xref: /freebsd/sys/netinet/in.c (revision a3cf0ef5a295c885c895fabfd56470c0d1db322d)
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_mpath.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/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/proc.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 
50 #include <net/if.h>
51 #include <net/if_var.h>
52 #include <net/if_dl.h>
53 #include <net/if_llatbl.h>
54 #include <net/if_types.h>
55 #include <net/route.h>
56 #include <net/vnet.h>
57 
58 #include <netinet/in.h>
59 #include <netinet/in_var.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/ip_var.h>
62 #include <netinet/igmp_var.h>
63 #include <netinet/udp.h>
64 #include <netinet/udp_var.h>
65 
66 static int in_mask2len(struct in_addr *);
67 static void in_len2mask(struct in_addr *, int);
68 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
69 	struct ifnet *, struct thread *);
70 
71 static int	in_addprefix(struct in_ifaddr *, int);
72 static int	in_scrubprefix(struct in_ifaddr *);
73 static void	in_socktrim(struct sockaddr_in *);
74 static int	in_ifinit(struct ifnet *,
75 	    struct in_ifaddr *, struct sockaddr_in *, int);
76 static void	in_purgemaddrs(struct ifnet *);
77 
78 static VNET_DEFINE(int, subnetsarelocal);
79 #define	V_subnetsarelocal		VNET(subnetsarelocal)
80 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
81 	&VNET_NAME(subnetsarelocal), 0,
82 	"Treat all subnets as directly connected");
83 static VNET_DEFINE(int, sameprefixcarponly);
84 #define	V_sameprefixcarponly		VNET(sameprefixcarponly)
85 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
86 	&VNET_NAME(sameprefixcarponly), 0,
87 	"Refuse to create same prefixes on different interfaces");
88 
89 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
90 #define	V_ripcbinfo			VNET(ripcbinfo)
91 
92 /*
93  * Return 1 if an internet address is for a ``local'' host
94  * (one to which we have a connection).  If subnetsarelocal
95  * is true, this includes other subnets of the local net.
96  * Otherwise, it includes only the directly-connected (sub)nets.
97  */
98 int
99 in_localaddr(struct in_addr in)
100 {
101 	register u_long i = ntohl(in.s_addr);
102 	register struct in_ifaddr *ia;
103 
104 	IN_IFADDR_RLOCK();
105 	if (V_subnetsarelocal) {
106 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
107 			if ((i & ia->ia_netmask) == ia->ia_net) {
108 				IN_IFADDR_RUNLOCK();
109 				return (1);
110 			}
111 		}
112 	} else {
113 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
114 			if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
115 				IN_IFADDR_RUNLOCK();
116 				return (1);
117 			}
118 		}
119 	}
120 	IN_IFADDR_RUNLOCK();
121 	return (0);
122 }
123 
124 /*
125  * Return 1 if an internet address is for the local host and configured
126  * on one of its interfaces.
127  */
128 int
129 in_localip(struct in_addr in)
130 {
131 	struct in_ifaddr *ia;
132 
133 	IN_IFADDR_RLOCK();
134 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
135 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
136 			IN_IFADDR_RUNLOCK();
137 			return (1);
138 		}
139 	}
140 	IN_IFADDR_RUNLOCK();
141 	return (0);
142 }
143 
144 /*
145  * Determine whether an IP address is in a reserved set of addresses
146  * that may not be forwarded, or whether datagrams to that destination
147  * may be forwarded.
148  */
149 int
150 in_canforward(struct in_addr in)
151 {
152 	register u_long i = ntohl(in.s_addr);
153 	register u_long net;
154 
155 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
156 		return (0);
157 	if (IN_CLASSA(i)) {
158 		net = i & IN_CLASSA_NET;
159 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
160 			return (0);
161 	}
162 	return (1);
163 }
164 
165 /*
166  * Trim a mask in a sockaddr
167  */
168 static void
169 in_socktrim(struct sockaddr_in *ap)
170 {
171     register char *cplim = (char *) &ap->sin_addr;
172     register char *cp = (char *) (&ap->sin_addr + 1);
173 
174     ap->sin_len = 0;
175     while (--cp >= cplim)
176 	if (*cp) {
177 	    (ap)->sin_len = cp - (char *) (ap) + 1;
178 	    break;
179 	}
180 }
181 
182 static int
183 in_mask2len(mask)
184 	struct in_addr *mask;
185 {
186 	int x, y;
187 	u_char *p;
188 
189 	p = (u_char *)mask;
190 	for (x = 0; x < sizeof(*mask); x++) {
191 		if (p[x] != 0xff)
192 			break;
193 	}
194 	y = 0;
195 	if (x < sizeof(*mask)) {
196 		for (y = 0; y < 8; y++) {
197 			if ((p[x] & (0x80 >> y)) == 0)
198 				break;
199 		}
200 	}
201 	return (x * 8 + y);
202 }
203 
204 static void
205 in_len2mask(struct in_addr *mask, int len)
206 {
207 	int i;
208 	u_char *p;
209 
210 	p = (u_char *)mask;
211 	bzero(mask, sizeof(*mask));
212 	for (i = 0; i < len / 8; i++)
213 		p[i] = 0xff;
214 	if (len % 8)
215 		p[i] = (0xff00 >> (len % 8)) & 0xff;
216 }
217 
218 /*
219  * Generic internet control operations (ioctl's).
220  *
221  * ifp is NULL if not an interface-specific ioctl.
