Home | History | Annotate | Line # | Download | only in netinet
in.c revision 1.158
      1 /*	$NetBSD: in.c,v 1.158 2015/08/31 08:02:44 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 1.158 2015/08/31 08:02:44 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_ether.h>
    125 #include <net/if_types.h>
    126 #include <net/if_llatbl.h>
    127 #include <net/if_dl.h>
    128 
    129 #include <netinet/in_systm.h>
    130 #include <netinet/in.h>
    131 #include <netinet/in_var.h>
    132 #include <netinet/ip.h>
    133 #include <netinet/ip_var.h>
    134 #include <netinet/in_ifattach.h>
    135 #include <netinet/in_pcb.h>
    136 #include <netinet/in_selsrc.h>
    137 #include <netinet/if_inarp.h>
    138 #include <netinet/ip_mroute.h>
    139 #include <netinet/igmp_var.h>
    140 
    141 #ifdef IPSELSRC
    142 #include <netinet/in_selsrc.h>
    143 #endif
    144 
    145 static u_int	in_mask2len(struct in_addr *);
    146 static void	in_len2mask(struct in_addr *, u_int);
    147 static int	in_lifaddr_ioctl(struct socket *, u_long, void *,
    148 	struct ifnet *);
    149 
    150 static int	in_addprefix(struct in_ifaddr *, int);
    151 static int	in_scrubprefix(struct in_ifaddr *);
    152 static void	in_sysctl_init(struct sysctllog **);
    153 
    154 #ifndef SUBNETSARELOCAL
    155 #define	SUBNETSARELOCAL	1
    156 #endif
    157 
    158 #ifndef HOSTZEROBROADCAST
    159 #define HOSTZEROBROADCAST 1
    160 #endif
    161 
    162 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
    163 #ifndef IN_MULTI_HASH_SIZE
    164 #define IN_MULTI_HASH_SIZE	509
    165 #endif
    166 
    167 static int			subnetsarelocal = SUBNETSARELOCAL;
    168 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
    169 
    170 /*
    171  * This list is used to keep track of in_multi chains which belong to
    172  * deleted interface addresses.  We use in_ifaddr so that a chain head
    173  * won't be deallocated until all multicast address record are deleted.
    174  */
    175 
    176 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
    177 
    178 static struct pool		inmulti_pool;
    179 static u_int			in_multientries;
    180 static struct in_multihashhead *in_multihashtbl;
    181 static u_long			in_multihash;
    182 static krwlock_t		in_multilock;
    183 
    184 #define IN_MULTI_HASH(x, ifp) \
    185     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
    186 
    187 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
    188 u_long				in_ifaddrhash;
    189 struct in_ifaddrhead		in_ifaddrhead;
    190 
    191 void
    192 in_init(void)
    193 {
    194 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
    195 	    NULL, IPL_SOFTNET);
    196 	TAILQ_INIT(&in_ifaddrhead);
    197 
    198 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    199 	    &in_ifaddrhash);
    200 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    201 	    &in_multihash);
    202 	rw_init(&in_multilock);
    203 
    204 	in_sysctl_init(NULL);
    205 }
    206 
    207 /*
    208  * Return 1 if an internet address is for a ``local'' host
    209  * (one to which we have a connection).  If subnetsarelocal
    210  * is true, this includes other subnets of the local net.
    211  * Otherwise, it includes only the directly-connected (sub)nets.
    212  */
    213 int
    214 in_localaddr(struct in_addr in)
    215 {
    216 	struct in_ifaddr *ia;
    217 
    218 	if (subnetsarelocal) {
    219 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    220 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net)
    221 				return (1);
    222 	} else {
    223 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    224 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet)
    225 				return (1);
    226 	}
    227 	return (0);
    228 }
    229 
    230 /*
    231  * Determine whether an IP address is in a reserved set of addresses
    232  * that may not be forwarded, or whether datagrams to that destination
    233  * may be forwarded.
