xref: /freebsd/sys/netinet/in.c (revision 9517e866259191fcd39434a97ad849a9b59b9b9f)
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 			in_ifscrub(ifp, ia);
540 			ia->ia_sockmask = ifra->ifra_mask;
541 			ia->ia_sockmask.sin_family = AF_INET;
542 			ia->ia_subnetmask =
543 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
544 			maskIsNew = 1;
545 		}
546 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
547 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
548 			in_ifscrub(ifp, ia);
549 			ia->ia_dstaddr = ifra->ifra_dstaddr;
550 			maskIsNew  = 1; /* We lie; but the effect's the same */
551 		}
552 		if (ifra->ifra_addr.sin_family == AF_INET &&
553 		    (hostIsNew || maskIsNew))
554 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
555 		if (error != 0 && iaIsNew)
556 			goto out;
557 
558 		if ((ifp->if_flags & IFF_BROADCAST) &&
559 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
560 			ia->ia_broadaddr = ifra->ifra_broadaddr;
561 		if (error == 0) {
562 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
563 			if (iaIsFirst &&
564 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
565 				error = in_joingroup(ifp, &allhosts_addr,
566 				    NULL, &ii->ii_allhosts);
567 			}
568 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
569 		}
570 		goto out;
571 
572 	case SIOCDIFADDR:
573 		/*
574 		 * in_ifscrub kills the interface route.
575 		 */
576 		in_ifscrub(ifp, ia);
577 
578 		/*
579 		 * in_ifadown gets rid of all the rest of
580 		 * the routes.  This is not quite the right
581 		 * thing to do, but at least if we are running
582 		 * a routing process they will come back.
583 		 */
584 		in_ifadown(&ia->ia_ifa, 1);
585 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
586 		error = 0;
587 		break;
588 
589 	default:
590 		panic("in_control: unsupported ioctl");
591 	}
592 
593 	IF_ADDR_LOCK(ifp);
594 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
595 	IF_ADDR_UNLOCK(ifp);
596 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
597 
598 	IN_IFADDR_WLOCK();
599 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
600 	if (ia->ia_addr.sin_family == AF_INET) {
601 		struct in_ifaddr *if_ia;
602 
603 		LIST_REMOVE(ia, ia_hash);
604 		IN_IFADDR_WUNLOCK();
605 		/*
606 		 * If this is the last IPv4 address configured on this
607 		 * interface, leave the all-hosts group.
608 		 * No state-change report need be transmitted.
609 		 */
610 		if_ia = NULL;
611 		IFP_TO_IA(ifp, if_ia);
612 		if (if_ia == NULL) {
613 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
614 			IN_MULTI_LOCK();
615 			if (ii->ii_allhosts) {
616 				(void)in_leavegroup_locked(ii->ii_allhosts,
617 				    NULL);
618 				ii->ii_allhosts = NULL;
619 			}
620 			IN_MULTI_UNLOCK();
621 		} else
622 			ifa_free(&if_ia->ia_ifa);
623 	} else
624 		IN_IFADDR_WUNLOCK();
625 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
626 out:
627 	if (ia != NULL)
628 		ifa_free(&ia->ia_ifa);
629 	return (error);
630 }
631 
632 /*
633  * SIOC[GAD]LIFADDR.
634  *	SIOCGLIFADDR: get first address. (?!?)
635  *	SIOCGLIFADDR with IFLR_PREFIX:
636  *		get first address that matches the specified prefix.
637  *	SIOCALIFADDR: add the specified address.
638  *	SIOCALIFADDR with IFLR_PREFIX:
639  *		EINVAL since we can't deduce hostid part of the address.
640  *	SIOCDLIFADDR: delete the specified address.
641  *	SIOCDLIFADDR with IFLR_PREFIX:
642  *		delete the first address that matches the specified prefix.
