xref: /netbsd-src/sys/netinet/in.c (revision 87d689fb734c654d2486f87f7be32f1b53ecdbec)
1 /*	$NetBSD: in.c,v 1.214 2018/01/10 10:56:30 knakahara 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 1.214 2018/01/10 10:56:30 knakahara 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)); XXX need ifa_not_held (psref_not_held) */
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 
916 	IFNET_LOCK(ifp);
917 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
918 	igmp_purgeif(ifp);		/* manipulates pools */
919 #ifdef MROUTING
920 	ip_mrouter_detach(ifp);
921 #endif
922 	IFNET_UNLOCK(ifp);
923 }
924 
925 /*
926  * SIOC[GAD]LIFADDR.
927  *	SIOCGLIFADDR: get first address. (???)
928  *	SIOCGLIFADDR with IFLR_PREFIX:
929  *		get first address that matches the specified prefix.
930  *	SIOCALIFADDR: add the specified address.
931  *	SIOCALIFADDR with IFLR_PREFIX:
932  *		EINVAL since we can't deduce hostid part of the address.
933  *	SIOCDLIFADDR: delete the specified address.
934  *	SIOCDLIFADDR with IFLR_PREFIX:
935  *		delete the first address that matches the specified prefix.
936  * return values:
937  *	EINVAL on invalid parameters
938  *	EADDRNOTAVAIL on prefix match failed/specified address not found
939  *	other values may be returned from in_ioctl()
940  */
941 static int
942 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
943     struct ifnet *ifp)
944 {
945 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
946 	struct ifaddr *ifa;
947 	struct sockaddr *sa;
948 
949 	/* sanity checks */
950 	if (data == NULL || ifp == NULL) {
951 		panic("invalid argument to in_lifaddr_ioctl");
952 		/*NOTRECHED*/
953 	}
954 
955 	switch (cmd) {
956 	case SIOCGLIFADDR:
957 		/* address must be specified on GET with IFLR_PREFIX */
958 		if ((iflr->flags & IFLR_PREFIX) == 0)
959 			break;
960 		/*FALLTHROUGH*/
961 	case SIOCALIFADDR:
962 	case SIOCDLIFADDR:
963 		/* address must be specified on ADD and DELETE */
964 		sa = (struct sockaddr *)&iflr->addr;
965 		if (sa->sa_family != AF_INET)
966 			return EINVAL;
967 		if (sa->sa_len != sizeof(struct sockaddr_in))
968 			return EINVAL;
969 		/* XXX need improvement */
970 		sa = (struct sockaddr *)&iflr->dstaddr;
971 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
972 			return EINVAL;
973 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
974 			return EINVAL;
975 		break;
976 	default: /*shouldn't happen*/
977 #if 0
978 		panic("invalid cmd to in_lifaddr_ioctl");
979 		/*NOTREACHED*/
980 #else
981 		return EOPNOTSUPP;
982 #endif
983 	}
984 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
985 		return EINVAL;
986 
987 	switch (cmd) {
988 	case SIOCALIFADDR:
989 	    {
990 		struct in_aliasreq ifra;
991 
992 		if (iflr->flags & IFLR_PREFIX)
993 			return EINVAL;
994 
995 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
996 		memset(&ifra, 0, sizeof(ifra));
997 		memcpy(ifra.ifra_name, iflr->iflr_name,
998 			sizeof(ifra.ifra_name));
999 
1000 		memcpy(&ifra.ifra_addr, &iflr->addr,
1001 			((struct sockaddr *)&iflr->addr)->sa_len);
1002 
1003 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1004 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1005 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1006 		}
1007 
1008 		ifra.ifra_mask.sin_family = AF_INET;
1009 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
1010 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
1011 
1012 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
1013 	    }
1014 	case SIOCGLIFADDR:
1015 	case SIOCDLIFADDR:
1016 	    {
1017 		struct in_ifaddr *ia;
1018 		struct in_addr mask, candidate, match;
1019 		struct sockaddr_in *sin;
1020 		int cmp, s;
1021 
1022 		memset(&mask, 0, sizeof(mask));
1023 		memset(&match, 0, sizeof(match));	/* XXX gcc */
1024 		if (iflr->flags & IFLR_PREFIX) {
1025 			/* lookup a prefix rather than address. */
1026 			in_len2mask(&mask, iflr->prefixlen);
1027 
1028 			sin = (struct sockaddr_in *)&iflr->addr;
1029 			match.s_addr = sin->sin_addr.s_addr;
1030 			match.s_addr &= mask.s_addr;
1031 
1032 			/* if you set extra bits, that's wrong */
1033 			if (match.s_addr != sin->sin_addr.s_addr)
1034 				return EINVAL;
1035 
1036 			cmp = 1;
1037 		} else {
1038 			if (cmd == SIOCGLIFADDR) {
1039 				/* on getting an address, take the 1st match */
1040 				cmp = 0;	/*XXX*/
1041 			} else {
1042 				/* on deleting an address, do exact match */
1043 				in_len2mask(&mask, 32);
1044 				sin = (struct sockaddr_in *)&iflr->addr;
1045 				match.s_addr = sin->sin_addr.s_addr;
1046 
1047 				cmp = 1;
1048 			}
1049 		}
1050 
1051 		s = pserialize_read_enter();
1052 		IFADDR_READER_FOREACH(ifa, ifp) {
1053 			if (ifa->ifa_addr->sa_family != AF_INET)
1054 				continue;
1055 			if (cmp == 0)
1056 				break;
1057 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
1058 			candidate.s_addr &= mask.s_addr;
1059 			if (candidate.s_addr == match.s_addr)
1060 				break;
1061 		}
1062 		if (ifa == NULL) {
1063 			pserialize_read_exit(s);
1064 			return EADDRNOTAVAIL;
1065 		}
1066 		ia = (struct in_ifaddr *)ifa;
1067 
1068 		if (cmd == SIOCGLIFADDR) {
1069 			/* fill in the if_laddrreq structure */
1070 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1071 
1072 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1073 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1074 					ia->ia_dstaddr.sin_len);
1075 			} else
1076 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1077 
1078 			iflr->prefixlen =
1079 				in_mask2len(&ia->ia_sockmask.sin_addr);
1080 
1081 			iflr->flags = 0;	/*XXX*/
1082 			pserialize_read_exit(s);
1083 
1084 			return 0;
1085 		} else {
1086 			struct in_aliasreq ifra;
1087 
1088 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1089 			memset(&ifra, 0, sizeof(ifra));
1090 			memcpy(ifra.ifra_name, iflr->iflr_name,
1091 				sizeof(ifra.ifra_name));
1092 
1093 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1094 				ia->ia_addr.sin_len);
1095 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1096 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1097 					ia->ia_dstaddr.sin_len);
1098 			}
1099 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1100 				ia->ia_sockmask.sin_len);
1101 			pserialize_read_exit(s);
1102 
1103 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1104 		}
1105 	    }
1106 	}
1107 
1108 	return EOPNOTSUPP;	/*just for safety*/
1109 }
1110 
1111 /*
1112  * Initialize an interface's internet address
1113  * and routing table entry.