222  */
223 /* ARGSUSED */
224 int
225 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
226     struct thread *td)
227 {
228 	register struct ifreq *ifr = (struct ifreq *)data;
229 	register struct in_ifaddr *ia, *iap;
230 	register struct ifaddr *ifa;
231 	struct in_addr allhosts_addr;
232 	struct in_addr dst;
233 	struct in_ifinfo *ii;
234 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
235 	struct sockaddr_in oldaddr;
236 	int error, hostIsNew, iaIsNew, maskIsNew;
237 	int iaIsFirst;
238 
239 	ia = NULL;
240 	iaIsFirst = 0;
241 	iaIsNew = 0;
242 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
243 
244 	/*
245 	 * Filter out ioctls we implement directly; forward the rest on to
246 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
247 	 */
248 	switch (cmd) {
249 	case SIOCAIFADDR:
250 	case SIOCDIFADDR:
251 	case SIOCGIFADDR:
252 	case SIOCGIFBRDADDR:
253 	case SIOCGIFDSTADDR:
254 	case SIOCGIFNETMASK:
255 	case SIOCSIFADDR:
256 	case SIOCSIFBRDADDR:
257 	case SIOCSIFDSTADDR:
258 	case SIOCSIFNETMASK:
259 		break;
260 
261 	case SIOCALIFADDR:
262 		if (td != NULL) {
263 			error = priv_check(td, PRIV_NET_ADDIFADDR);
264 			if (error)
265 				return (error);
266 		}
267 		if (ifp == NULL)
268 			return (EINVAL);
269 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
270 
271 	case SIOCDLIFADDR:
272 		if (td != NULL) {
273 			error = priv_check(td, PRIV_NET_DELIFADDR);
274 			if (error)
275 				return (error);
276 		}
277 		if (ifp == NULL)
278 			return (EINVAL);
279 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
280 
281 	case SIOCGLIFADDR:
282 		if (ifp == NULL)
283 			return (EINVAL);
284 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
285 
286 	default:
287 		if (ifp == NULL || ifp->if_ioctl == NULL)
288 			return (EOPNOTSUPP);
289 		return ((*ifp->if_ioctl)(ifp, cmd, data));
290 	}
291 
292 	if (ifp == NULL)
293 		return (EADDRNOTAVAIL);
294 
295 	/*
296 	 * Security checks before we get involved in any work.
297 	 */
298 	switch (cmd) {
299 	case SIOCAIFADDR:
300 	case SIOCSIFADDR:
301 	case SIOCSIFBRDADDR:
302 	case SIOCSIFNETMASK:
303 	case SIOCSIFDSTADDR:
304 		if (td != NULL) {
305 			error = priv_check(td, PRIV_NET_ADDIFADDR);
306 			if (error)
307 				return (error);
308 		}
309 		break;
310 
311 	case SIOCDIFADDR:
312 		if (td != NULL) {
313 			error = priv_check(td, PRIV_NET_DELIFADDR);
314 			if (error)
315 				return (error);
316 		}
317 		break;
318 	}
319 
320 	/*
321 	 * Find address for this interface, if it exists.
322 	 *
323 	 * If an alias address was specified, find that one instead of the
324 	 * first one on the interface, if possible.
325 	 */
326 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
327 	IN_IFADDR_RLOCK();
328 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
329 		if (iap->ia_ifp == ifp &&
330 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
331 			if (td == NULL || prison_check_ip4(td->td_ucred,
332 			    &dst) == 0)
333 				ia = iap;
334 			break;
335 		}
336 	}
337 	if (ia != NULL)
338 		ifa_ref(&ia->ia_ifa);
339 	IN_IFADDR_RUNLOCK();
340 	if (ia == NULL) {
341 		IF_ADDR_LOCK(ifp);
342 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
343 			iap = ifatoia(ifa);
344 			if (iap->ia_addr.sin_family == AF_INET) {
345 				if (td != NULL &&
346 				    prison_check_ip4(td->td_ucred,
347 				    &iap->ia_addr.sin_addr) != 0)
348 					continue;
349 				ia = iap;
350 				break;
351 			}
352 		}
353 		if (ia != NULL)
354 			ifa_ref(&ia->ia_ifa);
355 		IF_ADDR_UNLOCK(ifp);
356 	}
357 	if (ia == NULL)
358 		iaIsFirst = 1;
359 
360 	error = 0;
361 	switch (cmd) {
362 	case SIOCAIFADDR:
363 	case SIOCDIFADDR:
364 		if (ifra->ifra_addr.sin_family == AF_INET) {
365 			struct in_ifaddr *oia;
366 
367 			IN_IFADDR_RLOCK();
368 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
369 				if (ia->ia_ifp == ifp  &&
370 				    ia->ia_addr.sin_addr.s_addr ==
371 				    ifra->ifra_addr.sin_addr.s_addr)
372 					break;
373 			}
374 			if (ia != NULL && ia != oia)
375 				ifa_ref(&ia->ia_ifa);
376 			if (oia != NULL && ia != oia)
377 				ifa_free(&oia->ia_ifa);
378 			IN_IFADDR_RUNLOCK();
379 			if ((ifp->if_flags & IFF_POINTOPOINT)
380 			    && (cmd == SIOCAIFADDR)
381 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
382 				== INADDR_ANY)) {
383 				error = EDESTADDRREQ;
384 				goto out;
385 			}
386 		}
387 		if (cmd == SIOCDIFADDR && ia == NULL) {
388 			error = EADDRNOTAVAIL;
389 			goto out;
390 		}
391 		/* FALLTHROUGH */
392 	case SIOCSIFADDR:
393 	case SIOCSIFNETMASK:
394 	case SIOCSIFDSTADDR:
395 		if (ia == NULL) {
396 			ia = (struct in_ifaddr *)
397 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
398 				    M_ZERO);
399 			if (ia == NULL) {
400 				error = ENOBUFS;
401 				goto out;
402 			}
403 
404 			ifa = &ia->ia_ifa;
405 			ifa_init(ifa);
406 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
407 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
408 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
409 
410 			ia->ia_sockmask.sin_len = 8;
411 			ia->ia_sockmask.sin_family = AF_INET;
412 			if (ifp->if_flags & IFF_BROADCAST) {
413 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
414 				ia->ia_broadaddr.sin_family = AF_INET;
415 			}
416 			ia->ia_ifp = ifp;
417 
418 			ifa_ref(ifa);			/* if_addrhead */
419 			IF_ADDR_LOCK(ifp);
420 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
421 			IF_ADDR_UNLOCK(ifp);
422 			ifa_ref(ifa);			/* in_ifaddrhead */
423 			IN_IFADDR_WLOCK();
424 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
425 			IN_IFADDR_WUNLOCK();
426 			iaIsNew = 1;
427 		}
428 		break;
429 
430 	case SIOCSIFBRDADDR:
431 	case SIOCGIFADDR:
432 	case SIOCGIFNETMASK:
433 	case SIOCGIFDSTADDR:
434 	case SIOCGIFBRDADDR:
435 		if (ia == NULL) {
436 			error = EADDRNOTAVAIL;
437 			goto out;
438 		}
439 		break;
440 	}
441 
442 	/*
443 	 * Most paths in this switch return directly or via out.  Only paths
444 	 * that remove the address break in order to hit common removal code.
