xref: /freebsd/sys/netinet/in.c (revision 3e65b9c6e6b7b2081d54e1dc40983c3c00eaf738)
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/if_ether.h>
60 #include <netinet/in.h>
61 #include <netinet/in_var.h>
62 #include <netinet/in_pcb.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/ip_carp.h>
65 #include <netinet/igmp_var.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
68 
69 static int in_mask2len(struct in_addr *);
70 static void in_len2mask(struct in_addr *, int);
71 static int in_lifaddr_ioctl(struct socket *, u_long, caddr_t,
72 	struct ifnet *, struct thread *);
73 
74 static void	in_socktrim(struct sockaddr_in *);
75 static int	in_ifinit(struct ifnet *, struct in_ifaddr *,
76 		    struct sockaddr_in *, int, int, int);
77 static void	in_purgemaddrs(struct ifnet *);
78 
79 static VNET_DEFINE(int, nosameprefix);
80 #define	V_nosameprefix			VNET(nosameprefix)
81 SYSCTL_VNET_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_RW,
82 	&VNET_NAME(nosameprefix), 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 	u_long ocmd = cmd;
227 
228 	/*
229 	 * Pre-10.x compat: OSIOCAIFADDR passes a shorter
230 	 * struct in_aliasreq, without ifra_vhid.
231 	 */
232 	if (cmd == OSIOCAIFADDR)
233 		cmd = SIOCAIFADDR;
234 
235 	ia = NULL;
236 	iaIsFirst = 0;
237 	iaIsNew = 0;
238 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
239 
240 	/*
241 	 * Filter out ioctls we implement directly; forward the rest on to
242 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
243 	 */
244 	switch (cmd) {
245 	case SIOCGIFADDR:
246 	case SIOCGIFBRDADDR:
247 	case SIOCGIFDSTADDR:
248 	case SIOCGIFNETMASK:
249 	case SIOCDIFADDR:
250 		break;
251 	case SIOCAIFADDR:
252 		/*
253 		 * ifra_addr must be present and be of INET family.
254 		 * ifra_broadaddr and ifra_mask are optional.
255 		 */
256 		if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) ||
257 		    ifra->ifra_addr.sin_family != AF_INET)
258 			return (EINVAL);
259 		if (ifra->ifra_broadaddr.sin_len != 0 &&
260 		    (ifra->ifra_broadaddr.sin_len !=
261 		    sizeof(struct sockaddr_in) ||
262 		    ifra->ifra_broadaddr.sin_family != AF_INET))
263 			return (EINVAL);
264 #if 0
265 		/*
266 		 * ifconfig(8) in pre-10.x doesn't set sin_family for the
267 		 * mask. The code is disabled for the 10.x timeline, to
268 		 * make SIOCAIFADDR compatible with 9.x ifconfig(8).
269 		 * The code should be enabled in 11.x
270 		 */
271 		if (ifra->ifra_mask.sin_len != 0 &&
272 		    (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) ||
273 		    ifra->ifra_mask.sin_family != AF_INET))
274 			return (EINVAL);
275 #endif
276 		break;
277 	case SIOCSIFADDR:
278 	case SIOCSIFBRDADDR:
279 	case SIOCSIFDSTADDR:
280 	case SIOCSIFNETMASK:
281 		if (ifr->ifr_addr.sa_family != AF_INET ||
282 		    ifr->ifr_addr.sa_len != sizeof(struct sockaddr_in))
283 			return (EINVAL);
284 		break;
285 
286 	case SIOCALIFADDR:
287 		if (td != NULL) {
288 			error = priv_check(td, PRIV_NET_ADDIFADDR);
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 SIOCDLIFADDR:
297 		if (td != NULL) {
298 			error = priv_check(td, PRIV_NET_DELIFADDR);
299 			if (error)
300 				return (error);
301 		}
302 		if (ifp == NULL)
303 			return (EINVAL);
304 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
305 
306 	case SIOCGLIFADDR:
307 		if (ifp == NULL)
308 			return (EINVAL);
309 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
310 
311 	default:
312 		if (ifp == NULL || ifp->if_ioctl == NULL)
313 			return (EOPNOTSUPP);
314 		return ((*ifp->if_ioctl)(ifp, cmd, data));
315 	}
316 
317 	if (ifp == NULL)
318 		return (EADDRNOTAVAIL);
319 
320 	/*
321 	 * Security checks before we get involved in any work.
322 	 */
323 	switch (cmd) {
324 	case SIOCAIFADDR:
325 	case SIOCSIFADDR:
326 	case SIOCSIFBRDADDR:
327 	case SIOCSIFNETMASK:
328 	case SIOCSIFDSTADDR:
329 		if (td != NULL) {
330 			error = priv_check(td, PRIV_NET_ADDIFADDR);
331 			if (error)
332 				return (error);
333 		}
334 		break;
335 
336 	case SIOCDIFADDR:
337 		if (td != NULL) {
338 			error = priv_check(td, PRIV_NET_DELIFADDR);
339 			if (error)
340 				return (error);
341 		}
342 		break;
343 	}
344 
345 	/*
346 	 * Find address for this interface, if it exists.
347 	 *
348 	 * If an alias address was specified, find that one instead of the
349 	 * first one on the interface, if possible.
