xref: /freebsd/sys/netinet/in.c (revision a4bcd20486f8c20cc875b39bc75aa0d5a047373f)
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 		iaIsFirst = false;
381 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
382 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
383 			ia = it;
384 	}
385 	NET_EPOCH_EXIT(et);
386 
387 	if (ia != NULL)
388 		(void )in_difaddr_ioctl(cmd, data, ifp, td);
389 
390 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
391 	ia = (struct in_ifaddr *)ifa;
392 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
393 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
394 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
395 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
396 	    CALLOUT_RETURNUNLOCKED);
397 
398 	ia->ia_ifp = ifp;
399 	ia->ia_addr = *addr;
400 	if (mask->sin_len != 0) {
401 		ia->ia_sockmask = *mask;
402 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
403 	} else {
404 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
405 
406 		/*
407 	 	 * Be compatible with network classes, if netmask isn't
408 		 * supplied, guess it based on classes.
409 	 	 */
410 		if (IN_CLASSA(i))
411 			ia->ia_subnetmask = IN_CLASSA_NET;
412 		else if (IN_CLASSB(i))
413 			ia->ia_subnetmask = IN_CLASSB_NET;
414 		else
415 			ia->ia_subnetmask = IN_CLASSC_NET;
416 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
417 	}
418 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
419 	in_socktrim(&ia->ia_sockmask);
420 
421 	if (ifp->if_flags & IFF_BROADCAST) {
422 		if (broadaddr->sin_len != 0) {
423 			ia->ia_broadaddr = *broadaddr;
424 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
425 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
426 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
427 			ia->ia_broadaddr.sin_family = AF_INET;
428 		} else {
429 			ia->ia_broadaddr.sin_addr.s_addr =
430 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
431 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
432 			ia->ia_broadaddr.sin_family = AF_INET;
433 		}
434 	}
435 
436 	if (ifp->if_flags & IFF_POINTOPOINT)
437 		ia->ia_dstaddr = *dstaddr;
438 
439 	/* XXXGL: rtinit() needs this strange assignment. */
440 	if (ifp->if_flags & IFF_LOOPBACK)
441                 ia->ia_dstaddr = ia->ia_addr;
442 
443 	if (vhid != 0) {
444 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
445 		if (error)
446 			return (error);
447 	}
448 
449 	/* if_addrhead is already referenced by ifa_alloc() */
450 	IF_ADDR_WLOCK(ifp);
451 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
452 	IF_ADDR_WUNLOCK(ifp);
453 
454 	ifa_ref(ifa);			/* in_ifaddrhead */
455 	IN_IFADDR_WLOCK();
456 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
457 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
458 	IN_IFADDR_WUNLOCK();
459 
460 	/*
461 	 * Give the interface a chance to initialize
462 	 * if this is its first address,
463 	 * and to validate the address if necessary.
464 	 */
465 	if (ifp->if_ioctl != NULL) {
466 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
467 		if (error)
468 			goto fail1;
469 	}
470 
471 	/*
472 	 * Add route for the network.
473 	 */
474 	if (vhid == 0) {
475 		int flags = RTF_UP;
476 
477 		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
478 			flags |= RTF_HOST;
479 
480 		error = in_addprefix(ia, flags);
481 		if (error)
482 			goto fail1;
483 	}
484 
485 	/*
486 	 * Add a loopback route to self.
487 	 */
488 	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
489 	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
490 	    !((ifp->if_flags & IFF_POINTOPOINT) &&
491 	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
492 		struct in_ifaddr *eia;
493 
494 		eia = in_localip_more(ia);
495 
496 		if (eia == NULL) {
497 			error = ifa_add_loopback_route((struct ifaddr *)ia,
498 			    (struct sockaddr *)&ia->ia_addr);
499 			if (error)
500 				goto fail2;
501 		} else
502 			ifa_free(&eia->ia_ifa);
503 	}
504 
505 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
506 		struct in_addr allhosts_addr;
507 		struct in_ifinfo *ii;
508 
509 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
510 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
511 
512 		error = in_joingroup(ifp, &allhosts_addr, NULL,
513 			&ii->ii_allhosts);
514 	}
515 
516 	/*
517 	 * Note: we don't need extra reference for ifa, since we called
518 	 * with sx lock held, and ifaddr can not be deleted in concurrent
519 	 * thread.
