Home | History | Annotate | Line # | Download | only in netinet
in.c revision 1.161
      1 /*	$NetBSD: in.c,v 1.161 2015/08/31 16:46:14 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 1.161 2015/08/31 16:46:14 ozaki-r Exp $");
     95 
     96 #include "arp.h"
     97 #include "ether.h"
     98 
     99 #ifdef _KERNEL_OPT
    100 #include "opt_inet.h"
    101 #include "opt_inet_conf.h"
    102 #include "opt_mrouting.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 int	in_scrubprefix(struct in_ifaddr *);
    154 static void	in_sysctl_init(struct sysctllog **);
    155 
    156 #ifndef SUBNETSARELOCAL
    157 #define	SUBNETSARELOCAL	1
    158 #endif
    159 
    160 #ifndef HOSTZEROBROADCAST
    161 #define HOSTZEROBROADCAST 1
    162 #endif
    163 
    164 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
    165 #ifndef IN_MULTI_HASH_SIZE
    166 #define IN_MULTI_HASH_SIZE	509
    167 #endif
    168 
    169 static int			subnetsarelocal = SUBNETSARELOCAL;
    170 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
    171 
    172 /*
    173  * This list is used to keep track of in_multi chains which belong to
    174  * deleted interface addresses.  We use in_ifaddr so that a chain head
    175  * won't be deallocated until all multicast address record are deleted.
    176  */
    177 
    178 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
    179 
    180 static struct pool		inmulti_pool;
    181 static u_int			in_multientries;
    182 static struct in_multihashhead *in_multihashtbl;
    183 static u_long			in_multihash;
    184 static krwlock_t		in_multilock;
    185 
    186 #define IN_MULTI_HASH(x, ifp) \
    187     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
    188 
    189 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
    190 u_long				in_ifaddrhash;
    191 struct in_ifaddrhead		in_ifaddrhead;
    192 
    193 void
    194 in_init(void)
    195 {
    196 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
    197 	    NULL, IPL_SOFTNET);
    198 	TAILQ_INIT(&in_ifaddrhead);
    199 
    200 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    201 	    &in_ifaddrhash);
    202 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    203 	    &in_multihash);
    204 	rw_init(&in_multilock);
    205 
    206 	in_sysctl_init(NULL);
    207 }
    208 
    209 /*
    210  * Return 1 if an internet address is for a ``local'' host
    211  * (one to which we have a connection).  If subnetsarelocal
    212  * is true, this includes other subnets of the local net.
    213  * Otherwise, it includes only the directly-connected (sub)nets.
    214  */
    215 int
    216 in_localaddr(struct in_addr in)
    217 {
    218 	struct in_ifaddr *ia;
    219 
    220 	if (subnetsarelocal) {
    221 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    222 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net)
    223 				return (1);
    224 	} else {
    225 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    226 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet)
    227 				return (1);
    228 	}
    229 	return (0);
    230 }
    231 
    232 /*
    233  * Determine whether an IP address is in a reserved set of addresses
    234  * that may not be forwarded, or whether datagrams to that destination
    235  * may be forwarded.
    236  */
    237 int
    238 in_canforward(struct in_addr in)
    239 {
    240 	u_int32_t net;
    241 
    242 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
    243 		return (0);
    244 	if (IN_CLASSA(in.s_addr)) {
    245 		net = in.s_addr & IN_CLASSA_NET;
    246 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
    247 			return (0);
    248 	}
    249 	return (1);
    250 }
    251 
    252 /*
    253  * Trim a mask in a sockaddr
    254  */
    255 void
    256 in_socktrim(struct sockaddr_in *ap)
    257 {
    258 	char *cplim = (char *) &ap->sin_addr;
    259 	char *cp = (char *) (&ap->sin_addr + 1);
    260 
    261 	ap->sin_len = 0;
    262 	while (--cp >= cplim)
    263 		if (*cp) {
    264 			(ap)->sin_len = cp - (char *) (ap) + 1;
    265 			break;
    266 		}
    267 }
    268 
    269 /*
    270  *  Routine to take an Internet address and convert into a
    271  *  "dotted quad" representation for printing.
    272  */
    273 const char *
    274 in_fmtaddr(struct in_addr addr)
    275 {
    276 	static char buf[sizeof("123.456.789.123")];
    277 
    278 	addr.s_addr = ntohl(addr.s_addr);
    279 
    280 	snprintf(buf, sizeof(buf), "%d.%d.%d.%d",
    281 		(addr.s_addr >> 24) & 0xFF,
    282 		(addr.s_addr >> 16) & 0xFF,
    283 		(addr.s_addr >>  8) & 0xFF,
    284 		(addr.s_addr >>  0) & 0xFF);
    285 	return buf;
    286 }
    287 
    288 /*
    289  * Maintain the "in_maxmtu" variable, which is the largest
    290  * mtu for non-local interfaces with AF_INET addresses assigned
    291  * to them that are up.
    292  */
    293 unsigned long in_maxmtu;
    294 
    295 void
    296 in_setmaxmtu(void)
    297 {
    298 	struct in_ifaddr *ia;
    299 	struct ifnet *ifp;
    300 	unsigned long maxmtu = 0;
    301 
    302 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    303 		if ((ifp = ia->ia_ifp) == 0)
    304 			continue;
    305 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
    306 			continue;
    307 		if (ifp->if_mtu > maxmtu)
    308 			maxmtu = ifp->if_mtu;
    309 	}
    310 	if (maxmtu)
    311 		in_maxmtu = maxmtu;
    312 }
    313 
    314 static u_int
    315 in_mask2len(struct in_addr *mask)
    316 {
    317 	u_int x, y;
    318 	u_char *p;
    319 
    320 	p = (u_char *)mask;
    321 	for (x = 0; x < sizeof(*mask); x++) {
    322 		if (p[x] != 0xff)
    323 			break;
    324 	}
    325 	y = 0;
    326 	if (x < sizeof(*mask)) {
    327 		for (y = 0; y < NBBY; y++) {
    328 			if ((p[x] & (0x80 >> y)) == 0)
    329 				break;
    330 		}
    331 	}
    332 	return x * NBBY + y;
    333 }
    334 
    335 static void
    336 in_len2mask(struct in_addr *mask, u_int len)
    337 {
    338 	u_int i;
    339 	u_char *p;
    340 
    341 	p = (u_char *)mask;
    342 	memset(mask, 0, sizeof(*mask));
    343 	for (i = 0; i < len / NBBY; i++)
    344 		p[i] = 0xff;
    345 	if (len % NBBY)
    346 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
    347 }
    348 
    349 /*
    350  * Generic internet control operations (ioctl's).
    351  * Ifp is 0 if not an interface-specific ioctl.
    352  */
    353 /* ARGSUSED */
    354 int
    355 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
    356 {
    357 	struct ifreq *ifr = (struct ifreq *)data;
    358 	struct in_ifaddr *ia = NULL;
    359 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
    360 	struct sockaddr_in oldaddr;
    361 	int error, hostIsNew, maskIsNew;
    362 	int newifaddr = 0;
    363 
    364 	switch (cmd) {
    365 	case SIOCALIFADDR:
    366 	case SIOCDLIFADDR:
    367 	case SIOCGLIFADDR:
    368 		if (ifp == NULL)
    369 			return EINVAL;
    370 		return in_lifaddr_ioctl(so, cmd, data, ifp);
    371 	case SIOCGIFADDRPREF:
    372 	case SIOCSIFADDRPREF:
    373 		if (ifp == NULL)
    374 			return EINVAL;
    375 		return ifaddrpref_ioctl(so, cmd, data, ifp);
    376 	}
    377 
    378 	/*
    379 	 * Find address for this interface, if it exists.