643  * return values:
644  *	EINVAL on invalid parameters
645  *	EADDRNOTAVAIL on prefix match failed/specified address not found
646  *	other values may be returned from in_ioctl()
647  */
648 static int
649 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
650     struct ifnet *ifp, struct thread *td)
651 {
652 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
653 	struct ifaddr *ifa;
654 
655 	/* sanity checks */
656 	if (data == NULL || ifp == NULL) {
657 		panic("invalid argument to in_lifaddr_ioctl");
658 		/*NOTRECHED*/
659 	}
660 
661 	switch (cmd) {
662 	case SIOCGLIFADDR:
663 		/* address must be specified on GET with IFLR_PREFIX */
664 		if ((iflr->flags & IFLR_PREFIX) == 0)
665 			break;
666 		/*FALLTHROUGH*/
667 	case SIOCALIFADDR:
668 	case SIOCDLIFADDR:
669 		/* address must be specified on ADD and DELETE */
670 		if (iflr->addr.ss_family != AF_INET)
671 			return (EINVAL);
672 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
673 			return (EINVAL);
674 		/* XXX need improvement */
675 		if (iflr->dstaddr.ss_family
676 		 && iflr->dstaddr.ss_family != AF_INET)
677 			return (EINVAL);
678 		if (iflr->dstaddr.ss_family
679 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
680 			return (EINVAL);
681 		break;
682 	default: /*shouldn't happen*/
683 		return (EOPNOTSUPP);
684 	}
685 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
686 		return (EINVAL);
687 
688 	switch (cmd) {
689 	case SIOCALIFADDR:
690 	    {
691 		struct in_aliasreq ifra;
692 
693 		if (iflr->flags & IFLR_PREFIX)
694 			return (EINVAL);
695 
696 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
697 		bzero(&ifra, sizeof(ifra));
698 		bcopy(iflr->iflr_name, ifra.ifra_name,
699 			sizeof(ifra.ifra_name));
700 
701 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
702 
703 		if (iflr->dstaddr.ss_family) {	/*XXX*/
704 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
705 				iflr->dstaddr.ss_len);
706 		}
707 
708 		ifra.ifra_mask.sin_family = AF_INET;
709 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
710 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
711 
712 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
713 	    }
714 	case SIOCGLIFADDR:
715 	case SIOCDLIFADDR:
716 	    {
717 		struct in_ifaddr *ia;
718 		struct in_addr mask, candidate, match;
719 		struct sockaddr_in *sin;
720 
721 		bzero(&mask, sizeof(mask));
722 		bzero(&match, sizeof(match));
723 		if (iflr->flags & IFLR_PREFIX) {
724 			/* lookup a prefix rather than address. */
725 			in_len2mask(&mask, iflr->prefixlen);
726 
727 			sin = (struct sockaddr_in *)&iflr->addr;
728 			match.s_addr = sin->sin_addr.s_addr;
729 			match.s_addr &= mask.s_addr;
730 
731 			/* if you set extra bits, that's wrong */
732 			if (match.s_addr != sin->sin_addr.s_addr)
733 				return (EINVAL);
734 
735 		} else {
736 			/* on getting an address, take the 1st match */
737 			/* on deleting an address, do exact match */
738 			if (cmd != SIOCGLIFADDR) {
739 				in_len2mask(&mask, 32);
740 				sin = (struct sockaddr_in *)&iflr->addr;
741 				match.s_addr = sin->sin_addr.s_addr;
742 			}
743 		}
744 
745 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
746 			if (ifa->ifa_addr->sa_family != AF_INET6)
747 				continue;
748 			if (match.s_addr == 0)
749 				break;
750 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
751 			candidate.s_addr &= mask.s_addr;
752 			if (candidate.s_addr == match.s_addr)
753 				break;
754 		}
755 		if (ifa == NULL)
756 			return (EADDRNOTAVAIL);
757 		ia = (struct in_ifaddr *)ifa;
758 
759 		if (cmd == SIOCGLIFADDR) {
760 			/* fill in the if_laddrreq structure */
761 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
762 
763 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
764 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
765 					ia->ia_dstaddr.sin_len);
766 			} else
767 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
768 
769 			iflr->prefixlen =
770 				in_mask2len(&ia->ia_sockmask.sin_addr);
771 
772 			iflr->flags = 0;	/*XXX*/
773 
774 			return (0);
775 		} else {
776 			struct in_aliasreq ifra;
777 
778 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
779 			bzero(&ifra, sizeof(ifra));
780 			bcopy(iflr->iflr_name, ifra.ifra_name,
781 				sizeof(ifra.ifra_name));
782 
783 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
784 				ia->ia_addr.sin_len);
785 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
786 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
787 					ia->ia_dstaddr.sin_len);
788 			}
789 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
790 				ia->ia_sockmask.sin_len);
791 
792 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
793 			    ifp, td));
794 		}
795 	    }
796 	}
797 
798 	return (EOPNOTSUPP);	/*just for safety*/
799 }
800 
801 /*
802  * Delete any existing route for an interface.
