xref: /freebsd/sys/netinet/in.c (revision 6ab38b8e25f31df8140b99c09160d0207f912ef3)
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);
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 /*
89  * Return 1 if an internet address is for a ``local'' host
90  * (one to which we have a connection).
91  */
92 int
93 in_localaddr(struct in_addr in)
94 {
95 	register u_long i = ntohl(in.s_addr);
96 	register struct in_ifaddr *ia;
97 
98 	IN_IFADDR_RLOCK();
99 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
100 		if ((i & ia->ia_subnetmask) == ia->ia_subnet) {
101 			IN_IFADDR_RUNLOCK();
102 			return (1);
103 		}
104 	}
105 	IN_IFADDR_RUNLOCK();
106 	return (0);
107 }
108 
109 /*
110  * Return 1 if an internet address is for the local host and configured
111  * on one of its interfaces.
112  */
113 int
114 in_localip(struct in_addr in)
115 {
116 	struct in_ifaddr *ia;
117 
118 	IN_IFADDR_RLOCK();
119 	LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) {
120 		if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) {
121 			IN_IFADDR_RUNLOCK();
122 			return (1);
123 		}
124 	}
125 	IN_IFADDR_RUNLOCK();
126 	return (0);
127 }
128 
129 /*
130  * Determine whether an IP address is in a reserved set of addresses
131  * that may not be forwarded, or whether datagrams to that destination
132  * may be forwarded.
133  */
134 int
135 in_canforward(struct in_addr in)
136 {
137 	register u_long i = ntohl(in.s_addr);
138 	register u_long net;
139 
140 	if (IN_EXPERIMENTAL(i) || IN_MULTICAST(i) || IN_LINKLOCAL(i))
141 		return (0);
142 	if (IN_CLASSA(i)) {
143 		net = i & IN_CLASSA_NET;
144 		if (net == 0 || net == (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
145 			return (0);
146 	}
147 	return (1);
148 }
149 
150 /*
151  * Trim a mask in a sockaddr
152  */
153 static void
154 in_socktrim(struct sockaddr_in *ap)
155 {
156     register char *cplim = (char *) &ap->sin_addr;
157     register char *cp = (char *) (&ap->sin_addr + 1);
158 
159     ap->sin_len = 0;
160     while (--cp >= cplim)
161 	if (*cp) {
162 	    (ap)->sin_len = cp - (char *) (ap) + 1;
163 	    break;
164 	}
165 }
166 
167 static int
168 in_mask2len(mask)
169 	struct in_addr *mask;
170 {
171 	int x, y;
172 	u_char *p;
173 
174 	p = (u_char *)mask;
175 	for (x = 0; x < sizeof(*mask); x++) {
176 		if (p[x] != 0xff)
177 			break;
178 	}
179 	y = 0;
180 	if (x < sizeof(*mask)) {
181 		for (y = 0; y < 8; y++) {
182 			if ((p[x] & (0x80 >> y)) == 0)
183 				break;
184 		}
185 	}
186 	return (x * 8 + y);
187 }
188 
189 static void
190 in_len2mask(struct in_addr *mask, int len)
191 {
192 	int i;
193 	u_char *p;
194 
195 	p = (u_char *)mask;
196 	bzero(mask, sizeof(*mask));
197 	for (i = 0; i < len / 8; i++)
198 		p[i] = 0xff;
199 	if (len % 8)
200 		p[i] = (0xff00 >> (len % 8)) & 0xff;
201 }
202 
203 /*
204  * Generic internet control operations (ioctl's).
205  *
206  * ifp is NULL if not an interface-specific ioctl.
207  */
208 /* ARGSUSED */
209 int
210 in_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp,
211     struct thread *td)
212 {
213 	register struct ifreq *ifr = (struct ifreq *)data;
214 	register struct in_ifaddr *ia, *iap;
215 	register struct ifaddr *ifa;
216 	struct in_addr allhosts_addr;
217 	struct in_addr dst;
218 	struct in_ifinfo *ii;
219 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
220 	int error, hostIsNew, iaIsNew, maskIsNew;
221 	int iaIsFirst;
222 	u_long ocmd = cmd;
223 
224 	/*
225 	 * Pre-10.x compat: OSIOCAIFADDR passes a shorter
226 	 * struct in_aliasreq, without ifra_vhid.
227 	 */
228 	if (cmd == OSIOCAIFADDR)
229 		cmd = SIOCAIFADDR;
230 
231 	ia = NULL;
232 	iaIsFirst = 0;
233 	iaIsNew = 0;
234 	allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP);
235 
236 	/*
237 	 * Filter out ioctls we implement directly; forward the rest on to
238 	 * in_lifaddr_ioctl() and ifp->if_ioctl().
239 	 */
240 	switch (cmd) {
241 	case SIOCGIFADDR:
242 	case SIOCGIFBRDADDR:
243 	case SIOCGIFDSTADDR:
244 	case SIOCGIFNETMASK:
245 	case SIOCDIFADDR:
246 		break;
247 	case SIOCAIFADDR:
248 		/*
249 		 * ifra_addr must be present and be of INET family.
250 		 * ifra_broadaddr and ifra_mask are optional.
251 		 */
252 		if (ifra->ifra_addr.sin_len != sizeof(struct sockaddr_in) ||
253 		    ifra->ifra_addr.sin_family != AF_INET)
254 			return (EINVAL);
255 		if (ifra->ifra_broadaddr.sin_len != 0 &&
256 		    (ifra->ifra_broadaddr.sin_len !=
257 		    sizeof(struct sockaddr_in) ||
258 		    ifra->ifra_broadaddr.sin_family != AF_INET))
259 			return (EINVAL);
260 #if 0
261 		/*
262 		 * ifconfig(8) in pre-10.x doesn't set sin_family for the
263 		 * mask. The code is disabled for the 10.x timeline, to
264 		 * make SIOCAIFADDR compatible with 9.x ifconfig(8).
