xref: /freebsd/sys/netinet/in.c (revision 262e143bd46171a6415a5b28af260a5efa2a3db8)
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  * $FreeBSD$
32  */
33 
34 #include "opt_carp.h"
35 
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/sockio.h>
39 #include <sys/malloc.h>
40 #include <sys/socket.h>
41 #include <sys/kernel.h>
42 #include <sys/sysctl.h>
43 
44 #include <net/if.h>
45 #include <net/if_types.h>
46 #include <net/route.h>
47 
48 #include <netinet/in.h>
49 #include <netinet/in_var.h>
50 #include <netinet/in_pcb.h>
51 
52 #include <netinet/igmp_var.h>
53 
54 static MALLOC_DEFINE(M_IPMADDR, "in_multi", "internet multicast address");
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 
67 static int subnetsarelocal = 0;
68 SYSCTL_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
69 	&subnetsarelocal, 0, "Treat all subnets as directly connected");
70 static int sameprefixcarponly = 0;
71 SYSCTL_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
72 	&sameprefixcarponly, 0,
73 	"Refuse to create same prefixes on different interfaces");
74 
75 /*
76  * The IPv4 multicast list (in_multihead and associated structures) are
77  * protected by the global in_multi_mtx.  See in_var.h for more details.  For
78  * now, in_multi_mtx is marked as recursible due to IGMP's calling back into
79  * ip_output() to send IGMP packets while holding the lock; this probably is
80  * not quite desirable.
81  */
82 struct in_multihead in_multihead; /* XXX BSS initialization */
83 struct mtx in_multi_mtx;
84 MTX_SYSINIT(in_multi_mtx, &in_multi_mtx, "in_multi_mtx", MTX_DEF | MTX_RECURSE);
85 
86 extern struct inpcbinfo ripcbinfo;
87 extern struct inpcbinfo udbinfo;
88 
89 /*
90  * Return 1 if an internet address is for a ``local'' host
91  * (one to which we have a connection).  If subnetsarelocal
92  * is true, this includes other subnets of the local net.
93  * Otherwise, it includes only the directly-connected (sub)nets.
94  */
95 int
96 in_localaddr(in)
97 	struct in_addr in;
98 {
99 	register u_long i = ntohl(in.s_addr);
100 	register struct in_ifaddr *ia;
101 
102 	if (subnetsarelocal) {
103 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
104 			if ((i & ia->ia_netmask) == ia->ia_net)
105 				return (1);
106 	} else {
107 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
108 			if ((i & ia->ia_subnetmask) == ia->ia_subnet)
109 				return (1);
110 	}
111 	return (0);
112 }
113 
114 /*
115  * Return 1 if an internet address is for the local host and configured
116  * on one of its interfaces.
117  */
118 int
119 in_localip(in)
120 	struct in_addr in;
121 {
122 	struct in_ifaddr *ia;
123 
124 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
125 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr)
126 			return 1;
127 	}
128 	return 0;
129 }
130 
131 /*
132  * Determine whether an IP address is in a reserved set of addresses
133  * that may not be forwarded, or whether datagrams to that destination
134  * may be forwarded.
