xref: /netbsd-src/sys/netinet/in.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: in.c,v 1.178 2016/08/01 03:15:30 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 1.178 2016/08/01 03:15:30 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 #endif
103 
104 #include <sys/param.h>
105 #include <sys/ioctl.h>
106 #include <sys/errno.h>
107 #include <sys/kernel.h>
108 #include <sys/malloc.h>
109 #include <sys/socket.h>
110 #include <sys/socketvar.h>
111 #include <sys/sysctl.h>
112 #include <sys/systm.h>
113 #include <sys/proc.h>
114 #include <sys/syslog.h>
115 #include <sys/kauth.h>
116 #include <sys/kmem.h>
117 
118 #include <sys/cprng.h>
119 
120 #include <net/if.h>
121 #include <net/route.h>
122 #include <net/pfil.h>
123 
124 #include <net/if_arp.h>
125 #include <net/if_ether.h>
126 #include <net/if_types.h>
127 #include <net/if_llatbl.h>
128 #include <net/if_dl.h>
129 
130 #include <netinet/in_systm.h>
131 #include <netinet/in.h>
132 #include <netinet/in_var.h>
133 #include <netinet/ip.h>
134 #include <netinet/ip_var.h>
135 #include <netinet/in_ifattach.h>
136 #include <netinet/in_pcb.h>
137 #include <netinet/in_selsrc.h>
138 #include <netinet/if_inarp.h>
139 #include <netinet/ip_mroute.h>
140 #include <netinet/igmp_var.h>
141 
142 #ifdef IPSELSRC
143 #include <netinet/in_selsrc.h>
144 #endif
145 
146 static u_int	in_mask2len(struct in_addr *);
147 static void	in_len2mask(struct in_addr *, u_int);
148 static int	in_lifaddr_ioctl(struct socket *, u_long, void *,
149 	struct ifnet *);
150 
151 static int	in_addprefix(struct in_ifaddr *, int);
152 static int	in_scrubprefix(struct in_ifaddr *);
153 static void	in_sysctl_init(struct sysctllog **);
154 
155 #ifndef SUBNETSARELOCAL
156 #define	SUBNETSARELOCAL	1
157 #endif
158 
159 #ifndef HOSTZEROBROADCAST
160 #define HOSTZEROBROADCAST 0
161 #endif
162 
163 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
164 #ifndef IN_MULTI_HASH_SIZE
165 #define IN_MULTI_HASH_SIZE	509
166 #endif
167 
168 static int			subnetsarelocal = SUBNETSARELOCAL;
169 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
170 
171 /*
172  * This list is used to keep track of in_multi chains which belong to
173  * deleted interface addresses.  We use in_ifaddr so that a chain head
174  * won't be deallocated until all multicast address record are deleted.
175  */
176 
177 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
178 
179 static struct pool		inmulti_pool;
180 static u_int			in_multientries;
181 static struct in_multihashhead *in_multihashtbl;
182 static u_long			in_multihash;
183 static krwlock_t		in_multilock;
184 
185 #define IN_MULTI_HASH(x, ifp) \
186     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
187 
188 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */
189 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
190 u_long				in_ifaddrhash;
191 struct in_ifaddrhead		in_ifaddrhead;
192 static kmutex_t			in_ifaddr_lock;
193 
194 struct pslist_head *		in_ifaddrhashtbl_pslist;
195 u_long				in_ifaddrhash_pslist;
196 struct pslist_head		in_ifaddrhead_pslist;
197 
198 void
199 in_init(void)
200 {
201 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
202 	    NULL, IPL_SOFTNET);
203 	TAILQ_INIT(&in_ifaddrhead);
204 	PSLIST_INIT(&in_ifaddrhead_pslist);
205 
206 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
207 	    &in_ifaddrhash);
208 
209 	in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST,
210 	    true, &in_ifaddrhash_pslist);
211 	mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
212 
213 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
214 	    &in_multihash);
215 	rw_init(&in_multilock);
216 
217 	in_sysctl_init(NULL);
218 }
219 
220 /*
221  * Return 1 if an internet address is for a ``local'' host
222  * (one to which we have a connection).  If subnetsarelocal
223  * is true, this includes other subnets of the local net.
224  * Otherwise, it includes only the directly-connected (sub)nets.
225  */
226 int
227 in_localaddr(struct in_addr in)
228 {
229 	struct in_ifaddr *ia;
230 	int localaddr = 0;
231 	int s = pserialize_read_enter();
232 
233 	if (subnetsarelocal) {
234 		IN_ADDRLIST_READER_FOREACH(ia) {
235 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net) {
236 				localaddr = 1;
237 				break;
238 			}
239 		}
240 	} else {
241 		IN_ADDRLIST_READER_FOREACH(ia) {
242 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) {
243 				localaddr = 1;
244 				break;
245 			}
246 		}
247 	}
248 	pserialize_read_exit(s);
249 
250 	return localaddr;
251 }
252 
253 /*
254  * Determine whether an IP address is in a reserved set of addresses
255  * that may not be forwarded, or whether datagrams to that destination
256  * may be forwarded.
257  */
258 int
259 in_canforward(struct in_addr in)
260 {
261 	u_int32_t net;
262 
263 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
264 		return (0);
265 	if (IN_CLASSA(in.s_addr)) {
266 		net = in.s_addr & IN_CLASSA_NET;
267 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
268 			return (0);
269 	}
270 	return (1);
271 }
272 
273 /*
274  * Trim a mask in a sockaddr
275  */
276 void
277 in_socktrim(struct sockaddr_in *ap)
278 {
279 	char *cplim = (char *) &ap->sin_addr;
280 	char *cp = (char *) (&ap->sin_addr + 1);
281 
282 	ap->sin_len = 0;
283 	while (--cp >= cplim)
284 		if (*cp) {
285 			(ap)->sin_len = cp - (char *) (ap) + 1;
286 			break;
287 		}
288 }
289 
290 /*
291  *  Routine to take an Internet address and convert into a
292  *  "dotted quad" representation for printing.
293  */
294 const char *
295 in_fmtaddr(struct in_addr addr)
296 {
297 	static char buf[sizeof("123.456.789.123")];
298 
299 	addr.s_addr = ntohl(addr.s_addr);
300 
301 	snprintf(buf, sizeof(buf), "%d.%d.%d.%d",
302 		(addr.s_addr >> 24) & 0xFF,
303 		(addr.s_addr >> 16) & 0xFF,
304 		(addr.s_addr >>  8) & 0xFF,
305 		(addr.s_addr >>  0) & 0xFF);
306 	return buf;
307 }
308 
309 /*
310  * Maintain the "in_maxmtu" variable, which is the largest
311  * mtu for non-local interfaces with AF_INET addresses assigned
312  * to them that are up.
313  */
314 unsigned long in_maxmtu;
315 
316 void
317 in_setmaxmtu(void)
318 {
319 	struct in_ifaddr *ia;
320 	struct ifnet *ifp;
321 	unsigned long maxmtu = 0;
322 	int s = pserialize_read_enter();
323 
324 	IN_ADDRLIST_READER_FOREACH(ia) {
325 		if ((ifp = ia->ia_ifp) == 0)
326 			continue;
327 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
328 			continue;
329 		if (ifp->if_mtu > maxmtu)
330 			maxmtu = ifp->if_mtu;
331 	}
332 	if (maxmtu)
333 		in_maxmtu = maxmtu;
334 	pserialize_read_exit(s);
335 }
336 
337 static u_int
338 in_mask2len(struct in_addr *mask)
339 {
340 	u_int x, y;
341 	u_char *p;
342 
343 	p = (u_char *)mask;
344 	for (x = 0; x < sizeof(*mask); x++) {
345 		if (p[x] != 0xff)
346 			break;
347 	}
348 	y = 0;
349 	if (x < sizeof(*mask)) {
350 		for (y = 0; y < NBBY; y++) {
351 			if ((p[x] & (0x80 >> y)) == 0)
352 				break;
353 		}
354 	}
355 	return x * NBBY + y;
356 }
357 
358 static void
359 in_len2mask(struct in_addr *mask, u_int len)
360 {
361 	u_int i;
362 	u_char *p;
363 
364 	p = (u_char *)mask;
365 	memset(mask, 0, sizeof(*mask));
366 	for (i = 0; i < len / NBBY; i++)
367 		p[i] = 0xff;
368 	if (len % NBBY)
369 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
370 }
371 
372 /*
373  * Generic internet control operations (ioctl's).
374  * Ifp is 0 if not an interface-specific ioctl.