    380 	 */
    381 	if (ifp != NULL)
    382 		IFP_TO_IA(ifp, ia);
    383 
    384 	hostIsNew = 1;		/* moved here to appease gcc */
    385 	switch (cmd) {
    386 	case SIOCAIFADDR:
    387 	case SIOCDIFADDR:
    388 	case SIOCGIFALIAS:
    389 	case SIOCGIFAFLAG_IN:
    390 		if (ifra->ifra_addr.sin_family == AF_INET)
    391 			LIST_FOREACH(ia,
    392 			    &IN_IFADDR_HASH(ifra->ifra_addr.sin_addr.s_addr),
    393 			    ia_hash) {
    394 				if (ia->ia_ifp == ifp &&
    395 				    in_hosteq(ia->ia_addr.sin_addr,
    396 				    ifra->ifra_addr.sin_addr))
    397 					break;
    398 			}
    399 		if ((cmd == SIOCDIFADDR ||
    400 		    cmd == SIOCGIFALIAS ||
    401 		    cmd == SIOCGIFAFLAG_IN) &&
    402 		    ia == NULL)
    403 			return (EADDRNOTAVAIL);
    404 
    405 		if (cmd == SIOCDIFADDR &&
    406 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
    407 			ifra->ifra_addr.sin_family = AF_INET;
    408 		}
    409 		/* FALLTHROUGH */
    410 	case SIOCSIFADDR:
    411 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
    412 			;
    413 		else if (ifra->ifra_addr.sin_len == 0) {
    414 			ifra->ifra_addr = ia->ia_addr;
    415 			hostIsNew = 0;
    416 		} else if (in_hosteq(ia->ia_addr.sin_addr,
    417 		           ifra->ifra_addr.sin_addr))
    418 			hostIsNew = 0;
    419 		/* FALLTHROUGH */
    420 	case SIOCSIFDSTADDR:
    421 		if (ifra->ifra_addr.sin_family != AF_INET)
    422 			return (EAFNOSUPPORT);
    423 		/* FALLTHROUGH */
    424 	case SIOCSIFNETMASK:
    425 		if (ifp == NULL)
    426 			panic("in_control");
    427 
    428 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
    429 			break;
    430 
    431 		if (ia == NULL &&
    432 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR))
    433 			return (EADDRNOTAVAIL);
    434 
    435 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    436 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    437 		    NULL) != 0)
    438 			return (EPERM);
    439 
    440 		if (ia == NULL) {
    441 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
    442 			if (ia == NULL)
    443 				return (ENOBUFS);
    444 			TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
    445 			ifaref(&ia->ia_ifa);
    446 			ifa_insert(ifp, &ia->ia_ifa);
    447 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    448 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    449 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
    450 #ifdef IPSELSRC
    451 			ia->ia_ifa.ifa_getifa = in_getifa;
    452 #else /* IPSELSRC */
    453 			ia->ia_ifa.ifa_getifa = NULL;
    454 #endif /* IPSELSRC */
    455 			ia->ia_sockmask.sin_len = 8;
    456 			ia->ia_sockmask.sin_family = AF_INET;
    457 			if (ifp->if_flags & IFF_BROADCAST) {
    458 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
    459 				ia->ia_broadaddr.sin_family = AF_INET;
    460 			}
    461 			ia->ia_ifp = ifp;
    462 			ia->ia_idsalt = cprng_fast32() % 65535;
    463 			LIST_INIT(&ia->ia_multiaddrs);
    464 			newifaddr = 1;
    465 		}
    466 		break;
    467 
    468 	case SIOCSIFBRDADDR:
    469 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    470 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    471 		    NULL) != 0)
    472 			return (EPERM);
    473 		/* FALLTHROUGH */
    474 
    475 	case SIOCGIFADDR:
    476 	case SIOCGIFNETMASK:
    477 	case SIOCGIFDSTADDR:
    478 	case SIOCGIFBRDADDR:
    479 		if (ia == NULL)
    480 			return (EADDRNOTAVAIL);
    481 		break;
    482 	}
    483 	error = 0;
    484 	switch (cmd) {
    485 
    486 	case SIOCGIFADDR:
    487 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
    488 		break;
    489 
    490 	case SIOCGIFBRDADDR:
    491 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    492 			return (EINVAL);
    493 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
    494 		break;
    495 
    496 	case SIOCGIFDSTADDR:
    497 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    498 			return (EINVAL);
    499 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
    500 		break;
    501 
    502 	case SIOCGIFNETMASK:
    503 		/*
    504 		 * We keep the number of trailing zero bytes the sin_len field
    505 		 * of ia_sockmask, so we fix this before we pass it back to
    506 		 * userland.
    507 		 */
    508 		oldaddr = ia->ia_sockmask;
    509 		oldaddr.sin_len = sizeof(struct sockaddr_in);
    510 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
    511 		break;
    512 
    513 	case SIOCSIFDSTADDR:
    514 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    515 			return (EINVAL);
    516 		oldaddr = ia->ia_dstaddr;
    517 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
    518 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
    519 			ia->ia_dstaddr = oldaddr;
    520 			return error;
    521 		}
    522 		if (ia->ia_flags & IFA_ROUTE) {
    523 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
    524 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
    525 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    526 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
    527 		}
    528 		break;
    529 
    530 	case SIOCSIFBRDADDR:
    531 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    532 			return EINVAL;
    533 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
    534 		break;
    535 
    536 	case SIOCSIFADDR:
    537 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
    538 		    1, hostIsNew);
    539 		if (error == 0) {
    540 			(void)pfil_run_hooks(if_pfil,
    541 			    (struct mbuf **)SIOCSIFADDR, ifp, PFIL_IFADDR);
    542 		}
    543 		break;
    544 
    545 	case SIOCSIFNETMASK:
    546 		in_ifscrub(ifp, ia);
    547 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
    548 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    549 		error = in_ifinit(ifp, ia, NULL, 0, 0);
    550 		break;
    551 
    552 	case SIOCAIFADDR:
    553 		maskIsNew = 0;
    554 		if (ifra->ifra_mask.sin_len) {
    555 			/* Only scrub if we control the prefix route,
    556 			 * otherwise userland gets a bogus message */
    557 			if ((ia->ia_flags & IFA_ROUTE))
    558 				in_ifscrub(ifp, ia);
    559 			ia->ia_sockmask = ifra->ifra_mask;
    560 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    561 			maskIsNew = 1;
    562 		}
    563 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    564 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
    565 			/* Only scrub if we control the prefix route,
    566 			 * otherwise userland gets a bogus message */
    567 			if ((ia->ia_flags & IFA_ROUTE))
    568 				in_ifscrub(ifp, ia);
    569 			ia->ia_dstaddr = ifra->ifra_dstaddr;
    570 			maskIsNew  = 1; /* We lie; but the effect's the same */
    571 		}
    572 		if (ifra->ifra_addr.sin_family == AF_INET &&
    573 		    (hostIsNew || maskIsNew)) {
    574 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
    575 			    hostIsNew);
    576 		}
    577 		if ((ifp->if_flags & IFF_BROADCAST) &&
    578 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
    579 			ia->ia_broadaddr = ifra->ifra_broadaddr;
    580 		if (error == 0)
    581 			(void)pfil_run_hooks(if_pfil,
    582 			    (struct mbuf **)SIOCAIFADDR, ifp, PFIL_IFADDR);
