xref: /freebsd/sys/netinet/in.c (revision 675be9115aae86ad6b3d877155d4fd7822892105)
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 *, u_int);
74 static void	in_socktrim(struct sockaddr_in *);
75 static int	in_ifinit(struct ifnet *,
76 	    struct in_ifaddr *, struct sockaddr_in *, int, int);
77 static void	in_purgemaddrs(struct ifnet *);
78 
79 static VNET_DEFINE(int, sameprefixcarponly);
80 #define	V_sameprefixcarponly		VNET(sameprefixcarponly)
81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, same_prefix_carp_only, CTLFLAG_RW,
82 	&VNET_NAME(sameprefixcarponly), 0,
83 	"Refuse to create same prefixes on different interfaces");
84 
85 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
86 #define	V_ripcbinfo			VNET(ripcbinfo)
87 
88 VNET_DECLARE(struct arpstat, arpstat);  /* ARP statistics, see if_arp.h */
89 #define	V_arpstat		VNET(arpstat)
90 
91 /*
92  * Return 1 if an internet address is for a ``local'' host
93  * (one to which we have a connection).
94  */
95 int
96 in_localaddr(struct in_addr in)
97 {
98 	register u_long i = ntohl(in.s_addr);
99 	register struct in_ifaddr *ia;
100 
101 	IN_IFADDR_RLOCK();
102 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
103 		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
104 			IN_IFADDR_RUNLOCK();
105 			return (1);
106 		}
107 	}
108 	IN_IFADDR_RUNLOCK();
109 	return (0);
110 }
111 
112 /*
113  * Return 1 if an internet address is for the local host and configured
114  * on one of its interfaces.
115  */
116 int
117 in_localip(struct in_addr in)
118 {
119 	struct in_ifaddr *ia;
120 
121 	IN_IFADDR_RLOCK();
122 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
123 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
124 			IN_IFADDR_RUNLOCK();
125 			return (1);
126 		}
127 	}
128 	IN_IFADDR_RUNLOCK();
129 	return (0);
130 }
131 
132 /*
133  * Determine whether an IP address is in a reserved set of addresses
134  * that may not be forwarded, or whether datagrams to that destination
135  * may be forwarded.
136  */
137 int
138 in_canforward(struct in_addr in)
139 {
140 	register u_long i = ntohl(in.s_addr);
141 	register u_long net;
142 
143 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
144 		return (0);
145 	if (IN_CLASSA(i)) {
146 		net = i & IN_CLASSA_NET;
147 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
148 			return (0);
149 	}
150 	return (1);
151 }
152 
153 /*
154  * Trim a mask in a sockaddr
155  */
156 static void
157 in_socktrim(struct sockaddr_in *ap)
158 {
159     register char *cplim = (char *) &ap->sin_addr;
160     register char *cp = (char *) (&ap->sin_addr + 1);
161 
162     ap->sin_len = 0;
163     while (--cp >= cplim)
164 	if (*cp) {
165 	    (ap)->sin_len = cp - (char *) (ap) + 1;
166 	    break;
167 	}
168 }
169 
170 static int
171 in_mask2len(mask)
172 	struct in_addr *mask;
173 {
174 	int x, y;
175 	u_char *p;
176 
177 	p = (u_char *)mask;
178 	for (x = 0; x < sizeof(*mask); x++) {
179 		if (p[x] != 0xff)
180 			break;
181 	}
182 	y = 0;
183 	if (x < sizeof(*mask)) {
184 		for (y = 0; y < 8; y++) {
185 			if ((p[x] & (0x80 >> y)) == 0)
186 				break;
187 		}
188 	}
189 	return (x * 8 + y);
190 }
191 
192 static void
193 in_len2mask(struct in_addr *mask, int len)
194 {
195 	int i;
196 	u_char *p;
197 
198 	p = (u_char *)mask;
199 	bzero(mask, sizeof(*mask));
200 	for (i = 0; i < len / 8; i++)
201 		p[i] = 0xff;
202 	if (len % 8)
203 		p[i] = (0xff00 >> (len % 8)) & 0xff;
204 }
205 
206 /*
207  * Generic internet control operations (ioctl's).
208  *
209  * ifp is NULL if not an interface-specific ioctl.
210  */
211 /* ARGSUSED */
212 int
213 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
214     struct thread *td)
215 {
216 	register struct ifreq *ifr = (struct ifreq *)data;
217 	register struct in_ifaddr *ia, *iap;
218 	register struct ifaddr *ifa;
219 	struct in_addr allhosts_addr;
220 	struct in_addr dst;
221 	struct in_ifinfo *ii;
222 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
223 	struct sockaddr_in oldaddr;
224 	int error, hostIsNew, iaIsNew, maskIsNew;
225 	int iaIsFirst;
226 
227 	ia = NULL;
228 	iaIsFirst = 0;
229 	iaIsNew = 0;
230 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
231 
232 	/*
233 	 * Filter out ioctls we implement directly; forward the rest on to
234 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
235 	 */
236 	switch (cmd) {
237 	case SIOCGIFADDR:
238 	case SIOCGIFBRDADDR:
239 	case SIOCGIFDSTADDR:
240 	case SIOCGIFNETMASK:
241 	case SIOCDIFADDR:
242 		break;
243 	case SIOCAIFADDR:
244 		/*
245 		 * ifra_addr must be present and be of INET family.
246 		 * ifra_broadaddr and ifra_mask are optional.
247 		 */
248 		if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) ||
249 		    ifra->ifra_addr.sin_family != AF_INET)
250 			return (EINVAL);
251 		if (ifra->ifra_broadaddr.sin_len != 0 &&
252 		    (ifra->ifra_broadaddr.sin_len !=
253 		    sizeof(struct sockaddr_in) ||
254 		    ifra->ifra_broadaddr.sin_family != AF_INET))
255 			return (EINVAL);
256 #if 0
257 		/*
258 		 * ifconfig(8) historically doesn't set af_family for mask
259 		 * for unknown reason.
