xref: /netbsd-src/sys/netinet/in.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: in.c,v 1.210 2017/11/17 07:37:12 ozaki-r 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 1.210 2017/11/17 07:37:12 ozaki-r 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 		/* FALLTHROUGH */
483 	case SIOCSIFDSTADDR:
484 		if (ifra->ifra_addr.sin_family != AF_INET) {
485 			error = EAFNOSUPPORT;
486 			goto out;
487 		}
488 		/* FALLTHROUGH */
489 	case SIOCSIFNETMASK:
490 		if (ifp == NULL)
491 			panic("in_control");
492 
493 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
494 			break;
495 
496 		if (ia == NULL &&
497 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
498 			error = EADDRNOTAVAIL;
499 			goto out;
500 		}
501 
502 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
503 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
504 		    NULL) != 0) {
505 			error = EPERM;
506 			goto out;
507 		}
508 
509 		if (ia == NULL) {
510 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
511 			if (ia == NULL) {
512 				error = ENOBUFS;
513 				goto out;
514 			}
515 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
516 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
517 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
518 #ifdef IPSELSRC
519 			ia->ia_ifa.ifa_getifa = in_getifa;
520 #else /* IPSELSRC */
521 			ia->ia_ifa.ifa_getifa = NULL;
522 #endif /* IPSELSRC */
523 			ia->ia_sockmask.sin_len = 8;
524 			ia->ia_sockmask.sin_family = AF_INET;
525 			if (ifp->if_flags & IFF_BROADCAST) {
526 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
527 				ia->ia_broadaddr.sin_family = AF_INET;
528 			}
529 			ia->ia_ifp = ifp;
530 			ia->ia_idsalt = cprng_fast32() % 65535;
531 			LIST_INIT(&ia->ia_multiaddrs);
532 			IN_ADDRHASH_ENTRY_INIT(ia);
533 			IN_ADDRLIST_ENTRY_INIT(ia);
534 			ifa_psref_init(&ia->ia_ifa);
535 			/*
536 			 * We need a reference to make ia survive over in_ifinit
537 			 * that does ifaref and ifafree.
538 			 */
539 			ifaref(&ia->ia_ifa);
540 
541 			newifaddr = 1;
542 		}
543 		break;
544 
545 	case SIOCSIFBRDADDR:
546 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
547 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
548 		    NULL) != 0) {
549 			error = EPERM;
550 			goto out;
551 		}
552 		/* FALLTHROUGH */
553 
554 	case SIOCGIFADDR:
555 	case SIOCGIFNETMASK:
556 	case SIOCGIFDSTADDR:
557 	case SIOCGIFBRDADDR:
558 		if (ia == NULL) {
559 			error = EADDRNOTAVAIL;
560 			goto out;
561 		}
562 		break;
563 	}
564 	error = 0;
565 	switch (cmd) {
566 
567 	case SIOCGIFADDR:
568 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
569 		break;
570 
571 	case SIOCGIFBRDADDR:
572 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
573 			error = EINVAL;
574 			goto out;
575 		}
576 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
577 		break;
578 
579 	case SIOCGIFDSTADDR:
580 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
581 			error = EINVAL;
582 			goto out;
583 		}
584 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
585 		break;
586 
587 	case SIOCGIFNETMASK:
588 		/*
589 		 * We keep the number of trailing zero bytes the sin_len field
590 		 * of ia_sockmask, so we fix this before we pass it back to
591 		 * userland.
592 		 */
593 		oldaddr = ia->ia_sockmask;
594 		oldaddr.sin_len = sizeof(struct sockaddr_in);
595 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
596 		break;
597 
598 	case SIOCSIFDSTADDR:
599 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
600 			error = EINVAL;
601 			goto out;
602 		}
603 		oldaddr = ia->ia_dstaddr;
604 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
605 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
606 			ia->ia_dstaddr = oldaddr;
607 			goto out;
608 		}
609 		if (ia->ia_flags & IFA_ROUTE) {
610 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
611 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
612 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
613 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
614 		}
615 		break;
616 
617 	case SIOCSIFBRDADDR:
618 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
619 			error = EINVAL;
620 			goto out;
621 		}
622 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
623 		break;
624 
625 	case SIOCSIFADDR:
626 		if (!newifaddr) {
627 			in_addrhash_remove(ia);
628 			need_reinsert = true;
629 		}
630 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
631 		    NULL, 1);
632 
633 		run_hook = true;
634 		break;
635 
636 	case SIOCSIFNETMASK:
637 		in_scrubprefix(ia);
638 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
639 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
640 		if (!newifaddr) {
641 			in_addrhash_remove(ia);
642 			need_reinsert = true;
643 		}
644 		error = in_ifinit(ifp, ia, NULL, NULL, 0);
645 		break;
646 
647 	case SIOCAIFADDR:
648 		maskIsNew = 0;
649 		if (ifra->ifra_mask.sin_len) {
650 			in_scrubprefix(ia);
651 			ia->ia_sockmask = ifra->ifra_mask;
652 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
653 			maskIsNew = 1;
654 		}
655 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
656 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
657 			new_dstaddr = &ifra->ifra_dstaddr;
658 			maskIsNew  = 1; /* We lie; but the effect's the same */
659 		} else
660 			new_dstaddr = NULL;
661 		if (ifra->ifra_addr.sin_family == AF_INET &&
662 		    (hostIsNew || maskIsNew)) {
663 			if (!newifaddr) {
664 				in_addrhash_remove(ia);
665 				need_reinsert = true;
666 			}
667 			error = in_ifinit(ifp, ia, &ifra->ifra_addr,
668 			    new_dstaddr, 0);
669 		}
670 		if ((ifp->if_flags & IFF_BROADCAST) &&
671 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
672 			ia->ia_broadaddr = ifra->ifra_broadaddr;
673 		run_hook = true;
674 		break;
675 
676 	case SIOCGIFALIAS:
677 		ifra->ifra_mask = ia->ia_sockmask;
678 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
679 		    (ia->ia_dstaddr.sin_family == AF_INET))
680 			ifra->ifra_dstaddr = ia->ia_dstaddr;
681 		else if ((ifp->if_flags & IFF_BROADCAST) &&
682 		    (ia->ia_broadaddr.sin_family == AF_INET))
683 			ifra->ifra_broadaddr = ia->ia_broadaddr;
684 		else
685 			memset(&ifra->ifra_broadaddr, 0,
686 			      sizeof(ifra->ifra_broadaddr));
687 		break;
688 
689 	case SIOCGIFAFLAG_IN:
690 		ifr->ifr_addrflags = ia->ia4_flags;
691 		break;
692 
693 	case SIOCDIFADDR:
694 		ia4_release(ia, &psref);
695 		ifaref(&ia->ia_ifa);
696 		in_purgeaddr(&ia->ia_ifa);
697 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
698 		ifafree(&ia->ia_ifa);
699 		ia = NULL;
700 		break;
701 
702 #ifdef MROUTING
703 	case SIOCGETVIFCNT:
704 	case SIOCGETSGCNT:
705 		error = mrt_ioctl(so, cmd, data);
706 		break;
707 #endif /* MROUTING */
708 
709 	default:
710 		error = ENOTTY;
711 		goto out;
712 	}
713 
714 	/*
715 	 * XXX insert regardless of error to make in_purgeaddr below work.
716 	 * Need to improve.
