xref: /freebsd/sys/netinet/in.c (revision d8b878873e7aa8df1972cc6a642804b17eb61087)
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 #include "opt_mpath.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/sockio.h>
42 #include <sys/malloc.h>
43 #include <sys/priv.h>
44 #include <sys/socket.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/proc.h>
48 #include <sys/sysctl.h>
49 #include <sys/syslog.h>
50 
51 #include <net/if.h>
52 #include <net/if_var.h>
53 #include <net/if_dl.h>
54 #include <net/if_llatbl.h>
55 #include <net/if_types.h>
56 #include <net/route.h>
57 #include <net/vnet.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/in_var.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/igmp_var.h>
64 #include <netinet/udp.h>
65 #include <netinet/udp_var.h>
66 
67 static int in_mask2len(struct in_addr *);
68 static void in_len2mask(struct in_addr *, int);
69 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
70 	struct ifnet *, struct thread *);
71 
72 static int	in_addprefix(struct in_ifaddr *, int);
73 static int	in_scrubprefix(struct in_ifaddr *);
74 static void	in_socktrim(struct sockaddr_in *);
75 static int	in_ifinit(struct ifnet *,
76 	    struct in_ifaddr *, struct sockaddr_in *, int);
77 static void	in_purgemaddrs(struct ifnet *);
78 
79 static VNET_DEFINE(int, subnetsarelocal);
80 static VNET_DEFINE(int, sameprefixcarponly);
81 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
82 
83 #define	V_subnetsarelocal		VNET(subnetsarelocal)
84 #define	V_sameprefixcarponly		VNET(sameprefixcarponly)
85 #define	V_ripcbinfo			VNET(ripcbinfo)
86 
87 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
88 	&VNET_NAME(subnetsarelocal), 0,
89 	"Treat all subnets as directly connected");
90 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
91 	&VNET_NAME(sameprefixcarponly), 0,
92 	"Refuse to create same prefixes on different interfaces");
93 
94 /*
95  * Return 1 if an internet address is for a ``local'' host
96  * (one to which we have a connection).  If subnetsarelocal
97  * is true, this includes other subnets of the local net.
98  * Otherwise, it includes only the directly-connected (sub)nets.
99  */
100 int
101 in_localaddr(struct in_addr in)
102 {
103 	register u_long i = ntohl(in.s_addr);
104 	register struct in_ifaddr *ia;
105 
106 	IN_IFADDR_RLOCK();
107 	if (V_subnetsarelocal) {
108 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
109 			if ((i & ia->ia_netmask) == ia->ia_net) {
110 				IN_IFADDR_RUNLOCK();
111 				return (1);
112 			}
113 		}
114 	} else {
115 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
116 			if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
117 				IN_IFADDR_RUNLOCK();
118 				return (1);
119 			}
120 		}
121 	}
122 	IN_IFADDR_RUNLOCK();
123 	return (0);
124 }
125 
126 /*
127  * Return 1 if an internet address is for the local host and configured
128  * on one of its interfaces.
129  */
130 int
131 in_localip(struct in_addr in)
132 {
133 	struct in_ifaddr *ia;
134 
135 	IN_IFADDR_RLOCK();
136 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
137 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
138 			IN_IFADDR_RUNLOCK();
139 			return (1);
140 		}
141 	}
142 	IN_IFADDR_RUNLOCK();
143 	return (0);
144 }
145 
146 /*
147  * Determine whether an IP address is in a reserved set of addresses
148  * that may not be forwarded, or whether datagrams to that destination
149  * may be forwarded.
150  */
151 int
152 in_canforward(struct in_addr in)
153 {
154 	register u_long i = ntohl(in.s_addr);
155 	register u_long net;
156 
157 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
158 		return (0);
159 	if (IN_CLASSA(i)) {
160 		net = i & IN_CLASSA_NET;
161 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
162 			return (0);
163 	}
164 	return (1);
165 }
166 
167 /*
168  * Trim a mask in a sockaddr
169  */
170 static void
171 in_socktrim(struct sockaddr_in *ap)
172 {
173     register char *cplim = (char *) &ap->sin_addr;
174     register char *cp = (char *) (&ap->sin_addr + 1);
175 
176     ap->sin_len = 0;
177     while (--cp >= cplim)
178 	if (*cp) {
179 	    (ap)->sin_len = cp - (char *) (ap) + 1;
180 	    break;
181 	}
182 }
183 
184 static int
185 in_mask2len(mask)
186 	struct in_addr *mask;
187 {
188 	int x, y;
189 	u_char *p;
190 
191 	p = (u_char *)mask;
192 	for (x = 0; x < sizeof(*mask); x++) {
193 		if (p[x] != 0xff)
194 			break;
195 	}
196 	y = 0;
197 	if (x < sizeof(*mask)) {
198 		for (y = 0; y < 8; y++) {
199 			if ((p[x] & (0x80 >> y)) == 0)
200 				break;
201 		}
202 	}
203 	return (x * 8 + y);
204 }
205 
206 static void
207 in_len2mask(struct in_addr *mask, int len)
208 {
209 	int i;
210 	u_char *p;
211 
212 	p = (u_char *)mask;
213 	bzero(mask, sizeof(*mask));
214 	for (i = 0; i < len / 8; i++)
215 		p[i] = 0xff;
216 	if (len % 8)
217 		p[i] = (0xff00 >> (len % 8)) & 0xff;
218 }
219 
220 /*
221  * Generic internet control operations (ioctl's).
222  *
223  * ifp is NULL if not an interface-specific ioctl.
