xref: /netbsd-src/sys/netinet/in.c (revision 7863ba460b0a05b553c754e5dbc29247dddec322)
1 /*	$NetBSD: in.c,v 1.227 2018/04/06 16:03:16 ozaki-r Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1991, 1993
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)in.c	8.4 (Berkeley) 1/9/95
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: in.c,v 1.227 2018/04/06 16:03:16 ozaki-r Exp $");
95 
96 #include "arp.h"
97 
98 #ifdef _KERNEL_OPT
99 #include "opt_inet.h"
100 #include "opt_inet_conf.h"
101 #include "opt_mrouting.h"
102 #include "opt_net_mpsafe.h"
103 #endif
104 
105 #include <sys/param.h>
106 #include <sys/ioctl.h>
107 #include <sys/errno.h>
108 #include <sys/kernel.h>
109 #include <sys/malloc.h>
110 #include <sys/socket.h>
111 #include <sys/socketvar.h>
112 #include <sys/sysctl.h>
113 #include <sys/systm.h>
114 #include <sys/proc.h>
115 #include <sys/syslog.h>
116 #include <sys/kauth.h>
117 #include <sys/kmem.h>
118 
119 #include <sys/cprng.h>
120 
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <net/if_arp.h>
126 #include <net/if_ether.h>
127 #include <net/if_types.h>
128 #include <net/if_llatbl.h>
129 #include <net/if_dl.h>
130 
131 #include <netinet/in_systm.h>
132 #include <netinet/in.h>
133 #include <netinet/in_var.h>
134 #include <netinet/ip.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/in_ifattach.h>
137 #include <netinet/in_pcb.h>
138 #include <netinet/in_selsrc.h>
139 #include <netinet/if_inarp.h>
140 #include <netinet/ip_mroute.h>
141 #include <netinet/igmp_var.h>
142 
143 #ifdef IPSELSRC
144 #include <netinet/in_selsrc.h>
145 #endif
146 
147 static u_int	in_mask2len(struct in_addr *);
148 static void	in_len2mask(struct in_addr *, u_int);
149 static int	in_lifaddr_ioctl(struct socket *, u_long, void *,
150 	struct ifnet *);
151 
152 static void	in_addrhash_insert_locked(struct in_ifaddr *);
153 static void	in_addrhash_remove_locked(struct in_ifaddr *);
154 
155 static int	in_addprefix(struct in_ifaddr *, int);
156 static void	in_scrubaddr(struct in_ifaddr *);
157 static int	in_scrubprefix(struct in_ifaddr *);
158 static void	in_sysctl_init(struct sysctllog **);
159 
160 #ifndef SUBNETSARELOCAL
161 #define	SUBNETSARELOCAL	1
162 #endif
163 
164 #ifndef HOSTZEROBROADCAST
165 #define HOSTZEROBROADCAST 0
166 #endif
167 
168 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
169 #ifndef IN_MULTI_HASH_SIZE
170 #define IN_MULTI_HASH_SIZE	509
171 #endif
172 
173 static int			subnetsarelocal = SUBNETSARELOCAL;
174 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
175 
176 /*
177  * This list is used to keep track of in_multi chains which belong to
178  * deleted interface addresses.  We use in_ifaddr so that a chain head
179  * won't be deallocated until all multicast address record are deleted.
180  */
181 
182 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
183 
184 static struct pool		inmulti_pool;
185 static u_int			in_multientries;
186 static struct in_multihashhead *in_multihashtbl;
187 static u_long			in_multihash;
188 static krwlock_t		in_multilock;
189 
190 #define IN_MULTI_HASH(x, ifp) \
191     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
192 
193 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */
194 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
195 u_long				in_ifaddrhash;
196 struct in_ifaddrhead		in_ifaddrhead;
197 static kmutex_t			in_ifaddr_lock;
198 
199 pserialize_t			in_ifaddrhash_psz;
200 struct pslist_head *		in_ifaddrhashtbl_pslist;
201 u_long				in_ifaddrhash_pslist;
202 struct pslist_head		in_ifaddrhead_pslist;
203 
204 void
205 in_init(void)
206 {
207 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
208 	    NULL, IPL_SOFTNET);
209 	TAILQ_INIT(&in_ifaddrhead);
210 	PSLIST_INIT(&in_ifaddrhead_pslist);
211 
212 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
213 	    &in_ifaddrhash);
214 
215 	in_ifaddrhash_psz = pserialize_create();
216 	in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST,
217 	    true, &in_ifaddrhash_pslist);
218 	mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
219 
220 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
221 	    &in_multihash);
222 	rw_init(&in_multilock);
223 
224 	in_sysctl_init(NULL);
225 }
226 
227 /*
228  * Return 1 if an internet address is for a ``local'' host
229  * (one to which we have a connection).  If subnetsarelocal
230  * is true, this includes other subnets of the local net.
231  * Otherwise, it includes only the directly-connected (sub)nets.
232  */
233 int
234 in_localaddr(struct in_addr in)
235 {
236 	struct in_ifaddr *ia;
237 	int localaddr = 0;
238 	int s = pserialize_read_enter();
239 
240 	if (subnetsarelocal) {
241 		IN_ADDRLIST_READER_FOREACH(ia) {
242 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net) {
243 				localaddr = 1;
244 				break;
245 			}
246 		}
247 	} else {
248 		IN_ADDRLIST_READER_FOREACH(ia) {
249 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
250 				localaddr = 1;
251 				break;
252 			}
253 		}
254 	}
255 	pserialize_read_exit(s);
256 
257 	return localaddr;
258 }
259 
260 /*
261  * like in_localaddr() but can specify ifp.
262  */
263 int
264 in_direct(struct in_addr in, struct ifnet *ifp)
265 {
266 	struct ifaddr *ifa;
267 	int localaddr = 0;
268 	int s;
269 
270 	KASSERT(ifp != NULL);
271 
272 #define ia (ifatoia(ifa))
273 	s = pserialize_read_enter();
274 	if (subnetsarelocal) {
275 		IFADDR_READER_FOREACH(ifa, ifp) {
276 			if (ifa->ifa_addr->sa_family == AF_INET &&
277 			    ((in.s_addr & ia->ia_netmask) == ia->ia_net)) {
278 				localaddr = 1;
279 				break;
280 			}
281 		}
282 	} else {
283 		IFADDR_READER_FOREACH(ifa, ifp) {
284 			if (ifa->ifa_addr->sa_family == AF_INET &&
285 			    (in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
286 				localaddr = 1;
287 				break;
288 			}
289 		}
290 	}
291 	pserialize_read_exit(s);
292 
293 	return localaddr;
294 #undef ia
295 }
296 
297 /*
298  * Determine whether an IP address is in a reserved set of addresses
299  * that may not be forwarded, or whether datagrams to that destination
300  * may be forwarded.
301  */
302 int
303 in_canforward(struct in_addr in)
304 {
305 	u_int32_t net;
306 
307 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
308 		return (0);
309 	if (IN_CLASSA(in.s_addr)) {
310 		net = in.s_addr & IN_CLASSA_NET;
311 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
312 			return (0);
313 	}
314 	return (1);
315 }
316 
317 /*
318  * Trim a mask in a sockaddr
319  */
320 void
321 in_socktrim(struct sockaddr_in *ap)
322 {
323 	char *cplim = (char *) &ap->sin_addr;
324 	char *cp = (char *) (&ap->sin_addr + 1);
325 
326 	ap->sin_len = 0;
327 	while (--cp >= cplim)
328 		if (*cp) {
329 			(ap)->sin_len = cp - (char *) (ap) + 1;
330 			break;
331 		}
332 }
333 
334 /*
335  * Maintain the "in_maxmtu" variable, which is the largest
336  * mtu for non-local interfaces with AF_INET addresses assigned
337  * to them that are up.
338  */
339 unsigned long in_maxmtu;
340 
341 void
342 in_setmaxmtu(void)
343 {
344 	struct in_ifaddr *ia;
345 	struct ifnet *ifp;
346 	unsigned long maxmtu = 0;
347 	int s = pserialize_read_enter();
348 
349 	IN_ADDRLIST_READER_FOREACH(ia) {
350 		if ((ifp = ia->ia_ifp) == 0)
351 			continue;
352 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
353 			continue;
354 		if (ifp->if_mtu > maxmtu)
355 			maxmtu = ifp->if_mtu;
356 	}
357 	if (maxmtu)
358 		in_maxmtu = maxmtu;
359 	pserialize_read_exit(s);
360 }
361 
362 static u_int
363 in_mask2len(struct in_addr *mask)
364 {
365 	u_int x, y;
366 	u_char *p;
367 
368 	p = (u_char *)mask;
369 	for (x = 0; x < sizeof(*mask); x++) {
370 		if (p[x] != 0xff)
371 			break;
372 	}
373 	y = 0;
374 	if (x < sizeof(*mask)) {
375 		for (y = 0; y < NBBY; y++) {
376 			if ((p[x] & (0x80 >> y)) == 0)
377 				break;
378 		}
379 	}
380 	return x * NBBY + y;
381 }
382 
383 static void
384 in_len2mask(struct in_addr *mask, u_int len)
385 {
386 	u_int i;
387 	u_char *p;
388 
389 	p = (u_char *)mask;
390 	memset(mask, 0, sizeof(*mask));
391 	for (i = 0; i < len / NBBY; i++)
392 		p[i] = 0xff;
393 	if (len % NBBY)
394 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
395 }
396 
397 /*
398  * Generic internet control operations (ioctl's).
