xref: /netbsd-src/sys/netinet/in.c (revision 53d1339bf7f9c7367b35a9e1ebe693f9b047a47b)
1 /*	$NetBSD: in.c,v 1.241 2020/09/29 19:33:36 roy 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 1.241 2020/09/29 19:33:36 roy 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 int	in_lifaddr_ioctl(struct socket *, u_long, void *,
149 	struct ifnet *);
150 
151 static void	in_addrhash_insert_locked(struct in_ifaddr *);
152 static void	in_addrhash_remove_locked(struct in_ifaddr *);
153 
154 static int	in_addprefix(struct in_ifaddr *, int);
155 static void	in_scrubaddr(struct in_ifaddr *);
156 static int	in_scrubprefix(struct in_ifaddr *);
157 static void	in_sysctl_init(struct sysctllog **);
158 
159 #ifndef SUBNETSARELOCAL
160 #define	SUBNETSARELOCAL	1
161 #endif
162 
163 #ifndef HOSTZEROBROADCAST
164 #define HOSTZEROBROADCAST 0
165 #endif
166 
167 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
168 #ifndef IN_MULTI_HASH_SIZE
169 #define IN_MULTI_HASH_SIZE	509
170 #endif
171 
172 static int			subnetsarelocal = SUBNETSARELOCAL;
173 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
174 
175 /*
176  * This list is used to keep track of in_multi chains which belong to
177  * deleted interface addresses.  We use in_ifaddr so that a chain head
178  * won't be deallocated until all multicast address record are deleted.
179  */
180 
181 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
182 
183 static struct pool		inmulti_pool;
184 static u_int			in_multientries;
185 static struct in_multihashhead *in_multihashtbl;
186 static u_long			in_multihash;
187 static krwlock_t		in_multilock;
188 
189 #define IN_MULTI_HASH(x, ifp) \
190     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
191 
192 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */
193 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
194 u_long				in_ifaddrhash;
195 struct in_ifaddrhead		in_ifaddrhead;
196 static kmutex_t			in_ifaddr_lock;
197 
198 pserialize_t			in_ifaddrhash_psz;
199 struct pslist_head *		in_ifaddrhashtbl_pslist;
200 u_long				in_ifaddrhash_pslist;
201 struct pslist_head		in_ifaddrhead_pslist;
202 
203 void
204 in_init(void)
205 {
206 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
207 	    NULL, IPL_SOFTNET);
208 	TAILQ_INIT(&in_ifaddrhead);
209 	PSLIST_INIT(&in_ifaddrhead_pslist);
210 
211 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
212 	    &in_ifaddrhash);
213 
214 	in_ifaddrhash_psz = pserialize_create();
215 	in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST,
216 	    true, &in_ifaddrhash_pslist);
217 	mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
218 
219 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
220 	    &in_multihash);
221 	rw_init(&in_multilock);
222 
223 	in_sysctl_init(NULL);
224 }
225 
226 /*
227  * Return 1 if an internet address is for a ``local'' host
228  * (one to which we have a connection).  If subnetsarelocal
229  * is true, this includes other subnets of the local net.
230  * Otherwise, it includes only the directly-connected (sub)nets.
231  */
232 int
233 in_localaddr(struct in_addr in)
234 {
235 	struct in_ifaddr *ia;
236 	int localaddr = 0;
237 	int s = pserialize_read_enter();
238 
239 	if (subnetsarelocal) {
240 		IN_ADDRLIST_READER_FOREACH(ia) {
241 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net) {
242 				localaddr = 1;
243 				break;
244 			}
245 		}
246 	} else {
247 		IN_ADDRLIST_READER_FOREACH(ia) {
248 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
249 				localaddr = 1;
250 				break;
251 			}
252 		}
253 	}
254 	pserialize_read_exit(s);
255 
256 	return localaddr;
257 }
258 
259 /*
260  * like in_localaddr() but can specify ifp.
261  */
262 int
263 in_direct(struct in_addr in, struct ifnet *ifp)
264 {
265 	struct ifaddr *ifa;
266 	int localaddr = 0;
267 	int s;
268 
269 	KASSERT(ifp != NULL);
270 
271 #define ia (ifatoia(ifa))
272 	s = pserialize_read_enter();
273 	if (subnetsarelocal) {
274 		IFADDR_READER_FOREACH(ifa, ifp) {
275 			if (ifa->ifa_addr->sa_family == AF_INET &&
276 			    ((in.s_addr & ia->ia_netmask) == ia->ia_net)) {
277 				localaddr = 1;
278 				break;
279 			}
280 		}
281 	} else {
282 		IFADDR_READER_FOREACH(ifa, ifp) {
283 			if (ifa->ifa_addr->sa_family == AF_INET &&
284 			    (in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
285 				localaddr = 1;
286 				break;
287 			}
288 		}
289 	}
290 	pserialize_read_exit(s);
291 
292 	return localaddr;
293 #undef ia
294 }
295 
296 /*
297  * Determine whether an IP address is in a reserved set of addresses
298  * that may not be forwarded, or whether datagrams to that destination
299  * may be forwarded.
300  */
301 int
302 in_canforward(struct in_addr in)
303 {
304 	u_int32_t net;
305 
306 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
307 		return (0);
308 	if (IN_CLASSA(in.s_addr)) {
309 		net = in.s_addr & IN_CLASSA_NET;
310 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
311 			return (0);
312 	}
313 	return (1);
314 }
315 
316 /*
317  * Trim a mask in a sockaddr
318  */
319 void
320 in_socktrim(struct sockaddr_in *ap)
321 {
322 	char *cplim = (char *) &ap->sin_addr;
323 	char *cp = (char *) (&ap->sin_addr + 1);
324 
325 	ap->sin_len = 0;
326 	while (--cp >= cplim)
327 		if (*cp) {
328 			(ap)->sin_len = cp - (char *) (ap) + 1;
329 			break;
330 		}
331 }
332 
333 /*
334  * Maintain the "in_maxmtu" variable, which is the largest
335  * mtu for non-local interfaces with AF_INET addresses assigned
336  * to them that are up.
337  */
338 unsigned long in_maxmtu;
339 
340 void
341 in_setmaxmtu(void)
342 {
343 	struct in_ifaddr *ia;
344 	struct ifnet *ifp;
345 	unsigned long maxmtu = 0;
346 	int s = pserialize_read_enter();
347 
348 	IN_ADDRLIST_READER_FOREACH(ia) {
349 		if ((ifp = ia->ia_ifp) == 0)
350 			continue;
351 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
352 			continue;
353 		if (ifp->if_mtu > maxmtu)
354 			maxmtu = ifp->if_mtu;
355 	}
356 	if (maxmtu)
357 		in_maxmtu = maxmtu;
358 	pserialize_read_exit(s);
359 }
360 
361 static u_int
362 in_mask2len(struct in_addr *mask)
363 {
364 	u_int x, y;
365 	u_char *p;
366 
367 	p = (u_char *)mask;
368 	for (x = 0; x < sizeof(*mask); x++) {
369 		if (p[x] != 0xff)
370 			break;
371 	}
372 	y = 0;
373 	if (x < sizeof(*mask)) {
374 		for (y = 0; y < NBBY; y++) {
375 			if ((p[x] & (0x80 >> y)) == 0)
376 				break;
377 		}
378 	}
379 	return x * NBBY + y;
380 }
381 
382 void
383 in_len2mask(struct in_addr *mask, u_int len)
384 {
385 	u_int i;
386 	u_char *p;
387 
388 	p = (u_char *)mask;
389 	memset(mask, 0, sizeof(*mask));
390 	for (i = 0; i < len / NBBY; i++)
391 		p[i] = 0xff;
392 	if (len % NBBY)
393 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
394 }
395 
396 /*
397  * Generic internet control operations (ioctl's).
398  * Ifp is 0 if not an interface-specific ioctl.
399  */
400 /* ARGSUSED */
401 static int
402 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
403 {
404 	struct ifreq *ifr = (struct ifreq *)data;
405 	struct in_ifaddr *ia = NULL;
406 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
407 	struct sockaddr_in oldaddr, *new_dstaddr;
408 	int error, hostIsNew, maskIsNew;
409 	int newifaddr = 0;
410 	bool run_hook = false;
411 	bool need_reinsert = false;
412 	struct psref psref;
413 	int bound;
414 
415 	switch (cmd) {
416 	case SIOCALIFADDR:
417 	case SIOCDLIFADDR:
418 	case SIOCGLIFADDR:
419 		if (ifp == NULL)
420 			return EINVAL;
421 		return in_lifaddr_ioctl(so, cmd, data, ifp);
422 	case SIOCGIFADDRPREF:
423 	case SIOCSIFADDRPREF:
424 		if (ifp == NULL)
425 			return EINVAL;
426 		return ifaddrpref_ioctl(so, cmd, data, ifp);
427 #if NARP > 0
428 	case SIOCGNBRINFO:
429 	{
430 		struct in_nbrinfo *nbi = (struct in_nbrinfo *)data;
431 		struct llentry *ln;
432 		struct in_addr nb_addr = nbi->addr; /* make local for safety */
433 
434 		ln = arplookup(ifp, &nb_addr, NULL, 0);
435 		if (ln == NULL)
436 			return EINVAL;
437 		nbi->state = ln->ln_state;
438 		nbi->asked = ln->ln_asked;
439 		nbi->expire = ln->ln_expire ?
