xref: /freebsd/sys/netinet/in.c (revision 74ca7bf1d4c7173d5575ba168bc4b5f6d181ff5a)
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 static VNET_DEFINE(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 	TAILQ_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 	IF_ADDR_RLOCK(ifp);
146 	TAILQ_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 			IF_ADDR_RUNLOCK(ifp);
152 			return (1);
153 		}
154 	}
155 	IF_ADDR_RUNLOCK(ifp);
156 
157 	return (0);
158 }
159 
160 /*
161  * Return a reference to the interface address which is different to
162  * the supplied one but with same IP address value.
163  */
164 static struct in_ifaddr *
165 in_localip_more(struct in_ifaddr *ia)
166 {
167 	struct rm_priotracker in_ifa_tracker;
168 	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
169 	struct in_ifaddr *it;
170 
171 	IN_IFADDR_RLOCK(&in_ifa_tracker);
172 	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
173 		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
174 			ifa_ref(&it->ia_ifa);
175 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
176 			return (it);
177 		}
178 	}
179 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
180 
181 	return (NULL);
182 }
183 
184 /*
185  * Determine whether an IP address is in a reserved set of addresses
186  * that may not be forwarded, or whether datagrams to that destination
187  * may be forwarded.
188  */
189 int
190 in_canforward(struct in_addr in)
191 {
192 	u_long i = ntohl(in.s_addr);
193 	u_long net;
194 
195 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
196 		return (0);
197 	if (IN_CLASSA(i)) {
198 		net = i & IN_CLASSA_NET;
199 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
200 			return (0);
201 	}
202 	return (1);
203 }
204 
205 /*
206  * Trim a mask in a sockaddr
207  */
208 static void
209 in_socktrim(struct sockaddr_in *ap)
210 {
211     char *cplim = (char *) &ap->sin_addr;
212     char *cp = (char *) (&ap->sin_addr + 1);
213 
214     ap->sin_len = 0;
215     while (--cp >= cplim)
216 	if (*cp) {
217 	    (ap)->sin_len = cp - (char *) (ap) + 1;
218 	    break;
219 	}
220 }
221 
222 /*
223  * Generic internet control operations (ioctl's).
224  */
225 int
226 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
227     struct thread *td)
228 {
229 	struct ifreq *ifr = (struct ifreq *)data;
230 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
231 	struct ifaddr *ifa;
232 	struct in_ifaddr *ia;
233 	int error;
234 
235 	if (ifp == NULL)
236 		return (EADDRNOTAVAIL);
237 
238 	/*
239 	 * Filter out 4 ioctls we implement directly.  Forward the rest
240 	 * to specific functions and ifp->if_ioctl().
241 	 */
242 	switch (cmd) {
243 	case SIOCGIFADDR:
244 	case SIOCGIFBRDADDR:
245 	case SIOCGIFDSTADDR:
246 	case SIOCGIFNETMASK:
247 		break;
248 	case SIOCDIFADDR:
249 		sx_xlock(&in_control_sx);
250 		error = in_difaddr_ioctl(cmd, data, ifp, td);
251 		sx_xunlock(&in_control_sx);
252 		return (error);
253 	case OSIOCAIFADDR:	/* 9.x compat */
254 	case SIOCAIFADDR:
255 		sx_xlock(&in_control_sx);
256 		error = in_aifaddr_ioctl(cmd, data, ifp, td);
257 		sx_xunlock(&in_control_sx);
258 		return (error);
259 	case SIOCSIFADDR:
260 	case SIOCSIFBRDADDR:
261 	case SIOCSIFDSTADDR:
262 	case SIOCSIFNETMASK:
263 		/* We no longer support that old commands. */
264 		return (EINVAL);
265 	default:
266 		if (ifp->if_ioctl == NULL)
267 			return (EOPNOTSUPP);
268 		return ((*ifp->if_ioctl)(ifp, cmd, data));
269 	}
270 
271 	if (addr->sin_addr.s_addr != INADDR_ANY &&
272 	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
273 		return (EADDRNOTAVAIL);
274 
275 	/*
276 	 * Find address for this interface, if it exists.  If an
277 	 * address was specified, find that one instead of the
278 	 * first one on the interface, if possible.