399  * Ifp is 0 if not an interface-specific ioctl.
400  */
401 /* ARGSUSED */
402 static int
403 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
404 {
405 	struct ifreq *ifr = (struct ifreq *)data;
406 	struct in_ifaddr *ia = NULL;
407 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
408 	struct sockaddr_in oldaddr, *new_dstaddr;
409 	int error, hostIsNew, maskIsNew;
410 	int newifaddr = 0;
411 	bool run_hook = false;
412 	bool need_reinsert = false;
413 	struct psref psref;
414 	int bound;
415 
416 	switch (cmd) {
417 	case SIOCALIFADDR:
418 	case SIOCDLIFADDR:
419 	case SIOCGLIFADDR:
420 		if (ifp == NULL)
421 			return EINVAL;
422 		return in_lifaddr_ioctl(so, cmd, data, ifp);
423 	case SIOCGIFADDRPREF:
424 	case SIOCSIFADDRPREF:
425 		if (ifp == NULL)
426 			return EINVAL;
427 		return ifaddrpref_ioctl(so, cmd, data, ifp);
428 	}
429 
430 	bound = curlwp_bind();
431 	/*
432 	 * Find address for this interface, if it exists.
433 	 */
434 	if (ifp != NULL)
435 		ia = in_get_ia_from_ifp_psref(ifp, &psref);
436 
437 	hostIsNew = 1;		/* moved here to appease gcc */
438 	switch (cmd) {
439 	case SIOCAIFADDR:
440 	case SIOCDIFADDR:
441 	case SIOCGIFALIAS:
442 	case SIOCGIFAFLAG_IN:
443 		if (ifra->ifra_addr.sin_family == AF_INET) {
444 			int s;
445 
446 			if (ia != NULL)
447 				ia4_release(ia, &psref);
448 			s = pserialize_read_enter();
449 			IN_ADDRHASH_READER_FOREACH(ia,
450 			    ifra->ifra_addr.sin_addr.s_addr) {
451 				if (ia->ia_ifp == ifp &&
452 				    in_hosteq(ia->ia_addr.sin_addr,
453 				    ifra->ifra_addr.sin_addr))
454 					break;
455 			}
456 			if (ia != NULL)
457 				ia4_acquire(ia, &psref);
458 			pserialize_read_exit(s);
459 		}
460 		if ((cmd == SIOCDIFADDR ||
461 		    cmd == SIOCGIFALIAS ||
462 		    cmd == SIOCGIFAFLAG_IN) &&
463 		    ia == NULL) {
464 			error = EADDRNOTAVAIL;
465 			goto out;
466 		}
467 
468 		if (cmd == SIOCDIFADDR &&
469 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
470 			ifra->ifra_addr.sin_family = AF_INET;
471 		}
472 		/* FALLTHROUGH */
473 	case SIOCSIFADDR:
474 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
475 			;
476 		else if (ifra->ifra_addr.sin_len == 0) {
477 			ifra->ifra_addr = ia->ia_addr;
478 			hostIsNew = 0;
479 		} else if (in_hosteq(ia->ia_addr.sin_addr,
480 		           ifra->ifra_addr.sin_addr))
481 			hostIsNew = 0;
482 		/* FALLTHROUGH */
483 	case SIOCSIFDSTADDR:
484 		if (ifra->ifra_addr.sin_family != AF_INET) {
485 			error = EAFNOSUPPORT;
486 			goto out;
487 		}
488 		/* FALLTHROUGH */
489 	case SIOCSIFNETMASK:
490 		if (ifp == NULL)
491 			panic("in_control");
492 
493 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
494 			break;
495 
496 		if (ia == NULL &&
497 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
498 			error = EADDRNOTAVAIL;
499 			goto out;
500 		}
501 
502 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
503 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
504 		    NULL) != 0) {
505 			error = EPERM;
506 			goto out;
507 		}
508 
509 		if (ia == NULL) {
510 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
511 			if (ia == NULL) {
512 				error = ENOBUFS;
513 				goto out;
514 			}
515 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
516 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
517 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
518 #ifdef IPSELSRC
519 			ia->ia_ifa.ifa_getifa = in_getifa;
520 #else /* IPSELSRC */
521 			ia->ia_ifa.ifa_getifa = NULL;
522 #endif /* IPSELSRC */
523 			ia->ia_sockmask.sin_len = 8;
524 			ia->ia_sockmask.sin_family = AF_INET;
525 			if (ifp->if_flags & IFF_BROADCAST) {
526 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
527 				ia->ia_broadaddr.sin_family = AF_INET;
528 			}
529 			ia->ia_ifp = ifp;
530 			ia->ia_idsalt = cprng_fast32() % 65535;
531 			LIST_INIT(&ia->ia_multiaddrs);
532 			IN_ADDRHASH_ENTRY_INIT(ia);
533 			IN_ADDRLIST_ENTRY_INIT(ia);
534 			ifa_psref_init(&ia->ia_ifa);
535 			/*
536 			 * We need a reference to make ia survive over in_ifinit
537 			 * that does ifaref and ifafree.
538 			 */
539 			ifaref(&ia->ia_ifa);
540 
541 			newifaddr = 1;
542 		}
543 		break;
544 
545 	case SIOCSIFBRDADDR:
546 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
547 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
548 		    NULL) != 0) {
549 			error = EPERM;
550 			goto out;
551 		}
552 		/* FALLTHROUGH */
553 
554 	case SIOCGIFADDR:
555 	case SIOCGIFNETMASK:
556 	case SIOCGIFDSTADDR:
557 	case SIOCGIFBRDADDR:
558 		if (ia == NULL) {
559 			error = EADDRNOTAVAIL;
560 			goto out;
561 		}
562 		break;
563 	}
564 	error = 0;
565 	switch (cmd) {
566 
567 	case SIOCGIFADDR:
568 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
569 		break;
570 
571 	case SIOCGIFBRDADDR:
572 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
573 			error = EINVAL;
574 			goto out;
575 		}
576 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
577 		break;
578 
579 	case SIOCGIFDSTADDR:
580 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
581 			error = EINVAL;
582 			goto out;
583 		}
584 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
585 		break;
586 
587 	case SIOCGIFNETMASK:
588 		/*
589 		 * We keep the number of trailing zero bytes the sin_len field
590 		 * of ia_sockmask, so we fix this before we pass it back to
591 		 * userland.
592 		 */
593 		oldaddr = ia->ia_sockmask;
594 		oldaddr.sin_len = sizeof(struct sockaddr_in);
595 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
596 		break;
597 
598 	case SIOCSIFDSTADDR:
599 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
600 			error = EINVAL;
601 			goto out;
602 		}
603 		oldaddr = ia->ia_dstaddr;
604 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
605 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
606 			ia->ia_dstaddr = oldaddr;
607 			goto out;
608 		}
609 		if (ia->ia_flags & IFA_ROUTE) {
610 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
611 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
612 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
613 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
614 		}
615 		break;
616 
617 	case SIOCSIFBRDADDR:
618 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
619 			error = EINVAL;
620 			goto out;
621 		}
622 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
623 		break;
624 
625 	case SIOCSIFADDR:
626 		if (!newifaddr) {
627 			in_addrhash_remove(ia);
628 			need_reinsert = true;
629 		}
630 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
631 		    NULL, 1);
632 
633 		run_hook = true;
634 		break;
635 
636 	case SIOCSIFNETMASK:
637 		in_scrubprefix(ia);
638 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
639 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
640 		if (!newifaddr) {
641 			in_addrhash_remove(ia);
642 			need_reinsert = true;
643 		}
644 		error = in_ifinit(ifp, ia, NULL, NULL, 0);
645 		break;
646 
647 	case SIOCAIFADDR:
648 		maskIsNew = 0;
649 		if (ifra->ifra_mask.sin_len) {
650 			in_scrubprefix(ia);
651 			ia->ia_sockmask = ifra->ifra_mask;
652 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
653 			maskIsNew = 1;
654 		}
655 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
656 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
657 			new_dstaddr = &ifra->ifra_dstaddr;
658 			maskIsNew  = 1; /* We lie; but the effect's the same */
659 		} else
660 			new_dstaddr = NULL;
661 		if (ifra->ifra_addr.sin_family == AF_INET &&
662 		    (hostIsNew || maskIsNew)) {
663 			if (!newifaddr) {
664 				in_addrhash_remove(ia);
665 				need_reinsert = true;
666 			}
667 			error = in_ifinit(ifp, ia, &ifra->ifra_addr,
668 			    new_dstaddr, 0);
669 		}
670 		if ((ifp->if_flags & IFF_BROADCAST) &&
671 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
672 			ia->ia_broadaddr = ifra->ifra_broadaddr;
673 		run_hook = true;
674 		break;
675 
676 	case SIOCGIFALIAS:
677 		ifra->ifra_mask = ia->ia_sockmask;
678 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
679 		    (ia->ia_dstaddr.sin_family == AF_INET))
680 			ifra->ifra_dstaddr = ia->ia_dstaddr;
681 		else if ((ifp->if_flags & IFF_BROADCAST) &&
682 		    (ia->ia_broadaddr.sin_family == AF_INET))
683 			ifra->ifra_broadaddr = ia->ia_broadaddr;
684 		else
685 			memset(&ifra->ifra_broadaddr, 0,
686 			      sizeof(ifra->ifra_broadaddr));
687 		break;
688 
689 	case SIOCGIFAFLAG_IN:
690 		ifr->ifr_addrflags = ia->ia4_flags;
691 		break;
692 
693 	case SIOCDIFADDR:
694 		ia4_release(ia, &psref);
695 		ifaref(&ia->ia_ifa);
696 		in_purgeaddr(&ia->ia_ifa);
697 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
698 		ifafree(&ia->ia_ifa);
699 		ia = NULL;
700 		break;
701 
702 #ifdef MROUTING
703 	case SIOCGETVIFCNT:
704 	case SIOCGETSGCNT:
705 		error = mrt_ioctl(so, cmd, data);
706 		break;
707 #endif /* MROUTING */
708 
709 	default:
710 		error = ENOTTY;
711 		goto out;
712 	}
713 
714 	/*
715 	 * XXX insert regardless of error to make in_purgeaddr below work.
