xref: /freebsd/sys/netinet/in.c (revision 2ef9ff7dd34a78a7890ba4d6de64da34d9c10942)
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 epoch_tracker et;
143 	struct ifaddr *ifa;
144 	struct in_ifaddr *ia;
145 
146 	NET_EPOCH_ENTER(et);
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 			NET_EPOCH_EXIT(et);
153 			return (1);
154 		}
155 	}
156 	NET_EPOCH_EXIT(et);
157 
158 	return (0);
159 }
160 
161 /*
162  * Return a reference to the interface address which is different to
163  * the supplied one but with same IP address value.
164  */
165 static struct in_ifaddr *
166 in_localip_more(struct in_ifaddr *ia)
167 {
168 	struct rm_priotracker in_ifa_tracker;
169 	in_addr_t in = IA_SIN(ia)->sin_addr.s_addr;
170 	struct in_ifaddr *it;
171 
172 	IN_IFADDR_RLOCK(&in_ifa_tracker);
173 	LIST_FOREACH(it, INADDR_HASH(in), ia_hash) {
174 		if (it != ia && IA_SIN(it)->sin_addr.s_addr == in) {
175 			ifa_ref(&it->ia_ifa);
176 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
177 			return (it);
178 		}
179 	}
180 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
181 
182 	return (NULL);
183 }
184 
185 /*
186  * Determine whether an IP address is in a reserved set of addresses
187  * that may not be forwarded, or whether datagrams to that destination
188  * may be forwarded.
189  */
190 int
191 in_canforward(struct in_addr in)
192 {
193 	u_long i = ntohl(in.s_addr);
194 
195 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i) ||
196 	    IN_ZERONET(i) || IN_LOOPBACK(i))
197 		return (0);
198 	return (1);
199 }
200 
201 /*
202  * Trim a mask in a sockaddr
203  */
204 static void
205 in_socktrim(struct sockaddr_in *ap)
206 {
207     char *cplim = (char *) &ap->sin_addr;
208     char *cp = (char *) (&ap->sin_addr + 1);
209 
210     ap->sin_len = 0;
211     while (--cp >= cplim)
212 	if (*cp) {
213 	    (ap)->sin_len = cp - (char *) (ap) + 1;
214 	    break;
215 	}
216 }
217 
218 /*
219  * Generic internet control operations (ioctl's).
220  */
221 int
222 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
223     struct thread *td)
224 {
225 	struct ifreq *ifr = (struct ifreq *)data;
226 	struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr;
227 	struct epoch_tracker et;
228 	struct ifaddr *ifa;
229 	struct in_ifaddr *ia;
230 	int error;
231 
232 	if (ifp == NULL)
233 		return (EADDRNOTAVAIL);
234 
235 	/*
236 	 * Filter out 4 ioctls we implement directly.  Forward the rest
237 	 * to specific functions and ifp->if_ioctl().
238 	 */
239 	switch (cmd) {
240 	case SIOCGIFADDR:
241 	case SIOCGIFBRDADDR:
242 	case SIOCGIFDSTADDR:
243 	case SIOCGIFNETMASK:
244 		break;
245 	case SIOCDIFADDR:
246 		sx_xlock(&in_control_sx);
247 		error = in_difaddr_ioctl(cmd, data, ifp, td);
248 		sx_xunlock(&in_control_sx);
249 		return (error);
250 	case OSIOCAIFADDR:	/* 9.x compat */
251 	case SIOCAIFADDR:
252 		sx_xlock(&in_control_sx);
253 		error = in_aifaddr_ioctl(cmd, data, ifp, td);
254 		sx_xunlock(&in_control_sx);
255 		return (error);
256 	case SIOCSIFADDR:
257 	case SIOCSIFBRDADDR:
258 	case SIOCSIFDSTADDR:
259 	case SIOCSIFNETMASK:
260 		/* We no longer support that old commands. */
261 		return (EINVAL);
262 	default:
263 		if (ifp->if_ioctl == NULL)
264 			return (EOPNOTSUPP);
265 		return ((*ifp->if_ioctl)(ifp, cmd, data));
266 	}
267 
268 	if (addr->sin_addr.s_addr != INADDR_ANY &&
269 	    prison_check_ip4(td->td_ucred, &addr->sin_addr) != 0)
270 		return (EADDRNOTAVAIL);
271 
272 	/*
273 	 * Find address for this interface, if it exists.  If an
274 	 * address was specified, find that one instead of the
275 	 * first one on the interface, if possible.