279 	 */
280 	IF_ADDR_RLOCK(ifp);
281 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
282 		if (ifa->ifa_addr->sa_family != AF_INET)
283 			continue;
284 		ia = (struct in_ifaddr *)ifa;
285 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
286 			break;
287 	}
288 	if (ifa == NULL)
289 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
290 			if (ifa->ifa_addr->sa_family == AF_INET) {
291 				ia = (struct in_ifaddr *)ifa;
292 				if (prison_check_ip4(td->td_ucred,
293 				    &ia->ia_addr.sin_addr) == 0)
294 					break;
295 			}
296 
297 	if (ifa == NULL) {
298 		IF_ADDR_RUNLOCK(ifp);
299 		return (EADDRNOTAVAIL);
300 	}
301 
302 	error = 0;
303 	switch (cmd) {
304 	case SIOCGIFADDR:
305 		*addr = ia->ia_addr;
306 		break;
307 
308 	case SIOCGIFBRDADDR:
309 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
310 			error = EINVAL;
311 			break;
312 		}
313 		*addr = ia->ia_broadaddr;
314 		break;
315 
316 	case SIOCGIFDSTADDR:
317 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
318 			error = EINVAL;
319 			break;
320 		}
321 		*addr = ia->ia_dstaddr;
322 		break;
323 
324 	case SIOCGIFNETMASK:
325 		*addr = ia->ia_sockmask;
326 		break;
327 	}
328 
329 	IF_ADDR_RUNLOCK(ifp);
330 
331 	return (error);
332 }
333 
334 static int
335 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
336 {
337 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
338 	const struct sockaddr_in *addr = &ifra->ifra_addr;
339 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
340 	const struct sockaddr_in *mask = &ifra->ifra_mask;
341 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
342 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
343 	struct ifaddr *ifa;
344 	struct in_ifaddr *ia;
345 	bool iaIsFirst;
346 	int error = 0;
347 
348 	error = priv_check(td, PRIV_NET_ADDIFADDR);
349 	if (error)
350 		return (error);
351 
352 	/*
353 	 * ifra_addr must be present and be of INET family.
354 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
355 	 */
356 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
357 	    addr->sin_family != AF_INET)
358 		return (EINVAL);
359 	if (broadaddr->sin_len != 0 &&
360 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
361 	    broadaddr->sin_family != AF_INET))
362 		return (EINVAL);
363 	if (mask->sin_len != 0 &&
364 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
365 	    mask->sin_family != AF_INET))
366 		return (EINVAL);
367 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
368 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
369 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
370 		return (EDESTADDRREQ);
371 	if (vhid > 0 && carp_attach_p == NULL)
372 		return (EPROTONOSUPPORT);
373 
374 	/*
375 	 * See whether address already exist.
376 	 */
377 	iaIsFirst = true;
378 	ia = NULL;
379 	IF_ADDR_RLOCK(ifp);
380 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
381 		struct in_ifaddr *it;
382 
383 		if (ifa->ifa_addr->sa_family != AF_INET)
384 			continue;
385 
386 		it = (struct in_ifaddr *)ifa;
387 		iaIsFirst = false;
388 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
389 		    prison_check_ip4(td->td_ucred, &addr->sin_addr) == 0)
390 			ia = it;
391 	}
392 	IF_ADDR_RUNLOCK(ifp);
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 	TAILQ_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 	TAILQ_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 	EVENTHANDLER_INVOKE(ifaddr_event, ifp);
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 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
537 	IF_ADDR_WUNLOCK(ifp);
538 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
539 
540 	IN_IFADDR_WLOCK();
541 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, 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 	TAILQ_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 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
601 	IF_ADDR_WUNLOCK(ifp);
602 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
603 
604 	IN_IFADDR_WLOCK();
605 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, 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,
624 		    (cmd == SIOCDIFADDR) ? false : true);
625 
626 	/*
627 	 * If this is the last IPv4 address configured on this
628 	 * interface, leave the all-hosts group.
629 	 * No state-change report need be transmitted.
630 	 */
631 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
632 		struct in_ifinfo *ii;
633 
634 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
635 		IN_MULTI_LOCK();
636 		if (ii->ii_allhosts) {
637 			(void)in_leavegroup_locked(ii->ii_allhosts, NULL);
638 			ii->ii_allhosts = NULL;
639 		}
640 		IN_MULTI_UNLOCK();
641 	}
642 
643 	IF_ADDR_WLOCK(ifp);
644 	if (callout_stop(&ia->ia_garp_timer) == 1) {
645 		ifa_free(&ia->ia_ifa);
646 	}
647 	IF_ADDR_WUNLOCK(ifp);
648 
649 	EVENTHANDLER_INVOKE(ifaddr_event, ifp);
650 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
651 
652 	return (0);
653 }
654 
655 #define rtinitflags(x) \
656 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
657 	    ? RTF_HOST : 0)
658 
659 /*
660  * Check if we have a route for the given prefix already or add one accordingly.
