xref: /freebsd/sys/netinet/in.c (revision 2be1a816b9ff69588e55be0a84cbe2a31efc0f2f)
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 
47 #include <net/if.h>
48 #include <net/if_types.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_var.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/ip_var.h>
55 
56 static int in_mask2len(struct in_addr *);
57 static void in_len2mask(struct in_addr *, int);
58 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
59 	struct ifnet *, struct thread *);
60 
61 static int	in_addprefix(struct in_ifaddr *, int);
62 static int	in_scrubprefix(struct in_ifaddr *);
63 static void	in_socktrim(struct sockaddr_in *);
64 static int	in_ifinit(struct ifnet *,
65 	    struct in_ifaddr *, struct sockaddr_in *, int);
66 static void	in_purgemaddrs(struct ifnet *);
67 
68 static int subnetsarelocal = 0;
69 SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
70 	&subnetsarelocal, 0, "Treat all subnets as directly connected");
71 static int sameprefixcarponly = 0;
72 SYSCTL_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
73 	&sameprefixcarponly, 0,
74 	"Refuse to create same prefixes on different interfaces");
75 
76 extern struct inpcbinfo ripcbinfo;
77 extern struct inpcbinfo udbinfo;
78 
79 /*
80  * Return 1 if an internet address is for a ``local'' host
81  * (one to which we have a connection).  If subnetsarelocal
82  * is true, this includes other subnets of the local net.
83  * Otherwise, it includes only the directly-connected (sub)nets.
84  */
85 int
86 in_localaddr(struct in_addr in)
87 {
88 	register u_long i = ntohl(in.s_addr);
89 	register struct in_ifaddr *ia;
90 
91 	if (subnetsarelocal) {
92 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
93 			if ((i & ia->ia_netmask) == ia->ia_net)
94 				return (1);
95 	} else {
96 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
97 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
98 				return (1);
99 	}
100 	return (0);
101 }
102 
103 /*
104  * Return 1 if an internet address is for the local host and configured
105  * on one of its interfaces.
106  */
107 int
108 in_localip(struct in_addr in)
109 {
110 	struct in_ifaddr *ia;
111 
112 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
113 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
114 			return 1;
115 	}
116 	return 0;
117 }
118 
119 /*
120  * Determine whether an IP address is in a reserved set of addresses
121  * that may not be forwarded, or whether datagrams to that destination
122  * may be forwarded.
123  */
124 int
125 in_canforward(struct in_addr in)
126 {
127 	register u_long i = ntohl(in.s_addr);
128 	register u_long net;
129 
130 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
131 		return (0);
132 	if (IN_CLASSA(i)) {
133 		net = i & IN_CLASSA_NET;
134 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
135 			return (0);
136 	}
137 	return (1);
138 }
139 
140 /*
141  * Trim a mask in a sockaddr
142  */
143 static void
144 in_socktrim(struct sockaddr_in *ap)
145 {
146     register char *cplim = (char *) &ap->sin_addr;
147     register char *cp = (char *) (&ap->sin_addr + 1);
148 
149     ap->sin_len = 0;
150     while (--cp >= cplim)
151 	if (*cp) {
152 	    (ap)->sin_len = cp - (char *) (ap) + 1;
153 	    break;
154 	}
155 }
156 
157 static int
158 in_mask2len(mask)
159 	struct in_addr *mask;
160 {
161 	int x, y;
162 	u_char *p;
163 
164 	p = (u_char *)mask;
165 	for (x = 0; x < sizeof(*mask); x++) {
166 		if (p[x] != 0xff)
167 			break;
168 	}
169 	y = 0;
170 	if (x < sizeof(*mask)) {
171 		for (y = 0; y < 8; y++) {
172 			if ((p[x] & (0x80 >> y)) == 0)
173 				break;
174 		}
175 	}
176 	return x * 8 + y;
177 }
178 
179 static void
180 in_len2mask(struct in_addr *mask, int len)
181 {
182 	int i;
183 	u_char *p;
184 
185 	p = (u_char *)mask;
186 	bzero(mask, sizeof(*mask));
187 	for (i = 0; i < len / 8; i++)
188 		p[i] = 0xff;
189 	if (len % 8)
190 		p[i] = (0xff00 >> (len % 8)) & 0xff;
191 }
192 
193 /*
194  * Generic internet control operations (ioctl's).