224  */
225 /* ARGSUSED */
226 int
227 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
228     struct thread *td)
229 {
230 	register struct ifreq *ifr = (struct ifreq *)data;
231 	register struct in_ifaddr *ia, *iap;
232 	register struct ifaddr *ifa;
233 	struct in_addr allhosts_addr;
234 	struct in_addr dst;
235 	struct in_ifinfo *ii;
236 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
237 	struct sockaddr_in oldaddr;
238 	int error, hostIsNew, iaIsNew, maskIsNew;
239 	int iaIsFirst;
240 
241 	ia = NULL;
242 	iaIsFirst = 0;
243 	iaIsNew = 0;
244 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
245 
246 	/*
247 	 * Filter out ioctls we implement directly; forward the rest on to
248 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
249 	 */
250 	switch (cmd) {
251 	case SIOCAIFADDR:
252 	case SIOCDIFADDR:
253 	case SIOCGIFADDR:
254 	case SIOCGIFBRDADDR:
255 	case SIOCGIFDSTADDR:
256 	case SIOCGIFNETMASK:
257 	case SIOCSIFADDR:
258 	case SIOCSIFBRDADDR:
259 	case SIOCSIFDSTADDR:
260 	case SIOCSIFNETMASK:
261 		break;
262 
263 	case SIOCALIFADDR:
264 		if (td != NULL) {
265 			error = priv_check(td, PRIV_NET_ADDIFADDR);
266 			if (error)
267 				return (error);
268 		}
269 		if (ifp == NULL)
270 			return (EINVAL);
271 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
272 
273 	case SIOCDLIFADDR:
274 		if (td != NULL) {
275 			error = priv_check(td, PRIV_NET_DELIFADDR);
276 			if (error)
277 				return (error);
278 		}
279 		if (ifp == NULL)
280 			return (EINVAL);
281 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
282 
283 	case SIOCGLIFADDR:
284 		if (ifp == NULL)
285 			return (EINVAL);
286 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
287 
288 	default:
289 		if (ifp == NULL || ifp->if_ioctl == NULL)
290 			return (EOPNOTSUPP);
291 		return ((*ifp->if_ioctl)(ifp, cmd, data));
292 	}
293 
294 	if (ifp == NULL)
295 		return (EADDRNOTAVAIL);
296 
297 	/*
298 	 * Security checks before we get involved in any work.
299 	 */
300 	switch (cmd) {
301 	case SIOCAIFADDR:
302 	case SIOCSIFADDR:
303 	case SIOCSIFBRDADDR:
304 	case SIOCSIFNETMASK:
305 	case SIOCSIFDSTADDR:
306 		if (td != NULL) {
307 			error = priv_check(td, PRIV_NET_ADDIFADDR);
308 			if (error)
309 				return (error);
310 		}
311 		break;
312 
313 	case SIOCDIFADDR:
314 		if (td != NULL) {
315 			error = priv_check(td, PRIV_NET_DELIFADDR);
316 			if (error)
317 				return (error);
318 		}
319 		break;
320 	}
321 
322 	/*
323 	 * Find address for this interface, if it exists.
324 	 *
325 	 * If an alias address was specified, find that one instead of the
326 	 * first one on the interface, if possible.
327 	 */
328 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
329 	IN_IFADDR_RLOCK();
330 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
331 		if (iap->ia_ifp == ifp &&
332 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
333 			if (td == NULL || prison_check_ip4(td->td_ucred,
334 			    &dst) == 0)
335 				ia = iap;
336 			break;
337 		}
338 	}
339 	if (ia != NULL)
340 		ifa_ref(&ia->ia_ifa);
341 	IN_IFADDR_RUNLOCK();
342 	if (ia == NULL) {
343 		IF_ADDR_LOCK(ifp);
344 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
345 			iap = ifatoia(ifa);
346 			if (iap->ia_addr.sin_family == AF_INET) {
347 				if (td != NULL &&
348 				    prison_check_ip4(td->td_ucred,
349 				    &iap->ia_addr.sin_addr) != 0)
350 					continue;
351 				ia = iap;
352 				break;
353 			}
354 		}
355 		if (ia != NULL)
356 			ifa_ref(&ia->ia_ifa);
357 		IF_ADDR_UNLOCK(ifp);
358 	}
359 	if (ia == NULL)
360 		iaIsFirst = 1;
361 
362 	error = 0;
363 	switch (cmd) {
364 	case SIOCAIFADDR:
365 	case SIOCDIFADDR:
366 		if (ifra->ifra_addr.sin_family == AF_INET) {
367 			struct in_ifaddr *oia;
368 
369 			IN_IFADDR_RLOCK();
370 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
371 				if (ia->ia_ifp == ifp  &&
372 				    ia->ia_addr.sin_addr.s_addr ==
373 				    ifra->ifra_addr.sin_addr.s_addr)
374 					break;
375 			}
376 			if (ia != NULL && ia != oia)
377 				ifa_ref(&ia->ia_ifa);
378 			if (oia != NULL && ia != oia)
379 				ifa_free(&oia->ia_ifa);
380 			IN_IFADDR_RUNLOCK();
381 			if ((ifp->if_flags & IFF_POINTOPOINT)
382 			    && (cmd == SIOCAIFADDR)
383 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
384 				== INADDR_ANY)) {
385 				error = EDESTADDRREQ;
386 				goto out;
387 			}
388 		}
389 		if (cmd == SIOCDIFADDR && ia == NULL) {
390 			error = EADDRNOTAVAIL;
391 			goto out;
392 		}
393 		/* FALLTHROUGH */
394 	case SIOCSIFADDR:
395 	case SIOCSIFNETMASK:
396 	case SIOCSIFDSTADDR:
397 		if (ia == NULL) {
398 			ia = (struct in_ifaddr *)
399 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
400 				    M_ZERO);
401 			if (ia == NULL) {
402 				error = ENOBUFS;
403 				goto out;
404 			}
405 
406 			ifa = &ia->ia_ifa;
407 			ifa_init(ifa);
408 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
409 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
410 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
411 
412 			ia->ia_sockmask.sin_len = 8;
413 			ia->ia_sockmask.sin_family = AF_INET;
414 			if (ifp->if_flags & IFF_BROADCAST) {
415 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
416 				ia->ia_broadaddr.sin_family = AF_INET;
417 			}
418 			ia->ia_ifp = ifp;
419 
420 			ifa_ref(ifa);			/* if_addrhead */
421 			IF_ADDR_LOCK(ifp);
422 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
423 			IF_ADDR_UNLOCK(ifp);
424 			ifa_ref(ifa);			/* in_ifaddrhead */
425 			IN_IFADDR_WLOCK();
426 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
427 			IN_IFADDR_WUNLOCK();
428 			iaIsNew = 1;
429 		}
430 		break;
431 
432 	case SIOCSIFBRDADDR:
433 	case SIOCGIFADDR:
434 	case SIOCGIFNETMASK:
435 	case SIOCGIFDSTADDR:
436 	case SIOCGIFBRDADDR:
437 		if (ia == NULL) {
438 			error = EADDRNOTAVAIL;
439 			goto out;
440 		}
441 		break;
442 	}
443 
444 	/*
445 	 * Most paths in this switch return directly or via out.  Only paths
446 	 * that remove the address break in order to hit common removal code.
