_addr * mask)909 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
910     const struct in6_addr *mask)
911 {
912 
913 	memset(ifra, 0, sizeof(struct in6_aliasreq));
914 
915 	ifra->ifra_addr.sin6_family = AF_INET6;
916 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
917 	if (addr != NULL)
918 		ifra->ifra_addr.sin6_addr = *addr;
919 
920 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
921 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
922 	if (mask != NULL)
923 		ifra->ifra_prefixmask.sin6_addr = *mask;
924 }
925 
926 static int
in6_validate_ifra(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)927 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
928     struct in6_ifaddr *ia, int flags)
929 {
930 	int plen = -1;
931 	struct sockaddr_in6 dst6;
932 	struct in6_addrlifetime *lt;
933 	char ip6buf[INET6_ADDRSTRLEN];
934 
935 	/* Validate parameters */
936 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
937 		return (EINVAL);
938 
939 	/*
940 	 * The destination address for a p2p link must have a family
941 	 * of AF_UNSPEC or AF_INET6.
942 	 */
943 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
944 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
945 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
946 		return (EAFNOSUPPORT);
947 
948 	/*
949 	 * Validate address
950 	 */
951 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
952 	    ifra->ifra_addr.sin6_family != AF_INET6)
953 		return (EINVAL);
954 
955 	/*
956 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
957 	 * does not carry fields other than sin6_len.
958 	 */
959 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
960 		return (EINVAL);
961 	/*
962 	 * Because the IPv6 address architecture is classless, we require
963 	 * users to specify a (non 0) prefix length (mask) for a new address.
964 	 * We also require the prefix (when specified) mask is valid, and thus
965 	 * reject a non-consecutive mask.
966 	 */
967 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
968 		return (EINVAL);
969 	if (ifra->ifra_prefixmask.sin6_len != 0) {
970 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
971 		    (u_char *)&ifra->ifra_prefixmask +
972 		    ifra->ifra_prefixmask.sin6_len);
973 		if (plen <= 0)
974 			return (EINVAL);
975 	} else {
976 		/*
977 		 * In this case, ia must not be NULL.  We just use its prefix
978 		 * length.
979 		 */
980 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
981 	}
982 	/*
983 	 * If the destination address on a p2p interface is specified,
984 	 * and the address is a scoped one, validate/set the scope
985 	 * zone identifier.
986 	 */
987 	dst6 = ifra->ifra_dstaddr;
988 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
989 	    (dst6.sin6_family == AF_INET6)) {
990 		struct in6_addr in6_tmp;
991 		u_int32_t zoneid;
992 
993 		in6_tmp = dst6.sin6_addr;
994 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
995 			return (EINVAL); /* XXX: should be impossible */
996 
997 		if (dst6.sin6_scope_id != 0) {
998 			if (dst6.sin6_scope_id != zoneid)
999 				return (EINVAL);
1000 		} else		/* user omit to specify the ID. */
1001 			dst6.sin6_scope_id = zoneid;
1002 
1003 		/* convert into the internal form */
1004 		if (sa6_embedscope(&dst6, 0))
1005 			return (EINVAL); /* XXX: should be impossible */
1006 	}
1007 	/* Modify original ifra_dstaddr to reflect changes */
1008 	ifra->ifra_dstaddr = dst6;
1009 
1010 	/*
1011 	 * The destination address can be specified only for a p2p or a
1012 	 * loopback interface.  If specified, the corresponding prefix length
1013 	 * must be 128.
1014 	 */
1015 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1016 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1017 			/* XXX: noisy message */
1018 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1019 			    "be specified for a p2p or a loopback IF only\n"));
1020 			return (EINVAL);
1021 		}
1022 		if (plen != 128) {
1023 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1024 			    "be 128 when dstaddr is specified\n"));
1025 			return (EINVAL);
1026 		}
1027 	}
1028 	/* lifetime consistency check */
1029 	lt = &ifra->ifra_lifetime;
1030 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1031 		return (EINVAL);
1032 	if (lt->ia6t_vltime == 0) {
1033 		/*
1034 		 * the following log might be noisy, but this is a typical
1035 		 * configuration mistake or a tool's bug.
1036 		 */
1037 		nd6log((LOG_INFO,
1038 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1039 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1040 
1041 		if (ia == NULL)
1042 			return (0); /* there's nothing to do */
1043 	}
1044 
1045 	/* Check prefix mask */
1046 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1047 		/*
1048 		 * We prohibit changing the prefix length of an existing
1049 		 * address, because
1050 		 * + such an operation should be rare in IPv6, and
1051 		 * + the operation would confuse prefix management.
1052 		 */
1053 		if (ia->ia_prefixmask.sin6_len != 0 &&
1054 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1055 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1056 			    "of an existing %s address should not be changed\n",
1057 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1058 
1059 			return (EINVAL);
1060 		}
1061 	}
1062 
1063 	return (0);
1064 }
1065 
1066 /*
1067  * Allocate a new ifaddr and link it into chains.
1068  */
1069 static struct in6_ifaddr *
in6_alloc_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,int flags)1070 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1071 {
1072 	struct in6_ifaddr *ia;
1073 
1074 	/*
1075 	 * When in6_alloc_ifa() is called in a process of a received
1076 	 * RA, it is called under an interrupt context.  So, we should
1077 	 * call malloc with M_NOWAIT.
1078 	 */
1079 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1080 	if (ia == NULL)
1081 		return (NULL);
1082 	LIST_INIT(&ia->ia6_memberships);
1083 	/* Initialize the address and masks, and put time stamp */
1084 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1085 	ia->ia_addr.sin6_family = AF_INET6;
1086 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1087 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1088 	ia->ia_addr = ifra->ifra_addr;
1089 	ia->ia6_createtime = time_uptime;
1090 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1091 		/*
1092 		 * Some functions expect that ifa_dstaddr is not
1093 		 * NULL for p2p interfaces.
1094 		 */
1095 		ia->ia_ifa.ifa_dstaddr =
1096 		    (struct sockaddr *)&ia->ia_dstaddr;
1097 	} else {
1098 		ia->ia_ifa.ifa_dstaddr = NULL;
1099 	}
1100 
1101 	/* set prefix mask if any */
1102 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1103 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1104 		ia->ia_prefixmask.sin6_family = AF_INET6;
1105 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1106 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1107 	}
1108 
1109 	ia->ia_ifp = ifp;
1110 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1111 	IF_ADDR_WLOCK(ifp);
1112 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1113 	IF_ADDR_WUNLOCK(ifp);
1114 
1115 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1116 	IN6_IFADDR_WLOCK();
1117 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1118 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1119 	IN6_IFADDR_WUNLOCK();
1120 
1121 	return (ia);
1122 }
1123 
1124 /*
1125  * Update/configure interface address parameters:
1126  *
1127  * 1) Update lifetime
1128  * 2) Update interface metric ad flags
1129  * 3) Notify other subsystems
1130  */
1131 static int
in6_update_ifa_internal(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int hostIsNew,int flags)1132 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1133     struct in6_ifaddr *ia, int hostIsNew, int flags)
1134 {
1135 	int error;
1136 
1137 	/* update timestamp */
1138 	ia->ia6_updatetime = time_uptime;
1139 
1140 	/*
1141 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1142 	 * to see if the address is deprecated or invalidated, but initialize
1143 	 * these members for applications.