716 	 * Need to improve.
717 	 */
718 	if (newifaddr) {
719 		ifaref(&ia->ia_ifa);
720 		ifa_insert(ifp, &ia->ia_ifa);
721 
722 		mutex_enter(&in_ifaddr_lock);
723 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
724 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
725 		in_addrhash_insert_locked(ia);
726 		/* Release a reference that is held just after creation. */
727 		ifafree(&ia->ia_ifa);
728 		mutex_exit(&in_ifaddr_lock);
729 	} else if (need_reinsert) {
730 		in_addrhash_insert(ia);
731 	}
732 
733 	if (error == 0) {
734 		if (run_hook)
735 			pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
736 	} else if (newifaddr) {
737 		KASSERT(ia != NULL);
738 		in_purgeaddr(&ia->ia_ifa);
739 		ia = NULL;
740 	}
741 
742 out:
743 	if (!newifaddr && ia != NULL)
744 		ia4_release(ia, &psref);
745 	curlwp_bindx(bound);
746 	return error;
747 }
748 
749 int
750 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
751 {
752 	int error;
753 
754 #ifndef NET_MPSAFE
755 	KASSERT(KERNEL_LOCKED_P());
756 #endif
757 	error = in_control0(so, cmd, data, ifp);
758 
759 	return error;
760 }
761 
762 /* Add ownaddr as loopback rtentry. */
763 static void
764 in_ifaddlocal(struct ifaddr *ifa)
765 {
766 	struct in_ifaddr *ia;
767 
768 	ia = (struct in_ifaddr *)ifa;
769 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
770 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
771 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
772 	{
773 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
774 		return;
775 	}
776 
777 	rt_ifa_addlocal(ifa);
778 }
779 
780 /* Remove loopback entry of ownaddr */
781 static void
782 in_ifremlocal(struct ifaddr *ifa)
783 {
784 	struct in_ifaddr *ia, *p;
785 	struct ifaddr *alt_ifa = NULL;
786 	int ia_count = 0;
787 	int s;
788 	struct psref psref;
789 	int bound = curlwp_bind();
790 
791 	ia = (struct in_ifaddr *)ifa;
792 	/* Delete the entry if exactly one ifaddr matches the
793 	 * address, ifa->ifa_addr. */
794 	s = pserialize_read_enter();
795 	IN_ADDRLIST_READER_FOREACH(p) {
796 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
797 			continue;
798 		if (p->ia_ifp != ia->ia_ifp)
799 			alt_ifa = &p->ia_ifa;
800 		if (++ia_count > 1 && alt_ifa != NULL)
801 			break;
802 	}
803 	if (alt_ifa != NULL && ia_count > 1)
804 		ifa_acquire(alt_ifa, &psref);
805 	pserialize_read_exit(s);
806 
807 	if (ia_count == 0)
808 		goto out;
809 
810 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
811 	if (alt_ifa != NULL && ia_count > 1)
812 		ifa_release(alt_ifa, &psref);
813 out:
814 	curlwp_bindx(bound);
815 }
816 
817 static void
818 in_scrubaddr(struct in_ifaddr *ia)
819 {
820 
821 	/* stop DAD processing */
822 	if (ia->ia_dad_stop != NULL)
823 		ia->ia_dad_stop(&ia->ia_ifa);
824 
825 	in_scrubprefix(ia);
826 	in_ifremlocal(&ia->ia_ifa);
827 
828 	mutex_enter(&in_ifaddr_lock);
829 	if (ia->ia_allhosts != NULL) {
830 		in_delmulti(ia->ia_allhosts);
831 		ia->ia_allhosts = NULL;
832 	}
833 	mutex_exit(&in_ifaddr_lock);
834 }
835 
836 /*
837  * Depends on it isn't called in concurrent. It should be guaranteed
838  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
839  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
840  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
841  * never be executed in concurrent.
842  */
843 void
844 in_purgeaddr(struct ifaddr *ifa)
845 {
846 	struct in_ifaddr *ia = (void *) ifa;
847 	struct ifnet *ifp = ifa->ifa_ifp;
848 
849 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
850 
851 	ifa->ifa_flags |= IFA_DESTROYING;
852 	in_scrubaddr(ia);
853 
854 	mutex_enter(&in_ifaddr_lock);
855 	in_addrhash_remove_locked(ia);
856 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
857 	IN_ADDRLIST_WRITER_REMOVE(ia);
858 	ifa_remove(ifp, &ia->ia_ifa);
859 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
860 	mutex_exit(&in_ifaddr_lock);
861 	IN_ADDRHASH_ENTRY_DESTROY(ia);
862 	IN_ADDRLIST_ENTRY_DESTROY(ia);
863 	ifafree(&ia->ia_ifa);
864 	in_setmaxmtu();
865 }
866 
867 static void
868 in_addrhash_insert_locked(struct in_ifaddr *ia)
869 {
870 
871 	KASSERT(mutex_owned(&in_ifaddr_lock));
872 
873 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia,
874 	    ia_hash);
875 	IN_ADDRHASH_ENTRY_INIT(ia);
876 	IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
877 }
878 
879 void
880 in_addrhash_insert(struct in_ifaddr *ia)
881 {
882 
883 	mutex_enter(&in_ifaddr_lock);
884 	in_addrhash_insert_locked(ia);
885 	mutex_exit(&in_ifaddr_lock);
886 }
887 
888 static void
889 in_addrhash_remove_locked(struct in_ifaddr *ia)
890 {
891 
892 	KASSERT(mutex_owned(&in_ifaddr_lock));
893 
894 	LIST_REMOVE(ia, ia_hash);
895 	IN_ADDRHASH_WRITER_REMOVE(ia);
896 }
897 
898 void
899 in_addrhash_remove(struct in_ifaddr *ia)
900 {
901 
902 	mutex_enter(&in_ifaddr_lock);
903 	in_addrhash_remove_locked(ia);
904 #ifdef NET_MPSAFE
905 	pserialize_perform(in_ifaddrhash_psz);
906 #endif
907 	mutex_exit(&in_ifaddr_lock);
908 	IN_ADDRHASH_ENTRY_DESTROY(ia);
909 }
910 
911 void
912 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
913 {
914 
915 	IFNET_LOCK(ifp);
916 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
917 	igmp_purgeif(ifp);		/* manipulates pools */
918 #ifdef MROUTING
919 	ip_mrouter_detach(ifp);
920 #endif
921 	IFNET_UNLOCK(ifp);
922 }
923 
924 /*
925  * SIOC[GAD]LIFADDR.
926  *	SIOCGLIFADDR: get first address. (???)
927  *	SIOCGLIFADDR with IFLR_PREFIX:
928  *		get first address that matches the specified prefix.
929  *	SIOCALIFADDR: add the specified address.
930  *	SIOCALIFADDR with IFLR_PREFIX:
931  *		EINVAL since we can't deduce hostid part of the address.
932  *	SIOCDLIFADDR: delete the specified address.
933  *	SIOCDLIFADDR with IFLR_PREFIX:
934  *		delete the first address that matches the specified prefix.