265 		 * The code should be enabled in 11.x
266 		 */
267 		if (ifra->ifra_mask.sin_len != 0 &&
268 		    (ifra->ifra_mask.sin_len != sizeof(struct sockaddr_in) ||
269 		    ifra->ifra_mask.sin_family != AF_INET))
270 			return (EINVAL);
271 #endif
272 		break;
273 	case SIOCSIFADDR:
274 	case SIOCSIFBRDADDR:
275 	case SIOCSIFDSTADDR:
276 	case SIOCSIFNETMASK:
277 		/* We no longer support that old commands. */
278 		return (EINVAL);
279 
280 	case SIOCALIFADDR:
281 		if (td != NULL) {
282 			error = priv_check(td, PRIV_NET_ADDIFADDR);
283 			if (error)
284 				return (error);
285 		}
286 		if (ifp == NULL)
287 			return (EINVAL);
288 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
289 
290 	case SIOCDLIFADDR:
291 		if (td != NULL) {
292 			error = priv_check(td, PRIV_NET_DELIFADDR);
293 			if (error)
294 				return (error);
295 		}
296 		if (ifp == NULL)
297 			return (EINVAL);
298 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
299 
300 	case SIOCGLIFADDR:
301 		if (ifp == NULL)
302 			return (EINVAL);
303 		return in_lifaddr_ioctl(so, cmd, data, ifp, td);
304 
305 	default:
306 		if (ifp == NULL || ifp->if_ioctl == NULL)
307 			return (EOPNOTSUPP);
308 		return ((*ifp->if_ioctl)(ifp, cmd, data));
309 	}
310 
311 	if (ifp == NULL)
312 		return (EADDRNOTAVAIL);
313 
314 	/*
315 	 * Security checks before we get involved in any work.
316 	 */
317 	switch (cmd) {
318 	case SIOCAIFADDR:
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_RLOCK(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_RUNLOCK(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 		if (ia == NULL) {
407 			ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK);
408 			ia = (struct in_ifaddr *)ifa;
409 			ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr;
410 			ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
411 			ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask;
412 
413 			ia->ia_sockmask.sin_len = 8;
414 			ia->ia_sockmask.sin_family = AF_INET;
415 			if (ifp->if_flags & IFF_BROADCAST) {
416 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
417 				ia->ia_broadaddr.sin_family = AF_INET;
418 			}
419 			ia->ia_ifp = ifp;
420 
421 			ifa_ref(ifa);			/* if_addrhead */
422 			IF_ADDR_WLOCK(ifp);
423 			TAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link);
424 			IF_ADDR_WUNLOCK(ifp);
425 			ifa_ref(ifa);			/* in_ifaddrhead */
426 			IN_IFADDR_WLOCK();
427 			TAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link);
428 			IN_IFADDR_WUNLOCK();
429 			iaIsNew = 1;
430 		}
431 		break;
432 
433 	case SIOCGIFADDR:
434 	case SIOCGIFNETMASK:
435 	case SIOCGIFDSTADDR:
436 	case SIOCGIFBRDADDR:
437 		if (ia == NULL) {
438 			error = EADDRNOTAVAIL;
439 			goto out;
440 		}
441 		break;
442 	}
443 
444 	/*
445 	 * Most paths in this switch return directly or via out.  Only paths
446 	 * that remove the address break in order to hit common removal code.
447 	 */
448 	switch (cmd) {
449 	case SIOCGIFADDR:
450 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_addr;
451 		goto out;
452 
453 	case SIOCGIFBRDADDR:
454 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
455 			error = EINVAL;
456 			goto out;
457 		}
458 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_broadaddr;
459 		goto out;
460 
461 	case SIOCGIFDSTADDR:
462 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
463 			error = EINVAL;
464 			goto out;
465 		}
466 		*((struct sockaddr_in *)&ifr->ifr_dstaddr) = ia->ia_dstaddr;
467 		goto out;
468 
469 	case SIOCGIFNETMASK:
470 		*((struct sockaddr_in *)&ifr->ifr_addr) = ia->ia_sockmask;
471 		goto out;
472 
473 	case SIOCAIFADDR:
474 		maskIsNew = 0;
475 		hostIsNew = 1;
476 		error = 0;
477 		if (ifra->ifra_addr.sin_addr.s_addr ==
478 			    ia->ia_addr.sin_addr.s_addr)
479 			hostIsNew = 0;
480 		if (ifra->ifra_mask.sin_len) {
481 			/*
482 			 * QL: XXX
483 			 * Need to scrub the prefix here in case
484 			 * the issued command is SIOCAIFADDR with
485 			 * the same address, but with a different
486 			 * prefix length. And if the prefix length
487 			 * is the same as before, then the call is
488 			 * un-necessarily executed here.
489 			 */
490 			in_ifscrub(ifp, ia, LLE_STATIC);
491 			ia->ia_sockmask = ifra->ifra_mask;
492 			ia->ia_sockmask.sin_family = AF_INET;
493 			ia->ia_subnetmask =
494 			    ntohl(ia->ia_sockmask.sin_addr.s_addr);
495 			maskIsNew = 1;
496 		}
497 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
498 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
499 			in_ifscrub(ifp, ia, LLE_STATIC);
500 			ia->ia_dstaddr = ifra->ifra_dstaddr;
501 			maskIsNew  = 1; /* We lie; but the effect's the same */
502 		}
503 		if (hostIsNew || maskIsNew)
504 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, maskIsNew,
505 			    (ocmd == cmd ? ifra->ifra_vhid : 0));
506 		if (error != 0 && iaIsNew)
507 			break;
508 
509 		if ((ifp->if_flags & IFF_BROADCAST) &&
510 		    ifra->ifra_broadaddr.sin_len)
511 			ia->ia_broadaddr = ifra->ifra_broadaddr;
512 		if (error == 0) {
513 			ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
514 			if (iaIsFirst &&
515 			    (ifp->if_flags & IFF_MULTICAST) != 0) {
516 				error = in_joingroup(ifp, &allhosts_addr,
517 				    NULL, &ii->ii_allhosts);
518 			}
519 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
520 		}
521 		goto out;
522 
523 	case SIOCDIFADDR:
524 		/*
525 		 * in_ifscrub kills the interface route.
526 		 */
527 		in_ifscrub(ifp, ia, LLE_STATIC);
528 
529 		/*
530 		 * in_ifadown gets rid of all the rest of
531 		 * the routes.  This is not quite the right
532 		 * thing to do, but at least if we are running
533 		 * a routing process they will come back.
534 		 */
535 		in_ifadown(&ia->ia_ifa, 1);
536 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
537 		error = 0;
538 		break;
539 
540 	default:
541 		panic("in_control: unsupported ioctl");
542 	}
543 
544 	if (ia->ia_ifa.ifa_carp)
545 		(*carp_detach_p)(&ia->ia_ifa);
546 
547 	IF_ADDR_WLOCK(ifp);
548 	/* Re-check that ia is still part of the list. */
549 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
550 		if (ifa == &ia->ia_ifa)
551 			break;
552 	}
553 	if (ifa == NULL) {
554 		/*
555 		 * If we lost the race with another thread, there is no need to
556 		 * try it again for the next loop as there is no other exit
557 		 * path between here and out.