1144 	 */
1145 	ia->ia6_lifetime = ifra->ifra_lifetime;
1146 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1147 		ia->ia6_lifetime.ia6t_expire =
1148 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1149 	} else
1150 		ia->ia6_lifetime.ia6t_expire = 0;
1151 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1152 		ia->ia6_lifetime.ia6t_preferred =
1153 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1154 	} else
1155 		ia->ia6_lifetime.ia6t_preferred = 0;
1156 
1157 	/*
1158 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1159 	 * userland, make it deprecated.
1160 	 */
1161 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1162 		ia->ia6_lifetime.ia6t_pltime = 0;
1163 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1164 	}
1165 
1166 	/*
1167 	 * configure address flags.
1168 	 */
1169 	ia->ia6_flags = ifra->ifra_flags;
1170 
1171 	/*
1172 	 * Make the address tentative before joining multicast addresses,
1173 	 * so that corresponding MLD responses would not have a tentative
1174 	 * source address.
1175 	 */
1176 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1177 
1178 	/*
1179 	 * DAD should be performed for an new address or addresses on
1180 	 * an interface with ND6_IFF_IFDISABLED.
1181 	 */
1182 	if (in6if_do_dad(ifp) &&
1183 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1184 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1185 
1186 	/* notify other subsystems */
1187 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1188 
1189 	return (error);
1190 }
1191 
1192 /*
1193  * Do link-level ifa job:
1194  * 1) Add lle entry for added address
1195  * 2) Notifies routing socket users about new address
1196  * 3) join appropriate multicast group
1197  * 4) start DAD if enabled
1198  */
1199 static int
in6_broadcast_ifa(struct ifnet * ifp,struct in6_aliasreq * ifra,struct in6_ifaddr * ia,int flags)1200 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1201     struct in6_ifaddr *ia, int flags)
1202 {
1203 	struct in6_multi *in6m_sol;
1204 	int error = 0;
1205 
1206 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1207 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1208 		in6_purgeaddr(&ia->ia_ifa);
1209 		ifa_free(&ia->ia_ifa);
1210 		return (error);
1211 	}
1212 
1213 	/* Join necessary multicast groups. */
1214 	in6m_sol = NULL;
1215 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1216 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1217 		if (error != 0) {
1218 			in6_purgeaddr(&ia->ia_ifa);
1219 			ifa_free(&ia->ia_ifa);
1220 			return (error);
1221 		}
1222 	}
1223 
1224 	/* Perform DAD, if the address is TENTATIVE. */
1225 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1226 		int delay, mindelay, maxdelay;
1227 
1228 		delay = 0;
1229 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1230 			/*
1231 			 * We need to impose a delay before sending an NS
1232 			 * for DAD.  Check if we also needed a delay for the
1233 			 * corresponding MLD message.  If we did, the delay
1234 			 * should be larger than the MLD delay (this could be
1235 			 * relaxed a bit, but this simple logic is at least
1236 			 * safe).
1237 			 * XXX: Break data hiding guidelines and look at
1238 			 * state for the solicited multicast group.
1239 			 */
1240 			mindelay = 0;
1241 			if (in6m_sol != NULL &&
1242 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1243 				mindelay = in6m_sol->in6m_timer;
1244 			}
1245 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1246 			if (maxdelay - mindelay == 0)
1247 				delay = 0;
1248 			else {
1249 				delay =
1250 				    (arc4random() % (maxdelay - mindelay)) +
1251 				    mindelay;
1252 			}
1253 		}
1254 		nd6_dad_start((struct ifaddr *)ia, delay);
1255 	}
1256 
1257 	in6_newaddrmsg(ia, RTM_ADD);
1258 	ifa_free(&ia->ia_ifa);
1259 	return (error);
1260 }
1261 
1262 /*
1263  * Adds or deletes interface route for p2p ifa.
1264  * Returns 0 on success or errno.
1265  */
1266 static int
in6_handle_dstaddr_rtrequest(int cmd,struct in6_ifaddr * ia)1267 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1268 {
1269 	struct epoch_tracker et;
1270 	struct ifaddr *ifa = &ia->ia_ifa;
1271 	int error;
1272 
1273 	/* Prepare gateway */
1274 	struct sockaddr_dl_short sdl = {
1275 		.sdl_family = AF_LINK,
1276 		.sdl_len = sizeof(struct sockaddr_dl_short),
1277 		.sdl_type = ifa->ifa_ifp->if_type,
1278 		.sdl_index = ifa->ifa_ifp->if_index,
1279 	};
1280 
1281 	struct sockaddr_in6 dst = {
1282 		.sin6_family = AF_INET6,
1283 		.sin6_len = sizeof(struct sockaddr_in6),
1284 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1285 	};
1286 
1287 	struct rt_addrinfo info = {
1288 		.rti_ifa = ifa,
1289 		.rti_ifp = ifa->ifa_ifp,
1290 		.rti_flags = RTF_PINNED | RTF_HOST,
1291 		.rti_info = {
1292 			[RTAX_DST] = (struct sockaddr *)&dst,
1293 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1294 		},
1295 	};
1296 	/* Don't set additional per-gw filters on removal */
1297 
1298 	NET_EPOCH_ENTER(et);
1299 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1300 	NET_EPOCH_EXIT(et);
1301 
1302 	return (error);
1303 }
1304 
1305 static bool
ifa_is_p2p(struct in6_ifaddr * ia)1306 ifa_is_p2p(struct in6_ifaddr *ia)
1307 {
1308 	int plen;
1309 
1310 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1311 
1312 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1313 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1314 		return (true);
1315 
1316 	return (false);
1317 }
1318 
1319 void
in6_purgeaddr(struct ifaddr * ifa)1320 in6_purgeaddr(struct ifaddr *ifa)
1321 {
1322 	struct ifnet *ifp = ifa->ifa_ifp;
1323 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1324 	struct in6_multi_mship *imm;
1325 	int error;
1326 
1327 	if (ifa->ifa_carp)
1328 		(*carp_detach_p)(ifa, false);
1329 
1330 	/*
1331 	 * Remove the loopback route to the interface address.
1332 	 * The check for the current setting of "nd6_useloopback"
1333 	 * is not needed.