440 		    time_mono_to_wall(ln->ln_expire) : 0;
441 		LLE_RUNLOCK(ln);
442 		return 0;
443 	}
444 #endif
445 	}
446 
447 	bound = curlwp_bind();
448 	/*
449 	 * Find address for this interface, if it exists.
450 	 */
451 	if (ifp != NULL)
452 		ia = in_get_ia_from_ifp_psref(ifp, &psref);
453 
454 	hostIsNew = 1;		/* moved here to appease gcc */
455 	switch (cmd) {
456 	case SIOCAIFADDR:
457 	case SIOCDIFADDR:
458 	case SIOCGIFALIAS:
459 	case SIOCGIFAFLAG_IN:
460 		if (ifra->ifra_addr.sin_family == AF_INET) {
461 			int s;
462 
463 			if (ia != NULL)
464 				ia4_release(ia, &psref);
465 			s = pserialize_read_enter();
466 			IN_ADDRHASH_READER_FOREACH(ia,
467 			    ifra->ifra_addr.sin_addr.s_addr) {
468 				if (ia->ia_ifp == ifp &&
469 				    in_hosteq(ia->ia_addr.sin_addr,
470 				    ifra->ifra_addr.sin_addr))
471 					break;
472 			}
473 			if (ia != NULL)
474 				ia4_acquire(ia, &psref);
475 			pserialize_read_exit(s);
476 		}
477 		if ((cmd == SIOCDIFADDR ||
478 		    cmd == SIOCGIFALIAS ||
479 		    cmd == SIOCGIFAFLAG_IN) &&
480 		    ia == NULL) {
481 			error = EADDRNOTAVAIL;
482 			goto out;
483 		}
484 
485 		if (cmd == SIOCDIFADDR &&
486 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
487 			ifra->ifra_addr.sin_family = AF_INET;
488 		}
489 		/* FALLTHROUGH */
490 	case SIOCSIFADDR:
491 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
492 			;
493 		else if (ifra->ifra_addr.sin_len == 0) {
494 			ifra->ifra_addr = ia->ia_addr;
495 			hostIsNew = 0;
496 		} else if (in_hosteq(ia->ia_addr.sin_addr,
497 		           ifra->ifra_addr.sin_addr))
498 			hostIsNew = 0;
499 		if (ifra->ifra_addr.sin_family != AF_INET) {
500 			error = EAFNOSUPPORT;
501 			goto out;
502 		}
503 		/* FALLTHROUGH */
504 	case SIOCSIFDSTADDR:
505 		if (cmd == SIOCSIFDSTADDR &&
506 		    ifreq_getaddr(cmd, ifr)->sa_family != AF_INET) {
507 			error = EAFNOSUPPORT;
508 			goto out;
509 		}
510 		/* FALLTHROUGH */
511 	case SIOCSIFNETMASK:
512 		if (ifp == NULL)
513 			panic("in_control");
514 
515 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
516 			break;
517 
518 		if (ia == NULL &&
519 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
520 			error = EADDRNOTAVAIL;
521 			goto out;
522 		}
523 
524 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
525 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
526 		    NULL) != 0) {
527 			error = EPERM;
528 			goto out;
529 		}
530 
531 		if (ia == NULL) {
532 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
533 			if (ia == NULL) {
534 				error = ENOBUFS;
535 				goto out;
536 			}
537 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
538 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
539 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
540 #ifdef IPSELSRC
541 			ia->ia_ifa.ifa_getifa = in_getifa;
542 #else /* IPSELSRC */
543 			ia->ia_ifa.ifa_getifa = NULL;
544 #endif /* IPSELSRC */
545 			ia->ia_sockmask.sin_len = 8;
546 			ia->ia_sockmask.sin_family = AF_INET;
547 			if (ifp->if_flags & IFF_BROADCAST) {
548 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
549 				ia->ia_broadaddr.sin_family = AF_INET;
550 			}
551 			ia->ia_ifp = ifp;
552 			ia->ia_idsalt = cprng_fast32() % 65535;
553 			LIST_INIT(&ia->ia_multiaddrs);
554 			IN_ADDRHASH_ENTRY_INIT(ia);
555 			IN_ADDRLIST_ENTRY_INIT(ia);
556 			ifa_psref_init(&ia->ia_ifa);
557 			/*
558 			 * We need a reference to make ia survive over in_ifinit
559 			 * that does ifaref and ifafree.
560 			 */
561 			ifaref(&ia->ia_ifa);
562 
563 			newifaddr = 1;
564 		}
565 		break;
566 
567 	case SIOCSIFBRDADDR:
568 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
569 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
570 		    NULL) != 0) {
571 			error = EPERM;
572 			goto out;
573 		}
574 		/* FALLTHROUGH */
575 
576 	case SIOCGIFADDR:
577 	case SIOCGIFNETMASK:
578 	case SIOCGIFDSTADDR:
579 	case SIOCGIFBRDADDR:
580 		if (ia == NULL) {
581 			error = EADDRNOTAVAIL;
582 			goto out;
583 		}
584 		break;
585 	}
586 	error = 0;
587 	switch (cmd) {
588 
589 	case SIOCGIFADDR:
590 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
591 		break;
592 
593 	case SIOCGIFBRDADDR:
594 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
595 			error = EINVAL;
596 			goto out;
597 		}
598 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
599 		break;
600 
601 	case SIOCGIFDSTADDR:
602 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
603 			error = EINVAL;
604 			goto out;
605 		}
606 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
607 		break;
608 
609 	case SIOCGIFNETMASK:
610 		/*
611 		 * We keep the number of trailing zero bytes the sin_len field
612 		 * of ia_sockmask, so we fix this before we pass it back to
613 		 * userland.
614 		 */
615 		oldaddr = ia->ia_sockmask;
616 		oldaddr.sin_len = sizeof(struct sockaddr_in);
617 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
618 		break;
619 
620 	case SIOCSIFDSTADDR:
621 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
622 			error = EINVAL;
623 			goto out;
624 		}
625 		oldaddr = ia->ia_dstaddr;
626 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
627 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
628 			ia->ia_dstaddr = oldaddr;
629 			goto out;
630 		}
631 		if (ia->ia_flags & IFA_ROUTE) {
632 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
633 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
634 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
635 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
636 		}
637 		break;
638 
639 	case SIOCSIFBRDADDR:
640 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
641 			error = EINVAL;
642 			goto out;
643 		}
644 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
645 		break;
646 
647 	case SIOCSIFADDR:
648 		if (!newifaddr) {
649 			in_addrhash_remove(ia);
650 			need_reinsert = true;
651 		}
652 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
653 		    NULL, 1);
654 
655 		run_hook = true;
656 		break;
657 
658 	case SIOCSIFNETMASK:
659 		in_scrubprefix(ia);
660 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
661 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
662 		if (!newifaddr) {
663 			in_addrhash_remove(ia);
664 			need_reinsert = true;
665 		}
666 		error = in_ifinit(ifp, ia, NULL, NULL, 0);
667 		break;
668 
669 	case SIOCAIFADDR:
670 		maskIsNew = 0;
671 		if (ifra->ifra_mask.sin_len) {
672 			in_scrubprefix(ia);
673 			ia->ia_sockmask = ifra->ifra_mask;
674 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
675 			maskIsNew = 1;
676 		}
677 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
678 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
679 			new_dstaddr = &ifra->ifra_dstaddr;
680 			maskIsNew  = 1; /* We lie; but the effect's the same */
681 		} else
682 			new_dstaddr = NULL;
683 		if (ifra->ifra_addr.sin_family == AF_INET &&
684 		    (hostIsNew || maskIsNew)) {
685 			if (!newifaddr) {
686 				in_addrhash_remove(ia);
687 				need_reinsert = true;
688 			}
689 			error = in_ifinit(ifp, ia, &ifra->ifra_addr,
690 			    new_dstaddr, 0);
691 		}
692 		if ((ifp->if_flags & IFF_BROADCAST) &&
693 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
694 			ia->ia_broadaddr = ifra->ifra_broadaddr;
695 		run_hook = true;
696 		break;
697 
698 	case SIOCGIFALIAS:
699 		ifra->ifra_mask = ia->ia_sockmask;
700 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
701 		    (ia->ia_dstaddr.sin_family == AF_INET))
702 			ifra->ifra_dstaddr = ia->ia_dstaddr;
703 		else if ((ifp->if_flags & IFF_BROADCAST) &&
704 		    (ia->ia_broadaddr.sin_family == AF_INET))
705 			ifra->ifra_broadaddr = ia->ia_broadaddr;
706 		else
707 			memset(&ifra->ifra_broadaddr, 0,
708 			      sizeof(ifra->ifra_broadaddr));
709 		break;
710 
711 	case SIOCGIFAFLAG_IN:
712 		ifr->ifr_addrflags = ia->ia4_flags;
713 		break;
714 
715 	case SIOCDIFADDR:
716 		ia4_release(ia, &psref);
717 		ifaref(&ia->ia_ifa);
718 		in_purgeaddr(&ia->ia_ifa);
719 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
720 		ifafree(&ia->ia_ifa);
721 		ia = NULL;
722 		break;
723 
724 #ifdef MROUTING
725 	case SIOCGETVIFCNT:
726 	case SIOCGETSGCNT:
727 		error = mrt_ioctl(so, cmd, data);
728 		break;
729 #endif /* MROUTING */
730 
731 	default:
732 		error = ENOTTY;
733 		goto out;
734 	}
735 
736 	/*
737 	 * XXX insert regardless of error to make in_purgeaddr below work.