1114  */
1115 int
1116 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1117     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1118 {
1119 	u_int32_t i;
1120 	struct sockaddr_in oldaddr, olddst;
1121 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1122 
1123 	if (sin == NULL)
1124 		sin = &ia->ia_addr;
1125 	if (dst == NULL)
1126 		dst = &ia->ia_dstaddr;
1127 
1128 	/*
1129 	 * Set up new addresses.
1130 	 */
1131 	oldaddr = ia->ia_addr;
1132 	olddst = ia->ia_dstaddr;
1133 	oldflags = ia->ia4_flags;
1134 	ia->ia_addr = *sin;
1135 	ia->ia_dstaddr = *dst;
1136 	hostIsNew = oldaddr.sin_family != AF_INET ||
1137 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1138 	if (!scrub)
1139 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1140 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1141 
1142 	/*
1143 	 * Configure address flags.
1144 	 * We need to do this early because they maybe adjusted
1145 	 * by if_addr_init depending on the address.
1146 	 */
1147 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1148 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1149 		hostIsNew = 1;
1150 	}
1151 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1152 		ia->ia4_flags |= IN_IFF_DETACHED;
1153 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1154 	} else if (hostIsNew && if_do_dad(ifp))
1155 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1156 
1157 	/*
1158 	 * Give the interface a chance to initialize
1159 	 * if this is its first address,
1160 	 * and to validate the address if necessary.
1161 	 */
1162 	s = splsoftnet();
1163 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1164 	splx(s);
1165 	/* Now clear the try tentative flag, its job is done. */
1166 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1167 	if (error != 0) {
1168 		ia->ia_addr = oldaddr;
1169 		ia->ia_dstaddr = olddst;
1170 		ia->ia4_flags = oldflags;
1171 		return error;
1172 	}
1173 
1174 	if (scrub || hostIsNew) {
1175 		int newflags = ia->ia4_flags;
1176 
1177 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1178 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1179 		ia->ia4_flags = oldflags;
1180 		if (hostIsNew)
1181 			in_scrubaddr(ia);
1182 		else if (scrub)
1183 			in_scrubprefix(ia);
1184 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1185 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1186 		ia->ia4_flags = newflags;
1187 	}
1188 
1189 	i = ia->ia_addr.sin_addr.s_addr;
1190 	if (ifp->if_flags & IFF_POINTOPOINT)
1191 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1192 	else if (IN_CLASSA(i))
1193 		ia->ia_netmask = IN_CLASSA_NET;
1194 	else if (IN_CLASSB(i))
1195 		ia->ia_netmask = IN_CLASSB_NET;
1196 	else
1197 		ia->ia_netmask = IN_CLASSC_NET;
1198 	/*
1199 	 * The subnet mask usually includes at least the standard network part,
1200 	 * but may may be smaller in the case of supernetting.
1201 	 * If it is set, we believe it.
1202 	 */
1203 	if (ia->ia_subnetmask == 0) {
1204 		ia->ia_subnetmask = ia->ia_netmask;
1205 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1206 	} else
1207 		ia->ia_netmask &= ia->ia_subnetmask;
1208 
1209 	ia->ia_net = i & ia->ia_netmask;
1210 	ia->ia_subnet = i & ia->ia_subnetmask;
1211 	in_socktrim(&ia->ia_sockmask);
1212 
1213 	/* re-calculate the "in_maxmtu" value */
1214 	in_setmaxmtu();
1215 
1216 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1217 	if (ifp->if_flags & IFF_BROADCAST) {
1218 		ia->ia_broadaddr.sin_addr.s_addr =
1219 			ia->ia_subnet | ~ia->ia_subnetmask;
1220 		ia->ia_netbroadcast.s_addr =
1221 			ia->ia_net | ~ia->ia_netmask;
1222 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1223 		ia->ia_dstaddr = ia->ia_addr;
1224 		flags |= RTF_HOST;
1225 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1226 		if (ia->ia_dstaddr.sin_family != AF_INET)
1227 			return (0);
1228 		flags |= RTF_HOST;
1229 	}
1230 
1231 	/* Add the local route to the address */
1232 	in_ifaddlocal(&ia->ia_ifa);
1233 
1234 	/* Add the prefix route for the address */
1235 	error = in_addprefix(ia, flags);
1236 
1237 	/*
1238 	 * If the interface supports multicast, join the "all hosts"
1239 	 * multicast group on that interface.
1240 	 */
1241 	mutex_enter(&in_ifaddr_lock);
1242 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1243 		struct in_addr addr;
1244 
1245 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1246 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1247 	}
1248 	mutex_exit(&in_ifaddr_lock);
1249 
1250 	if (hostIsNew &&
1251 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1252 	    if_do_dad(ifp))
1253 		ia->ia_dad_start((struct ifaddr *)ia);
1254 
1255 	return error;
1256 }
1257 
1258 #define rtinitflags(x) \
1259 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1260 	    ? RTF_HOST : 0)
1261 
1262 /*
1263  * add a route to prefix ("connected route" in cisco terminology).
1264  * does nothing if there's some interface address with the same prefix already.
