xref: /freebsd/sys/netinet/in.c (revision c6989859ae9388eeb46a24fe88f9b8d07101c710)
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 "opt_mpath.h"
39 
40 #include <sys/param.h>
41 #include <sys/eventhandler.h>
42 #include <sys/systm.h>
43 #include <sys/sockio.h>
44 #include <sys/malloc.h>
45 #include <sys/priv.h>
46 #include <sys/socket.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/proc.h>
51 #include <sys/rmlock.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/sx.h>
55 
56 #include <net/if.h>
57 #include <net/if_var.h>
58 #include <net/if_arp.h>
59 #include <net/if_dl.h>
60 #include <net/if_llatbl.h>
61 #include <net/if_types.h>
62 #include <net/route.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/if_ether.h>
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_carp.h>
71 #include <netinet/igmp_var.h>
72 #include <netinet/udp.h>
73 #include <netinet/udp_var.h>
74 
75 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
76 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
77 
78 static void	in_socktrim(struct sockaddr_in *);
79 static void	in_purgemaddrs(struct ifnet *);
80 
81 VNET_DEFINE_STATIC(int, nosameprefix);
82 #define	V_nosameprefix			VNET(nosameprefix)
83 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
84 	&VNET_NAME(nosameprefix), 0,
85 	"Refuse to create same prefixes on different interfaces");
86 
87 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
88 #define	V_ripcbinfo			VNET(ripcbinfo)
89 
90 static struct sx in_control_sx;
91 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
92 
93 /*
94  * Return 1 if an internet address is for a ``local'' host
95  * (one to which we have a connection).
96  */
97 int
98 in_localaddr(struct in_addr in)
99 {
100 	struct rm_priotracker in_ifa_tracker;
101 	u_long i = ntohl(in.s_addr);
102 	struct in_ifaddr *ia;
103 
104 	IN_IFADDR_RLOCK(&in_ifa_tracker);
105 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
106 		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
107 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
108 			return (1);
109 		}
110 	}
111 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
112 	return (0);
113 }
114 
115 /*
116  * Return 1 if an internet address is for the local host and configured
117  * on one of its interfaces.
118  */
119 int
120 in_localip(struct in_addr in)
121 {
122 	struct rm_priotracker in_ifa_tracker;
123 	struct in_ifaddr *ia;
124 
125 	IN_IFADDR_RLOCK(&in_ifa_tracker);
126 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
127 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
128 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
129 			return (1);
130 		}
131 	}
132 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
133 	return (0);
134 }
135 
136 /*
137  * Return 1 if an internet address is configured on an interface.
138  */
139 int
140 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
141 {
142 	struct ifaddr *ifa;
143 	struct in_ifaddr *ia;
144 
145 	NET_EPOCH_ASSERT();
146 
147 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
148 		if (ifa->ifa_addr->sa_family != AF_INET)
149 			continue;
150 		ia = (struct in_ifaddr *)ifa;
151 		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
152 			return (1);
153 	}
154 
155 	return (0);
156 }
157 
158 /*
159  * Return a reference to the interface address which is different to
160  * the supplied one but with same IP address value.
161  */
162 static struct in_ifaddr *
163 in_localip_more(struct in_ifaddr *ia)
164 {
165 	struct rm_priotracker in_ifa_tracker;
166 	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
167 	struct in_ifaddr *it;
168 
169 	IN_IFADDR_RLOCK(&in_ifa_tracker);
170 	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
171 		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
172 			ifa_ref(&it->ia_ifa);
173 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
174 			return (it);
175 		}
176 	}
177 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
178 
179 	return (NULL);
180 }
181 
182 /*
183  * Determine whether an IP address is in a reserved set of addresses
184  * that may not be forwarded, or whether datagrams to that destination
185  * may be forwarded.
186  */
187 int
188 in_canforward(struct in_addr in)
189 {
190 	u_long i = ntohl(in.s_addr);
191 
192 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
193 	    IN_ZERONET(i) || IN_LOOPBACK(i))
194 		return (0);
195 	return (1);
196 }
197 
198 /*
199  * Trim a mask in a sockaddr
200  */
201 static void
202 in_socktrim(struct sockaddr_in *ap)
203 {
204     char *cplim = (char *) &ap->sin_addr;
205     char *cp = (char *) (&ap->sin_addr + 1);
206 
207     ap->sin_len = 0;
208     while (--cp >= cplim)
209 	if (*cp) {
210 	    (ap)->sin_len = cp - (char *) (ap) + 1;
211 	    break;
212 	}
213 }
214 
215 /*
216  * Generic internet control operations (ioctl's).
217  */
218 int
219 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
220     struct thread *td)
221 {
222 	struct ifreq *ifr = (struct ifreq *)data;
223 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
224 	struct epoch_tracker et;
225 	struct ifaddr *ifa;
226 	struct in_ifaddr *ia;
227 	int error;
228 
229 	if (ifp == NULL)
230 		return (EADDRNOTAVAIL);
231 
232 	/*
233 	 * Filter out 4 ioctls we implement directly.  Forward the rest
234 	 * to specific functions and ifp->if_ioctl().
