xref: /freebsd/sys/netinet/in.c (revision eb6d21b4ca6d668cf89afd99eef7baeafa712197)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  * Copyright (C) 2001 WIDE Project.  All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 4. Neither the name of the University nor the names of its contributors
15  *    may be used to endorse or promote products derived from this software
16  *    without specific prior written permission.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  *
30  *	@(#)in.c	8.4 (Berkeley) 1/9/95
31  */
32 
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35 
36 #include "opt_carp.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/sockio.h>
41 #include <sys/malloc.h>
42 #include <sys/priv.h>
43 #include <sys/socket.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/proc.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 
50 #include <net/if.h>
51 #include <net/if_var.h>
52 #include <net/if_dl.h>
53 #include <net/if_llatbl.h>
54 #include <net/if_types.h>
55 #include <net/route.h>
56 #include <net/vnet.h>
57 
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/ip_var.h>
62 #include <netinet/igmp_var.h>
63 #include <netinet/udp.h>
64 #include <netinet/udp_var.h>
65 
66 static int in_mask2len(struct in_addr *);
67 static void in_len2mask(struct in_addr *, int);
68 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
69 	struct ifnet *, struct thread *);
70 
71 static int	in_addprefix(struct in_ifaddr *, int);
72 static int	in_scrubprefix(struct in_ifaddr *);
73 static void	in_socktrim(struct sockaddr_in *);
74 static int	in_ifinit(struct ifnet *,
75 	    struct in_ifaddr *, struct sockaddr_in *, int);
76 static void	in_purgemaddrs(struct ifnet *);
77 
78 static VNET_DEFINE(int, subnetsarelocal);
79 static VNET_DEFINE(int, sameprefixcarponly);
80 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
81 
82 #define	V_subnetsarelocal		VNET(subnetsarelocal)
83 #define	V_sameprefixcarponly		VNET(sameprefixcarponly)
84 #define	V_ripcbinfo			VNET(ripcbinfo)
85 
86 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
87 	&VNET_NAME(subnetsarelocal), 0,
88 	"Treat all subnets as directly connected");
89 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
90 	&VNET_NAME(sameprefixcarponly), 0,
91 	"Refuse to create same prefixes on different interfaces");
92 
93 /*
94  * Return 1 if an internet address is for a ``local'' host
95  * (one to which we have a connection).  If subnetsarelocal
96  * is true, this includes other subnets of the local net.
97  * Otherwise, it includes only the directly-connected (sub)nets.
98  */
99 int
100 in_localaddr(struct in_addr in)
101 {
102 	register u_long i = ntohl(in.s_addr);
103 	register struct in_ifaddr *ia;
104 
105 	IN_IFADDR_RLOCK();
106 	if (V_subnetsarelocal) {
107 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
108 			if ((i & ia->ia_netmask) == ia->ia_net) {
109 				IN_IFADDR_RUNLOCK();
110 				return (1);
111 			}
112 		}
113 	} else {
114 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
115 			if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
116 				IN_IFADDR_RUNLOCK();
117 				return (1);
118 			}
119 		}
120 	}
121 	IN_IFADDR_RUNLOCK();
122 	return (0);
123 }
124 
125 /*
126  * Return 1 if an internet address is for the local host and configured
127  * on one of its interfaces.
128  */
129 int
130 in_localip(struct in_addr in)
131 {
132 	struct in_ifaddr *ia;
133 
134 	IN_IFADDR_RLOCK();
135 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
136 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
137 			IN_IFADDR_RUNLOCK();
138 			return (1);
139 		}
140 	}
141 	IN_IFADDR_RUNLOCK();
142 	return (0);
143 }
144 
145 /*
146  * Determine whether an IP address is in a reserved set of addresses
147  * that may not be forwarded, or whether datagrams to that destination
148  * may be forwarded.
149  */
150 int
151 in_canforward(struct in_addr in)
152 {
153 	register u_long i = ntohl(in.s_addr);
154 	register u_long net;
155 
156 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
157 		return (0);
158 	if (IN_CLASSA(i)) {
159 		net = i & IN_CLASSA_NET;
160 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
161 			return (0);
162 	}
163 	return (1);
164 }
165 
166 /*
167  * Trim a mask in a sockaddr
168  */
169 static void
170 in_socktrim(struct sockaddr_in *ap)
171 {
172     register char *cplim = (char *) &ap->sin_addr;
173     register char *cp = (char *) (&ap->sin_addr + 1);
174 
175     ap->sin_len = 0;
176     while (--cp >= cplim)
177 	if (*cp) {
178 	    (ap)->sin_len = cp - (char *) (ap) + 1;
179 	    break;
180 	}
181 }
182 
183 static int
184 in_mask2len(mask)
185 	struct in_addr *mask;
186 {
187 	int x, y;
188 	u_char *p;
189 
190 	p = (u_char *)mask;
191 	for (x = 0; x < sizeof(*mask); x++) {
192 		if (p[x] != 0xff)
193 			break;
194 	}
195 	y = 0;
196 	if (x < sizeof(*mask)) {
197 		for (y = 0; y < 8; y++) {
198 			if ((p[x] & (0x80 >> y)) == 0)
199 				break;
200 		}
201 	}
202 	return (x * 8 + y);
203 }
204 
205 static void
206 in_len2mask(struct in_addr *mask, int len)
207 {
208 	int i;
209 	u_char *p;
210 
211 	p = (u_char *)mask;
212 	bzero(mask, sizeof(*mask));
213 	for (i = 0; i < len / 8; i++)
214 		p[i] = 0xff;
215 	if (len % 8)
216 		p[i] = (0xff00 >> (len % 8)) & 0xff;
217 }
218 
219 /*
220  * Generic internet control operations (ioctl's).
221  *
222  * ifp is NULL if not an interface-specific ioctl.