661  */
662 int
663 in_addprefix(struct in_ifaddr *target, int flags)
664 {
665 	struct rm_priotracker in_ifa_tracker;
666 	struct in_ifaddr *ia;
667 	struct in_addr prefix, mask, p, m;
668 	int error;
669 
670 	if ((flags & RTF_HOST) != 0) {
671 		prefix = target->ia_dstaddr.sin_addr;
672 		mask.s_addr = 0;
673 	} else {
674 		prefix = target->ia_addr.sin_addr;
675 		mask = target->ia_sockmask.sin_addr;
676 		prefix.s_addr &= mask.s_addr;
677 	}
678 
679 	IN_IFADDR_RLOCK(&in_ifa_tracker);
680 	/* Look for an existing address with the same prefix, mask, and fib */
681 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
682 		if (rtinitflags(ia)) {
683 			p = ia->ia_dstaddr.sin_addr;
684 
685 			if (prefix.s_addr != p.s_addr)
686 				continue;
687 		} else {
688 			p = ia->ia_addr.sin_addr;
689 			m = ia->ia_sockmask.sin_addr;
690 			p.s_addr &= m.s_addr;
691 
692 			if (prefix.s_addr != p.s_addr ||
693 			    mask.s_addr != m.s_addr)
694 				continue;
695 		}
696 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
697 			continue;
698 
699 		/*
700 		 * If we got a matching prefix route inserted by other
701 		 * interface address, we are done here.
702 		 */
703 		if (ia->ia_flags & IFA_ROUTE) {
704 #ifdef RADIX_MPATH
705 			if (ia->ia_addr.sin_addr.s_addr ==
706 			    target->ia_addr.sin_addr.s_addr) {
707 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
708 				return (EEXIST);
709 			} else
710 				break;
711 #endif
712 			if (V_nosameprefix) {
713 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
714 				return (EEXIST);
715 			} else {
716 				int fibnum;
717 
718 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
719 					target->ia_ifp->if_fib;
720 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
721 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
722 				return (0);
723 			}
724 		}
725 	}
726 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
727 
728 	/*
729 	 * No-one seem to have this prefix route, so we try to insert it.
730 	 */
731 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
732 	if (!error)
733 		target->ia_flags |= IFA_ROUTE;
734 	return (error);
735 }
736 
737 /*
738  * Removes either all lle entries for given @ia, or lle
739  * corresponding to @ia address.
740  */
741 static void
742 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
743 {
744 	struct sockaddr_in addr, mask;
745 	struct sockaddr *saddr, *smask;
746 	struct ifnet *ifp;
747 
748 	saddr = (struct sockaddr *)&addr;
749 	bzero(&addr, sizeof(addr));
750 	addr.sin_len = sizeof(addr);
751 	addr.sin_family = AF_INET;
752 	smask = (struct sockaddr *)&mask;
753 	bzero(&mask, sizeof(mask));
754 	mask.sin_len = sizeof(mask);
755 	mask.sin_family = AF_INET;
756 	mask.sin_addr.s_addr = ia->ia_subnetmask;
757 	ifp = ia->ia_ifp;
758 
759 	if (all) {
760 
761 		/*
762 		 * Remove all L2 entries matching given prefix.
763 		 * Convert address to host representation to avoid
764 		 * doing this on every callback. ia_subnetmask is already
765 		 * stored in host representation.
766 		 */
767 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
768 		lltable_prefix_free(AF_INET, saddr, smask, flags);
769 	} else {
770 		/* Remove interface address only */
771 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
772 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
773 	}
774 }
775 
776 /*
777  * If there is no other address in the system that can serve a route to the
778  * same prefix, remove the route.  Hand over the route to the new address
779  * otherwise.
780  */
781 int
782 in_scrubprefix(struct in_ifaddr *target, u_int flags)
783 {
784 	struct rm_priotracker in_ifa_tracker;
785 	struct in_ifaddr *ia;
786 	struct in_addr prefix, mask, p, m;
787 	int error = 0;
788 
789 	/*
790 	 * Remove the loopback route to the interface address.
791 	 */
792 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
793 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
794 	    (flags & LLE_STATIC)) {
795 		struct in_ifaddr *eia;
796 
797 		/*
798 		 * XXXME: add fib-aware in_localip.
