xref: /freebsd/sys/netinet/in.c (revision c3f8f86efd5fdbf5701d41c47a7fc4cd3ccdc802)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * Copyright (C) 2001 WIDE Project.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)in.c	8.4 (Berkeley) 1/9/95
33  */
34 
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
37 
38 #include <sys/param.h>
39 #include <sys/eventhandler.h>
40 #include <sys/systm.h>
41 #include <sys/sockio.h>
42 #include <sys/malloc.h>
43 #include <sys/priv.h>
44 #include <sys/socket.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/proc.h>
49 #include <sys/rmlock.h>
50 #include <sys/sysctl.h>
51 #include <sys/syslog.h>
52 #include <sys/sx.h>
53 
54 #include <net/if.h>
55 #include <net/if_var.h>
56 #include <net/if_arp.h>
57 #include <net/if_dl.h>
58 #include <net/if_llatbl.h>
59 #include <net/if_types.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/if_ether.h>
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <netinet/in_pcb.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/ip_carp.h>
69 #include <netinet/igmp_var.h>
70 #include <netinet/udp.h>
71 #include <netinet/udp_var.h>
72 
73 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
74 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
75 
76 static void	in_socktrim(struct sockaddr_in *);
77 static void	in_purgemaddrs(struct ifnet *);
78 
79 VNET_DEFINE_STATIC(int, nosameprefix);
80 #define	V_nosameprefix			VNET(nosameprefix)
81 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
82 	&VNET_NAME(nosameprefix), 0,
83 	"Refuse to create same prefixes on different interfaces");
84 
85 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
86 #define	V_ripcbinfo			VNET(ripcbinfo)
87 
88 static struct sx in_control_sx;
89 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
90 
91 /*
92  * Return 1 if an internet address is for a ``local'' host
93  * (one to which we have a connection).
94  */
95 int
96 in_localaddr(struct in_addr in)
97 {
98 	struct rm_priotracker in_ifa_tracker;
99 	u_long i = ntohl(in.s_addr);
100 	struct in_ifaddr *ia;
101 
102 	IN_IFADDR_RLOCK(&in_ifa_tracker);
103 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
104 		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
105 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
106 			return (1);
107 		}
108 	}
109 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
110 	return (0);
111 }
112 
113 /*
114  * Return 1 if an internet address is for the local host and configured
115  * on one of its interfaces.
116  */
117 int
118 in_localip(struct in_addr in)
119 {
120 	struct rm_priotracker in_ifa_tracker;
121 	struct in_ifaddr *ia;
122 
123 	IN_IFADDR_RLOCK(&in_ifa_tracker);
124 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
125 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
126 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
127 			return (1);
128 		}
129 	}
130 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
131 	return (0);
132 }
133 
134 /*
135  * Return 1 if an internet address is configured on an interface.
136  */
137 int
138 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
139 {
140 	struct ifaddr *ifa;
141 	struct in_ifaddr *ia;
142 
143 	NET_EPOCH_ASSERT();
144 
145 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
146 		if (ifa->ifa_addr->sa_family != AF_INET)
147 			continue;
148 		ia = (struct in_ifaddr *)ifa;
149 		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
150 			return (1);
151 	}
152 
153 	return (0);
154 }
155 
156 /*
157  * Return a reference to the interface address which is different to
158  * the supplied one but with same IP address value.
159  */
160 static struct in_ifaddr *
161 in_localip_more(struct in_ifaddr *ia)
162 {
163 	struct rm_priotracker in_ifa_tracker;
164 	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
165 	struct in_ifaddr *it;
166 
167 	IN_IFADDR_RLOCK(&in_ifa_tracker);
168 	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
169 		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
170 			ifa_ref(&it->ia_ifa);
171 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
172 			return (it);
173 		}
174 	}
175 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
176 
177 	return (NULL);
178 }
179 
180 /*
181  * Determine whether an IP address is in a reserved set of addresses
182  * that may not be forwarded, or whether datagrams to that destination
183  * may be forwarded.
184  */
185 int
186 in_canforward(struct in_addr in)
187 {
188 	u_long i = ntohl(in.s_addr);
189 
190 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
191 	    IN_ZERONET(i) || IN_LOOPBACK(i))
192 		return (0);
193 	return (1);
194 }
195 
196 /*
197  * Trim a mask in a sockaddr
198  */
199 static void
200 in_socktrim(struct sockaddr_in *ap)
201 {
202     char *cplim = (char *) &ap->sin_addr;
203     char *cp = (char *) (&ap->sin_addr + 1);
204 
205     ap->sin_len = 0;
206     while (--cp >= cplim)
207 	if (*cp) {
208 	    (ap)->sin_len = cp - (char *) (ap) + 1;
209 	    break;
210 	}
211 }
212 
213 /*
214  * Generic internet control operations (ioctl's).
215  */
216 int
217 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
218     struct thread *td)
219 {
220 	struct ifreq *ifr = (struct ifreq *)data;
221 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
222 	struct epoch_tracker et;
223 	struct ifaddr *ifa;
224 	struct in_ifaddr *ia;
225 	int error;
226 
227 	if (ifp == NULL)
228 		return (EADDRNOTAVAIL);
229 
230 	/*
231 	 * Filter out 4 ioctls we implement directly.  Forward the rest
232 	 * to specific functions and ifp->if_ioctl().