276 	 */
277 	NET_EPOCH_ENTER(et);
278 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
279 		if (ifa->ifa_addr->sa_family != AF_INET)
280 			continue;
281 		ia = (struct in_ifaddr *)ifa;
282 		if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr)
283 			break;
284 	}
285 	if (ifa == NULL)
286 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
287 			if (ifa->ifa_addr->sa_family == AF_INET) {
288 				ia = (struct in_ifaddr *)ifa;
289 				if (prison_check_ip4(td->td_ucred,
290 				    &ia->ia_addr.sin_addr) == 0)
291 					break;
292 			}
293 
294 	if (ifa == NULL) {
295 		NET_EPOCH_EXIT(et);
296 		return (EADDRNOTAVAIL);
297 	}
298 
299 	error = 0;
300 	switch (cmd) {
301 	case SIOCGIFADDR:
302 		*addr = ia->ia_addr;
303 		break;
304 
305 	case SIOCGIFBRDADDR:
306 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
307 			error = EINVAL;
308 			break;
309 		}
310 		*addr = ia->ia_broadaddr;
311 		break;
312 
313 	case SIOCGIFDSTADDR:
314 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
315 			error = EINVAL;
316 			break;
317 		}
318 		*addr = ia->ia_dstaddr;
319 		break;
320 
321 	case SIOCGIFNETMASK:
322 		*addr = ia->ia_sockmask;
323 		break;
324 	}
325 
326 	NET_EPOCH_EXIT(et);
327 
328 	return (error);
329 }
330 
331 static int
332 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
333 {
334 	const struct in_aliasreq *ifra = (struct in_aliasreq *)data;
335 	const struct sockaddr_in *addr = &ifra->ifra_addr;
336 	const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr;
337 	const struct sockaddr_in *mask = &ifra->ifra_mask;
338 	const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr;
339 	const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0;
340 	struct epoch_tracker et;
341 	struct ifaddr *ifa;
342 	struct in_ifaddr *ia;
343 	bool iaIsFirst;
344 	int error = 0;
345 
346 	error = priv_check(td, PRIV_NET_ADDIFADDR);
347 	if (error)
348 		return (error);
349 
350 	/*
351 	 * ifra_addr must be present and be of INET family.
352 	 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional.
353 	 */
354 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
355 	    addr->sin_family != AF_INET)
356 		return (EINVAL);
357 	if (broadaddr->sin_len != 0 &&
358 	    (broadaddr->sin_len != sizeof(struct sockaddr_in) ||
359 	    broadaddr->sin_family != AF_INET))
360 		return (EINVAL);
361 	if (mask->sin_len != 0 &&
362 	    (mask->sin_len != sizeof(struct sockaddr_in) ||
363 	    mask->sin_family != AF_INET))
364 		return (EINVAL);
365 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
366 	    (dstaddr->sin_len != sizeof(struct sockaddr_in) ||
367 	     dstaddr->sin_addr.s_addr == INADDR_ANY))
368 		return (EDESTADDRREQ);
369 	if (vhid > 0 && carp_attach_p == NULL)
370 		return (EPROTONOSUPPORT);
371 
372 	/*
373 	 * See whether address already exist.
374 	 */
375 	iaIsFirst = true;
376 	ia = NULL;
377 	NET_EPOCH_ENTER(et);
378 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
379 		struct in_ifaddr *it;
380 
381 		if (ifa->ifa_addr->sa_family != AF_INET)
382 			continue;
383 
384 		it = (struct in_ifaddr *)ifa;
385 		iaIsFirst = false;
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 	}
390 	NET_EPOCH_EXIT(et);
391 
392 	if (ia != NULL)
393 		(void )in_difaddr_ioctl(cmd, data, ifp, td);
394 
395 	ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
396 	ia = (struct in_ifaddr *)ifa;
397 	ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
398 	ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
399 	ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
400 	callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock,
401 	    CALLOUT_RETURNUNLOCKED);
402 
403 	ia->ia_ifp = ifp;
404 	ia->ia_addr = *addr;
405 	if (mask->sin_len != 0) {
406 		ia->ia_sockmask = *mask;
407 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
408 	} else {
409 		in_addr_t i = ntohl(addr->sin_addr.s_addr);
410 
411 		/*
412 	 	 * Be compatible with network classes, if netmask isn't
413 		 * supplied, guess it based on classes.
414 	 	 */
415 		if (IN_CLASSA(i))
416 			ia->ia_subnetmask = IN_CLASSA_NET;
417 		else if (IN_CLASSB(i))
418 			ia->ia_subnetmask = IN_CLASSB_NET;
419 		else
420 			ia->ia_subnetmask = IN_CLASSC_NET;
421 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
422 	}
423 	ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask;
424 	in_socktrim(&ia->ia_sockmask);
425 
426 	if (ifp->if_flags & IFF_BROADCAST) {
427 		if (broadaddr->sin_len != 0) {
428 			ia->ia_broadaddr = *broadaddr;
429 		} else if (ia->ia_subnetmask == IN_RFC3021_MASK) {
430 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
431 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
432 			ia->ia_broadaddr.sin_family = AF_INET;
433 		} else {
434 			ia->ia_broadaddr.sin_addr.s_addr =
435 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
436 			ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in);
437 			ia->ia_broadaddr.sin_family = AF_INET;
438 		}
439 	}
440 
441 	if (ifp->if_flags & IFF_POINTOPOINT)
442 		ia->ia_dstaddr = *dstaddr;
443 
444 	/* XXXGL: rtinit() needs this strange assignment. */
445 	if (ifp->if_flags & IFF_LOOPBACK)
446                 ia->ia_dstaddr = ia->ia_addr;
447 
448 	if (vhid != 0) {
449 		error = (*carp_attach_p)(&ia->ia_ifa, vhid);
450 		if (error)
451 			return (error);
452 	}
453 
454 	/* if_addrhead is already referenced by ifa_alloc() */
455 	IF_ADDR_WLOCK(ifp);
456 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
457 	IF_ADDR_WUNLOCK(ifp);
458 
459 	ifa_ref(ifa);			/* in_ifaddrhead */
460 	IN_IFADDR_WLOCK();
461 	CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
462 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
463 	IN_IFADDR_WUNLOCK();
464 
465 	/*
466 	 * Give the interface a chance to initialize
467 	 * if this is its first address,
468 	 * and to validate the address if necessary.
469 	 */
470 	if (ifp->if_ioctl != NULL) {
471 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
472 		if (error)
473 			goto fail1;
474 	}
475 
476 	/*
477 	 * Add route for the network.