935  * return values:
936  *	EINVAL on invalid parameters
937  *	EADDRNOTAVAIL on prefix match failed/specified address not found
938  *	other values may be returned from in_ioctl()
939  */
940 static int
941 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
942     struct ifnet *ifp)
943 {
944 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
945 	struct ifaddr *ifa;
946 	struct sockaddr *sa;
947 
948 	/* sanity checks */
949 	if (data == NULL || ifp == NULL) {
950 		panic("invalid argument to in_lifaddr_ioctl");
951 		/*NOTRECHED*/
952 	}
953 
954 	switch (cmd) {
955 	case SIOCGLIFADDR:
956 		/* address must be specified on GET with IFLR_PREFIX */
957 		if ((iflr->flags & IFLR_PREFIX) == 0)
958 			break;
959 		/*FALLTHROUGH*/
960 	case SIOCALIFADDR:
961 	case SIOCDLIFADDR:
962 		/* address must be specified on ADD and DELETE */
963 		sa = (struct sockaddr *)&iflr->addr;
964 		if (sa->sa_family != AF_INET)
965 			return EINVAL;
966 		if (sa->sa_len != sizeof(struct sockaddr_in))
967 			return EINVAL;
968 		/* XXX need improvement */
969 		sa = (struct sockaddr *)&iflr->dstaddr;
970 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
971 			return EINVAL;
972 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
973 			return EINVAL;
974 		break;
975 	default: /*shouldn't happen*/
976 #if 0
977 		panic("invalid cmd to in_lifaddr_ioctl");
978 		/*NOTREACHED*/
979 #else
980 		return EOPNOTSUPP;
981 #endif
982 	}
983 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
984 		return EINVAL;
985 
986 	switch (cmd) {
987 	case SIOCALIFADDR:
988 	    {
989 		struct in_aliasreq ifra;
990 
991 		if (iflr->flags & IFLR_PREFIX)
992 			return EINVAL;
993 
994 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
995 		memset(&ifra, 0, sizeof(ifra));
996 		memcpy(ifra.ifra_name, iflr->iflr_name,
997 			sizeof(ifra.ifra_name));
998 
999 		memcpy(&ifra.ifra_addr, &iflr->addr,
1000 			((struct sockaddr *)&iflr->addr)->sa_len);
1001 
1002 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1003 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1004 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1005 		}
1006 
1007 		ifra.ifra_mask.sin_family = AF_INET;
1008 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
1009 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
1010 
1011 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
1012 	    }
1013 	case SIOCGLIFADDR:
1014 	case SIOCDLIFADDR:
1015 	    {
1016 		struct in_ifaddr *ia;
1017 		struct in_addr mask, candidate, match;
1018 		struct sockaddr_in *sin;
1019 		int cmp, s;
1020 
1021 		memset(&mask, 0, sizeof(mask));
1022 		memset(&match, 0, sizeof(match));	/* XXX gcc */
1023 		if (iflr->flags & IFLR_PREFIX) {
1024 			/* lookup a prefix rather than address. */
1025 			in_len2mask(&mask, iflr->prefixlen);
1026 
1027 			sin = (struct sockaddr_in *)&iflr->addr;
1028 			match.s_addr = sin->sin_addr.s_addr;
1029 			match.s_addr &= mask.s_addr;
1030 
1031 			/* if you set extra bits, that's wrong */
1032 			if (match.s_addr != sin->sin_addr.s_addr)
1033 				return EINVAL;
1034 
1035 			cmp = 1;
1036 		} else {
1037 			if (cmd == SIOCGLIFADDR) {
1038 				/* on getting an address, take the 1st match */
1039 				cmp = 0;	/*XXX*/
1040 			} else {
1041 				/* on deleting an address, do exact match */
1042 				in_len2mask(&mask, 32);
1043 				sin = (struct sockaddr_in *)&iflr->addr;
1044 				match.s_addr = sin->sin_addr.s_addr;
1045 
1046 				cmp = 1;
1047 			}
1048 		}
1049 
1050 		s = pserialize_read_enter();
1051 		IFADDR_READER_FOREACH(ifa, ifp) {
1052 			if (ifa->ifa_addr->sa_family != AF_INET)
1053 				continue;
1054 			if (cmp == 0)
1055 				break;
1056 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
1057 			candidate.s_addr &= mask.s_addr;
1058 			if (candidate.s_addr == match.s_addr)
1059 				break;
1060 		}
1061 		if (ifa == NULL) {
1062 			pserialize_read_exit(s);
1063 			return EADDRNOTAVAIL;
1064 		}
1065 		ia = (struct in_ifaddr *)ifa;
1066 
1067 		if (cmd == SIOCGLIFADDR) {
1068 			/* fill in the if_laddrreq structure */
1069 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1070 
1071 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1072 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1073 					ia->ia_dstaddr.sin_len);
1074 			} else
1075 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1076 
1077 			iflr->prefixlen =
1078 				in_mask2len(&ia->ia_sockmask.sin_addr);
1079 
1080 			iflr->flags = 0;	/*XXX*/
1081 			pserialize_read_exit(s);
1082 
1083 			return 0;
1084 		} else {
1085 			struct in_aliasreq ifra;
1086 
1087 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1088 			memset(&ifra, 0, sizeof(ifra));
1089 			memcpy(ifra.ifra_name, iflr->iflr_name,
1090 				sizeof(ifra.ifra_name));
1091 
1092 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1093 				ia->ia_addr.sin_len);
1094 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1095 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1096 					ia->ia_dstaddr.sin_len);
1097 			}
1098 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1099 				ia->ia_sockmask.sin_len);
1100 			pserialize_read_exit(s);
1101 
1102 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1103 		}
1104 	    }
1105 	}
1106 
1107 	return EOPNOTSUPP;	/*just for safety*/
1108 }
1109 
1110 /*
1111  * Initialize an interface's internet address
1112  * and routing table entry.
1113  */
1114 int
1115 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1116     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1117 {
1118 	u_int32_t i;
1119 	struct sockaddr_in oldaddr, olddst;
1120 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1121 
1122 	if (sin == NULL)
1123 		sin = &ia->ia_addr;
1124 	if (dst == NULL)
1125 		dst = &ia->ia_dstaddr;
1126 
1127 	/*
1128 	 * Set up new addresses.
1129 	 */
1130 	oldaddr = ia->ia_addr;
1131 	olddst = ia->ia_dstaddr;
1132 	oldflags = ia->ia4_flags;
1133 	ia->ia_addr = *sin;
1134 	ia->ia_dstaddr = *dst;
1135 	hostIsNew = oldaddr.sin_family != AF_INET ||
1136 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1137 	if (!scrub)
1138 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1139 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1140 
1141 	/*
1142 	 * Configure address flags.
1143 	 * We need to do this early because they maybe adjusted
1144 	 * by if_addr_init depending on the address.
1145 	 */
1146 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1147 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1148 		hostIsNew = 1;
1149 	}
1150 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1151 		ia->ia4_flags |= IN_IFF_DETACHED;
1152 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1153 	} else if (hostIsNew && if_do_dad(ifp) && ip_dad_count > 0)
1154 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1155 
1156 	/*
1157 	 * Give the interface a chance to initialize
1158 	 * if this is its first address,
1159 	 * and to validate the address if necessary.
1160 	 */
1161 	s = splsoftnet();
1162 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1163 	splx(s);
1164 	/* Now clear the try tentative flag, its job is done. */
1165 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1166 	if (error != 0) {
1167 		ia->ia_addr = oldaddr;
1168 		ia->ia_dstaddr = olddst;
1169 		ia->ia4_flags = oldflags;
1170 		return error;
1171 	}
1172 
1173 	if (scrub || hostIsNew) {
1174 		int newflags = ia->ia4_flags;
1175 
1176 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1177 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1178 		ia->ia4_flags = oldflags;
1179 		if (hostIsNew)
1180 			in_scrubaddr(ia);
1181 		else if (scrub)
1182 			in_scrubprefix(ia);
1183 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1184 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1185 		ia->ia4_flags = newflags;
1186 	}
1187 
1188 	i = ia->ia_addr.sin_addr.s_addr;
1189 	if (ifp->if_flags & IFF_POINTOPOINT)
1190 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1191 	else if (IN_CLASSA(i))
1192 		ia->ia_netmask = IN_CLASSA_NET;
1193 	else if (IN_CLASSB(i))
1194 		ia->ia_netmask = IN_CLASSB_NET;
1195 	else
1196 		ia->ia_netmask = IN_CLASSC_NET;
1197 	/*
1198 	 * The subnet mask usually includes at least the standard network part,
1199 	 * but may may be smaller in the case of supernetting.
1200 	 * If it is set, we believe it.
1201 	 */
1202 	if (ia->ia_subnetmask == 0) {
1203 		ia->ia_subnetmask = ia->ia_netmask;
1204 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1205 	} else
1206 		ia->ia_netmask &= ia->ia_subnetmask;
1207 
1208 	ia->ia_net = i & ia->ia_netmask;
1209 	ia->ia_subnet = i & ia->ia_subnetmask;
1210 	in_socktrim(&ia->ia_sockmask);
1211 
1212 	/* re-calculate the "in_maxmtu" value */
1213 	in_setmaxmtu();
1214 
1215 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1216 	if (ifp->if_flags & IFF_BROADCAST) {
1217 		ia->ia_broadaddr.sin_addr.s_addr =
1218 			ia->ia_subnet | ~ia->ia_subnetmask;
1219 		ia->ia_netbroadcast.s_addr =
1220 			ia->ia_net | ~ia->ia_netmask;
1221 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1222 		ia->ia_dstaddr = ia->ia_addr;
1223 		flags |= RTF_HOST;
1224 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1225 		if (ia->ia_dstaddr.sin_family != AF_INET)
1226 			return (0);
1227 		flags |= RTF_HOST;
1228 	}
1229 
1230 	/* Add the local route to the address */
1231 	in_ifaddlocal(&ia->ia_ifa);
1232 
1233 	/* Add the prefix route for the address */
1234 	error = in_addprefix(ia, flags);
1235 
1236 	/*
1237 	 * If the interface supports multicast, join the "all hosts"
1238 	 * multicast group on that interface.
