xref: /freebsd/sys/netinet/in.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
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 #ifdef VIMAGE_GLOBALS
70 static int subnetsarelocal;
71 static int sameprefixcarponly;
72 extern struct inpcbinfo ripcbinfo;
73 #endif
74 
75 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, subnets_are_local,
76 	CTLFLAG_RW, subnetsarelocal, 0,
77 	"Treat all subnets as directly connected");
78 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, same_prefix_carp_only,
79 	CTLFLAG_RW, sameprefixcarponly, 0,
80 	"Refuse to create same prefixes on different interfaces");
81 
82 /*
83  * Return 1 if an internet address is for a ``local'' host
84  * (one to which we have a connection).  If subnetsarelocal
85  * is true, this includes other subnets of the local net.
86  * Otherwise, it includes only the directly-connected (sub)nets.
87  */
88 int
89 in_localaddr(struct in_addr in)
90 {
91 	INIT_VNET_INET(curvnet);
92 	register u_long i = ntohl(in.s_addr);
93 	register struct in_ifaddr *ia;
94 
95 	if (V_subnetsarelocal) {
96 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
97 			if ((i & ia->ia_netmask) == ia->ia_net)
98 				return (1);
99 	} else {
100 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
101 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
102 				return (1);
103 	}
104 	return (0);
105 }
106 
107 /*
108  * Return 1 if an internet address is for the local host and configured
109  * on one of its interfaces.
110  */
111 int
112 in_localip(struct in_addr in)
113 {
114 	INIT_VNET_INET(curvnet);
115 	struct in_ifaddr *ia;
116 
117 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
118 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
119 			return (1);
120 	}
121 	return (0);
122 }
123 
124 /*
125  * Determine whether an IP address is in a reserved set of addresses
126  * that may not be forwarded, or whether datagrams to that destination
127  * may be forwarded.
128  */
129 int
130 in_canforward(struct in_addr in)
131 {
132 	register u_long i = ntohl(in.s_addr);
133 	register u_long net;
134 
135 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
136 		return (0);
137 	if (IN_CLASSA(i)) {
138 		net = i & IN_CLASSA_NET;
139 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
140 			return (0);
141 	}
142 	return (1);
143 }
144 
145 /*
146  * Trim a mask in a sockaddr
147  */
148 static void
149 in_socktrim(struct sockaddr_in *ap)
150 {
151     register char *cplim = (char *) &ap->sin_addr;
152     register char *cp = (char *) (&ap->sin_addr + 1);
153 
154     ap->sin_len = 0;
155     while (--cp >= cplim)
156 	if (*cp) {
157 	    (ap)->sin_len = cp - (char *) (ap) + 1;
158 	    break;
159 	}
160 }
161 
162 static int
163 in_mask2len(mask)
164 	struct in_addr *mask;
165 {
166 	int x, y;
167 	u_char *p;
168 
169 	p = (u_char *)mask;
170 	for (x = 0; x < sizeof(*mask); x++) {
171 		if (p[x] != 0xff)
172 			break;
173 	}
174 	y = 0;
175 	if (x < sizeof(*mask)) {
176 		for (y = 0; y < 8; y++) {
177 			if ((p[x] & (0x80 >> y)) == 0)
178 				break;
179 		}
180 	}
181 	return (x * 8 + y);
182 }
183 
184 static void
185 in_len2mask(struct in_addr *mask, int len)
186 {
187 	int i;
188 	u_char *p;
189 
190 	p = (u_char *)mask;
191 	bzero(mask, sizeof(*mask));
192 	for (i = 0; i < len / 8; i++)
193 		p[i] = 0xff;
194 	if (len % 8)
195 		p[i] = (0xff00 >> (len % 8)) & 0xff;
196 }
197 
198 /*
199  * Generic internet control operations (ioctl's).
200  * Ifp is 0 if not an interface-specific ioctl.