445 	 */
446 	switch (cmd) {
447 	case SIOCGIFADDR:
448 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
449 		goto out;
450 
451 	case SIOCGIFBRDADDR:
452 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
453 			error = EINVAL;
454 			goto out;
455 		}
456 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
457 		goto out;
458 
459 	case SIOCGIFDSTADDR:
460 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
461 			error = EINVAL;
462 			goto out;
463 		}
464 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
465 		goto out;
466 
467 	case SIOCGIFNETMASK:
468 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
469 		goto out;
470 
471 	case SIOCSIFDSTADDR:
472 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
473 			error = EINVAL;
474 			goto out;
475 		}
476 		oldaddr = ia->ia_dstaddr;
477 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
478 		if (ifp->if_ioctl != NULL) {
479 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
480 			    (caddr_t)ia);
481 			if (error) {
482 				ia->ia_dstaddr = oldaddr;
483 				goto out;
484 			}
485 		}
486 		if (ia->ia_flags & IFA_ROUTE) {
487 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
488 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
489 			ia->ia_ifa.ifa_dstaddr =
490 					(struct sockaddr *)&ia->ia_dstaddr;
491 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
492 		}
493 		goto out;
494 
495 	case SIOCSIFBRDADDR:
496 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
497 			error = EINVAL;
498 			goto out;
499 		}
500 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
501 		goto out;
502 
503 	case SIOCSIFADDR:
504 		error = in_ifinit(ifp, ia,
505 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
506 		if (error != 0 && iaIsNew)
507 			break;
508 		if (error == 0) {
509 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
510 			if (iaIsFirst &&
511 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
512 				error = in_joingroup(ifp, &allhosts_addr,
513 				    NULL, &ii->ii_allhosts);
514 			}
515 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
516 		}
517 		error = 0;
518 		goto out;
519 
520 	case SIOCSIFNETMASK:
521 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
522 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
523 		goto out;
524 
525 	case SIOCAIFADDR:
526 		maskIsNew = 0;
527 		hostIsNew = 1;
528 		error = 0;
529 		if (ia->ia_addr.sin_family == AF_INET) {
530 			if (ifra->ifra_addr.sin_len == 0) {
531 				ifra->ifra_addr = ia->ia_addr;
532 				hostIsNew = 0;
533 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
534 					       ia->ia_addr.sin_addr.s_addr)
535 				hostIsNew = 0;
536 		}
537 		if (ifra->ifra_mask.sin_len) {
538 			/*
539 			 * QL: XXX
540 			 * Need to scrub the prefix here in case
541 			 * the issued command is SIOCAIFADDR with
542 			 * the same address, but with a different
543 			 * prefix length. And if the prefix length
544 			 * is the same as before, then the call is
545 			 * un-necessarily executed here.
546 			 */
547 			in_ifscrub(ifp, ia);
548 			ia->ia_sockmask = ifra->ifra_mask;
549 			ia->ia_sockmask.sin_family = AF_INET;
550 			ia->ia_subnetmask =
551 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
552 			maskIsNew = 1;
553 		}
554 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
555 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
556 			in_ifscrub(ifp, ia);
557 			ia->ia_dstaddr = ifra->ifra_dstaddr;
558 			maskIsNew  = 1; /* We lie; but the effect's the same */
559 		}
560 		if (ifra->ifra_addr.sin_family == AF_INET &&
561 		    (hostIsNew || maskIsNew))
562 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
563 		if (error != 0 && iaIsNew)
564 			break;
565 
566 		if ((ifp->if_flags & IFF_BROADCAST) &&
567 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
568 			ia->ia_broadaddr = ifra->ifra_broadaddr;
569 		if (error == 0) {
570 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
571 			if (iaIsFirst &&
572 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
573 				error = in_joingroup(ifp, &allhosts_addr,
574 				    NULL, &ii->ii_allhosts);
575 			}
576 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
577 		}
578 		goto out;
579 
580 	case SIOCDIFADDR:
581 		/*
582 		 * in_ifscrub kills the interface route.
583 		 */
584 		in_ifscrub(ifp, ia);
585 
586 		/*
587 		 * in_ifadown gets rid of all the rest of
588 		 * the routes.  This is not quite the right
589 		 * thing to do, but at least if we are running
590 		 * a routing process they will come back.
591 		 */
592 		in_ifadown(&ia->ia_ifa, 1);
593 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
594 		error = 0;
595 		break;
596 
597 	default:
598 		panic("in_control: unsupported ioctl");
599 	}
600 
601 	IF_ADDR_LOCK(ifp);
602 	/* Re-check that ia is still part of the list. */
603 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
604 		if (ifa == &ia->ia_ifa)
605 			break;
606 	}
607 	if (ifa == NULL) {
608 		/*
609 		 * If we lost the race with another thread, there is no need to
610 		 * try it again for the next loop as there is no other exit
611 		 * path between here and out.
612 		 */
613 		IF_ADDR_UNLOCK(ifp);
614 		error = EADDRNOTAVAIL;
615 		goto out;
616 	}
617 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
618 	IF_ADDR_UNLOCK(ifp);
619 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
620 
621 	IN_IFADDR_WLOCK();
622 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
623 	if (ia->ia_addr.sin_family == AF_INET) {
624 		struct in_ifaddr *if_ia;
625 
626 		LIST_REMOVE(ia, ia_hash);
627 		IN_IFADDR_WUNLOCK();
628 		/*
629 		 * If this is the last IPv4 address configured on this
630 		 * interface, leave the all-hosts group.
631 		 * No state-change report need be transmitted.
632 		 */
633 		if_ia = NULL;
634 		IFP_TO_IA(ifp, if_ia);
635 		if (if_ia == NULL) {
636 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
637 			IN_MULTI_LOCK();
638 			if (ii->ii_allhosts) {
639 				(void)in_leavegroup_locked(ii->ii_allhosts,
640 				    NULL);
641 				ii->ii_allhosts = NULL;
642 			}
643 			IN_MULTI_UNLOCK();
644 		} else
645 			ifa_free(&if_ia->ia_ifa);
646 	} else
647 		IN_IFADDR_WUNLOCK();
648 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
649 out:
650 	if (ia != NULL)
651 		ifa_free(&ia->ia_ifa);
652 	return (error);
653 }
654 
655 /*
656  * SIOC[GAD]LIFADDR.