260 		 */
261 		if (ifra->ifra_mask.sin_len != 0 &&
262 		    (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) ||
263 		    ifra->ifra_mask.sin_family != AF_INET))
264 			return (EINVAL);
265 #endif
266 		break;
267 	case SIOCSIFADDR:
268 	case SIOCSIFBRDADDR:
269 	case SIOCSIFDSTADDR:
270 	case SIOCSIFNETMASK:
271 		if (ifr->ifr_addr.sa_family != AF_INET ||
272 		    ifr->ifr_addr.sa_len != sizeof(struct sockaddr_in))
273 			return (EINVAL);
274 		break;
275 
276 	case SIOCALIFADDR:
277 		if (td != NULL) {
278 			error = priv_check(td, PRIV_NET_ADDIFADDR);
279 			if (error)
280 				return (error);
281 		}
282 		if (ifp == NULL)
283 			return (EINVAL);
284 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
285 
286 	case SIOCDLIFADDR:
287 		if (td != NULL) {
288 			error = priv_check(td, PRIV_NET_DELIFADDR);
289 			if (error)
290 				return (error);
291 		}
292 		if (ifp == NULL)
293 			return (EINVAL);
294 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
295 
296 	case SIOCGLIFADDR:
297 		if (ifp == NULL)
298 			return (EINVAL);
299 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
300 
301 	default:
302 		if (ifp == NULL || ifp->if_ioctl == NULL)
303 			return (EOPNOTSUPP);
304 		return ((*ifp->if_ioctl)(ifp, cmd, data));
305 	}
306 
307 	if (ifp == NULL)
308 		return (EADDRNOTAVAIL);
309 
310 	/*
311 	 * Security checks before we get involved in any work.
312 	 */
313 	switch (cmd) {
314 	case SIOCAIFADDR:
315 	case SIOCSIFADDR:
316 	case SIOCSIFBRDADDR:
317 	case SIOCSIFNETMASK:
318 	case SIOCSIFDSTADDR:
319 		if (td != NULL) {
320 			error = priv_check(td, PRIV_NET_ADDIFADDR);
321 			if (error)
322 				return (error);
323 		}
324 		break;
325 
326 	case SIOCDIFADDR:
327 		if (td != NULL) {
328 			error = priv_check(td, PRIV_NET_DELIFADDR);
329 			if (error)
330 				return (error);
331 		}
332 		break;
333 	}
334 
335 	/*
336 	 * Find address for this interface, if it exists.
337 	 *
338 	 * If an alias address was specified, find that one instead of the
339 	 * first one on the interface, if possible.
340 	 */
341 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
342 	IN_IFADDR_RLOCK();
343 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
344 		if (iap->ia_ifp == ifp &&
345 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
346 			if (td == NULL || prison_check_ip4(td->td_ucred,
347 			    &dst) == 0)
348 				ia = iap;
349 			break;
350 		}
351 	}
352 	if (ia != NULL)
353 		ifa_ref(&ia->ia_ifa);
354 	IN_IFADDR_RUNLOCK();
355 	if (ia == NULL) {
356 		IF_ADDR_LOCK(ifp);
357 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
358 			iap = ifatoia(ifa);
359 			if (iap->ia_addr.sin_family == AF_INET) {
360 				if (td != NULL &&
361 				    prison_check_ip4(td->td_ucred,
362 				    &iap->ia_addr.sin_addr) != 0)
363 					continue;
364 				ia = iap;
365 				break;
366 			}
367 		}
368 		if (ia != NULL)
369 			ifa_ref(&ia->ia_ifa);
370 		IF_ADDR_UNLOCK(ifp);
371 	}
372 	if (ia == NULL)
373 		iaIsFirst = 1;
374 
375 	error = 0;
376 	switch (cmd) {
377 	case SIOCAIFADDR:
378 	case SIOCDIFADDR:
379 		if (ifra->ifra_addr.sin_family == AF_INET) {
380 			struct in_ifaddr *oia;
381 
382 			IN_IFADDR_RLOCK();
383 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
384 				if (ia->ia_ifp == ifp  &&
385 				    ia->ia_addr.sin_addr.s_addr ==
386 				    ifra->ifra_addr.sin_addr.s_addr)
387 					break;
388 			}
389 			if (ia != NULL && ia != oia)
390 				ifa_ref(&ia->ia_ifa);
391 			if (oia != NULL && ia != oia)
392 				ifa_free(&oia->ia_ifa);
393 			IN_IFADDR_RUNLOCK();
394 			if ((ifp->if_flags & IFF_POINTOPOINT)
395 			    && (cmd == SIOCAIFADDR)
396 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
397 				== INADDR_ANY)) {
398 				error = EDESTADDRREQ;
399 				goto out;
400 			}
401 		}
402 		if (cmd == SIOCDIFADDR && ia == NULL) {
403 			error = EADDRNOTAVAIL;
404 			goto out;
405 		}
406 		/* FALLTHROUGH */
407 	case SIOCSIFADDR:
408 	case SIOCSIFNETMASK:
409 	case SIOCSIFDSTADDR:
410 		if (ia == NULL) {
411 			ia = (struct in_ifaddr *)
412 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
413 				    M_ZERO);
414 			if (ia == NULL) {
415 				error = ENOBUFS;
416 				goto out;
417 			}
418 
419 			ifa = &ia->ia_ifa;
420 			ifa_init(ifa);
421 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
422 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
423 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
424 
425 			ia->ia_sockmask.sin_len = 8;
426 			ia->ia_sockmask.sin_family = AF_INET;
427 			if (ifp->if_flags & IFF_BROADCAST) {
428 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
429 				ia->ia_broadaddr.sin_family = AF_INET;
430 			}
431 			ia->ia_ifp = ifp;
432 
433 			ifa_ref(ifa);			/* if_addrhead */
434 			IF_ADDR_LOCK(ifp);
435 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
436 			IF_ADDR_UNLOCK(ifp);
437 			ifa_ref(ifa);			/* in_ifaddrhead */
438 			IN_IFADDR_WLOCK();
439 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
440 			IN_IFADDR_WUNLOCK();
441 			iaIsNew = 1;
442 		}
443 		break;
444 
445 	case SIOCSIFBRDADDR:
446 	case SIOCGIFADDR:
447 	case SIOCGIFNETMASK:
448 	case SIOCGIFDSTADDR:
449 	case SIOCGIFBRDADDR:
450 		if (ia == NULL) {
451 			error = EADDRNOTAVAIL;
452 			goto out;
453 		}
454 		break;
455 	}
456 
457 	/*
458 	 * Most paths in this switch return directly or via out.  Only paths
459 	 * that remove the address break in order to hit common removal code.
