xref: /netbsd-src/sys/netinet/in.c (revision 8450a7c42673d65e3b1f6560d3b6ecd317a6cbe8)
1 /*	$NetBSD: in.c,v 1.187 2016/10/18 07:30:31 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 1.187 2016/10/18 07:30:31 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 int	in_addprefix(struct in_ifaddr *, int);
153 static void	in_scrubaddr(struct in_ifaddr *);
154 static int	in_scrubprefix(struct in_ifaddr *);
155 static void	in_sysctl_init(struct sysctllog **);
156 
157 #ifndef SUBNETSARELOCAL
158 #define	SUBNETSARELOCAL	1
159 #endif
160 
161 #ifndef HOSTZEROBROADCAST
162 #define HOSTZEROBROADCAST 0
163 #endif
164 
165 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
166 #ifndef IN_MULTI_HASH_SIZE
167 #define IN_MULTI_HASH_SIZE	509
168 #endif
169 
170 static int			subnetsarelocal = SUBNETSARELOCAL;
171 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
172 
173 /*
174  * This list is used to keep track of in_multi chains which belong to
175  * deleted interface addresses.  We use in_ifaddr so that a chain head
176  * won't be deallocated until all multicast address record are deleted.
177  */
178 
179 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
180 
181 static struct pool		inmulti_pool;
182 static u_int			in_multientries;
183 static struct in_multihashhead *in_multihashtbl;
184 static u_long			in_multihash;
185 static krwlock_t		in_multilock;
186 
187 #define IN_MULTI_HASH(x, ifp) \
188     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
189 
190 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */
191 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
192 u_long				in_ifaddrhash;
193 struct in_ifaddrhead		in_ifaddrhead;
194 static kmutex_t			in_ifaddr_lock;
195 
196 struct pslist_head *		in_ifaddrhashtbl_pslist;
197 u_long				in_ifaddrhash_pslist;
198 struct pslist_head		in_ifaddrhead_pslist;
199 
200 void
201 in_init(void)
202 {
203 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
204 	    NULL, IPL_SOFTNET);
205 	TAILQ_INIT(&in_ifaddrhead);
206 	PSLIST_INIT(&in_ifaddrhead_pslist);
207 
208 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
209 	    &in_ifaddrhash);
210 
211 	in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST,
212 	    true, &in_ifaddrhash_pslist);
213 	mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
214 
215 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
216 	    &in_multihash);
217 	rw_init(&in_multilock);
218 
219 	in_sysctl_init(NULL);
220 }
221 
222 /*
223  * Return 1 if an internet address is for a ``local'' host
224  * (one to which we have a connection).  If subnetsarelocal
225  * is true, this includes other subnets of the local net.
226  * Otherwise, it includes only the directly-connected (sub)nets.
227  */
228 int
229 in_localaddr(struct in_addr in)
230 {
231 	struct in_ifaddr *ia;
232 	int localaddr = 0;
233 	int s = pserialize_read_enter();
234 
235 	if (subnetsarelocal) {
236 		IN_ADDRLIST_READER_FOREACH(ia) {
237 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net) {
238 				localaddr = 1;
239 				break;
240 			}
241 		}
242 	} else {
243 		IN_ADDRLIST_READER_FOREACH(ia) {
244 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
245 				localaddr = 1;
246 				break;
247 			}
248 		}
249 	}
250 	pserialize_read_exit(s);
251 
252 	return localaddr;
253 }
254 
255 /*
256  * Determine whether an IP address is in a reserved set of addresses
257  * that may not be forwarded, or whether datagrams to that destination
258  * may be forwarded.
259  */
260 int
261 in_canforward(struct in_addr in)
262 {
263 	u_int32_t net;
264 
265 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
266 		return (0);
267 	if (IN_CLASSA(in.s_addr)) {
268 		net = in.s_addr & IN_CLASSA_NET;
269 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
270 			return (0);
271 	}
272 	return (1);
273 }
274 
275 /*
276  * Trim a mask in a sockaddr
277  */
278 void
279 in_socktrim(struct sockaddr_in *ap)
280 {
281 	char *cplim = (char *) &ap->sin_addr;
282 	char *cp = (char *) (&ap->sin_addr + 1);
283 
284 	ap->sin_len = 0;
285 	while (--cp >= cplim)
286 		if (*cp) {
287 			(ap)->sin_len = cp - (char *) (ap) + 1;
288 			break;
289 		}
290 }
291 
292 /*
293  *  Routine to take an Internet address and convert into a
294  *  "dotted quad" representation for printing.
295  */
296 const char *
297 in_fmtaddr(struct in_addr addr)
298 {
299 	static char buf[sizeof("123.456.789.123")];
300 
301 	addr.s_addr = ntohl(addr.s_addr);
302 
303 	snprintf(buf, sizeof(buf), "%d.%d.%d.%d",
304 		(addr.s_addr >> 24) & 0xFF,
305 		(addr.s_addr >> 16) & 0xFF,
306 		(addr.s_addr >>  8) & 0xFF,
307 		(addr.s_addr >>  0) & 0xFF);
308 	return buf;
309 }
310 
311 /*
312  * Maintain the "in_maxmtu" variable, which is the largest
313  * mtu for non-local interfaces with AF_INET addresses assigned
314  * to them that are up.
315  */
316 unsigned long in_maxmtu;
317 
318 void
319 in_setmaxmtu(void)
320 {
321 	struct in_ifaddr *ia;
322 	struct ifnet *ifp;
323 	unsigned long maxmtu = 0;
324 	int s = pserialize_read_enter();
325 
326 	IN_ADDRLIST_READER_FOREACH(ia) {
327 		if ((ifp = ia->ia_ifp) == 0)
328 			continue;
329 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
330 			continue;
331 		if (ifp->if_mtu > maxmtu)
332 			maxmtu = ifp->if_mtu;
333 	}
334 	if (maxmtu)
335 		in_maxmtu = maxmtu;
336 	pserialize_read_exit(s);
337 }
338 
339 static u_int
340 in_mask2len(struct in_addr *mask)
341 {
342 	u_int x, y;
343 	u_char *p;
344 
345 	p = (u_char *)mask;
346 	for (x = 0; x < sizeof(*mask); x++) {
347 		if (p[x] != 0xff)
348 			break;
349 	}
350 	y = 0;
351 	if (x < sizeof(*mask)) {
352 		for (y = 0; y < NBBY; y++) {
353 			if ((p[x] & (0x80 >> y)) == 0)
354 				break;
355 		}
356 	}
357 	return x * NBBY + y;
358 }
359 
360 static void
361 in_len2mask(struct in_addr *mask, u_int len)
362 {
363 	u_int i;
364 	u_char *p;
365 
366 	p = (u_char *)mask;
367 	memset(mask, 0, sizeof(*mask));
368 	for (i = 0; i < len / NBBY; i++)
369 		p[i] = 0xff;
370 	if (len % NBBY)
371 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
372 }
373 
374 /*
375  * Generic internet control operations (ioctl's).
376  * Ifp is 0 if not an interface-specific ioctl.
377  */
378 /* ARGSUSED */
379 static int
380 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
381 {
382 	struct ifreq *ifr = (struct ifreq *)data;
383 	struct in_ifaddr *ia = NULL;
384 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
385 	struct sockaddr_in oldaddr, *new_dstaddr;
386 	int error, hostIsNew, maskIsNew;
387 	int newifaddr = 0;
388 	bool run_hook = false;
389 	bool need_reinsert = false;
390 	struct psref psref;
391 	int bound;
392 
393 	switch (cmd) {
394 	case SIOCALIFADDR:
395 	case SIOCDLIFADDR:
396 	case SIOCGLIFADDR:
397 		if (ifp == NULL)
398 			return EINVAL;
399 		return in_lifaddr_ioctl(so, cmd, data, ifp);
400 	case SIOCGIFADDRPREF:
401 	case SIOCSIFADDRPREF:
402 		if (ifp == NULL)
403 			return EINVAL;
404 		return ifaddrpref_ioctl(so, cmd, data, ifp);
405 	}
406 
407 	bound = curlwp_bind();
408 	/*
409 	 * Find address for this interface, if it exists.