375  */
376 /* ARGSUSED */
377 static int
378 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
379 {
380 	struct ifreq *ifr = (struct ifreq *)data;
381 	struct in_ifaddr *ia = NULL;
382 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
383 	struct sockaddr_in oldaddr;
384 	int error, hostIsNew, maskIsNew;
385 	int newifaddr = 0;
386 	bool run_hook = false;
387 	bool need_reinsert = false;
388 	struct psref psref;
389 	int bound;
390 
391 	switch (cmd) {
392 	case SIOCALIFADDR:
393 	case SIOCDLIFADDR:
394 	case SIOCGLIFADDR:
395 		if (ifp == NULL)
396 			return EINVAL;
397 		return in_lifaddr_ioctl(so, cmd, data, ifp);
398 	case SIOCGIFADDRPREF:
399 	case SIOCSIFADDRPREF:
400 		if (ifp == NULL)
401 			return EINVAL;
402 		return ifaddrpref_ioctl(so, cmd, data, ifp);
403 	}
404 
405 	bound = curlwp_bind();
406 	/*
407 	 * Find address for this interface, if it exists.
408 	 */
409 	if (ifp != NULL)
410 		ia = in_get_ia_from_ifp_psref(ifp, &psref);
411 
412 	hostIsNew = 1;		/* moved here to appease gcc */
413 	switch (cmd) {
414 	case SIOCAIFADDR:
415 	case SIOCDIFADDR:
416 	case SIOCGIFALIAS:
417 	case SIOCGIFAFLAG_IN:
418 		if (ifra->ifra_addr.sin_family == AF_INET) {
419 			int s;
420 
421 			if (ia != NULL)
422 				ia4_release(ia, &psref);
423 			s = pserialize_read_enter();
424 			IN_ADDRHASH_READER_FOREACH(ia,
425 			    ifra->ifra_addr.sin_addr.s_addr) {
426 				if (ia->ia_ifp == ifp &&
427 				    in_hosteq(ia->ia_addr.sin_addr,
428 				    ifra->ifra_addr.sin_addr))
429 					break;
430 			}
431 			if (ia != NULL)
432 				ia4_acquire(ia, &psref);
433 			pserialize_read_exit(s);
434 		}
435 		if ((cmd == SIOCDIFADDR ||
436 		    cmd == SIOCGIFALIAS ||
437 		    cmd == SIOCGIFAFLAG_IN) &&
438 		    ia == NULL) {
439 			error = EADDRNOTAVAIL;
440 			goto out;
441 		}
442 
443 		if (cmd == SIOCDIFADDR &&
444 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
445 			ifra->ifra_addr.sin_family = AF_INET;
446 		}
447 		/* FALLTHROUGH */
448 	case SIOCSIFADDR:
449 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
450 			;
451 		else if (ifra->ifra_addr.sin_len == 0) {
452 			ifra->ifra_addr = ia->ia_addr;
453 			hostIsNew = 0;
454 		} else if (in_hosteq(ia->ia_addr.sin_addr,
455 		           ifra->ifra_addr.sin_addr))
456 			hostIsNew = 0;
457 		/* FALLTHROUGH */
458 	case SIOCSIFDSTADDR:
459 		if (ifra->ifra_addr.sin_family != AF_INET) {
460 			error = EAFNOSUPPORT;
461 			goto out;
462 		}
463 		/* FALLTHROUGH */
464 	case SIOCSIFNETMASK:
465 		if (ifp == NULL)
466 			panic("in_control");
467 
468 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
469 			break;
470 
471 		if (ia == NULL &&
472 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) {
473 			error = EADDRNOTAVAIL;
474 			goto out;
475 		}
476 
477 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
478 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
479 		    NULL) != 0) {
480 			error = EPERM;
481 			goto out;
482 		}
483 
484 		if (ia == NULL) {
485 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
486 			if (ia == NULL) {
487 				error = ENOBUFS;
488 				goto out;
489 			}
490 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
491 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
492 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
493 #ifdef IPSELSRC
494 			ia->ia_ifa.ifa_getifa = in_getifa;
495 #else /* IPSELSRC */
496 			ia->ia_ifa.ifa_getifa = NULL;
497 #endif /* IPSELSRC */
498 			ia->ia_sockmask.sin_len = 8;
499 			ia->ia_sockmask.sin_family = AF_INET;
500 			if (ifp->if_flags & IFF_BROADCAST) {
501 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
502 				ia->ia_broadaddr.sin_family = AF_INET;
503 			}
504 			ia->ia_ifp = ifp;
505 			ia->ia_idsalt = cprng_fast32() % 65535;
506 			LIST_INIT(&ia->ia_multiaddrs);
507 			IN_ADDRHASH_ENTRY_INIT(ia);
508 			IN_ADDRLIST_ENTRY_INIT(ia);
509 			ifa_psref_init(&ia->ia_ifa);
510 
511 			newifaddr = 1;
512 		}
513 		break;
514 
515 	case SIOCSIFBRDADDR:
516 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
517 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
518 		    NULL) != 0) {
519 			error = EPERM;
520 			goto out;
521 		}
522 		/* FALLTHROUGH */
523 
524 	case SIOCGIFADDR:
525 	case SIOCGIFNETMASK:
526 	case SIOCGIFDSTADDR:
527 	case SIOCGIFBRDADDR:
528 		if (ia == NULL) {
529 			error = EADDRNOTAVAIL;
530 			goto out;
531 		}
532 		break;
533 	}
534 	error = 0;
535 	switch (cmd) {
536 
537 	case SIOCGIFADDR:
538 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
539 		break;
540 
541 	case SIOCGIFBRDADDR:
542 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
543 			error = EINVAL;
544 			goto out;
545 		}
546 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
547 		break;
548 
549 	case SIOCGIFDSTADDR:
550 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
551 			error = EINVAL;
552 			goto out;
553 		}
554 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
555 		break;
556 
557 	case SIOCGIFNETMASK:
558 		/*
559 		 * We keep the number of trailing zero bytes the sin_len field
560 		 * of ia_sockmask, so we fix this before we pass it back to
561 		 * userland.
562 		 */
563 		oldaddr = ia->ia_sockmask;
564 		oldaddr.sin_len = sizeof(struct sockaddr_in);
565 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
566 		break;
567 
568 	case SIOCSIFDSTADDR:
569 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
570 			error = EINVAL;
571 			goto out;
572 		}
573 		oldaddr = ia->ia_dstaddr;
574 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
575 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
576 			ia->ia_dstaddr = oldaddr;
577 			goto out;
578 		}
579 		if (ia->ia_flags & IFA_ROUTE) {
580 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
581 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
582 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
583 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
584 		}
585 		break;
586 
587 	case SIOCSIFBRDADDR:
588 		if ((ifp->if_flags & IFF_BROADCAST) == 0) {
589 			error = EINVAL;
590 			goto out;
591 		}
592 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
593 		break;
594 
595 	case SIOCSIFADDR:
596 		if (!newifaddr) {
597 			mutex_enter(&in_ifaddr_lock);
598 			LIST_REMOVE(ia, ia_hash);
599 			IN_ADDRHASH_WRITER_REMOVE(ia);
600 			mutex_exit(&in_ifaddr_lock);
601 			need_reinsert = true;
602 		}
603 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
604 		    1, hostIsNew);
605 
606 		run_hook = true;
607 		break;
608 
609 	case SIOCSIFNETMASK:
610 		in_ifscrub(ifp, ia);
611 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
612 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
613 		if (!newifaddr) {
614 			mutex_enter(&in_ifaddr_lock);
615 			LIST_REMOVE(ia, ia_hash);
616 			IN_ADDRHASH_WRITER_REMOVE(ia);
617 			mutex_exit(&in_ifaddr_lock);
618 			need_reinsert = true;
619 		}
620 		error = in_ifinit(ifp, ia, NULL, 0, 0);
621 		break;
622 
623 	case SIOCAIFADDR:
624 		maskIsNew = 0;
625 		if (ifra->ifra_mask.sin_len) {
626 			/* Only scrub if we control the prefix route,
627 			 * otherwise userland gets a bogus message */
628 			if ((ia->ia_flags & IFA_ROUTE))
629 				in_ifscrub(ifp, ia);
630 			ia->ia_sockmask = ifra->ifra_mask;
631 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
632 			maskIsNew = 1;
633 		}
634 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
635 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
636 			/* Only scrub if we control the prefix route,
637 			 * otherwise userland gets a bogus message */
638 			if ((ia->ia_flags & IFA_ROUTE))
639 				in_ifscrub(ifp, ia);
640 			ia->ia_dstaddr = ifra->ifra_dstaddr;
641 			maskIsNew  = 1; /* We lie; but the effect's the same */
642 		}
643 		if (ifra->ifra_addr.sin_family == AF_INET &&
644 		    (hostIsNew || maskIsNew)) {
645 			if (!newifaddr) {
646 				mutex_enter(&in_ifaddr_lock);
647 				LIST_REMOVE(ia, ia_hash);
648 				IN_ADDRHASH_WRITER_REMOVE(ia);
649 				mutex_exit(&in_ifaddr_lock);
650 				need_reinsert = true;
651 			}
652 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
653 			    hostIsNew);
654 		}
655 		if ((ifp->if_flags & IFF_BROADCAST) &&
656 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
657 			ia->ia_broadaddr = ifra->ifra_broadaddr;
658 		run_hook = true;
659 		break;
660 
661 	case SIOCGIFALIAS:
662 		ifra->ifra_mask = ia->ia_sockmask;
663 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
664 		    (ia->ia_dstaddr.sin_family == AF_INET))
665 			ifra->ifra_dstaddr = ia->ia_dstaddr;
666 		else if ((ifp->if_flags & IFF_BROADCAST) &&
667 		    (ia->ia_broadaddr.sin_family == AF_INET))
668 			ifra->ifra_broadaddr = ia->ia_broadaddr;
669 		else
670 			memset(&ifra->ifra_broadaddr, 0,
671 			      sizeof(ifra->ifra_broadaddr));
672 		break;
673 
674 	case SIOCGIFAFLAG_IN:
675 		ifr->ifr_addrflags = ia->ia4_flags;
676 		break;
677 
678 	case SIOCDIFADDR:
679 		ia4_release(ia, &psref);
680 		in_purgeaddr(&ia->ia_ifa);
681 		ia = NULL;
682 		run_hook = true;
683 		break;
684 
685 #ifdef MROUTING
686 	case SIOCGETVIFCNT:
687 	case SIOCGETSGCNT:
688 		error = mrt_ioctl(so, cmd, data);
689 		break;
690 #endif /* MROUTING */
691 
692 	default:
693 		error = ENOTTY;
694 		goto out;
695 	}
696 
697 	/*
698 	 * XXX insert regardless of error to make in_purgeaddr below work.