478 	 */
479 	if (vhid == 0) {
480 		int flags = RTF_UP;
481 
482 		if (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))
483 			flags |= RTF_HOST;
484 
485 		error = in_addprefix(ia, flags);
486 		if (error)
487 			goto fail1;
488 	}
489 
490 	/*
491 	 * Add a loopback route to self.
492 	 */
493 	if (vhid == 0 && (ifp->if_flags & IFF_LOOPBACK) == 0 &&
494 	    ia->ia_addr.sin_addr.s_addr != INADDR_ANY &&
495 	    !((ifp->if_flags & IFF_POINTOPOINT) &&
496 	     ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)) {
497 		struct in_ifaddr *eia;
498 
499 		eia = in_localip_more(ia);
500 
501 		if (eia == NULL) {
502 			error = ifa_add_loopback_route((struct ifaddr *)ia,
503 			    (struct sockaddr *)&ia->ia_addr);
504 			if (error)
505 				goto fail2;
506 		} else
507 			ifa_free(&eia->ia_ifa);
508 	}
509 
510 	if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) {
511 		struct in_addr allhosts_addr;
512 		struct in_ifinfo *ii;
513 
514 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
515 		allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
516 
517 		error = in_joingroup(ifp, &allhosts_addr, NULL,
518 			&ii->ii_allhosts);
519 	}
520 
521 	/*
522 	 * Note: we don't need extra reference for ifa, since we called
523 	 * with sx lock held, and ifaddr can not be deleted in concurrent
524 	 * thread.
525 	 */
526 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD);
527 
528 	return (error);
529 
530 fail2:
531 	if (vhid == 0)
532 		(void )in_scrubprefix(ia, LLE_STATIC);
533 
534 fail1:
535 	if (ia->ia_ifa.ifa_carp)
536 		(*carp_detach_p)(&ia->ia_ifa, false);
537 
538 	IF_ADDR_WLOCK(ifp);
539 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
540 	IF_ADDR_WUNLOCK(ifp);
541 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
542 
543 	IN_IFADDR_WLOCK();
544 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
545 	LIST_REMOVE(ia, ia_hash);
546 	IN_IFADDR_WUNLOCK();
547 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
548 
549 	return (error);
550 }
551 
552 static int
553 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
554 {
555 	const struct ifreq *ifr = (struct ifreq *)data;
556 	const struct sockaddr_in *addr = (const struct sockaddr_in *)
557 	    &ifr->ifr_addr;
558 	struct ifaddr *ifa;
559 	struct in_ifaddr *ia;
560 	bool deleteAny, iaIsLast;
561 	int error;
562 
563 	if (td != NULL) {
564 		error = priv_check(td, PRIV_NET_DELIFADDR);
565 		if (error)
566 			return (error);
567 	}
568 
569 	if (addr->sin_len != sizeof(struct sockaddr_in) ||
570 	    addr->sin_family != AF_INET)
571 		deleteAny = true;
572 	else
573 		deleteAny = false;
574 
575 	iaIsLast = true;
576 	ia = NULL;
577 	IF_ADDR_WLOCK(ifp);
578 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
579 		struct in_ifaddr *it;
580 
581 		if (ifa->ifa_addr->sa_family != AF_INET)
582 			continue;
583 
584 		it = (struct in_ifaddr *)ifa;
585 		if (deleteAny && ia == NULL && (td == NULL ||
586 		    prison_check_ip4(td->td_ucred, &it->ia_addr.sin_addr) == 0))
587 			ia = it;
588 
589 		if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr &&
590 		    (td == NULL || prison_check_ip4(td->td_ucred,
591 		    &addr->sin_addr) == 0))
592 			ia = it;
593 
594 		if (it != ia)
595 			iaIsLast = false;
596 	}
597 
598 	if (ia == NULL) {
599 		IF_ADDR_WUNLOCK(ifp);
600 		return (EADDRNOTAVAIL);
601 	}
602 
603 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
604 	IF_ADDR_WUNLOCK(ifp);
605 	ifa_free(&ia->ia_ifa);		/* if_addrhead */
606 
607 	IN_IFADDR_WLOCK();
608 	CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link);
609 	LIST_REMOVE(ia, ia_hash);
610 	IN_IFADDR_WUNLOCK();
611 
612 	/*
613 	 * in_scrubprefix() kills the interface route.
614 	 */
615 	in_scrubprefix(ia, LLE_STATIC);
616 
617 	/*
618 	 * in_ifadown gets rid of all the rest of
619 	 * the routes.  This is not quite the right
620 	 * thing to do, but at least if we are running
621 	 * a routing process they will come back.
622 	 */
623 	in_ifadown(&ia->ia_ifa, 1);
624 
625 	if (ia->ia_ifa.ifa_carp)
626 		(*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR);
627 
628 	/*
629 	 * If this is the last IPv4 address configured on this
630 	 * interface, leave the all-hosts group.
631 	 * No state-change report need be transmitted.
632 	 */
633 	if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) {
634 		struct in_ifinfo *ii;
635 
636 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
637 		if (ii->ii_allhosts) {
638 			(void)in_leavegroup(ii->ii_allhosts, NULL);
639 			ii->ii_allhosts = NULL;
640 		}
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_ext, ifp, &ia->ia_ifa,
650 	    IFADDR_EVENT_DEL);
651 	ifa_free(&ia->ia_ifa);		/* in_ifaddrhead */
652 
653 	return (0);
654 }
655 
656 #define rtinitflags(x) \
657 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
658 	    ? RTF_HOST : 0)
659 
660 /*
661  * Check if we have a route for the given prefix already or add one accordingly.