717 	 */
718 	if (newifaddr) {
719 		ifaref(&ia->ia_ifa);
720 		ifa_insert(ifp, &ia->ia_ifa);
721 
722 		mutex_enter(&in_ifaddr_lock);
723 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
724 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
725 		in_addrhash_insert_locked(ia);
726 		/* Release a reference that is held just after creation. */
727 		ifafree(&ia->ia_ifa);
728 		mutex_exit(&in_ifaddr_lock);
729 	} else if (need_reinsert) {
730 		in_addrhash_insert(ia);
731 	}
732 
733 	if (error == 0) {
734 		if (run_hook)
735 			pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
736 	} else if (newifaddr) {
737 		KASSERT(ia != NULL);
738 		in_purgeaddr(&ia->ia_ifa);
739 		ia = NULL;
740 	}
741 
742 out:
743 	if (!newifaddr && ia != NULL)
744 		ia4_release(ia, &psref);
745 	curlwp_bindx(bound);
746 	return error;
747 }
748 
749 int
750 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
751 {
752 	int error;
753 
754 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
755 	error = in_control0(so, cmd, data, ifp);
756 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
757 
758 	return error;
759 }
760 
761 /* Add ownaddr as loopback rtentry. */
762 static void
763 in_ifaddlocal(struct ifaddr *ifa)
764 {
765 	struct in_ifaddr *ia;
766 
767 	ia = (struct in_ifaddr *)ifa;
768 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
769 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
770 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
771 	{
772 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
773 		return;
774 	}
775 
776 	rt_ifa_addlocal(ifa);
777 }
778 
779 /* Remove loopback entry of ownaddr */
780 static void
781 in_ifremlocal(struct ifaddr *ifa)
782 {
783 	struct in_ifaddr *ia, *p;
784 	struct ifaddr *alt_ifa = NULL;
785 	int ia_count = 0;
786 	int s;
787 	struct psref psref;
788 	int bound = curlwp_bind();
789 
790 	ia = (struct in_ifaddr *)ifa;
791 	/* Delete the entry if exactly one ifaddr matches the
792 	 * address, ifa->ifa_addr. */
793 	s = pserialize_read_enter();
794 	IN_ADDRLIST_READER_FOREACH(p) {
795 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
796 			continue;
797 		if (p->ia_ifp != ia->ia_ifp)
798 			alt_ifa = &p->ia_ifa;
799 		if (++ia_count > 1 && alt_ifa != NULL)
800 			break;
801 	}
802 	if (alt_ifa != NULL && ia_count > 1)
803 		ifa_acquire(alt_ifa, &psref);
804 	pserialize_read_exit(s);
805 
806 	if (ia_count == 0)
807 		goto out;
808 
809 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
810 	if (alt_ifa != NULL && ia_count > 1)
811 		ifa_release(alt_ifa, &psref);
812 out:
813 	curlwp_bindx(bound);
814 }
815 
816 static void
817 in_scrubaddr(struct in_ifaddr *ia)
818 {
819 
820 	/* stop DAD processing */
821 	if (ia->ia_dad_stop != NULL)
822 		ia->ia_dad_stop(&ia->ia_ifa);
823 
824 	in_scrubprefix(ia);
825 	in_ifremlocal(&ia->ia_ifa);
826 
827 	mutex_enter(&in_ifaddr_lock);
828 	if (ia->ia_allhosts != NULL) {
829 		in_delmulti(ia->ia_allhosts);
830 		ia->ia_allhosts = NULL;
831 	}
832 	mutex_exit(&in_ifaddr_lock);
833 }
834 
835 /*
836  * Depends on it isn't called in concurrent. It should be guaranteed
837  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
838  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
839  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
840  * never be executed in concurrent.
841  */
842 void
843 in_purgeaddr(struct ifaddr *ifa)
844 {
845 	struct in_ifaddr *ia = (void *) ifa;
846 	struct ifnet *ifp = ifa->ifa_ifp;
847 
848 	KASSERT(!ifa_held(ifa));
849 
850 	ifa->ifa_flags |= IFA_DESTROYING;
851 	in_scrubaddr(ia);
852 
853 	mutex_enter(&in_ifaddr_lock);
854 	in_addrhash_remove_locked(ia);
855 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
856 	IN_ADDRLIST_WRITER_REMOVE(ia);
857 	ifa_remove(ifp, &ia->ia_ifa);
858 #ifdef NET_MPSAFE
859 	pserialize_perform(in_ifaddrhash_psz);
860 #endif
861 	mutex_exit(&in_ifaddr_lock);
862 	IN_ADDRHASH_ENTRY_DESTROY(ia);
863 	IN_ADDRLIST_ENTRY_DESTROY(ia);
864 	ifafree(&ia->ia_ifa);
865 	in_setmaxmtu();
866 }
867 
868 static void
869 in_addrhash_insert_locked(struct in_ifaddr *ia)
870 {
871 
872 	KASSERT(mutex_owned(&in_ifaddr_lock));
873 
874 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
875 	    ia_hash);
876 	IN_ADDRHASH_ENTRY_INIT(ia);
877 	IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
878 }
879 
880 void
881 in_addrhash_insert(struct in_ifaddr *ia)
882 {
883 
884 	mutex_enter(&in_ifaddr_lock);
885 	in_addrhash_insert_locked(ia);
886 	mutex_exit(&in_ifaddr_lock);
887 }
888 
889 static void
890 in_addrhash_remove_locked(struct in_ifaddr *ia)
891 {
892 
893 	KASSERT(mutex_owned(&in_ifaddr_lock));
894 
895 	LIST_REMOVE(ia, ia_hash);
896 	IN_ADDRHASH_WRITER_REMOVE(ia);
897 }
898 
899 void
900 in_addrhash_remove(struct in_ifaddr *ia)
901 {
902 
903 	mutex_enter(&in_ifaddr_lock);
904 	in_addrhash_remove_locked(ia);
905 #ifdef NET_MPSAFE
906 	pserialize_perform(in_ifaddrhash_psz);
907 #endif
908 	mutex_exit(&in_ifaddr_lock);
909 	IN_ADDRHASH_ENTRY_DESTROY(ia);
910 }
911 
912 void
913 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
914 {
915 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
916 	igmp_purgeif(ifp);		/* manipulates pools */
917 #ifdef MROUTING
918 	ip_mrouter_detach(ifp);
919 #endif
920 }
921 
922 /*
923  * SIOC[GAD]LIFADDR.
924  *	SIOCGLIFADDR: get first address. (???)
925  *	SIOCGLIFADDR with IFLR_PREFIX:
926  *		get first address that matches the specified prefix.
927  *	SIOCALIFADDR: add the specified address.
928  *	SIOCALIFADDR with IFLR_PREFIX:
929  *		EINVAL since we can't deduce hostid part of the address.
930  *	SIOCDLIFADDR: delete the specified address.
931  *	SIOCDLIFADDR with IFLR_PREFIX:
932  *		delete the first address that matches the specified prefix.