    234  */
    235 int
    236 in_canforward(struct in_addr in)
    237 {
    238 	u_int32_t net;
    239 
    240 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
    241 		return (0);
    242 	if (IN_CLASSA(in.s_addr)) {
    243 		net = in.s_addr & IN_CLASSA_NET;
    244 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
    245 			return (0);
    246 	}
    247 	return (1);
    248 }
    249 
    250 /*
    251  * Trim a mask in a sockaddr
    252  */
    253 void
    254 in_socktrim(struct sockaddr_in *ap)
    255 {
    256 	char *cplim = (char *) &ap->sin_addr;
    257 	char *cp = (char *) (&ap->sin_addr + 1);
    258 
    259 	ap->sin_len = 0;
    260 	while (--cp >= cplim)
    261 		if (*cp) {
    262 			(ap)->sin_len = cp - (char *) (ap) + 1;
    263 			break;
    264 		}
    265 }
    266 
    267 /*
    268  *  Routine to take an Internet address and convert into a
    269  *  "dotted quad" representation for printing.
    270  */
    271 const char *
    272 in_fmtaddr(struct in_addr addr)
    273 {
    274 	static char buf[sizeof("123.456.789.123")];
    275 
    276 	addr.s_addr = ntohl(addr.s_addr);
    277 
    278 	snprintf(buf, sizeof(buf), "%d.%d.%d.%d",
    279 		(addr.s_addr >> 24) & 0xFF,
    280 		(addr.s_addr >> 16) & 0xFF,
    281 		(addr.s_addr >>  8) & 0xFF,
    282 		(addr.s_addr >>  0) & 0xFF);
    283 	return buf;
    284 }
    285 
    286 /*
    287  * Maintain the "in_maxmtu" variable, which is the largest
    288  * mtu for non-local interfaces with AF_INET addresses assigned
    289  * to them that are up.
    290  */
    291 unsigned long in_maxmtu;
    292 
    293 void
    294 in_setmaxmtu(void)
    295 {
    296 	struct in_ifaddr *ia;
    297 	struct ifnet *ifp;
    298 	unsigned long maxmtu = 0;
    299 
    300 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    301 		if ((ifp = ia->ia_ifp) == 0)
    302 			continue;
    303 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
    304 			continue;
    305 		if (ifp->if_mtu > maxmtu)
    306 			maxmtu = ifp->if_mtu;
    307 	}
    308 	if (maxmtu)
    309 		in_maxmtu = maxmtu;
    310 }
    311 
    312 static u_int
    313 in_mask2len(struct in_addr *mask)
    314 {
    315 	u_int x, y;
    316 	u_char *p;
    317 
    318 	p = (u_char *)mask;
    319 	for (x = 0; x < sizeof(*mask); x++) {
    320 		if (p[x] != 0xff)
    321 			break;
    322 	}
    323 	y = 0;
    324 	if (x < sizeof(*mask)) {
    325 		for (y = 0; y < NBBY; y++) {
    326 			if ((p[x] & (0x80 >> y)) == 0)
    327 				break;
    328 		}
    329 	}
    330 	return x * NBBY + y;
    331 }
    332 
    333 static void
    334 in_len2mask(struct in_addr *mask, u_int len)
    335 {
    336 	u_int i;
    337 	u_char *p;
    338 
    339 	p = (u_char *)mask;
    340 	memset(mask, 0, sizeof(*mask));
    341 	for (i = 0; i < len / NBBY; i++)
    342 		p[i] = 0xff;
    343 	if (len % NBBY)
    344 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
    345 }
    346 
    347 /*
    348  * Generic internet control operations (ioctl's).
    349  * Ifp is 0 if not an interface-specific ioctl.
    350  */
    351 /* ARGSUSED */
    352 int
    353 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
    354 {
    355 	struct ifreq *ifr = (struct ifreq *)data;
    356 	struct in_ifaddr *ia = NULL;
    357 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
    358 	struct sockaddr_in oldaddr;
    359 	int error, hostIsNew, maskIsNew;
    360 	int newifaddr = 0;
    361 
    362 	switch (cmd) {
    363 	case SIOCALIFADDR:
    364 	case SIOCDLIFADDR:
    365 	case SIOCGLIFADDR:
    366 		if (ifp == NULL)
    367 			return EINVAL;
    368 		return in_lifaddr_ioctl(so, cmd, data, ifp);
    369 	case SIOCGIFADDRPREF:
    370 	case SIOCSIFADDRPREF:
    371 		if (ifp == NULL)
    372 			return EINVAL;
    373 		return ifaddrpref_ioctl(so, cmd, data, ifp);
    374 	}
    375 
    376 	/*
    377 	 * Find address for this interface, if it exists.