235 	 */
236 	switch (cmd) {
237 	case SIOCGIFADDR:
238 	case SIOCGIFBRDADDR:
239 	case SIOCGIFDSTADDR:
240 	case SIOCGIFNETMASK:
241 		break;
242 	case SIOCDIFADDR:
243 		sx_xlock(&in_control_sx);
244 		error = in_difaddr_ioctl(cmd, data, ifp, td);
245 		sx_xunlock(&in_control_sx);
246 		return (error);
247 	case OSIOCAIFADDR:	/* 9.x compat */
248 	case SIOCAIFADDR:
249 		sx_xlock(&in_control_sx);
250 		error = in_aifaddr_ioctl(cmd, data, ifp, td);
251 		sx_xunlock(&in_control_sx);
252 		return (error);
253 	case SIOCSIFADDR:
254 	case SIOCSIFBRDADDR:
255 	case SIOCSIFDSTADDR:
256 	case SIOCSIFNETMASK:
257 		/* We no longer support that old commands. */
258 		return (EINVAL);
259 	default:
260 		if (ifp->if_ioctl == NULL)
261 			return (EOPNOTSUPP);
262 		return ((*ifp->if_ioctl)(ifp, cmd, data));
263 	}
264 
265 	if (addr->sin_addr.s_addr != INADDR_ANY &&
266 	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
267 		return (EADDRNOTAVAIL);
268 
269 	/*
270 	 * Find address for this interface, if it exists.  If an
271 	 * address was specified, find that one instead of the
272 	 * first one on the interface, if possible.
273 	 */
274 	NET_EPOCH_ENTER(et);
275 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
276 		if (ifa->ifa_addr->sa_family != AF_INET)
277 			continue;
278 		ia = (struct in_ifaddr *)ifa;
279 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
280 			break;
281 	}
282 	if (ifa == NULL)
283 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
284 			if (ifa->ifa_addr->sa_family == AF_INET) {
285 				ia = (struct in_ifaddr *)ifa;
286 				if (prison_check_ip4(td->td_ucred,
287 				    &ia->ia_addr.sin_addr) == 0)
288 					break;
289 			}
290 
291 	if (ifa == NULL) {
292 		NET_EPOCH_EXIT(et);
293 		return (EADDRNOTAVAIL);
294 	}
295 
296 	error = 0;
297 	switch (cmd) {
298 	case SIOCGIFADDR:
299 		*addr = ia->ia_addr;
300 		break;
301 
302 	case SIOCGIFBRDADDR:
303 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
304 			error = EINVAL;
305 			break;
306 		}
307 		*addr = ia->ia_broadaddr;
308 		break;
309 
310 	case SIOCGIFDSTADDR:
311 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
312 			error = EINVAL;
313 			break;
314 		}
315 		*addr = ia->ia_dstaddr;
316 		break;
317 
318 	case SIOCGIFNETMASK:
319 		*addr = ia->ia_sockmask;
320 		break;
321 	}
322 
323 	NET_EPOCH_EXIT(et);
324 
325 	return (error);
326 }
327 
328 static int
329 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
330 {
331 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
332 	const struct sockaddr_in *addr = &ifra->ifra_addr;
333 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
334 	const struct sockaddr_in *mask = &ifra->ifra_mask;
335 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
336 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
337 	struct epoch_tracker et;
338 	struct ifaddr *ifa;
339 	struct in_ifaddr *ia;
340 	bool iaIsFirst;
341 	int error = 0;
342 
343 	error = priv_check(td, PRIV_NET_ADDIFADDR);
344 	if (error)
345 		return (error);
346 
347 	/*
348 	 * ifra_addr must be present and be of INET family.
349 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
350 	 */
351 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
352 	    addr->sin_family != AF_INET)
353 		return (EINVAL);
354 	if (broadaddr->sin_len != 0 &&
355 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
356 	    broadaddr->sin_family != AF_INET))
357 		return (EINVAL);
358 	if (mask->sin_len != 0 &&
359 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
360 	    mask->sin_family != AF_INET))
361 		return (EINVAL);
362 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
363 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
364 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
365 		return (EDESTADDRREQ);
366 	if (vhid > 0 && carp_attach_p == NULL)
367 		return (EPROTONOSUPPORT);
368 
369 	/*
370 	 * See whether address already exist.
371 	 */
372 	iaIsFirst = true;
373 	ia = NULL;
374 	NET_EPOCH_ENTER(et);
375 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
376 		struct in_ifaddr *it;
377 
378 		if (ifa->ifa_addr->sa_family != AF_INET)
379 			continue;
380 
381 		it = (struct in_ifaddr *)ifa;
382 		iaIsFirst = false;
383 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
384 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
385 			ia = it;
386 	}
387 	NET_EPOCH_EXIT(et);
388 
389 	if (ia != NULL)
390 		(void )in_difaddr_ioctl(cmd, data, ifp, td);
391 
392 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
393 	ia = (struct in_ifaddr *)ifa;
394 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
395 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
396 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
397 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
398 	    CALLOUT_RETURNUNLOCKED);
399 
400 	ia->ia_ifp = ifp;
401 	ia->ia_addr = *addr;
402 	if (mask->sin_len != 0) {
403 		ia->ia_sockmask = *mask;
404 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
405 	} else {
406 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
407 
408 		/*
409 	 	 * Be compatible with network classes, if netmask isn't
410 		 * supplied, guess it based on classes.
