xref: /netbsd-src/sys/netinet/in.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: in.c,v 1.231 2018/05/13 22:42:51 khorben Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1991, 1993
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)in.c	8.4 (Berkeley) 1/9/95
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: in.c,v 1.231 2018/05/13 22:42:51 khorben Exp $");
95 
96 #include "arp.h"
97 
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_inet_conf.h"
101 #include "opt_mrouting.h"
102 #include "opt_net_mpsafe.h"
103 #endif
104 
105 #include <sys/param.h>
106 #include <sys/ioctl.h>
107 #include <sys/errno.h>
108 #include <sys/kernel.h>
109 #include <sys/malloc.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/sysctl.h>
113 #include <sys/systm.h>
114 #include <sys/proc.h>
115 #include <sys/syslog.h>
116 #include <sys/kauth.h>
117 #include <sys/kmem.h>
118 
119 #include <sys/cprng.h>
120 
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <net/if_arp.h>
126 #include <net/if_ether.h>
127 #include <net/if_types.h>
128 #include <net/if_llatbl.h>
129 #include <net/if_dl.h>
130 
131 #include <netinet/in_systm.h>
132 #include <netinet/in.h>
133 #include <netinet/in_var.h>
134 #include <netinet/ip.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/in_ifattach.h>
137 #include <netinet/in_pcb.h>
138 #include <netinet/in_selsrc.h>
139 #include <netinet/if_inarp.h>
140 #include <netinet/ip_mroute.h>
141 #include <netinet/igmp_var.h>
142 
143 #ifdef IPSELSRC
144 #include <netinet/in_selsrc.h>
145 #endif
146 
147 static u_int	in_mask2len(struct in_addr *);
148 static void	in_len2mask(struct in_addr *, u_int);
149 static int	in_lifaddr_ioctl(struct socket *, u_long, void *,
150 	struct ifnet *);
151 
152 static void	in_addrhash_insert_locked(struct in_ifaddr *);
153 static void	in_addrhash_remove_locked(struct in_ifaddr *);
154 
155 static int	in_addprefix(struct in_ifaddr *, int);
156 static void	in_scrubaddr(struct in_ifaddr *);
157 static int	in_scrubprefix(struct in_ifaddr *);
158 static void	in_sysctl_init(struct sysctllog **);
159 
160 #ifndef SUBNETSARELOCAL
161 #define	SUBNETSARELOCAL	1
162 #endif
163 
164 #ifndef HOSTZEROBROADCAST
165 #define HOSTZEROBROADCAST 0
166 #endif
167 
168 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
169 #ifndef IN_MULTI_HASH_SIZE
170 #define IN_MULTI_HASH_SIZE	509
171 #endif
172 
173 static int			subnetsarelocal = SUBNETSARELOCAL;
174 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
175 
176 /*
177  * This list is used to keep track of in_multi chains which belong to
178  * deleted interface addresses.  We use in_ifaddr so that a chain head
179  * won't be deallocated until all multicast address record are deleted.
180  */
181 
182 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
183 
184 static struct pool		inmulti_pool;
185 static u_int			in_multientries;
186 static struct in_multihashhead *in_multihashtbl;
187 static u_long			in_multihash;
188 static krwlock_t		in_multilock;
189 
190 #define IN_MULTI_HASH(x, ifp) \
191     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
192 
193 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */
194 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
195 u_long				in_ifaddrhash;
196 struct in_ifaddrhead		in_ifaddrhead;
197 static kmutex_t			in_ifaddr_lock;
198 
199 pserialize_t			in_ifaddrhash_psz;
200 struct pslist_head *		in_ifaddrhashtbl_pslist;
201 u_long				in_ifaddrhash_pslist;
202 struct pslist_head		in_ifaddrhead_pslist;
203 
204 void
205 in_init(void)
206 {
207 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
208 	    NULL, IPL_SOFTNET);
209 	TAILQ_INIT(&in_ifaddrhead);
210 	PSLIST_INIT(&in_ifaddrhead_pslist);
211 
212 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
213 	    &in_ifaddrhash);
214 
215 	in_ifaddrhash_psz = pserialize_create();
216 	in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST,
217 	    true, &in_ifaddrhash_pslist);
218 	mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
219 
220 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
221 	    &in_multihash);
222 	rw_init(&in_multilock);
223 
224 	in_sysctl_init(NULL);
225 }
226 
227 /*
228  * Return 1 if an internet address is for a ``local'' host
229  * (one to which we have a connection).  If subnetsarelocal
230  * is true, this includes other subnets of the local net.
231  * Otherwise, it includes only the directly-connected (sub)nets.
232  */
233 int
234 in_localaddr(struct in_addr in)
235 {
236 	struct in_ifaddr *ia;
237 	int localaddr = 0;
238 	int s = pserialize_read_enter();
239 
240 	if (subnetsarelocal) {
241 		IN_ADDRLIST_READER_FOREACH(ia) {
242 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net) {
243 				localaddr = 1;
244 				break;
245 			}
246 		}
247 	} else {
248 		IN_ADDRLIST_READER_FOREACH(ia) {
249 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
250 				localaddr = 1;
251 				break;
252 			}
253 		}
254 	}
255 	pserialize_read_exit(s);
256 
257 	return localaddr;
258 }
259 
260 /*
261  * like in_localaddr() but can specify ifp.
262  */
263 int
264 in_direct(struct in_addr in, struct ifnet *ifp)
265 {
266 	struct ifaddr *ifa;
267 	int localaddr = 0;
268 	int s;
269 
270 	KASSERT(ifp != NULL);
271 
272 #define ia (ifatoia(ifa))
273 	s = pserialize_read_enter();
274 	if (subnetsarelocal) {
275 		IFADDR_READER_FOREACH(ifa, ifp) {
276 			if (ifa->ifa_addr->sa_family == AF_INET &&
277 			    ((in.s_addr & ia->ia_netmask) == ia->ia_net)) {
278 				localaddr = 1;
279 				break;
280 			}
281 		}
282 	} else {
283 		IFADDR_READER_FOREACH(ifa, ifp) {
284 			if (ifa->ifa_addr->sa_family == AF_INET &&
285 			    (in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
286 				localaddr = 1;
287 				break;
288 			}
289 		}
290 	}
291 	pserialize_read_exit(s);
292 
293 	return localaddr;
294 #undef ia
295 }
296 
297 /*
298  * Determine whether an IP address is in a reserved set of addresses
299  * that may not be forwarded, or whether datagrams to that destination
300  * may be forwarded.
301  */
302 int
303 in_canforward(struct in_addr in)
304 {
305 	u_int32_t net;
306 
307 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
308 		return (0);
309 	if (IN_CLASSA(in.s_addr)) {
310 		net = in.s_addr & IN_CLASSA_NET;
311 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
312 			return (0);
313 	}
314 	return (1);
315 }
316 
317 /*
318  * Trim a mask in a sockaddr
319  */
320 void
321 in_socktrim(struct sockaddr_in *ap)
322 {
323 	char *cplim = (char *) &ap->sin_addr;
324 	char *cp = (char *) (&ap->sin_addr + 1);
325 
326 	ap->sin_len = 0;
327 	while (--cp >= cplim)
328 		if (*cp) {
329 			(ap)->sin_len = cp - (char *) (ap) + 1;
330 			break;
331 		}
332 }
333 
334 /*
335  * Maintain the "in_maxmtu" variable, which is the largest
336  * mtu for non-local interfaces with AF_INET addresses assigned
337  * to them that are up.
338  */
339 unsigned long in_maxmtu;
340 
341 void
342 in_setmaxmtu(void)
343 {
344 	struct in_ifaddr *ia;
345 	struct ifnet *ifp;
346 	unsigned long maxmtu = 0;
347 	int s = pserialize_read_enter();
348 
349 	IN_ADDRLIST_READER_FOREACH(ia) {
350 		if ((ifp = ia->ia_ifp) == 0)
351 			continue;
352 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
353 			continue;
354 		if (ifp->if_mtu > maxmtu)
355 			maxmtu = ifp->if_mtu;
356 	}
357 	if (maxmtu)
358 		in_maxmtu = maxmtu;
359 	pserialize_read_exit(s);
360 }
361 
362 static u_int
363 in_mask2len(struct in_addr *mask)
364 {
365 	u_int x, y;
366 	u_char *p;
367 
368 	p = (u_char *)mask;
369 	for (x = 0; x < sizeof(*mask); x++) {
370 		if (p[x] != 0xff)
371 			break;
372 	}
373 	y = 0;
374 	if (x < sizeof(*mask)) {
375 		for (y = 0; y < NBBY; y++) {
376 			if ((p[x] & (0x80 >> y)) == 0)
377 				break;
378 		}
379 	}
380 	return x * NBBY + y;
381 }
382 
383 static void
384 in_len2mask(struct in_addr *mask, u_int len)
385 {
386 	u_int i;
387 	u_char *p;
388 
389 	p = (u_char *)mask;
390 	memset(mask, 0, sizeof(*mask));
391 	for (i = 0; i < len / NBBY; i++)
392 		p[i] = 0xff;
393 	if (len % NBBY)
394 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
395 }
396 
397 /*
398  * Generic internet control operations (ioctl's).