350 	 */
351 	dst = ((struct sockaddr_in *)&ifr->ifr_addr)->sin_addr;
352 	IN_IFADDR_RLOCK();
353 	LIST_FOREACH(iap, INADDR_HASH(dst.s_addr), ia_hash) {
354 		if (iap->ia_ifp == ifp &&
355 		    iap->ia_addr.sin_addr.s_addr == dst.s_addr) {
356 			if (td == NULL || prison_check_ip4(td->td_ucred,
357 			    &dst) == 0)
358 				ia = iap;
359 			break;
360 		}
361 	}
362 	if (ia != NULL)
363 		ifa_ref(&ia->ia_ifa);
364 	IN_IFADDR_RUNLOCK();
365 	if (ia == NULL) {
366 		IF_ADDR_LOCK(ifp);
367 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
368 			iap = ifatoia(ifa);
369 			if (iap->ia_addr.sin_family == AF_INET) {
370 				if (td != NULL &&
371 				    prison_check_ip4(td->td_ucred,
372 				    &iap->ia_addr.sin_addr) != 0)
373 					continue;
374 				ia = iap;
375 				break;
376 			}
377 		}
378 		if (ia != NULL)
379 			ifa_ref(&ia->ia_ifa);
380 		IF_ADDR_UNLOCK(ifp);
381 	}
382 	if (ia == NULL)
383 		iaIsFirst = 1;
384 
385 	error = 0;
386 	switch (cmd) {
387 	case SIOCAIFADDR:
388 	case SIOCDIFADDR:
389 		if (ifra->ifra_addr.sin_family == AF_INET) {
390 			struct in_ifaddr *oia;
391 
392 			IN_IFADDR_RLOCK();
393 			for (oia = ia; ia; ia = TAILQ_NEXT(ia, ia_link)) {
394 				if (ia->ia_ifp == ifp  &&
395 				    ia->ia_addr.sin_addr.s_addr ==
396 				    ifra->ifra_addr.sin_addr.s_addr)
397 					break;
398 			}
399 			if (ia != NULL && ia != oia)
400 				ifa_ref(&ia->ia_ifa);
401 			if (oia != NULL && ia != oia)
402 				ifa_free(&oia->ia_ifa);
403 			IN_IFADDR_RUNLOCK();
404 			if ((ifp->if_flags & IFF_POINTOPOINT)
405 			    && (cmd == SIOCAIFADDR)
406 			    && (ifra->ifra_dstaddr.sin_addr.s_addr
407 				== INADDR_ANY)) {
408 				error = EDESTADDRREQ;
409 				goto out;
410 			}
411 		}
412 		if (cmd == SIOCDIFADDR && ia == NULL) {
413 			error = EADDRNOTAVAIL;
414 			goto out;
415 		}
416 		/* FALLTHROUGH */
417 	case SIOCSIFADDR:
418 	case SIOCSIFNETMASK:
419 	case SIOCSIFDSTADDR:
420 		if (ia == NULL) {
421 			ia = (struct in_ifaddr *)
422 				malloc(sizeof *ia, M_IFADDR, M_NOWAIT |
423 				    M_ZERO);
424 			if (ia == NULL) {
425 				error = ENOBUFS;
426 				goto out;
427 			}
428 
429 			ifa = &ia->ia_ifa;
430 			ifa_init(ifa);
431 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
432 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
433 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
434 
435 			ia->ia_sockmask.sin_len = 8;
436 			ia->ia_sockmask.sin_family = AF_INET;
437 			if (ifp->if_flags & IFF_BROADCAST) {
438 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
439 				ia->ia_broadaddr.sin_family = AF_INET;
440 			}
441 			ia->ia_ifp = ifp;
442 
443 			ifa_ref(ifa);			/* if_addrhead */
444 			IF_ADDR_LOCK(ifp);
445 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
446 			IF_ADDR_UNLOCK(ifp);
447 			ifa_ref(ifa);			/* in_ifaddrhead */
448 			IN_IFADDR_WLOCK();
449 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
450 			IN_IFADDR_WUNLOCK();
451 			iaIsNew = 1;
452 		}
453 		break;
454 
455 	case SIOCSIFBRDADDR:
456 	case SIOCGIFADDR:
457 	case SIOCGIFNETMASK:
458 	case SIOCGIFDSTADDR:
459 	case SIOCGIFBRDADDR:
460 		if (ia == NULL) {
461 			error = EADDRNOTAVAIL;
462 			goto out;
463 		}
464 		break;
465 	}
466 
467 	/*
468 	 * Most paths in this switch return directly or via out.  Only paths
469 	 * that remove the address break in order to hit common removal code.
470 	 */
471 	switch (cmd) {
472 	case SIOCGIFADDR:
473 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
474 		goto out;
475 
476 	case SIOCGIFBRDADDR:
477 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
478 			error = EINVAL;
479 			goto out;
480 		}
481 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
482 		goto out;
483 
484 	case SIOCGIFDSTADDR:
485 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
486 			error = EINVAL;
487 			goto out;
488 		}
489 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
490 		goto out;
491 
492 	case SIOCGIFNETMASK:
493 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
494 		goto out;
495 
496 	case SIOCSIFDSTADDR:
497 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
498 			error = EINVAL;
499 			goto out;
500 		}
501 		oldaddr = ia->ia_dstaddr;
502 		ia->ia_dstaddr = *(struct sockaddr_in *)&ifr->ifr_dstaddr;
503 		if (ifp->if_ioctl != NULL) {
504 			error = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR,
505 			    (caddr_t)ia);
506 			if (error) {
507 				ia->ia_dstaddr = oldaddr;
508 				goto out;
509 			}
510 		}
511 		if (ia->ia_flags & IFA_ROUTE) {
512 			ia->ia_ifa.ifa_dstaddr = (struct sockaddr *)&oldaddr;
513 			rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST);
514 			ia->ia_ifa.ifa_dstaddr =
515 					(struct sockaddr *)&ia->ia_dstaddr;
516 			rtinit(&(ia->ia_ifa), (int)RTM_ADD, RTF_HOST|RTF_UP);
517 		}
518 		goto out;
519 
520 	case SIOCSIFBRDADDR:
521 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
522 			error = EINVAL;
523 			goto out;
524 		}
525 		ia->ia_broadaddr = *(struct sockaddr_in *)&ifr->ifr_broadaddr;
526 		goto out;
527 
528 	case SIOCSIFADDR:
529 		error = in_ifinit(ifp, ia,
530 		    (struct sockaddr_in *) &ifr->ifr_addr, 1, 0, 0);
531 		if (error != 0 && iaIsNew)
532 			break;
533 		if (error == 0) {
534 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
535 			if (iaIsFirst &&
536 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
537 				error = in_joingroup(ifp, &allhosts_addr,
538 				    NULL, &ii->ii_allhosts);
539 			}
540 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
541 		}
542 		error = 0;
543 		goto out;
544 
545 	case SIOCSIFNETMASK:
546 		ia->ia_sockmask.sin_addr = ((struct sockaddr_in *)
547 		    &ifr->ifr_addr)->sin_addr;
548 		ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr);
549 		goto out;
550 
551 	case SIOCAIFADDR:
552 		maskIsNew = 0;
553 		hostIsNew = 1;
554 		error = 0;
555 		if (ifra->ifra_addr.sin_addr.s_addr ==
556 			    ia->ia_addr.sin_addr.s_addr)
557 			hostIsNew = 0;
558 		if (ifra->ifra_mask.sin_len) {
559 			/*
560 			 * QL: XXX
561 			 * Need to scrub the prefix here in case
562 			 * the issued command is SIOCAIFADDR with
563 			 * the same address, but with a different
564 			 * prefix length. And if the prefix length
565 			 * is the same as before, then the call is
566 			 * un-necessarily executed here.