662  */
663 int
664 in_addprefix(struct in_ifaddr *target, int flags)
665 {
666 	struct rm_priotracker in_ifa_tracker;
667 	struct in_ifaddr *ia;
668 	struct in_addr prefix, mask, p, m;
669 	int error;
670 
671 	if ((flags & RTF_HOST) != 0) {
672 		prefix = target->ia_dstaddr.sin_addr;
673 		mask.s_addr = 0;
674 	} else {
675 		prefix = target->ia_addr.sin_addr;
676 		mask = target->ia_sockmask.sin_addr;
677 		prefix.s_addr &= mask.s_addr;
678 	}
679 
680 	IN_IFADDR_RLOCK(&in_ifa_tracker);
681 	/* Look for an existing address with the same prefix, mask, and fib */
682 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
683 		if (rtinitflags(ia)) {
684 			p = ia->ia_dstaddr.sin_addr;
685 
686 			if (prefix.s_addr != p.s_addr)
687 				continue;
688 		} else {
689 			p = ia->ia_addr.sin_addr;
690 			m = ia->ia_sockmask.sin_addr;
691 			p.s_addr &= m.s_addr;
692 
693 			if (prefix.s_addr != p.s_addr ||
694 			    mask.s_addr != m.s_addr)
695 				continue;
696 		}
697 		if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib)
698 			continue;
699 
700 		/*
701 		 * If we got a matching prefix route inserted by other
702 		 * interface address, we are done here.
703 		 */
704 		if (ia->ia_flags & IFA_ROUTE) {
705 #ifdef RADIX_MPATH
706 			if (ia->ia_addr.sin_addr.s_addr ==
707 			    target->ia_addr.sin_addr.s_addr) {
708 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
709 				return (EEXIST);
710 			} else
711 				break;
712 #endif
713 			if (V_nosameprefix) {
714 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
715 				return (EEXIST);
716 			} else {
717 				int fibnum;
718 
719 				fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
720 					target->ia_ifp->if_fib;
721 				rt_addrmsg(RTM_ADD, &target->ia_ifa, fibnum);
722 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
723 				return (0);
724 			}
725 		}
726 	}
727 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
728 
729 	/*
730 	 * No-one seem to have this prefix route, so we try to insert it.
731 	 */
732 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
733 	if (!error)
734 		target->ia_flags |= IFA_ROUTE;
735 	return (error);
736 }
737 
738 /*
739  * Removes either all lle entries for given @ia, or lle
740  * corresponding to @ia address.
741  */
742 static void
743 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags)
744 {
745 	struct sockaddr_in addr, mask;
746 	struct sockaddr *saddr, *smask;
747 	struct ifnet *ifp;
748 
749 	saddr = (struct sockaddr *)&addr;
750 	bzero(&addr, sizeof(addr));
751 	addr.sin_len = sizeof(addr);
752 	addr.sin_family = AF_INET;
753 	smask = (struct sockaddr *)&mask;
754 	bzero(&mask, sizeof(mask));
755 	mask.sin_len = sizeof(mask);
756 	mask.sin_family = AF_INET;
757 	mask.sin_addr.s_addr = ia->ia_subnetmask;
758 	ifp = ia->ia_ifp;
759 
760 	if (all) {
761 
762 		/*
763 		 * Remove all L2 entries matching given prefix.
764 		 * Convert address to host representation to avoid
765 		 * doing this on every callback. ia_subnetmask is already
766 		 * stored in host representation.
767 		 */
768 		addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr);
769 		lltable_prefix_free(AF_INET, saddr, smask, flags);
770 	} else {
771 		/* Remove interface address only */
772 		addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr;
773 		lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr);
774 	}
775 }
776 
777 /*
778  * If there is no other address in the system that can serve a route to the
779  * same prefix, remove the route.  Hand over the route to the new address
780  * otherwise.
781  */
782 int
783 in_scrubprefix(struct in_ifaddr *target, u_int flags)
784 {
785 	struct rm_priotracker in_ifa_tracker;
786 	struct in_ifaddr *ia;
787 	struct in_addr prefix, mask, p, m;
788 	int error = 0;
789 
790 	/*
791 	 * Remove the loopback route to the interface address.
792 	 */
793 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
794 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
795 	    (flags & LLE_STATIC)) {
796 		struct in_ifaddr *eia;
797 
798 		/*
799 		 * XXXME: add fib-aware in_localip.
800 		 * We definitely don't want to switch between
801 		 * prefixes in different fibs.
802 		 */
803 		eia = in_localip_more(target);
804 
805 		if (eia != NULL) {
806 			error = ifa_switch_loopback_route((struct ifaddr *)eia,
807 			    (struct sockaddr *)&target->ia_addr);
808 			ifa_free(&eia->ia_ifa);
809 		} else {
810 			error = ifa_del_loopback_route((struct ifaddr *)target,
811 			    (struct sockaddr *)&target->ia_addr);
812 		}
813 	}
814 
815 	if (rtinitflags(target)) {
816 		prefix = target->ia_dstaddr.sin_addr;
817 		mask.s_addr = 0;
818 	} else {
819 		prefix = target->ia_addr.sin_addr;
820 		mask = target->ia_sockmask.sin_addr;
821 		prefix.s_addr &= mask.s_addr;
822 	}
823 
824 	if ((target->ia_flags & IFA_ROUTE) == 0) {
825 		int fibnum;
826 
827 		fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS :
828 			target->ia_ifp->if_fib;
829 		rt_addrmsg(RTM_DELETE, &target->ia_ifa, fibnum);
830 
831 		/*
832 		 * Removing address from !IFF_UP interface or
833 		 * prefix which exists on other interface (along with route).