    378 	 */
    379 	if (ifp != NULL)
    380 		IFP_TO_IA(ifp, ia);
    381 
    382 	hostIsNew = 1;		/* moved here to appease gcc */
    383 	switch (cmd) {
    384 	case SIOCAIFADDR:
    385 	case SIOCDIFADDR:
    386 	case SIOCGIFALIAS:
    387 	case SIOCGIFAFLAG_IN:
    388 		if (ifra->ifra_addr.sin_family == AF_INET)
    389 			LIST_FOREACH(ia,
    390 			    &IN_IFADDR_HASH(ifra->ifra_addr.sin_addr.s_addr),
    391 			    ia_hash) {
    392 				if (ia->ia_ifp == ifp &&
    393 				    in_hosteq(ia->ia_addr.sin_addr,
    394 				    ifra->ifra_addr.sin_addr))
    395 					break;
    396 			}
    397 		if ((cmd == SIOCDIFADDR ||
    398 		    cmd == SIOCGIFALIAS ||
    399 		    cmd == SIOCGIFAFLAG_IN) &&
    400 		    ia == NULL)
    401 			return (EADDRNOTAVAIL);
    402 
    403 		if (cmd == SIOCDIFADDR &&
    404 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
    405 			ifra->ifra_addr.sin_family = AF_INET;
    406 		}
    407 		/* FALLTHROUGH */
    408 	case SIOCSIFADDR:
    409 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
    410 			;
    411 		else if (ifra->ifra_addr.sin_len == 0) {
    412 			ifra->ifra_addr = ia->ia_addr;
    413 			hostIsNew = 0;
    414 		} else if (in_hosteq(ia->ia_addr.sin_addr,
    415 		           ifra->ifra_addr.sin_addr))
    416 			hostIsNew = 0;
    417 		/* FALLTHROUGH */
    418 	case SIOCSIFDSTADDR:
    419 		if (ifra->ifra_addr.sin_family != AF_INET)
    420 			return (EAFNOSUPPORT);
    421 		/* FALLTHROUGH */
    422 	case SIOCSIFNETMASK:
    423 		if (ifp == NULL)
    424 			panic("in_control");
    425 
    426 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
    427 			break;
    428 
    429 		if (ia == NULL &&
    430 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR))
    431 			return (EADDRNOTAVAIL);
    432 
    433 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    434 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    435 		    NULL) != 0)
    436 			return (EPERM);
    437 
    438 		if (ia == NULL) {
    439 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
    440 			if (ia == NULL)
    441 				return (ENOBUFS);
    442 			TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
    443 			ifaref(&ia->ia_ifa);
    444 			ifa_insert(ifp, &ia->ia_ifa);
    445 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    446 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    447 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
    448 #ifdef IPSELSRC
    449 			ia->ia_ifa.ifa_getifa = in_getifa;
    450 #else /* IPSELSRC */
    451 			ia->ia_ifa.ifa_getifa = NULL;
    452 #endif /* IPSELSRC */
    453 			ia->ia_sockmask.sin_len = 8;
    454 			ia->ia_sockmask.sin_family = AF_INET;
    455 			if (ifp->if_flags & IFF_BROADCAST) {
    456 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
    457 				ia->ia_broadaddr.sin_family = AF_INET;
    458 			}
    459 			ia->ia_ifp = ifp;
    460 			ia->ia_idsalt = cprng_fast32() % 65535;
    461 			LIST_INIT(&ia->ia_multiaddrs);
    462 			newifaddr = 1;
    463 		}
    464 		break;
    465 
    466 	case SIOCSIFBRDADDR:
    467 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    468 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    469 		    NULL) != 0)
    470 			return (EPERM);
    471 		/* FALLTHROUGH */
    472 
    473 	case SIOCGIFADDR:
    474 	case SIOCGIFNETMASK:
    475 	case SIOCGIFDSTADDR:
    476 	case SIOCGIFBRDADDR:
    477 		if (ia == NULL)
    478 			return (EADDRNOTAVAIL);
    479 		break;
    480 	}
    481 	error = 0;
    482 	switch (cmd) {
    483 
    484 	case SIOCGIFADDR:
    485 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
    486 		break;
    487 
    488 	case SIOCGIFBRDADDR:
    489 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    490 			return (EINVAL);
    491 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
    492 		break;
    493 
    494 	case SIOCGIFDSTADDR:
    495 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    496 			return (EINVAL);
    497 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
    498 		break;
    499 
    500 	case SIOCGIFNETMASK:
    501 		/*
    502 		 * We keep the number of trailing zero bytes the sin_len field
    503 		 * of ia_sockmask, so we fix this before we pass it back to
    504 		 * userland.
