xref: /freebsd/sys/netinet/in.c (revision 7a1c0d963366a31363d3705697a083dd8efee077)
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_arp.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 #define	V_subnetsarelocal		VNET(subnetsarelocal)
81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, subnets_are_local, CTLFLAG_RW,
82 	&VNET_NAME(subnetsarelocal), 0,
83 	"Treat all subnets as directly connected");
84 static VNET_DEFINE(int, sameprefixcarponly);
85 #define	V_sameprefixcarponly		VNET(sameprefixcarponly)
86 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
87 	&VNET_NAME(sameprefixcarponly), 0,
88 	"Refuse to create same prefixes on different interfaces");
89 
90 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
91 #define	V_ripcbinfo			VNET(ripcbinfo)
92 
93 VNET_DECLARE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
94 #define	V_arpstat		VNET(arpstat)
95 
96 /*
97  * Return 1 if an internet address is for a ``local'' host
98  * (one to which we have a connection).  If subnetsarelocal
99  * is true, this includes other subnets of the local net.
100  * Otherwise, it includes only the directly-connected (sub)nets.
101  */
102 int
103 in_localaddr(struct in_addr in)
104 {
105 	register u_long i = ntohl(in.s_addr);
106 	register struct in_ifaddr *ia;
107 
108 	IN_IFADDR_RLOCK();
109 	if (V_subnetsarelocal) {
110 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
111 			if ((i & ia->ia_netmask) == ia->ia_net) {
112 				IN_IFADDR_RUNLOCK();
113 				return (1);
114 			}
115 		}
116 	} else {
117 		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
118 			if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
119 				IN_IFADDR_RUNLOCK();
120 				return (1);
121 			}
122 		}
123 	}
124 	IN_IFADDR_RUNLOCK();
125 	return (0);
126 }
127 
128 /*
129  * Return 1 if an internet address is for the local host and configured
130  * on one of its interfaces.
131  */
132 int
133 in_localip(struct in_addr in)
134 {
135 	struct in_ifaddr *ia;
136 
137 	IN_IFADDR_RLOCK();
138 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
139 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
140 			IN_IFADDR_RUNLOCK();
141 			return (1);
142 		}
143 	}
144 	IN_IFADDR_RUNLOCK();
145 	return (0);
146 }
147 
148 /*
149  * Determine whether an IP address is in a reserved set of addresses
150  * that may not be forwarded, or whether datagrams to that destination
151  * may be forwarded.
152  */
153 int
154 in_canforward(struct in_addr in)
155 {
156 	register u_long i = ntohl(in.s_addr);
157 	register u_long net;
158 
159 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
160 		return (0);
161 	if (IN_CLASSA(i)) {
162 		net = i & IN_CLASSA_NET;
163 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
164 			return (0);
165 	}
166 	return (1);
167 }
168 
169 /*
170  * Trim a mask in a sockaddr
171  */
172 static void
173 in_socktrim(struct sockaddr_in *ap)
174 {
175     register char *cplim = (char *) &ap->sin_addr;
176     register char *cp = (char *) (&ap->sin_addr + 1);
177 
178     ap->sin_len = 0;
179     while (--cp >= cplim)
180 	if (*cp) {
181 	    (ap)->sin_len = cp - (char *) (ap) + 1;
182 	    break;
183 	}
184 }
185 
186 static int
187 in_mask2len(mask)
188 	struct in_addr *mask;
189 {
190 	int x, y;
191 	u_char *p;
192 
193 	p = (u_char *)mask;
194 	for (x = 0; x < sizeof(*mask); x++) {
195 		if (p[x] != 0xff)
196 			break;
197 	}
198 	y = 0;
199 	if (x < sizeof(*mask)) {
200 		for (y = 0; y < 8; y++) {
201 			if ((p[x] & (0x80 >> y)) == 0)
202 				break;
203 		}
204 	}
205 	return (x * 8 + y);
206 }
207 
208 static void
209 in_len2mask(struct in_addr *mask, int len)
210 {
211 	int i;
212 	u_char *p;
213 
214 	p = (u_char *)mask;
215 	bzero(mask, sizeof(*mask));
216 	for (i = 0; i < len / 8; i++)
217 		p[i] = 0xff;
218 	if (len % 8)
219 		p[i] = (0xff00 >> (len % 8)) & 0xff;
220 }
221 
222 /*
223  * Generic internet control operations (ioctl's).
224  *
225  * ifp is NULL if not an interface-specific ioctl.
226  */
227 /* ARGSUSED */
228 int
229 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
230     struct thread *td)
231 {
232 	register struct ifreq *ifr = (struct ifreq *)data;
233 	register struct in_ifaddr *ia, *iap;
234 	register struct ifaddr *ifa;
235 	struct in_addr allhosts_addr;
236 	struct in_addr dst;
237 	struct in_ifinfo *ii;
238 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
239 	struct sockaddr_in oldaddr;
240 	int error, hostIsNew, iaIsNew, maskIsNew;
241 	int iaIsFirst;
242 
243 	ia = NULL;
244 	iaIsFirst = 0;
245 	iaIsNew = 0;
246 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
247 
248 	/*
249 	 * Filter out ioctls we implement directly; forward the rest on to
250 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
251 	 */
252 	switch (cmd) {
253 	case SIOCAIFADDR:
254 	case SIOCDIFADDR:
255 	case SIOCGIFADDR:
256 	case SIOCGIFBRDADDR:
257 	case SIOCGIFDSTADDR:
258 	case SIOCGIFNETMASK:
259 	case SIOCSIFADDR:
260 	case SIOCSIFBRDADDR:
261 	case SIOCSIFDSTADDR:
262 	case SIOCSIFNETMASK:
263 		break;
264 
265 	case SIOCALIFADDR:
266 		if (td != NULL) {
267 			error = priv_check(td, PRIV_NET_ADDIFADDR);
268 			if (error)
269 				return (error);
270 		}
271 		if (ifp == NULL)
272 			return (EINVAL);
273 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
274 
275 	case SIOCDLIFADDR:
276 		if (td != NULL) {
277 			error = priv_check(td, PRIV_NET_DELIFADDR);
278 			if (error)
279 				return (error);
280 		}
281 		if (ifp == NULL)
282 			return (EINVAL);
283 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
284 
285 	case SIOCGLIFADDR:
286 		if (ifp == NULL)
287 			return (EINVAL);
288 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
289 
290 	default:
291 		if (ifp == NULL || ifp->if_ioctl == NULL)
292 			return (EOPNOTSUPP);
293 		return ((*ifp->if_ioctl)(ifp, cmd, data));
294 	}
295 
296 	if (ifp == NULL)
297 		return (EADDRNOTAVAIL);
298 
299 	/*
300 	 * Security checks before we get involved in any work.