803  */
804 void
805 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
806 {
807 
808 	in_scrubprefix(ia);
809 }
810 
811 /*
812  * Initialize an interface's internet address
813  * and routing table entry.
814  */
815 static int
816 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
817     int scrub)
818 {
819 	register u_long i = ntohl(sin->sin_addr.s_addr);
820 	struct sockaddr_in oldaddr;
821 	struct rtentry *rt = NULL;
822 	struct rt_addrinfo info;
823 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
824 	int s = splimp(), flags = RTF_UP, error = 0;
825 
826 	oldaddr = ia->ia_addr;
827 	if (oldaddr.sin_family == AF_INET)
828 		LIST_REMOVE(ia, ia_hash);
829 	ia->ia_addr = *sin;
830 	if (ia->ia_addr.sin_family == AF_INET) {
831 		IN_IFADDR_WLOCK();
832 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
833 		    ia, ia_hash);
834 		IN_IFADDR_WUNLOCK();
835 	}
836 	/*
837 	 * Give the interface a chance to initialize
838 	 * if this is its first address,
839 	 * and to validate the address if necessary.
840 	 */
841 	if (ifp->if_ioctl != NULL) {
842 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
843 		if (error) {
844 			splx(s);
845 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
846 			ia->ia_addr = oldaddr;
847 			IN_IFADDR_WLOCK();
848 			if (ia->ia_addr.sin_family == AF_INET)
849 				LIST_INSERT_HEAD(INADDR_HASH(
850 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
851 			else
852 				/*
853 				 * If oldaddr family is not AF_INET (e.g.
854 				 * interface has been just created) in_control
855 				 * does not call LIST_REMOVE, and we end up
856 				 * with bogus ia entries in hash
857 				 */
858 				LIST_REMOVE(ia, ia_hash);
859 			IN_IFADDR_WUNLOCK();
860 			return (error);
861 		}
862 	}
863 	splx(s);
864 	if (scrub) {
865 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
866 		in_ifscrub(ifp, ia);
867 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
868 	}
869 	if (IN_CLASSA(i))
870 		ia->ia_netmask = IN_CLASSA_NET;
871 	else if (IN_CLASSB(i))
872 		ia->ia_netmask = IN_CLASSB_NET;
873 	else
874 		ia->ia_netmask = IN_CLASSC_NET;
875 	/*
876 	 * The subnet mask usually includes at least the standard network part,
877 	 * but may may be smaller in the case of supernetting.
878 	 * If it is set, we believe it.
879 	 */
880 	if (ia->ia_subnetmask == 0) {
881 		ia->ia_subnetmask = ia->ia_netmask;
882 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
883 	} else
884 		ia->ia_netmask &= ia->ia_subnetmask;
885 	ia->ia_net = i & ia->ia_netmask;
886 	ia->ia_subnet = i & ia->ia_subnetmask;
887 	in_socktrim(&ia->ia_sockmask);
888 #ifdef DEV_CARP
889 	/*
890 	 * XXX: carp(4) does not have interface route
891 	 */
892 	if (ifp->if_type == IFT_CARP)
893 		return (0);
894 #endif
895 	/*
896 	 * Add route for the network.