411 	 	 */
412 		if (IN_CLASSA(i))
413 			ia->ia_subnetmask = IN_CLASSA_NET;
414 		else if (IN_CLASSB(i))
415 			ia->ia_subnetmask = IN_CLASSB_NET;
416 		else
417 			ia->ia_subnetmask = IN_CLASSC_NET;
418 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
419 	}
420 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
421 	in_socktrim(&ia->ia_sockmask);
422 
423 	if (ifp->if_flags & IFF_BROADCAST) {
424 		if (broadaddr->sin_len != 0) {
425 			ia->ia_broadaddr = *broadaddr;
426 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
427 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
428 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
429 			ia->ia_broadaddr.sin_family = AF_INET;
430 		} else {
431 			ia->ia_broadaddr.sin_addr.s_addr =
432 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
433 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
434 			ia->ia_broadaddr.sin_family = AF_INET;
435 		}
436 	}
437 
438 	if (ifp->if_flags & IFF_POINTOPOINT)
439 		ia->ia_dstaddr = *dstaddr;
440 
441 	/* XXXGL: rtinit() needs this strange assignment. */
442 	if (ifp->if_flags & IFF_LOOPBACK)
443                 ia->ia_dstaddr = ia->ia_addr;
444 
445 	if (vhid != 0) {
446 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
447 		if (error)
448 			return (error);
449 	}
450 
451 	/* if_addrhead is already referenced by ifa_alloc() */
452 	IF_ADDR_WLOCK(ifp);
453 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
454 	IF_ADDR_WUNLOCK(ifp);
455 
456 	ifa_ref(ifa);			/* in_ifaddrhead */
457 	IN_IFADDR_WLOCK();
458 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
459 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
460 	IN_IFADDR_WUNLOCK();
461 
462 	/*
463 	 * Give the interface a chance to initialize
464 	 * if this is its first address,
465 	 * and to validate the address if necessary.
466 	 */
467 	if (ifp->if_ioctl != NULL) {
468 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
469 		if (error)
470 			goto fail1;
471 	}
472 
473 	/*
474 	 * Add route for the network.
475 	 */
476 	if (vhid == 0) {
477 		int flags = RTF_UP;
478 
479 		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
480 			flags |= RTF_HOST;
481 
482 		error = in_addprefix(ia, flags);
483 		if (error)
484 			goto fail1;
485 	}
486 
487 	/*
488 	 * Add a loopback route to self.
489 	 */
490 	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
491 	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
492 	    !((ifp->if_flags & IFF_POINTOPOINT) &&
493 	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
494 		struct in_ifaddr *eia;
495 
496 		eia = in_localip_more(ia);
497 
498 		if (eia == NULL) {
499 			error = ifa_add_loopback_route((struct ifaddr *)ia,
500 			    (struct sockaddr *)&ia->ia_addr);
501 			if (error)
502 				goto fail2;
503 		} else
504 			ifa_free(&eia->ia_ifa);
505 	}
506 
507 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
508 		struct in_addr allhosts_addr;
509 		struct in_ifinfo *ii;
510 
511 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
512 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
513 
514 		error = in_joingroup(ifp, &allhosts_addr, NULL,
515 			&ii->ii_allhosts);
516 	}
517 
518 	/*
519 	 * Note: we don't need extra reference for ifa, since we called
520 	 * with sx lock held, and ifaddr can not be deleted in concurrent
521 	 * thread.
522 	 */
523 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
524 
525 	return (error);
526 
527 fail2:
528 	if (vhid == 0)
529 		(void )in_scrubprefix(ia, LLE_STATIC);
530 
531 fail1:
532 	if (ia->ia_ifa.ifa_carp)
533 		(*carp_detach_p)(&ia->ia_ifa, false);
534 
535 	IF_ADDR_WLOCK(ifp);
536 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
537 	IF_ADDR_WUNLOCK(ifp);
538 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
539 
540 	IN_IFADDR_WLOCK();
541 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
542 	LIST_REMOVE(ia, ia_hash);
543 	IN_IFADDR_WUNLOCK();
544 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
545 
546 	return (error);
547 }
548 
549 static int
550 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
551 {
552 	const struct ifreq *ifr = (struct ifreq *)data;
553 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
554 	    &ifr->ifr_addr;
555 	struct ifaddr *ifa;
556 	struct in_ifaddr *ia;
557 	bool deleteAny, iaIsLast;
558 	int error;
559 
560 	if (td != NULL) {
561 		error = priv_check(td, PRIV_NET_DELIFADDR);
562 		if (error)
563 			return (error);
564 	}
565 
566 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
567 	    addr->sin_family != AF_INET)
568 		deleteAny = true;
569 	else
570 		deleteAny = false;
571 
572 	iaIsLast = true;
573 	ia = NULL;
574 	IF_ADDR_WLOCK(ifp);
575 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
576 		struct in_ifaddr *it;
577 
578 		if (ifa->ifa_addr->sa_family != AF_INET)
579 			continue;
580 
581 		it = (struct in_ifaddr *)ifa;
582 		if (deleteAny && ia == NULL && (td == NULL ||
583 		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
584 			ia = it;
585 
586 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
587 		    (td == NULL || prison_check_ip4(td->td_ucred,
588 		    &addr->sin_addr) == 0))
589 			ia = it;
590 
591 		if (it != ia)
592 			iaIsLast = false;
593 	}
594 
595 	if (ia == NULL) {
596 		IF_ADDR_WUNLOCK(ifp);
597 		return (EADDRNOTAVAIL);
598 	}
599 
600 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
601 	IF_ADDR_WUNLOCK(ifp);
602 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
603 
604 	IN_IFADDR_WLOCK();
605 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
606 	LIST_REMOVE(ia, ia_hash);
607 	IN_IFADDR_WUNLOCK();
608 
609 	/*
610 	 * in_scrubprefix() kills the interface route.
611 	 */
612 	in_scrubprefix(ia, LLE_STATIC);
613 
614 	/*
615 	 * in_ifadown gets rid of all the rest of
616 	 * the routes.  This is not quite the right
617 	 * thing to do, but at least if we are running
618 	 * a routing process they will come back.