233 	 */
234 	switch (cmd) {
235 	case SIOCGIFADDR:
236 	case SIOCGIFBRDADDR:
237 	case SIOCGIFDSTADDR:
238 	case SIOCGIFNETMASK:
239 		break;
240 	case SIOCDIFADDR:
241 		sx_xlock(&in_control_sx);
242 		error = in_difaddr_ioctl(cmd, data, ifp, td);
243 		sx_xunlock(&in_control_sx);
244 		return (error);
245 	case OSIOCAIFADDR:	/* 9.x compat */
246 	case SIOCAIFADDR:
247 		sx_xlock(&in_control_sx);
248 		error = in_aifaddr_ioctl(cmd, data, ifp, td);
249 		sx_xunlock(&in_control_sx);
250 		return (error);
251 	case SIOCSIFADDR:
252 	case SIOCSIFBRDADDR:
253 	case SIOCSIFDSTADDR:
254 	case SIOCSIFNETMASK:
255 		/* We no longer support that old commands. */
256 		return (EINVAL);
257 	default:
258 		if (ifp->if_ioctl == NULL)
259 			return (EOPNOTSUPP);
260 		return ((*ifp->if_ioctl)(ifp, cmd, data));
261 	}
262 
263 	if (addr->sin_addr.s_addr != INADDR_ANY &&
264 	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
265 		return (EADDRNOTAVAIL);
266 
267 	/*
268 	 * Find address for this interface, if it exists.  If an
269 	 * address was specified, find that one instead of the
270 	 * first one on the interface, if possible.
271 	 */
272 	NET_EPOCH_ENTER(et);
273 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
274 		if (ifa->ifa_addr->sa_family != AF_INET)
275 			continue;
276 		ia = (struct in_ifaddr *)ifa;
277 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
278 			break;
279 	}
280 	if (ifa == NULL)
281 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
282 			if (ifa->ifa_addr->sa_family == AF_INET) {
283 				ia = (struct in_ifaddr *)ifa;
284 				if (prison_check_ip4(td->td_ucred,
285 				    &ia->ia_addr.sin_addr) == 0)
286 					break;
287 			}
288 
289 	if (ifa == NULL) {
290 		NET_EPOCH_EXIT(et);
291 		return (EADDRNOTAVAIL);
292 	}
293 
294 	error = 0;
295 	switch (cmd) {
296 	case SIOCGIFADDR:
297 		*addr = ia->ia_addr;
298 		break;
299 
300 	case SIOCGIFBRDADDR:
301 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
302 			error = EINVAL;
303 			break;
304 		}
305 		*addr = ia->ia_broadaddr;
306 		break;
307 
308 	case SIOCGIFDSTADDR:
309 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
310 			error = EINVAL;
311 			break;
312 		}
313 		*addr = ia->ia_dstaddr;
314 		break;
315 
316 	case SIOCGIFNETMASK:
317 		*addr = ia->ia_sockmask;
318 		break;
319 	}
320 
321 	NET_EPOCH_EXIT(et);
322 
323 	return (error);
324 }
325 
326 static int
327 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
328 {
329 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
330 	const struct sockaddr_in *addr = &ifra->ifra_addr;
331 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
332 	const struct sockaddr_in *mask = &ifra->ifra_mask;
333 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
334 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
335 	struct epoch_tracker et;
336 	struct ifaddr *ifa;
337 	struct in_ifaddr *ia;
338 	bool iaIsFirst;
339 	int error = 0;
340 
341 	error = priv_check(td, PRIV_NET_ADDIFADDR);
342 	if (error)
343 		return (error);
344 
345 	/*
346 	 * ifra_addr must be present and be of INET family.
347 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
348 	 */
349 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
350 	    addr->sin_family != AF_INET)
351 		return (EINVAL);
352 	if (broadaddr->sin_len != 0 &&
353 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
354 	    broadaddr->sin_family != AF_INET))
355 		return (EINVAL);
356 	if (mask->sin_len != 0 &&
357 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
358 	    mask->sin_family != AF_INET))
359 		return (EINVAL);
360 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
361 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
362 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
363 		return (EDESTADDRREQ);
364 	if (vhid > 0 && carp_attach_p == NULL)
365 		return (EPROTONOSUPPORT);
366 
367 	/*
368 	 * See whether address already exist.
369 	 */
370 	iaIsFirst = true;
371 	ia = NULL;
372 	NET_EPOCH_ENTER(et);
373 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
374 		struct in_ifaddr *it;
375 
376 		if (ifa->ifa_addr->sa_family != AF_INET)
377 			continue;
378 
379 		it = (struct in_ifaddr *)ifa;
380 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
381 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
382 			ia = it;
383 		else
384 			iaIsFirst = false;
385 	}
386 	NET_EPOCH_EXIT(et);
387 
388 	if (ia != NULL)
389 		(void )in_difaddr_ioctl(cmd, data, ifp, td);
390 
391 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
392 	ia = (struct in_ifaddr *)ifa;
393 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
394 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
395 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
396 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
397 	    CALLOUT_RETURNUNLOCKED);
398 
399 	ia->ia_ifp = ifp;
400 	ia->ia_addr = *addr;
401 	if (mask->sin_len != 0) {
402 		ia->ia_sockmask = *mask;
403 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
404 	} else {
405 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
406 
407 		/*
408 	 	 * Be compatible with network classes, if netmask isn't
409 		 * supplied, guess it based on classes.