447 	 */
448 	switch (cmd) {
449 	case SIOCGIFADDR:
450 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
451 		goto out;
452 
453 	case SIOCGIFBRDADDR:
454 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
455 			error = EINVAL;
456 			goto out;
457 		}
458 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
459 		goto out;
460 
461 	case SIOCGIFDSTADDR:
462 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
463 			error = EINVAL;
464 			goto out;
465 		}
466 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
467 		goto out;
468 
469 	case SIOCGIFNETMASK:
470 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
471 		goto out;
472 
473 	case SIOCSIFDSTADDR:
474 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
475 			error = EINVAL;
476 			goto out;
477 		}
478 		oldaddr = ia->ia_dstaddr;
479 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
480 		if (ifp->if_ioctl != NULL) {
481 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
482 			    (caddr_t)ia);
483 			if (error) {
484 				ia->ia_dstaddr = oldaddr;
485 				goto out;
486 			}
487 		}
488 		if (ia->ia_flags & IFA_ROUTE) {
489 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
490 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
491 			ia->ia_ifa.ifa_dstaddr =
492 					(struct sockaddr *)&ia->ia_dstaddr;
493 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
494 		}
495 		goto out;
496 
497 	case SIOCSIFBRDADDR:
498 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
499 			error = EINVAL;
500 			goto out;
501 		}
502 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
503 		goto out;
504 
505 	case SIOCSIFADDR:
506 		error = in_ifinit(ifp, ia,
507 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
508 		if (error != 0 && iaIsNew)
509 			break;
510 		if (error == 0) {
511 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
512 			if (iaIsFirst &&
513 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
514 				error = in_joingroup(ifp, &allhosts_addr,
515 				    NULL, &ii->ii_allhosts);
516 			}
517 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
518 		}
519 		error = 0;
520 		goto out;
521 
522 	case SIOCSIFNETMASK:
523 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
524 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
525 		goto out;
526 
527 	case SIOCAIFADDR:
528 		maskIsNew = 0;
529 		hostIsNew = 1;
530 		error = 0;
531 		if (ia->ia_addr.sin_family == AF_INET) {
532 			if (ifra->ifra_addr.sin_len == 0) {
533 				ifra->ifra_addr = ia->ia_addr;
534 				hostIsNew = 0;
535 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
536 					       ia->ia_addr.sin_addr.s_addr)
537 				hostIsNew = 0;
538 		}
539 		if (ifra->ifra_mask.sin_len) {
540 			/*
541 			 * QL: XXX
542 			 * Need to scrub the prefix here in case
543 			 * the issued command is SIOCAIFADDR with
544 			 * the same address, but with a different
545 			 * prefix length. And if the prefix length
546 			 * is the same as before, then the call is
547 			 * un-necessarily executed here.
548 			 */
549 			in_ifscrub(ifp, ia);
550 			ia->ia_sockmask = ifra->ifra_mask;
551 			ia->ia_sockmask.sin_family = AF_INET;
552 			ia->ia_subnetmask =
553 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
554 			maskIsNew = 1;
555 		}
556 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
557 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
558 			in_ifscrub(ifp, ia);
559 			ia->ia_dstaddr = ifra->ifra_dstaddr;
560 			maskIsNew  = 1; /* We lie; but the effect's the same */
561 		}
562 		if (ifra->ifra_addr.sin_family == AF_INET &&
563 		    (hostIsNew || maskIsNew))
564 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
565 		if (error != 0 && iaIsNew)
566 			break;
567 
568 		if ((ifp->if_flags & IFF_BROADCAST) &&
569 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
570 			ia->ia_broadaddr = ifra->ifra_broadaddr;
571 		if (error == 0) {
572 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
573 			if (iaIsFirst &&
574 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
575 				error = in_joingroup(ifp, &allhosts_addr,
576 				    NULL, &ii->ii_allhosts);
577 			}
578 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
579 		}
580 		goto out;
581 
582 	case SIOCDIFADDR:
583 		/*
584 		 * in_ifscrub kills the interface route.
585 		 */
586 		in_ifscrub(ifp, ia);
587 
588 		/*
589 		 * in_ifadown gets rid of all the rest of
590 		 * the routes.  This is not quite the right
591 		 * thing to do, but at least if we are running
592 		 * a routing process they will come back.
593 		 */
594 		in_ifadown(&ia->ia_ifa, 1);
595 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
596 		error = 0;
597 		break;
598 
599 	default:
600 		panic("in_control: unsupported ioctl");
601 	}
602 
603 	IF_ADDR_LOCK(ifp);
604 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
605 	IF_ADDR_UNLOCK(ifp);
606 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
607 
608 	IN_IFADDR_WLOCK();
609 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
610 	if (ia->ia_addr.sin_family == AF_INET) {
611 		struct in_ifaddr *if_ia;
612 
613 		LIST_REMOVE(ia, ia_hash);
614 		IN_IFADDR_WUNLOCK();
615 		/*
616 		 * If this is the last IPv4 address configured on this
617 		 * interface, leave the all-hosts group.
618 		 * No state-change report need be transmitted.
619 		 */
620 		if_ia = NULL;
621 		IFP_TO_IA(ifp, if_ia);
622 		if (if_ia == NULL) {
623 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
624 			IN_MULTI_LOCK();
625 			if (ii->ii_allhosts) {
626 				(void)in_leavegroup_locked(ii->ii_allhosts,
627 				    NULL);
628 				ii->ii_allhosts = NULL;
629 			}
630 			IN_MULTI_UNLOCK();
631 		} else
632 			ifa_free(&if_ia->ia_ifa);
633 	} else
634 		IN_IFADDR_WUNLOCK();
635 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
636 out:
637 	if (ia != NULL)
638 		ifa_free(&ia->ia_ifa);
639 	return (error);
640 }
641 
642 /*
643  * SIOC[GAD]LIFADDR.
644  *	SIOCGLIFADDR: get first address. (?!?)
