xref: /freebsd/sys/netinet/in.c (revision e2eeea75eb8b6dd50c1298067a0655880d186734)
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 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct thread *);
76 
77 static void	in_socktrim(struct sockaddr_in *);
78 static void	in_purgemaddrs(struct ifnet *);
79 
80 VNET_DEFINE_STATIC(int, nosameprefix);
81 #define	V_nosameprefix			VNET(nosameprefix)
82 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW,
83 	&VNET_NAME(nosameprefix), 0,
84 	"Refuse to create same prefixes on different interfaces");
85 
86 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
87 #define	V_ripcbinfo			VNET(ripcbinfo)
88 
89 static struct sx in_control_sx;
90 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control");
91 
92 /*
93  * Return 1 if an internet address is for a ``local'' host
94  * (one to which we have a connection).
95  */
96 int
97 in_localaddr(struct in_addr in)
98 {
99 	struct rm_priotracker in_ifa_tracker;
100 	u_long i = ntohl(in.s_addr);
101 	struct in_ifaddr *ia;
102 
103 	IN_IFADDR_RLOCK(&in_ifa_tracker);
104 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
105 		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
106 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
107 			return (1);
108 		}
109 	}
110 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
111 	return (0);
112 }
113 
114 /*
115  * Return 1 if an internet address is for the local host and configured
116  * on one of its interfaces.
117  */
118 int
119 in_localip(struct in_addr in)
120 {
121 	struct rm_priotracker in_ifa_tracker;
122 	struct in_ifaddr *ia;
123 
124 	IN_IFADDR_RLOCK(&in_ifa_tracker);
125 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
126 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
127 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
128 			return (1);
129 		}
130 	}
131 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
132 	return (0);
133 }
134 
135 /*
136  * Return 1 if an internet address is configured on an interface.
137  */
138 int
139 in_ifhasaddr(struct ifnet *ifp, struct in_addr in)
140 {
141 	struct ifaddr *ifa;
142 	struct in_ifaddr *ia;
143 
144 	NET_EPOCH_ASSERT();
145 
146 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
147 		if (ifa->ifa_addr->sa_family != AF_INET)
148 			continue;
149 		ia = (struct in_ifaddr *)ifa;
150 		if (ia->ia_addr.sin_addr.s_addr == in.s_addr)
151 			return (1);
152 	}
153 
154 	return (0);
155 }
156 
157 /*
158  * Return a reference to the interface address which is different to
159  * the supplied one but with same IP address value.
160  */
161 static struct in_ifaddr *
162 in_localip_more(struct in_ifaddr *ia)
163 {
164 	struct rm_priotracker in_ifa_tracker;
165 	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
166 	struct in_ifaddr *it;
167 
168 	IN_IFADDR_RLOCK(&in_ifa_tracker);
169 	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
170 		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
171 			ifa_ref(&it->ia_ifa);
172 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
173 			return (it);
174 		}
175 	}
176 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
177 
178 	return (NULL);
179 }
180 
181 /*
182  * Determine whether an IP address is in a reserved set of addresses
183  * that may not be forwarded, or whether datagrams to that destination
184  * may be forwarded.
185  */
186 int
187 in_canforward(struct in_addr in)
188 {
189 	u_long i = ntohl(in.s_addr);
190 
191 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
192 	    IN_ZERONET(i) || IN_LOOPBACK(i))
193 		return (0);
194 	return (1);
195 }
196 
197 /*
198  * Trim a mask in a sockaddr
199  */
200 static void
201 in_socktrim(struct sockaddr_in *ap)
202 {
203     char *cplim = (char *) &ap->sin_addr;
204     char *cp = (char *) (&ap->sin_addr + 1);
205 
206     ap->sin_len = 0;
207     while (--cp >= cplim)
208 	if (*cp) {
209 	    (ap)->sin_len = cp - (char *) (ap) + 1;
210 	    break;
211 	}
212 }
213 
214 /*
215  * Generic internet control operations (ioctl's).
216  */
217 int
218 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
219     struct thread *td)
220 {
221 	struct ifreq *ifr = (struct ifreq *)data;
222 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
223 	struct epoch_tracker et;
224 	struct ifaddr *ifa;
225 	struct in_ifaddr *ia;
226 	int error;
227 
228 	if (ifp == NULL)
229 		return (EADDRNOTAVAIL);
230 
231 	/*
232 	 * Filter out 4 ioctls we implement directly.  Forward the rest
233 	 * to specific functions and ifp->if_ioctl().
234 	 */
235 	switch (cmd) {
236 	case SIOCGIFADDR:
237 	case SIOCGIFBRDADDR:
238 	case SIOCGIFDSTADDR:
239 	case SIOCGIFNETMASK:
240 		break;
241 	case SIOCGIFALIAS:
242 		sx_xlock(&in_control_sx);
243 		error = in_gifaddr_ioctl(cmd, data, ifp, td);
244 		sx_xunlock(&in_control_sx);
245 		return (error);
246 	case SIOCDIFADDR:
247 		sx_xlock(&in_control_sx);
248 		error = in_difaddr_ioctl(cmd, data, ifp, td);
249 		sx_xunlock(&in_control_sx);
250 		return (error);
251 	case OSIOCAIFADDR:	/* 9.x compat */
252 	case SIOCAIFADDR:
253 		sx_xlock(&in_control_sx);
254 		error = in_aifaddr_ioctl(cmd, data, ifp, td);
255 		sx_xunlock(&in_control_sx);
256 		return (error);
257 	case SIOCSIFADDR:
258 	case SIOCSIFBRDADDR:
259 	case SIOCSIFDSTADDR:
260 	case SIOCSIFNETMASK:
261 		/* We no longer support that old commands. */
262 		return (EINVAL);
263 	default:
264 		if (ifp->if_ioctl == NULL)
265 			return (EOPNOTSUPP);
266 		return ((*ifp->if_ioctl)(ifp, cmd, data));
267 	}
268 
269 	if (addr->sin_addr.s_addr != INADDR_ANY &&
270 	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
271 		return (EADDRNOTAVAIL);
272 
273 	/*
274 	 * Find address for this interface, if it exists.  If an
275 	 * address was specified, find that one instead of the
276 	 * first one on the interface, if possible.