301 	 */
302 	switch (cmd) {
303 	case SIOCAIFADDR:
304 	case SIOCSIFADDR:
305 	case SIOCSIFBRDADDR:
306 	case SIOCSIFNETMASK:
307 	case SIOCSIFDSTADDR:
308 		if (td != NULL) {
309 			error = priv_check(td, PRIV_NET_ADDIFADDR);
310 			if (error)
311 				return (error);
312 		}
313 		break;
314 
315 	case SIOCDIFADDR:
316 		if (td != NULL) {
317 			error = priv_check(td, PRIV_NET_DELIFADDR);
318 			if (error)
319 				return (error);
320 		}
321 		break;
322 	}
323 
324 	/*
325 	 * Find address for this interface, if it exists.
326 	 *
327 	 * If an alias address was specified, find that one instead of the
328 	 * first one on the interface, if possible.
329 	 */
330 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
331 	IN_IFADDR_RLOCK();
332 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
333 		if (iap->ia_ifp == ifp &&
334 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
335 			if (td == NULL || prison_check_ip4(td->td_ucred,
336 			    &dst) == 0)
337 				ia = iap;
338 			break;
339 		}
340 	}
341 	if (ia != NULL)
342 		ifa_ref(&ia->ia_ifa);
343 	IN_IFADDR_RUNLOCK();
344 	if (ia == NULL) {
345 		IF_ADDR_LOCK(ifp);
346 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
347 			iap = ifatoia(ifa);
348 			if (iap->ia_addr.sin_family == AF_INET) {
349 				if (td != NULL &&
350 				    prison_check_ip4(td->td_ucred,
351 				    &iap->ia_addr.sin_addr) != 0)
352 					continue;
353 				ia = iap;
354 				break;
355 			}
356 		}
357 		if (ia != NULL)
358 			ifa_ref(&ia->ia_ifa);
359 		IF_ADDR_UNLOCK(ifp);
360 	}
361 	if (ia == NULL)
362 		iaIsFirst = 1;
363 
364 	error = 0;
365 	switch (cmd) {
366 	case SIOCAIFADDR:
367 	case SIOCDIFADDR:
368 		if (ifra->ifra_addr.sin_family == AF_INET) {
369 			struct in_ifaddr *oia;
370 
371 			IN_IFADDR_RLOCK();
372 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
373 				if (ia->ia_ifp == ifp  &&
374 				    ia->ia_addr.sin_addr.s_addr ==
375 				    ifra->ifra_addr.sin_addr.s_addr)
376 					break;
377 			}
378 			if (ia != NULL && ia != oia)
379 				ifa_ref(&ia->ia_ifa);
380 			if (oia != NULL && ia != oia)
381 				ifa_free(&oia->ia_ifa);
382 			IN_IFADDR_RUNLOCK();
383 			if ((ifp->if_flags & IFF_POINTOPOINT)
384 			    && (cmd == SIOCAIFADDR)
385 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
386 				== INADDR_ANY)) {
387 				error = EDESTADDRREQ;
388 				goto out;
389 			}
390 		}
391 		if (cmd == SIOCDIFADDR && ia == NULL) {
392 			error = EADDRNOTAVAIL;
393 			goto out;
394 		}
395 		/* FALLTHROUGH */
396 	case SIOCSIFADDR:
397 	case SIOCSIFNETMASK:
398 	case SIOCSIFDSTADDR:
399 		if (ia == NULL) {
400 			ia = (struct in_ifaddr *)
401 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
402 				    M_ZERO);
403 			if (ia == NULL) {
404 				error = ENOBUFS;
405 				goto out;
406 			}
407 
408 			ifa = &ia->ia_ifa;
409 			ifa_init(ifa);
410 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
411 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
412 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
413 
414 			ia->ia_sockmask.sin_len = 8;
415 			ia->ia_sockmask.sin_family = AF_INET;
416 			if (ifp->if_flags & IFF_BROADCAST) {
417 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
418 				ia->ia_broadaddr.sin_family = AF_INET;
419 			}
420 			ia->ia_ifp = ifp;
421 
422 			ifa_ref(ifa);			/* if_addrhead */
423 			IF_ADDR_LOCK(ifp);
424 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
425 			IF_ADDR_UNLOCK(ifp);
426 			ifa_ref(ifa);			/* in_ifaddrhead */
427 			IN_IFADDR_WLOCK();
428 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
429 			IN_IFADDR_WUNLOCK();
430 			iaIsNew = 1;
431 		}
432 		break;
433 
434 	case SIOCSIFBRDADDR:
435 	case SIOCGIFADDR:
436 	case SIOCGIFNETMASK:
437 	case SIOCGIFDSTADDR:
438 	case SIOCGIFBRDADDR:
439 		if (ia == NULL) {
440 			error = EADDRNOTAVAIL;
441 			goto out;
442 		}
443 		break;
444 	}
445 
446 	/*
447 	 * Most paths in this switch return directly or via out.  Only paths
448 	 * that remove the address break in order to hit common removal code.
