xref: /freebsd/sys/netinet/in.c (revision 6c6c03be2ddb04c54e455122799923deaefa4114)
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_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, subnets_are_local,
71 	CTLFLAG_RW, subnetsarelocal, 0,
72 	"Treat all subnets as directly connected");
73 static int sameprefixcarponly = 0;
74 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, same_prefix_carp_only,
75 	CTLFLAG_RW, sameprefixcarponly, 0,
76 	"Refuse to create same prefixes on different interfaces");
77 
78 extern struct inpcbinfo ripcbinfo;
79 extern struct inpcbinfo udbinfo;
80 
81 /*
82  * Return 1 if an internet address is for a ``local'' host
83  * (one to which we have a connection).  If subnetsarelocal
84  * is true, this includes other subnets of the local net.
85  * Otherwise, it includes only the directly-connected (sub)nets.
86  */
87 int
88 in_localaddr(struct in_addr in)
89 {
90 	INIT_VNET_INET(curvnet);
91 	register u_long i = ntohl(in.s_addr);
92 	register struct in_ifaddr *ia;
93 
94 	if (V_subnetsarelocal) {
95 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
96 			if ((i & ia->ia_netmask) == ia->ia_net)
97 				return (1);
98 	} else {
99 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
100 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
101 				return (1);
102 	}
103 	return (0);
104 }
105 
106 /*
107  * Return 1 if an internet address is for the local host and configured
108  * on one of its interfaces.
109  */
110 int
111 in_localip(struct in_addr in)
112 {
113 	INIT_VNET_INET(curvnet);
114 	struct in_ifaddr *ia;
115 
116 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
117 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
118 			return (1);
119 	}
120 	return (0);
121 }
122 
123 /*
124  * Determine whether an IP address is in a reserved set of addresses
125  * that may not be forwarded, or whether datagrams to that destination
126  * may be forwarded.
127  */
128 int
129 in_canforward(struct in_addr in)
130 {
131 	register u_long i = ntohl(in.s_addr);
132 	register u_long net;
133 
134 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
135 		return (0);
136 	if (IN_CLASSA(i)) {
137 		net = i & IN_CLASSA_NET;
138 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
139 			return (0);
140 	}
141 	return (1);
142 }
143 
144 /*
145  * Trim a mask in a sockaddr
146  */
147 static void
148 in_socktrim(struct sockaddr_in *ap)
149 {
150     register char *cplim = (char *) &ap->sin_addr;
151     register char *cp = (char *) (&ap->sin_addr + 1);
152 
153     ap->sin_len = 0;
154     while (--cp >= cplim)
155 	if (*cp) {
156 	    (ap)->sin_len = cp - (char *) (ap) + 1;
157 	    break;
158 	}
159 }
160 
161 static int
162 in_mask2len(mask)
163 	struct in_addr *mask;
164 {
165 	int x, y;
166 	u_char *p;
167 
168 	p = (u_char *)mask;
169 	for (x = 0; x < sizeof(*mask); x++) {
170 		if (p[x] != 0xff)
171 			break;
172 	}
173 	y = 0;
174 	if (x < sizeof(*mask)) {
175 		for (y = 0; y < 8; y++) {
176 			if ((p[x] & (0x80 >> y)) == 0)
177 				break;
178 		}
179 	}
180 	return (x * 8 + y);
181 }
182 
183 static void
184 in_len2mask(struct in_addr *mask, int len)
185 {
186 	int i;
187 	u_char *p;
188 
189 	p = (u_char *)mask;
190 	bzero(mask, sizeof(*mask));
191 	for (i = 0; i < len / 8; i++)
192 		p[i] = 0xff;
193 	if (len % 8)
194 		p[i] = (0xff00 >> (len % 8)) & 0xff;
195 }
196 
197 /*
198  * Generic internet control operations (ioctl's).
199  * Ifp is 0 if not an interface-specific ioctl.