277 	 */
278 	NET_EPOCH_ENTER(et);
279 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
280 		if (ifa->ifa_addr->sa_family != AF_INET)
281 			continue;
282 		ia = (struct in_ifaddr *)ifa;
283 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
284 			break;
285 	}
286 	if (ifa == NULL)
287 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
288 			if (ifa->ifa_addr->sa_family == AF_INET) {
289 				ia = (struct in_ifaddr *)ifa;
290 				if (prison_check_ip4(td->td_ucred,
291 				    &ia->ia_addr.sin_addr) == 0)
292 					break;
293 			}
294 
295 	if (ifa == NULL) {
296 		NET_EPOCH_EXIT(et);
297 		return (EADDRNOTAVAIL);
298 	}
299 
300 	error = 0;
301 	switch (cmd) {
302 	case SIOCGIFADDR:
303 		*addr = ia->ia_addr;
304 		break;
305 
306 	case SIOCGIFBRDADDR:
307 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
308 			error = EINVAL;
309 			break;
310 		}
311 		*addr = ia->ia_broadaddr;
312 		break;
313 
314 	case SIOCGIFDSTADDR:
315 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
316 			error = EINVAL;
317 			break;
318 		}
319 		*addr = ia->ia_dstaddr;
320 		break;
321 
322 	case SIOCGIFNETMASK:
323 		*addr = ia->ia_sockmask;
324 		break;
325 	}
326 
327 	NET_EPOCH_EXIT(et);
328 
329 	return (error);
330 }
331 
332 static int
333 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
334 {
335 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
336 	const struct sockaddr_in *addr = &ifra->ifra_addr;
337 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
338 	const struct sockaddr_in *mask = &ifra->ifra_mask;
339 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
340 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
341 	struct epoch_tracker et;
342 	struct ifaddr *ifa;
343 	struct in_ifaddr *ia;
344 	bool iaIsFirst;
345 	int error = 0;
346 
347 	error = priv_check(td, PRIV_NET_ADDIFADDR);
348 	if (error)
349 		return (error);
350 
351 	/*
352 	 * ifra_addr must be present and be of INET family.
353 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
354 	 */
355 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
356 	    addr->sin_family != AF_INET)
357 		return (EINVAL);
358 	if (broadaddr->sin_len != 0 &&
359 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
360 	    broadaddr->sin_family != AF_INET))
361 		return (EINVAL);
362 	if (mask->sin_len != 0 &&
363 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
364 	    mask->sin_family != AF_INET))
365 		return (EINVAL);
366 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
367 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
368 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
369 		return (EDESTADDRREQ);
370 	if (vhid > 0 && carp_attach_p == NULL)
371 		return (EPROTONOSUPPORT);
372 
373 	/*
374 	 * See whether address already exist.
375 	 */
376 	iaIsFirst = true;
377 	ia = NULL;
378 	NET_EPOCH_ENTER(et);
379 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
380 		struct in_ifaddr *it;
381 
382 		if (ifa->ifa_addr->sa_family != AF_INET)
383 			continue;
384 
385 		it = (struct in_ifaddr *)ifa;
386 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
387 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
388 			ia = it;
389 		else
390 			iaIsFirst = false;
391 	}
392 	NET_EPOCH_EXIT(et);
393 
394 	if (ia != NULL)
395 		(void )in_difaddr_ioctl(cmd, data, ifp, td);
396 
397 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
398 	ia = (struct in_ifaddr *)ifa;
399 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
400 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
401 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
402 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
403 	    CALLOUT_RETURNUNLOCKED);
404 
405 	ia->ia_ifp = ifp;
406 	ia->ia_addr = *addr;
407 	if (mask->sin_len != 0) {
408 		ia->ia_sockmask = *mask;
409 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
410 	} else {
411 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
412 
413 		/*
414 	 	 * Be compatible with network classes, if netmask isn't
415 		 * supplied, guess it based on classes.
416 	 	 */
417 		if (IN_CLASSA(i))
418 			ia->ia_subnetmask = IN_CLASSA_NET;
419 		else if (IN_CLASSB(i))
420 			ia->ia_subnetmask = IN_CLASSB_NET;
421 		else
422 			ia->ia_subnetmask = IN_CLASSC_NET;
423 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
424 	}
425 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
426 	in_socktrim(&ia->ia_sockmask);
427 
428 	if (ifp->if_flags & IFF_BROADCAST) {
429 		if (broadaddr->sin_len != 0) {
430 			ia->ia_broadaddr = *broadaddr;
431 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
432 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
433 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
434 			ia->ia_broadaddr.sin_family = AF_INET;
435 		} else {
436 			ia->ia_broadaddr.sin_addr.s_addr =
437 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
438 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
439 			ia->ia_broadaddr.sin_family = AF_INET;
440 		}
441 	}
442 
443 	if (ifp->if_flags & IFF_POINTOPOINT)
444 		ia->ia_dstaddr = *dstaddr;
445 
446 	/* XXXGL: rtinit() needs this strange assignment. */
447 	if (ifp->if_flags & IFF_LOOPBACK)
448                 ia->ia_dstaddr = ia->ia_addr;
449 
450 	if (vhid != 0) {
451 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
452 		if (error)
453 			return (error);
454 	}
455 
456 	/* if_addrhead is already referenced by ifa_alloc() */
457 	IF_ADDR_WLOCK(ifp);
458 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
459 	IF_ADDR_WUNLOCK(ifp);
460 
461 	ifa_ref(ifa);			/* in_ifaddrhead */
462 	IN_IFADDR_WLOCK();
463 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
464 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
465 	IN_IFADDR_WUNLOCK();
466 
467 	/*
468 	 * Give the interface a chance to initialize
469 	 * if this is its first address,
470 	 * and to validate the address if necessary.