933  * return values:
934  *	EINVAL on invalid parameters
935  *	EADDRNOTAVAIL on prefix match failed/specified address not found
936  *	other values may be returned from in_ioctl()
937  */
938 static int
939 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
940     struct ifnet *ifp)
941 {
942 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
943 	struct ifaddr *ifa;
944 	struct sockaddr *sa;
945 
946 	/* sanity checks */
947 	if (data == NULL || ifp == NULL) {
948 		panic("invalid argument to in_lifaddr_ioctl");
949 		/*NOTRECHED*/
950 	}
951 
952 	switch (cmd) {
953 	case SIOCGLIFADDR:
954 		/* address must be specified on GET with IFLR_PREFIX */
955 		if ((iflr->flags & IFLR_PREFIX) == 0)
956 			break;
957 		/*FALLTHROUGH*/
958 	case SIOCALIFADDR:
959 	case SIOCDLIFADDR:
960 		/* address must be specified on ADD and DELETE */
961 		sa = (struct sockaddr *)&iflr->addr;
962 		if (sa->sa_family != AF_INET)
963 			return EINVAL;
964 		if (sa->sa_len != sizeof(struct sockaddr_in))
965 			return EINVAL;
966 		/* XXX need improvement */
967 		sa = (struct sockaddr *)&iflr->dstaddr;
968 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
969 			return EINVAL;
970 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
971 			return EINVAL;
972 		break;
973 	default: /*shouldn't happen*/
974 #if 0
975 		panic("invalid cmd to in_lifaddr_ioctl");
976 		/*NOTREACHED*/
977 #else
978 		return EOPNOTSUPP;
979 #endif
980 	}
981 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
982 		return EINVAL;
983 
984 	switch (cmd) {
985 	case SIOCALIFADDR:
986 	    {
987 		struct in_aliasreq ifra;
988 
989 		if (iflr->flags & IFLR_PREFIX)
990 			return EINVAL;
991 
992 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
993 		memset(&ifra, 0, sizeof(ifra));
994 		memcpy(ifra.ifra_name, iflr->iflr_name,
995 			sizeof(ifra.ifra_name));
996 
997 		memcpy(&ifra.ifra_addr, &iflr->addr,
998 			((struct sockaddr *)&iflr->addr)->sa_len);
999 
1000 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1001 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1002 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1003 		}
1004 
1005 		ifra.ifra_mask.sin_family = AF_INET;
1006 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
1007 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
1008 
1009 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
1010 	    }
1011 	case SIOCGLIFADDR:
1012 	case SIOCDLIFADDR:
1013 	    {
1014 		struct in_ifaddr *ia;
1015 		struct in_addr mask, candidate, match;
1016 		struct sockaddr_in *sin;
1017 		int cmp, s;
1018 
1019 		memset(&mask, 0, sizeof(mask));
1020 		memset(&match, 0, sizeof(match));	/* XXX gcc */
1021 		if (iflr->flags & IFLR_PREFIX) {
1022 			/* lookup a prefix rather than address. */
1023 			in_len2mask(&mask, iflr->prefixlen);
1024 
1025 			sin = (struct sockaddr_in *)&iflr->addr;
1026 			match.s_addr = sin->sin_addr.s_addr;
1027 			match.s_addr &= mask.s_addr;
1028 
1029 			/* if you set extra bits, that's wrong */
1030 			if (match.s_addr != sin->sin_addr.s_addr)
1031 				return EINVAL;
1032 
1033 			cmp = 1;
1034 		} else {
1035 			if (cmd == SIOCGLIFADDR) {
1036 				/* on getting an address, take the 1st match */
1037 				cmp = 0;	/*XXX*/
1038 			} else {
1039 				/* on deleting an address, do exact match */
1040 				in_len2mask(&mask, 32);
1041 				sin = (struct sockaddr_in *)&iflr->addr;
1042 				match.s_addr = sin->sin_addr.s_addr;
1043 
1044 				cmp = 1;
1045 			}
1046 		}
1047 
1048 		s = pserialize_read_enter();
1049 		IFADDR_READER_FOREACH(ifa, ifp) {
1050 			if (ifa->ifa_addr->sa_family != AF_INET)
1051 				continue;
1052 			if (cmp == 0)
1053 				break;
1054 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
1055 			candidate.s_addr &= mask.s_addr;
1056 			if (candidate.s_addr == match.s_addr)
1057 				break;
1058 		}
1059 		if (ifa == NULL) {
1060 			pserialize_read_exit(s);
1061 			return EADDRNOTAVAIL;
1062 		}
1063 		ia = (struct in_ifaddr *)ifa;
1064 
1065 		if (cmd == SIOCGLIFADDR) {
1066 			/* fill in the if_laddrreq structure */
1067 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1068 
1069 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1070 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1071 					ia->ia_dstaddr.sin_len);
1072 			} else
1073 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1074 
1075 			iflr->prefixlen =
1076 				in_mask2len(&ia->ia_sockmask.sin_addr);
1077 
1078 			iflr->flags = 0;	/*XXX*/
1079 			pserialize_read_exit(s);
1080 
1081 			return 0;
1082 		} else {
1083 			struct in_aliasreq ifra;
1084 
1085 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1086 			memset(&ifra, 0, sizeof(ifra));
1087 			memcpy(ifra.ifra_name, iflr->iflr_name,
1088 				sizeof(ifra.ifra_name));
1089 
1090 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1091 				ia->ia_addr.sin_len);
1092 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1093 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1094 					ia->ia_dstaddr.sin_len);
1095 			}
1096 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1097 				ia->ia_sockmask.sin_len);
1098 			pserialize_read_exit(s);
1099 
1100 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1101 		}
1102 	    }
1103 	}
1104 
1105 	return EOPNOTSUPP;	/*just for safety*/
1106 }
1107 
1108 /*
1109  * Initialize an interface's internet address
1110  * and routing table entry.
1111  */
1112 int
1113 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1114     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1115 {
1116 	u_int32_t i;
1117 	struct sockaddr_in oldaddr, olddst;
1118 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1119 
1120 	if (sin == NULL)
1121 		sin = &ia->ia_addr;
1122 	if (dst == NULL)
1123 		dst = &ia->ia_dstaddr;
1124 
1125 	/*
1126 	 * Set up new addresses.
1127 	 */
1128 	oldaddr = ia->ia_addr;
1129 	olddst = ia->ia_dstaddr;
1130 	oldflags = ia->ia4_flags;
1131 	ia->ia_addr = *sin;
1132 	ia->ia_dstaddr = *dst;
1133 	hostIsNew = oldaddr.sin_family != AF_INET ||
1134 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1135 	if (!scrub)
1136 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1137 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1138 
1139 	/*
1140 	 * Configure address flags.
1141 	 * We need to do this early because they maybe adjusted
1142 	 * by if_addr_init depending on the address.
1143 	 */
1144 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1145 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1146 		hostIsNew = 1;
1147 	}
1148 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1149 		ia->ia4_flags |= IN_IFF_DETACHED;
1150 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1151 	} else if (hostIsNew && if_do_dad(ifp))
1152 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1153 
1154 	/*
1155 	 * Give the interface a chance to initialize
1156 	 * if this is its first address,
1157 	 * and to validate the address if necessary.
1158 	 */
1159 	s = splsoftnet();
1160 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1161 	splx(s);
1162 	/* Now clear the try tentative flag, its job is done. */
1163 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1164 	if (error != 0) {
1165 		ia->ia_addr = oldaddr;
1166 		ia->ia_dstaddr = olddst;
1167 		ia->ia4_flags = oldflags;
1168 		return error;
1169 	}
1170 
1171 	if (scrub || hostIsNew) {
1172 		int newflags = ia->ia4_flags;
1173 
1174 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1175 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1176 		ia->ia4_flags = oldflags;
1177 		if (hostIsNew)
1178 			in_scrubaddr(ia);
1179 		else if (scrub)
1180 			in_scrubprefix(ia);
1181 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1182 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1183 		ia->ia4_flags = newflags;
1184 	}
1185 
1186 	i = ia->ia_addr.sin_addr.s_addr;
1187 	if (ifp->if_flags & IFF_POINTOPOINT)
1188 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1189 	else if (IN_CLASSA(i))
1190 		ia->ia_netmask = IN_CLASSA_NET;
1191 	else if (IN_CLASSB(i))
1192 		ia->ia_netmask = IN_CLASSB_NET;
1193 	else
1194 		ia->ia_netmask = IN_CLASSC_NET;
1195 	/*
1196 	 * The subnet mask usually includes at least the standard network part,
1197 	 * but may may be smaller in the case of supernetting.