410 	 	 */
411 		if (IN_CLASSA(i))
412 			ia->ia_subnetmask = IN_CLASSA_NET;
413 		else if (IN_CLASSB(i))
414 			ia->ia_subnetmask = IN_CLASSB_NET;
415 		else
416 			ia->ia_subnetmask = IN_CLASSC_NET;
417 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
418 	}
419 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
420 	in_socktrim(&ia->ia_sockmask);
421 
422 	if (ifp->if_flags & IFF_BROADCAST) {
423 		if (broadaddr->sin_len != 0) {
424 			ia->ia_broadaddr = *broadaddr;
425 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
426 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
427 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
428 			ia->ia_broadaddr.sin_family = AF_INET;
429 		} else {
430 			ia->ia_broadaddr.sin_addr.s_addr =
431 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
432 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
433 			ia->ia_broadaddr.sin_family = AF_INET;
434 		}
435 	}
436 
437 	if (ifp->if_flags & IFF_POINTOPOINT)
438 		ia->ia_dstaddr = *dstaddr;
439 
440 	/* XXXGL: rtinit() needs this strange assignment. */
441 	if (ifp->if_flags & IFF_LOOPBACK)
442                 ia->ia_dstaddr = ia->ia_addr;
443 
444 	if (vhid != 0) {
445 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
446 		if (error)
447 			return (error);
448 	}
449 
450 	/* if_addrhead is already referenced by ifa_alloc() */
451 	IF_ADDR_WLOCK(ifp);
452 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
453 	IF_ADDR_WUNLOCK(ifp);
454 
455 	ifa_ref(ifa);			/* in_ifaddrhead */
456 	IN_IFADDR_WLOCK();
457 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
458 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
459 	IN_IFADDR_WUNLOCK();
460 
461 	/*
462 	 * Give the interface a chance to initialize
463 	 * if this is its first address,
464 	 * and to validate the address if necessary.
465 	 */
466 	if (ifp->if_ioctl != NULL) {
467 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
468 		if (error)
469 			goto fail1;
470 	}
471 
472 	/*
473 	 * Add route for the network.
474 	 */
475 	if (vhid == 0) {
476 		int flags = RTF_UP;
477 
478 		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
479 			flags |= RTF_HOST;
480 
481 		error = in_addprefix(ia, flags);
482 		if (error)
483 			goto fail1;
484 	}
485 
486 	/*
487 	 * Add a loopback route to self.
488 	 */
489 	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
490 	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
491 	    !((ifp->if_flags & IFF_POINTOPOINT) &&
492 	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
493 		struct in_ifaddr *eia;
494 
495 		eia = in_localip_more(ia);
496 
497 		if (eia == NULL) {
498 			error = ifa_add_loopback_route((struct ifaddr *)ia,
499 			    (struct sockaddr *)&ia->ia_addr);
500 			if (error)
501 				goto fail2;
502 		} else
503 			ifa_free(&eia->ia_ifa);
504 	}
505 
506 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
507 		struct in_addr allhosts_addr;
508 		struct in_ifinfo *ii;
509 
510 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
511 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
512 
513 		error = in_joingroup(ifp, &allhosts_addr, NULL,
514 			&ii->ii_allhosts);
515 	}
516 
517 	/*
518 	 * Note: we don't need extra reference for ifa, since we called
519 	 * with sx lock held, and ifaddr can not be deleted in concurrent
520 	 * thread.
521 	 */
522 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
523 
524 	return (error);
525 
526 fail2:
527 	if (vhid == 0)
528 		(void )in_scrubprefix(ia, LLE_STATIC);
529 
530 fail1:
531 	if (ia->ia_ifa.ifa_carp)
532 		(*carp_detach_p)(&ia->ia_ifa, false);
533 
534 	IF_ADDR_WLOCK(ifp);
535 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
536 	IF_ADDR_WUNLOCK(ifp);
537 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
538 
539 	IN_IFADDR_WLOCK();
540 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
541 	LIST_REMOVE(ia, ia_hash);
542 	IN_IFADDR_WUNLOCK();
543 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
544 
545 	return (error);
546 }
547 
548 static int
549 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
550 {
551 	const struct ifreq *ifr = (struct ifreq *)data;
552 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
553 	    &ifr->ifr_addr;
554 	struct ifaddr *ifa;
555 	struct in_ifaddr *ia;
556 	bool deleteAny, iaIsLast;
557 	int error;
558 
559 	if (td != NULL) {
560 		error = priv_check(td, PRIV_NET_DELIFADDR);
561 		if (error)
562 			return (error);
563 	}
564 
565 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
566 	    addr->sin_family != AF_INET)
567 		deleteAny = true;
568 	else
569 		deleteAny = false;
570 
571 	iaIsLast = true;
572 	ia = NULL;
573 	IF_ADDR_WLOCK(ifp);
574 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
575 		struct in_ifaddr *it;
576 
577 		if (ifa->ifa_addr->sa_family != AF_INET)
578 			continue;
579 
580 		it = (struct in_ifaddr *)ifa;
581 		if (deleteAny && ia == NULL && (td == NULL ||
582 		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
583 			ia = it;
584 
585 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
586 		    (td == NULL || prison_check_ip4(td->td_ucred,
587 		    &addr->sin_addr) == 0))
588 			ia = it;
589 
590 		if (it != ia)
591 			iaIsLast = false;
592 	}
593 
594 	if (ia == NULL) {
595 		IF_ADDR_WUNLOCK(ifp);
596 		return (EADDRNOTAVAIL);
597 	}
598 
599 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
600 	IF_ADDR_WUNLOCK(ifp);
601 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
602 
603 	IN_IFADDR_WLOCK();
604 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
605 	LIST_REMOVE(ia, ia_hash);
606 	IN_IFADDR_WUNLOCK();
607 
608 	/*
609 	 * in_scrubprefix() kills the interface route.