619 	 */
620 	in_ifadown(&ia->ia_ifa, 1);
621 
622 	if (ia->ia_ifa.ifa_carp)
623 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
624 
625 	/*
626 	 * If this is the last IPv4 address configured on this
627 	 * interface, leave the all-hosts group.
628 	 * No state-change report need be transmitted.
629 	 */
630 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
631 		struct in_ifinfo *ii;
632 
633 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
634 		if (ii->ii_allhosts) {
635 			(void)in_leavegroup(ii->ii_allhosts, NULL);
636 			ii->ii_allhosts = NULL;
637 		}
638 	}
639 
640 	IF_ADDR_WLOCK(ifp);
641 	if (callout_stop(&ia->ia_garp_timer) == 1) {
642 		ifa_free(&ia->ia_ifa);
643 	}
644 	IF_ADDR_WUNLOCK(ifp);
645 
646 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
647 	    IFADDR_EVENT_DEL);
648 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
649 
650 	return (0);
651 }
652 
653 #define rtinitflags(x) \
654 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
655 	    ? RTF_HOST : 0)
656 
657 /*
658  * Check if we have a route for the given prefix already or add one accordingly.
659  */
660 int
661 in_addprefix(struct in_ifaddr *target, int flags)
662 {
663 	struct rm_priotracker in_ifa_tracker;
664 	struct in_ifaddr *ia;
665 	struct in_addr prefix, mask, p, m;
666 	int error;
667 
668 	if ((flags & RTF_HOST) != 0) {
669 		prefix = target->ia_dstaddr.sin_addr;
670 		mask.s_addr = 0;
671 	} else {
672 		prefix = target->ia_addr.sin_addr;
673 		mask = target->ia_sockmask.sin_addr;
674 		prefix.s_addr &= mask.s_addr;
675 	}
676 
677 	IN_IFADDR_RLOCK(&in_ifa_tracker);
678 	/* Look for an existing address with the same prefix, mask, and fib */
679 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
680 		if (rtinitflags(ia)) {
681 			p = ia->ia_dstaddr.sin_addr;
682 
683 			if (prefix.s_addr != p.s_addr)
684 				continue;
685 		} else {
686 			p = ia->ia_addr.sin_addr;
687 			m = ia->ia_sockmask.sin_addr;
688 			p.s_addr &= m.s_addr;
689 
690 			if (prefix.s_addr != p.s_addr ||
691 			    mask.s_addr != m.s_addr)
692 				continue;
693 		}
694 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
695 			continue;
696 
697 		/*
698 		 * If we got a matching prefix route inserted by other
699 		 * interface address, we are done here.
700 		 */
701 		if (ia->ia_flags & IFA_ROUTE) {
702 #ifdef RADIX_MPATH
703 			if (ia->ia_addr.sin_addr.s_addr ==
704 			    target->ia_addr.sin_addr.s_addr) {
705 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
706 				return (EEXIST);
707 			} else
708 				break;
709 #endif
710 			if (V_nosameprefix) {
711 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
712 				return (EEXIST);
713 			} else {
714 				int fibnum;
715 
716 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
717 					target->ia_ifp->if_fib;
718 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
719 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
720 				return (0);
721 			}
722 		}
723 	}
724 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
725 
726 	/*
727 	 * No-one seem to have this prefix route, so we try to insert it.
728 	 */
729 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
730 	if (!error)
731 		target->ia_flags |= IFA_ROUTE;
732 	return (error);
733 }
734 
735 /*
736  * Removes either all lle entries for given @ia, or lle
737  * corresponding to @ia address.
738  */
739 static void
740 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
741 {
742 	struct sockaddr_in addr, mask;
743 	struct sockaddr *saddr, *smask;
744 	struct ifnet *ifp;
745 
746 	saddr = (struct sockaddr *)&addr;
747 	bzero(&addr, sizeof(addr));
748 	addr.sin_len = sizeof(addr);
749 	addr.sin_family = AF_INET;
750 	smask = (struct sockaddr *)&mask;
751 	bzero(&mask, sizeof(mask));
752 	mask.sin_len = sizeof(mask);
753 	mask.sin_family = AF_INET;
754 	mask.sin_addr.s_addr = ia->ia_subnetmask;
755 	ifp = ia->ia_ifp;
756 
757 	if (all) {
758 		/*
759 		 * Remove all L2 entries matching given prefix.
760 		 * Convert address to host representation to avoid
761 		 * doing this on every callback. ia_subnetmask is already
762 		 * stored in host representation.
763 		 */
764 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
765 		lltable_prefix_free(AF_INET, saddr, smask, flags);
766 	} else {
767 		/* Remove interface address only */
768 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
769 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
770 	}
771 }
772 
773 /*
774  * If there is no other address in the system that can serve a route to the
775  * same prefix, remove the route.  Hand over the route to the new address
776  * otherwise.
777  */
778 int
779 in_scrubprefix(struct in_ifaddr *target, u_int flags)
780 {
781 	struct rm_priotracker in_ifa_tracker;
782 	struct in_ifaddr *ia;
783 	struct in_addr prefix, mask, p, m;
784 	int error = 0;
785 
786 	/*
787 	 * Remove the loopback route to the interface address.
788 	 */
789 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
790 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
791 	    (flags & LLE_STATIC)) {
792 		struct in_ifaddr *eia;
793 
794 		/*
795 		 * XXXME: add fib-aware in_localip.
796 		 * We definitely don't want to switch between
797 		 * prefixes in different fibs.