223  */
224 /* ARGSUSED */
225 int
226 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
227     struct thread *td)
228 {
229 	register struct ifreq *ifr = (struct ifreq *)data;
230 	register struct in_ifaddr *ia, *iap;
231 	register struct ifaddr *ifa;
232 	struct in_addr allhosts_addr;
233 	struct in_addr dst;
234 	struct in_ifinfo *ii;
235 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
236 	struct sockaddr_in oldaddr;
237 	int error, hostIsNew, iaIsNew, maskIsNew;
238 	int iaIsFirst;
239 
240 	ia = NULL;
241 	iaIsFirst = 0;
242 	iaIsNew = 0;
243 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
244 
245 	/*
246 	 * Filter out ioctls we implement directly; forward the rest on to
247 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
248 	 */
249 	switch (cmd) {
250 	case SIOCAIFADDR:
251 	case SIOCDIFADDR:
252 	case SIOCGIFADDR:
253 	case SIOCGIFBRDADDR:
254 	case SIOCGIFDSTADDR:
255 	case SIOCGIFNETMASK:
256 	case SIOCSIFADDR:
257 	case SIOCSIFBRDADDR:
258 	case SIOCSIFDSTADDR:
259 	case SIOCSIFNETMASK:
260 		break;
261 
262 	case SIOCALIFADDR:
263 		if (td != NULL) {
264 			error = priv_check(td, PRIV_NET_ADDIFADDR);
265 			if (error)
266 				return (error);
267 		}
268 		if (ifp == NULL)
269 			return (EINVAL);
270 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
271 
272 	case SIOCDLIFADDR:
273 		if (td != NULL) {
274 			error = priv_check(td, PRIV_NET_DELIFADDR);
275 			if (error)
276 				return (error);
277 		}
278 		if (ifp == NULL)
279 			return (EINVAL);
280 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
281 
282 	case SIOCGLIFADDR:
283 		if (ifp == NULL)
284 			return (EINVAL);
285 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
286 
287 	default:
288 		if (ifp == NULL || ifp->if_ioctl == NULL)
289 			return (EOPNOTSUPP);
290 		return ((*ifp->if_ioctl)(ifp, cmd, data));
291 	}
292 
293 	if (ifp == NULL)
294 		return (EADDRNOTAVAIL);
295 
296 	/*
297 	 * Security checks before we get involved in any work.
298 	 */
299 	switch (cmd) {
300 	case SIOCAIFADDR:
301 	case SIOCSIFADDR:
302 	case SIOCSIFBRDADDR:
303 	case SIOCSIFNETMASK:
304 	case SIOCSIFDSTADDR:
305 		if (td != NULL) {
306 			error = priv_check(td, PRIV_NET_ADDIFADDR);
307 			if (error)
308 				return (error);
309 		}
310 		break;
311 
312 	case SIOCDIFADDR:
313 		if (td != NULL) {
314 			error = priv_check(td, PRIV_NET_DELIFADDR);
315 			if (error)
316 				return (error);
317 		}
318 		break;
319 	}
320 
321 	/*
322 	 * Find address for this interface, if it exists.
323 	 *
324 	 * If an alias address was specified, find that one instead of the
325 	 * first one on the interface, if possible.
326 	 */
327 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
328 	IN_IFADDR_RLOCK();
329 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
330 		if (iap->ia_ifp == ifp &&
331 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
332 			if (td == NULL || prison_check_ip4(td->td_ucred,
333 			    &dst) == 0)
334 				ia = iap;
335 			break;
336 		}
337 	}
338 	if (ia != NULL)
339 		ifa_ref(&ia->ia_ifa);
340 	IN_IFADDR_RUNLOCK();
341 	if (ia == NULL) {
342 		IF_ADDR_LOCK(ifp);
343 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
344 			iap = ifatoia(ifa);
345 			if (iap->ia_addr.sin_family == AF_INET) {
346 				if (td != NULL &&
347 				    prison_check_ip4(td->td_ucred,
348 				    &iap->ia_addr.sin_addr) != 0)
349 					continue;
350 				ia = iap;
351 				break;
352 			}
353 		}
354 		if (ia != NULL)
355 			ifa_ref(&ia->ia_ifa);
356 		IF_ADDR_UNLOCK(ifp);
357 	}
358 	if (ia == NULL)
359 		iaIsFirst = 1;
360 
361 	error = 0;
362 	switch (cmd) {
363 	case SIOCAIFADDR:
364 	case SIOCDIFADDR:
365 		if (ifra->ifra_addr.sin_family == AF_INET) {
366 			struct in_ifaddr *oia;
367 
368 			IN_IFADDR_RLOCK();
369 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
370 				if (ia->ia_ifp == ifp  &&
371 				    ia->ia_addr.sin_addr.s_addr ==
372 				    ifra->ifra_addr.sin_addr.s_addr)
373 					break;
374 			}
375 			if (ia != NULL && ia != oia)
376 				ifa_ref(&ia->ia_ifa);
377 			if (oia != NULL && ia != oia)
378 				ifa_free(&oia->ia_ifa);
379 			IN_IFADDR_RUNLOCK();
380 			if ((ifp->if_flags & IFF_POINTOPOINT)
381 			    && (cmd == SIOCAIFADDR)
382 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
383 				== INADDR_ANY)) {
384 				error = EDESTADDRREQ;
385 				goto out;
386 			}
387 		}
388 		if (cmd == SIOCDIFADDR && ia == NULL) {
389 			error = EADDRNOTAVAIL;
390 			goto out;
391 		}
392 		/* FALLTHROUGH */
393 	case SIOCSIFADDR:
394 	case SIOCSIFNETMASK:
395 	case SIOCSIFDSTADDR:
396 		if (ia == NULL) {
397 			ia = (struct in_ifaddr *)
398 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
399 				    M_ZERO);
400 			if (ia == NULL) {
401 				error = ENOBUFS;
402 				goto out;
403 			}
404 
405 			ifa = &ia->ia_ifa;
406 			ifa_init(ifa);
407 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
408 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
409 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
410 
411 			ia->ia_sockmask.sin_len = 8;
412 			ia->ia_sockmask.sin_family = AF_INET;
413 			if (ifp->if_flags & IFF_BROADCAST) {
414 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
415 				ia->ia_broadaddr.sin_family = AF_INET;
416 			}
417 			ia->ia_ifp = ifp;
418 
419 			ifa_ref(ifa);			/* if_addrhead */
420 			IF_ADDR_LOCK(ifp);
421 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
422 			IF_ADDR_UNLOCK(ifp);
423 			ifa_ref(ifa);			/* in_ifaddrhead */
424 			IN_IFADDR_WLOCK();
425 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
426 			IN_IFADDR_WUNLOCK();
427 			iaIsNew = 1;
428 		}
429 		break;
430 
431 	case SIOCSIFBRDADDR:
432 	case SIOCGIFADDR:
433 	case SIOCGIFNETMASK:
434 	case SIOCGIFDSTADDR:
435 	case SIOCGIFBRDADDR:
436 		if (ia == NULL) {
437 			error = EADDRNOTAVAIL;
438 			goto out;
439 		}
440 		break;
441 	}
442 
443 	/*
444 	 * Most paths in this switch return directly or via out.  Only paths
445 	 * that remove the address break in order to hit common removal code.