610 	 */
611 	in_scrubprefix(ia, LLE_STATIC);
612 
613 	/*
614 	 * in_ifadown gets rid of all the rest of
615 	 * the routes.  This is not quite the right
616 	 * thing to do, but at least if we are running
617 	 * a routing process they will come back.
618 	 */
619 	in_ifadown(&ia->ia_ifa, 1);
620 
621 	if (ia->ia_ifa.ifa_carp)
622 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
623 
624 	/*
625 	 * If this is the last IPv4 address configured on this
626 	 * interface, leave the all-hosts group.
627 	 * No state-change report need be transmitted.
628 	 */
629 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
630 		struct in_ifinfo *ii;
631 
632 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
633 		if (ii->ii_allhosts) {
634 			(void)in_leavegroup(ii->ii_allhosts, NULL);
635 			ii->ii_allhosts = NULL;
636 		}
637 	}
638 
639 	IF_ADDR_WLOCK(ifp);
640 	if (callout_stop(&ia->ia_garp_timer) == 1) {
641 		ifa_free(&ia->ia_ifa);
642 	}
643 	IF_ADDR_WUNLOCK(ifp);
644 
645 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
646 	    IFADDR_EVENT_DEL);
647 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
648 
649 	return (0);
650 }
651 
652 #define rtinitflags(x) \
653 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
654 	    ? RTF_HOST : 0)
655 
656 /*
657  * Check if we have a route for the given prefix already or add one accordingly.
658  */
659 int
660 in_addprefix(struct in_ifaddr *target, int flags)
661 {
662 	struct rm_priotracker in_ifa_tracker;
663 	struct in_ifaddr *ia;
664 	struct in_addr prefix, mask, p, m;
665 	int error;
666 
667 	if ((flags & RTF_HOST) != 0) {
668 		prefix = target->ia_dstaddr.sin_addr;
669 		mask.s_addr = 0;
670 	} else {
671 		prefix = target->ia_addr.sin_addr;
672 		mask = target->ia_sockmask.sin_addr;
673 		prefix.s_addr &= mask.s_addr;
674 	}
675 
676 	IN_IFADDR_RLOCK(&in_ifa_tracker);
677 	/* Look for an existing address with the same prefix, mask, and fib */
678 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
679 		if (rtinitflags(ia)) {
680 			p = ia->ia_dstaddr.sin_addr;
681 
682 			if (prefix.s_addr != p.s_addr)
683 				continue;
684 		} else {
685 			p = ia->ia_addr.sin_addr;
686 			m = ia->ia_sockmask.sin_addr;
687 			p.s_addr &= m.s_addr;
688 
689 			if (prefix.s_addr != p.s_addr ||
690 			    mask.s_addr != m.s_addr)
691 				continue;
692 		}
693 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
694 			continue;
695 
696 		/*
697 		 * If we got a matching prefix route inserted by other
698 		 * interface address, we are done here.
699 		 */
700 		if (ia->ia_flags & IFA_ROUTE) {
701 			if (V_nosameprefix) {
702 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
703 				return (EEXIST);
704 			} else {
705 				int fibnum;
706 
707 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
708 					target->ia_ifp->if_fib;
709 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
710 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
711 				return (0);
712 			}
713 		}
714 	}
715 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
716 
717 	/*
718 	 * No-one seem to have this prefix route, so we try to insert it.
719 	 */
720 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
721 	if (!error)
722 		target->ia_flags |= IFA_ROUTE;
723 	return (error);
724 }
725 
726 /*
727  * Removes either all lle entries for given @ia, or lle
728  * corresponding to @ia address.
729  */
730 static void
731 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
732 {
733 	struct sockaddr_in addr, mask;
734 	struct sockaddr *saddr, *smask;
735 	struct ifnet *ifp;
736 
737 	saddr = (struct sockaddr *)&addr;
738 	bzero(&addr, sizeof(addr));
739 	addr.sin_len = sizeof(addr);
740 	addr.sin_family = AF_INET;
741 	smask = (struct sockaddr *)&mask;
742 	bzero(&mask, sizeof(mask));
743 	mask.sin_len = sizeof(mask);
744 	mask.sin_family = AF_INET;
745 	mask.sin_addr.s_addr = ia->ia_subnetmask;
746 	ifp = ia->ia_ifp;
747 
748 	if (all) {
749 		/*
750 		 * Remove all L2 entries matching given prefix.
751 		 * Convert address to host representation to avoid
752 		 * doing this on every callback. ia_subnetmask is already
753 		 * stored in host representation.
754 		 */
755 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
756 		lltable_prefix_free(AF_INET, saddr, smask, flags);
757 	} else {
758 		/* Remove interface address only */
759 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
760 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
761 	}
762 }
763 
764 /*
765  * If there is no other address in the system that can serve a route to the
766  * same prefix, remove the route.  Hand over the route to the new address
767  * otherwise.
768  */
769 int
770 in_scrubprefix(struct in_ifaddr *target, u_int flags)
771 {
772 	struct rm_priotracker in_ifa_tracker;
773 	struct in_ifaddr *ia;
774 	struct in_addr prefix, mask, p, m;
775 	int error = 0;
776 
777 	/*
778 	 * Remove the loopback route to the interface address.
779 	 */
780 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
781 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
782 	    (flags & LLE_STATIC)) {
783 		struct in_ifaddr *eia;
784 
785 		/*
786 		 * XXXME: add fib-aware in_localip.
787 		 * We definitely don't want to switch between
788 		 * prefixes in different fibs.