399  * Ifp is 0 if not an interface-specific ioctl.
400  */
401 /* ARGSUSED */
402 static int
403 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
404 {
405 	struct ifreq *ifr = (struct ifreq *)data;
406 	struct in_ifaddr *ia = NULL;
407 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
408 	struct sockaddr_in oldaddr, *new_dstaddr;
409 	int error, hostIsNew, maskIsNew;
410 	int newifaddr = 0;
411 	bool run_hook = false;
412 	bool need_reinsert = false;
413 	struct psref psref;
414 	int bound;
415 
416 	switch (cmd) {
417 	case SIOCALIFADDR:
418 	case SIOCDLIFADDR:
419 	case SIOCGLIFADDR:
420 		if (ifp == NULL)
421 			return EINVAL;
422 		return in_lifaddr_ioctl(so, cmd, data, ifp);
423 	case SIOCGIFADDRPREF:
424 	case SIOCSIFADDRPREF:
425 		if (ifp == NULL)
426 			return EINVAL;
427 		return ifaddrpref_ioctl(so, cmd, data, ifp);
428 	}
429 
430 	bound = curlwp_bind();
431 	/*
432 	 * Find address for this interface, if it exists.
433 	 */
434 	if (ifp != NULL)
435 		ia = in_get_ia_from_ifp_psref(ifp, &psref);
436 
437 	hostIsNew = 1;		/* moved here to appease gcc */
438 	switch (cmd) {
439 	case SIOCAIFADDR:
440 	case SIOCDIFADDR:
441 	case SIOCGIFALIAS:
442 	case SIOCGIFAFLAG_IN:
443 		if (ifra->ifra_addr.sin_family == AF_INET) {
444 			int s;
445 
446 			if (ia != NULL)
447 				ia4_release(ia, &psref);
448 			s = pserialize_read_enter();
449 			IN_ADDRHASH_READER_FOREACH(ia,
450 			    ifra->ifra_addr.sin_addr.s_addr) {
451 				if (ia->ia_ifp == ifp &&
452 				    in_hosteq(ia->ia_addr.sin_addr,
453 				    ifra->ifra_addr.sin_addr))
454 					break;
455 			}
456 			if (ia != NULL)
457 				ia4_acquire(ia, &psref);
458 			pserialize_read_exit(s);
459 		}
460 		if ((cmd == SIOCDIFADDR ||
461 		    cmd == SIOCGIFALIAS ||
462 		    cmd == SIOCGIFAFLAG_IN) &&
463 		    ia == NULL) {
464 			error = EADDRNOTAVAIL;
465 			goto out;
466 		}
467 
468 		if (cmd == SIOCDIFADDR &&
469 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
470 			ifra->ifra_addr.sin_family = AF_INET;
471 		}
472 		/* FALLTHROUGH */
473 	case SIOCSIFADDR:
474 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
475 			;
476 		else if (ifra->ifra_addr.sin_len == 0) {
477 			ifra->ifra_addr = ia->ia_addr;
478 			hostIsNew = 0;
479 		} else if (in_hosteq(ia->ia_addr.sin_addr,
480 		           ifra->ifra_addr.sin_addr))
481 			hostIsNew = 0;
482 		if (ifra->ifra_addr.sin_family != AF_INET) {
483 			error = EAFNOSUPPORT;
484 			goto out;
485 		}
486 		/* FALLTHROUGH */
487 	case SIOCSIFDSTADDR:
488 		if (cmd == SIOCSIFDSTADDR &&
489 		    ifreq_getaddr(cmd, ifr)->sa_family != AF_INET) {
490 			error = EAFNOSUPPORT;
491 			goto out;
492 		}
493 		/* FALLTHROUGH */
494 	case SIOCSIFNETMASK:
495 		if (ifp == NULL)
496 			panic("in_control");
497 
498 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
499 			break;
500 
501 		if (ia == NULL &&
502 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
503 			error = EADDRNOTAVAIL;
504 			goto out;
505 		}
506 
507 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
508 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
509 		    NULL) != 0) {
510 			error = EPERM;
511 			goto out;
512 		}
513 
514 		if (ia == NULL) {
515 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
516 			if (ia == NULL) {
517 				error = ENOBUFS;
518 				goto out;
519 			}
520 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
521 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
522 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
523 #ifdef IPSELSRC
524 			ia->ia_ifa.ifa_getifa = in_getifa;
525 #else /* IPSELSRC */
526 			ia->ia_ifa.ifa_getifa = NULL;
527 #endif /* IPSELSRC */
528 			ia->ia_sockmask.sin_len = 8;
529 			ia->ia_sockmask.sin_family = AF_INET;
530 			if (ifp->if_flags & IFF_BROADCAST) {
531 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
532 				ia->ia_broadaddr.sin_family = AF_INET;
533 			}
534 			ia->ia_ifp = ifp;
535 			ia->ia_idsalt = cprng_fast32() % 65535;
536 			LIST_INIT(&ia->ia_multiaddrs);
537 			IN_ADDRHASH_ENTRY_INIT(ia);
538 			IN_ADDRLIST_ENTRY_INIT(ia);
539 			ifa_psref_init(&ia->ia_ifa);
540 			/*
541 			 * We need a reference to make ia survive over in_ifinit
542 			 * that does ifaref and ifafree.
543 			 */
544 			ifaref(&ia->ia_ifa);
545 
546 			newifaddr = 1;
547 		}
548 		break;
549 
550 	case SIOCSIFBRDADDR:
551 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
552 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
553 		    NULL) != 0) {
554 			error = EPERM;
555 			goto out;
556 		}
557 		/* FALLTHROUGH */
558 
559 	case SIOCGIFADDR:
560 	case SIOCGIFNETMASK:
561 	case SIOCGIFDSTADDR:
562 	case SIOCGIFBRDADDR:
563 		if (ia == NULL) {
564 			error = EADDRNOTAVAIL;
565 			goto out;
566 		}
567 		break;
568 	}
569 	error = 0;
570 	switch (cmd) {
571 
572 	case SIOCGIFADDR:
573 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
574 		break;
575 
576 	case SIOCGIFBRDADDR:
577 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
578 			error = EINVAL;
579 			goto out;
580 		}
581 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
582 		break;
583 
584 	case SIOCGIFDSTADDR:
585 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
586 			error = EINVAL;
587 			goto out;
588 		}
589 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
590 		break;
591 
592 	case SIOCGIFNETMASK:
593 		/*
594 		 * We keep the number of trailing zero bytes the sin_len field
595 		 * of ia_sockmask, so we fix this before we pass it back to
596 		 * userland.
597 		 */
598 		oldaddr = ia->ia_sockmask;
599 		oldaddr.sin_len = sizeof(struct sockaddr_in);
600 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
601 		break;
602 
603 	case SIOCSIFDSTADDR:
604 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
605 			error = EINVAL;
606 			goto out;
607 		}
608 		oldaddr = ia->ia_dstaddr;
609 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
610 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
611 			ia->ia_dstaddr = oldaddr;
612 			goto out;
613 		}
614 		if (ia->ia_flags & IFA_ROUTE) {
615 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
616 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
617 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
618 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
619 		}
620 		break;
621 
622 	case SIOCSIFBRDADDR:
623 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
624 			error = EINVAL;
625 			goto out;
626 		}
627 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
628 		break;
629 
630 	case SIOCSIFADDR:
631 		if (!newifaddr) {
632 			in_addrhash_remove(ia);
633 			need_reinsert = true;
634 		}
635 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
636 		    NULL, 1);
637 
638 		run_hook = true;
639 		break;
640 
641 	case SIOCSIFNETMASK:
642 		in_scrubprefix(ia);
643 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
644 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
645 		if (!newifaddr) {
646 			in_addrhash_remove(ia);
647 			need_reinsert = true;
648 		}
649 		error = in_ifinit(ifp, ia, NULL, NULL, 0);
650 		break;
651 
652 	case SIOCAIFADDR:
653 		maskIsNew = 0;
654 		if (ifra->ifra_mask.sin_len) {
655 			in_scrubprefix(ia);
656 			ia->ia_sockmask = ifra->ifra_mask;
657 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
658 			maskIsNew = 1;
659 		}
660 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
661 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
662 			new_dstaddr = &ifra->ifra_dstaddr;
663 			maskIsNew  = 1; /* We lie; but the effect's the same */
664 		} else
665 			new_dstaddr = NULL;
666 		if (ifra->ifra_addr.sin_family == AF_INET &&
667 		    (hostIsNew || maskIsNew)) {
668 			if (!newifaddr) {
669 				in_addrhash_remove(ia);
670 				need_reinsert = true;
671 			}
672 			error = in_ifinit(ifp, ia, &ifra->ifra_addr,
673 			    new_dstaddr, 0);
674 		}
675 		if ((ifp->if_flags & IFF_BROADCAST) &&
676 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
677 			ia->ia_broadaddr = ifra->ifra_broadaddr;
678 		run_hook = true;
679 		break;
680 
681 	case SIOCGIFALIAS:
682 		ifra->ifra_mask = ia->ia_sockmask;
683 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
684 		    (ia->ia_dstaddr.sin_family == AF_INET))
685 			ifra->ifra_dstaddr = ia->ia_dstaddr;
686 		else if ((ifp->if_flags & IFF_BROADCAST) &&
687 		    (ia->ia_broadaddr.sin_family == AF_INET))
688 			ifra->ifra_broadaddr = ia->ia_broadaddr;
689 		else
690 			memset(&ifra->ifra_broadaddr, 0,
691 			      sizeof(ifra->ifra_broadaddr));
692 		break;
693 
694 	case SIOCGIFAFLAG_IN:
695 		ifr->ifr_addrflags = ia->ia4_flags;
696 		break;
697 
698 	case SIOCDIFADDR:
699 		ia4_release(ia, &psref);
700 		ifaref(&ia->ia_ifa);
701 		in_purgeaddr(&ia->ia_ifa);
702 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
703 		ifafree(&ia->ia_ifa);
704 		ia = NULL;
705 		break;
706 
707 #ifdef MROUTING
708 	case SIOCGETVIFCNT:
709 	case SIOCGETSGCNT:
710 		error = mrt_ioctl(so, cmd, data);
711 		break;
712 #endif /* MROUTING */
713 
714 	default:
715 		error = ENOTTY;
716 		goto out;
717 	}
718 
719 	/*
720 	 * XXX insert regardless of error to make in_purgeaddr below work.