1265  */
1266 static int
1267 in_addprefix(struct in_ifaddr *target, int flags)
1268 {
1269 	struct in_ifaddr *ia;
1270 	struct in_addr prefix, mask, p;
1271 	int error;
1272 	int s;
1273 
1274 	if ((flags & RTF_HOST) != 0)
1275 		prefix = target->ia_dstaddr.sin_addr;
1276 	else {
1277 		prefix = target->ia_addr.sin_addr;
1278 		mask = target->ia_sockmask.sin_addr;
1279 		prefix.s_addr &= mask.s_addr;
1280 	}
1281 
1282 	s = pserialize_read_enter();
1283 	IN_ADDRLIST_READER_FOREACH(ia) {
1284 		if (rtinitflags(ia))
1285 			p = ia->ia_dstaddr.sin_addr;
1286 		else {
1287 			p = ia->ia_addr.sin_addr;
1288 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1289 		}
1290 
1291 		if (prefix.s_addr != p.s_addr)
1292 			continue;
1293 
1294 		/*
1295 		 * if we got a matching prefix route inserted by other
1296 		 * interface address, we don't need to bother
1297 		 *
1298 		 * XXX RADIX_MPATH implications here? -dyoung
1299 		 */
1300 		if (ia->ia_flags & IFA_ROUTE) {
1301 			pserialize_read_exit(s);
1302 			return 0;
1303 		}
1304 	}
1305 	pserialize_read_exit(s);
1306 
1307 	/*
1308 	 * noone seem to have prefix route.  insert it.
1309 	 */
1310 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1311 	if (error == 0)
1312 		target->ia_flags |= IFA_ROUTE;
1313 	else if (error == EEXIST) {
1314 		/*
1315 		 * the fact the route already exists is not an error.
1316 		 */
1317 		error = 0;
1318 	}
1319 	return error;
1320 }
1321 
1322 /*
1323  * remove a route to prefix ("connected route" in cisco terminology).
1324  * re-installs the route by using another interface address, if there's one
1325  * with the same prefix (otherwise we lose the route mistakenly).
1326  */
1327 static int
1328 in_scrubprefix(struct in_ifaddr *target)
1329 {
1330 	struct in_ifaddr *ia;
1331 	struct in_addr prefix, mask, p;
1332 	int error;
1333 	int s;
1334 
1335 	/* If we don't have IFA_ROUTE we have nothing to do */
1336 	if ((target->ia_flags & IFA_ROUTE) == 0)
1337 		return 0;
1338 
1339 	if (rtinitflags(target))
1340 		prefix = target->ia_dstaddr.sin_addr;
1341 	else {
1342 		prefix = target->ia_addr.sin_addr;
1343 		mask = target->ia_sockmask.sin_addr;
1344 		prefix.s_addr &= mask.s_addr;
1345 	}
1346 
1347 	s = pserialize_read_enter();
1348 	IN_ADDRLIST_READER_FOREACH(ia) {
1349 		if (rtinitflags(ia))
1350 			p = ia->ia_dstaddr.sin_addr;
1351 		else {
1352 			p = ia->ia_addr.sin_addr;
1353 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1354 		}
1355 
1356 		if (prefix.s_addr != p.s_addr)
1357 			continue;
1358 
1359 		/*
1360 		 * if we got a matching prefix route, move IFA_ROUTE to him
1361 		 */
1362 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1363 			struct psref psref;
1364 			int bound = curlwp_bind();
1365 
1366 			ia4_acquire(ia, &psref);
1367 			pserialize_read_exit(s);
1368 
1369 			rtinit(&target->ia_ifa, RTM_DELETE,
1370 			    rtinitflags(target));
1371 			target->ia_flags &= ~IFA_ROUTE;
1372 
1373 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1374 			    rtinitflags(ia) | RTF_UP);
1375 			if (error == 0)
1376 				ia->ia_flags |= IFA_ROUTE;
1377 
1378 			ia4_release(ia, &psref);
1379 			curlwp_bindx(bound);
1380 
1381 			return error;
1382 		}
1383 	}
1384 	pserialize_read_exit(s);
1385 
1386 	/*
1387 	 * noone seem to have prefix route.  remove it.
1388 	 */
1389 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1390 	target->ia_flags &= ~IFA_ROUTE;
1391 	return 0;
1392 }
1393 
1394 #undef rtinitflags
1395 
1396 /*
1397  * Return 1 if the address might be a local broadcast address.
1398  */
1399 int
1400 in_broadcast(struct in_addr in, struct ifnet *ifp)
1401 {
1402 	struct ifaddr *ifa;
1403 	int s;
1404 
1405 	KASSERT(ifp != NULL);
1406 
1407 	if (in.s_addr == INADDR_BROADCAST ||
1408 	    in_nullhost(in))
1409 		return 1;
1410 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1411 		return 0;
1412 	/*
1413 	 * Look through the list of addresses for a match
1414 	 * with a broadcast address.
1415 	 */
1416 #define ia (ifatoia(ifa))
1417 	s = pserialize_read_enter();
1418 	IFADDR_READER_FOREACH(ifa, ifp) {
1419 		if (ifa->ifa_addr->sa_family == AF_INET &&
1420 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1421 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1422 		     in_hosteq(in, ia->ia_netbroadcast) ||
1423 		     (hostzeroisbroadcast &&
1424 		      /*
1425 		       * Check for old-style (host 0) broadcast.
1426 		       */
1427 		      (in.s_addr == ia->ia_subnet ||
1428 		       in.s_addr == ia->ia_net)))) {
1429 			pserialize_read_exit(s);
1430 			return 1;
1431 		}
1432 	}
1433 	pserialize_read_exit(s);
1434 	return (0);
1435 #undef ia
1436 }
1437 
1438 /*
1439  * perform DAD when interface becomes IFF_UP.
