xref: /freebsd/sys/netinet/in.c (revision 6356dba0b403daa023dec24559ab1f8e602e4f14)
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/kernel.h>
45 #include <sys/sysctl.h>
46 #include <sys/vimage.h>
47 
48 #include <net/if.h>
49 #include <net/if_types.h>
50 #include <net/route.h>
51 
52 #include <netinet/in.h>
53 #include <netinet/in_var.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/ip_var.h>
56 
57 static int in_mask2len(struct in_addr *);
58 static void in_len2mask(struct in_addr *, int);
59 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
60 	struct ifnet *, struct thread *);
61 
62 static int	in_addprefix(struct in_ifaddr *, int);
63 static int	in_scrubprefix(struct in_ifaddr *);
64 static void	in_socktrim(struct sockaddr_in *);
65 static int	in_ifinit(struct ifnet *,
66 	    struct in_ifaddr *, struct sockaddr_in *, int);
67 static void	in_purgemaddrs(struct ifnet *);
68 
69 static int subnetsarelocal = 0;
70 SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
71 	&subnetsarelocal, 0, "Treat all subnets as directly connected");
72 static int sameprefixcarponly = 0;
73 SYSCTL_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
74 	&sameprefixcarponly, 0,
75 	"Refuse to create same prefixes on different interfaces");
76 
77 extern struct inpcbinfo ripcbinfo;
78 extern struct inpcbinfo udbinfo;
79 
80 /*
81  * Return 1 if an internet address is for a ``local'' host
82  * (one to which we have a connection).  If subnetsarelocal
83  * is true, this includes other subnets of the local net.
84  * Otherwise, it includes only the directly-connected (sub)nets.
85  */
86 int
87 in_localaddr(struct in_addr in)
88 {
89 	register u_long i = ntohl(in.s_addr);
90 	register struct in_ifaddr *ia;
91 
92 	if (V_subnetsarelocal) {
93 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
94 			if ((i & ia->ia_netmask) == ia->ia_net)
95 				return (1);
96 	} else {
97 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
98 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
99 				return (1);
100 	}
101 	return (0);
102 }
103 
104 /*
105  * Return 1 if an internet address is for the local host and configured
106  * on one of its interfaces.
107  */
108 int
109 in_localip(struct in_addr in)
110 {
111 	struct in_ifaddr *ia;
112 
113 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
114 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
115 			return 1;
116 	}
117 	return 0;
118 }
119 
120 /*
121  * Determine whether an IP address is in a reserved set of addresses
122  * that may not be forwarded, or whether datagrams to that destination
123  * may be forwarded.
124  */
125 int
126 in_canforward(struct in_addr in)
127 {
128 	register u_long i = ntohl(in.s_addr);
129 	register u_long net;
130 
131 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
132 		return (0);
133 	if (IN_CLASSA(i)) {
134 		net = i & IN_CLASSA_NET;
135 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
136 			return (0);
137 	}
138 	return (1);
139 }
140 
141 /*
142  * Trim a mask in a sockaddr
143  */
144 static void
145 in_socktrim(struct sockaddr_in *ap)
146 {
147     register char *cplim = (char *) &ap->sin_addr;
148     register char *cp = (char *) (&ap->sin_addr + 1);
149 
150     ap->sin_len = 0;
151     while (--cp >= cplim)
152 	if (*cp) {
153 	    (ap)->sin_len = cp - (char *) (ap) + 1;
154 	    break;
155 	}
156 }
157 
158 static int
159 in_mask2len(mask)
160 	struct in_addr *mask;
161 {
162 	int x, y;
163 	u_char *p;
164 
165 	p = (u_char *)mask;
166 	for (x = 0; x < sizeof(*mask); x++) {
167 		if (p[x] != 0xff)
168 			break;
169 	}
170 	y = 0;
171 	if (x < sizeof(*mask)) {
172 		for (y = 0; y < 8; y++) {
173 			if ((p[x] & (0x80 >> y)) == 0)
174 				break;
175 		}
176 	}
177 	return x * 8 + y;
178 }
179 
180 static void
181 in_len2mask(struct in_addr *mask, int len)
182 {
183 	int i;
184 	u_char *p;
185 
186 	p = (u_char *)mask;
187 	bzero(mask, sizeof(*mask));
188 	for (i = 0; i < len / 8; i++)
189 		p[i] = 0xff;
190 	if (len % 8)
191 		p[i] = (0xff00 >> (len % 8)) & 0xff;
192 }
193 
194 /*
195  * Generic internet control operations (ioctl's).