471 	 */
472 	if (ifp->if_ioctl != NULL) {
473 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
474 		if (error)
475 			goto fail1;
476 	}
477 
478 	/*
479 	 * Add route for the network.
480 	 */
481 	if (vhid == 0) {
482 		int flags = RTF_UP;
483 
484 		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
485 			flags |= RTF_HOST;
486 
487 		error = in_addprefix(ia, flags);
488 		if (error)
489 			goto fail1;
490 	}
491 
492 	/*
493 	 * Add a loopback route to self.
494 	 */
495 	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
496 	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
497 	    !((ifp->if_flags & IFF_POINTOPOINT) &&
498 	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
499 		struct in_ifaddr *eia;
500 
501 		eia = in_localip_more(ia);
502 
503 		if (eia == NULL) {
504 			error = ifa_add_loopback_route((struct ifaddr *)ia,
505 			    (struct sockaddr *)&ia->ia_addr);
506 			if (error)
507 				goto fail2;
508 		} else
509 			ifa_free(&eia->ia_ifa);
510 	}
511 
512 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
513 		struct in_addr allhosts_addr;
514 		struct in_ifinfo *ii;
515 
516 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
517 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
518 
519 		error = in_joingroup(ifp, &allhosts_addr, NULL,
520 			&ii->ii_allhosts);
521 	}
522 
523 	/*
524 	 * Note: we don't need extra reference for ifa, since we called
525 	 * with sx lock held, and ifaddr can not be deleted in concurrent
526 	 * thread.
527 	 */
528 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
529 
530 	return (error);
531 
532 fail2:
533 	if (vhid == 0)
534 		(void )in_scrubprefix(ia, LLE_STATIC);
535 
536 fail1:
537 	if (ia->ia_ifa.ifa_carp)
538 		(*carp_detach_p)(&ia->ia_ifa, false);
539 
540 	IF_ADDR_WLOCK(ifp);
541 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
542 	IF_ADDR_WUNLOCK(ifp);
543 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
544 
545 	IN_IFADDR_WLOCK();
546 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
547 	LIST_REMOVE(ia, ia_hash);
548 	IN_IFADDR_WUNLOCK();
549 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
550 
551 	return (error);
552 }
553 
554 static int
555 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
556 {
557 	const struct ifreq *ifr = (struct ifreq *)data;
558 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
559 	    &ifr->ifr_addr;
560 	struct ifaddr *ifa;
561 	struct in_ifaddr *ia;
562 	bool deleteAny, iaIsLast;
563 	int error;
564 
565 	if (td != NULL) {
566 		error = priv_check(td, PRIV_NET_DELIFADDR);
567 		if (error)
568 			return (error);
569 	}
570 
571 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
572 	    addr->sin_family != AF_INET)
573 		deleteAny = true;
574 	else
575 		deleteAny = false;
576 
577 	iaIsLast = true;
578 	ia = NULL;
579 	IF_ADDR_WLOCK(ifp);
580 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
581 		struct in_ifaddr *it;
582 
583 		if (ifa->ifa_addr->sa_family != AF_INET)
584 			continue;
585 
586 		it = (struct in_ifaddr *)ifa;
587 		if (deleteAny && ia == NULL && (td == NULL ||
588 		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
589 			ia = it;
590 
591 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
592 		    (td == NULL || prison_check_ip4(td->td_ucred,
593 		    &addr->sin_addr) == 0))
594 			ia = it;
595 
596 		if (it != ia)
597 			iaIsLast = false;
598 	}
599 
600 	if (ia == NULL) {
601 		IF_ADDR_WUNLOCK(ifp);
602 		return (EADDRNOTAVAIL);
603 	}
604 
605 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
606 	IF_ADDR_WUNLOCK(ifp);
607 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
608 
609 	IN_IFADDR_WLOCK();
610 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
611 	LIST_REMOVE(ia, ia_hash);
612 	IN_IFADDR_WUNLOCK();
613 
614 	/*
615 	 * in_scrubprefix() kills the interface route.
616 	 */
617 	in_scrubprefix(ia, LLE_STATIC);
618 
619 	/*
620 	 * in_ifadown gets rid of all the rest of
621 	 * the routes.  This is not quite the right
622 	 * thing to do, but at least if we are running
623 	 * a routing process they will come back.
624 	 */
625 	in_ifadown(&ia->ia_ifa, 1);
626 
627 	if (ia->ia_ifa.ifa_carp)
628 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
629 
630 	/*
631 	 * If this is the last IPv4 address configured on this
632 	 * interface, leave the all-hosts group.
633 	 * No state-change report need be transmitted.