1334 	 */
1335 	if (ia->ia_flags & IFA_RTSELF) {
1336 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1337 		    (struct sockaddr *)&ia->ia_addr);
1338 		if (error == 0)
1339 			ia->ia_flags &= ~IFA_RTSELF;
1340 	}
1341 
1342 	/* stop DAD processing */
1343 	nd6_dad_stop(ifa);
1344 
1345 	/* Leave multicast groups. */
1346 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1347 		LIST_REMOVE(imm, i6mm_chain);
1348 		if (imm->i6mm_maddr != NULL)
1349 			in6_leavegroup(imm->i6mm_maddr, NULL);
1350 		free(imm, M_IP6MADDR);
1351 	}
1352 	/* Check if we need to remove p2p route */
1353 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1354 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1355 		if (error != 0)
1356 			log(LOG_INFO, "%s: err=%d, destination address delete "
1357 			    "failed\n", __func__, error);
1358 		ia->ia_flags &= ~IFA_ROUTE;
1359 	}
1360 
1361 	in6_newaddrmsg(ia, RTM_DELETE);
1362 	in6_unlink_ifa(ia, ifp);
1363 }
1364 
1365 /*
1366  * Removes @ia from the corresponding interfaces and unlinks corresponding
1367  *  prefix if no addresses are using it anymore.
1368  */
1369 void
in6_purgeifaddr(struct in6_ifaddr * ia)1370 in6_purgeifaddr(struct in6_ifaddr *ia)
1371 {
1372 	struct nd_prefix *pr;
1373 
1374 	/*
1375 	 * If the address being deleted is the only one that owns
1376 	 * the corresponding prefix, expire the prefix as well.
1377 	 * XXX: theoretically, we don't have to worry about such
1378 	 * relationship, since we separate the address management
1379 	 * and the prefix management.  We do this, however, to provide
1380 	 * as much backward compatibility as possible in terms of
1381 	 * the ioctl operation.
1382 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1383 	 */
1384 	pr = ia->ia6_ndpr;
1385 	in6_purgeaddr(&ia->ia_ifa);
1386 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1387 		ND6_WLOCK();
1388 		nd6_prefix_unlink(pr, NULL);
1389 		ND6_WUNLOCK();
1390 		nd6_prefix_del(pr);
1391 	}
1392 }
1393 
1394 
1395 static void
in6_unlink_ifa(struct in6_ifaddr * ia,struct ifnet * ifp)1396 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1397 {
1398 	char ip6buf[INET6_ADDRSTRLEN];
1399 	int remove_lle;
1400 
1401 	IF_ADDR_WLOCK(ifp);
1402 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1403 	IF_ADDR_WUNLOCK(ifp);
1404 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1405 
1406 	/*
1407 	 * Defer the release of what might be the last reference to the
1408 	 * in6_ifaddr so that it can't be freed before the remainder of the
1409 	 * cleanup.
1410 	 */
1411 	IN6_IFADDR_WLOCK();
1412 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1413 	CK_LIST_REMOVE(ia, ia6_hash);
1414 	IN6_IFADDR_WUNLOCK();
1415 
1416 	/*
1417 	 * Release the reference to the base prefix.  There should be a
1418 	 * positive reference.
1419 	 */
1420 	remove_lle = 0;
1421 	if (ia->ia6_ndpr == NULL) {
1422 		nd6log((LOG_NOTICE,
1423 		    "in6_unlink_ifa: autoconf'ed address "
1424 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1425 	} else {
1426 		ia->ia6_ndpr->ndpr_addrcnt--;
1427 		/* Do not delete lles within prefix if refcont != 0 */
1428 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1429 			remove_lle = 1;
1430 		ia->ia6_ndpr = NULL;
1431 	}
1432 
1433 	nd6_rem_ifa_lle(ia, remove_lle);
1434 
1435 	/*
1436 	 * Also, if the address being removed is autoconf'ed, call
1437 	 * pfxlist_onlink_check() since the release might affect the status of
1438 	 * other (detached) addresses.
1439 	 */
1440 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1441 		pfxlist_onlink_check();
1442 	}
1443 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1444 }
1445 
1446 /*
1447  * Notifies other subsystems about address change/arrival:
1448  * 1) Notifies device handler on the first IPv6 address assignment
1449  * 2) Handle routing table changes for P2P links and route
1450  * 3) Handle routing table changes for address host route
1451  */
1452 static int
in6_notify_ifa(struct ifnet * ifp,struct in6_ifaddr * ia,struct in6_aliasreq * ifra,int hostIsNew)1453 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1454     struct in6_aliasreq *ifra, int hostIsNew)
1455 {
1456 	int	error = 0, ifacount = 0;
1457 	struct ifaddr *ifa;
1458 	struct sockaddr_in6 *pdst;
1459 	char ip6buf[INET6_ADDRSTRLEN];
1460 
1461 	/*
1462 	 * Give the interface a chance to initialize
1463 	 * if this is its first address,
1464 	 */
1465 	if (hostIsNew != 0) {
1466 		struct epoch_tracker et;
1467 
1468 		NET_EPOCH_ENTER(et);
1469 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1470 			if (ifa->ifa_addr->sa_family != AF_INET6)
1471 				continue;
1472 			ifacount++;
1473 		}
1474 		NET_EPOCH_EXIT(et);
1475 	}
1476 
1477 	if (ifacount <= 1 && ifp->if_ioctl) {
1478 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1479 		if (error)
1480 			goto done;
1481 	}
1482 
1483 	/*
1484 	 * If a new destination address is specified, scrub the old one and
1485 	 * install the new destination.  Note that the interface must be
1486 	 * p2p or loopback.
1487 	 */
1488 	pdst = &ifra->ifra_dstaddr;
1489 	if (pdst->sin6_family == AF_INET6 &&
1490 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1491 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1492 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1493 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1494 			    "remove a route to the old destination: %s\n",
1495 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1496 			/* proceed anyway... */
1497 		} else
1498 			ia->ia_flags &= ~IFA_ROUTE;
1499 		ia->ia_dstaddr = *pdst;
1500 	}
1501 
1502 	/*
1503 	 * If a new destination address is specified for a point-to-point
1504 	 * interface, install a route to the destination as an interface
1505 	 * direct route.
1506 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1507 	 * interface that share the same destination.
1508 	 */
1509 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1510 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1511 		if (error)
1512 			goto done;
1513 		ia->ia_flags |= IFA_ROUTE;
1514 	}
1515 
1516 	/*
1517 	 * add a loopback route to self if not exists
1518 	 */
1519 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1520 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1521 		    (struct sockaddr *)&ia->ia_addr);
1522 		if (error == 0)
1523 			ia->ia_flags |= IFA_RTSELF;
1524 	}
1525 done:
1526 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1527 	    "Invoking IPv6 network device address event may sleep");
1528 
1529 	ifa_ref(&ia->ia_ifa);
1530 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1531 	    IFADDR_EVENT_ADD);
1532 	ifa_free(&ia->ia_ifa);
1533 
1534 	return (error);
1535 }
1536 
1537 /*
1538  * Find an IPv6 interface link-local address specific to an interface.