558 		 */
559 		IF_ADDR_WUNLOCK(ifp);
560 		error = EADDRNOTAVAIL;
561 		goto out;
562 	}
563 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
564 	IF_ADDR_WUNLOCK(ifp);
565 	ifa_free(&ia->ia_ifa);		      /* if_addrhead */
566 
567 	IN_IFADDR_WLOCK();
568 	TAILQ_REMOVE(&V_in_ifaddrhead, ia, ia_link);
569 
570 	LIST_REMOVE(ia, ia_hash);
571 	IN_IFADDR_WUNLOCK();
572 	/*
573 	 * If this is the last IPv4 address configured on this
574 	 * interface, leave the all-hosts group.
575 	 * No state-change report need be transmitted.
576 	 */
577 	IFP_TO_IA(ifp, iap);
578 	if (iap == NULL) {
579 		ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]);
580 		IN_MULTI_LOCK();
581 		if (ii->ii_allhosts) {
582 			(void)in_leavegroup_locked(ii->ii_allhosts, NULL);
583 			ii->ii_allhosts = NULL;
584 		}
585 		IN_MULTI_UNLOCK();
586 	} else
587 		ifa_free(&iap->ia_ifa);
588 
589 	ifa_free(&ia->ia_ifa);				/* in_ifaddrhead */
590 out:
591 	if (ia != NULL)
592 		ifa_free(&ia->ia_ifa);
593 	return (error);
594 }
595 
596 /*
597  * SIOC[GAD]LIFADDR.
598  *	SIOCGLIFADDR: get first address. (?!?)
599  *	SIOCGLIFADDR with IFLR_PREFIX:
600  *		get first address that matches the specified prefix.
601  *	SIOCALIFADDR: add the specified address.
602  *	SIOCALIFADDR with IFLR_PREFIX:
603  *		EINVAL since we can't deduce hostid part of the address.
604  *	SIOCDLIFADDR: delete the specified address.
605  *	SIOCDLIFADDR with IFLR_PREFIX:
606  *		delete the first address that matches the specified prefix.
607  * return values:
608  *	EINVAL on invalid parameters
609  *	EADDRNOTAVAIL on prefix match failed/specified address not found
610  *	other values may be returned from in_ioctl()
611  */
612 static int
613 in_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
614     struct ifnet *ifp, struct thread *td)
615 {
616 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
617 	struct ifaddr *ifa;
618 
619 	/* sanity checks */
620 	if (data == NULL || ifp == NULL) {
621 		panic("invalid argument to in_lifaddr_ioctl");
622 		/*NOTRECHED*/
623 	}
624 
625 	switch (cmd) {
626 	case SIOCGLIFADDR:
627 		/* address must be specified on GET with IFLR_PREFIX */
628 		if ((iflr->flags & IFLR_PREFIX) == 0)
629 			break;
630 		/*FALLTHROUGH*/
631 	case SIOCALIFADDR:
632 	case SIOCDLIFADDR:
633 		/* address must be specified on ADD and DELETE */
634 		if (iflr->addr.ss_family != AF_INET)
635 			return (EINVAL);
636 		if (iflr->addr.ss_len != sizeof(struct sockaddr_in))
637 			return (EINVAL);
638 		/* XXX need improvement */
639 		if (iflr->dstaddr.ss_family
640 		 && iflr->dstaddr.ss_family != AF_INET)
641 			return (EINVAL);
642 		if (iflr->dstaddr.ss_family
643 		 && iflr->dstaddr.ss_len != sizeof(struct sockaddr_in))
644 			return (EINVAL);
645 		break;
646 	default: /*shouldn't happen*/
647 		return (EOPNOTSUPP);
648 	}
649 	if (sizeof(struct in_addr) * 8 < iflr->prefixlen)
650 		return (EINVAL);
651 
652 	switch (cmd) {
653 	case SIOCALIFADDR:
654 	    {
655 		struct in_aliasreq ifra;
656 
657 		if (iflr->flags & IFLR_PREFIX)
658 			return (EINVAL);
659 
660 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
661 		bzero(&ifra, sizeof(ifra));
662 		bcopy(iflr->iflr_name, ifra.ifra_name,
663 			sizeof(ifra.ifra_name));
664 
665 		bcopy(&iflr->addr, &ifra.ifra_addr, iflr->addr.ss_len);
666 
667 		if (iflr->dstaddr.ss_family) {	/*XXX*/
668 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
669 				iflr->dstaddr.ss_len);
670 		}
671 
672 		ifra.ifra_mask.sin_family = AF_INET;
673 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
674 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
675 
676 		return (in_control(so, SIOCAIFADDR, (caddr_t)&ifra, ifp, td));
677 	    }
678 	case SIOCGLIFADDR:
679 	case SIOCDLIFADDR:
680 	    {
681 		struct in_ifaddr *ia;
682 		struct in_addr mask, candidate, match;
683 		struct sockaddr_in *sin;
684 
685 		bzero(&mask, sizeof(mask));
686 		bzero(&match, sizeof(match));
687 		if (iflr->flags & IFLR_PREFIX) {
688 			/* lookup a prefix rather than address. */
689 			in_len2mask(&mask, iflr->prefixlen);
690 
691 			sin = (struct sockaddr_in *)&iflr->addr;
692 			match.s_addr = sin->sin_addr.s_addr;
693 			match.s_addr &= mask.s_addr;
694 
695 			/* if you set extra bits, that's wrong */
696 			if (match.s_addr != sin->sin_addr.s_addr)
697 				return (EINVAL);
698 
699 		} else {
700 			/* on getting an address, take the 1st match */
701 			/* on deleting an address, do exact match */
702 			if (cmd != SIOCGLIFADDR) {
703 				in_len2mask(&mask, 32);
704 				sin = (struct sockaddr_in *)&iflr->addr;
705 				match.s_addr = sin->sin_addr.s_addr;
706 			}
707 		}
708 
709 		IF_ADDR_RLOCK(ifp);
710 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)	{
711 			if (ifa->ifa_addr->sa_family != AF_INET)
712 				continue;
713 			if (match.s_addr == 0)
714 				break;
715 			sin = (struct sockaddr_in *)&ifa->ifa_addr;
716 			candidate.s_addr = sin->sin_addr.s_addr;
717 			candidate.s_addr &= mask.s_addr;
718 			if (candidate.s_addr == match.s_addr)
719 				break;
720 		}
721 		if (ifa != NULL)
722 			ifa_ref(ifa);
723 		IF_ADDR_RUNLOCK(ifp);
724 		if (ifa == NULL)
725 			return (EADDRNOTAVAIL);
726 		ia = (struct in_ifaddr *)ifa;
727 
728 		if (cmd == SIOCGLIFADDR) {
729 			/* fill in the if_laddrreq structure */
730 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin_len);
731 
732 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
733 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
734 					ia->ia_dstaddr.sin_len);
735 			} else
736 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
737 
738 			iflr->prefixlen =
739 				in_mask2len(&ia->ia_sockmask.sin_addr);
740 
741 			iflr->flags = 0;	/*XXX*/
742 			ifa_free(ifa);
743 
744 			return (0);
745 		} else {
746 			struct in_aliasreq ifra;
747 
748 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
749 			bzero(&ifra, sizeof(ifra));
750 			bcopy(iflr->iflr_name, ifra.ifra_name,
751 				sizeof(ifra.ifra_name));
752 
753 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
754 				ia->ia_addr.sin_len);
755 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
756 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
757 					ia->ia_dstaddr.sin_len);
758 			}
759 			bcopy(&ia->ia_sockmask, &ifra.ifra_dstaddr,
760 				ia->ia_sockmask.sin_len);
761 			ifa_free(ifa);
762 
763 			return (in_control(so, SIOCDIFADDR, (caddr_t)&ifra,
764 			    ifp, td));
765 		}
766 	    }
767 	}
768 
769 	return (EOPNOTSUPP);	/*just for safety*/
770 }
771 
772 /*
773  * Delete any existing route for an interface.