460 	 */
461 	switch (cmd) {
462 	case SIOCGIFADDR:
463 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
464 		goto out;
465 
466 	case SIOCGIFBRDADDR:
467 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
468 			error = EINVAL;
469 			goto out;
470 		}
471 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
472 		goto out;
473 
474 	case SIOCGIFDSTADDR:
475 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
476 			error = EINVAL;
477 			goto out;
478 		}
479 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
480 		goto out;
481 
482 	case SIOCGIFNETMASK:
483 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
484 		goto out;
485 
486 	case SIOCSIFDSTADDR:
487 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
488 			error = EINVAL;
489 			goto out;
490 		}
491 		oldaddr = ia->ia_dstaddr;
492 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
493 		if (ifp->if_ioctl != NULL) {
494 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
495 			    (caddr_t)ia);
496 			if (error) {
497 				ia->ia_dstaddr = oldaddr;
498 				goto out;
499 			}
500 		}
501 		if (ia->ia_flags & IFA_ROUTE) {
502 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
503 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
504 			ia->ia_ifa.ifa_dstaddr =
505 					(struct sockaddr *)&ia->ia_dstaddr;
506 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
507 		}
508 		goto out;
509 
510 	case SIOCSIFBRDADDR:
511 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
512 			error = EINVAL;
513 			goto out;
514 		}
515 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
516 		goto out;
517 
518 	case SIOCSIFADDR:
519 		error = in_ifinit(ifp, ia,
520 		    (struct sockaddr_in *) &ifr->ifr_addr, 1, 0);
521 		if (error != 0 && iaIsNew)
522 			break;
523 		if (error == 0) {
524 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
525 			if (iaIsFirst &&
526 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
527 				error = in_joingroup(ifp, &allhosts_addr,
528 				    NULL, &ii->ii_allhosts);
529 			}
530 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
531 		}
532 		error = 0;
533 		goto out;
534 
535 	case SIOCSIFNETMASK:
536 		ia->ia_sockmask.sin_addr = ((struct sockaddr_in *)
537 		    &ifr->ifr_addr)->sin_addr;
538 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
539 		goto out;
540 
541 	case SIOCAIFADDR:
542 		maskIsNew = 0;
543 		hostIsNew = 1;
544 		error = 0;
545 		if (ifra->ifra_addr.sin_addr.s_addr ==
546 			    ia->ia_addr.sin_addr.s_addr)
547 			hostIsNew = 0;
548 		if (ifra->ifra_mask.sin_len) {
549 			/*
550 			 * QL: XXX
551 			 * Need to scrub the prefix here in case
552 			 * the issued command is SIOCAIFADDR with
553 			 * the same address, but with a different
554 			 * prefix length. And if the prefix length
555 			 * is the same as before, then the call is
556 			 * un-necessarily executed here.
557 			 */
558 			in_ifscrub(ifp, ia, LLE_STATIC);
559 			ia->ia_sockmask = ifra->ifra_mask;
560 			ia->ia_sockmask.sin_family = AF_INET;
561 			ia->ia_subnetmask =
562 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
563 			maskIsNew = 1;
564 		}
565 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
566 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
567 			in_ifscrub(ifp, ia, LLE_STATIC);
568 			ia->ia_dstaddr = ifra->ifra_dstaddr;
569 			maskIsNew  = 1; /* We lie; but the effect's the same */
570 		}
571 		if (hostIsNew || maskIsNew)
572 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
573 			    maskIsNew);
574 		if (error != 0 && iaIsNew)
575 			break;
576 
577 		if ((ifp->if_flags & IFF_BROADCAST) &&
578 		    ifra->ifra_broadaddr.sin_len)
579 			ia->ia_broadaddr = ifra->ifra_broadaddr;
580 		if (error == 0) {
581 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
582 			if (iaIsFirst &&
583 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
584 				error = in_joingroup(ifp, &allhosts_addr,
585 				    NULL, &ii->ii_allhosts);
586 			}
587 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
588 		}
589 		goto out;
590 
591 	case SIOCDIFADDR:
592 		/*
593 		 * in_ifscrub kills the interface route.
594 		 */
595 		in_ifscrub(ifp, ia, LLE_STATIC);
596 
597 		/*
598 		 * in_ifadown gets rid of all the rest of
599 		 * the routes.  This is not quite the right
600 		 * thing to do, but at least if we are running
601 		 * a routing process they will come back.
602 		 */
603 		in_ifadown(&ia->ia_ifa, 1);
604 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
605 		error = 0;
606 		break;
607 
608 	default:
609 		panic("in_control: unsupported ioctl");
610 	}
611 
612 	IF_ADDR_LOCK(ifp);
613 	/* Re-check that ia is still part of the list. */
614 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
615 		if (ifa == &ia->ia_ifa)
616 			break;
617 	}
618 	if (ifa == NULL) {
619 		/*
620 		 * If we lost the race with another thread, there is no need to
621 		 * try it again for the next loop as there is no other exit
622 		 * path between here and out.
623 		 */
624 		IF_ADDR_UNLOCK(ifp);
625 		error = EADDRNOTAVAIL;
626 		goto out;
627 	}
628 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
629 	IF_ADDR_UNLOCK(ifp);
630 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
631 
632 	IN_IFADDR_WLOCK();
633 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
634 
635 	LIST_REMOVE(ia, ia_hash);
636 	IN_IFADDR_WUNLOCK();
637 	/*
638 	 * If this is the last IPv4 address configured on this
639 	 * interface, leave the all-hosts group.
640 	 * No state-change report need be transmitted.
641 	 */
642 	IFP_TO_IA(ifp, iap);
643 	if (iap == NULL) {
644 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
645 		IN_MULTI_LOCK();
646 		if (ii->ii_allhosts) {
647 			(void)in_leavegroup_locked(ii->ii_allhosts, NULL);
648 			ii->ii_allhosts = NULL;
649 		}
650 		IN_MULTI_UNLOCK();
651 	} else
652 		ifa_free(&iap->ia_ifa);
653 
654 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
655 out:
656 	if (ia != NULL)
657 		ifa_free(&ia->ia_ifa);
658 	return (error);
659 }
660 
661 /*
662  * SIOC[GAD]LIFADDR.
663  *	SIOCGLIFADDR: get first address. (?!?)
664  *	SIOCGLIFADDR with IFLR_PREFIX:
665  *		get first address that matches the specified prefix.
666  *	SIOCALIFADDR: add the specified address.
667  *	SIOCALIFADDR with IFLR_PREFIX:
668  *		EINVAL since we can't deduce hostid part of the address.
669  *	SIOCDLIFADDR: delete the specified address.
670  *	SIOCDLIFADDR with IFLR_PREFIX:
671  *		delete the first address that matches the specified prefix.