    583 		break;
    584 
    585 	case SIOCGIFALIAS:
    586 		ifra->ifra_mask = ia->ia_sockmask;
    587 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    588 		    (ia->ia_dstaddr.sin_family == AF_INET))
    589 			ifra->ifra_dstaddr = ia->ia_dstaddr;
    590 		else if ((ifp->if_flags & IFF_BROADCAST) &&
    591 		    (ia->ia_broadaddr.sin_family == AF_INET))
    592 			ifra->ifra_broadaddr = ia->ia_broadaddr;
    593 		else
    594 			memset(&ifra->ifra_broadaddr, 0,
    595 			      sizeof(ifra->ifra_broadaddr));
    596 		break;
    597 
    598 	case SIOCGIFAFLAG_IN:
    599 		ifr->ifr_addrflags = ia->ia4_flags;
    600 		break;
    601 
    602 	case SIOCDIFADDR:
    603 		in_purgeaddr(&ia->ia_ifa);
    604 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR,
    605 		    ifp, PFIL_IFADDR);
    606 		break;
    607 
    608 #ifdef MROUTING
    609 	case SIOCGETVIFCNT:
    610 	case SIOCGETSGCNT:
    611 		error = mrt_ioctl(so, cmd, data);
    612 		break;
    613 #endif /* MROUTING */
    614 
    615 	default:
    616 		return ENOTTY;
    617 	}
    618 
    619 	if (error != 0 && newifaddr) {
    620 		KASSERT(ia != NULL);
    621 		in_purgeaddr(&ia->ia_ifa);
    622 	}
    623 
    624 	return error;
    625 }
    626 
    627 /* Add ownaddr as loopback rtentry. */
    628 static void
    629 in_ifaddlocal(struct ifaddr *ifa)
    630 {
    631 	struct in_ifaddr *ia;
    632 
    633 	ia = (struct in_ifaddr *)ifa;
    634 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
    635 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
    636 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
    637 	{
    638 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
    639 		return;
    640 	}
    641 
    642 	rt_ifa_addlocal(ifa);
    643 }
    644 
    645 /* Rempve loopback entry of ownaddr */
    646 static void
    647 in_ifremlocal(struct ifaddr *ifa)
    648 {
    649 	struct in_ifaddr *ia, *p;
    650 	struct ifaddr *alt_ifa = NULL;
    651 	int ia_count = 0;
    652 
    653 	ia = (struct in_ifaddr *)ifa;
    654 	/* Delete the entry if exactly one ifaddr matches the
    655 	 * address, ifa->ifa_addr. */
    656 	TAILQ_FOREACH(p, &in_ifaddrhead, ia_list) {
    657 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
    658 			continue;
    659 		if (p->ia_ifp != ia->ia_ifp)
    660 			alt_ifa = &p->ia_ifa;
    661 		if (++ia_count > 1 && alt_ifa != NULL)
    662 			break;
    663 	}
    664 
    665 	if (ia_count == 0)
    666 		return;
    667 
    668 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
    669 }
    670 
    671 void
    672 in_purgeaddr(struct ifaddr *ifa)
    673 {
    674 	struct ifnet *ifp = ifa->ifa_ifp;
    675 	struct in_ifaddr *ia = (void *) ifa;
    676 
    677 	/* stop DAD processing */
    678 	if (ia->ia_dad_stop != NULL)
    679 		ia->ia_dad_stop(ifa);
    680 
    681 	in_ifscrub(ifp, ia);
    682 	in_ifremlocal(ifa);
    683 	LIST_REMOVE(ia, ia_hash);
    684 	ifa_remove(ifp, &ia->ia_ifa);
    685 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
    686 	if (ia->ia_allhosts != NULL)
    687 		in_delmulti(ia->ia_allhosts);
    688 	ifafree(&ia->ia_ifa);
    689 	in_setmaxmtu();
    690 }
    691 
    692 void
    693 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
    694 {
    695 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
    696 	igmp_purgeif(ifp);		/* manipulates pools */
    697 #ifdef MROUTING
    698 	ip_mrouter_detach(ifp);
    699 #endif
    700 }
    701 
    702 /*
    703  * SIOC[GAD]LIFADDR.
    704  *	SIOCGLIFADDR: get first address. (???)
    705  *	SIOCGLIFADDR with IFLR_PREFIX:
    706  *		get first address that matches the specified prefix.
    707  *	SIOCALIFADDR: add the specified address.
    708  *	SIOCALIFADDR with IFLR_PREFIX:
    709  *		EINVAL since we can't deduce hostid part of the address.
    710  *	SIOCDLIFADDR: delete the specified address.
    711  *	SIOCDLIFADDR with IFLR_PREFIX:
    712  *		delete the first address that matches the specified prefix.
    713  * return values:
    714  *	EINVAL on invalid parameters
    715  *	EADDRNOTAVAIL on prefix match failed/specified address not found
    716  *	other values may be returned from in_ioctl()
    717  */
    718 static int
    719 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
    720     struct ifnet *ifp)
    721 {
    722 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
    723 	struct ifaddr *ifa;
    724 	struct sockaddr *sa;
    725 
    726 	/* sanity checks */
    727 	if (data == NULL || ifp == NULL) {
    728 		panic("invalid argument to in_lifaddr_ioctl");
    729 		/*NOTRECHED*/
    730 	}
    731 
    732 	switch (cmd) {
    733 	case SIOCGLIFADDR:
    734 		/* address must be specified on GET with IFLR_PREFIX */
    735 		if ((iflr->flags & IFLR_PREFIX) == 0)
    736 			break;
    737 		/*FALLTHROUGH*/
    738 	case SIOCALIFADDR:
    739 	case SIOCDLIFADDR:
    740 		/* address must be specified on ADD and DELETE */
    741 		sa = (struct sockaddr *)&iflr->addr;
    742 		if (sa->sa_family != AF_INET)
    743 			return EINVAL;
    744 		if (sa->sa_len != sizeof(struct sockaddr_in))
    745 			return EINVAL;
    746 		/* XXX need improvement */
    747 		sa = (struct sockaddr *)&iflr->dstaddr;
    748 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
    749 			return EINVAL;
    750 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
    751 			return EINVAL;
    752 		break;
    753 	default: /*shouldn't happen*/
    754 #if 0
    755 		panic("invalid cmd to in_lifaddr_ioctl");
    756 		/*NOTREACHED*/
    757 #else
    758 		return EOPNOTSUPP;
    759 #endif
    760 	}
    761 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
    762 		return EINVAL;
    763 
    764 	switch (cmd) {
    765 	case SIOCALIFADDR:
    766 	    {
    767 		struct in_aliasreq ifra;
    768 
    769 		if (iflr->flags & IFLR_PREFIX)
    770 			return EINVAL;
    771 
    772 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
    773 		memset(&ifra, 0, sizeof(ifra));
    774 		memcpy(ifra.ifra_name, iflr->iflr_name,
    775 			sizeof(ifra.ifra_name));
    776 
    777 		memcpy(&ifra.ifra_addr, &iflr->addr,
    778 			((struct sockaddr *)&iflr->addr)->sa_len);
    779 
    780 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
    781 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
    782 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
    783 		}
    784 
    785 		ifra.ifra_mask.sin_family = AF_INET;
    786 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
    787 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
    788 
    789 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
    790 	    }
    791 	case SIOCGLIFADDR:
    792 	case SIOCDLIFADDR:
    793 	    {
    794 		struct in_ifaddr *ia;
    795 		struct in_addr mask, candidate, match;
    796 		struct sockaddr_in *sin;
    797 		int cmp;
    798 
    799 		memset(&mask, 0, sizeof(mask));
    800 		memset(&match, 0, sizeof(match));	/* XXX gcc */
    801 		if (iflr->flags & IFLR_PREFIX) {
    802 			/* lookup a prefix rather than address. */
    803 			in_len2mask(&mask, iflr->prefixlen);
    804 
    805 			sin = (struct sockaddr_in *)&iflr->addr;