789 		 */
790 		eia = in_localip_more(target);
791 
792 		if (eia != NULL) {
793 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
794 			    (struct sockaddr *)&target->ia_addr);
795 			ifa_free(&eia->ia_ifa);
796 		} else {
797 			error = ifa_del_loopback_route((struct ifaddr *)target,
798 			    (struct sockaddr *)&target->ia_addr);
799 		}
800 	}
801 
802 	if (rtinitflags(target)) {
803 		prefix = target->ia_dstaddr.sin_addr;
804 		mask.s_addr = 0;
805 	} else {
806 		prefix = target->ia_addr.sin_addr;
807 		mask = target->ia_sockmask.sin_addr;
808 		prefix.s_addr &= mask.s_addr;
809 	}
810 
811 	if ((target->ia_flags & IFA_ROUTE) == 0) {
812 		int fibnum;
813 
814 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
815 			target->ia_ifp->if_fib;
816 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
817 
818 		/*
819 		 * Removing address from !IFF_UP interface or
820 		 * prefix which exists on other interface (along with route).
821 		 * No entries should exist here except target addr.
822 		 * Given that, delete this entry only.
823 		 */
824 		in_scrubprefixlle(target, 0, flags);
825 		return (0);
826 	}
827 
828 	IN_IFADDR_RLOCK(&in_ifa_tracker);
829 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
830 		if (rtinitflags(ia)) {
831 			p = ia->ia_dstaddr.sin_addr;
832 
833 			if (prefix.s_addr != p.s_addr)
834 				continue;
835 		} else {
836 			p = ia->ia_addr.sin_addr;
837 			m = ia->ia_sockmask.sin_addr;
838 			p.s_addr &= m.s_addr;
839 
840 			if (prefix.s_addr != p.s_addr ||
841 			    mask.s_addr != m.s_addr)
842 				continue;
843 		}
844 
845 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
846 			continue;
847 
848 		/*
849 		 * If we got a matching prefix address, move IFA_ROUTE and
850 		 * the route itself to it.  Make sure that routing daemons
851 		 * get a heads-up.
852 		 */
853 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
854 			ifa_ref(&ia->ia_ifa);
855 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
856 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
857 			    rtinitflags(target));
858 			if (error == 0)
859 				target->ia_flags &= ~IFA_ROUTE;
860 			else
861 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
862 					error);
863 			/* Scrub all entries IFF interface is different */
864 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
865 			    flags);
866 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
867 			    rtinitflags(ia) | RTF_UP);
868 			if (error == 0)
869 				ia->ia_flags |= IFA_ROUTE;
870 			else
871 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
872 					error);
873 			ifa_free(&ia->ia_ifa);
874 			return (error);
875 		}
876 	}
877 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
878 
879 	/*
880 	 * remove all L2 entries on the given prefix
881 	 */
882 	in_scrubprefixlle(target, 1, flags);
883 
884 	/*
885 	 * As no-one seem to have this prefix, we can remove the route.
886 	 */
887 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
888 	if (error == 0)
889 		target->ia_flags &= ~IFA_ROUTE;
890 	else
891 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
892 	return (error);
893 }
894 
895 #undef rtinitflags
896 
897 void
898 in_ifscrub_all(void)
899 {
900 	struct ifnet *ifp;
901 	struct ifaddr *ifa, *nifa;
902 	struct ifaliasreq ifr;
903 
904 	IFNET_RLOCK();
905 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
906 		/* Cannot lock here - lock recursion. */
907 		/* NET_EPOCH_ENTER(et); */
908 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
909 			if (ifa->ifa_addr->sa_family != AF_INET)
910 				continue;
911 
912 			/*
913 			 * This is ugly but the only way for legacy IP to
914 			 * cleanly remove addresses and everything attached.
915 			 */
916 			bzero(&ifr, sizeof(ifr));
917 			ifr.ifra_addr = *ifa->ifa_addr;
918 			if (ifa->ifa_dstaddr)
919 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
920 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
921 			    ifp, NULL);
922 		}
923 		/* NET_EPOCH_EXIT(et); */
924 		in_purgemaddrs(ifp);
925 		igmp_domifdetach(ifp);
926 	}
927 	IFNET_RUNLOCK();
928 }
929 
930 int
931 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
932 {
933 
934 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
935 	     /*
936 	      * Check for old-style (host 0) broadcast, but
937 	      * taking into account that RFC 3021 obsoletes it.
938 	      */
939 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
940 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
941 	     /*
942 	      * Check for an all one subnetmask. These
943 	      * only exist when an interface gets a secondary
944 	      * address.
945 	      */
946 	    ia->ia_subnetmask != (u_long)0xffffffff);
947 }
948 
949 /*
950  * Return 1 if the address might be a local broadcast address.
951  */
952 int
953 in_broadcast(struct in_addr in, struct ifnet *ifp)
954 {
955 	struct ifaddr *ifa;
956 	int found;
957 
958 	NET_EPOCH_ASSERT();
959 
960 	if (in.s_addr == INADDR_BROADCAST ||
961 	    in.s_addr == INADDR_ANY)
962 		return (1);
963 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
964 		return (0);
965 	found = 0;
966 	/*
967 	 * Look through the list of addresses for a match
968 	 * with a broadcast address.