201  */
202 /* ARGSUSED */
203 int
204 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
205     struct thread *td)
206 {
207 	INIT_VNET_INET(curvnet); /* both so and ifp can be NULL here! */
208 	register struct ifreq *ifr = (struct ifreq *)data;
209 	register struct in_ifaddr *ia, *iap;
210 	register struct ifaddr *ifa;
211 	struct in_addr allhosts_addr;
212 	struct in_addr dst;
213 	struct in_ifaddr *oia;
214 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
215 	struct sockaddr_in oldaddr;
216 	int error, hostIsNew, iaIsNew, maskIsNew, s;
217 	int iaIsFirst;
218 
219 	ia = NULL;
220 	iaIsFirst = 0;
221 	iaIsNew = 0;
222 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
223 
224 	switch (cmd) {
225 	case SIOCALIFADDR:
226 		if (td != NULL) {
227 			error = priv_check(td, PRIV_NET_ADDIFADDR);
228 			if (error)
229 				return (error);
230 		}
231 		if (ifp == NULL)
232 			return (EINVAL);
233 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
234 
235 	case SIOCDLIFADDR:
236 		if (td != NULL) {
237 			error = priv_check(td, PRIV_NET_DELIFADDR);
238 			if (error)
239 				return (error);
240 		}
241 		if (ifp == NULL)
242 			return (EINVAL);
243 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
244 
245 	case SIOCGLIFADDR:
246 		if (ifp == NULL)
247 			return (EINVAL);
248 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
249 	}
250 
251 	/*
252 	 * Find address for this interface, if it exists.
253 	 *
254 	 * If an alias address was specified, find that one instead of
255 	 * the first one on the interface, if possible.
256 	 */
257 	if (ifp != NULL) {
258 		dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
259 		LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash)
260 			if (iap->ia_ifp == ifp &&
261 			    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
262 				ia = iap;
263 				break;
264 			}
265 		if (ia == NULL)
266 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
267 				iap = ifatoia(ifa);
268 				if (iap->ia_addr.sin_family == AF_INET) {
269 					ia = iap;
270 					break;
271 				}
272 			}
273 		if (ia == NULL)
274 			iaIsFirst = 1;
275 	}
276 
277 	switch (cmd) {
278 
279 	case SIOCAIFADDR:
280 	case SIOCDIFADDR:
281 		if (ifp == NULL)
282 			return (EADDRNOTAVAIL);
283 		if (ifra->ifra_addr.sin_family == AF_INET) {
284 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
285 				if (ia->ia_ifp == ifp  &&
286 				    ia->ia_addr.sin_addr.s_addr ==
287 				    ifra->ifra_addr.sin_addr.s_addr)
288 					break;
289 			}
290 			if ((ifp->if_flags & IFF_POINTOPOINT)
291 			    && (cmd == SIOCAIFADDR)
292 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
293 				== INADDR_ANY)) {
294 				return (EDESTADDRREQ);
295 			}
296 		}
297 		if (cmd == SIOCDIFADDR && ia == NULL)
298 			return (EADDRNOTAVAIL);
299 		/* FALLTHROUGH */
300 	case SIOCSIFADDR:
301 	case SIOCSIFNETMASK:
302 	case SIOCSIFDSTADDR:
303 		if (td != NULL) {
304 			error = priv_check(td, (cmd == SIOCDIFADDR) ?
305 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
306 			if (error)
307 				return (error);
308 		}
309 
310 		if (ifp == NULL)
311 			return (EADDRNOTAVAIL);
312 		if (ia == NULL) {
313 			ia = (struct in_ifaddr *)
314 				malloc(sizeof *ia, M_IFADDR, M_WAITOK | M_ZERO);
315 			if (ia == NULL)
316 				return (ENOBUFS);
317 			/*
318 			 * Protect from ipintr() traversing address list
319 			 * while we're modifying it.