634 	 */
635 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
636 		struct in_ifinfo *ii;
637 
638 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
639 		if (ii->ii_allhosts) {
640 			(void)in_leavegroup(ii->ii_allhosts, NULL);
641 			ii->ii_allhosts = NULL;
642 		}
643 	}
644 
645 	IF_ADDR_WLOCK(ifp);
646 	if (callout_stop(&ia->ia_garp_timer) == 1) {
647 		ifa_free(&ia->ia_ifa);
648 	}
649 	IF_ADDR_WUNLOCK(ifp);
650 
651 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
652 	    IFADDR_EVENT_DEL);
653 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
654 
655 	return (0);
656 }
657 
658 static int
659 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
660 {
661 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
662 	const struct sockaddr_in *addr = &ifra->ifra_addr;
663 	struct epoch_tracker et;
664 	struct ifaddr *ifa;
665 	struct in_ifaddr *ia;
666 
667 	/*
668 	 * ifra_addr must be present and be of INET family.
669 	 */
670 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
671 	    addr->sin_family != AF_INET)
672 		return (EINVAL);
673 
674 	/*
675 	 * See whether address exist.
676 	 */
677 	ia = NULL;
678 	NET_EPOCH_ENTER(et);
679 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
680 		struct in_ifaddr *it;
681 
682 		if (ifa->ifa_addr->sa_family != AF_INET)
683 			continue;
684 
685 		it = (struct in_ifaddr *)ifa;
686 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
687 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0) {
688 			ia = it;
689 			break;
690 		}
691 	}
692 	if (ia == NULL) {
693 		NET_EPOCH_EXIT(et);
694 		return (EADDRNOTAVAIL);
695 	}
696 
697 	ifra->ifra_mask = ia->ia_sockmask;
698 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
699 	    ia->ia_dstaddr.sin_family == AF_INET)
700 		ifra->ifra_dstaddr = ia->ia_dstaddr;
701 	else if ((ifp->if_flags & IFF_BROADCAST) &&
702 	    ia->ia_broadaddr.sin_family == AF_INET)
703 		ifra->ifra_broadaddr = ia->ia_broadaddr;
704 	else
705 		memset(&ifra->ifra_broadaddr, 0,
706 		    sizeof(ifra->ifra_broadaddr));
707 
708 	NET_EPOCH_EXIT(et);
709 	return (0);
710 }
711 
712 #define rtinitflags(x) \
713 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
714 	    ? RTF_HOST : 0)
715 
716 /*
717  * Check if we have a route for the given prefix already or add one accordingly.
718  */
719 int
720 in_addprefix(struct in_ifaddr *target, int flags)
721 {
722 	struct rm_priotracker in_ifa_tracker;
723 	struct in_ifaddr *ia;
724 	struct in_addr prefix, mask, p, m;
725 	int error;
726 
727 	if ((flags & RTF_HOST) != 0) {
728 		prefix = target->ia_dstaddr.sin_addr;
729 		mask.s_addr = 0;
730 	} else {
731 		prefix = target->ia_addr.sin_addr;
732 		mask = target->ia_sockmask.sin_addr;
733 		prefix.s_addr &= mask.s_addr;
734 	}
735 
736 	IN_IFADDR_RLOCK(&in_ifa_tracker);
737 	/* Look for an existing address with the same prefix, mask, and fib */
738 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
739 		if (rtinitflags(ia)) {
740 			p = ia->ia_dstaddr.sin_addr;
741 
742 			if (prefix.s_addr != p.s_addr)
743 				continue;
744 		} else {
745 			p = ia->ia_addr.sin_addr;
746 			m = ia->ia_sockmask.sin_addr;
747 			p.s_addr &= m.s_addr;
748 
749 			if (prefix.s_addr != p.s_addr ||
750 			    mask.s_addr != m.s_addr)
751 				continue;
752 		}
753 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
754 			continue;
755 
756 		/*
757 		 * If we got a matching prefix route inserted by other
758 		 * interface address, we are done here.
759 		 */
760 		if (ia->ia_flags & IFA_ROUTE) {
761 			if (V_nosameprefix) {
762 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
763 				return (EEXIST);
764 			} else {
765 				int fibnum;
766 
767 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
768 					target->ia_ifp->if_fib;
769 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
770 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
771 				return (0);
772 			}
773 		}
774 	}
775 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
776 
777 	/*
778 	 * No-one seem to have this prefix route, so we try to insert it.
779 	 */
780 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
781 	if (!error)
782 		target->ia_flags |= IFA_ROUTE;
783 	return (error);
784 }
785 
786 /*
787  * Removes either all lle entries for given @ia, or lle
788  * corresponding to @ia address.
789  */
790 static void
791 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
792 {
793 	struct sockaddr_in addr, mask;
794 	struct sockaddr *saddr, *smask;
795 	struct ifnet *ifp;
796 
797 	saddr = (struct sockaddr *)&addr;
798 	bzero(&addr, sizeof(addr));
799 	addr.sin_len = sizeof(addr);
800 	addr.sin_family = AF_INET;
801 	smask = (struct sockaddr *)&mask;
802 	bzero(&mask, sizeof(mask));
803 	mask.sin_len = sizeof(mask);
804 	mask.sin_family = AF_INET;
805 	mask.sin_addr.s_addr = ia->ia_subnetmask;
806 	ifp = ia->ia_ifp;
807 
808 	if (all) {
809 		/*
810 		 * Remove all L2 entries matching given prefix.
811 		 * Convert address to host representation to avoid
812 		 * doing this on every callback. ia_subnetmask is already
813 		 * stored in host representation.
814 		 */
815 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
816 		lltable_prefix_free(AF_INET, saddr, smask, flags);
817 	} else {
818 		/* Remove interface address only */
819 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
820 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
821 	}
822 }
823 
824 /*
825  * If there is no other address in the system that can serve a route to the
826  * same prefix, remove the route.  Hand over the route to the new address
827  * otherwise.
828  */
829 int
830 in_scrubprefix(struct in_ifaddr *target, u_int flags)
831 {
832 	struct rm_priotracker in_ifa_tracker;
833 	struct in_ifaddr *ia;
834 	struct in_addr prefix, mask, p, m;
835 	int error = 0;
836 
837 	/*
838 	 * Remove the loopback route to the interface address.