520 	 */
521 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
522 
523 	return (error);
524 
525 fail2:
526 	if (vhid == 0)
527 		(void )in_scrubprefix(ia, LLE_STATIC);
528 
529 fail1:
530 	if (ia->ia_ifa.ifa_carp)
531 		(*carp_detach_p)(&ia->ia_ifa, false);
532 
533 	IF_ADDR_WLOCK(ifp);
534 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
535 	IF_ADDR_WUNLOCK(ifp);
536 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
537 
538 	IN_IFADDR_WLOCK();
539 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
540 	LIST_REMOVE(ia, ia_hash);
541 	IN_IFADDR_WUNLOCK();
542 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
543 
544 	return (error);
545 }
546 
547 static int
548 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
549 {
550 	const struct ifreq *ifr = (struct ifreq *)data;
551 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
552 	    &ifr->ifr_addr;
553 	struct ifaddr *ifa;
554 	struct in_ifaddr *ia;
555 	bool deleteAny, iaIsLast;
556 	int error;
557 
558 	if (td != NULL) {
559 		error = priv_check(td, PRIV_NET_DELIFADDR);
560 		if (error)
561 			return (error);
562 	}
563 
564 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
565 	    addr->sin_family != AF_INET)
566 		deleteAny = true;
567 	else
568 		deleteAny = false;
569 
570 	iaIsLast = true;
571 	ia = NULL;
572 	IF_ADDR_WLOCK(ifp);
573 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
574 		struct in_ifaddr *it;
575 
576 		if (ifa->ifa_addr->sa_family != AF_INET)
577 			continue;
578 
579 		it = (struct in_ifaddr *)ifa;
580 		if (deleteAny && ia == NULL && (td == NULL ||
581 		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
582 			ia = it;
583 
584 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
585 		    (td == NULL || prison_check_ip4(td->td_ucred,
586 		    &addr->sin_addr) == 0))
587 			ia = it;
588 
589 		if (it != ia)
590 			iaIsLast = false;
591 	}
592 
593 	if (ia == NULL) {
594 		IF_ADDR_WUNLOCK(ifp);
595 		return (EADDRNOTAVAIL);
596 	}
597 
598 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
599 	IF_ADDR_WUNLOCK(ifp);
600 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
601 
602 	IN_IFADDR_WLOCK();
603 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
604 	LIST_REMOVE(ia, ia_hash);
605 	IN_IFADDR_WUNLOCK();
606 
607 	/*
608 	 * in_scrubprefix() kills the interface route.
609 	 */
610 	in_scrubprefix(ia, LLE_STATIC);
611 
612 	/*
613 	 * in_ifadown gets rid of all the rest of
614 	 * the routes.  This is not quite the right
615 	 * thing to do, but at least if we are running
616 	 * a routing process they will come back.
617 	 */
618 	in_ifadown(&ia->ia_ifa, 1);
619 
620 	if (ia->ia_ifa.ifa_carp)
621 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
622 
623 	/*
624 	 * If this is the last IPv4 address configured on this
625 	 * interface, leave the all-hosts group.
626 	 * No state-change report need be transmitted.
627 	 */
628 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
629 		struct in_ifinfo *ii;
630 
631 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
632 		if (ii->ii_allhosts) {
633 			(void)in_leavegroup(ii->ii_allhosts, NULL);
634 			ii->ii_allhosts = NULL;
635 		}
636 	}
637 
638 	IF_ADDR_WLOCK(ifp);
639 	if (callout_stop(&ia->ia_garp_timer) == 1) {
640 		ifa_free(&ia->ia_ifa);
641 	}
642 	IF_ADDR_WUNLOCK(ifp);
643 
644 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
645 	    IFADDR_EVENT_DEL);
646 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
647 
648 	return (0);
649 }
650 
651 #define rtinitflags(x) \
652 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
653 	    ? RTF_HOST : 0)
654 
655 /*
656  * Check if we have a route for the given prefix already or add one accordingly.