1440  */
1441 void
1442 in_if_link_up(struct ifnet *ifp)
1443 {
1444 	struct ifaddr *ifa;
1445 	struct in_ifaddr *ia;
1446 	int s, bound;
1447 
1448 	/* Ensure it's sane to run DAD */
1449 	if (ifp->if_link_state == LINK_STATE_DOWN)
1450 		return;
1451 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1452 		return;
1453 
1454 	bound = curlwp_bind();
1455 	s = pserialize_read_enter();
1456 	IFADDR_READER_FOREACH(ifa, ifp) {
1457 		struct psref psref;
1458 
1459 		if (ifa->ifa_addr->sa_family != AF_INET)
1460 			continue;
1461 		ifa_acquire(ifa, &psref);
1462 		pserialize_read_exit(s);
1463 
1464 		ia = (struct in_ifaddr *)ifa;
1465 
1466 		/* If detached then mark as tentative */
1467 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1468 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1469 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1470 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1471 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1472 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1473 		}
1474 
1475 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1476 			/* Clear the duplicated flag as we're starting DAD. */
1477 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1478 			ia->ia_dad_start(ifa);
1479 		}
1480 
1481 		s = pserialize_read_enter();
1482 		ifa_release(ifa, &psref);
1483 	}
1484 	pserialize_read_exit(s);
1485 	curlwp_bindx(bound);
1486 }
1487 
1488 void
1489 in_if_up(struct ifnet *ifp)
1490 {
1491 
1492 	/* interface may not support link state, so bring it up also */
1493 	in_if_link_up(ifp);
1494 }
1495 
1496 /*
1497  * Mark all addresses as detached.
1498  */
1499 void
1500 in_if_link_down(struct ifnet *ifp)
1501 {
1502 	struct ifaddr *ifa;
1503 	struct in_ifaddr *ia;
1504 	int s, bound;
1505 
1506 	bound = curlwp_bind();
1507 	s = pserialize_read_enter();
1508 	IFADDR_READER_FOREACH(ifa, ifp) {
1509 		struct psref psref;
1510 
1511 		if (ifa->ifa_addr->sa_family != AF_INET)
1512 			continue;
1513 		ifa_acquire(ifa, &psref);
1514 		pserialize_read_exit(s);
1515 
1516 		ia = (struct in_ifaddr *)ifa;
1517 
1518 		/* Stop DAD processing */
1519 		if (ia->ia_dad_stop != NULL)
1520 			ia->ia_dad_stop(ifa);
1521 
1522 		/*
1523 		 * Mark the address as detached.
1524 		 */
1525 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1526 			ia->ia4_flags |= IN_IFF_DETACHED;
1527 			ia->ia4_flags &=
1528 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1529 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1530 		}
1531 
1532 		s = pserialize_read_enter();
1533 		ifa_release(ifa, &psref);
1534 	}
1535 	pserialize_read_exit(s);
1536 	curlwp_bindx(bound);
1537 }
1538 
1539 void
1540 in_if_down(struct ifnet *ifp)
1541 {
1542 
1543 	in_if_link_down(ifp);
1544 	lltable_purge_entries(LLTABLE(ifp));
1545 }
1546 
1547 void
1548 in_if_link_state_change(struct ifnet *ifp, int link_state)
1549 {
1550 
1551 	switch (link_state) {
1552 	case LINK_STATE_DOWN:
1553 		in_if_link_down(ifp);
1554 		break;
1555 	case LINK_STATE_UP:
1556 		in_if_link_up(ifp);
1557 		break;
1558 	}
1559 }
1560 
1561 /*
1562  * in_lookup_multi: look up the in_multi record for a given IP
1563  * multicast address on a given interface.  If no matching record is
1564  * found, return NULL.
1565  */
1566 struct in_multi *
1567 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1568 {
1569 	struct in_multi *inm;
1570 
1571 	KASSERT(rw_lock_held(&in_multilock));
1572 
1573 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1574 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1575 			break;
1576 	}
1577 	return inm;
1578 }
1579 
1580 /*
1581  * in_multi_group: check whether the address belongs to an IP multicast
1582  * group we are joined on this interface.  Returns true or false.
1583  */
1584 bool
1585 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1586 {
1587 	bool ingroup;
1588 
1589 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1590 		rw_enter(&in_multilock, RW_READER);
1591 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1592 		rw_exit(&in_multilock);
1593 	} else {
1594 		/* XXX Recursive call from ip_output(). */
1595 		KASSERT(rw_lock_held(&in_multilock));
1596 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1597 	}
1598 	return ingroup;
1599 }
1600 
1601 /*
1602  * Add an address to the list of IP multicast addresses for a given interface.
1603  */
1604 struct in_multi *
1605 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1606 {
1607 	struct sockaddr_in sin;
1608 	struct in_multi *inm;
1609 
1610 	/*
1611 	 * See if address already in list.
1612 	 */
1613 	rw_enter(&in_multilock, RW_WRITER);
1614 	inm = in_lookup_multi(*ap, ifp);
1615 	if (inm != NULL) {
1616 		/*
1617 		 * Found it; just increment the reference count.
1618 		 */
1619 		inm->inm_refcount++;
1620 		rw_exit(&in_multilock);
1621 		return inm;
1622 	}
1623 
1624 	/*
1625 	 * New address; allocate a new multicast record.
1626 	 */
1627 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1628 	if (inm == NULL) {
1629 		rw_exit(&in_multilock);
1630 		return NULL;
1631 	}
1632 	inm->inm_addr = *ap;
1633 	inm->inm_ifp = ifp;
1634 	inm->inm_refcount = 1;
1635 
1636 	/*
1637 	 * Ask the network driver to update its multicast reception
1638 	 * filter appropriately for the new address.
1639 	 */
1640 	sockaddr_in_init(&sin, ap, 0);
1641 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1642 		rw_exit(&in_multilock);
1643 		pool_put(&inmulti_pool, inm);
1644 		return NULL;
1645 	}
1646 
1647 	/*
1648 	 * Let IGMP know that we have joined a new IP multicast group.
1649 	 */
1650 	if (igmp_joingroup(inm) != 0) {
1651 		rw_exit(&in_multilock);
1652 		pool_put(&inmulti_pool, inm);
1653 		return NULL;
1654 	}
1655 	LIST_INSERT_HEAD(
1656 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1657 	    inm, inm_list);
1658 	in_multientries++;
1659 	rw_exit(&in_multilock);
1660 
1661 	return inm;
1662 }
1663 
1664 /*
1665  * Delete a multicast address record.