320 			 */
321 			s = splnet();
322 			ifa = &ia->ia_ifa;
323 			IFA_LOCK_INIT(ifa);
324 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
325 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
326 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
327 			ifa->ifa_refcnt = 1;
328 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
329 
330 			ia->ia_sockmask.sin_len = 8;
331 			ia->ia_sockmask.sin_family = AF_INET;
332 			if (ifp->if_flags & IFF_BROADCAST) {
333 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
334 				ia->ia_broadaddr.sin_family = AF_INET;
335 			}
336 			ia->ia_ifp = ifp;
337 
338 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
339 			splx(s);
340 			iaIsNew = 1;
341 		}
342 		break;
343 
344 	case SIOCSIFBRDADDR:
345 		if (td != NULL) {
346 			error = priv_check(td, PRIV_NET_ADDIFADDR);
347 			if (error)
348 				return (error);
349 		}
350 		/* FALLTHROUGH */
351 
352 	case SIOCGIFADDR:
353 	case SIOCGIFNETMASK:
354 	case SIOCGIFDSTADDR:
355 	case SIOCGIFBRDADDR:
356 		if (ia == NULL)
357 			return (EADDRNOTAVAIL);
358 		break;
359 	}
360 	switch (cmd) {
361 
362 	case SIOCGIFADDR:
363 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
364 		return (0);
365 
366 	case SIOCGIFBRDADDR:
367 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
368 			return (EINVAL);
369 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
370 		return (0);
371 
372 	case SIOCGIFDSTADDR:
373 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
374 			return (EINVAL);
375 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
376 		return (0);
377 
378 	case SIOCGIFNETMASK:
379 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
380 		return (0);
381 
382 	case SIOCSIFDSTADDR:
383 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
384 			return (EINVAL);
385 		oldaddr = ia->ia_dstaddr;
386 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
387 		if (ifp->if_ioctl != NULL) {
388 			IFF_LOCKGIANT(ifp);
389 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
390 			    (caddr_t)ia);
391 			IFF_UNLOCKGIANT(ifp);
392 			if (error) {
393 				ia->ia_dstaddr = oldaddr;
394 				return (error);
395 			}
396 		}
397 		if (ia->ia_flags & IFA_ROUTE) {
398 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
399 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
400 			ia->ia_ifa.ifa_dstaddr =
401 					(struct sockaddr *)&ia->ia_dstaddr;
402 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
403 		}
404 		return (0);
405 
406 	case SIOCSIFBRDADDR:
407 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
408 			return (EINVAL);
409 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
410 		return (0);
411 
412 	case SIOCSIFADDR:
413 		error = in_ifinit(ifp, ia,
414 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
415 		if (error != 0 && iaIsNew)
416 			break;
417 		if (error == 0) {
418 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
419 				in_addmulti(&allhosts_addr, ifp);
420 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
421 		}
422 		return (0);
423 
424 	case SIOCSIFNETMASK:
425 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
426 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
427 		return (0);
428 
429 	case SIOCAIFADDR:
430 		maskIsNew = 0;
431 		hostIsNew = 1;
432 		error = 0;
433 		if (ia->ia_addr.sin_family == AF_INET) {
434 			if (ifra->ifra_addr.sin_len == 0) {
435 				ifra->ifra_addr = ia->ia_addr;
436 				hostIsNew = 0;
437 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
438 					       ia->ia_addr.sin_addr.s_addr)
439 				hostIsNew = 0;
440 		}
441 		if (ifra->ifra_mask.sin_len) {
442 			in_ifscrub(ifp, ia);
443 			ia->ia_sockmask = ifra->ifra_mask;
444 			ia->ia_sockmask.sin_family = AF_INET;
445 			ia->ia_subnetmask =
446 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
447 			maskIsNew = 1;
448 		}
449 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
450 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
451 			in_ifscrub(ifp, ia);
452 			ia->ia_dstaddr = ifra->ifra_dstaddr;
453 			maskIsNew  = 1; /* We lie; but the effect's the same */
454 		}
455 		if (ifra->ifra_addr.sin_family == AF_INET &&
456 		    (hostIsNew || maskIsNew))
457 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
458 		if (error != 0 && iaIsNew)
459 			break;
460 
461 		if ((ifp->if_flags & IFF_BROADCAST) &&
462 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
463 			ia->ia_broadaddr = ifra->ifra_broadaddr;
464 		if (error == 0) {
465 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
466 				in_addmulti(&allhosts_addr, ifp);
467 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
468 		}
469 		return (error);
470 
471 	case SIOCDIFADDR:
472 		/*
473 		 * in_ifscrub kills the interface route.
474 		 */
475 		in_ifscrub(ifp, ia);
476 		/*
477 		 * in_ifadown gets rid of all the rest of
478 		 * the routes.  This is not quite the right
479 		 * thing to do, but at least if we are running
480 		 * a routing process they will come back.
481 		 */
482 		in_ifadown(&ia->ia_ifa, 1);
483 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
484 		error = 0;
485 		break;
486 
487 	default:
488 		if (ifp == NULL || ifp->if_ioctl == NULL)
489 			return (EOPNOTSUPP);
490 		IFF_LOCKGIANT(ifp);
491 		error = (*ifp->if_ioctl)(ifp, cmd, data);
492 		IFF_UNLOCKGIANT(ifp);
493 		return (error);
494 	}
495 
496 	/*
497 	 * Protect from ipintr() traversing address list while we're modifying
498 	 * it.