    806 			match.s_addr = sin->sin_addr.s_addr;
    807 			match.s_addr &= mask.s_addr;
    808 
    809 			/* if you set extra bits, that's wrong */
    810 			if (match.s_addr != sin->sin_addr.s_addr)
    811 				return EINVAL;
    812 
    813 			cmp = 1;
    814 		} else {
    815 			if (cmd == SIOCGLIFADDR) {
    816 				/* on getting an address, take the 1st match */
    817 				cmp = 0;	/*XXX*/
    818 			} else {
    819 				/* on deleting an address, do exact match */
    820 				in_len2mask(&mask, 32);
    821 				sin = (struct sockaddr_in *)&iflr->addr;
    822 				match.s_addr = sin->sin_addr.s_addr;
    823 
    824 				cmp = 1;
    825 			}
    826 		}
    827 
    828 		IFADDR_FOREACH(ifa, ifp) {
    829 			if (ifa->ifa_addr->sa_family != AF_INET)
    830 				continue;
    831 			if (cmp == 0)
    832 				break;
    833 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
    834 			candidate.s_addr &= mask.s_addr;
    835 			if (candidate.s_addr == match.s_addr)
    836 				break;
    837 		}
    838 		if (ifa == NULL)
    839 			return EADDRNOTAVAIL;
    840 		ia = (struct in_ifaddr *)ifa;
    841 
    842 		if (cmd == SIOCGLIFADDR) {
    843 			/* fill in the if_laddrreq structure */
    844 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
    845 
    846 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    847 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
    848 					ia->ia_dstaddr.sin_len);
    849 			} else
    850 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
    851 
    852 			iflr->prefixlen =
    853 				in_mask2len(&ia->ia_sockmask.sin_addr);
    854 
    855 			iflr->flags = 0;	/*XXX*/
    856 
    857 			return 0;
    858 		} else {
    859 			struct in_aliasreq ifra;
    860 
    861 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
    862 			memset(&ifra, 0, sizeof(ifra));
    863 			memcpy(ifra.ifra_name, iflr->iflr_name,
    864 				sizeof(ifra.ifra_name));
    865 
    866 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
    867 				ia->ia_addr.sin_len);
    868 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    869 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
    870 					ia->ia_dstaddr.sin_len);
    871 			}
    872 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
    873 				ia->ia_sockmask.sin_len);
    874 
    875 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
    876 		}
    877 	    }
    878 	}
    879 
    880 	return EOPNOTSUPP;	/*just for safety*/
    881 }
    882 
    883 /*
    884  * Delete any existing route for an interface.
    885  */
    886 void
    887 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
    888 {
    889 
    890 	in_scrubprefix(ia);
    891 }
    892 
    893 /*
    894  * Initialize an interface's internet address
    895  * and routing table entry.
    896  */
    897 int
    898 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
    899     const struct sockaddr_in *sin, int scrub, int hostIsNew)
    900 {
    901 	u_int32_t i;
    902 	struct sockaddr_in oldaddr;
    903 	int s = splnet(), flags = RTF_UP, error;
    904 
    905 	if (sin == NULL)
    906 		sin = &ia->ia_addr;
    907 
    908 	/*
    909 	 * Set up new addresses.
    910 	 */
    911 	oldaddr = ia->ia_addr;
    912 	if (ia->ia_addr.sin_family == AF_INET)
    913 		LIST_REMOVE(ia, ia_hash);
    914 	ia->ia_addr = *sin;
    915 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
    916 
    917 	/* Set IN_IFF flags early for if_addr_init() */
    918 	if (hostIsNew && if_do_dad(ifp) && !in_nullhost(ia->ia_addr.sin_addr)) {
    919 		if (ifp->if_link_state == LINK_STATE_DOWN)
    920 			ia->ia4_flags |= IN_IFF_DETACHED;
    921 		else
    922 			/* State the intent to try DAD if possible */
    923 			ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
    924 	}
    925 
    926 	/*
    927 	 * Give the interface a chance to initialize
    928 	 * if this is its first address,
    929 	 * and to validate the address if necessary.
    930 	 */
    931 	if ((error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0)
    932 		goto bad;
    933 	/* Now clear the try tentative flag, it's job is done. */
    934 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
    935 	splx(s);
    936 
    937 	if (scrub) {
    938 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
    939 		in_ifscrub(ifp, ia);
    940 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    941 	}
    942 
    943 	/* Add the local route to the address */
    944 	in_ifaddlocal(&ia->ia_ifa);
    945 
    946 	i = ia->ia_addr.sin_addr.s_addr;
    947 	if (IN_CLASSA(i))
    948 		ia->ia_netmask = IN_CLASSA_NET;
    949 	else if (IN_CLASSB(i))
    950 		ia->ia_netmask = IN_CLASSB_NET;
    951 	else
    952 		ia->ia_netmask = IN_CLASSC_NET;
    953 	/*
    954 	 * The subnet mask usually includes at least the standard network part,
    955 	 * but may may be smaller in the case of supernetting.
    956 	 * If it is set, we believe it.
    957 	 */
    958 	if (ia->ia_subnetmask == 0) {
    959 		ia->ia_subnetmask = ia->ia_netmask;
    960 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
    961 	} else
    962 		ia->ia_netmask &= ia->ia_subnetmask;
    963 
    964 	ia->ia_net = i & ia->ia_netmask;
    965 	ia->ia_subnet = i & ia->ia_subnetmask;
    966 	in_socktrim(&ia->ia_sockmask);
    967 	/* re-calculate the "in_maxmtu" value */
    968 	in_setmaxmtu();
    969 	/*
    970 	 * Add route for the network.
    971 	 */
    972 	ia->ia_ifa.ifa_metric = ifp->if_metric;
    973 	if (ifp->if_flags & IFF_BROADCAST) {
    974 		ia->ia_broadaddr.sin_addr.s_addr =
    975 			ia->ia_subnet | ~ia->ia_subnetmask;
    976 		ia->ia_netbroadcast.s_addr =
    977 			ia->ia_net | ~ia->ia_netmask;
    978 	} else if (ifp->if_flags & IFF_LOOPBACK) {
    979 		ia->ia_dstaddr = ia->ia_addr;
    980 		flags |= RTF_HOST;
    981 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
    982 		if (ia->ia_dstaddr.sin_family != AF_INET)
    983 			return (0);
    984 		flags |= RTF_HOST;
    985 	}
    986 	error = in_addprefix(ia, flags);
    987 	/*
    988 	 * If the interface supports multicast, join the "all hosts"
    989 	 * multicast group on that interface.
    990 	 */
    991 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
    992 		struct in_addr addr;
    993 
    994 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
    995 		ia->ia_allhosts = in_addmulti(&addr, ifp);
    996 	}
    997 
    998 	if (hostIsNew && if_do_dad(ifp) &&
    999 	    !in_nullhost(ia->ia_addr.sin_addr) &&
   1000 	    ia->ia4_flags & IN_IFF_TENTATIVE)
   1001 		ia->ia_dad_start((struct ifaddr *)ia);
   1002 
   1003 	return (error);
   1004 bad:
   1005 	splx(s);
   1006 	LIST_REMOVE(ia, ia_hash);
   1007 	ia->ia_addr = oldaddr;
   1008 	if (ia->ia_addr.sin_family == AF_INET)
   1009 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
   1010 		    ia, ia_hash);
   1011 	return (error);
   1012 }
   1013 
   1014 #define rtinitflags(x) \
   1015 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
   1016 	    ? RTF_HOST : 0)
   1017 
   1018 /*
   1019  * add a route to prefix ("connected route" in cisco terminology).
   1020  * does nothing if there's some interface address with the same prefix already.