567 			 */
568 			in_ifscrub(ifp, ia, LLE_STATIC);
569 			ia->ia_sockmask = ifra->ifra_mask;
570 			ia->ia_sockmask.sin_family = AF_INET;
571 			ia->ia_subnetmask =
572 			     ntohl(ia->ia_sockmask.sin_addr.s_addr);
573 			maskIsNew = 1;
574 		}
575 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
576 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
577 			in_ifscrub(ifp, ia, LLE_STATIC);
578 			ia->ia_dstaddr = ifra->ifra_dstaddr;
579 			maskIsNew  = 1; /* We lie; but the effect's the same */
580 		}
581 		if (hostIsNew || maskIsNew)
582 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
583 			    maskIsNew, (ocmd == cmd ? ifra->ifra_vhid : 0));
584 		if (error != 0 && iaIsNew)
585 			break;
586 
587 		if ((ifp->if_flags & IFF_BROADCAST) &&
588 		    ifra->ifra_broadaddr.sin_len)
589 			ia->ia_broadaddr = ifra->ifra_broadaddr;
590 		if (error == 0) {
591 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
592 			if (iaIsFirst &&
593 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
594 				error = in_joingroup(ifp, &allhosts_addr,
595 				    NULL, &ii->ii_allhosts);
596 			}
597 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
598 		}
599 		goto out;
600 
601 	case SIOCDIFADDR:
602 		/*
603 		 * in_ifscrub kills the interface route.
604 		 */
605 		in_ifscrub(ifp, ia, LLE_STATIC);
606 
607 		/*
608 		 * in_ifadown gets rid of all the rest of
609 		 * the routes.  This is not quite the right
610 		 * thing to do, but at least if we are running
611 		 * a routing process they will come back.
612 		 */
613 		in_ifadown(&ia->ia_ifa, 1);
614 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
615 		error = 0;
616 		break;
617 
618 	default:
619 		panic("in_control: unsupported ioctl");
620 	}
621 
622 	if (ia->ia_ifa.ifa_carp)
623 		(*carp_detach_p)(&ia->ia_ifa);
624 
625 	IF_ADDR_LOCK(ifp);
626 	/* Re-check that ia is still part of the list. */
627 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
628 		if (ifa == &ia->ia_ifa)
629 			break;
630 	}
631 	if (ifa == NULL) {
632 		/*
633 		 * If we lost the race with another thread, there is no need to
634 		 * try it again for the next loop as there is no other exit
635 		 * path between here and out.
636 		 */
637 		IF_ADDR_UNLOCK(ifp);
638 		error = EADDRNOTAVAIL;
639 		goto out;
640 	}
641 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
642 	IF_ADDR_UNLOCK(ifp);
643 	ifa_free(&ia->ia_ifa);				/* if_addrhead */
644 
645 	IN_IFADDR_WLOCK();
646 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
647 
648 	LIST_REMOVE(ia, ia_hash);
649 	IN_IFADDR_WUNLOCK();
650 	/*
651 	 * If this is the last IPv4 address configured on this
652 	 * interface, leave the all-hosts group.
653 	 * No state-change report need be transmitted.
654 	 */
655 	IFP_TO_IA(ifp, iap);
656 	if (iap == NULL) {
657 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
658 		IN_MULTI_LOCK();
659 		if (ii->ii_allhosts) {
660 			(void)in_leavegroup_locked(ii->ii_allhosts, NULL);
661 			ii->ii_allhosts = NULL;
662 		}
663 		IN_MULTI_UNLOCK();
664 	} else
665 		ifa_free(&iap->ia_ifa);
666 
667 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
668 out:
669 	if (ia != NULL)
670 		ifa_free(&ia->ia_ifa);
671 	return (error);
672 }
673 
674 /*
675  * SIOC[GAD]LIFADDR.
676  *	SIOCGLIFADDR: get first address. (?!?)
677  *	SIOCGLIFADDR with IFLR_PREFIX:
678  *		get first address that matches the specified prefix.
679  *	SIOCALIFADDR: add the specified address.
680  *	SIOCALIFADDR with IFLR_PREFIX:
681  *		EINVAL since we can't deduce hostid part of the address.
682  *	SIOCDLIFADDR: delete the specified address.
683  *	SIOCDLIFADDR with IFLR_PREFIX:
684  *		delete the first address that matches the specified prefix.