721 	 * Need to improve.
722 	 */
723 	if (newifaddr) {
724 		ifaref(&ia->ia_ifa);
725 		ifa_insert(ifp, &ia->ia_ifa);
726 
727 		mutex_enter(&in_ifaddr_lock);
728 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
729 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
730 		in_addrhash_insert_locked(ia);
731 		/* Release a reference that is held just after creation. */
732 		ifafree(&ia->ia_ifa);
733 		mutex_exit(&in_ifaddr_lock);
734 	} else if (need_reinsert) {
735 		in_addrhash_insert(ia);
736 	}
737 
738 	if (error == 0) {
739 		if (run_hook)
740 			pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
741 	} else if (newifaddr) {
742 		KASSERT(ia != NULL);
743 		in_purgeaddr(&ia->ia_ifa);
744 		ia = NULL;
745 	}
746 
747 out:
748 	if (!newifaddr && ia != NULL)
749 		ia4_release(ia, &psref);
750 	curlwp_bindx(bound);
751 	return error;
752 }
753 
754 int
755 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
756 {
757 	int error;
758 
759 #ifndef NET_MPSAFE
760 	KASSERT(KERNEL_LOCKED_P());
761 #endif
762 	error = in_control0(so, cmd, data, ifp);
763 
764 	return error;
765 }
766 
767 /* Add ownaddr as loopback rtentry. */
768 static void
769 in_ifaddlocal(struct ifaddr *ifa)
770 {
771 	struct in_ifaddr *ia;
772 
773 	ia = (struct in_ifaddr *)ifa;
774 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
775 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
776 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
777 	{
778 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
779 		return;
780 	}
781 
782 	rt_ifa_addlocal(ifa);
783 }
784 
785 /* Remove loopback entry of ownaddr */
786 static void
787 in_ifremlocal(struct ifaddr *ifa)
788 {
789 	struct in_ifaddr *ia, *p;
790 	struct ifaddr *alt_ifa = NULL;
791 	int ia_count = 0;
792 	int s;
793 	struct psref psref;
794 	int bound = curlwp_bind();
795 
796 	ia = (struct in_ifaddr *)ifa;
797 	/* Delete the entry if exactly one ifaddr matches the
798 	 * address, ifa->ifa_addr. */
799 	s = pserialize_read_enter();
800 	IN_ADDRLIST_READER_FOREACH(p) {
801 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
802 			continue;
803 		if (p->ia_ifp != ia->ia_ifp)
804 			alt_ifa = &p->ia_ifa;
805 		if (++ia_count > 1 && alt_ifa != NULL)
806 			break;
807 	}
808 	if (alt_ifa != NULL && ia_count > 1)
809 		ifa_acquire(alt_ifa, &psref);
810 	pserialize_read_exit(s);
811 
812 	if (ia_count == 0)
813 		goto out;
814 
815 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
816 	if (alt_ifa != NULL && ia_count > 1)
817 		ifa_release(alt_ifa, &psref);
818 out:
819 	curlwp_bindx(bound);
820 }
821 
822 static void
823 in_scrubaddr(struct in_ifaddr *ia)
824 {
825 
826 	/* stop DAD processing */
827 	if (ia->ia_dad_stop != NULL)
828 		ia->ia_dad_stop(&ia->ia_ifa);
829 
830 	in_scrubprefix(ia);
831 	in_ifremlocal(&ia->ia_ifa);
832 
833 	mutex_enter(&in_ifaddr_lock);
834 	if (ia->ia_allhosts != NULL) {
835 		in_delmulti(ia->ia_allhosts);
836 		ia->ia_allhosts = NULL;
837 	}
838 	mutex_exit(&in_ifaddr_lock);
839 }
840 
841 /*
842  * Depends on it isn't called in concurrent. It should be guaranteed
843  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
844  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
845  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
846  * never be executed in concurrent.
847  */
848 void
849 in_purgeaddr(struct ifaddr *ifa)
850 {
851 	struct in_ifaddr *ia = (void *) ifa;
852 	struct ifnet *ifp = ifa->ifa_ifp;
853 
854 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
855 
856 	ifa->ifa_flags |= IFA_DESTROYING;
857 	in_scrubaddr(ia);
858 
859 	mutex_enter(&in_ifaddr_lock);
860 	in_addrhash_remove_locked(ia);
861 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
862 	IN_ADDRLIST_WRITER_REMOVE(ia);
863 	ifa_remove(ifp, &ia->ia_ifa);
864 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
865 	mutex_exit(&in_ifaddr_lock);
866 	IN_ADDRHASH_ENTRY_DESTROY(ia);
867 	IN_ADDRLIST_ENTRY_DESTROY(ia);
868 	ifafree(&ia->ia_ifa);
869 	in_setmaxmtu();
870 }
871 
872 static void
873 in_addrhash_insert_locked(struct in_ifaddr *ia)
874 {
875 
876 	KASSERT(mutex_owned(&in_ifaddr_lock));
877 
878 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
879 	    ia_hash);
880 	IN_ADDRHASH_ENTRY_INIT(ia);
881 	IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
882 }
883 
884 void
885 in_addrhash_insert(struct in_ifaddr *ia)
886 {
887 
888 	mutex_enter(&in_ifaddr_lock);
889 	in_addrhash_insert_locked(ia);
890 	mutex_exit(&in_ifaddr_lock);
891 }
892 
893 static void
894 in_addrhash_remove_locked(struct in_ifaddr *ia)
895 {
896 
897 	KASSERT(mutex_owned(&in_ifaddr_lock));
898 
899 	LIST_REMOVE(ia, ia_hash);
900 	IN_ADDRHASH_WRITER_REMOVE(ia);
901 }
902 
903 void
904 in_addrhash_remove(struct in_ifaddr *ia)
905 {
906 
907 	mutex_enter(&in_ifaddr_lock);
908 	in_addrhash_remove_locked(ia);
909 #ifdef NET_MPSAFE
910 	pserialize_perform(in_ifaddrhash_psz);
911 #endif
912 	mutex_exit(&in_ifaddr_lock);
913 	IN_ADDRHASH_ENTRY_DESTROY(ia);
914 }
915 
916 void
917 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
918 {
919 
920 	IFNET_LOCK(ifp);
921 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
922 	igmp_purgeif(ifp);		/* manipulates pools */
923 #ifdef MROUTING
924 	ip_mrouter_detach(ifp);
925 #endif
926 	IFNET_UNLOCK(ifp);
927 }
928 
929 /*
930  * SIOC[GAD]LIFADDR.
931  *	SIOCGLIFADDR: get first address. (???)
932  *	SIOCGLIFADDR with IFLR_PREFIX:
933  *		get first address that matches the specified prefix.
934  *	SIOCALIFADDR: add the specified address.
935  *	SIOCALIFADDR with IFLR_PREFIX:
936  *		EINVAL since we can't deduce hostid part of the address.
937  *	SIOCDLIFADDR: delete the specified address.
938  *	SIOCDLIFADDR with IFLR_PREFIX:
939  *		delete the first address that matches the specified prefix.