135  */
136 int
137 in_canforward(in)
138 	struct in_addr in;
139 {
140 	register u_long i = ntohl(in.s_addr);
141 	register u_long net;
142 
143 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i))
144 		return (0);
145 	if (IN_CLASSA(i)) {
146 		net = i & IN_CLASSA_NET;
147 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
148 			return (0);
149 	}
150 	return (1);
151 }
152 
153 /*
154  * Trim a mask in a sockaddr
155  */
156 static void
157 in_socktrim(ap)
158 struct sockaddr_in *ap;
159 {
160     register char *cplim = (char *) &ap->sin_addr;
161     register char *cp = (char *) (&ap->sin_addr + 1);
162 
163     ap->sin_len = 0;
164     while (--cp >= cplim)
165 	if (*cp) {
166 	    (ap)->sin_len = cp - (char *) (ap) + 1;
167 	    break;
168 	}
169 }
170 
171 static int
172 in_mask2len(mask)
173 	struct in_addr *mask;
174 {
175 	int x, y;
176 	u_char *p;
177 
178 	p = (u_char *)mask;
179 	for (x = 0; x < sizeof(*mask); x++) {
180 		if (p[x] != 0xff)
181 			break;
182 	}
183 	y = 0;
184 	if (x < sizeof(*mask)) {
185 		for (y = 0; y < 8; y++) {
186 			if ((p[x] & (0x80 >> y)) == 0)
187 				break;
188 		}
189 	}
190 	return x * 8 + y;
191 }
192 
193 static void
194 in_len2mask(mask, len)
195 	struct in_addr *mask;
196 	int len;
197 {
198 	int i;
199 	u_char *p;
200 
201 	p = (u_char *)mask;
202 	bzero(mask, sizeof(*mask));
203 	for (i = 0; i < len / 8; i++)
204 		p[i] = 0xff;
205 	if (len % 8)
206 		p[i] = (0xff00 >> (len % 8)) & 0xff;
207 }
208 
209 /*
210  * Generic internet control operations (ioctl's).
211  * Ifp is 0 if not an interface-specific ioctl.
212  */
213 /* ARGSUSED */
214 int
215 in_control(so, cmd, data, ifp, td)
216 	struct socket *so;
217 	u_long cmd;
218 	caddr_t data;
219 	register struct ifnet *ifp;
220 	struct thread *td;
221 {
222 	register struct ifreq *ifr = (struct ifreq *)data;
223 	register struct in_ifaddr *ia = 0, *iap;
224 	register struct ifaddr *ifa;
225 	struct in_addr dst;
226 	struct in_ifaddr *oia;
227 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
228 	struct sockaddr_in oldaddr;
229 	int error, hostIsNew, iaIsNew, maskIsNew, s;
230 
231 	iaIsNew = 0;
232 
233 	switch (cmd) {
234 	case SIOCALIFADDR:
235 	case SIOCDLIFADDR:
236 		if (td && (error = suser(td)) != 0)
237 			return error;
238 		/*fall through*/
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 	}
268 
269 	switch (cmd) {
270 
271 	case SIOCAIFADDR:
272 	case SIOCDIFADDR:
273 		if (ifp == 0)
274 			return (EADDRNOTAVAIL);
275 		if (ifra->ifra_addr.sin_family == AF_INET) {
276 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
277 				if (ia->ia_ifp == ifp  &&
278 				    ia->ia_addr.sin_addr.s_addr ==
279 				    ifra->ifra_addr.sin_addr.s_addr)
280 					break;
281 			}
282 			if ((ifp->if_flags & IFF_POINTOPOINT)
283 			    && (cmd == SIOCAIFADDR)
284 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
285 				== INADDR_ANY)) {
286 				return EDESTADDRREQ;
287 			}
288 		}
289 		if (cmd == SIOCDIFADDR && ia == 0)
290 			return (EADDRNOTAVAIL);
291 		/* FALLTHROUGH */
292 	case SIOCSIFADDR:
293 	case SIOCSIFNETMASK:
294 	case SIOCSIFDSTADDR:
295 		if (td && (error = suser(td)) != 0)
296 			return error;
297 
298 		if (ifp == 0)
299 			return (EADDRNOTAVAIL);
300 		if (ia == (struct in_ifaddr *)0) {
301 			ia = (struct in_ifaddr *)
302 				malloc(sizeof *ia, M_IFADDR, M_WAITOK | M_ZERO);
303 			if (ia == (struct in_ifaddr *)NULL)
304 				return (ENOBUFS);
305 			/*
306 			 * Protect from ipintr() traversing address list
307 			 * while we're modifying it.