    505 		 */
    506 		oldaddr = ia->ia_sockmask;
    507 		oldaddr.sin_len = sizeof(struct sockaddr_in);
    508 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
    509 		break;
    510 
    511 	case SIOCSIFDSTADDR:
    512 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    513 			return (EINVAL);
    514 		oldaddr = ia->ia_dstaddr;
    515 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
    516 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
    517 			ia->ia_dstaddr = oldaddr;
    518 			return error;
    519 		}
    520 		if (ia->ia_flags & IFA_ROUTE) {
    521 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
    522 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
    523 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    524 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
    525 		}
    526 		break;
    527 
    528 	case SIOCSIFBRDADDR:
    529 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    530 			return EINVAL;
    531 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
    532 		break;
    533 
    534 	case SIOCSIFADDR:
    535 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
    536 		    1, hostIsNew);
    537 		if (error == 0) {
    538 			(void)pfil_run_hooks(if_pfil,
    539 			    (struct mbuf **)SIOCSIFADDR, ifp, PFIL_IFADDR);
    540 		}
    541 		break;
    542 
    543 	case SIOCSIFNETMASK:
    544 		in_ifscrub(ifp, ia);
    545 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
    546 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    547 		error = in_ifinit(ifp, ia, NULL, 0, 0);
    548 		break;
    549 
    550 	case SIOCAIFADDR:
    551 		maskIsNew = 0;
    552 		if (ifra->ifra_mask.sin_len) {
    553 			/* Only scrub if we control the prefix route,
    554 			 * otherwise userland gets a bogus message */
    555 			if ((ia->ia_flags & IFA_ROUTE))
    556 				in_ifscrub(ifp, ia);
    557 			ia->ia_sockmask = ifra->ifra_mask;
    558 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    559 			maskIsNew = 1;
    560 		}
    561 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    562 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
    563 			/* Only scrub if we control the prefix route,
    564 			 * otherwise userland gets a bogus message */
    565 			if ((ia->ia_flags & IFA_ROUTE))
    566 				in_ifscrub(ifp, ia);
    567 			ia->ia_dstaddr = ifra->ifra_dstaddr;
    568 			maskIsNew  = 1; /* We lie; but the effect's the same */
    569 		}
    570 		if (ifra->ifra_addr.sin_family == AF_INET &&
    571 		    (hostIsNew || maskIsNew)) {
    572 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
    573 			    hostIsNew);
    574 		}
    575 		if ((ifp->if_flags & IFF_BROADCAST) &&
    576 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
    577 			ia->ia_broadaddr = ifra->ifra_broadaddr;
    578 		if (error == 0)
    579 			(void)pfil_run_hooks(if_pfil,
    580 			    (struct mbuf **)SIOCAIFADDR, ifp, PFIL_IFADDR);
    581 		break;
    582 
    583 	case SIOCGIFALIAS:
    584 		ifra->ifra_mask = ia->ia_sockmask;
    585 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    586 		    (ia->ia_dstaddr.sin_family == AF_INET))
    587 			ifra->ifra_dstaddr = ia->ia_dstaddr;
    588 		else if ((ifp->if_flags & IFF_BROADCAST) &&
    589 		    (ia->ia_broadaddr.sin_family == AF_INET))
    590 			ifra->ifra_broadaddr = ia->ia_broadaddr;
    591 		else
    592 			memset(&ifra->ifra_broadaddr, 0,
    593 			      sizeof(ifra->ifra_broadaddr));