657  */
658 int
659 in_addprefix(struct in_ifaddr *target, int flags)
660 {
661 	struct rm_priotracker in_ifa_tracker;
662 	struct in_ifaddr *ia;
663 	struct in_addr prefix, mask, p, m;
664 	int error;
665 
666 	if ((flags & RTF_HOST) != 0) {
667 		prefix = target->ia_dstaddr.sin_addr;
668 		mask.s_addr = 0;
669 	} else {
670 		prefix = target->ia_addr.sin_addr;
671 		mask = target->ia_sockmask.sin_addr;
672 		prefix.s_addr &= mask.s_addr;
673 	}
674 
675 	IN_IFADDR_RLOCK(&in_ifa_tracker);
676 	/* Look for an existing address with the same prefix, mask, and fib */
677 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
678 		if (rtinitflags(ia)) {
679 			p = ia->ia_dstaddr.sin_addr;
680 
681 			if (prefix.s_addr != p.s_addr)
682 				continue;
683 		} else {
684 			p = ia->ia_addr.sin_addr;
685 			m = ia->ia_sockmask.sin_addr;
686 			p.s_addr &= m.s_addr;
687 
688 			if (prefix.s_addr != p.s_addr ||
689 			    mask.s_addr != m.s_addr)
690 				continue;
691 		}
692 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
693 			continue;
694 
695 		/*
696 		 * If we got a matching prefix route inserted by other
697 		 * interface address, we are done here.
698 		 */
699 		if (ia->ia_flags & IFA_ROUTE) {
700 			if (V_nosameprefix) {
701 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
702 				return (EEXIST);
703 			} else {
704 				int fibnum;
705 
706 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
707 					target->ia_ifp->if_fib;
708 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
709 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
710 				return (0);
711 			}
712 		}
713 	}
714 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
715 
716 	/*
717 	 * No-one seem to have this prefix route, so we try to insert it.
718 	 */
719 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
720 	if (!error)
721 		target->ia_flags |= IFA_ROUTE;
722 	return (error);
723 }
724 
725 /*
726  * Removes either all lle entries for given @ia, or lle
727  * corresponding to @ia address.
728  */
729 static void
730 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
731 {
732 	struct sockaddr_in addr, mask;
733 	struct sockaddr *saddr, *smask;
734 	struct ifnet *ifp;
735 
736 	saddr = (struct sockaddr *)&addr;
737 	bzero(&addr, sizeof(addr));
738 	addr.sin_len = sizeof(addr);
739 	addr.sin_family = AF_INET;
740 	smask = (struct sockaddr *)&mask;
741 	bzero(&mask, sizeof(mask));
742 	mask.sin_len = sizeof(mask);
743 	mask.sin_family = AF_INET;
744 	mask.sin_addr.s_addr = ia->ia_subnetmask;
745 	ifp = ia->ia_ifp;
746 
747 	if (all) {
748 		/*
749 		 * Remove all L2 entries matching given prefix.
750 		 * Convert address to host representation to avoid
751 		 * doing this on every callback. ia_subnetmask is already
752 		 * stored in host representation.
753 		 */
754 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
755 		lltable_prefix_free(AF_INET, saddr, smask, flags);
756 	} else {
757 		/* Remove interface address only */
758 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
759 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
760 	}
761 }
762 
763 /*
764  * If there is no other address in the system that can serve a route to the
765  * same prefix, remove the route.  Hand over the route to the new address
766  * otherwise.
767  */
768 int
769 in_scrubprefix(struct in_ifaddr *target, u_int flags)
770 {
771 	struct rm_priotracker in_ifa_tracker;
772 	struct in_ifaddr *ia;
773 	struct in_addr prefix, mask, p, m;
774 	int error = 0;
775 
776 	/*
777 	 * Remove the loopback route to the interface address.
778 	 */
779 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
780 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
781 	    (flags & LLE_STATIC)) {
782 		struct in_ifaddr *eia;
783 
784 		/*
785 		 * XXXME: add fib-aware in_localip.
786 		 * We definitely don't want to switch between
787 		 * prefixes in different fibs.
788 		 */
789 		eia = in_localip_more(target);
790 
791 		if (eia != NULL) {
792 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
793 			    (struct sockaddr *)&target->ia_addr);
794 			ifa_free(&eia->ia_ifa);
795 		} else {
796 			error = ifa_del_loopback_route((struct ifaddr *)target,
797 			    (struct sockaddr *)&target->ia_addr);
798 		}
799 	}
800 
801 	if (rtinitflags(target)) {
802 		prefix = target->ia_dstaddr.sin_addr;
803 		mask.s_addr = 0;
804 	} else {
805 		prefix = target->ia_addr.sin_addr;
806 		mask = target->ia_sockmask.sin_addr;
807 		prefix.s_addr &= mask.s_addr;
808 	}
809 
810 	if ((target->ia_flags & IFA_ROUTE) == 0) {
811 		int fibnum;
812 
813 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
814 			target->ia_ifp->if_fib;
815 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
816 
817 		/*
818 		 * Removing address from !IFF_UP interface or
819 		 * prefix which exists on other interface (along with route).