410 	 */
411 	if (ifp != NULL)
412 		ia = in_get_ia_from_ifp_psref(ifp, &psref);
413 
414 	hostIsNew = 1;		/* moved here to appease gcc */
415 	switch (cmd) {
416 	case SIOCAIFADDR:
417 	case SIOCDIFADDR:
418 	case SIOCGIFALIAS:
419 	case SIOCGIFAFLAG_IN:
420 		if (ifra->ifra_addr.sin_family == AF_INET) {
421 			int s;
422 
423 			if (ia != NULL)
424 				ia4_release(ia, &psref);
425 			s = pserialize_read_enter();
426 			IN_ADDRHASH_READER_FOREACH(ia,
427 			    ifra->ifra_addr.sin_addr.s_addr) {
428 				if (ia->ia_ifp == ifp &&
429 				    in_hosteq(ia->ia_addr.sin_addr,
430 				    ifra->ifra_addr.sin_addr))
431 					break;
432 			}
433 			if (ia != NULL)
434 				ia4_acquire(ia, &psref);
435 			pserialize_read_exit(s);
436 		}
437 		if ((cmd == SIOCDIFADDR ||
438 		    cmd == SIOCGIFALIAS ||
439 		    cmd == SIOCGIFAFLAG_IN) &&
440 		    ia == NULL) {
441 			error = EADDRNOTAVAIL;
442 			goto out;
443 		}
444 
445 		if (cmd == SIOCDIFADDR &&
446 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
447 			ifra->ifra_addr.sin_family = AF_INET;
448 		}
449 		/* FALLTHROUGH */
450 	case SIOCSIFADDR:
451 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
452 			;
453 		else if (ifra->ifra_addr.sin_len == 0) {
454 			ifra->ifra_addr = ia->ia_addr;
455 			hostIsNew = 0;
456 		} else if (in_hosteq(ia->ia_addr.sin_addr,
457 		           ifra->ifra_addr.sin_addr))
458 			hostIsNew = 0;
459 		/* FALLTHROUGH */
460 	case SIOCSIFDSTADDR:
461 		if (ifra->ifra_addr.sin_family != AF_INET) {
462 			error = EAFNOSUPPORT;
463 			goto out;
464 		}
465 		/* FALLTHROUGH */
466 	case SIOCSIFNETMASK:
467 		if (ifp == NULL)
468 			panic("in_control");
469 
470 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
471 			break;
472 
473 		if (ia == NULL &&
474 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
475 			error = EADDRNOTAVAIL;
476 			goto out;
477 		}
478 
479 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
480 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
481 		    NULL) != 0) {
482 			error = EPERM;
483 			goto out;
484 		}
485 
486 		if (ia == NULL) {
487 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
488 			if (ia == NULL) {
489 				error = ENOBUFS;
490 				goto out;
491 			}
492 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
493 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
494 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
495 #ifdef IPSELSRC
496 			ia->ia_ifa.ifa_getifa = in_getifa;
497 #else /* IPSELSRC */
498 			ia->ia_ifa.ifa_getifa = NULL;
499 #endif /* IPSELSRC */
500 			ia->ia_sockmask.sin_len = 8;
501 			ia->ia_sockmask.sin_family = AF_INET;
502 			if (ifp->if_flags & IFF_BROADCAST) {
503 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
504 				ia->ia_broadaddr.sin_family = AF_INET;
505 			}
506 			ia->ia_ifp = ifp;
507 			ia->ia_idsalt = cprng_fast32() % 65535;
508 			LIST_INIT(&ia->ia_multiaddrs);
509 			IN_ADDRHASH_ENTRY_INIT(ia);
510 			IN_ADDRLIST_ENTRY_INIT(ia);
511 			ifa_psref_init(&ia->ia_ifa);
512 
513 			newifaddr = 1;
514 		}
515 		break;
516 
517 	case SIOCSIFBRDADDR:
518 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
519 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
520 		    NULL) != 0) {
521 			error = EPERM;
522 			goto out;
523 		}
524 		/* FALLTHROUGH */
525 
526 	case SIOCGIFADDR:
527 	case SIOCGIFNETMASK:
528 	case SIOCGIFDSTADDR:
529 	case SIOCGIFBRDADDR:
530 		if (ia == NULL) {
531 			error = EADDRNOTAVAIL;
532 			goto out;
533 		}
534 		break;
535 	}
536 	error = 0;
537 	switch (cmd) {
538 
539 	case SIOCGIFADDR:
540 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
541 		break;
542 
543 	case SIOCGIFBRDADDR:
544 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
545 			error = EINVAL;
546 			goto out;
547 		}
548 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
549 		break;
550 
551 	case SIOCGIFDSTADDR:
552 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
553 			error = EINVAL;
554 			goto out;
555 		}
556 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
557 		break;
558 
559 	case SIOCGIFNETMASK:
560 		/*
561 		 * We keep the number of trailing zero bytes the sin_len field
562 		 * of ia_sockmask, so we fix this before we pass it back to
563 		 * userland.
564 		 */
565 		oldaddr = ia->ia_sockmask;
566 		oldaddr.sin_len = sizeof(struct sockaddr_in);
567 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
568 		break;
569 
570 	case SIOCSIFDSTADDR:
571 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
572 			error = EINVAL;
573 			goto out;
574 		}
575 		oldaddr = ia->ia_dstaddr;
576 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
577 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
578 			ia->ia_dstaddr = oldaddr;
579 			goto out;
580 		}
581 		if (ia->ia_flags & IFA_ROUTE) {
582 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
583 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
584 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
585 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
586 		}
587 		break;
588 
589 	case SIOCSIFBRDADDR:
590 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
591 			error = EINVAL;
592 			goto out;
593 		}
594 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
595 		break;
596 
597 	case SIOCSIFADDR:
598 		if (!newifaddr) {
599 			mutex_enter(&in_ifaddr_lock);
600 			LIST_REMOVE(ia, ia_hash);
601 			IN_ADDRHASH_WRITER_REMOVE(ia);
602 			mutex_exit(&in_ifaddr_lock);
603 			need_reinsert = true;
604 		}
605 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
606 		    NULL, 1);
607 
608 		run_hook = true;
609 		break;
610 
611 	case SIOCSIFNETMASK:
612 		in_scrubprefix(ia);
613 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
614 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
615 		if (!newifaddr) {
616 			mutex_enter(&in_ifaddr_lock);
617 			LIST_REMOVE(ia, ia_hash);
618 			IN_ADDRHASH_WRITER_REMOVE(ia);
619 			mutex_exit(&in_ifaddr_lock);
620 			need_reinsert = true;
621 		}
622 		error = in_ifinit(ifp, ia, NULL, NULL, 0);
623 		break;
624 
625 	case SIOCAIFADDR:
626 		maskIsNew = 0;
627 		if (ifra->ifra_mask.sin_len) {
628 			in_scrubprefix(ia);
629 			ia->ia_sockmask = ifra->ifra_mask;
630 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
631 			maskIsNew = 1;
632 		}
633 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
634 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
635 			new_dstaddr = &ifra->ifra_dstaddr;
636 			maskIsNew  = 1; /* We lie; but the effect's the same */
637 		} else
638 			new_dstaddr = NULL;
639 		if (ifra->ifra_addr.sin_family == AF_INET &&
640 		    (hostIsNew || maskIsNew)) {
641 			if (!newifaddr) {
642 				mutex_enter(&in_ifaddr_lock);
643 				LIST_REMOVE(ia, ia_hash);
644 				IN_ADDRHASH_WRITER_REMOVE(ia);
645 				mutex_exit(&in_ifaddr_lock);
646 				need_reinsert = true;
647 			}
648 			error = in_ifinit(ifp, ia, &ifra->ifra_addr,
649 			    new_dstaddr, 0);
650 		}
651 		if ((ifp->if_flags & IFF_BROADCAST) &&
652 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
653 			ia->ia_broadaddr = ifra->ifra_broadaddr;
654 		run_hook = true;
655 		break;
656 
657 	case SIOCGIFALIAS:
658 		ifra->ifra_mask = ia->ia_sockmask;
659 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
660 		    (ia->ia_dstaddr.sin_family == AF_INET))
661 			ifra->ifra_dstaddr = ia->ia_dstaddr;
662 		else if ((ifp->if_flags & IFF_BROADCAST) &&
663 		    (ia->ia_broadaddr.sin_family == AF_INET))
664 			ifra->ifra_broadaddr = ia->ia_broadaddr;
665 		else
666 			memset(&ifra->ifra_broadaddr, 0,
667 			      sizeof(ifra->ifra_broadaddr));
668 		break;
669 
670 	case SIOCGIFAFLAG_IN:
671 		ifr->ifr_addrflags = ia->ia4_flags;
672 		break;
673 
674 	case SIOCDIFADDR:
675 		ia4_release(ia, &psref);
676 		in_purgeaddr(&ia->ia_ifa);
677 		ia = NULL;
678 		run_hook = true;
679 		break;
680 
681 #ifdef MROUTING
682 	case SIOCGETVIFCNT:
683 	case SIOCGETSGCNT:
684 		error = mrt_ioctl(so, cmd, data);
685 		break;
686 #endif /* MROUTING */
687 
688 	default:
689 		error = ENOTTY;
690 		goto out;
691 	}
692 
693 	/*
694 	 * XXX insert regardless of error to make in_purgeaddr below work.