    594 		break;
    595 
    596 	case SIOCGIFAFLAG_IN:
    597 		ifr->ifr_addrflags = ia->ia4_flags;
    598 		break;
    599 
    600 	case SIOCDIFADDR:
    601 		in_purgeaddr(&ia->ia_ifa);
    602 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR,
    603 		    ifp, PFIL_IFADDR);
    604 		break;
    605 
    606 #ifdef MROUTING
    607 	case SIOCGETVIFCNT:
    608 	case SIOCGETSGCNT:
    609 		error = mrt_ioctl(so, cmd, data);
    610 		break;
    611 #endif /* MROUTING */
    612 
    613 	default:
    614 		return ENOTTY;
    615 	}
    616 
    617 	if (error != 0 && newifaddr) {
    618 		KASSERT(ia != NULL);
    619 		in_purgeaddr(&ia->ia_ifa);
    620 	}
    621 
    622 	return error;
    623 }
    624 
    625 /* Add ownaddr as loopback rtentry. */
    626 static void
    627 in_ifaddlocal(struct ifaddr *ifa)
    628 {
    629 	struct in_ifaddr *ia;
    630 
    631 	ia = (struct in_ifaddr *)ifa;
    632 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
    633 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
    634 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
    635 	{
    636 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
    637 		return;
    638 	}
    639 
    640 	rt_ifa_addlocal(ifa);
    641 }
    642 
    643 /* Rempve loopback entry of ownaddr */
    644 static void
    645 in_ifremlocal(struct ifaddr *ifa)
    646 {
    647 	struct in_ifaddr *ia, *p;
    648 	struct ifaddr *alt_ifa = NULL;
    649 	int ia_count = 0;
    650 
    651 	ia = (struct in_ifaddr *)ifa;
    652 	/* Delete the entry if exactly one ifaddr matches the
    653 	 * address, ifa->ifa_addr. */
    654 	TAILQ_FOREACH(p, &in_ifaddrhead, ia_list) {
    655 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
    656 			continue;
    657 		if (p->ia_ifp != ia->ia_ifp)
    658 			alt_ifa = &p->ia_ifa;
    659 		if (++ia_count > 1 && alt_ifa != NULL)
    660 			break;
    661 	}
    662 
    663 	if (ia_count == 0)
    664 		return;
    665 
    666 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
    667 }
    668 
    669 void
    670 in_purgeaddr(struct ifaddr *ifa)
    671 {
    672 	struct ifnet *ifp = ifa->ifa_ifp;
    673 	struct in_ifaddr *ia = (void *) ifa;
    674 
    675 	/* stop DAD processing */
    676 	if (ia->ia_dad_stop != NULL)
    677 		ia->ia_dad_stop(ifa);
    678 
    679 	in_ifscrub(ifp, ia);
    680 	in_ifremlocal(ifa);
    681 	LIST_REMOVE(ia, ia_hash);
    682 	ifa_remove(ifp, &ia->ia_ifa);
    683 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
    684 	if (ia->ia_allhosts != NULL)
    685 		in_delmulti(ia->ia_allhosts);
    686 	ifafree(&ia->ia_ifa);
    687 	in_setmaxmtu();
    688 }
    689 
    690 void
    691 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
    692 {
    693 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
    694 	igmp_purgeif(ifp);		/* manipulates pools */
    695 #ifdef MROUTING
    696 	ip_mrouter_detach(ifp);
    697 #endif
    698 }
    699 
    700 /*
    701  * SIOC[GAD]LIFADDR.
    702  *	SIOCGLIFADDR: get first address. (???)
    703  *	SIOCGLIFADDR with IFLR_PREFIX:
    704  *		get first address that matches the specified prefix.
    705  *	SIOCALIFADDR: add the specified address.
    706  *	SIOCALIFADDR with IFLR_PREFIX:
    707  *		EINVAL since we can't deduce hostid part of the address.
    708  *	SIOCDLIFADDR: delete the specified address.
    709  *	SIOCDLIFADDR with IFLR_PREFIX:
    710  *		delete the first address that matches the specified prefix.