   1021  */
   1022 static int
   1023 in_addprefix(struct in_ifaddr *target, int flags)
   1024 {
   1025 	struct in_ifaddr *ia;
   1026 	struct in_addr prefix, mask, p;
   1027 	int error;
   1028 
   1029 	if ((flags & RTF_HOST) != 0)
   1030 		prefix = target->ia_dstaddr.sin_addr;
   1031 	else {
   1032 		prefix = target->ia_addr.sin_addr;
   1033 		mask = target->ia_sockmask.sin_addr;
   1034 		prefix.s_addr &= mask.s_addr;
   1035 	}
   1036 
   1037 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1038 		if (rtinitflags(ia))
   1039 			p = ia->ia_dstaddr.sin_addr;
   1040 		else {
   1041 			p = ia->ia_addr.sin_addr;
   1042 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1043 		}
   1044 
   1045 		if (prefix.s_addr != p.s_addr)
   1046 			continue;
   1047 
   1048 		/*
   1049 		 * if we got a matching prefix route inserted by other
   1050 		 * interface address, we don't need to bother
   1051 		 *
   1052 		 * XXX RADIX_MPATH implications here? -dyoung
   1053 		 */
   1054 		if (ia->ia_flags & IFA_ROUTE)
   1055 			return 0;
   1056 	}
   1057 
   1058 	/*
   1059 	 * noone seem to have prefix route.  insert it.
   1060 	 */
   1061 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
   1062 	if (error == 0)
   1063 		target->ia_flags |= IFA_ROUTE;
   1064 	else if (error == EEXIST) {
   1065 		/*
   1066 		 * the fact the route already exists is not an error.
   1067 		 */
   1068 		error = 0;
   1069 	}
   1070 	return error;
   1071 }
   1072 
   1073 /*
   1074  * remove a route to prefix ("connected route" in cisco terminology).
   1075  * re-installs the route by using another interface address, if there's one
   1076  * with the same prefix (otherwise we lose the route mistakenly).
   1077  */
   1078 static int
   1079 in_scrubprefix(struct in_ifaddr *target)
   1080 {
   1081 	struct in_ifaddr *ia;
   1082 	struct in_addr prefix, mask, p;
   1083 	int error;
   1084 
   1085 	/* If we don't have IFA_ROUTE we should still inform userland */
   1086 	if ((target->ia_flags & IFA_ROUTE) == 0)
   1087 		return 0;
   1088 
   1089 	if (rtinitflags(target))
   1090 		prefix = target->ia_dstaddr.sin_addr;
   1091 	else {
   1092 		prefix = target->ia_addr.sin_addr;
   1093 		mask = target->ia_sockmask.sin_addr;
   1094 		prefix.s_addr &= mask.s_addr;
   1095 	}
   1096 
   1097 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1098 		if (rtinitflags(ia))
   1099 			p = ia->ia_dstaddr.sin_addr;
   1100 		else {
   1101 			p = ia->ia_addr.sin_addr;
   1102 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1103 		}
   1104 
   1105 		if (prefix.s_addr != p.s_addr)
   1106 			continue;
   1107 
   1108 		/*
   1109 		 * if we got a matching prefix route, move IFA_ROUTE to him
   1110 		 */
   1111 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
   1112 			rtinit(&target->ia_ifa, RTM_DELETE,
   1113 			    rtinitflags(target));
   1114 			target->ia_flags &= ~IFA_ROUTE;
   1115 
   1116 			error = rtinit(&ia->ia_ifa, RTM_ADD,
   1117 			    rtinitflags(ia) | RTF_UP);
   1118 			if (error == 0)
   1119 				ia->ia_flags |= IFA_ROUTE;
   1120 			return error;
   1121 		}
   1122 	}
   1123 
   1124 	/*
   1125 	 * noone seem to have prefix route.  remove it.
   1126 	 */
   1127 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
   1128 	target->ia_flags &= ~IFA_ROUTE;
   1129 	return 0;
   1130 }
   1131 
   1132 #undef rtinitflags
   1133 
   1134 /*
   1135  * Return 1 if the address might be a local broadcast address.
   1136  */
   1137 int
   1138 in_broadcast(struct in_addr in, struct ifnet *ifp)
   1139 {
   1140 	struct ifaddr *ifa;
   1141 
   1142 	if (in.s_addr == INADDR_BROADCAST ||
   1143 	    in_nullhost(in))
   1144 		return 1;
   1145 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
   1146 		return 0;
   1147 	/*
   1148 	 * Look through the list of addresses for a match
   1149 	 * with a broadcast address.
   1150 	 */
   1151 #define ia (ifatoia(ifa))
   1152 	IFADDR_FOREACH(ifa, ifp)
   1153 		if (ifa->ifa_addr->sa_family == AF_INET &&
   1154 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
   1155 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
   1156 		     in_hosteq(in, ia->ia_netbroadcast) ||
   1157 		     (hostzeroisbroadcast &&
   1158 		      /*
   1159 		       * Check for old-style (host 0) broadcast.
   1160 		       */
   1161 		      (in.s_addr == ia->ia_subnet ||
   1162 		       in.s_addr == ia->ia_net))))
   1163 			return 1;
   1164 	return (0);
   1165 #undef ia
   1166 }
   1167 
   1168 /*
   1169  * perform DAD when interface becomes IFF_UP.
   1170  */
   1171 void
   1172 in_if_link_up(struct ifnet *ifp)
   1173 {
   1174 	struct ifaddr *ifa;
   1175 	struct in_ifaddr *ia;
   1176 
   1177 	/* Ensure it's sane to run DAD */
   1178 	if (ifp->if_link_state == LINK_STATE_DOWN)
   1179 		return;
   1180 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
   1181 		return;
   1182 
   1183 	IFADDR_FOREACH(ifa, ifp) {
   1184 		if (ifa->ifa_addr->sa_family != AF_INET)
   1185 			continue;
   1186 		ia = (struct in_ifaddr *)ifa;
   1187 
   1188 		/* If detached then mark as tentative */
   1189 		if (ia->ia4_flags & IN_IFF_DETACHED) {
   1190 			ia->ia4_flags &= ~IN_IFF_DETACHED;
   1191 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
   1192 				ia->ia4_flags |= IN_IFF_TENTATIVE;
   1193 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
   1194 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1195 		}
   1196 
   1197 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
   1198 			/* Clear the duplicated flag as we're starting DAD. */
   1199 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
   1200 			ia->ia_dad_start(ifa);
   1201 		}
   1202 	}
   1203 }
   1204 
   1205 void
   1206 in_if_up(struct ifnet *ifp)
   1207 {
   1208 
   1209 	/* interface may not support link state, so bring it up also */
   1210 	in_if_link_up(ifp);
   1211 }
   1212 
   1213 /*
   1214  * Mark all addresses as detached.
   1215  */
   1216 void
   1217 in_if_link_down(struct ifnet *ifp)
   1218 {
   1219 	struct ifaddr *ifa;
   1220 	struct in_ifaddr *ia;
   1221 
   1222 	IFADDR_FOREACH(ifa, ifp) {
   1223 		if (ifa->ifa_addr->sa_family != AF_INET)
   1224 			continue;
   1225 		ia = (struct in_ifaddr *)ifa;
   1226 
   1227 		/* Stop DAD processing */
   1228 		if (ia->ia_dad_stop != NULL)
   1229 			ia->ia_dad_stop(ifa);
   1230 
   1231 		/*
   1232 		 * Mark the address as detached.
   1233 		 */
   1234 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
   1235 			ia->ia4_flags |= IN_IFF_DETACHED;
   1236 			ia->ia4_flags &=
   1237 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
   1238 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1239 		}
   1240 	}
   1241 }
   1242 
   1243 void
   1244 in_if_down(struct ifnet *ifp)
   1245 {
   1246 
   1247 	in_if_link_down(ifp);
   1248 }
   1249 
   1250 void
   1251 in_if_link_state_change(struct ifnet *ifp, int link_state)
   1252 {
   1253 
   1254 	switch (link_state) {
   1255 	case LINK_STATE_DOWN:
   1256 		in_if_link_down(ifp);
   1257 		break;
   1258 	case LINK_STATE_UP:
   1259 		in_if_link_up(ifp);
   1260 		break;
   1261 	}
   1262 }
   1263 
   1264 /*
   1265  * in_lookup_multi: look up the in_multi record for a given IP
   1266  * multicast address on a given interface.  If no matching record is
   1267  * found, return NULL.