446 	 */
447 	switch (cmd) {
448 	case SIOCGIFADDR:
449 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
450 		goto out;
451 
452 	case SIOCGIFBRDADDR:
453 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
454 			error = EINVAL;
455 			goto out;
456 		}
457 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
458 		goto out;
459 
460 	case SIOCGIFDSTADDR:
461 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
462 			error = EINVAL;
463 			goto out;
464 		}
465 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
466 		goto out;
467 
468 	case SIOCGIFNETMASK:
469 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
470 		goto out;
471 
472 	case SIOCSIFDSTADDR:
473 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
474 			error = EINVAL;
475 			goto out;
476 		}
477 		oldaddr = ia->ia_dstaddr;
478 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
479 		if (ifp->if_ioctl != NULL) {
480 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
481 			    (caddr_t)ia);
482 			if (error) {
483 				ia->ia_dstaddr = oldaddr;
484 				goto out;
485 			}
486 		}
487 		if (ia->ia_flags & IFA_ROUTE) {
488 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
489 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
490 			ia->ia_ifa.ifa_dstaddr =
491 					(struct sockaddr *)&ia->ia_dstaddr;
492 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
493 		}
494 		goto out;
495 
496 	case SIOCSIFBRDADDR:
497 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
498 			error = EINVAL;
499 			goto out;
500 		}
501 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
502 		goto out;
503 
504 	case SIOCSIFADDR:
505 		error = in_ifinit(ifp, ia,
506 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
507 		if (error != 0 && iaIsNew)
508 			break;
509 		if (error == 0) {
510 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
511 			if (iaIsFirst &&
512 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
513 				error = in_joingroup(ifp, &allhosts_addr,
514 				    NULL, &ii->ii_allhosts);
515 			}
516 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
517 		}
518 		error = 0;
519 		goto out;
520 
521 	case SIOCSIFNETMASK:
522 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
523 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
524 		goto out;
525 
526 	case SIOCAIFADDR:
527 		maskIsNew = 0;
528 		hostIsNew = 1;
529 		error = 0;
530 		if (ia->ia_addr.sin_family == AF_INET) {
531 			if (ifra->ifra_addr.sin_len == 0) {
532 				ifra->ifra_addr = ia->ia_addr;
533 				hostIsNew = 0;
534 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
535 					       ia->ia_addr.sin_addr.s_addr)
536 				hostIsNew = 0;
537 		}
538 		if (ifra->ifra_mask.sin_len) {
539 			/*
540 			 * QL: XXX
541 			 * Need to scrub the prefix here in case
542 			 * the issued command is SIOCAIFADDR with
543 			 * the same address, but with a different
544 			 * prefix length. And if the prefix length
545 			 * is the same as before, then the call is
546 			 * un-necessarily executed here.
547 			 */
548 			in_ifscrub(ifp, ia);
549 			ia->ia_sockmask = ifra->ifra_mask;
550 			ia->ia_sockmask.sin_family = AF_INET;
551 			ia->ia_subnetmask =
552 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
553 			maskIsNew = 1;
554 		}
555 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
556 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
557 			in_ifscrub(ifp, ia);
558 			ia->ia_dstaddr = ifra->ifra_dstaddr;
559 			maskIsNew  = 1; /* We lie; but the effect's the same */
560 		}
561 		if (ifra->ifra_addr.sin_family == AF_INET &&
562 		    (hostIsNew || maskIsNew))
563 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
564 		if (error != 0 && iaIsNew)
565 			goto out;
566 
567 		if ((ifp->if_flags & IFF_BROADCAST) &&
568 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
569 			ia->ia_broadaddr = ifra->ifra_broadaddr;
570 		if (error == 0) {
571 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
572 			if (iaIsFirst &&
573 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
574 				error = in_joingroup(ifp, &allhosts_addr,
575 				    NULL, &ii->ii_allhosts);
576 			}
577 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
578 		}
579 		goto out;
580 
581 	case SIOCDIFADDR:
582 		/*
583 		 * in_ifscrub kills the interface route.
584 		 */
585 		in_ifscrub(ifp, ia);
586 
587 		/*
588 		 * in_ifadown gets rid of all the rest of
589 		 * the routes.  This is not quite the right
590 		 * thing to do, but at least if we are running
591 		 * a routing process they will come back.
592 		 */
593 		in_ifadown(&ia->ia_ifa, 1);
594 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
595 		error = 0;
596 		break;
597 
598 	default:
599 		panic("in_control: unsupported ioctl");
600 	}
601 
602 	IF_ADDR_LOCK(ifp);
603 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
604 	IF_ADDR_UNLOCK(ifp);
605 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
606 
607 	IN_IFADDR_WLOCK();
608 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
609 	if (ia->ia_addr.sin_family == AF_INET) {
610 		struct in_ifaddr *if_ia;
611 
612 		LIST_REMOVE(ia, ia_hash);
613 		IN_IFADDR_WUNLOCK();
614 		/*
615 		 * If this is the last IPv4 address configured on this
616 		 * interface, leave the all-hosts group.
617 		 * No state-change report need be transmitted.