200  */
201 /* ARGSUSED */
202 int
203 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
204     struct thread *td)
205 {
206 	INIT_VNET_INET(curvnet); /* both so and ifp can be NULL here! */
207 	register struct ifreq *ifr = (struct ifreq *)data;
208 	register struct in_ifaddr *ia, *iap;
209 	register struct ifaddr *ifa;
210 	struct in_addr allhosts_addr;
211 	struct in_addr dst;
212 	struct in_ifaddr *oia;
213 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
214 	struct sockaddr_in oldaddr;
215 	int error, hostIsNew, iaIsNew, maskIsNew, s;
216 	int iaIsFirst;
217 
218 	ia = NULL;
219 	iaIsFirst = 0;
220 	iaIsNew = 0;
221 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
222 
223 	switch (cmd) {
224 	case SIOCALIFADDR:
225 		if (td != NULL) {
226 			error = priv_check(td, PRIV_NET_ADDIFADDR);
227 			if (error)
228 				return (error);
229 		}
230 		if (ifp == NULL)
231 			return (EINVAL);
232 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
233 
234 	case SIOCDLIFADDR:
235 		if (td != NULL) {
236 			error = priv_check(td, PRIV_NET_DELIFADDR);
237 			if (error)
238 				return (error);
239 		}
240 		if (ifp == NULL)
241 			return (EINVAL);
242 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
243 
244 	case SIOCGLIFADDR:
245 		if (ifp == NULL)
246 			return (EINVAL);
247 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
248 	}
249 
250 	/*
251 	 * Find address for this interface, if it exists.
252 	 *
253 	 * If an alias address was specified, find that one instead of
254 	 * the first one on the interface, if possible.
255 	 */
256 	if (ifp != NULL) {
257 		dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
258 		LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash)
259 			if (iap->ia_ifp == ifp &&
260 			    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
261 				ia = iap;
262 				break;
263 			}
264 		if (ia == NULL)
265 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
266 				iap = ifatoia(ifa);
267 				if (iap->ia_addr.sin_family == AF_INET) {
268 					ia = iap;
269 					break;
270 				}
271 			}
272 		if (ia == NULL)
273 			iaIsFirst = 1;
274 	}
275 
276 	switch (cmd) {
277 
278 	case SIOCAIFADDR:
279 	case SIOCDIFADDR:
280 		if (ifp == NULL)
281 			return (EADDRNOTAVAIL);
282 		if (ifra->ifra_addr.sin_family == AF_INET) {
283 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
284 				if (ia->ia_ifp == ifp  &&
285 				    ia->ia_addr.sin_addr.s_addr ==
286 				    ifra->ifra_addr.sin_addr.s_addr)
287 					break;
288 			}
289 			if ((ifp->if_flags & IFF_POINTOPOINT)
290 			    && (cmd == SIOCAIFADDR)
291 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
292 				== INADDR_ANY)) {
293 				return (EDESTADDRREQ);
294 			}
295 		}
296 		if (cmd == SIOCDIFADDR && ia == NULL)
297 			return (EADDRNOTAVAIL);
298 		/* FALLTHROUGH */
299 	case SIOCSIFADDR:
300 	case SIOCSIFNETMASK:
301 	case SIOCSIFDSTADDR:
302 		if (td != NULL) {
303 			error = priv_check(td, (cmd == SIOCDIFADDR) ?
304 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
305 			if (error)
306 				return (error);
307 		}
308 
309 		if (ifp == NULL)
310 			return (EADDRNOTAVAIL);
311 		if (ia == NULL) {
312 			ia = (struct in_ifaddr *)
313 				malloc(sizeof *ia, M_IFADDR, M_WAITOK | M_ZERO);
314 			if (ia == NULL)
315 				return (ENOBUFS);
316 			/*
317 			 * Protect from ipintr() traversing address list
318 			 * while we're modifying it.