695 	 * Need to improve.
696 	 */
697 	if (newifaddr) {
698 		ifaref(&ia->ia_ifa);
699 		ifa_insert(ifp, &ia->ia_ifa);
700 
701 		mutex_enter(&in_ifaddr_lock);
702 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
703 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
704 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
705 		    ia, ia_hash);
706 		IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
707 		mutex_exit(&in_ifaddr_lock);
708 	} else if (need_reinsert) {
709 		mutex_enter(&in_ifaddr_lock);
710 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
711 		    ia, ia_hash);
712 		IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
713 		mutex_exit(&in_ifaddr_lock);
714 	}
715 
716 	if (error == 0) {
717 		if (run_hook)
718 			(void)pfil_run_hooks(if_pfil,
719 			    (struct mbuf **)cmd, ifp, PFIL_IFADDR);
720 	} else if (newifaddr) {
721 		KASSERT(ia != NULL);
722 		in_purgeaddr(&ia->ia_ifa);
723 		ia = NULL;
724 	}
725 
726 out:
727 	if (!newifaddr && ia != NULL)
728 		ia4_release(ia, &psref);
729 	curlwp_bindx(bound);
730 	return error;
731 }
732 
733 int
734 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
735 {
736 	int error;
737 
738 #ifndef NET_MPSAFE
739 	mutex_enter(softnet_lock);
740 #endif
741 	error = in_control0(so, cmd, data, ifp);
742 #ifndef NET_MPSAFE
743 	mutex_exit(softnet_lock);
744 #endif
745 
746 	return error;
747 }
748 
749 /* Add ownaddr as loopback rtentry. */
750 static void
751 in_ifaddlocal(struct ifaddr *ifa)
752 {
753 	struct in_ifaddr *ia;
754 
755 	ia = (struct in_ifaddr *)ifa;
756 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
757 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
758 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
759 	{
760 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
761 		return;
762 	}
763 
764 	rt_ifa_addlocal(ifa);
765 }
766 
767 /* Remove loopback entry of ownaddr */
768 static void
769 in_ifremlocal(struct ifaddr *ifa)
770 {
771 	struct in_ifaddr *ia, *p;
772 	struct ifaddr *alt_ifa = NULL;
773 	int ia_count = 0;
774 	int s;
775 	struct psref psref;
776 	int bound = curlwp_bind();
777 
778 	ia = (struct in_ifaddr *)ifa;
779 	/* Delete the entry if exactly one ifaddr matches the
780 	 * address, ifa->ifa_addr. */
781 	s = pserialize_read_enter();
782 	IN_ADDRLIST_READER_FOREACH(p) {
783 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
784 			continue;
785 		if (p->ia_ifp != ia->ia_ifp)
786 			alt_ifa = &p->ia_ifa;
787 		if (++ia_count > 1 && alt_ifa != NULL)
788 			break;
789 	}
790 	if (alt_ifa != NULL && ia_count > 1)
791 		ifa_acquire(alt_ifa, &psref);
792 	pserialize_read_exit(s);
793 
794 	if (ia_count == 0)
795 		goto out;
796 
797 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
798 	if (alt_ifa != NULL && ia_count > 1)
799 		ifa_release(alt_ifa, &psref);
800 out:
801 	curlwp_bindx(bound);
802 }
803 
804 static void
805 in_scrubaddr(struct in_ifaddr *ia)
806 {
807 
808 	/* stop DAD processing */
809 	if (ia->ia_dad_stop != NULL)
810 		ia->ia_dad_stop(&ia->ia_ifa);
811 
812 	in_scrubprefix(ia);
813 	in_ifremlocal(&ia->ia_ifa);
814 
815 	if (ia->ia_allhosts != NULL) {
816 		in_delmulti(ia->ia_allhosts);
817 		ia->ia_allhosts = NULL;
818 	}
819 }
820 
821 /*
822  * Depends on it isn't called in concurrent. It should be guaranteed
823  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
824  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
825  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
826  * never be executed in concurrent.
827  */
828 void
829 in_purgeaddr(struct ifaddr *ifa)
830 {
831 	struct in_ifaddr *ia = (void *) ifa;
832 	struct ifnet *ifp = ifa->ifa_ifp;
833 
834 	KASSERT(!ifa_held(ifa));
835 
836 	in_scrubaddr(ia);
837 
838 	mutex_enter(&in_ifaddr_lock);
839 	LIST_REMOVE(ia, ia_hash);
840 	IN_ADDRHASH_WRITER_REMOVE(ia);
841 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
842 	IN_ADDRLIST_WRITER_REMOVE(ia);
843 	ifa_remove(ifp, &ia->ia_ifa);
844 	mutex_exit(&in_ifaddr_lock);
845 
846 	IN_ADDRHASH_ENTRY_DESTROY(ia);
847 	IN_ADDRLIST_ENTRY_DESTROY(ia);
848 	ifafree(&ia->ia_ifa);
849 	in_setmaxmtu();
850 }
851 
852 void
853 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
854 {
855 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
856 	igmp_purgeif(ifp);		/* manipulates pools */
857 #ifdef MROUTING
858 	ip_mrouter_detach(ifp);
859 #endif
860 }
861 
862 /*
863  * SIOC[GAD]LIFADDR.
864  *	SIOCGLIFADDR: get first address. (???)
865  *	SIOCGLIFADDR with IFLR_PREFIX:
866  *		get first address that matches the specified prefix.
867  *	SIOCALIFADDR: add the specified address.
868  *	SIOCALIFADDR with IFLR_PREFIX:
869  *		EINVAL since we can't deduce hostid part of the address.
870  *	SIOCDLIFADDR: delete the specified address.
871  *	SIOCDLIFADDR with IFLR_PREFIX:
872  *		delete the first address that matches the specified prefix.