1539  * ifaddr is returned referenced.
1540  */
1541 struct in6_ifaddr *
in6ifa_ifpforlinklocal(struct ifnet * ifp,int ignoreflags)1542 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1543 {
1544 	struct ifaddr *ifa;
1545 
1546 	NET_EPOCH_ASSERT();
1547 
1548 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1549 		if (ifa->ifa_addr->sa_family != AF_INET6)
1550 			continue;
1551 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1552 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1553 			    ignoreflags) != 0)
1554 				continue;
1555 			ifa_ref(ifa);
1556 			break;
1557 		}
1558 	}
1559 
1560 	return ((struct in6_ifaddr *)ifa);
1561 }
1562 
1563 /*
1564  * find the interface address corresponding to a given IPv6 address.
1565  * ifaddr is returned referenced if @referenced flag is set.
1566  */
1567 struct in6_ifaddr *
in6ifa_ifwithaddr(const struct in6_addr * addr,uint32_t zoneid,bool referenced)1568 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1569 {
1570 	struct rm_priotracker in6_ifa_tracker;
1571 	struct in6_ifaddr *ia;
1572 
1573 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1574 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1575 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1576 			if (zoneid != 0 &&
1577 			    zoneid != ia->ia_addr.sin6_scope_id)
1578 				continue;
1579 			if (referenced)
1580 				ifa_ref(&ia->ia_ifa);
1581 			break;
1582 		}
1583 	}
1584 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1585 	return (ia);
1586 }
1587 
1588 /*
1589  * find the internet address corresponding to a given interface and address.
1590  * ifaddr is returned referenced.
1591  */
1592 struct in6_ifaddr *
in6ifa_ifpwithaddr(struct ifnet * ifp,const struct in6_addr * addr)1593 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1594 {
1595 	struct epoch_tracker et;
1596 	struct ifaddr *ifa;
1597 
1598 	NET_EPOCH_ENTER(et);
1599 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1600 		if (ifa->ifa_addr->sa_family != AF_INET6)
1601 			continue;
1602 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1603 			ifa_ref(ifa);
1604 			break;
1605 		}
1606 	}
1607 	NET_EPOCH_EXIT(et);
1608 
1609 	return ((struct in6_ifaddr *)ifa);
1610 }
1611 
1612 /*
1613  * Find a link-local scoped address on ifp and return it if any.
1614  */
1615 struct in6_ifaddr *
in6ifa_llaonifp(struct ifnet * ifp)1616 in6ifa_llaonifp(struct ifnet *ifp)
1617 {
1618 	struct epoch_tracker et;
1619 	struct sockaddr_in6 *sin6;
1620 	struct ifaddr *ifa;
1621 
1622 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1623 		return (NULL);
1624 	NET_EPOCH_ENTER(et);
1625 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1626 		if (ifa->ifa_addr->sa_family != AF_INET6)
1627 			continue;
1628 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1629 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1630 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1631 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1632 			break;
1633 	}
1634 	NET_EPOCH_EXIT(et);
1635 
1636 	return ((struct in6_ifaddr *)ifa);
1637 }
1638 
1639 /*
1640  * Convert IP6 address to printable (loggable) representation. Caller
1641  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1642  */
1643 static char digits[] = "0123456789abcdef";
1644 char *
ip6_sprintf(char * ip6buf,const struct in6_addr * addr)1645 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1646 {
1647 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1648 	char *cp;
1649 	const u_int16_t *a = (const u_int16_t *)addr;
1650 	const u_int8_t *d;
1651 	int dcolon = 0, zero = 0;
1652 
1653 	cp = ip6buf;
1654 
1655 	for (i = 0; i < 8; i++) {
1656 		if (*(a + i) == 0) {
1657 			cnt++;
1658 			if (cnt == 1)
1659 				idx = i;
1660 		}
1661 		else if (maxcnt < cnt) {
1662 			maxcnt = cnt;
1663 			index = idx;
1664 			cnt = 0;
1665 		}
1666 	}
1667 	if (maxcnt < cnt) {
1668 		maxcnt = cnt;
1669 		index = idx;
1670 	}
1671 
1672 	for (i = 0; i < 8; i++) {
1673 		if (dcolon == 1) {
1674 			if (*a == 0) {
1675 				if (i == 7)
1676 					*cp++ = ':';
1677 				a++;
1678 				continue;
1679 			} else
1680 				dcolon = 2;
1681 		}
1682 		if (*a == 0) {
1683 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1684 				if (i == 0)
1685 					*cp++ = ':';
1686 				*cp++ = ':';
1687 				dcolon = 1;
1688 			} else {
1689 				*cp++ = '0';
1690 				*cp++ = ':';
1691 			}
1692 			a++;
1693 			continue;
1694 		}
1695 		d = (const u_char *)a;
1696 		/* Try to eliminate leading zeros in printout like in :0001. */
1697 		zero = 1;
1698 		*cp = digits[*d >> 4];
1699 		if (*cp != '0') {
1700 			zero = 0;
1701 			cp++;
1702 		}
1703 		*cp = digits[*d++ & 0xf];
1704 		if (zero == 0 || (*cp != '0')) {
1705 			zero = 0;
1706 			cp++;
1707 		}
1708 		*cp = digits[*d >> 4];
1709 		if (zero == 0 || (*cp != '0')) {
1710 			zero = 0;
1711 			cp++;
1712 		}
1713 		*cp++ = digits[*d & 0xf];
1714 		*cp++ = ':';
1715 		a++;
1716 	}
1717 	*--cp = '\0';
1718 	return (ip6buf);
1719 }
1720 
1721 int
in6_localaddr(struct in6_addr * in6)1722 in6_localaddr(struct in6_addr *in6)
1723 {
1724 	struct rm_priotracker in6_ifa_tracker;
1725 	struct in6_ifaddr *ia;
1726 
1727 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1728 		return 1;
1729 
1730 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1731 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1732 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1733 		    &ia->ia_prefixmask.sin6_addr)) {
1734 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1735 			return 1;
1736 		}
1737 	}
1738 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1739 
1740 	return (0);
1741 }
1742 
1743 /*
1744  * Return 1 if an internet address is for the local host and configured
1745  * on one of its interfaces.
1746  */
1747 int
in6_localip(struct in6_addr * in6)1748 in6_localip(struct in6_addr *in6)
1749 {
1750 	struct rm_priotracker in6_ifa_tracker;
1751 	struct in6_ifaddr *ia;
1752 
1753 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1754 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1755 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1756 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1757 			return (1);
1758 		}
1759 	}
1760 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1761 	return (0);
1762 }
1763 
1764 /*
1765  * Return 1 if an internet address is configured on an interface.