834 		 * No entries should exist here except target addr.
835 		 * Given that, delete this entry only.
836 		 */
837 		in_scrubprefixlle(target, 0, flags);
838 		return (0);
839 	}
840 
841 	IN_IFADDR_RLOCK(&in_ifa_tracker);
842 	CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
843 		if (rtinitflags(ia)) {
844 			p = ia->ia_dstaddr.sin_addr;
845 
846 			if (prefix.s_addr != p.s_addr)
847 				continue;
848 		} else {
849 			p = ia->ia_addr.sin_addr;
850 			m = ia->ia_sockmask.sin_addr;
851 			p.s_addr &= m.s_addr;
852 
853 			if (prefix.s_addr != p.s_addr ||
854 			    mask.s_addr != m.s_addr)
855 				continue;
856 		}
857 
858 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
859 			continue;
860 
861 		/*
862 		 * If we got a matching prefix address, move IFA_ROUTE and
863 		 * the route itself to it.  Make sure that routing daemons
864 		 * get a heads-up.
865 		 */
866 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
867 			ifa_ref(&ia->ia_ifa);
868 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
869 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
870 			    rtinitflags(target));
871 			if (error == 0)
872 				target->ia_flags &= ~IFA_ROUTE;
873 			else
874 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
875 					error);
876 			/* Scrub all entries IFF interface is different */
877 			in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp,
878 			    flags);
879 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
880 			    rtinitflags(ia) | RTF_UP);
881 			if (error == 0)
882 				ia->ia_flags |= IFA_ROUTE;
883 			else
884 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
885 					error);
886 			ifa_free(&ia->ia_ifa);
887 			return (error);
888 		}
889 	}
890 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
891 
892 	/*
893 	 * remove all L2 entries on the given prefix
894 	 */
895 	in_scrubprefixlle(target, 1, flags);
896 
897 	/*
898 	 * As no-one seem to have this prefix, we can remove the route.
899 	 */
900 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
901 	if (error == 0)
902 		target->ia_flags &= ~IFA_ROUTE;
903 	else
904 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
905 	return (error);
906 }
907 
908 #undef rtinitflags
909 
910 void
911 in_ifscrub_all(void)
912 {
913 	struct ifnet *ifp;
914 	struct ifaddr *ifa, *nifa;
915 	struct ifaliasreq ifr;
916 
917 	IFNET_RLOCK();
918 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
919 		/* Cannot lock here - lock recursion. */
920 		/* NET_EPOCH_ENTER(et); */
921 		CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
922 			if (ifa->ifa_addr->sa_family != AF_INET)
923 				continue;
924 
925 			/*
926 			 * This is ugly but the only way for legacy IP to
927 			 * cleanly remove addresses and everything attached.
928 			 */
929 			bzero(&ifr, sizeof(ifr));
930 			ifr.ifra_addr = *ifa->ifa_addr;
931 			if (ifa->ifa_dstaddr)
932 			ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
933 			(void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr,
934 			    ifp, NULL);
935 		}
936 		/* NET_EPOCH_EXIT(et); */
937 		in_purgemaddrs(ifp);
938 		igmp_domifdetach(ifp);
939 	}
940 	IFNET_RUNLOCK();
941 }
942 
943 int
944 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia)
945 {
946 
947 	return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
948 	     /*
949 	      * Check for old-style (host 0) broadcast, but
950 	      * taking into account that RFC 3021 obsoletes it.
951 	      */
952 	    (ia->ia_subnetmask != IN_RFC3021_MASK &&
953 	    ntohl(in.s_addr) == ia->ia_subnet)) &&
954 	     /*
955 	      * Check for an all one subnetmask. These
956 	      * only exist when an interface gets a secondary
957 	      * address.
958 	      */
959 	    ia->ia_subnetmask != (u_long)0xffffffff);
960 }
961 
962 /*
963  * Return 1 if the address might be a local broadcast address.
964  */
965 int
966 in_broadcast(struct in_addr in, struct ifnet *ifp)
967 {
968 	struct epoch_tracker et;
969 	struct ifaddr *ifa;
970 	int found;
971 
972 	if (in.s_addr == INADDR_BROADCAST ||
973 	    in.s_addr == INADDR_ANY)
974 		return (1);
975 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
976 		return (0);
977 	found = 0;
978 	/*
979 	 * Look through the list of addresses for a match
980 	 * with a broadcast address.
981 	 */
982 	NET_EPOCH_ENTER(et);
983 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
984 		if (ifa->ifa_addr->sa_family == AF_INET &&
985 		    in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) {
986 			found = 1;
987 			break;
988 		}
989 	NET_EPOCH_EXIT(et);
990 	return (found);
991 }
992 
993 /*
994  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
995  */
996 void
997 in_ifdetach(struct ifnet *ifp)
998 {
999 	IN_MULTI_LOCK();
1000 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1001 	in_pcbpurgeif0(&V_udbinfo, ifp);
1002 	in_pcbpurgeif0(&V_ulitecbinfo, ifp);
1003 	in_purgemaddrs(ifp);
1004 	IN_MULTI_UNLOCK();
1005 }
1006 
1007 /*
1008  * Delete all IPv4 multicast address records, and associated link-layer
1009  * multicast address records, associated with ifp.
1010  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1011  * XXX This should not race with ifma_protospec being set during
1012  * a new allocation, if it does, we have bigger problems.