308 			 */
309 			s = splnet();
310 			ifa = &ia->ia_ifa;
311 			IFA_LOCK_INIT(ifa);
312 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
313 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
314 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
315 			ifa->ifa_refcnt = 1;
316 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
317 
318 			ia->ia_sockmask.sin_len = 8;
319 			ia->ia_sockmask.sin_family = AF_INET;
320 			if (ifp->if_flags & IFF_BROADCAST) {
321 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
322 				ia->ia_broadaddr.sin_family = AF_INET;
323 			}
324 			ia->ia_ifp = ifp;
325 
326 			TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_link);
327 			splx(s);
328 			iaIsNew = 1;
329 		}
330 		break;
331 
332 	case SIOCSIFBRDADDR:
333 		if (td && (error = suser(td)) != 0)
334 			return error;
335 		/* FALLTHROUGH */
336 
337 	case SIOCGIFADDR:
338 	case SIOCGIFNETMASK:
339 	case SIOCGIFDSTADDR:
340 	case SIOCGIFBRDADDR:
341 		if (ia == (struct in_ifaddr *)0)
342 			return (EADDRNOTAVAIL);
343 		break;
344 	}
345 	switch (cmd) {
346 
347 	case SIOCGIFADDR:
348 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
349 		return (0);
350 
351 	case SIOCGIFBRDADDR:
352 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
353 			return (EINVAL);
354 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
355 		return (0);
356 
357 	case SIOCGIFDSTADDR:
358 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
359 			return (EINVAL);
360 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
361 		return (0);
362 
363 	case SIOCGIFNETMASK:
364 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
365 		return (0);
366 
367 	case SIOCSIFDSTADDR:
368 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
369 			return (EINVAL);
370 		oldaddr = ia->ia_dstaddr;
371 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
372 		if (ifp->if_ioctl) {
373 			IFF_LOCKGIANT(ifp);
374 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
375 			    (caddr_t)ia);
376 			IFF_UNLOCKGIANT(ifp);
377 			if (error) {
378 				ia->ia_dstaddr = oldaddr;
379 				return (error);
380 			}
381 		}
382 		if (ia->ia_flags & IFA_ROUTE) {
383 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
384 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
385 			ia->ia_ifa.ifa_dstaddr =
386 					(struct sockaddr *)&ia->ia_dstaddr;
387 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
388 		}
389 		return (0);
390 
391 	case SIOCSIFBRDADDR:
392 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
393 			return (EINVAL);
394 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
395 		return (0);
396 
397 	case SIOCSIFADDR:
398 		error = in_ifinit(ifp, ia,
399 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
400 		if (error != 0 && iaIsNew)
401 			break;
402 		if (error == 0)
403 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
404 		return (0);
405 
406 	case SIOCSIFNETMASK:
407 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
408 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
409 		return (0);
410 
411 	case SIOCAIFADDR:
412 		maskIsNew = 0;
413 		hostIsNew = 1;
414 		error = 0;
415 		if (ia->ia_addr.sin_family == AF_INET) {
416 			if (ifra->ifra_addr.sin_len == 0) {
417 				ifra->ifra_addr = ia->ia_addr;
418 				hostIsNew = 0;
419 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
420 					       ia->ia_addr.sin_addr.s_addr)
421 				hostIsNew = 0;
422 		}
423 		if (ifra->ifra_mask.sin_len) {
424 			in_ifscrub(ifp, ia);
425 			ia->ia_sockmask = ifra->ifra_mask;
426 			ia->ia_sockmask.sin_family = AF_INET;
427 			ia->ia_subnetmask =
428 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
429 			maskIsNew = 1;
430 		}
431 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
432 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
433 			in_ifscrub(ifp, ia);
434 			ia->ia_dstaddr = ifra->ifra_dstaddr;
435 			maskIsNew  = 1; /* We lie; but the effect's the same */
436 		}
437 		if (ifra->ifra_addr.sin_family == AF_INET &&
438 		    (hostIsNew || maskIsNew))
439 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
440 		if (error != 0 && iaIsNew)
441 			break;
442 
443 		if ((ifp->if_flags & IFF_BROADCAST) &&
444 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
445 			ia->ia_broadaddr = ifra->ifra_broadaddr;
446 		if (error == 0)
447 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
448 		return (error);
449 
450 	case SIOCDIFADDR:
451 		/*
452 		 * in_ifscrub kills the interface route.
453 		 */
454 		in_ifscrub(ifp, ia);
455 		/*
456 		 * in_ifadown gets rid of all the rest of
457 		 * the routes.  This is not quite the right
458 		 * thing to do, but at least if we are running
459 		 * a routing process they will come back.