618 		 */
619 		if_ia = NULL;
620 		IFP_TO_IA(ifp, if_ia);
621 		if (if_ia == NULL) {
622 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
623 			IN_MULTI_LOCK();
624 			if (ii->ii_allhosts) {
625 				(void)in_leavegroup_locked(ii->ii_allhosts,
626 				    NULL);
627 				ii->ii_allhosts = NULL;
628 			}
629 			IN_MULTI_UNLOCK();
630 		} else
631 			ifa_free(&if_ia->ia_ifa);
632 	} else
633 		IN_IFADDR_WUNLOCK();
634 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
635 out:
636 	if (ia != NULL)
637 		ifa_free(&ia->ia_ifa);
638 	return (error);
639 }
640 
641 /*
642  * SIOC[GAD]LIFADDR.
643  *	SIOCGLIFADDR: get first address. (?!?)
644  *	SIOCGLIFADDR with IFLR_PREFIX:
645  *		get first address that matches the specified prefix.
646  *	SIOCALIFADDR: add the specified address.
647  *	SIOCALIFADDR with IFLR_PREFIX:
648  *		EINVAL since we can't deduce hostid part of the address.
649  *	SIOCDLIFADDR: delete the specified address.
650  *	SIOCDLIFADDR with IFLR_PREFIX:
651  *		delete the first address that matches the specified prefix.
652  * return values:
653  *	EINVAL on invalid parameters
654  *	EADDRNOTAVAIL on prefix match failed/specified address not found
655  *	other values may be returned from in_ioctl()
656  */
657 static int
658 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
659     struct ifnet *ifp, struct thread *td)
660 {
661 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
662 	struct ifaddr *ifa;
663 
664 	/* sanity checks */
665 	if (data == NULL || ifp == NULL) {
666 		panic("invalid argument to in_lifaddr_ioctl");
667 		/*NOTRECHED*/
668 	}
669 
670 	switch (cmd) {
671 	case SIOCGLIFADDR:
672 		/* address must be specified on GET with IFLR_PREFIX */
673 		if ((iflr->flags & IFLR_PREFIX) == 0)
674 			break;
675 		/*FALLTHROUGH*/
676 	case SIOCALIFADDR:
677 	case SIOCDLIFADDR:
678 		/* address must be specified on ADD and DELETE */
679 		if (iflr->addr.ss_family != AF_INET)
680 			return (EINVAL);
681 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
682 			return (EINVAL);
683 		/* XXX need improvement */
684 		if (iflr->dstaddr.ss_family
685 		 && iflr->dstaddr.ss_family != AF_INET)
686 			return (EINVAL);
687 		if (iflr->dstaddr.ss_family
688 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
689 			return (EINVAL);
690 		break;
691 	default: /*shouldn't happen*/
692 		return (EOPNOTSUPP);
693 	}
694 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
695 		return (EINVAL);
696 
697 	switch (cmd) {
698 	case SIOCALIFADDR:
699 	    {
700 		struct in_aliasreq ifra;
701 
702 		if (iflr->flags & IFLR_PREFIX)
703 			return (EINVAL);
704 
705 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
706 		bzero(&ifra, sizeof(ifra));
707 		bcopy(iflr->iflr_name, ifra.ifra_name,
708 			sizeof(ifra.ifra_name));
709 
710 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
711 
712 		if (iflr->dstaddr.ss_family) {	/*XXX*/
713 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
714 				iflr->dstaddr.ss_len);
715 		}
716 
717 		ifra.ifra_mask.sin_family = AF_INET;
718 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
719 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
720 
721 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
722 	    }
723 	case SIOCGLIFADDR:
724 	case SIOCDLIFADDR:
725 	    {
726 		struct in_ifaddr *ia;
727 		struct in_addr mask, candidate, match;
728 		struct sockaddr_in *sin;
729 
730 		bzero(&mask, sizeof(mask));
731 		bzero(&match, sizeof(match));
732 		if (iflr->flags & IFLR_PREFIX) {
733 			/* lookup a prefix rather than address. */
734 			in_len2mask(&mask, iflr->prefixlen);
735 
736 			sin = (struct sockaddr_in *)&iflr->addr;
737 			match.s_addr = sin->sin_addr.s_addr;
738 			match.s_addr &= mask.s_addr;
739 
740 			/* if you set extra bits, that's wrong */
741 			if (match.s_addr != sin->sin_addr.s_addr)
742 				return (EINVAL);
743 
744 		} else {
745 			/* on getting an address, take the 1st match */
746 			/* on deleting an address, do exact match */
747 			if (cmd != SIOCGLIFADDR) {
748 				in_len2mask(&mask, 32);
749 				sin = (struct sockaddr_in *)&iflr->addr;
750 				match.s_addr = sin->sin_addr.s_addr;
751 			}
752 		}
753 
754 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
755 			if (ifa->ifa_addr->sa_family != AF_INET6)
756 				continue;
757 			if (match.s_addr == 0)
758 				break;
759 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
760 			candidate.s_addr &= mask.s_addr;
761 			if (candidate.s_addr == match.s_addr)
762 				break;
763 		}
764 		if (ifa == NULL)
765 			return (EADDRNOTAVAIL);
766 		ia = (struct in_ifaddr *)ifa;
767 
768 		if (cmd == SIOCGLIFADDR) {
769 			/* fill in the if_laddrreq structure */
770 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
771 
772 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
773 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
774 					ia->ia_dstaddr.sin_len);
775 			} else
776 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
777 
778 			iflr->prefixlen =
779 				in_mask2len(&ia->ia_sockmask.sin_addr);
780 
781 			iflr->flags = 0;	/*XXX*/
782 
783 			return (0);
784 		} else {
785 			struct in_aliasreq ifra;
786 
787 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
788 			bzero(&ifra, sizeof(ifra));
789 			bcopy(iflr->iflr_name, ifra.ifra_name,
790 				sizeof(ifra.ifra_name));
791 
792 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
793 				ia->ia_addr.sin_len);
794 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
795 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
796 					ia->ia_dstaddr.sin_len);
797 			}
798 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
799 				ia->ia_sockmask.sin_len);
800 
801 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
802 			    ifp, td));
803 		}
804 	    }
805 	}
806 
807 	return (EOPNOTSUPP);	/*just for safety*/
808 }
809 
810 /*
811  * Delete any existing route for an interface.