738 	 * Need to improve.
739 	 */
740 	if (newifaddr) {
741 		ifaref(&ia->ia_ifa);
742 		ifa_insert(ifp, &ia->ia_ifa);
743 
744 		mutex_enter(&in_ifaddr_lock);
745 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
746 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
747 		in_addrhash_insert_locked(ia);
748 		/* Release a reference that is held just after creation. */
749 		ifafree(&ia->ia_ifa);
750 		mutex_exit(&in_ifaddr_lock);
751 	} else if (need_reinsert) {
752 		in_addrhash_insert(ia);
753 	}
754 
755 	if (error == 0) {
756 		if (run_hook)
757 			pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
758 	} else if (newifaddr) {
759 		KASSERT(ia != NULL);
760 		in_purgeaddr(&ia->ia_ifa);
761 		ia = NULL;
762 	}
763 
764 out:
765 	if (!newifaddr && ia != NULL)
766 		ia4_release(ia, &psref);
767 	curlwp_bindx(bound);
768 	return error;
769 }
770 
771 int
772 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
773 {
774 	int error;
775 
776 #ifndef NET_MPSAFE
777 	KASSERT(KERNEL_LOCKED_P());
778 #endif
779 	error = in_control0(so, cmd, data, ifp);
780 
781 	return error;
782 }
783 
784 /* Add ownaddr as loopback rtentry. */
785 static void
786 in_ifaddlocal(struct ifaddr *ifa)
787 {
788 	struct in_ifaddr *ia;
789 
790 	ia = (struct in_ifaddr *)ifa;
791 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
792 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
793 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
794 	{
795 		rt_addrmsg(RTM_NEWADDR, ifa);
796 		return;
797 	}
798 
799 	rt_ifa_addlocal(ifa);
800 }
801 
802 /* Remove loopback entry of ownaddr */
803 static void
804 in_ifremlocal(struct ifaddr *ifa)
805 {
806 	struct in_ifaddr *ia, *p;
807 	struct ifaddr *alt_ifa = NULL;
808 	int ia_count = 0;
809 	int s;
810 	struct psref psref;
811 	int bound = curlwp_bind();
812 
813 	ia = (struct in_ifaddr *)ifa;
814 	/* Delete the entry if exactly one ifaddr matches the
815 	 * address, ifa->ifa_addr. */
816 	s = pserialize_read_enter();
817 	IN_ADDRLIST_READER_FOREACH(p) {
818 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
819 			continue;
820 		if (p->ia_ifp != ia->ia_ifp)
821 			alt_ifa = &p->ia_ifa;
822 		if (++ia_count > 1 && alt_ifa != NULL)
823 			break;
824 	}
825 	if (alt_ifa != NULL && ia_count > 1)
826 		ifa_acquire(alt_ifa, &psref);
827 	pserialize_read_exit(s);
828 
829 	if (ia_count == 0)
830 		goto out;
831 
832 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
833 	if (alt_ifa != NULL && ia_count > 1)
834 		ifa_release(alt_ifa, &psref);
835 out:
836 	curlwp_bindx(bound);
837 }
838 
839 static void
840 in_scrubaddr(struct in_ifaddr *ia)
841 {
842 
843 	/* stop DAD processing */
844 	if (ia->ia_dad_stop != NULL)
845 		ia->ia_dad_stop(&ia->ia_ifa);
846 
847 	in_scrubprefix(ia);
848 	in_ifremlocal(&ia->ia_ifa);
849 
850 	mutex_enter(&in_ifaddr_lock);
851 	if (ia->ia_allhosts != NULL) {
852 		in_delmulti(ia->ia_allhosts);
853 		ia->ia_allhosts = NULL;
854 	}
855 	mutex_exit(&in_ifaddr_lock);
856 }
857 
858 /*
859  * Depends on it isn't called in concurrent. It should be guaranteed
860  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
861  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
862  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
863  * never be executed in concurrent.
864  */
865 void
866 in_purgeaddr(struct ifaddr *ifa)
867 {
868 	struct in_ifaddr *ia = (void *) ifa;
869 	struct ifnet *ifp = ifa->ifa_ifp;
870 
871 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
872 
873 	ifa->ifa_flags |= IFA_DESTROYING;
874 	in_scrubaddr(ia);
875 
876 	mutex_enter(&in_ifaddr_lock);
877 	in_addrhash_remove_locked(ia);
878 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
879 	IN_ADDRLIST_WRITER_REMOVE(ia);
880 	ifa_remove(ifp, &ia->ia_ifa);
881 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
882 	mutex_exit(&in_ifaddr_lock);
883 	IN_ADDRHASH_ENTRY_DESTROY(ia);
884 	IN_ADDRLIST_ENTRY_DESTROY(ia);
885 	ifafree(&ia->ia_ifa);
886 	in_setmaxmtu();
887 }
888 
889 static void
890 in_addrhash_insert_locked(struct in_ifaddr *ia)
891 {
892 
893 	KASSERT(mutex_owned(&in_ifaddr_lock));
894 
895 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
896 	    ia_hash);
897 	IN_ADDRHASH_ENTRY_INIT(ia);
898 	IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
899 }
900 
901 void
902 in_addrhash_insert(struct in_ifaddr *ia)
903 {
904 
905 	mutex_enter(&in_ifaddr_lock);
906 	in_addrhash_insert_locked(ia);
907 	mutex_exit(&in_ifaddr_lock);
908 }
909 
910 static void
911 in_addrhash_remove_locked(struct in_ifaddr *ia)
912 {
913 
914 	KASSERT(mutex_owned(&in_ifaddr_lock));
915 
916 	LIST_REMOVE(ia, ia_hash);
917 	IN_ADDRHASH_WRITER_REMOVE(ia);
918 }
919 
920 void
921 in_addrhash_remove(struct in_ifaddr *ia)
922 {
923 
924 	mutex_enter(&in_ifaddr_lock);
925 	in_addrhash_remove_locked(ia);
926 #ifdef NET_MPSAFE
927 	pserialize_perform(in_ifaddrhash_psz);
928 #endif
929 	mutex_exit(&in_ifaddr_lock);
930 	IN_ADDRHASH_ENTRY_DESTROY(ia);
931 }
932 
933 void
934 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
935 {
936 
937 	IFNET_LOCK(ifp);
938 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
939 	igmp_purgeif(ifp);		/* manipulates pools */
940 #ifdef MROUTING
941 	ip_mrouter_detach(ifp);
942 #endif
943 	IFNET_UNLOCK(ifp);
944 }
945 
946 /*
947  * SIOC[GAD]LIFADDR.
948  *	SIOCGLIFADDR: get first address. (???)
949  *	SIOCGLIFADDR with IFLR_PREFIX:
950  *		get first address that matches the specified prefix.
951  *	SIOCALIFADDR: add the specified address.
952  *	SIOCALIFADDR with IFLR_PREFIX:
953  *		EINVAL since we can't deduce hostid part of the address.
954  *	SIOCDLIFADDR: delete the specified address.
955  *	SIOCDLIFADDR with IFLR_PREFIX:
956  *		delete the first address that matches the specified prefix.