774  */
775 void
776 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia, u_int flags)
777 {
778 
779 	in_scrubprefix(ia, flags);
780 }
781 
782 /*
783  * Initialize an interface's internet address
784  * and routing table entry.
785  */
786 static int
787 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, struct sockaddr_in *sin,
788     int masksupplied, int vhid)
789 {
790 	register u_long i = ntohl(sin->sin_addr.s_addr);
791 	int flags, error = 0;
792 
793 	IN_IFADDR_WLOCK();
794 	if (ia->ia_addr.sin_family == AF_INET)
795 		LIST_REMOVE(ia, ia_hash);
796 	ia->ia_addr = *sin;
797 	LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr),
798 	    ia, ia_hash);
799 	IN_IFADDR_WUNLOCK();
800 
801 	if (vhid > 0) {
802 		if (carp_attach_p != NULL)
803 			error = (*carp_attach_p)(&ia->ia_ifa, vhid);
804 		else
805 			error = EPROTONOSUPPORT;
806 	}
807 	if (error)
808 		return (error);
809 
810 	/*
811 	 * Give the interface a chance to initialize
812 	 * if this is its first address,
813 	 * and to validate the address if necessary.
814 	 */
815 	if (ifp->if_ioctl != NULL &&
816 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia)) != 0)
817 			/* LIST_REMOVE(ia, ia_hash) is done in in_control */
818 			return (error);
819 
820 	/*
821 	 * Be compatible with network classes, if netmask isn't supplied,
822 	 * guess it based on classes.
823 	 */
824 	if (!masksupplied) {
825 		if (IN_CLASSA(i))
826 			ia->ia_subnetmask = IN_CLASSA_NET;
827 		else if (IN_CLASSB(i))
828 			ia->ia_subnetmask = IN_CLASSB_NET;
829 		else
830 			ia->ia_subnetmask = IN_CLASSC_NET;
831 		ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask);
832 	}
833 	ia->ia_subnet = i & ia->ia_subnetmask;
834 	in_socktrim(&ia->ia_sockmask);
835 
836 	/*
837 	 * Add route for the network.
838 	 */
839 	flags = RTF_UP;
840 	ia->ia_ifa.ifa_metric = ifp->if_metric;
841 	if (ifp->if_flags & IFF_BROADCAST) {
842 		if (ia->ia_subnetmask == IN_RFC3021_MASK)
843 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
844 		else
845 			ia->ia_broadaddr.sin_addr.s_addr =
846 			    htonl(ia->ia_subnet | ~ia->ia_subnetmask);
847 	} else if (ifp->if_flags & IFF_LOOPBACK) {
848 		ia->ia_dstaddr = ia->ia_addr;
849 		flags |= RTF_HOST;
850 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
851 		if (ia->ia_dstaddr.sin_family != AF_INET)
852 			return (0);
853 		flags |= RTF_HOST;
854 	}
855 	if (!vhid && (error = in_addprefix(ia, flags)) != 0)
856 		return (error);
857 
858 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)
859 		return (0);
860 
861 	if (ifp->if_flags & IFF_POINTOPOINT &&
862 	    ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
863 			return (0);
864 
865 	/*
866 	 * add a loopback route to self
867 	 */
868 	if (V_useloopback && !vhid && !(ifp->if_flags & IFF_LOOPBACK)) {
869 		struct route ia_ro;
870 
871 		bzero(&ia_ro, sizeof(ia_ro));
872 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = ia->ia_addr;
873 		rtalloc_ign_fib(&ia_ro, 0, RT_DEFAULT_FIB);
874 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
875 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
876 			RT_LOCK(ia_ro.ro_rt);
877 			RT_ADDREF(ia_ro.ro_rt);
878 			RTFREE_LOCKED(ia_ro.ro_rt);
879 		} else
880 			error = ifa_add_loopback_route((struct ifaddr *)ia,
881 			    (struct sockaddr *)&ia->ia_addr);
882 		if (error == 0)
883 			ia->ia_flags |= IFA_RTSELF;
884 		if (ia_ro.ro_rt != NULL)
885 			RTFREE(ia_ro.ro_rt);
886 	}
887 
888 	return (error);
889 }
890 
891 #define rtinitflags(x) \
892 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
893 	    ? RTF_HOST : 0)
894 
895 /*
896  * Generate a routing message when inserting or deleting
897  * an interface address alias.