940  * return values:
941  *	EINVAL on invalid parameters
942  *	EADDRNOTAVAIL on prefix match failed/specified address not found
943  *	other values may be returned from in_ioctl()
944  */
945 static int
946 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
947     struct ifnet *ifp)
948 {
949 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
950 	struct ifaddr *ifa;
951 	struct sockaddr *sa;
952 
953 	/* sanity checks */
954 	if (data == NULL || ifp == NULL) {
955 		panic("invalid argument to in_lifaddr_ioctl");
956 		/*NOTRECHED*/
957 	}
958 
959 	switch (cmd) {
960 	case SIOCGLIFADDR:
961 		/* address must be specified on GET with IFLR_PREFIX */
962 		if ((iflr->flags & IFLR_PREFIX) == 0)
963 			break;
964 		/*FALLTHROUGH*/
965 	case SIOCALIFADDR:
966 	case SIOCDLIFADDR:
967 		/* address must be specified on ADD and DELETE */
968 		sa = (struct sockaddr *)&iflr->addr;
969 		if (sa->sa_family != AF_INET)
970 			return EINVAL;
971 		if (sa->sa_len != sizeof(struct sockaddr_in))
972 			return EINVAL;
973 		/* XXX need improvement */
974 		sa = (struct sockaddr *)&iflr->dstaddr;
975 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
976 			return EINVAL;
977 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
978 			return EINVAL;
979 		break;
980 	default: /*shouldn't happen*/
981 #if 0
982 		panic("invalid cmd to in_lifaddr_ioctl");
983 		/*NOTREACHED*/
984 #else
985 		return EOPNOTSUPP;
986 #endif
987 	}
988 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
989 		return EINVAL;
990 
991 	switch (cmd) {
992 	case SIOCALIFADDR:
993 	    {
994 		struct in_aliasreq ifra;
995 
996 		if (iflr->flags & IFLR_PREFIX)
997 			return EINVAL;
998 
999 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
1000 		memset(&ifra, 0, sizeof(ifra));
1001 		memcpy(ifra.ifra_name, iflr->iflr_name,
1002 			sizeof(ifra.ifra_name));
1003 
1004 		memcpy(&ifra.ifra_addr, &iflr->addr,
1005 			((struct sockaddr *)&iflr->addr)->sa_len);
1006 
1007 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1008 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1009 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1010 		}
1011 
1012 		ifra.ifra_mask.sin_family = AF_INET;
1013 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
1014 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
1015 
1016 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
1017 	    }
1018 	case SIOCGLIFADDR:
1019 	case SIOCDLIFADDR:
1020 	    {
1021 		struct in_ifaddr *ia;
1022 		struct in_addr mask, candidate, match;
1023 		struct sockaddr_in *sin;
1024 		int cmp, s;
1025 
1026 		memset(&mask, 0, sizeof(mask));
1027 		memset(&match, 0, sizeof(match));	/* XXX gcc */
1028 		if (iflr->flags & IFLR_PREFIX) {
1029 			/* lookup a prefix rather than address. */
1030 			in_len2mask(&mask, iflr->prefixlen);
1031 
1032 			sin = (struct sockaddr_in *)&iflr->addr;
1033 			match.s_addr = sin->sin_addr.s_addr;
1034 			match.s_addr &= mask.s_addr;
1035 
1036 			/* if you set extra bits, that's wrong */
1037 			if (match.s_addr != sin->sin_addr.s_addr)
1038 				return EINVAL;
1039 
1040 			cmp = 1;
1041 		} else {
1042 			if (cmd == SIOCGLIFADDR) {
1043 				/* on getting an address, take the 1st match */
1044 				cmp = 0;	/*XXX*/
1045 			} else {
1046 				/* on deleting an address, do exact match */
1047 				in_len2mask(&mask, 32);
1048 				sin = (struct sockaddr_in *)&iflr->addr;
1049 				match.s_addr = sin->sin_addr.s_addr;
1050 
1051 				cmp = 1;
1052 			}
1053 		}
1054 
1055 		s = pserialize_read_enter();
1056 		IFADDR_READER_FOREACH(ifa, ifp) {
1057 			if (ifa->ifa_addr->sa_family != AF_INET)
1058 				continue;
1059 			if (cmp == 0)
1060 				break;
1061 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
1062 			candidate.s_addr &= mask.s_addr;
1063 			if (candidate.s_addr == match.s_addr)
1064 				break;
1065 		}
1066 		if (ifa == NULL) {
1067 			pserialize_read_exit(s);
1068 			return EADDRNOTAVAIL;
1069 		}
1070 		ia = (struct in_ifaddr *)ifa;
1071 
1072 		if (cmd == SIOCGLIFADDR) {
1073 			/* fill in the if_laddrreq structure */
1074 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1075 
1076 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1077 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1078 					ia->ia_dstaddr.sin_len);
1079 			} else
1080 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1081 
1082 			iflr->prefixlen =
1083 				in_mask2len(&ia->ia_sockmask.sin_addr);
1084 
1085 			iflr->flags = 0;	/*XXX*/
1086 			pserialize_read_exit(s);
1087 
1088 			return 0;
1089 		} else {
1090 			struct in_aliasreq ifra;
1091 
1092 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1093 			memset(&ifra, 0, sizeof(ifra));
1094 			memcpy(ifra.ifra_name, iflr->iflr_name,
1095 				sizeof(ifra.ifra_name));
1096 
1097 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1098 				ia->ia_addr.sin_len);
1099 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1100 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1101 					ia->ia_dstaddr.sin_len);
1102 			}
1103 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1104 				ia->ia_sockmask.sin_len);
1105 			pserialize_read_exit(s);
1106 
1107 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1108 		}
1109 	    }
1110 	}
1111 
1112 	return EOPNOTSUPP;	/*just for safety*/
1113 }
1114 
1115 /*
1116  * Initialize an interface's internet address
1117  * and routing table entry.
1118  */
1119 int
1120 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1121     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1122 {
1123 	u_int32_t i;
1124 	struct sockaddr_in oldaddr, olddst;
1125 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1126 
1127 	if (sin == NULL)
1128 		sin = &ia->ia_addr;
1129 	if (dst == NULL)
1130 		dst = &ia->ia_dstaddr;
1131 
1132 	/*
1133 	 * Set up new addresses.
1134 	 */
1135 	oldaddr = ia->ia_addr;
1136 	olddst = ia->ia_dstaddr;
1137 	oldflags = ia->ia4_flags;
1138 	ia->ia_addr = *sin;
1139 	ia->ia_dstaddr = *dst;
1140 	hostIsNew = oldaddr.sin_family != AF_INET ||
1141 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1142 	if (!scrub)
1143 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1144 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1145 
1146 	/*
1147 	 * Configure address flags.
1148 	 * We need to do this early because they may be adjusted
1149 	 * by if_addr_init depending on the address.
1150 	 */
1151 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1152 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1153 		hostIsNew = 1;
1154 	}
1155 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1156 		ia->ia4_flags |= IN_IFF_DETACHED;
1157 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1158 	} else if (hostIsNew && if_do_dad(ifp)
1159 #if NARP > 0
1160 	    && ip_dad_count > 0
1161 #endif
1162 	)
1163 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1164 
1165 	/*
1166 	 * Give the interface a chance to initialize
1167 	 * if this is its first address,
1168 	 * and to validate the address if necessary.
1169 	 */
1170 	s = splsoftnet();
1171 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1172 	splx(s);
1173 	/* Now clear the try tentative flag, its job is done. */
1174 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1175 	if (error != 0) {
1176 		ia->ia_addr = oldaddr;
1177 		ia->ia_dstaddr = olddst;
1178 		ia->ia4_flags = oldflags;
1179 		return error;
1180 	}
1181 
1182 	if (scrub || hostIsNew) {
1183 		int newflags = ia->ia4_flags;
1184 
1185 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1186 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1187 		ia->ia4_flags = oldflags;
1188 		if (hostIsNew)
1189 			in_scrubaddr(ia);
1190 		else if (scrub)
1191 			in_scrubprefix(ia);
1192 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1193 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1194 		ia->ia4_flags = newflags;
1195 	}
1196 
1197 	i = ia->ia_addr.sin_addr.s_addr;
1198 	if (ifp->if_flags & IFF_POINTOPOINT)
1199 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1200 	else if (IN_CLASSA(i))
1201 		ia->ia_netmask = IN_CLASSA_NET;
1202 	else if (IN_CLASSB(i))
1203 		ia->ia_netmask = IN_CLASSB_NET;
1204 	else
1205 		ia->ia_netmask = IN_CLASSC_NET;
1206 	/*
1207 	 * The subnet mask usually includes at least the standard network part,
1208 	 * but may may be smaller in the case of supernetting.
1209 	 * If it is set, we believe it.
1210 	 */
1211 	if (ia->ia_subnetmask == 0) {
1212 		ia->ia_subnetmask = ia->ia_netmask;
1213 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1214 	} else
1215 		ia->ia_netmask &= ia->ia_subnetmask;
1216 
1217 	ia->ia_net = i & ia->ia_netmask;
1218 	ia->ia_subnet = i & ia->ia_subnetmask;
1219 	in_socktrim(&ia->ia_sockmask);
1220 
1221 	/* re-calculate the "in_maxmtu" value */
1222 	in_setmaxmtu();
1223 
1224 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1225 	if (ifp->if_flags & IFF_BROADCAST) {
1226 		ia->ia_broadaddr.sin_addr.s_addr =
1227 			ia->ia_subnet | ~ia->ia_subnetmask;
1228 		ia->ia_netbroadcast.s_addr =
1229 			ia->ia_net | ~ia->ia_netmask;
1230 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1231 		ia->ia_dstaddr = ia->ia_addr;
1232 		flags |= RTF_HOST;
1233 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1234 		if (ia->ia_dstaddr.sin_family != AF_INET)
1235 			return (0);
1236 		flags |= RTF_HOST;
1237 	}
1238 
1239 	/* Add the local route to the address */
1240 	in_ifaddlocal(&ia->ia_ifa);
1241 
1242 	/* Add the prefix route for the address */
1243 	error = in_addprefix(ia, flags);
1244 
1245 	/*
1246 	 * If the interface supports multicast, join the "all hosts"
1247 	 * multicast group on that interface.