672  * return values:
673  *	EINVAL on invalid parameters
674  *	EADDRNOTAVAIL on prefix match failed/specified address not found
675  *	other values may be returned from in_ioctl()
676  */
677 static int
678 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
679     struct ifnet *ifp, struct thread *td)
680 {
681 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
682 	struct ifaddr *ifa;
683 
684 	/* sanity checks */
685 	if (data == NULL || ifp == NULL) {
686 		panic("invalid argument to in_lifaddr_ioctl");
687 		/*NOTRECHED*/
688 	}
689 
690 	switch (cmd) {
691 	case SIOCGLIFADDR:
692 		/* address must be specified on GET with IFLR_PREFIX */
693 		if ((iflr->flags & IFLR_PREFIX) == 0)
694 			break;
695 		/*FALLTHROUGH*/
696 	case SIOCALIFADDR:
697 	case SIOCDLIFADDR:
698 		/* address must be specified on ADD and DELETE */
699 		if (iflr->addr.ss_family != AF_INET)
700 			return (EINVAL);
701 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
702 			return (EINVAL);
703 		/* XXX need improvement */
704 		if (iflr->dstaddr.ss_family
705 		 && iflr->dstaddr.ss_family != AF_INET)
706 			return (EINVAL);
707 		if (iflr->dstaddr.ss_family
708 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
709 			return (EINVAL);
710 		break;
711 	default: /*shouldn't happen*/
712 		return (EOPNOTSUPP);
713 	}
714 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
715 		return (EINVAL);
716 
717 	switch (cmd) {
718 	case SIOCALIFADDR:
719 	    {
720 		struct in_aliasreq ifra;
721 
722 		if (iflr->flags & IFLR_PREFIX)
723 			return (EINVAL);
724 
725 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
726 		bzero(&ifra, sizeof(ifra));
727 		bcopy(iflr->iflr_name, ifra.ifra_name,
728 			sizeof(ifra.ifra_name));
729 
730 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
731 
732 		if (iflr->dstaddr.ss_family) {	/*XXX*/
733 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
734 				iflr->dstaddr.ss_len);
735 		}
736 
737 		ifra.ifra_mask.sin_family = AF_INET;
738 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
739 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
740 
741 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
742 	    }
743 	case SIOCGLIFADDR:
744 	case SIOCDLIFADDR:
745 	    {
746 		struct in_ifaddr *ia;
747 		struct in_addr mask, candidate, match;
748 		struct sockaddr_in *sin;
749 
750 		bzero(&mask, sizeof(mask));
751 		bzero(&match, sizeof(match));
752 		if (iflr->flags & IFLR_PREFIX) {
753 			/* lookup a prefix rather than address. */
754 			in_len2mask(&mask, iflr->prefixlen);
755 
756 			sin = (struct sockaddr_in *)&iflr->addr;
757 			match.s_addr = sin->sin_addr.s_addr;
758 			match.s_addr &= mask.s_addr;
759 
760 			/* if you set extra bits, that's wrong */
761 			if (match.s_addr != sin->sin_addr.s_addr)
762 				return (EINVAL);
763 
764 		} else {
765 			/* on getting an address, take the 1st match */
766 			/* on deleting an address, do exact match */
767 			if (cmd != SIOCGLIFADDR) {
768 				in_len2mask(&mask, 32);
769 				sin = (struct sockaddr_in *)&iflr->addr;
770 				match.s_addr = sin->sin_addr.s_addr;
771 			}
772 		}
773 
774 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
775 			if (ifa->ifa_addr->sa_family != AF_INET6)
776 				continue;
777 			if (match.s_addr == 0)
778 				break;
779 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
780 			candidate.s_addr &= mask.s_addr;
781 			if (candidate.s_addr == match.s_addr)
782 				break;
783 		}
784 		if (ifa == NULL)
785 			return (EADDRNOTAVAIL);
786 		ia = (struct in_ifaddr *)ifa;
787 
788 		if (cmd == SIOCGLIFADDR) {
789 			/* fill in the if_laddrreq structure */
790 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
791 
792 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
793 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
794 					ia->ia_dstaddr.sin_len);
795 			} else
796 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
797 
798 			iflr->prefixlen =
799 				in_mask2len(&ia->ia_sockmask.sin_addr);
800 
801 			iflr->flags = 0;	/*XXX*/
802 
803 			return (0);
804 		} else {
805 			struct in_aliasreq ifra;
806 
807 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
808 			bzero(&ifra, sizeof(ifra));
809 			bcopy(iflr->iflr_name, ifra.ifra_name,
810 				sizeof(ifra.ifra_name));
811 
812 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
813 				ia->ia_addr.sin_len);
814 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
815 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
816 					ia->ia_dstaddr.sin_len);
817 			}
818 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
819 				ia->ia_sockmask.sin_len);
820 
821 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
822 			    ifp, td));
823 		}
824 	    }
825 	}
826 
827 	return (EOPNOTSUPP);	/*just for safety*/
828 }
829 
830 /*
831  * Delete any existing route for an interface.
832  */
833 void
834 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
835 {
836 
837 	in_scrubprefix(ia, flags);
838 }
839 
840 /*
841  * Initialize an interface's internet address
842  * and routing table entry.
843  */
844 static int
845 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
846     int scrub, int masksupplied)
847 {
848 	register u_long i = ntohl(sin->sin_addr.s_addr);
849 	int flags = RTF_UP, error = 0;
850 
851 	if (scrub)
852 		in_scrubprefix(ia, LLE_STATIC);
853 
854 	IN_IFADDR_WLOCK();
855 	if (ia->ia_addr.sin_family == AF_INET)
856 		LIST_REMOVE(ia, ia_hash);
857 	ia->ia_addr = *sin;
858 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
859 	    ia, ia_hash);
860 	IN_IFADDR_WUNLOCK();
861 
862 	/*
863 	 * Give the interface a chance to initialize
864 	 * if this is its first address,
865 	 * and to validate the address if necessary.
866 	 */
867 	if (ifp->if_ioctl != NULL &&
868 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0)
869 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
870 			return (error);
871 
872 	/*
873 	 * Be compatible with network classes, if netmask isn't supplied,
874 	 * guess it based on classes.
875 	 */
876 	if (!masksupplied) {
877 		if (IN_CLASSA(i))
878 			ia->ia_subnetmask = IN_CLASSA_NET;
879 		else if (IN_CLASSB(i))
880 			ia->ia_subnetmask = IN_CLASSB_NET;
881 		else
882 			ia->ia_subnetmask = IN_CLASSC_NET;
883 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
884 	}
885 	ia->ia_subnet = i & ia->ia_subnetmask;
886 	in_socktrim(&ia->ia_sockmask);
887 	/*
888 	 * XXX: carp(4) does not have interface route
889 	 */
890 	if (ifp->if_type == IFT_CARP)
891 		return (0);
892 	/*
893 	 * Add route for the network.