1198 	 * If it is set, we believe it.
1199 	 */
1200 	if (ia->ia_subnetmask == 0) {
1201 		ia->ia_subnetmask = ia->ia_netmask;
1202 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1203 	} else
1204 		ia->ia_netmask &= ia->ia_subnetmask;
1205 
1206 	ia->ia_net = i & ia->ia_netmask;
1207 	ia->ia_subnet = i & ia->ia_subnetmask;
1208 	in_socktrim(&ia->ia_sockmask);
1209 
1210 	/* re-calculate the "in_maxmtu" value */
1211 	in_setmaxmtu();
1212 
1213 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1214 	if (ifp->if_flags & IFF_BROADCAST) {
1215 		ia->ia_broadaddr.sin_addr.s_addr =
1216 			ia->ia_subnet | ~ia->ia_subnetmask;
1217 		ia->ia_netbroadcast.s_addr =
1218 			ia->ia_net | ~ia->ia_netmask;
1219 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1220 		ia->ia_dstaddr = ia->ia_addr;
1221 		flags |= RTF_HOST;
1222 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1223 		if (ia->ia_dstaddr.sin_family != AF_INET)
1224 			return (0);
1225 		flags |= RTF_HOST;
1226 	}
1227 
1228 	/* Add the local route to the address */
1229 	in_ifaddlocal(&ia->ia_ifa);
1230 
1231 	/* Add the prefix route for the address */
1232 	error = in_addprefix(ia, flags);
1233 
1234 	/*
1235 	 * If the interface supports multicast, join the "all hosts"
1236 	 * multicast group on that interface.
1237 	 */
1238 	mutex_enter(&in_ifaddr_lock);
1239 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1240 		struct in_addr addr;
1241 
1242 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1243 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1244 	}
1245 	mutex_exit(&in_ifaddr_lock);
1246 
1247 	if (hostIsNew &&
1248 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1249 	    if_do_dad(ifp))
1250 		ia->ia_dad_start((struct ifaddr *)ia);
1251 
1252 	return error;
1253 }
1254 
1255 #define rtinitflags(x) \
1256 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1257 	    ? RTF_HOST : 0)
1258 
1259 /*
1260  * add a route to prefix ("connected route" in cisco terminology).
1261  * does nothing if there's some interface address with the same prefix already.
1262  */
1263 static int
1264 in_addprefix(struct in_ifaddr *target, int flags)
1265 {
1266 	struct in_ifaddr *ia;
1267 	struct in_addr prefix, mask, p;
1268 	int error;
1269 	int s;
1270 
1271 	if ((flags & RTF_HOST) != 0)
1272 		prefix = target->ia_dstaddr.sin_addr;
1273 	else {
1274 		prefix = target->ia_addr.sin_addr;
1275 		mask = target->ia_sockmask.sin_addr;
1276 		prefix.s_addr &= mask.s_addr;
1277 	}
1278 
1279 	s = pserialize_read_enter();
1280 	IN_ADDRLIST_READER_FOREACH(ia) {
1281 		if (rtinitflags(ia))
1282 			p = ia->ia_dstaddr.sin_addr;
1283 		else {
1284 			p = ia->ia_addr.sin_addr;
1285 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1286 		}
1287 
1288 		if (prefix.s_addr != p.s_addr)
1289 			continue;
1290 
1291 		/*
1292 		 * if we got a matching prefix route inserted by other
1293 		 * interface address, we don't need to bother
1294 		 *
1295 		 * XXX RADIX_MPATH implications here? -dyoung
1296 		 */
1297 		if (ia->ia_flags & IFA_ROUTE) {
1298 			pserialize_read_exit(s);
1299 			return 0;
1300 		}
1301 	}
1302 	pserialize_read_exit(s);
1303 
1304 	/*
1305 	 * noone seem to have prefix route.  insert it.
1306 	 */
1307 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1308 	if (error == 0)
1309 		target->ia_flags |= IFA_ROUTE;
1310 	else if (error == EEXIST) {
1311 		/*
1312 		 * the fact the route already exists is not an error.
1313 		 */
1314 		error = 0;
1315 	}
1316 	return error;
1317 }
1318 
1319 /*
1320  * remove a route to prefix ("connected route" in cisco terminology).
1321  * re-installs the route by using another interface address, if there's one
1322  * with the same prefix (otherwise we lose the route mistakenly).
1323  */
1324 static int
1325 in_scrubprefix(struct in_ifaddr *target)
1326 {
1327 	struct in_ifaddr *ia;
1328 	struct in_addr prefix, mask, p;
1329 	int error;
1330 	int s;
1331 
1332 	/* If we don't have IFA_ROUTE we have nothing to do */
1333 	if ((target->ia_flags & IFA_ROUTE) == 0)
1334 		return 0;
1335 
1336 	if (rtinitflags(target))
1337 		prefix = target->ia_dstaddr.sin_addr;
1338 	else {
1339 		prefix = target->ia_addr.sin_addr;
1340 		mask = target->ia_sockmask.sin_addr;
1341 		prefix.s_addr &= mask.s_addr;
1342 	}
1343 
1344 	s = pserialize_read_enter();
1345 	IN_ADDRLIST_READER_FOREACH(ia) {
1346 		if (rtinitflags(ia))
1347 			p = ia->ia_dstaddr.sin_addr;
1348 		else {
1349 			p = ia->ia_addr.sin_addr;
1350 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1351 		}
1352 
1353 		if (prefix.s_addr != p.s_addr)
1354 			continue;
1355 
1356 		/*
1357 		 * if we got a matching prefix route, move IFA_ROUTE to him
1358 		 */
1359 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1360 			struct psref psref;
1361 			int bound = curlwp_bind();
1362 
1363 			ia4_acquire(ia, &psref);
1364 			pserialize_read_exit(s);
1365 
1366 			rtinit(&target->ia_ifa, RTM_DELETE,
1367 			    rtinitflags(target));
1368 			target->ia_flags &= ~IFA_ROUTE;
1369 
1370 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1371 			    rtinitflags(ia) | RTF_UP);
1372 			if (error == 0)
1373 				ia->ia_flags |= IFA_ROUTE;
1374 
1375 			ia4_release(ia, &psref);
1376 			curlwp_bindx(bound);
1377 
1378 			return error;
1379 		}
1380 	}
1381 	pserialize_read_exit(s);
1382 
1383 	/*
1384 	 * noone seem to have prefix route.  remove it.
1385 	 */
1386 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1387 	target->ia_flags &= ~IFA_ROUTE;
1388 	return 0;
1389 }
1390 
1391 #undef rtinitflags
1392 
1393 /*
1394  * Return 1 if the address might be a local broadcast address.
1395  */
1396 int
1397 in_broadcast(struct in_addr in, struct ifnet *ifp)
1398 {
1399 	struct ifaddr *ifa;
1400 	int s;
1401 
1402 	KASSERT(ifp != NULL);
1403 
1404 	if (in.s_addr == INADDR_BROADCAST ||
1405 	    in_nullhost(in))
1406 		return 1;
1407 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1408 		return 0;
1409 	/*
1410 	 * Look through the list of addresses for a match
1411 	 * with a broadcast address.