897 	 */
898 	ia->ia_ifa.ifa_metric = ifp->if_metric;
899 	if (ifp->if_flags & IFF_BROADCAST) {
900 		ia->ia_broadaddr.sin_addr.s_addr =
901 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
902 		ia->ia_netbroadcast.s_addr =
903 			htonl(ia->ia_net | ~ ia->ia_netmask);
904 	} else if (ifp->if_flags & IFF_LOOPBACK) {
905 		ia->ia_dstaddr = ia->ia_addr;
906 		flags |= RTF_HOST;
907 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
908 		if (ia->ia_dstaddr.sin_family != AF_INET)
909 			return (0);
910 		flags |= RTF_HOST;
911 	}
912 	if ((error = in_addprefix(ia, flags)) != 0)
913 		return (error);
914 
915 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
916 		return (0);
917 
918 	/*
919 	 * add a loopback route to self
920 	 */
921 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
922 		bzero(&info, sizeof(info));
923 		info.rti_ifp = V_loif;
924 		info.rti_flags = ia->ia_flags | RTF_HOST | RTF_STATIC;
925 		info.rti_info[RTAX_DST] = (struct sockaddr *)&ia->ia_addr;
926 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
927 		error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
928 
929 		if (error == 0 && rt != NULL) {
930 			RT_LOCK(rt);
931 			((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
932 				rt->rt_ifp->if_type;
933 			((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
934 				rt->rt_ifp->if_index;
935 			RT_REMREF(rt);
936 			RT_UNLOCK(rt);
937 		} else if (error != 0)
938 			log(LOG_INFO, "in_ifinit: insertion failed\n");
939 	}
940 
941 	return (error);
942 }
943 
944 #define rtinitflags(x) \
945 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
946 	    ? RTF_HOST : 0)
947 /*
948  * Check if we have a route for the given prefix already or add one accordingly.
949  */
950 static int
951 in_addprefix(struct in_ifaddr *target, int flags)
952 {
953 	struct in_ifaddr *ia;
954 	struct in_addr prefix, mask, p, m;
955 	int error;
956 
957 	if ((flags & RTF_HOST) != 0) {
958 		prefix = target->ia_dstaddr.sin_addr;
959 		mask.s_addr = 0;
960 	} else {
961 		prefix = target->ia_addr.sin_addr;
962 		mask = target->ia_sockmask.sin_addr;
963 		prefix.s_addr &= mask.s_addr;
964 	}
965 
966 	IN_IFADDR_RLOCK();
967 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
968 		if (rtinitflags(ia)) {
969 			p = ia->ia_addr.sin_addr;
970 
971 			if (prefix.s_addr != p.s_addr)
972 				continue;
973 		} else {
974 			p = ia->ia_addr.sin_addr;
975 			m = ia->ia_sockmask.sin_addr;
976 			p.s_addr &= m.s_addr;
977 
978 			if (prefix.s_addr != p.s_addr ||
979 			    mask.s_addr != m.s_addr)
980 				continue;
981 		}
982 
983 		/*
984 		 * If we got a matching prefix route inserted by other
985 		 * interface address, we are done here.
986 		 */
987 		if (ia->ia_flags & IFA_ROUTE) {
988 			if (V_sameprefixcarponly &&
989 			    target->ia_ifp->if_type != IFT_CARP &&
990 			    ia->ia_ifp->if_type != IFT_CARP) {
991 				IN_IFADDR_RUNLOCK();
992 				return (EEXIST);
993 			} else {
994 				IN_IFADDR_RUNLOCK();
995 				return (0);
996 			}
997 		}
998 	}
999 	IN_IFADDR_RUNLOCK();
1000 
1001 	/*
1002 	 * No-one seem to have this prefix route, so we try to insert it.
1003 	 */
1004 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1005 	if (!error)
1006 		target->ia_flags |= IFA_ROUTE;
1007 	return (error);
1008 }
1009 
1010 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1011 
1012 /*
1013  * If there is no other address in the system that can serve a route to the
1014  * same prefix, remove the route.  Hand over the route to the new address
1015  * otherwise.