   1268  */
   1269 struct in_multi *
   1270 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
   1271 {
   1272 	struct in_multi *inm;
   1273 
   1274 	KASSERT(rw_lock_held(&in_multilock));
   1275 
   1276 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
   1277 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
   1278 			break;
   1279 	}
   1280 	return inm;
   1281 }
   1282 
   1283 /*
   1284  * in_multi_group: check whether the address belongs to an IP multicast
   1285  * group we are joined on this interface.  Returns true or false.
   1286  */
   1287 bool
   1288 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
   1289 {
   1290 	bool ingroup;
   1291 
   1292 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
   1293 		rw_enter(&in_multilock, RW_READER);
   1294 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1295 		rw_exit(&in_multilock);
   1296 	} else {
   1297 		/* XXX Recursive call from ip_output(). */
   1298 		KASSERT(rw_lock_held(&in_multilock));
   1299 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1300 	}
   1301 	return ingroup;
   1302 }
   1303 
   1304 /*
   1305  * Add an address to the list of IP multicast addresses for a given interface.
   1306  */
   1307 struct in_multi *
   1308 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
   1309 {
   1310 	struct sockaddr_in sin;
   1311 	struct in_multi *inm;
   1312 
   1313 	/*
   1314 	 * See if address already in list.
   1315 	 */
   1316 	rw_enter(&in_multilock, RW_WRITER);
   1317 	inm = in_lookup_multi(*ap, ifp);
   1318 	if (inm != NULL) {
   1319 		/*
   1320 		 * Found it; just increment the reference count.
   1321 		 */
   1322 		inm->inm_refcount++;
   1323 		rw_exit(&in_multilock);
   1324 		return inm;
   1325 	}
   1326 
   1327 	/*
   1328 	 * New address; allocate a new multicast record.
   1329 	 */
   1330 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
   1331 	if (inm == NULL) {
   1332 		rw_exit(&in_multilock);
   1333 		return NULL;
   1334 	}
   1335 	inm->inm_addr = *ap;
   1336 	inm->inm_ifp = ifp;
   1337 	inm->inm_refcount = 1;
   1338 
   1339 	/*
   1340 	 * Ask the network driver to update its multicast reception
   1341 	 * filter appropriately for the new address.
   1342 	 */
   1343 	sockaddr_in_init(&sin, ap, 0);
   1344 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
   1345 		rw_exit(&in_multilock);
   1346 		pool_put(&inmulti_pool, inm);
   1347 		return NULL;
   1348 	}
   1349 
   1350 	/*
   1351 	 * Let IGMP know that we have joined a new IP multicast group.
   1352 	 */
   1353 	if (igmp_joingroup(inm) != 0) {
   1354 		rw_exit(&in_multilock);
   1355 		pool_put(&inmulti_pool, inm);
   1356 		return NULL;
   1357 	}
   1358 	LIST_INSERT_HEAD(
   1359 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
   1360 	    inm, inm_list);
   1361 	in_multientries++;
   1362 	rw_exit(&in_multilock);
   1363 
   1364 	return inm;
   1365 }
   1366 
   1367 /*
   1368  * Delete a multicast address record.
   1369  */
   1370 void
   1371 in_delmulti(struct in_multi *inm)
   1372 {
   1373 	struct sockaddr_in sin;
   1374 
   1375 	rw_enter(&in_multilock, RW_WRITER);
   1376 	if (--inm->inm_refcount > 0) {
   1377 		rw_exit(&in_multilock);
   1378 		return;
   1379 	}
   1380 
   1381 	/*
   1382 	 * No remaining claims to this record; let IGMP know that
   1383 	 * we are leaving the multicast group.
   1384 	 */
   1385 	igmp_leavegroup(inm);
   1386 
   1387 	/*
   1388 	 * Notify the network driver to update its multicast reception
   1389 	 * filter.
   1390 	 */
   1391 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
   1392 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
   1393 
   1394 	/*
   1395 	 * Unlink from list.
   1396 	 */
   1397 	LIST_REMOVE(inm, inm_list);
   1398 	in_multientries--;
   1399 	rw_exit(&in_multilock);
   1400 
   1401 	pool_put(&inmulti_pool, inm);
   1402 }
   1403 
   1404 /*
   1405  * in_next_multi: step through all of the in_multi records, one at a time.
   1406  * The current position is remembered in "step", which the caller must
   1407  * provide.  in_first_multi(), below, must be called to initialize "step"
   1408  * and get the first record.  Both macros return a NULL "inm" when there
   1409  * are no remaining records.
   1410  */
   1411 struct in_multi *
   1412 in_next_multi(struct in_multistep *step)
   1413 {
   1414 	struct in_multi *inm;
   1415 
   1416 	KASSERT(rw_lock_held(&in_multilock));
   1417 
   1418 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
   1419 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
   1420 	}
   1421 	if ((inm = step->i_inm) != NULL) {
   1422 		step->i_inm = LIST_NEXT(inm, inm_list);
   1423 	}
   1424 	return inm;
   1425 }
   1426 
   1427 struct in_multi *
   1428 in_first_multi(struct in_multistep *step)
   1429 {
   1430 	KASSERT(rw_lock_held(&in_multilock));
   1431 
   1432 	step->i_n = 0;
   1433 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
   1434 	return in_next_multi(step);
   1435 }
   1436 
   1437 void
   1438 in_multi_lock(int op)
   1439 {
   1440 	rw_enter(&in_multilock, op);
   1441 }
   1442 
   1443 void
   1444 in_multi_unlock(void)
   1445 {
   1446 	rw_exit(&in_multilock);
   1447 }
   1448 
   1449 int
   1450 in_multi_lock_held(void)
   1451 {
   1452 	return rw_lock_held(&in_multilock);
   1453 }
   1454 
   1455 struct sockaddr_in *
   1456 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
   1457     int soopts, struct ip_moptions *mopts, int *errorp)
   1458 {
   1459 	struct rtentry *rt = NULL;
   1460 	struct in_ifaddr *ia = NULL;
   1461 
   1462 	/*
   1463          * If route is known or can be allocated now, take the
   1464          * source address from the interface.  Otherwise, punt.
   1465 	 */
   1466 	if ((soopts & SO_DONTROUTE) != 0)
   1467 		rtcache_free(ro);
   1468 	else {
   1469 		union {
   1470 			struct sockaddr		dst;
   1471 			struct sockaddr_in	dst4;
   1472 		} u;
   1473 
   1474 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
   1475 		rt = rtcache_lookup(ro, &u.dst);
   1476 	}
   1477 	/*
   1478 	 * If we found a route, use the address
   1479 	 * corresponding to the outgoing interface
   1480 	 * unless it is the loopback (in case a route
   1481 	 * to our address on another net goes to loopback).
   1482 	 *
   1483 	 * XXX Is this still true?  Do we care?
   1484 	 */
   1485 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1486 		ia = ifatoia(rt->rt_ifa);
   1487 	if (ia == NULL) {
   1488 		u_int16_t fport = sin->sin_port;
   1489 
   1490 		sin->sin_port = 0;
   1491 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
   1492 		sin->sin_port = fport;
   1493 		if (ia == NULL) {
   1494 			/* Find 1st non-loopback AF_INET address */
   1495 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1496 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
   1497 					break;
   1498 			}
   1499 		}
   1500 		if (ia == NULL) {
   1501 			*errorp = EADDRNOTAVAIL;
   1502 			return NULL;
   1503 		}
   1504 	}
   1505 	/*
   1506 	 * If the destination address is multicast and an outgoing
   1507 	 * interface has been set as a multicast option, use the
   1508 	 * address of that interface as our source address.