873  * return values:
874  *	EINVAL on invalid parameters
875  *	EADDRNOTAVAIL on prefix match failed/specified address not found
876  *	other values may be returned from in_ioctl()
877  */
878 static int
879 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
880     struct ifnet *ifp)
881 {
882 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
883 	struct ifaddr *ifa;
884 	struct sockaddr *sa;
885 
886 	/* sanity checks */
887 	if (data == NULL || ifp == NULL) {
888 		panic("invalid argument to in_lifaddr_ioctl");
889 		/*NOTRECHED*/
890 	}
891 
892 	switch (cmd) {
893 	case SIOCGLIFADDR:
894 		/* address must be specified on GET with IFLR_PREFIX */
895 		if ((iflr->flags & IFLR_PREFIX) == 0)
896 			break;
897 		/*FALLTHROUGH*/
898 	case SIOCALIFADDR:
899 	case SIOCDLIFADDR:
900 		/* address must be specified on ADD and DELETE */
901 		sa = (struct sockaddr *)&iflr->addr;
902 		if (sa->sa_family != AF_INET)
903 			return EINVAL;
904 		if (sa->sa_len != sizeof(struct sockaddr_in))
905 			return EINVAL;
906 		/* XXX need improvement */
907 		sa = (struct sockaddr *)&iflr->dstaddr;
908 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
909 			return EINVAL;
910 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
911 			return EINVAL;
912 		break;
913 	default: /*shouldn't happen*/
914 #if 0
915 		panic("invalid cmd to in_lifaddr_ioctl");
916 		/*NOTREACHED*/
917 #else
918 		return EOPNOTSUPP;
919 #endif
920 	}
921 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
922 		return EINVAL;
923 
924 	switch (cmd) {
925 	case SIOCALIFADDR:
926 	    {
927 		struct in_aliasreq ifra;
928 
929 		if (iflr->flags & IFLR_PREFIX)
930 			return EINVAL;
931 
932 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
933 		memset(&ifra, 0, sizeof(ifra));
934 		memcpy(ifra.ifra_name, iflr->iflr_name,
935 			sizeof(ifra.ifra_name));
936 
937 		memcpy(&ifra.ifra_addr, &iflr->addr,
938 			((struct sockaddr *)&iflr->addr)->sa_len);
939 
940 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
941 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
942 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
943 		}
944 
945 		ifra.ifra_mask.sin_family = AF_INET;
946 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
947 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
948 
949 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
950 	    }
951 	case SIOCGLIFADDR:
952 	case SIOCDLIFADDR:
953 	    {
954 		struct in_ifaddr *ia;
955 		struct in_addr mask, candidate, match;
956 		struct sockaddr_in *sin;
957 		int cmp, s;
958 
959 		memset(&mask, 0, sizeof(mask));
960 		memset(&match, 0, sizeof(match));	/* XXX gcc */
961 		if (iflr->flags & IFLR_PREFIX) {
962 			/* lookup a prefix rather than address. */
963 			in_len2mask(&mask, iflr->prefixlen);
964 
965 			sin = (struct sockaddr_in *)&iflr->addr;
966 			match.s_addr = sin->sin_addr.s_addr;
967 			match.s_addr &= mask.s_addr;
968 
969 			/* if you set extra bits, that's wrong */
970 			if (match.s_addr != sin->sin_addr.s_addr)
971 				return EINVAL;
972 
973 			cmp = 1;
974 		} else {
975 			if (cmd == SIOCGLIFADDR) {
976 				/* on getting an address, take the 1st match */
977 				cmp = 0;	/*XXX*/
978 			} else {
979 				/* on deleting an address, do exact match */
980 				in_len2mask(&mask, 32);
981 				sin = (struct sockaddr_in *)&iflr->addr;
982 				match.s_addr = sin->sin_addr.s_addr;
983 
984 				cmp = 1;
985 			}
986 		}
987 
988 		s = pserialize_read_enter();
989 		IFADDR_READER_FOREACH(ifa, ifp) {
990 			if (ifa->ifa_addr->sa_family != AF_INET)
991 				continue;
992 			if (cmp == 0)
993 				break;
994 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
995 			candidate.s_addr &= mask.s_addr;
996 			if (candidate.s_addr == match.s_addr)
997 				break;
998 		}
999 		if (ifa == NULL) {
1000 			pserialize_read_exit(s);
1001 			return EADDRNOTAVAIL;
1002 		}
1003 		ia = (struct in_ifaddr *)ifa;
1004 
1005 		if (cmd == SIOCGLIFADDR) {
1006 			/* fill in the if_laddrreq structure */
1007 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
1008 
1009 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1010 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1011 					ia->ia_dstaddr.sin_len);
1012 			} else
1013 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1014 
1015 			iflr->prefixlen =
1016 				in_mask2len(&ia->ia_sockmask.sin_addr);
1017 
1018 			iflr->flags = 0;	/*XXX*/
1019 			pserialize_read_exit(s);
1020 
1021 			return 0;
1022 		} else {
1023 			struct in_aliasreq ifra;
1024 
1025 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1026 			memset(&ifra, 0, sizeof(ifra));
1027 			memcpy(ifra.ifra_name, iflr->iflr_name,
1028 				sizeof(ifra.ifra_name));
1029 
1030 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1031 				ia->ia_addr.sin_len);
1032 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1033 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1034 					ia->ia_dstaddr.sin_len);
1035 			}
1036 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1037 				ia->ia_sockmask.sin_len);
1038 			pserialize_read_exit(s);
1039 
1040 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1041 		}
1042 	    }
1043 	}
1044 
1045 	return EOPNOTSUPP;	/*just for safety*/
1046 }
1047 
1048 /*
1049  * Initialize an interface's internet address
1050  * and routing table entry.
1051  */
1052 int
1053 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1054     const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub)
1055 {
1056 	u_int32_t i;
1057 	struct sockaddr_in oldaddr, olddst;
1058 	int s, oldflags, flags = RTF_UP, error, hostIsNew;
1059 
1060 	if (sin == NULL)
1061 		sin = &ia->ia_addr;
1062 	if (dst == NULL)
1063 		dst = &ia->ia_dstaddr;
1064 
1065 	/*
1066 	 * Set up new addresses.
1067 	 */
1068 	oldaddr = ia->ia_addr;
1069 	olddst = ia->ia_dstaddr;
1070 	oldflags = ia->ia4_flags;
1071 	ia->ia_addr = *sin;
1072 	ia->ia_dstaddr = *dst;
1073 	hostIsNew = oldaddr.sin_family != AF_INET ||
1074 	    !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr);
1075 	if (!scrub)
1076 		scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family ||
1077 		    !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr);
1078 
1079 	/*
1080 	 * Configure address flags.
1081 	 * We need to do this early because they maybe adjusted
1082 	 * by if_addr_init depending on the address.
1083 	 */
1084 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
1085 		ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1086 		hostIsNew = 1;
1087 	}
1088 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1089 		ia->ia4_flags |= IN_IFF_DETACHED;
1090 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
1091 	} else if (hostIsNew && if_do_dad(ifp))
1092 		ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1093 
1094 	/*
1095 	 * Give the interface a chance to initialize
1096 	 * if this is its first address,
1097 	 * and to validate the address if necessary.
1098 	 */
1099 	s = splnet();
1100 	error = if_addr_init(ifp, &ia->ia_ifa, true);
1101 	splx(s);
1102 	/* Now clear the try tentative flag, it's job is done. */
1103 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1104 	if (error != 0) {
1105 		ia->ia_addr = oldaddr;
1106 		ia->ia_dstaddr = olddst;
1107 		ia->ia4_flags = oldflags;
1108 		return error;
1109 	}
1110 
1111 	if (scrub || hostIsNew) {
1112 		int newflags = ia->ia4_flags;
1113 
1114 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1115 		ia->ia_ifa.ifa_dstaddr = sintosa(&olddst);
1116 		ia->ia4_flags = oldflags;
1117 		if (hostIsNew)
1118 			in_scrubaddr(ia);
1119 		else if (scrub)
1120 			in_scrubprefix(ia);
1121 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1122 		ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
1123 		ia->ia4_flags = newflags;
1124 	}
1125 
1126 	/* Add the local route to the address */
1127 	in_ifaddlocal(&ia->ia_ifa);
1128 
1129 	i = ia->ia_addr.sin_addr.s_addr;
1130 	if (ifp->if_flags & IFF_POINTOPOINT)
1131 		ia->ia_netmask = INADDR_BROADCAST;	/* default to /32 */
1132 	else if (IN_CLASSA(i))
1133 		ia->ia_netmask = IN_CLASSA_NET;
1134 	else if (IN_CLASSB(i))
1135 		ia->ia_netmask = IN_CLASSB_NET;
1136 	else
1137 		ia->ia_netmask = IN_CLASSC_NET;
1138 	/*
1139 	 * The subnet mask usually includes at least the standard network part,
1140 	 * but may may be smaller in the case of supernetting.
1141 	 * If it is set, we believe it.
1142 	 */
1143 	if (ia->ia_subnetmask == 0) {
1144 		ia->ia_subnetmask = ia->ia_netmask;
1145 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1146 	} else
1147 		ia->ia_netmask &= ia->ia_subnetmask;
1148 
1149 	ia->ia_net = i & ia->ia_netmask;
1150 	ia->ia_subnet = i & ia->ia_subnetmask;
1151 	in_socktrim(&ia->ia_sockmask);
1152 	/* re-calculate the "in_maxmtu" value */
1153 	in_setmaxmtu();
1154 	/*
1155 	 * Add route for the network.
1156 	 */
1157 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1158 	if (ifp->if_flags & IFF_BROADCAST) {
1159 		ia->ia_broadaddr.sin_addr.s_addr =
1160 			ia->ia_subnet | ~ia->ia_subnetmask;
1161 		ia->ia_netbroadcast.s_addr =
1162 			ia->ia_net | ~ia->ia_netmask;
1163 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1164 		ia->ia_dstaddr = ia->ia_addr;
1165 		flags |= RTF_HOST;
1166 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1167 		if (ia->ia_dstaddr.sin_family != AF_INET)
1168 			return (0);
1169 		flags |= RTF_HOST;
1170 	}
1171 	error = in_addprefix(ia, flags);
1172 	/*
1173 	 * If the interface supports multicast, join the "all hosts"
1174 	 * multicast group on that interface.
1175 	 */
1176 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1177 		struct in_addr addr;
1178 
1179 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1180 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1181 	}
1182 
1183 	if (hostIsNew &&
1184 	    ia->ia4_flags & IN_IFF_TENTATIVE &&
1185 	    if_do_dad(ifp))
1186 		ia->ia_dad_start((struct ifaddr *)ia);
1187 
1188 	return error;
1189 }
1190 
1191 #define rtinitflags(x) \
1192 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1193 	    ? RTF_HOST : 0)
1194 
1195 /*
1196  * add a route to prefix ("connected route" in cisco terminology).
1197  * does nothing if there's some interface address with the same prefix already.