1239 	 */
1240 	mutex_enter(&in_ifaddr_lock);
1241 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1242 		struct in_addr addr;
1243 
1244 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1245 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1246 	}
1247 	mutex_exit(&in_ifaddr_lock);
1248 
1249 	if (hostIsNew &&
1250 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1251 	    if_do_dad(ifp))
1252 		ia->ia_dad_start((struct ifaddr *)ia);
1253 
1254 	return error;
1255 }
1256 
1257 #define rtinitflags(x) \
1258 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1259 	    ? RTF_HOST : 0)
1260 
1261 /*
1262  * add a route to prefix ("connected route" in cisco terminology).
1263  * does nothing if there's some interface address with the same prefix already.
1264  */
1265 static int
1266 in_addprefix(struct in_ifaddr *target, int flags)
1267 {
1268 	struct in_ifaddr *ia;
1269 	struct in_addr prefix, mask, p;
1270 	int error;
1271 	int s;
1272 
1273 	if ((flags & RTF_HOST) != 0)
1274 		prefix = target->ia_dstaddr.sin_addr;
1275 	else {
1276 		prefix = target->ia_addr.sin_addr;
1277 		mask = target->ia_sockmask.sin_addr;
1278 		prefix.s_addr &= mask.s_addr;
1279 	}
1280 
1281 	s = pserialize_read_enter();
1282 	IN_ADDRLIST_READER_FOREACH(ia) {
1283 		if (rtinitflags(ia))
1284 			p = ia->ia_dstaddr.sin_addr;
1285 		else {
1286 			p = ia->ia_addr.sin_addr;
1287 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1288 		}
1289 
1290 		if (prefix.s_addr != p.s_addr)
1291 			continue;
1292 
1293 		/*
1294 		 * if we got a matching prefix route inserted by other
1295 		 * interface address, we don't need to bother
1296 		 *
1297 		 * XXX RADIX_MPATH implications here? -dyoung
1298 		 */
1299 		if (ia->ia_flags & IFA_ROUTE) {
1300 			pserialize_read_exit(s);
1301 			return 0;
1302 		}
1303 	}
1304 	pserialize_read_exit(s);
1305 
1306 	/*
1307 	 * noone seem to have prefix route.  insert it.
1308 	 */
1309 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1310 	if (error == 0)
1311 		target->ia_flags |= IFA_ROUTE;
1312 	else if (error == EEXIST) {
1313 		/*
1314 		 * the fact the route already exists is not an error.
1315 		 */
1316 		error = 0;
1317 	}
1318 	return error;
1319 }
1320 
1321 /*
1322  * remove a route to prefix ("connected route" in cisco terminology).
1323  * re-installs the route by using another interface address, if there's one
1324  * with the same prefix (otherwise we lose the route mistakenly).
1325  */
1326 static int
1327 in_scrubprefix(struct in_ifaddr *target)
1328 {
1329 	struct in_ifaddr *ia;
1330 	struct in_addr prefix, mask, p;
1331 	int error;
1332 	int s;
1333 
1334 	/* If we don't have IFA_ROUTE we have nothing to do */
1335 	if ((target->ia_flags & IFA_ROUTE) == 0)
1336 		return 0;
1337 
1338 	if (rtinitflags(target))
1339 		prefix = target->ia_dstaddr.sin_addr;
1340 	else {
1341 		prefix = target->ia_addr.sin_addr;
1342 		mask = target->ia_sockmask.sin_addr;
1343 		prefix.s_addr &= mask.s_addr;
1344 	}
1345 
1346 	s = pserialize_read_enter();
1347 	IN_ADDRLIST_READER_FOREACH(ia) {
1348 		if (rtinitflags(ia))
1349 			p = ia->ia_dstaddr.sin_addr;
1350 		else {
1351 			p = ia->ia_addr.sin_addr;
1352 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1353 		}
1354 
1355 		if (prefix.s_addr != p.s_addr)
1356 			continue;
1357 
1358 		/*
1359 		 * if we got a matching prefix route, move IFA_ROUTE to him
1360 		 */
1361 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1362 			struct psref psref;
1363 			int bound = curlwp_bind();
1364 
1365 			ia4_acquire(ia, &psref);
1366 			pserialize_read_exit(s);
1367 
1368 			rtinit(&target->ia_ifa, RTM_DELETE,
1369 			    rtinitflags(target));
1370 			target->ia_flags &= ~IFA_ROUTE;
1371 
1372 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1373 			    rtinitflags(ia) | RTF_UP);
1374 			if (error == 0)
1375 				ia->ia_flags |= IFA_ROUTE;
1376 
1377 			ia4_release(ia, &psref);
1378 			curlwp_bindx(bound);
1379 
1380 			return error;
1381 		}
1382 	}
1383 	pserialize_read_exit(s);
1384 
1385 	/*
1386 	 * noone seem to have prefix route.  remove it.
1387 	 */
1388 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1389 	target->ia_flags &= ~IFA_ROUTE;
1390 	return 0;
1391 }
1392 
1393 #undef rtinitflags
1394 
1395 /*
1396  * Return 1 if the address might be a local broadcast address.
1397  */
1398 int
1399 in_broadcast(struct in_addr in, struct ifnet *ifp)
1400 {
1401 	struct ifaddr *ifa;
1402 	int s;
1403 
1404 	KASSERT(ifp != NULL);
1405 
1406 	if (in.s_addr == INADDR_BROADCAST ||
1407 	    in_nullhost(in))
1408 		return 1;
1409 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1410 		return 0;
1411 	/*
1412 	 * Look through the list of addresses for a match
1413 	 * with a broadcast address.
1414 	 */
1415 #define ia (ifatoia(ifa))
1416 	s = pserialize_read_enter();
1417 	IFADDR_READER_FOREACH(ifa, ifp) {
1418 		if (ifa->ifa_addr->sa_family == AF_INET &&
1419 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1420 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1421 		     in_hosteq(in, ia->ia_netbroadcast) ||
1422 		     (hostzeroisbroadcast &&
1423 		      /*
1424 		       * Check for old-style (host 0) broadcast.
1425 		       */
1426 		      (in.s_addr == ia->ia_subnet ||
1427 		       in.s_addr == ia->ia_net)))) {
1428 			pserialize_read_exit(s);
1429 			return 1;
1430 		}
1431 	}
1432 	pserialize_read_exit(s);
1433 	return (0);
1434 #undef ia
1435 }
1436 
1437 /*
1438  * perform DAD when interface becomes IFF_UP.
1439  */
1440 void
1441 in_if_link_up(struct ifnet *ifp)
1442 {
1443 	struct ifaddr *ifa;
1444 	struct in_ifaddr *ia;
1445 	int s, bound;
1446 
1447 	/* Ensure it's sane to run DAD */
1448 	if (ifp->if_link_state == LINK_STATE_DOWN)
1449 		return;
1450 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1451 		return;
1452 
1453 	bound = curlwp_bind();
1454 	s = pserialize_read_enter();
1455 	IFADDR_READER_FOREACH(ifa, ifp) {
1456 		struct psref psref;
1457 
1458 		if (ifa->ifa_addr->sa_family != AF_INET)
1459 			continue;
1460 		ifa_acquire(ifa, &psref);
1461 		pserialize_read_exit(s);
1462 
1463 		ia = (struct in_ifaddr *)ifa;
1464 
1465 		/* If detached then mark as tentative */
1466 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1467 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1468 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1469 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1470 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1471 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1472 		}
1473 
1474 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1475 			/* Clear the duplicated flag as we're starting DAD. */
1476 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1477 			ia->ia_dad_start(ifa);
1478 		}
1479 
1480 		s = pserialize_read_enter();
1481 		ifa_release(ifa, &psref);
1482 	}
1483 	pserialize_read_exit(s);
1484 	curlwp_bindx(bound);
1485 }
1486 
1487 void
1488 in_if_up(struct ifnet *ifp)
1489 {
1490 
1491 	/* interface may not support link state, so bring it up also */
1492 	in_if_link_up(ifp);
1493 }
1494 
1495 /*
1496  * Mark all addresses as detached.
1497  */
1498 void
1499 in_if_link_down(struct ifnet *ifp)
1500 {
1501 	struct ifaddr *ifa;
1502 	struct in_ifaddr *ia;
1503 	int s, bound;
1504 
1505 	bound = curlwp_bind();
1506 	s = pserialize_read_enter();
1507 	IFADDR_READER_FOREACH(ifa, ifp) {
1508 		struct psref psref;
1509 
1510 		if (ifa->ifa_addr->sa_family != AF_INET)
1511 			continue;
1512 		ifa_acquire(ifa, &psref);
1513 		pserialize_read_exit(s);
1514 
1515 		ia = (struct in_ifaddr *)ifa;
1516 
1517 		/* Stop DAD processing */
1518 		if (ia->ia_dad_stop != NULL)
1519 			ia->ia_dad_stop(ifa);
1520 
1521 		/*
1522 		 * Mark the address as detached.