685  * return values:
686  *	EINVAL on invalid parameters
687  *	EADDRNOTAVAIL on prefix match failed/specified address not found
688  *	other values may be returned from in_ioctl()
689  */
690 static int
691 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
692     struct ifnet *ifp, struct thread *td)
693 {
694 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
695 	struct ifaddr *ifa;
696 
697 	/* sanity checks */
698 	if (data == NULL || ifp == NULL) {
699 		panic("invalid argument to in_lifaddr_ioctl");
700 		/*NOTRECHED*/
701 	}
702 
703 	switch (cmd) {
704 	case SIOCGLIFADDR:
705 		/* address must be specified on GET with IFLR_PREFIX */
706 		if ((iflr->flags & IFLR_PREFIX) == 0)
707 			break;
708 		/*FALLTHROUGH*/
709 	case SIOCALIFADDR:
710 	case SIOCDLIFADDR:
711 		/* address must be specified on ADD and DELETE */
712 		if (iflr->addr.ss_family != AF_INET)
713 			return (EINVAL);
714 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
715 			return (EINVAL);
716 		/* XXX need improvement */
717 		if (iflr->dstaddr.ss_family
718 		 && iflr->dstaddr.ss_family != AF_INET)
719 			return (EINVAL);
720 		if (iflr->dstaddr.ss_family
721 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
722 			return (EINVAL);
723 		break;
724 	default: /*shouldn't happen*/
725 		return (EOPNOTSUPP);
726 	}
727 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
728 		return (EINVAL);
729 
730 	switch (cmd) {
731 	case SIOCALIFADDR:
732 	    {
733 		struct in_aliasreq ifra;
734 
735 		if (iflr->flags & IFLR_PREFIX)
736 			return (EINVAL);
737 
738 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
739 		bzero(&ifra, sizeof(ifra));
740 		bcopy(iflr->iflr_name, ifra.ifra_name,
741 			sizeof(ifra.ifra_name));
742 
743 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
744 
745 		if (iflr->dstaddr.ss_family) {	/*XXX*/
746 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
747 				iflr->dstaddr.ss_len);
748 		}
749 
750 		ifra.ifra_mask.sin_family = AF_INET;
751 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
752 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
753 
754 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
755 	    }
756 	case SIOCGLIFADDR:
757 	case SIOCDLIFADDR:
758 	    {
759 		struct in_ifaddr *ia;
760 		struct in_addr mask, candidate, match;
761 		struct sockaddr_in *sin;
762 
763 		bzero(&mask, sizeof(mask));
764 		bzero(&match, sizeof(match));
765 		if (iflr->flags & IFLR_PREFIX) {
766 			/* lookup a prefix rather than address. */
767 			in_len2mask(&mask, iflr->prefixlen);
768 
769 			sin = (struct sockaddr_in *)&iflr->addr;
770 			match.s_addr = sin->sin_addr.s_addr;
771 			match.s_addr &= mask.s_addr;
772 
773 			/* if you set extra bits, that's wrong */
774 			if (match.s_addr != sin->sin_addr.s_addr)
775 				return (EINVAL);
776 
777 		} else {
778 			/* on getting an address, take the 1st match */
779 			/* on deleting an address, do exact match */
780 			if (cmd != SIOCGLIFADDR) {
781 				in_len2mask(&mask, 32);
782 				sin = (struct sockaddr_in *)&iflr->addr;
783 				match.s_addr = sin->sin_addr.s_addr;
784 			}
785 		}
786 
787 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
788 			if (ifa->ifa_addr->sa_family != AF_INET6)
789 				continue;
790 			if (match.s_addr == 0)
791 				break;
792 			candidate.s_addr = ((struct sockaddr_in *)&ifa->ifa_addr)->sin_addr.s_addr;
793 			candidate.s_addr &= mask.s_addr;
794 			if (candidate.s_addr == match.s_addr)
795 				break;
796 		}
797 		if (ifa == NULL)
798 			return (EADDRNOTAVAIL);
799 		ia = (struct in_ifaddr *)ifa;
800 
801 		if (cmd == SIOCGLIFADDR) {
802 			/* fill in the if_laddrreq structure */
803 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
804 
805 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
806 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
807 					ia->ia_dstaddr.sin_len);
808 			} else
809 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
810 
811 			iflr->prefixlen =
812 				in_mask2len(&ia->ia_sockmask.sin_addr);
813 
814 			iflr->flags = 0;	/*XXX*/
815 
816 			return (0);
817 		} else {
818 			struct in_aliasreq ifra;
819 
820 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
821 			bzero(&ifra, sizeof(ifra));
822 			bcopy(iflr->iflr_name, ifra.ifra_name,
823 				sizeof(ifra.ifra_name));
824 
825 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
826 				ia->ia_addr.sin_len);
827 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
828 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
829 					ia->ia_dstaddr.sin_len);
830 			}
831 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
832 				ia->ia_sockmask.sin_len);
833 
834 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
835 			    ifp, td));
836 		}
837 	    }
838 	}
839 
840 	return (EOPNOTSUPP);	/*just for safety*/
841 }
842 
843 /*
844  * Delete any existing route for an interface.
845  */
846 void
847 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
848 {
849 
850 	in_scrubprefix(ia, flags);
851 }
852 
853 /*
854  * Initialize an interface's internet address
855  * and routing table entry.
856  */
857 static int
858 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
859     int scrub, int masksupplied, int vhid)
860 {
861 	register u_long i = ntohl(sin->sin_addr.s_addr);
862 	int flags = RTF_UP, error = 0;
863 
864 	if (scrub)
865 		in_scrubprefix(ia, LLE_STATIC);
866 
867 	IN_IFADDR_WLOCK();
868 	if (ia->ia_addr.sin_family == AF_INET)
869 		LIST_REMOVE(ia, ia_hash);
870 	ia->ia_addr = *sin;
871 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
872 	    ia, ia_hash);
873 	IN_IFADDR_WUNLOCK();
874 
875 	if (vhid > 0) {
876 		if (carp_attach_p != NULL)
877 			error = (*carp_attach_p)(&ia->ia_ifa, vhid);
878 		else
879 			error = EPROTONOSUPPORT;
880 	}
881 	if (error)
882 		return (error);
883 
884 	/*
885 	 * Give the interface a chance to initialize
886 	 * if this is its first address,
887 	 * and to validate the address if necessary.
888 	 */
889 	if (ifp->if_ioctl != NULL &&
890 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0)
891 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
892 			return (error);
893 
894 	/*
895 	 * Be compatible with network classes, if netmask isn't supplied,
896 	 * guess it based on classes.
897 	 */
898 	if (!masksupplied) {
899 		if (IN_CLASSA(i))
900 			ia->ia_subnetmask = IN_CLASSA_NET;
901 		else if (IN_CLASSB(i))
902 			ia->ia_subnetmask = IN_CLASSB_NET;
903 		else
904 			ia->ia_subnetmask = IN_CLASSC_NET;
905 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
906 	}
907 	ia->ia_subnet = i & ia->ia_subnetmask;
908 	in_socktrim(&ia->ia_sockmask);
909 	/*
910 	 * Add route for the network.