499 	 */
500 	s = splnet();
501 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
502 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
503 	if (ia->ia_addr.sin_family == AF_INET) {
504 		LIST_REMOVE(ia, ia_hash);
505 		/*
506 		 * If this is the last IPv4 address configured on this
507 		 * interface, leave the all-hosts group.
508 		 * XXX: This is quite ugly because of locking and structure.
509 		 */
510 		oia = NULL;
511 		IFP_TO_IA(ifp, oia);
512 		if (oia == NULL) {
513 			struct in_multi *inm;
514 
515 			IFF_LOCKGIANT(ifp);
516 			IN_MULTI_LOCK();
517 			IN_LOOKUP_MULTI(allhosts_addr, ifp, inm);
518 			if (inm != NULL)
519 				in_delmulti_locked(inm);
520 			IN_MULTI_UNLOCK();
521 			IFF_UNLOCKGIANT(ifp);
522 		}
523 	}
524 	IFAFREE(&ia->ia_ifa);
525 	splx(s);
526 
527 	return (error);
528 }
529 
530 /*
531  * SIOC[GAD]LIFADDR.
532  *	SIOCGLIFADDR: get first address. (?!?)
533  *	SIOCGLIFADDR with IFLR_PREFIX:
534  *		get first address that matches the specified prefix.
535  *	SIOCALIFADDR: add the specified address.
536  *	SIOCALIFADDR with IFLR_PREFIX:
537  *		EINVAL since we can't deduce hostid part of the address.
538  *	SIOCDLIFADDR: delete the specified address.
539  *	SIOCDLIFADDR with IFLR_PREFIX:
540  *		delete the first address that matches the specified prefix.
541  * return values:
542  *	EINVAL on invalid parameters
543  *	EADDRNOTAVAIL on prefix match failed/specified address not found
544  *	other values may be returned from in_ioctl()
545  */
546 static int
547 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
548     struct ifnet *ifp, struct thread *td)
549 {
550 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
551 	struct ifaddr *ifa;
552 
553 	/* sanity checks */
554 	if (data == NULL || ifp == NULL) {
555 		panic("invalid argument to in_lifaddr_ioctl");
556 		/*NOTRECHED*/
557 	}
558 
559 	switch (cmd) {
560 	case SIOCGLIFADDR:
561 		/* address must be specified on GET with IFLR_PREFIX */
562 		if ((iflr->flags & IFLR_PREFIX) == 0)
563 			break;
564 		/*FALLTHROUGH*/
565 	case SIOCALIFADDR:
566 	case SIOCDLIFADDR:
567 		/* address must be specified on ADD and DELETE */
568 		if (iflr->addr.ss_family != AF_INET)
569 			return (EINVAL);
570 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
571 			return (EINVAL);
572 		/* XXX need improvement */
573 		if (iflr->dstaddr.ss_family
574 		 && iflr->dstaddr.ss_family != AF_INET)
575 			return (EINVAL);
576 		if (iflr->dstaddr.ss_family
577 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
578 			return (EINVAL);
579 		break;
580 	default: /*shouldn't happen*/
581 		return (EOPNOTSUPP);
582 	}
583 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
584 		return (EINVAL);
585 
586 	switch (cmd) {
587 	case SIOCALIFADDR:
588 	    {
589 		struct in_aliasreq ifra;
590 
591 		if (iflr->flags & IFLR_PREFIX)
592 			return (EINVAL);
593 
594 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
595 		bzero(&ifra, sizeof(ifra));
596 		bcopy(iflr->iflr_name, ifra.ifra_name,
597 			sizeof(ifra.ifra_name));
598 
599 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
600 
601 		if (iflr->dstaddr.ss_family) {	/*XXX*/
602 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
603 				iflr->dstaddr.ss_len);
604 		}
605 
606 		ifra.ifra_mask.sin_family = AF_INET;
607 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
608 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
609 
610 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
611 	    }
612 	case SIOCGLIFADDR:
613 	case SIOCDLIFADDR:
614 	    {
615 		struct in_ifaddr *ia;
616 		struct in_addr mask, candidate, match;
617 		struct sockaddr_in *sin;
618 
619 		bzero(&mask, sizeof(mask));
620 		bzero(&match, sizeof(match));
621 		if (iflr->flags & IFLR_PREFIX) {