1523 		 */
1524 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1525 			ia->ia4_flags |= IN_IFF_DETACHED;
1526 			ia->ia4_flags &=
1527 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1528 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1529 		}
1530 
1531 		s = pserialize_read_enter();
1532 		ifa_release(ifa, &psref);
1533 	}
1534 	pserialize_read_exit(s);
1535 	curlwp_bindx(bound);
1536 }
1537 
1538 void
1539 in_if_down(struct ifnet *ifp)
1540 {
1541 
1542 	in_if_link_down(ifp);
1543 #if NARP > 0
1544 	lltable_purge_entries(LLTABLE(ifp));
1545 #endif
1546 }
1547 
1548 void
1549 in_if_link_state_change(struct ifnet *ifp, int link_state)
1550 {
1551 
1552 	switch (link_state) {
1553 	case LINK_STATE_DOWN:
1554 		in_if_link_down(ifp);
1555 		break;
1556 	case LINK_STATE_UP:
1557 		in_if_link_up(ifp);
1558 		break;
1559 	}
1560 }
1561 
1562 /*
1563  * in_lookup_multi: look up the in_multi record for a given IP
1564  * multicast address on a given interface.  If no matching record is
1565  * found, return NULL.
1566  */
1567 struct in_multi *
1568 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1569 {
1570 	struct in_multi *inm;
1571 
1572 	KASSERT(rw_lock_held(&in_multilock));
1573 
1574 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1575 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1576 			break;
1577 	}
1578 	return inm;
1579 }
1580 
1581 /*
1582  * in_multi_group: check whether the address belongs to an IP multicast
1583  * group we are joined on this interface.  Returns true or false.
1584  */
1585 bool
1586 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1587 {
1588 	bool ingroup;
1589 
1590 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1591 		rw_enter(&in_multilock, RW_READER);
1592 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1593 		rw_exit(&in_multilock);
1594 	} else {
1595 		/* XXX Recursive call from ip_output(). */
1596 		KASSERT(rw_lock_held(&in_multilock));
1597 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1598 	}
1599 	return ingroup;
1600 }
1601 
1602 /*
1603  * Add an address to the list of IP multicast addresses for a given interface.
1604  */
1605 struct in_multi *
1606 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1607 {
1608 	struct sockaddr_in sin;
1609 	struct in_multi *inm;
1610 
1611 	/*
1612 	 * See if address already in list.
1613 	 */
1614 	rw_enter(&in_multilock, RW_WRITER);
1615 	inm = in_lookup_multi(*ap, ifp);
1616 	if (inm != NULL) {
1617 		/*
1618 		 * Found it; just increment the reference count.
1619 		 */
1620 		inm->inm_refcount++;
1621 		rw_exit(&in_multilock);
1622 		return inm;
1623 	}
1624 
1625 	/*
1626 	 * New address; allocate a new multicast record.
1627 	 */
1628 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1629 	if (inm == NULL) {
1630 		rw_exit(&in_multilock);
1631 		return NULL;
1632 	}
1633 	inm->inm_addr = *ap;
1634 	inm->inm_ifp = ifp;
1635 	inm->inm_refcount = 1;
1636 
1637 	/*
1638 	 * Ask the network driver to update its multicast reception
1639 	 * filter appropriately for the new address.
1640 	 */
1641 	sockaddr_in_init(&sin, ap, 0);
1642 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1643 		rw_exit(&in_multilock);
1644 		pool_put(&inmulti_pool, inm);
1645 		return NULL;
1646 	}
1647 
1648 	/*
1649 	 * Let IGMP know that we have joined a new IP multicast group.
1650 	 */
1651 	if (igmp_joingroup(inm) != 0) {
1652 		rw_exit(&in_multilock);
1653 		pool_put(&inmulti_pool, inm);
1654 		return NULL;
1655 	}
1656 	LIST_INSERT_HEAD(
1657 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1658 	    inm, inm_list);
1659 	in_multientries++;
1660 	rw_exit(&in_multilock);
1661 
1662 	return inm;
1663 }
1664 
1665 /*
1666  * Delete a multicast address record.
1667  */
1668 void
1669 in_delmulti(struct in_multi *inm)
1670 {
1671 	struct sockaddr_in sin;
1672 
1673 	rw_enter(&in_multilock, RW_WRITER);
1674 	if (--inm->inm_refcount > 0) {
1675 		rw_exit(&in_multilock);
1676 		return;
1677 	}
1678 
1679 	/*
1680 	 * No remaining claims to this record; let IGMP know that
1681 	 * we are leaving the multicast group.
1682 	 */
1683 	igmp_leavegroup(inm);
1684 
1685 	/*
1686 	 * Notify the network driver to update its multicast reception
1687 	 * filter.
1688 	 */
1689 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1690 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1691 
1692 	/*
1693 	 * Unlink from list.
1694 	 */
1695 	LIST_REMOVE(inm, inm_list);
1696 	in_multientries--;
1697 	rw_exit(&in_multilock);
1698 
1699 	pool_put(&inmulti_pool, inm);
1700 }
1701 
1702 /*
1703  * in_next_multi: step through all of the in_multi records, one at a time.
1704  * The current position is remembered in "step", which the caller must
1705  * provide.  in_first_multi(), below, must be called to initialize "step"
1706  * and get the first record.  Both macros return a NULL "inm" when there
1707  * are no remaining records.
1708  */
1709 struct in_multi *
1710 in_next_multi(struct in_multistep *step)
1711 {
1712 	struct in_multi *inm;
1713 
1714 	KASSERT(rw_lock_held(&in_multilock));
1715 
1716 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1717 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1718 	}
1719 	if ((inm = step->i_inm) != NULL) {
1720 		step->i_inm = LIST_NEXT(inm, inm_list);
1721 	}
1722 	return inm;
1723 }
1724 
1725 struct in_multi *
1726 in_first_multi(struct in_multistep *step)
1727 {
1728 	KASSERT(rw_lock_held(&in_multilock));
1729 
1730 	step->i_n = 0;
1731 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1732 	return in_next_multi(step);
1733 }
1734 
1735 void
1736 in_multi_lock(int op)
1737 {
1738 	rw_enter(&in_multilock, op);
1739 }
1740 
1741 void
1742 in_multi_unlock(void)
1743 {
1744 	rw_exit(&in_multilock);
1745 }
1746 
1747 int
1748 in_multi_lock_held(void)
1749 {
1750 	return rw_lock_held(&in_multilock);
1751 }
1752 
1753 struct in_ifaddr *
1754 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1755     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1756 {
1757 	struct rtentry *rt = NULL;
1758 	struct in_ifaddr *ia = NULL;
1759 
1760 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1761 	/*
1762          * If route is known or can be allocated now, take the
1763          * source address from the interface.  Otherwise, punt.
1764 	 */
1765 	if ((soopts & SO_DONTROUTE) != 0)
1766 		rtcache_free(ro);
1767 	else {
1768 		union {
1769 			struct sockaddr		dst;
1770 			struct sockaddr_in	dst4;
1771 		} u;
1772 
1773 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1774 		rt = rtcache_lookup(ro, &u.dst);
1775 	}
1776 	/*
1777 	 * If we found a route, use the address
1778 	 * corresponding to the outgoing interface
1779 	 * unless it is the loopback (in case a route
1780 	 * to our address on another net goes to loopback).
1781 	 *
1782 	 * XXX Is this still true?  Do we care?
1783 	 */
1784 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1785 		int s;
1786 		struct ifaddr *ifa;
1787 		/*
1788 		 * Just in case. May not need to do this workaround.
1789 		 * Revisit when working on rtentry MP-ification.
1790 		 */
1791 		s = pserialize_read_enter();
1792 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1793 			if (ifa == rt->rt_ifa)
1794 				break;
1795 		}
1796 		if (ifa != NULL)
1797 			ifa_acquire(ifa, psref);
1798 		pserialize_read_exit(s);
1799 
1800 		ia = ifatoia(ifa);
1801 	}
1802 	if (ia == NULL) {
1803 		u_int16_t fport = sin->sin_port;
1804 		struct ifaddr *ifa;
1805 		int s;
1806 
1807 		sin->sin_port = 0;
1808 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1809 		sin->sin_port = fport;
1810 		if (ifa == NULL) {
1811 			/* Find 1st non-loopback AF_INET address */
1812 			s = pserialize_read_enter();
1813 			IN_ADDRLIST_READER_FOREACH(ia) {
1814 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1815 					break;
1816 			}
1817 			if (ia != NULL)
1818 				ia4_acquire(ia, psref);
1819 			pserialize_read_exit(s);
1820 		} else {
1821 			/* ia is already referenced by psref */
1822 			ia = ifatoia(ifa);
1823 		}
1824 		if (ia == NULL) {
1825 			*errorp = EADDRNOTAVAIL;
1826 			goto out;
1827 		}
1828 	}
1829 	/*
1830 	 * If the destination address is multicast and an outgoing
1831 	 * interface has been set as a multicast option, use the
1832 	 * address of that interface as our source address.