957  * return values:
958  *	EINVAL on invalid parameters
959  *	EADDRNOTAVAIL on prefix match failed/specified address not found
960  *	other values may be returned from in_ioctl()
961  */
962 static int
963 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
964     struct ifnet *ifp)
965 {
966 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
967 	struct ifaddr *ifa;
968 	struct sockaddr *sa;
969 
970 	/* sanity checks */
971 	if (data == NULL || ifp == NULL) {
972 		panic("invalid argument to in_lifaddr_ioctl");
973 		/*NOTRECHED*/
974 	}
975 
976 	switch (cmd) {
977 	case SIOCGLIFADDR:
978 		/* address must be specified on GET with IFLR_PREFIX */
979 		if ((iflr->flags & IFLR_PREFIX) == 0)
980 			break;
981 		/*FALLTHROUGH*/
982 	case SIOCALIFADDR:
983 	case SIOCDLIFADDR:
984 		/* address must be specified on ADD and DELETE */
985 		sa = (struct sockaddr *)&iflr->addr;
986 		if (sa->sa_family != AF_INET)
987 			return EINVAL;
988 		if (sa->sa_len != sizeof(struct sockaddr_in))
989 			return EINVAL;
990 		/* XXX need improvement */
991 		sa = (struct sockaddr *)&iflr->dstaddr;
992 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
993 			return EINVAL;
994 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
995 			return EINVAL;
996 		break;
997 	default: /*shouldn't happen*/
998 #if 0
999 		panic("invalid cmd to in_lifaddr_ioctl");
1000 		/*NOTREACHED*/
1001 #else
1002 		return EOPNOTSUPP;
1003 #endif
1004 	}
1005 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
1006 		return EINVAL;
1007 
1008 	switch (cmd) {
1009 	case SIOCALIFADDR:
1010 	    {
1011 		struct in_aliasreq ifra;
1012 
1013 		if (iflr->flags & IFLR_PREFIX)
1014 			return EINVAL;
1015 
1016 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
1017 		memset(&ifra, 0, sizeof(ifra));
1018 		memcpy(ifra.ifra_name, iflr->iflr_name,
1019 			sizeof(ifra.ifra_name));
1020 
1021 		memcpy(&ifra.ifra_addr, &iflr->addr,
1022 			((struct sockaddr *)&iflr->addr)->sa_len);
1023 
1024 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1025 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1026 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1027 		}
1028 
1029 		ifra.ifra_mask.sin_family = AF_INET;
1030 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
1031 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
1032 
1033 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
1034 	    }
1035 	case SIOCGLIFADDR:
1036 	case SIOCDLIFADDR:
1037 	    {
1038 		struct in_ifaddr *ia;
1039 		struct in_addr mask, candidate, match;
1040 		struct sockaddr_in *sin;
1041 		int cmp, s;
1042 
1043 		memset(&mask, 0, sizeof(mask));
1044 		memset(&match, 0, sizeof(match));	/* XXX gcc */
1045 		if (iflr->flags & IFLR_PREFIX) {
1046 			/* lookup a prefix rather than address. */
1047 			in_len2mask(&mask, iflr->prefixlen);
1048 
1049 			sin = (struct sockaddr_in *)&iflr->addr;
1050 			match.s_addr = sin->sin_addr.s_addr;
1051 			match.s_addr &= mask.s_addr;
1052 
1053 			/* if you set extra bits, that's wrong */
1054 			if (match.s_addr != sin->sin_addr.s_addr)
1055 				return EINVAL;
1056 
1057 			cmp = 1;
1058 		} else {
1059 			if (cmd == SIOCGLIFADDR) {
1060 				/* on getting an address, take the 1st match */
1061 				cmp = 0;	/*XXX*/
1062 			} else {
1063 				/* on deleting an address, do exact match */
1064 				in_len2mask(&mask, 32);
1065 				sin = (struct sockaddr_in *)&iflr->addr;
1066 				match.s_addr = sin->sin_addr.s_addr;
1067 
1068 				cmp = 1;
1069 			}
1070 		}
1071 
1072 		s = pserialize_read_enter();
1073 		IFADDR_READER_FOREACH(ifa, ifp) {
1074 			if (ifa->ifa_addr->sa_family != AF_INET)
1075 				continue;
1076 			if (cmp == 0)
1077 				break;
1078 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
1079 			candidate.s_addr &= mask.s_addr;
1080 			if (candidate.s_addr == match.s_addr)
1081 				break;
1082 		}
1083 		if (ifa == NULL) {
1084 			pserialize_read_exit(s);
1085 			return EADDRNOTAVAIL;
1086 		}
1087 		ia = (struct in_ifaddr *)ifa;
1088 
1089 		if (cmd == SIOCGLIFADDR) {
1090 			/* fill in the if_laddrreq structure */
1091 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1092 
1093 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1094 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1095 					ia->ia_dstaddr.sin_len);
1096 			} else
1097 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1098 
1099 			iflr->prefixlen =
1100 				in_mask2len(&ia->ia_sockmask.sin_addr);
1101 
1102 			iflr->flags = 0;	/*XXX*/
1103 			pserialize_read_exit(s);
1104 
1105 			return 0;
1106 		} else {
1107 			struct in_aliasreq ifra;
1108 
1109 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1110 			memset(&ifra, 0, sizeof(ifra));
1111 			memcpy(ifra.ifra_name, iflr->iflr_name,
1112 				sizeof(ifra.ifra_name));
1113 
1114 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1115 				ia->ia_addr.sin_len);
1116 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1117 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1118 					ia->ia_dstaddr.sin_len);
1119 			}
1120 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1121 				ia->ia_sockmask.sin_len);
1122 			pserialize_read_exit(s);
1123 
1124 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1125 		}
1126 	    }
1127 	}
1128 
1129 	return EOPNOTSUPP;	/*just for safety*/
1130 }
1131 
1132 /*
1133  * Initialize an interface's internet address
1134  * and routing table entry.
1135  */
1136 int
1137 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1138     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1139 {
1140 	u_int32_t i;
1141 	struct sockaddr_in oldaddr, olddst;
1142 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1143 
1144 	if (sin == NULL)
1145 		sin = &ia->ia_addr;
1146 	if (dst == NULL)
1147 		dst = &ia->ia_dstaddr;
1148 
1149 	/*
1150 	 * Set up new addresses.
1151 	 */
1152 	oldaddr = ia->ia_addr;
1153 	olddst = ia->ia_dstaddr;
1154 	oldflags = ia->ia4_flags;
1155 	ia->ia_addr = *sin;
1156 	ia->ia_dstaddr = *dst;
1157 	hostIsNew = oldaddr.sin_family != AF_INET ||
1158 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1159 	if (!scrub)
1160 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1161 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1162 
1163 	/*
1164 	 * Configure address flags.
1165 	 * We need to do this early because they may be adjusted
1166 	 * by if_addr_init depending on the address.
1167 	 */
1168 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1169 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1170 		hostIsNew = 1;
1171 	}
1172 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1173 		ia->ia4_flags |= IN_IFF_DETACHED;
1174 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1175 	} else if (hostIsNew && if_do_dad(ifp) && ip_dad_enabled())
1176 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1177 
1178 	/*
1179 	 * Give the interface a chance to initialize
1180 	 * if this is its first address,
1181 	 * and to validate the address if necessary.
1182 	 */
1183 	s = splsoftnet();
1184 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1185 	splx(s);
1186 	/* Now clear the try tentative flag, its job is done. */
1187 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1188 	if (error != 0) {
1189 		ia->ia_addr = oldaddr;
1190 		ia->ia_dstaddr = olddst;
1191 		ia->ia4_flags = oldflags;
1192 		return error;
1193 	}
1194 
1195 	if (scrub || hostIsNew) {
1196 		int newflags = ia->ia4_flags;
1197 
1198 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1199 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1200 		ia->ia4_flags = oldflags;
1201 		if (hostIsNew)
1202 			in_scrubaddr(ia);
1203 		else if (scrub)
1204 			in_scrubprefix(ia);
1205 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1206 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1207 		ia->ia4_flags = newflags;
1208 	}
1209 
1210 	i = ia->ia_addr.sin_addr.s_addr;
1211 	if (ifp->if_flags & IFF_POINTOPOINT)
1212 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1213 	else if (IN_CLASSA(i))
1214 		ia->ia_netmask = IN_CLASSA_NET;
1215 	else if (IN_CLASSB(i))
1216 		ia->ia_netmask = IN_CLASSB_NET;
1217 	else
1218 		ia->ia_netmask = IN_CLASSC_NET;
1219 	/*
1220 	 * The subnet mask usually includes at least the standard network part,
1221 	 * but may may be smaller in the case of supernetting.
1222 	 * If it is set, we believe it.
1223 	 */
1224 	if (ia->ia_subnetmask == 0) {
1225 		ia->ia_subnetmask = ia->ia_netmask;
1226 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1227 	} else
1228 		ia->ia_netmask &= ia->ia_subnetmask;
1229 
1230 	ia->ia_net = i & ia->ia_netmask;
1231 	ia->ia_subnet = i & ia->ia_subnetmask;
1232 	in_socktrim(&ia->ia_sockmask);
1233 
1234 	/* re-calculate the "in_maxmtu" value */
1235 	in_setmaxmtu();
1236 
1237 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1238 	if (ifp->if_flags & IFF_BROADCAST) {
1239 		if (ia->ia_subnetmask == IN_RFC3021_MASK) {
1240 			ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST;
1241 			ia->ia_netbroadcast.s_addr = INADDR_BROADCAST;
1242 		} else {
1243 			ia->ia_broadaddr.sin_addr.s_addr =
1244 				ia->ia_subnet | ~ia->ia_subnetmask;
1245 			ia->ia_netbroadcast.s_addr =
1246 				ia->ia_net | ~ia->ia_netmask;
1247 		}
1248 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1249 		ia->ia_dstaddr = ia->ia_addr;
1250 		flags |= RTF_HOST;
1251 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1252 		if (ia->ia_dstaddr.sin_family != AF_INET)
1253 			return (0);
1254 		flags |= RTF_HOST;
1255 	}
1256 
1257 	/* Add the local route to the address */
1258 	in_ifaddlocal(&ia->ia_ifa);
1259 
1260 	/* Add the prefix route for the address */
1261 	error = in_addprefix(ia, flags);
1262 
1263 	/*
1264 	 * If the interface supports multicast, join the "all hosts"
1265 	 * multicast group on that interface.