812  */
813 void
814 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
815 {
816 
817 	in_scrubprefix(ia);
818 }
819 
820 /*
821  * Initialize an interface's internet address
822  * and routing table entry.
823  */
824 static int
825 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
826     int scrub)
827 {
828 	register u_long i = ntohl(sin->sin_addr.s_addr);
829 	struct sockaddr_in oldaddr;
830 	int s = splimp(), flags = RTF_UP, error = 0;
831 
832 	oldaddr = ia->ia_addr;
833 	if (oldaddr.sin_family == AF_INET)
834 		LIST_REMOVE(ia, ia_hash);
835 	ia->ia_addr = *sin;
836 	if (ia->ia_addr.sin_family == AF_INET) {
837 		IN_IFADDR_WLOCK();
838 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
839 		    ia, ia_hash);
840 		IN_IFADDR_WUNLOCK();
841 	}
842 	/*
843 	 * Give the interface a chance to initialize
844 	 * if this is its first address,
845 	 * and to validate the address if necessary.
846 	 */
847 	if (ifp->if_ioctl != NULL) {
848 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
849 		if (error) {
850 			splx(s);
851 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
852 			ia->ia_addr = oldaddr;
853 			IN_IFADDR_WLOCK();
854 			if (ia->ia_addr.sin_family == AF_INET)
855 				LIST_INSERT_HEAD(INADDR_HASH(
856 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
857 			else
858 				/*
859 				 * If oldaddr family is not AF_INET (e.g.
860 				 * interface has been just created) in_control
861 				 * does not call LIST_REMOVE, and we end up
862 				 * with bogus ia entries in hash
863 				 */
864 				LIST_REMOVE(ia, ia_hash);
865 			IN_IFADDR_WUNLOCK();
866 			return (error);
867 		}
868 	}
869 	splx(s);
870 	if (scrub) {
871 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
872 		in_ifscrub(ifp, ia);
873 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
874 	}
875 	if (IN_CLASSA(i))
876 		ia->ia_netmask = IN_CLASSA_NET;
877 	else if (IN_CLASSB(i))
878 		ia->ia_netmask = IN_CLASSB_NET;
879 	else
880 		ia->ia_netmask = IN_CLASSC_NET;
881 	/*
882 	 * The subnet mask usually includes at least the standard network part,
883 	 * but may may be smaller in the case of supernetting.
884 	 * If it is set, we believe it.
885 	 */
886 	if (ia->ia_subnetmask == 0) {
887 		ia->ia_subnetmask = ia->ia_netmask;
888 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
889 	} else
890 		ia->ia_netmask &= ia->ia_subnetmask;
891 	ia->ia_net = i & ia->ia_netmask;
892 	ia->ia_subnet = i & ia->ia_subnetmask;
893 	in_socktrim(&ia->ia_sockmask);
894 #ifdef DEV_CARP
895 	/*
896 	 * XXX: carp(4) does not have interface route
897 	 */
898 	if (ifp->if_type == IFT_CARP)
899 		return (0);
900 #endif
901 	/*
902 	 * Add route for the network.
903 	 */
904 	ia->ia_ifa.ifa_metric = ifp->if_metric;
905 	if (ifp->if_flags & IFF_BROADCAST) {
906 		ia->ia_broadaddr.sin_addr.s_addr =
907 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
908 		ia->ia_netbroadcast.s_addr =
909 			htonl(ia->ia_net | ~ ia->ia_netmask);
910 	} else if (ifp->if_flags & IFF_LOOPBACK) {
911 		ia->ia_dstaddr = ia->ia_addr;
912 		flags |= RTF_HOST;
913 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
914 		if (ia->ia_dstaddr.sin_family != AF_INET)
915 			return (0);
916 		flags |= RTF_HOST;
917 	}
918 	if ((error = in_addprefix(ia, flags)) != 0)
919 		return (error);
920 
921 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
922 		return (0);
923 
924 	/*
925 	 * add a loopback route to self
926 	 */
927 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK))
928 		error = ifa_add_loopback_route((struct ifaddr *)ia,
929 				       (struct sockaddr *)&ia->ia_addr);
930 
931 	return (error);
932 }
933 
934 #define rtinitflags(x) \
935 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
936 	    ? RTF_HOST : 0)
937 /*
938  * Check if we have a route for the given prefix already or add one accordingly.
939  */
940 static int
941 in_addprefix(struct in_ifaddr *target, int flags)
942 {
943 	struct in_ifaddr *ia;
944 	struct in_addr prefix, mask, p, m;
945 	int error;
946 
947 	if ((flags & RTF_HOST) != 0) {
948 		prefix = target->ia_dstaddr.sin_addr;
949 		mask.s_addr = 0;
950 	} else {
951 		prefix = target->ia_addr.sin_addr;
952 		mask = target->ia_sockmask.sin_addr;
953 		prefix.s_addr &= mask.s_addr;
954 	}
955 
956 	IN_IFADDR_RLOCK();
957 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
958 		if (rtinitflags(ia)) {
959 			p = ia->ia_addr.sin_addr;
960 
961 			if (prefix.s_addr != p.s_addr)
962 				continue;
963 		} else {
964 			p = ia->ia_addr.sin_addr;
965 			m = ia->ia_sockmask.sin_addr;
966 			p.s_addr &= m.s_addr;
967 
968 			if (prefix.s_addr != p.s_addr ||
969 			    mask.s_addr != m.s_addr)
970 				continue;
971 		}
972 
973 		/*
974 		 * If we got a matching prefix route inserted by other
975 		 * interface address, we are done here.