699 	 * Need to improve.
700 	 */
701 	if (newifaddr) {
702 		ifaref(&ia->ia_ifa);
703 		ifa_insert(ifp, &ia->ia_ifa);
704 
705 		mutex_enter(&in_ifaddr_lock);
706 		TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
707 		IN_ADDRLIST_WRITER_INSERT_TAIL(ia);
708 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
709 		    ia, ia_hash);
710 		IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
711 		mutex_exit(&in_ifaddr_lock);
712 	} else if (need_reinsert) {
713 		mutex_enter(&in_ifaddr_lock);
714 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
715 		    ia, ia_hash);
716 		IN_ADDRHASH_WRITER_INSERT_HEAD(ia);
717 		mutex_exit(&in_ifaddr_lock);
718 	}
719 
720 	if (error == 0) {
721 		if (run_hook)
722 			(void)pfil_run_hooks(if_pfil,
723 			    (struct mbuf **)cmd, ifp, PFIL_IFADDR);
724 	} else if (newifaddr) {
725 		KASSERT(ia != NULL);
726 		in_purgeaddr(&ia->ia_ifa);
727 		ia = NULL;
728 	}
729 
730 out:
731 	if (!newifaddr && ia != NULL)
732 		ia4_release(ia, &psref);
733 	curlwp_bindx(bound);
734 	return error;
735 }
736 
737 int
738 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
739 {
740 	int error;
741 
742 	mutex_enter(softnet_lock);
743 	error = in_control0(so, cmd, data, ifp);
744 	mutex_exit(softnet_lock);
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 /*
805  * Depends on it isn't called in concurrent. It should be guaranteed
806  * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control
807  * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl
808  * and the latter is called via ifa->ifa_ifp's if_detach. The functions
809  * never be executed in concurrent.
810  */
811 void
812 in_purgeaddr(struct ifaddr *ifa)
813 {
814 	struct ifnet *ifp = ifa->ifa_ifp;
815 	struct in_ifaddr *ia = (void *) ifa;
816 
817 	KASSERT(!ifa_held(ifa));
818 
819 	/* stop DAD processing */
820 	if (ia->ia_dad_stop != NULL)
821 		ia->ia_dad_stop(ifa);
822 
823 	in_ifscrub(ifp, ia);
824 	in_ifremlocal(ifa);
825 	if (ia->ia_allhosts != NULL)
826 		in_delmulti(ia->ia_allhosts);
827 
828 	mutex_enter(&in_ifaddr_lock);
829 	LIST_REMOVE(ia, ia_hash);
830 	IN_ADDRHASH_WRITER_REMOVE(ia);
831 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
832 	IN_ADDRLIST_WRITER_REMOVE(ia);
833 	ifa_remove(ifp, &ia->ia_ifa);
834 	mutex_exit(&in_ifaddr_lock);
835 
836 	IN_ADDRHASH_ENTRY_DESTROY(ia);
837 	IN_ADDRLIST_ENTRY_DESTROY(ia);
838 	ifafree(&ia->ia_ifa);
839 	in_setmaxmtu();
840 }
841 
842 void
843 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
844 {
845 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
846 	igmp_purgeif(ifp);		/* manipulates pools */
847 #ifdef MROUTING
848 	ip_mrouter_detach(ifp);
849 #endif
850 }
851 
852 /*
853  * SIOC[GAD]LIFADDR.
854  *	SIOCGLIFADDR: get first address. (???)
855  *	SIOCGLIFADDR with IFLR_PREFIX:
856  *		get first address that matches the specified prefix.
857  *	SIOCALIFADDR: add the specified address.
858  *	SIOCALIFADDR with IFLR_PREFIX:
859  *		EINVAL since we can't deduce hostid part of the address.
860  *	SIOCDLIFADDR: delete the specified address.
861  *	SIOCDLIFADDR with IFLR_PREFIX:
862  *		delete the first address that matches the specified prefix.