622 			/* lookup a prefix rather than address. */
623 			in_len2mask(&mask, iflr->prefixlen);
624 
625 			sin = (struct sockaddr_in *)&iflr->addr;
626 			match.s_addr = sin->sin_addr.s_addr;
627 			match.s_addr &= mask.s_addr;
628 
629 			/* if you set extra bits, that's wrong */
630 			if (match.s_addr != sin->sin_addr.s_addr)
631 				return (EINVAL);
632 
633 		} else {
634 			/* on getting an address, take the 1st match */
635 			/* on deleting an address, do exact match */
636 			if (cmd != SIOCGLIFADDR) {
637 				in_len2mask(&mask, 32);
638 				sin = (struct sockaddr_in *)&iflr->addr;
639 				match.s_addr = sin->sin_addr.s_addr;
640 			}
641 		}
642 
643 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
644 			if (ifa->ifa_addr->sa_family != AF_INET6)
645 				continue;
646 			if (match.s_addr == 0)
647 				break;
648 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
649 			candidate.s_addr &= mask.s_addr;
650 			if (candidate.s_addr == match.s_addr)
651 				break;
652 		}
653 		if (ifa == NULL)
654 			return (EADDRNOTAVAIL);
655 		ia = (struct in_ifaddr *)ifa;
656 
657 		if (cmd == SIOCGLIFADDR) {
658 			/* fill in the if_laddrreq structure */
659 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
660 
661 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
662 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
663 					ia->ia_dstaddr.sin_len);
664 			} else
665 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
666 
667 			iflr->prefixlen =
668 				in_mask2len(&ia->ia_sockmask.sin_addr);
669 
670 			iflr->flags = 0;	/*XXX*/
671 
672 			return (0);
673 		} else {
674 			struct in_aliasreq ifra;
675 
676 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
677 			bzero(&ifra, sizeof(ifra));
678 			bcopy(iflr->iflr_name, ifra.ifra_name,
679 				sizeof(ifra.ifra_name));
680 
681 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
682 				ia->ia_addr.sin_len);
683 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
684 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
685 					ia->ia_dstaddr.sin_len);
686 			}
687 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
688 				ia->ia_sockmask.sin_len);
689 
690 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
691 			    ifp, td));
692 		}
693 	    }
694 	}
695 
696 	return (EOPNOTSUPP);	/*just for safety*/
697 }
698 
699 /*
700  * Delete any existing route for an interface.
701  */
702 void
703 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
704 {
705 
706 	in_scrubprefix(ia);
707 }
708 
709 /*
710  * Initialize an interface's internet address
711  * and routing table entry.
712  */
713 static int
714 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
715     int scrub)
716 {
717 	INIT_VNET_INET(ifp->if_vnet);
718 	register u_long i = ntohl(sin->sin_addr.s_addr);
719 	struct sockaddr_in oldaddr;
720 	int s = splimp(), flags = RTF_UP, error = 0;
721 
722 	oldaddr = ia->ia_addr;
723 	if (oldaddr.sin_family == AF_INET)
724 		LIST_REMOVE(ia, ia_hash);
725 	ia->ia_addr = *sin;
726 	if (ia->ia_addr.sin_family == AF_INET)
727 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
728 		    ia, ia_hash);
729 	/*
730 	 * Give the interface a chance to initialize
731 	 * if this is its first address,
732 	 * and to validate the address if necessary.
733 	 */
734 	if (ifp->if_ioctl != NULL) {
735 		IFF_LOCKGIANT(ifp);
736 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
737 		IFF_UNLOCKGIANT(ifp);
738 		if (error) {
739 			splx(s);
740 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
741 			ia->ia_addr = oldaddr;
742 			if (ia->ia_addr.sin_family == AF_INET)
743 				LIST_INSERT_HEAD(INADDR_HASH(
744 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
745 			else
746 				/*
747 				 * If oldaddr family is not AF_INET (e.g.