976 		 */
977 		if (ia->ia_flags & IFA_ROUTE) {
978 			if (V_sameprefixcarponly &&
979 			    target->ia_ifp->if_type != IFT_CARP &&
980 			    ia->ia_ifp->if_type != IFT_CARP) {
981 				IN_IFADDR_RUNLOCK();
982 				return (EEXIST);
983 			} else {
984 				struct route pfx_ro;
985 				struct sockaddr_in *pfx_addr;
986 				struct rtentry msg_rt;
987 
988 				/* QL: XXX
989 				 * This is a bit questionable because there is no
990 				 * additional route entry added for an address alias.
991 				 * Therefore this route report is inaccurate. Perhaps
992 				 * it's better to supply a empty rtentry as how it
993 				 * is done in in_scrubprefix().
994 				 */
995 				bzero(&pfx_ro, sizeof(pfx_ro));
996 				pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
997 				pfx_addr->sin_len = sizeof(*pfx_addr);
998 				pfx_addr->sin_family = AF_INET;
999 				pfx_addr->sin_addr = prefix;
1000 				rtalloc_ign_fib(&pfx_ro, 0, 0);
1001 				if (pfx_ro.ro_rt != NULL) {
1002 					msg_rt = *pfx_ro.ro_rt;
1003 					/* QL: XXX
1004 					 * Point the gateway to the given interface
1005 					 * address as if a new prefix route entry has
1006 					 * been added through the new address alias.
1007 					 * All other parts of the rtentry is accurate,
1008 					 * e.g., rt_key, rt_mask, rt_ifp etc.
1009 					 */
1010 					msg_rt.rt_gateway =
1011 						(struct sockaddr *)&ia->ia_addr;
1012 					rt_newaddrmsg(RTM_ADD,
1013 						      (struct ifaddr *)target,
1014 						      0, &msg_rt);
1015 					RTFREE(pfx_ro.ro_rt);
1016 				}
1017 
1018 				IN_IFADDR_RUNLOCK();
1019 				return (0);
1020 			}
1021 		}
1022 	}
1023 	IN_IFADDR_RUNLOCK();
1024 
1025 	/*
1026 	 * No-one seem to have this prefix route, so we try to insert it.
1027 	 */
1028 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1029 	if (!error)
1030 		target->ia_flags |= IFA_ROUTE;
1031 	return (error);
1032 }
1033 
1034 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1035 
1036 /*
1037  * If there is no other address in the system that can serve a route to the
1038  * same prefix, remove the route.  Hand over the route to the new address
1039  * otherwise.
1040  */
1041 static int
1042 in_scrubprefix(struct in_ifaddr *target)
1043 {
1044 	struct in_ifaddr *ia;
1045 	struct in_addr prefix, mask, p;
1046 	int error;
1047 	struct sockaddr_in prefix0, mask0;
1048 
1049 	/*
1050 	 * Remove the loopback route to the interface address.
1051 	 * The "useloopback" setting is not consulted because if the
1052 	 * user configures an interface address, turns off this
1053 	 * setting, and then tries to delete that interface address,
1054 	 * checking the current setting of "useloopback" would leave
1055 	 * that interface address loopback route untouched, which
1056 	 * would be wrong. Therefore the interface address loopback route
1057 	 * deletion is unconditional.
1058 	 */
1059 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1060 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK)) {
1061 		error = ifa_del_loopback_route((struct ifaddr *)target,
1062 				       (struct sockaddr *)&target->ia_addr);
1063 		/* remove arp cache */
1064 		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1065 	}
1066 
1067 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1068 		struct rtentry rt;
1069 
1070 		/* QL: XXX
1071 		 * Report a blank rtentry when a route has not been
1072 		 * installed for the given interface address.
1073 		 */
1074 		bzero(&rt, sizeof(rt));
1075 		rt_newaddrmsg(RTM_DELETE,
1076 			      (struct ifaddr *)target,
1077 			      0, &rt);
1078 		return (0);
1079 	}
1080 
1081 	if (rtinitflags(target))
1082 		prefix = target->ia_dstaddr.sin_addr;
1083 	else {
1084 		prefix = target->ia_addr.sin_addr;
1085 		mask = target->ia_sockmask.sin_addr;
1086 		prefix.s_addr &= mask.s_addr;
1087 	}
1088 
1089 	IN_IFADDR_RLOCK();
1090 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1091 		if (rtinitflags(ia))
1092 			p = ia->ia_dstaddr.sin_addr;
1093 		else {
1094 			p = ia->ia_addr.sin_addr;
1095 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1096 		}
1097 
1098 		if (prefix.s_addr != p.s_addr)
1099 			continue;
1100 
1101 		/*
1102 		 * If we got a matching prefix address, move IFA_ROUTE and
1103 		 * the route itself to it.  Make sure that routing daemons
1104 		 * get a heads-up.
1105 		 *
1106 		 * XXX: a special case for carp(4) interface
1107 		 */
1108 		if ((ia->ia_flags & IFA_ROUTE) == 0
1109 #ifdef DEV_CARP
1110 		    && (ia->ia_ifp->if_type != IFT_CARP)
1111 #endif
1112 							) {
1113 			IN_IFADDR_RUNLOCK();
1114 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1115 			    rtinitflags(target));
1116 			target->ia_flags &= ~IFA_ROUTE;
1117 
1118 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1119 			    rtinitflags(ia) | RTF_UP);
1120 			if (error == 0)
1121 				ia->ia_flags |= IFA_ROUTE;
1122 			return (error);
1123 		}
1124 	}
1125 	IN_IFADDR_RUNLOCK();
1126 
1127 	/*
1128 	 * remove all L2 entries on the given prefix
1129 	 */
1130 	bzero(&prefix0, sizeof(prefix0));
1131 	prefix0.sin_len = sizeof(prefix0);
1132 	prefix0.sin_family = AF_INET;
1133 	prefix0.sin_addr.s_addr = target->ia_subnet;
1134 	bzero(&mask0, sizeof(mask0));
1135 	mask0.sin_len = sizeof(mask0);
1136 	mask0.sin_family = AF_INET;
1137 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1138 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1139 			    (struct sockaddr *)&mask0);
1140 
1141 	/*
1142 	 * As no-one seem to have this prefix, we can remove the route.