1198  */
1199 static int
1200 in_addprefix(struct in_ifaddr *target, int flags)
1201 {
1202 	struct in_ifaddr *ia;
1203 	struct in_addr prefix, mask, p;
1204 	int error;
1205 	int s;
1206 
1207 	if ((flags & RTF_HOST) != 0)
1208 		prefix = target->ia_dstaddr.sin_addr;
1209 	else {
1210 		prefix = target->ia_addr.sin_addr;
1211 		mask = target->ia_sockmask.sin_addr;
1212 		prefix.s_addr &= mask.s_addr;
1213 	}
1214 
1215 	s = pserialize_read_enter();
1216 	IN_ADDRLIST_READER_FOREACH(ia) {
1217 		if (rtinitflags(ia))
1218 			p = ia->ia_dstaddr.sin_addr;
1219 		else {
1220 			p = ia->ia_addr.sin_addr;
1221 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1222 		}
1223 
1224 		if (prefix.s_addr != p.s_addr)
1225 			continue;
1226 
1227 		/*
1228 		 * if we got a matching prefix route inserted by other
1229 		 * interface address, we don't need to bother
1230 		 *
1231 		 * XXX RADIX_MPATH implications here? -dyoung
1232 		 */
1233 		if (ia->ia_flags & IFA_ROUTE) {
1234 			pserialize_read_exit(s);
1235 			return 0;
1236 		}
1237 	}
1238 	pserialize_read_exit(s);
1239 
1240 	/*
1241 	 * noone seem to have prefix route.  insert it.
1242 	 */
1243 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1244 	if (error == 0)
1245 		target->ia_flags |= IFA_ROUTE;
1246 	else if (error == EEXIST) {
1247 		/*
1248 		 * the fact the route already exists is not an error.
1249 		 */
1250 		error = 0;
1251 	}
1252 	return error;
1253 }
1254 
1255 /*
1256  * remove a route to prefix ("connected route" in cisco terminology).
1257  * re-installs the route by using another interface address, if there's one
1258  * with the same prefix (otherwise we lose the route mistakenly).
1259  */
1260 static int
1261 in_scrubprefix(struct in_ifaddr *target)
1262 {
1263 	struct in_ifaddr *ia;
1264 	struct in_addr prefix, mask, p;
1265 	int error;
1266 	int s;
1267 
1268 	/* If we don't have IFA_ROUTE we have nothing to do */
1269 	if ((target->ia_flags & IFA_ROUTE) == 0)
1270 		return 0;
1271 
1272 	if (rtinitflags(target))
1273 		prefix = target->ia_dstaddr.sin_addr;
1274 	else {
1275 		prefix = target->ia_addr.sin_addr;
1276 		mask = target->ia_sockmask.sin_addr;
1277 		prefix.s_addr &= mask.s_addr;
1278 	}
1279 
1280 	s = pserialize_read_enter();
1281 	IN_ADDRLIST_READER_FOREACH(ia) {
1282 		if (rtinitflags(ia))
1283 			p = ia->ia_dstaddr.sin_addr;
1284 		else {
1285 			p = ia->ia_addr.sin_addr;
1286 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1287 		}
1288 
1289 		if (prefix.s_addr != p.s_addr)
1290 			continue;
1291 
1292 		/*
1293 		 * if we got a matching prefix route, move IFA_ROUTE to him
1294 		 */
1295 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1296 			struct psref psref;
1297 			int bound = curlwp_bind();
1298 
1299 			ia4_acquire(ia, &psref);
1300 			pserialize_read_exit(s);
1301 
1302 			rtinit(&target->ia_ifa, RTM_DELETE,
1303 			    rtinitflags(target));
1304 			target->ia_flags &= ~IFA_ROUTE;
1305 
1306 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1307 			    rtinitflags(ia) | RTF_UP);
1308 			if (error == 0)
1309 				ia->ia_flags |= IFA_ROUTE;
1310 
1311 			ia4_release(ia, &psref);
1312 			curlwp_bindx(bound);
1313 
1314 			return error;
1315 		}
1316 	}
1317 	pserialize_read_exit(s);
1318 
1319 	/*
1320 	 * noone seem to have prefix route.  remove it.
1321 	 */
1322 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1323 	target->ia_flags &= ~IFA_ROUTE;
1324 	return 0;
1325 }
1326 
1327 #undef rtinitflags
1328 
1329 /*
1330  * Return 1 if the address might be a local broadcast address.
1331  */
1332 int
1333 in_broadcast(struct in_addr in, struct ifnet *ifp)
1334 {
1335 	struct ifaddr *ifa;
1336 	int s;
1337 
1338 	KASSERT(ifp != NULL);
1339 
1340 	if (in.s_addr == INADDR_BROADCAST ||
1341 	    in_nullhost(in))
1342 		return 1;
1343 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1344 		return 0;
1345 	/*
1346 	 * Look through the list of addresses for a match
1347 	 * with a broadcast address.
1348 	 */
1349 #define ia (ifatoia(ifa))
1350 	s = pserialize_read_enter();
1351 	IFADDR_READER_FOREACH(ifa, ifp) {
1352 		if (ifa->ifa_addr->sa_family == AF_INET &&
1353 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1354 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1355 		     in_hosteq(in, ia->ia_netbroadcast) ||
1356 		     (hostzeroisbroadcast &&
1357 		      /*
1358 		       * Check for old-style (host 0) broadcast.
1359 		       */
1360 		      (in.s_addr == ia->ia_subnet ||
1361 		       in.s_addr == ia->ia_net)))) {
1362 			pserialize_read_exit(s);
1363 			return 1;
1364 		}
1365 	}
1366 	pserialize_read_exit(s);
1367 	return (0);
1368 #undef ia
1369 }
1370 
1371 /*
1372  * perform DAD when interface becomes IFF_UP.
1373  */
1374 void
1375 in_if_link_up(struct ifnet *ifp)
1376 {
1377 	struct ifaddr *ifa;
1378 	struct in_ifaddr *ia;
1379 	int s, bound;
1380 
1381 	/* Ensure it's sane to run DAD */
1382 	if (ifp->if_link_state == LINK_STATE_DOWN)
1383 		return;
1384 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1385 		return;
1386 
1387 	bound = curlwp_bind();
1388 	s = pserialize_read_enter();
1389 	IFADDR_READER_FOREACH(ifa, ifp) {
1390 		struct psref psref;
1391 
1392 		if (ifa->ifa_addr->sa_family != AF_INET)
1393 			continue;
1394 		ifa_acquire(ifa, &psref);
1395 		pserialize_read_exit(s);
1396 
1397 		ia = (struct in_ifaddr *)ifa;
1398 
1399 		/* If detached then mark as tentative */
1400 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1401 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1402 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1403 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1404 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1405 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1406 		}
1407 
1408 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1409 			/* Clear the duplicated flag as we're starting DAD. */
1410 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1411 			ia->ia_dad_start(ifa);
1412 		}
1413 
1414 		s = pserialize_read_enter();
1415 		ifa_release(ifa, &psref);
1416 	}
1417 	pserialize_read_exit(s);
1418 	curlwp_bindx(bound);
1419 }
1420 
1421 void
1422 in_if_up(struct ifnet *ifp)
1423 {
1424 
1425 	/* interface may not support link state, so bring it up also */
1426 	in_if_link_up(ifp);
1427 }
1428 
1429 /*
1430  * Mark all addresses as detached.
1431  */
1432 void
1433 in_if_link_down(struct ifnet *ifp)
1434 {
1435 	struct ifaddr *ifa;
1436 	struct in_ifaddr *ia;
1437 	int s, bound;
1438 
1439 	bound = curlwp_bind();
1440 	s = pserialize_read_enter();
1441 	IFADDR_READER_FOREACH(ifa, ifp) {
1442 		struct psref psref;
1443 
1444 		if (ifa->ifa_addr->sa_family != AF_INET)
1445 			continue;
1446 		ifa_acquire(ifa, &psref);
1447 		pserialize_read_exit(s);
1448 
1449 		ia = (struct in_ifaddr *)ifa;
1450 
1451 		/* Stop DAD processing */
1452 		if (ia->ia_dad_stop != NULL)
1453 			ia->ia_dad_stop(ifa);
1454 
1455 		/*
1456 		 * Mark the address as detached.