449 	 */
450 	switch (cmd) {
451 	case SIOCGIFADDR:
452 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
453 		goto out;
454 
455 	case SIOCGIFBRDADDR:
456 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
457 			error = EINVAL;
458 			goto out;
459 		}
460 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
461 		goto out;
462 
463 	case SIOCGIFDSTADDR:
464 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
465 			error = EINVAL;
466 			goto out;
467 		}
468 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
469 		goto out;
470 
471 	case SIOCGIFNETMASK:
472 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
473 		goto out;
474 
475 	case SIOCSIFDSTADDR:
476 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
477 			error = EINVAL;
478 			goto out;
479 		}
480 		oldaddr = ia->ia_dstaddr;
481 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
482 		if (ifp->if_ioctl != NULL) {
483 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
484 			    (caddr_t)ia);
485 			if (error) {
486 				ia->ia_dstaddr = oldaddr;
487 				goto out;
488 			}
489 		}
490 		if (ia->ia_flags & IFA_ROUTE) {
491 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
492 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
493 			ia->ia_ifa.ifa_dstaddr =
494 					(struct sockaddr *)&ia->ia_dstaddr;
495 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
496 		}
497 		goto out;
498 
499 	case SIOCSIFBRDADDR:
500 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
501 			error = EINVAL;
502 			goto out;
503 		}
504 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
505 		goto out;
506 
507 	case SIOCSIFADDR:
508 		error = in_ifinit(ifp, ia,
509 		    (struct sockaddr_in *) &ifr->ifr_addr, 1);
510 		if (error != 0 && iaIsNew)
511 			break;
512 		if (error == 0) {
513 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
514 			if (iaIsFirst &&
515 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
516 				error = in_joingroup(ifp, &allhosts_addr,
517 				    NULL, &ii->ii_allhosts);
518 			}
519 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
520 		}
521 		error = 0;
522 		goto out;
523 
524 	case SIOCSIFNETMASK:
525 		ia->ia_sockmask.sin_addr = ifra->ifra_addr.sin_addr;
526 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
527 		goto out;
528 
529 	case SIOCAIFADDR:
530 		maskIsNew = 0;
531 		hostIsNew = 1;
532 		error = 0;
533 		if (ia->ia_addr.sin_family == AF_INET) {
534 			if (ifra->ifra_addr.sin_len == 0) {
535 				ifra->ifra_addr = ia->ia_addr;
536 				hostIsNew = 0;
537 			} else if (ifra->ifra_addr.sin_addr.s_addr ==
538 					       ia->ia_addr.sin_addr.s_addr)
539 				hostIsNew = 0;
540 		}
541 		if (ifra->ifra_mask.sin_len) {
542 			/*
543 			 * QL: XXX
544 			 * Need to scrub the prefix here in case
545 			 * the issued command is SIOCAIFADDR with
546 			 * the same address, but with a different
547 			 * prefix length. And if the prefix length
548 			 * is the same as before, then the call is
549 			 * un-necessarily executed here.
550 			 */
551 			in_ifscrub(ifp, ia);
552 			ia->ia_sockmask = ifra->ifra_mask;
553 			ia->ia_sockmask.sin_family = AF_INET;
554 			ia->ia_subnetmask =
555 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
556 			maskIsNew = 1;
557 		}
558 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
559 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
560 			in_ifscrub(ifp, ia);
561 			ia->ia_dstaddr = ifra->ifra_dstaddr;
562 			maskIsNew  = 1; /* We lie; but the effect's the same */
563 		}
564 		if (ifra->ifra_addr.sin_family == AF_INET &&
565 		    (hostIsNew || maskIsNew))
566 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0);
567 		if (error != 0 && iaIsNew)
568 			break;
569 
570 		if ((ifp->if_flags & IFF_BROADCAST) &&
571 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
572 			ia->ia_broadaddr = ifra->ifra_broadaddr;
573 		if (error == 0) {
574 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
575 			if (iaIsFirst &&
576 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
577 				error = in_joingroup(ifp, &allhosts_addr,
578 				    NULL, &ii->ii_allhosts);
579 			}
580 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
581 		}
582 		goto out;
583 
584 	case SIOCDIFADDR:
585 		/*
586 		 * in_ifscrub kills the interface route.
587 		 */
588 		in_ifscrub(ifp, ia);
589 
590 		/*
591 		 * in_ifadown gets rid of all the rest of
592 		 * the routes.  This is not quite the right
593 		 * thing to do, but at least if we are running
594 		 * a routing process they will come back.
595 		 */
596 		in_ifadown(&ia->ia_ifa, 1);
597 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
598 		error = 0;
599 		break;
600 
601 	default:
602 		panic("in_control: unsupported ioctl");
603 	}
604 
605 	IF_ADDR_LOCK(ifp);
606 	/* Re-check that ia is still part of the list. */
607 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
608 		if (ifa == &ia->ia_ifa)
609 			break;
610 	}
611 	if (ifa == NULL) {
612 		/*
613 		 * If we lost the race with another thread, there is no need to
614 		 * try it again for the next loop as there is no other exit
615 		 * path between here and out.
616 		 */
617 		IF_ADDR_UNLOCK(ifp);
618 		error = EADDRNOTAVAIL;
619 		goto out;
620 	}
621 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
622 	IF_ADDR_UNLOCK(ifp);
623 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
624 
625 	IN_IFADDR_WLOCK();
626 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
627 	if (ia->ia_addr.sin_family == AF_INET) {
628 		struct in_ifaddr *if_ia;
629 
630 		LIST_REMOVE(ia, ia_hash);
631 		IN_IFADDR_WUNLOCK();
632 		/*
633 		 * If this is the last IPv4 address configured on this
634 		 * interface, leave the all-hosts group.
635 		 * No state-change report need be transmitted.
636 		 */
637 		if_ia = NULL;
638 		IFP_TO_IA(ifp, if_ia);
639 		if (if_ia == NULL) {
640 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
641 			IN_MULTI_LOCK();
642 			if (ii->ii_allhosts) {
643 				(void)in_leavegroup_locked(ii->ii_allhosts,
644 				    NULL);
645 				ii->ii_allhosts = NULL;
646 			}
647 			IN_MULTI_UNLOCK();
648 		} else
649 			ifa_free(&if_ia->ia_ifa);
650 	} else
651 		IN_IFADDR_WUNLOCK();
652 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
653 out:
654 	if (ia != NULL)
655 		ifa_free(&ia->ia_ifa);
656 	return (error);
657 }
658 
659 /*
660  * SIOC[GAD]LIFADDR.
661  *	SIOCGLIFADDR: get first address. (?!?)
662  *	SIOCGLIFADDR with IFLR_PREFIX:
663  *		get first address that matches the specified prefix.
664  *	SIOCALIFADDR: add the specified address.
665  *	SIOCALIFADDR with IFLR_PREFIX:
666  *		EINVAL since we can't deduce hostid part of the address.
667  *	SIOCDLIFADDR: delete the specified address.
668  *	SIOCDLIFADDR with IFLR_PREFIX:
669  *		delete the first address that matches the specified prefix.