319 			 */
320 			s = splnet();
321 			ifa = &ia->ia_ifa;
322 			IFA_LOCK_INIT(ifa);
323 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
324 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
325 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
326 			ifa->ifa_refcnt = 1;
327 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
328 
329 			ia->ia_sockmask.sin_len = 8;
330 			ia->ia_sockmask.sin_family = AF_INET;
331 			if (ifp->if_flags & IFF_BROADCAST) {
332 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
333 				ia->ia_broadaddr.sin_family = AF_INET;
334 			}
335 			ia->ia_ifp = ifp;
336 
337 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
338 			splx(s);
339 			iaIsNew = 1;
340 		}
341 		break;
342 
343 	case SIOCSIFBRDADDR:
344 		if (td != NULL) {
345 			error = priv_check(td, PRIV_NET_ADDIFADDR);
346 			if (error)
347 				return (error);
348 		}
349 		/* FALLTHROUGH */
350 
351 	case SIOCGIFADDR:
352 	case SIOCGIFNETMASK:
353 	case SIOCGIFDSTADDR:
354 	case SIOCGIFBRDADDR:
355 		if (ia == NULL)
356 			return (EADDRNOTAVAIL);
357 		break;
358 	}
359 	switch (cmd) {
360 
361 	case SIOCGIFADDR:
362 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
363 		return (0);
364 
365 	case SIOCGIFBRDADDR:
366 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
367 			return (EINVAL);
368 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
369 		return (0);
370 
371 	case SIOCGIFDSTADDR:
372 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
373 			return (EINVAL);
374 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
375 		return (0);
376 
377 	case SIOCGIFNETMASK:
378 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
379 		return (0);
380 
381 	case SIOCSIFDSTADDR:
382 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
383 			return (EINVAL);
384 		oldaddr = ia->ia_dstaddr;
385 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
386 		if (ifp->if_ioctl != NULL) {
387 			IFF_LOCKGIANT(ifp);
388 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
389 			    (caddr_t)ia);
390 			IFF_UNLOCKGIANT(ifp);
391 			if (error) {
392 				ia->ia_dstaddr = oldaddr;
393 				return (error);
394 			}
395 		}
396 		if (ia->ia_flags & IFA_ROUTE) {
397 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
398 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
399 			ia->ia_ifa.ifa_dstaddr =
400 					(struct sockaddr *)&ia->ia_dstaddr;
401 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
402 		}
403 		return (0);
404 
405 	case SIOCSIFBRDADDR:
406 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
407 			return (EINVAL);
408 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
409 		return (0);
410 
411 	case SIOCSIFADDR:
412 		error = in_ifinit(ifp, ia,
413 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
414 		if (error != 0 && iaIsNew)
415 			break;
416 		if (error == 0) {
417 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
418 				in_addmulti(&allhosts_addr, ifp);
419 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
420 		}
421 		return (0);
422 
423 	case SIOCSIFNETMASK:
424 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
425 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
426 		return (0);
427 
428 	case SIOCAIFADDR:
429 		maskIsNew = 0;
430 		hostIsNew = 1;
431 		error = 0;
432 		if (ia->ia_addr.sin_family == AF_INET) {
433 			if (ifra->ifra_addr.sin_len == 0) {
434 				ifra->ifra_addr = ia->ia_addr;
435 				hostIsNew = 0;
436 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
437 					       ia->ia_addr.sin_addr.s_addr)
438 				hostIsNew = 0;
439 		}
440 		if (ifra->ifra_mask.sin_len) {
441 			in_ifscrub(ifp, ia);
442 			ia->ia_sockmask = ifra->ifra_mask;
443 			ia->ia_sockmask.sin_family = AF_INET;
444 			ia->ia_subnetmask =
445 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
446 			maskIsNew = 1;
447 		}
448 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
449 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
450 			in_ifscrub(ifp, ia);
451 			ia->ia_dstaddr = ifra->ifra_dstaddr;
452 			maskIsNew  = 1; /* We lie; but the effect's the same */
453 		}
454 		if (ifra->ifra_addr.sin_family == AF_INET &&
455 		    (hostIsNew || maskIsNew))
456 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
457 		if (error != 0 && iaIsNew)
458 			break;
459 
460 		if ((ifp->if_flags & IFF_BROADCAST) &&
461 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
462 			ia->ia_broadaddr = ifra->ifra_broadaddr;
463 		if (error == 0) {
464 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
465 				in_addmulti(&allhosts_addr, ifp);
466 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
467 		}
468 		return (error);
469 
470 	case SIOCDIFADDR:
471 		/*
472 		 * in_ifscrub kills the interface route.
473 		 */
474 		in_ifscrub(ifp, ia);
475 		/*
476 		 * in_ifadown gets rid of all the rest of
477 		 * the routes.  This is not quite the right
478 		 * thing to do, but at least if we are running
479 		 * a routing process they will come back.
480 		 */
481 		in_ifadown(&ia->ia_ifa, 1);
482 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
483 		error = 0;
484 		break;
485 
486 	default:
487 		if (ifp == NULL || ifp->if_ioctl == NULL)
488 			return (EOPNOTSUPP);
489 		IFF_LOCKGIANT(ifp);
490 		error = (*ifp->if_ioctl)(ifp, cmd, data);
491 		IFF_UNLOCKGIANT(ifp);
492 		return (error);
493 	}
494 
495 	/*
496 	 * Protect from ipintr() traversing address list while we're modifying
497 	 * it.