1766  */
1767 int
in6_ifhasaddr(struct ifnet * ifp,struct in6_addr * addr)1768 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1769 {
1770 	struct in6_addr in6;
1771 	struct ifaddr *ifa;
1772 	struct in6_ifaddr *ia6;
1773 
1774 	NET_EPOCH_ASSERT();
1775 
1776 	in6 = *addr;
1777 	if (in6_clearscope(&in6))
1778 		return (0);
1779 	in6_setscope(&in6, ifp, NULL);
1780 
1781 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1782 		if (ifa->ifa_addr->sa_family != AF_INET6)
1783 			continue;
1784 		ia6 = (struct in6_ifaddr *)ifa;
1785 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1786 			return (1);
1787 	}
1788 
1789 	return (0);
1790 }
1791 
1792 int
in6_is_addr_deprecated(struct sockaddr_in6 * sa6)1793 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1794 {
1795 	struct rm_priotracker in6_ifa_tracker;
1796 	struct in6_ifaddr *ia;
1797 
1798 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1799 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1800 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1801 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1802 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1803 				return (1); /* true */
1804 			}
1805 			break;
1806 		}
1807 	}
1808 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1809 
1810 	return (0);		/* false */
1811 }
1812 
1813 /*
1814  * return length of part which dst and src are equal
1815  * hard coding...
1816  */
1817 int
in6_matchlen(struct in6_addr * src,struct in6_addr * dst)1818 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1819 {
1820 	int match = 0;
1821 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1822 	u_char *lim = s + 16, r;
1823 
1824 	while (s < lim)
1825 		if ((r = (*d++ ^ *s++)) != 0) {
1826 			while (r < 128) {
1827 				match++;
1828 				r <<= 1;
1829 			}
1830 			break;
1831 		} else
1832 			match += 8;
1833 	return match;
1834 }
1835 
1836 /* XXX: to be scope conscious */
1837 int
in6_are_prefix_equal(struct in6_addr * p1,struct in6_addr * p2,int len)1838 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1839 {
1840 	int bytelen, bitlen;
1841 
1842 	/* sanity check */
1843 	if (0 > len || len > 128) {
1844 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1845 		    len);
1846 		return (0);
1847 	}
1848 
1849 	bytelen = len / 8;
1850 	bitlen = len % 8;
1851 
1852 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1853 		return (0);
1854 	if (bitlen != 0 &&
1855 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1856 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1857 		return (0);
1858 
1859 	return (1);
1860 }
1861 
1862 void
in6_prefixlen2mask(struct in6_addr * maskp,int len)1863 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1864 {
1865 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1866 	int bytelen, bitlen, i;
1867 
1868 	/* sanity check */
1869 	if (0 > len || len > 128) {
1870 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1871 		    len);
1872 		return;
1873 	}
1874 
1875 	bzero(maskp, sizeof(*maskp));
1876 	bytelen = len / 8;
1877 	bitlen = len % 8;
1878 	for (i = 0; i < bytelen; i++)
1879 		maskp->s6_addr[i] = 0xff;
1880 	if (bitlen)
1881 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1882 }
1883 
1884 /*
1885  * return the best address out of the same scope. if no address was
1886  * found, return the first valid address from designated IF.
1887  */
1888 struct in6_ifaddr *
in6_ifawithifp(struct ifnet * ifp,struct in6_addr * dst)1889 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1890 {
1891 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1892 	struct ifaddr *ifa;
1893 	struct in6_ifaddr *besta = NULL;
1894 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1895 
1896 	NET_EPOCH_ASSERT();
1897 
1898 	dep[0] = dep[1] = NULL;
1899 
1900 	/*
1901 	 * We first look for addresses in the same scope.
1902 	 * If there is one, return it.
1903 	 * If two or more, return one which matches the dst longest.
1904 	 * If none, return one of global addresses assigned other ifs.
1905 	 */
1906 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1907 		if (ifa->ifa_addr->sa_family != AF_INET6)
1908 			continue;
1909 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1910 			continue; /* XXX: is there any case to allow anycast? */
1911 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1912 			continue; /* don't use this interface */
1913 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1914 			continue;
1915 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1916 			if (V_ip6_use_deprecated)
1917 				dep[0] = (struct in6_ifaddr *)ifa;
1918 			continue;
1919 		}
1920 
1921 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1922 			/*
1923 			 * call in6_matchlen() as few as possible
1924 			 */
1925 			if (besta) {
1926 				if (blen == -1)
1927 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1928 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1929 				if (tlen > blen) {
1930 					blen = tlen;
1931 					besta = (struct in6_ifaddr *)ifa;
1932 				}
1933 			} else
1934 				besta = (struct in6_ifaddr *)ifa;
1935 		}
1936 	}
1937 	if (besta)
1938 		return (besta);
1939 
1940 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1941 		if (ifa->ifa_addr->sa_family != AF_INET6)
1942 			continue;
1943 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1944 			continue; /* XXX: is there any case to allow anycast? */
1945 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1946 			continue; /* don't use this interface */
1947 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1948 			continue;
1949 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1950 			if (V_ip6_use_deprecated)
1951 				dep[1] = (struct in6_ifaddr *)ifa;
1952 			continue;
1953 		}
1954 
1955 		return (struct in6_ifaddr *)ifa;
1956 	}
1957 
1958 	/* use the last-resort values, that are, deprecated addresses */
1959 	if (dep[0])
1960 		return dep[0];
1961 	if (dep[1])
1962 		return dep[1];
1963 
1964 	return NULL;
1965 }
1966 
1967 /*
1968  * perform DAD when interface becomes IFF_UP.
1969  */
1970 void
in6_if_up(struct ifnet * ifp)1971 in6_if_up(struct ifnet *ifp)
1972 {
1973 	struct epoch_tracker et;
1974 	struct ifaddr *ifa;
1975 	struct in6_ifaddr *ia;
1976 
1977 	NET_EPOCH_ENTER(et);
1978 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1979 		if (ifa->ifa_addr->sa_family != AF_INET6)
1980 			continue;
1981 		ia = (struct in6_ifaddr *)ifa;
1982 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1983 			/*
1984 			 * The TENTATIVE flag was likely set by hand
1985 			 * beforehand, implicitly indicating the need for DAD.
1986 			 * We may be able to skip the random delay in this
1987 			 * case, but we impose delays just in case.