670  * return values:
671  *	EINVAL on invalid parameters
672  *	EADDRNOTAVAIL on prefix match failed/specified address not found
673  *	other values may be returned from in_ioctl()
674  */
675 static int
676 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
677     struct ifnet *ifp, struct thread *td)
678 {
679 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
680 	struct ifaddr *ifa;
681 
682 	/* sanity checks */
683 	if (data == NULL || ifp == NULL) {
684 		panic("invalid argument to in_lifaddr_ioctl");
685 		/*NOTRECHED*/
686 	}
687 
688 	switch (cmd) {
689 	case SIOCGLIFADDR:
690 		/* address must be specified on GET with IFLR_PREFIX */
691 		if ((iflr->flags & IFLR_PREFIX) == 0)
692 			break;
693 		/*FALLTHROUGH*/
694 	case SIOCALIFADDR:
695 	case SIOCDLIFADDR:
696 		/* address must be specified on ADD and DELETE */
697 		if (iflr->addr.ss_family != AF_INET)
698 			return (EINVAL);
699 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
700 			return (EINVAL);
701 		/* XXX need improvement */
702 		if (iflr->dstaddr.ss_family
703 		 && iflr->dstaddr.ss_family != AF_INET)
704 			return (EINVAL);
705 		if (iflr->dstaddr.ss_family
706 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
707 			return (EINVAL);
708 		break;
709 	default: /*shouldn't happen*/
710 		return (EOPNOTSUPP);
711 	}
712 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
713 		return (EINVAL);
714 
715 	switch (cmd) {
716 	case SIOCALIFADDR:
717 	    {
718 		struct in_aliasreq ifra;
719 
720 		if (iflr->flags & IFLR_PREFIX)
721 			return (EINVAL);
722 
723 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
724 		bzero(&ifra, sizeof(ifra));
725 		bcopy(iflr->iflr_name, ifra.ifra_name,
726 			sizeof(ifra.ifra_name));
727 
728 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
729 
730 		if (iflr->dstaddr.ss_family) {	/*XXX*/
731 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
732 				iflr->dstaddr.ss_len);
733 		}
734 
735 		ifra.ifra_mask.sin_family = AF_INET;
736 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
737 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
738 
739 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
740 	    }
741 	case SIOCGLIFADDR:
742 	case SIOCDLIFADDR:
743 	    {
744 		struct in_ifaddr *ia;
745 		struct in_addr mask, candidate, match;
746 		struct sockaddr_in *sin;
747 
748 		bzero(&mask, sizeof(mask));
749 		bzero(&match, sizeof(match));
750 		if (iflr->flags & IFLR_PREFIX) {
751 			/* lookup a prefix rather than address. */
752 			in_len2mask(&mask, iflr->prefixlen);
753 
754 			sin = (struct sockaddr_in *)&iflr->addr;
755 			match.s_addr = sin->sin_addr.s_addr;
756 			match.s_addr &= mask.s_addr;
757 
758 			/* if you set extra bits, that's wrong */
759 			if (match.s_addr != sin->sin_addr.s_addr)
760 				return (EINVAL);
761 
762 		} else {
763 			/* on getting an address, take the 1st match */
764 			/* on deleting an address, do exact match */
765 			if (cmd != SIOCGLIFADDR) {
766 				in_len2mask(&mask, 32);
767 				sin = (struct sockaddr_in *)&iflr->addr;
768 				match.s_addr = sin->sin_addr.s_addr;
769 			}
770 		}
771 
772 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
773 			if (ifa->ifa_addr->sa_family != AF_INET6)
774 				continue;
775 			if (match.s_addr == 0)
776 				break;
777 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
778 			candidate.s_addr &= mask.s_addr;
779 			if (candidate.s_addr == match.s_addr)
780 				break;
781 		}
782 		if (ifa == NULL)
783 			return (EADDRNOTAVAIL);
784 		ia = (struct in_ifaddr *)ifa;
785 
786 		if (cmd == SIOCGLIFADDR) {
787 			/* fill in the if_laddrreq structure */
788 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
789 
790 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
791 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
792 					ia->ia_dstaddr.sin_len);
793 			} else
794 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
795 
796 			iflr->prefixlen =
797 				in_mask2len(&ia->ia_sockmask.sin_addr);
798 
799 			iflr->flags = 0;	/*XXX*/
800 
801 			return (0);
802 		} else {
803 			struct in_aliasreq ifra;
804 
805 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
806 			bzero(&ifra, sizeof(ifra));
807 			bcopy(iflr->iflr_name, ifra.ifra_name,
808 				sizeof(ifra.ifra_name));
809 
810 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
811 				ia->ia_addr.sin_len);
812 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
813 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
814 					ia->ia_dstaddr.sin_len);
815 			}
816 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
817 				ia->ia_sockmask.sin_len);
818 
819 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
820 			    ifp, td));
821 		}
822 	    }
823 	}
824 
825 	return (EOPNOTSUPP);	/*just for safety*/
826 }
827 
828 /*
829  * Delete any existing route for an interface.
830  */
831 void
832 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
833 {
834 
835 	in_scrubprefix(ia);
836 }
837 
838 /*
839  * Initialize an interface's internet address
840  * and routing table entry.
841  */
842 static int
843 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
844     int scrub)
845 {
846 	register u_long i = ntohl(sin->sin_addr.s_addr);
847 	struct sockaddr_in oldaddr;
848 	int s = splimp(), flags = RTF_UP, error = 0;
849 
850 	oldaddr = ia->ia_addr;
851 	if (oldaddr.sin_family == AF_INET)
852 		LIST_REMOVE(ia, ia_hash);
853 	ia->ia_addr = *sin;
854 	if (ia->ia_addr.sin_family == AF_INET) {
855 		IN_IFADDR_WLOCK();
856 		LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
857 		    ia, ia_hash);
858 		IN_IFADDR_WUNLOCK();
859 	}
860 	/*
861 	 * Give the interface a chance to initialize
862 	 * if this is its first address,
863 	 * and to validate the address if necessary.