839 	 */
840 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
841 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
842 	    (flags & LLE_STATIC)) {
843 		struct in_ifaddr *eia;
844 
845 		/*
846 		 * XXXME: add fib-aware in_localip.
847 		 * We definitely don't want to switch between
848 		 * prefixes in different fibs.
849 		 */
850 		eia = in_localip_more(target);
851 
852 		if (eia != NULL) {
853 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
854 			    (struct sockaddr *)&target->ia_addr);
855 			ifa_free(&eia->ia_ifa);
856 		} else {
857 			error = ifa_del_loopback_route((struct ifaddr *)target,
858 			    (struct sockaddr *)&target->ia_addr);
859 		}
860 	}
861 
862 	if (rtinitflags(target)) {
863 		prefix = target->ia_dstaddr.sin_addr;
864 		mask.s_addr = 0;
865 	} else {
866 		prefix = target->ia_addr.sin_addr;
867 		mask = target->ia_sockmask.sin_addr;
868 		prefix.s_addr &= mask.s_addr;
869 	}
870 
871 	if ((target->ia_flags & IFA_ROUTE) == 0) {
872 		int fibnum;
873 
874 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
875 			target->ia_ifp->if_fib;
876 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
877 
878 		/*
879 		 * Removing address from !IFF_UP interface or
880 		 * prefix which exists on other interface (along with route).
881 		 * No entries should exist here except target addr.
882 		 * Given that, delete this entry only.
883 		 */
884 		in_scrubprefixlle(target, 0, flags);
885 		return (0);
886 	}
887 
888 	IN_IFADDR_RLOCK(&in_ifa_tracker);
889 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
890 		if (rtinitflags(ia)) {
891 			p = ia->ia_dstaddr.sin_addr;
892 
893 			if (prefix.s_addr != p.s_addr)
894 				continue;
895 		} else {
896 			p = ia->ia_addr.sin_addr;
897 			m = ia->ia_sockmask.sin_addr;
898 			p.s_addr &= m.s_addr;
899 
900 			if (prefix.s_addr != p.s_addr ||
901 			    mask.s_addr != m.s_addr)
902 				continue;
903 		}
904 
905 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
906 			continue;
907 
908 		/*
909 		 * If we got a matching prefix address, move IFA_ROUTE and
910 		 * the route itself to it.  Make sure that routing daemons
911 		 * get a heads-up.
912 		 */
913 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
914 			ifa_ref(&ia->ia_ifa);
915 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
916 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
917 			    rtinitflags(target));
918 			if (error == 0)
919 				target->ia_flags &= ~IFA_ROUTE;
920 			else
921 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
922 					error);
923 			/* Scrub all entries IFF interface is different */
924 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
925 			    flags);
926 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
927 			    rtinitflags(ia) | RTF_UP);
928 			if (error == 0)
929 				ia->ia_flags |= IFA_ROUTE;
930 			else
931 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
932 					error);
933 			ifa_free(&ia->ia_ifa);
934 			return (error);
935 		}
936 	}
937 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
938 
939 	/*
940 	 * remove all L2 entries on the given prefix
941 	 */
942 	in_scrubprefixlle(target, 1, flags);
943 
944 	/*
945 	 * As no-one seem to have this prefix, we can remove the route.
946 	 */
947 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
948 	if (error == 0)
949 		target->ia_flags &= ~IFA_ROUTE;
950 	else
951 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
952 	return (error);
953 }
954 
955 #undef rtinitflags
956 
957 void
958 in_ifscrub_all(void)
959 {
960 	struct ifnet *ifp;
961 	struct ifaddr *ifa, *nifa;
962 	struct ifaliasreq ifr;
963 
964 	IFNET_RLOCK();
965 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
966 		/* Cannot lock here - lock recursion. */
967 		/* NET_EPOCH_ENTER(et); */
968 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
969 			if (ifa->ifa_addr->sa_family != AF_INET)
970 				continue;
971 
972 			/*
973 			 * This is ugly but the only way for legacy IP to
974 			 * cleanly remove addresses and everything attached.
975 			 */
976 			bzero(&ifr, sizeof(ifr));
977 			ifr.ifra_addr = *ifa->ifa_addr;
978 			if (ifa->ifa_dstaddr)
979 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
980 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
981 			    ifp, NULL);
982 		}
983 		/* NET_EPOCH_EXIT(et); */
984 		in_purgemaddrs(ifp);
985 		igmp_domifdetach(ifp);
986 	}
987 	IFNET_RUNLOCK();
988 }
989 
990 int
991 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
992 {
993 
994 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
995 	     /*
996 	      * Check for old-style (host 0) broadcast, but
997 	      * taking into account that RFC 3021 obsoletes it.
998 	      */
999 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
1000 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
1001 	     /*
1002 	      * Check for an all one subnetmask. These
1003 	      * only exist when an interface gets a secondary
1004 	      * address.
1005 	      */
1006 	    ia->ia_subnetmask != (u_long)0xffffffff);
1007 }
1008 
1009 /*
1010  * Return 1 if the address might be a local broadcast address.
1011  */
1012 int
1013 in_broadcast(struct in_addr in, struct ifnet *ifp)
1014 {
1015 	struct ifaddr *ifa;
1016 	int found;
1017 
1018 	NET_EPOCH_ASSERT();
1019 
1020 	if (in.s_addr == INADDR_BROADCAST ||
1021 	    in.s_addr == INADDR_ANY)
1022 		return (1);
1023 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1024 		return (0);
1025 	found = 0;
1026 	/*
1027 	 * Look through the list of addresses for a match
1028 	 * with a broadcast address.