1988 			 */
1989 			nd6_dad_start(ifa,
1990 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1991 		}
1992 	}
1993 	NET_EPOCH_EXIT(et);
1994 
1995 	/*
1996 	 * special cases, like 6to4, are handled in in6_ifattach
1997 	 */
1998 	in6_ifattach(ifp, NULL);
1999 }
2000 
2001 int
in6if_do_dad(struct ifnet * ifp)2002 in6if_do_dad(struct ifnet *ifp)
2003 {
2004 
2005 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2006 		return (0);
2007 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2008 		return (0);
2009 	if ((ND_IFINFO(ifp)->flags &
2010 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2011 		return (0);
2012 	return (1);
2013 }
2014 
2015 /*
2016  * Calculate max IPv6 MTU through all the interfaces and store it
2017  * to in6_maxmtu.
2018  */
2019 void
in6_setmaxmtu(void)2020 in6_setmaxmtu(void)
2021 {
2022 	struct epoch_tracker et;
2023 	unsigned long maxmtu = 0;
2024 	struct ifnet *ifp;
2025 
2026 	NET_EPOCH_ENTER(et);
2027 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2028 		/* this function can be called during ifnet initialization */
2029 		if (!ifp->if_afdata[AF_INET6])
2030 			continue;
2031 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2032 		    IN6_LINKMTU(ifp) > maxmtu)
2033 			maxmtu = IN6_LINKMTU(ifp);
2034 	}
2035 	NET_EPOCH_EXIT(et);
2036 	if (maxmtu)	/* update only when maxmtu is positive */
2037 		V_in6_maxmtu = maxmtu;
2038 }
2039 
2040 /*
2041  * Provide the length of interface identifiers to be used for the link attached
2042  * to the given interface.  The length should be defined in "IPv6 over
2043  * xxx-link" document.  Note that address architecture might also define
2044  * the length for a particular set of address prefixes, regardless of the
2045  * link type.  As clarified in rfc2462bis, those two definitions should be
2046  * consistent, and those really are as of August 2004.
2047  */
2048 int
in6_if2idlen(struct ifnet * ifp)2049 in6_if2idlen(struct ifnet *ifp)
2050 {
2051 	switch (ifp->if_type) {
2052 	case IFT_ETHER:		/* RFC2464 */
2053 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2054 	case IFT_L2VLAN:	/* ditto */
2055 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2056 	case IFT_INFINIBAND:
2057 		return (64);
2058 	case IFT_PPP:		/* RFC2472 */
2059 		return (64);
2060 	case IFT_FRELAY:	/* RFC2590 */
2061 		return (64);
2062 	case IFT_IEEE1394:	/* RFC3146 */
2063 		return (64);
2064 	case IFT_GIF:
2065 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2066 	case IFT_LOOP:
2067 		return (64);	/* XXX: is this really correct? */
2068 	default:
2069 		/*
2070 		 * Unknown link type:
2071 		 * It might be controversial to use the today's common constant
2072 		 * of 64 for these cases unconditionally.  For full compliance,
2073 		 * we should return an error in this case.  On the other hand,
2074 		 * if we simply miss the standard for the link type or a new
2075 		 * standard is defined for a new link type, the IFID length
2076 		 * is very likely to be the common constant.  As a compromise,
2077 		 * we always use the constant, but make an explicit notice
2078 		 * indicating the "unknown" case.
2079 		 */
2080 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2081 		return (64);
2082 	}
2083 }
2084 
2085 struct in6_llentry {
2086 	struct llentry		base;
2087 };
2088 
2089 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2090 #define	IN6_LLTBL_HASH(k, h) \
2091 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2092 
2093 /*
2094  * Do actual deallocation of @lle.
2095  */
2096 static void
in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)2097 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2098 {
2099 	struct llentry *lle;
2100 
2101 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2102 	LLE_LOCK_DESTROY(lle);
2103 	LLE_REQ_DESTROY(lle);
2104 	free(lle, M_LLTABLE);
2105 }
2106 
2107 /*
2108  * Called by LLE_FREE_LOCKED when number of references
2109  * drops to zero.
2110  */
2111 static void
in6_lltable_destroy_lle(struct llentry * lle)2112 in6_lltable_destroy_lle(struct llentry *lle)
2113 {
2114 
2115 	LLE_WUNLOCK(lle);
2116 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2117 }
2118 
2119 static struct llentry *
in6_lltable_new(const struct in6_addr * addr6,u_int flags)2120 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2121 {
2122 	struct in6_llentry *lle;
2123 
2124 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2125 	if (lle == NULL)		/* NB: caller generates msg */
2126 		return NULL;
2127 
2128 	lle->base.r_l3addr.addr6 = *addr6;
2129 	lle->base.lle_refcnt = 1;
2130 	lle->base.lle_free = in6_lltable_destroy_lle;
2131 	LLE_LOCK_INIT(&lle->base);
2132 	LLE_REQ_INIT(&lle->base);
2133 	callout_init(&lle->base.lle_timer, 1);
2134 
2135 	return (&lle->base);
2136 }
2137 
2138 static int
in6_lltable_match_prefix(const struct sockaddr * saddr,const struct sockaddr * smask,u_int flags,struct llentry * lle)2139 in6_lltable_match_prefix(const struct sockaddr *saddr,
2140     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2141 {
2142 	const struct in6_addr *addr, *mask, *lle_addr;
2143 
2144 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2145 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2146 	lle_addr = &lle->r_l3addr.addr6;
2147 
2148 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2149 		return (0);
2150 
2151 	if (lle->la_flags & LLE_IFADDR) {
2152 		/*
2153 		 * Delete LLE_IFADDR records IFF address & flag matches.
2154 		 * Note that addr is the interface address within prefix
2155 		 * being matched.
2156 		 */
2157 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2158 		    (flags & LLE_STATIC) != 0)
2159 			return (1);
2160 		return (0);
2161 	}
2162 
2163 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2164 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2165 		return (1);
2166 
2167 	return (0);
2168 }
2169 
2170 static void
in6_lltable_free_entry(struct lltable * llt,struct llentry * lle)2171 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2172 {
2173 	struct ifnet *ifp;
2174 
2175 	LLE_WLOCK_ASSERT(lle);
2176 	KASSERT(llt != NULL, ("lltable is NULL"));
2177 
2178 	/* Unlink entry from table */
2179 	if ((lle->la_flags & LLE_LINKED) != 0) {
2180 		ifp = llt->llt_ifp;
2181 		IF_AFDATA_WLOCK_ASSERT(ifp);
2182 		lltable_unlink_entry(llt, lle);
2183 	}
2184 
2185 	llentry_free(lle);
2186 }
2187 
2188 static int
in6_lltable_rtcheck(struct ifnet * ifp,u_int flags,const struct sockaddr * l3addr)2189 in6_lltable_rtcheck(struct ifnet *ifp,
2190 		    u_int flags,
2191 		    const struct sockaddr *l3addr)
2192 {
2193 	const struct sockaddr_in6 *sin6;
2194 	struct nhop_object *nh;
2195 	struct in6_addr dst;
2196 	uint32_t scopeid;
2197 	char ip6buf[INET6_ADDRSTRLEN];
2198 	int fibnum;
2199 
2200 	NET_EPOCH_ASSERT();
2201 	KASSERT(l3addr->sa_family == AF_INET6,
2202 	    ("sin_family %d", l3addr->sa_family));
2203 
2204 	sin6 = (const struct sockaddr_in6 *)l3addr;
2205 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2206 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2207 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2208 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2209 		struct ifaddr *ifa;
2210 		/*
2211 		 * Create an ND6 cache for an IPv6 neighbor
2212 		 * that is not covered by our own prefix.