799 		 * We definitely don't want to switch between
800 		 * prefixes in different fibs.
801 		 */
802 		eia = in_localip_more(target);
803 
804 		if (eia != NULL) {
805 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
806 			    (struct sockaddr *)&target->ia_addr);
807 			ifa_free(&eia->ia_ifa);
808 		} else {
809 			error = ifa_del_loopback_route((struct ifaddr *)target,
810 			    (struct sockaddr *)&target->ia_addr);
811 		}
812 	}
813 
814 	if (rtinitflags(target)) {
815 		prefix = target->ia_dstaddr.sin_addr;
816 		mask.s_addr = 0;
817 	} else {
818 		prefix = target->ia_addr.sin_addr;
819 		mask = target->ia_sockmask.sin_addr;
820 		prefix.s_addr &= mask.s_addr;
821 	}
822 
823 	if ((target->ia_flags & IFA_ROUTE) == 0) {
824 		int fibnum;
825 
826 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
827 			target->ia_ifp->if_fib;
828 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
829 
830 		/*
831 		 * Removing address from !IFF_UP interface or
832 		 * prefix which exists on other interface (along with route).
833 		 * No entries should exist here except target addr.
834 		 * Given that, delete this entry only.
835 		 */
836 		in_scrubprefixlle(target, 0, flags);
837 		return (0);
838 	}
839 
840 	IN_IFADDR_RLOCK(&in_ifa_tracker);
841 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
842 		if (rtinitflags(ia)) {
843 			p = ia->ia_dstaddr.sin_addr;
844 
845 			if (prefix.s_addr != p.s_addr)
846 				continue;
847 		} else {
848 			p = ia->ia_addr.sin_addr;
849 			m = ia->ia_sockmask.sin_addr;
850 			p.s_addr &= m.s_addr;
851 
852 			if (prefix.s_addr != p.s_addr ||
853 			    mask.s_addr != m.s_addr)
854 				continue;
855 		}
856 
857 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
858 			continue;
859 
860 		/*
861 		 * If we got a matching prefix address, move IFA_ROUTE and
862 		 * the route itself to it.  Make sure that routing daemons
863 		 * get a heads-up.
864 		 */
865 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
866 			ifa_ref(&ia->ia_ifa);
867 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
868 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
869 			    rtinitflags(target));
870 			if (error == 0)
871 				target->ia_flags &= ~IFA_ROUTE;
872 			else
873 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
874 					error);
875 			/* Scrub all entries IFF interface is different */
876 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
877 			    flags);
878 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
879 			    rtinitflags(ia) | RTF_UP);
880 			if (error == 0)
881 				ia->ia_flags |= IFA_ROUTE;
882 			else
883 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
884 					error);
885 			ifa_free(&ia->ia_ifa);
886 			return (error);
887 		}
888 	}
889 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
890 
891 	/*
892 	 * remove all L2 entries on the given prefix
893 	 */
894 	in_scrubprefixlle(target, 1, flags);
895 
896 	/*
897 	 * As no-one seem to have this prefix, we can remove the route.
898 	 */
899 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
900 	if (error == 0)
901 		target->ia_flags &= ~IFA_ROUTE;
902 	else
903 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
904 	return (error);
905 }
906 
907 #undef rtinitflags
908 
909 void
910 in_ifscrub_all(void)
911 {
912 	struct ifnet *ifp;
913 	struct ifaddr *ifa, *nifa;
914 	struct ifaliasreq ifr;
915 
916 	IFNET_RLOCK();
917 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
918 		/* Cannot lock here - lock recursion. */
919 		/* IF_ADDR_RLOCK(ifp); */
920 		TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
921 			if (ifa->ifa_addr->sa_family != AF_INET)
922 				continue;
923 
924 			/*
925 			 * This is ugly but the only way for legacy IP to
926 			 * cleanly remove addresses and everything attached.
927 			 */
928 			bzero(&ifr, sizeof(ifr));
929 			ifr.ifra_addr = *ifa->ifa_addr;
930 			if (ifa->ifa_dstaddr)
931 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
932 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
933 			    ifp, NULL);
934 		}
935 		/* IF_ADDR_RUNLOCK(ifp); */
936 		in_purgemaddrs(ifp);
937 		igmp_domifdetach(ifp);
938 	}
939 	IFNET_RUNLOCK();
940 }
941 
942 int
943 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
944 {
945 
946 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
947 	     /*
948 	      * Check for old-style (host 0) broadcast, but
949 	      * taking into account that RFC 3021 obsoletes it.