    711  * return values:
    712  *	EINVAL on invalid parameters
    713  *	EADDRNOTAVAIL on prefix match failed/specified address not found
    714  *	other values may be returned from in_ioctl()
    715  */
    716 static int
    717 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
    718     struct ifnet *ifp)
    719 {
    720 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
    721 	struct ifaddr *ifa;
    722 	struct sockaddr *sa;
    723 
    724 	/* sanity checks */
    725 	if (data == NULL || ifp == NULL) {
    726 		panic("invalid argument to in_lifaddr_ioctl");
    727 		/*NOTRECHED*/
    728 	}
    729 
    730 	switch (cmd) {
    731 	case SIOCGLIFADDR:
    732 		/* address must be specified on GET with IFLR_PREFIX */
    733 		if ((iflr->flags & IFLR_PREFIX) == 0)
    734 			break;
    735 		/*FALLTHROUGH*/
    736 	case SIOCALIFADDR:
    737 	case SIOCDLIFADDR:
    738 		/* address must be specified on ADD and DELETE */
    739 		sa = (struct sockaddr *)&iflr->addr;
    740 		if (sa->sa_family != AF_INET)
    741 			return EINVAL;
    742 		if (sa->sa_len != sizeof(struct sockaddr_in))
    743 			return EINVAL;
    744 		/* XXX need improvement */
    745 		sa = (struct sockaddr *)&iflr->dstaddr;
    746 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
    747 			return EINVAL;
    748 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
    749 			return EINVAL;
    750 		break;
    751 	default: /*shouldn't happen*/
    752 #if 0
    753 		panic("invalid cmd to in_lifaddr_ioctl");
    754 		/*NOTREACHED*/
    755 #else
    756 		return EOPNOTSUPP;
    757 #endif
    758 	}
    759 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
    760 		return EINVAL;
    761 
    762 	switch (cmd) {
    763 	case SIOCALIFADDR:
    764 	    {
    765 		struct in_aliasreq ifra;
    766 
    767 		if (iflr->flags & IFLR_PREFIX)
    768 			return EINVAL;
    769 
    770 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
    771 		memset(&ifra, 0, sizeof(ifra));
    772 		memcpy(ifra.ifra_name, iflr->iflr_name,
    773 			sizeof(ifra.ifra_name));
    774 
    775 		memcpy(&ifra.ifra_addr, &iflr->addr,
    776 			((struct sockaddr *)&iflr->addr)->sa_len);
    777 
    778 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
    779 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
    780 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
    781 		}
    782 
    783 		ifra.ifra_mask.sin_family = AF_INET;
    784 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
    785 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
    786 
    787 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
    788 	    }
    789 	case SIOCGLIFADDR:
    790 	case SIOCDLIFADDR:
    791 	    {
    792 		struct in_ifaddr *ia;
    793 		struct in_addr mask, candidate, match;
    794 		struct sockaddr_in *sin;
    795 		int cmp;
    796 
    797 		memset(&mask, 0, sizeof(mask));
    798 		memset(&match, 0, sizeof(match));	/* XXX gcc */
    799 		if (iflr->flags & IFLR_PREFIX) {
    800 			/* lookup a prefix rather than address. */
    801 			in_len2mask(&mask, iflr->prefixlen);
    802 
    803 			sin = (struct sockaddr_in *)&iflr->addr;
    804 			match.s_addr = sin->sin_addr.s_addr;
    805 			match.s_addr &= mask.s_addr;
    806 
    807 			/* if you set extra bits, that's wrong */
    808 			if (match.s_addr != sin->sin_addr.s_addr)
    809 				return EINVAL;
    810 
    811 			cmp = 1;
    812 		} else {
    813 			if (cmd == SIOCGLIFADDR) {
    814 				/* on getting an address, take the 1st match */
    815 				cmp = 0;	/*XXX*/
    816 			} else {
    817 				/* on deleting an address, do exact match */
    818 				in_len2mask(&mask, 32);
    819 				sin = (struct sockaddr_in *)&iflr->addr;
    820 				match.s_addr = sin->sin_addr.s_addr;
    821 
    822 				cmp = 1;
    823 			}
    824 		}
    825 
    826 		IFADDR_FOREACH(ifa, ifp) {
    827 			if (ifa->ifa_addr->sa_family != AF_INET)
    828 				continue;
    829 			if (cmp == 0)
    830 				break;
    831 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
    832 			candidate.s_addr &= mask.s_addr;
    833 			if (candidate.s_addr == match.s_addr)
    834 				break;
    835 		}
    836 		if (ifa == NULL)
    837 			return EADDRNOTAVAIL;
    838 		ia = (struct in_ifaddr *)ifa;
    839 
    840 		if (cmd == SIOCGLIFADDR) {
    841 			/* fill in the if_laddrreq structure */
    842 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
    843 
    844 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    845 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
    846 					ia->ia_dstaddr.sin_len);
    847 			} else
    848 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
    849 
    850 			iflr->prefixlen =
    851 				in_mask2len(&ia->ia_sockmask.sin_addr);
    852 
    853 			iflr->flags = 0;	/*XXX*/
    854 
    855 			return 0;
    856 		} else {
    857 			struct in_aliasreq ifra;
    858 
    859 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
    860 			memset(&ifra, 0, sizeof(ifra));
    861 			memcpy(ifra.ifra_name, iflr->iflr_name,
    862 				sizeof(ifra.ifra_name));
    863 
    864 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
    865 				ia->ia_addr.sin_len);
    866 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    867 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
    868 					ia->ia_dstaddr.sin_len);
    869 			}
    870 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
    871 				ia->ia_sockmask.sin_len);
    872 
    873 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
    874 		}
    875 	    }
    876 	}
    877 
    878 	return EOPNOTSUPP;	/*just for safety*/
    879 }
    880 
    881 /*
    882  * Delete any existing route for an interface.