1013  */
1014 static void
1015 in_purgemaddrs(struct ifnet *ifp)
1016 {
1017 	struct in_multi_head purgeinms;
1018 	struct in_multi		*inm;
1019 	struct ifmultiaddr	*ifma, *next;
1020 
1021 	SLIST_INIT(&purgeinms);
1022 	IN_MULTI_LIST_LOCK();
1023 
1024 	/*
1025 	 * Extract list of in_multi associated with the detaching ifp
1026 	 * which the PF_INET layer is about to release.
1027 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1028 	 * by code further down.
1029 	 */
1030 	IF_ADDR_WLOCK(ifp);
1031  restart:
1032 	CK_STAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next) {
1033 		if (ifma->ifma_addr->sa_family != AF_INET ||
1034 		    ifma->ifma_protospec == NULL)
1035 			continue;
1036 		inm = (struct in_multi *)ifma->ifma_protospec;
1037 		inm_rele_locked(&purgeinms, inm);
1038 		if (__predict_false(ifma_restart)) {
1039 			ifma_restart = true;
1040 			goto restart;
1041 		}
1042 	}
1043 	IF_ADDR_WUNLOCK(ifp);
1044 
1045 	inm_release_list_deferred(&purgeinms);
1046 	igmp_ifdetach(ifp);
1047 	IN_MULTI_LIST_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(epoch_context_t ctx)
1063 {
1064 	struct llentry *lle;
1065 
1066 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
1067 	LLE_LOCK_DESTROY(lle);
1068 	LLE_REQ_DESTROY(lle);
1069 	free(lle, M_LLTABLE);
1070 }
1071 
1072 /*
1073  * Called by the datapath to indicate that
1074  * the entry was used.
1075  */
1076 static void
1077 in_lltable_mark_used(struct llentry *lle)
1078 {
1079 
1080 	LLE_REQ_LOCK(lle);
1081 	lle->r_skip_req = 0;
1082 	LLE_REQ_UNLOCK(lle);
1083 }
1084 
1085 /*
1086  * Called by LLE_FREE_LOCKED when number of references
1087  * drops to zero.
1088  */
1089 static void
1090 in_lltable_destroy_lle(struct llentry *lle)
1091 {
1092 
1093 	LLE_WUNLOCK(lle);
1094 	epoch_call(net_epoch_preempt,  &lle->lle_epoch_ctx, in_lltable_destroy_lle_unlocked);
1095 }
1096 
1097 static struct llentry *
1098 in_lltable_new(struct in_addr addr4, u_int flags)
1099 {
1100 	struct in_llentry *lle;
1101 
1102 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1103 	if (lle == NULL)		/* NB: caller generates msg */
1104 		return NULL;
1105 
1106 	/*
1107 	 * For IPv4 this will trigger "arpresolve" to generate
1108 	 * an ARP request.
1109 	 */
1110 	lle->base.la_expire = time_uptime; /* mark expired */
1111 	lle->base.r_l3addr.addr4 = addr4;
1112 	lle->base.lle_refcnt = 1;
1113 	lle->base.lle_free = in_lltable_destroy_lle;
1114 	LLE_LOCK_INIT(&lle->base);
1115 	LLE_REQ_INIT(&lle->base);
1116 	callout_init(&lle->base.lle_timer, 1);
1117 
1118 	return (&lle->base);
1119 }
1120 
1121 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(		\
1122 	((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 )
1123 
1124 static int
1125 in_lltable_match_prefix(const struct sockaddr *saddr,
1126     const struct sockaddr *smask, u_int flags, struct llentry *lle)
1127 {
1128 	struct in_addr addr, mask, lle_addr;
1129 
1130 	addr = ((const struct sockaddr_in *)saddr)->sin_addr;
1131 	mask = ((const struct sockaddr_in *)smask)->sin_addr;
1132 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1133 
1134 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
1135 		return (0);
1136 
1137 	if (lle->la_flags & LLE_IFADDR) {
1138 
1139 		/*
1140 		 * Delete LLE_IFADDR records IFF address & flag matches.
1141 		 * Note that addr is the interface address within prefix
1142 		 * being matched.
1143 		 * Note also we should handle 'ifdown' cases without removing
1144 		 * ifaddr macs.
1145 		 */
1146 		if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0)
1147 			return (1);
1148 		return (0);
1149 	}
1150 
1151 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
1152 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
1153 		return (1);
1154 
1155 	return (0);
1156 }
1157 
1158 static void
1159 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1160 {
1161 	size_t pkts_dropped;
1162 
1163 	LLE_WLOCK_ASSERT(lle);
1164 	KASSERT(llt != NULL, ("lltable is NULL"));
1165 
1166 	/* Unlink entry from table if not already */
1167 	if ((lle->la_flags & LLE_LINKED) != 0) {
1168 		IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
1169 		lltable_unlink_entry(llt, lle);
1170 	}
1171 
1172 	/* Drop hold queue */
1173 	pkts_dropped = llentry_free(lle);
1174 	ARPSTAT_ADD(dropped, pkts_dropped);
1175 }
1176 
1177 static int
1178 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1179 {
1180 	struct rt_addrinfo info;
1181 	struct sockaddr_in rt_key, rt_mask;
1182 	struct sockaddr rt_gateway;
1183 	int rt_flags;
1184 
1185 	KASSERT(l3addr->sa_family == AF_INET,
1186 	    ("sin_family %d", l3addr->sa_family));
1187 
1188 	bzero(&rt_key, sizeof(rt_key));
1189 	rt_key.sin_len = sizeof(rt_key);
1190 	bzero(&rt_mask, sizeof(rt_mask));
1191 	rt_mask.sin_len = sizeof(rt_mask);
1192 	bzero(&rt_gateway, sizeof(rt_gateway));
1193 	rt_gateway.sa_len = sizeof(rt_gateway);
1194 
1195 	bzero(&info, sizeof(info));
1196 	info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
1197 	info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&rt_mask;
1198 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
1199 
1200 	if (rib_lookup_info(ifp->if_fib, l3addr, NHR_REF, 0, &info) != 0)
1201 		return (EINVAL);
1202 
1203 	rt_flags = info.rti_flags;
1204 
1205 	/*
1206 	 * If the gateway for an existing host route matches the target L3
1207 	 * address, which is a special route inserted by some implementation
1208 	 * such as MANET, and the interface is of the correct type, then
1209 	 * allow for ARP to proceed.