1266 	 */
1267 	mutex_enter(&in_ifaddr_lock);
1268 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1269 		struct in_addr addr;
1270 
1271 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1272 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1273 	}
1274 	mutex_exit(&in_ifaddr_lock);
1275 
1276 	if (hostIsNew &&
1277 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1278 	    if_do_dad(ifp))
1279 		ia->ia_dad_start((struct ifaddr *)ia);
1280 
1281 	return error;
1282 }
1283 
1284 #define rtinitflags(x) \
1285 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1286 	    ? RTF_HOST : 0)
1287 
1288 /*
1289  * add a route to prefix ("connected route" in cisco terminology).
1290  * does nothing if there's some interface address with the same prefix already.
1291  */
1292 static int
1293 in_addprefix(struct in_ifaddr *target, int flags)
1294 {
1295 	struct in_ifaddr *ia;
1296 	struct in_addr prefix, mask, p;
1297 	int error;
1298 	int s;
1299 
1300 	if ((flags & RTF_HOST) != 0)
1301 		prefix = target->ia_dstaddr.sin_addr;
1302 	else {
1303 		prefix = target->ia_addr.sin_addr;
1304 		mask = target->ia_sockmask.sin_addr;
1305 		prefix.s_addr &= mask.s_addr;
1306 	}
1307 
1308 	s = pserialize_read_enter();
1309 	IN_ADDRLIST_READER_FOREACH(ia) {
1310 		if (rtinitflags(ia))
1311 			p = ia->ia_dstaddr.sin_addr;
1312 		else {
1313 			p = ia->ia_addr.sin_addr;
1314 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1315 		}
1316 
1317 		if (prefix.s_addr != p.s_addr)
1318 			continue;
1319 
1320 		/*
1321 		 * if we got a matching prefix route inserted by other
1322 		 * interface address, we don't need to bother
1323 		 *
1324 		 * XXX RADIX_MPATH implications here? -dyoung
1325 		 */
1326 		if (ia->ia_flags & IFA_ROUTE) {
1327 			pserialize_read_exit(s);
1328 			return 0;
1329 		}
1330 	}
1331 	pserialize_read_exit(s);
1332 
1333 	/*
1334 	 * noone seem to have prefix route.  insert it.
1335 	 */
1336 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1337 	if (error == 0)
1338 		target->ia_flags |= IFA_ROUTE;
1339 	else if (error == EEXIST) {
1340 		/*
1341 		 * the fact the route already exists is not an error.
1342 		 */
1343 		error = 0;
1344 	}
1345 	return error;
1346 }
1347 
1348 /*
1349  * remove a route to prefix ("connected route" in cisco terminology).
1350  * re-installs the route by using another interface address, if there's one
1351  * with the same prefix (otherwise we lose the route mistakenly).
1352  */
1353 static int
1354 in_scrubprefix(struct in_ifaddr *target)
1355 {
1356 	struct in_ifaddr *ia;
1357 	struct in_addr prefix, mask, p;
1358 	int error;
1359 	int s;
1360 
1361 	/* If we don't have IFA_ROUTE we have nothing to do */
1362 	if ((target->ia_flags & IFA_ROUTE) == 0)
1363 		return 0;
1364 
1365 	if (rtinitflags(target))
1366 		prefix = target->ia_dstaddr.sin_addr;
1367 	else {
1368 		prefix = target->ia_addr.sin_addr;
1369 		mask = target->ia_sockmask.sin_addr;
1370 		prefix.s_addr &= mask.s_addr;
1371 	}
1372 
1373 	s = pserialize_read_enter();
1374 	IN_ADDRLIST_READER_FOREACH(ia) {
1375 		if (rtinitflags(ia))
1376 			p = ia->ia_dstaddr.sin_addr;
1377 		else {
1378 			p = ia->ia_addr.sin_addr;
1379 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1380 		}
1381 
1382 		if (prefix.s_addr != p.s_addr)
1383 			continue;
1384 
1385 		/*
1386 		 * if we got a matching prefix route, move IFA_ROUTE to him
1387 		 */
1388 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1389 			struct psref psref;
1390 			int bound = curlwp_bind();
1391 
1392 			ia4_acquire(ia, &psref);
1393 			pserialize_read_exit(s);
1394 
1395 			rtinit(&target->ia_ifa, RTM_DELETE,
1396 			    rtinitflags(target));
1397 			target->ia_flags &= ~IFA_ROUTE;
1398 
1399 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1400 			    rtinitflags(ia) | RTF_UP);
1401 			if (error == 0)
1402 				ia->ia_flags |= IFA_ROUTE;
1403 
1404 			ia4_release(ia, &psref);
1405 			curlwp_bindx(bound);
1406 
1407 			return error;
1408 		}
1409 	}
1410 	pserialize_read_exit(s);
1411 
1412 	/*
1413 	 * noone seem to have prefix route.  remove it.
1414 	 */
1415 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1416 	target->ia_flags &= ~IFA_ROUTE;
1417 	return 0;
1418 }
1419 
1420 #undef rtinitflags
1421 
1422 /*
1423  * Return 1 if the address might be a local broadcast address.
1424  */
1425 int
1426 in_broadcast(struct in_addr in, struct ifnet *ifp)
1427 {
1428 	struct ifaddr *ifa;
1429 	int s;
1430 
1431 	KASSERT(ifp != NULL);
1432 
1433 	if (in.s_addr == INADDR_BROADCAST ||
1434 	    in_nullhost(in))
1435 		return 1;
1436 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1437 		return 0;
1438 	/*
1439 	 * Look through the list of addresses for a match
1440 	 * with a broadcast address.
1441 	 */
1442 #define ia (ifatoia(ifa))
1443 	s = pserialize_read_enter();
1444 	IFADDR_READER_FOREACH(ifa, ifp) {
1445 		if (ifa->ifa_addr->sa_family == AF_INET &&
1446 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1447 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1448 		     in_hosteq(in, ia->ia_netbroadcast) ||
1449 		     (hostzeroisbroadcast &&
1450 		      /*
1451 		       * Check for old-style (host 0) broadcast, but
1452 		       * taking into account that RFC 3021 obsoletes it.
1453 		       */
1454 		      ia->ia_subnetmask != IN_RFC3021_MASK &&
1455 		      (in.s_addr == ia->ia_subnet ||
1456 		       in.s_addr == ia->ia_net)))) {
1457 			pserialize_read_exit(s);
1458 			return 1;
1459 		}
1460 	}
1461 	pserialize_read_exit(s);
1462 	return (0);
1463 #undef ia
1464 }
1465 
1466 /*
1467  * perform DAD when interface becomes IFF_UP.
1468  */
1469 void
1470 in_if_link_up(struct ifnet *ifp)
1471 {
1472 	struct ifaddr *ifa;
1473 	struct in_ifaddr *ia;
1474 	int s, bound;
1475 
1476 	/* Ensure it's sane to run DAD */
1477 	if (ifp->if_link_state == LINK_STATE_DOWN)
1478 		return;
1479 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1480 		return;
1481 
1482 	bound = curlwp_bind();
1483 	s = pserialize_read_enter();
1484 	IFADDR_READER_FOREACH(ifa, ifp) {
1485 		struct psref psref;
1486 
1487 		if (ifa->ifa_addr->sa_family != AF_INET)
1488 			continue;
1489 		ifa_acquire(ifa, &psref);
1490 		pserialize_read_exit(s);
1491 
1492 		ia = (struct in_ifaddr *)ifa;
1493 
1494 		/* If detached then mark as tentative */
1495 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1496 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1497 			if (ip_dad_enabled() && if_do_dad(ifp) &&
1498 			    ia->ia_dad_start != NULL)
1499 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1500 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1501 				rt_addrmsg(RTM_NEWADDR, ifa);
1502 		}
1503 
1504 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1505 			/* Clear the duplicated flag as we're starting DAD. */
1506 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1507 			ia->ia_dad_start(ifa);
1508 		}
1509 
1510 		s = pserialize_read_enter();
1511 		ifa_release(ifa, &psref);
1512 	}
1513 	pserialize_read_exit(s);
1514 	curlwp_bindx(bound);
1515 }
1516 
1517 void
1518 in_if_up(struct ifnet *ifp)
1519 {
1520 
1521 	/* interface may not support link state, so bring it up also */
1522 	in_if_link_up(ifp);
1523 }
1524 
1525 /*
1526  * Mark all addresses as detached.
1527  */
1528 void
1529 in_if_link_down(struct ifnet *ifp)
1530 {
1531 	struct ifaddr *ifa;
1532 	struct in_ifaddr *ia;
1533 	int s, bound;
1534 
1535 	bound = curlwp_bind();
1536 	s = pserialize_read_enter();
1537 	IFADDR_READER_FOREACH(ifa, ifp) {
1538 		struct psref psref;
1539 
1540 		if (ifa->ifa_addr->sa_family != AF_INET)
1541 			continue;
1542 		ifa_acquire(ifa, &psref);
1543 		pserialize_read_exit(s);
1544 
1545 		ia = (struct in_ifaddr *)ifa;
1546 
1547 		/* Stop DAD processing */
1548 		if (ia->ia_dad_stop != NULL)
1549 			ia->ia_dad_stop(ifa);
1550 
1551 		/*
1552 		 * Mark the address as detached.