657  *	SIOCGLIFADDR: get first address. (?!?)
658  *	SIOCGLIFADDR with IFLR_PREFIX:
659  *		get first address that matches the specified prefix.
660  *	SIOCALIFADDR: add the specified address.
661  *	SIOCALIFADDR with IFLR_PREFIX:
662  *		EINVAL since we can't deduce hostid part of the address.
663  *	SIOCDLIFADDR: delete the specified address.
664  *	SIOCDLIFADDR with IFLR_PREFIX:
665  *		delete the first address that matches the specified prefix.
666  * return values:
667  *	EINVAL on invalid parameters
668  *	EADDRNOTAVAIL on prefix match failed/specified address not found
669  *	other values may be returned from in_ioctl()
670  */
671 static int
672 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
673     struct ifnet *ifp, struct thread *td)
674 {
675 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
676 	struct ifaddr *ifa;
677 
678 	/* sanity checks */
679 	if (data == NULL || ifp == NULL) {
680 		panic("invalid argument to in_lifaddr_ioctl");
681 		/*NOTRECHED*/
682 	}
683 
684 	switch (cmd) {
685 	case SIOCGLIFADDR:
686 		/* address must be specified on GET with IFLR_PREFIX */
687 		if ((iflr->flags & IFLR_PREFIX) == 0)
688 			break;
689 		/*FALLTHROUGH*/
690 	case SIOCALIFADDR:
691 	case SIOCDLIFADDR:
692 		/* address must be specified on ADD and DELETE */
693 		if (iflr->addr.ss_family != AF_INET)
694 			return (EINVAL);
695 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
696 			return (EINVAL);
697 		/* XXX need improvement */
698 		if (iflr->dstaddr.ss_family
699 		 && iflr->dstaddr.ss_family != AF_INET)
700 			return (EINVAL);
701 		if (iflr->dstaddr.ss_family
702 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
703 			return (EINVAL);
704 		break;
705 	default: /*shouldn't happen*/
706 		return (EOPNOTSUPP);
707 	}
708 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
709 		return (EINVAL);
710 
711 	switch (cmd) {
712 	case SIOCALIFADDR:
713 	    {
714 		struct in_aliasreq ifra;
715 
716 		if (iflr->flags & IFLR_PREFIX)
717 			return (EINVAL);
718 
719 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
720 		bzero(&ifra, sizeof(ifra));
721 		bcopy(iflr->iflr_name, ifra.ifra_name,
722 			sizeof(ifra.ifra_name));
723 
724 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
725 
726 		if (iflr->dstaddr.ss_family) {	/*XXX*/
727 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
728 				iflr->dstaddr.ss_len);
729 		}
730 
731 		ifra.ifra_mask.sin_family = AF_INET;
732 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
733 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
734 
735 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
736 	    }
737 	case SIOCGLIFADDR:
738 	case SIOCDLIFADDR:
739 	    {
740 		struct in_ifaddr *ia;
741 		struct in_addr mask, candidate, match;
742 		struct sockaddr_in *sin;
743 
744 		bzero(&mask, sizeof(mask));
745 		bzero(&match, sizeof(match));
746 		if (iflr->flags & IFLR_PREFIX) {
747 			/* lookup a prefix rather than address. */
748 			in_len2mask(&mask, iflr->prefixlen);
749 
750 			sin = (struct sockaddr_in *)&iflr->addr;
751 			match.s_addr = sin->sin_addr.s_addr;
752 			match.s_addr &= mask.s_addr;
753 
754 			/* if you set extra bits, that's wrong */
755 			if (match.s_addr != sin->sin_addr.s_addr)
756 				return (EINVAL);
757 
758 		} else {
759 			/* on getting an address, take the 1st match */
760 			/* on deleting an address, do exact match */
761 			if (cmd != SIOCGLIFADDR) {
762 				in_len2mask(&mask, 32);
763 				sin = (struct sockaddr_in *)&iflr->addr;
764 				match.s_addr = sin->sin_addr.s_addr;
765 			}
766 		}
767 
768 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
769 			if (ifa->ifa_addr->sa_family != AF_INET6)
770 				continue;
771 			if (match.s_addr == 0)
772 				break;
773 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
774 			candidate.s_addr &= mask.s_addr;
775 			if (candidate.s_addr == match.s_addr)
776 				break;
777 		}
778 		if (ifa == NULL)
779 			return (EADDRNOTAVAIL);
780 		ia = (struct in_ifaddr *)ifa;
781 
782 		if (cmd == SIOCGLIFADDR) {
783 			/* fill in the if_laddrreq structure */
784 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
785 
786 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
787 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
788 					ia->ia_dstaddr.sin_len);
789 			} else
790 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
791 
792 			iflr->prefixlen =
793 				in_mask2len(&ia->ia_sockmask.sin_addr);
794 
795 			iflr->flags = 0;	/*XXX*/
796 
797 			return (0);
798 		} else {
799 			struct in_aliasreq ifra;
800 
801 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
802 			bzero(&ifra, sizeof(ifra));
803 			bcopy(iflr->iflr_name, ifra.ifra_name,
804 				sizeof(ifra.ifra_name));
805 
806 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
807 				ia->ia_addr.sin_len);
808 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
809 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
810 					ia->ia_dstaddr.sin_len);
811 			}
812 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
813 				ia->ia_sockmask.sin_len);
814 
815 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
816 			    ifp, td));
817 		}
818 	    }
819 	}
820 
821 	return (EOPNOTSUPP);	/*just for safety*/
822 }
823 
824 /*
825  * Delete any existing route for an interface.
826  */
827 void
828 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
829 {
830 
831 	in_scrubprefix(ia);
832 }
833 
834 /*
835  * Initialize an interface's internet address
836  * and routing table entry.
837  */
838 static int
839 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
840     int scrub)
841 {
842 	register u_long i = ntohl(sin->sin_addr.s_addr);
843 	struct sockaddr_in oldaddr;
844 	int s = splimp(), flags = RTF_UP, error = 0;
845 
846 	oldaddr = ia->ia_addr;
847 	if (oldaddr.sin_family == AF_INET)
848 		LIST_REMOVE(ia, ia_hash);
849 	ia->ia_addr = *sin;
850 	if (ia->ia_addr.sin_family == AF_INET) {
851 		IN_IFADDR_WLOCK();
852 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
853 		    ia, ia_hash);
854 		IN_IFADDR_WUNLOCK();
855 	}
856 	/*
857 	 * Give the interface a chance to initialize
858 	 * if this is its first address,
859 	 * and to validate the address if necessary.