1833 	 */
1834 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1835 		struct ip_moptions *imo;
1836 
1837 		imo = mopts;
1838 		if (imo->imo_multicast_if_index != 0) {
1839 			struct ifnet *ifp;
1840 			int s;
1841 
1842 			if (ia != NULL)
1843 				ia4_release(ia, psref);
1844 			s = pserialize_read_enter();
1845 			ifp = if_byindex(imo->imo_multicast_if_index);
1846 			if (ifp != NULL) {
1847 				/* XXX */
1848 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1849 			} else
1850 				ia = NULL;
1851 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1852 				pserialize_read_exit(s);
1853 				if (ia != NULL)
1854 					ia4_release(ia, psref);
1855 				*errorp = EADDRNOTAVAIL;
1856 				ia = NULL;
1857 				goto out;
1858 			}
1859 			pserialize_read_exit(s);
1860 		}
1861 	}
1862 	if (ia->ia_ifa.ifa_getifa != NULL) {
1863 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1864 		                                      sintosa(sin)));
1865 		if (ia == NULL) {
1866 			*errorp = EADDRNOTAVAIL;
1867 			goto out;
1868 		}
1869 		/* FIXME NOMPSAFE */
1870 		ia4_acquire(ia, psref);
1871 	}
1872 #ifdef GETIFA_DEBUG
1873 	else
1874 		printf("%s: missing ifa_getifa\n", __func__);
1875 #endif
1876 out:
1877 	rtcache_unref(rt, ro);
1878 	return ia;
1879 }
1880 
1881 int
1882 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst)
1883 {
1884 	struct in_ifaddr *ia4;
1885 	int s;
1886 
1887 	/* check for address match */
1888 	if (src.s_addr != ip->ip_dst.s_addr ||
1889 	    dst.s_addr != ip->ip_src.s_addr)
1890 		return 0;
1891 
1892 	/* martian filters on outer source - NOT done in ip_input! */
1893 	if (IN_MULTICAST(ip->ip_src.s_addr))
1894 		return 0;
1895 	switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) {
1896 	case 0:
1897 	case 127:
1898 	case 255:
1899 		return 0;
1900 	}
1901 	/* reject packets with broadcast on source */
1902 	s = pserialize_read_enter();
1903 	IN_ADDRLIST_READER_FOREACH(ia4) {
1904 		if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0)
1905 			continue;
1906 		if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) {
1907 			pserialize_read_exit(s);
1908 			return 0;
1909 		}
1910 	}
1911 	pserialize_read_exit(s);
1912 
1913 	/* NOTE: packet may dropped by uRPF */
1914 
1915 	/* return valid bytes length */
1916 	return sizeof(src) + sizeof(dst);
1917 }
1918 
1919 #if NARP > 0
1920 
1921 #define	IN_LLTBL_DEFAULT_HSIZE	32
1922 #define	IN_LLTBL_HASH(k, h) \
1923 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1924 
1925 /*
1926  * Do actual deallocation of @lle.
1927  * Called by LLE_FREE_LOCKED when number of references
1928  * drops to zero.
1929  */
1930 static void
1931 in_lltable_destroy_lle(struct llentry *lle)
1932 {
1933 
1934 	KASSERT(lle->la_numheld == 0);
1935 
1936 	LLE_WUNLOCK(lle);
1937 	LLE_LOCK_DESTROY(lle);
1938 	llentry_pool_put(lle);
1939 }
1940 
1941 static struct llentry *
1942 in_lltable_new(struct in_addr addr4, u_int flags)
1943 {
1944 	struct llentry *lle;
1945 
1946 	lle = llentry_pool_get(PR_NOWAIT);
1947 	if (lle == NULL)		/* NB: caller generates msg */
1948 		return NULL;
1949 
1950 	/*
1951 	 * For IPv4 this will trigger "arpresolve" to generate
1952 	 * an ARP request.
1953 	 */
1954 	lle->la_expire = time_uptime; /* mark expired */
1955 	lle->r_l3addr.addr4 = addr4;
1956 	lle->lle_refcnt = 1;
1957 	lle->lle_free = in_lltable_destroy_lle;
1958 	LLE_LOCK_INIT(lle);
1959 	callout_init(&lle->la_timer, CALLOUT_MPSAFE);
1960 
1961 	return lle;
1962 }
1963 
1964 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1965 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1966 
1967 static int
1968 in_lltable_match_prefix(const struct sockaddr *prefix,
1969     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1970 {
1971 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1972 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1973 	struct in_addr lle_addr;
1974 
1975 	lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr);
1976 
1977 	/*
1978 	 * (flags & LLE_STATIC) means deleting all entries
1979 	 * including static ARP entries.
1980 	 */
1981 	if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) &&
1982 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1983 		return (1);
1984 
1985 	return (0);
1986 }
1987 
1988 static void
1989 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1990 {
1991 	size_t pkts_dropped;
1992 
1993 	LLE_WLOCK_ASSERT(lle);
1994 	KASSERT(llt != NULL);
1995 
1996 	pkts_dropped = llentry_free(lle);
1997 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
1998 }
1999 
2000 static int
2001 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr,
2002     const struct rtentry *rt)
2003 {
2004 	int error = EINVAL;
2005 
2006 	if (rt == NULL)
2007 		return error;
2008 
2009 	/*
2010 	 * If the gateway for an existing host route matches the target L3
2011 	 * address, which is a special route inserted by some implementation
2012 	 * such as MANET, and the interface is of the correct type, then
2013 	 * allow for ARP to proceed.
2014 	 */
2015 	if (rt->rt_flags & RTF_GATEWAY) {
2016 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
2017 		    rt->rt_ifp->if_type != IFT_ETHER ||
2018 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
2019 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
2020 		    sizeof(in_addr_t)) != 0) {
2021 			goto error;
2022 		}
2023 	}
2024 
2025 	/*
2026 	 * Make sure that at least the destination address is covered
2027 	 * by the route. This is for handling the case where 2 or more
2028 	 * interfaces have the same prefix. An incoming packet arrives
2029 	 * on one interface and the corresponding outgoing packet leaves
2030 	 * another interface.
2031 	 */
2032 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
2033 		const char *sa, *mask, *addr, *lim;
2034 		int len;
2035 
2036 		mask = (const char *)rt_mask(rt);
2037 		/*
2038 		 * Just being extra cautious to avoid some custom
2039 		 * code getting into trouble.
2040 		 */
2041 		if (mask == NULL)
2042 			goto error;
2043 
2044 		sa = (const char *)rt_getkey(rt);
2045 		addr = (const char *)l3addr;
2046 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
2047 		lim = addr + len;
2048 
2049 		for ( ; addr < lim; sa++, mask++, addr++) {
2050 			if ((*sa ^ *addr) & *mask) {
2051 #ifdef DIAGNOSTIC
2052 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
2053 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
2054 #endif
2055 				goto error;
2056 			}
2057 		}
2058 	}
2059 
2060 	error = 0;
2061 error:
2062 	return error;
2063 }
2064 
2065 static inline uint32_t
2066 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
2067 {
2068 
2069 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
2070 }
2071 
2072 static uint32_t
2073 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
2074 {
2075 
2076 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
2077 }
2078 
2079 static void
2080 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2081 {
2082 	struct sockaddr_in *sin;
2083 
2084 	sin = (struct sockaddr_in *)sa;
2085 	memset(sin, 0, sizeof(*sin));
2086 	sin->sin_family = AF_INET;
2087 	sin->sin_len = sizeof(*sin);
2088 	sin->sin_addr = lle->r_l3addr.addr4;
2089 }
2090 
2091 static inline struct llentry *
2092 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2093 {
2094 	struct llentry *lle;
2095 	struct llentries *lleh;
2096 	u_int hashidx;
2097 
2098 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2099 	lleh = &llt->lle_head[hashidx];
2100 	LIST_FOREACH(lle, lleh, lle_next) {
2101 		if (lle->la_flags & LLE_DELETED)
2102 			continue;
2103 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2104 			break;
2105 	}
2106 
2107 	return (lle);
2108 }
2109 
2110 static int
2111 in_lltable_delete(struct lltable *llt, u_int flags,
2112     const struct sockaddr *l3addr)
2113 {
2114 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2115 	struct ifnet *ifp __diagused = llt->llt_ifp;
2116 	struct llentry *lle;
2117 
2118 	IF_AFDATA_WLOCK_ASSERT(ifp);
2119 	KASSERTMSG(l3addr->sa_family == AF_INET,
2120 	    "sin_family %d", l3addr->sa_family);
2121 
2122 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2123 	if (lle == NULL) {
2124 #ifdef LLTABLE_DEBUG
2125 		char buf[64];
2126 		sockaddr_format(l3addr, buf, sizeof(buf));
2127 		log(LOG_INFO, "%s: cache for %s is not found\n",
2128 		    __func__, buf);
2129 #endif
2130 		return (ENOENT);
2131 	}
2132 
2133 	LLE_WLOCK(lle);
2134 #ifdef LLTABLE_DEBUG
2135 	{
2136 		char buf[64];
2137 		sockaddr_format(l3addr, buf, sizeof(buf));
2138 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2139 		    __func__, buf, lle);
2140 	}
2141 #endif
2142 	llentry_free(lle);
2143 
2144 	return (0);
2145 }
2146 
2147 static struct llentry *
2148 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr,
2149     const struct rtentry *rt)
2150 {
2151 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2152 	struct ifnet *ifp = llt->llt_ifp;
2153 	struct llentry *lle;
2154 
2155 	IF_AFDATA_WLOCK_ASSERT(ifp);
2156 	KASSERTMSG(l3addr->sa_family == AF_INET,
2157 	    "sin_family %d", l3addr->sa_family);
2158 
2159 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2160 
2161 	if (lle != NULL) {
2162 		LLE_WLOCK(lle);
2163 		return (lle);
2164 	}
2165 
2166 	/* no existing record, we need to create new one */
2167 
2168 	/*
2169 	 * A route that covers the given address must have
2170 	 * been installed 1st because we are doing a resolution,
2171 	 * verify this.