196  * Ifp is 0 if not an interface-specific ioctl.
197  */
198 /* ARGSUSED */
199 int
200 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
201     struct thread *td)
202 {
203 	register struct ifreq *ifr = (struct ifreq *)data;
204 	register struct in_ifaddr *ia = 0, *iap;
205 	register struct ifaddr *ifa;
206 	struct in_addr allhosts_addr;
207 	struct in_addr dst;
208 	struct in_ifaddr *oia;
209 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
210 	struct sockaddr_in oldaddr;
211 	int error, hostIsNew, iaIsNew, maskIsNew, s;
212 	int iaIsFirst;
213 
214 	iaIsFirst = 0;
215 	iaIsNew = 0;
216 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
217 
218 	switch (cmd) {
219 	case SIOCALIFADDR:
220 		if (td != NULL) {
221 			error = priv_check(td, PRIV_NET_ADDIFADDR);
222 			if (error)
223 				return (error);
224 		}
225 		if (!ifp)
226 			return EINVAL;
227 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
228 
229 	case SIOCDLIFADDR:
230 		if (td != NULL) {
231 			error = priv_check(td, PRIV_NET_DELIFADDR);
232 			if (error)
233 				return (error);
234 		}
235 		if (!ifp)
236 			return EINVAL;
237 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
238 
239 	case SIOCGLIFADDR:
240 		if (!ifp)
241 			return EINVAL;
242 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
243 	}
244 
245 	/*
246 	 * Find address for this interface, if it exists.
247 	 *
248 	 * If an alias address was specified, find that one instead of
249 	 * the first one on the interface, if possible.
250 	 */
251 	if (ifp) {
252 		dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
253 		LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash)
254 			if (iap->ia_ifp == ifp &&
255 			    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
256 				ia = iap;
257 				break;
258 			}
259 		if (ia == NULL)
260 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
261 				iap = ifatoia(ifa);
262 				if (iap->ia_addr.sin_family == AF_INET) {
263 					ia = iap;
264 					break;
265 				}
266 			}
267 		if (ia == NULL)
268 			iaIsFirst = 1;
269 	}
270 
271 	switch (cmd) {
272 
273 	case SIOCAIFADDR:
274 	case SIOCDIFADDR:
275 		if (ifp == 0)
276 			return (EADDRNOTAVAIL);
277 		if (ifra->ifra_addr.sin_family == AF_INET) {
278 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
279 				if (ia->ia_ifp == ifp  &&
280 				    ia->ia_addr.sin_addr.s_addr ==
281 				    ifra->ifra_addr.sin_addr.s_addr)
282 					break;
283 			}
284 			if ((ifp->if_flags & IFF_POINTOPOINT)
285 			    && (cmd == SIOCAIFADDR)
286 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
287 				== INADDR_ANY)) {
288 				return EDESTADDRREQ;
289 			}
290 		}
291 		if (cmd == SIOCDIFADDR && ia == 0)
292 			return (EADDRNOTAVAIL);
293 		/* FALLTHROUGH */
294 	case SIOCSIFADDR:
295 	case SIOCSIFNETMASK:
296 	case SIOCSIFDSTADDR:
297 		if (td != NULL) {
298 			error = priv_check(td, (cmd == SIOCDIFADDR) ?
299 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
300 			if (error)
301 				return (error);
302 		}
303 
304 		if (ifp == 0)
305 			return (EADDRNOTAVAIL);
306 		if (ia == (struct in_ifaddr *)0) {
307 			ia = (struct in_ifaddr *)
308 				malloc(sizeof *ia, M_IFADDR, M_WAITOK | M_ZERO);
309 			if (ia == (struct in_ifaddr *)NULL)
310 				return (ENOBUFS);
311 			/*
312 			 * Protect from ipintr() traversing address list
313 			 * while we're modifying it.