195  * Ifp is 0 if not an interface-specific ioctl.
196  */
197 /* ARGSUSED */
198 int
199 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
200     struct thread *td)
201 {
202 	register struct ifreq *ifr = (struct ifreq *)data;
203 	register struct in_ifaddr *ia = 0, *iap;
204 	register struct ifaddr *ifa;
205 	struct in_addr allhosts_addr;
206 	struct in_addr dst;
207 	struct in_ifaddr *oia;
208 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
209 	struct sockaddr_in oldaddr;
210 	int error, hostIsNew, iaIsNew, maskIsNew, s;
211 	int iaIsFirst;
212 
213 	iaIsFirst = 0;
214 	iaIsNew = 0;
215 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
216 
217 	switch (cmd) {
218 	case SIOCALIFADDR:
219 		if (td != NULL) {
220 			error = priv_check(td, PRIV_NET_ADDIFADDR);
221 			if (error)
222 				return (error);
223 		}
224 		if (!ifp)
225 			return EINVAL;
226 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
227 
228 	case SIOCDLIFADDR:
229 		if (td != NULL) {
230 			error = priv_check(td, PRIV_NET_DELIFADDR);
231 			if (error)
232 				return (error);
233 		}
234 		if (!ifp)
235 			return EINVAL;
236 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
237 
238 	case SIOCGLIFADDR:
239 		if (!ifp)
240 			return EINVAL;
241 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
242 	}
243 
244 	/*
245 	 * Find address for this interface, if it exists.
246 	 *
247 	 * If an alias address was specified, find that one instead of
248 	 * the first one on the interface, if possible.
249 	 */
250 	if (ifp) {
251 		dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
252 		LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash)
253 			if (iap->ia_ifp == ifp &&
254 			    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
255 				ia = iap;
256 				break;
257 			}
258 		if (ia == NULL)
259 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
260 				iap = ifatoia(ifa);
261 				if (iap->ia_addr.sin_family == AF_INET) {
262 					ia = iap;
263 					break;
264 				}
265 			}
266 		if (ia == NULL)
267 			iaIsFirst = 1;
268 	}
269 
270 	switch (cmd) {
271 
272 	case SIOCAIFADDR:
273 	case SIOCDIFADDR:
274 		if (ifp == 0)
275 			return (EADDRNOTAVAIL);
276 		if (ifra->ifra_addr.sin_family == AF_INET) {
277 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
278 				if (ia->ia_ifp == ifp  &&
279 				    ia->ia_addr.sin_addr.s_addr ==
280 				    ifra->ifra_addr.sin_addr.s_addr)
281 					break;
282 			}
283 			if ((ifp->if_flags & IFF_POINTOPOINT)
284 			    && (cmd == SIOCAIFADDR)
285 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
286 				== INADDR_ANY)) {
287 				return EDESTADDRREQ;
288 			}
289 		}
290 		if (cmd == SIOCDIFADDR && ia == 0)
291 			return (EADDRNOTAVAIL);
292 		/* FALLTHROUGH */
293 	case SIOCSIFADDR:
294 	case SIOCSIFNETMASK:
295 	case SIOCSIFDSTADDR:
296 		if (td != NULL) {
297 			error = priv_check(td, (cmd == SIOCDIFADDR) ?
298 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
299 			if (error)
300 				return (error);
301 		}
302 
303 		if (ifp == 0)
304 			return (EADDRNOTAVAIL);
305 		if (ia == (struct in_ifaddr *)0) {
306 			ia = (struct in_ifaddr *)
307 				malloc(sizeof *ia, M_IFADDR, M_WAITOK | M_ZERO);
308 			if (ia == (struct in_ifaddr *)NULL)
309 				return (ENOBUFS);
310 			/*
311 			 * Protect from ipintr() traversing address list
312 			 * while we're modifying it.