1457 		 */
1458 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1459 			ia->ia4_flags |= IN_IFF_DETACHED;
1460 			ia->ia4_flags &=
1461 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1462 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1463 		}
1464 
1465 		s = pserialize_read_enter();
1466 		ifa_release(ifa, &psref);
1467 	}
1468 	pserialize_read_exit(s);
1469 	curlwp_bindx(bound);
1470 }
1471 
1472 void
1473 in_if_down(struct ifnet *ifp)
1474 {
1475 
1476 	in_if_link_down(ifp);
1477 }
1478 
1479 void
1480 in_if_link_state_change(struct ifnet *ifp, int link_state)
1481 {
1482 
1483 	switch (link_state) {
1484 	case LINK_STATE_DOWN:
1485 		in_if_link_down(ifp);
1486 		break;
1487 	case LINK_STATE_UP:
1488 		in_if_link_up(ifp);
1489 		break;
1490 	}
1491 }
1492 
1493 /*
1494  * in_lookup_multi: look up the in_multi record for a given IP
1495  * multicast address on a given interface.  If no matching record is
1496  * found, return NULL.
1497  */
1498 struct in_multi *
1499 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1500 {
1501 	struct in_multi *inm;
1502 
1503 	KASSERT(rw_lock_held(&in_multilock));
1504 
1505 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1506 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1507 			break;
1508 	}
1509 	return inm;
1510 }
1511 
1512 /*
1513  * in_multi_group: check whether the address belongs to an IP multicast
1514  * group we are joined on this interface.  Returns true or false.
1515  */
1516 bool
1517 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1518 {
1519 	bool ingroup;
1520 
1521 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1522 		rw_enter(&in_multilock, RW_READER);
1523 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1524 		rw_exit(&in_multilock);
1525 	} else {
1526 		/* XXX Recursive call from ip_output(). */
1527 		KASSERT(rw_lock_held(&in_multilock));
1528 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1529 	}
1530 	return ingroup;
1531 }
1532 
1533 /*
1534  * Add an address to the list of IP multicast addresses for a given interface.
1535  */
1536 struct in_multi *
1537 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1538 {
1539 	struct sockaddr_in sin;
1540 	struct in_multi *inm;
1541 
1542 	/*
1543 	 * See if address already in list.
1544 	 */
1545 	rw_enter(&in_multilock, RW_WRITER);
1546 	inm = in_lookup_multi(*ap, ifp);
1547 	if (inm != NULL) {
1548 		/*
1549 		 * Found it; just increment the reference count.
1550 		 */
1551 		inm->inm_refcount++;
1552 		rw_exit(&in_multilock);
1553 		return inm;
1554 	}
1555 
1556 	/*
1557 	 * New address; allocate a new multicast record.
1558 	 */
1559 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1560 	if (inm == NULL) {
1561 		rw_exit(&in_multilock);
1562 		return NULL;
1563 	}
1564 	inm->inm_addr = *ap;
1565 	inm->inm_ifp = ifp;
1566 	inm->inm_refcount = 1;
1567 
1568 	/*
1569 	 * Ask the network driver to update its multicast reception
1570 	 * filter appropriately for the new address.
1571 	 */
1572 	sockaddr_in_init(&sin, ap, 0);
1573 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1574 		rw_exit(&in_multilock);
1575 		pool_put(&inmulti_pool, inm);
1576 		return NULL;
1577 	}
1578 
1579 	/*
1580 	 * Let IGMP know that we have joined a new IP multicast group.
1581 	 */
1582 	if (igmp_joingroup(inm) != 0) {
1583 		rw_exit(&in_multilock);
1584 		pool_put(&inmulti_pool, inm);
1585 		return NULL;
1586 	}
1587 	LIST_INSERT_HEAD(
1588 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1589 	    inm, inm_list);
1590 	in_multientries++;
1591 	rw_exit(&in_multilock);
1592 
1593 	return inm;
1594 }
1595 
1596 /*
1597  * Delete a multicast address record.
1598  */
1599 void
1600 in_delmulti(struct in_multi *inm)
1601 {
1602 	struct sockaddr_in sin;
1603 
1604 	rw_enter(&in_multilock, RW_WRITER);
1605 	if (--inm->inm_refcount > 0) {
1606 		rw_exit(&in_multilock);
1607 		return;
1608 	}
1609 
1610 	/*
1611 	 * No remaining claims to this record; let IGMP know that
1612 	 * we are leaving the multicast group.
1613 	 */
1614 	igmp_leavegroup(inm);
1615 
1616 	/*
1617 	 * Notify the network driver to update its multicast reception
1618 	 * filter.
1619 	 */
1620 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1621 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1622 
1623 	/*
1624 	 * Unlink from list.
1625 	 */
1626 	LIST_REMOVE(inm, inm_list);
1627 	in_multientries--;
1628 	rw_exit(&in_multilock);
1629 
1630 	pool_put(&inmulti_pool, inm);
1631 }
1632 
1633 /*
1634  * in_next_multi: step through all of the in_multi records, one at a time.
1635  * The current position is remembered in "step", which the caller must
1636  * provide.  in_first_multi(), below, must be called to initialize "step"
1637  * and get the first record.  Both macros return a NULL "inm" when there
1638  * are no remaining records.
1639  */
1640 struct in_multi *
1641 in_next_multi(struct in_multistep *step)
1642 {
1643 	struct in_multi *inm;
1644 
1645 	KASSERT(rw_lock_held(&in_multilock));
1646 
1647 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1648 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1649 	}
1650 	if ((inm = step->i_inm) != NULL) {
1651 		step->i_inm = LIST_NEXT(inm, inm_list);
1652 	}
1653 	return inm;
1654 }
1655 
1656 struct in_multi *
1657 in_first_multi(struct in_multistep *step)
1658 {
1659 	KASSERT(rw_lock_held(&in_multilock));
1660 
1661 	step->i_n = 0;
1662 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1663 	return in_next_multi(step);
1664 }
1665 
1666 void
1667 in_multi_lock(int op)
1668 {
1669 	rw_enter(&in_multilock, op);
1670 }
1671 
1672 void
1673 in_multi_unlock(void)
1674 {
1675 	rw_exit(&in_multilock);
1676 }
1677 
1678 int
1679 in_multi_lock_held(void)
1680 {
1681 	return rw_lock_held(&in_multilock);
1682 }
1683 
1684 struct in_ifaddr *
1685 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1686     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1687 {
1688 	struct rtentry *rt = NULL;
1689 	struct in_ifaddr *ia = NULL;
1690 
1691 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1692 	/*
1693          * If route is known or can be allocated now, take the
1694          * source address from the interface.  Otherwise, punt.
1695 	 */
1696 	if ((soopts & SO_DONTROUTE) != 0)
1697 		rtcache_free(ro);
1698 	else {
1699 		union {
1700 			struct sockaddr		dst;
1701 			struct sockaddr_in	dst4;
1702 		} u;
1703 
1704 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1705 		rt = rtcache_lookup(ro, &u.dst);
1706 	}
1707 	/*
1708 	 * If we found a route, use the address
1709 	 * corresponding to the outgoing interface
1710 	 * unless it is the loopback (in case a route
1711 	 * to our address on another net goes to loopback).
1712 	 *
1713 	 * XXX Is this still true?  Do we care?
1714 	 */
1715 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1716 		int s;
1717 		struct ifaddr *ifa;
1718 		/*
1719 		 * Just in case. May not need to do this workaround.
1720 		 * Revisit when working on rtentry MP-ification.
1721 		 */
1722 		s = pserialize_read_enter();
1723 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1724 			if (ifa == rt->rt_ifa)
1725 				break;
1726 		}
1727 		if (ifa != NULL)
1728 			ifa_acquire(ifa, psref);
1729 		pserialize_read_exit(s);
1730 
1731 		ia = ifatoia(ifa);
1732 	}
1733 	if (ia == NULL) {
1734 		u_int16_t fport = sin->sin_port;
1735 		struct ifaddr *ifa;
1736 		int s;
1737 
1738 		sin->sin_port = 0;
1739 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1740 		sin->sin_port = fport;
1741 		if (ifa == NULL) {
1742 			/* Find 1st non-loopback AF_INET address */
1743 			s = pserialize_read_enter();
1744 			IN_ADDRLIST_READER_FOREACH(ia) {
1745 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1746 					break;
1747 			}
1748 			if (ia != NULL)
1749 				ia4_acquire(ia, psref);
1750 			pserialize_read_exit(s);
1751 		} else {
1752 			/* ia is already referenced by psref */
1753 			ia = ifatoia(ifa);
1754 		}
1755 		if (ia == NULL) {
1756 			*errorp = EADDRNOTAVAIL;
1757 			return NULL;
1758 		}
1759 	}
1760 	/*
1761 	 * If the destination address is multicast and an outgoing
1762 	 * interface has been set as a multicast option, use the
1763 	 * address of that interface as our source address.