1553 		 */
1554 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1555 			ia->ia4_flags |= IN_IFF_DETACHED;
1556 			ia->ia4_flags &=
1557 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1558 			rt_addrmsg(RTM_NEWADDR, ifa);
1559 		}
1560 
1561 		s = pserialize_read_enter();
1562 		ifa_release(ifa, &psref);
1563 	}
1564 	pserialize_read_exit(s);
1565 	curlwp_bindx(bound);
1566 }
1567 
1568 void
1569 in_if_down(struct ifnet *ifp)
1570 {
1571 
1572 	in_if_link_down(ifp);
1573 #if NARP > 0
1574 	lltable_purge_entries(LLTABLE(ifp));
1575 #endif
1576 }
1577 
1578 void
1579 in_if_link_state_change(struct ifnet *ifp, int link_state)
1580 {
1581 
1582 	/*
1583 	 * Treat LINK_STATE_UNKNOWN as UP.
1584 	 * LINK_STATE_UNKNOWN transitions to LINK_STATE_DOWN when
1585 	 * if_link_state_change() transitions to LINK_STATE_UP.
1586 	 */
1587 	if (link_state == LINK_STATE_DOWN)
1588 		in_if_link_down(ifp);
1589 	else
1590 		in_if_link_up(ifp);
1591 }
1592 
1593 /*
1594  * in_lookup_multi: look up the in_multi record for a given IP
1595  * multicast address on a given interface.  If no matching record is
1596  * found, return NULL.
1597  */
1598 struct in_multi *
1599 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1600 {
1601 	struct in_multi *inm;
1602 
1603 	KASSERT(rw_lock_held(&in_multilock));
1604 
1605 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1606 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1607 			break;
1608 	}
1609 	return inm;
1610 }
1611 
1612 /*
1613  * in_multi_group: check whether the address belongs to an IP multicast
1614  * group we are joined on this interface.  Returns true or false.
1615  */
1616 bool
1617 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1618 {
1619 	bool ingroup;
1620 
1621 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1622 		rw_enter(&in_multilock, RW_READER);
1623 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1624 		rw_exit(&in_multilock);
1625 	} else {
1626 		/* XXX Recursive call from ip_output(). */
1627 		KASSERT(rw_lock_held(&in_multilock));
1628 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1629 	}
1630 	return ingroup;
1631 }
1632 
1633 /*
1634  * Add an address to the list of IP multicast addresses for a given interface.
1635  */
1636 struct in_multi *
1637 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1638 {
1639 	struct sockaddr_in sin;
1640 	struct in_multi *inm;
1641 
1642 	/*
1643 	 * See if address already in list.
1644 	 */
1645 	rw_enter(&in_multilock, RW_WRITER);
1646 	inm = in_lookup_multi(*ap, ifp);
1647 	if (inm != NULL) {
1648 		/*
1649 		 * Found it; just increment the reference count.
1650 		 */
1651 		inm->inm_refcount++;
1652 		rw_exit(&in_multilock);
1653 		return inm;
1654 	}
1655 
1656 	/*
1657 	 * New address; allocate a new multicast record.
1658 	 */
1659 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1660 	if (inm == NULL) {
1661 		rw_exit(&in_multilock);
1662 		return NULL;
1663 	}
1664 	inm->inm_addr = *ap;
1665 	inm->inm_ifp = ifp;
1666 	inm->inm_refcount = 1;
1667 
1668 	/*
1669 	 * Ask the network driver to update its multicast reception
1670 	 * filter appropriately for the new address.
1671 	 */
1672 	sockaddr_in_init(&sin, ap, 0);
1673 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1674 		rw_exit(&in_multilock);
1675 		pool_put(&inmulti_pool, inm);
1676 		return NULL;
1677 	}
1678 
1679 	/*
1680 	 * Let IGMP know that we have joined a new IP multicast group.
1681 	 */
1682 	if (igmp_joingroup(inm) != 0) {
1683 		rw_exit(&in_multilock);
1684 		pool_put(&inmulti_pool, inm);
1685 		return NULL;
1686 	}
1687 	LIST_INSERT_HEAD(
1688 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1689 	    inm, inm_list);
1690 	in_multientries++;
1691 	rw_exit(&in_multilock);
1692 
1693 	return inm;
1694 }
1695 
1696 /*
1697  * Delete a multicast address record.
1698  */
1699 void
1700 in_delmulti(struct in_multi *inm)
1701 {
1702 	struct sockaddr_in sin;
1703 
1704 	rw_enter(&in_multilock, RW_WRITER);
1705 	if (--inm->inm_refcount > 0) {
1706 		rw_exit(&in_multilock);
1707 		return;
1708 	}
1709 
1710 	/*
1711 	 * No remaining claims to this record; let IGMP know that
1712 	 * we are leaving the multicast group.
1713 	 */
1714 	igmp_leavegroup(inm);
1715 
1716 	/*
1717 	 * Notify the network driver to update its multicast reception
1718 	 * filter.
1719 	 */
1720 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1721 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1722 
1723 	/*
1724 	 * Unlink from list.
1725 	 */
1726 	LIST_REMOVE(inm, inm_list);
1727 	in_multientries--;
1728 	rw_exit(&in_multilock);
1729 
1730 	pool_put(&inmulti_pool, inm);
1731 }
1732 
1733 /*
1734  * in_next_multi: step through all of the in_multi records, one at a time.
1735  * The current position is remembered in "step", which the caller must
1736  * provide.  in_first_multi(), below, must be called to initialize "step"
1737  * and get the first record.  Both macros return a NULL "inm" when there
1738  * are no remaining records.
1739  */
1740 struct in_multi *
1741 in_next_multi(struct in_multistep *step)
1742 {
1743 	struct in_multi *inm;
1744 
1745 	KASSERT(rw_lock_held(&in_multilock));
1746 
1747 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1748 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1749 	}
1750 	if ((inm = step->i_inm) != NULL) {
1751 		step->i_inm = LIST_NEXT(inm, inm_list);
1752 	}
1753 	return inm;
1754 }
1755 
1756 struct in_multi *
1757 in_first_multi(struct in_multistep *step)
1758 {
1759 	KASSERT(rw_lock_held(&in_multilock));
1760 
1761 	step->i_n = 0;
1762 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1763 	return in_next_multi(step);
1764 }
1765 
1766 void
1767 in_multi_lock(int op)
1768 {
1769 	rw_enter(&in_multilock, op);
1770 }
1771 
1772 void
1773 in_multi_unlock(void)
1774 {
1775 	rw_exit(&in_multilock);
1776 }
1777 
1778 int
1779 in_multi_lock_held(void)
1780 {
1781 	return rw_lock_held(&in_multilock);
1782 }
1783 
1784 struct in_ifaddr *
1785 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1786     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1787 {
1788 	struct rtentry *rt = NULL;
1789 	struct in_ifaddr *ia = NULL;
1790 
1791 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1792 	/*
1793          * If route is known or can be allocated now, take the
1794          * source address from the interface.  Otherwise, punt.
1795 	 */
1796 	if ((soopts & SO_DONTROUTE) != 0)
1797 		rtcache_free(ro);
1798 	else {
1799 		union {
1800 			struct sockaddr		dst;
1801 			struct sockaddr_in	dst4;
1802 		} u;
1803 
1804 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1805 		rt = rtcache_lookup(ro, &u.dst);
1806 	}
1807 	/*
1808 	 * If we found a route, use the address
1809 	 * corresponding to the outgoing interface
1810 	 * unless it is the loopback (in case a route
1811 	 * to our address on another net goes to loopback).
1812 	 *
1813 	 * XXX Is this still true?  Do we care?
1814 	 */
1815 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1816 		int s;
1817 		struct ifaddr *ifa;
1818 		/*
1819 		 * Just in case. May not need to do this workaround.
1820 		 * Revisit when working on rtentry MP-ification.
1821 		 */
1822 		s = pserialize_read_enter();
1823 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1824 			if (ifa == rt->rt_ifa)
1825 				break;
1826 		}
1827 		if (ifa != NULL)
1828 			ifa_acquire(ifa, psref);
1829 		pserialize_read_exit(s);
1830 
1831 		ia = ifatoia(ifa);
1832 	}
1833 	if (ia == NULL) {
1834 		u_int16_t fport = sin->sin_port;
1835 		struct ifaddr *ifa;
1836 		int s;
1837 
1838 		sin->sin_port = 0;
1839 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1840 		sin->sin_port = fport;
1841 		if (ifa == NULL) {
1842 			/* Find 1st non-loopback AF_INET address */
1843 			s = pserialize_read_enter();
1844 			IN_ADDRLIST_READER_FOREACH(ia) {
1845 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1846 					break;
1847 			}
1848 			if (ia != NULL)
1849 				ia4_acquire(ia, psref);
1850 			pserialize_read_exit(s);
1851 		} else {
1852 			/* ia is already referenced by psref */
1853 			ia = ifatoia(ifa);
1854 		}
1855 		if (ia == NULL) {
1856 			*errorp = EADDRNOTAVAIL;
1857 			goto out;
1858 		}
1859 	}
1860 	/*
1861 	 * If the destination address is multicast and an outgoing
1862 	 * interface has been set as a multicast option, use the
1863 	 * address of that interface as our source address.