498 	 */
499 	s = splnet();
500 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
501 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
502 	if (ia->ia_addr.sin_family == AF_INET) {
503 		LIST_REMOVE(ia, ia_hash);
504 		/*
505 		 * If this is the last IPv4 address configured on this
506 		 * interface, leave the all-hosts group.
507 		 * XXX: This is quite ugly because of locking and structure.
508 		 */
509 		oia = NULL;
510 		IFP_TO_IA(ifp, oia);
511 		if (oia == NULL) {
512 			struct in_multi *inm;
513 
514 			IFF_LOCKGIANT(ifp);
515 			IN_MULTI_LOCK();
516 			IN_LOOKUP_MULTI(allhosts_addr, ifp, inm);
517 			if (inm != NULL)
518 				in_delmulti_locked(inm);
519 			IN_MULTI_UNLOCK();
520 			IFF_UNLOCKGIANT(ifp);
521 		}
522 	}
523 	IFAFREE(&ia->ia_ifa);
524 	splx(s);
525 
526 	return (error);
527 }
528 
529 /*
530  * SIOC[GAD]LIFADDR.
531  *	SIOCGLIFADDR: get first address. (?!?)
532  *	SIOCGLIFADDR with IFLR_PREFIX:
533  *		get first address that matches the specified prefix.
534  *	SIOCALIFADDR: add the specified address.
535  *	SIOCALIFADDR with IFLR_PREFIX:
536  *		EINVAL since we can't deduce hostid part of the address.
537  *	SIOCDLIFADDR: delete the specified address.
538  *	SIOCDLIFADDR with IFLR_PREFIX:
539  *		delete the first address that matches the specified prefix.
540  * return values:
541  *	EINVAL on invalid parameters
542  *	EADDRNOTAVAIL on prefix match failed/specified address not found
543  *	other values may be returned from in_ioctl()
544  */
545 static int
546 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
547     struct ifnet *ifp, struct thread *td)
548 {
549 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
550 	struct ifaddr *ifa;
551 
552 	/* sanity checks */
553 	if (data == NULL || ifp == NULL) {
554 		panic("invalid argument to in_lifaddr_ioctl");
555 		/*NOTRECHED*/
556 	}
557 
558 	switch (cmd) {
559 	case SIOCGLIFADDR:
560 		/* address must be specified on GET with IFLR_PREFIX */
561 		if ((iflr->flags & IFLR_PREFIX) == 0)
562 			break;
563 		/*FALLTHROUGH*/
564 	case SIOCALIFADDR:
565 	case SIOCDLIFADDR:
566 		/* address must be specified on ADD and DELETE */
567 		if (iflr->addr.ss_family != AF_INET)
568 			return (EINVAL);
569 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
570 			return (EINVAL);
571 		/* XXX need improvement */
572 		if (iflr->dstaddr.ss_family
573 		 && iflr->dstaddr.ss_family != AF_INET)
574 			return (EINVAL);
575 		if (iflr->dstaddr.ss_family
576 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
577 			return (EINVAL);
578 		break;
579 	default: /*shouldn't happen*/
580 		return (EOPNOTSUPP);
581 	}
582 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
583 		return (EINVAL);
584 
585 	switch (cmd) {
586 	case SIOCALIFADDR:
587 	    {
588 		struct in_aliasreq ifra;
589 
590 		if (iflr->flags & IFLR_PREFIX)
591 			return (EINVAL);
592 
593 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
594 		bzero(&ifra, sizeof(ifra));
595 		bcopy(iflr->iflr_name, ifra.ifra_name,
596 			sizeof(ifra.ifra_name));
597 
598 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
599 
600 		if (iflr->dstaddr.ss_family) {	/*XXX*/
601 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
602 				iflr->dstaddr.ss_len);
603 		}
604 
605 		ifra.ifra_mask.sin_family = AF_INET;
606 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
607 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
608 
609 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
610 	    }
611 	case SIOCGLIFADDR:
612 	case SIOCDLIFADDR:
613 	    {
614 		struct in_ifaddr *ia;
615 		struct in_addr mask, candidate, match;
616 		struct sockaddr_in *sin;
617 
618 		bzero(&mask, sizeof(mask));
619 		bzero(&match, sizeof(match));
620 		if (iflr->flags & IFLR_PREFIX) {