314 			 */
315 			s = splnet();
316 			ifa = &ia->ia_ifa;
317 			IFA_LOCK_INIT(ifa);
318 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
319 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
320 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
321 			ifa->ifa_refcnt = 1;
322 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
323 
324 			ia->ia_sockmask.sin_len = 8;
325 			ia->ia_sockmask.sin_family = AF_INET;
326 			if (ifp->if_flags & IFF_BROADCAST) {
327 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
328 				ia->ia_broadaddr.sin_family = AF_INET;
329 			}
330 			ia->ia_ifp = ifp;
331 
332 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
333 			splx(s);
334 			iaIsNew = 1;
335 		}
336 		break;
337 
338 	case SIOCSIFBRDADDR:
339 		if (td != NULL) {
340 			error = priv_check(td, PRIV_NET_ADDIFADDR);
341 			if (error)
342 				return (error);
343 		}
344 		/* FALLTHROUGH */
345 
346 	case SIOCGIFADDR:
347 	case SIOCGIFNETMASK:
348 	case SIOCGIFDSTADDR:
349 	case SIOCGIFBRDADDR:
350 		if (ia == (struct in_ifaddr *)0)
351 			return (EADDRNOTAVAIL);
352 		break;
353 	}
354 	switch (cmd) {
355 
356 	case SIOCGIFADDR:
357 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
358 		return (0);
359 
360 	case SIOCGIFBRDADDR:
361 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
362 			return (EINVAL);
363 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
364 		return (0);
365 
366 	case SIOCGIFDSTADDR:
367 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
368 			return (EINVAL);
369 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
370 		return (0);
371 
372 	case SIOCGIFNETMASK:
373 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
374 		return (0);
375 
376 	case SIOCSIFDSTADDR:
377 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
378 			return (EINVAL);
379 		oldaddr = ia->ia_dstaddr;
380 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
381 		if (ifp->if_ioctl) {
382 			IFF_LOCKGIANT(ifp);
383 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
384 			    (caddr_t)ia);
385 			IFF_UNLOCKGIANT(ifp);
386 			if (error) {
387 				ia->ia_dstaddr = oldaddr;
388 				return (error);
389 			}
390 		}
391 		if (ia->ia_flags & IFA_ROUTE) {
392 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
393 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
394 			ia->ia_ifa.ifa_dstaddr =
395 					(struct sockaddr *)&ia->ia_dstaddr;
396 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
397 		}
398 		return (0);
399 
400 	case SIOCSIFBRDADDR:
401 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
402 			return (EINVAL);
403 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
404 		return (0);
405 
406 	case SIOCSIFADDR:
407 		error = in_ifinit(ifp, ia,
408 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
409 		if (error != 0 && iaIsNew)
410 			break;
411 		if (error == 0) {
412 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
413 				in_addmulti(&allhosts_addr, ifp);
414 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
415 		}
416 		return (0);
417 
418 	case SIOCSIFNETMASK:
419 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
420 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
421 		return (0);
422 
423 	case SIOCAIFADDR:
424 		maskIsNew = 0;
425 		hostIsNew = 1;
426 		error = 0;
427 		if (ia->ia_addr.sin_family == AF_INET) {
428 			if (ifra->ifra_addr.sin_len == 0) {
429 				ifra->ifra_addr = ia->ia_addr;
430 				hostIsNew = 0;
431 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
432 					       ia->ia_addr.sin_addr.s_addr)
433 				hostIsNew = 0;
434 		}
435 		if (ifra->ifra_mask.sin_len) {
436 			in_ifscrub(ifp, ia);
437 			ia->ia_sockmask = ifra->ifra_mask;
438 			ia->ia_sockmask.sin_family = AF_INET;
439 			ia->ia_subnetmask =
440 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
441 			maskIsNew = 1;
442 		}
443 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
444 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
445 			in_ifscrub(ifp, ia);
446 			ia->ia_dstaddr = ifra->ifra_dstaddr;
447 			maskIsNew  = 1; /* We lie; but the effect's the same */
448 		}
449 		if (ifra->ifra_addr.sin_family == AF_INET &&
450 		    (hostIsNew || maskIsNew))
451 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
452 		if (error != 0 && iaIsNew)
453 			break;
454 
455 		if ((ifp->if_flags & IFF_BROADCAST) &&
456 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
457 			ia->ia_broadaddr = ifra->ifra_broadaddr;
458 		if (error == 0) {
459 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
460 				in_addmulti(&allhosts_addr, ifp);
461 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
462 		}
463 		return (error);
464 
465 	case SIOCDIFADDR:
466 		/*
467 		 * in_ifscrub kills the interface route.
468 		 */
469 		in_ifscrub(ifp, ia);
470 		/*
471 		 * in_ifadown gets rid of all the rest of
472 		 * the routes.  This is not quite the right
473 		 * thing to do, but at least if we are running
474 		 * a routing process they will come back.