863  * return values:
864  *	EINVAL on invalid parameters
865  *	EADDRNOTAVAIL on prefix match failed/specified address not found
866  *	other values may be returned from in_ioctl()
867  */
868 static int
869 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
870     struct ifnet *ifp)
871 {
872 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
873 	struct ifaddr *ifa;
874 	struct sockaddr *sa;
875 
876 	/* sanity checks */
877 	if (data == NULL || ifp == NULL) {
878 		panic("invalid argument to in_lifaddr_ioctl");
879 		/*NOTRECHED*/
880 	}
881 
882 	switch (cmd) {
883 	case SIOCGLIFADDR:
884 		/* address must be specified on GET with IFLR_PREFIX */
885 		if ((iflr->flags & IFLR_PREFIX) == 0)
886 			break;
887 		/*FALLTHROUGH*/
888 	case SIOCALIFADDR:
889 	case SIOCDLIFADDR:
890 		/* address must be specified on ADD and DELETE */
891 		sa = (struct sockaddr *)&iflr->addr;
892 		if (sa->sa_family != AF_INET)
893 			return EINVAL;
894 		if (sa->sa_len != sizeof(struct sockaddr_in))
895 			return EINVAL;
896 		/* XXX need improvement */
897 		sa = (struct sockaddr *)&iflr->dstaddr;
898 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
899 			return EINVAL;
900 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
901 			return EINVAL;
902 		break;
903 	default: /*shouldn't happen*/
904 #if 0
905 		panic("invalid cmd to in_lifaddr_ioctl");
906 		/*NOTREACHED*/
907 #else
908 		return EOPNOTSUPP;
909 #endif
910 	}
911 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
912 		return EINVAL;
913 
914 	switch (cmd) {
915 	case SIOCALIFADDR:
916 	    {
917 		struct in_aliasreq ifra;
918 
919 		if (iflr->flags & IFLR_PREFIX)
920 			return EINVAL;
921 
922 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
923 		memset(&ifra, 0, sizeof(ifra));
924 		memcpy(ifra.ifra_name, iflr->iflr_name,
925 			sizeof(ifra.ifra_name));
926 
927 		memcpy(&ifra.ifra_addr, &iflr->addr,
928 			((struct sockaddr *)&iflr->addr)->sa_len);
929 
930 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
931 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
932 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
933 		}
934 
935 		ifra.ifra_mask.sin_family = AF_INET;
936 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
937 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
938 
939 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
940 	    }
941 	case SIOCGLIFADDR:
942 	case SIOCDLIFADDR:
943 	    {
944 		struct in_ifaddr *ia;
945 		struct in_addr mask, candidate, match;
946 		struct sockaddr_in *sin;
947 		int cmp;
948 
949 		memset(&mask, 0, sizeof(mask));
950 		memset(&match, 0, sizeof(match));	/* XXX gcc */
951 		if (iflr->flags & IFLR_PREFIX) {
952 			/* lookup a prefix rather than address. */
953 			in_len2mask(&mask, iflr->prefixlen);
954 
955 			sin = (struct sockaddr_in *)&iflr->addr;
956 			match.s_addr = sin->sin_addr.s_addr;
957 			match.s_addr &= mask.s_addr;
958 
959 			/* if you set extra bits, that's wrong */
960 			if (match.s_addr != sin->sin_addr.s_addr)
961 				return EINVAL;
962 
963 			cmp = 1;
964 		} else {
965 			if (cmd == SIOCGLIFADDR) {
966 				/* on getting an address, take the 1st match */
967 				cmp = 0;	/*XXX*/
968 			} else {
969 				/* on deleting an address, do exact match */
970 				in_len2mask(&mask, 32);
971 				sin = (struct sockaddr_in *)&iflr->addr;
972 				match.s_addr = sin->sin_addr.s_addr;
973 
974 				cmp = 1;
975 			}
976 		}
977 
978 		IFADDR_READER_FOREACH(ifa, ifp) {
979 			if (ifa->ifa_addr->sa_family != AF_INET)
980 				continue;
981 			if (cmp == 0)
982 				break;
983 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
984 			candidate.s_addr &= mask.s_addr;
985 			if (candidate.s_addr == match.s_addr)
986 				break;
987 		}
988 		if (ifa == NULL)
989 			return EADDRNOTAVAIL;
990 		ia = (struct in_ifaddr *)ifa;
991 
992 		if (cmd == SIOCGLIFADDR) {
993 			/* fill in the if_laddrreq structure */
994 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
995 
996 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
997 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
998 					ia->ia_dstaddr.sin_len);
999 			} else
1000 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1001 
1002 			iflr->prefixlen =
1003 				in_mask2len(&ia->ia_sockmask.sin_addr);
1004 
1005 			iflr->flags = 0;	/*XXX*/
1006 
1007 			return 0;
1008 		} else {
1009 			struct in_aliasreq ifra;
1010 
1011 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
1012 			memset(&ifra, 0, sizeof(ifra));
1013 			memcpy(ifra.ifra_name, iflr->iflr_name,
1014 				sizeof(ifra.ifra_name));
1015 
1016 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1017 				ia->ia_addr.sin_len);
1018 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1019 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1020 					ia->ia_dstaddr.sin_len);
1021 			}
1022 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
1023 				ia->ia_sockmask.sin_len);
1024 
1025 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
1026 		}
1027 	    }
1028 	}
1029 
1030 	return EOPNOTSUPP;	/*just for safety*/
1031 }
1032 
1033 /*
1034  * Delete any existing route for an interface.
1035  */
1036 void
1037 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
1038 {
1039 
1040 	in_scrubprefix(ia);
1041 }
1042 
1043 /*
1044  * Initialize an interface's internet address
1045  * and routing table entry.
1046  */
1047 int
1048 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
1049     const struct sockaddr_in *sin, int scrub, int hostIsNew)
1050 {
1051 	u_int32_t i;
1052 	struct sockaddr_in oldaddr;
1053 	int s = splnet(), flags = RTF_UP, error;
1054 
1055 	if (sin == NULL)
1056 		sin = &ia->ia_addr;
1057 
1058 	/*
1059 	 * Set up new addresses.
1060 	 */
1061 	oldaddr = ia->ia_addr;
1062 	ia->ia_addr = *sin;
1063 
1064 	/* Set IN_IFF flags early for if_addr_init() */
1065 	if (hostIsNew && if_do_dad(ifp) && !in_nullhost(ia->ia_addr.sin_addr)) {
1066 		if (ifp->if_link_state == LINK_STATE_DOWN)
1067 			ia->ia4_flags |= IN_IFF_DETACHED;
1068 		else
1069 			/* State the intent to try DAD if possible */
1070 			ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
1071 	}
1072 
1073 	/*
1074 	 * Give the interface a chance to initialize
1075 	 * if this is its first address,
1076 	 * and to validate the address if necessary.
1077 	 */
1078 	if ((error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0)
1079 		goto bad;
1080 	/* Now clear the try tentative flag, it's job is done. */
1081 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
1082 	splx(s);
1083 
1084 	if (scrub) {
1085 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
1086 		in_ifscrub(ifp, ia);
1087 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
1088 	}
1089 
1090 	/* Add the local route to the address */
1091 	in_ifaddlocal(&ia->ia_ifa);
1092 
1093 	i = ia->ia_addr.sin_addr.s_addr;
1094 	if (IN_CLASSA(i))
1095 		ia->ia_netmask = IN_CLASSA_NET;
1096 	else if (IN_CLASSB(i))
1097 		ia->ia_netmask = IN_CLASSB_NET;
1098 	else
1099 		ia->ia_netmask = IN_CLASSC_NET;
1100 	/*
1101 	 * The subnet mask usually includes at least the standard network part,
1102 	 * but may may be smaller in the case of supernetting.
1103 	 * If it is set, we believe it.
1104 	 */
1105 	if (ia->ia_subnetmask == 0) {
1106 		ia->ia_subnetmask = ia->ia_netmask;
1107 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
1108 	} else
1109 		ia->ia_netmask &= ia->ia_subnetmask;
1110 
1111 	ia->ia_net = i & ia->ia_netmask;
1112 	ia->ia_subnet = i & ia->ia_subnetmask;
1113 	in_socktrim(&ia->ia_sockmask);
1114 	/* re-calculate the "in_maxmtu" value */
1115 	in_setmaxmtu();
1116 	/*
1117 	 * Add route for the network.
1118 	 */
1119 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1120 	if (ifp->if_flags & IFF_BROADCAST) {
1121 		ia->ia_broadaddr.sin_addr.s_addr =
1122 			ia->ia_subnet | ~ia->ia_subnetmask;
1123 		ia->ia_netbroadcast.s_addr =
1124 			ia->ia_net | ~ia->ia_netmask;
1125 	} else if (ifp->if_flags & IFF_LOOPBACK) {
1126 		ia->ia_dstaddr = ia->ia_addr;
1127 		flags |= RTF_HOST;
1128 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
1129 		if (ia->ia_dstaddr.sin_family != AF_INET)
1130 			return (0);
1131 		flags |= RTF_HOST;
1132 	}
1133 	error = in_addprefix(ia, flags);
1134 	/*
1135 	 * If the interface supports multicast, join the "all hosts"
1136 	 * multicast group on that interface.
1137 	 */
1138 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
1139 		struct in_addr addr;
1140 
1141 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
1142 		ia->ia_allhosts = in_addmulti(&addr, ifp);
1143 	}
1144 
1145 	if (hostIsNew && if_do_dad(ifp) &&
1146 	    !in_nullhost(ia->ia_addr.sin_addr) &&
1147 	    ia->ia4_flags & IN_IFF_TENTATIVE)
1148 		ia->ia_dad_start((struct ifaddr *)ia);
1149 
1150 	return (error);
1151 bad:
1152 	splx(s);
1153 	ia->ia_addr = oldaddr;
1154 	return (error);
1155 }
1156 
1157 #define rtinitflags(x) \
1158 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
1159 	    ? RTF_HOST : 0)
1160 
1161 /*
1162  * add a route to prefix ("connected route" in cisco terminology).
1163  * does nothing if there's some interface address with the same prefix already.