1864 	 */
1865 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1866 		struct ip_moptions *imo;
1867 
1868 		imo = mopts;
1869 		if (imo->imo_multicast_if_index != 0) {
1870 			struct ifnet *ifp;
1871 			int s;
1872 
1873 			if (ia != NULL)
1874 				ia4_release(ia, psref);
1875 			s = pserialize_read_enter();
1876 			ifp = if_byindex(imo->imo_multicast_if_index);
1877 			if (ifp != NULL) {
1878 				/* XXX */
1879 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1880 			} else
1881 				ia = NULL;
1882 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1883 				pserialize_read_exit(s);
1884 				if (ia != NULL)
1885 					ia4_release(ia, psref);
1886 				*errorp = EADDRNOTAVAIL;
1887 				ia = NULL;
1888 				goto out;
1889 			}
1890 			pserialize_read_exit(s);
1891 		}
1892 	}
1893 	if (ia->ia_ifa.ifa_getifa != NULL) {
1894 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1895 		                                      sintosa(sin)));
1896 		if (ia == NULL) {
1897 			*errorp = EADDRNOTAVAIL;
1898 			goto out;
1899 		}
1900 		/* FIXME NOMPSAFE */
1901 		ia4_acquire(ia, psref);
1902 	}
1903 #ifdef GETIFA_DEBUG
1904 	else
1905 		printf("%s: missing ifa_getifa\n", __func__);
1906 #endif
1907 out:
1908 	rtcache_unref(rt, ro);
1909 	return ia;
1910 }
1911 
1912 int
1913 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst)
1914 {
1915 	struct in_ifaddr *ia4;
1916 	int s;
1917 
1918 	/* check for address match */
1919 	if (src.s_addr != ip->ip_dst.s_addr ||
1920 	    dst.s_addr != ip->ip_src.s_addr)
1921 		return 0;
1922 
1923 	/* martian filters on outer source - NOT done in ip_input! */
1924 	if (IN_MULTICAST(ip->ip_src.s_addr))
1925 		return 0;
1926 	switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) {
1927 	case 0:
1928 	case 127:
1929 	case 255:
1930 		return 0;
1931 	}
1932 	/* reject packets with broadcast on source */
1933 	s = pserialize_read_enter();
1934 	IN_ADDRLIST_READER_FOREACH(ia4) {
1935 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
1936 			continue;
1937 		if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
1938 			pserialize_read_exit(s);
1939 			return 0;
1940 		}
1941 	}
1942 	pserialize_read_exit(s);
1943 
1944 	/* NOTE: packet may dropped by uRPF */
1945 
1946 	/* return valid bytes length */
1947 	return sizeof(src) + sizeof(dst);
1948 }
1949 
1950 #if NARP > 0
1951 
1952 #define	IN_LLTBL_DEFAULT_HSIZE	32
1953 #define	IN_LLTBL_HASH(k, h) \
1954 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1955 
1956 /*
1957  * Do actual deallocation of @lle.
1958  * Called by LLE_FREE_LOCKED when number of references
1959  * drops to zero.
1960  */
1961 static void
1962 in_lltable_destroy_lle(struct llentry *lle)
1963 {
1964 
1965 	KASSERTMSG(lle->la_numheld == 0, "la_numheld=%d", lle->la_numheld);
1966 
1967 	LLE_WUNLOCK(lle);
1968 	LLE_LOCK_DESTROY(lle);
1969 	llentry_pool_put(lle);
1970 }
1971 
1972 static struct llentry *
1973 in_lltable_new(struct in_addr addr4, u_int flags)
1974 {
1975 	struct llentry *lle;
1976 
1977 	lle = llentry_pool_get(PR_NOWAIT);
1978 	if (lle == NULL)		/* NB: caller generates msg */
1979 		return NULL;
1980 
1981 	lle->r_l3addr.addr4 = addr4;
1982 	lle->lle_refcnt = 1;
1983 	lle->lle_free = in_lltable_destroy_lle;
1984 	LLE_LOCK_INIT(lle);
1985 	callout_init(&lle->la_timer, CALLOUT_MPSAFE);
1986 
1987 	return lle;
1988 }
1989 
1990 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1991 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1992 
1993 static int
1994 in_lltable_match_prefix(const struct sockaddr *prefix,
1995     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1996 {
1997 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1998 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1999 	struct in_addr lle_addr;
2000 
2001 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
2002 
2003 	/*
2004 	 * (flags & LLE_STATIC) means deleting all entries
2005 	 * including static ARP entries.
2006 	 */
2007 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) &&
2008 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2009 		return (1);
2010 
2011 	return (0);
2012 }
2013 
2014 static void
2015 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2016 {
2017 	size_t pkts_dropped;
2018 
2019 	LLE_WLOCK_ASSERT(lle);
2020 	KASSERT(llt != NULL);
2021 
2022 	pkts_dropped = llentry_free(lle);
2023 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
2024 }
2025 
2026 static int
2027 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr,
2028     const struct rtentry *rt)
2029 {
2030 	int error = EINVAL;
2031 
2032 	if (rt == NULL)
2033 		return error;
2034 
2035 	/*
2036 	 * If the gateway for an existing host route matches the target L3
2037 	 * address, which is a special route inserted by some implementation
2038 	 * such as MANET, and the interface is of the correct type, then
2039 	 * allow for ARP to proceed.
2040 	 */
2041 	if (rt->rt_flags & RTF_GATEWAY) {
2042 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
2043 		    rt->rt_ifp->if_type != IFT_ETHER ||
2044 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
2045 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
2046 		    sizeof(in_addr_t)) != 0) {
2047 			goto error;
2048 		}
2049 	}
2050 
2051 	/*
2052 	 * Make sure that at least the destination address is covered
2053 	 * by the route. This is for handling the case where 2 or more
2054 	 * interfaces have the same prefix. An incoming packet arrives
2055 	 * on one interface and the corresponding outgoing packet leaves
2056 	 * another interface.
2057 	 */
2058 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
2059 		const char *sa, *mask, *addr, *lim;
2060 		int len;
2061 
2062 		mask = (const char *)rt_mask(rt);
2063 		/*
2064 		 * Just being extra cautious to avoid some custom
2065 		 * code getting into trouble.
2066 		 */
2067 		if (mask == NULL)
2068 			goto error;
2069 
2070 		sa = (const char *)rt_getkey(rt);
2071 		addr = (const char *)l3addr;
2072 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
2073 		lim = addr + len;
2074 
2075 		for ( ; addr < lim; sa++, mask++, addr++) {
2076 			if ((*sa ^ *addr) & *mask) {
2077 #ifdef DIAGNOSTIC
2078 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
2079 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
2080 #endif
2081 				goto error;
2082 			}
2083 		}
2084 	}
2085 
2086 	error = 0;
2087 error:
2088 	return error;
2089 }
2090 
2091 static inline uint32_t
2092 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
2093 {
2094 
2095 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
2096 }
2097 
2098 static uint32_t
2099 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
2100 {
2101 
2102 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
2103 }
2104 
2105 static void
2106 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2107 {
2108 	struct sockaddr_in *sin;
2109 
2110 	sin = (struct sockaddr_in *)sa;
2111 	memset(sin, 0, sizeof(*sin));
2112 	sin->sin_family = AF_INET;
2113 	sin->sin_len = sizeof(*sin);
2114 	sin->sin_addr = lle->r_l3addr.addr4;
2115 }
2116 
2117 static inline struct llentry *
2118 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2119 {
2120 	struct llentry *lle;
2121 	struct llentries *lleh;
2122 	u_int hashidx;
2123 
2124 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2125 	lleh = &llt->lle_head[hashidx];
2126 	LIST_FOREACH(lle, lleh, lle_next) {
2127 		if (lle->la_flags & LLE_DELETED)
2128 			continue;
2129 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2130 			break;
2131 	}
2132 
2133 	return (lle);
2134 }
2135 
2136 static int
2137 in_lltable_delete(struct lltable *llt, u_int flags,
2138     const struct sockaddr *l3addr)
2139 {
2140 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2141 	struct ifnet *ifp __diagused = llt->llt_ifp;
2142 	struct llentry *lle;
2143 
2144 	IF_AFDATA_WLOCK_ASSERT(ifp);
2145 	KASSERTMSG(l3addr->sa_family == AF_INET,
2146 	    "sin_family %d", l3addr->sa_family);
2147 
2148 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2149 	if (lle == NULL) {
2150 #ifdef LLTABLE_DEBUG
2151 		char buf[64];
2152 		sockaddr_format(l3addr, buf, sizeof(buf));
2153 		log(LOG_INFO, "%s: cache for %s is not found\n",
2154 		    __func__, buf);
2155 #endif
2156 		return (ENOENT);
2157 	}
2158 
2159 	LLE_WLOCK(lle);
2160 #ifdef LLTABLE_DEBUG
2161 	{
2162 		char buf[64];
2163 		sockaddr_format(l3addr, buf, sizeof(buf));
2164 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2165 		    __func__, buf, lle);
2166 	}
2167 #endif
2168 	llentry_free(lle);
2169 
2170 	return (0);
2171 }
2172 
2173 static struct llentry *
2174 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr,
2175     const struct rtentry *rt)
2176 {
2177 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2178 	struct ifnet *ifp = llt->llt_ifp;
2179 	struct llentry *lle;
2180 
2181 	IF_AFDATA_WLOCK_ASSERT(ifp);
2182 	KASSERTMSG(l3addr->sa_family == AF_INET,
2183 	    "sin_family %d", l3addr->sa_family);
2184 
2185 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2186 
2187 	if (lle != NULL) {
2188 		LLE_WLOCK(lle);
2189 		return (lle);
2190 	}
2191 
2192 	/* no existing record, we need to create new one */
2193 
2194 	/*
2195 	 * A route that covers the given address must have
2196 	 * been installed 1st because we are doing a resolution,
2197 	 * verify this.