   1509 	 */
   1510 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
   1511 		struct ip_moptions *imo;
   1512 		struct ifnet *ifp;
   1513 
   1514 		imo = mopts;
   1515 		if (imo->imo_multicast_ifp != NULL) {
   1516 			ifp = imo->imo_multicast_ifp;
   1517 			IFP_TO_IA(ifp, ia);		/* XXX */
   1518 			if (ia == 0 || ia->ia4_flags & IN_IFF_NOTREADY) {
   1519 				*errorp = EADDRNOTAVAIL;
   1520 				return NULL;
   1521 			}
   1522 		}
   1523 	}
   1524 	if (ia->ia_ifa.ifa_getifa != NULL) {
   1525 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
   1526 		                                      sintosa(sin)));
   1527 		if (ia == NULL) {
   1528 			*errorp = EADDRNOTAVAIL;
   1529 			return NULL;
   1530 		}
   1531 	}
   1532 #ifdef GETIFA_DEBUG
   1533 	else
   1534 		printf("%s: missing ifa_getifa\n", __func__);
   1535 #endif
   1536 	return satosin(&ia->ia_addr);
   1537 }
   1538 
   1539 #if NETHER > 0
   1540 
   1541 struct in_llentry {
   1542 	struct llentry		base;
   1543 };
   1544 
   1545 #define	IN_LLTBL_DEFAULT_HSIZE	32
   1546 #define	IN_LLTBL_HASH(k, h) \
   1547 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
   1548 
   1549 /*
   1550  * Do actual deallocation of @lle.
   1551  * Called by LLE_FREE_LOCKED when number of references
   1552  * drops to zero.
   1553  */
   1554 static void
   1555 in_lltable_destroy_lle(struct llentry *lle)
   1556 {
   1557 
   1558 	LLE_WUNLOCK(lle);
   1559 	LLE_LOCK_DESTROY(lle);
   1560 	kmem_intr_free(lle, sizeof(*lle));
   1561 }
   1562 
   1563 static struct llentry *
   1564 in_lltable_new(struct in_addr addr4, u_int flags)
   1565 {
   1566 	struct in_llentry *lle;
   1567 
   1568 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
   1569 	if (lle == NULL)		/* NB: caller generates msg */
   1570 		return NULL;
   1571 
   1572 	/*
   1573 	 * For IPv4 this will trigger "arpresolve" to generate
   1574 	 * an ARP request.
   1575 	 */
   1576 	lle->base.la_expire = time_uptime; /* mark expired */
   1577 	lle->base.r_l3addr.addr4 = addr4;
   1578 	lle->base.lle_refcnt = 1;
   1579 	lle->base.lle_free = in_lltable_destroy_lle;
   1580 	LLE_LOCK_INIT(&lle->base);
   1581 	callout_init(&lle->base.la_timer, CALLOUT_MPSAFE);
   1582 
   1583 	return (&lle->base);
   1584 }
   1585 
   1586 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
   1587 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
   1588 
   1589 static int
   1590 in_lltable_match_prefix(const struct sockaddr *prefix,
   1591     const struct sockaddr *mask, u_int flags, struct llentry *lle)
   1592 {
   1593 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
   1594 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
   1595 
   1596 	/*
   1597 	 * (flags & LLE_STATIC) means deleting all entries
   1598 	 * including static ARP entries.
   1599 	 */
   1600 	if (IN_ARE_MASKED_ADDR_EQUAL(lle->r_l3addr.addr4, pfx, msk) &&
   1601 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
   1602 		return (1);
   1603 
   1604 	return (0);
   1605 }
   1606 
   1607 static void
   1608 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
   1609 {
   1610 	struct ifnet *ifp __diagused;
   1611 	size_t pkts_dropped;
   1612 
   1613 	LLE_WLOCK_ASSERT(lle);
   1614 	KASSERT(llt != NULL);
   1615 
   1616 	/* Unlink entry from table if not already */
   1617 	if ((lle->la_flags & LLE_LINKED) != 0) {
   1618 		ifp = llt->llt_ifp;
   1619 		IF_AFDATA_WLOCK_ASSERT(ifp);
   1620 		lltable_unlink_entry(llt, lle);
   1621 	}
   1622 
   1623 	/* cancel timer */
   1624 	if (callout_stop(&lle->lle_timer))
   1625 		LLE_REMREF(lle);
   1626 
   1627 	/* Drop hold queue */
   1628 	pkts_dropped = llentry_free(lle);
   1629 	arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped);
   1630 }
   1631 
   1632 static int
   1633 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr)
   1634 {
   1635 	struct rtentry *rt;
   1636 	int error = EINVAL;
   1637 
   1638 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1639 	    "sin_family %d", l3addr->sa_family);
   1640 
   1641 	rt = rtalloc1(l3addr, 0);
   1642 	if (rt == NULL)
   1643 		return error;
   1644 
   1645 	/*
   1646 	 * If the gateway for an existing host route matches the target L3
   1647 	 * address, which is a special route inserted by some implementation
   1648 	 * such as MANET, and the interface is of the correct type, then
   1649 	 * allow for ARP to proceed.
   1650 	 */
   1651 	if (rt->rt_flags & RTF_GATEWAY) {
   1652 		if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp ||
   1653 		    rt->rt_ifp->if_type != IFT_ETHER ||
   1654 #ifdef __FreeBSD__
   1655 		    (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 ||
   1656 #else
   1657 		    (rt->rt_ifp->if_flags & IFF_NOARP) != 0 ||
   1658 #endif
   1659 		    memcmp(rt->rt_gateway->sa_data, l3addr->sa_data,
   1660 		    sizeof(in_addr_t)) != 0) {
   1661 			goto error;
   1662 		}
   1663 	}
   1664 
   1665 	/*
   1666 	 * Make sure that at least the destination address is covered
   1667 	 * by the route. This is for handling the case where 2 or more
   1668 	 * interfaces have the same prefix. An incoming packet arrives
   1669 	 * on one interface and the corresponding outgoing packet leaves
   1670 	 * another interface.
   1671 	 */
   1672 	if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) {
   1673 		const char *sa, *mask, *addr, *lim;
   1674 		int len;
   1675 
   1676 		mask = (const char *)rt_mask(rt);
   1677 		/*
   1678 		 * Just being extra cautious to avoid some custom
   1679 		 * code getting into trouble.