645  *	SIOCGLIFADDR with IFLR_PREFIX:
646  *		get first address that matches the specified prefix.
647  *	SIOCALIFADDR: add the specified address.
648  *	SIOCALIFADDR with IFLR_PREFIX:
649  *		EINVAL since we can't deduce hostid part of the address.
650  *	SIOCDLIFADDR: delete the specified address.
651  *	SIOCDLIFADDR with IFLR_PREFIX:
652  *		delete the first address that matches the specified prefix.
653  * return values:
654  *	EINVAL on invalid parameters
655  *	EADDRNOTAVAIL on prefix match failed/specified address not found
656  *	other values may be returned from in_ioctl()
657  */
658 static int
659 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
660     struct ifnet *ifp, struct thread *td)
661 {
662 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
663 	struct ifaddr *ifa;
664 
665 	/* sanity checks */
666 	if (data == NULL || ifp == NULL) {
667 		panic("invalid argument to in_lifaddr_ioctl");
668 		/*NOTRECHED*/
669 	}
670 
671 	switch (cmd) {
672 	case SIOCGLIFADDR:
673 		/* address must be specified on GET with IFLR_PREFIX */
674 		if ((iflr->flags & IFLR_PREFIX) == 0)
675 			break;
676 		/*FALLTHROUGH*/
677 	case SIOCALIFADDR:
678 	case SIOCDLIFADDR:
679 		/* address must be specified on ADD and DELETE */
680 		if (iflr->addr.ss_family != AF_INET)
681 			return (EINVAL);
682 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
683 			return (EINVAL);
684 		/* XXX need improvement */
685 		if (iflr->dstaddr.ss_family
686 		 && iflr->dstaddr.ss_family != AF_INET)
687 			return (EINVAL);
688 		if (iflr->dstaddr.ss_family
689 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
690 			return (EINVAL);
691 		break;
692 	default: /*shouldn't happen*/
693 		return (EOPNOTSUPP);
694 	}
695 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
696 		return (EINVAL);
697 
698 	switch (cmd) {
699 	case SIOCALIFADDR:
700 	    {
701 		struct in_aliasreq ifra;
702 
703 		if (iflr->flags & IFLR_PREFIX)
704 			return (EINVAL);
705 
706 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
707 		bzero(&ifra, sizeof(ifra));
708 		bcopy(iflr->iflr_name, ifra.ifra_name,
709 			sizeof(ifra.ifra_name));
710 
711 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
712 
713 		if (iflr->dstaddr.ss_family) {	/*XXX*/
714 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
715 				iflr->dstaddr.ss_len);
716 		}
717 
718 		ifra.ifra_mask.sin_family = AF_INET;
719 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
720 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
721 
722 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
723 	    }
724 	case SIOCGLIFADDR:
725 	case SIOCDLIFADDR:
726 	    {
727 		struct in_ifaddr *ia;
728 		struct in_addr mask, candidate, match;
729 		struct sockaddr_in *sin;
730 
731 		bzero(&mask, sizeof(mask));
732 		bzero(&match, sizeof(match));
733 		if (iflr->flags & IFLR_PREFIX) {
734 			/* lookup a prefix rather than address. */
735 			in_len2mask(&mask, iflr->prefixlen);
736 
737 			sin = (struct sockaddr_in *)&iflr->addr;
738 			match.s_addr = sin->sin_addr.s_addr;
739 			match.s_addr &= mask.s_addr;
740 
741 			/* if you set extra bits, that's wrong */
742 			if (match.s_addr != sin->sin_addr.s_addr)
743 				return (EINVAL);
744 
745 		} else {
746 			/* on getting an address, take the 1st match */
747 			/* on deleting an address, do exact match */
748 			if (cmd != SIOCGLIFADDR) {
749 				in_len2mask(&mask, 32);
750 				sin = (struct sockaddr_in *)&iflr->addr;
751 				match.s_addr = sin->sin_addr.s_addr;
752 			}
753 		}
754 
755 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
756 			if (ifa->ifa_addr->sa_family != AF_INET6)
757 				continue;
758 			if (match.s_addr == 0)
759 				break;
760 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
761 			candidate.s_addr &= mask.s_addr;
762 			if (candidate.s_addr == match.s_addr)
763 				break;
764 		}
765 		if (ifa == NULL)
766 			return (EADDRNOTAVAIL);
767 		ia = (struct in_ifaddr *)ifa;
768 
769 		if (cmd == SIOCGLIFADDR) {
770 			/* fill in the if_laddrreq structure */
771 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
772 
773 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
774 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
775 					ia->ia_dstaddr.sin_len);
776 			} else
777 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
778 
779 			iflr->prefixlen =
780 				in_mask2len(&ia->ia_sockmask.sin_addr);
781 
782 			iflr->flags = 0;	/*XXX*/
783 
784 			return (0);
785 		} else {
786 			struct in_aliasreq ifra;
787 
788 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
789 			bzero(&ifra, sizeof(ifra));
790 			bcopy(iflr->iflr_name, ifra.ifra_name,
791 				sizeof(ifra.ifra_name));
792 
793 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
794 				ia->ia_addr.sin_len);
795 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
796 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
797 					ia->ia_dstaddr.sin_len);
798 			}
799 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
800 				ia->ia_sockmask.sin_len);
801 
802 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
803 			    ifp, td));
804 		}
805 	    }
806 	}
807 
808 	return (EOPNOTSUPP);	/*just for safety*/
809 }
810 
811 /*
812  * Delete any existing route for an interface.
813  */
814 void
815 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
816 {
817 
818 	in_scrubprefix(ia);
819 }
820 
821 /*
822  * Initialize an interface's internet address
823  * and routing table entry.
824  */
825 static int
826 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
827     int scrub)
828 {
829 	register u_long i = ntohl(sin->sin_addr.s_addr);
830 	struct sockaddr_in oldaddr;
831 	int s = splimp(), flags = RTF_UP, error = 0;
832 
833 	oldaddr = ia->ia_addr;
834 	if (oldaddr.sin_family == AF_INET)
835 		LIST_REMOVE(ia, ia_hash);
836 	ia->ia_addr = *sin;
837 	if (ia->ia_addr.sin_family == AF_INET) {
838 		IN_IFADDR_WLOCK();
839 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
840 		    ia, ia_hash);
841 		IN_IFADDR_WUNLOCK();
842 	}
843 	/*
844 	 * Give the interface a chance to initialize
845 	 * if this is its first address,
846 	 * and to validate the address if necessary.