475 		 */
476 		in_ifadown(&ia->ia_ifa, 1);
477 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
478 		error = 0;
479 		break;
480 
481 	default:
482 		if (ifp == 0 || ifp->if_ioctl == 0)
483 			return (EOPNOTSUPP);
484 		IFF_LOCKGIANT(ifp);
485 		error = (*ifp->if_ioctl)(ifp, cmd, data);
486 		IFF_UNLOCKGIANT(ifp);
487 		return (error);
488 	}
489 
490 	/*
491 	 * Protect from ipintr() traversing address list while we're modifying
492 	 * it.
493 	 */
494 	s = splnet();
495 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
496 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
497 	if (ia->ia_addr.sin_family == AF_INET) {
498 		LIST_REMOVE(ia, ia_hash);
499 		/*
500 		 * If this is the last IPv4 address configured on this
501 		 * interface, leave the all-hosts group.
502 		 * XXX: This is quite ugly because of locking and structure.
503 		 */
504 		oia = NULL;
505 		IFP_TO_IA(ifp, oia);
506 		if (oia == NULL) {
507 			struct in_multi *inm;
508 
509 			IFF_LOCKGIANT(ifp);
510 			IN_MULTI_LOCK();
511 			IN_LOOKUP_MULTI(allhosts_addr, ifp, inm);
512 			if (inm != NULL)
513 				in_delmulti_locked(inm);
514 			IN_MULTI_UNLOCK();
515 			IFF_UNLOCKGIANT(ifp);
516 		}
517 	}
518 	IFAFREE(&ia->ia_ifa);
519 	splx(s);
520 
521 	return (error);
522 }
523 
524 /*
525  * SIOC[GAD]LIFADDR.
526  *	SIOCGLIFADDR: get first address. (?!?)
527  *	SIOCGLIFADDR with IFLR_PREFIX:
528  *		get first address that matches the specified prefix.
529  *	SIOCALIFADDR: add the specified address.
530  *	SIOCALIFADDR with IFLR_PREFIX:
531  *		EINVAL since we can't deduce hostid part of the address.
532  *	SIOCDLIFADDR: delete the specified address.
533  *	SIOCDLIFADDR with IFLR_PREFIX:
534  *		delete the first address that matches the specified prefix.
535  * return values:
536  *	EINVAL on invalid parameters
537  *	EADDRNOTAVAIL on prefix match failed/specified address not found
538  *	other values may be returned from in_ioctl()
539  */
540 static int
541 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
542     struct ifnet *ifp, struct thread *td)
543 {
544 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
545 	struct ifaddr *ifa;
546 
547 	/* sanity checks */
548 	if (!data || !ifp) {
549 		panic("invalid argument to in_lifaddr_ioctl");
550 		/*NOTRECHED*/
551 	}
552 
553 	switch (cmd) {
554 	case SIOCGLIFADDR:
555 		/* address must be specified on GET with IFLR_PREFIX */
556 		if ((iflr->flags & IFLR_PREFIX) == 0)
557 			break;
558 		/*FALLTHROUGH*/
559 	case SIOCALIFADDR:
560 	case SIOCDLIFADDR:
561 		/* address must be specified on ADD and DELETE */
562 		if (iflr->addr.ss_family != AF_INET)
563 			return EINVAL;
564 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
565 			return EINVAL;
566 		/* XXX need improvement */
567 		if (iflr->dstaddr.ss_family
568 		 && iflr->dstaddr.ss_family != AF_INET)
569 			return EINVAL;
570 		if (iflr->dstaddr.ss_family
571 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
572 			return EINVAL;
573 		break;
574 	default: /*shouldn't happen*/
575 		return EOPNOTSUPP;
576 	}
577 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
578 		return EINVAL;
579 
580 	switch (cmd) {
581 	case SIOCALIFADDR:
582 	    {
583 		struct in_aliasreq ifra;
584 
585 		if (iflr->flags & IFLR_PREFIX)
586 			return EINVAL;
587 
588 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
589 		bzero(&ifra, sizeof(ifra));
590 		bcopy(iflr->iflr_name, ifra.ifra_name,
591 			sizeof(ifra.ifra_name));
592 
593 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
594 
595 		if (iflr->dstaddr.ss_family) {	/*XXX*/
596 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
597 				iflr->dstaddr.ss_len);
598 		}
599 
600 		ifra.ifra_mask.sin_family = AF_INET;
601 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
602 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
603 
604 		return in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td);
605 	    }
606 	case SIOCGLIFADDR:
607 	case SIOCDLIFADDR:
608 	    {
609 		struct in_ifaddr *ia;
610 		struct in_addr mask, candidate, match;
611 		struct sockaddr_in *sin;
612 
613 		bzero(&mask, sizeof(mask));
614 		bzero(&match, sizeof(match));
615 		if (iflr->flags & IFLR_PREFIX) {
616 			/* lookup a prefix rather than address. */
617 			in_len2mask(&mask, iflr->prefixlen);
618 
619 			sin = (struct sockaddr_in *)&iflr->addr;
620 			match.s_addr = sin->sin_addr.s_addr;
621 			match.s_addr &= mask.s_addr;
622 
623 			/* if you set extra bits, that's wrong */
624 			if (match.s_addr != sin->sin_addr.s_addr)
625 				return EINVAL;
626 
627 		} else {
628 			/* on getting an address, take the 1st match */
629 			/* on deleting an address, do exact match */
630 			if (cmd != SIOCGLIFADDR) {
631 				in_len2mask(&mask, 32);
632 				sin = (struct sockaddr_in *)&iflr->addr;
633 				match.s_addr = sin->sin_addr.s_addr;
634 			}
635 		}
636 
637 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
638 			if (ifa->ifa_addr->sa_family != AF_INET6)
639 				continue;
640 			if (match.s_addr == 0)
641 				break;
642 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
643 			candidate.s_addr &= mask.s_addr;
644 			if (candidate.s_addr == match.s_addr)
645 				break;
646 		}
647 		if (!ifa)
648 			return EADDRNOTAVAIL;
649 		ia = (struct in_ifaddr *)ifa;
650 
651 		if (cmd == SIOCGLIFADDR) {
652 			/* fill in the if_laddrreq structure */
653 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
654 
655 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
656 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
657 					ia->ia_dstaddr.sin_len);
658 			} else
659 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
660 
661 			iflr->prefixlen =
662 				in_mask2len(&ia->ia_sockmask.sin_addr);
663 
664 			iflr->flags = 0;	/*XXX*/
665 
666 			return 0;
667 		} else {
668 			struct in_aliasreq ifra;
669 
670 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
671 			bzero(&ifra, sizeof(ifra));
672 			bcopy(iflr->iflr_name, ifra.ifra_name,
673 				sizeof(ifra.ifra_name));
674 
675 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
676 				ia->ia_addr.sin_len);
677 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
678 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
679 					ia->ia_dstaddr.sin_len);
680 			}
681 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
682 				ia->ia_sockmask.sin_len);
683 
684 			return in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
685 					  ifp, td);
686 		}
687 	    }
688 	}
689 
690 	return EOPNOTSUPP;	/*just for safety*/
691 }
692 
693 /*
694  * Delete any existing route for an interface.
695  */
696 void
697 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
698 {
699 
700 	in_scrubprefix(ia);
701 }
702 
703 /*
704  * Initialize an interface's internet address
705  * and routing table entry.
706  */
707 static int
708 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
709     int scrub)
710 {
711 	register u_long i = ntohl(sin->sin_addr.s_addr);
712 	struct sockaddr_in oldaddr;
713 	int s = splimp(), flags = RTF_UP, error = 0;
714 
715 	oldaddr = ia->ia_addr;
716 	if (oldaddr.sin_family == AF_INET)
717 		LIST_REMOVE(ia, ia_hash);
718 	ia->ia_addr = *sin;
719 	if (ia->ia_addr.sin_family == AF_INET)
720 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
721 		    ia, ia_hash);
722 	/*
723 	 * Give the interface a chance to initialize
724 	 * if this is its first address,
725 	 * and to validate the address if necessary.
726 	 */
727 	if (ifp->if_ioctl) {
728 		IFF_LOCKGIANT(ifp);
729 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
730 		IFF_UNLOCKGIANT(ifp);
731 		if (error) {
732 			splx(s);
733 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
734 			ia->ia_addr = oldaddr;
735 			if (ia->ia_addr.sin_family == AF_INET)
736 				LIST_INSERT_HEAD(INADDR_HASH(
737 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
738 			else
739 				/*
740 				 * If oldaddr family is not AF_INET (e.g.