1210 	 */
1211 	if (rt_flags & RTF_GATEWAY) {
1212 		if (!(rt_flags & RTF_HOST) || !info.rti_ifp ||
1213 		    info.rti_ifp->if_type != IFT_ETHER ||
1214 		    (info.rti_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1215 		    memcmp(rt_gateway.sa_data, l3addr->sa_data,
1216 		    sizeof(in_addr_t)) != 0) {
1217 			rib_free_info(&info);
1218 			return (EINVAL);
1219 		}
1220 	}
1221 	rib_free_info(&info);
1222 
1223 	/*
1224 	 * Make sure that at least the destination address is covered
1225 	 * by the route. This is for handling the case where 2 or more
1226 	 * interfaces have the same prefix. An incoming packet arrives
1227 	 * on one interface and the corresponding outgoing packet leaves
1228 	 * another interface.
1229 	 */
1230 	if (!(rt_flags & RTF_HOST) && info.rti_ifp != ifp) {
1231 		const char *sa, *mask, *addr, *lim;
1232 		const struct sockaddr_in *l3sin;
1233 
1234 		mask = (const char *)&rt_mask;
1235 		/*
1236 		 * Just being extra cautious to avoid some custom
1237 		 * code getting into trouble.
1238 		 */
1239 		if ((info.rti_addrs & RTA_NETMASK) == 0)
1240 			return (EINVAL);
1241 
1242 		sa = (const char *)&rt_key;
1243 		addr = (const char *)l3addr;
1244 		l3sin = (const struct sockaddr_in *)l3addr;
1245 		lim = addr + l3sin->sin_len;
1246 
1247 		for ( ; addr < lim; sa++, mask++, addr++) {
1248 			if ((*sa ^ *addr) & *mask) {
1249 #ifdef DIAGNOSTIC
1250 				char addrbuf[INET_ADDRSTRLEN];
1251 
1252 				log(LOG_INFO, "IPv4 address: \"%s\" "
1253 				    "is not on the network\n",
1254 				    inet_ntoa_r(l3sin->sin_addr, addrbuf));
1255 #endif
1256 				return (EINVAL);
1257 			}
1258 		}
1259 	}
1260 
1261 	return (0);
1262 }
1263 
1264 static inline uint32_t
1265 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1266 {
1267 
1268 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1269 }
1270 
1271 static uint32_t
1272 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1273 {
1274 
1275 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1276 }
1277 
1278 static void
1279 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1280 {
1281 	struct sockaddr_in *sin;
1282 
1283 	sin = (struct sockaddr_in *)sa;
1284 	bzero(sin, sizeof(*sin));
1285 	sin->sin_family = AF_INET;
1286 	sin->sin_len = sizeof(*sin);
1287 	sin->sin_addr = lle->r_l3addr.addr4;
1288 }
1289 
1290 static inline struct llentry *
1291 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1292 {
1293 	struct llentry *lle;
1294 	struct llentries *lleh;
1295 	u_int hashidx;
1296 
1297 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1298 	lleh = &llt->lle_head[hashidx];
1299 	CK_LIST_FOREACH(lle, lleh, lle_next) {
1300 		if (lle->la_flags & LLE_DELETED)
1301 			continue;
1302 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1303 			break;
1304 	}
1305 
1306 	return (lle);
1307 }
1308 
1309 static void
1310 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
1311 {
1312 
1313 	lle->la_flags |= LLE_DELETED;
1314 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1315 #ifdef DIAGNOSTIC
1316 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1317 #endif
1318 	llentry_free(lle);
1319 }
1320 
1321 static struct llentry *
1322 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1323 {
1324 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1325 	struct ifnet *ifp = llt->llt_ifp;
1326 	struct llentry *lle;
1327 	char linkhdr[LLE_MAX_LINKHDR];
1328 	size_t linkhdrsize;
1329 	int lladdr_off;
1330 
1331 	KASSERT(l3addr->sa_family == AF_INET,
1332 	    ("sin_family %d", l3addr->sa_family));
1333 
1334 	/*
1335 	 * A route that covers the given address must have
1336 	 * been installed 1st because we are doing a resolution,
1337 	 * verify this.
1338 	 */
1339 	if (!(flags & LLE_IFADDR) &&
1340 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1341 		return (NULL);
1342 
1343 	lle = in_lltable_new(sin->sin_addr, flags);
1344 	if (lle == NULL) {
1345 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1346 		return (NULL);
1347 	}
1348 	lle->la_flags = flags;
1349 	if (flags & LLE_STATIC)
1350 		lle->r_flags |= RLLE_VALID;
1351 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
1352 		linkhdrsize = LLE_MAX_LINKHDR;
1353 		if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp),
1354 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
1355 			epoch_call(net_epoch_preempt,  &lle->lle_epoch_ctx, in_lltable_destroy_lle_unlocked);
1356 			return (NULL);
1357 		}
1358 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
1359 		    lladdr_off);
1360 		lle->la_flags |= LLE_STATIC;
1361 		lle->r_flags |= (RLLE_VALID | RLLE_IFADDR);
1362 	}
1363 
1364 	return (lle);
1365 }
1366 
1367 /*
1368  * Return NULL if not found or marked for deletion.