847 	 */
848 	if (ifp->if_ioctl != NULL) {
849 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
850 		if (error) {
851 			splx(s);
852 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
853 			ia->ia_addr = oldaddr;
854 			IN_IFADDR_WLOCK();
855 			if (ia->ia_addr.sin_family == AF_INET)
856 				LIST_INSERT_HEAD(INADDR_HASH(
857 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
858 			else
859 				/*
860 				 * If oldaddr family is not AF_INET (e.g.
861 				 * interface has been just created) in_control
862 				 * does not call LIST_REMOVE, and we end up
863 				 * with bogus ia entries in hash
864 				 */
865 				LIST_REMOVE(ia, ia_hash);
866 			IN_IFADDR_WUNLOCK();
867 			return (error);
868 		}
869 	}
870 	splx(s);
871 	if (scrub) {
872 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
873 		in_ifscrub(ifp, ia);
874 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
875 	}
876 	if (IN_CLASSA(i))
877 		ia->ia_netmask = IN_CLASSA_NET;
878 	else if (IN_CLASSB(i))
879 		ia->ia_netmask = IN_CLASSB_NET;
880 	else
881 		ia->ia_netmask = IN_CLASSC_NET;
882 	/*
883 	 * The subnet mask usually includes at least the standard network part,
884 	 * but may may be smaller in the case of supernetting.
885 	 * If it is set, we believe it.
886 	 */
887 	if (ia->ia_subnetmask == 0) {
888 		ia->ia_subnetmask = ia->ia_netmask;
889 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
890 	} else
891 		ia->ia_netmask &= ia->ia_subnetmask;
892 	ia->ia_net = i & ia->ia_netmask;
893 	ia->ia_subnet = i & ia->ia_subnetmask;
894 	in_socktrim(&ia->ia_sockmask);
895 #ifdef DEV_CARP
896 	/*
897 	 * XXX: carp(4) does not have interface route
898 	 */
899 	if (ifp->if_type == IFT_CARP)
900 		return (0);
901 #endif
902 	/*
903 	 * Add route for the network.
904 	 */
905 	ia->ia_ifa.ifa_metric = ifp->if_metric;
906 	if (ifp->if_flags & IFF_BROADCAST) {
907 		ia->ia_broadaddr.sin_addr.s_addr =
908 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
909 		ia->ia_netbroadcast.s_addr =
910 			htonl(ia->ia_net | ~ ia->ia_netmask);
911 	} else if (ifp->if_flags & IFF_LOOPBACK) {
912 		ia->ia_dstaddr = ia->ia_addr;
913 		flags |= RTF_HOST;
914 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
915 		if (ia->ia_dstaddr.sin_family != AF_INET)
916 			return (0);
917 		flags |= RTF_HOST;
918 	}
919 	if ((error = in_addprefix(ia, flags)) != 0)
920 		return (error);
921 
922 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
923 		return (0);
924 
925 	if (ifp->if_flags & IFF_POINTOPOINT) {
926 		if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
927 			return (0);
928 	}
929 
930 
931 	/*
932 	 * add a loopback route to self
933 	 */
934 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
935 		struct route ia_ro;
936 
937 		bzero(&ia_ro, sizeof(ia_ro));
938 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
939 		rtalloc_ign_fib(&ia_ro, 0, 0);
940 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
941 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
942 			RT_LOCK(ia_ro.ro_rt);
943 			RT_ADDREF(ia_ro.ro_rt);
944 			RTFREE_LOCKED(ia_ro.ro_rt);
945 		} else
946 			error = ifa_add_loopback_route((struct ifaddr *)ia,
947 				       (struct sockaddr *)&ia->ia_addr);
948 		if (error == 0)
949 			ia->ia_flags |= IFA_RTSELF;
950 		if (ia_ro.ro_rt != NULL)
951 			RTFREE(ia_ro.ro_rt);
952 	}
953 
954 	return (error);
955 }
956 
957 #define rtinitflags(x) \
958 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
959 	    ? RTF_HOST : 0)
960 
961 /*
962  * Generate a routing message when inserting or deleting
963  * an interface address alias.
964  */
965 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
966     struct in_ifaddr *target)
967 {
968 	struct route pfx_ro;
969 	struct sockaddr_in *pfx_addr;
970 	struct rtentry msg_rt;
971 
972 	/* QL: XXX
973 	 * This is a bit questionable because there is no
974 	 * additional route entry added/deleted for an address
975 	 * alias. Therefore this route report is inaccurate.
976 	 */
977 	bzero(&pfx_ro, sizeof(pfx_ro));
978 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
979 	pfx_addr->sin_len = sizeof(*pfx_addr);
980 	pfx_addr->sin_family = AF_INET;
981 	pfx_addr->sin_addr = *prefix;
982 	rtalloc_ign_fib(&pfx_ro, 0, 0);
983 	if (pfx_ro.ro_rt != NULL) {
984 		msg_rt = *pfx_ro.ro_rt;
985 
986 		/* QL: XXX
987 		 * Point the gateway to the new interface
988 		 * address as if a new prefix route entry has
989 		 * been added through the new address alias.
990 		 * All other parts of the rtentry is accurate,
991 		 * e.g., rt_key, rt_mask, rt_ifp etc.
992 		 */
993 		msg_rt.rt_gateway =
994 			(struct sockaddr *)&target->ia_addr;
995 		rt_newaddrmsg(cmd,
996 			      (struct ifaddr *)target,
997 			      0, &msg_rt);
998 		RTFREE(pfx_ro.ro_rt);
999 	}
1000 	return;
1001 }
1002 
1003 /*
1004  * Check if we have a route for the given prefix already or add one accordingly.