1016  */
1017 static int
1018 in_scrubprefix(struct in_ifaddr *target)
1019 {
1020 	struct in_ifaddr *ia;
1021 	struct in_addr prefix, mask, p;
1022 	int error;
1023 	struct sockaddr_in prefix0, mask0;
1024 	struct rt_addrinfo info;
1025 	struct sockaddr_dl null_sdl;
1026 
1027 	if ((target->ia_flags & IFA_ROUTE) == 0)
1028 		return (0);
1029 
1030 	/*
1031 	 * Remove the loopback route to the interface address.
1032 	 * The "useloopback" setting is not consulted because if the
1033 	 * user configures an interface address, turns off this
1034 	 * setting, and then tries to delete that interface address,
1035 	 * checking the current setting of "useloopback" would leave
1036 	 * that interface address loopback route untouched, which
1037 	 * would be wrong. Therefore the interface address loopback route
1038 	 * deletion is unconditional.
1039 	 */
1040 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1041 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK)) {
1042 		bzero(&null_sdl, sizeof(null_sdl));
1043 		null_sdl.sdl_len = sizeof(null_sdl);
1044 		null_sdl.sdl_family = AF_LINK;
1045 		null_sdl.sdl_type = V_loif->if_type;
1046 		null_sdl.sdl_index = V_loif->if_index;
1047 		bzero(&info, sizeof(info));
1048 		info.rti_flags = target->ia_flags | RTF_HOST | RTF_STATIC;
1049 		info.rti_info[RTAX_DST] = (struct sockaddr *)&target->ia_addr;
1050 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1051 		error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1052 
1053 		if (error != 0)
1054 			log(LOG_INFO, "in_scrubprefix: deletion failed\n");
1055 	}
1056 
1057 	if (rtinitflags(target))
1058 		prefix = target->ia_dstaddr.sin_addr;
1059 	else {
1060 		prefix = target->ia_addr.sin_addr;
1061 		mask = target->ia_sockmask.sin_addr;
1062 		prefix.s_addr &= mask.s_addr;
1063 		/* remove arp cache */
1064 		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1065 	}
1066 
1067 	IN_IFADDR_RLOCK();
1068 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1069 		if (rtinitflags(ia))
1070 			p = ia->ia_dstaddr.sin_addr;
1071 		else {
1072 			p = ia->ia_addr.sin_addr;
1073 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1074 		}
1075 
1076 		if (prefix.s_addr != p.s_addr)
1077 			continue;
1078 
1079 		/*
1080 		 * If we got a matching prefix address, move IFA_ROUTE and
1081 		 * the route itself to it.  Make sure that routing daemons
1082 		 * get a heads-up.
1083 		 *
1084 		 * XXX: a special case for carp(4) interface
1085 		 */
1086 		if ((ia->ia_flags & IFA_ROUTE) == 0
1087 #ifdef DEV_CARP
1088 		    && (ia->ia_ifp->if_type != IFT_CARP)
1089 #endif
1090 							) {
1091 			IN_IFADDR_RUNLOCK();
1092 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1093 			    rtinitflags(target));
1094 			target->ia_flags &= ~IFA_ROUTE;
1095 
1096 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1097 			    rtinitflags(ia) | RTF_UP);
1098 			if (error == 0)
1099 				ia->ia_flags |= IFA_ROUTE;
1100 			return (error);
1101 		}
1102 	}
1103 	IN_IFADDR_RUNLOCK();
1104 
1105 	/*
1106 	 * remove all L2 entries on the given prefix
1107 	 */
1108 	bzero(&prefix0, sizeof(prefix0));
1109 	prefix0.sin_len = sizeof(prefix0);
1110 	prefix0.sin_family = AF_INET;
1111 	prefix0.sin_addr.s_addr = target->ia_subnet;
1112 	bzero(&mask0, sizeof(mask0));
1113 	mask0.sin_len = sizeof(mask0);
1114 	mask0.sin_family = AF_INET;
1115 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1116 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1117 			    (struct sockaddr *)&mask0);
1118 
1119 	/*
1120 	 * As no-one seem to have this prefix, we can remove the route.