1764 	 */
1765 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1766 		struct ip_moptions *imo;
1767 
1768 		imo = mopts;
1769 		if (imo->imo_multicast_if_index != 0) {
1770 			struct ifnet *ifp;
1771 			int s;
1772 
1773 			if (ia != NULL)
1774 				ia4_release(ia, psref);
1775 			s = pserialize_read_enter();
1776 			ifp = if_byindex(imo->imo_multicast_if_index);
1777 			if (ifp != NULL) {
1778 				/* XXX */
1779 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1780 			} else
1781 				ia = NULL;
1782 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1783 				pserialize_read_exit(s);
1784 				if (ia != NULL)
1785 					ia4_release(ia, psref);
1786 				*errorp = EADDRNOTAVAIL;
1787 				return NULL;
1788 			}
1789 			pserialize_read_exit(s);
1790 		}
1791 	}
1792 	if (ia->ia_ifa.ifa_getifa != NULL) {
1793 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1794 		                                      sintosa(sin)));
1795 		if (ia == NULL) {
1796 			*errorp = EADDRNOTAVAIL;
1797 			return NULL;
1798 		}
1799 		/* FIXME NOMPSAFE */
1800 		ia4_acquire(ia, psref);
1801 	}
1802 #ifdef GETIFA_DEBUG
1803 	else
1804 		printf("%s: missing ifa_getifa\n", __func__);
1805 #endif
1806 	return ia;
1807 }
1808 
1809 #if NARP > 0
1810 
1811 struct in_llentry {
1812 	struct llentry		base;
1813 };
1814 
1815 #define	IN_LLTBL_DEFAULT_HSIZE	32
1816 #define	IN_LLTBL_HASH(k, h) \
1817 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1818 
1819 /*
1820  * Do actual deallocation of @lle.
1821  * Called by LLE_FREE_LOCKED when number of references
1822  * drops to zero.
1823  */
1824 static void
1825 in_lltable_destroy_lle(struct llentry *lle)
1826 {
1827 
1828 	LLE_WUNLOCK(lle);
1829 	LLE_LOCK_DESTROY(lle);
1830 	kmem_intr_free(lle, sizeof(*lle));
1831 }
1832 
1833 static struct llentry *
1834 in_lltable_new(struct in_addr addr4, u_int flags)
1835 {
1836 	struct in_llentry *lle;
1837 
1838 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
1839 	if (lle == NULL)		/* NB: caller generates msg */
1840 		return NULL;
1841 
1842 	/*
1843 	 * For IPv4 this will trigger "arpresolve" to generate
1844 	 * an ARP request.
1845 	 */
1846 	lle->base.la_expire = time_uptime; /* mark expired */
1847 	lle->base.r_l3addr.addr4 = addr4;
1848 	lle->base.lle_refcnt = 1;
1849 	lle->base.lle_free = in_lltable_destroy_lle;
1850 	LLE_LOCK_INIT(&lle->base);
1851 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1852 
1853 	return (&lle->base);
1854 }
1855 
1856 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1857 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1858 
1859 static int
1860 in_lltable_match_prefix(const struct sockaddr *prefix,
1861     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1862 {
1863 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1864 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1865 
1866 	/*
1867 	 * (flags & LLE_STATIC) means deleting all entries
1868 	 * including static ARP entries.
1869 	 */
1870 	if (IN_ARE_MASKED_ADDR_EQUAL(lle->r_l3addr.addr4, pfx, msk) &&
1871 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1872 		return (1);
1873 
1874 	return (0);
1875 }
1876 
1877 static void
1878 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1879 {
1880 	struct ifnet *ifp __diagused;
1881 	size_t pkts_dropped;
1882 
1883 	LLE_WLOCK_ASSERT(lle);
1884 	KASSERT(llt != NULL);
1885 
1886 	/* Unlink entry from table if not already */
1887 	if ((lle->la_flags & LLE_LINKED) != 0) {
1888 		ifp = llt->llt_ifp;
1889 		IF_AFDATA_WLOCK_ASSERT(ifp);
1890 		lltable_unlink_entry(llt, lle);
1891 	}
1892 
1893 	/* cancel timer */
1894 	if (callout_halt(&lle->lle_timer, &lle->lle_lock))
1895 		LLE_REMREF(lle);
1896 
1897 	/* Drop hold queue */
1898 	pkts_dropped = llentry_free(lle);
1899 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
1900 }
1901 
1902 static int
1903 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1904 {
1905 	struct rtentry *rt;
1906 	int error = EINVAL;
1907 
1908 	KASSERTMSG(l3addr->sa_family == AF_INET,
1909 	    "sin_family %d", l3addr->sa_family);
1910 
1911 	rt = rtalloc1(l3addr, 0);
1912 	if (rt == NULL)
1913 		return error;
1914 
1915 	/*
1916 	 * If the gateway for an existing host route matches the target L3
1917 	 * address, which is a special route inserted by some implementation
1918 	 * such as MANET, and the interface is of the correct type, then
1919 	 * allow for ARP to proceed.
1920 	 */
1921 	if (rt->rt_flags & RTF_GATEWAY) {
1922 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1923 		    rt->rt_ifp->if_type != IFT_ETHER ||
1924 #ifdef __FreeBSD__
1925 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1926 #else
1927 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
1928 #endif
1929 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1930 		    sizeof(in_addr_t)) != 0) {
1931 			goto error;
1932 		}
1933 	}
1934 
1935 	/*
1936 	 * Make sure that at least the destination address is covered
1937 	 * by the route. This is for handling the case where 2 or more
1938 	 * interfaces have the same prefix. An incoming packet arrives
1939 	 * on one interface and the corresponding outgoing packet leaves
1940 	 * another interface.
1941 	 */
1942 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1943 		const char *sa, *mask, *addr, *lim;
1944 		int len;
1945 
1946 		mask = (const char *)rt_mask(rt);
1947 		/*
1948 		 * Just being extra cautious to avoid some custom
1949 		 * code getting into trouble.
1950 		 */
1951 		if (mask == NULL)
1952 			goto error;
1953 
1954 		sa = (const char *)rt_getkey(rt);
1955 		addr = (const char *)l3addr;
1956 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
1957 		lim = addr + len;
1958 
1959 		for ( ; addr < lim; sa++, mask++, addr++) {
1960 			if ((*sa ^ *addr) & *mask) {
1961 #ifdef DIAGNOSTIC
1962 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1963 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1964 #endif
1965 				goto error;
1966 			}
1967 		}
1968 	}
1969 
1970 	error = 0;
1971 error:
1972 	rtfree(rt);
1973 	return error;
1974 }
1975 
1976 static inline uint32_t
1977 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1978 {
1979 
1980 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1981 }
1982 
1983 static uint32_t
1984 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1985 {
1986 
1987 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1988 }
1989 
1990 static void
1991 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1992 {
1993 	struct sockaddr_in *sin;
1994 
1995 	sin = (struct sockaddr_in *)sa;
1996 	memset(sin, 0, sizeof(*sin));
1997 	sin->sin_family = AF_INET;
1998 	sin->sin_len = sizeof(*sin);
1999 	sin->sin_addr = lle->r_l3addr.addr4;
2000 }
2001 
2002 static inline struct llentry *
2003 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
2004 {
2005 	struct llentry *lle;
2006 	struct llentries *lleh;
2007 	u_int hashidx;
2008 
2009 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
2010 	lleh = &llt->lle_head[hashidx];
2011 	LIST_FOREACH(lle, lleh, lle_next) {
2012 		if (lle->la_flags & LLE_DELETED)
2013 			continue;
2014 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
2015 			break;
2016 	}
2017 
2018 	return (lle);
2019 }
2020 
2021 static int
2022 in_lltable_delete(struct lltable *llt, u_int flags,
2023     const struct sockaddr *l3addr)
2024 {
2025 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2026 	struct ifnet *ifp __diagused = llt->llt_ifp;
2027 	struct llentry *lle;
2028 
2029 	IF_AFDATA_WLOCK_ASSERT(ifp);
2030 	KASSERTMSG(l3addr->sa_family == AF_INET,
2031 	    "sin_family %d", l3addr->sa_family);
2032 
2033 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2034 	if (lle == NULL) {
2035 #ifdef DIAGNOSTIC
2036 		log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
2037 #endif
2038 		return (ENOENT);
2039 	}
2040 
2041 	LLE_WLOCK(lle);
2042 	lle->la_flags |= LLE_DELETED;
2043 #ifdef DIAGNOSTIC
2044 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2045 #endif
2046 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2047 		llentry_free(lle);
2048 	else
2049 		LLE_WUNLOCK(lle);
2050 
2051 	return (0);
2052 }
2053 
2054 static struct llentry *
2055 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2056 {
2057 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2058 	struct ifnet *ifp = llt->llt_ifp;
2059 	struct llentry *lle;
2060 
2061 	IF_AFDATA_WLOCK_ASSERT(ifp);
2062 	KASSERTMSG(l3addr->sa_family == AF_INET,
2063 	    "sin_family %d", l3addr->sa_family);
2064 
2065 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2066 
2067 	if (lle != NULL) {
2068 		LLE_WLOCK(lle);
2069 		return (lle);
2070 	}
2071 
2072 	/* no existing record, we need to create new one */
2073 
2074 	/*
2075 	 * A route that covers the given address must have
2076 	 * been installed 1st because we are doing a resolution,
2077 	 * verify this.