894 	 */
895 	ia->ia_ifa.ifa_metric = ifp->if_metric;
896 	if (ifp->if_flags & IFF_BROADCAST) {
897 		if (ia->ia_subnetmask == IN_RFC3021_MASK)
898 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
899 		else
900 			ia->ia_broadaddr.sin_addr.s_addr =
901 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
902 	} else if (ifp->if_flags & IFF_LOOPBACK) {
903 		ia->ia_dstaddr = ia->ia_addr;
904 		flags |= RTF_HOST;
905 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
906 		if (ia->ia_dstaddr.sin_family != AF_INET)
907 			return (0);
908 		flags |= RTF_HOST;
909 	}
910 	if ((error = in_addprefix(ia, flags)) != 0)
911 		return (error);
912 
913 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
914 		return (0);
915 
916 	if (ifp->if_flags & IFF_POINTOPOINT &&
917 	    ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
918 			return (0);
919 
920 	/*
921 	 * add a loopback route to self
922 	 */
923 	if (V_useloopback && !(ifp->if_flags & IFF_LOOPBACK)) {
924 		struct route ia_ro;
925 
926 		bzero(&ia_ro, sizeof(ia_ro));
927 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
928 		rtalloc_ign_fib(&ia_ro, 0, 0);
929 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
930 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
931 			RT_LOCK(ia_ro.ro_rt);
932 			RT_ADDREF(ia_ro.ro_rt);
933 			RTFREE_LOCKED(ia_ro.ro_rt);
934 		} else
935 			error = ifa_add_loopback_route((struct ifaddr *)ia,
936 				       (struct sockaddr *)&ia->ia_addr);
937 		if (error == 0)
938 			ia->ia_flags |= IFA_RTSELF;
939 		if (ia_ro.ro_rt != NULL)
940 			RTFREE(ia_ro.ro_rt);
941 	}
942 
943 	return (error);
944 }
945 
946 #define rtinitflags(x) \
947 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
948 	    ? RTF_HOST : 0)
949 
950 /*
951  * Generate a routing message when inserting or deleting
952  * an interface address alias.
953  */
954 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
955     struct in_ifaddr *target)
956 {
957 	struct route pfx_ro;
958 	struct sockaddr_in *pfx_addr;
959 	struct rtentry msg_rt;
960 
961 	/* QL: XXX
962 	 * This is a bit questionable because there is no
963 	 * additional route entry added/deleted for an address
964 	 * alias. Therefore this route report is inaccurate.
965 	 */
966 	bzero(&pfx_ro, sizeof(pfx_ro));
967 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
968 	pfx_addr->sin_len = sizeof(*pfx_addr);
969 	pfx_addr->sin_family = AF_INET;
970 	pfx_addr->sin_addr = *prefix;
971 	rtalloc_ign_fib(&pfx_ro, 0, 0);
972 	if (pfx_ro.ro_rt != NULL) {
973 		msg_rt = *pfx_ro.ro_rt;
974 
975 		/* QL: XXX
976 		 * Point the gateway to the new interface
977 		 * address as if a new prefix route entry has
978 		 * been added through the new address alias.
979 		 * All other parts of the rtentry is accurate,
980 		 * e.g., rt_key, rt_mask, rt_ifp etc.
981 		 */
982 		msg_rt.rt_gateway =
983 			(struct sockaddr *)&target->ia_addr;
984 		rt_newaddrmsg(cmd,
985 			      (struct ifaddr *)target,
986 			      0, &msg_rt);
987 		RTFREE(pfx_ro.ro_rt);
988 	}
989 	return;
990 }
991 
992 /*
993  * Check if we have a route for the given prefix already or add one accordingly.
994  */
995 static int
996 in_addprefix(struct in_ifaddr *target, int flags)
997 {
998 	struct in_ifaddr *ia;
999 	struct in_addr prefix, mask, p, m;
1000 	int error;
1001 
1002 	if ((flags & RTF_HOST) != 0) {
1003 		prefix = target->ia_dstaddr.sin_addr;
1004 		mask.s_addr = 0;
1005 	} else {
1006 		prefix = target->ia_addr.sin_addr;
1007 		mask = target->ia_sockmask.sin_addr;
1008 		prefix.s_addr &= mask.s_addr;
1009 	}
1010 
1011 	IN_IFADDR_RLOCK();
1012 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1013 		if (rtinitflags(ia)) {
1014 			p = ia->ia_dstaddr.sin_addr;
1015 
1016 			if (prefix.s_addr != p.s_addr)
1017 				continue;
1018 		} else {
1019 			p = ia->ia_addr.sin_addr;
1020 			m = ia->ia_sockmask.sin_addr;
1021 			p.s_addr &= m.s_addr;
1022 
1023 			if (prefix.s_addr != p.s_addr ||
1024 			    mask.s_addr != m.s_addr)
1025 				continue;
1026 		}
1027 
1028 		/*
1029 		 * If we got a matching prefix route inserted by other
1030 		 * interface address, we are done here.
1031 		 */
1032 		if (ia->ia_flags & IFA_ROUTE) {
1033 #ifdef RADIX_MPATH
1034 			if (ia->ia_addr.sin_addr.s_addr ==
1035 			    target->ia_addr.sin_addr.s_addr) {
1036 				IN_IFADDR_RUNLOCK();
1037 				return (EEXIST);
1038 			} else
1039 				break;
1040 #endif
1041 			if (V_sameprefixcarponly &&
1042 			    target->ia_ifp->if_type != IFT_CARP &&
1043 			    ia->ia_ifp->if_type != IFT_CARP) {
1044 				IN_IFADDR_RUNLOCK();
1045 				return (EEXIST);
1046 			} else {
1047 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
1048 				IN_IFADDR_RUNLOCK();
1049 				return (0);
1050 			}
1051 		}
1052 	}
1053 	IN_IFADDR_RUNLOCK();
1054 
1055 	/*
1056 	 * No-one seem to have this prefix route, so we try to insert it.
1057 	 */
1058 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1059 	if (!error)
1060 		target->ia_flags |= IFA_ROUTE;
1061 	return (error);
1062 }
1063 
1064 extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr);
1065 
1066 /*
1067  * If there is no other address in the system that can serve a route to the
1068  * same prefix, remove the route.  Hand over the route to the new address
1069  * otherwise.