460 		 */
461 		in_ifadown(&ia->ia_ifa, 1);
462 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
463 		error = 0;
464 		break;
465 
466 	default:
467 		if (ifp == 0 || ifp->if_ioctl == 0)
468 			return (EOPNOTSUPP);
469 		IFF_LOCKGIANT(ifp);
470 		error = (*ifp->if_ioctl)(ifp, cmd, data);
471 		IFF_UNLOCKGIANT(ifp);
472 		return (error);
473 	}
474 
475 	/*
476 	 * Protect from ipintr() traversing address list while we're modifying
477 	 * it.
478 	 */
479 	s = splnet();
480 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
481 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_link);
482 	LIST_REMOVE(ia, ia_hash);
483 	IFAFREE(&ia->ia_ifa);
484 	splx(s);
485 
486 	return (error);
487 }
488 
489 /*
490  * SIOC[GAD]LIFADDR.
491  *	SIOCGLIFADDR: get first address. (?!?)
492  *	SIOCGLIFADDR with IFLR_PREFIX:
493  *		get first address that matches the specified prefix.
494  *	SIOCALIFADDR: add the specified address.
495  *	SIOCALIFADDR with IFLR_PREFIX:
496  *		EINVAL since we can't deduce hostid part of the address.
497  *	SIOCDLIFADDR: delete the specified address.
498  *	SIOCDLIFADDR with IFLR_PREFIX:
499  *		delete the first address that matches the specified prefix.
500  * return values:
501  *	EINVAL on invalid parameters
502  *	EADDRNOTAVAIL on prefix match failed/specified address not found
503  *	other values may be returned from in_ioctl()
504  */
505 static int
506 in_lifaddr_ioctl(so, cmd, data, ifp, td)
507 	struct socket *so;
508 	u_long cmd;
509 	caddr_t	data;
510 	struct ifnet *ifp;
511 	struct thread *td;
512 {
513 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
514 	struct ifaddr *ifa;
515 
516 	/* sanity checks */
517 	if (!data || !ifp) {
518 		panic("invalid argument to in_lifaddr_ioctl");
519 		/*NOTRECHED*/
520 	}
521 
522 	switch (cmd) {
523 	case SIOCGLIFADDR:
524 		/* address must be specified on GET with IFLR_PREFIX */
525 		if ((iflr->flags & IFLR_PREFIX) == 0)
526 			break;
527 		/*FALLTHROUGH*/
528 	case SIOCALIFADDR:
529 	case SIOCDLIFADDR:
530 		/* address must be specified on ADD and DELETE */
531 		if (iflr->addr.ss_family != AF_INET)
532 			return EINVAL;
533 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
534 			return EINVAL;
535 		/* XXX need improvement */
536 		if (iflr->dstaddr.ss_family
537 		 && iflr->dstaddr.ss_family != AF_INET)
538 			return EINVAL;
539 		if (iflr->dstaddr.ss_family
540 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
541 			return EINVAL;
542 		break;
543 	default: /*shouldn't happen*/
544 		return EOPNOTSUPP;
545 	}
546 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
547 		return EINVAL;
548 
549 	switch (cmd) {
550 	case SIOCALIFADDR:
551 	    {
552 		struct in_aliasreq ifra;
553 
554 		if (iflr->flags & IFLR_PREFIX)
555 			return EINVAL;
556 
557 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
558 		bzero(&ifra, sizeof(ifra));
559 		bcopy(iflr->iflr_name, ifra.ifra_name,
560 			sizeof(ifra.ifra_name));
561 
562 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
563 
564 		if (iflr->dstaddr.ss_family) {	/*XXX*/
565 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
566 				iflr->dstaddr.ss_len);
567 		}
568 
569 		ifra.ifra_mask.sin_family = AF_INET;
570 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
571 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
572 
573 		return in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td);
574 	    }
575 	case SIOCGLIFADDR:
576 	case SIOCDLIFADDR:
577 	    {
578 		struct in_ifaddr *ia;
579 		struct in_addr mask, candidate, match;