820 		 * No entries should exist here except target addr.
821 		 * Given that, delete this entry only.
822 		 */
823 		in_scrubprefixlle(target, 0, flags);
824 		return (0);
825 	}
826 
827 	IN_IFADDR_RLOCK(&in_ifa_tracker);
828 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
829 		if (rtinitflags(ia)) {
830 			p = ia->ia_dstaddr.sin_addr;
831 
832 			if (prefix.s_addr != p.s_addr)
833 				continue;
834 		} else {
835 			p = ia->ia_addr.sin_addr;
836 			m = ia->ia_sockmask.sin_addr;
837 			p.s_addr &= m.s_addr;
838 
839 			if (prefix.s_addr != p.s_addr ||
840 			    mask.s_addr != m.s_addr)
841 				continue;
842 		}
843 
844 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
845 			continue;
846 
847 		/*
848 		 * If we got a matching prefix address, move IFA_ROUTE and
849 		 * the route itself to it.  Make sure that routing daemons
850 		 * get a heads-up.
851 		 */
852 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
853 			ifa_ref(&ia->ia_ifa);
854 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
855 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
856 			    rtinitflags(target));
857 			if (error == 0)
858 				target->ia_flags &= ~IFA_ROUTE;
859 			else
860 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
861 					error);
862 			/* Scrub all entries IFF interface is different */
863 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
864 			    flags);
865 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
866 			    rtinitflags(ia) | RTF_UP);
867 			if (error == 0)
868 				ia->ia_flags |= IFA_ROUTE;
869 			else
870 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
871 					error);
872 			ifa_free(&ia->ia_ifa);
873 			return (error);
874 		}
875 	}
876 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
877 
878 	/*
879 	 * remove all L2 entries on the given prefix
880 	 */
881 	in_scrubprefixlle(target, 1, flags);
882 
883 	/*
884 	 * As no-one seem to have this prefix, we can remove the route.
885 	 */
886 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
887 	if (error == 0)
888 		target->ia_flags &= ~IFA_ROUTE;
889 	else
890 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
891 	return (error);
892 }
893 
894 #undef rtinitflags
895 
896 void
897 in_ifscrub_all(void)
898 {
899 	struct ifnet *ifp;
900 	struct ifaddr *ifa, *nifa;
901 	struct ifaliasreq ifr;
902 
903 	IFNET_RLOCK();
904 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
905 		/* Cannot lock here - lock recursion. */
906 		/* NET_EPOCH_ENTER(et); */
907 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
908 			if (ifa->ifa_addr->sa_family != AF_INET)
909 				continue;
910 
911 			/*
912 			 * This is ugly but the only way for legacy IP to
913 			 * cleanly remove addresses and everything attached.
914 			 */
915 			bzero(&ifr, sizeof(ifr));
916 			ifr.ifra_addr = *ifa->ifa_addr;
917 			if (ifa->ifa_dstaddr)
918 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
919 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
920 			    ifp, NULL);
921 		}
922 		/* NET_EPOCH_EXIT(et); */
923 		in_purgemaddrs(ifp);
924 		igmp_domifdetach(ifp);
925 	}
926 	IFNET_RUNLOCK();
927 }
928 
929 int
930 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
931 {
932 
933 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
934 	     /*
935 	      * Check for old-style (host 0) broadcast, but
936 	      * taking into account that RFC 3021 obsoletes it.
937 	      */
938 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
939 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
940 	     /*
941 	      * Check for an all one subnetmask. These
942 	      * only exist when an interface gets a secondary
943 	      * address.
944 	      */
945 	    ia->ia_subnetmask != (u_long)0xffffffff);
946 }
947 
948 /*
949  * Return 1 if the address might be a local broadcast address.