1412 	 */
1413 #define ia (ifatoia(ifa))
1414 	s = pserialize_read_enter();
1415 	IFADDR_READER_FOREACH(ifa, ifp) {
1416 		if (ifa->ifa_addr->sa_family == AF_INET &&
1417 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1418 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1419 		     in_hosteq(in, ia->ia_netbroadcast) ||
1420 		     (hostzeroisbroadcast &&
1421 		      /*
1422 		       * Check for old-style (host 0) broadcast.
1423 		       */
1424 		      (in.s_addr == ia->ia_subnet ||
1425 		       in.s_addr == ia->ia_net)))) {
1426 			pserialize_read_exit(s);
1427 			return 1;
1428 		}
1429 	}
1430 	pserialize_read_exit(s);
1431 	return (0);
1432 #undef ia
1433 }
1434 
1435 /*
1436  * perform DAD when interface becomes IFF_UP.
1437  */
1438 void
1439 in_if_link_up(struct ifnet *ifp)
1440 {
1441 	struct ifaddr *ifa;
1442 	struct in_ifaddr *ia;
1443 	int s, bound;
1444 
1445 	/* Ensure it's sane to run DAD */
1446 	if (ifp->if_link_state == LINK_STATE_DOWN)
1447 		return;
1448 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1449 		return;
1450 
1451 	bound = curlwp_bind();
1452 	s = pserialize_read_enter();
1453 	IFADDR_READER_FOREACH(ifa, ifp) {
1454 		struct psref psref;
1455 
1456 		if (ifa->ifa_addr->sa_family != AF_INET)
1457 			continue;
1458 		ifa_acquire(ifa, &psref);
1459 		pserialize_read_exit(s);
1460 
1461 		ia = (struct in_ifaddr *)ifa;
1462 
1463 		/* If detached then mark as tentative */
1464 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1465 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1466 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1467 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1468 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1469 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1470 		}
1471 
1472 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1473 			/* Clear the duplicated flag as we're starting DAD. */
1474 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1475 			ia->ia_dad_start(ifa);
1476 		}
1477 
1478 		s = pserialize_read_enter();
1479 		ifa_release(ifa, &psref);
1480 	}
1481 	pserialize_read_exit(s);
1482 	curlwp_bindx(bound);
1483 }
1484 
1485 void
1486 in_if_up(struct ifnet *ifp)
1487 {
1488 
1489 	/* interface may not support link state, so bring it up also */
1490 	in_if_link_up(ifp);
1491 }
1492 
1493 /*
1494  * Mark all addresses as detached.
1495  */
1496 void
1497 in_if_link_down(struct ifnet *ifp)
1498 {
1499 	struct ifaddr *ifa;
1500 	struct in_ifaddr *ia;
1501 	int s, bound;
1502 
1503 	bound = curlwp_bind();
1504 	s = pserialize_read_enter();
1505 	IFADDR_READER_FOREACH(ifa, ifp) {
1506 		struct psref psref;
1507 
1508 		if (ifa->ifa_addr->sa_family != AF_INET)
1509 			continue;
1510 		ifa_acquire(ifa, &psref);
1511 		pserialize_read_exit(s);
1512 
1513 		ia = (struct in_ifaddr *)ifa;
1514 
1515 		/* Stop DAD processing */
1516 		if (ia->ia_dad_stop != NULL)
1517 			ia->ia_dad_stop(ifa);
1518 
1519 		/*
1520 		 * Mark the address as detached.
1521 		 */
1522 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1523 			ia->ia4_flags |= IN_IFF_DETACHED;
1524 			ia->ia4_flags &=
1525 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1526 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1527 		}
1528 
1529 		s = pserialize_read_enter();
1530 		ifa_release(ifa, &psref);
1531 	}
1532 	pserialize_read_exit(s);
1533 	curlwp_bindx(bound);
1534 }
1535 
1536 void
1537 in_if_down(struct ifnet *ifp)
1538 {
1539 
1540 	in_if_link_down(ifp);
1541 	lltable_purge_entries(LLTABLE(ifp));
1542 }
1543 
1544 void
1545 in_if_link_state_change(struct ifnet *ifp, int link_state)
1546 {
1547 
1548 	switch (link_state) {
1549 	case LINK_STATE_DOWN:
1550 		in_if_link_down(ifp);
1551 		break;
1552 	case LINK_STATE_UP:
1553 		in_if_link_up(ifp);
1554 		break;
1555 	}
1556 }
1557 
1558 /*
1559  * in_lookup_multi: look up the in_multi record for a given IP
1560  * multicast address on a given interface.  If no matching record is
1561  * found, return NULL.
1562  */
1563 struct in_multi *
1564 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1565 {
1566 	struct in_multi *inm;
1567 
1568 	KASSERT(rw_lock_held(&in_multilock));
1569 
1570 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1571 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1572 			break;
1573 	}
1574 	return inm;
1575 }
1576 
1577 /*
1578  * in_multi_group: check whether the address belongs to an IP multicast
1579  * group we are joined on this interface.  Returns true or false.
1580  */
1581 bool
1582 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1583 {
1584 	bool ingroup;
1585 
1586 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1587 		rw_enter(&in_multilock, RW_READER);
1588 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1589 		rw_exit(&in_multilock);
1590 	} else {
1591 		/* XXX Recursive call from ip_output(). */
1592 		KASSERT(rw_lock_held(&in_multilock));
1593 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1594 	}
1595 	return ingroup;
1596 }
1597 
1598 /*
1599  * Add an address to the list of IP multicast addresses for a given interface.
1600  */
1601 struct in_multi *
1602 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1603 {
1604 	struct sockaddr_in sin;
1605 	struct in_multi *inm;
1606 
1607 	/*
1608 	 * See if address already in list.
1609 	 */
1610 	rw_enter(&in_multilock, RW_WRITER);
1611 	inm = in_lookup_multi(*ap, ifp);
1612 	if (inm != NULL) {
1613 		/*
1614 		 * Found it; just increment the reference count.
1615 		 */
1616 		inm->inm_refcount++;
1617 		rw_exit(&in_multilock);
1618 		return inm;
1619 	}
1620 
1621 	/*
1622 	 * New address; allocate a new multicast record.
1623 	 */
1624 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1625 	if (inm == NULL) {
1626 		rw_exit(&in_multilock);
1627 		return NULL;
1628 	}
1629 	inm->inm_addr = *ap;
1630 	inm->inm_ifp = ifp;
1631 	inm->inm_refcount = 1;
1632 
1633 	/*
1634 	 * Ask the network driver to update its multicast reception
1635 	 * filter appropriately for the new address.
1636 	 */
1637 	sockaddr_in_init(&sin, ap, 0);
1638 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1639 		rw_exit(&in_multilock);
1640 		pool_put(&inmulti_pool, inm);
1641 		return NULL;
1642 	}
1643 
1644 	/*
1645 	 * Let IGMP know that we have joined a new IP multicast group.
1646 	 */
1647 	if (igmp_joingroup(inm) != 0) {
1648 		rw_exit(&in_multilock);
1649 		pool_put(&inmulti_pool, inm);
1650 		return NULL;
1651 	}
1652 	LIST_INSERT_HEAD(
1653 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1654 	    inm, inm_list);
1655 	in_multientries++;
1656 	rw_exit(&in_multilock);
1657 
1658 	return inm;
1659 }
1660 
1661 /*
1662  * Delete a multicast address record.