    883  */
    884 void
    885 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
    886 {
    887 
    888 	in_scrubprefix(ia);
    889 }
    890 
    891 /*
    892  * Initialize an interface's internet address
    893  * and routing table entry.
    894  */
    895 int
    896 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
    897     const struct sockaddr_in *sin, int scrub, int hostIsNew)
    898 {
    899 	u_int32_t i;
    900 	struct sockaddr_in oldaddr;
    901 	int s = splnet(), flags = RTF_UP, error;
    902 
    903 	if (sin == NULL)
    904 		sin = &ia->ia_addr;
    905 
    906 	/*
    907 	 * Set up new addresses.
    908 	 */
    909 	oldaddr = ia->ia_addr;
    910 	if (ia->ia_addr.sin_family == AF_INET)
    911 		LIST_REMOVE(ia, ia_hash);
    912 	ia->ia_addr = *sin;
    913 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
    914 
    915 	/* Set IN_IFF flags early for if_addr_init() */
    916 	if (hostIsNew && if_do_dad(ifp) && !in_nullhost(ia->ia_addr.sin_addr)) {
    917 		if (ifp->if_link_state == LINK_STATE_DOWN)
    918 			ia->ia4_flags |= IN_IFF_DETACHED;
    919 		else
    920 			/* State the intent to try DAD if possible */
    921 			ia->ia4_flags |= IN_IFF_TRYTENTATIVE;
    922 	}
    923 
    924 	/*
    925 	 * Give the interface a chance to initialize
    926 	 * if this is its first address,
    927 	 * and to validate the address if necessary.
    928 	 */
    929 	if ((error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0)
    930 		goto bad;
    931 	/* Now clear the try tentative flag, it's job is done. */
    932 	ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE;
    933 	splx(s);
    934 
    935 	if (scrub) {
    936 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
    937 		in_ifscrub(ifp, ia);
    938 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    939 	}
    940 
    941 	/* Add the local route to the address */
    942 	in_ifaddlocal(&ia->ia_ifa);
    943 
    944 	i = ia->ia_addr.sin_addr.s_addr;
    945 	if (IN_CLASSA(i))
    946 		ia->ia_netmask = IN_CLASSA_NET;
    947 	else if (IN_CLASSB(i))
    948 		ia->ia_netmask = IN_CLASSB_NET;
    949 	else
    950 		ia->ia_netmask = IN_CLASSC_NET;
    951 	/*
    952 	 * The subnet mask usually includes at least the standard network part,
    953 	 * but may may be smaller in the case of supernetting.
    954 	 * If it is set, we believe it.
    955 	 */
    956 	if (ia->ia_subnetmask == 0) {
    957 		ia->ia_subnetmask = ia->ia_netmask;
    958 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
    959 	} else
    960 		ia->ia_netmask &= ia->ia_subnetmask;
    961 
    962 	ia->ia_net = i & ia->ia_netmask;
    963 	ia->ia_subnet = i & ia->ia_subnetmask;
    964 	in_socktrim(&ia->ia_sockmask);
    965 	/* re-calculate the "in_maxmtu" value */
    966 	in_setmaxmtu();
    967 	/*
    968 	 * Add route for the network.