969 	 */
970 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
971 		if (ifa->ifa_addr->sa_family == AF_INET &&
972 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
973 			found = 1;
974 			break;
975 		}
976 	return (found);
977 }
978 
979 /*
980  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
981  */
982 void
983 in_ifdetach(struct ifnet *ifp)
984 {
985 	IN_MULTI_LOCK();
986 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
987 	in_pcbpurgeif0(&V_udbinfo, ifp);
988 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
989 	in_purgemaddrs(ifp);
990 	IN_MULTI_UNLOCK();
991 
992 	/*
993 	 * Make sure all multicast deletions invoking if_ioctl() are
994 	 * completed before returning. Else we risk accessing a freed
995 	 * ifnet structure pointer.
996 	 */
997 	inm_release_wait(NULL);
998 }
999 
1000 /*
1001  * Delete all IPv4 multicast address records, and associated link-layer
1002  * multicast address records, associated with ifp.
1003  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1004  * XXX This should not race with ifma_protospec being set during
1005  * a new allocation, if it does, we have bigger problems.
1006  */
1007 static void
1008 in_purgemaddrs(struct ifnet *ifp)
1009 {
1010 	struct in_multi_head purgeinms;
1011 	struct in_multi		*inm;
1012 	struct ifmultiaddr	*ifma, *next;
1013 
1014 	SLIST_INIT(&purgeinms);
1015 	IN_MULTI_LIST_LOCK();
1016 
1017 	/*
1018 	 * Extract list of in_multi associated with the detaching ifp
1019 	 * which the PF_INET layer is about to release.
1020 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1021 	 * by code further down.
1022 	 */
1023 	IF_ADDR_WLOCK(ifp);
1024  restart:
1025 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1026 		if (ifma->ifma_addr->sa_family != AF_INET ||
1027 		    ifma->ifma_protospec == NULL)
1028 			continue;
1029 		inm = (struct in_multi *)ifma->ifma_protospec;
1030 		inm_rele_locked(&purgeinms, inm);
1031 		if (__predict_false(ifma_restart)) {
1032 			ifma_restart = true;
1033 			goto restart;
1034 		}
1035 	}
1036 	IF_ADDR_WUNLOCK(ifp);
1037 
1038 	inm_release_list_deferred(&purgeinms);
1039 	igmp_ifdetach(ifp);
1040 	IN_MULTI_LIST_UNLOCK();
1041 }
1042 
1043 struct in_llentry {
1044 	struct llentry		base;
1045 };
1046 
1047 #define	IN_LLTBL_DEFAULT_HSIZE	32
1048 #define	IN_LLTBL_HASH(k, h) \
1049 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1050 
1051 /*
1052  * Do actual deallocation of @lle.
1053  */
1054 static void
1055 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1056 {
1057 	struct llentry *lle;
1058 
1059 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1060 	LLE_LOCK_DESTROY(lle);
1061 	LLE_REQ_DESTROY(lle);
1062 	free(lle, M_LLTABLE);
1063 }
1064 
1065 /*
1066  * Called by the datapath to indicate that
1067  * the entry was used.
1068  */
1069 static void
1070 in_lltable_mark_used(struct llentry *lle)
1071 {
1072 
1073 	LLE_REQ_LOCK(lle);
1074 	lle->r_skip_req = 0;
1075 	LLE_REQ_UNLOCK(lle);
1076 }
1077 
1078 /*
1079  * Called by LLE_FREE_LOCKED when number of references
1080  * drops to zero.
1081  */
1082 static void
1083 in_lltable_destroy_lle(struct llentry *lle)
1084 {
1085 
1086 	LLE_WUNLOCK(lle);
1087 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1088 }
1089 
1090 static struct llentry *
1091 in_lltable_new(struct in_addr addr4, u_int flags)
1092 {
1093 	struct in_llentry *lle;
1094 
1095 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1096 	if (lle == NULL)		/* NB: caller generates msg */
1097 		return NULL;
1098 
1099 	/*
1100 	 * For IPv4 this will trigger "arpresolve" to generate
1101 	 * an ARP request.
1102 	 */
1103 	lle->base.la_expire = time_uptime; /* mark expired */
1104 	lle->base.r_l3addr.addr4 = addr4;
1105 	lle->base.lle_refcnt = 1;
1106 	lle->base.lle_free = in_lltable_destroy_lle;
1107 	LLE_LOCK_INIT(&lle->base);
1108 	LLE_REQ_INIT(&lle->base);
1109 	callout_init(&lle->base.lle_timer, 1);
1110 
1111 	return (&lle->base);
1112 }
1113 
1114 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1115 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1116 
1117 static int
1118 in_lltable_match_prefix(const struct sockaddr *saddr,
1119     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1120 {
1121 	struct in_addr addr, mask, lle_addr;
1122 
1123 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1124 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1125 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1126 
1127 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1128 		return (0);
1129 
1130 	if (lle->la_flags & LLE_IFADDR) {
1131 		/*
1132 		 * Delete LLE_IFADDR records IFF address & flag matches.
1133 		 * Note that addr is the interface address within prefix
1134 		 * being matched.
1135 		 * Note also we should handle 'ifdown' cases without removing
1136 		 * ifaddr macs.