864 	 */
865 	if (ifp->if_ioctl != NULL) {
866 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
867 		if (error) {
868 			splx(s);
869 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
870 			ia->ia_addr = oldaddr;
871 			IN_IFADDR_WLOCK();
872 			if (ia->ia_addr.sin_family == AF_INET)
873 				LIST_INSERT_HEAD(INADDR_HASH(
874 				    ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
875 			else
876 				/*
877 				 * If oldaddr family is not AF_INET (e.g.
878 				 * interface has been just created) in_control
879 				 * does not call LIST_REMOVE, and we end up
880 				 * with bogus ia entries in hash
881 				 */
882 				LIST_REMOVE(ia, ia_hash);
883 			IN_IFADDR_WUNLOCK();
884 			return (error);
885 		}
886 	}
887 	splx(s);
888 	if (scrub) {
889 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&oldaddr;
890 		in_ifscrub(ifp, ia);
891 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
892 	}
893 	if (IN_CLASSA(i))
894 		ia->ia_netmask = IN_CLASSA_NET;
895 	else if (IN_CLASSB(i))
896 		ia->ia_netmask = IN_CLASSB_NET;
897 	else
898 		ia->ia_netmask = IN_CLASSC_NET;
899 	/*
900 	 * The subnet mask usually includes at least the standard network part,
901 	 * but may may be smaller in the case of supernetting.
902 	 * If it is set, we believe it.
903 	 */
904 	if (ia->ia_subnetmask == 0) {
905 		ia->ia_subnetmask = ia->ia_netmask;
906 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
907 	} else
908 		ia->ia_netmask &= ia->ia_subnetmask;
909 	ia->ia_net = i & ia->ia_netmask;
910 	ia->ia_subnet = i & ia->ia_subnetmask;
911 	in_socktrim(&ia->ia_sockmask);
912 	/*
913 	 * XXX: carp(4) does not have interface route
914 	 */
915 	if (ifp->if_type == IFT_CARP)
916 		return (0);
917 	/*
918 	 * Add route for the network.
919 	 */
920 	ia->ia_ifa.ifa_metric = ifp->if_metric;
921 	if (ifp->if_flags & IFF_BROADCAST) {
922 		ia->ia_broadaddr.sin_addr.s_addr =
923 			htonl(ia->ia_subnet | ~ia->ia_subnetmask);
924 		ia->ia_netbroadcast.s_addr =
925 			htonl(ia->ia_net | ~ ia->ia_netmask);
926 	} else if (ifp->if_flags & IFF_LOOPBACK) {
927 		ia->ia_dstaddr = ia->ia_addr;
928 		flags |= RTF_HOST;
929 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
930 		if (ia->ia_dstaddr.sin_family != AF_INET)
931 			return (0);
932 		flags |= RTF_HOST;
933 	}
934 	if ((error = in_addprefix(ia, flags)) != 0)
935 		return (error);
936 
937 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
938 		return (0);
939 
940 	if (ifp->if_flags & IFF_POINTOPOINT) {
941 		if (ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
942 			return (0);
943 	}
944 
945 
946 	/*
947 	 * add a loopback route to self
948 	 */
949 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
950 		struct route ia_ro;
951 
952 		bzero(&ia_ro, sizeof(ia_ro));
953 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
954 		rtalloc_ign_fib(&ia_ro, 0, 0);
955 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
956 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
957 			RT_LOCK(ia_ro.ro_rt);
958 			RT_ADDREF(ia_ro.ro_rt);
959 			RTFREE_LOCKED(ia_ro.ro_rt);
960 		} else
961 			error = ifa_add_loopback_route((struct ifaddr *)ia,
962 				       (struct sockaddr *)&ia->ia_addr);
963 		if (error == 0)
964 			ia->ia_flags |= IFA_RTSELF;
965 		if (ia_ro.ro_rt != NULL)
966 			RTFREE(ia_ro.ro_rt);
967 	}
968 
969 	return (error);
970 }
971 
972 #define rtinitflags(x) \
973 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
974 	    ? RTF_HOST : 0)
975 
976 /*
977  * Generate a routing message when inserting or deleting
978  * an interface address alias.
979  */
980 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
981     struct in_ifaddr *target)
982 {
983 	struct route pfx_ro;
984 	struct sockaddr_in *pfx_addr;
985 	struct rtentry msg_rt;
986 
987 	/* QL: XXX
988 	 * This is a bit questionable because there is no
989 	 * additional route entry added/deleted for an address
990 	 * alias. Therefore this route report is inaccurate.
991 	 */
992 	bzero(&pfx_ro, sizeof(pfx_ro));
993 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
994 	pfx_addr->sin_len = sizeof(*pfx_addr);
995 	pfx_addr->sin_family = AF_INET;
996 	pfx_addr->sin_addr = *prefix;
997 	rtalloc_ign_fib(&pfx_ro, 0, 0);
998 	if (pfx_ro.ro_rt != NULL) {
999 		msg_rt = *pfx_ro.ro_rt;
1000 
1001 		/* QL: XXX
1002 		 * Point the gateway to the new interface
1003 		 * address as if a new prefix route entry has
1004 		 * been added through the new address alias.
1005 		 * All other parts of the rtentry is accurate,
1006 		 * e.g., rt_key, rt_mask, rt_ifp etc.
1007 		 */
1008 		msg_rt.rt_gateway =
1009 			(struct sockaddr *)&target->ia_addr;
1010 		rt_newaddrmsg(cmd,
1011 			      (struct ifaddr *)target,
1012 			      0, &msg_rt);
1013 		RTFREE(pfx_ro.ro_rt);
1014 	}
1015 	return;
1016 }
1017 
1018 /*
1019  * Check if we have a route for the given prefix already or add one accordingly.
1020  */
1021 static int
1022 in_addprefix(struct in_ifaddr *target, int flags)
1023 {
1024 	struct in_ifaddr *ia;
1025 	struct in_addr prefix, mask, p, m;
1026 	int error;
1027 
1028 	if ((flags & RTF_HOST) != 0) {
1029 		prefix = target->ia_dstaddr.sin_addr;
1030 		mask.s_addr = 0;
1031 	} else {
1032 		prefix = target->ia_addr.sin_addr;
1033 		mask = target->ia_sockmask.sin_addr;
1034 		prefix.s_addr &= mask.s_addr;
1035 	}
1036 
1037 	IN_IFADDR_RLOCK();
1038 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1039 		if (rtinitflags(ia)) {
1040 			p = ia->ia_addr.sin_addr;
1041 
1042 			if (prefix.s_addr != p.s_addr)
1043 				continue;
1044 		} else {
1045 			p = ia->ia_addr.sin_addr;
1046 			m = ia->ia_sockmask.sin_addr;
1047 			p.s_addr &= m.s_addr;
1048 
1049 			if (prefix.s_addr != p.s_addr ||
1050 			    mask.s_addr != m.s_addr)
1051 				continue;
1052 		}
1053 
1054 		/*
1055 		 * If we got a matching prefix route inserted by other
1056 		 * interface address, we are done here.