1143 	 */
1144 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1145 	target->ia_flags &= ~IFA_ROUTE;
1146 	return (0);
1147 }
1148 
1149 #undef rtinitflags
1150 
1151 /*
1152  * Return 1 if the address might be a local broadcast address.
1153  */
1154 int
1155 in_broadcast(struct in_addr in, struct ifnet *ifp)
1156 {
1157 	register struct ifaddr *ifa;
1158 	u_long t;
1159 
1160 	if (in.s_addr == INADDR_BROADCAST ||
1161 	    in.s_addr == INADDR_ANY)
1162 		return (1);
1163 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1164 		return (0);
1165 	t = ntohl(in.s_addr);
1166 	/*
1167 	 * Look through the list of addresses for a match
1168 	 * with a broadcast address.
1169 	 */
1170 #define ia ((struct in_ifaddr *)ifa)
1171 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1172 		if (ifa->ifa_addr->sa_family == AF_INET &&
1173 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1174 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1175 		     /*
1176 		      * Check for old-style (host 0) broadcast.
1177 		      */
1178 		     t == ia->ia_subnet || t == ia->ia_net) &&
1179 		     /*
1180 		      * Check for an all one subnetmask. These
1181 		      * only exist when an interface gets a secondary
1182 		      * address.
1183 		      */
1184 		     ia->ia_subnetmask != (u_long)0xffffffff)
1185 			    return (1);
1186 	return (0);
1187 #undef ia
1188 }
1189 
1190 /*
1191  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1192  */
1193 void
1194 in_ifdetach(struct ifnet *ifp)
1195 {
1196 
1197 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1198 	in_pcbpurgeif0(&V_udbinfo, ifp);
1199 	in_purgemaddrs(ifp);
1200 }
1201 
1202 /*
1203  * Delete all IPv4 multicast address records, and associated link-layer
1204  * multicast address records, associated with ifp.
1205  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1206  * XXX This should not race with ifma_protospec being set during
1207  * a new allocation, if it does, we have bigger problems.
1208  */
1209 static void
1210 in_purgemaddrs(struct ifnet *ifp)
1211 {
1212 	LIST_HEAD(,in_multi) purgeinms;
1213 	struct in_multi		*inm, *tinm;
1214 	struct ifmultiaddr	*ifma;
1215 
1216 	LIST_INIT(&purgeinms);
1217 	IN_MULTI_LOCK();
1218 
1219 	/*
1220 	 * Extract list of in_multi associated with the detaching ifp
1221 	 * which the PF_INET layer is about to release.
1222 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1223 	 * by code further down.
1224 	 */
1225 	IF_ADDR_LOCK(ifp);
1226 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1227 		if (ifma->ifma_addr->sa_family != AF_INET ||
1228 		    ifma->ifma_protospec == NULL)
1229 			continue;
1230 #if 0
1231 		KASSERT(ifma->ifma_protospec != NULL,
1232 		    ("%s: ifma_protospec is NULL", __func__));
1233 #endif
1234 		inm = (struct in_multi *)ifma->ifma_protospec;
1235 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1236 	}
1237 	IF_ADDR_UNLOCK(ifp);
1238 
1239 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1240 		LIST_REMOVE(inm, inm_link);
1241 		inm_release_locked(inm);
1242 	}
1243 	igmp_ifdetach(ifp);
1244 
1245 	IN_MULTI_UNLOCK();
1246 }
1247 
1248 #include <net/if_dl.h>
1249 #include <netinet/if_ether.h>
1250 
1251 struct in_llentry {
1252 	struct llentry		base;
1253 	struct sockaddr_in	l3_addr4;
1254 };
1255 
1256 static struct llentry *
1257 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1258 {
1259 	struct in_llentry *lle;
1260 
1261 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1262 	if (lle == NULL)		/* NB: caller generates msg */
1263 		return NULL;
1264 
1265 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1266 	/*
1267 	 * For IPv4 this will trigger "arpresolve" to generate
1268 	 * an ARP request.
1269 	 */
1270 	lle->base.la_expire = time_second; /* mark expired */
1271 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1272 	lle->base.lle_refcnt = 1;
1273 	LLE_LOCK_INIT(&lle->base);
1274 	return &lle->base;
1275 }
1276 
1277 /*
1278  * Deletes an address from the address table.
1279  * This function is called by the timer functions
1280  * such as arptimer() and nd6_llinfo_timer(), and
1281  * the caller does the locking.
1282  */
1283 static void
1284 in_lltable_free(struct lltable *llt, struct llentry *lle)
1285 {
1286 	LLE_WUNLOCK(lle);
1287 	LLE_LOCK_DESTROY(lle);
1288 	free(lle, M_LLTABLE);
1289 }
1290 
1291 
1292 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1293 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1294 
1295 static void
1296 in_lltable_prefix_free(struct lltable *llt,
1297 		       const struct sockaddr *prefix,
1298 		       const struct sockaddr *mask)
1299 {
1300 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1301 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1302 	struct llentry *lle, *next;
1303 	register int i;
1304 
1305 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1306 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1307 
1308 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1309 						     pfx, msk)) {
1310 				callout_drain(&lle->la_timer);
1311 				LLE_WLOCK(lle);
1312 				llentry_free(lle);
1313 			}
1314 		}
1315 	}
1316 }
1317 
1318 
1319 static int
1320 in_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
1321 {
1322 	struct rtentry *rt;
1323 
1324 	KASSERT(l3addr->sa_family == AF_INET,
1325 	    ("sin_family %d", l3addr->sa_family));
1326 
1327 	/* XXX rtalloc1 should take a const param */
1328 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1329 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
1330 #ifdef DIAGNOSTIC
1331 		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1332 		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1333 #endif
1334 		if (rt != NULL)
1335 			RTFREE_LOCKED(rt);
1336 		return (EINVAL);
1337 	}
1338 	RTFREE_LOCKED(rt);
1339 	return 0;
1340 }
1341 
1342 /*
1343  * Return NULL if not found or marked for deletion.
1344  * If found return lle read locked.