950  */
951 int
952 in_broadcast(struct in_addr in, struct ifnet *ifp)
953 {
954 	struct ifaddr *ifa;
955 	int found;
956 
957 	NET_EPOCH_ASSERT();
958 
959 	if (in.s_addr == INADDR_BROADCAST ||
960 	    in.s_addr == INADDR_ANY)
961 		return (1);
962 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
963 		return (0);
964 	found = 0;
965 	/*
966 	 * Look through the list of addresses for a match
967 	 * with a broadcast address.
968 	 */
969 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
970 		if (ifa->ifa_addr->sa_family == AF_INET &&
971 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
972 			found = 1;
973 			break;
974 		}
975 	return (found);
976 }
977 
978 /*
979  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
980  */
981 void
982 in_ifdetach(struct ifnet *ifp)
983 {
984 	IN_MULTI_LOCK();
985 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
986 	in_pcbpurgeif0(&V_udbinfo, ifp);
987 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
988 	in_purgemaddrs(ifp);
989 	IN_MULTI_UNLOCK();
990 
991 	/*
992 	 * Make sure all multicast deletions invoking if_ioctl() are
993 	 * completed before returning. Else we risk accessing a freed
994 	 * ifnet structure pointer.
995 	 */
996 	inm_release_wait(NULL);
997 }
998 
999 /*
1000  * Delete all IPv4 multicast address records, and associated link-layer
1001  * multicast address records, associated with ifp.
1002  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1003  * XXX This should not race with ifma_protospec being set during
1004  * a new allocation, if it does, we have bigger problems.
1005  */
1006 static void
1007 in_purgemaddrs(struct ifnet *ifp)
1008 {
1009 	struct in_multi_head purgeinms;
1010 	struct in_multi		*inm;
1011 	struct ifmultiaddr	*ifma, *next;
1012 
1013 	SLIST_INIT(&purgeinms);
1014 	IN_MULTI_LIST_LOCK();
1015 
1016 	/*
1017 	 * Extract list of in_multi associated with the detaching ifp
1018 	 * which the PF_INET layer is about to release.
1019 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1020 	 * by code further down.
1021 	 */
1022 	IF_ADDR_WLOCK(ifp);
1023  restart:
1024 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1025 		if (ifma->ifma_addr->sa_family != AF_INET ||
1026 		    ifma->ifma_protospec == NULL)
1027 			continue;
1028 		inm = (struct in_multi *)ifma->ifma_protospec;
1029 		inm_rele_locked(&purgeinms, inm);
1030 		if (__predict_false(ifma_restart)) {
1031 			ifma_restart = true;
1032 			goto restart;
1033 		}
1034 	}
1035 	IF_ADDR_WUNLOCK(ifp);
1036 
1037 	inm_release_list_deferred(&purgeinms);
1038 	igmp_ifdetach(ifp);
1039 	IN_MULTI_LIST_UNLOCK();
1040 }
1041 
1042 struct in_llentry {
1043 	struct llentry		base;
1044 };
1045 
1046 #define	IN_LLTBL_DEFAULT_HSIZE	32
1047 #define	IN_LLTBL_HASH(k, h) \
1048 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1049 
1050 /*
1051  * Do actual deallocation of @lle.
1052  */
1053 static void
1054 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1055 {
1056 	struct llentry *lle;
1057 
1058 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1059 	LLE_LOCK_DESTROY(lle);
1060 	LLE_REQ_DESTROY(lle);
1061 	free(lle, M_LLTABLE);
1062 }
1063 
1064 /*
1065  * Called by the datapath to indicate that
1066  * the entry was used.
1067  */
1068 static void
1069 in_lltable_mark_used(struct llentry *lle)
1070 {
1071 
1072 	LLE_REQ_LOCK(lle);
1073 	lle->r_skip_req = 0;
1074 	LLE_REQ_UNLOCK(lle);
1075 }
1076 
1077 /*
1078  * Called by LLE_FREE_LOCKED when number of references
1079  * drops to zero.
1080  */
1081 static void
1082 in_lltable_destroy_lle(struct llentry *lle)
1083 {
1084 
1085 	LLE_WUNLOCK(lle);
1086 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1087 }
1088 
1089 static struct llentry *
1090 in_lltable_new(struct in_addr addr4, u_int flags)
1091 {
1092 	struct in_llentry *lle;
1093 
1094 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1095 	if (lle == NULL)		/* NB: caller generates msg */
1096 		return NULL;
1097 
1098 	/*
1099 	 * For IPv4 this will trigger "arpresolve" to generate
1100 	 * an ARP request.