1248 	 */
1249 	mutex_enter(&in_ifaddr_lock);
1250 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1251 		struct in_addr addr;
1252 
1253 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1254 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1255 	}
1256 	mutex_exit(&in_ifaddr_lock);
1257 
1258 	if (hostIsNew &&
1259 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1260 	    if_do_dad(ifp))
1261 		ia->ia_dad_start((struct ifaddr *)ia);
1262 
1263 	return error;
1264 }
1265 
1266 #define rtinitflags(x) \
1267 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1268 	    ? RTF_HOST : 0)
1269 
1270 /*
1271  * add a route to prefix ("connected route" in cisco terminology).
1272  * does nothing if there's some interface address with the same prefix already.
1273  */
1274 static int
1275 in_addprefix(struct in_ifaddr *target, int flags)
1276 {
1277 	struct in_ifaddr *ia;
1278 	struct in_addr prefix, mask, p;
1279 	int error;
1280 	int s;
1281 
1282 	if ((flags & RTF_HOST) != 0)
1283 		prefix = target->ia_dstaddr.sin_addr;
1284 	else {
1285 		prefix = target->ia_addr.sin_addr;
1286 		mask = target->ia_sockmask.sin_addr;
1287 		prefix.s_addr &= mask.s_addr;
1288 	}
1289 
1290 	s = pserialize_read_enter();
1291 	IN_ADDRLIST_READER_FOREACH(ia) {
1292 		if (rtinitflags(ia))
1293 			p = ia->ia_dstaddr.sin_addr;
1294 		else {
1295 			p = ia->ia_addr.sin_addr;
1296 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1297 		}
1298 
1299 		if (prefix.s_addr != p.s_addr)
1300 			continue;
1301 
1302 		/*
1303 		 * if we got a matching prefix route inserted by other
1304 		 * interface address, we don't need to bother
1305 		 *
1306 		 * XXX RADIX_MPATH implications here? -dyoung
1307 		 */
1308 		if (ia->ia_flags & IFA_ROUTE) {
1309 			pserialize_read_exit(s);
1310 			return 0;
1311 		}
1312 	}
1313 	pserialize_read_exit(s);
1314 
1315 	/*
1316 	 * noone seem to have prefix route.  insert it.
1317 	 */
1318 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1319 	if (error == 0)
1320 		target->ia_flags |= IFA_ROUTE;
1321 	else if (error == EEXIST) {
1322 		/*
1323 		 * the fact the route already exists is not an error.
1324 		 */
1325 		error = 0;
1326 	}
1327 	return error;
1328 }
1329 
1330 /*
1331  * remove a route to prefix ("connected route" in cisco terminology).
1332  * re-installs the route by using another interface address, if there's one
1333  * with the same prefix (otherwise we lose the route mistakenly).
1334  */
1335 static int
1336 in_scrubprefix(struct in_ifaddr *target)
1337 {
1338 	struct in_ifaddr *ia;
1339 	struct in_addr prefix, mask, p;
1340 	int error;
1341 	int s;
1342 
1343 	/* If we don't have IFA_ROUTE we have nothing to do */
1344 	if ((target->ia_flags & IFA_ROUTE) == 0)
1345 		return 0;
1346 
1347 	if (rtinitflags(target))
1348 		prefix = target->ia_dstaddr.sin_addr;
1349 	else {
1350 		prefix = target->ia_addr.sin_addr;
1351 		mask = target->ia_sockmask.sin_addr;
1352 		prefix.s_addr &= mask.s_addr;
1353 	}
1354 
1355 	s = pserialize_read_enter();
1356 	IN_ADDRLIST_READER_FOREACH(ia) {
1357 		if (rtinitflags(ia))
1358 			p = ia->ia_dstaddr.sin_addr;
1359 		else {
1360 			p = ia->ia_addr.sin_addr;
1361 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1362 		}
1363 
1364 		if (prefix.s_addr != p.s_addr)
1365 			continue;
1366 
1367 		/*
1368 		 * if we got a matching prefix route, move IFA_ROUTE to him
1369 		 */
1370 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1371 			struct psref psref;
1372 			int bound = curlwp_bind();
1373 
1374 			ia4_acquire(ia, &psref);
1375 			pserialize_read_exit(s);
1376 
1377 			rtinit(&target->ia_ifa, RTM_DELETE,
1378 			    rtinitflags(target));
1379 			target->ia_flags &= ~IFA_ROUTE;
1380 
1381 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1382 			    rtinitflags(ia) | RTF_UP);
1383 			if (error == 0)
1384 				ia->ia_flags |= IFA_ROUTE;
1385 
1386 			ia4_release(ia, &psref);
1387 			curlwp_bindx(bound);
1388 
1389 			return error;
1390 		}
1391 	}
1392 	pserialize_read_exit(s);
1393 
1394 	/*
1395 	 * noone seem to have prefix route.  remove it.
1396 	 */
1397 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1398 	target->ia_flags &= ~IFA_ROUTE;
1399 	return 0;
1400 }
1401 
1402 #undef rtinitflags
1403 
1404 /*
1405  * Return 1 if the address might be a local broadcast address.
1406  */
1407 int
1408 in_broadcast(struct in_addr in, struct ifnet *ifp)
1409 {
1410 	struct ifaddr *ifa;
1411 	int s;
1412 
1413 	KASSERT(ifp != NULL);
1414 
1415 	if (in.s_addr == INADDR_BROADCAST ||
1416 	    in_nullhost(in))
1417 		return 1;
1418 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1419 		return 0;
1420 	/*
1421 	 * Look through the list of addresses for a match
1422 	 * with a broadcast address.
1423 	 */
1424 #define ia (ifatoia(ifa))
1425 	s = pserialize_read_enter();
1426 	IFADDR_READER_FOREACH(ifa, ifp) {
1427 		if (ifa->ifa_addr->sa_family == AF_INET &&
1428 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1429 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1430 		     in_hosteq(in, ia->ia_netbroadcast) ||
1431 		     (hostzeroisbroadcast &&
1432 		      /*
1433 		       * Check for old-style (host 0) broadcast.
1434 		       */
1435 		      (in.s_addr == ia->ia_subnet ||
1436 		       in.s_addr == ia->ia_net)))) {
1437 			pserialize_read_exit(s);
1438 			return 1;
1439 		}
1440 	}
1441 	pserialize_read_exit(s);
1442 	return (0);
1443 #undef ia
1444 }
1445 
1446 /*
1447  * perform DAD when interface becomes IFF_UP.
1448  */
1449 void
1450 in_if_link_up(struct ifnet *ifp)
1451 {
1452 	struct ifaddr *ifa;
1453 	struct in_ifaddr *ia;
1454 	int s, bound;
1455 
1456 	/* Ensure it's sane to run DAD */
1457 	if (ifp->if_link_state == LINK_STATE_DOWN)
1458 		return;
1459 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1460 		return;
1461 
1462 	bound = curlwp_bind();
1463 	s = pserialize_read_enter();
1464 	IFADDR_READER_FOREACH(ifa, ifp) {
1465 		struct psref psref;
1466 
1467 		if (ifa->ifa_addr->sa_family != AF_INET)
1468 			continue;
1469 		ifa_acquire(ifa, &psref);
1470 		pserialize_read_exit(s);
1471 
1472 		ia = (struct in_ifaddr *)ifa;
1473 
1474 		/* If detached then mark as tentative */
1475 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1476 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1477 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1478 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1479 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1480 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1481 		}
1482 
1483 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1484 			/* Clear the duplicated flag as we're starting DAD. */
1485 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1486 			ia->ia_dad_start(ifa);
1487 		}
1488 
1489 		s = pserialize_read_enter();
1490 		ifa_release(ifa, &psref);
1491 	}
1492 	pserialize_read_exit(s);
1493 	curlwp_bindx(bound);
1494 }
1495 
1496 void
1497 in_if_up(struct ifnet *ifp)
1498 {
1499 
1500 	/* interface may not support link state, so bring it up also */
1501 	in_if_link_up(ifp);
1502 }
1503 
1504 /*
1505  * Mark all addresses as detached.
1506  */
1507 void
1508 in_if_link_down(struct ifnet *ifp)
1509 {
1510 	struct ifaddr *ifa;
1511 	struct in_ifaddr *ia;
1512 	int s, bound;
1513 
1514 	bound = curlwp_bind();
1515 	s = pserialize_read_enter();
1516 	IFADDR_READER_FOREACH(ifa, ifp) {
1517 		struct psref psref;
1518 
1519 		if (ifa->ifa_addr->sa_family != AF_INET)
1520 			continue;
1521 		ifa_acquire(ifa, &psref);
1522 		pserialize_read_exit(s);
1523 
1524 		ia = (struct in_ifaddr *)ifa;
1525 
1526 		/* Stop DAD processing */
1527 		if (ia->ia_dad_stop != NULL)
1528 			ia->ia_dad_stop(ifa);
1529 
1530 		/*
1531 		 * Mark the address as detached.