1345  */
1346 static struct llentry *
1347 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1348 {
1349 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1350 	struct ifnet *ifp = llt->llt_ifp;
1351 	struct llentry *lle;
1352 	struct llentries *lleh;
1353 	u_int hashkey;
1354 
1355 	IF_AFDATA_LOCK_ASSERT(ifp);
1356 	KASSERT(l3addr->sa_family == AF_INET,
1357 	    ("sin_family %d", l3addr->sa_family));
1358 
1359 	hashkey = sin->sin_addr.s_addr;
1360 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1361 	LIST_FOREACH(lle, lleh, lle_next) {
1362 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1363 		if (lle->la_flags & LLE_DELETED)
1364 			continue;
1365 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1366 			break;
1367 	}
1368 	if (lle == NULL) {
1369 #ifdef DIAGNOSTIC
1370 		if (flags & LLE_DELETE)
1371 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1372 #endif
1373 		if (!(flags & LLE_CREATE))
1374 			return (NULL);
1375 		/*
1376 		 * A route that covers the given address must have
1377 		 * been installed 1st because we are doing a resolution,
1378 		 * verify this.
1379 		 */
1380 		if (!(flags & LLE_IFADDR) &&
1381 		    in_lltable_rtcheck(ifp, l3addr) != 0)
1382 			goto done;
1383 
1384 		lle = in_lltable_new(l3addr, flags);
1385 		if (lle == NULL) {
1386 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1387 			goto done;
1388 		}
1389 		lle->la_flags = flags & ~LLE_CREATE;
1390 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1391 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1392 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1393 		}
1394 
1395 		lle->lle_tbl  = llt;
1396 		lle->lle_head = lleh;
1397 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1398 	} else if (flags & LLE_DELETE) {
1399 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1400 			LLE_WLOCK(lle);
1401 			lle->la_flags = LLE_DELETED;
1402 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1403 			LLE_WUNLOCK(lle);
1404 #ifdef DIAGNOSTIC
1405 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1406 #endif
1407 		}
1408 		lle = (void *)-1;
1409 
1410 	}
1411 	if (LLE_IS_VALID(lle)) {
1412 		if (flags & LLE_EXCLUSIVE)
1413 			LLE_WLOCK(lle);
1414 		else
1415 			LLE_RLOCK(lle);
1416 	}
1417 done:
1418 	return (lle);
1419 }
1420 
1421 static int
1422 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1423 {
1424 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1425 	struct ifnet *ifp = llt->llt_ifp;
1426 	struct llentry *lle;
1427 	/* XXX stack use */
1428 	struct {
1429 		struct rt_msghdr	rtm;
1430 		struct sockaddr_inarp	sin;
1431 		struct sockaddr_dl	sdl;
1432 	} arpc;
1433 	int error, i;
1434 
1435 	LLTABLE_LOCK_ASSERT();
1436 
1437 	error = 0;
1438 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1439 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1440 			struct sockaddr_dl *sdl;
1441 
1442 			/* skip deleted entries */
1443 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1444 				continue;
1445 			/* Skip if jailed and not a valid IP of the prison. */
1446 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1447 				continue;
1448 			/*
1449 			 * produce a msg made of:
1450 			 *  struct rt_msghdr;
1451 			 *  struct sockaddr_inarp; (IPv4)
1452 			 *  struct sockaddr_dl;
1453 			 */
1454 			bzero(&arpc, sizeof(arpc));
1455 			arpc.rtm.rtm_msglen = sizeof(arpc);
1456 			arpc.rtm.rtm_version = RTM_VERSION;
1457 			arpc.rtm.rtm_type = RTM_GET;
1458 			arpc.rtm.rtm_flags = RTF_UP;
1459 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1460 			arpc.sin.sin_family = AF_INET;
1461 			arpc.sin.sin_len = sizeof(arpc.sin);
1462 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1463 
1464 			/* publish */
1465 			if (lle->la_flags & LLE_PUB) {
1466 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1467 				/* proxy only */
1468 				if (lle->la_flags & LLE_PROXY)
1469 					arpc.sin.sin_other = SIN_PROXY;
1470 			}
1471 
1472 			sdl = &arpc.sdl;
1473 			sdl->sdl_family = AF_LINK;
1474 			sdl->sdl_len = sizeof(*sdl);
1475 			sdl->sdl_index = ifp->if_index;
1476 			sdl->sdl_type = ifp->if_type;
1477 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1478 				sdl->sdl_alen = ifp->if_addrlen;
1479 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1480 			} else {
1481 				sdl->sdl_alen = 0;
1482 				bzero(LLADDR(sdl), ifp->if_addrlen);
1483 			}
1484 
1485 			arpc.rtm.rtm_rmx.rmx_expire =
1486 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1487 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1488 			if (lle->la_flags & LLE_STATIC)
1489 				arpc.rtm.rtm_flags |= RTF_STATIC;
1490 			arpc.rtm.rtm_index = ifp->if_index;
1491 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1492 			if (error)
1493 				break;
1494 		}
1495 	}
1496 	return error;
1497 #undef SIN
1498 }
1499 
1500 void *
1501 in_domifattach(struct ifnet *ifp)
1502 {
1503 	struct in_ifinfo *ii;
1504 	struct lltable *llt;
1505 
1506 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1507 
1508 	llt = lltable_init(ifp, AF_INET);
1509 	if (llt != NULL) {
1510 		llt->llt_new = in_lltable_new;
1511 		llt->llt_free = in_lltable_free;
1512 		llt->llt_prefix_free = in_lltable_prefix_free;
1513 		llt->llt_rtcheck = in_lltable_rtcheck;
1514 		llt->llt_lookup = in_lltable_lookup;
1515 		llt->llt_dump = in_lltable_dump;
1516 	}
1517 	ii->ii_llt = llt;
1518 
1519 	ii->ii_igmp = igmp_domifattach(ifp);
1520 
1521 	return ii;
1522 }
1523 
1524 void
1525 in_domifdetach(struct ifnet *ifp, void *aux)
1526 {
1527 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1528 
1529 	igmp_domifdetach(ifp);
1530 	lltable_free(ii->ii_llt);
1531 	free(ii, M_IFADDR);
1532 }
1533