1137 		 */
1138 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1139 			return (1);
1140 		return (0);
1141 	}
1142 
1143 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1144 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1145 		return (1);
1146 
1147 	return (0);
1148 }
1149 
1150 static void
1151 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1152 {
1153 	size_t pkts_dropped;
1154 
1155 	LLE_WLOCK_ASSERT(lle);
1156 	KASSERT(llt != NULL, ("lltable is NULL"));
1157 
1158 	/* Unlink entry from table if not already */
1159 	if ((lle->la_flags & LLE_LINKED) != 0) {
1160 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1161 		lltable_unlink_entry(llt, lle);
1162 	}
1163 
1164 	/* Drop hold queue */
1165 	pkts_dropped = llentry_free(lle);
1166 	ARPSTAT_ADD(dropped, pkts_dropped);
1167 }
1168 
1169 static int
1170 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1171 {
1172 	struct rt_addrinfo info;
1173 	struct sockaddr_in rt_key, rt_mask;
1174 	struct sockaddr rt_gateway;
1175 	int rt_flags;
1176 
1177 	KASSERT(l3addr->sa_family == AF_INET,
1178 	    ("sin_family %d", l3addr->sa_family));
1179 
1180 	bzero(&rt_key, sizeof(rt_key));
1181 	rt_key.sin_len = sizeof(rt_key);
1182 	bzero(&rt_mask, sizeof(rt_mask));
1183 	rt_mask.sin_len = sizeof(rt_mask);
1184 	bzero(&rt_gateway, sizeof(rt_gateway));
1185 	rt_gateway.sa_len = sizeof(rt_gateway);
1186 
1187 	bzero(&info, sizeof(info));
1188 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1189 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1190 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1191 
1192 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1193 		return (EINVAL);
1194 
1195 	rt_flags = info.rti_flags;
1196 
1197 	/*
1198 	 * If the gateway for an existing host route matches the target L3
1199 	 * address, which is a special route inserted by some implementation
1200 	 * such as MANET, and the interface is of the correct type, then
1201 	 * allow for ARP to proceed.
1202 	 */
1203 	if (rt_flags & RTF_GATEWAY) {
1204 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1205 		    info.rti_ifp->if_type != IFT_ETHER ||
1206 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1207 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1208 		    sizeof(in_addr_t)) != 0) {
1209 			rib_free_info(&info);
1210 			return (EINVAL);
1211 		}
1212 	}
1213 	rib_free_info(&info);
1214 
1215 	/*
1216 	 * Make sure that at least the destination address is covered
1217 	 * by the route. This is for handling the case where 2 or more
1218 	 * interfaces have the same prefix. An incoming packet arrives
1219 	 * on one interface and the corresponding outgoing packet leaves
1220 	 * another interface.
1221 	 */
1222 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1223 		const char *sa, *mask, *addr, *lim;
1224 		const struct sockaddr_in *l3sin;
1225 
1226 		mask = (const char *)&rt_mask;
1227 		/*
1228 		 * Just being extra cautious to avoid some custom
1229 		 * code getting into trouble.
1230 		 */
1231 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1232 			return (EINVAL);
1233 
1234 		sa = (const char *)&rt_key;
1235 		addr = (const char *)l3addr;
1236 		l3sin = (const struct sockaddr_in *)l3addr;
1237 		lim = addr + l3sin->sin_len;
1238 
1239 		for ( ; addr < lim; sa++, mask++, addr++) {
1240 			if ((*sa ^ *addr) & *mask) {
1241 #ifdef DIAGNOSTIC
1242 				char addrbuf[INET_ADDRSTRLEN];
1243 
1244 				log(LOG_INFO, "IPv4 address: \"%s\" "
1245 				    "is not on the network\n",
1246 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1247 #endif
1248 				return (EINVAL);
1249 			}
1250 		}
1251 	}
1252 
1253 	return (0);
1254 }
1255 
1256 static inline uint32_t
1257 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1258 {
1259 
1260 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1261 }
1262 
1263 static uint32_t
1264 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1265 {
1266 
1267 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1268 }
1269 
1270 static void
1271 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1272 {
1273 	struct sockaddr_in *sin;
1274 
1275 	sin = (struct sockaddr_in *)sa;
1276 	bzero(sin, sizeof(*sin));
1277 	sin->sin_family = AF_INET;
1278 	sin->sin_len = sizeof(*sin);
1279 	sin->sin_addr = lle->r_l3addr.addr4;
1280 }
1281 
1282 static inline struct llentry *
1283 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1284 {
1285 	struct llentry *lle;
1286 	struct llentries *lleh;
1287 	u_int hashidx;
1288 
1289 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1290 	lleh = &llt->lle_head[hashidx];
1291 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1292 		if (lle->la_flags & LLE_DELETED)
1293 			continue;
1294 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1295 			break;
1296 	}
1297 
1298 	return (lle);
1299 }
1300 
1301 static void
1302 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1303 {
1304 
1305 	lle->la_flags |= LLE_DELETED;
1306 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1307 #ifdef DIAGNOSTIC
1308 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1309 #endif
1310 	llentry_free(lle);
1311 }
1312 
1313 static struct llentry *
1314 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1315 {
1316 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1317 	struct ifnet *ifp = llt->llt_ifp;
1318 	struct llentry *lle;
1319 	char linkhdr[LLE_MAX_LINKHDR];
1320 	size_t linkhdrsize;
1321 	int lladdr_off;
1322 
1323 	KASSERT(l3addr->sa_family == AF_INET,
1324 	    ("sin_family %d", l3addr->sa_family));
1325 
1326 	/*
1327 	 * A route that covers the given address must have
1328 	 * been installed 1st because we are doing a resolution,
1329 	 * verify this.