1164  */
1165 static int
1166 in_addprefix(struct in_ifaddr *target, int flags)
1167 {
1168 	struct in_ifaddr *ia;
1169 	struct in_addr prefix, mask, p;
1170 	int error;
1171 	int s;
1172 
1173 	if ((flags & RTF_HOST) != 0)
1174 		prefix = target->ia_dstaddr.sin_addr;
1175 	else {
1176 		prefix = target->ia_addr.sin_addr;
1177 		mask = target->ia_sockmask.sin_addr;
1178 		prefix.s_addr &= mask.s_addr;
1179 	}
1180 
1181 	s = pserialize_read_enter();
1182 	IN_ADDRLIST_READER_FOREACH(ia) {
1183 		if (rtinitflags(ia))
1184 			p = ia->ia_dstaddr.sin_addr;
1185 		else {
1186 			p = ia->ia_addr.sin_addr;
1187 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1188 		}
1189 
1190 		if (prefix.s_addr != p.s_addr)
1191 			continue;
1192 
1193 		/*
1194 		 * if we got a matching prefix route inserted by other
1195 		 * interface address, we don't need to bother
1196 		 *
1197 		 * XXX RADIX_MPATH implications here? -dyoung
1198 		 */
1199 		if (ia->ia_flags & IFA_ROUTE) {
1200 			pserialize_read_exit(s);
1201 			return 0;
1202 		}
1203 	}
1204 	pserialize_read_exit(s);
1205 
1206 	/*
1207 	 * noone seem to have prefix route.  insert it.
1208 	 */
1209 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
1210 	if (error == 0)
1211 		target->ia_flags |= IFA_ROUTE;
1212 	else if (error == EEXIST) {
1213 		/*
1214 		 * the fact the route already exists is not an error.
1215 		 */
1216 		error = 0;
1217 	}
1218 	return error;
1219 }
1220 
1221 /*
1222  * remove a route to prefix ("connected route" in cisco terminology).
1223  * re-installs the route by using another interface address, if there's one
1224  * with the same prefix (otherwise we lose the route mistakenly).
1225  */
1226 static int
1227 in_scrubprefix(struct in_ifaddr *target)
1228 {
1229 	struct in_ifaddr *ia;
1230 	struct in_addr prefix, mask, p;
1231 	int error;
1232 	int s;
1233 
1234 	/* If we don't have IFA_ROUTE we should still inform userland */
1235 	if ((target->ia_flags & IFA_ROUTE) == 0)
1236 		return 0;
1237 
1238 	if (rtinitflags(target))
1239 		prefix = target->ia_dstaddr.sin_addr;
1240 	else {
1241 		prefix = target->ia_addr.sin_addr;
1242 		mask = target->ia_sockmask.sin_addr;
1243 		prefix.s_addr &= mask.s_addr;
1244 	}
1245 
1246 	s = pserialize_read_enter();
1247 	IN_ADDRLIST_READER_FOREACH(ia) {
1248 		if (rtinitflags(ia))
1249 			p = ia->ia_dstaddr.sin_addr;
1250 		else {
1251 			p = ia->ia_addr.sin_addr;
1252 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
1253 		}
1254 
1255 		if (prefix.s_addr != p.s_addr)
1256 			continue;
1257 
1258 		/*
1259 		 * if we got a matching prefix route, move IFA_ROUTE to him
1260 		 */
1261 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
1262 			struct psref psref;
1263 			int bound = curlwp_bind();
1264 
1265 			ia4_acquire(ia, &psref);
1266 			pserialize_read_exit(s);
1267 
1268 			rtinit(&target->ia_ifa, RTM_DELETE,
1269 			    rtinitflags(target));
1270 			target->ia_flags &= ~IFA_ROUTE;
1271 
1272 			error = rtinit(&ia->ia_ifa, RTM_ADD,
1273 			    rtinitflags(ia) | RTF_UP);
1274 			if (error == 0)
1275 				ia->ia_flags |= IFA_ROUTE;
1276 
1277 			ia4_release(ia, &psref);
1278 			curlwp_bindx(bound);
1279 
1280 			return error;
1281 		}
1282 	}
1283 	pserialize_read_exit(s);
1284 
1285 	/*
1286 	 * noone seem to have prefix route.  remove it.
1287 	 */
1288 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
1289 	target->ia_flags &= ~IFA_ROUTE;
1290 	return 0;
1291 }
1292 
1293 #undef rtinitflags
1294 
1295 /*
1296  * Return 1 if the address might be a local broadcast address.
1297  */
1298 int
1299 in_broadcast(struct in_addr in, struct ifnet *ifp)
1300 {
1301 	struct ifaddr *ifa;
1302 	int s;
1303 
1304 	KASSERT(ifp != NULL);
1305 
1306 	if (in.s_addr == INADDR_BROADCAST ||
1307 	    in_nullhost(in))
1308 		return 1;
1309 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
1310 		return 0;
1311 	/*
1312 	 * Look through the list of addresses for a match
1313 	 * with a broadcast address.
1314 	 */
1315 #define ia (ifatoia(ifa))
1316 	s = pserialize_read_enter();
1317 	IFADDR_READER_FOREACH(ifa, ifp) {
1318 		if (ifa->ifa_addr->sa_family == AF_INET &&
1319 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
1320 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
1321 		     in_hosteq(in, ia->ia_netbroadcast) ||
1322 		     (hostzeroisbroadcast &&
1323 		      /*
1324 		       * Check for old-style (host 0) broadcast.
1325 		       */
1326 		      (in.s_addr == ia->ia_subnet ||
1327 		       in.s_addr == ia->ia_net)))) {
1328 			pserialize_read_exit(s);
1329 			return 1;
1330 		}
1331 	}
1332 	pserialize_read_exit(s);
1333 	return (0);
1334 #undef ia
1335 }
1336 
1337 /*
1338  * perform DAD when interface becomes IFF_UP.
1339  */
1340 void
1341 in_if_link_up(struct ifnet *ifp)
1342 {
1343 	struct ifaddr *ifa;
1344 	struct in_ifaddr *ia;
1345 
1346 	/* Ensure it's sane to run DAD */
1347 	if (ifp->if_link_state == LINK_STATE_DOWN)
1348 		return;
1349 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
1350 		return;
1351 
1352 	IFADDR_READER_FOREACH(ifa, ifp) {
1353 		if (ifa->ifa_addr->sa_family != AF_INET)
1354 			continue;
1355 		ia = (struct in_ifaddr *)ifa;
1356 
1357 		/* If detached then mark as tentative */
1358 		if (ia->ia4_flags & IN_IFF_DETACHED) {
1359 			ia->ia4_flags &= ~IN_IFF_DETACHED;
1360 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
1361 				ia->ia4_flags |= IN_IFF_TENTATIVE;
1362 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
1363 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1364 		}
1365 
1366 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
1367 			/* Clear the duplicated flag as we're starting DAD. */
1368 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
1369 			ia->ia_dad_start(ifa);
1370 		}
1371 	}
1372 }
1373 
1374 void
1375 in_if_up(struct ifnet *ifp)
1376 {
1377 
1378 	/* interface may not support link state, so bring it up also */
1379 	in_if_link_up(ifp);
1380 }
1381 
1382 /*
1383  * Mark all addresses as detached.
1384  */
1385 void
1386 in_if_link_down(struct ifnet *ifp)
1387 {
1388 	struct ifaddr *ifa;
1389 	struct in_ifaddr *ia;
1390 
1391 	IFADDR_READER_FOREACH(ifa, ifp) {
1392 		if (ifa->ifa_addr->sa_family != AF_INET)
1393 			continue;
1394 		ia = (struct in_ifaddr *)ifa;
1395 
1396 		/* Stop DAD processing */
1397 		if (ia->ia_dad_stop != NULL)
1398 			ia->ia_dad_stop(ifa);
1399 
1400 		/*
1401 		 * Mark the address as detached.
1402 		 */
1403 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
1404 			ia->ia4_flags |= IN_IFF_DETACHED;
1405 			ia->ia4_flags &=
1406 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
1407 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1408 		}
1409 	}
1410 }
1411 
1412 void
1413 in_if_down(struct ifnet *ifp)
1414 {
1415 
1416 	in_if_link_down(ifp);
1417 }
1418 
1419 void
1420 in_if_link_state_change(struct ifnet *ifp, int link_state)
1421 {
1422 
1423 	switch (link_state) {
1424 	case LINK_STATE_DOWN:
1425 		in_if_link_down(ifp);
1426 		break;
1427 	case LINK_STATE_UP:
1428 		in_if_link_up(ifp);
1429 		break;
1430 	}
1431 }
1432 
1433 /*
1434  * in_lookup_multi: look up the in_multi record for a given IP
1435  * multicast address on a given interface.  If no matching record is
1436  * found, return NULL.
1437  */
1438 struct in_multi *
1439 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
1440 {
1441 	struct in_multi *inm;
1442 
1443 	KASSERT(rw_lock_held(&in_multilock));
1444 
1445 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
1446 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
1447 			break;
1448 	}
1449 	return inm;
1450 }
1451 
1452 /*
1453  * in_multi_group: check whether the address belongs to an IP multicast
1454  * group we are joined on this interface.  Returns true or false.