1666  */
1667 void
1668 in_delmulti(struct in_multi *inm)
1669 {
1670 	struct sockaddr_in sin;
1671 
1672 	rw_enter(&in_multilock, RW_WRITER);
1673 	if (--inm->inm_refcount > 0) {
1674 		rw_exit(&in_multilock);
1675 		return;
1676 	}
1677 
1678 	/*
1679 	 * No remaining claims to this record; let IGMP know that
1680 	 * we are leaving the multicast group.
1681 	 */
1682 	igmp_leavegroup(inm);
1683 
1684 	/*
1685 	 * Notify the network driver to update its multicast reception
1686 	 * filter.
1687 	 */
1688 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1689 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1690 
1691 	/*
1692 	 * Unlink from list.
1693 	 */
1694 	LIST_REMOVE(inm, inm_list);
1695 	in_multientries--;
1696 	rw_exit(&in_multilock);
1697 
1698 	pool_put(&inmulti_pool, inm);
1699 }
1700 
1701 /*
1702  * in_next_multi: step through all of the in_multi records, one at a time.
1703  * The current position is remembered in "step", which the caller must
1704  * provide.  in_first_multi(), below, must be called to initialize "step"
1705  * and get the first record.  Both macros return a NULL "inm" when there
1706  * are no remaining records.
1707  */
1708 struct in_multi *
1709 in_next_multi(struct in_multistep *step)
1710 {
1711 	struct in_multi *inm;
1712 
1713 	KASSERT(rw_lock_held(&in_multilock));
1714 
1715 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1716 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1717 	}
1718 	if ((inm = step->i_inm) != NULL) {
1719 		step->i_inm = LIST_NEXT(inm, inm_list);
1720 	}
1721 	return inm;
1722 }
1723 
1724 struct in_multi *
1725 in_first_multi(struct in_multistep *step)
1726 {
1727 	KASSERT(rw_lock_held(&in_multilock));
1728 
1729 	step->i_n = 0;
1730 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1731 	return in_next_multi(step);
1732 }
1733 
1734 void
1735 in_multi_lock(int op)
1736 {
1737 	rw_enter(&in_multilock, op);
1738 }
1739 
1740 void
1741 in_multi_unlock(void)
1742 {
1743 	rw_exit(&in_multilock);
1744 }
1745 
1746 int
1747 in_multi_lock_held(void)
1748 {
1749 	return rw_lock_held(&in_multilock);
1750 }
1751 
1752 struct in_ifaddr *
1753 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1754     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1755 {
1756 	struct rtentry *rt = NULL;
1757 	struct in_ifaddr *ia = NULL;
1758 
1759 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1760 	/*
1761          * If route is known or can be allocated now, take the
1762          * source address from the interface.  Otherwise, punt.
1763 	 */
1764 	if ((soopts & SO_DONTROUTE) != 0)
1765 		rtcache_free(ro);
1766 	else {
1767 		union {
1768 			struct sockaddr		dst;
1769 			struct sockaddr_in	dst4;
1770 		} u;
1771 
1772 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1773 		rt = rtcache_lookup(ro, &u.dst);
1774 	}
1775 	/*
1776 	 * If we found a route, use the address
1777 	 * corresponding to the outgoing interface
1778 	 * unless it is the loopback (in case a route
1779 	 * to our address on another net goes to loopback).
1780 	 *
1781 	 * XXX Is this still true?  Do we care?
1782 	 */
1783 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1784 		int s;
1785 		struct ifaddr *ifa;
1786 		/*
1787 		 * Just in case. May not need to do this workaround.
1788 		 * Revisit when working on rtentry MP-ification.
1789 		 */
1790 		s = pserialize_read_enter();
1791 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1792 			if (ifa == rt->rt_ifa)
1793 				break;
1794 		}
1795 		if (ifa != NULL)
1796 			ifa_acquire(ifa, psref);
1797 		pserialize_read_exit(s);
1798 
1799 		ia = ifatoia(ifa);
1800 	}
1801 	if (ia == NULL) {
1802 		u_int16_t fport = sin->sin_port;
1803 		struct ifaddr *ifa;
1804 		int s;
1805 
1806 		sin->sin_port = 0;
1807 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1808 		sin->sin_port = fport;
1809 		if (ifa == NULL) {
1810 			/* Find 1st non-loopback AF_INET address */
1811 			s = pserialize_read_enter();
1812 			IN_ADDRLIST_READER_FOREACH(ia) {
1813 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1814 					break;
1815 			}
1816 			if (ia != NULL)
1817 				ia4_acquire(ia, psref);
1818 			pserialize_read_exit(s);
1819 		} else {
1820 			/* ia is already referenced by psref */
1821 			ia = ifatoia(ifa);
1822 		}
1823 		if (ia == NULL) {
1824 			*errorp = EADDRNOTAVAIL;
1825 			goto out;
1826 		}
1827 	}
1828 	/*
1829 	 * If the destination address is multicast and an outgoing
1830 	 * interface has been set as a multicast option, use the
1831 	 * address of that interface as our source address.