   1680 		 */
   1681 		if (mask == NULL)
   1682 			goto error;
   1683 
   1684 		sa = (const char *)rt_getkey(rt);
   1685 		addr = (const char *)l3addr;
   1686 		len = ((const struct sockaddr_in *)l3addr)->sin_len;
   1687 		lim = addr + len;
   1688 
   1689 		for ( ; addr < lim; sa++, mask++, addr++) {
   1690 			if ((*sa ^ *addr) & *mask) {
   1691 #ifdef DIAGNOSTIC
   1692 				log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n",
   1693 				    inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr));
   1694 #endif
   1695 				goto error;
   1696 			}
   1697 		}
   1698 	}
   1699 
   1700 	error = 0;
   1701 error:
   1702 	return error;
   1703 }
   1704 
   1705 static inline uint32_t
   1706 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
   1707 {
   1708 
   1709 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
   1710 }
   1711 
   1712 static uint32_t
   1713 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
   1714 {
   1715 
   1716 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
   1717 }
   1718 
   1719 static void
   1720 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
   1721 {
   1722 	struct sockaddr_in *sin;
   1723 
   1724 	sin = (struct sockaddr_in *)sa;
   1725 	memset(sin, 0, sizeof(*sin));
   1726 	sin->sin_family = AF_INET;
   1727 	sin->sin_len = sizeof(*sin);
   1728 	sin->sin_addr = lle->r_l3addr.addr4;
   1729 }
   1730 
   1731 static inline struct llentry *
   1732 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
   1733 {
   1734 	struct llentry *lle;
   1735 	struct llentries *lleh;
   1736 	u_int hashidx;
   1737 
   1738 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
   1739 	lleh = &llt->lle_head[hashidx];
   1740 	LIST_FOREACH(lle, lleh, lle_next) {
   1741 		if (lle->la_flags & LLE_DELETED)
   1742 			continue;
   1743 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
   1744 			break;
   1745 	}
   1746 
   1747 	return (lle);
   1748 }
   1749 
   1750 static int
   1751 in_lltable_delete(struct lltable *llt, u_int flags,
   1752     const struct sockaddr *l3addr)
   1753 {
   1754 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1755 	struct ifnet *ifp __diagused = llt->llt_ifp;
   1756 	struct llentry *lle;
   1757 
   1758 	IF_AFDATA_WLOCK_ASSERT(ifp);
   1759 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1760 	    "sin_family %d", l3addr->sa_family);
   1761 
   1762 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1763 	if (lle == NULL) {
   1764 #ifdef DIAGNOSTIC
   1765 		log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
   1766 #endif
   1767 		return (ENOENT);
   1768 	}
   1769 
   1770 	if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
   1771 		LLE_WLOCK(lle);
   1772 		lle->la_flags |= LLE_DELETED;
   1773 #ifdef DIAGNOSTIC
   1774 		log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
   1775 #endif
   1776 		if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
   1777 			llentry_free(lle);
   1778 		else
   1779 			LLE_WUNLOCK(lle);
   1780 	}
   1781 
   1782 	return (0);
   1783 }
   1784 
   1785 static struct llentry *
   1786 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
   1787 {
   1788 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1789 	struct ifnet *ifp = llt->llt_ifp;
   1790 	struct llentry *lle;
   1791 
   1792 	IF_AFDATA_WLOCK_ASSERT(ifp);
   1793 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1794 	    "sin_family %d", l3addr->sa_family);
   1795 
   1796 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1797 
   1798 	if (lle != NULL) {
   1799 		LLE_WLOCK(lle);
   1800 		return (lle);
   1801 	}
   1802 
   1803 	/* no existing record, we need to create new one */
   1804 
   1805 	/*
   1806 	 * A route that covers the given address must have
   1807 	 * been installed 1st because we are doing a resolution,
   1808 	 * verify this.
   1809 	 */
   1810 	if (!(flags & LLE_IFADDR) &&
   1811 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
   1812 		return (NULL);
   1813 
   1814 	lle = in_lltable_new(sin->sin_addr, flags);
   1815 	if (lle == NULL) {
   1816 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
   1817 		return (NULL);
   1818 	}
   1819 	lle->la_flags = flags;
   1820 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
   1821 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
   1822 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
   1823 	}
   1824 
   1825 	lltable_link_entry(llt, lle);
   1826 	LLE_WLOCK(lle);
   1827 
   1828 	return (lle);
   1829 }
   1830 
   1831 /*
   1832  * Return NULL if not found or marked for deletion.
   1833  * If found return lle read locked.
   1834  */
   1835 static struct llentry *
   1836 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
   1837 {
   1838 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1839 	struct llentry *lle;
   1840 
   1841 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
   1842 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1843 	    "sin_family %d", l3addr->sa_family);
   1844 
   1845 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1846 
   1847 	if (lle == NULL)
   1848 		return NULL;
   1849 
   1850 	if (flags & LLE_EXCLUSIVE)
   1851 		LLE_WLOCK(lle);
   1852 	else
   1853 		LLE_RLOCK(lle);
   1854 
   1855 	return lle;
   1856 }
   1857 
   1858 #endif /* NETHER > 0 */
   1859 
   1860 static void
   1861 in_sysctl_init(struct sysctllog **clog)
   1862 {
   1863 	sysctl_createv(clog, 0, NULL, NULL,
   1864 		       CTLFLAG_PERMANENT,
   1865 		       CTLTYPE_NODE, "inet",
   1866 		       SYSCTL_DESCR("PF_INET related settings"),
   1867 		       NULL, 0, NULL, 0,
   1868 		       CTL_NET, PF_INET, CTL_EOL);
   1869 	sysctl_createv(clog, 0, NULL, NULL,
   1870 		       CTLFLAG_PERMANENT,
   1871 		       CTLTYPE_NODE, "ip",
   1872 		       SYSCTL_DESCR("IPv4 related settings"),
   1873 		       NULL, 0, NULL, 0,
   1874 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
   1875 
   1876 	sysctl_createv(clog, 0, NULL, NULL,
   1877 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1878 		       CTLTYPE_INT, "subnetsarelocal",
   1879 		       SYSCTL_DESCR("Whether logical subnets are considered "
   1880 				    "local"),
   1881 		       NULL, 0, &subnetsarelocal, 0,
   1882 		       CTL_NET, PF_INET, IPPROTO_IP,
   1883 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
   1884 	sysctl_createv(clog, 0, NULL, NULL,
   1885 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1886 		       CTLTYPE_INT, "hostzerobroadcast",
   1887 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
   1888 		       NULL, 0, &hostzeroisbroadcast, 0,
   1889 		       CTL_NET, PF_INET, IPPROTO_IP,
   1890 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
   1891 }
   1892 
   1893 #if NETHER > 0
   1894 static struct lltable *
   1895 in_lltattach(struct ifnet *ifp)
   1896 {
   1897 	struct lltable *llt;
   1898 
   1899 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
   1900 	llt->llt_af = AF_INET;
   1901 	llt->llt_ifp = ifp;
   1902 
   1903 	llt->llt_lookup = in_lltable_lookup;
   1904 	llt->llt_create = in_lltable_create;
   1905 	llt->llt_delete = in_lltable_delete;
   1906 #if 0
   1907 	llt->llt_dump_entry = in_lltable_dump_entry;
   1908 #endif
   1909 	llt->llt_hash = in_lltable_hash;
   1910 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
   1911 	llt->llt_free_entry = in_lltable_free_entry;
   1912 	llt->llt_match_prefix = in_lltable_match_prefix;
   1913 	lltable_link(llt);
   1914 
   1915 	return (llt);
   1916 }
   1917 #endif /* NETHER > 0 */
   1918 
   1919 void *
   1920 in_domifattach(struct ifnet *ifp)
   1921 {
   1922 	struct in_ifinfo *ii;
   1923 
   1924 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
   1925 	KASSERT(ii != NULL);
   1926 
   1927 #if NETHER > 0
   1928 	ii->ii_llt = in_lltattach(ifp);
   1929 #endif
   1930 
   1931 #ifdef IPSELSRC
   1932 	ii->ii_selsrc = in_selsrc_domifattach(ifp);
   1933 	KASSERT(ii->ii_selsrc != NULL);
   1934 #endif
   1935 
   1936 	return ii;
   1937 }
   1938 
   1939 void
   1940 in_domifdetach(struct ifnet *ifp, void *aux)
   1941 {
   1942 	struct in_ifinfo *ii = aux;
   1943 
   1944 #ifdef IPSELSRC
   1945 	in_selsrc_domifdetach(ifp, ii->ii_selsrc);
   1946 #endif
   1947 #if NETHER > 0
   1948 	lltable_free(ii->ii_llt);
   1949 #endif
   1950 	kmem_free(ii, sizeof(struct in_ifinfo));
   1951 }
   1952