1663  */
1664 void
1665 in_delmulti(struct in_multi *inm)
1666 {
1667 	struct sockaddr_in sin;
1668 
1669 	rw_enter(&in_multilock, RW_WRITER);
1670 	if (--inm->inm_refcount > 0) {
1671 		rw_exit(&in_multilock);
1672 		return;
1673 	}
1674 
1675 	/*
1676 	 * No remaining claims to this record; let IGMP know that
1677 	 * we are leaving the multicast group.
1678 	 */
1679 	igmp_leavegroup(inm);
1680 
1681 	/*
1682 	 * Notify the network driver to update its multicast reception
1683 	 * filter.
1684 	 */
1685 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1686 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1687 
1688 	/*
1689 	 * Unlink from list.
1690 	 */
1691 	LIST_REMOVE(inm, inm_list);
1692 	in_multientries--;
1693 	rw_exit(&in_multilock);
1694 
1695 	pool_put(&inmulti_pool, inm);
1696 }
1697 
1698 /*
1699  * in_next_multi: step through all of the in_multi records, one at a time.
1700  * The current position is remembered in "step", which the caller must
1701  * provide.  in_first_multi(), below, must be called to initialize "step"
1702  * and get the first record.  Both macros return a NULL "inm" when there
1703  * are no remaining records.
1704  */
1705 struct in_multi *
1706 in_next_multi(struct in_multistep *step)
1707 {
1708 	struct in_multi *inm;
1709 
1710 	KASSERT(rw_lock_held(&in_multilock));
1711 
1712 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1713 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1714 	}
1715 	if ((inm = step->i_inm) != NULL) {
1716 		step->i_inm = LIST_NEXT(inm, inm_list);
1717 	}
1718 	return inm;
1719 }
1720 
1721 struct in_multi *
1722 in_first_multi(struct in_multistep *step)
1723 {
1724 	KASSERT(rw_lock_held(&in_multilock));
1725 
1726 	step->i_n = 0;
1727 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1728 	return in_next_multi(step);
1729 }
1730 
1731 void
1732 in_multi_lock(int op)
1733 {
1734 	rw_enter(&in_multilock, op);
1735 }
1736 
1737 void
1738 in_multi_unlock(void)
1739 {
1740 	rw_exit(&in_multilock);
1741 }
1742 
1743 int
1744 in_multi_lock_held(void)
1745 {
1746 	return rw_lock_held(&in_multilock);
1747 }
1748 
1749 struct in_ifaddr *
1750 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1751     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1752 {
1753 	struct rtentry *rt = NULL;
1754 	struct in_ifaddr *ia = NULL;
1755 
1756 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1757 	/*
1758          * If route is known or can be allocated now, take the
1759          * source address from the interface.  Otherwise, punt.
1760 	 */
1761 	if ((soopts & SO_DONTROUTE) != 0)
1762 		rtcache_free(ro);
1763 	else {
1764 		union {
1765 			struct sockaddr		dst;
1766 			struct sockaddr_in	dst4;
1767 		} u;
1768 
1769 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1770 		rt = rtcache_lookup(ro, &u.dst);
1771 	}
1772 	/*
1773 	 * If we found a route, use the address
1774 	 * corresponding to the outgoing interface
1775 	 * unless it is the loopback (in case a route
1776 	 * to our address on another net goes to loopback).
1777 	 *
1778 	 * XXX Is this still true?  Do we care?
1779 	 */
1780 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1781 		int s;
1782 		struct ifaddr *ifa;
1783 		/*
1784 		 * Just in case. May not need to do this workaround.
1785 		 * Revisit when working on rtentry MP-ification.
1786 		 */
1787 		s = pserialize_read_enter();
1788 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1789 			if (ifa == rt->rt_ifa)
1790 				break;
1791 		}
1792 		if (ifa != NULL)
1793 			ifa_acquire(ifa, psref);
1794 		pserialize_read_exit(s);
1795 
1796 		ia = ifatoia(ifa);
1797 	}
1798 	if (ia == NULL) {
1799 		u_int16_t fport = sin->sin_port;
1800 		struct ifaddr *ifa;
1801 		int s;
1802 
1803 		sin->sin_port = 0;
1804 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1805 		sin->sin_port = fport;
1806 		if (ifa == NULL) {
1807 			/* Find 1st non-loopback AF_INET address */
1808 			s = pserialize_read_enter();
1809 			IN_ADDRLIST_READER_FOREACH(ia) {
1810 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1811 					break;
1812 			}
1813 			if (ia != NULL)
1814 				ia4_acquire(ia, psref);
1815 			pserialize_read_exit(s);
1816 		} else {
1817 			/* ia is already referenced by psref */
1818 			ia = ifatoia(ifa);
1819 		}
1820 		if (ia == NULL) {
1821 			*errorp = EADDRNOTAVAIL;
1822 			goto out;
1823 		}
1824 	}
1825 	/*
1826 	 * If the destination address is multicast and an outgoing
1827 	 * interface has been set as a multicast option, use the
1828 	 * address of that interface as our source address.
1829 	 */
1830 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1831 		struct ip_moptions *imo;
1832 
1833 		imo = mopts;
1834 		if (imo->imo_multicast_if_index != 0) {
1835 			struct ifnet *ifp;
1836 			int s;
1837 
1838 			if (ia != NULL)
1839 				ia4_release(ia, psref);
1840 			s = pserialize_read_enter();
1841 			ifp = if_byindex(imo->imo_multicast_if_index);
1842 			if (ifp != NULL) {
1843 				/* XXX */
1844 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1845 			} else
1846 				ia = NULL;
1847 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1848 				pserialize_read_exit(s);
1849 				if (ia != NULL)
1850 					ia4_release(ia, psref);
1851 				*errorp = EADDRNOTAVAIL;
1852 				ia = NULL;
1853 				goto out;
1854 			}
1855 			pserialize_read_exit(s);
1856 		}
1857 	}
1858 	if (ia->ia_ifa.ifa_getifa != NULL) {
1859 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1860 		                                      sintosa(sin)));
1861 		if (ia == NULL) {
1862 			*errorp = EADDRNOTAVAIL;
1863 			goto out;
1864 		}
1865 		/* FIXME NOMPSAFE */
1866 		ia4_acquire(ia, psref);
1867 	}
1868 #ifdef GETIFA_DEBUG
1869 	else
1870 		printf("%s: missing ifa_getifa\n", __func__);
1871 #endif
1872 out:
1873 	rtcache_unref(rt, ro);
1874 	return ia;
1875 }
1876 
1877 #if NARP > 0
1878 
1879 struct in_llentry {
1880 	struct llentry		base;
1881 };
1882 
1883 #define	IN_LLTBL_DEFAULT_HSIZE	32
1884 #define	IN_LLTBL_HASH(k, h) \
1885 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1886 
1887 /*
1888  * Do actual deallocation of @lle.
1889  * Called by LLE_FREE_LOCKED when number of references
1890  * drops to zero.
1891  */
1892 static void
1893 in_lltable_destroy_lle(struct llentry *lle)
1894 {
1895 
1896 	LLE_WUNLOCK(lle);
1897 	LLE_LOCK_DESTROY(lle);
1898 	kmem_intr_free(lle, sizeof(*lle));
1899 }
1900 
1901 static struct llentry *
1902 in_lltable_new(struct in_addr addr4, u_int flags)
1903 {
1904 	struct in_llentry *lle;
1905 
1906 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
1907 	if (lle == NULL)		/* NB: caller generates msg */
1908 		return NULL;
1909 
1910 	/*
1911 	 * For IPv4 this will trigger "arpresolve" to generate
1912 	 * an ARP request.