950 	      */
951 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
952 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
953 	     /*
954 	      * Check for an all one subnetmask. These
955 	      * only exist when an interface gets a secondary
956 	      * address.
957 	      */
958 	    ia->ia_subnetmask != (u_long)0xffffffff);
959 }
960 
961 /*
962  * Return 1 if the address might be a local broadcast address.
963  */
964 int
965 in_broadcast(struct in_addr in, struct ifnet *ifp)
966 {
967 	struct ifaddr *ifa;
968 	int found;
969 
970 	if (in.s_addr == INADDR_BROADCAST ||
971 	    in.s_addr == INADDR_ANY)
972 		return (1);
973 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
974 		return (0);
975 	found = 0;
976 	/*
977 	 * Look through the list of addresses for a match
978 	 * with a broadcast address.
979 	 */
980 	IF_ADDR_RLOCK(ifp);
981 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
982 		if (ifa->ifa_addr->sa_family == AF_INET &&
983 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
984 			found = 1;
985 			break;
986 		}
987 	IF_ADDR_RUNLOCK(ifp);
988 	return (found);
989 }
990 
991 /*
992  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
993  */
994 void
995 in_ifdetach(struct ifnet *ifp)
996 {
997 
998 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
999 	in_pcbpurgeif0(&V_udbinfo, ifp);
1000 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1001 	in_purgemaddrs(ifp);
1002 }
1003 
1004 /*
1005  * Delete all IPv4 multicast address records, and associated link-layer
1006  * multicast address records, associated with ifp.
1007  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1008  * XXX This should not race with ifma_protospec being set during
1009  * a new allocation, if it does, we have bigger problems.
1010  */
1011 static void
1012 in_purgemaddrs(struct ifnet *ifp)
1013 {
1014 	LIST_HEAD(,in_multi) purgeinms;
1015 	struct in_multi		*inm, *tinm;
1016 	struct ifmultiaddr	*ifma;
1017 
1018 	LIST_INIT(&purgeinms);
1019 	IN_MULTI_LOCK();
1020 
1021 	/*
1022 	 * Extract list of in_multi associated with the detaching ifp
1023 	 * which the PF_INET layer is about to release.
1024 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1025 	 * by code further down.
1026 	 */
1027 	IF_ADDR_RLOCK(ifp);
1028 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1029 		if (ifma->ifma_addr->sa_family != AF_INET ||
1030 		    ifma->ifma_protospec == NULL)
1031 			continue;
1032 #if 0
1033 		KASSERT(ifma->ifma_protospec != NULL,
1034 		    ("%s: ifma_protospec is NULL", __func__));
1035 #endif
1036 		inm = (struct in_multi *)ifma->ifma_protospec;
1037 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1038 	}
1039 	IF_ADDR_RUNLOCK(ifp);
1040 
1041 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1042 		LIST_REMOVE(inm, inm_link);
1043 		inm_release_locked(inm);
1044 	}
1045 	igmp_ifdetach(ifp);
1046 
1047 	IN_MULTI_UNLOCK();
1048 }
1049 
1050 struct in_llentry {
1051 	struct llentry		base;
1052 };
1053 
1054 #define	IN_LLTBL_DEFAULT_HSIZE	32
1055 #define	IN_LLTBL_HASH(k, h) \
1056 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1057 
1058 /*
1059  * Do actual deallocation of @lle.
1060  */
1061 static void
1062 in_lltable_destroy_lle_unlocked(struct llentry *lle)
1063 {
1064 
1065 	LLE_LOCK_DESTROY(lle);
1066 	LLE_REQ_DESTROY(lle);
1067 	free(lle, M_LLTABLE);
1068 }
1069 
1070 /*
1071  * Called by the datapath to indicate that
1072  * the entry was used.
1073  */
1074 static void
1075 in_lltable_mark_used(struct llentry *lle)
1076 {
1077 
1078 	LLE_REQ_LOCK(lle);
1079 	lle->r_skip_req = 0;
1080 	LLE_REQ_UNLOCK(lle);
1081 }
1082 
1083 /*
1084  * Called by LLE_FREE_LOCKED when number of references
1085  * drops to zero.