2213 		 */
2214 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2215 		if (ifa != NULL) {
2216 			return 0;
2217 		}
2218 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2219 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2220 		return EINVAL;
2221 	}
2222 	return 0;
2223 }
2224 
2225 static inline uint32_t
in6_lltable_hash_dst(const struct in6_addr * dst,uint32_t hsize)2226 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2227 {
2228 
2229 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2230 }
2231 
2232 static uint32_t
in6_lltable_hash(const struct llentry * lle,uint32_t hsize)2233 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2234 {
2235 
2236 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2237 }
2238 
2239 static void
in6_lltable_fill_sa_entry(const struct llentry * lle,struct sockaddr * sa)2240 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2241 {
2242 	struct sockaddr_in6 *sin6;
2243 
2244 	sin6 = (struct sockaddr_in6 *)sa;
2245 	bzero(sin6, sizeof(*sin6));
2246 	sin6->sin6_family = AF_INET6;
2247 	sin6->sin6_len = sizeof(*sin6);
2248 	sin6->sin6_addr = lle->r_l3addr.addr6;
2249 }
2250 
2251 static inline struct llentry *
in6_lltable_find_dst(struct lltable * llt,const struct in6_addr * dst)2252 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2253 {
2254 	struct llentry *lle;
2255 	struct llentries *lleh;
2256 	u_int hashidx;
2257 
2258 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2259 	lleh = &llt->lle_head[hashidx];
2260 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2261 		if (lle->la_flags & LLE_DELETED)
2262 			continue;
2263 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2264 			break;
2265 	}
2266 
2267 	return (lle);
2268 }
2269 
2270 static void
in6_lltable_delete_entry(struct lltable * llt,struct llentry * lle)2271 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2272 {
2273 
2274 	lle->la_flags |= LLE_DELETED;
2275 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2276 #ifdef DIAGNOSTIC
2277 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2278 #endif
2279 	llentry_free(lle);
2280 }
2281 
2282 static struct llentry *
in6_lltable_alloc(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)2283 in6_lltable_alloc(struct lltable *llt, u_int flags,
2284 	const struct sockaddr *l3addr)
2285 {
2286 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2287 	struct ifnet *ifp = llt->llt_ifp;
2288 	struct llentry *lle;
2289 	char linkhdr[LLE_MAX_LINKHDR];
2290 	size_t linkhdrsize;
2291 	int lladdr_off;
2292 
2293 	KASSERT(l3addr->sa_family == AF_INET6,
2294 	    ("sin_family %d", l3addr->sa_family));
2295 
2296 	/*
2297 	 * A route that covers the given address must have
2298 	 * been installed 1st because we are doing a resolution,
2299 	 * verify this.
2300 	 */
2301 	if (!(flags & LLE_IFADDR) &&
2302 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2303 		return (NULL);
2304 
2305 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2306 	if (lle == NULL) {
2307 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2308 		return (NULL);
2309 	}
2310 	lle->la_flags = flags;
2311 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2312 		linkhdrsize = LLE_MAX_LINKHDR;
2313 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2314 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2315 			in6_lltable_free_entry(llt, lle);
2316 			return (NULL);
2317 		}
2318 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2319 		    lladdr_off);
2320 		lle->la_flags |= LLE_STATIC;
2321 	}
2322 
2323 	if ((lle->la_flags & LLE_STATIC) != 0)
2324 		lle->ln_state = ND6_LLINFO_REACHABLE;
2325 
2326 	return (lle);
2327 }
2328 
2329 static struct llentry *
in6_lltable_lookup(struct lltable * llt,u_int flags,const struct sockaddr * l3addr)2330 in6_lltable_lookup(struct lltable *llt, u_int flags,
2331 	const struct sockaddr *l3addr)
2332 {
2333 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2334 	int family = flags >> 16;
2335 	struct llentry *lle;
2336 
2337 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2338 	KASSERT(l3addr->sa_family == AF_INET6,
2339 	    ("sin_family %d", l3addr->sa_family));
2340 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2341 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2342 	    ("wrong lle request flags: %#x", flags));
2343 
2344 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2345 
2346 	if (__predict_false(family != AF_INET6))
2347 		lle = llentry_lookup_family(lle, family);
2348 
2349 	if (lle == NULL)
2350 		return (NULL);
2351 
2352 	if (flags & LLE_UNLOCKED)
2353 		return (lle);
2354 
2355 	if (flags & LLE_EXCLUSIVE)
2356 		LLE_WLOCK(lle);
2357 	else
2358 		LLE_RLOCK(lle);
2359 
2360 	/*
2361 	 * If the afdata lock is not held, the LLE may have been unlinked while
2362 	 * we were blocked on the LLE lock.  Check for this case.