1005  */
1006 static int
1007 in_addprefix(struct in_ifaddr *target, int flags)
1008 {
1009 	struct in_ifaddr *ia;
1010 	struct in_addr prefix, mask, p, m;
1011 	int error;
1012 
1013 	if ((flags & RTF_HOST) != 0) {
1014 		prefix = target->ia_dstaddr.sin_addr;
1015 		mask.s_addr = 0;
1016 	} else {
1017 		prefix = target->ia_addr.sin_addr;
1018 		mask = target->ia_sockmask.sin_addr;
1019 		prefix.s_addr &= mask.s_addr;
1020 	}
1021 
1022 	IN_IFADDR_RLOCK();
1023 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1024 		if (rtinitflags(ia)) {
1025 			p = ia->ia_addr.sin_addr;
1026 
1027 			if (prefix.s_addr != p.s_addr)
1028 				continue;
1029 		} else {
1030 			p = ia->ia_addr.sin_addr;
1031 			m = ia->ia_sockmask.sin_addr;
1032 			p.s_addr &= m.s_addr;
1033 
1034 			if (prefix.s_addr != p.s_addr ||
1035 			    mask.s_addr != m.s_addr)
1036 				continue;
1037 		}
1038 
1039 		/*
1040 		 * If we got a matching prefix route inserted by other
1041 		 * interface address, we are done here.
1042 		 */
1043 		if (ia->ia_flags & IFA_ROUTE) {
1044 #ifdef RADIX_MPATH
1045 			if (ia->ia_addr.sin_addr.s_addr ==
1046 			    target->ia_addr.sin_addr.s_addr)
1047 				return (EEXIST);
1048 			else
1049 				break;
1050 #endif
1051 			if (V_sameprefixcarponly &&
1052 			    target->ia_ifp->if_type != IFT_CARP &&
1053 			    ia->ia_ifp->if_type != IFT_CARP) {
1054 				IN_IFADDR_RUNLOCK();
1055 				return (EEXIST);
1056 			} else {
1057 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
1058 				IN_IFADDR_RUNLOCK();
1059 				return (0);
1060 			}
1061 		}
1062 	}
1063 	IN_IFADDR_RUNLOCK();
1064 
1065 	/*
1066 	 * No-one seem to have this prefix route, so we try to insert it.
1067 	 */
1068 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1069 	if (!error)
1070 		target->ia_flags |= IFA_ROUTE;
1071 	return (error);
1072 }
1073 
1074 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1075 
1076 /*
1077  * If there is no other address in the system that can serve a route to the
1078  * same prefix, remove the route.  Hand over the route to the new address
1079  * otherwise.
1080  */
1081 static int
1082 in_scrubprefix(struct in_ifaddr *target)
1083 {
1084 	struct in_ifaddr *ia;
1085 	struct in_addr prefix, mask, p;
1086 	int error = 0;
1087 	struct sockaddr_in prefix0, mask0;
1088 
1089 	/*
1090 	 * Remove the loopback route to the interface address.
1091 	 * The "useloopback" setting is not consulted because if the
1092 	 * user configures an interface address, turns off this
1093 	 * setting, and then tries to delete that interface address,
1094 	 * checking the current setting of "useloopback" would leave
1095 	 * that interface address loopback route untouched, which
1096 	 * would be wrong. Therefore the interface address loopback route
1097 	 * deletion is unconditional.
1098 	 */
1099 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1100 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1101 	    (target->ia_flags & IFA_RTSELF)) {
1102 		struct route ia_ro;
1103 		int freeit = 0;
1104 
1105 		bzero(&ia_ro, sizeof(ia_ro));
1106 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1107 		rtalloc_ign_fib(&ia_ro, 0, 0);
1108 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1109 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1110 			RT_LOCK(ia_ro.ro_rt);
1111 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1112 				freeit = 1;
1113 			else
1114 				RT_REMREF(ia_ro.ro_rt);
1115 			RTFREE_LOCKED(ia_ro.ro_rt);
1116 		}
1117 		if (freeit)
1118 			error = ifa_del_loopback_route((struct ifaddr *)target,
1119 				       (struct sockaddr *)&target->ia_addr);
1120 		if (error == 0)
1121 			target->ia_flags &= ~IFA_RTSELF;
1122 		/* remove arp cache */
1123 		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1124 	}
1125 
1126 	if (rtinitflags(target))
1127 		prefix = target->ia_dstaddr.sin_addr;
1128 	else {
1129 		prefix = target->ia_addr.sin_addr;
1130 		mask = target->ia_sockmask.sin_addr;
1131 		prefix.s_addr &= mask.s_addr;
1132 	}
1133 
1134 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1135 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1136 		return (0);
1137 	}
1138 
1139 	IN_IFADDR_RLOCK();
1140 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1141 		if (rtinitflags(ia))
1142 			p = ia->ia_dstaddr.sin_addr;
1143 		else {
1144 			p = ia->ia_addr.sin_addr;
1145 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1146 		}
1147 
1148 		if (prefix.s_addr != p.s_addr)
1149 			continue;
1150 
1151 		/*
1152 		 * If we got a matching prefix address, move IFA_ROUTE and
1153 		 * the route itself to it.  Make sure that routing daemons
1154 		 * get a heads-up.
1155 		 *
1156 		 * XXX: a special case for carp(4) interface
1157 		 */
1158 		if ((ia->ia_flags & IFA_ROUTE) == 0
1159 #ifdef DEV_CARP
1160 		    && (ia->ia_ifp->if_type != IFT_CARP)
1161 #endif
1162 							) {
1163 			IN_IFADDR_RUNLOCK();
1164 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1165 			    rtinitflags(target));
1166 			target->ia_flags &= ~IFA_ROUTE;
1167 
1168 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1169 			    rtinitflags(ia) | RTF_UP);
1170 			if (error == 0)
1171 				ia->ia_flags |= IFA_ROUTE;
1172 			return (error);
1173 		}
1174 	}
1175 	IN_IFADDR_RUNLOCK();
1176 
1177 	/*
1178 	 * remove all L2 entries on the given prefix
1179 	 */
1180 	bzero(&prefix0, sizeof(prefix0));
1181 	prefix0.sin_len = sizeof(prefix0);
1182 	prefix0.sin_family = AF_INET;
1183 	prefix0.sin_addr.s_addr = target->ia_subnet;
1184 	bzero(&mask0, sizeof(mask0));
1185 	mask0.sin_len = sizeof(mask0);
1186 	mask0.sin_family = AF_INET;
1187 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1188 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1189 			    (struct sockaddr *)&mask0);
1190 
1191 	/*
1192 	 * As no-one seem to have this prefix, we can remove the route.