898  */
899 static void in_addralias_rtmsg(int cmd, struct in_addr *prefix,
900     struct in_ifaddr *target)
901 {
902 	struct route pfx_ro;
903 	struct sockaddr_in *pfx_addr;
904 	struct rtentry msg_rt;
905 
906 	/* QL: XXX
907 	 * This is a bit questionable because there is no
908 	 * additional route entry added/deleted for an address
909 	 * alias. Therefore this route report is inaccurate.
910 	 */
911 	bzero(&pfx_ro, sizeof(pfx_ro));
912 	pfx_addr = (struct sockaddr_in *)(&pfx_ro.ro_dst);
913 	pfx_addr->sin_len = sizeof(*pfx_addr);
914 	pfx_addr->sin_family = AF_INET;
915 	pfx_addr->sin_addr = *prefix;
916 	rtalloc_ign_fib(&pfx_ro, 0, 0);
917 	if (pfx_ro.ro_rt != NULL) {
918 		msg_rt = *pfx_ro.ro_rt;
919 
920 		/* QL: XXX
921 		 * Point the gateway to the new interface
922 		 * address as if a new prefix route entry has
923 		 * been added through the new address alias.
924 		 * All other parts of the rtentry is accurate,
925 		 * e.g., rt_key, rt_mask, rt_ifp etc.
926 		 */
927 		msg_rt.rt_gateway = (struct sockaddr *)&target->ia_addr;
928 		rt_newaddrmsg(cmd, (struct ifaddr *)target, 0, &msg_rt);
929 		RTFREE(pfx_ro.ro_rt);
930 	}
931 	return;
932 }
933 
934 /*
935  * Check if we have a route for the given prefix already or add one accordingly.
936  */
937 int
938 in_addprefix(struct in_ifaddr *target, int flags)
939 {
940 	struct in_ifaddr *ia;
941 	struct in_addr prefix, mask, p, m;
942 	int error;
943 
944 	if ((flags & RTF_HOST) != 0) {
945 		prefix = target->ia_dstaddr.sin_addr;
946 		mask.s_addr = 0;
947 	} else {
948 		prefix = target->ia_addr.sin_addr;
949 		mask = target->ia_sockmask.sin_addr;
950 		prefix.s_addr &= mask.s_addr;
951 	}
952 
953 	IN_IFADDR_RLOCK();
954 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
955 		if (rtinitflags(ia)) {
956 			p = ia->ia_dstaddr.sin_addr;
957 
958 			if (prefix.s_addr != p.s_addr)
959 				continue;
960 		} else {
961 			p = ia->ia_addr.sin_addr;
962 			m = ia->ia_sockmask.sin_addr;
963 			p.s_addr &= m.s_addr;
964 
965 			if (prefix.s_addr != p.s_addr ||
966 			    mask.s_addr != m.s_addr)
967 				continue;
968 		}
969 
970 		/*
971 		 * If we got a matching prefix route inserted by other
972 		 * interface address, we are done here.
973 		 */
974 		if (ia->ia_flags & IFA_ROUTE) {
975 #ifdef RADIX_MPATH
976 			if (ia->ia_addr.sin_addr.s_addr ==
977 			    target->ia_addr.sin_addr.s_addr) {
978 				IN_IFADDR_RUNLOCK();
979 				return (EEXIST);
980 			} else
981 				break;
982 #endif
983 			if (V_nosameprefix) {
984 				IN_IFADDR_RUNLOCK();
985 				return (EEXIST);
986 			} else {
987 				in_addralias_rtmsg(RTM_ADD, &prefix, target);
988 				IN_IFADDR_RUNLOCK();
989 				return (0);
990 			}
991 		}
992 	}
993 	IN_IFADDR_RUNLOCK();
994 
995 	/*
996 	 * No-one seem to have this prefix route, so we try to insert it.
997 	 */
998 	error = rtinit(&target->ia_ifa, (int)RTM_ADD, flags);
999 	if (!error)
1000 		target->ia_flags |= IFA_ROUTE;
1001 	return (error);
1002 }
1003 
1004 /*
1005  * If there is no other address in the system that can serve a route to the
1006  * same prefix, remove the route.  Hand over the route to the new address
1007  * otherwise.
1008  */
1009 int
1010 in_scrubprefix(struct in_ifaddr *target, u_int flags)
1011 {
1012 	struct in_ifaddr *ia;
1013 	struct in_addr prefix, mask, p, m;
1014 	int error = 0;
1015 	struct sockaddr_in prefix0, mask0;
1016 
1017 	/*
1018 	 * Remove the loopback route to the interface address.
1019 	 * The "useloopback" setting is not consulted because if the
1020 	 * user configures an interface address, turns off this
1021 	 * setting, and then tries to delete that interface address,
1022 	 * checking the current setting of "useloopback" would leave
1023 	 * that interface address loopback route untouched, which
1024 	 * would be wrong. Therefore the interface address loopback route
1025 	 * deletion is unconditional.
1026 	 */
1027 	if ((target->ia_addr.sin_addr.s_addr != INADDR_ANY) &&
1028 	    !(target->ia_ifp->if_flags & IFF_LOOPBACK) &&
1029 	    (target->ia_flags & IFA_RTSELF)) {
1030 		struct route ia_ro;
1031 		int freeit = 0;
1032 
1033 		bzero(&ia_ro, sizeof(ia_ro));
1034 		*((struct sockaddr_in *)(&ia_ro.ro_dst)) = target->ia_addr;
1035 		rtalloc_ign_fib(&ia_ro, 0, 0);
1036 		if ((ia_ro.ro_rt != NULL) && (ia_ro.ro_rt->rt_ifp != NULL) &&
1037 		    (ia_ro.ro_rt->rt_ifp == V_loif)) {
1038 			RT_LOCK(ia_ro.ro_rt);
1039 			if (ia_ro.ro_rt->rt_refcnt <= 1)
1040 				freeit = 1;
1041 			else if (flags & LLE_STATIC) {
1042 				RT_REMREF(ia_ro.ro_rt);
1043 				target->ia_flags &= ~IFA_RTSELF;
1044 			}
1045 			RTFREE_LOCKED(ia_ro.ro_rt);
1046 		}
1047 		if (freeit && (flags & LLE_STATIC)) {
1048 			error = ifa_del_loopback_route((struct ifaddr *)target,
1049 			    (struct sockaddr *)&target->ia_addr);
1050 			if (error == 0)
1051 				target->ia_flags &= ~IFA_RTSELF;
1052 		}
1053 		if ((flags & LLE_STATIC) &&
1054 			!(target->ia_ifp->if_flags & IFF_NOARP))
1055 			/* remove arp cache */
1056 			arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr);
1057 	}
1058 
1059 	if (rtinitflags(target)) {
1060 		prefix = target->ia_dstaddr.sin_addr;
1061 		mask.s_addr = 0;
1062 	} else {
1063 		prefix = target->ia_addr.sin_addr;
1064 		mask = target->ia_sockmask.sin_addr;
1065 		prefix.s_addr &= mask.s_addr;
1066 	}
1067 
1068 	if ((target->ia_flags & IFA_ROUTE) == 0) {
1069 		in_addralias_rtmsg(RTM_DELETE, &prefix, target);
1070 		return (0);
1071 	}
1072 
1073 	IN_IFADDR_RLOCK();
1074 	TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1075 		if (rtinitflags(ia)) {
1076 			p = ia->ia_dstaddr.sin_addr;
1077 
1078 			if (prefix.s_addr != p.s_addr)
1079 				continue;
1080 		} else {
1081 			p = ia->ia_addr.sin_addr;
1082 			m = ia->ia_sockmask.sin_addr;
1083 			p.s_addr &= m.s_addr;
1084 
1085 			if (prefix.s_addr != p.s_addr ||
1086 			    mask.s_addr != m.s_addr)
1087 				continue;
1088 		}
1089 
1090 		if ((ia->ia_ifp->if_flags & IFF_UP) == 0)
1091 			continue;
1092 
1093 		/*
1094 		 * If we got a matching prefix address, move IFA_ROUTE and
1095 		 * the route itself to it.  Make sure that routing daemons
1096 		 * get a heads-up.