1101 	 */
1102 	lle->base.la_expire = time_uptime; /* mark expired */
1103 	lle->base.r_l3addr.addr4 = addr4;
1104 	lle->base.lle_refcnt = 1;
1105 	lle->base.lle_free = in_lltable_destroy_lle;
1106 	LLE_LOCK_INIT(&lle->base);
1107 	LLE_REQ_INIT(&lle->base);
1108 	callout_init(&lle->base.lle_timer, 1);
1109 
1110 	return (&lle->base);
1111 }
1112 
1113 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1114 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1115 
1116 static int
1117 in_lltable_match_prefix(const struct sockaddr *saddr,
1118     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1119 {
1120 	struct in_addr addr, mask, lle_addr;
1121 
1122 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1123 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1124 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1125 
1126 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1127 		return (0);
1128 
1129 	if (lle->la_flags & LLE_IFADDR) {
1130 		/*
1131 		 * Delete LLE_IFADDR records IFF address & flag matches.
1132 		 * Note that addr is the interface address within prefix
1133 		 * being matched.
1134 		 * Note also we should handle 'ifdown' cases without removing
1135 		 * ifaddr macs.
1136 		 */
1137 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1138 			return (1);
1139 		return (0);
1140 	}
1141 
1142 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1143 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1144 		return (1);
1145 
1146 	return (0);
1147 }
1148 
1149 static void
1150 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1151 {
1152 	size_t pkts_dropped;
1153 
1154 	LLE_WLOCK_ASSERT(lle);
1155 	KASSERT(llt != NULL, ("lltable is NULL"));
1156 
1157 	/* Unlink entry from table if not already */
1158 	if ((lle->la_flags & LLE_LINKED) != 0) {
1159 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1160 		lltable_unlink_entry(llt, lle);
1161 	}
1162 
1163 	/* Drop hold queue */
1164 	pkts_dropped = llentry_free(lle);
1165 	ARPSTAT_ADD(dropped, pkts_dropped);
1166 }
1167 
1168 static int
1169 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1170 {
1171 	struct rt_addrinfo info;
1172 	struct sockaddr_in rt_key, rt_mask;
1173 	struct sockaddr rt_gateway;
1174 	int rt_flags;
1175 
1176 	KASSERT(l3addr->sa_family == AF_INET,
1177 	    ("sin_family %d", l3addr->sa_family));
1178 
1179 	bzero(&rt_key, sizeof(rt_key));
1180 	rt_key.sin_len = sizeof(rt_key);
1181 	bzero(&rt_mask, sizeof(rt_mask));
1182 	rt_mask.sin_len = sizeof(rt_mask);
1183 	bzero(&rt_gateway, sizeof(rt_gateway));
1184 	rt_gateway.sa_len = sizeof(rt_gateway);
1185 
1186 	bzero(&info, sizeof(info));
1187 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1188 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1189 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1190 
1191 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1192 		return (EINVAL);
1193 
1194 	rt_flags = info.rti_flags;
1195 
1196 	/*
1197 	 * If the gateway for an existing host route matches the target L3
1198 	 * address, which is a special route inserted by some implementation
1199 	 * such as MANET, and the interface is of the correct type, then
1200 	 * allow for ARP to proceed.
1201 	 */
1202 	if (rt_flags & RTF_GATEWAY) {
1203 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1204 		    info.rti_ifp->if_type != IFT_ETHER ||
1205 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1206 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1207 		    sizeof(in_addr_t)) != 0) {
1208 			rib_free_info(&info);
1209 			return (EINVAL);
1210 		}
1211 	}
1212 	rib_free_info(&info);
1213 
1214 	/*
1215 	 * Make sure that at least the destination address is covered
1216 	 * by the route. This is for handling the case where 2 or more
1217 	 * interfaces have the same prefix. An incoming packet arrives
1218 	 * on one interface and the corresponding outgoing packet leaves
1219 	 * another interface.
1220 	 */
1221 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1222 		const char *sa, *mask, *addr, *lim;
1223 		const struct sockaddr_in *l3sin;
1224 
1225 		mask = (const char *)&rt_mask;
1226 		/*
1227 		 * Just being extra cautious to avoid some custom
1228 		 * code getting into trouble.