313 			 */
314 			s = splnet();
315 			ifa = &ia->ia_ifa;
316 			IFA_LOCK_INIT(ifa);
317 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
318 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
319 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
320 			ifa->ifa_refcnt = 1;
321 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
322 
323 			ia->ia_sockmask.sin_len = 8;
324 			ia->ia_sockmask.sin_family = AF_INET;
325 			if (ifp->if_flags & IFF_BROADCAST) {
326 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
327 				ia->ia_broadaddr.sin_family = AF_INET;
328 			}
329 			ia->ia_ifp = ifp;
330 
331 			TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_link);
332 			splx(s);
333 			iaIsNew = 1;
334 		}
335 		break;
336 
337 	case SIOCSIFBRDADDR:
338 		if (td != NULL) {
339 			error = priv_check(td, PRIV_NET_ADDIFADDR);
340 			if (error)
341 				return (error);
342 		}
343 		/* FALLTHROUGH */
344 
345 	case SIOCGIFADDR:
346 	case SIOCGIFNETMASK:
347 	case SIOCGIFDSTADDR:
348 	case SIOCGIFBRDADDR:
349 		if (ia == (struct in_ifaddr *)0)
350 			return (EADDRNOTAVAIL);
351 		break;
352 	}
353 	switch (cmd) {
354 
355 	case SIOCGIFADDR:
356 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
357 		return (0);
358 
359 	case SIOCGIFBRDADDR:
360 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
361 			return (EINVAL);
362 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
363 		return (0);
364 
365 	case SIOCGIFDSTADDR:
366 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
367 			return (EINVAL);
368 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
369 		return (0);
370 
371 	case SIOCGIFNETMASK:
372 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
373 		return (0);
374 
375 	case SIOCSIFDSTADDR:
376 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
377 			return (EINVAL);
378 		oldaddr = ia->ia_dstaddr;
379 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
380 		if (ifp->if_ioctl) {
381 			IFF_LOCKGIANT(ifp);
382 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
383 			    (caddr_t)ia);
384 			IFF_UNLOCKGIANT(ifp);
385 			if (error) {
386 				ia->ia_dstaddr = oldaddr;
387 				return (error);
388 			}
389 		}
390 		if (ia->ia_flags & IFA_ROUTE) {
391 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
392 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
393 			ia->ia_ifa.ifa_dstaddr =
394 					(struct sockaddr *)&ia->ia_dstaddr;
395 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
396 		}
397 		return (0);
398 
399 	case SIOCSIFBRDADDR:
400 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
401 			return (EINVAL);
402 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
403 		return (0);
404 
405 	case SIOCSIFADDR:
406 		error = in_ifinit(ifp, ia,
407 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
408 		if (error != 0 && iaIsNew)
409 			break;
410 		if (error == 0) {
411 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
412 				in_addmulti(&allhosts_addr, ifp);
413 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
414 		}
415 		return (0);
416 
417 	case SIOCSIFNETMASK:
418 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
419 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
420 		return (0);
421 
422 	case SIOCAIFADDR:
423 		maskIsNew = 0;
424 		hostIsNew = 1;
425 		error = 0;
426 		if (ia->ia_addr.sin_family == AF_INET) {
427 			if (ifra->ifra_addr.sin_len == 0) {
428 				ifra->ifra_addr = ia->ia_addr;
429 				hostIsNew = 0;
430 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
431 					       ia->ia_addr.sin_addr.s_addr)
432 				hostIsNew = 0;
433 		}
434 		if (ifra->ifra_mask.sin_len) {
435 			in_ifscrub(ifp, ia);
436 			ia->ia_sockmask = ifra->ifra_mask;
437 			ia->ia_sockmask.sin_family = AF_INET;
438 			ia->ia_subnetmask =
439 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
440 			maskIsNew = 1;
441 		}
442 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
443 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
444 			in_ifscrub(ifp, ia);
445 			ia->ia_dstaddr = ifra->ifra_dstaddr;
446 			maskIsNew  = 1; /* We lie; but the effect's the same */
447 		}
448 		if (ifra->ifra_addr.sin_family == AF_INET &&
449 		    (hostIsNew || maskIsNew))
450 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
451 		if (error != 0 && iaIsNew)
452 			break;
453 
454 		if ((ifp->if_flags & IFF_BROADCAST) &&
455 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
456 			ia->ia_broadaddr = ifra->ifra_broadaddr;
457 		if (error == 0) {
458 			if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST) != 0)
459 				in_addmulti(&allhosts_addr, ifp);
460 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
461 		}
462 		return (error);
463 
464 	case SIOCDIFADDR:
465 		/*
466 		 * in_ifscrub kills the interface route.
467 		 */
468 		in_ifscrub(ifp, ia);
469 		/*
470 		 * in_ifadown gets rid of all the rest of
471 		 * the routes.  This is not quite the right
472 		 * thing to do, but at least if we are running
473 		 * a routing process they will come back.