1369  * If found return lle read locked.
1370  */
1371 static struct llentry *
1372 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1373 {
1374 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1375 	struct llentry *lle;
1376 
1377 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
1378 	KASSERT(l3addr->sa_family == AF_INET,
1379 	    ("sin_family %d", l3addr->sa_family));
1380 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
1381 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
1382 	    ("wrong lle request flags: %#x", flags));
1383 
1384 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1385 	if (lle == NULL)
1386 		return (NULL);
1387 	if (flags & LLE_UNLOCKED)
1388 		return (lle);
1389 
1390 	if (flags & LLE_EXCLUSIVE)
1391 		LLE_WLOCK(lle);
1392 	else
1393 		LLE_RLOCK(lle);
1394 
1395 	/*
1396 	 * If the afdata lock is not held, the LLE may have been unlinked while
1397 	 * we were blocked on the LLE lock.  Check for this case.
1398 	 */
1399 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
1400 		if (flags & LLE_EXCLUSIVE)
1401 			LLE_WUNLOCK(lle);
1402 		else
1403 			LLE_RUNLOCK(lle);
1404 		return (NULL);
1405 	}
1406 	return (lle);
1407 }
1408 
1409 static int
1410 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
1411     struct sysctl_req *wr)
1412 {
1413 	struct ifnet *ifp = llt->llt_ifp;
1414 	/* XXX stack use */
1415 	struct {
1416 		struct rt_msghdr	rtm;
1417 		struct sockaddr_in	sin;
1418 		struct sockaddr_dl	sdl;
1419 	} arpc;
1420 	struct sockaddr_dl *sdl;
1421 	int error;
1422 
1423 	bzero(&arpc, sizeof(arpc));
1424 	/* skip deleted entries */
1425 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1426 		return (0);
1427 	/* Skip if jailed and not a valid IP of the prison. */
1428 	lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin);
1429 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0)
1430 		return (0);
1431 	/*
1432 	 * produce a msg made of:
1433 	 *  struct rt_msghdr;
1434 	 *  struct sockaddr_in; (IPv4)
1435 	 *  struct sockaddr_dl;
1436 	 */
1437 	arpc.rtm.rtm_msglen = sizeof(arpc);
1438 	arpc.rtm.rtm_version = RTM_VERSION;
1439 	arpc.rtm.rtm_type = RTM_GET;
1440 	arpc.rtm.rtm_flags = RTF_UP;
1441 	arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1442 
1443 	/* publish */
1444 	if (lle->la_flags & LLE_PUB)
1445 		arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1446 
1447 	sdl = &arpc.sdl;
1448 	sdl->sdl_family = AF_LINK;
1449 	sdl->sdl_len = sizeof(*sdl);
1450 	sdl->sdl_index = ifp->if_index;
1451 	sdl->sdl_type = ifp->if_type;
1452 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1453 		sdl->sdl_alen = ifp->if_addrlen;
1454 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1455 	} else {
1456 		sdl->sdl_alen = 0;
1457 		bzero(LLADDR(sdl), ifp->if_addrlen);
1458 	}
1459 
1460 	arpc.rtm.rtm_rmx.rmx_expire =
1461 	    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1462 	arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1463 	if (lle->la_flags & LLE_STATIC)
1464 		arpc.rtm.rtm_flags |= RTF_STATIC;
1465 	if (lle->la_flags & LLE_IFADDR)
1466 		arpc.rtm.rtm_flags |= RTF_PINNED;
1467 	arpc.rtm.rtm_index = ifp->if_index;
1468 	error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1469 
1470 	return (error);
1471 }
1472 
1473 static struct lltable *
1474 in_lltattach(struct ifnet *ifp)
1475 {
1476 	struct lltable *llt;
1477 
1478 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
1479  	llt->llt_af = AF_INET;
1480  	llt->llt_ifp = ifp;
1481 
1482 	llt->llt_lookup = in_lltable_lookup;
1483 	llt->llt_alloc_entry = in_lltable_alloc;
1484 	llt->llt_delete_entry = in_lltable_delete_entry;
1485 	llt->llt_dump_entry = in_lltable_dump_entry;
1486 	llt->llt_hash = in_lltable_hash;
1487 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
1488 	llt->llt_free_entry = in_lltable_free_entry;
1489 	llt->llt_match_prefix = in_lltable_match_prefix;
1490 	llt->llt_mark_used = in_lltable_mark_used;
1491  	lltable_link(llt);
1492 
1493 	return (llt);
1494 }
1495 
1496 void *
1497 in_domifattach(struct ifnet *ifp)
1498 {
1499 	struct in_ifinfo *ii;
1500 
1501 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1502 
1503 	ii->ii_llt = in_lltattach(ifp);
1504 	ii->ii_igmp = igmp_domifattach(ifp);
1505 
1506 	return (ii);
1507 }
1508 
1509 void
1510 in_domifdetach(struct ifnet *ifp, void *aux)
1511 {
1512 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1513 
1514 	igmp_domifdetach(ifp);
1515 	lltable_free(ii->ii_llt);
1516 	free(ii, M_IFADDR);
1517 }
1518