1121 	 */
1122 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1123 	target->ia_flags &= ~IFA_ROUTE;
1124 	return (0);
1125 }
1126 
1127 #undef rtinitflags
1128 
1129 /*
1130  * Return 1 if the address might be a local broadcast address.
1131  */
1132 int
1133 in_broadcast(struct in_addr in, struct ifnet *ifp)
1134 {
1135 	register struct ifaddr *ifa;
1136 	u_long t;
1137 
1138 	if (in.s_addr == INADDR_BROADCAST ||
1139 	    in.s_addr == INADDR_ANY)
1140 		return (1);
1141 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1142 		return (0);
1143 	t = ntohl(in.s_addr);
1144 	/*
1145 	 * Look through the list of addresses for a match
1146 	 * with a broadcast address.
1147 	 */
1148 #define ia ((struct in_ifaddr *)ifa)
1149 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1150 		if (ifa->ifa_addr->sa_family == AF_INET &&
1151 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1152 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1153 		     /*
1154 		      * Check for old-style (host 0) broadcast.
1155 		      */
1156 		     t == ia->ia_subnet || t == ia->ia_net) &&
1157 		     /*
1158 		      * Check for an all one subnetmask. These
1159 		      * only exist when an interface gets a secondary
1160 		      * address.
1161 		      */
1162 		     ia->ia_subnetmask != (u_long)0xffffffff)
1163 			    return (1);
1164 	return (0);
1165 #undef ia
1166 }
1167 
1168 /*
1169  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1170  */
1171 void
1172 in_ifdetach(struct ifnet *ifp)
1173 {
1174 
1175 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1176 	in_pcbpurgeif0(&V_udbinfo, ifp);
1177 	in_purgemaddrs(ifp);
1178 }
1179 
1180 /*
1181  * Delete all IPv4 multicast address records, and associated link-layer
1182  * multicast address records, associated with ifp.
1183  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1184  * XXX This should not race with ifma_protospec being set during
1185  * a new allocation, if it does, we have bigger problems.
1186  */
1187 static void
1188 in_purgemaddrs(struct ifnet *ifp)
1189 {
1190 	LIST_HEAD(,in_multi) purgeinms;
1191 	struct in_multi		*inm, *tinm;
1192 	struct ifmultiaddr	*ifma;
1193 
1194 	LIST_INIT(&purgeinms);
1195 	IN_MULTI_LOCK();
1196 
1197 	/*
1198 	 * Extract list of in_multi associated with the detaching ifp
1199 	 * which the PF_INET layer is about to release.
1200 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1201 	 * by code further down.
1202 	 */
1203 	IF_ADDR_LOCK(ifp);
1204 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1205 		if (ifma->ifma_addr->sa_family != AF_INET ||
1206 		    ifma->ifma_protospec == NULL)
1207 			continue;
1208 #if 0
1209 		KASSERT(ifma->ifma_protospec != NULL,
1210 		    ("%s: ifma_protospec is NULL", __func__));
1211 #endif
1212 		inm = (struct in_multi *)ifma->ifma_protospec;
1213 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1214 	}
1215 	IF_ADDR_UNLOCK(ifp);
1216 
1217 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1218 		LIST_REMOVE(inm, inm_link);
1219 		inm_release_locked(inm);
1220 	}
1221 	igmp_ifdetach(ifp);
1222 
1223 	IN_MULTI_UNLOCK();
1224 }
1225 
1226 #include <net/if_dl.h>
1227 #include <netinet/if_ether.h>
1228 
1229 struct in_llentry {
1230 	struct llentry		base;
1231 	struct sockaddr_in	l3_addr4;
1232 };
1233 
1234 static struct llentry *
1235 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1236 {
1237 	struct in_llentry *lle;
1238 
1239 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1240 	if (lle == NULL)		/* NB: caller generates msg */
1241 		return NULL;
1242 
1243 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1244 	/*
1245 	 * For IPv4 this will trigger "arpresolve" to generate
1246 	 * an ARP request.