1057 		 */
1058 		if (ia->ia_flags & IFA_ROUTE) {
1059 #ifdef RADIX_MPATH
1060 			if (ia->ia_addr.sin_addr.s_addr ==
1061 			    target->ia_addr.sin_addr.s_addr) {
1062 				IN_IFADDR_RUNLOCK();
1063 				return (EEXIST);
1064 			} else
1065 				break;
1066 #endif
1067 			if (V_sameprefixcarponly &&
1068 			    target->ia_ifp->if_type != IFT_CARP &&
1069 			    ia->ia_ifp->if_type != IFT_CARP) {
1070 				IN_IFADDR_RUNLOCK();
1071 				return (EEXIST);
1072 			} else {
1073 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
1074 				IN_IFADDR_RUNLOCK();
1075 				return (0);
1076 			}
1077 		}
1078 	}
1079 	IN_IFADDR_RUNLOCK();
1080 
1081 	/*
1082 	 * No-one seem to have this prefix route, so we try to insert it.
1083 	 */
1084 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1085 	if (!error)
1086 		target->ia_flags |= IFA_ROUTE;
1087 	return (error);
1088 }
1089 
1090 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1091 
1092 /*
1093  * If there is no other address in the system that can serve a route to the
1094  * same prefix, remove the route.  Hand over the route to the new address
1095  * otherwise.
1096  */
1097 static int
1098 in_scrubprefix(struct in_ifaddr *target)
1099 {
1100 	struct in_ifaddr *ia;
1101 	struct in_addr prefix, mask, p;
1102 	int error = 0;
1103 	struct sockaddr_in prefix0, mask0;
1104 
1105 	/*
1106 	 * Remove the loopback route to the interface address.
1107 	 * The "useloopback" setting is not consulted because if the
1108 	 * user configures an interface address, turns off this
1109 	 * setting, and then tries to delete that interface address,
1110 	 * checking the current setting of "useloopback" would leave
1111 	 * that interface address loopback route untouched, which
1112 	 * would be wrong. Therefore the interface address loopback route
1113 	 * deletion is unconditional.
1114 	 */
1115 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1116 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1117 	    (target->ia_flags & IFA_RTSELF)) {
1118 		struct route ia_ro;
1119 		int freeit = 0;
1120 
1121 		bzero(&ia_ro, sizeof(ia_ro));
1122 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1123 		rtalloc_ign_fib(&ia_ro, 0, 0);
1124 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1125 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1126 			RT_LOCK(ia_ro.ro_rt);
1127 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1128 				freeit = 1;
1129 			else
1130 				RT_REMREF(ia_ro.ro_rt);
1131 			RTFREE_LOCKED(ia_ro.ro_rt);
1132 		}
1133 		if (freeit)
1134 			error = ifa_del_loopback_route((struct ifaddr *)target,
1135 				       (struct sockaddr *)&target->ia_addr);
1136 		if (error == 0)
1137 			target->ia_flags &= ~IFA_RTSELF;
1138 		/* remove arp cache */
1139 		arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1140 	}
1141 
1142 	if (rtinitflags(target))
1143 		prefix = target->ia_dstaddr.sin_addr;
1144 	else {
1145 		prefix = target->ia_addr.sin_addr;
1146 		mask = target->ia_sockmask.sin_addr;
1147 		prefix.s_addr &= mask.s_addr;
1148 	}
1149 
1150 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1151 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1152 		return (0);
1153 	}
1154 
1155 	IN_IFADDR_RLOCK();
1156 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1157 		if (rtinitflags(ia))
1158 			p = ia->ia_dstaddr.sin_addr;
1159 		else {
1160 			p = ia->ia_addr.sin_addr;
1161 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1162 		}
1163 
1164 		if (prefix.s_addr != p.s_addr)
1165 			continue;
1166 
1167 		/*
1168 		 * If we got a matching prefix address, move IFA_ROUTE and
1169 		 * the route itself to it.  Make sure that routing daemons
1170 		 * get a heads-up.
1171 		 *
1172 		 * XXX: a special case for carp(4) interface - this should
1173 		 *      be more generally specified as an interface that
1174 		 *      doesn't support such action.
1175 		 */
1176 		if ((ia->ia_flags & IFA_ROUTE) == 0
1177 		    && (ia->ia_ifp->if_type != IFT_CARP)) {
1178 			ifa_ref(&ia->ia_ifa);
1179 			IN_IFADDR_RUNLOCK();
1180 			rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1181 			    rtinitflags(target));
1182 			target->ia_flags &= ~IFA_ROUTE;
1183 
1184 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1185 			    rtinitflags(ia) | RTF_UP);
1186 			if (error == 0)
1187 				ia->ia_flags |= IFA_ROUTE;
1188 			ifa_free(&ia->ia_ifa);
1189 			return (error);
1190 		}
1191 	}
1192 	IN_IFADDR_RUNLOCK();
1193 
1194 	/*
1195 	 * remove all L2 entries on the given prefix
1196 	 */
1197 	bzero(&prefix0, sizeof(prefix0));
1198 	prefix0.sin_len = sizeof(prefix0);
1199 	prefix0.sin_family = AF_INET;
1200 	prefix0.sin_addr.s_addr = target->ia_subnet;
1201 	bzero(&mask0, sizeof(mask0));
1202 	mask0.sin_len = sizeof(mask0);
1203 	mask0.sin_family = AF_INET;
1204 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1205 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1206 			    (struct sockaddr *)&mask0);
1207 
1208 	/*
1209 	 * As no-one seem to have this prefix, we can remove the route.
1210 	 */
1211 	rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1212 	target->ia_flags &= ~IFA_ROUTE;
1213 	return (0);
1214 }
1215 
1216 #undef rtinitflags
1217 
1218 /*
1219  * Return 1 if the address might be a local broadcast address.