1086  */
1087 static void
1088 in_lltable_destroy_lle(struct llentry *lle)
1089 {
1090 
1091 	LLE_WUNLOCK(lle);
1092 	in_lltable_destroy_lle_unlocked(lle);
1093 }
1094 
1095 static struct llentry *
1096 in_lltable_new(struct in_addr addr4, u_int flags)
1097 {
1098 	struct in_llentry *lle;
1099 
1100 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1101 	if (lle == NULL)		/* NB: caller generates msg */
1102 		return NULL;
1103 
1104 	/*
1105 	 * For IPv4 this will trigger "arpresolve" to generate
1106 	 * an ARP request.
1107 	 */
1108 	lle->base.la_expire = time_uptime; /* mark expired */
1109 	lle->base.r_l3addr.addr4 = addr4;
1110 	lle->base.lle_refcnt = 1;
1111 	lle->base.lle_free = in_lltable_destroy_lle;
1112 	LLE_LOCK_INIT(&lle->base);
1113 	LLE_REQ_INIT(&lle->base);
1114 	callout_init(&lle->base.lle_timer, 1);
1115 
1116 	return (&lle->base);
1117 }
1118 
1119 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1120 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1121 
1122 static int
1123 in_lltable_match_prefix(const struct sockaddr *saddr,
1124     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1125 {
1126 	struct in_addr addr, mask, lle_addr;
1127 
1128 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1129 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1130 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1131 
1132 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1133 		return (0);
1134 
1135 	if (lle->la_flags & LLE_IFADDR) {
1136 
1137 		/*
1138 		 * Delete LLE_IFADDR records IFF address & flag matches.
1139 		 * Note that addr is the interface address within prefix
1140 		 * being matched.
1141 		 * Note also we should handle 'ifdown' cases without removing
1142 		 * ifaddr macs.
1143 		 */
1144 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1145 			return (1);
1146 		return (0);
1147 	}
1148 
1149 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1150 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1151 		return (1);
1152 
1153 	return (0);
1154 }
1155 
1156 static void
1157 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1158 {
1159 	struct ifnet *ifp;
1160 	size_t pkts_dropped;
1161 
1162 	LLE_WLOCK_ASSERT(lle);
1163 	KASSERT(llt != NULL, ("lltable is NULL"));
1164 
1165 	/* Unlink entry from table if not already */
1166 	if ((lle->la_flags & LLE_LINKED) != 0) {
1167 		ifp = llt->llt_ifp;
1168 		IF_AFDATA_WLOCK_ASSERT(ifp);
1169 		lltable_unlink_entry(llt, lle);
1170 	}
1171 
1172 	/* cancel timer */
1173 	if (callout_stop(&lle->lle_timer) > 0)
1174 		LLE_REMREF(lle);
1175 
1176 	/* Drop hold queue */
1177 	pkts_dropped = llentry_free(lle);
1178 	ARPSTAT_ADD(dropped, pkts_dropped);
1179 }
1180 
1181 static int
1182 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1183 {
1184 	struct rt_addrinfo info;
1185 	struct sockaddr_in rt_key, rt_mask;
1186 	struct sockaddr rt_gateway;
1187 	int rt_flags;
1188 
1189 	KASSERT(l3addr->sa_family == AF_INET,
1190 	    ("sin_family %d", l3addr->sa_family));
1191 
1192 	bzero(&rt_key, sizeof(rt_key));
1193 	rt_key.sin_len = sizeof(rt_key);
1194 	bzero(&rt_mask, sizeof(rt_mask));
1195 	rt_mask.sin_len = sizeof(rt_mask);
1196 	bzero(&rt_gateway, sizeof(rt_gateway));
1197 	rt_gateway.sa_len = sizeof(rt_gateway);
1198 
1199 	bzero(&info, sizeof(info));
1200 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1201 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1202 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1203 
1204 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1205 		return (EINVAL);
1206 
1207 	rt_flags = info.rti_flags;
1208 
1209 	/*
1210 	 * If the gateway for an existing host route matches the target L3
1211 	 * address, which is a special route inserted by some implementation
1212 	 * such as MANET, and the interface is of the correct type, then
1213 	 * allow for ARP to proceed.
1214 	 */
1215 	if (rt_flags & RTF_GATEWAY) {
1216 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1217 		    info.rti_ifp->if_type != IFT_ETHER ||
1218 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1219 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1220 		    sizeof(in_addr_t)) != 0) {
1221 			rib_free_info(&info);
1222 			return (EINVAL);
1223 		}
1224 	}
1225 	rib_free_info(&info);
1226 
1227 	/*
1228 	 * Make sure that at least the destination address is covered
1229 	 * by the route. This is for handling the case where 2 or more
1230 	 * interfaces have the same prefix. An incoming packet arrives
1231 	 * on one interface and the corresponding outgoing packet leaves
1232 	 * another interface.