911 	 */
912 	ia->ia_ifa.ifa_metric = ifp->if_metric;
913 	if (ifp->if_flags & IFF_BROADCAST) {
914 		if (ia->ia_subnetmask == IN_RFC3021_MASK)
915 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
916 		else
917 			ia->ia_broadaddr.sin_addr.s_addr =
918 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
919 	} else if (ifp->if_flags & IFF_LOOPBACK) {
920 		ia->ia_dstaddr = ia->ia_addr;
921 		flags |= RTF_HOST;
922 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
923 		if (ia->ia_dstaddr.sin_family != AF_INET)
924 			return (0);
925 		flags |= RTF_HOST;
926 	}
927 	if (!vhid && (error = in_addprefix(ia, flags)) != 0)
928 		return (error);
929 
930 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
931 		return (0);
932 
933 	if (ifp->if_flags & IFF_POINTOPOINT &&
934 	    ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
935 			return (0);
936 
937 	/*
938 	 * add a loopback route to self
939 	 */
940 	if (V_useloopback && !vhid && !(ifp->if_flags & IFF_LOOPBACK)) {
941 		struct route ia_ro;
942 
943 		bzero(&ia_ro, sizeof(ia_ro));
944 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
945 		rtalloc_ign_fib(&ia_ro, 0, 0);
946 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
947 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
948 			RT_LOCK(ia_ro.ro_rt);
949 			RT_ADDREF(ia_ro.ro_rt);
950 			RTFREE_LOCKED(ia_ro.ro_rt);
951 		} else
952 			error = ifa_add_loopback_route((struct ifaddr *)ia,
953 				       (struct sockaddr *)&ia->ia_addr);
954 		if (error == 0)
955 			ia->ia_flags |= IFA_RTSELF;
956 		if (ia_ro.ro_rt != NULL)
957 			RTFREE(ia_ro.ro_rt);
958 	}
959 
960 	return (error);
961 }
962 
963 #define rtinitflags(x) \
964 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
965 	    ? RTF_HOST : 0)
966 
967 /*
968  * Generate a routing message when inserting or deleting
969  * an interface address alias.
970  */
971 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
972     struct in_ifaddr *target)
973 {
974 	struct route pfx_ro;
975 	struct sockaddr_in *pfx_addr;
976 	struct rtentry msg_rt;
977 
978 	/* QL: XXX
979 	 * This is a bit questionable because there is no
980 	 * additional route entry added/deleted for an address
981 	 * alias. Therefore this route report is inaccurate.
982 	 */
983 	bzero(&pfx_ro, sizeof(pfx_ro));
984 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
985 	pfx_addr->sin_len = sizeof(*pfx_addr);
986 	pfx_addr->sin_family = AF_INET;
987 	pfx_addr->sin_addr = *prefix;
988 	rtalloc_ign_fib(&pfx_ro, 0, 0);
989 	if (pfx_ro.ro_rt != NULL) {
990 		msg_rt = *pfx_ro.ro_rt;
991 
992 		/* QL: XXX
993 		 * Point the gateway to the new interface
994 		 * address as if a new prefix route entry has
995 		 * been added through the new address alias.
996 		 * All other parts of the rtentry is accurate,
997 		 * e.g., rt_key, rt_mask, rt_ifp etc.
998 		 */
999 		msg_rt.rt_gateway =
1000 			(struct sockaddr *)&target->ia_addr;
1001 		rt_newaddrmsg(cmd,
1002 			      (struct ifaddr *)target,
1003 			      0, &msg_rt);
1004 		RTFREE(pfx_ro.ro_rt);
1005 	}
1006 	return;
1007 }
1008 
1009 /*
1010  * Check if we have a route for the given prefix already or add one accordingly.
1011  */
1012 int
1013 in_addprefix(struct in_ifaddr *target, int flags)
1014 {
1015 	struct in_ifaddr *ia;
1016 	struct in_addr prefix, mask, p, m;
1017 	int error;
1018 
1019 	if ((flags & RTF_HOST) != 0) {
1020 		prefix = target->ia_dstaddr.sin_addr;
1021 		mask.s_addr = 0;
1022 	} else {
1023 		prefix = target->ia_addr.sin_addr;
1024 		mask = target->ia_sockmask.sin_addr;
1025 		prefix.s_addr &= mask.s_addr;
1026 	}
1027 
1028 	IN_IFADDR_RLOCK();
1029 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1030 		if (rtinitflags(ia)) {
1031 			p = ia->ia_dstaddr.sin_addr;
1032 
1033 			if (prefix.s_addr != p.s_addr)
1034 				continue;
1035 		} else {
1036 			p = ia->ia_addr.sin_addr;
1037 			m = ia->ia_sockmask.sin_addr;
1038 			p.s_addr &= m.s_addr;
1039 
1040 			if (prefix.s_addr != p.s_addr ||
1041 			    mask.s_addr != m.s_addr)
1042 				continue;
1043 		}
1044 
1045 		/*
1046 		 * If we got a matching prefix route inserted by other
1047 		 * interface address, we are done here.
1048 		 */
1049 		if (ia->ia_flags & IFA_ROUTE) {
1050 #ifdef RADIX_MPATH
1051 			if (ia->ia_addr.sin_addr.s_addr ==
1052 			    target->ia_addr.sin_addr.s_addr) {
1053 				IN_IFADDR_RUNLOCK();
1054 				return (EEXIST);
1055 			} else
1056 				break;
1057 #endif
1058 			if (V_nosameprefix) {
1059 				IN_IFADDR_RUNLOCK();
1060 				return (EEXIST);
1061 			} else {
1062 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
1063 				IN_IFADDR_RUNLOCK();
1064 				return (0);
1065 			}
1066 		}
1067 	}
1068 	IN_IFADDR_RUNLOCK();
1069 
1070 	/*
1071 	 * No-one seem to have this prefix route, so we try to insert it.
1072 	 */
1073 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
1074 	if (!error)
1075 		target->ia_flags |= IFA_ROUTE;
1076 	return (error);
1077 }
1078 
1079 /*
1080  * If there is no other address in the system that can serve a route to the
1081  * same prefix, remove the route.  Hand over the route to the new address
1082  * otherwise.
1083  */
1084 int
1085 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1086 {
1087 	struct in_ifaddr *ia;
1088 	struct in_addr prefix, mask, p, m;
1089 	int error = 0;
1090 	struct sockaddr_in prefix0, mask0;
1091 
1092 	/*
1093 	 * Remove the loopback route to the interface address.
1094 	 * The "useloopback" setting is not consulted because if the
1095 	 * user configures an interface address, turns off this
1096 	 * setting, and then tries to delete that interface address,
1097 	 * checking the current setting of "useloopback" would leave
1098 	 * that interface address loopback route untouched, which
1099 	 * would be wrong. Therefore the interface address loopback route
1100 	 * deletion is unconditional.