580 		struct sockaddr_in *sin;
581 		int cmp;
582 
583 		bzero(&mask, sizeof(mask));
584 		if (iflr->flags & IFLR_PREFIX) {
585 			/* lookup a prefix rather than address. */
586 			in_len2mask(&mask, iflr->prefixlen);
587 
588 			sin = (struct sockaddr_in *)&iflr->addr;
589 			match.s_addr = sin->sin_addr.s_addr;
590 			match.s_addr &= mask.s_addr;
591 
592 			/* if you set extra bits, that's wrong */
593 			if (match.s_addr != sin->sin_addr.s_addr)
594 				return EINVAL;
595 
596 			cmp = 1;
597 		} else {
598 			if (cmd == SIOCGLIFADDR) {
599 				/* on getting an address, take the 1st match */
600 				cmp = 0;	/*XXX*/
601 			} else {
602 				/* on deleting an address, do exact match */
603 				in_len2mask(&mask, 32);
604 				sin = (struct sockaddr_in *)&iflr->addr;
605 				match.s_addr = sin->sin_addr.s_addr;
606 
607 				cmp = 1;
608 			}
609 		}
610 
611 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
612 			if (ifa->ifa_addr->sa_family != AF_INET6)
613 				continue;
614 			if (!cmp)
615 				break;
616 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
617 			candidate.s_addr &= mask.s_addr;
618 			if (candidate.s_addr == match.s_addr)
619 				break;
620 		}
621 		if (!ifa)
622 			return EADDRNOTAVAIL;
623 		ia = (struct in_ifaddr *)ifa;
624 
625 		if (cmd == SIOCGLIFADDR) {
626 			/* fill in the if_laddrreq structure */
627 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
628 
629 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
630 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
631 					ia->ia_dstaddr.sin_len);
632 			} else
633 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
634 
635 			iflr->prefixlen =
636 				in_mask2len(&ia->ia_sockmask.sin_addr);
637 
638 			iflr->flags = 0;	/*XXX*/
639 
640 			return 0;
641 		} else {
642 			struct in_aliasreq ifra;
643 
644 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
645 			bzero(&ifra, sizeof(ifra));
646 			bcopy(iflr->iflr_name, ifra.ifra_name,
647 				sizeof(ifra.ifra_name));
648 
649 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
650 				ia->ia_addr.sin_len);
651 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
652 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
653 					ia->ia_dstaddr.sin_len);
654 			}
655 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
656 				ia->ia_sockmask.sin_len);
657 
658 			return in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
659 					  ifp, td);
660 		}
661 	    }
662 	}
663 
664 	return EOPNOTSUPP;	/*just for safety*/
665 }
666 
667 /*
668  * Delete any existing route for an interface.
669  */
670 void
671 in_ifscrub(ifp, ia)
672 	register struct ifnet *ifp;
673 	register struct in_ifaddr *ia;
674 {
675 	in_scrubprefix(ia);
676 }
677 
678 /*
679  * Initialize an interface's internet address
680  * and routing table entry.
681  */
682 static int
683 in_ifinit(ifp, ia, sin, scrub)
684 	register struct ifnet *ifp;
685 	register struct in_ifaddr *ia;
686 	struct sockaddr_in *sin;
687 	int scrub;
688 {
689 	register u_long i = ntohl(sin->sin_addr.s_addr);
690 	struct sockaddr_in oldaddr;
691 	int s = splimp(), flags = RTF_UP, error = 0;
692 
693 	oldaddr = ia->ia_addr;
694 	if (oldaddr.sin_family == AF_INET)
695 		LIST_REMOVE(ia, ia_hash);
696 	ia->ia_addr = *sin;
697 	if (ia->ia_addr.sin_family == AF_INET)
698 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
699 		    ia, ia_hash);
700 	/*
701 	 * Give the interface a chance to initialize
702 	 * if this is its first address,
703 	 * and to validate the address if necessary.