1247 	 */
1248 	lle->base.la_expire = time_second; /* mark expired */
1249 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1250 	lle->base.lle_refcnt = 1;
1251 	LLE_LOCK_INIT(&lle->base);
1252 	return &lle->base;
1253 }
1254 
1255 /*
1256  * Deletes an address from the address table.
1257  * This function is called by the timer functions
1258  * such as arptimer() and nd6_llinfo_timer(), and
1259  * the caller does the locking.
1260  */
1261 static void
1262 in_lltable_free(struct lltable *llt, struct llentry *lle)
1263 {
1264 	LLE_WUNLOCK(lle);
1265 	LLE_LOCK_DESTROY(lle);
1266 	free(lle, M_LLTABLE);
1267 }
1268 
1269 
1270 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1271 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1272 
1273 static void
1274 in_lltable_prefix_free(struct lltable *llt,
1275 		       const struct sockaddr *prefix,
1276 		       const struct sockaddr *mask)
1277 {
1278 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1279 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1280 	struct llentry *lle, *next;
1281 	register int i;
1282 
1283 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1284 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1285 
1286 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1287 						     pfx, msk)) {
1288 				callout_drain(&lle->la_timer);
1289 				LLE_WLOCK(lle);
1290 				llentry_free(lle);
1291 			}
1292 		}
1293 	}
1294 }
1295 
1296 
1297 static int
1298 in_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
1299 {
1300 	struct rtentry *rt;
1301 
1302 	KASSERT(l3addr->sa_family == AF_INET,
1303 	    ("sin_family %d", l3addr->sa_family));
1304 
1305 	/* XXX rtalloc1 should take a const param */
1306 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1307 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
1308 		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1309 		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1310 		if (rt != NULL)
1311 			RTFREE_LOCKED(rt);
1312 		return (EINVAL);
1313 	}
1314 	RTFREE_LOCKED(rt);
1315 	return 0;
1316 }
1317 
1318 /*
1319  * Return NULL if not found or marked for deletion.
1320  * If found return lle read locked.
1321  */
1322 static struct llentry *
1323 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1324 {
1325 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1326 	struct ifnet *ifp = llt->llt_ifp;
1327 	struct llentry *lle;
1328 	struct llentries *lleh;
1329 	u_int hashkey;
1330 
1331 	IF_AFDATA_LOCK_ASSERT(ifp);
1332 	KASSERT(l3addr->sa_family == AF_INET,
1333 	    ("sin_family %d", l3addr->sa_family));
1334 
1335 	hashkey = sin->sin_addr.s_addr;
1336 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1337 	LIST_FOREACH(lle, lleh, lle_next) {
1338 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1339 		if (lle->la_flags & LLE_DELETED)
1340 			continue;
1341 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1342 			break;
1343 	}
1344 	if (lle == NULL) {
1345 #ifdef DIAGNOSTICS
1346 		if (flags & LLE_DELETE)
1347 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1348 #endif
1349 		if (!(flags & LLE_CREATE))
1350 			return (NULL);
1351 		/*
1352 		 * A route that covers the given address must have
1353 		 * been installed 1st because we are doing a resolution,
1354 		 * verify this.