474 		 */
475 		in_ifadown(&ia->ia_ifa, 1);
476 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
477 		error = 0;
478 		break;
479 
480 	default:
481 		if (ifp == 0 || ifp->if_ioctl == 0)
482 			return (EOPNOTSUPP);
483 		IFF_LOCKGIANT(ifp);
484 		error = (*ifp->if_ioctl)(ifp, cmd, data);
485 		IFF_UNLOCKGIANT(ifp);
486 		return (error);
487 	}
488 
489 	/*
490 	 * Protect from ipintr() traversing address list while we're modifying
491 	 * it.
492 	 */
493 	s = splnet();
494 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
495 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_link);
496 	if (ia->ia_addr.sin_family == AF_INET) {
497 		LIST_REMOVE(ia, ia_hash);
498 		/*
499 		 * If this is the last IPv4 address configured on this
500 		 * interface, leave the all-hosts group.
501 		 * XXX: This is quite ugly because of locking and structure.
502 		 */
503 		oia = NULL;
504 		IFP_TO_IA(ifp, oia);
505 		if (oia == NULL) {
506 			struct in_multi *inm;
507 
508 			IFF_LOCKGIANT(ifp);
509 			IN_MULTI_LOCK();
510 			IN_LOOKUP_MULTI(allhosts_addr, ifp, inm);
511 			if (inm != NULL)
512 				in_delmulti_locked(inm);
513 			IN_MULTI_UNLOCK();
514 			IFF_UNLOCKGIANT(ifp);
515 		}
516 	}
517 	IFAFREE(&ia->ia_ifa);
518 	splx(s);
519 
520 	return (error);
521 }
522 
523 /*
524  * SIOC[GAD]LIFADDR.
525  *	SIOCGLIFADDR: get first address. (?!?)
526  *	SIOCGLIFADDR with IFLR_PREFIX:
527  *		get first address that matches the specified prefix.
528  *	SIOCALIFADDR: add the specified address.
529  *	SIOCALIFADDR with IFLR_PREFIX:
530  *		EINVAL since we can't deduce hostid part of the address.
531  *	SIOCDLIFADDR: delete the specified address.
532  *	SIOCDLIFADDR with IFLR_PREFIX:
533  *		delete the first address that matches the specified prefix.
534  * return values:
535  *	EINVAL on invalid parameters
536  *	EADDRNOTAVAIL on prefix match failed/specified address not found
537  *	other values may be returned from in_ioctl()
538  */
539 static int
540 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
541     struct ifnet *ifp, struct thread *td)
542 {
543 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
544 	struct ifaddr *ifa;
545 
546 	/* sanity checks */
547 	if (!data || !ifp) {
548 		panic("invalid argument to in_lifaddr_ioctl");
549 		/*NOTRECHED*/
550 	}
551 
552 	switch (cmd) {
553 	case SIOCGLIFADDR:
554 		/* address must be specified on GET with IFLR_PREFIX */
555 		if ((iflr->flags & IFLR_PREFIX) == 0)
556 			break;
557 		/*FALLTHROUGH*/
558 	case SIOCALIFADDR:
559 	case SIOCDLIFADDR:
560 		/* address must be specified on ADD and DELETE */
561 		if (iflr->addr.ss_family != AF_INET)
562 			return EINVAL;
563 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
564 			return EINVAL;
565 		/* XXX need improvement */
566 		if (iflr->dstaddr.ss_family
567 		 && iflr->dstaddr.ss_family != AF_INET)
568 			return EINVAL;
569 		if (iflr->dstaddr.ss_family
570 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
571 			return EINVAL;
572 		break;
573 	default: /*shouldn't happen*/
574 		return EOPNOTSUPP;
575 	}
576 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
577 		return EINVAL;
578 
579 	switch (cmd) {
580 	case SIOCALIFADDR:
581 	    {
582 		struct in_aliasreq ifra;
583 
584 		if (iflr->flags & IFLR_PREFIX)
585 			return EINVAL;
586 
587 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
588 		bzero(&ifra, sizeof(ifra));
589 		bcopy(iflr->iflr_name, ifra.ifra_name,
590 			sizeof(ifra.ifra_name));
591 
592 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
593 
594 		if (iflr->dstaddr.ss_family) {	/*XXX*/
595 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
596 				iflr->dstaddr.ss_len);
597 		}
598 
599 		ifra.ifra_mask.sin_family = AF_INET;