798 		 */
799 		eia = in_localip_more(target);
800 
801 		if (eia != NULL) {
802 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
803 			    (struct sockaddr *)&target->ia_addr);
804 			ifa_free(&eia->ia_ifa);
805 		} else {
806 			error = ifa_del_loopback_route((struct ifaddr *)target,
807 			    (struct sockaddr *)&target->ia_addr);
808 		}
809 	}
810 
811 	if (rtinitflags(target)) {
812 		prefix = target->ia_dstaddr.sin_addr;
813 		mask.s_addr = 0;
814 	} else {
815 		prefix = target->ia_addr.sin_addr;
816 		mask = target->ia_sockmask.sin_addr;
817 		prefix.s_addr &= mask.s_addr;
818 	}
819 
820 	if ((target->ia_flags & IFA_ROUTE) == 0) {
821 		int fibnum;
822 
823 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
824 			target->ia_ifp->if_fib;
825 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
826 
827 		/*
828 		 * Removing address from !IFF_UP interface or
829 		 * prefix which exists on other interface (along with route).
830 		 * No entries should exist here except target addr.
831 		 * Given that, delete this entry only.
832 		 */
833 		in_scrubprefixlle(target, 0, flags);
834 		return (0);
835 	}
836 
837 	IN_IFADDR_RLOCK(&in_ifa_tracker);
838 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
839 		if (rtinitflags(ia)) {
840 			p = ia->ia_dstaddr.sin_addr;
841 
842 			if (prefix.s_addr != p.s_addr)
843 				continue;
844 		} else {
845 			p = ia->ia_addr.sin_addr;
846 			m = ia->ia_sockmask.sin_addr;
847 			p.s_addr &= m.s_addr;
848 
849 			if (prefix.s_addr != p.s_addr ||
850 			    mask.s_addr != m.s_addr)
851 				continue;
852 		}
853 
854 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
855 			continue;
856 
857 		/*
858 		 * If we got a matching prefix address, move IFA_ROUTE and
859 		 * the route itself to it.  Make sure that routing daemons
860 		 * get a heads-up.
861 		 */
862 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
863 			ifa_ref(&ia->ia_ifa);
864 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
865 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
866 			    rtinitflags(target));
867 			if (error == 0)
868 				target->ia_flags &= ~IFA_ROUTE;
869 			else
870 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
871 					error);
872 			/* Scrub all entries IFF interface is different */
873 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
874 			    flags);
875 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
876 			    rtinitflags(ia) | RTF_UP);
877 			if (error == 0)
878 				ia->ia_flags |= IFA_ROUTE;
879 			else
880 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
881 					error);
882 			ifa_free(&ia->ia_ifa);
883 			return (error);
884 		}
885 	}
886 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
887 
888 	/*
889 	 * remove all L2 entries on the given prefix
890 	 */
891 	in_scrubprefixlle(target, 1, flags);
892 
893 	/*
894 	 * As no-one seem to have this prefix, we can remove the route.
895 	 */
896 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
897 	if (error == 0)
898 		target->ia_flags &= ~IFA_ROUTE;
899 	else
900 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
901 	return (error);
902 }
903 
904 #undef rtinitflags
905 
906 void
907 in_ifscrub_all(void)
908 {
909 	struct ifnet *ifp;
910 	struct ifaddr *ifa, *nifa;
911 	struct ifaliasreq ifr;
912 
913 	IFNET_RLOCK();
914 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
915 		/* Cannot lock here - lock recursion. */
916 		/* NET_EPOCH_ENTER(et); */
917 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
918 			if (ifa->ifa_addr->sa_family != AF_INET)
919 				continue;
920 
921 			/*
922 			 * This is ugly but the only way for legacy IP to
923 			 * cleanly remove addresses and everything attached.
924 			 */
925 			bzero(&ifr, sizeof(ifr));
926 			ifr.ifra_addr = *ifa->ifa_addr;
927 			if (ifa->ifa_dstaddr)
928 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
929 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
930 			    ifp, NULL);
931 		}
932 		/* NET_EPOCH_EXIT(et); */
933 		in_purgemaddrs(ifp);
934 		igmp_domifdetach(ifp);
935 	}
936 	IFNET_RUNLOCK();
937 }
938 
939 int
940 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
941 {
942 
943 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
944 	     /*
945 	      * Check for old-style (host 0) broadcast, but
946 	      * taking into account that RFC 3021 obsoletes it.
947 	      */
948 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
949 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
950 	     /*
951 	      * Check for an all one subnetmask. These
952 	      * only exist when an interface gets a secondary
953 	      * address.
954 	      */
955 	    ia->ia_subnetmask != (u_long)0xffffffff);
956 }
957 
958 /*
959  * Return 1 if the address might be a local broadcast address.
960  */
961 int
962 in_broadcast(struct in_addr in, struct ifnet *ifp)
963 {
964 	struct ifaddr *ifa;
965 	int found;
966 
967 	NET_EPOCH_ASSERT();
968 
969 	if (in.s_addr == INADDR_BROADCAST ||
970 	    in.s_addr == INADDR_ANY)
971 		return (1);
972 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
973 		return (0);
974 	found = 0;
975 	/*
976 	 * Look through the list of addresses for a match
977 	 * with a broadcast address.