1532 		 */
1533 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1534 			ia->ia4_flags |= IN_IFF_DETACHED;
1535 			ia->ia4_flags &=
1536 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1537 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1538 		}
1539 
1540 		s = pserialize_read_enter();
1541 		ifa_release(ifa, &psref);
1542 	}
1543 	pserialize_read_exit(s);
1544 	curlwp_bindx(bound);
1545 }
1546 
1547 void
1548 in_if_down(struct ifnet *ifp)
1549 {
1550 
1551 	in_if_link_down(ifp);
1552 #if NARP > 0
1553 	lltable_purge_entries(LLTABLE(ifp));
1554 #endif
1555 }
1556 
1557 void
1558 in_if_link_state_change(struct ifnet *ifp, int link_state)
1559 {
1560 
1561 	switch (link_state) {
1562 	case LINK_STATE_DOWN:
1563 		in_if_link_down(ifp);
1564 		break;
1565 	case LINK_STATE_UP:
1566 		in_if_link_up(ifp);
1567 		break;
1568 	}
1569 }
1570 
1571 /*
1572  * in_lookup_multi: look up the in_multi record for a given IP
1573  * multicast address on a given interface.  If no matching record is
1574  * found, return NULL.
1575  */
1576 struct in_multi *
1577 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1578 {
1579 	struct in_multi *inm;
1580 
1581 	KASSERT(rw_lock_held(&in_multilock));
1582 
1583 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1584 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1585 			break;
1586 	}
1587 	return inm;
1588 }
1589 
1590 /*
1591  * in_multi_group: check whether the address belongs to an IP multicast
1592  * group we are joined on this interface.  Returns true or false.
1593  */
1594 bool
1595 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1596 {
1597 	bool ingroup;
1598 
1599 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1600 		rw_enter(&in_multilock, RW_READER);
1601 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1602 		rw_exit(&in_multilock);
1603 	} else {
1604 		/* XXX Recursive call from ip_output(). */
1605 		KASSERT(rw_lock_held(&in_multilock));
1606 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1607 	}
1608 	return ingroup;
1609 }
1610 
1611 /*
1612  * Add an address to the list of IP multicast addresses for a given interface.
1613  */
1614 struct in_multi *
1615 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1616 {
1617 	struct sockaddr_in sin;
1618 	struct in_multi *inm;
1619 
1620 	/*
1621 	 * See if address already in list.
1622 	 */
1623 	rw_enter(&in_multilock, RW_WRITER);
1624 	inm = in_lookup_multi(*ap, ifp);
1625 	if (inm != NULL) {
1626 		/*
1627 		 * Found it; just increment the reference count.
1628 		 */
1629 		inm->inm_refcount++;
1630 		rw_exit(&in_multilock);
1631 		return inm;
1632 	}
1633 
1634 	/*
1635 	 * New address; allocate a new multicast record.
1636 	 */
1637 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1638 	if (inm == NULL) {
1639 		rw_exit(&in_multilock);
1640 		return NULL;
1641 	}
1642 	inm->inm_addr = *ap;
1643 	inm->inm_ifp = ifp;
1644 	inm->inm_refcount = 1;
1645 
1646 	/*
1647 	 * Ask the network driver to update its multicast reception
1648 	 * filter appropriately for the new address.
1649 	 */
1650 	sockaddr_in_init(&sin, ap, 0);
1651 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1652 		rw_exit(&in_multilock);
1653 		pool_put(&inmulti_pool, inm);
1654 		return NULL;
1655 	}
1656 
1657 	/*
1658 	 * Let IGMP know that we have joined a new IP multicast group.
1659 	 */
1660 	if (igmp_joingroup(inm) != 0) {
1661 		rw_exit(&in_multilock);
1662 		pool_put(&inmulti_pool, inm);
1663 		return NULL;
1664 	}
1665 	LIST_INSERT_HEAD(
1666 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1667 	    inm, inm_list);
1668 	in_multientries++;
1669 	rw_exit(&in_multilock);
1670 
1671 	return inm;
1672 }
1673 
1674 /*
1675  * Delete a multicast address record.
1676  */
1677 void
1678 in_delmulti(struct in_multi *inm)
1679 {
1680 	struct sockaddr_in sin;
1681 
1682 	rw_enter(&in_multilock, RW_WRITER);
1683 	if (--inm->inm_refcount > 0) {
1684 		rw_exit(&in_multilock);
1685 		return;
1686 	}
1687 
1688 	/*
1689 	 * No remaining claims to this record; let IGMP know that
1690 	 * we are leaving the multicast group.
1691 	 */
1692 	igmp_leavegroup(inm);
1693 
1694 	/*
1695 	 * Notify the network driver to update its multicast reception
1696 	 * filter.
1697 	 */
1698 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1699 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1700 
1701 	/*
1702 	 * Unlink from list.
1703 	 */
1704 	LIST_REMOVE(inm, inm_list);
1705 	in_multientries--;
1706 	rw_exit(&in_multilock);
1707 
1708 	pool_put(&inmulti_pool, inm);
1709 }
1710 
1711 /*
1712  * in_next_multi: step through all of the in_multi records, one at a time.
1713  * The current position is remembered in "step", which the caller must
1714  * provide.  in_first_multi(), below, must be called to initialize "step"
1715  * and get the first record.  Both macros return a NULL "inm" when there
1716  * are no remaining records.
1717  */
1718 struct in_multi *
1719 in_next_multi(struct in_multistep *step)
1720 {
1721 	struct in_multi *inm;
1722 
1723 	KASSERT(rw_lock_held(&in_multilock));
1724 
1725 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1726 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1727 	}
1728 	if ((inm = step->i_inm) != NULL) {
1729 		step->i_inm = LIST_NEXT(inm, inm_list);
1730 	}
1731 	return inm;
1732 }
1733 
1734 struct in_multi *
1735 in_first_multi(struct in_multistep *step)
1736 {
1737 	KASSERT(rw_lock_held(&in_multilock));
1738 
1739 	step->i_n = 0;
1740 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1741 	return in_next_multi(step);
1742 }
1743 
1744 void
1745 in_multi_lock(int op)
1746 {
1747 	rw_enter(&in_multilock, op);
1748 }
1749 
1750 void
1751 in_multi_unlock(void)
1752 {
1753 	rw_exit(&in_multilock);
1754 }
1755 
1756 int
1757 in_multi_lock_held(void)
1758 {
1759 	return rw_lock_held(&in_multilock);
1760 }
1761 
1762 struct in_ifaddr *
1763 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1764     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1765 {
1766 	struct rtentry *rt = NULL;
1767 	struct in_ifaddr *ia = NULL;
1768 
1769 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1770 	/*
1771          * If route is known or can be allocated now, take the
1772          * source address from the interface.  Otherwise, punt.
1773 	 */
1774 	if ((soopts & SO_DONTROUTE) != 0)
1775 		rtcache_free(ro);
1776 	else {
1777 		union {
1778 			struct sockaddr		dst;
1779 			struct sockaddr_in	dst4;
1780 		} u;
1781 
1782 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1783 		rt = rtcache_lookup(ro, &u.dst);
1784 	}
1785 	/*
1786 	 * If we found a route, use the address
1787 	 * corresponding to the outgoing interface
1788 	 * unless it is the loopback (in case a route
1789 	 * to our address on another net goes to loopback).
1790 	 *
1791 	 * XXX Is this still true?  Do we care?
1792 	 */
1793 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1794 		int s;
1795 		struct ifaddr *ifa;
1796 		/*
1797 		 * Just in case. May not need to do this workaround.
1798 		 * Revisit when working on rtentry MP-ification.
1799 		 */
1800 		s = pserialize_read_enter();
1801 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1802 			if (ifa == rt->rt_ifa)
1803 				break;
1804 		}
1805 		if (ifa != NULL)
1806 			ifa_acquire(ifa, psref);
1807 		pserialize_read_exit(s);
1808 
1809 		ia = ifatoia(ifa);
1810 	}
1811 	if (ia == NULL) {
1812 		u_int16_t fport = sin->sin_port;
1813 		struct ifaddr *ifa;
1814 		int s;
1815 
1816 		sin->sin_port = 0;
1817 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1818 		sin->sin_port = fport;
1819 		if (ifa == NULL) {
1820 			/* Find 1st non-loopback AF_INET address */
1821 			s = pserialize_read_enter();
1822 			IN_ADDRLIST_READER_FOREACH(ia) {
1823 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1824 					break;
1825 			}
1826 			if (ia != NULL)
1827 				ia4_acquire(ia, psref);
1828 			pserialize_read_exit(s);
1829 		} else {
1830 			/* ia is already referenced by psref */
1831 			ia = ifatoia(ifa);
1832 		}
1833 		if (ia == NULL) {
1834 			*errorp = EADDRNOTAVAIL;
1835 			goto out;
1836 		}
1837 	}
1838 	/*
1839 	 * If the destination address is multicast and an outgoing
1840 	 * interface has been set as a multicast option, use the
1841 	 * address of that interface as our source address.