1097 		 */
1098 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1099 			ifa_ref(&ia->ia_ifa);
1100 			IN_IFADDR_RUNLOCK();
1101 			error = rtinit(&(target->ia_ifa), (int)RTM_DELETE,
1102 			    rtinitflags(target));
1103 			if (error == 0)
1104 				target->ia_flags &= ~IFA_ROUTE;
1105 			else
1106 				log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n",
1107 					error);
1108 			error = rtinit(&ia->ia_ifa, (int)RTM_ADD,
1109 			    rtinitflags(ia) | RTF_UP);
1110 			if (error == 0)
1111 				ia->ia_flags |= IFA_ROUTE;
1112 			else
1113 				log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n",
1114 					error);
1115 			ifa_free(&ia->ia_ifa);
1116 			return (error);
1117 		}
1118 	}
1119 	IN_IFADDR_RUNLOCK();
1120 
1121 	/*
1122 	 * remove all L2 entries on the given prefix
1123 	 */
1124 	bzero(&prefix0, sizeof(prefix0));
1125 	prefix0.sin_len = sizeof(prefix0);
1126 	prefix0.sin_family = AF_INET;
1127 	prefix0.sin_addr.s_addr = target->ia_subnet;
1128 	bzero(&mask0, sizeof(mask0));
1129 	mask0.sin_len = sizeof(mask0);
1130 	mask0.sin_family = AF_INET;
1131 	mask0.sin_addr.s_addr = target->ia_subnetmask;
1132 	lltable_prefix_free(AF_INET, (struct sockaddr *)&prefix0,
1133 	    (struct sockaddr *)&mask0, flags);
1134 
1135 	/*
1136 	 * As no-one seem to have this prefix, we can remove the route.
1137 	 */
1138 	error = rtinit(&(target->ia_ifa), (int)RTM_DELETE, rtinitflags(target));
1139 	if (error == 0)
1140 		target->ia_flags &= ~IFA_ROUTE;
1141 	else
1142 		log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error);
1143 	return (error);
1144 }
1145 
1146 #undef rtinitflags
1147 
1148 /*
1149  * Return 1 if the address might be a local broadcast address.
1150  */
1151 int
1152 in_broadcast(struct in_addr in, struct ifnet *ifp)
1153 {
1154 	register struct ifaddr *ifa;
1155 	u_long t;
1156 
1157 	if (in.s_addr == INADDR_BROADCAST ||
1158 	    in.s_addr == INADDR_ANY)
1159 		return (1);
1160 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1161 		return (0);
1162 	t = ntohl(in.s_addr);
1163 	/*
1164 	 * Look through the list of addresses for a match
1165 	 * with a broadcast address.
1166 	 */
1167 #define ia ((struct in_ifaddr *)ifa)
1168 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1169 		if (ifa->ifa_addr->sa_family == AF_INET &&
1170 		    (in.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
1171 		     /*
1172 		      * Check for old-style (host 0) broadcast, but
1173 		      * taking into account that RFC 3021 obsoletes it.
1174 		      */
1175 		    (ia->ia_subnetmask != IN_RFC3021_MASK &&
1176 		    t == ia->ia_subnet)) &&
1177 		     /*
1178 		      * Check for an all one subnetmask. These
1179 		      * only exist when an interface gets a secondary
1180 		      * address.
1181 		      */
1182 		    ia->ia_subnetmask != (u_long)0xffffffff)
1183 			    return (1);
1184 	return (0);
1185 #undef ia
1186 }
1187 
1188 /*
1189  * On interface removal, clean up IPv4 data structures hung off of the ifnet.
1190  */
1191 void
1192 in_ifdetach(struct ifnet *ifp)
1193 {
1194 
1195 	in_pcbpurgeif0(&V_ripcbinfo, ifp);
1196 	in_pcbpurgeif0(&V_udbinfo, ifp);
1197 	in_purgemaddrs(ifp);
1198 }
1199 
1200 /*
1201  * Delete all IPv4 multicast address records, and associated link-layer
1202  * multicast address records, associated with ifp.
1203  * XXX It looks like domifdetach runs AFTER the link layer cleanup.
1204  * XXX This should not race with ifma_protospec being set during
1205  * a new allocation, if it does, we have bigger problems.
1206  */
1207 static void
1208 in_purgemaddrs(struct ifnet *ifp)
1209 {
1210 	LIST_HEAD(,in_multi) purgeinms;
1211 	struct in_multi		*inm, *tinm;
1212 	struct ifmultiaddr	*ifma;
1213 
1214 	LIST_INIT(&purgeinms);
1215 	IN_MULTI_LOCK();
1216 
1217 	/*
1218 	 * Extract list of in_multi associated with the detaching ifp
1219 	 * which the PF_INET layer is about to release.
1220 	 * We need to do this as IF_ADDR_LOCK() may be re-acquired
1221 	 * by code further down.