704 	 */
705 	if (ifp->if_ioctl) {
706 		IFF_LOCKGIANT(ifp);
707 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
708 		IFF_UNLOCKGIANT(ifp);
709 		if (error) {
710 			splx(s);
711 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
712 			ia->ia_addr = oldaddr;
713 			if (ia->ia_addr.sin_family == AF_INET)
714 				LIST_INSERT_HEAD(INADDR_HASH(
715 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
716 			return (error);
717 		}
718 	}
719 	splx(s);
720 	if (scrub) {
721 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
722 		in_ifscrub(ifp, ia);
723 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
724 	}
725 	if (IN_CLASSA(i))
726 		ia->ia_netmask = IN_CLASSA_NET;
727 	else if (IN_CLASSB(i))
728 		ia->ia_netmask = IN_CLASSB_NET;
729 	else
730 		ia->ia_netmask = IN_CLASSC_NET;
731 	/*
732 	 * The subnet mask usually includes at least the standard network part,
733 	 * but may may be smaller in the case of supernetting.
734 	 * If it is set, we believe it.
735 	 */
736 	if (ia->ia_subnetmask == 0) {
737 		ia->ia_subnetmask = ia->ia_netmask;
738 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
739 	} else
740 		ia->ia_netmask &= ia->ia_subnetmask;
741 	ia->ia_net = i & ia->ia_netmask;
742 	ia->ia_subnet = i & ia->ia_subnetmask;
743 	in_socktrim(&ia->ia_sockmask);
744 #ifdef DEV_CARP
745 	/*
746 	 * XXX: carp(4) does not have interface route
747 	 */
748 	if (ifp->if_type == IFT_CARP)
749 		return (0);
750 #endif
751 	/*
752 	 * Add route for the network.
753 	 */
754 	ia->ia_ifa.ifa_metric = ifp->if_metric;
755 	if (ifp->if_flags & IFF_BROADCAST) {
756 		ia->ia_broadaddr.sin_addr.s_addr =
757 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
758 		ia->ia_netbroadcast.s_addr =
759 			htonl(ia->ia_net | ~ ia->ia_netmask);
760 	} else if (ifp->if_flags & IFF_LOOPBACK) {
761 		ia->ia_dstaddr = ia->ia_addr;
762 		flags |= RTF_HOST;
763 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
764 		if (ia->ia_dstaddr.sin_family != AF_INET)
765 			return (0);
766 		flags |= RTF_HOST;
767 	}
768 	if ((error = in_addprefix(ia, flags)) != 0)
769 		return (error);
770 
771 	/*
772 	 * If the interface supports multicast, join the "all hosts"
773 	 * multicast group on that interface.
774 	 */
775 	if (ifp->if_flags & IFF_MULTICAST) {
776 		struct in_addr addr;
777 
778 		addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
779 		in_addmulti(&addr, ifp);
780 	}
781 	return (error);
782 }
783 
784 #define rtinitflags(x) \
785 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
786 	    ? RTF_HOST : 0)
787 /*
788  * Check if we have a route for the given prefix already or add a one
789  * accordingly.
790  */
791 static int
792 in_addprefix(target, flags)
793 	struct in_ifaddr *target;
794 	int flags;
795 {
796 	struct in_ifaddr *ia;
797 	struct in_addr prefix, mask, p, m;
798 	int error;
799 
800 	if ((flags & RTF_HOST) != 0)
801 		prefix = target->ia_dstaddr.sin_addr;
802 	else {
803 		prefix = target->ia_addr.sin_addr;
804 		mask = target->ia_sockmask.sin_addr;
805 		prefix.s_addr &= mask.s_addr;
806 	}
807 
808 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
809 		if (rtinitflags(ia)) {
810 			p = ia->ia_addr.sin_addr;
811 
812 			if (prefix.s_addr != p.s_addr)
813 				continue;
814 		} else {
815 			p = ia->ia_addr.sin_addr;
816 			m = ia->ia_sockmask.sin_addr;
817 			p.s_addr &= m.s_addr;
818 
819 			if (prefix.s_addr != p.s_addr ||
820 			    mask.s_addr != m.s_addr)
821 				continue;
822 		}
823 
824 		/*
825 		 * If we got a matching prefix route inserted by other
826 		 * interface address, we are done here.