1070  */
1071 static int
1072 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1073 {
1074 	struct in_ifaddr *ia;
1075 	struct in_addr prefix, mask, p, m;
1076 	int error = 0;
1077 	struct sockaddr_in prefix0, mask0;
1078 
1079 	/*
1080 	 * Remove the loopback route to the interface address.
1081 	 * The "useloopback" setting is not consulted because if the
1082 	 * user configures an interface address, turns off this
1083 	 * setting, and then tries to delete that interface address,
1084 	 * checking the current setting of "useloopback" would leave
1085 	 * that interface address loopback route untouched, which
1086 	 * would be wrong. Therefore the interface address loopback route
1087 	 * deletion is unconditional.
1088 	 */
1089 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1090 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1091 	    (target->ia_flags & IFA_RTSELF)) {
1092 		struct route ia_ro;
1093 		int freeit = 0;
1094 
1095 		bzero(&ia_ro, sizeof(ia_ro));
1096 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1097 		rtalloc_ign_fib(&ia_ro, 0, 0);
1098 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1099 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1100 			RT_LOCK(ia_ro.ro_rt);
1101 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1102 				freeit = 1;
1103 			else if (flags & LLE_STATIC) {
1104 				RT_REMREF(ia_ro.ro_rt);
1105 				target->ia_flags &= ~IFA_RTSELF;
1106 			}
1107 			RTFREE_LOCKED(ia_ro.ro_rt);
1108 		}
1109 		if (freeit && (flags & LLE_STATIC)) {
1110 			error = ifa_del_loopback_route((struct ifaddr *)target,
1111 				       (struct sockaddr *)&target->ia_addr);
1112 			if (error == 0)
1113 				target->ia_flags &= ~IFA_RTSELF;
1114 		}
1115 		if ((flags & LLE_STATIC) &&
1116 			!(target->ia_ifp->if_flags & IFF_NOARP))
1117 			/* remove arp cache */
1118 			arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1119 	}
1120 
1121 	if (rtinitflags(target)) {
1122 		prefix = target->ia_dstaddr.sin_addr;
1123 		mask.s_addr = 0;
1124 	} else {
1125 		prefix = target->ia_addr.sin_addr;
1126 		mask = target->ia_sockmask.sin_addr;
1127 		prefix.s_addr &= mask.s_addr;
1128 	}
1129 
1130 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1131 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1132 		return (0);
1133 	}
1134 
1135 	IN_IFADDR_RLOCK();
1136 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1137 		if (rtinitflags(ia)) {
1138 			p = ia->ia_dstaddr.sin_addr;
1139 
1140 			if (prefix.s_addr != p.s_addr)
1141 				continue;
1142 		} else {
1143 			p = ia->ia_addr.sin_addr;
1144 			m = ia->ia_sockmask.sin_addr;
1145 			p.s_addr &= m.s_addr;
1146 
1147 			if (prefix.s_addr != p.s_addr ||
1148 			    mask.s_addr != m.s_addr)
1149 				continue;
1150 		}
1151 
1152 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1153 			continue;
1154 
1155 		/*
1156 		 * If we got a matching prefix address, move IFA_ROUTE and
1157 		 * the route itself to it.  Make sure that routing daemons
1158 		 * get a heads-up.
1159 		 *
1160 		 * XXX: a special case for carp(4) interface - this should
1161 		 *      be more generally specified as an interface that
1162 		 *      doesn't support such action.
1163 		 */
1164 		if ((ia->ia_flags & IFA_ROUTE) == 0
1165 		    && (ia->ia_ifp->if_type != IFT_CARP)) {
1166 			ifa_ref(&ia->ia_ifa);
1167 			IN_IFADDR_RUNLOCK();
1168 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1169 			    rtinitflags(target));
1170 			if (error == 0)
1171 				target->ia_flags &= ~IFA_ROUTE;
1172 			else
1173 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1174 					error);
1175 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1176 			    rtinitflags(ia) | RTF_UP);
1177 			if (error == 0)
1178 				ia->ia_flags |= IFA_ROUTE;
1179 			else
1180 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1181 					error);
1182 			ifa_free(&ia->ia_ifa);
1183 			return (error);
1184 		}
1185 	}
1186 	IN_IFADDR_RUNLOCK();
1187 
1188 	/*
1189 	 * remove all L2 entries on the given prefix
1190 	 */
1191 	bzero(&prefix0, sizeof(prefix0));
1192 	prefix0.sin_len = sizeof(prefix0);
1193 	prefix0.sin_family = AF_INET;
1194 	prefix0.sin_addr.s_addr = target->ia_subnet;
1195 	bzero(&mask0, sizeof(mask0));
1196 	mask0.sin_len = sizeof(mask0);
1197 	mask0.sin_family = AF_INET;
1198 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1199 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1200 			    (struct sockaddr *)&mask0, flags);
1201 
1202 	/*
1203 	 * As no-one seem to have this prefix, we can remove the route.
1204 	 */
1205 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1206 	if (error == 0)
1207 		target->ia_flags &= ~IFA_ROUTE;
1208 	else
1209 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1210 	return (error);
1211 }
1212 
1213 #undef rtinitflags
1214 
1215 /*
1216  * Return 1 if the address might be a local broadcast address.
1217  */
1218 int
1219 in_broadcast(struct in_addr in, struct ifnet *ifp)
1220 {
1221 	register struct ifaddr *ifa;
1222 	u_long t;
1223 
1224 	if (in.s_addr == INADDR_BROADCAST ||
1225 	    in.s_addr == INADDR_ANY)
1226 		return (1);
1227 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1228 		return (0);
1229 	t = ntohl(in.s_addr);
1230 	/*
1231 	 * Look through the list of addresses for a match
1232 	 * with a broadcast address.
1233 	 */
1234 #define ia ((struct in_ifaddr *)ifa)
1235 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1236 		if (ifa->ifa_addr->sa_family == AF_INET &&
1237 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1238 		     /*
1239 		      * Check for old-style (host 0) broadcast, but
1240 		      * taking into account that RFC 3021 obsoletes it.
1241 		      */
1242 		     (ia->ia_subnetmask != IN_RFC3021_MASK &&
1243 		     t == ia->ia_subnet)) &&
1244 		     /*
1245 		      * Check for an all one subnetmask. These
1246 		      * only exist when an interface gets a secondary
1247 		      * address.