1101 	 */
1102 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1103 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1104 	    (target->ia_flags & IFA_RTSELF)) {
1105 		struct route ia_ro;
1106 		int freeit = 0;
1107 
1108 		bzero(&ia_ro, sizeof(ia_ro));
1109 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1110 		rtalloc_ign_fib(&ia_ro, 0, 0);
1111 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1112 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1113 			RT_LOCK(ia_ro.ro_rt);
1114 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1115 				freeit = 1;
1116 			else if (flags & LLE_STATIC) {
1117 				RT_REMREF(ia_ro.ro_rt);
1118 				target->ia_flags &= ~IFA_RTSELF;
1119 			}
1120 			RTFREE_LOCKED(ia_ro.ro_rt);
1121 		}
1122 		if (freeit && (flags & LLE_STATIC)) {
1123 			error = ifa_del_loopback_route((struct ifaddr *)target,
1124 				       (struct sockaddr *)&target->ia_addr);
1125 			if (error == 0)
1126 				target->ia_flags &= ~IFA_RTSELF;
1127 		}
1128 		if ((flags & LLE_STATIC) &&
1129 			!(target->ia_ifp->if_flags & IFF_NOARP))
1130 			/* remove arp cache */
1131 			arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1132 	}
1133 
1134 	if (rtinitflags(target)) {
1135 		prefix = target->ia_dstaddr.sin_addr;
1136 		mask.s_addr = 0;
1137 	} else {
1138 		prefix = target->ia_addr.sin_addr;
1139 		mask = target->ia_sockmask.sin_addr;
1140 		prefix.s_addr &= mask.s_addr;
1141 	}
1142 
1143 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1144 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1145 		return (0);
1146 	}
1147 
1148 	IN_IFADDR_RLOCK();
1149 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1150 		if (rtinitflags(ia)) {
1151 			p = ia->ia_dstaddr.sin_addr;
1152 
1153 			if (prefix.s_addr != p.s_addr)
1154 				continue;
1155 		} else {
1156 			p = ia->ia_addr.sin_addr;
1157 			m = ia->ia_sockmask.sin_addr;
1158 			p.s_addr &= m.s_addr;
1159 
1160 			if (prefix.s_addr != p.s_addr ||
1161 			    mask.s_addr != m.s_addr)
1162 				continue;
1163 		}
1164 
1165 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1166 			continue;
1167 
1168 		/*
1169 		 * If we got a matching prefix address, move IFA_ROUTE and
1170 		 * the route itself to it.  Make sure that routing daemons
1171 		 * get a heads-up.
1172 		 */
1173 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1174 			ifa_ref(&ia->ia_ifa);
1175 			IN_IFADDR_RUNLOCK();
1176 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1177 			    rtinitflags(target));
1178 			if (error == 0)
1179 				target->ia_flags &= ~IFA_ROUTE;
1180 			else
1181 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1182 					error);
1183 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1184 			    rtinitflags(ia) | RTF_UP);
1185 			if (error == 0)
1186 				ia->ia_flags |= IFA_ROUTE;
1187 			else
1188 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1189 					error);
1190 			ifa_free(&ia->ia_ifa);
1191 			return (error);
1192 		}
1193 	}
1194 	IN_IFADDR_RUNLOCK();
1195 
1196 	/*
1197 	 * remove all L2 entries on the given prefix
1198 	 */
1199 	bzero(&prefix0, sizeof(prefix0));
1200 	prefix0.sin_len = sizeof(prefix0);
1201 	prefix0.sin_family = AF_INET;
1202 	prefix0.sin_addr.s_addr = target->ia_subnet;
1203 	bzero(&mask0, sizeof(mask0));
1204 	mask0.sin_len = sizeof(mask0);
1205 	mask0.sin_family = AF_INET;
1206 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1207 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1208 			    (struct sockaddr *)&mask0, flags);
1209 
1210 	/*
1211 	 * As no-one seem to have this prefix, we can remove the route.
1212 	 */
1213 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1214 	if (error == 0)
1215 		target->ia_flags &= ~IFA_ROUTE;
1216 	else
1217 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1218 	return (error);
1219 }
1220 
1221 #undef rtinitflags
1222 
1223 /*
1224  * Return 1 if the address might be a local broadcast address.
1225  */
1226 int
1227 in_broadcast(struct in_addr in, struct ifnet *ifp)
1228 {
1229 	register struct ifaddr *ifa;
1230 	u_long t;
1231 
1232 	if (in.s_addr == INADDR_BROADCAST ||
1233 	    in.s_addr == INADDR_ANY)
1234 		return (1);
1235 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1236 		return (0);
1237 	t = ntohl(in.s_addr);
1238 	/*
1239 	 * Look through the list of addresses for a match
1240 	 * with a broadcast address.
1241 	 */
1242 #define ia ((struct in_ifaddr *)ifa)
1243 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1244 		if (ifa->ifa_addr->sa_family == AF_INET &&
1245 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1246 		     /*
1247 		      * Check for old-style (host 0) broadcast, but
1248 		      * taking into account that RFC 3021 obsoletes it.
1249 		      */
1250 		     (ia->ia_subnetmask != IN_RFC3021_MASK &&
1251 		     t == ia->ia_subnet)) &&
1252 		     /*
1253 		      * Check for an all one subnetmask. These
1254 		      * only exist when an interface gets a secondary
1255 		      * address.
1256 		      */
1257 		     ia->ia_subnetmask != (u_long)0xffffffff)
1258 			    return (1);
1259 	return (0);
1260 #undef ia
1261 }
1262 
1263 /*
1264  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1265  */
1266 void
1267 in_ifdetach(struct ifnet *ifp)
1268 {
1269 
1270 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1271 	in_pcbpurgeif0(&V_udbinfo, ifp);
1272 	in_purgemaddrs(ifp);
1273 }
1274 
1275 /*
1276  * Delete all IPv4 multicast address records, and associated link-layer
1277  * multicast address records, associated with ifp.
1278  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1279  * XXX This should not race with ifma_protospec being set during
1280  * a new allocation, if it does, we have bigger problems.