741 				 * interface has been just created) in_control
742 				 * does not call LIST_REMOVE, and we end up
743 				 * with bogus ia entries in hash
744 				 */
745 				LIST_REMOVE(ia, ia_hash);
746 			return (error);
747 		}
748 	}
749 	splx(s);
750 	if (scrub) {
751 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
752 		in_ifscrub(ifp, ia);
753 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
754 	}
755 	if (IN_CLASSA(i))
756 		ia->ia_netmask = IN_CLASSA_NET;
757 	else if (IN_CLASSB(i))
758 		ia->ia_netmask = IN_CLASSB_NET;
759 	else
760 		ia->ia_netmask = IN_CLASSC_NET;
761 	/*
762 	 * The subnet mask usually includes at least the standard network part,
763 	 * but may may be smaller in the case of supernetting.
764 	 * If it is set, we believe it.
765 	 */
766 	if (ia->ia_subnetmask == 0) {
767 		ia->ia_subnetmask = ia->ia_netmask;
768 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
769 	} else
770 		ia->ia_netmask &= ia->ia_subnetmask;
771 	ia->ia_net = i & ia->ia_netmask;
772 	ia->ia_subnet = i & ia->ia_subnetmask;
773 	in_socktrim(&ia->ia_sockmask);
774 #ifdef DEV_CARP
775 	/*
776 	 * XXX: carp(4) does not have interface route
777 	 */
778 	if (ifp->if_type == IFT_CARP)
779 		return (0);
780 #endif
781 	/*
782 	 * Add route for the network.
783 	 */
784 	ia->ia_ifa.ifa_metric = ifp->if_metric;
785 	if (ifp->if_flags & IFF_BROADCAST) {
786 		ia->ia_broadaddr.sin_addr.s_addr =
787 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
788 		ia->ia_netbroadcast.s_addr =
789 			htonl(ia->ia_net | ~ ia->ia_netmask);
790 	} else if (ifp->if_flags & IFF_LOOPBACK) {
791 		ia->ia_dstaddr = ia->ia_addr;
792 		flags |= RTF_HOST;
793 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
794 		if (ia->ia_dstaddr.sin_family != AF_INET)
795 			return (0);
796 		flags |= RTF_HOST;
797 	}
798 	if ((error = in_addprefix(ia, flags)) != 0)
799 		return (error);
800 
801 	return (error);
802 }
803 
804 #define rtinitflags(x) \
805 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
806 	    ? RTF_HOST : 0)
807 /*
808  * Check if we have a route for the given prefix already or add one accordingly.
809  */
810 static int
811 in_addprefix(struct in_ifaddr *target, int flags)
812 {
813 	struct in_ifaddr *ia;
814 	struct in_addr prefix, mask, p, m;
815 	int error;
816 
817 	if ((flags & RTF_HOST) != 0) {
818 		prefix = target->ia_dstaddr.sin_addr;
819 		mask.s_addr = 0;
820 	} else {
821 		prefix = target->ia_addr.sin_addr;
822 		mask = target->ia_sockmask.sin_addr;
823 		prefix.s_addr &= mask.s_addr;
824 	}
825 
826 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
827 		if (rtinitflags(ia)) {
828 			p = ia->ia_addr.sin_addr;
829 
830 			if (prefix.s_addr != p.s_addr)
831 				continue;
832 		} else {
833 			p = ia->ia_addr.sin_addr;
834 			m = ia->ia_sockmask.sin_addr;
835 			p.s_addr &= m.s_addr;
836 
837 			if (prefix.s_addr != p.s_addr ||
838 			    mask.s_addr != m.s_addr)
839 				continue;
840 		}
841 
842 		/*
843 		 * If we got a matching prefix route inserted by other
844 		 * interface address, we are done here.
845 		 */
846 		if (ia->ia_flags & IFA_ROUTE) {
847 			if (V_sameprefixcarponly &&
848 			    target->ia_ifp->if_type != IFT_CARP &&
849 			    ia->ia_ifp->if_type != IFT_CARP)
850 				return (EEXIST);
851 			else
852 				return (0);
853 		}
854 	}
855 
856 	/*
857 	 * No-one seem to have this prefix route, so we try to insert it.
858 	 */
859 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
860 	if (!error)
861 		target->ia_flags |= IFA_ROUTE;
862 	return error;
863 }
864 
865 /*
866  * If there is no other address in the system that can serve a route to the
867  * same prefix, remove the route.  Hand over the route to the new address
868  * otherwise.