1029 	 */
1030 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1031 		if (ifa->ifa_addr->sa_family == AF_INET &&
1032 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
1033 			found = 1;
1034 			break;
1035 		}
1036 	return (found);
1037 }
1038 
1039 /*
1040  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1041  */
1042 void
1043 in_ifdetach(struct ifnet *ifp)
1044 {
1045 	IN_MULTI_LOCK();
1046 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1047 	in_pcbpurgeif0(&V_udbinfo, ifp);
1048 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1049 	in_purgemaddrs(ifp);
1050 	IN_MULTI_UNLOCK();
1051 
1052 	/*
1053 	 * Make sure all multicast deletions invoking if_ioctl() are
1054 	 * completed before returning. Else we risk accessing a freed
1055 	 * ifnet structure pointer.
1056 	 */
1057 	inm_release_wait(NULL);
1058 }
1059 
1060 /*
1061  * Delete all IPv4 multicast address records, and associated link-layer
1062  * multicast address records, associated with ifp.
1063  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1064  * XXX This should not race with ifma_protospec being set during
1065  * a new allocation, if it does, we have bigger problems.
1066  */
1067 static void
1068 in_purgemaddrs(struct ifnet *ifp)
1069 {
1070 	struct in_multi_head purgeinms;
1071 	struct in_multi		*inm;
1072 	struct ifmultiaddr	*ifma, *next;
1073 
1074 	SLIST_INIT(&purgeinms);
1075 	IN_MULTI_LIST_LOCK();
1076 
1077 	/*
1078 	 * Extract list of in_multi associated with the detaching ifp
1079 	 * which the PF_INET layer is about to release.
1080 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1081 	 * by code further down.
1082 	 */
1083 	IF_ADDR_WLOCK(ifp);
1084  restart:
1085 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1086 		if (ifma->ifma_addr->sa_family != AF_INET ||
1087 		    ifma->ifma_protospec == NULL)
1088 			continue;
1089 		inm = (struct in_multi *)ifma->ifma_protospec;
1090 		inm_rele_locked(&purgeinms, inm);
1091 		if (__predict_false(ifma_restart)) {
1092 			ifma_restart = true;
1093 			goto restart;
1094 		}
1095 	}
1096 	IF_ADDR_WUNLOCK(ifp);
1097 
1098 	inm_release_list_deferred(&purgeinms);
1099 	igmp_ifdetach(ifp);
1100 	IN_MULTI_LIST_UNLOCK();
1101 }
1102 
1103 struct in_llentry {
1104 	struct llentry		base;
1105 };
1106 
1107 #define	IN_LLTBL_DEFAULT_HSIZE	32
1108 #define	IN_LLTBL_HASH(k, h) \
1109 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1110 
1111 /*
1112  * Do actual deallocation of @lle.
1113  */
1114 static void
1115 in_lltable_destroy_lle_unlocked(epoch_context_t ctx)
1116 {
1117 	struct llentry *lle;
1118 
1119 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1120 	LLE_LOCK_DESTROY(lle);
1121 	LLE_REQ_DESTROY(lle);
1122 	free(lle, M_LLTABLE);
1123 }
1124 
1125 /*
1126  * Called by the datapath to indicate that
1127  * the entry was used.
1128  */
1129 static void
1130 in_lltable_mark_used(struct llentry *lle)
1131 {
1132 
1133 	LLE_REQ_LOCK(lle);
1134 	lle->r_skip_req = 0;
1135 	LLE_REQ_UNLOCK(lle);
1136 }
1137 
1138 /*
1139  * Called by LLE_FREE_LOCKED when number of references
1140  * drops to zero.
1141  */
1142 static void
1143 in_lltable_destroy_lle(struct llentry *lle)
1144 {
1145 
1146 	LLE_WUNLOCK(lle);
1147 	NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1148 }
1149 
1150 static struct llentry *
1151 in_lltable_new(struct in_addr addr4, u_int flags)
1152 {
1153 	struct in_llentry *lle;
1154 
1155 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1156 	if (lle == NULL)		/* NB: caller generates msg */
1157 		return NULL;
1158 
1159 	/*
1160 	 * For IPv4 this will trigger "arpresolve" to generate
1161 	 * an ARP request.
1162 	 */
1163 	lle->base.la_expire = time_uptime; /* mark expired */
1164 	lle->base.r_l3addr.addr4 = addr4;
1165 	lle->base.lle_refcnt = 1;
1166 	lle->base.lle_free = in_lltable_destroy_lle;
1167 	LLE_LOCK_INIT(&lle->base);
1168 	LLE_REQ_INIT(&lle->base);
1169 	callout_init(&lle->base.lle_timer, 1);
1170 
1171 	return (&lle->base);
1172 }
1173 
1174 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1175 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1176 
1177 static int
1178 in_lltable_match_prefix(const struct sockaddr *saddr,
1179     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1180 {
1181 	struct in_addr addr, mask, lle_addr;
1182 
1183 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1184 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1185 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1186 
1187 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1188 		return (0);
1189 
1190 	if (lle->la_flags & LLE_IFADDR) {
1191 		/*
1192 		 * Delete LLE_IFADDR records IFF address & flag matches.
1193 		 * Note that addr is the interface address within prefix
1194 		 * being matched.
1195 		 * Note also we should handle 'ifdown' cases without removing
1196 		 * ifaddr macs.