1281  */
1282 static void
1283 in_purgemaddrs(struct ifnet *ifp)
1284 {
1285 	LIST_HEAD(,in_multi) purgeinms;
1286 	struct in_multi		*inm, *tinm;
1287 	struct ifmultiaddr	*ifma;
1288 
1289 	LIST_INIT(&purgeinms);
1290 	IN_MULTI_LOCK();
1291 
1292 	/*
1293 	 * Extract list of in_multi associated with the detaching ifp
1294 	 * which the PF_INET layer is about to release.
1295 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1296 	 * by code further down.
1297 	 */
1298 	IF_ADDR_LOCK(ifp);
1299 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1300 		if (ifma->ifma_addr->sa_family != AF_INET ||
1301 		    ifma->ifma_protospec == NULL)
1302 			continue;
1303 #if 0
1304 		KASSERT(ifma->ifma_protospec != NULL,
1305 		    ("%s: ifma_protospec is NULL", __func__));
1306 #endif
1307 		inm = (struct in_multi *)ifma->ifma_protospec;
1308 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1309 	}
1310 	IF_ADDR_UNLOCK(ifp);
1311 
1312 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1313 		LIST_REMOVE(inm, inm_link);
1314 		inm_release_locked(inm);
1315 	}
1316 	igmp_ifdetach(ifp);
1317 
1318 	IN_MULTI_UNLOCK();
1319 }
1320 
1321 struct in_llentry {
1322 	struct llentry		base;
1323 	struct sockaddr_in	l3_addr4;
1324 };
1325 
1326 static struct llentry *
1327 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1328 {
1329 	struct in_llentry *lle;
1330 
1331 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO);
1332 	if (lle == NULL)		/* NB: caller generates msg */
1333 		return NULL;
1334 
1335 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1336 	/*
1337 	 * For IPv4 this will trigger "arpresolve" to generate
1338 	 * an ARP request.
1339 	 */
1340 	lle->base.la_expire = time_uptime; /* mark expired */
1341 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1342 	lle->base.lle_refcnt = 1;
1343 	LLE_LOCK_INIT(&lle->base);
1344 	return &lle->base;
1345 }
1346 
1347 /*
1348  * Deletes an address from the address table.
1349  * This function is called by the timer functions
1350  * such as arptimer() and nd6_llinfo_timer(), and
1351  * the caller does the locking.
1352  */
1353 static void
1354 in_lltable_free(struct lltable *llt, struct llentry *lle)
1355 {
1356 	LLE_WUNLOCK(lle);
1357 	LLE_LOCK_DESTROY(lle);
1358 	free(lle, M_LLTABLE);
1359 }
1360 
1361 
1362 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1363 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1364 
1365 static void
1366 in_lltable_prefix_free(struct lltable *llt,
1367 		       const struct sockaddr *prefix,
1368 		       const struct sockaddr *mask,
1369 		       u_int flags)
1370 {
1371 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1372 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1373 	struct llentry *lle, *next;
1374 	register int i;
1375 	size_t pkts_dropped;
1376 
1377 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
1378 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1379 
1380 		        /*
1381 			 * (flags & LLE_STATIC) means deleting all entries
1382 			 * including static ARP entries
1383 			 */
1384 			if (IN_ARE_MASKED_ADDR_EQUAL((struct sockaddr_in *)L3_ADDR(lle),
1385 						     pfx, msk) &&
1386 			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
1387 				int canceled;
1388 
1389 				canceled = callout_drain(&lle->la_timer);
1390 				LLE_WLOCK(lle);
1391 				if (canceled)
1392 					LLE_REMREF(lle);
1393 				pkts_dropped = llentry_free(lle);
1394 				ARPSTAT_ADD(dropped, pkts_dropped);
1395 			}
1396 		}
1397 	}
1398 }
1399 
1400 
1401 static int
1402 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1403 {
1404 	struct rtentry *rt;
1405 
1406 	KASSERT(l3addr->sa_family == AF_INET,
1407 	    ("sin_family %d", l3addr->sa_family));
1408 
1409 	/* XXX rtalloc1 should take a const param */
1410 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1411 
1412 	if (rt == NULL)
1413 		return (EINVAL);
1414 
1415 	/*
1416 	 * If the gateway for an existing host route matches the target L3
1417 	 * address, which is a special route inserted by some implementation
1418 	 * such as MANET, and the interface is of the correct type, then
1419 	 * allow for ARP to proceed.
1420 	 */
1421 	if (rt->rt_flags & RTF_GATEWAY) {
1422 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1423 			rt->rt_ifp->if_type != IFT_ETHER ||
1424 			  (rt->rt_ifp->if_flags &
1425 			   (IFF_NOARP | IFF_STATICARP)) != 0 ||
1426 			  memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1427 				 sizeof(in_addr_t)) != 0) {
1428 			RTFREE_LOCKED(rt);
1429 			return (EINVAL);
1430 		}
1431 	}
1432 
1433 	/*
1434 	 * Make sure that at least the destination address is covered
1435 	 * by the route. This is for handling the case where 2 or more
1436 	 * interfaces have the same prefix. An incoming packet arrives
1437 	 * on one interface and the corresponding outgoing packet leaves
1438 	 * another interface.
1439 	 */
1440 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1441 		const char *sa, *mask, *addr, *lim;
1442 		int len;
1443 
1444 		mask = (const char *)rt_mask(rt);
1445 		/*
1446 		 * Just being extra cautious to avoid some custom
1447 		 * code getting into trouble.
1448 		 */
1449 		if (mask == NULL) {
1450 			RTFREE_LOCKED(rt);
1451 			return (EINVAL);
1452 		}
1453 
1454 		sa = (const char *)rt_key(rt);
1455 		addr = (const char *)l3addr;
1456 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
1457 		lim = addr + len;
1458 
1459 		for ( ; addr < lim; sa++, mask++, addr++) {
1460 			if ((*sa ^ *addr) & *mask) {
1461 #ifdef DIAGNOSTIC
1462 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1463 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1464 #endif
1465 				RTFREE_LOCKED(rt);
1466 				return (EINVAL);
1467 			}
1468 		}
1469 	}
1470 
1471 	RTFREE_LOCKED(rt);
1472 	return (0);
1473 }
1474 
1475 /*
1476  * Return NULL if not found or marked for deletion.