621 			/* lookup a prefix rather than address. */
622 			in_len2mask(&mask, iflr->prefixlen);
623 
624 			sin = (struct sockaddr_in *)&iflr->addr;
625 			match.s_addr = sin->sin_addr.s_addr;
626 			match.s_addr &= mask.s_addr;
627 
628 			/* if you set extra bits, that's wrong */
629 			if (match.s_addr != sin->sin_addr.s_addr)
630 				return (EINVAL);
631 
632 		} else {
633 			/* on getting an address, take the 1st match */
634 			/* on deleting an address, do exact match */
635 			if (cmd != SIOCGLIFADDR) {
636 				in_len2mask(&mask, 32);
637 				sin = (struct sockaddr_in *)&iflr->addr;
638 				match.s_addr = sin->sin_addr.s_addr;
639 			}
640 		}
641 
642 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
643 			if (ifa->ifa_addr->sa_family != AF_INET6)
644 				continue;
645 			if (match.s_addr == 0)
646 				break;
647 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
648 			candidate.s_addr &= mask.s_addr;
649 			if (candidate.s_addr == match.s_addr)
650 				break;
651 		}
652 		if (ifa == NULL)
653 			return (EADDRNOTAVAIL);
654 		ia = (struct in_ifaddr *)ifa;
655 
656 		if (cmd == SIOCGLIFADDR) {
657 			/* fill in the if_laddrreq structure */
658 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
659 
660 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
661 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
662 					ia->ia_dstaddr.sin_len);
663 			} else
664 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
665 
666 			iflr->prefixlen =
667 				in_mask2len(&ia->ia_sockmask.sin_addr);
668 
669 			iflr->flags = 0;	/*XXX*/
670 
671 			return (0);
672 		} else {
673 			struct in_aliasreq ifra;
674 
675 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
676 			bzero(&ifra, sizeof(ifra));
677 			bcopy(iflr->iflr_name, ifra.ifra_name,
678 				sizeof(ifra.ifra_name));
679 
680 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
681 				ia->ia_addr.sin_len);
682 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
683 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
684 					ia->ia_dstaddr.sin_len);
685 			}
686 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
687 				ia->ia_sockmask.sin_len);
688 
689 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
690 			    ifp, td));
691 		}
692 	    }
693 	}
694 
695 	return (EOPNOTSUPP);	/*just for safety*/
696 }
697 
698 /*
699  * Delete any existing route for an interface.
700  */
701 void
702 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
703 {
704 
705 	in_scrubprefix(ia);
706 }
707 
708 /*
709  * Initialize an interface's internet address
710  * and routing table entry.
711  */
712 static int
713 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
714     int scrub)
715 {
716 	INIT_VNET_INET(ifp->if_vnet);
717 	register u_long i = ntohl(sin->sin_addr.s_addr);
718 	struct sockaddr_in oldaddr;
719 	int s = splimp(), flags = RTF_UP, error = 0;
720 
721 	oldaddr = ia->ia_addr;
722 	if (oldaddr.sin_family == AF_INET)
723 		LIST_REMOVE(ia, ia_hash);
724 	ia->ia_addr = *sin;
725 	if (ia->ia_addr.sin_family == AF_INET)
726 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
727 		    ia, ia_hash);
728 	/*
729 	 * Give the interface a chance to initialize
730 	 * if this is its first address,
731 	 * and to validate the address if necessary.
732 	 */
733 	if (ifp->if_ioctl != NULL) {
734 		IFF_LOCKGIANT(ifp);
735 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
736 		IFF_UNLOCKGIANT(ifp);
737 		if (error) {
738 			splx(s);
739 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
740 			ia->ia_addr = oldaddr;
741 			if (ia->ia_addr.sin_family == AF_INET)
742 				LIST_INSERT_HEAD(INADDR_HASH(
743 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
744 			else
745 				/*
746 				 * If oldaddr family is not AF_INET (e.g.