1193 	 */
1194 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1195 	target->ia_flags &= ~IFA_ROUTE;
1196 	return (0);
1197 }
1198 
1199 #undef rtinitflags
1200 
1201 /*
1202  * Return 1 if the address might be a local broadcast address.
1203  */
1204 int
1205 in_broadcast(struct in_addr in, struct ifnet *ifp)
1206 {
1207 	register struct ifaddr *ifa;
1208 	u_long t;
1209 
1210 	if (in.s_addr == INADDR_BROADCAST ||
1211 	    in.s_addr == INADDR_ANY)
1212 		return (1);
1213 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1214 		return (0);
1215 	t = ntohl(in.s_addr);
1216 	/*
1217 	 * Look through the list of addresses for a match
1218 	 * with a broadcast address.
1219 	 */
1220 #define ia ((struct in_ifaddr *)ifa)
1221 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1222 		if (ifa->ifa_addr->sa_family == AF_INET &&
1223 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1224 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1225 		     /*
1226 		      * Check for old-style (host 0) broadcast.
1227 		      */
1228 		     t == ia->ia_subnet || t == ia->ia_net) &&
1229 		     /*
1230 		      * Check for an all one subnetmask. These
1231 		      * only exist when an interface gets a secondary
1232 		      * address.
1233 		      */
1234 		     ia->ia_subnetmask != (u_long)0xffffffff)
1235 			    return (1);
1236 	return (0);
1237 #undef ia
1238 }
1239 
1240 /*
1241  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1242  */
1243 void
1244 in_ifdetach(struct ifnet *ifp)
1245 {
1246 
1247 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1248 	in_pcbpurgeif0(&V_udbinfo, ifp);
1249 	in_purgemaddrs(ifp);
1250 }
1251 
1252 /*
1253  * Delete all IPv4 multicast address records, and associated link-layer
1254  * multicast address records, associated with ifp.
1255  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1256  * XXX This should not race with ifma_protospec being set during
1257  * a new allocation, if it does, we have bigger problems.
1258  */
1259 static void
1260 in_purgemaddrs(struct ifnet *ifp)
1261 {
1262 	LIST_HEAD(,in_multi) purgeinms;
1263 	struct in_multi		*inm, *tinm;
1264 	struct ifmultiaddr	*ifma;
1265 
1266 	LIST_INIT(&purgeinms);
1267 	IN_MULTI_LOCK();
1268 
1269 	/*
1270 	 * Extract list of in_multi associated with the detaching ifp
1271 	 * which the PF_INET layer is about to release.
1272 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1273 	 * by code further down.
1274 	 */
1275 	IF_ADDR_LOCK(ifp);
1276 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1277 		if (ifma->ifma_addr->sa_family != AF_INET ||
1278 		    ifma->ifma_protospec == NULL)
1279 			continue;
1280 #if 0
1281 		KASSERT(ifma->ifma_protospec != NULL,
1282 		    ("%s: ifma_protospec is NULL", __func__));
1283 #endif
1284 		inm = (struct in_multi *)ifma->ifma_protospec;
1285 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1286 	}
1287 	IF_ADDR_UNLOCK(ifp);
1288 
1289 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1290 		LIST_REMOVE(inm, inm_link);
1291 		inm_release_locked(inm);
1292 	}
1293 	igmp_ifdetach(ifp);
1294 
1295 	IN_MULTI_UNLOCK();
1296 }
1297 
1298 #include <net/if_dl.h>
1299 #include <netinet/if_ether.h>
1300 
1301 struct in_llentry {
1302 	struct llentry		base;
1303 	struct sockaddr_in	l3_addr4;
1304 };
1305 
1306 static struct llentry *
1307 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1308 {
1309 	struct in_llentry *lle;
1310 
1311 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1312 	if (lle == NULL)		/* NB: caller generates msg */
1313 		return NULL;
1314 
1315 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1316 	/*
1317 	 * For IPv4 this will trigger "arpresolve" to generate
1318 	 * an ARP request.
1319 	 */
1320 	lle->base.la_expire = time_second; /* mark expired */
1321 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1322 	lle->base.lle_refcnt = 1;
1323 	LLE_LOCK_INIT(&lle->base);
1324 	return &lle->base;
1325 }
1326 
1327 /*
1328  * Deletes an address from the address table.
1329  * This function is called by the timer functions
1330  * such as arptimer() and nd6_llinfo_timer(), and
1331  * the caller does the locking.
1332  */
1333 static void
1334 in_lltable_free(struct lltable *llt, struct llentry *lle)
1335 {
1336 	LLE_WUNLOCK(lle);
1337 	LLE_LOCK_DESTROY(lle);
1338 	free(lle, M_LLTABLE);
1339 }
1340 
1341 
1342 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1343 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1344 
1345 static void
1346 in_lltable_prefix_free(struct lltable *llt,
1347 		       const struct sockaddr *prefix,
1348 		       const struct sockaddr *mask)
1349 {
1350 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1351 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1352 	struct llentry *lle, *next;
1353 	register int i;
1354 
1355 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1356 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1357 
1358 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1359 						     pfx, msk)) {
1360 				int canceled;
1361 
1362 				canceled = callout_drain(&lle->la_timer);
1363 				LLE_WLOCK(lle);
1364 				if (canceled)
1365 					LLE_REMREF(lle);
1366 				llentry_free(lle);
1367 			}
1368 		}
1369 	}
1370 }
1371 
1372 
1373 static int
1374 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1375 {
1376 	struct rtentry *rt;
1377 
1378 	KASSERT(l3addr->sa_family == AF_INET,
1379 	    ("sin_family %d", l3addr->sa_family));
1380 
1381 	/* XXX rtalloc1 should take a const param */
1382 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1383 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) ||
1384 	    ((rt->rt_ifp != ifp) && !(flags & LLE_PUB))) {
1385 #ifdef DIAGNOSTIC
1386 		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1387 		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1388 #endif
1389 		if (rt != NULL)
1390 			RTFREE_LOCKED(rt);
1391 		return (EINVAL);
1392 	}
1393 	RTFREE_LOCKED(rt);
1394 	return 0;
1395 }
1396 
1397 /*
1398  * Return NULL if not found or marked for deletion.