1832 	 */
1833 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1834 		struct ip_moptions *imo;
1835 
1836 		imo = mopts;
1837 		if (imo->imo_multicast_if_index != 0) {
1838 			struct ifnet *ifp;
1839 			int s;
1840 
1841 			if (ia != NULL)
1842 				ia4_release(ia, psref);
1843 			s = pserialize_read_enter();
1844 			ifp = if_byindex(imo->imo_multicast_if_index);
1845 			if (ifp != NULL) {
1846 				/* XXX */
1847 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1848 			} else
1849 				ia = NULL;
1850 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1851 				pserialize_read_exit(s);
1852 				if (ia != NULL)
1853 					ia4_release(ia, psref);
1854 				*errorp = EADDRNOTAVAIL;
1855 				ia = NULL;
1856 				goto out;
1857 			}
1858 			pserialize_read_exit(s);
1859 		}
1860 	}
1861 	if (ia->ia_ifa.ifa_getifa != NULL) {
1862 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1863 		                                      sintosa(sin)));
1864 		if (ia == NULL) {
1865 			*errorp = EADDRNOTAVAIL;
1866 			goto out;
1867 		}
1868 		/* FIXME NOMPSAFE */
1869 		ia4_acquire(ia, psref);
1870 	}
1871 #ifdef GETIFA_DEBUG
1872 	else
1873 		printf("%s: missing ifa_getifa\n", __func__);
1874 #endif
1875 out:
1876 	rtcache_unref(rt, ro);
1877 	return ia;
1878 }
1879 
1880 int
1881 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst)
1882 {
1883 	struct in_ifaddr *ia4;
1884 	int s;
1885 
1886 	/* check for address match */
1887 	if (src.s_addr != ip->ip_dst.s_addr ||
1888 	    dst.s_addr != ip->ip_src.s_addr)
1889 		return 0;
1890 
1891 	/* martian filters on outer source - NOT done in ip_input! */
1892 	if (IN_MULTICAST(ip->ip_src.s_addr))
1893 		return 0;
1894 	switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) {
1895 	case 0:
1896 	case 127:
1897 	case 255:
1898 		return 0;
1899 	}
1900 	/* reject packets with broadcast on source */
1901 	s = pserialize_read_enter();
1902 	IN_ADDRLIST_READER_FOREACH(ia4) {
1903 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
1904 			continue;
1905 		if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
1906 			pserialize_read_exit(s);
1907 			return 0;
1908 		}
1909 	}
1910 	pserialize_read_exit(s);
1911 
1912 	/* NOTE: packet may dropped by uRPF */
1913 
1914 	/* return valid bytes length */
1915 	return sizeof(src) + sizeof(dst);
1916 }
1917 
1918 #if NARP > 0
1919 
1920 struct in_llentry {
1921 	struct llentry		base;
1922 };
1923 
1924 #define	IN_LLTBL_DEFAULT_HSIZE	32
1925 #define	IN_LLTBL_HASH(k, h) \
1926 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1927 
1928 /*
1929  * Do actual deallocation of @lle.
1930  * Called by LLE_FREE_LOCKED when number of references
1931  * drops to zero.
1932  */
1933 static void
1934 in_lltable_destroy_lle(struct llentry *lle)
1935 {
1936 
1937 	LLE_WUNLOCK(lle);
1938 	LLE_LOCK_DESTROY(lle);
1939 	kmem_intr_free(lle, sizeof(*lle));
1940 }
1941 
1942 static struct llentry *
1943 in_lltable_new(struct in_addr addr4, u_int flags)
1944 {
1945 	struct in_llentry *lle;
1946 
1947 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
1948 	if (lle == NULL)		/* NB: caller generates msg */
1949 		return NULL;
1950 
1951 	/*
1952 	 * For IPv4 this will trigger "arpresolve" to generate
1953 	 * an ARP request.
1954 	 */
1955 	lle->base.la_expire = time_uptime; /* mark expired */
1956 	lle->base.r_l3addr.addr4 = addr4;
1957 	lle->base.lle_refcnt = 1;
1958 	lle->base.lle_free = in_lltable_destroy_lle;
1959 	LLE_LOCK_INIT(&lle->base);
1960 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1961 
1962 	return (&lle->base);
1963 }
1964 
1965 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1966 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1967 
1968 static int
1969 in_lltable_match_prefix(const struct sockaddr *prefix,
1970     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1971 {
1972 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1973 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1974 	struct in_addr lle_addr;
1975 
1976 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1977 
1978 	/*
1979 	 * (flags & LLE_STATIC) means deleting all entries
1980 	 * including static ARP entries.
1981 	 */
1982 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) &&
1983 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1984 		return (1);
1985 
1986 	return (0);
1987 }
1988 
1989 static void
1990 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1991 {
1992 	struct ifnet *ifp __diagused;
1993 	size_t pkts_dropped;
1994 	bool locked = false;
1995 
1996 	LLE_WLOCK_ASSERT(lle);
1997 	KASSERT(llt != NULL);
1998 
1999 	/* Unlink entry from table if not already */
2000 	if ((lle->la_flags & LLE_LINKED) != 0) {
2001 		ifp = llt->llt_ifp;
2002 		IF_AFDATA_WLOCK_ASSERT(ifp);
2003 		lltable_unlink_entry(llt, lle);
2004 		locked = true;
2005 	}
2006 
2007 	/*
2008 	 * We need to release the lock here to lle_timer proceeds;
2009 	 * lle_timer should stop immediately if LLE_LINKED isn't set.
2010 	 * Note that we cannot pass lle->lle_lock to callout_halt
2011 	 * because it's a rwlock.
2012 	 */
2013 	LLE_ADDREF(lle);
2014 	LLE_WUNLOCK(lle);
2015 	if (locked)
2016 		IF_AFDATA_WUNLOCK(ifp);
2017 
2018 	/* cancel timer */
2019 	callout_halt(&lle->lle_timer, NULL);
2020 
2021 	LLE_WLOCK(lle);
2022 	LLE_REMREF(lle);
2023 
2024 	/* Drop hold queue */
2025 	pkts_dropped = llentry_free(lle);
2026 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
2027 
2028 	if (locked)
2029 		IF_AFDATA_WLOCK(ifp);
2030 }
2031 
2032 static int
2033 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr,
2034     const struct rtentry *rt)
2035 {
2036 	int error = EINVAL;
2037 
2038 	if (rt == NULL)
2039 		return error;
2040 
2041 	/*
2042 	 * If the gateway for an existing host route matches the target L3
2043 	 * address, which is a special route inserted by some implementation
2044 	 * such as MANET, and the interface is of the correct type, then
2045 	 * allow for ARP to proceed.
2046 	 */
2047 	if (rt->rt_flags & RTF_GATEWAY) {
2048 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
2049 		    rt->rt_ifp->if_type != IFT_ETHER ||
2050 #ifdef __FreeBSD__
2051 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
2052 #else
2053 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
2054 #endif
2055 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
2056 		    sizeof(in_addr_t)) != 0) {
2057 			goto error;
2058 		}
2059 	}
2060 
2061 	/*
2062 	 * Make sure that at least the destination address is covered
2063 	 * by the route. This is for handling the case where 2 or more
2064 	 * interfaces have the same prefix. An incoming packet arrives
2065 	 * on one interface and the corresponding outgoing packet leaves
2066 	 * another interface.