1220  */
1221 int
1222 in_broadcast(struct in_addr in, struct ifnet *ifp)
1223 {
1224 	register struct ifaddr *ifa;
1225 	u_long t;
1226 
1227 	if (in.s_addr == INADDR_BROADCAST ||
1228 	    in.s_addr == INADDR_ANY)
1229 		return (1);
1230 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1231 		return (0);
1232 	t = ntohl(in.s_addr);
1233 	/*
1234 	 * Look through the list of addresses for a match
1235 	 * with a broadcast address.
1236 	 */
1237 #define ia ((struct in_ifaddr *)ifa)
1238 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1239 		if (ifa->ifa_addr->sa_family == AF_INET &&
1240 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1241 		     in.s_addr == ia->ia_netbroadcast.s_addr ||
1242 		     /*
1243 		      * Check for old-style (host 0) broadcast.
1244 		      */
1245 		     t == ia->ia_subnet || t == ia->ia_net) &&
1246 		     /*
1247 		      * Check for an all one subnetmask. These
1248 		      * only exist when an interface gets a secondary
1249 		      * address.
1250 		      */
1251 		     ia->ia_subnetmask != (u_long)0xffffffff)
1252 			    return (1);
1253 	return (0);
1254 #undef ia
1255 }
1256 
1257 /*
1258  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1259  */
1260 void
1261 in_ifdetach(struct ifnet *ifp)
1262 {
1263 
1264 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1265 	in_pcbpurgeif0(&V_udbinfo, ifp);
1266 	in_purgemaddrs(ifp);
1267 }
1268 
1269 /*
1270  * Delete all IPv4 multicast address records, and associated link-layer
1271  * multicast address records, associated with ifp.
1272  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1273  * XXX This should not race with ifma_protospec being set during
1274  * a new allocation, if it does, we have bigger problems.
1275  */
1276 static void
1277 in_purgemaddrs(struct ifnet *ifp)
1278 {
1279 	LIST_HEAD(,in_multi) purgeinms;
1280 	struct in_multi		*inm, *tinm;
1281 	struct ifmultiaddr	*ifma;
1282 
1283 	LIST_INIT(&purgeinms);
1284 	IN_MULTI_LOCK();
1285 
1286 	/*
1287 	 * Extract list of in_multi associated with the detaching ifp
1288 	 * which the PF_INET layer is about to release.
1289 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1290 	 * by code further down.
1291 	 */
1292 	IF_ADDR_LOCK(ifp);
1293 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1294 		if (ifma->ifma_addr->sa_family != AF_INET ||
1295 		    ifma->ifma_protospec == NULL)
1296 			continue;
1297 #if 0
1298 		KASSERT(ifma->ifma_protospec != NULL,
1299 		    ("%s: ifma_protospec is NULL", __func__));
1300 #endif
1301 		inm = (struct in_multi *)ifma->ifma_protospec;
1302 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1303 	}
1304 	IF_ADDR_UNLOCK(ifp);
1305 
1306 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1307 		LIST_REMOVE(inm, inm_link);
1308 		inm_release_locked(inm);
1309 	}
1310 	igmp_ifdetach(ifp);
1311 
1312 	IN_MULTI_UNLOCK();
1313 }
1314 
1315 #include <net/if_dl.h>
1316 #include <netinet/if_ether.h>
1317 
1318 struct in_llentry {
1319 	struct llentry		base;
1320 	struct sockaddr_in	l3_addr4;
1321 };
1322 
1323 static struct llentry *
1324 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1325 {
1326 	struct in_llentry *lle;
1327 
1328 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1329 	if (lle == NULL)		/* NB: caller generates msg */
1330 		return NULL;
1331 
1332 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1333 	/*
1334 	 * For IPv4 this will trigger "arpresolve" to generate
1335 	 * an ARP request.
1336 	 */
1337 	lle->base.la_expire = time_uptime; /* mark expired */
1338 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1339 	lle->base.lle_refcnt = 1;
1340 	LLE_LOCK_INIT(&lle->base);
1341 	return &lle->base;
1342 }
1343 
1344 /*
1345  * Deletes an address from the address table.
1346  * This function is called by the timer functions
1347  * such as arptimer() and nd6_llinfo_timer(), and
1348  * the caller does the locking.
1349  */
1350 static void
1351 in_lltable_free(struct lltable *llt, struct llentry *lle)
1352 {
1353 	LLE_WUNLOCK(lle);
1354 	LLE_LOCK_DESTROY(lle);
1355 	free(lle, M_LLTABLE);
1356 }
1357 
1358 
1359 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1360 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1361 
1362 static void
1363 in_lltable_prefix_free(struct lltable *llt,
1364 		       const struct sockaddr *prefix,
1365 		       const struct sockaddr *mask)
1366 {
1367 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1368 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1369 	struct llentry *lle, *next;
1370 	register int i;
1371 	size_t pkts_dropped;
1372 
1373 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1374 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1375 
1376 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1377 						     pfx, msk)) {
1378 				int canceled;
1379 
1380 				canceled = callout_drain(&lle->la_timer);
1381 				LLE_WLOCK(lle);
1382 				if (canceled)
1383 					LLE_REMREF(lle);
1384 				pkts_dropped = llentry_free(lle);
1385 				ARPSTAT_ADD(dropped, pkts_dropped);
1386 			}
1387 		}
1388 	}
1389 }
1390 
1391 
1392 static int
1393 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1394 {
1395 	struct rtentry *rt;
1396 
1397 	KASSERT(l3addr->sa_family == AF_INET,
1398 	    ("sin_family %d", l3addr->sa_family));
1399 
1400 	/* XXX rtalloc1 should take a const param */
1401 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1402 	if (rt == NULL || (!(flags & LLE_PUB) &&
1403 			   ((rt->rt_flags & RTF_GATEWAY) ||
1404 			    (rt->rt_ifp != ifp)))) {
1405 #ifdef DIAGNOSTIC
1406 		log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1407 		    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1408 #endif
1409 		if (rt != NULL)
1410 			RTFREE_LOCKED(rt);
1411 		return (EINVAL);
1412 	}
1413 	RTFREE_LOCKED(rt);
1414 	return 0;
1415 }
1416 
1417 /*
1418  * Return NULL if not found or marked for deletion.