827 		 */
828 		if (ia->ia_flags & IFA_ROUTE) {
829 			if (sameprefixcarponly &&
830 			    target->ia_ifp->if_type != IFT_CARP &&
831 			    ia->ia_ifp->if_type != IFT_CARP)
832 				return (EEXIST);
833 			else
834 				return (0);
835 		}
836 	}
837 
838 	/*
839 	 * No-one seem to have this prefix route, so we try to insert it.
840 	 */
841 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
842 	if (!error)
843 		target->ia_flags |= IFA_ROUTE;
844 	return error;
845 }
846 
847 /*
848  * If there is no other address in the system that can serve a route to the
849  * same prefix, remove the route.  Hand over the route to the new address
850  * otherwise.
851  */
852 static int
853 in_scrubprefix(target)
854 	struct in_ifaddr *target;
855 {
856 	struct in_ifaddr *ia;
857 	struct in_addr prefix, mask, p;
858 	int error;
859 
860 	if ((target->ia_flags & IFA_ROUTE) == 0)
861 		return 0;
862 
863 	if (rtinitflags(target))
864 		prefix = target->ia_dstaddr.sin_addr;
865 	else {
866 		prefix = target->ia_addr.sin_addr;
867 		mask = target->ia_sockmask.sin_addr;
868 		prefix.s_addr &= mask.s_addr;
869 	}
870 
871 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
872 		if (rtinitflags(ia))
873 			p = ia->ia_dstaddr.sin_addr;
874 		else {
875 			p = ia->ia_addr.sin_addr;
876 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
877 		}
878 
879 		if (prefix.s_addr != p.s_addr)
880 			continue;
881 
882 		/*
883 		 * If we got a matching prefix address, move IFA_ROUTE and
884 		 * the route itself to it.  Make sure that routing daemons
885 		 * get a heads-up.
886 		 *
887 		 * XXX: a special case for carp(4) interface
888 		 */
889 		if ((ia->ia_flags & IFA_ROUTE) == 0
890 #ifdef DEV_CARP
891 		    && (ia->ia_ifp->if_type != IFT_CARP)
892 #endif
893 							) {
894 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
895 			    rtinitflags(target));
896 			target->ia_flags &= ~IFA_ROUTE;
897 
898 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
899 			    rtinitflags(ia) | RTF_UP);
900 			if (error == 0)
901 				ia->ia_flags |= IFA_ROUTE;
902 			return error;
903 		}
904 	}
905 
906 	/*
907 	 * As no-one seem to have this prefix, we can remove the route.
908 	 */
909 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
910 	target->ia_flags &= ~IFA_ROUTE;
911 	return 0;
912 }
913 
914 #undef rtinitflags
915 
916 /*
917  * Return 1 if the address might be a local broadcast address.
918  */
919 int
920 in_broadcast(in, ifp)
921 	struct in_addr in;
922 	struct ifnet *ifp;
923 {
924 	register struct ifaddr *ifa;
925 	u_long t;
926 
927 	if (in.s_addr == INADDR_BROADCAST ||
928 	    in.s_addr == INADDR_ANY)
929 		return 1;
930 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
931 		return 0;
932 	t = ntohl(in.s_addr);
933 	/*
934 	 * Look through the list of addresses for a match
935 	 * with a broadcast address.
936 	 */
937 #define ia ((struct in_ifaddr *)ifa)
938 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
939 		if (ifa->ifa_addr->sa_family == AF_INET &&
940 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
941 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
942 		     /*
943 		      * Check for old-style (host 0) broadcast.
944 		      */
945 		     t == ia->ia_subnet || t == ia->ia_net) &&
946 		     /*
947 		      * Check for an all one subnetmask. These
948 		      * only exist when an interface gets a secondary
949 		      * address.
950 		      */
951 		     ia->ia_subnetmask != (u_long)0xffffffff)
952 			    return 1;
953 	return (0);
954 #undef ia
955 }
956 /*
957  * Add an address to the list of IP multicast addresses for a given interface.