978 	 */
979 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
980 		if (ifa->ifa_addr->sa_family == AF_INET &&
981 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
982 			found = 1;
983 			break;
984 		}
985 	return (found);
986 }
987 
988 /*
989  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
990  */
991 void
992 in_ifdetach(struct ifnet *ifp)
993 {
994 	IN_MULTI_LOCK();
995 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
996 	in_pcbpurgeif0(&V_udbinfo, ifp);
997 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
998 	in_purgemaddrs(ifp);
999 	IN_MULTI_UNLOCK();
1000 
1001 	/*
1002 	 * Make sure all multicast deletions invoking if_ioctl() are
1003 	 * completed before returning. Else we risk accessing a freed
1004 	 * ifnet structure pointer.
1005 	 */
1006 	inm_release_wait(NULL);
1007 }
1008 
1009 /*
1010  * Delete all IPv4 multicast address records, and associated link-layer
1011  * multicast address records, associated with ifp.
1012  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1013  * XXX This should not race with ifma_protospec being set during
1014  * a new allocation, if it does, we have bigger problems.
1015  */
1016 static void
1017 in_purgemaddrs(struct ifnet *ifp)
1018 {
1019 	struct in_multi_head purgeinms;
1020 	struct in_multi		*inm;
1021 	struct ifmultiaddr	*ifma, *next;
1022 
1023 	SLIST_INIT(&purgeinms);
1024 	IN_MULTI_LIST_LOCK();
1025 
1026 	/*
1027 	 * Extract list of in_multi associated with the detaching ifp
1028 	 * which the PF_INET layer is about to release.
1029 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1030 	 * by code further down.
1031 	 */
1032 	IF_ADDR_WLOCK(ifp);
1033  restart:
1034 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1035 		if (ifma->ifma_addr->sa_family != AF_INET ||
1036 		    ifma->ifma_protospec == NULL)
1037 			continue;
1038 		inm = (struct in_multi *)ifma->ifma_protospec;
1039 		inm_rele_locked(&purgeinms, inm);
1040 		if (__predict_false(ifma_restart)) {
1041 			ifma_restart = true;
1042 			goto restart;
1043 		}
1044 	}
1045 	IF_ADDR_WUNLOCK(ifp);
1046 
1047 	inm_release_list_deferred(&purgeinms);
1048 	igmp_ifdetach(ifp);
1049 	IN_MULTI_LIST_UNLOCK();
1050 }
1051 
1052 struct in_llentry {
1053 	struct llentry		base;
1054 };
1055 
1056 #define	IN_LLTBL_DEFAULT_HSIZE	32
1057 #define	IN_LLTBL_HASH(k, h) \
1058 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1059 
1060 /*
1061  * Do actual deallocation of @lle.
1062  */
1063 static void
1064 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1065 {
1066 	struct llentry *lle;
1067 
1068 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1069 	LLE_LOCK_DESTROY(lle);
1070 	LLE_REQ_DESTROY(lle);
1071 	free(lle, M_LLTABLE);
1072 }
1073 
1074 /*
1075  * Called by the datapath to indicate that
1076  * the entry was used.
1077  */
1078 static void
1079 in_lltable_mark_used(struct llentry *lle)
1080 {
1081 
1082 	LLE_REQ_LOCK(lle);
1083 	lle->r_skip_req = 0;
1084 	LLE_REQ_UNLOCK(lle);
1085 }
1086 
1087 /*
1088  * Called by LLE_FREE_LOCKED when number of references
1089  * drops to zero.
1090  */
1091 static void
1092 in_lltable_destroy_lle(struct llentry *lle)
1093 {
1094 
1095 	LLE_WUNLOCK(lle);
1096 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1097 }
1098 
1099 static struct llentry *
1100 in_lltable_new(struct in_addr addr4, u_int flags)
1101 {
1102 	struct in_llentry *lle;
1103 
1104 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1105 	if (lle == NULL)		/* NB: caller generates msg */
1106 		return NULL;
1107 
1108 	/*
1109 	 * For IPv4 this will trigger "arpresolve" to generate
1110 	 * an ARP request.
1111 	 */
1112 	lle->base.la_expire = time_uptime; /* mark expired */
1113 	lle->base.r_l3addr.addr4 = addr4;
1114 	lle->base.lle_refcnt = 1;
1115 	lle->base.lle_free = in_lltable_destroy_lle;
1116 	LLE_LOCK_INIT(&lle->base);
1117 	LLE_REQ_INIT(&lle->base);
1118 	callout_init(&lle->base.lle_timer, 1);
1119 
1120 	return (&lle->base);
1121 }
1122 
1123 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1124 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1125 
1126 static int
1127 in_lltable_match_prefix(const struct sockaddr *saddr,
1128     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1129 {
1130 	struct in_addr addr, mask, lle_addr;
1131 
1132 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1133 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1134 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1135 
1136 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1137 		return (0);
1138 
1139 	if (lle->la_flags & LLE_IFADDR) {
1140 		/*
1141 		 * Delete LLE_IFADDR records IFF address & flag matches.
1142 		 * Note that addr is the interface address within prefix
1143 		 * being matched.
1144 		 * Note also we should handle 'ifdown' cases without removing
1145 		 * ifaddr macs.