1477  * If found return lle read locked.
1478  */
1479 static struct llentry *
1480 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1481 {
1482 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1483 	struct ifnet *ifp = llt->llt_ifp;
1484 	struct llentry *lle;
1485 	struct llentries *lleh;
1486 	u_int hashkey;
1487 
1488 	IF_AFDATA_LOCK_ASSERT(ifp);
1489 	KASSERT(l3addr->sa_family == AF_INET,
1490 	    ("sin_family %d", l3addr->sa_family));
1491 
1492 	hashkey = sin->sin_addr.s_addr;
1493 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1494 	LIST_FOREACH(lle, lleh, lle_next) {
1495 		struct sockaddr_in *sa2 = (struct sockaddr_in *)L3_ADDR(lle);
1496 		if (lle->la_flags & LLE_DELETED)
1497 			continue;
1498 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1499 			break;
1500 	}
1501 	if (lle == NULL) {
1502 #ifdef DIAGNOSTIC
1503 		if (flags & LLE_DELETE)
1504 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1505 #endif
1506 		if (!(flags & LLE_CREATE))
1507 			return (NULL);
1508 		/*
1509 		 * A route that covers the given address must have
1510 		 * been installed 1st because we are doing a resolution,
1511 		 * verify this.
1512 		 */
1513 		if (!(flags & LLE_IFADDR) &&
1514 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1515 			goto done;
1516 
1517 		lle = in_lltable_new(l3addr, flags);
1518 		if (lle == NULL) {
1519 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1520 			goto done;
1521 		}
1522 		lle->la_flags = flags & ~LLE_CREATE;
1523 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1524 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1525 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1526 		}
1527 
1528 		lle->lle_tbl  = llt;
1529 		lle->lle_head = lleh;
1530 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1531 	} else if (flags & LLE_DELETE) {
1532 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1533 			LLE_WLOCK(lle);
1534 			lle->la_flags = LLE_DELETED;
1535 			EVENTHANDLER_INVOKE(arp_update_event, lle);
1536 			LLE_WUNLOCK(lle);
1537 #ifdef DIAGNOSTIC
1538 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
1539 #endif
1540 		}
1541 		lle = (void *)-1;
1542 
1543 	}
1544 	if (LLE_IS_VALID(lle)) {
1545 		if (flags & LLE_EXCLUSIVE)
1546 			LLE_WLOCK(lle);
1547 		else
1548 			LLE_RLOCK(lle);
1549 	}
1550 done:
1551 	return (lle);
1552 }
1553 
1554 static int
1555 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1556 {
1557 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1558 	struct ifnet *ifp = llt->llt_ifp;
1559 	struct llentry *lle;
1560 	/* XXX stack use */
1561 	struct {
1562 		struct rt_msghdr	rtm;
1563 		struct sockaddr_inarp	sin;
1564 		struct sockaddr_dl	sdl;
1565 	} arpc;
1566 	int error, i;
1567 
1568 	LLTABLE_LOCK_ASSERT();
1569 
1570 	error = 0;
1571 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1572 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1573 			struct sockaddr_dl *sdl;
1574 
1575 			/* skip deleted entries */
1576 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1577 				continue;
1578 			/* Skip if jailed and not a valid IP of the prison. */
1579 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1580 				continue;
1581 			/*
1582 			 * produce a msg made of:
1583 			 *  struct rt_msghdr;
1584 			 *  struct sockaddr_inarp; (IPv4)
1585 			 *  struct sockaddr_dl;
1586 			 */
1587 			bzero(&arpc, sizeof(arpc));
1588 			arpc.rtm.rtm_msglen = sizeof(arpc);
1589 			arpc.rtm.rtm_version = RTM_VERSION;
1590 			arpc.rtm.rtm_type = RTM_GET;
1591 			arpc.rtm.rtm_flags = RTF_UP;
1592 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1593 			arpc.sin.sin_family = AF_INET;
1594 			arpc.sin.sin_len = sizeof(arpc.sin);
1595 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1596 
1597 			/* publish */
1598 			if (lle->la_flags & LLE_PUB) {
1599 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1600 				/* proxy only */
1601 				if (lle->la_flags & LLE_PROXY)
1602 					arpc.sin.sin_other = SIN_PROXY;
1603 			}
1604 
1605 			sdl = &arpc.sdl;
1606 			sdl->sdl_family = AF_LINK;
1607 			sdl->sdl_len = sizeof(*sdl);
1608 			sdl->sdl_index = ifp->if_index;
1609 			sdl->sdl_type = ifp->if_type;
1610 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1611 				sdl->sdl_alen = ifp->if_addrlen;
1612 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1613 			} else {
1614 				sdl->sdl_alen = 0;
1615 				bzero(LLADDR(sdl), ifp->if_addrlen);
1616 			}
1617 
1618 			arpc.rtm.rtm_rmx.rmx_expire =
1619 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1620 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1621 			if (lle->la_flags & LLE_STATIC)
1622 				arpc.rtm.rtm_flags |= RTF_STATIC;
1623 			arpc.rtm.rtm_index = ifp->if_index;
1624 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1625 			if (error)
1626 				break;
1627 		}
1628 	}
1629 	return error;
1630 #undef SIN
1631 }
1632 
1633 void *
1634 in_domifattach(struct ifnet *ifp)
1635 {
1636 	struct in_ifinfo *ii;
1637 	struct lltable *llt;
1638 
1639 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1640 
1641 	llt = lltable_init(ifp, AF_INET);
1642 	if (llt != NULL) {
1643 		llt->llt_free = in_lltable_free;
1644 		llt->llt_prefix_free = in_lltable_prefix_free;
1645 		llt->llt_lookup = in_lltable_lookup;
1646 		llt->llt_dump = in_lltable_dump;
1647 	}
1648 	ii->ii_llt = llt;
1649 
1650 	ii->ii_igmp = igmp_domifattach(ifp);
1651 
1652 	return ii;
1653 }
1654 
1655 void
1656 in_domifdetach(struct ifnet *ifp, void *aux)
1657 {
1658 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1659 
1660 	igmp_domifdetach(ifp);
1661 	lltable_free(ii->ii_llt);
1662 	free(ii, M_IFADDR);
1663 }
1664