1222 	 */
1223 	IF_ADDR_RLOCK(ifp);
1224 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1225 		if (ifma->ifma_addr->sa_family != AF_INET ||
1226 		    ifma->ifma_protospec == NULL)
1227 			continue;
1228 #if 0
1229 		KASSERT(ifma->ifma_protospec != NULL,
1230 		    ("%s: ifma_protospec is NULL", __func__));
1231 #endif
1232 		inm = (struct in_multi *)ifma->ifma_protospec;
1233 		LIST_INSERT_HEAD(&purgeinms, inm, inm_link);
1234 	}
1235 	IF_ADDR_RUNLOCK(ifp);
1236 
1237 	LIST_FOREACH_SAFE(inm, &purgeinms, inm_link, tinm) {
1238 		LIST_REMOVE(inm, inm_link);
1239 		inm_release_locked(inm);
1240 	}
1241 	igmp_ifdetach(ifp);
1242 
1243 	IN_MULTI_UNLOCK();
1244 }
1245 
1246 struct in_llentry {
1247 	struct llentry		base;
1248 	struct sockaddr_in	l3_addr4;
1249 };
1250 
1251 /*
1252  * Deletes an address from the address table.
1253  * This function is called by the timer functions
1254  * such as arptimer() and nd6_llinfo_timer(), and
1255  * the caller does the locking.
1256  */
1257 static void
1258 in_lltable_free(struct lltable *llt, struct llentry *lle)
1259 {
1260 	LLE_WUNLOCK(lle);
1261 	LLE_LOCK_DESTROY(lle);
1262 	free(lle, M_LLTABLE);
1263 }
1264 
1265 static struct llentry *
1266 in_lltable_new(const struct sockaddr *l3addr, u_int flags)
1267 {
1268 	struct in_llentry *lle;
1269 
1270 	lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
1271 	if (lle == NULL)		/* NB: caller generates msg */
1272 		return NULL;
1273 
1274 	/*
1275 	 * For IPv4 this will trigger "arpresolve" to generate
1276 	 * an ARP request.
1277 	 */
1278 	lle->base.la_expire = time_uptime; /* mark expired */
1279 	lle->l3_addr4 = *(const struct sockaddr_in *)l3addr;
1280 	lle->base.lle_refcnt = 1;
1281 	lle->base.lle_free = in_lltable_free;
1282 	LLE_LOCK_INIT(&lle->base);
1283 	callout_init_rw(&lle->base.la_timer, &lle->base.lle_lock,
1284 	    CALLOUT_RETURNUNLOCKED);
1285 
1286 	return (&lle->base);
1287 }
1288 
1289 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1290 	    (((ntohl((d)->sin_addr.s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1291 
1292 static void
1293 in_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
1294     const struct sockaddr *mask, u_int flags)
1295 {
1296 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1297 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1298 	struct llentry *lle, *next;
1299 	int i;
1300 	size_t pkts_dropped;
1301 
1302 	IF_AFDATA_WLOCK(llt->llt_ifp);
1303 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1304 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
1305 			/*
1306 			 * (flags & LLE_STATIC) means deleting all entries
1307 			 * including static ARP entries.
1308 			 */
1309 			if (IN_ARE_MASKED_ADDR_EQUAL(satosin(L3_ADDR(lle)),
1310 			    pfx, msk) && ((flags & LLE_STATIC) ||
1311 			    !(lle->la_flags & LLE_STATIC))) {
1312 				LLE_WLOCK(lle);
1313 				if (callout_stop(&lle->la_timer))
1314 					LLE_REMREF(lle);
1315 				pkts_dropped = llentry_free(lle);
1316 				ARPSTAT_ADD(dropped, pkts_dropped);
1317 			}
1318 		}
1319 	}
1320 	IF_AFDATA_WUNLOCK(llt->llt_ifp);
1321 }
1322 
1323 
1324 static int
1325 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1326 {
1327 	struct rtentry *rt;
1328 
1329 	KASSERT(l3addr->sa_family == AF_INET,
1330 	    ("sin_family %d", l3addr->sa_family));
1331 
1332 	/* XXX rtalloc1 should take a const param */
1333 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
1334 
1335 	if (rt == NULL)
1336 		return (EINVAL);
1337 
1338 	/*
1339 	 * If the gateway for an existing host route matches the target L3
1340 	 * address, which is a special route inserted by some implementation
1341 	 * such as MANET, and the interface is of the correct type, then
1342 	 * allow for ARP to proceed.
1343 	 */
1344 	if (rt->rt_flags & RTF_GATEWAY) {
1345 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1346 		    rt->rt_ifp->if_type != IFT_ETHER ||
1347 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1348 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1349 		    sizeof(in_addr_t)) != 0) {
1350 			RTFREE_LOCKED(rt);
1351 			return (EINVAL);
1352 		}
1353 	}
1354 
1355 	/*
1356 	 * Make sure that at least the destination address is covered
1357 	 * by the route. This is for handling the case where 2 or more
1358 	 * interfaces have the same prefix. An incoming packet arrives
1359 	 * on one interface and the corresponding outgoing packet leaves
1360 	 * another interface.
1361 	 */
1362 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1363 		const char *sa, *mask, *addr, *lim;
1364 		int len;
1365 
1366 		mask = (const char *)rt_mask(rt);
1367 		/*
1368 		 * Just being extra cautious to avoid some custom
1369 		 * code getting into trouble.
1370 		 */
1371 		if (mask == NULL) {
1372 			RTFREE_LOCKED(rt);
1373 			return (EINVAL);
1374 		}
1375 
1376 		sa = (const char *)rt_key(rt);
1377 		addr = (const char *)l3addr;
1378 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
1379 		lim = addr + len;
1380 
1381 		for ( ; addr < lim; sa++, mask++, addr++) {
1382 			if ((*sa ^ *addr) & *mask) {
1383 #ifdef DIAGNOSTIC
1384 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1385 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1386 #endif
1387 				RTFREE_LOCKED(rt);
1388 				return (EINVAL);
1389 			}
1390 		}
1391 	}
1392 
1393 	RTFREE_LOCKED(rt);
1394 	return (0);
1395 }
1396 
1397 /*
1398  * Return NULL if not found or marked for deletion.
1399  * If found return lle read locked.