869  */
870 static int
871 in_scrubprefix(struct in_ifaddr *target)
872 {
873 	struct in_ifaddr *ia;
874 	struct in_addr prefix, mask, p;
875 	int error;
876 
877 	if ((target->ia_flags & IFA_ROUTE) == 0)
878 		return 0;
879 
880 	if (rtinitflags(target))
881 		prefix = target->ia_dstaddr.sin_addr;
882 	else {
883 		prefix = target->ia_addr.sin_addr;
884 		mask = target->ia_sockmask.sin_addr;
885 		prefix.s_addr &= mask.s_addr;
886 	}
887 
888 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
889 		if (rtinitflags(ia))
890 			p = ia->ia_dstaddr.sin_addr;
891 		else {
892 			p = ia->ia_addr.sin_addr;
893 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
894 		}
895 
896 		if (prefix.s_addr != p.s_addr)
897 			continue;
898 
899 		/*
900 		 * If we got a matching prefix address, move IFA_ROUTE and
901 		 * the route itself to it.  Make sure that routing daemons
902 		 * get a heads-up.
903 		 *
904 		 * XXX: a special case for carp(4) interface
905 		 */
906 		if ((ia->ia_flags & IFA_ROUTE) == 0
907 #ifdef DEV_CARP
908 		    && (ia->ia_ifp->if_type != IFT_CARP)
909 #endif
910 							) {
911 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
912 			    rtinitflags(target));
913 			target->ia_flags &= ~IFA_ROUTE;
914 
915 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
916 			    rtinitflags(ia) | RTF_UP);
917 			if (error == 0)
918 				ia->ia_flags |= IFA_ROUTE;
919 			return error;
920 		}
921 	}
922 
923 	/*
924 	 * As no-one seem to have this prefix, we can remove the route.
925 	 */
926 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
927 	target->ia_flags &= ~IFA_ROUTE;
928 	return 0;
929 }
930 
931 #undef rtinitflags
932 
933 /*
934  * Return 1 if the address might be a local broadcast address.
935  */
936 int
937 in_broadcast(struct in_addr in, struct ifnet *ifp)
938 {
939 	register struct ifaddr *ifa;
940 	u_long t;
941 
942 	if (in.s_addr == INADDR_BROADCAST ||
943 	    in.s_addr == INADDR_ANY)
944 		return 1;
945 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
946 		return 0;
947 	t = ntohl(in.s_addr);
948 	/*
949 	 * Look through the list of addresses for a match
950 	 * with a broadcast address.
951 	 */
952 #define ia ((struct in_ifaddr *)ifa)
953 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
954 		if (ifa->ifa_addr->sa_family == AF_INET &&
955 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
956 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
957 		     /*
958 		      * Check for old-style (host 0) broadcast.
959 		      */
960 		     t == ia->ia_subnet || t == ia->ia_net) &&
961 		     /*
962 		      * Check for an all one subnetmask. These
963 		      * only exist when an interface gets a secondary
964 		      * address.
965 		      */
966 		     ia->ia_subnetmask != (u_long)0xffffffff)
967 			    return 1;
968 	return (0);
969 #undef ia
970 }
971 
972 /*
973  * Delete all IPv4 multicast address records, and associated link-layer
974  * multicast address records, associated with ifp.
975  */
976 static void
977 in_purgemaddrs(struct ifnet *ifp)
978 {
979 	struct in_multi *inm;
980 	struct in_multi *oinm;
981 
982 #ifdef DIAGNOSTIC
983 	printf("%s: purging ifp %p\n", __func__, ifp);
984 #endif
985 	IFF_LOCKGIANT(ifp);
986 	IN_MULTI_LOCK();
987 	LIST_FOREACH_SAFE(inm, &V_in_multihead, inm_link, oinm) {
988 		if (inm->inm_ifp == ifp)
989 			in_delmulti_locked(inm);
990 	}
991 	IN_MULTI_UNLOCK();
992 	IFF_UNLOCKGIANT(ifp);
993 }
994 
995 /*
996  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
997  */
998 void
999 in_ifdetach(struct ifnet *ifp)
1000 {
1001 
1002 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1003 	in_pcbpurgeif0(&V_udbinfo, ifp);
1004 	in_purgemaddrs(ifp);
1005 }
1006