860 	 */
861 	if (ifp->if_ioctl != NULL) {
862 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
863 		if (error) {
864 			splx(s);
865 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
866 			ia->ia_addr = oldaddr;
867 			IN_IFADDR_WLOCK();
868 			if (ia->ia_addr.sin_family == AF_INET)
869 				LIST_INSERT_HEAD(INADDR_HASH(
870 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
871 			else
872 				/*
873 				 * If oldaddr family is not AF_INET (e.g.
874 				 * interface has been just created) in_control
875 				 * does not call LIST_REMOVE, and we end up
876 				 * with bogus ia entries in hash
877 				 */
878 				LIST_REMOVE(ia, ia_hash);
879 			IN_IFADDR_WUNLOCK();
880 			return (error);
881 		}
882 	}
883 	splx(s);
884 	if (scrub) {
885 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
886 		in_ifscrub(ifp, ia);
887 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
888 	}
889 	if (IN_CLASSA(i))
890 		ia->ia_netmask = IN_CLASSA_NET;
891 	else if (IN_CLASSB(i))
892 		ia->ia_netmask = IN_CLASSB_NET;
893 	else
894 		ia->ia_netmask = IN_CLASSC_NET;
895 	/*
896 	 * The subnet mask usually includes at least the standard network part,
897 	 * but may may be smaller in the case of supernetting.
898 	 * If it is set, we believe it.
899 	 */
900 	if (ia->ia_subnetmask == 0) {
901 		ia->ia_subnetmask = ia->ia_netmask;
902 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
903 	} else
904 		ia->ia_netmask &= ia->ia_subnetmask;
905 	ia->ia_net = i & ia->ia_netmask;
906 	ia->ia_subnet = i & ia->ia_subnetmask;
907 	in_socktrim(&ia->ia_sockmask);
908 	/*
909 	 * XXX: carp(4) does not have interface route
910 	 */
911 	if (ifp->if_type == IFT_CARP)
912 		return (0);
913 	/*
914 	 * Add route for the network.
915 	 */
916 	ia->ia_ifa.ifa_metric = ifp->if_metric;
917 	if (ifp->if_flags & IFF_BROADCAST) {
918 		ia->ia_broadaddr.sin_addr.s_addr =
919 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
920 		ia->ia_netbroadcast.s_addr =
921 			htonl(ia->ia_net | ~ ia->ia_netmask);
922 	} else if (ifp->if_flags & IFF_LOOPBACK) {
923 		ia->ia_dstaddr = ia->ia_addr;
924 		flags |= RTF_HOST;
925 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
926 		if (ia->ia_dstaddr.sin_family != AF_INET)
927 			return (0);
928 		flags |= RTF_HOST;
929 	}
930 	if ((error = in_addprefix(ia, flags)) != 0)
931 		return (error);
932 
933 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
934 		return (0);
935 
936 	if (ifp->if_flags & IFF_POINTOPOINT) {
937 		if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
938 			return (0);
939 	}
940 
941 
942 	/*
943 	 * add a loopback route to self
944 	 */
945 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
946 		struct route ia_ro;
947 
948 		bzero(&ia_ro, sizeof(ia_ro));
949 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
950 		rtalloc_ign_fib(&ia_ro, 0, 0);
951 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
952 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
953 			RT_LOCK(ia_ro.ro_rt);
954 			RT_ADDREF(ia_ro.ro_rt);
955 			RTFREE_LOCKED(ia_ro.ro_rt);
956 		} else
957 			error = ifa_add_loopback_route((struct ifaddr *)ia,
958 				       (struct sockaddr *)&ia->ia_addr);
959 		if (error == 0)
960 			ia->ia_flags |= IFA_RTSELF;
961 		if (ia_ro.ro_rt != NULL)
962 			RTFREE(ia_ro.ro_rt);
963 	}
964 
965 	return (error);
966 }
967 
968 #define rtinitflags(x) \
969 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
970 	    ? RTF_HOST : 0)
971 
972 /*
973  * Generate a routing message when inserting or deleting
974  * an interface address alias.
975  */
976 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
977     struct in_ifaddr *target)
978 {
979 	struct route pfx_ro;
980 	struct sockaddr_in *pfx_addr;
981 	struct rtentry msg_rt;
982 
983 	/* QL: XXX
984 	 * This is a bit questionable because there is no
985 	 * additional route entry added/deleted for an address
986 	 * alias. Therefore this route report is inaccurate.
987 	 */
988 	bzero(&pfx_ro, sizeof(pfx_ro));
989 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
990 	pfx_addr->sin_len = sizeof(*pfx_addr);
991 	pfx_addr->sin_family = AF_INET;
992 	pfx_addr->sin_addr = *prefix;
993 	rtalloc_ign_fib(&pfx_ro, 0, 0);
994 	if (pfx_ro.ro_rt != NULL) {
995 		msg_rt = *pfx_ro.ro_rt;
996 
997 		/* QL: XXX
998 		 * Point the gateway to the new interface
999 		 * address as if a new prefix route entry has
1000 		 * been added through the new address alias.
1001 		 * All other parts of the rtentry is accurate,
1002 		 * e.g., rt_key, rt_mask, rt_ifp etc.
1003 		 */
1004 		msg_rt.rt_gateway =
1005 			(struct sockaddr *)&target->ia_addr;
1006 		rt_newaddrmsg(cmd,
1007 			      (struct ifaddr *)target,
1008 			      0, &msg_rt);
1009 		RTFREE(pfx_ro.ro_rt);
1010 	}
1011 	return;
1012 }
1013 
1014 /*
1015  * Check if we have a route for the given prefix already or add one accordingly.