1248 		      */
1249 		     ia->ia_subnetmask != (u_long)0xffffffff)
1250 			    return (1);
1251 	return (0);
1252 #undef ia
1253 }
1254 
1255 /*
1256  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1257  */
1258 void
1259 in_ifdetach(struct ifnet *ifp)
1260 {
1261 
1262 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1263 	in_pcbpurgeif0(&V_udbinfo, ifp);
1264 	in_purgemaddrs(ifp);
1265 }
1266 
1267 /*
1268  * Delete all IPv4 multicast address records, and associated link-layer
1269  * multicast address records, associated with ifp.
1270  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1271  * XXX This should not race with ifma_protospec being set during
1272  * a new allocation, if it does, we have bigger problems.
1273  */
1274 static void
1275 in_purgemaddrs(struct ifnet *ifp)
1276 {
1277 	LIST_HEAD(,in_multi) purgeinms;
1278 	struct in_multi		*inm, *tinm;
1279 	struct ifmultiaddr	*ifma;
1280 
1281 	LIST_INIT(&purgeinms);
1282 	IN_MULTI_LOCK();
1283 
1284 	/*
1285 	 * Extract list of in_multi associated with the detaching ifp
1286 	 * which the PF_INET layer is about to release.
1287 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1288 	 * by code further down.
1289 	 */
1290 	IF_ADDR_LOCK(ifp);
1291 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1292 		if (ifma->ifma_addr->sa_family != AF_INET ||
1293 		    ifma->ifma_protospec == NULL)
1294 			continue;
1295 #if 0
1296 		KASSERT(ifma->ifma_protospec != NULL,
1297 		    ("%s: ifma_protospec is NULL", __func__));
1298 #endif
1299 		inm = (struct in_multi *)ifma->ifma_protospec;
1300 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1301 	}
1302 	IF_ADDR_UNLOCK(ifp);
1303 
1304 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1305 		LIST_REMOVE(inm, inm_link);
1306 		inm_release_locked(inm);
1307 	}
1308 	igmp_ifdetach(ifp);
1309 
1310 	IN_MULTI_UNLOCK();
1311 }
1312 
1313 #include <net/if_dl.h>
1314 #include <netinet/if_ether.h>
1315 
1316 struct in_llentry {
1317 	struct llentry		base;
1318 	struct sockaddr_in	l3_addr4;
1319 };
1320 
1321 static struct llentry *
1322 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1323 {
1324 	struct in_llentry *lle;
1325 
1326 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1327 	if (lle == NULL)		/* NB: caller generates msg */
1328 		return NULL;
1329 
1330 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1331 	/*
1332 	 * For IPv4 this will trigger "arpresolve" to generate
1333 	 * an ARP request.
1334 	 */
1335 	lle->base.la_expire = time_uptime; /* mark expired */
1336 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1337 	lle->base.lle_refcnt = 1;
1338 	LLE_LOCK_INIT(&lle->base);
1339 	return &lle->base;
1340 }
1341 
1342 /*
1343  * Deletes an address from the address table.
1344  * This function is called by the timer functions
1345  * such as arptimer() and nd6_llinfo_timer(), and
1346  * the caller does the locking.
1347  */
1348 static void
1349 in_lltable_free(struct lltable *llt, struct llentry *lle)
1350 {
1351 	LLE_WUNLOCK(lle);
1352 	LLE_LOCK_DESTROY(lle);
1353 	free(lle, M_LLTABLE);
1354 }
1355 
1356 
1357 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1358 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1359 
1360 static void
1361 in_lltable_prefix_free(struct lltable *llt,
1362 		       const struct sockaddr *prefix,
1363 		       const struct sockaddr *mask,
1364 		       u_int flags)
1365 {
1366 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1367 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1368 	struct llentry *lle, *next;
1369 	register int i;
1370 	size_t pkts_dropped;
1371 
1372 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1373 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1374 
1375 		        /*
1376 			 * (flags & LLE_STATIC) means deleting all entries
1377 			 * including static ARP entries
1378 			 */
1379 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1380 						     pfx, msk) &&
1381 			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
1382 				int canceled;
1383 
1384 				canceled = callout_drain(&lle->la_timer);
1385 				LLE_WLOCK(lle);
1386 				if (canceled)
1387 					LLE_REMREF(lle);
1388 				pkts_dropped = llentry_free(lle);
1389 				ARPSTAT_ADD(dropped, pkts_dropped);
1390 			}
1391 		}
1392 	}
1393 }
1394 
1395 
1396 static int
1397 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1398 {
1399 	struct rtentry *rt;
1400 
1401 	KASSERT(l3addr->sa_family == AF_INET,
1402 	    ("sin_family %d", l3addr->sa_family));
1403 
1404 	/* XXX rtalloc1 should take a const param */
1405 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1406 
1407 	if (rt == NULL)
1408 		return (EINVAL);
1409 
1410 	/*
1411 	 * If the gateway for an existing host route matches the target L3
1412 	 * address, which is a special route inserted by some implementation
1413 	 * such as MANET, and the interface is of the correct type, then
1414 	 * allow for ARP to proceed.
1415 	 */
1416 	if (rt->rt_flags & RTF_GATEWAY) {
1417 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1418 			rt->rt_ifp->if_type != IFT_ETHER ||
1419 			  (rt->rt_ifp->if_flags &
1420 			   (IFF_NOARP | IFF_STATICARP)) != 0 ||
1421 			  memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1422 				 sizeof(in_addr_t)) != 0) {
1423 			RTFREE_LOCKED(rt);
1424 			return (EINVAL);
1425 		}
1426 	}
1427 
1428 	/*
1429 	 * Make sure that at least the destination address is covered
1430 	 * by the route. This is for handling the case where 2 or more
1431 	 * interfaces have the same prefix. An incoming packet arrives
1432 	 * on one interface and the corresponding outgoing packet leaves
1433 	 * another interface.
1434 	 */
1435 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1436 		const char *sa, *mask, *addr, *lim;
1437 		int len;
1438 
1439 		mask = (const char *)rt_mask(rt);
1440 		/*
1441 		 * Just being extra cautious to avoid some custom
1442 		 * code getting into trouble.