1842 	 */
1843 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1844 		struct ip_moptions *imo;
1845 
1846 		imo = mopts;
1847 		if (imo->imo_multicast_if_index != 0) {
1848 			struct ifnet *ifp;
1849 			int s;
1850 
1851 			if (ia != NULL)
1852 				ia4_release(ia, psref);
1853 			s = pserialize_read_enter();
1854 			ifp = if_byindex(imo->imo_multicast_if_index);
1855 			if (ifp != NULL) {
1856 				/* XXX */
1857 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1858 			} else
1859 				ia = NULL;
1860 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1861 				pserialize_read_exit(s);
1862 				if (ia != NULL)
1863 					ia4_release(ia, psref);
1864 				*errorp = EADDRNOTAVAIL;
1865 				ia = NULL;
1866 				goto out;
1867 			}
1868 			pserialize_read_exit(s);
1869 		}
1870 	}
1871 	if (ia->ia_ifa.ifa_getifa != NULL) {
1872 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1873 		                                      sintosa(sin)));
1874 		if (ia == NULL) {
1875 			*errorp = EADDRNOTAVAIL;
1876 			goto out;
1877 		}
1878 		/* FIXME NOMPSAFE */
1879 		ia4_acquire(ia, psref);
1880 	}
1881 #ifdef GETIFA_DEBUG
1882 	else
1883 		printf("%s: missing ifa_getifa\n", __func__);
1884 #endif
1885 out:
1886 	rtcache_unref(rt, ro);
1887 	return ia;
1888 }
1889 
1890 int
1891 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst)
1892 {
1893 	struct in_ifaddr *ia4;
1894 	int s;
1895 
1896 	/* check for address match */
1897 	if (src.s_addr != ip->ip_dst.s_addr ||
1898 	    dst.s_addr != ip->ip_src.s_addr)
1899 		return 0;
1900 
1901 	/* martian filters on outer source - NOT done in ip_input! */
1902 	if (IN_MULTICAST(ip->ip_src.s_addr))
1903 		return 0;
1904 	switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) {
1905 	case 0:
1906 	case 127:
1907 	case 255:
1908 		return 0;
1909 	}
1910 	/* reject packets with broadcast on source */
1911 	s = pserialize_read_enter();
1912 	IN_ADDRLIST_READER_FOREACH(ia4) {
1913 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
1914 			continue;
1915 		if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
1916 			pserialize_read_exit(s);
1917 			return 0;
1918 		}
1919 	}
1920 	pserialize_read_exit(s);
1921 
1922 	/* NOTE: packet may dropped by uRPF */
1923 
1924 	/* return valid bytes length */
1925 	return sizeof(src) + sizeof(dst);
1926 }
1927 
1928 #if NARP > 0
1929 
1930 #define	IN_LLTBL_DEFAULT_HSIZE	32
1931 #define	IN_LLTBL_HASH(k, h) \
1932 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1933 
1934 /*
1935  * Do actual deallocation of @lle.
1936  * Called by LLE_FREE_LOCKED when number of references
1937  * drops to zero.
1938  */
1939 static void
1940 in_lltable_destroy_lle(struct llentry *lle)
1941 {
1942 
1943 	KASSERT(lle->la_numheld == 0);
1944 
1945 	LLE_WUNLOCK(lle);
1946 	LLE_LOCK_DESTROY(lle);
1947 	llentry_pool_put(lle);
1948 }
1949 
1950 static struct llentry *
1951 in_lltable_new(struct in_addr addr4, u_int flags)
1952 {
1953 	struct llentry *lle;
1954 
1955 	lle = llentry_pool_get(PR_NOWAIT);
1956 	if (lle == NULL)		/* NB: caller generates msg */
1957 		return NULL;
1958 
1959 	/*
1960 	 * For IPv4 this will trigger "arpresolve" to generate
1961 	 * an ARP request.
1962 	 */
1963 	lle->la_expire = time_uptime; /* mark expired */
1964 	lle->r_l3addr.addr4 = addr4;
1965 	lle->lle_refcnt = 1;
1966 	lle->lle_free = in_lltable_destroy_lle;
1967 	LLE_LOCK_INIT(lle);
1968 	callout_init(&lle->la_timer, CALLOUT_MPSAFE);
1969 
1970 	return lle;
1971 }
1972 
1973 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1974 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1975 
1976 static int
1977 in_lltable_match_prefix(const struct sockaddr *prefix,
1978     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1979 {
1980 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1981 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1982 	struct in_addr lle_addr;
1983 
1984 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1985 
1986 	/*
1987 	 * (flags & LLE_STATIC) means deleting all entries
1988 	 * including static ARP entries.
1989 	 */
1990 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) &&
1991 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1992 		return (1);
1993 
1994 	return (0);
1995 }
1996 
1997 static void
1998 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1999 {
2000 	size_t pkts_dropped;
2001 
2002 	LLE_WLOCK_ASSERT(lle);
2003 	KASSERT(llt != NULL);
2004 
2005 	pkts_dropped = llentry_free(lle);
2006 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
2007 }
2008 
2009 static int
2010 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr,
2011     const struct rtentry *rt)
2012 {
2013 	int error = EINVAL;
2014 
2015 	if (rt == NULL)
2016 		return error;
2017 
2018 	/*
2019 	 * If the gateway for an existing host route matches the target L3
2020 	 * address, which is a special route inserted by some implementation
2021 	 * such as MANET, and the interface is of the correct type, then
2022 	 * allow for ARP to proceed.
2023 	 */
2024 	if (rt->rt_flags & RTF_GATEWAY) {
2025 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
2026 		    rt->rt_ifp->if_type != IFT_ETHER ||
2027 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
2028 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
2029 		    sizeof(in_addr_t)) != 0) {
2030 			goto error;
2031 		}
2032 	}
2033 
2034 	/*
2035 	 * Make sure that at least the destination address is covered
2036 	 * by the route. This is for handling the case where 2 or more
2037 	 * interfaces have the same prefix. An incoming packet arrives
2038 	 * on one interface and the corresponding outgoing packet leaves
2039 	 * another interface.
2040 	 */
2041 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
2042 		const char *sa, *mask, *addr, *lim;
2043 		int len;
2044 
2045 		mask = (const char *)rt_mask(rt);
2046 		/*
2047 		 * Just being extra cautious to avoid some custom
2048 		 * code getting into trouble.
2049 		 */
2050 		if (mask == NULL)
2051 			goto error;
2052 
2053 		sa = (const char *)rt_getkey(rt);
2054 		addr = (const char *)l3addr;
2055 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
2056 		lim = addr + len;
2057 
2058 		for ( ; addr < lim; sa++, mask++, addr++) {
2059 			if ((*sa ^ *addr) & *mask) {
2060 #ifdef DIAGNOSTIC
2061 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
2062 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
2063 #endif
2064 				goto error;
2065 			}
2066 		}
2067 	}
2068 
2069 	error = 0;
2070 error:
2071 	return error;
2072 }
2073 
2074 static inline uint32_t
2075 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
2076 {
2077 
2078 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
2079 }
2080 
2081 static uint32_t
2082 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
2083 {
2084 
2085 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
2086 }
2087 
2088 static void
2089 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2090 {
2091 	struct sockaddr_in *sin;
2092 
2093 	sin = (struct sockaddr_in *)sa;
2094 	memset(sin, 0, sizeof(*sin));
2095 	sin->sin_family = AF_INET;
2096 	sin->sin_len = sizeof(*sin);
2097 	sin->sin_addr = lle->r_l3addr.addr4;
2098 }
2099 
2100 static inline struct llentry *
2101 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2102 {
2103 	struct llentry *lle;
2104 	struct llentries *lleh;
2105 	u_int hashidx;
2106 
2107 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2108 	lleh = &llt->lle_head[hashidx];
2109 	LIST_FOREACH(lle, lleh, lle_next) {
2110 		if (lle->la_flags & LLE_DELETED)
2111 			continue;
2112 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2113 			break;
2114 	}
2115 
2116 	return (lle);
2117 }
2118 
2119 static int
2120 in_lltable_delete(struct lltable *llt, u_int flags,
2121     const struct sockaddr *l3addr)
2122 {
2123 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2124 	struct ifnet *ifp __diagused = llt->llt_ifp;
2125 	struct llentry *lle;
2126 
2127 	IF_AFDATA_WLOCK_ASSERT(ifp);
2128 	KASSERTMSG(l3addr->sa_family == AF_INET,
2129 	    "sin_family %d", l3addr->sa_family);
2130 
2131 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2132 	if (lle == NULL) {
2133 #ifdef LLTABLE_DEBUG
2134 		char buf[64];
2135 		sockaddr_format(l3addr, buf, sizeof(buf));
2136 		log(LOG_INFO, "%s: cache for %s is not found\n",
2137 		    __func__, buf);
2138 #endif
2139 		return (ENOENT);
2140 	}
2141 
2142 	LLE_WLOCK(lle);
2143 #ifdef LLTABLE_DEBUG
2144 	{
2145 		char buf[64];
2146 		sockaddr_format(l3addr, buf, sizeof(buf));
2147 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2148 		    __func__, buf, lle);
2149 	}
2150 #endif
2151 	llentry_free(lle);
2152 
2153 	return (0);
2154 }
2155 
2156 static struct llentry *
2157 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr,
2158     const struct rtentry *rt)
2159 {
2160 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2161 	struct ifnet *ifp = llt->llt_ifp;
2162 	struct llentry *lle;
2163 
2164 	IF_AFDATA_WLOCK_ASSERT(ifp);
2165 	KASSERTMSG(l3addr->sa_family == AF_INET,
2166 	    "sin_family %d", l3addr->sa_family);
2167 
2168 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2169 
2170 	if (lle != NULL) {
2171 		LLE_WLOCK(lle);
2172 		return (lle);
2173 	}
2174 
2175 	/* no existing record, we need to create new one */
2176 
2177 	/*
2178 	 * A route that covers the given address must have
2179 	 * been installed 1st because we are doing a resolution,
2180 	 * verify this.