747 				 * interface has been just created) in_control
748 				 * does not call LIST_REMOVE, and we end up
749 				 * with bogus ia entries in hash
750 				 */
751 				LIST_REMOVE(ia, ia_hash);
752 			return (error);
753 		}
754 	}
755 	splx(s);
756 	if (scrub) {
757 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
758 		in_ifscrub(ifp, ia);
759 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
760 	}
761 	if (IN_CLASSA(i))
762 		ia->ia_netmask = IN_CLASSA_NET;
763 	else if (IN_CLASSB(i))
764 		ia->ia_netmask = IN_CLASSB_NET;
765 	else
766 		ia->ia_netmask = IN_CLASSC_NET;
767 	/*
768 	 * The subnet mask usually includes at least the standard network part,
769 	 * but may may be smaller in the case of supernetting.
770 	 * If it is set, we believe it.
771 	 */
772 	if (ia->ia_subnetmask == 0) {
773 		ia->ia_subnetmask = ia->ia_netmask;
774 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
775 	} else
776 		ia->ia_netmask &= ia->ia_subnetmask;
777 	ia->ia_net = i & ia->ia_netmask;
778 	ia->ia_subnet = i & ia->ia_subnetmask;
779 	in_socktrim(&ia->ia_sockmask);
780 #ifdef DEV_CARP
781 	/*
782 	 * XXX: carp(4) does not have interface route
783 	 */
784 	if (ifp->if_type == IFT_CARP)
785 		return (0);
786 #endif
787 	/*
788 	 * Add route for the network.
789 	 */
790 	ia->ia_ifa.ifa_metric = ifp->if_metric;
791 	if (ifp->if_flags & IFF_BROADCAST) {
792 		ia->ia_broadaddr.sin_addr.s_addr =
793 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
794 		ia->ia_netbroadcast.s_addr =
795 			htonl(ia->ia_net | ~ ia->ia_netmask);
796 	} else if (ifp->if_flags & IFF_LOOPBACK) {
797 		ia->ia_dstaddr = ia->ia_addr;
798 		flags |= RTF_HOST;
799 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
800 		if (ia->ia_dstaddr.sin_family != AF_INET)
801 			return (0);
802 		flags |= RTF_HOST;
803 	}
804 	if ((error = in_addprefix(ia, flags)) != 0)
805 		return (error);
806 
807 	return (error);
808 }
809 
810 #define rtinitflags(x) \
811 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
812 	    ? RTF_HOST : 0)
813 /*
814  * Check if we have a route for the given prefix already or add one accordingly.
815  */
816 static int
817 in_addprefix(struct in_ifaddr *target, int flags)
818 {
819 	INIT_VNET_INET(curvnet);
820 	struct in_ifaddr *ia;
821 	struct in_addr prefix, mask, p, m;
822 	int error;
823 
824 	if ((flags & RTF_HOST) != 0) {
825 		prefix = target->ia_dstaddr.sin_addr;
826 		mask.s_addr = 0;
827 	} else {
828 		prefix = target->ia_addr.sin_addr;
829 		mask = target->ia_sockmask.sin_addr;
830 		prefix.s_addr &= mask.s_addr;
831 	}
832 
833 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
834 		if (rtinitflags(ia)) {
835 			p = ia->ia_addr.sin_addr;
836 
837 			if (prefix.s_addr != p.s_addr)
838 				continue;
839 		} else {
840 			p = ia->ia_addr.sin_addr;
841 			m = ia->ia_sockmask.sin_addr;
842 			p.s_addr &= m.s_addr;
843 
844 			if (prefix.s_addr != p.s_addr ||
845 			    mask.s_addr != m.s_addr)
846 				continue;
847 		}
848 
849 		/*
850 		 * If we got a matching prefix route inserted by other
851 		 * interface address, we are done here.
852 		 */
853 		if (ia->ia_flags & IFA_ROUTE) {
854 			if (V_sameprefixcarponly &&
855 			    target->ia_ifp->if_type != IFT_CARP &&
856 			    ia->ia_ifp->if_type != IFT_CARP)
857 				return (EEXIST);
858 			else
859 				return (0);
860 		}
861 	}
862 
863 	/*
864 	 * No-one seem to have this prefix route, so we try to insert it.
865 	 */
866 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
867 	if (!error)
868 		target->ia_flags |= IFA_ROUTE;
869 	return (error);
870 }
871 
872 /*
873  * If there is no other address in the system that can serve a route to the
874  * same prefix, remove the route.  Hand over the route to the new address
875  * otherwise.