1400  */
1401 static struct llentry *
1402 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1403 {
1404 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1405 	struct ifnet *ifp = llt->llt_ifp;
1406 	struct llentry *lle;
1407 	struct llentries *lleh;
1408 	u_int hashkey;
1409 
1410 	IF_AFDATA_LOCK_ASSERT(ifp);
1411 	KASSERT(l3addr->sa_family == AF_INET,
1412 	    ("sin_family %d", l3addr->sa_family));
1413 
1414 	hashkey = sin->sin_addr.s_addr;
1415 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
1416 	LIST_FOREACH(lle, lleh, lle_next) {
1417 		struct sockaddr_in *sa2 = satosin(L3_ADDR(lle));
1418 		if (lle->la_flags & LLE_DELETED)
1419 			continue;
1420 		if (sa2->sin_addr.s_addr == sin->sin_addr.s_addr)
1421 			break;
1422 	}
1423 	if (lle == NULL) {
1424 #ifdef DIAGNOSTIC
1425 		if (flags & LLE_DELETE)
1426 			log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1427 #endif
1428 		if (!(flags & LLE_CREATE))
1429 			return (NULL);
1430 		/*
1431 		 * A route that covers the given address must have
1432 		 * been installed 1st because we are doing a resolution,
1433 		 * verify this.
1434 		 */
1435 		if (!(flags & LLE_IFADDR) &&
1436 		    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
1437 			goto done;
1438 
1439 		lle = in_lltable_new(l3addr, flags);
1440 		if (lle == NULL) {
1441 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
1442 			goto done;
1443 		}
1444 		lle->la_flags = flags & ~LLE_CREATE;
1445 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
1446 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
1447 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
1448 		}
1449 
1450 		lle->lle_tbl  = llt;
1451 		lle->lle_head = lleh;
1452 		lle->la_flags |= LLE_LINKED;
1453 		LIST_INSERT_HEAD(lleh, lle, lle_next);
1454 	} else if (flags & LLE_DELETE) {
1455 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
1456 			LLE_WLOCK(lle);
1457 			lle->la_flags |= LLE_DELETED;
1458 			EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
1459 #ifdef DIAGNOSTIC
1460 			log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1461 #endif
1462 			if ((lle->la_flags &
1463 			    (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
1464 				llentry_free(lle);
1465 			else
1466 				LLE_WUNLOCK(lle);
1467 		}
1468 		lle = (void *)-1;
1469 
1470 	}
1471 	if (LLE_IS_VALID(lle)) {
1472 		if (flags & LLE_EXCLUSIVE)
1473 			LLE_WLOCK(lle);
1474 		else
1475 			LLE_RLOCK(lle);
1476 	}
1477 done:
1478 	return (lle);
1479 }
1480 
1481 static int
1482 in_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
1483 {
1484 #define	SIN(lle)	((struct sockaddr_in *) L3_ADDR(lle))
1485 	struct ifnet *ifp = llt->llt_ifp;
1486 	struct llentry *lle;
1487 	/* XXX stack use */
1488 	struct {
1489 		struct rt_msghdr	rtm;
1490 		struct sockaddr_in	sin;
1491 		struct sockaddr_dl	sdl;
1492 	} arpc;
1493 	int error, i;
1494 
1495 	LLTABLE_LOCK_ASSERT();
1496 
1497 	error = 0;
1498 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
1499 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
1500 			struct sockaddr_dl *sdl;
1501 
1502 			/* skip deleted entries */
1503 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
1504 				continue;
1505 			/* Skip if jailed and not a valid IP of the prison. */
1506 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
1507 				continue;
1508 			/*
1509 			 * produce a msg made of:
1510 			 *  struct rt_msghdr;
1511 			 *  struct sockaddr_in; (IPv4)
1512 			 *  struct sockaddr_dl;
1513 			 */
1514 			bzero(&arpc, sizeof(arpc));
1515 			arpc.rtm.rtm_msglen = sizeof(arpc);
1516 			arpc.rtm.rtm_version = RTM_VERSION;
1517 			arpc.rtm.rtm_type = RTM_GET;
1518 			arpc.rtm.rtm_flags = RTF_UP;
1519 			arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
1520 			arpc.sin.sin_family = AF_INET;
1521 			arpc.sin.sin_len = sizeof(arpc.sin);
1522 			arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr;
1523 
1524 			/* publish */
1525 			if (lle->la_flags & LLE_PUB)
1526 				arpc.rtm.rtm_flags |= RTF_ANNOUNCE;
1527 
1528 			sdl = &arpc.sdl;
1529 			sdl->sdl_family = AF_LINK;
1530 			sdl->sdl_len = sizeof(*sdl);
1531 			sdl->sdl_index = ifp->if_index;
1532 			sdl->sdl_type = ifp->if_type;
1533 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
1534 				sdl->sdl_alen = ifp->if_addrlen;
1535 				bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
1536 			} else {
1537 				sdl->sdl_alen = 0;
1538 				bzero(LLADDR(sdl), ifp->if_addrlen);
1539 			}
1540 
1541 			arpc.rtm.rtm_rmx.rmx_expire =
1542 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
1543 			arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
1544 			if (lle->la_flags & LLE_STATIC)
1545 				arpc.rtm.rtm_flags |= RTF_STATIC;
1546 			arpc.rtm.rtm_index = ifp->if_index;
1547 			error = SYSCTL_OUT(wr, &arpc, sizeof(arpc));
1548 			if (error)
1549 				break;
1550 		}
1551 	}
1552 	return error;
1553 #undef SIN
1554 }
1555 
1556 void *
1557 in_domifattach(struct ifnet *ifp)
1558 {
1559 	struct in_ifinfo *ii;
1560 	struct lltable *llt;
1561 
1562 	ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO);
1563 
1564 	llt = lltable_init(ifp, AF_INET);
1565 	if (llt != NULL) {
1566 		llt->llt_prefix_free = in_lltable_prefix_free;
1567 		llt->llt_lookup = in_lltable_lookup;
1568 		llt->llt_dump = in_lltable_dump;
1569 	}
1570 	ii->ii_llt = llt;
1571 
1572 	ii->ii_igmp = igmp_domifattach(ifp);
1573 
1574 	return ii;
1575 }
1576 
1577 void
1578 in_domifdetach(struct ifnet *ifp, void *aux)
1579 {
1580 	struct in_ifinfo *ii = (struct in_ifinfo *)aux;
1581 
1582 	igmp_domifdetach(ifp);
1583 	lltable_free(ii->ii_llt);
1584 	free(ii, M_IFADDR);
1585 }
1586