1399  * If found return lle read locked.
1400  */
1401 static struct llentry *
1402 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1403 {
1404 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1405 	struct ifnet *ifp = llt->llt_ifp;
1406 	struct llentry *lle;
1407 	struct llentries *lleh;
1408 	u_int hashkey;
1409 
1410 	IF_AFDATA_LOCK_ASSERT(ifp);
1411 	KASSERT(l3addr->sa_family == AF_INET,
1412 	    ("sin_family %d", l3addr->sa_family));
1413 
1414 	hashkey = sin->sin_addr.s_addr;
1415 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1416 	LIST_FOREACH(lle, lleh, lle_next) {
1417 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1418 		if (lle->la_flags & LLE_DELETED)
1419 			continue;
1420 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1421 			break;
1422 	}
1423 	if (lle == NULL) {
1424 #ifdef DIAGNOSTIC
1425 		if (flags & LLE_DELETE)
1426 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1427 #endif
1428 		if (!(flags & LLE_CREATE))
1429 			return (NULL);
1430 		/*
1431 		 * A route that covers the given address must have
1432 		 * been installed 1st because we are doing a resolution,
1433 		 * verify this.
1434 		 */
1435 		if (!(flags & LLE_IFADDR) &&
1436 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1437 			goto done;
1438 
1439 		lle = in_lltable_new(l3addr, flags);
1440 		if (lle == NULL) {
1441 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1442 			goto done;
1443 		}
1444 		lle->la_flags = flags & ~LLE_CREATE;
1445 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1446 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1447 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1448 		}
1449 
1450 		lle->lle_tbl  = llt;
1451 		lle->lle_head = lleh;
1452 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1453 	} else if (flags & LLE_DELETE) {
1454 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1455 			LLE_WLOCK(lle);
1456 			lle->la_flags = LLE_DELETED;
1457 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1458 			LLE_WUNLOCK(lle);
1459 #ifdef DIAGNOSTIC
1460 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1461 #endif
1462 		}
1463 		lle = (void *)-1;
1464 
1465 	}
1466 	if (LLE_IS_VALID(lle)) {
1467 		if (flags & LLE_EXCLUSIVE)
1468 			LLE_WLOCK(lle);
1469 		else
1470 			LLE_RLOCK(lle);
1471 	}
1472 done:
1473 	return (lle);
1474 }
1475 
1476 static int
1477 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1478 {
1479 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1480 	struct ifnet *ifp = llt->llt_ifp;
1481 	struct llentry *lle;
1482 	/* XXX stack use */
1483 	struct {
1484 		struct rt_msghdr	rtm;
1485 		struct sockaddr_inarp	sin;
1486 		struct sockaddr_dl	sdl;
1487 	} arpc;
1488 	int error, i;
1489 
1490 	LLTABLE_LOCK_ASSERT();
1491 
1492 	error = 0;
1493 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1494 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1495 			struct sockaddr_dl *sdl;
1496 
1497 			/* skip deleted entries */
1498 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1499 				continue;
1500 			/* Skip if jailed and not a valid IP of the prison. */
1501 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1502 				continue;
1503 			/*
1504 			 * produce a msg made of:
1505 			 *  struct rt_msghdr;
1506 			 *  struct sockaddr_inarp; (IPv4)
1507 			 *  struct sockaddr_dl;
1508 			 */
1509 			bzero(&arpc, sizeof(arpc));
1510 			arpc.rtm.rtm_msglen = sizeof(arpc);
1511 			arpc.rtm.rtm_version = RTM_VERSION;
1512 			arpc.rtm.rtm_type = RTM_GET;
1513 			arpc.rtm.rtm_flags = RTF_UP;
1514 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1515 			arpc.sin.sin_family = AF_INET;
1516 			arpc.sin.sin_len = sizeof(arpc.sin);
1517 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1518 
1519 			/* publish */
1520 			if (lle->la_flags & LLE_PUB) {
1521 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1522 				/* proxy only */
1523 				if (lle->la_flags & LLE_PROXY)
1524 					arpc.sin.sin_other = SIN_PROXY;
1525 			}
1526 
1527 			sdl = &arpc.sdl;
1528 			sdl->sdl_family = AF_LINK;
1529 			sdl->sdl_len = sizeof(*sdl);
1530 			sdl->sdl_index = ifp->if_index;
1531 			sdl->sdl_type = ifp->if_type;
1532 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1533 				sdl->sdl_alen = ifp->if_addrlen;
1534 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1535 			} else {
1536 				sdl->sdl_alen = 0;
1537 				bzero(LLADDR(sdl), ifp->if_addrlen);
1538 			}
1539 
1540 			arpc.rtm.rtm_rmx.rmx_expire =
1541 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1542 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1543 			if (lle->la_flags & LLE_STATIC)
1544 				arpc.rtm.rtm_flags |= RTF_STATIC;
1545 			arpc.rtm.rtm_index = ifp->if_index;
1546 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1547 			if (error)
1548 				break;
1549 		}
1550 	}
1551 	return error;
1552 #undef SIN
1553 }
1554 
1555 void *
1556 in_domifattach(struct ifnet *ifp)
1557 {
1558 	struct in_ifinfo *ii;
1559 	struct lltable *llt;
1560 
1561 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1562 
1563 	llt = lltable_init(ifp, AF_INET);
1564 	if (llt != NULL) {
1565 		llt->llt_new = in_lltable_new;
1566 		llt->llt_free = in_lltable_free;
1567 		llt->llt_prefix_free = in_lltable_prefix_free;
1568 		llt->llt_rtcheck = in_lltable_rtcheck;
1569 		llt->llt_lookup = in_lltable_lookup;
1570 		llt->llt_dump = in_lltable_dump;
1571 	}
1572 	ii->ii_llt = llt;
1573 
1574 	ii->ii_igmp = igmp_domifattach(ifp);
1575 
1576 	return ii;
1577 }
1578 
1579 void
1580 in_domifdetach(struct ifnet *ifp, void *aux)
1581 {
1582 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1583 
1584 	igmp_domifdetach(ifp);
1585 	lltable_free(ii->ii_llt);
1586 	free(ii, M_IFADDR);
1587 }
1588