876  */
877 static int
878 in_scrubprefix(struct in_ifaddr *target)
879 {
880 	INIT_VNET_INET(curvnet);
881 	struct in_ifaddr *ia;
882 	struct in_addr prefix, mask, p;
883 	int error;
884 
885 	if ((target->ia_flags & IFA_ROUTE) == 0)
886 		return (0);
887 
888 	if (rtinitflags(target))
889 		prefix = target->ia_dstaddr.sin_addr;
890 	else {
891 		prefix = target->ia_addr.sin_addr;
892 		mask = target->ia_sockmask.sin_addr;
893 		prefix.s_addr &= mask.s_addr;
894 	}
895 
896 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
897 		if (rtinitflags(ia))
898 			p = ia->ia_dstaddr.sin_addr;
899 		else {
900 			p = ia->ia_addr.sin_addr;
901 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
902 		}
903 
904 		if (prefix.s_addr != p.s_addr)
905 			continue;
906 
907 		/*
908 		 * If we got a matching prefix address, move IFA_ROUTE and
909 		 * the route itself to it.  Make sure that routing daemons
910 		 * get a heads-up.
911 		 *
912 		 * XXX: a special case for carp(4) interface
913 		 */
914 		if ((ia->ia_flags & IFA_ROUTE) == 0
915 #ifdef DEV_CARP
916 		    && (ia->ia_ifp->if_type != IFT_CARP)
917 #endif
918 							) {
919 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
920 			    rtinitflags(target));
921 			target->ia_flags &= ~IFA_ROUTE;
922 
923 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
924 			    rtinitflags(ia) | RTF_UP);
925 			if (error == 0)
926 				ia->ia_flags |= IFA_ROUTE;
927 			return (error);
928 		}
929 	}
930 
931 	/*
932 	 * As no-one seem to have this prefix, we can remove the route.
933 	 */
934 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
935 	target->ia_flags &= ~IFA_ROUTE;
936 	return (0);
937 }
938 
939 #undef rtinitflags
940 
941 /*
942  * Return 1 if the address might be a local broadcast address.
943  */
944 int
945 in_broadcast(struct in_addr in, struct ifnet *ifp)
946 {
947 	register struct ifaddr *ifa;
948 	u_long t;
949 
950 	if (in.s_addr == INADDR_BROADCAST ||
951 	    in.s_addr == INADDR_ANY)
952 		return (1);
953 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
954 		return (0);
955 	t = ntohl(in.s_addr);
956 	/*
957 	 * Look through the list of addresses for a match
958 	 * with a broadcast address.
959 	 */
960 #define ia ((struct in_ifaddr *)ifa)
961 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
962 		if (ifa->ifa_addr->sa_family == AF_INET &&
963 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
964 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
965 		     /*
966 		      * Check for old-style (host 0) broadcast.
967 		      */
968 		     t == ia->ia_subnet || t == ia->ia_net) &&
969 		     /*
970 		      * Check for an all one subnetmask. These
971 		      * only exist when an interface gets a secondary
972 		      * address.
973 		      */
974 		     ia->ia_subnetmask != (u_long)0xffffffff)
975 			    return (1);
976 	return (0);
977 #undef ia
978 }
979 
980 /*
981  * Delete all IPv4 multicast address records, and associated link-layer
982  * multicast address records, associated with ifp.
983  */
984 static void
985 in_purgemaddrs(struct ifnet *ifp)
986 {
987 	INIT_VNET_INET(ifp->if_vnet);
988 	struct in_multi *inm;
989 	struct in_multi *oinm;
990 
991 #ifdef DIAGNOSTIC
992 	printf("%s: purging ifp %p\n", __func__, ifp);
993 #endif
994 	IFF_LOCKGIANT(ifp);
995 	IN_MULTI_LOCK();
996 	LIST_FOREACH_SAFE(inm, &V_in_multihead, inm_link, oinm) {
997 		if (inm->inm_ifp == ifp)
998 			in_delmulti_locked(inm);
999 	}
1000 	IN_MULTI_UNLOCK();
1001 	IFF_UNLOCKGIANT(ifp);
1002 }
1003 
1004 /*
1005  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1006  */
1007 void
1008 in_ifdetach(struct ifnet *ifp)
1009 {
1010 	INIT_VNET_INET(ifp->if_vnet);
1011 
1012 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1013 	in_pcbpurgeif0(&V_udbinfo, ifp);
1014 	in_purgemaddrs(ifp);
1015 }
1016