1455  */
1456 bool
1457 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
1458 {
1459 	bool ingroup;
1460 
1461 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
1462 		rw_enter(&in_multilock, RW_READER);
1463 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1464 		rw_exit(&in_multilock);
1465 	} else {
1466 		/* XXX Recursive call from ip_output(). */
1467 		KASSERT(rw_lock_held(&in_multilock));
1468 		ingroup = in_lookup_multi(addr, ifp) != NULL;
1469 	}
1470 	return ingroup;
1471 }
1472 
1473 /*
1474  * Add an address to the list of IP multicast addresses for a given interface.
1475  */
1476 struct in_multi *
1477 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
1478 {
1479 	struct sockaddr_in sin;
1480 	struct in_multi *inm;
1481 
1482 	/*
1483 	 * See if address already in list.
1484 	 */
1485 	rw_enter(&in_multilock, RW_WRITER);
1486 	inm = in_lookup_multi(*ap, ifp);
1487 	if (inm != NULL) {
1488 		/*
1489 		 * Found it; just increment the reference count.
1490 		 */
1491 		inm->inm_refcount++;
1492 		rw_exit(&in_multilock);
1493 		return inm;
1494 	}
1495 
1496 	/*
1497 	 * New address; allocate a new multicast record.
1498 	 */
1499 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
1500 	if (inm == NULL) {
1501 		rw_exit(&in_multilock);
1502 		return NULL;
1503 	}
1504 	inm->inm_addr = *ap;
1505 	inm->inm_ifp = ifp;
1506 	inm->inm_refcount = 1;
1507 
1508 	/*
1509 	 * Ask the network driver to update its multicast reception
1510 	 * filter appropriately for the new address.
1511 	 */
1512 	sockaddr_in_init(&sin, ap, 0);
1513 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
1514 		rw_exit(&in_multilock);
1515 		pool_put(&inmulti_pool, inm);
1516 		return NULL;
1517 	}
1518 
1519 	/*
1520 	 * Let IGMP know that we have joined a new IP multicast group.
1521 	 */
1522 	if (igmp_joingroup(inm) != 0) {
1523 		rw_exit(&in_multilock);
1524 		pool_put(&inmulti_pool, inm);
1525 		return NULL;
1526 	}
1527 	LIST_INSERT_HEAD(
1528 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
1529 	    inm, inm_list);
1530 	in_multientries++;
1531 	rw_exit(&in_multilock);
1532 
1533 	return inm;
1534 }
1535 
1536 /*
1537  * Delete a multicast address record.
1538  */
1539 void
1540 in_delmulti(struct in_multi *inm)
1541 {
1542 	struct sockaddr_in sin;
1543 
1544 	rw_enter(&in_multilock, RW_WRITER);
1545 	if (--inm->inm_refcount > 0) {
1546 		rw_exit(&in_multilock);
1547 		return;
1548 	}
1549 
1550 	/*
1551 	 * No remaining claims to this record; let IGMP know that
1552 	 * we are leaving the multicast group.
1553 	 */
1554 	igmp_leavegroup(inm);
1555 
1556 	/*
1557 	 * Notify the network driver to update its multicast reception
1558 	 * filter.
1559 	 */
1560 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
1561 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
1562 
1563 	/*
1564 	 * Unlink from list.
1565 	 */
1566 	LIST_REMOVE(inm, inm_list);
1567 	in_multientries--;
1568 	rw_exit(&in_multilock);
1569 
1570 	pool_put(&inmulti_pool, inm);
1571 }
1572 
1573 /*
1574  * in_next_multi: step through all of the in_multi records, one at a time.
1575  * The current position is remembered in "step", which the caller must
1576  * provide.  in_first_multi(), below, must be called to initialize "step"
1577  * and get the first record.  Both macros return a NULL "inm" when there
1578  * are no remaining records.
1579  */
1580 struct in_multi *
1581 in_next_multi(struct in_multistep *step)
1582 {
1583 	struct in_multi *inm;
1584 
1585 	KASSERT(rw_lock_held(&in_multilock));
1586 
1587 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
1588 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
1589 	}
1590 	if ((inm = step->i_inm) != NULL) {
1591 		step->i_inm = LIST_NEXT(inm, inm_list);
1592 	}
1593 	return inm;
1594 }
1595 
1596 struct in_multi *
1597 in_first_multi(struct in_multistep *step)
1598 {
1599 	KASSERT(rw_lock_held(&in_multilock));
1600 
1601 	step->i_n = 0;
1602 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
1603 	return in_next_multi(step);
1604 }
1605 
1606 void
1607 in_multi_lock(int op)
1608 {
1609 	rw_enter(&in_multilock, op);
1610 }
1611 
1612 void
1613 in_multi_unlock(void)
1614 {
1615 	rw_exit(&in_multilock);
1616 }
1617 
1618 int
1619 in_multi_lock_held(void)
1620 {
1621 	return rw_lock_held(&in_multilock);
1622 }
1623 
1624 struct in_ifaddr *
1625 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
1626     int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref)
1627 {
1628 	struct rtentry *rt = NULL;
1629 	struct in_ifaddr *ia = NULL;
1630 
1631 	KASSERT(ISSET(curlwp->l_pflag, LP_BOUND));
1632 	/*
1633          * If route is known or can be allocated now, take the
1634          * source address from the interface.  Otherwise, punt.
1635 	 */
1636 	if ((soopts & SO_DONTROUTE) != 0)
1637 		rtcache_free(ro);
1638 	else {
1639 		union {
1640 			struct sockaddr		dst;
1641 			struct sockaddr_in	dst4;
1642 		} u;
1643 
1644 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
1645 		rt = rtcache_lookup(ro, &u.dst);
1646 	}
1647 	/*
1648 	 * If we found a route, use the address
1649 	 * corresponding to the outgoing interface
1650 	 * unless it is the loopback (in case a route
1651 	 * to our address on another net goes to loopback).
1652 	 *
1653 	 * XXX Is this still true?  Do we care?
1654 	 */
1655 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
1656 		int s;
1657 		struct ifaddr *ifa;
1658 		/*
1659 		 * Just in case. May not need to do this workaround.
1660 		 * Revisit when working on rtentry MP-ification.
1661 		 */
1662 		s = pserialize_read_enter();
1663 		IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
1664 			if (ifa == rt->rt_ifa)
1665 				break;
1666 		}
1667 		if (ifa != NULL)
1668 			ifa_acquire(ifa, psref);
1669 		pserialize_read_exit(s);
1670 
1671 		ia = ifatoia(ifa);
1672 	}
1673 	if (ia == NULL) {
1674 		u_int16_t fport = sin->sin_port;
1675 		struct ifaddr *ifa;
1676 		int s;
1677 
1678 		sin->sin_port = 0;
1679 		ifa = ifa_ifwithladdr_psref(sintosa(sin), psref);
1680 		sin->sin_port = fport;
1681 		if (ifa == NULL) {
1682 			/* Find 1st non-loopback AF_INET address */
1683 			s = pserialize_read_enter();
1684 			IN_ADDRLIST_READER_FOREACH(ia) {
1685 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
1686 					break;
1687 			}
1688 			if (ia != NULL)
1689 				ia4_acquire(ia, psref);
1690 			pserialize_read_exit(s);
1691 		} else {
1692 			/* ia is already referenced by psref */
1693 			ia = ifatoia(ifa);
1694 		}
1695 		if (ia == NULL) {
1696 			*errorp = EADDRNOTAVAIL;
1697 			return NULL;
1698 		}
1699 	}
1700 	/*
1701 	 * If the destination address is multicast and an outgoing
1702 	 * interface has been set as a multicast option, use the
1703 	 * address of that interface as our source address.