1233 	 */
1234 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1235 		const char *sa, *mask, *addr, *lim;
1236 		const struct sockaddr_in *l3sin;
1237 
1238 		mask = (const char *)&rt_mask;
1239 		/*
1240 		 * Just being extra cautious to avoid some custom
1241 		 * code getting into trouble.
1242 		 */
1243 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1244 			return (EINVAL);
1245 
1246 		sa = (const char *)&rt_key;
1247 		addr = (const char *)l3addr;
1248 		l3sin = (const struct sockaddr_in *)l3addr;
1249 		lim = addr + l3sin->sin_len;
1250 
1251 		for ( ; addr < lim; sa++, mask++, addr++) {
1252 			if ((*sa ^ *addr) & *mask) {
1253 #ifdef DIAGNOSTIC
1254 				char addrbuf[INET_ADDRSTRLEN];
1255 
1256 				log(LOG_INFO, "IPv4 address: \"%s\" "
1257 				    "is not on the network\n",
1258 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1259 #endif
1260 				return (EINVAL);
1261 			}
1262 		}
1263 	}
1264 
1265 	return (0);
1266 }
1267 
1268 static inline uint32_t
1269 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1270 {
1271 
1272 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1273 }
1274 
1275 static uint32_t
1276 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1277 {
1278 
1279 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1280 }
1281 
1282 static void
1283 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1284 {
1285 	struct sockaddr_in *sin;
1286 
1287 	sin = (struct sockaddr_in *)sa;
1288 	bzero(sin, sizeof(*sin));
1289 	sin->sin_family = AF_INET;
1290 	sin->sin_len = sizeof(*sin);
1291 	sin->sin_addr = lle->r_l3addr.addr4;
1292 }
1293 
1294 static inline struct llentry *
1295 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1296 {
1297 	struct llentry *lle;
1298 	struct llentries *lleh;
1299 	u_int hashidx;
1300 
1301 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1302 	lleh = &llt->lle_head[hashidx];
1303 	LIST_FOREACH(lle, lleh, lle_next) {
1304 		if (lle->la_flags & LLE_DELETED)
1305 			continue;
1306 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1307 			break;
1308 	}
1309 
1310 	return (lle);
1311 }
1312 
1313 static void
1314 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1315 {
1316 
1317 	lle->la_flags |= LLE_DELETED;
1318 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1319 #ifdef DIAGNOSTIC
1320 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1321 #endif
1322 	llentry_free(lle);
1323 }
1324 
1325 static struct llentry *
1326 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1327 {
1328 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1329 	struct ifnet *ifp = llt->llt_ifp;
1330 	struct llentry *lle;
1331 	char linkhdr[LLE_MAX_LINKHDR];
1332 	size_t linkhdrsize;
1333 	int lladdr_off;
1334 
1335 	KASSERT(l3addr->sa_family == AF_INET,
1336 	    ("sin_family %d", l3addr->sa_family));
1337 
1338 	/*
1339 	 * A route that covers the given address must have
1340 	 * been installed 1st because we are doing a resolution,
1341 	 * verify this.
1342 	 */
1343 	if (!(flags & LLE_IFADDR) &&
1344 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1345 		return (NULL);
1346 
1347 	lle = in_lltable_new(sin->sin_addr, flags);
1348 	if (lle == NULL) {
1349 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1350 		return (NULL);
1351 	}
1352 	lle->la_flags = flags;
1353 	if (flags & LLE_STATIC)
1354 		lle->r_flags |= RLLE_VALID;
1355 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1356 		linkhdrsize = LLE_MAX_LINKHDR;
1357 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1358 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1359 			in_lltable_destroy_lle_unlocked(lle);
1360 			return (NULL);
1361 		}
1362 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1363 		    lladdr_off);
1364 		lle->la_flags |= LLE_STATIC;
1365 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1366 	}
1367 
1368 	return (lle);
1369 }
1370 
1371 /*
1372  * Return NULL if not found or marked for deletion.
1373  * If found return lle read locked.