2067 	 */
2068 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
2069 		const char *sa, *mask, *addr, *lim;
2070 		int len;
2071 
2072 		mask = (const char *)rt_mask(rt);
2073 		/*
2074 		 * Just being extra cautious to avoid some custom
2075 		 * code getting into trouble.
2076 		 */
2077 		if (mask == NULL)
2078 			goto error;
2079 
2080 		sa = (const char *)rt_getkey(rt);
2081 		addr = (const char *)l3addr;
2082 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
2083 		lim = addr + len;
2084 
2085 		for ( ; addr < lim; sa++, mask++, addr++) {
2086 			if ((*sa ^ *addr) & *mask) {
2087 #ifdef DIAGNOSTIC
2088 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
2089 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
2090 #endif
2091 				goto error;
2092 			}
2093 		}
2094 	}
2095 
2096 	error = 0;
2097 error:
2098 	return error;
2099 }
2100 
2101 static inline uint32_t
2102 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
2103 {
2104 
2105 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
2106 }
2107 
2108 static uint32_t
2109 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
2110 {
2111 
2112 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
2113 }
2114 
2115 static void
2116 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2117 {
2118 	struct sockaddr_in *sin;
2119 
2120 	sin = (struct sockaddr_in *)sa;
2121 	memset(sin, 0, sizeof(*sin));
2122 	sin->sin_family = AF_INET;
2123 	sin->sin_len = sizeof(*sin);
2124 	sin->sin_addr = lle->r_l3addr.addr4;
2125 }
2126 
2127 static inline struct llentry *
2128 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2129 {
2130 	struct llentry *lle;
2131 	struct llentries *lleh;
2132 	u_int hashidx;
2133 
2134 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2135 	lleh = &llt->lle_head[hashidx];
2136 	LIST_FOREACH(lle, lleh, lle_next) {
2137 		if (lle->la_flags & LLE_DELETED)
2138 			continue;
2139 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2140 			break;
2141 	}
2142 
2143 	return (lle);
2144 }
2145 
2146 static int
2147 in_lltable_delete(struct lltable *llt, u_int flags,
2148     const struct sockaddr *l3addr)
2149 {
2150 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2151 	struct ifnet *ifp __diagused = llt->llt_ifp;
2152 	struct llentry *lle;
2153 
2154 	IF_AFDATA_WLOCK_ASSERT(ifp);
2155 	KASSERTMSG(l3addr->sa_family == AF_INET,
2156 	    "sin_family %d", l3addr->sa_family);
2157 
2158 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2159 	if (lle == NULL) {
2160 #ifdef DEBUG
2161 		char buf[64];
2162 		sockaddr_format(l3addr, buf, sizeof(buf));
2163 		log(LOG_INFO, "%s: cache for %s is not found\n",
2164 		    __func__, buf);
2165 #endif
2166 		return (ENOENT);
2167 	}
2168 
2169 	LLE_WLOCK(lle);
2170 	lle->la_flags |= LLE_DELETED;
2171 #ifdef DEBUG
2172 	{
2173 		char buf[64];
2174 		sockaddr_format(l3addr, buf, sizeof(buf));
2175 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2176 		    __func__, buf, lle);
2177 	}
2178 #endif
2179 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2180 		llentry_free(lle);
2181 	else
2182 		LLE_WUNLOCK(lle);
2183 
2184 	return (0);
2185 }
2186 
2187 static struct llentry *
2188 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr,
2189     const struct rtentry *rt)
2190 {
2191 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2192 	struct ifnet *ifp = llt->llt_ifp;
2193 	struct llentry *lle;
2194 
2195 	IF_AFDATA_WLOCK_ASSERT(ifp);
2196 	KASSERTMSG(l3addr->sa_family == AF_INET,
2197 	    "sin_family %d", l3addr->sa_family);
2198 
2199 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2200 
2201 	if (lle != NULL) {
2202 		LLE_WLOCK(lle);
2203 		return (lle);
2204 	}
2205 
2206 	/* no existing record, we need to create new one */
2207 
2208 	/*
2209 	 * A route that covers the given address must have
2210 	 * been installed 1st because we are doing a resolution,
2211 	 * verify this.
2212 	 */
2213 	if (!(flags & LLE_IFADDR) &&
2214 	    in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2215 		return (NULL);
2216 
2217 	lle = in_lltable_new(sin->sin_addr, flags);
2218 	if (lle == NULL) {
2219 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2220 		return (NULL);
2221 	}
2222 	lle->la_flags = flags;
2223 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2224 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2225 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2226 	}
2227 
2228 	lltable_link_entry(llt, lle);
2229 	LLE_WLOCK(lle);
2230 
2231 	return (lle);
2232 }
2233 
2234 /*
2235  * Return NULL if not found or marked for deletion.
2236  * If found return lle read locked.