1146 		 */
1147 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1148 			return (1);
1149 		return (0);
1150 	}
1151 
1152 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1153 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1154 		return (1);
1155 
1156 	return (0);
1157 }
1158 
1159 static void
1160 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1161 {
1162 	size_t pkts_dropped;
1163 
1164 	LLE_WLOCK_ASSERT(lle);
1165 	KASSERT(llt != NULL, ("lltable is NULL"));
1166 
1167 	/* Unlink entry from table if not already */
1168 	if ((lle->la_flags & LLE_LINKED) != 0) {
1169 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1170 		lltable_unlink_entry(llt, lle);
1171 	}
1172 
1173 	/* Drop hold queue */
1174 	pkts_dropped = llentry_free(lle);
1175 	ARPSTAT_ADD(dropped, pkts_dropped);
1176 }
1177 
1178 static int
1179 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1180 {
1181 	struct rt_addrinfo info;
1182 	struct sockaddr_in rt_key, rt_mask;
1183 	struct sockaddr rt_gateway;
1184 	int rt_flags;
1185 
1186 	KASSERT(l3addr->sa_family == AF_INET,
1187 	    ("sin_family %d", l3addr->sa_family));
1188 
1189 	bzero(&rt_key, sizeof(rt_key));
1190 	rt_key.sin_len = sizeof(rt_key);
1191 	bzero(&rt_mask, sizeof(rt_mask));
1192 	rt_mask.sin_len = sizeof(rt_mask);
1193 	bzero(&rt_gateway, sizeof(rt_gateway));
1194 	rt_gateway.sa_len = sizeof(rt_gateway);
1195 
1196 	bzero(&info, sizeof(info));
1197 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1198 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1199 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1200 
1201 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1202 		return (EINVAL);
1203 
1204 	rt_flags = info.rti_flags;
1205 
1206 	/*
1207 	 * If the gateway for an existing host route matches the target L3
1208 	 * address, which is a special route inserted by some implementation
1209 	 * such as MANET, and the interface is of the correct type, then
1210 	 * allow for ARP to proceed.
1211 	 */
1212 	if (rt_flags & RTF_GATEWAY) {
1213 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1214 		    info.rti_ifp->if_type != IFT_ETHER ||
1215 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1216 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1217 		    sizeof(in_addr_t)) != 0) {
1218 			rib_free_info(&info);
1219 			return (EINVAL);
1220 		}
1221 	}
1222 	rib_free_info(&info);
1223 
1224 	/*
1225 	 * Make sure that at least the destination address is covered
1226 	 * by the route. This is for handling the case where 2 or more
1227 	 * interfaces have the same prefix. An incoming packet arrives
1228 	 * on one interface and the corresponding outgoing packet leaves
1229 	 * another interface.
1230 	 */
1231 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1232 		const char *sa, *mask, *addr, *lim;
1233 		const struct sockaddr_in *l3sin;
1234 
1235 		mask = (const char *)&rt_mask;
1236 		/*
1237 		 * Just being extra cautious to avoid some custom
1238 		 * code getting into trouble.
1239 		 */
1240 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1241 			return (EINVAL);
1242 
1243 		sa = (const char *)&rt_key;
1244 		addr = (const char *)l3addr;
1245 		l3sin = (const struct sockaddr_in *)l3addr;
1246 		lim = addr + l3sin->sin_len;
1247 
1248 		for ( ; addr < lim; sa++, mask++, addr++) {
1249 			if ((*sa ^ *addr) & *mask) {
1250 #ifdef DIAGNOSTIC
1251 				char addrbuf[INET_ADDRSTRLEN];
1252 
1253 				log(LOG_INFO, "IPv4 address: \"%s\" "
1254 				    "is not on the network\n",
1255 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1256 #endif
1257 				return (EINVAL);
1258 			}
1259 		}
1260 	}
1261 
1262 	return (0);
1263 }
1264 
1265 static inline uint32_t
1266 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1267 {
1268 
1269 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1270 }
1271 
1272 static uint32_t
1273 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1274 {
1275 
1276 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1277 }
1278 
1279 static void
1280 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1281 {
1282 	struct sockaddr_in *sin;
1283 
1284 	sin = (struct sockaddr_in *)sa;
1285 	bzero(sin, sizeof(*sin));
1286 	sin->sin_family = AF_INET;
1287 	sin->sin_len = sizeof(*sin);
1288 	sin->sin_addr = lle->r_l3addr.addr4;
1289 }
1290 
1291 static inline struct llentry *
1292 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1293 {
1294 	struct llentry *lle;
1295 	struct llentries *lleh;
1296 	u_int hashidx;
1297 
1298 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1299 	lleh = &llt->lle_head[hashidx];
1300 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1301 		if (lle->la_flags & LLE_DELETED)
1302 			continue;
1303 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1304 			break;
1305 	}
1306 
1307 	return (lle);
1308 }
1309 
1310 static void
1311 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1312 {
1313 
1314 	lle->la_flags |= LLE_DELETED;
1315 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1316 #ifdef DIAGNOSTIC
1317 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1318 #endif
1319 	llentry_free(lle);
1320 }
1321 
1322 static struct llentry *
1323 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1324 {
1325 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1326 	struct ifnet *ifp = llt->llt_ifp;
1327 	struct llentry *lle;
1328 	char linkhdr[LLE_MAX_LINKHDR];
1329 	size_t linkhdrsize;
1330 	int lladdr_off;
1331 
1332 	KASSERT(l3addr->sa_family == AF_INET,
1333 	    ("sin_family %d", l3addr->sa_family));
1334 
1335 	/*
1336 	 * A route that covers the given address must have
1337 	 * been installed 1st because we are doing a resolution,
1338 	 * verify this.