2198 	 */
2199 	if (!(flags & LLE_IFADDR) &&
2200 	    in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2201 		return (NULL);
2202 
2203 	lle = in_lltable_new(sin->sin_addr, flags);
2204 	if (lle == NULL) {
2205 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2206 		return (NULL);
2207 	}
2208 	lle->la_flags = flags;
2209 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2210 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2211 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2212 	}
2213 
2214 	lltable_link_entry(llt, lle);
2215 	LLE_WLOCK(lle);
2216 
2217 	return (lle);
2218 }
2219 
2220 /*
2221  * Return NULL if not found or marked for deletion.
2222  * If found return lle read locked.
2223  */
2224 static struct llentry *
2225 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2226 {
2227 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2228 	struct llentry *lle;
2229 
2230 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2231 	KASSERTMSG(l3addr->sa_family == AF_INET,
2232 	    "sin_family %d", l3addr->sa_family);
2233 
2234 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2235 
2236 	if (lle == NULL)
2237 		return NULL;
2238 
2239 	if (flags & LLE_EXCLUSIVE)
2240 		LLE_WLOCK(lle);
2241 	else
2242 		LLE_RLOCK(lle);
2243 
2244 	return lle;
2245 }
2246 
2247 static int
2248 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2249     struct rt_walkarg *w)
2250 {
2251 	struct sockaddr_in sin;
2252 
2253 	LLTABLE_LOCK_ASSERT();
2254 
2255 	/* skip deleted entries */
2256 	if (lle->la_flags & LLE_DELETED)
2257 		return 0;
2258 
2259 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2260 
2261 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2262 }
2263 
2264 #endif /* NARP > 0 */
2265 
2266 static int
2267 in_multicast_sysctl(SYSCTLFN_ARGS)
2268 {
2269 	struct ifnet *ifp;
2270 	struct ifaddr *ifa;
2271 	struct in_ifaddr *ifa4;
2272 	struct in_multi *inm;
2273 	uint32_t tmp;
2274 	int error;
2275 	size_t written;
2276 	struct psref psref;
2277 	int bound;
2278 
2279 	if (namelen != 1)
2280 		return EINVAL;
2281 
2282 	bound = curlwp_bind();
2283 	ifp = if_get_byindex(name[0], &psref);
2284 	if (ifp == NULL) {
2285 		curlwp_bindx(bound);
2286 		return ENODEV;
2287 	}
2288 
2289 	if (oldp == NULL) {
2290 		*oldlenp = 0;
2291 		IFADDR_FOREACH(ifa, ifp) {
2292 			if (ifa->ifa_addr->sa_family != AF_INET)
2293 				continue;
2294 			ifa4 = (void *)ifa;
2295 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2296 				*oldlenp += 2 * sizeof(struct in_addr) +
2297 				    sizeof(uint32_t);
2298 			}
2299 		}
2300 		if_put(ifp, &psref);
2301 		curlwp_bindx(bound);
2302 		return 0;
2303 	}
2304 
2305 	error = 0;
2306 	written = 0;
2307 	IFADDR_FOREACH(ifa, ifp) {
2308 		if (ifa->ifa_addr->sa_family != AF_INET)
2309 			continue;
2310 		ifa4 = (void *)ifa;
2311 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2312 			if (written + 2 * sizeof(struct in_addr) +
2313 			    sizeof(uint32_t) > *oldlenp)
2314 				goto done;
2315 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2316 			    oldp, sizeof(struct in_addr));
2317 			if (error)
2318 				goto done;
2319 			oldp = (char *)oldp + sizeof(struct in_addr);
2320 			written += sizeof(struct in_addr);
2321 			error = sysctl_copyout(l, &inm->inm_addr,
2322 			    oldp, sizeof(struct in_addr));
2323 			if (error)
2324 				goto done;
2325 			oldp = (char *)oldp + sizeof(struct in_addr);
2326 			written += sizeof(struct in_addr);
2327 			tmp = inm->inm_refcount;
2328 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2329 			if (error)
2330 				goto done;
2331 			oldp = (char *)oldp + sizeof(tmp);
2332 			written += sizeof(tmp);
2333 		}
2334 	}
2335 done:
2336 	if_put(ifp, &psref);
2337 	curlwp_bindx(bound);
2338 	*oldlenp = written;
2339 	return error;
2340 }
2341 
2342 static void
2343 in_sysctl_init(struct sysctllog **clog)
2344 {
2345 	sysctl_createv(clog, 0, NULL, NULL,
2346 		       CTLFLAG_PERMANENT,
2347 		       CTLTYPE_NODE, "inet",
2348 		       SYSCTL_DESCR("PF_INET related settings"),
2349 		       NULL, 0, NULL, 0,
2350 		       CTL_NET, PF_INET, CTL_EOL);
2351 	sysctl_createv(clog, 0, NULL, NULL,
2352 		       CTLFLAG_PERMANENT,
2353 		       CTLTYPE_NODE, "multicast",
2354 		       SYSCTL_DESCR("Multicast information"),
2355 		       in_multicast_sysctl, 0, NULL, 0,
2356 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2357 	sysctl_createv(clog, 0, NULL, NULL,
2358 		       CTLFLAG_PERMANENT,
2359 		       CTLTYPE_NODE, "ip",
2360 		       SYSCTL_DESCR("IPv4 related settings"),
2361 		       NULL, 0, NULL, 0,
2362 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2363 
2364 	sysctl_createv(clog, 0, NULL, NULL,
2365 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2366 		       CTLTYPE_INT, "subnetsarelocal",
2367 		       SYSCTL_DESCR("Whether logical subnets are considered "
2368 				    "local"),
2369 		       NULL, 0, &subnetsarelocal, 0,
2370 		       CTL_NET, PF_INET, IPPROTO_IP,
2371 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2372 	sysctl_createv(clog, 0, NULL, NULL,
2373 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2374 		       CTLTYPE_INT, "hostzerobroadcast",
2375 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2376 		       NULL, 0, &hostzeroisbroadcast, 0,
2377 		       CTL_NET, PF_INET, IPPROTO_IP,
2378 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2379 }
2380 
2381 #if NARP > 0
2382 
2383 static struct lltable *
2384 in_lltattach(struct ifnet *ifp)
2385 {
2386 	struct lltable *llt;
2387 
2388 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2389 	llt->llt_af = AF_INET;
2390 	llt->llt_ifp = ifp;
2391 
2392 	llt->llt_lookup = in_lltable_lookup;
2393 	llt->llt_create = in_lltable_create;
2394 	llt->llt_delete = in_lltable_delete;
2395 	llt->llt_dump_entry = in_lltable_dump_entry;
2396 	llt->llt_hash = in_lltable_hash;
2397 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2398 	llt->llt_free_entry = in_lltable_free_entry;
2399 	llt->llt_match_prefix = in_lltable_match_prefix;
2400 	lltable_link(llt);
2401 
2402 	return (llt);
2403 }
2404 
2405 #endif /* NARP > 0 */
2406 
2407 void *
2408 in_domifattach(struct ifnet *ifp)
2409 {
2410 	struct in_ifinfo *ii;
2411 
2412 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2413 
2414 #if NARP > 0
2415 	ii->ii_llt = in_lltattach(ifp);
2416 #endif
2417 
2418 #ifdef IPSELSRC
2419 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2420 	KASSERT(ii->ii_selsrc != NULL);
2421 #endif
2422 
2423 	return ii;
2424 }
2425 
2426 void
2427 in_domifdetach(struct ifnet *ifp, void *aux)
2428 {
2429 	struct in_ifinfo *ii = aux;
2430 
2431 #ifdef IPSELSRC
2432 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2433 #endif
2434 #if NARP > 0
2435 	lltable_free(ii->ii_llt);
2436 #endif
2437 	kmem_free(ii, sizeof(struct in_ifinfo));
2438 }
2439