1913 	 */
1914 	lle->base.la_expire = time_uptime; /* mark expired */
1915 	lle->base.r_l3addr.addr4 = addr4;
1916 	lle->base.lle_refcnt = 1;
1917 	lle->base.lle_free = in_lltable_destroy_lle;
1918 	LLE_LOCK_INIT(&lle->base);
1919 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1920 
1921 	return (&lle->base);
1922 }
1923 
1924 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1925 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1926 
1927 static int
1928 in_lltable_match_prefix(const struct sockaddr *prefix,
1929     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1930 {
1931 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1932 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1933 	struct in_addr lle_addr;
1934 
1935 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1936 
1937 	/*
1938 	 * (flags & LLE_STATIC) means deleting all entries
1939 	 * including static ARP entries.
1940 	 */
1941 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) &&
1942 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1943 		return (1);
1944 
1945 	return (0);
1946 }
1947 
1948 static void
1949 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1950 {
1951 	struct ifnet *ifp __diagused;
1952 	size_t pkts_dropped;
1953 
1954 	LLE_WLOCK_ASSERT(lle);
1955 	KASSERT(llt != NULL);
1956 
1957 	/* Unlink entry from table if not already */
1958 	if ((lle->la_flags & LLE_LINKED) != 0) {
1959 		ifp = llt->llt_ifp;
1960 		IF_AFDATA_WLOCK_ASSERT(ifp);
1961 		lltable_unlink_entry(llt, lle);
1962 	}
1963 
1964 	/* cancel timer */
1965 	if (callout_halt(&lle->lle_timer, &lle->lle_lock))
1966 		LLE_REMREF(lle);
1967 
1968 	/* Drop hold queue */
1969 	pkts_dropped = llentry_free(lle);
1970 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
1971 }
1972 
1973 static int
1974 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr,
1975     const struct rtentry *rt)
1976 {
1977 	int error = EINVAL;
1978 
1979 	if (rt == NULL)
1980 		return error;
1981 
1982 	/*
1983 	 * If the gateway for an existing host route matches the target L3
1984 	 * address, which is a special route inserted by some implementation
1985 	 * such as MANET, and the interface is of the correct type, then
1986 	 * allow for ARP to proceed.
1987 	 */
1988 	if (rt->rt_flags & RTF_GATEWAY) {
1989 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1990 		    rt->rt_ifp->if_type != IFT_ETHER ||
1991 #ifdef __FreeBSD__
1992 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1993 #else
1994 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
1995 #endif
1996 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1997 		    sizeof(in_addr_t)) != 0) {
1998 			goto error;
1999 		}
2000 	}
2001 
2002 	/*
2003 	 * Make sure that at least the destination address is covered
2004 	 * by the route. This is for handling the case where 2 or more
2005 	 * interfaces have the same prefix. An incoming packet arrives
2006 	 * on one interface and the corresponding outgoing packet leaves
2007 	 * another interface.
2008 	 */
2009 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
2010 		const char *sa, *mask, *addr, *lim;
2011 		int len;
2012 
2013 		mask = (const char *)rt_mask(rt);
2014 		/*
2015 		 * Just being extra cautious to avoid some custom
2016 		 * code getting into trouble.
2017 		 */
2018 		if (mask == NULL)
2019 			goto error;
2020 
2021 		sa = (const char *)rt_getkey(rt);
2022 		addr = (const char *)l3addr;
2023 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
2024 		lim = addr + len;
2025 
2026 		for ( ; addr < lim; sa++, mask++, addr++) {
2027 			if ((*sa ^ *addr) & *mask) {
2028 #ifdef DIAGNOSTIC
2029 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
2030 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
2031 #endif
2032 				goto error;
2033 			}
2034 		}
2035 	}
2036 
2037 	error = 0;
2038 error:
2039 	return error;
2040 }
2041 
2042 static inline uint32_t
2043 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
2044 {
2045 
2046 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
2047 }
2048 
2049 static uint32_t
2050 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
2051 {
2052 
2053 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
2054 }
2055 
2056 static void
2057 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2058 {
2059 	struct sockaddr_in *sin;
2060 
2061 	sin = (struct sockaddr_in *)sa;
2062 	memset(sin, 0, sizeof(*sin));
2063 	sin->sin_family = AF_INET;
2064 	sin->sin_len = sizeof(*sin);
2065 	sin->sin_addr = lle->r_l3addr.addr4;
2066 }
2067 
2068 static inline struct llentry *
2069 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2070 {
2071 	struct llentry *lle;
2072 	struct llentries *lleh;
2073 	u_int hashidx;
2074 
2075 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2076 	lleh = &llt->lle_head[hashidx];
2077 	LIST_FOREACH(lle, lleh, lle_next) {
2078 		if (lle->la_flags & LLE_DELETED)
2079 			continue;
2080 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2081 			break;
2082 	}
2083 
2084 	return (lle);
2085 }
2086 
2087 static int
2088 in_lltable_delete(struct lltable *llt, u_int flags,
2089     const struct sockaddr *l3addr)
2090 {
2091 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2092 	struct ifnet *ifp __diagused = llt->llt_ifp;
2093 	struct llentry *lle;
2094 
2095 	IF_AFDATA_WLOCK_ASSERT(ifp);
2096 	KASSERTMSG(l3addr->sa_family == AF_INET,
2097 	    "sin_family %d", l3addr->sa_family);
2098 
2099 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2100 	if (lle == NULL) {
2101 #ifdef DEBUG
2102 		char buf[64];
2103 		sockaddr_format(l3addr, buf, sizeof(buf));
2104 		log(LOG_INFO, "%s: cache for %s is not found\n",
2105 		    __func__, buf);
2106 #endif
2107 		return (ENOENT);
2108 	}
2109 
2110 	LLE_WLOCK(lle);
2111 	lle->la_flags |= LLE_DELETED;
2112 #ifdef DEBUG
2113 	{
2114 		char buf[64];
2115 		sockaddr_format(l3addr, buf, sizeof(buf));
2116 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2117 		    __func__, buf, lle);
2118 	}
2119 #endif
2120 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2121 		llentry_free(lle);
2122 	else
2123 		LLE_WUNLOCK(lle);
2124 
2125 	return (0);
2126 }
2127 
2128 static struct llentry *
2129 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr,
2130     const struct rtentry *rt)
2131 {
2132 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2133 	struct ifnet *ifp = llt->llt_ifp;
2134 	struct llentry *lle;
2135 
2136 	IF_AFDATA_WLOCK_ASSERT(ifp);
2137 	KASSERTMSG(l3addr->sa_family == AF_INET,
2138 	    "sin_family %d", l3addr->sa_family);
2139 
2140 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2141 
2142 	if (lle != NULL) {
2143 		LLE_WLOCK(lle);
2144 		return (lle);
2145 	}
2146 
2147 	/* no existing record, we need to create new one */
2148 
2149 	/*
2150 	 * A route that covers the given address must have
2151 	 * been installed 1st because we are doing a resolution,
2152 	 * verify this.
2153 	 */
2154 	if (!(flags & LLE_IFADDR) &&
2155 	    in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2156 		return (NULL);
2157 
2158 	lle = in_lltable_new(sin->sin_addr, flags);
2159 	if (lle == NULL) {
2160 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2161 		return (NULL);
2162 	}
2163 	lle->la_flags = flags;
2164 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2165 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2166 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2167 	}
2168 
2169 	lltable_link_entry(llt, lle);
2170 	LLE_WLOCK(lle);
2171 
2172 	return (lle);
2173 }
2174 
2175 /*
2176  * Return NULL if not found or marked for deletion.