2363 	 */
2364 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2365 		if (flags & LLE_EXCLUSIVE)
2366 			LLE_WUNLOCK(lle);
2367 		else
2368 			LLE_RUNLOCK(lle);
2369 		return (NULL);
2370 	}
2371 	return (lle);
2372 }
2373 
2374 static int
in6_lltable_dump_entry(struct lltable * llt,struct llentry * lle,struct sysctl_req * wr)2375 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2376     struct sysctl_req *wr)
2377 {
2378 	struct ifnet *ifp = llt->llt_ifp;
2379 	/* XXX stack use */
2380 	struct {
2381 		struct rt_msghdr	rtm;
2382 		struct sockaddr_in6	sin6;
2383 		/*
2384 		 * ndp.c assumes that sdl is word aligned
2385 		 */
2386 #ifdef __LP64__
2387 		uint32_t		pad;
2388 #endif
2389 		struct sockaddr_dl	sdl;
2390 	} ndpc;
2391 	struct sockaddr_dl *sdl;
2392 	int error;
2393 
2394 	bzero(&ndpc, sizeof(ndpc));
2395 	/* skip deleted entries */
2396 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2397 		return (0);
2398 	/* Skip if jailed and not a valid IP of the prison. */
2399 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2400 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2401 		return (0);
2402 	/*
2403 	 * produce a msg made of:
2404 	 *  struct rt_msghdr;
2405 	 *  struct sockaddr_in6 (IPv6)
2406 	 *  struct sockaddr_dl;
2407 	 */
2408 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2409 	ndpc.rtm.rtm_version = RTM_VERSION;
2410 	ndpc.rtm.rtm_type = RTM_GET;
2411 	ndpc.rtm.rtm_flags = RTF_UP;
2412 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2413 	sa6_recoverscope(&ndpc.sin6);
2414 
2415 	/* publish */
2416 	if (lle->la_flags & LLE_PUB)
2417 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2418 
2419 	sdl = &ndpc.sdl;
2420 	sdl->sdl_family = AF_LINK;
2421 	sdl->sdl_len = sizeof(*sdl);
2422 	sdl->sdl_index = ifp->if_index;
2423 	sdl->sdl_type = ifp->if_type;
2424 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2425 		sdl->sdl_alen = ifp->if_addrlen;
2426 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2427 	} else {
2428 		sdl->sdl_alen = 0;
2429 		bzero(LLADDR(sdl), ifp->if_addrlen);
2430 	}
2431 	if (lle->la_expire != 0)
2432 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2433 		    lle->lle_remtime / hz + time_second - time_uptime;
2434 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2435 	if (lle->la_flags & LLE_STATIC)
2436 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2437 	if (lle->la_flags & LLE_IFADDR)
2438 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2439 	if (lle->ln_router != 0)
2440 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2441 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2442 	/* Store state in rmx_weight value */
2443 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2444 	ndpc.rtm.rtm_index = ifp->if_index;
2445 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2446 
2447 	return (error);
2448 }
2449 
2450 static struct lltable *
in6_lltattach(struct ifnet * ifp)2451 in6_lltattach(struct ifnet *ifp)
2452 {
2453 	struct lltable *llt;
2454 
2455 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2456 	llt->llt_af = AF_INET6;
2457 	llt->llt_ifp = ifp;
2458 
2459 	llt->llt_lookup = in6_lltable_lookup;
2460 	llt->llt_alloc_entry = in6_lltable_alloc;
2461 	llt->llt_delete_entry = in6_lltable_delete_entry;
2462 	llt->llt_dump_entry = in6_lltable_dump_entry;
2463 	llt->llt_hash = in6_lltable_hash;
2464 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2465 	llt->llt_free_entry = in6_lltable_free_entry;
2466 	llt->llt_match_prefix = in6_lltable_match_prefix;
2467 	llt->llt_mark_used = llentry_mark_used;
2468  	lltable_link(llt);
2469 
2470 	return (llt);
2471 }
2472 
2473 struct lltable *
in6_lltable_get(struct ifnet * ifp)2474 in6_lltable_get(struct ifnet *ifp)
2475 {
2476 	struct lltable *llt = NULL;
2477 
2478 	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2479 	if (afdata_ptr != NULL)
2480 		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2481 	return (llt);
2482 }
2483 
2484 void *
in6_domifattach(struct ifnet * ifp)2485 in6_domifattach(struct ifnet *ifp)
2486 {
2487 	struct in6_ifextra *ext;
2488 
2489 	/* There are not IPv6-capable interfaces. */
2490 	switch (ifp->if_type) {
2491 	case IFT_PFLOG:
2492 	case IFT_PFSYNC:
2493 	case IFT_USB:
2494 		return (NULL);
2495 	}
2496 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2497 	bzero(ext, sizeof(*ext));
2498 
2499 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2500 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2501 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2502 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2503 
2504 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2505 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2506 	    M_WAITOK);
2507 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2508 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2509 
2510 	ext->nd_ifinfo = nd6_ifattach(ifp);
2511 	ext->scope6_id = scope6_ifattach(ifp);
2512 	ext->lltable = in6_lltattach(ifp);
2513 
2514 	ext->mld_ifinfo = mld_domifattach(ifp);
2515 
2516 	return ext;
2517 }
2518 
2519 int
in6_domifmtu(struct ifnet * ifp)2520 in6_domifmtu(struct ifnet *ifp)
2521 {
2522 	if (ifp->if_afdata[AF_INET6] == NULL)
2523 		return ifp->if_mtu;
2524 
2525 	return (IN6_LINKMTU(ifp));
2526 }
2527 
2528 void
in6_domifdetach(struct ifnet * ifp,void * aux)2529 in6_domifdetach(struct ifnet *ifp, void *aux)
2530 {
2531 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2532 
2533 	mld_domifdetach(ifp);
2534 	scope6_ifdetach(ext->scope6_id);
2535 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2536 	lltable_free(ext->lltable);
2537 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2538 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2539 	free(ext->in6_ifstat, M_IFADDR);
2540 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2541 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2542 	free(ext->icmp6_ifstat, M_IFADDR);
2543 	free(ext, M_IFADDR);
2544 }
2545 
2546 /*
2547  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2548  * v4 mapped addr or v4 compat addr
2549  */
2550 void
in6_sin6_2_sin(struct sockaddr_in * sin,struct sockaddr_in6 * sin6)2551 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2552 {
2553 
2554 	bzero(sin, sizeof(*sin));
2555 	sin->sin_len = sizeof(struct sockaddr_in);
2556 	sin->sin_family = AF_INET;
2557 	sin->sin_port = sin6->sin6_port;
2558 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2559 }
2560 
2561 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2562 void
in6_sin_2_v4mapsin6(struct sockaddr_in * sin,struct sockaddr_in6 * sin6)2563 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2564 {
2565 	bzero(sin6, sizeof(*sin6));
2566 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2567 	sin6->sin6_family = AF_INET6;
2568 	sin6->sin6_port = sin->sin_port;
2569 	sin6->sin6_addr.s6_addr32[0] = 0;
2570 	sin6->sin6_addr.s6_addr32[1] = 0;
2571 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2572 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2573 }
2574 
2575 /* Convert sockaddr_in6 into sockaddr_in. */
2576 void
in6_sin6_2_sin_in_sock(struct sockaddr * nam)2577 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2578 {
2579 	struct sockaddr_in *sin_p;
2580 	struct sockaddr_in6 sin6;
2581 
2582 	/*
2583 	 * Save original sockaddr_in6 addr and convert it
2584 	 * to sockaddr_in.
2585 	 */
2586 	sin6 = *(struct sockaddr_in6 *)nam;
2587 	sin_p = (struct sockaddr_in *)nam;
2588 	in6_sin6_2_sin(sin_p, &sin6);
2589 }
2590 
2591 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2592 void
in6_sin_2_v4mapsin6_in_sock(struct sockaddr ** nam)2593 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2594 {
2595 	struct sockaddr_in *sin_p;
2596 	struct sockaddr_in6 *sin6_p;
2597 
2598 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2599 	sin_p = (struct sockaddr_in *)*nam;
2600 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2601 	free(*nam, M_SONAME);
2602 	*nam = (struct sockaddr *)sin6_p;
2603 }
2604