2078 	 */
2079 	if (!(flags & LLE_IFADDR) &&
2080 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
2081 		return (NULL);
2082 
2083 	lle = in_lltable_new(sin->sin_addr, flags);
2084 	if (lle == NULL) {
2085 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2086 		return (NULL);
2087 	}
2088 	lle->la_flags = flags;
2089 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2090 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2091 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2092 	}
2093 
2094 	lltable_link_entry(llt, lle);
2095 	LLE_WLOCK(lle);
2096 
2097 	return (lle);
2098 }
2099 
2100 /*
2101  * Return NULL if not found or marked for deletion.
2102  * If found return lle read locked.
2103  */
2104 static struct llentry *
2105 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2106 {
2107 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2108 	struct llentry *lle;
2109 
2110 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2111 	KASSERTMSG(l3addr->sa_family == AF_INET,
2112 	    "sin_family %d", l3addr->sa_family);
2113 
2114 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2115 
2116 	if (lle == NULL)
2117 		return NULL;
2118 
2119 	if (flags & LLE_EXCLUSIVE)
2120 		LLE_WLOCK(lle);
2121 	else
2122 		LLE_RLOCK(lle);
2123 
2124 	return lle;
2125 }
2126 
2127 static int
2128 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2129     struct rt_walkarg *w)
2130 {
2131 	struct sockaddr_in sin;
2132 
2133 	LLTABLE_LOCK_ASSERT();
2134 
2135 	/* skip deleted entries */
2136 	if (lle->la_flags & LLE_DELETED)
2137 		return 0;
2138 
2139 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2140 
2141 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2142 }
2143 
2144 #endif /* NARP > 0 */
2145 
2146 static int
2147 in_multicast_sysctl(SYSCTLFN_ARGS)
2148 {
2149 	struct ifnet *ifp;
2150 	struct ifaddr *ifa;
2151 	struct in_ifaddr *ifa4;
2152 	struct in_multi *inm;
2153 	uint32_t tmp;
2154 	int error;
2155 	size_t written;
2156 	struct psref psref;
2157 	int bound;
2158 
2159 	if (namelen != 1)
2160 		return EINVAL;
2161 
2162 	bound = curlwp_bind();
2163 	ifp = if_get_byindex(name[0], &psref);
2164 	if (ifp == NULL) {
2165 		curlwp_bindx(bound);
2166 		return ENODEV;
2167 	}
2168 
2169 	if (oldp == NULL) {
2170 		*oldlenp = 0;
2171 		IFADDR_FOREACH(ifa, ifp) {
2172 			if (ifa->ifa_addr->sa_family != AF_INET)
2173 				continue;
2174 			ifa4 = (void *)ifa;
2175 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2176 				*oldlenp += 2 * sizeof(struct in_addr) +
2177 				    sizeof(uint32_t);
2178 			}
2179 		}
2180 		if_put(ifp, &psref);
2181 		curlwp_bindx(bound);
2182 		return 0;
2183 	}
2184 
2185 	error = 0;
2186 	written = 0;
2187 	IFADDR_FOREACH(ifa, ifp) {
2188 		if (ifa->ifa_addr->sa_family != AF_INET)
2189 			continue;
2190 		ifa4 = (void *)ifa;
2191 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2192 			if (written + 2 * sizeof(struct in_addr) +
2193 			    sizeof(uint32_t) > *oldlenp)
2194 				goto done;
2195 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2196 			    oldp, sizeof(struct in_addr));
2197 			if (error)
2198 				goto done;
2199 			oldp = (char *)oldp + sizeof(struct in_addr);
2200 			written += sizeof(struct in_addr);
2201 			error = sysctl_copyout(l, &inm->inm_addr,
2202 			    oldp, sizeof(struct in_addr));
2203 			if (error)
2204 				goto done;
2205 			oldp = (char *)oldp + sizeof(struct in_addr);
2206 			written += sizeof(struct in_addr);
2207 			tmp = inm->inm_refcount;
2208 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2209 			if (error)
2210 				goto done;
2211 			oldp = (char *)oldp + sizeof(tmp);
2212 			written += sizeof(tmp);
2213 		}
2214 	}
2215 done:
2216 	if_put(ifp, &psref);
2217 	curlwp_bindx(bound);
2218 	*oldlenp = written;
2219 	return error;
2220 }
2221 
2222 static void
2223 in_sysctl_init(struct sysctllog **clog)
2224 {
2225 	sysctl_createv(clog, 0, NULL, NULL,
2226 		       CTLFLAG_PERMANENT,
2227 		       CTLTYPE_NODE, "inet",
2228 		       SYSCTL_DESCR("PF_INET related settings"),
2229 		       NULL, 0, NULL, 0,
2230 		       CTL_NET, PF_INET, CTL_EOL);
2231 	sysctl_createv(clog, 0, NULL, NULL,
2232 		       CTLFLAG_PERMANENT,
2233 		       CTLTYPE_NODE, "multicast",
2234 		       SYSCTL_DESCR("Multicast information"),
2235 		       in_multicast_sysctl, 0, NULL, 0,
2236 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2237 	sysctl_createv(clog, 0, NULL, NULL,
2238 		       CTLFLAG_PERMANENT,
2239 		       CTLTYPE_NODE, "ip",
2240 		       SYSCTL_DESCR("IPv4 related settings"),
2241 		       NULL, 0, NULL, 0,
2242 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2243 
2244 	sysctl_createv(clog, 0, NULL, NULL,
2245 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2246 		       CTLTYPE_INT, "subnetsarelocal",
2247 		       SYSCTL_DESCR("Whether logical subnets are considered "
2248 				    "local"),
2249 		       NULL, 0, &subnetsarelocal, 0,
2250 		       CTL_NET, PF_INET, IPPROTO_IP,
2251 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2252 	sysctl_createv(clog, 0, NULL, NULL,
2253 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2254 		       CTLTYPE_INT, "hostzerobroadcast",
2255 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2256 		       NULL, 0, &hostzeroisbroadcast, 0,
2257 		       CTL_NET, PF_INET, IPPROTO_IP,
2258 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2259 }
2260 
2261 #if NARP > 0
2262 
2263 static struct lltable *
2264 in_lltattach(struct ifnet *ifp)
2265 {
2266 	struct lltable *llt;
2267 
2268 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2269 	llt->llt_af = AF_INET;
2270 	llt->llt_ifp = ifp;
2271 
2272 	llt->llt_lookup = in_lltable_lookup;
2273 	llt->llt_create = in_lltable_create;
2274 	llt->llt_delete = in_lltable_delete;
2275 	llt->llt_dump_entry = in_lltable_dump_entry;
2276 	llt->llt_hash = in_lltable_hash;
2277 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2278 	llt->llt_free_entry = in_lltable_free_entry;
2279 	llt->llt_match_prefix = in_lltable_match_prefix;
2280 	lltable_link(llt);
2281 
2282 	return (llt);
2283 }
2284 
2285 #endif /* NARP > 0 */
2286 
2287 void *
2288 in_domifattach(struct ifnet *ifp)
2289 {
2290 	struct in_ifinfo *ii;
2291 
2292 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2293 	KASSERT(ii != NULL);
2294 
2295 #if NARP > 0
2296 	ii->ii_llt = in_lltattach(ifp);
2297 #endif
2298 
2299 #ifdef IPSELSRC
2300 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2301 	KASSERT(ii->ii_selsrc != NULL);
2302 #endif
2303 
2304 	return ii;
2305 }
2306 
2307 void
2308 in_domifdetach(struct ifnet *ifp, void *aux)
2309 {
2310 	struct in_ifinfo *ii = aux;
2311 
2312 #ifdef IPSELSRC
2313 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2314 #endif
2315 #if NARP > 0
2316 	lltable_free(ii->ii_llt);
2317 #endif
2318 	kmem_free(ii, sizeof(struct in_ifinfo));
2319 }
2320