1419  * If found return lle read locked.
1420  */
1421 static struct llentry *
1422 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1423 {
1424 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1425 	struct ifnet *ifp = llt->llt_ifp;
1426 	struct llentry *lle;
1427 	struct llentries *lleh;
1428 	u_int hashkey;
1429 
1430 	IF_AFDATA_LOCK_ASSERT(ifp);
1431 	KASSERT(l3addr->sa_family == AF_INET,
1432 	    ("sin_family %d", l3addr->sa_family));
1433 
1434 	hashkey = sin->sin_addr.s_addr;
1435 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1436 	LIST_FOREACH(lle, lleh, lle_next) {
1437 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1438 		if (lle->la_flags & LLE_DELETED)
1439 			continue;
1440 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1441 			break;
1442 	}
1443 	if (lle == NULL) {
1444 #ifdef DIAGNOSTIC
1445 		if (flags & LLE_DELETE)
1446 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1447 #endif
1448 		if (!(flags & LLE_CREATE))
1449 			return (NULL);
1450 		/*
1451 		 * A route that covers the given address must have
1452 		 * been installed 1st because we are doing a resolution,
1453 		 * verify this.
1454 		 */
1455 		if (!(flags & LLE_IFADDR) &&
1456 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1457 			goto done;
1458 
1459 		lle = in_lltable_new(l3addr, flags);
1460 		if (lle == NULL) {
1461 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1462 			goto done;
1463 		}
1464 		lle->la_flags = flags & ~LLE_CREATE;
1465 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1466 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1467 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1468 		}
1469 
1470 		lle->lle_tbl  = llt;
1471 		lle->lle_head = lleh;
1472 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1473 	} else if (flags & LLE_DELETE) {
1474 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1475 			LLE_WLOCK(lle);
1476 			lle->la_flags = LLE_DELETED;
1477 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1478 			LLE_WUNLOCK(lle);
1479 #ifdef DIAGNOSTIC
1480 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1481 #endif
1482 		}
1483 		lle = (void *)-1;
1484 
1485 	}
1486 	if (LLE_IS_VALID(lle)) {
1487 		if (flags & LLE_EXCLUSIVE)
1488 			LLE_WLOCK(lle);
1489 		else
1490 			LLE_RLOCK(lle);
1491 	}
1492 done:
1493 	return (lle);
1494 }
1495 
1496 static int
1497 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1498 {
1499 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1500 	struct ifnet *ifp = llt->llt_ifp;
1501 	struct llentry *lle;
1502 	/* XXX stack use */
1503 	struct {
1504 		struct rt_msghdr	rtm;
1505 		struct sockaddr_inarp	sin;
1506 		struct sockaddr_dl	sdl;
1507 	} arpc;
1508 	int error, i;
1509 
1510 	LLTABLE_LOCK_ASSERT();
1511 
1512 	error = 0;
1513 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1514 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1515 			struct sockaddr_dl *sdl;
1516 
1517 			/* skip deleted entries */
1518 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1519 				continue;
1520 			/* Skip if jailed and not a valid IP of the prison. */
1521 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1522 				continue;
1523 			/*
1524 			 * produce a msg made of:
1525 			 *  struct rt_msghdr;
1526 			 *  struct sockaddr_inarp; (IPv4)
1527 			 *  struct sockaddr_dl;
1528 			 */
1529 			bzero(&arpc, sizeof(arpc));
1530 			arpc.rtm.rtm_msglen = sizeof(arpc);
1531 			arpc.rtm.rtm_version = RTM_VERSION;
1532 			arpc.rtm.rtm_type = RTM_GET;
1533 			arpc.rtm.rtm_flags = RTF_UP;
1534 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1535 			arpc.sin.sin_family = AF_INET;
1536 			arpc.sin.sin_len = sizeof(arpc.sin);
1537 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1538 
1539 			/* publish */
1540 			if (lle->la_flags & LLE_PUB) {
1541 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1542 				/* proxy only */
1543 				if (lle->la_flags & LLE_PROXY)
1544 					arpc.sin.sin_other = SIN_PROXY;
1545 			}
1546 
1547 			sdl = &arpc.sdl;
1548 			sdl->sdl_family = AF_LINK;
1549 			sdl->sdl_len = sizeof(*sdl);
1550 			sdl->sdl_index = ifp->if_index;
1551 			sdl->sdl_type = ifp->if_type;
1552 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1553 				sdl->sdl_alen = ifp->if_addrlen;
1554 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1555 			} else {
1556 				sdl->sdl_alen = 0;
1557 				bzero(LLADDR(sdl), ifp->if_addrlen);
1558 			}
1559 
1560 			arpc.rtm.rtm_rmx.rmx_expire =
1561 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1562 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1563 			if (lle->la_flags & LLE_STATIC)
1564 				arpc.rtm.rtm_flags |= RTF_STATIC;
1565 			arpc.rtm.rtm_index = ifp->if_index;
1566 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1567 			if (error)
1568 				break;
1569 		}
1570 	}
1571 	return error;
1572 #undef SIN
1573 }
1574 
1575 void *
1576 in_domifattach(struct ifnet *ifp)
1577 {
1578 	struct in_ifinfo *ii;
1579 	struct lltable *llt;
1580 
1581 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1582 
1583 	llt = lltable_init(ifp, AF_INET);
1584 	if (llt != NULL) {
1585 		llt->llt_new = in_lltable_new;
1586 		llt->llt_free = in_lltable_free;
1587 		llt->llt_prefix_free = in_lltable_prefix_free;
1588 		llt->llt_rtcheck = in_lltable_rtcheck;
1589 		llt->llt_lookup = in_lltable_lookup;
1590 		llt->llt_dump = in_lltable_dump;
1591 	}
1592 	ii->ii_llt = llt;
1593 
1594 	ii->ii_igmp = igmp_domifattach(ifp);
1595 
1596 	return ii;
1597 }
1598 
1599 void
1600 in_domifdetach(struct ifnet *ifp, void *aux)
1601 {
1602 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1603 
1604 	igmp_domifdetach(ifp);
1605 	lltable_free(ii->ii_llt);
1606 	free(ii, M_IFADDR);
1607 }
1608