1704 	 */
1705 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
1706 		struct ip_moptions *imo;
1707 
1708 		imo = mopts;
1709 		if (imo->imo_multicast_if_index != 0) {
1710 			struct ifnet *ifp;
1711 			int s;
1712 
1713 			if (ia != NULL)
1714 				ia4_release(ia, psref);
1715 			s = pserialize_read_enter();
1716 			ifp = if_byindex(imo->imo_multicast_if_index);
1717 			if (ifp != NULL) {
1718 				/* XXX */
1719 				ia = in_get_ia_from_ifp_psref(ifp, psref);
1720 			} else
1721 				ia = NULL;
1722 			if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) {
1723 				pserialize_read_exit(s);
1724 				if (ia != NULL)
1725 					ia4_release(ia, psref);
1726 				*errorp = EADDRNOTAVAIL;
1727 				return NULL;
1728 			}
1729 			pserialize_read_exit(s);
1730 		}
1731 	}
1732 	if (ia->ia_ifa.ifa_getifa != NULL) {
1733 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
1734 		                                      sintosa(sin)));
1735 		if (ia == NULL) {
1736 			*errorp = EADDRNOTAVAIL;
1737 			return NULL;
1738 		}
1739 		/* FIXME NOMPSAFE */
1740 		ia4_acquire(ia, psref);
1741 	}
1742 #ifdef GETIFA_DEBUG
1743 	else
1744 		printf("%s: missing ifa_getifa\n", __func__);
1745 #endif
1746 	return ia;
1747 }
1748 
1749 #if NARP > 0
1750 
1751 struct in_llentry {
1752 	struct llentry		base;
1753 };
1754 
1755 #define	IN_LLTBL_DEFAULT_HSIZE	32
1756 #define	IN_LLTBL_HASH(k, h) \
1757 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
1758 
1759 /*
1760  * Do actual deallocation of @lle.
1761  * Called by LLE_FREE_LOCKED when number of references
1762  * drops to zero.
1763  */
1764 static void
1765 in_lltable_destroy_lle(struct llentry *lle)
1766 {
1767 
1768 	LLE_WUNLOCK(lle);
1769 	LLE_LOCK_DESTROY(lle);
1770 	kmem_intr_free(lle, sizeof(*lle));
1771 }
1772 
1773 static struct llentry *
1774 in_lltable_new(struct in_addr addr4, u_int flags)
1775 {
1776 	struct in_llentry *lle;
1777 
1778 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
1779 	if (lle == NULL)		/* NB: caller generates msg */
1780 		return NULL;
1781 
1782 	/*
1783 	 * For IPv4 this will trigger "arpresolve" to generate
1784 	 * an ARP request.
1785 	 */
1786 	lle->base.la_expire = time_uptime; /* mark expired */
1787 	lle->base.r_l3addr.addr4 = addr4;
1788 	lle->base.lle_refcnt = 1;
1789 	lle->base.lle_free = in_lltable_destroy_lle;
1790 	LLE_LOCK_INIT(&lle->base);
1791 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
1792 
1793 	return (&lle->base);
1794 }
1795 
1796 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
1797 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
1798 
1799 static int
1800 in_lltable_match_prefix(const struct sockaddr *prefix,
1801     const struct sockaddr *mask, u_int flags, struct llentry *lle)
1802 {
1803 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
1804 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
1805 
1806 	/*
1807 	 * (flags & LLE_STATIC) means deleting all entries
1808 	 * including static ARP entries.
1809 	 */
1810 	if (IN_ARE_MASKED_ADDR_EQUAL(lle->r_l3addr.addr4, pfx, msk) &&
1811 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
1812 		return (1);
1813 
1814 	return (0);
1815 }
1816 
1817 static void
1818 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
1819 {
1820 	struct ifnet *ifp __diagused;
1821 	size_t pkts_dropped;
1822 
1823 	LLE_WLOCK_ASSERT(lle);
1824 	KASSERT(llt != NULL);
1825 
1826 	/* Unlink entry from table if not already */
1827 	if ((lle->la_flags & LLE_LINKED) != 0) {
1828 		ifp = llt->llt_ifp;
1829 		IF_AFDATA_WLOCK_ASSERT(ifp);
1830 		lltable_unlink_entry(llt, lle);
1831 	}
1832 
1833 	/* cancel timer */
1834 	if (callout_halt(&lle->lle_timer, &lle->lle_lock))
1835 		LLE_REMREF(lle);
1836 
1837 	/* Drop hold queue */
1838 	pkts_dropped = llentry_free(lle);
1839 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
1840 }
1841 
1842 static int
1843 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
1844 {
1845 	struct rtentry *rt;
1846 	int error = EINVAL;
1847 
1848 	KASSERTMSG(l3addr->sa_family == AF_INET,
1849 	    "sin_family %d", l3addr->sa_family);
1850 
1851 	rt = rtalloc1(l3addr, 0);
1852 	if (rt == NULL)
1853 		return error;
1854 
1855 	/*
1856 	 * If the gateway for an existing host route matches the target L3
1857 	 * address, which is a special route inserted by some implementation
1858 	 * such as MANET, and the interface is of the correct type, then
1859 	 * allow for ARP to proceed.
1860 	 */
1861 	if (rt->rt_flags & RTF_GATEWAY) {
1862 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
1863 		    rt->rt_ifp->if_type != IFT_ETHER ||
1864 #ifdef __FreeBSD__
1865 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
1866 #else
1867 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
1868 #endif
1869 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
1870 		    sizeof(in_addr_t)) != 0) {
1871 			goto error;
1872 		}
1873 	}
1874 
1875 	/*
1876 	 * Make sure that at least the destination address is covered
1877 	 * by the route. This is for handling the case where 2 or more
1878 	 * interfaces have the same prefix. An incoming packet arrives
1879 	 * on one interface and the corresponding outgoing packet leaves
1880 	 * another interface.
1881 	 */
1882 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
1883 		const char *sa, *mask, *addr, *lim;
1884 		int len;
1885 
1886 		mask = (const char *)rt_mask(rt);
1887 		/*
1888 		 * Just being extra cautious to avoid some custom
1889 		 * code getting into trouble.
1890 		 */
1891 		if (mask == NULL)
1892 			goto error;
1893 
1894 		sa = (const char *)rt_getkey(rt);
1895 		addr = (const char *)l3addr;
1896 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
1897 		lim = addr + len;
1898 
1899 		for ( ; addr < lim; sa++, mask++, addr++) {
1900 			if ((*sa ^ *addr) & *mask) {
1901 #ifdef DIAGNOSTIC
1902 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
1903 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
1904 #endif
1905 				goto error;
1906 			}
1907 		}
1908 	}
1909 
1910 	error = 0;
1911 error:
1912 	rtfree(rt);
1913 	return error;
1914 }
1915 
1916 static inline uint32_t
1917 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
1918 {
1919 
1920 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
1921 }
1922 
1923 static uint32_t
1924 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
1925 {
1926 
1927 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
1928 }
1929 
1930 static void
1931 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
1932 {
1933 	struct sockaddr_in *sin;
1934 
1935 	sin = (struct sockaddr_in *)sa;
1936 	memset(sin, 0, sizeof(*sin));
1937 	sin->sin_family = AF_INET;
1938 	sin->sin_len = sizeof(*sin);
1939 	sin->sin_addr = lle->r_l3addr.addr4;
1940 }
1941 
1942 static inline struct llentry *
1943 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
1944 {
1945 	struct llentry *lle;
1946 	struct llentries *lleh;
1947 	u_int hashidx;
1948 
1949 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
1950 	lleh = &llt->lle_head[hashidx];
1951 	LIST_FOREACH(lle, lleh, lle_next) {
1952 		if (lle->la_flags & LLE_DELETED)
1953 			continue;
1954 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
1955 			break;
1956 	}
1957 
1958 	return (lle);
1959 }
1960 
1961 static int
1962 in_lltable_delete(struct lltable *llt, u_int flags,
1963     const struct sockaddr *l3addr)
1964 {
1965 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1966 	struct ifnet *ifp __diagused = llt->llt_ifp;
1967 	struct llentry *lle;
1968 
1969 	IF_AFDATA_WLOCK_ASSERT(ifp);
1970 	KASSERTMSG(l3addr->sa_family == AF_INET,
1971 	    "sin_family %d", l3addr->sa_family);
1972 
1973 	lle = in_lltable_find_dst(llt, sin->sin_addr);
1974 	if (lle == NULL) {
1975 #ifdef DIAGNOSTIC
1976 		log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
1977 #endif
1978 		return (ENOENT);
1979 	}
1980 
1981 	LLE_WLOCK(lle);
1982 	lle->la_flags |= LLE_DELETED;
1983 #ifdef DIAGNOSTIC
1984 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
1985 #endif
1986 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
1987 		llentry_free(lle);
1988 	else
1989 		LLE_WUNLOCK(lle);
1990 
1991 	return (0);
1992 }
1993 
1994 static struct llentry *
1995 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
1996 {
1997 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
1998 	struct ifnet *ifp = llt->llt_ifp;
1999 	struct llentry *lle;
2000 
2001 	IF_AFDATA_WLOCK_ASSERT(ifp);
2002 	KASSERTMSG(l3addr->sa_family == AF_INET,
2003 	    "sin_family %d", l3addr->sa_family);
2004 
2005 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2006 
2007 	if (lle != NULL) {
2008 		LLE_WLOCK(lle);
2009 		return (lle);
2010 	}
2011 
2012 	/* no existing record, we need to create new one */
2013 
2014 	/*
2015 	 * A route that covers the given address must have
2016 	 * been installed 1st because we are doing a resolution,
2017 	 * verify this.