1197 		 */
1198 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1199 			return (1);
1200 		return (0);
1201 	}
1202 
1203 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1204 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1205 		return (1);
1206 
1207 	return (0);
1208 }
1209 
1210 static void
1211 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1212 {
1213 	size_t pkts_dropped;
1214 
1215 	LLE_WLOCK_ASSERT(lle);
1216 	KASSERT(llt != NULL, ("lltable is NULL"));
1217 
1218 	/* Unlink entry from table if not already */
1219 	if ((lle->la_flags & LLE_LINKED) != 0) {
1220 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1221 		lltable_unlink_entry(llt, lle);
1222 	}
1223 
1224 	/* Drop hold queue */
1225 	pkts_dropped = llentry_free(lle);
1226 	ARPSTAT_ADD(dropped, pkts_dropped);
1227 }
1228 
1229 static int
1230 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1231 {
1232 	struct rt_addrinfo info;
1233 	struct sockaddr_in rt_key, rt_mask;
1234 	struct sockaddr rt_gateway;
1235 	int rt_flags;
1236 
1237 	KASSERT(l3addr->sa_family == AF_INET,
1238 	    ("sin_family %d", l3addr->sa_family));
1239 
1240 	bzero(&rt_key, sizeof(rt_key));
1241 	rt_key.sin_len = sizeof(rt_key);
1242 	bzero(&rt_mask, sizeof(rt_mask));
1243 	rt_mask.sin_len = sizeof(rt_mask);
1244 	bzero(&rt_gateway, sizeof(rt_gateway));
1245 	rt_gateway.sa_len = sizeof(rt_gateway);
1246 
1247 	bzero(&info, sizeof(info));
1248 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1249 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1250 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1251 
1252 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1253 		return (EINVAL);
1254 
1255 	rt_flags = info.rti_flags;
1256 
1257 	/*
1258 	 * If the gateway for an existing host route matches the target L3
1259 	 * address, which is a special route inserted by some implementation
1260 	 * such as MANET, and the interface is of the correct type, then
1261 	 * allow for ARP to proceed.
1262 	 */
1263 	if (rt_flags & RTF_GATEWAY) {
1264 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1265 		    info.rti_ifp->if_type != IFT_ETHER ||
1266 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1267 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1268 		    sizeof(in_addr_t)) != 0) {
1269 			rib_free_info(&info);
1270 			return (EINVAL);
1271 		}
1272 	}
1273 	rib_free_info(&info);
1274 
1275 	/*
1276 	 * Make sure that at least the destination address is covered
1277 	 * by the route. This is for handling the case where 2 or more
1278 	 * interfaces have the same prefix. An incoming packet arrives
1279 	 * on one interface and the corresponding outgoing packet leaves
1280 	 * another interface.
1281 	 */
1282 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1283 		const char *sa, *mask, *addr, *lim;
1284 		const struct sockaddr_in *l3sin;
1285 
1286 		mask = (const char *)&rt_mask;
1287 		/*
1288 		 * Just being extra cautious to avoid some custom
1289 		 * code getting into trouble.
1290 		 */
1291 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1292 			return (EINVAL);
1293 
1294 		sa = (const char *)&rt_key;
1295 		addr = (const char *)l3addr;
1296 		l3sin = (const struct sockaddr_in *)l3addr;
1297 		lim = addr + l3sin->sin_len;
1298 
1299 		for ( ; addr < lim; sa++, mask++, addr++) {
1300 			if ((*sa ^ *addr) & *mask) {
1301 #ifdef DIAGNOSTIC
1302 				char addrbuf[INET_ADDRSTRLEN];
1303 
1304 				log(LOG_INFO, "IPv4 address: \"%s\" "
1305 				    "is not on the network\n",
1306 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1307 #endif
1308 				return (EINVAL);
1309 			}
1310 		}
1311 	}
1312 
1313 	return (0);
1314 }
1315 
1316 static inline uint32_t
1317 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1318 {
1319 
1320 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1321 }
1322 
1323 static uint32_t
1324 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1325 {
1326 
1327 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1328 }
1329 
1330 static void
1331 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1332 {
1333 	struct sockaddr_in *sin;
1334 
1335 	sin = (struct sockaddr_in *)sa;
1336 	bzero(sin, sizeof(*sin));
1337 	sin->sin_family = AF_INET;
1338 	sin->sin_len = sizeof(*sin);
1339 	sin->sin_addr = lle->r_l3addr.addr4;
1340 }
1341 
1342 static inline struct llentry *
1343 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1344 {
1345 	struct llentry *lle;
1346 	struct llentries *lleh;
1347 	u_int hashidx;
1348 
1349 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1350 	lleh = &llt->lle_head[hashidx];
1351 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1352 		if (lle->la_flags & LLE_DELETED)
1353 			continue;
1354 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1355 			break;
1356 	}
1357 
1358 	return (lle);
1359 }
1360 
1361 static void
1362 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1363 {
1364 
1365 	lle->la_flags |= LLE_DELETED;
1366 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1367 #ifdef DIAGNOSTIC
1368 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1369 #endif
1370 	llentry_free(lle);
1371 }
1372 
1373 static struct llentry *
1374 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1375 {
1376 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1377 	struct ifnet *ifp = llt->llt_ifp;
1378 	struct llentry *lle;
1379 	char linkhdr[LLE_MAX_LINKHDR];
1380 	size_t linkhdrsize;
1381 	int lladdr_off;
1382 
1383 	KASSERT(l3addr->sa_family == AF_INET,
1384 	    ("sin_family %d", l3addr->sa_family));
1385 
1386 	/*
1387 	 * A route that covers the given address must have
1388 	 * been installed 1st because we are doing a resolution,
1389 	 * verify this.