2237  */
2238 static struct llentry *
2239 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2240 {
2241 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2242 	struct llentry *lle;
2243 
2244 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2245 	KASSERTMSG(l3addr->sa_family == AF_INET,
2246 	    "sin_family %d", l3addr->sa_family);
2247 
2248 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2249 
2250 	if (lle == NULL)
2251 		return NULL;
2252 
2253 	if (flags & LLE_EXCLUSIVE)
2254 		LLE_WLOCK(lle);
2255 	else
2256 		LLE_RLOCK(lle);
2257 
2258 	return lle;
2259 }
2260 
2261 static int
2262 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2263     struct rt_walkarg *w)
2264 {
2265 	struct sockaddr_in sin;
2266 
2267 	LLTABLE_LOCK_ASSERT();
2268 
2269 	/* skip deleted entries */
2270 	if (lle->la_flags & LLE_DELETED)
2271 		return 0;
2272 
2273 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2274 
2275 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2276 }
2277 
2278 #endif /* NARP > 0 */
2279 
2280 static int
2281 in_multicast_sysctl(SYSCTLFN_ARGS)
2282 {
2283 	struct ifnet *ifp;
2284 	struct ifaddr *ifa;
2285 	struct in_ifaddr *ifa4;
2286 	struct in_multi *inm;
2287 	uint32_t tmp;
2288 	int error;
2289 	size_t written;
2290 	struct psref psref;
2291 	int bound;
2292 
2293 	if (namelen != 1)
2294 		return EINVAL;
2295 
2296 	bound = curlwp_bind();
2297 	ifp = if_get_byindex(name[0], &psref);
2298 	if (ifp == NULL) {
2299 		curlwp_bindx(bound);
2300 		return ENODEV;
2301 	}
2302 
2303 	if (oldp == NULL) {
2304 		*oldlenp = 0;
2305 		IFADDR_FOREACH(ifa, ifp) {
2306 			if (ifa->ifa_addr->sa_family != AF_INET)
2307 				continue;
2308 			ifa4 = (void *)ifa;
2309 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2310 				*oldlenp += 2 * sizeof(struct in_addr) +
2311 				    sizeof(uint32_t);
2312 			}
2313 		}
2314 		if_put(ifp, &psref);
2315 		curlwp_bindx(bound);
2316 		return 0;
2317 	}
2318 
2319 	error = 0;
2320 	written = 0;
2321 	IFADDR_FOREACH(ifa, ifp) {
2322 		if (ifa->ifa_addr->sa_family != AF_INET)
2323 			continue;
2324 		ifa4 = (void *)ifa;
2325 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2326 			if (written + 2 * sizeof(struct in_addr) +
2327 			    sizeof(uint32_t) > *oldlenp)
2328 				goto done;
2329 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2330 			    oldp, sizeof(struct in_addr));
2331 			if (error)
2332 				goto done;
2333 			oldp = (char *)oldp + sizeof(struct in_addr);
2334 			written += sizeof(struct in_addr);
2335 			error = sysctl_copyout(l, &inm->inm_addr,
2336 			    oldp, sizeof(struct in_addr));
2337 			if (error)
2338 				goto done;
2339 			oldp = (char *)oldp + sizeof(struct in_addr);
2340 			written += sizeof(struct in_addr);
2341 			tmp = inm->inm_refcount;
2342 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2343 			if (error)
2344 				goto done;
2345 			oldp = (char *)oldp + sizeof(tmp);
2346 			written += sizeof(tmp);
2347 		}
2348 	}
2349 done:
2350 	if_put(ifp, &psref);
2351 	curlwp_bindx(bound);
2352 	*oldlenp = written;
2353 	return error;
2354 }
2355 
2356 static void
2357 in_sysctl_init(struct sysctllog **clog)
2358 {
2359 	sysctl_createv(clog, 0, NULL, NULL,
2360 		       CTLFLAG_PERMANENT,
2361 		       CTLTYPE_NODE, "inet",
2362 		       SYSCTL_DESCR("PF_INET related settings"),
2363 		       NULL, 0, NULL, 0,
2364 		       CTL_NET, PF_INET, CTL_EOL);
2365 	sysctl_createv(clog, 0, NULL, NULL,
2366 		       CTLFLAG_PERMANENT,
2367 		       CTLTYPE_NODE, "multicast",
2368 		       SYSCTL_DESCR("Multicast information"),
2369 		       in_multicast_sysctl, 0, NULL, 0,
2370 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2371 	sysctl_createv(clog, 0, NULL, NULL,
2372 		       CTLFLAG_PERMANENT,
2373 		       CTLTYPE_NODE, "ip",
2374 		       SYSCTL_DESCR("IPv4 related settings"),
2375 		       NULL, 0, NULL, 0,
2376 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2377 
2378 	sysctl_createv(clog, 0, NULL, NULL,
2379 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2380 		       CTLTYPE_INT, "subnetsarelocal",
2381 		       SYSCTL_DESCR("Whether logical subnets are considered "
2382 				    "local"),
2383 		       NULL, 0, &subnetsarelocal, 0,
2384 		       CTL_NET, PF_INET, IPPROTO_IP,
2385 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2386 	sysctl_createv(clog, 0, NULL, NULL,
2387 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2388 		       CTLTYPE_INT, "hostzerobroadcast",
2389 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2390 		       NULL, 0, &hostzeroisbroadcast, 0,
2391 		       CTL_NET, PF_INET, IPPROTO_IP,
2392 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2393 }
2394 
2395 #if NARP > 0
2396 
2397 static struct lltable *
2398 in_lltattach(struct ifnet *ifp)
2399 {
2400 	struct lltable *llt;
2401 
2402 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2403 	llt->llt_af = AF_INET;
2404 	llt->llt_ifp = ifp;
2405 
2406 	llt->llt_lookup = in_lltable_lookup;
2407 	llt->llt_create = in_lltable_create;
2408 	llt->llt_delete = in_lltable_delete;
2409 	llt->llt_dump_entry = in_lltable_dump_entry;
2410 	llt->llt_hash = in_lltable_hash;
2411 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2412 	llt->llt_free_entry = in_lltable_free_entry;
2413 	llt->llt_match_prefix = in_lltable_match_prefix;
2414 	lltable_link(llt);
2415 
2416 	return (llt);
2417 }
2418 
2419 #endif /* NARP > 0 */
2420 
2421 void *
2422 in_domifattach(struct ifnet *ifp)
2423 {
2424 	struct in_ifinfo *ii;
2425 
2426 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2427 
2428 #if NARP > 0
2429 	ii->ii_llt = in_lltattach(ifp);
2430 #endif
2431 
2432 #ifdef IPSELSRC
2433 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2434 	KASSERT(ii->ii_selsrc != NULL);
2435 #endif
2436 
2437 	return ii;
2438 }
2439 
2440 void
2441 in_domifdetach(struct ifnet *ifp, void *aux)
2442 {
2443 	struct in_ifinfo *ii = aux;
2444 
2445 #ifdef IPSELSRC
2446 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2447 #endif
2448 #if NARP > 0
2449 	lltable_free(ii->ii_llt);
2450 #endif
2451 	kmem_free(ii, sizeof(struct in_ifinfo));
2452 }
2453