1016  */
1017 static int
1018 in_addprefix(struct in_ifaddr *target, int flags)
1019 {
1020 	struct in_ifaddr *ia;
1021 	struct in_addr prefix, mask, p, m;
1022 	int error;
1023 
1024 	if ((flags & RTF_HOST) != 0) {
1025 		prefix = target->ia_dstaddr.sin_addr;
1026 		mask.s_addr = 0;
1027 	} else {
1028 		prefix = target->ia_addr.sin_addr;
1029 		mask = target->ia_sockmask.sin_addr;
1030 		prefix.s_addr &= mask.s_addr;
1031 	}
1032 
1033 	IN_IFADDR_RLOCK();
1034 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1035 		if (rtinitflags(ia)) {
1036 			p = ia->ia_addr.sin_addr;
1037 
1038 			if (prefix.s_addr != p.s_addr)
1039 				continue;
1040 		} else {
1041 			p = ia->ia_addr.sin_addr;
1042 			m = ia->ia_sockmask.sin_addr;
1043 			p.s_addr &= m.s_addr;
1044 
1045 			if (prefix.s_addr != p.s_addr ||
1046 			    mask.s_addr != m.s_addr)
1047 				continue;
1048 		}
1049 
1050 		/*
1051 		 * If we got a matching prefix route inserted by other
1052 		 * interface address, we are done here.
1053 		 */
1054 		if (ia->ia_flags & IFA_ROUTE) {
1055 #ifdef RADIX_MPATH
1056 			if (ia->ia_addr.sin_addr.s_addr ==
1057 			    target->ia_addr.sin_addr.s_addr) {
1058 				IN_IFADDR_RUNLOCK();
1059 				return (EEXIST);
1060 			} else
1061 				break;
1062 #endif
1063 			if (V_sameprefixcarponly &&
1064 			    target->ia_ifp->if_type != IFT_CARP &&
1065 			    ia->ia_ifp->if_type != IFT_CARP) {
1066 				IN_IFADDR_RUNLOCK();
1067 				return (EEXIST);
1068 			} else {
1069 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
1070 				IN_IFADDR_RUNLOCK();
1071 				return (0);
1072 			}
1073 		}
1074 	}
1075 	IN_IFADDR_RUNLOCK();
1076 
1077 	/*
1078 	 * No-one seem to have this prefix route, so we try to insert it.
1079 	 */
1080 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1081 	if (!error)
1082 		target->ia_flags |= IFA_ROUTE;
1083 	return (error);
1084 }
1085 
1086 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1087 
1088 /*
1089  * If there is no other address in the system that can serve a route to the
1090  * same prefix, remove the route.  Hand over the route to the new address
1091  * otherwise.
1092  */
1093 static int
1094 in_scrubprefix(struct in_ifaddr *target)
1095 {
1096 	struct in_ifaddr *ia;
1097 	struct in_addr prefix, mask, p;
1098 	int error = 0;
1099 	struct sockaddr_in prefix0, mask0;
1100 
1101 	/*
1102 	 * Remove the loopback route to the interface address.
1103 	 * The "useloopback" setting is not consulted because if the
1104 	 * user configures an interface address, turns off this
1105 	 * setting, and then tries to delete that interface address,
1106 	 * checking the current setting of "useloopback" would leave
1107 	 * that interface address loopback route untouched, which
1108 	 * would be wrong. Therefore the interface address loopback route
1109 	 * deletion is unconditional.
1110 	 */
1111 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1112 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1113 	    (target->ia_flags & IFA_RTSELF)) {
1114 		struct route ia_ro;
1115 		int freeit = 0;
1116 
1117 		bzero(&ia_ro, sizeof(ia_ro));
1118 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1119 		rtalloc_ign_fib(&ia_ro, 0, 0);
1120 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1121 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1122 			RT_LOCK(ia_ro.ro_rt);
1123 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1124 				freeit = 1;
1125 			else
1126 				RT_REMREF(ia_ro.ro_rt);
1127 			RTFREE_LOCKED(ia_ro.ro_rt);
1128 		}
1129 		if (freeit)
1130 			error = ifa_del_loopback_route((struct ifaddr *)target,
1131 				       (struct sockaddr *)&target->ia_addr);
1132 		if (error == 0)
1133 			target->ia_flags &= ~IFA_RTSELF;
1134 		/* remove arp cache */
1135 		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1136 	}
1137 
1138 	if (rtinitflags(target))
1139 		prefix = target->ia_dstaddr.sin_addr;
1140 	else {
1141 		prefix = target->ia_addr.sin_addr;
1142 		mask = target->ia_sockmask.sin_addr;
1143 		prefix.s_addr &= mask.s_addr;
1144 	}
1145 
1146 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1147 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1148 		return (0);
1149 	}
1150 
1151 	IN_IFADDR_RLOCK();
1152 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1153 		if (rtinitflags(ia))
1154 			p = ia->ia_dstaddr.sin_addr;
1155 		else {
1156 			p = ia->ia_addr.sin_addr;
1157 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1158 		}
1159 
1160 		if (prefix.s_addr != p.s_addr)
1161 			continue;
1162 
1163 		/*
1164 		 * If we got a matching prefix address, move IFA_ROUTE and
1165 		 * the route itself to it.  Make sure that routing daemons
1166 		 * get a heads-up.
1167 		 *
1168 		 * XXX: a special case for carp(4) interface - this should
1169 		 *      be more generally specified as an interface that
1170 		 *      doesn't support such action.
1171 		 */
1172 		if ((ia->ia_flags & IFA_ROUTE) == 0
1173 		    && (ia->ia_ifp->if_type != IFT_CARP)
1174 							) {
1175 			IN_IFADDR_RUNLOCK();
1176 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1177 			    rtinitflags(target));
1178 			target->ia_flags &= ~IFA_ROUTE;
1179 
1180 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1181 			    rtinitflags(ia) | RTF_UP);
1182 			if (error == 0)
1183 				ia->ia_flags |= IFA_ROUTE;
1184 			return (error);
1185 		}
1186 	}
1187 	IN_IFADDR_RUNLOCK();
1188 
1189 	/*
1190 	 * remove all L2 entries on the given prefix
1191 	 */
1192 	bzero(&prefix0, sizeof(prefix0));
1193 	prefix0.sin_len = sizeof(prefix0);
1194 	prefix0.sin_family = AF_INET;
1195 	prefix0.sin_addr.s_addr = target->ia_subnet;
1196 	bzero(&mask0, sizeof(mask0));
1197 	mask0.sin_len = sizeof(mask0);
1198 	mask0.sin_family = AF_INET;
1199 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1200 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1201 			    (struct sockaddr *)&mask0);
1202 
1203 	/*
1204 	 * As no-one seem to have this prefix, we can remove the route.