2018 	 */
2019 	if (!(flags & LLE_IFADDR) &&
2020 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
2021 		return (NULL);
2022 
2023 	lle = in_lltable_new(sin->sin_addr, flags);
2024 	if (lle == NULL) {
2025 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2026 		return (NULL);
2027 	}
2028 	lle->la_flags = flags;
2029 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2030 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2031 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2032 	}
2033 
2034 	lltable_link_entry(llt, lle);
2035 	LLE_WLOCK(lle);
2036 
2037 	return (lle);
2038 }
2039 
2040 /*
2041  * Return NULL if not found or marked for deletion.
2042  * If found return lle read locked.
2043  */
2044 static struct llentry *
2045 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
2046 {
2047 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
2048 	struct llentry *lle;
2049 
2050 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2051 	KASSERTMSG(l3addr->sa_family == AF_INET,
2052 	    "sin_family %d", l3addr->sa_family);
2053 
2054 	lle = in_lltable_find_dst(llt, sin->sin_addr);
2055 
2056 	if (lle == NULL)
2057 		return NULL;
2058 
2059 	if (flags & LLE_EXCLUSIVE)
2060 		LLE_WLOCK(lle);
2061 	else
2062 		LLE_RLOCK(lle);
2063 
2064 	return lle;
2065 }
2066 
2067 static int
2068 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2069     struct rt_walkarg *w)
2070 {
2071 	struct sockaddr_in sin;
2072 
2073 	LLTABLE_LOCK_ASSERT();
2074 
2075 	/* skip deleted entries */
2076 	if (lle->la_flags & LLE_DELETED)
2077 		return 0;
2078 
2079 	sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0);
2080 
2081 	return lltable_dump_entry(llt, lle, w, sintosa(&sin));
2082 }
2083 
2084 #endif /* NARP > 0 */
2085 
2086 static int
2087 in_multicast_sysctl(SYSCTLFN_ARGS)
2088 {
2089 	struct ifnet *ifp;
2090 	struct ifaddr *ifa;
2091 	struct in_ifaddr *ifa4;
2092 	struct in_multi *inm;
2093 	uint32_t tmp;
2094 	int error;
2095 	size_t written;
2096 	struct psref psref;
2097 	int bound;
2098 
2099 	if (namelen != 1)
2100 		return EINVAL;
2101 
2102 	bound = curlwp_bind();
2103 	ifp = if_get_byindex(name[0], &psref);
2104 	if (ifp == NULL) {
2105 		curlwp_bindx(bound);
2106 		return ENODEV;
2107 	}
2108 
2109 	if (oldp == NULL) {
2110 		*oldlenp = 0;
2111 		IFADDR_FOREACH(ifa, ifp) {
2112 			if (ifa->ifa_addr->sa_family != AF_INET)
2113 				continue;
2114 			ifa4 = (void *)ifa;
2115 			LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2116 				*oldlenp += 2 * sizeof(struct in_addr) +
2117 				    sizeof(uint32_t);
2118 			}
2119 		}
2120 		if_put(ifp, &psref);
2121 		curlwp_bindx(bound);
2122 		return 0;
2123 	}
2124 
2125 	error = 0;
2126 	written = 0;
2127 	IFADDR_FOREACH(ifa, ifp) {
2128 		if (ifa->ifa_addr->sa_family != AF_INET)
2129 			continue;
2130 		ifa4 = (void *)ifa;
2131 		LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) {
2132 			if (written + 2 * sizeof(struct in_addr) +
2133 			    sizeof(uint32_t) > *oldlenp)
2134 				goto done;
2135 			error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr,
2136 			    oldp, sizeof(struct in_addr));
2137 			if (error)
2138 				goto done;
2139 			oldp = (char *)oldp + sizeof(struct in_addr);
2140 			written += sizeof(struct in_addr);
2141 			error = sysctl_copyout(l, &inm->inm_addr,
2142 			    oldp, sizeof(struct in_addr));
2143 			if (error)
2144 				goto done;
2145 			oldp = (char *)oldp + sizeof(struct in_addr);
2146 			written += sizeof(struct in_addr);
2147 			tmp = inm->inm_refcount;
2148 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
2149 			if (error)
2150 				goto done;
2151 			oldp = (char *)oldp + sizeof(tmp);
2152 			written += sizeof(tmp);
2153 		}
2154 	}
2155 done:
2156 	if_put(ifp, &psref);
2157 	curlwp_bindx(bound);
2158 	*oldlenp = written;
2159 	return error;
2160 }
2161 
2162 static void
2163 in_sysctl_init(struct sysctllog **clog)
2164 {
2165 	sysctl_createv(clog, 0, NULL, NULL,
2166 		       CTLFLAG_PERMANENT,
2167 		       CTLTYPE_NODE, "inet",
2168 		       SYSCTL_DESCR("PF_INET related settings"),
2169 		       NULL, 0, NULL, 0,
2170 		       CTL_NET, PF_INET, CTL_EOL);
2171 	sysctl_createv(clog, 0, NULL, NULL,
2172 		       CTLFLAG_PERMANENT,
2173 		       CTLTYPE_NODE, "multicast",
2174 		       SYSCTL_DESCR("Multicast information"),
2175 		       in_multicast_sysctl, 0, NULL, 0,
2176 		       CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
2177 	sysctl_createv(clog, 0, NULL, NULL,
2178 		       CTLFLAG_PERMANENT,
2179 		       CTLTYPE_NODE, "ip",
2180 		       SYSCTL_DESCR("IPv4 related settings"),
2181 		       NULL, 0, NULL, 0,
2182 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2183 
2184 	sysctl_createv(clog, 0, NULL, NULL,
2185 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2186 		       CTLTYPE_INT, "subnetsarelocal",
2187 		       SYSCTL_DESCR("Whether logical subnets are considered "
2188 				    "local"),
2189 		       NULL, 0, &subnetsarelocal, 0,
2190 		       CTL_NET, PF_INET, IPPROTO_IP,
2191 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2192 	sysctl_createv(clog, 0, NULL, NULL,
2193 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2194 		       CTLTYPE_INT, "hostzerobroadcast",
2195 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2196 		       NULL, 0, &hostzeroisbroadcast, 0,
2197 		       CTL_NET, PF_INET, IPPROTO_IP,
2198 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2199 }
2200 
2201 #if NARP > 0
2202 
2203 static struct lltable *
2204 in_lltattach(struct ifnet *ifp)
2205 {
2206 	struct lltable *llt;
2207 
2208 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
2209 	llt->llt_af = AF_INET;
2210 	llt->llt_ifp = ifp;
2211 
2212 	llt->llt_lookup = in_lltable_lookup;
2213 	llt->llt_create = in_lltable_create;
2214 	llt->llt_delete = in_lltable_delete;
2215 	llt->llt_dump_entry = in_lltable_dump_entry;
2216 	llt->llt_hash = in_lltable_hash;
2217 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
2218 	llt->llt_free_entry = in_lltable_free_entry;
2219 	llt->llt_match_prefix = in_lltable_match_prefix;
2220 	lltable_link(llt);
2221 
2222 	return (llt);
2223 }
2224 
2225 #endif /* NARP > 0 */
2226 
2227 void *
2228 in_domifattach(struct ifnet *ifp)
2229 {
2230 	struct in_ifinfo *ii;
2231 
2232 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
2233 	KASSERT(ii != NULL);
2234 
2235 #if NARP > 0
2236 	ii->ii_llt = in_lltattach(ifp);
2237 #endif
2238 
2239 #ifdef IPSELSRC
2240 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
2241 	KASSERT(ii->ii_selsrc != NULL);
2242 #endif
2243 
2244 	return ii;
2245 }
2246 
2247 void
2248 in_domifdetach(struct ifnet *ifp, void *aux)
2249 {
2250 	struct in_ifinfo *ii = aux;
2251 
2252 #ifdef IPSELSRC
2253 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
2254 #endif
2255 #if NARP > 0
2256 	lltable_free(ii->ii_llt);
2257 #endif
2258 	kmem_free(ii, sizeof(struct in_ifinfo));
2259 }
2260