1229 		 */
1230 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1231 			return (EINVAL);
1232 
1233 		sa = (const char *)&rt_key;
1234 		addr = (const char *)l3addr;
1235 		l3sin = (const struct sockaddr_in *)l3addr;
1236 		lim = addr + l3sin->sin_len;
1237 
1238 		for ( ; addr < lim; sa++, mask++, addr++) {
1239 			if ((*sa ^ *addr) & *mask) {
1240 #ifdef DIAGNOSTIC
1241 				char addrbuf[INET_ADDRSTRLEN];
1242 
1243 				log(LOG_INFO, "IPv4 address: \"%s\" "
1244 				    "is not on the network\n",
1245 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1246 #endif
1247 				return (EINVAL);
1248 			}
1249 		}
1250 	}
1251 
1252 	return (0);
1253 }
1254 
1255 static inline uint32_t
1256 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1257 {
1258 
1259 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1260 }
1261 
1262 static uint32_t
1263 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1264 {
1265 
1266 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1267 }
1268 
1269 static void
1270 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1271 {
1272 	struct sockaddr_in *sin;
1273 
1274 	sin = (struct sockaddr_in *)sa;
1275 	bzero(sin, sizeof(*sin));
1276 	sin->sin_family = AF_INET;
1277 	sin->sin_len = sizeof(*sin);
1278 	sin->sin_addr = lle->r_l3addr.addr4;
1279 }
1280 
1281 static inline struct llentry *
1282 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1283 {
1284 	struct llentry *lle;
1285 	struct llentries *lleh;
1286 	u_int hashidx;
1287 
1288 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1289 	lleh = &llt->lle_head[hashidx];
1290 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1291 		if (lle->la_flags & LLE_DELETED)
1292 			continue;
1293 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1294 			break;
1295 	}
1296 
1297 	return (lle);
1298 }
1299 
1300 static void
1301 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1302 {
1303 
1304 	lle->la_flags |= LLE_DELETED;
1305 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1306 #ifdef DIAGNOSTIC
1307 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1308 #endif
1309 	llentry_free(lle);
1310 }
1311 
1312 static struct llentry *
1313 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1314 {
1315 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1316 	struct ifnet *ifp = llt->llt_ifp;
1317 	struct llentry *lle;
1318 	char linkhdr[LLE_MAX_LINKHDR];
1319 	size_t linkhdrsize;
1320 	int lladdr_off;
1321 
1322 	KASSERT(l3addr->sa_family == AF_INET,
1323 	    ("sin_family %d", l3addr->sa_family));
1324 
1325 	/*
1326 	 * A route that covers the given address must have
1327 	 * been installed 1st because we are doing a resolution,
1328 	 * verify this.
1329 	 */
1330 	if (!(flags & LLE_IFADDR) &&
1331 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1332 		return (NULL);
1333 
1334 	lle = in_lltable_new(sin->sin_addr, flags);
1335 	if (lle == NULL) {
1336 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1337 		return (NULL);
1338 	}
1339 	lle->la_flags = flags;
1340 	if (flags & LLE_STATIC)
1341 		lle->r_flags |= RLLE_VALID;
1342 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1343 		linkhdrsize = LLE_MAX_LINKHDR;
1344 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1345 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1346 			NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1347 			return (NULL);
1348 		}
1349 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1350 		    lladdr_off);
1351 		lle->la_flags |= LLE_STATIC;
1352 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1353 	}
1354 
1355 	return (lle);
1356 }
1357 
1358 /*
1359  * Return NULL if not found or marked for deletion.
1360  * If found return lle read locked.
1361  */
1362 static struct llentry *
1363 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1364 {
1365 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1366 	struct llentry *lle;
1367 
1368 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1369 	KASSERT(l3addr->sa_family == AF_INET,
1370 	    ("sin_family %d", l3addr->sa_family));
1371 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1372 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1373 	    ("wrong lle request flags: %#x", flags));
1374 
1375 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1376 	if (lle == NULL)
1377 		return (NULL);
1378 	if (flags & LLE_UNLOCKED)
1379 		return (lle);
1380 
1381 	if (flags & LLE_EXCLUSIVE)
1382 		LLE_WLOCK(lle);
1383 	else
1384 		LLE_RLOCK(lle);
1385 
1386 	/*
1387 	 * If the afdata lock is not held, the LLE may have been unlinked while
1388 	 * we were blocked on the LLE lock.  Check for this case.