748 				 * interface has been just created) in_control
749 				 * does not call LIST_REMOVE, and we end up
750 				 * with bogus ia entries in hash
751 				 */
752 				LIST_REMOVE(ia, ia_hash);
753 			return (error);
754 		}
755 	}
756 	splx(s);
757 	if (scrub) {
758 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
759 		in_ifscrub(ifp, ia);
760 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
761 	}
762 	if (IN_CLASSA(i))
763 		ia->ia_netmask = IN_CLASSA_NET;
764 	else if (IN_CLASSB(i))
765 		ia->ia_netmask = IN_CLASSB_NET;
766 	else
767 		ia->ia_netmask = IN_CLASSC_NET;
768 	/*
769 	 * The subnet mask usually includes at least the standard network part,
770 	 * but may may be smaller in the case of supernetting.
771 	 * If it is set, we believe it.
772 	 */
773 	if (ia->ia_subnetmask == 0) {
774 		ia->ia_subnetmask = ia->ia_netmask;
775 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
776 	} else
777 		ia->ia_netmask &= ia->ia_subnetmask;
778 	ia->ia_net = i & ia->ia_netmask;
779 	ia->ia_subnet = i & ia->ia_subnetmask;
780 	in_socktrim(&ia->ia_sockmask);
781 #ifdef DEV_CARP
782 	/*
783 	 * XXX: carp(4) does not have interface route
784 	 */
785 	if (ifp->if_type == IFT_CARP)
786 		return (0);
787 #endif
788 	/*
789 	 * Add route for the network.
790 	 */
791 	ia->ia_ifa.ifa_metric = ifp->if_metric;
792 	if (ifp->if_flags & IFF_BROADCAST) {
793 		ia->ia_broadaddr.sin_addr.s_addr =
794 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
795 		ia->ia_netbroadcast.s_addr =
796 			htonl(ia->ia_net | ~ ia->ia_netmask);
797 	} else if (ifp->if_flags & IFF_LOOPBACK) {
798 		ia->ia_dstaddr = ia->ia_addr;
799 		flags |= RTF_HOST;
800 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
801 		if (ia->ia_dstaddr.sin_family != AF_INET)
802 			return (0);
803 		flags |= RTF_HOST;
804 	}
805 	if ((error = in_addprefix(ia, flags)) != 0)
806 		return (error);
807 
808 	return (error);
809 }
810 
811 #define rtinitflags(x) \
812 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
813 	    ? RTF_HOST : 0)
814 /*
815  * Check if we have a route for the given prefix already or add one accordingly.
816  */
817 static int
818 in_addprefix(struct in_ifaddr *target, int flags)
819 {
820 	INIT_VNET_INET(curvnet);
821 	struct in_ifaddr *ia;
822 	struct in_addr prefix, mask, p, m;
823 	int error;
824 
825 	if ((flags & RTF_HOST) != 0) {
826 		prefix = target->ia_dstaddr.sin_addr;
827 		mask.s_addr = 0;
828 	} else {
829 		prefix = target->ia_addr.sin_addr;
830 		mask = target->ia_sockmask.sin_addr;
831 		prefix.s_addr &= mask.s_addr;
832 	}
833 
834 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
835 		if (rtinitflags(ia)) {
836 			p = ia->ia_addr.sin_addr;
837 
838 			if (prefix.s_addr != p.s_addr)
839 				continue;
840 		} else {
841 			p = ia->ia_addr.sin_addr;
842 			m = ia->ia_sockmask.sin_addr;
843 			p.s_addr &= m.s_addr;
844 
845 			if (prefix.s_addr != p.s_addr ||
846 			    mask.s_addr != m.s_addr)
847 				continue;
848 		}
849 
850 		/*
851 		 * If we got a matching prefix route inserted by other
852 		 * interface address, we are done here.
853 		 */
854 		if (ia->ia_flags & IFA_ROUTE) {
855 			if (V_sameprefixcarponly &&
856 			    target->ia_ifp->if_type != IFT_CARP &&
857 			    ia->ia_ifp->if_type != IFT_CARP)
858 				return (EEXIST);
859 			else
860 				return (0);
861 		}
862 	}
863 
864 	/*
865 	 * No-one seem to have this prefix route, so we try to insert it.
866 	 */
867 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
868 	if (!error)
869 		target->ia_flags |= IFA_ROUTE;
870 	return (error);
871 }
872 
873 /*
874  * If there is no other address in the system that can serve a route to the
875  * same prefix, remove the route.  Hand over the route to the new address
876  * otherwise.