600 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
601 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
602 
603 		return in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td);
604 	    }
605 	case SIOCGLIFADDR:
606 	case SIOCDLIFADDR:
607 	    {
608 		struct in_ifaddr *ia;
609 		struct in_addr mask, candidate, match;
610 		struct sockaddr_in *sin;
611 
612 		bzero(&mask, sizeof(mask));
613 		bzero(&match, sizeof(match));
614 		if (iflr->flags & IFLR_PREFIX) {
615 			/* lookup a prefix rather than address. */
616 			in_len2mask(&mask, iflr->prefixlen);
617 
618 			sin = (struct sockaddr_in *)&iflr->addr;
619 			match.s_addr = sin->sin_addr.s_addr;
620 			match.s_addr &= mask.s_addr;
621 
622 			/* if you set extra bits, that's wrong */
623 			if (match.s_addr != sin->sin_addr.s_addr)
624 				return EINVAL;
625 
626 		} else {
627 			/* on getting an address, take the 1st match */
628 			/* on deleting an address, do exact match */
629 			if (cmd != SIOCGLIFADDR) {
630 				in_len2mask(&mask, 32);
631 				sin = (struct sockaddr_in *)&iflr->addr;
632 				match.s_addr = sin->sin_addr.s_addr;
633 			}
634 		}
635 
636 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
637 			if (ifa->ifa_addr->sa_family != AF_INET6)
638 				continue;
639 			if (match.s_addr == 0)
640 				break;
641 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
642 			candidate.s_addr &= mask.s_addr;
643 			if (candidate.s_addr == match.s_addr)
644 				break;
645 		}
646 		if (!ifa)
647 			return EADDRNOTAVAIL;
648 		ia = (struct in_ifaddr *)ifa;
649 
650 		if (cmd == SIOCGLIFADDR) {
651 			/* fill in the if_laddrreq structure */
652 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
653 
654 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
655 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
656 					ia->ia_dstaddr.sin_len);
657 			} else
658 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
659 
660 			iflr->prefixlen =
661 				in_mask2len(&ia->ia_sockmask.sin_addr);
662 
663 			iflr->flags = 0;	/*XXX*/
664 
665 			return 0;
666 		} else {
667 			struct in_aliasreq ifra;
668 
669 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
670 			bzero(&ifra, sizeof(ifra));
671 			bcopy(iflr->iflr_name, ifra.ifra_name,
672 				sizeof(ifra.ifra_name));
673 
674 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
675 				ia->ia_addr.sin_len);
676 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
677 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
678 					ia->ia_dstaddr.sin_len);
679 			}
680 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
681 				ia->ia_sockmask.sin_len);
682 
683 			return in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
684 					  ifp, td);
685 		}
686 	    }
687 	}
688 
689 	return EOPNOTSUPP;	/*just for safety*/
690 }
691 
692 /*
693  * Delete any existing route for an interface.
694  */
695 void
696 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
697 {
698 
699 	in_scrubprefix(ia);
700 }
701 
702 /*
703  * Initialize an interface's internet address
704  * and routing table entry.
705  */
706 static int
707 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
708     int scrub)
709 {
710 	register u_long i = ntohl(sin->sin_addr.s_addr);
711 	struct sockaddr_in oldaddr;
712 	int s = splimp(), flags = RTF_UP, error = 0;
713 
714 	oldaddr = ia->ia_addr;
715 	if (oldaddr.sin_family == AF_INET)
716 		LIST_REMOVE(ia, ia_hash);
717 	ia->ia_addr = *sin;
718 	if (ia->ia_addr.sin_family == AF_INET)
719 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
720 		    ia, ia_hash);
721 	/*
722 	 * Give the interface a chance to initialize
723 	 * if this is its first address,
724 	 * and to validate the address if necessary.