2172 	 */
2173 	if (!(flags & LLE_IFADDR) &&
2174 	    in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2175 		return (NULL);
2176 
2177 	lle = in_lltable_new(sin->sin_addr, flags);
2178 	if (lle == NULL) {
2179 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2180 		return (NULL);
2181 	}
2182 	lle->la_flags = flags;
2183 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2184 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2185 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2186 	}
2187 
2188 	lltable_link_entry(llt, lle);
2189 	LLE_WLOCK(lle);
2190 
2191 	return (lle);
2192 }
2193 
2194 /*
2195  * Return NULL if not found or marked for deletion.
2196  * If found return lle read locked.
2197  */
2198 static struct llentry *
2199 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2200 {
2201 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2202 	struct llentry *lle;
2203 
2204 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2205 	KASSERTMSG(l3addr->sa_family == AF_INET,
2206 	    "sin_family %d", l3addr->sa_family);
2207 
2208 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2209 
2210 	if (lle == NULL)
2211 		return NULL;
2212 
2213 	if (flags & LLE_EXCLUSIVE)
2214 		LLE_WLOCK(lle);
2215 	else
2216 		LLE_RLOCK(lle);
2217 
2218 	return lle;
2219 }
2220 
2221 static int
2222 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2223     struct rt_walkarg *w)
2224 {
2225 	struct sockaddr_in sin;
2226 
2227 	LLTABLE_LOCK_ASSERT();
2228 
2229 	/* skip deleted entries */
2230 	if (lle->la_flags & LLE_DELETED)
2231 		return 0;
2232 
2233 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2234 
2235 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2236 }
2237 
2238 #endif /* NARP > 0 */
2239 
2240 static int
2241 in_multicast_sysctl(SYSCTLFN_ARGS)
2242 {
2243 	struct ifnet *ifp;
2244 	struct ifaddr *ifa;
2245 	struct in_ifaddr *ifa4;
2246 	struct in_multi *inm;
2247 	uint32_t tmp;
2248 	int error;
2249 	size_t written;
2250 	struct psref psref;
2251 	int bound;
2252 
2253 	if (namelen != 1)
2254 		return EINVAL;
2255 
2256 	bound = curlwp_bind();
2257 	ifp = if_get_byindex(name[0], &psref);
2258 	if (ifp == NULL) {
2259 		curlwp_bindx(bound);
2260 		return ENODEV;
2261 	}
2262 
2263 	if (oldp == NULL) {
2264 		*oldlenp = 0;
2265 		IFADDR_FOREACH(ifa, ifp) {
2266 			if (ifa->ifa_addr->sa_family != AF_INET)
2267 				continue;
2268 			ifa4 = (void *)ifa;
2269 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2270 				*oldlenp += 2 * sizeof(struct in_addr) +
2271 				    sizeof(uint32_t);
2272 			}
2273 		}
2274 		if_put(ifp, &psref);
2275 		curlwp_bindx(bound);
2276 		return 0;
2277 	}
2278 
2279 	error = 0;
2280 	written = 0;
2281 	IFADDR_FOREACH(ifa, ifp) {
2282 		if (ifa->ifa_addr->sa_family != AF_INET)
2283 			continue;
2284 		ifa4 = (void *)ifa;
2285 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2286 			if (written + 2 * sizeof(struct in_addr) +
2287 			    sizeof(uint32_t) > *oldlenp)
2288 				goto done;
2289 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2290 			    oldp, sizeof(struct in_addr));
2291 			if (error)
2292 				goto done;
2293 			oldp = (char *)oldp + sizeof(struct in_addr);
2294 			written += sizeof(struct in_addr);
2295 			error = sysctl_copyout(l, &inm->inm_addr,
2296 			    oldp, sizeof(struct in_addr));
2297 			if (error)
2298 				goto done;
2299 			oldp = (char *)oldp + sizeof(struct in_addr);
2300 			written += sizeof(struct in_addr);
2301 			tmp = inm->inm_refcount;
2302 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2303 			if (error)
2304 				goto done;
2305 			oldp = (char *)oldp + sizeof(tmp);
2306 			written += sizeof(tmp);
2307 		}
2308 	}
2309 done:
2310 	if_put(ifp, &psref);
2311 	curlwp_bindx(bound);
2312 	*oldlenp = written;
2313 	return error;
2314 }
2315 
2316 static void
2317 in_sysctl_init(struct sysctllog **clog)
2318 {
2319 	sysctl_createv(clog, 0, NULL, NULL,
2320 		       CTLFLAG_PERMANENT,
2321 		       CTLTYPE_NODE, "inet",
2322 		       SYSCTL_DESCR("PF_INET related settings"),
2323 		       NULL, 0, NULL, 0,
2324 		       CTL_NET, PF_INET, CTL_EOL);
2325 	sysctl_createv(clog, 0, NULL, NULL,
2326 		       CTLFLAG_PERMANENT,
2327 		       CTLTYPE_NODE, "multicast",
2328 		       SYSCTL_DESCR("Multicast information"),
2329 		       in_multicast_sysctl, 0, NULL, 0,
2330 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2331 	sysctl_createv(clog, 0, NULL, NULL,
2332 		       CTLFLAG_PERMANENT,
2333 		       CTLTYPE_NODE, "ip",
2334 		       SYSCTL_DESCR("IPv4 related settings"),
2335 		       NULL, 0, NULL, 0,
2336 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2337 
2338 	sysctl_createv(clog, 0, NULL, NULL,
2339 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2340 		       CTLTYPE_INT, "subnetsarelocal",
2341 		       SYSCTL_DESCR("Whether logical subnets are considered "
2342 				    "local"),
2343 		       NULL, 0, &subnetsarelocal, 0,
2344 		       CTL_NET, PF_INET, IPPROTO_IP,
2345 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2346 	sysctl_createv(clog, 0, NULL, NULL,
2347 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2348 		       CTLTYPE_INT, "hostzerobroadcast",
2349 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2350 		       NULL, 0, &hostzeroisbroadcast, 0,
2351 		       CTL_NET, PF_INET, IPPROTO_IP,
2352 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2353 }
2354 
2355 #if NARP > 0
2356 
2357 static struct lltable *
2358 in_lltattach(struct ifnet *ifp)
2359 {
2360 	struct lltable *llt;
2361 
2362 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2363 	llt->llt_af = AF_INET;
2364 	llt->llt_ifp = ifp;
2365 
2366 	llt->llt_lookup = in_lltable_lookup;
2367 	llt->llt_create = in_lltable_create;
2368 	llt->llt_delete = in_lltable_delete;
2369 	llt->llt_dump_entry = in_lltable_dump_entry;
2370 	llt->llt_hash = in_lltable_hash;
2371 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2372 	llt->llt_free_entry = in_lltable_free_entry;
2373 	llt->llt_match_prefix = in_lltable_match_prefix;
2374 	lltable_link(llt);
2375 
2376 	return (llt);
2377 }
2378 
2379 #endif /* NARP > 0 */
2380 
2381 void *
2382 in_domifattach(struct ifnet *ifp)
2383 {
2384 	struct in_ifinfo *ii;
2385 
2386 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2387 
2388 #if NARP > 0
2389 	ii->ii_llt = in_lltattach(ifp);
2390 #endif
2391 
2392 #ifdef IPSELSRC
2393 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2394 	KASSERT(ii->ii_selsrc != NULL);
2395 #endif
2396 
2397 	return ii;
2398 }
2399 
2400 void
2401 in_domifdetach(struct ifnet *ifp, void *aux)
2402 {
2403 	struct in_ifinfo *ii = aux;
2404 
2405 #ifdef IPSELSRC
2406 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2407 #endif
2408 #if NARP > 0
2409 	lltable_free(ii->ii_llt);
2410 #endif
2411 	kmem_free(ii, sizeof(struct in_ifinfo));
2412 }
2413