877  */
878 static int
879 in_scrubprefix(struct in_ifaddr *target)
880 {
881 	INIT_VNET_INET(curvnet);
882 	struct in_ifaddr *ia;
883 	struct in_addr prefix, mask, p;
884 	int error;
885 
886 	if ((target->ia_flags & IFA_ROUTE) == 0)
887 		return (0);
888 
889 	if (rtinitflags(target))
890 		prefix = target->ia_dstaddr.sin_addr;
891 	else {
892 		prefix = target->ia_addr.sin_addr;
893 		mask = target->ia_sockmask.sin_addr;
894 		prefix.s_addr &= mask.s_addr;
895 	}
896 
897 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
898 		if (rtinitflags(ia))
899 			p = ia->ia_dstaddr.sin_addr;
900 		else {
901 			p = ia->ia_addr.sin_addr;
902 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
903 		}
904 
905 		if (prefix.s_addr != p.s_addr)
906 			continue;
907 
908 		/*
909 		 * If we got a matching prefix address, move IFA_ROUTE and
910 		 * the route itself to it.  Make sure that routing daemons
911 		 * get a heads-up.
912 		 *
913 		 * XXX: a special case for carp(4) interface
914 		 */
915 		if ((ia->ia_flags & IFA_ROUTE) == 0
916 #ifdef DEV_CARP
917 		    && (ia->ia_ifp->if_type != IFT_CARP)
918 #endif
919 							) {
920 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
921 			    rtinitflags(target));
922 			target->ia_flags &= ~IFA_ROUTE;
923 
924 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
925 			    rtinitflags(ia) | RTF_UP);
926 			if (error == 0)
927 				ia->ia_flags |= IFA_ROUTE;
928 			return (error);
929 		}
930 	}
931 
932 	/*
933 	 * As no-one seem to have this prefix, we can remove the route.
934 	 */
935 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
936 	target->ia_flags &= ~IFA_ROUTE;
937 	return (0);
938 }
939 
940 #undef rtinitflags
941 
942 /*
943  * Return 1 if the address might be a local broadcast address.
944  */
945 int
946 in_broadcast(struct in_addr in, struct ifnet *ifp)
947 {
948 	register struct ifaddr *ifa;
949 	u_long t;
950 
951 	if (in.s_addr == INADDR_BROADCAST ||
952 	    in.s_addr == INADDR_ANY)
953 		return (1);
954 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
955 		return (0);
956 	t = ntohl(in.s_addr);
957 	/*
958 	 * Look through the list of addresses for a match
959 	 * with a broadcast address.
960 	 */
961 #define ia ((struct in_ifaddr *)ifa)
962 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
963 		if (ifa->ifa_addr->sa_family == AF_INET &&
964 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
965 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
966 		     /*
967 		      * Check for old-style (host 0) broadcast.
968 		      */
969 		     t == ia->ia_subnet || t == ia->ia_net) &&
970 		     /*
971 		      * Check for an all one subnetmask. These
972 		      * only exist when an interface gets a secondary
973 		      * address.
974 		      */
975 		     ia->ia_subnetmask != (u_long)0xffffffff)
976 			    return (1);
977 	return (0);
978 #undef ia
979 }
980 
981 /*
982  * Delete all IPv4 multicast address records, and associated link-layer
983  * multicast address records, associated with ifp.
984  */
985 static void
986 in_purgemaddrs(struct ifnet *ifp)
987 {
988 	INIT_VNET_INET(ifp->if_vnet);
989 	struct in_multi *inm;
990 	struct in_multi *oinm;
991 
992 #ifdef DIAGNOSTIC
993 	printf("%s: purging ifp %p\n", __func__, ifp);
994 #endif
995 	IFF_LOCKGIANT(ifp);
996 	IN_MULTI_LOCK();
997 	LIST_FOREACH_SAFE(inm, &V_in_multihead, inm_link, oinm) {
998 		if (inm->inm_ifp == ifp)
999 			in_delmulti_locked(inm);
1000 	}
1001 	IN_MULTI_UNLOCK();
1002 	IFF_UNLOCKGIANT(ifp);
1003 }
1004 
1005 /*
1006  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1007  */
1008 void
1009 in_ifdetach(struct ifnet *ifp)
1010 {
1011 	INIT_VNET_INET(ifp->if_vnet);
1012 
1013 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1014 	in_pcbpurgeif0(&V_udbinfo, ifp);
1015 	in_purgemaddrs(ifp);
1016 }
1017