1205 	 */
1206 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1207 	target->ia_flags &= ~IFA_ROUTE;
1208 	return (0);
1209 }
1210 
1211 #undef rtinitflags
1212 
1213 /*
1214  * Return 1 if the address might be a local broadcast address.
1215  */
1216 int
1217 in_broadcast(struct in_addr in, struct ifnet *ifp)
1218 {
1219 	register struct ifaddr *ifa;
1220 	u_long t;
1221 
1222 	if (in.s_addr == INADDR_BROADCAST ||
1223 	    in.s_addr == INADDR_ANY)
1224 		return (1);
1225 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1226 		return (0);
1227 	t = ntohl(in.s_addr);
1228 	/*
1229 	 * Look through the list of addresses for a match
1230 	 * with a broadcast address.
1231 	 */
1232 #define ia ((struct in_ifaddr *)ifa)
1233 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1234 		if (ifa->ifa_addr->sa_family == AF_INET &&
1235 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1236 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1237 		     /*
1238 		      * Check for old-style (host 0) broadcast.
1239 		      */
1240 		     t == ia->ia_subnet || t == ia->ia_net) &&
1241 		     /*
1242 		      * Check for an all one subnetmask. These
1243 		      * only exist when an interface gets a secondary
1244 		      * address.
1245 		      */
1246 		     ia->ia_subnetmask != (u_long)0xffffffff)
1247 			    return (1);
1248 	return (0);
1249 #undef ia
1250 }
1251 
1252 /*
1253  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1254  */
1255 void
1256 in_ifdetach(struct ifnet *ifp)
1257 {
1258 
1259 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1260 	in_pcbpurgeif0(&V_udbinfo, ifp);
1261 	in_purgemaddrs(ifp);
1262 }
1263 
1264 /*
1265  * Delete all IPv4 multicast address records, and associated link-layer
1266  * multicast address records, associated with ifp.
1267  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1268  * XXX This should not race with ifma_protospec being set during
1269  * a new allocation, if it does, we have bigger problems.
1270  */
1271 static void
1272 in_purgemaddrs(struct ifnet *ifp)
1273 {
1274 	LIST_HEAD(,in_multi) purgeinms;
1275 	struct in_multi		*inm, *tinm;
1276 	struct ifmultiaddr	*ifma;
1277 
1278 	LIST_INIT(&purgeinms);
1279 	IN_MULTI_LOCK();
1280 
1281 	/*
1282 	 * Extract list of in_multi associated with the detaching ifp
1283 	 * which the PF_INET layer is about to release.
1284 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1285 	 * by code further down.
1286 	 */
1287 	IF_ADDR_LOCK(ifp);
1288 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1289 		if (ifma->ifma_addr->sa_family != AF_INET ||
1290 		    ifma->ifma_protospec == NULL)
1291 			continue;
1292 #if 0
1293 		KASSERT(ifma->ifma_protospec != NULL,
1294 		    ("%s: ifma_protospec is NULL", __func__));
1295 #endif
1296 		inm = (struct in_multi *)ifma->ifma_protospec;
1297 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1298 	}
1299 	IF_ADDR_UNLOCK(ifp);
1300 
1301 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1302 		LIST_REMOVE(inm, inm_link);
1303 		inm_release_locked(inm);
1304 	}
1305 	igmp_ifdetach(ifp);
1306 
1307 	IN_MULTI_UNLOCK();
1308 }
1309 
1310 #include <net/if_dl.h>
1311 #include <netinet/if_ether.h>
1312 
1313 struct in_llentry {
1314 	struct llentry		base;
1315 	struct sockaddr_in	l3_addr4;
1316 };
1317 
1318 static struct llentry *
1319 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1320 {
1321 	struct in_llentry *lle;
1322 
1323 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1324 	if (lle == NULL)		/* NB: caller generates msg */
1325 		return NULL;
1326 
1327 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1328 	/*
1329 	 * For IPv4 this will trigger "arpresolve" to generate
1330 	 * an ARP request.
1331 	 */
1332 	lle->base.la_expire = time_second; /* mark expired */
1333 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1334 	lle->base.lle_refcnt = 1;
1335 	LLE_LOCK_INIT(&lle->base);
1336 	return &lle->base;
1337 }
1338 
1339 /*
1340  * Deletes an address from the address table.
1341  * This function is called by the timer functions
1342  * such as arptimer() and nd6_llinfo_timer(), and
1343  * the caller does the locking.
1344  */
1345 static void
1346 in_lltable_free(struct lltable *llt, struct llentry *lle)
1347 {
1348 	LLE_WUNLOCK(lle);
1349 	LLE_LOCK_DESTROY(lle);
1350 	free(lle, M_LLTABLE);
1351 }
1352 
1353 
1354 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1355 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1356 
1357 static void
1358 in_lltable_prefix_free(struct lltable *llt,
1359 		       const struct sockaddr *prefix,
1360 		       const struct sockaddr *mask)
1361 {
1362 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1363 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1364 	struct llentry *lle, *next;
1365 	register int i;
1366 
1367 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1368 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1369 
1370 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1371 						     pfx, msk)) {
1372 				int canceled;
1373 
1374 				canceled = callout_drain(&lle->la_timer);
1375 				LLE_WLOCK(lle);
1376 				if (canceled)
1377 					LLE_REMREF(lle);
1378 				llentry_free(lle);
1379 			}
1380 		}
1381 	}
1382 }
1383 
1384 
1385 static int
1386 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1387 {
1388 	struct rtentry *rt;
1389 
1390 	KASSERT(l3addr->sa_family == AF_INET,
1391 	    ("sin_family %d", l3addr->sa_family));
1392 
1393 	/* XXX rtalloc1 should take a const param */
1394 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1395 	if (rt == NULL || (!(flags & LLE_PUB) &&
1396 			   ((rt->rt_flags & RTF_GATEWAY) ||
1397 			    (rt->rt_ifp != ifp)))) {
1398 #ifdef DIAGNOSTIC
1399 		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1400 		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1401 #endif
1402 		if (rt != NULL)
1403 			RTFREE_LOCKED(rt);
1404 		return (EINVAL);
1405 	}
1406 	RTFREE_LOCKED(rt);
1407 	return 0;
1408 }
1409 
1410 /*
1411  * Return NULL if not found or marked for deletion.