1374  */
1375 static struct llentry *
1376 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1377 {
1378 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1379 	struct llentry *lle;
1380 
1381 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1382 	KASSERT(l3addr->sa_family == AF_INET,
1383 	    ("sin_family %d", l3addr->sa_family));
1384 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1385 
1386 	if (lle == NULL)
1387 		return (NULL);
1388 
1389 	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
1390 	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
1391 	    flags));
1392 
1393 	if (flags & LLE_UNLOCKED)
1394 		return (lle);
1395 
1396 	if (flags & LLE_EXCLUSIVE)
1397 		LLE_WLOCK(lle);
1398 	else
1399 		LLE_RLOCK(lle);
1400 
1401 	return (lle);
1402 }
1403 
1404 static int
1405 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1406     struct sysctl_req *wr)
1407 {
1408 	struct ifnet *ifp = llt->llt_ifp;
1409 	/* XXX stack use */
1410 	struct {
1411 		struct rt_msghdr	rtm;
1412 		struct sockaddr_in	sin;
1413 		struct sockaddr_dl	sdl;
1414 	} arpc;
1415 	struct sockaddr_dl *sdl;
1416 	int error;
1417 
1418 	bzero(&arpc, sizeof(arpc));
1419 			/* skip deleted entries */
1420 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1421 				return (0);
1422 			/* Skip if jailed and not a valid IP of the prison. */
1423 			lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1424 			if (prison_if(wr->td->td_ucred,
1425 			    (struct sockaddr *)&arpc.sin) != 0)
1426 				return (0);
1427 			/*
1428 			 * produce a msg made of:
1429 			 *  struct rt_msghdr;
1430 			 *  struct sockaddr_in; (IPv4)
1431 			 *  struct sockaddr_dl;
1432 			 */
1433 			arpc.rtm.rtm_msglen = sizeof(arpc);
1434 			arpc.rtm.rtm_version = RTM_VERSION;
1435 			arpc.rtm.rtm_type = RTM_GET;
1436 			arpc.rtm.rtm_flags = RTF_UP;
1437 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1438 
1439 			/* publish */
1440 			if (lle->la_flags & LLE_PUB)
1441 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1442 
1443 			sdl = &arpc.sdl;
1444 			sdl->sdl_family = AF_LINK;
1445 			sdl->sdl_len = sizeof(*sdl);
1446 			sdl->sdl_index = ifp->if_index;
1447 			sdl->sdl_type = ifp->if_type;
1448 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1449 				sdl->sdl_alen = ifp->if_addrlen;
1450 				bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1451 			} else {
1452 				sdl->sdl_alen = 0;
1453 				bzero(LLADDR(sdl), ifp->if_addrlen);
1454 			}
1455 
1456 			arpc.rtm.rtm_rmx.rmx_expire =
1457 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1458 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1459 			if (lle->la_flags & LLE_STATIC)
1460 				arpc.rtm.rtm_flags |= RTF_STATIC;
1461 			if (lle->la_flags & LLE_IFADDR)
1462 				arpc.rtm.rtm_flags |= RTF_PINNED;
1463 			arpc.rtm.rtm_index = ifp->if_index;
1464 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1465 
1466 	return (error);
1467 }
1468 
1469 static struct lltable *
1470 in_lltattach(struct ifnet *ifp)
1471 {
1472 	struct lltable *llt;
1473 
1474 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1475  	llt->llt_af = AF_INET;
1476  	llt->llt_ifp = ifp;
1477 
1478 	llt->llt_lookup = in_lltable_lookup;
1479 	llt->llt_alloc_entry = in_lltable_alloc;
1480 	llt->llt_delete_entry = in_lltable_delete_entry;
1481 	llt->llt_dump_entry = in_lltable_dump_entry;
1482 	llt->llt_hash = in_lltable_hash;
1483 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1484 	llt->llt_free_entry = in_lltable_free_entry;
1485 	llt->llt_match_prefix = in_lltable_match_prefix;
1486 	llt->llt_mark_used = in_lltable_mark_used;
1487  	lltable_link(llt);
1488 
1489 	return (llt);
1490 }
1491 
1492 void *
1493 in_domifattach(struct ifnet *ifp)
1494 {
1495 	struct in_ifinfo *ii;
1496 
1497 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1498 
1499 	ii->ii_llt = in_lltattach(ifp);
1500 	ii->ii_igmp = igmp_domifattach(ifp);
1501 
1502 	return (ii);
1503 }
1504 
1505 void
1506 in_domifdetach(struct ifnet *ifp, void *aux)
1507 {
1508 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1509 
1510 	igmp_domifdetach(ifp);
1511 	lltable_free(ii->ii_llt);
1512 	free(ii, M_IFADDR);
1513 }
1514