2177  * If found return lle read locked.
2178  */
2179 static struct llentry *
2180 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2181 {
2182 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2183 	struct llentry *lle;
2184 
2185 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2186 	KASSERTMSG(l3addr->sa_family == AF_INET,
2187 	    "sin_family %d", l3addr->sa_family);
2188 
2189 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2190 
2191 	if (lle == NULL)
2192 		return NULL;
2193 
2194 	if (flags & LLE_EXCLUSIVE)
2195 		LLE_WLOCK(lle);
2196 	else
2197 		LLE_RLOCK(lle);
2198 
2199 	return lle;
2200 }
2201 
2202 static int
2203 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2204     struct rt_walkarg *w)
2205 {
2206 	struct sockaddr_in sin;
2207 
2208 	LLTABLE_LOCK_ASSERT();
2209 
2210 	/* skip deleted entries */
2211 	if (lle->la_flags & LLE_DELETED)
2212 		return 0;
2213 
2214 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2215 
2216 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2217 }
2218 
2219 #endif /* NARP > 0 */
2220 
2221 static int
2222 in_multicast_sysctl(SYSCTLFN_ARGS)
2223 {
2224 	struct ifnet *ifp;
2225 	struct ifaddr *ifa;
2226 	struct in_ifaddr *ifa4;
2227 	struct in_multi *inm;
2228 	uint32_t tmp;
2229 	int error;
2230 	size_t written;
2231 	struct psref psref;
2232 	int bound;
2233 
2234 	if (namelen != 1)
2235 		return EINVAL;
2236 
2237 	bound = curlwp_bind();
2238 	ifp = if_get_byindex(name[0], &psref);
2239 	if (ifp == NULL) {
2240 		curlwp_bindx(bound);
2241 		return ENODEV;
2242 	}
2243 
2244 	if (oldp == NULL) {
2245 		*oldlenp = 0;
2246 		IFADDR_FOREACH(ifa, ifp) {
2247 			if (ifa->ifa_addr->sa_family != AF_INET)
2248 				continue;
2249 			ifa4 = (void *)ifa;
2250 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2251 				*oldlenp += 2 * sizeof(struct in_addr) +
2252 				    sizeof(uint32_t);
2253 			}
2254 		}
2255 		if_put(ifp, &psref);
2256 		curlwp_bindx(bound);
2257 		return 0;
2258 	}
2259 
2260 	error = 0;
2261 	written = 0;
2262 	IFADDR_FOREACH(ifa, ifp) {
2263 		if (ifa->ifa_addr->sa_family != AF_INET)
2264 			continue;
2265 		ifa4 = (void *)ifa;
2266 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2267 			if (written + 2 * sizeof(struct in_addr) +
2268 			    sizeof(uint32_t) > *oldlenp)
2269 				goto done;
2270 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2271 			    oldp, sizeof(struct in_addr));
2272 			if (error)
2273 				goto done;
2274 			oldp = (char *)oldp + sizeof(struct in_addr);
2275 			written += sizeof(struct in_addr);
2276 			error = sysctl_copyout(l, &inm->inm_addr,
2277 			    oldp, sizeof(struct in_addr));
2278 			if (error)
2279 				goto done;
2280 			oldp = (char *)oldp + sizeof(struct in_addr);
2281 			written += sizeof(struct in_addr);
2282 			tmp = inm->inm_refcount;
2283 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2284 			if (error)
2285 				goto done;
2286 			oldp = (char *)oldp + sizeof(tmp);
2287 			written += sizeof(tmp);
2288 		}
2289 	}
2290 done:
2291 	if_put(ifp, &psref);
2292 	curlwp_bindx(bound);
2293 	*oldlenp = written;
2294 	return error;
2295 }
2296 
2297 static void
2298 in_sysctl_init(struct sysctllog **clog)
2299 {
2300 	sysctl_createv(clog, 0, NULL, NULL,
2301 		       CTLFLAG_PERMANENT,
2302 		       CTLTYPE_NODE, "inet",
2303 		       SYSCTL_DESCR("PF_INET related settings"),
2304 		       NULL, 0, NULL, 0,
2305 		       CTL_NET, PF_INET, CTL_EOL);
2306 	sysctl_createv(clog, 0, NULL, NULL,
2307 		       CTLFLAG_PERMANENT,
2308 		       CTLTYPE_NODE, "multicast",
2309 		       SYSCTL_DESCR("Multicast information"),
2310 		       in_multicast_sysctl, 0, NULL, 0,
2311 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2312 	sysctl_createv(clog, 0, NULL, NULL,
2313 		       CTLFLAG_PERMANENT,
2314 		       CTLTYPE_NODE, "ip",
2315 		       SYSCTL_DESCR("IPv4 related settings"),
2316 		       NULL, 0, NULL, 0,
2317 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2318 
2319 	sysctl_createv(clog, 0, NULL, NULL,
2320 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2321 		       CTLTYPE_INT, "subnetsarelocal",
2322 		       SYSCTL_DESCR("Whether logical subnets are considered "
2323 				    "local"),
2324 		       NULL, 0, &subnetsarelocal, 0,
2325 		       CTL_NET, PF_INET, IPPROTO_IP,
2326 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2327 	sysctl_createv(clog, 0, NULL, NULL,
2328 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2329 		       CTLTYPE_INT, "hostzerobroadcast",
2330 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2331 		       NULL, 0, &hostzeroisbroadcast, 0,
2332 		       CTL_NET, PF_INET, IPPROTO_IP,
2333 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2334 }
2335 
2336 #if NARP > 0
2337 
2338 static struct lltable *
2339 in_lltattach(struct ifnet *ifp)
2340 {
2341 	struct lltable *llt;
2342 
2343 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2344 	llt->llt_af = AF_INET;
2345 	llt->llt_ifp = ifp;
2346 
2347 	llt->llt_lookup = in_lltable_lookup;
2348 	llt->llt_create = in_lltable_create;
2349 	llt->llt_delete = in_lltable_delete;
2350 	llt->llt_dump_entry = in_lltable_dump_entry;
2351 	llt->llt_hash = in_lltable_hash;
2352 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2353 	llt->llt_free_entry = in_lltable_free_entry;
2354 	llt->llt_match_prefix = in_lltable_match_prefix;
2355 	lltable_link(llt);
2356 
2357 	return (llt);
2358 }
2359 
2360 #endif /* NARP > 0 */
2361 
2362 void *
2363 in_domifattach(struct ifnet *ifp)
2364 {
2365 	struct in_ifinfo *ii;
2366 
2367 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2368 
2369 #if NARP > 0
2370 	ii->ii_llt = in_lltattach(ifp);
2371 #endif
2372 
2373 #ifdef IPSELSRC
2374 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2375 	KASSERT(ii->ii_selsrc != NULL);
2376 #endif
2377 
2378 	return ii;
2379 }
2380 
2381 void
2382 in_domifdetach(struct ifnet *ifp, void *aux)
2383 {
2384 	struct in_ifinfo *ii = aux;
2385 
2386 #ifdef IPSELSRC
2387 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2388 #endif
2389 #if NARP > 0
2390 	lltable_free(ii->ii_llt);
2391 #endif
2392 	kmem_free(ii, sizeof(struct in_ifinfo));
2393 }
2394