1412  * If found return lle read locked.
1413  */
1414 static struct llentry *
1415 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1416 {
1417 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1418 	struct ifnet *ifp = llt->llt_ifp;
1419 	struct llentry *lle;
1420 	struct llentries *lleh;
1421 	u_int hashkey;
1422 
1423 	IF_AFDATA_LOCK_ASSERT(ifp);
1424 	KASSERT(l3addr->sa_family == AF_INET,
1425 	    ("sin_family %d", l3addr->sa_family));
1426 
1427 	hashkey = sin->sin_addr.s_addr;
1428 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1429 	LIST_FOREACH(lle, lleh, lle_next) {
1430 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1431 		if (lle->la_flags & LLE_DELETED)
1432 			continue;
1433 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1434 			break;
1435 	}
1436 	if (lle == NULL) {
1437 #ifdef DIAGNOSTIC
1438 		if (flags & LLE_DELETE)
1439 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1440 #endif
1441 		if (!(flags & LLE_CREATE))
1442 			return (NULL);
1443 		/*
1444 		 * A route that covers the given address must have
1445 		 * been installed 1st because we are doing a resolution,
1446 		 * verify this.
1447 		 */
1448 		if (!(flags & LLE_IFADDR) &&
1449 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1450 			goto done;
1451 
1452 		lle = in_lltable_new(l3addr, flags);
1453 		if (lle == NULL) {
1454 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1455 			goto done;
1456 		}
1457 		lle->la_flags = flags & ~LLE_CREATE;
1458 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1459 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1460 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1461 		}
1462 
1463 		lle->lle_tbl  = llt;
1464 		lle->lle_head = lleh;
1465 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1466 	} else if (flags & LLE_DELETE) {
1467 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1468 			LLE_WLOCK(lle);
1469 			lle->la_flags = LLE_DELETED;
1470 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1471 			LLE_WUNLOCK(lle);
1472 #ifdef DIAGNOSTIC
1473 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1474 #endif
1475 		}
1476 		lle = (void *)-1;
1477 
1478 	}
1479 	if (LLE_IS_VALID(lle)) {
1480 		if (flags & LLE_EXCLUSIVE)
1481 			LLE_WLOCK(lle);
1482 		else
1483 			LLE_RLOCK(lle);
1484 	}
1485 done:
1486 	return (lle);
1487 }
1488 
1489 static int
1490 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1491 {
1492 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1493 	struct ifnet *ifp = llt->llt_ifp;
1494 	struct llentry *lle;
1495 	/* XXX stack use */
1496 	struct {
1497 		struct rt_msghdr	rtm;
1498 		struct sockaddr_inarp	sin;
1499 		struct sockaddr_dl	sdl;
1500 	} arpc;
1501 	int error, i;
1502 
1503 	LLTABLE_LOCK_ASSERT();
1504 
1505 	error = 0;
1506 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1507 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1508 			struct sockaddr_dl *sdl;
1509 
1510 			/* skip deleted entries */
1511 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1512 				continue;
1513 			/* Skip if jailed and not a valid IP of the prison. */
1514 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1515 				continue;
1516 			/*
1517 			 * produce a msg made of:
1518 			 *  struct rt_msghdr;
1519 			 *  struct sockaddr_inarp; (IPv4)
1520 			 *  struct sockaddr_dl;
1521 			 */
1522 			bzero(&arpc, sizeof(arpc));
1523 			arpc.rtm.rtm_msglen = sizeof(arpc);
1524 			arpc.rtm.rtm_version = RTM_VERSION;
1525 			arpc.rtm.rtm_type = RTM_GET;
1526 			arpc.rtm.rtm_flags = RTF_UP;
1527 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1528 			arpc.sin.sin_family = AF_INET;
1529 			arpc.sin.sin_len = sizeof(arpc.sin);
1530 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1531 
1532 			/* publish */
1533 			if (lle->la_flags & LLE_PUB) {
1534 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1535 				/* proxy only */
1536 				if (lle->la_flags & LLE_PROXY)
1537 					arpc.sin.sin_other = SIN_PROXY;
1538 			}
1539 
1540 			sdl = &arpc.sdl;
1541 			sdl->sdl_family = AF_LINK;
1542 			sdl->sdl_len = sizeof(*sdl);
1543 			sdl->sdl_index = ifp->if_index;
1544 			sdl->sdl_type = ifp->if_type;
1545 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1546 				sdl->sdl_alen = ifp->if_addrlen;
1547 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1548 			} else {
1549 				sdl->sdl_alen = 0;
1550 				bzero(LLADDR(sdl), ifp->if_addrlen);
1551 			}
1552 
1553 			arpc.rtm.rtm_rmx.rmx_expire =
1554 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1555 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1556 			if (lle->la_flags & LLE_STATIC)
1557 				arpc.rtm.rtm_flags |= RTF_STATIC;
1558 			arpc.rtm.rtm_index = ifp->if_index;
1559 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1560 			if (error)
1561 				break;
1562 		}
1563 	}
1564 	return error;
1565 #undef SIN
1566 }
1567 
1568 void *
1569 in_domifattach(struct ifnet *ifp)
1570 {
1571 	struct in_ifinfo *ii;
1572 	struct lltable *llt;
1573 
1574 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1575 
1576 	llt = lltable_init(ifp, AF_INET);
1577 	if (llt != NULL) {
1578 		llt->llt_new = in_lltable_new;
1579 		llt->llt_free = in_lltable_free;
1580 		llt->llt_prefix_free = in_lltable_prefix_free;
1581 		llt->llt_rtcheck = in_lltable_rtcheck;
1582 		llt->llt_lookup = in_lltable_lookup;
1583 		llt->llt_dump = in_lltable_dump;
1584 	}
1585 	ii->ii_llt = llt;
1586 
1587 	ii->ii_igmp = igmp_domifattach(ifp);
1588 
1589 	return ii;
1590 }
1591 
1592 void
1593 in_domifdetach(struct ifnet *ifp, void *aux)
1594 {
1595 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1596 
1597 	igmp_domifdetach(ifp);
1598 	lltable_free(ii->ii_llt);
1599 	free(ii, M_IFADDR);
1600 }
1601