1443 		 */
1444 		if (mask == NULL) {
1445 			RTFREE_LOCKED(rt);
1446 			return (EINVAL);
1447 		}
1448 
1449 		sa = (const char *)rt_key(rt);
1450 		addr = (const char *)l3addr;
1451 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
1452 		lim = addr + len;
1453 
1454 		for ( ; addr < lim; sa++, mask++, addr++) {
1455 			if ((*sa ^ *addr) & *mask) {
1456 #ifdef DIAGNOSTIC
1457 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1458 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1459 #endif
1460 				RTFREE_LOCKED(rt);
1461 				return (EINVAL);
1462 			}
1463 		}
1464 	}
1465 
1466 	RTFREE_LOCKED(rt);
1467 	return (0);
1468 }
1469 
1470 /*
1471  * Return NULL if not found or marked for deletion.
1472  * If found return lle read locked.
1473  */
1474 static struct llentry *
1475 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1476 {
1477 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1478 	struct ifnet *ifp = llt->llt_ifp;
1479 	struct llentry *lle;
1480 	struct llentries *lleh;
1481 	u_int hashkey;
1482 
1483 	IF_AFDATA_LOCK_ASSERT(ifp);
1484 	KASSERT(l3addr->sa_family == AF_INET,
1485 	    ("sin_family %d", l3addr->sa_family));
1486 
1487 	hashkey = sin->sin_addr.s_addr;
1488 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1489 	LIST_FOREACH(lle, lleh, lle_next) {
1490 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1491 		if (lle->la_flags & LLE_DELETED)
1492 			continue;
1493 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1494 			break;
1495 	}
1496 	if (lle == NULL) {
1497 #ifdef DIAGNOSTIC
1498 		if (flags & LLE_DELETE)
1499 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1500 #endif
1501 		if (!(flags & LLE_CREATE))
1502 			return (NULL);
1503 		/*
1504 		 * A route that covers the given address must have
1505 		 * been installed 1st because we are doing a resolution,
1506 		 * verify this.
1507 		 */
1508 		if (!(flags & LLE_IFADDR) &&
1509 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1510 			goto done;
1511 
1512 		lle = in_lltable_new(l3addr, flags);
1513 		if (lle == NULL) {
1514 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1515 			goto done;
1516 		}
1517 		lle->la_flags = flags & ~LLE_CREATE;
1518 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1519 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1520 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1521 		}
1522 
1523 		lle->lle_tbl  = llt;
1524 		lle->lle_head = lleh;
1525 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1526 	} else if (flags & LLE_DELETE) {
1527 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1528 			LLE_WLOCK(lle);
1529 			lle->la_flags = LLE_DELETED;
1530 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1531 			LLE_WUNLOCK(lle);
1532 #ifdef DIAGNOSTIC
1533 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1534 #endif
1535 		}
1536 		lle = (void *)-1;
1537 
1538 	}
1539 	if (LLE_IS_VALID(lle)) {
1540 		if (flags & LLE_EXCLUSIVE)
1541 			LLE_WLOCK(lle);
1542 		else
1543 			LLE_RLOCK(lle);
1544 	}
1545 done:
1546 	return (lle);
1547 }
1548 
1549 static int
1550 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1551 {
1552 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1553 	struct ifnet *ifp = llt->llt_ifp;
1554 	struct llentry *lle;
1555 	/* XXX stack use */
1556 	struct {
1557 		struct rt_msghdr	rtm;
1558 		struct sockaddr_inarp	sin;
1559 		struct sockaddr_dl	sdl;
1560 	} arpc;
1561 	int error, i;
1562 
1563 	LLTABLE_LOCK_ASSERT();
1564 
1565 	error = 0;
1566 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1567 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1568 			struct sockaddr_dl *sdl;
1569 
1570 			/* skip deleted entries */
1571 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1572 				continue;
1573 			/* Skip if jailed and not a valid IP of the prison. */
1574 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1575 				continue;
1576 			/*
1577 			 * produce a msg made of:
1578 			 *  struct rt_msghdr;
1579 			 *  struct sockaddr_inarp; (IPv4)
1580 			 *  struct sockaddr_dl;
1581 			 */
1582 			bzero(&arpc, sizeof(arpc));
1583 			arpc.rtm.rtm_msglen = sizeof(arpc);
1584 			arpc.rtm.rtm_version = RTM_VERSION;
1585 			arpc.rtm.rtm_type = RTM_GET;
1586 			arpc.rtm.rtm_flags = RTF_UP;
1587 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1588 			arpc.sin.sin_family = AF_INET;
1589 			arpc.sin.sin_len = sizeof(arpc.sin);
1590 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1591 
1592 			/* publish */
1593 			if (lle->la_flags & LLE_PUB) {
1594 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1595 				/* proxy only */
1596 				if (lle->la_flags & LLE_PROXY)
1597 					arpc.sin.sin_other = SIN_PROXY;
1598 			}
1599 
1600 			sdl = &arpc.sdl;
1601 			sdl->sdl_family = AF_LINK;
1602 			sdl->sdl_len = sizeof(*sdl);
1603 			sdl->sdl_index = ifp->if_index;
1604 			sdl->sdl_type = ifp->if_type;
1605 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1606 				sdl->sdl_alen = ifp->if_addrlen;
1607 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1608 			} else {
1609 				sdl->sdl_alen = 0;
1610 				bzero(LLADDR(sdl), ifp->if_addrlen);
1611 			}
1612 
1613 			arpc.rtm.rtm_rmx.rmx_expire =
1614 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1615 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1616 			if (lle->la_flags & LLE_STATIC)
1617 				arpc.rtm.rtm_flags |= RTF_STATIC;
1618 			arpc.rtm.rtm_index = ifp->if_index;
1619 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1620 			if (error)
1621 				break;
1622 		}
1623 	}
1624 	return error;
1625 #undef SIN
1626 }
1627 
1628 void *
1629 in_domifattach(struct ifnet *ifp)
1630 {
1631 	struct in_ifinfo *ii;
1632 	struct lltable *llt;
1633 
1634 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1635 
1636 	llt = lltable_init(ifp, AF_INET);
1637 	if (llt != NULL) {
1638 		llt->llt_free = in_lltable_free;
1639 		llt->llt_prefix_free = in_lltable_prefix_free;
1640 		llt->llt_lookup = in_lltable_lookup;
1641 		llt->llt_dump = in_lltable_dump;
1642 	}
1643 	ii->ii_llt = llt;
1644 
1645 	ii->ii_igmp = igmp_domifattach(ifp);
1646 
1647 	return ii;
1648 }
1649 
1650 void
1651 in_domifdetach(struct ifnet *ifp, void *aux)
1652 {
1653 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1654 
1655 	igmp_domifdetach(ifp);
1656 	lltable_free(ii->ii_llt);
1657 	free(ii, M_IFADDR);
1658 }
1659