1339 	 */
1340 	if (!(flags & LLE_IFADDR) &&
1341 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1342 		return (NULL);
1343 
1344 	lle = in_lltable_new(sin->sin_addr, flags);
1345 	if (lle == NULL) {
1346 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1347 		return (NULL);
1348 	}
1349 	lle->la_flags = flags;
1350 	if (flags & LLE_STATIC)
1351 		lle->r_flags |= RLLE_VALID;
1352 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1353 		linkhdrsize = LLE_MAX_LINKHDR;
1354 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1355 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1356 			NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1357 			return (NULL);
1358 		}
1359 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1360 		    lladdr_off);
1361 		lle->la_flags |= LLE_STATIC;
1362 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1363 	}
1364 
1365 	return (lle);
1366 }
1367 
1368 /*
1369  * Return NULL if not found or marked for deletion.
1370  * If found return lle read locked.
1371  */
1372 static struct llentry *
1373 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1374 {
1375 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1376 	struct llentry *lle;
1377 
1378 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1379 	KASSERT(l3addr->sa_family == AF_INET,
1380 	    ("sin_family %d", l3addr->sa_family));
1381 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1382 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1383 	    ("wrong lle request flags: %#x", flags));
1384 
1385 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1386 	if (lle == NULL)
1387 		return (NULL);
1388 	if (flags & LLE_UNLOCKED)
1389 		return (lle);
1390 
1391 	if (flags & LLE_EXCLUSIVE)
1392 		LLE_WLOCK(lle);
1393 	else
1394 		LLE_RLOCK(lle);
1395 
1396 	/*
1397 	 * If the afdata lock is not held, the LLE may have been unlinked while
1398 	 * we were blocked on the LLE lock.  Check for this case.
1399 	 */
1400 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1401 		if (flags & LLE_EXCLUSIVE)
1402 			LLE_WUNLOCK(lle);
1403 		else
1404 			LLE_RUNLOCK(lle);
1405 		return (NULL);
1406 	}
1407 	return (lle);
1408 }
1409 
1410 static int
1411 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1412     struct sysctl_req *wr)
1413 {
1414 	struct ifnet *ifp = llt->llt_ifp;
1415 	/* XXX stack use */
1416 	struct {
1417 		struct rt_msghdr	rtm;
1418 		struct sockaddr_in	sin;
1419 		struct sockaddr_dl	sdl;
1420 	} arpc;
1421 	struct sockaddr_dl *sdl;
1422 	int error;
1423 
1424 	bzero(&arpc, sizeof(arpc));
1425 	/* skip deleted entries */
1426 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1427 		return (0);
1428 	/* Skip if jailed and not a valid IP of the prison. */
1429 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1430 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1431 		return (0);
1432 	/*
1433 	 * produce a msg made of:
1434 	 *  struct rt_msghdr;
1435 	 *  struct sockaddr_in; (IPv4)
1436 	 *  struct sockaddr_dl;
1437 	 */
1438 	arpc.rtm.rtm_msglen = sizeof(arpc);
1439 	arpc.rtm.rtm_version = RTM_VERSION;
1440 	arpc.rtm.rtm_type = RTM_GET;
1441 	arpc.rtm.rtm_flags = RTF_UP;
1442 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1443 
1444 	/* publish */
1445 	if (lle->la_flags & LLE_PUB)
1446 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1447 
1448 	sdl = &arpc.sdl;
1449 	sdl->sdl_family = AF_LINK;
1450 	sdl->sdl_len = sizeof(*sdl);
1451 	sdl->sdl_index = ifp->if_index;
1452 	sdl->sdl_type = ifp->if_type;
1453 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1454 		sdl->sdl_alen = ifp->if_addrlen;
1455 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1456 	} else {
1457 		sdl->sdl_alen = 0;
1458 		bzero(LLADDR(sdl), ifp->if_addrlen);
1459 	}
1460 
1461 	arpc.rtm.rtm_rmx.rmx_expire =
1462 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1463 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1464 	if (lle->la_flags & LLE_STATIC)
1465 		arpc.rtm.rtm_flags |= RTF_STATIC;
1466 	if (lle->la_flags & LLE_IFADDR)
1467 		arpc.rtm.rtm_flags |= RTF_PINNED;
1468 	arpc.rtm.rtm_index = ifp->if_index;
1469 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1470 
1471 	return (error);
1472 }
1473 
1474 static struct lltable *
1475 in_lltattach(struct ifnet *ifp)
1476 {
1477 	struct lltable *llt;
1478 
1479 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1480  	llt->llt_af = AF_INET;
1481  	llt->llt_ifp = ifp;
1482 
1483 	llt->llt_lookup = in_lltable_lookup;
1484 	llt->llt_alloc_entry = in_lltable_alloc;
1485 	llt->llt_delete_entry = in_lltable_delete_entry;
1486 	llt->llt_dump_entry = in_lltable_dump_entry;
1487 	llt->llt_hash = in_lltable_hash;
1488 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1489 	llt->llt_free_entry = in_lltable_free_entry;
1490 	llt->llt_match_prefix = in_lltable_match_prefix;
1491 	llt->llt_mark_used = in_lltable_mark_used;
1492  	lltable_link(llt);
1493 
1494 	return (llt);
1495 }
1496 
1497 void *
1498 in_domifattach(struct ifnet *ifp)
1499 {
1500 	struct in_ifinfo *ii;
1501 
1502 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1503 
1504 	ii->ii_llt = in_lltattach(ifp);
1505 	ii->ii_igmp = igmp_domifattach(ifp);
1506 
1507 	return (ii);
1508 }
1509 
1510 void
1511 in_domifdetach(struct ifnet *ifp, void *aux)
1512 {
1513 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1514 
1515 	igmp_domifdetach(ifp);
1516 	lltable_free(ii->ii_llt);
1517 	free(ii, M_IFADDR);
1518 }
1519