1330 	 */
1331 	if (!(flags & LLE_IFADDR) &&
1332 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1333 		return (NULL);
1334 
1335 	lle = in_lltable_new(sin->sin_addr, flags);
1336 	if (lle == NULL) {
1337 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1338 		return (NULL);
1339 	}
1340 	lle->la_flags = flags;
1341 	if (flags & LLE_STATIC)
1342 		lle->r_flags |= RLLE_VALID;
1343 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1344 		linkhdrsize = LLE_MAX_LINKHDR;
1345 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1346 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1347 			NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1348 			return (NULL);
1349 		}
1350 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1351 		    lladdr_off);
1352 		lle->la_flags |= LLE_STATIC;
1353 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1354 	}
1355 
1356 	return (lle);
1357 }
1358 
1359 /*
1360  * Return NULL if not found or marked for deletion.
1361  * If found return lle read locked.
1362  */
1363 static struct llentry *
1364 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1365 {
1366 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1367 	struct llentry *lle;
1368 
1369 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1370 	KASSERT(l3addr->sa_family == AF_INET,
1371 	    ("sin_family %d", l3addr->sa_family));
1372 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1373 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1374 	    ("wrong lle request flags: %#x", flags));
1375 
1376 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1377 	if (lle == NULL)
1378 		return (NULL);
1379 	if (flags & LLE_UNLOCKED)
1380 		return (lle);
1381 
1382 	if (flags & LLE_EXCLUSIVE)
1383 		LLE_WLOCK(lle);
1384 	else
1385 		LLE_RLOCK(lle);
1386 
1387 	/*
1388 	 * If the afdata lock is not held, the LLE may have been unlinked while
1389 	 * we were blocked on the LLE lock.  Check for this case.
1390 	 */
1391 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1392 		if (flags & LLE_EXCLUSIVE)
1393 			LLE_WUNLOCK(lle);
1394 		else
1395 			LLE_RUNLOCK(lle);
1396 		return (NULL);
1397 	}
1398 	return (lle);
1399 }
1400 
1401 static int
1402 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1403     struct sysctl_req *wr)
1404 {
1405 	struct ifnet *ifp = llt->llt_ifp;
1406 	/* XXX stack use */
1407 	struct {
1408 		struct rt_msghdr	rtm;
1409 		struct sockaddr_in	sin;
1410 		struct sockaddr_dl	sdl;
1411 	} arpc;
1412 	struct sockaddr_dl *sdl;
1413 	int error;
1414 
1415 	bzero(&arpc, sizeof(arpc));
1416 	/* skip deleted entries */
1417 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1418 		return (0);
1419 	/* Skip if jailed and not a valid IP of the prison. */
1420 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1421 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1422 		return (0);
1423 	/*
1424 	 * produce a msg made of:
1425 	 *  struct rt_msghdr;
1426 	 *  struct sockaddr_in; (IPv4)
1427 	 *  struct sockaddr_dl;
1428 	 */
1429 	arpc.rtm.rtm_msglen = sizeof(arpc);
1430 	arpc.rtm.rtm_version = RTM_VERSION;
1431 	arpc.rtm.rtm_type = RTM_GET;
1432 	arpc.rtm.rtm_flags = RTF_UP;
1433 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1434 
1435 	/* publish */
1436 	if (lle->la_flags & LLE_PUB)
1437 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1438 
1439 	sdl = &arpc.sdl;
1440 	sdl->sdl_family = AF_LINK;
1441 	sdl->sdl_len = sizeof(*sdl);
1442 	sdl->sdl_index = ifp->if_index;
1443 	sdl->sdl_type = ifp->if_type;
1444 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1445 		sdl->sdl_alen = ifp->if_addrlen;
1446 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1447 	} else {
1448 		sdl->sdl_alen = 0;
1449 		bzero(LLADDR(sdl), ifp->if_addrlen);
1450 	}
1451 
1452 	arpc.rtm.rtm_rmx.rmx_expire =
1453 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1454 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1455 	if (lle->la_flags & LLE_STATIC)
1456 		arpc.rtm.rtm_flags |= RTF_STATIC;
1457 	if (lle->la_flags & LLE_IFADDR)
1458 		arpc.rtm.rtm_flags |= RTF_PINNED;
1459 	arpc.rtm.rtm_index = ifp->if_index;
1460 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1461 
1462 	return (error);
1463 }
1464 
1465 static struct lltable *
1466 in_lltattach(struct ifnet *ifp)
1467 {
1468 	struct lltable *llt;
1469 
1470 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1471  	llt->llt_af = AF_INET;
1472  	llt->llt_ifp = ifp;
1473 
1474 	llt->llt_lookup = in_lltable_lookup;
1475 	llt->llt_alloc_entry = in_lltable_alloc;
1476 	llt->llt_delete_entry = in_lltable_delete_entry;
1477 	llt->llt_dump_entry = in_lltable_dump_entry;
1478 	llt->llt_hash = in_lltable_hash;
1479 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1480 	llt->llt_free_entry = in_lltable_free_entry;
1481 	llt->llt_match_prefix = in_lltable_match_prefix;
1482 	llt->llt_mark_used = in_lltable_mark_used;
1483  	lltable_link(llt);
1484 
1485 	return (llt);
1486 }
1487 
1488 void *
1489 in_domifattach(struct ifnet *ifp)
1490 {
1491 	struct in_ifinfo *ii;
1492 
1493 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1494 
1495 	ii->ii_llt = in_lltattach(ifp);
1496 	ii->ii_igmp = igmp_domifattach(ifp);
1497 
1498 	return (ii);
1499 }
1500 
1501 void
1502 in_domifdetach(struct ifnet *ifp, void *aux)
1503 {
1504 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1505 
1506 	igmp_domifdetach(ifp);
1507 	lltable_free(ii->ii_llt);
1508 	free(ii, M_IFADDR);
1509 }
1510