2181 	 */
2182 	if (!(flags & LLE_IFADDR) &&
2183 	    in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2184 		return (NULL);
2185 
2186 	lle = in_lltable_new(sin->sin_addr, flags);
2187 	if (lle == NULL) {
2188 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2189 		return (NULL);
2190 	}
2191 	lle->la_flags = flags;
2192 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2193 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2194 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2195 	}
2196 
2197 	lltable_link_entry(llt, lle);
2198 	LLE_WLOCK(lle);
2199 
2200 	return (lle);
2201 }
2202 
2203 /*
2204  * Return NULL if not found or marked for deletion.
2205  * If found return lle read locked.
2206  */
2207 static struct llentry *
2208 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2209 {
2210 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2211 	struct llentry *lle;
2212 
2213 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2214 	KASSERTMSG(l3addr->sa_family == AF_INET,
2215 	    "sin_family %d", l3addr->sa_family);
2216 
2217 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2218 
2219 	if (lle == NULL)
2220 		return NULL;
2221 
2222 	if (flags & LLE_EXCLUSIVE)
2223 		LLE_WLOCK(lle);
2224 	else
2225 		LLE_RLOCK(lle);
2226 
2227 	return lle;
2228 }
2229 
2230 static int
2231 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2232     struct rt_walkarg *w)
2233 {
2234 	struct sockaddr_in sin;
2235 
2236 	LLTABLE_LOCK_ASSERT();
2237 
2238 	/* skip deleted entries */
2239 	if (lle->la_flags & LLE_DELETED)
2240 		return 0;
2241 
2242 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2243 
2244 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2245 }
2246 
2247 #endif /* NARP > 0 */
2248 
2249 static int
2250 in_multicast_sysctl(SYSCTLFN_ARGS)
2251 {
2252 	struct ifnet *ifp;
2253 	struct ifaddr *ifa;
2254 	struct in_ifaddr *ifa4;
2255 	struct in_multi *inm;
2256 	uint32_t tmp;
2257 	int error;
2258 	size_t written;
2259 	struct psref psref;
2260 	int bound;
2261 
2262 	if (namelen != 1)
2263 		return EINVAL;
2264 
2265 	bound = curlwp_bind();
2266 	ifp = if_get_byindex(name[0], &psref);
2267 	if (ifp == NULL) {
2268 		curlwp_bindx(bound);
2269 		return ENODEV;
2270 	}
2271 
2272 	if (oldp == NULL) {
2273 		*oldlenp = 0;
2274 		IFADDR_FOREACH(ifa, ifp) {
2275 			if (ifa->ifa_addr->sa_family != AF_INET)
2276 				continue;
2277 			ifa4 = (void *)ifa;
2278 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2279 				*oldlenp += 2 * sizeof(struct in_addr) +
2280 				    sizeof(uint32_t);
2281 			}
2282 		}
2283 		if_put(ifp, &psref);
2284 		curlwp_bindx(bound);
2285 		return 0;
2286 	}
2287 
2288 	error = 0;
2289 	written = 0;
2290 	IFADDR_FOREACH(ifa, ifp) {
2291 		if (ifa->ifa_addr->sa_family != AF_INET)
2292 			continue;
2293 		ifa4 = (void *)ifa;
2294 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2295 			if (written + 2 * sizeof(struct in_addr) +
2296 			    sizeof(uint32_t) > *oldlenp)
2297 				goto done;
2298 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2299 			    oldp, sizeof(struct in_addr));
2300 			if (error)
2301 				goto done;
2302 			oldp = (char *)oldp + sizeof(struct in_addr);
2303 			written += sizeof(struct in_addr);
2304 			error = sysctl_copyout(l, &inm->inm_addr,
2305 			    oldp, sizeof(struct in_addr));
2306 			if (error)
2307 				goto done;
2308 			oldp = (char *)oldp + sizeof(struct in_addr);
2309 			written += sizeof(struct in_addr);
2310 			tmp = inm->inm_refcount;
2311 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2312 			if (error)
2313 				goto done;
2314 			oldp = (char *)oldp + sizeof(tmp);
2315 			written += sizeof(tmp);
2316 		}
2317 	}
2318 done:
2319 	if_put(ifp, &psref);
2320 	curlwp_bindx(bound);
2321 	*oldlenp = written;
2322 	return error;
2323 }
2324 
2325 static void
2326 in_sysctl_init(struct sysctllog **clog)
2327 {
2328 	sysctl_createv(clog, 0, NULL, NULL,
2329 		       CTLFLAG_PERMANENT,
2330 		       CTLTYPE_NODE, "inet",
2331 		       SYSCTL_DESCR("PF_INET related settings"),
2332 		       NULL, 0, NULL, 0,
2333 		       CTL_NET, PF_INET, CTL_EOL);
2334 	sysctl_createv(clog, 0, NULL, NULL,
2335 		       CTLFLAG_PERMANENT,
2336 		       CTLTYPE_NODE, "multicast",
2337 		       SYSCTL_DESCR("Multicast information"),
2338 		       in_multicast_sysctl, 0, NULL, 0,
2339 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2340 	sysctl_createv(clog, 0, NULL, NULL,
2341 		       CTLFLAG_PERMANENT,
2342 		       CTLTYPE_NODE, "ip",
2343 		       SYSCTL_DESCR("IPv4 related settings"),
2344 		       NULL, 0, NULL, 0,
2345 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2346 
2347 	sysctl_createv(clog, 0, NULL, NULL,
2348 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2349 		       CTLTYPE_INT, "subnetsarelocal",
2350 		       SYSCTL_DESCR("Whether logical subnets are considered "
2351 				    "local"),
2352 		       NULL, 0, &subnetsarelocal, 0,
2353 		       CTL_NET, PF_INET, IPPROTO_IP,
2354 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2355 	sysctl_createv(clog, 0, NULL, NULL,
2356 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2357 		       CTLTYPE_INT, "hostzerobroadcast",
2358 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2359 		       NULL, 0, &hostzeroisbroadcast, 0,
2360 		       CTL_NET, PF_INET, IPPROTO_IP,
2361 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2362 }
2363 
2364 #if NARP > 0
2365 
2366 static struct lltable *
2367 in_lltattach(struct ifnet *ifp)
2368 {
2369 	struct lltable *llt;
2370 
2371 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2372 	llt->llt_af = AF_INET;
2373 	llt->llt_ifp = ifp;
2374 
2375 	llt->llt_lookup = in_lltable_lookup;
2376 	llt->llt_create = in_lltable_create;
2377 	llt->llt_delete = in_lltable_delete;
2378 	llt->llt_dump_entry = in_lltable_dump_entry;
2379 	llt->llt_hash = in_lltable_hash;
2380 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2381 	llt->llt_free_entry = in_lltable_free_entry;
2382 	llt->llt_match_prefix = in_lltable_match_prefix;
2383 	lltable_link(llt);
2384 
2385 	return (llt);
2386 }
2387 
2388 #endif /* NARP > 0 */
2389 
2390 void *
2391 in_domifattach(struct ifnet *ifp)
2392 {
2393 	struct in_ifinfo *ii;
2394 
2395 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2396 
2397 #if NARP > 0
2398 	ii->ii_llt = in_lltattach(ifp);
2399 #endif
2400 
2401 #ifdef IPSELSRC
2402 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2403 	KASSERT(ii->ii_selsrc != NULL);
2404 #endif
2405 
2406 	return ii;
2407 }
2408 
2409 void
2410 in_domifdetach(struct ifnet *ifp, void *aux)
2411 {
2412 	struct in_ifinfo *ii = aux;
2413 
2414 #ifdef IPSELSRC
2415 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2416 #endif
2417 #if NARP > 0
2418 	lltable_free(ii->ii_llt);
2419 #endif
2420 	kmem_free(ii, sizeof(struct in_ifinfo));
2421 }
2422