725 	 */
726 	if (ifp->if_ioctl) {
727 		IFF_LOCKGIANT(ifp);
728 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
729 		IFF_UNLOCKGIANT(ifp);
730 		if (error) {
731 			splx(s);
732 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
733 			ia->ia_addr = oldaddr;
734 			if (ia->ia_addr.sin_family == AF_INET)
735 				LIST_INSERT_HEAD(INADDR_HASH(
736 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
737 			return (error);
738 		}
739 	}
740 	splx(s);
741 	if (scrub) {
742 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
743 		in_ifscrub(ifp, ia);
744 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
745 	}
746 	if (IN_CLASSA(i))
747 		ia->ia_netmask = IN_CLASSA_NET;
748 	else if (IN_CLASSB(i))
749 		ia->ia_netmask = IN_CLASSB_NET;
750 	else
751 		ia->ia_netmask = IN_CLASSC_NET;
752 	/*
753 	 * The subnet mask usually includes at least the standard network part,
754 	 * but may may be smaller in the case of supernetting.
755 	 * If it is set, we believe it.
756 	 */
757 	if (ia->ia_subnetmask == 0) {
758 		ia->ia_subnetmask = ia->ia_netmask;
759 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
760 	} else
761 		ia->ia_netmask &= ia->ia_subnetmask;
762 	ia->ia_net = i & ia->ia_netmask;
763 	ia->ia_subnet = i & ia->ia_subnetmask;
764 	in_socktrim(&ia->ia_sockmask);
765 #ifdef DEV_CARP
766 	/*
767 	 * XXX: carp(4) does not have interface route
768 	 */
769 	if (ifp->if_type == IFT_CARP)
770 		return (0);
771 #endif
772 	/*
773 	 * Add route for the network.
774 	 */
775 	ia->ia_ifa.ifa_metric = ifp->if_metric;
776 	if (ifp->if_flags & IFF_BROADCAST) {
777 		ia->ia_broadaddr.sin_addr.s_addr =
778 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
779 		ia->ia_netbroadcast.s_addr =
780 			htonl(ia->ia_net | ~ ia->ia_netmask);
781 	} else if (ifp->if_flags & IFF_LOOPBACK) {
782 		ia->ia_dstaddr = ia->ia_addr;
783 		flags |= RTF_HOST;
784 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
785 		if (ia->ia_dstaddr.sin_family != AF_INET)
786 			return (0);
787 		flags |= RTF_HOST;
788 	}
789 	if ((error = in_addprefix(ia, flags)) != 0)
790 		return (error);
791 
792 	return (error);
793 }
794 
795 #define rtinitflags(x) \
796 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
797 	    ? RTF_HOST : 0)
798 /*
799  * Check if we have a route for the given prefix already or add one accordingly.
800  */
801 static int
802 in_addprefix(struct in_ifaddr *target, int flags)
803 {
804 	struct in_ifaddr *ia;
805 	struct in_addr prefix, mask, p, m;
806 	int error;
807 
808 	if ((flags & RTF_HOST) != 0) {
809 		prefix = target->ia_dstaddr.sin_addr;
810 		mask.s_addr = 0;
811 	} else {
812 		prefix = target->ia_addr.sin_addr;
813 		mask = target->ia_sockmask.sin_addr;
814 		prefix.s_addr &= mask.s_addr;
815 	}
816 
817 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
818 		if (rtinitflags(ia)) {
819 			p = ia->ia_addr.sin_addr;
820 
821 			if (prefix.s_addr != p.s_addr)
822 				continue;
823 		} else {
824 			p = ia->ia_addr.sin_addr;
825 			m = ia->ia_sockmask.sin_addr;
826 			p.s_addr &= m.s_addr;
827 
828 			if (prefix.s_addr != p.s_addr ||
829 			    mask.s_addr != m.s_addr)
830 				continue;
831 		}
832 
833 		/*
834 		 * If we got a matching prefix route inserted by other
835 		 * interface address, we are done here.
836 		 */
837 		if (ia->ia_flags & IFA_ROUTE) {
838 			if (sameprefixcarponly &&
839 			    target->ia_ifp->if_type != IFT_CARP &&
840 			    ia->ia_ifp->if_type != IFT_CARP)
841 				return (EEXIST);
842 			else
843 				return (0);
844 		}
845 	}
846 
847 	/*
848 	 * No-one seem to have this prefix route, so we try to insert it.
849 	 */
850 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
851 	if (!error)
852 		target->ia_flags |= IFA_ROUTE;
853 	return error;
854 }
855 
856 /*
857  * If there is no other address in the system that can serve a route to the
858  * same prefix, remove the route.  Hand over the route to the new address
859  * otherwise.