1355 		 */
1356 		if (!(flags & LLE_IFADDR) &&
1357 		    in_lltable_rtcheck(ifp, l3addr) != 0)
1358 			goto done;
1359 
1360 		lle = in_lltable_new(l3addr, flags);
1361 		if (lle == NULL) {
1362 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1363 			goto done;
1364 		}
1365 		lle->la_flags = flags & ~LLE_CREATE;
1366 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1367 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1368 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1369 		}
1370 
1371 		lle->lle_tbl  = llt;
1372 		lle->lle_head = lleh;
1373 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1374 	} else if (flags & LLE_DELETE) {
1375 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1376 			LLE_WLOCK(lle);
1377 			lle->la_flags = LLE_DELETED;
1378 			LLE_WUNLOCK(lle);
1379 #ifdef DIAGNOSTICS
1380 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1381 #endif
1382 		}
1383 		lle = (void *)-1;
1384 
1385 	}
1386 	if (LLE_IS_VALID(lle)) {
1387 		if (flags & LLE_EXCLUSIVE)
1388 			LLE_WLOCK(lle);
1389 		else
1390 			LLE_RLOCK(lle);
1391 	}
1392 done:
1393 	return (lle);
1394 }
1395 
1396 static int
1397 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1398 {
1399 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1400 	struct ifnet *ifp = llt->llt_ifp;
1401 	struct llentry *lle;
1402 	/* XXX stack use */
1403 	struct {
1404 		struct rt_msghdr	rtm;
1405 		struct sockaddr_inarp	sin;
1406 		struct sockaddr_dl	sdl;
1407 	} arpc;
1408 	int error, i;
1409 
1410 	/* XXXXX
1411 	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
1412 	 * so it is okay to use this ASSERT, change it when
1413 	 * IFNET lock is finalized
1414 	 */
1415 	IFNET_WLOCK_ASSERT();
1416 
1417 	error = 0;
1418 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1419 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1420 			struct sockaddr_dl *sdl;
1421 
1422 			/* skip deleted entries */
1423 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
1424 				continue;
1425 			/* Skip if jailed and not a valid IP of the prison. */
1426 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1427 				continue;
1428 			/*
1429 			 * produce a msg made of:
1430 			 *  struct rt_msghdr;
1431 			 *  struct sockaddr_inarp; (IPv4)
1432 			 *  struct sockaddr_dl;
1433 			 */
1434 			bzero(&arpc, sizeof(arpc));
1435 			arpc.rtm.rtm_msglen = sizeof(arpc);
1436 			arpc.rtm.rtm_version = RTM_VERSION;
1437 			arpc.rtm.rtm_type = RTM_GET;
1438 			arpc.rtm.rtm_flags = RTF_UP;
1439 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1440 			arpc.sin.sin_family = AF_INET;
1441 			arpc.sin.sin_len = sizeof(arpc.sin);
1442 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1443 
1444 			/* publish */
1445 			if (lle->la_flags & LLE_PUB) {
1446 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1447 				/* proxy only */
1448 				if (lle->la_flags & LLE_PROXY)
1449 					arpc.sin.sin_other = SIN_PROXY;
1450 			}
1451 
1452 			sdl = &arpc.sdl;
1453 			sdl->sdl_family = AF_LINK;
1454 			sdl->sdl_len = sizeof(*sdl);
1455 			sdl->sdl_alen = ifp->if_addrlen;
1456 			sdl->sdl_index = ifp->if_index;
1457 			sdl->sdl_type = ifp->if_type;
1458 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
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 			arpc.rtm.rtm_index = ifp->if_index;
1466 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1467 			if (error)
1468 				break;
1469 		}
1470 	}
1471 	return error;
1472 #undef SIN
1473 }
1474 
1475 void *
1476 in_domifattach(struct ifnet *ifp)
1477 {
1478 	struct in_ifinfo *ii;
1479 	struct lltable *llt;
1480 
1481 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1482 
1483 	llt = lltable_init(ifp, AF_INET);
1484 	if (llt != NULL) {
1485 		llt->llt_new = in_lltable_new;
1486 		llt->llt_free = in_lltable_free;
1487 		llt->llt_prefix_free = in_lltable_prefix_free;
1488 		llt->llt_rtcheck = in_lltable_rtcheck;
1489 		llt->llt_lookup = in_lltable_lookup;
1490 		llt->llt_dump = in_lltable_dump;
1491 	}
1492 	ii->ii_llt = llt;
1493 
1494 	ii->ii_igmp = igmp_domifattach(ifp);
1495 
1496 	return ii;
1497 }
1498 
1499 void
1500 in_domifdetach(struct ifnet *ifp, void *aux)
1501 {
1502 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1503 
1504 	igmp_domifdetach(ifp);
1505 	lltable_free(ii->ii_llt);
1506 	free(ii, M_IFADDR);
1507 }
1508