1390 	 */
1391 	if (!(flags & LLE_IFADDR) &&
1392 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1393 		return (NULL);
1394 
1395 	lle = in_lltable_new(sin->sin_addr, flags);
1396 	if (lle == NULL) {
1397 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1398 		return (NULL);
1399 	}
1400 	lle->la_flags = flags;
1401 	if (flags & LLE_STATIC)
1402 		lle->r_flags |= RLLE_VALID;
1403 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1404 		linkhdrsize = LLE_MAX_LINKHDR;
1405 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1406 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1407 			NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
1408 			return (NULL);
1409 		}
1410 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1411 		    lladdr_off);
1412 		lle->la_flags |= LLE_STATIC;
1413 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1414 	}
1415 
1416 	return (lle);
1417 }
1418 
1419 /*
1420  * Return NULL if not found or marked for deletion.
1421  * If found return lle read locked.
1422  */
1423 static struct llentry *
1424 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1425 {
1426 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1427 	struct llentry *lle;
1428 
1429 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1430 	KASSERT(l3addr->sa_family == AF_INET,
1431 	    ("sin_family %d", l3addr->sa_family));
1432 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1433 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1434 	    ("wrong lle request flags: %#x", flags));
1435 
1436 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1437 	if (lle == NULL)
1438 		return (NULL);
1439 	if (flags & LLE_UNLOCKED)
1440 		return (lle);
1441 
1442 	if (flags & LLE_EXCLUSIVE)
1443 		LLE_WLOCK(lle);
1444 	else
1445 		LLE_RLOCK(lle);
1446 
1447 	/*
1448 	 * If the afdata lock is not held, the LLE may have been unlinked while
1449 	 * we were blocked on the LLE lock.  Check for this case.
1450 	 */
1451 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1452 		if (flags & LLE_EXCLUSIVE)
1453 			LLE_WUNLOCK(lle);
1454 		else
1455 			LLE_RUNLOCK(lle);
1456 		return (NULL);
1457 	}
1458 	return (lle);
1459 }
1460 
1461 static int
1462 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1463     struct sysctl_req *wr)
1464 {
1465 	struct ifnet *ifp = llt->llt_ifp;
1466 	/* XXX stack use */
1467 	struct {
1468 		struct rt_msghdr	rtm;
1469 		struct sockaddr_in	sin;
1470 		struct sockaddr_dl	sdl;
1471 	} arpc;
1472 	struct sockaddr_dl *sdl;
1473 	int error;
1474 
1475 	bzero(&arpc, sizeof(arpc));
1476 	/* skip deleted entries */
1477 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1478 		return (0);
1479 	/* Skip if jailed and not a valid IP of the prison. */
1480 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1481 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1482 		return (0);
1483 	/*
1484 	 * produce a msg made of:
1485 	 *  struct rt_msghdr;
1486 	 *  struct sockaddr_in; (IPv4)
1487 	 *  struct sockaddr_dl;
1488 	 */
1489 	arpc.rtm.rtm_msglen = sizeof(arpc);
1490 	arpc.rtm.rtm_version = RTM_VERSION;
1491 	arpc.rtm.rtm_type = RTM_GET;
1492 	arpc.rtm.rtm_flags = RTF_UP;
1493 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1494 
1495 	/* publish */
1496 	if (lle->la_flags & LLE_PUB)
1497 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1498 
1499 	sdl = &arpc.sdl;
1500 	sdl->sdl_family = AF_LINK;
1501 	sdl->sdl_len = sizeof(*sdl);
1502 	sdl->sdl_index = ifp->if_index;
1503 	sdl->sdl_type = ifp->if_type;
1504 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1505 		sdl->sdl_alen = ifp->if_addrlen;
1506 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1507 	} else {
1508 		sdl->sdl_alen = 0;
1509 		bzero(LLADDR(sdl), ifp->if_addrlen);
1510 	}
1511 
1512 	arpc.rtm.rtm_rmx.rmx_expire =
1513 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1514 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1515 	if (lle->la_flags & LLE_STATIC)
1516 		arpc.rtm.rtm_flags |= RTF_STATIC;
1517 	if (lle->la_flags & LLE_IFADDR)
1518 		arpc.rtm.rtm_flags |= RTF_PINNED;
1519 	arpc.rtm.rtm_index = ifp->if_index;
1520 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1521 
1522 	return (error);
1523 }
1524 
1525 static struct lltable *
1526 in_lltattach(struct ifnet *ifp)
1527 {
1528 	struct lltable *llt;
1529 
1530 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1531  	llt->llt_af = AF_INET;
1532  	llt->llt_ifp = ifp;
1533 
1534 	llt->llt_lookup = in_lltable_lookup;
1535 	llt->llt_alloc_entry = in_lltable_alloc;
1536 	llt->llt_delete_entry = in_lltable_delete_entry;
1537 	llt->llt_dump_entry = in_lltable_dump_entry;
1538 	llt->llt_hash = in_lltable_hash;
1539 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1540 	llt->llt_free_entry = in_lltable_free_entry;
1541 	llt->llt_match_prefix = in_lltable_match_prefix;
1542 	llt->llt_mark_used = in_lltable_mark_used;
1543  	lltable_link(llt);
1544 
1545 	return (llt);
1546 }
1547 
1548 void *
1549 in_domifattach(struct ifnet *ifp)
1550 {
1551 	struct in_ifinfo *ii;
1552 
1553 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1554 
1555 	ii->ii_llt = in_lltattach(ifp);
1556 	ii->ii_igmp = igmp_domifattach(ifp);
1557 
1558 	return (ii);
1559 }
1560 
1561 void
1562 in_domifdetach(struct ifnet *ifp, void *aux)
1563 {
1564 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1565 
1566 	igmp_domifdetach(ifp);
1567 	lltable_free(ii->ii_llt);
1568 	free(ii, M_IFADDR);
1569 }
1570