860  */
861 static int
862 in_scrubprefix(struct in_ifaddr *target)
863 {
864 	struct in_ifaddr *ia;
865 	struct in_addr prefix, mask, p;
866 	int error;
867 
868 	if ((target->ia_flags & IFA_ROUTE) == 0)
869 		return 0;
870 
871 	if (rtinitflags(target))
872 		prefix = target->ia_dstaddr.sin_addr;
873 	else {
874 		prefix = target->ia_addr.sin_addr;
875 		mask = target->ia_sockmask.sin_addr;
876 		prefix.s_addr &= mask.s_addr;
877 	}
878 
879 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
880 		if (rtinitflags(ia))
881 			p = ia->ia_dstaddr.sin_addr;
882 		else {
883 			p = ia->ia_addr.sin_addr;
884 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
885 		}
886 
887 		if (prefix.s_addr != p.s_addr)
888 			continue;
889 
890 		/*
891 		 * If we got a matching prefix address, move IFA_ROUTE and
892 		 * the route itself to it.  Make sure that routing daemons
893 		 * get a heads-up.
894 		 *
895 		 * XXX: a special case for carp(4) interface
896 		 */
897 		if ((ia->ia_flags & IFA_ROUTE) == 0
898 #ifdef DEV_CARP
899 		    && (ia->ia_ifp->if_type != IFT_CARP)
900 #endif
901 							) {
902 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
903 			    rtinitflags(target));
904 			target->ia_flags &= ~IFA_ROUTE;
905 
906 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
907 			    rtinitflags(ia) | RTF_UP);
908 			if (error == 0)
909 				ia->ia_flags |= IFA_ROUTE;
910 			return error;
911 		}
912 	}
913 
914 	/*
915 	 * As no-one seem to have this prefix, we can remove the route.
916 	 */
917 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
918 	target->ia_flags &= ~IFA_ROUTE;
919 	return 0;
920 }
921 
922 #undef rtinitflags
923 
924 /*
925  * Return 1 if the address might be a local broadcast address.
926  */
927 int
928 in_broadcast(struct in_addr in, struct ifnet *ifp)
929 {
930 	register struct ifaddr *ifa;
931 	u_long t;
932 
933 	if (in.s_addr == INADDR_BROADCAST ||
934 	    in.s_addr == INADDR_ANY)
935 		return 1;
936 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
937 		return 0;
938 	t = ntohl(in.s_addr);
939 	/*
940 	 * Look through the list of addresses for a match
941 	 * with a broadcast address.
942 	 */
943 #define ia ((struct in_ifaddr *)ifa)
944 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
945 		if (ifa->ifa_addr->sa_family == AF_INET &&
946 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
947 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
948 		     /*
949 		      * Check for old-style (host 0) broadcast.
950 		      */
951 		     t == ia->ia_subnet || t == ia->ia_net) &&
952 		     /*
953 		      * Check for an all one subnetmask. These
954 		      * only exist when an interface gets a secondary
955 		      * address.
956 		      */
957 		     ia->ia_subnetmask != (u_long)0xffffffff)
958 			    return 1;
959 	return (0);
960 #undef ia
961 }
962 
963 /*
964  * Delete all IPv4 multicast address records, and associated link-layer
965  * multicast address records, associated with ifp.
966  */
967 static void
968 in_purgemaddrs(struct ifnet *ifp)
969 {
970 	struct in_multi *inm;
971 	struct in_multi *oinm;
972 
973 #ifdef DIAGNOSTIC
974 	printf("%s: purging ifp %p\n", __func__, ifp);
975 #endif
976 	IFF_LOCKGIANT(ifp);
977 	IN_MULTI_LOCK();
978 	LIST_FOREACH_SAFE(inm, &in_multihead, inm_link, oinm) {
979 		if (inm->inm_ifp == ifp)
980 			in_delmulti_locked(inm);
981 	}
982 	IN_MULTI_UNLOCK();
983 	IFF_UNLOCKGIANT(ifp);
984 }
985 
986 /*
987  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
988  */
989 void
990 in_ifdetach(struct ifnet *ifp)
991 {
992 
993 	in_pcbpurgeif0(&ripcbinfo, ifp);
994 	in_pcbpurgeif0(&udbinfo, ifp);
995 	in_purgemaddrs(ifp);
996 }
997