958  */
959 struct in_multi *
960 in_addmulti(ap, ifp)
961 	register struct in_addr *ap;
962 	register struct ifnet *ifp;
963 {
964 	register struct in_multi *inm;
965 	int error;
966 	struct sockaddr_in sin;
967 	struct ifmultiaddr *ifma;
968 
969 	IFF_LOCKGIANT(ifp);
970 	IN_MULTI_LOCK();
971 	/*
972 	 * Call generic routine to add membership or increment
973 	 * refcount.  It wants addresses in the form of a sockaddr,
974 	 * so we build one here (being careful to zero the unused bytes).
975 	 */
976 	bzero(&sin, sizeof sin);
977 	sin.sin_family = AF_INET;
978 	sin.sin_len = sizeof sin;
979 	sin.sin_addr = *ap;
980 	error = if_addmulti(ifp, (struct sockaddr *)&sin, &ifma);
981 	if (error) {
982 		IN_MULTI_UNLOCK();
983 		IFF_UNLOCKGIANT(ifp);
984 		return 0;
985 	}
986 
987 	/*
988 	 * If ifma->ifma_protospec is null, then if_addmulti() created
989 	 * a new record.  Otherwise, we are done.
990 	 */
991 	if (ifma->ifma_protospec != NULL) {
992 		IN_MULTI_UNLOCK();
993 		IFF_UNLOCKGIANT(ifp);
994 		return ifma->ifma_protospec;
995 	}
996 
997 	inm = (struct in_multi *)malloc(sizeof(*inm), M_IPMADDR,
998 	    M_NOWAIT | M_ZERO);
999 	if (inm == NULL) {
1000 		IN_MULTI_UNLOCK();
1001 		IFF_UNLOCKGIANT(ifp);
1002 		return (NULL);
1003 	}
1004 
1005 	inm->inm_addr = *ap;
1006 	inm->inm_ifp = ifp;
1007 	inm->inm_ifma = ifma;
1008 	ifma->ifma_protospec = inm;
1009 	LIST_INSERT_HEAD(&in_multihead, inm, inm_link);
1010 
1011 	/*
1012 	 * Let IGMP know that we have joined a new IP multicast group.
1013 	 */
1014 	igmp_joingroup(inm);
1015 	IN_MULTI_UNLOCK();
1016 	IFF_UNLOCKGIANT(ifp);
1017 	return (inm);
1018 }
1019 
1020 /*
1021  * Delete a multicast address record.
1022  */
1023 void
1024 in_delmulti(inm)
1025 	register struct in_multi *inm;
1026 {
1027 	struct ifmultiaddr *ifma;
1028 	struct in_multi my_inm;
1029 	struct ifnet *ifp;
1030 
1031 	ifp = inm->inm_ifp;
1032 	IFF_LOCKGIANT(ifp);
1033 	IN_MULTI_LOCK();
1034 	ifma = inm->inm_ifma;
1035 	my_inm.inm_ifp = NULL ; /* don't send the leave msg */
1036 	if (ifma->ifma_refcount == 1) {
1037 		/*
1038 		 * No remaining claims to this record; let IGMP know that
1039 		 * we are leaving the multicast group.
1040 		 * But do it after the if_delmulti() which might reset
1041 		 * the interface and nuke the packet.
1042 		 */
1043 		my_inm = *inm ;
1044 		ifma->ifma_protospec = NULL;
1045 		LIST_REMOVE(inm, inm_link);
1046 		free(inm, M_IPMADDR);
1047 	}
1048 	/* XXX - should be separate API for when we have an ifma? */
1049 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1050 	if (my_inm.inm_ifp != NULL)
1051 		igmp_leavegroup(&my_inm);
1052 	IN_MULTI_UNLOCK();
1053 	IFF_UNLOCKGIANT(ifp);
1054 }
1055 
1056 /*
1057  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1058  */
1059 void
1060 in_ifdetach(ifp)
1061 	struct ifnet *ifp;
1062 {
1063 
1064 	in_pcbpurgeif0(&ripcbinfo, ifp);
1065 	in_pcbpurgeif0(&udbinfo, ifp);
1066 }
1067