1389 	 */
1390 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1391 		if (flags & LLE_EXCLUSIVE)
1392 			LLE_WUNLOCK(lle);
1393 		else
1394 			LLE_RUNLOCK(lle);
1395 		return (NULL);
1396 	}
1397 	return (lle);
1398 }
1399 
1400 static int
1401 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1402     struct sysctl_req *wr)
1403 {
1404 	struct ifnet *ifp = llt->llt_ifp;
1405 	/* XXX stack use */
1406 	struct {
1407 		struct rt_msghdr	rtm;
1408 		struct sockaddr_in	sin;
1409 		struct sockaddr_dl	sdl;
1410 	} arpc;
1411 	struct sockaddr_dl *sdl;
1412 	int error;
1413 
1414 	bzero(&arpc, sizeof(arpc));
1415 	/* skip deleted entries */
1416 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1417 		return (0);
1418 	/* Skip if jailed and not a valid IP of the prison. */
1419 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1420 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1421 		return (0);
1422 	/*
1423 	 * produce a msg made of:
1424 	 *  struct rt_msghdr;
1425 	 *  struct sockaddr_in; (IPv4)
1426 	 *  struct sockaddr_dl;
1427 	 */
1428 	arpc.rtm.rtm_msglen = sizeof(arpc);
1429 	arpc.rtm.rtm_version = RTM_VERSION;
1430 	arpc.rtm.rtm_type = RTM_GET;
1431 	arpc.rtm.rtm_flags = RTF_UP;
1432 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1433 
1434 	/* publish */
1435 	if (lle->la_flags & LLE_PUB)
1436 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1437 
1438 	sdl = &arpc.sdl;
1439 	sdl->sdl_family = AF_LINK;
1440 	sdl->sdl_len = sizeof(*sdl);
1441 	sdl->sdl_index = ifp->if_index;
1442 	sdl->sdl_type = ifp->if_type;
1443 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1444 		sdl->sdl_alen = ifp->if_addrlen;
1445 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1446 	} else {
1447 		sdl->sdl_alen = 0;
1448 		bzero(LLADDR(sdl), ifp->if_addrlen);
1449 	}
1450 
1451 	arpc.rtm.rtm_rmx.rmx_expire =
1452 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1453 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1454 	if (lle->la_flags & LLE_STATIC)
1455 		arpc.rtm.rtm_flags |= RTF_STATIC;
1456 	if (lle->la_flags & LLE_IFADDR)
1457 		arpc.rtm.rtm_flags |= RTF_PINNED;
1458 	arpc.rtm.rtm_index = ifp->if_index;
1459 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1460 
1461 	return (error);
1462 }
1463 
1464 static struct lltable *
1465 in_lltattach(struct ifnet *ifp)
1466 {
1467 	struct lltable *llt;
1468 
1469 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1470  	llt->llt_af = AF_INET;
1471  	llt->llt_ifp = ifp;
1472 
1473 	llt->llt_lookup = in_lltable_lookup;
1474 	llt->llt_alloc_entry = in_lltable_alloc;
1475 	llt->llt_delete_entry = in_lltable_delete_entry;
1476 	llt->llt_dump_entry = in_lltable_dump_entry;
1477 	llt->llt_hash = in_lltable_hash;
1478 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1479 	llt->llt_free_entry = in_lltable_free_entry;
1480 	llt->llt_match_prefix = in_lltable_match_prefix;
1481 	llt->llt_mark_used = in_lltable_mark_used;
1482  	lltable_link(llt);
1483 
1484 	return (llt);
1485 }
1486 
1487 void *
1488 in_domifattach(struct ifnet *ifp)
1489 {
1490 	struct in_ifinfo *ii;
1491 
1492 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1493 
1494 	ii->ii_llt = in_lltattach(ifp);
1495 	ii->ii_igmp = igmp_domifattach(ifp);
1496 
1497 	return (ii);
1498 }
1499 
1500 void
1501 in_domifdetach(struct ifnet *ifp, void *aux)
1502 {
1503 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1504 
1505 	igmp_domifdetach(ifp);
1506 	lltable_free(ii->ii_llt);
1507 	free(ii, M_IFADDR);
1508 }
1509