    969 	 */
    970 	ia->ia_ifa.ifa_metric = ifp->if_metric;
    971 	if (ifp->if_flags & IFF_BROADCAST) {
    972 		ia->ia_broadaddr.sin_addr.s_addr =
    973 			ia->ia_subnet | ~ia->ia_subnetmask;
    974 		ia->ia_netbroadcast.s_addr =
    975 			ia->ia_net | ~ia->ia_netmask;
    976 	} else if (ifp->if_flags & IFF_LOOPBACK) {
    977 		ia->ia_dstaddr = ia->ia_addr;
    978 		flags |= RTF_HOST;
    979 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
    980 		if (ia->ia_dstaddr.sin_family != AF_INET)
    981 			return (0);
    982 		flags |= RTF_HOST;
    983 	}
    984 	error = in_addprefix(ia, flags);
    985 	/*
    986 	 * If the interface supports multicast, join the "all hosts"
    987 	 * multicast group on that interface.
    988 	 */
    989 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
    990 		struct in_addr addr;
    991 
    992 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
    993 		ia->ia_allhosts = in_addmulti(&addr, ifp);
    994 	}
    995 
    996 	if (hostIsNew && if_do_dad(ifp) &&
    997 	    !in_nullhost(ia->ia_addr.sin_addr) &&
    998 	    ia->ia4_flags & IN_IFF_TENTATIVE)
    999 		ia->ia_dad_start((struct ifaddr *)ia);
   1000 
   1001 	return (error);
   1002 bad:
   1003 	splx(s);
   1004 	LIST_REMOVE(ia, ia_hash);
   1005 	ia->ia_addr = oldaddr;
   1006 	if (ia->ia_addr.sin_family == AF_INET)
   1007 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
   1008 		    ia, ia_hash);
   1009 	return (error);
   1010 }
   1011 
   1012 #define rtinitflags(x) \
   1013 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
   1014 	    ? RTF_HOST : 0)
   1015 
   1016 /*
   1017  * add a route to prefix ("connected route" in cisco terminology).
   1018  * does nothing if there's some interface address with the same prefix already.
   1019  */
   1020 static int
   1021 in_addprefix(struct in_ifaddr *target, int flags)
   1022 {
   1023 	struct in_ifaddr *ia;
   1024 	struct in_addr prefix, mask, p;
   1025 	int error;
   1026 
   1027 	if ((flags & RTF_HOST) != 0)
   1028 		prefix = target->ia_dstaddr.sin_addr;
   1029 	else {
   1030 		prefix = target->ia_addr.sin_addr;
   1031 		mask = target->ia_sockmask.sin_addr;
   1032 		prefix.s_addr &= mask.s_addr;
   1033 	}
   1034 
   1035 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1036 		if (rtinitflags(ia))
   1037 			p = ia->ia_dstaddr.sin_addr;
   1038 		else {
   1039 			p = ia->ia_addr.sin_addr;
   1040 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1041 		}
   1042 
   1043 		if (prefix.s_addr != p.s_addr)
   1044 			continue;
   1045 
   1046 		/*
   1047 		 * if we got a matching prefix route inserted by other
   1048 		 * interface address, we don't need to bother
   1049 		 *
   1050 		 * XXX RADIX_MPATH implications here? -dyoung
   1051 		 */
   1052 		if (ia->ia_flags & IFA_ROUTE)
   1053 			return 0;
   1054 	}
   1055 
   1056 	/*
   1057 	 * noone seem to have prefix route.  insert it.
   1058 	 */
   1059 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
   1060 	if (error == 0)
   1061 		target->ia_flags |= IFA_ROUTE;
   1062 	else if (error == EEXIST) {
   1063 		/*
   1064 		 * the fact the route already exists is not an error.
   1065 		 */
   1066 		error = 0;
   1067 	}
   1068 	return error;
   1069 }
   1070 
   1071 /*
   1072  * remove a route to prefix ("connected route" in cisco terminology).
   1073  * re-installs the route by using another interface address, if there's one
   1074  * with the same prefix (otherwise we lose the route mistakenly).
   1075  */
   1076 static int
   1077 in_scrubprefix(struct in_ifaddr *target)
   1078 {
   1079 	struct in_ifaddr *ia;
   1080 	struct in_addr prefix, mask, p;
   1081 	int error;
   1082 
   1083 	/* If we don't have IFA_ROUTE we should still inform userland */
   1084 	if ((target->ia_flags & IFA_ROUTE) == 0)
   1085 		return 0;
   1086 
   1087 	if (rtinitflags(target))
   1088 		prefix = target->ia_dstaddr.sin_addr;
   1089 	else {
   1090 		prefix = target->ia_addr.sin_addr;
   1091 		mask = target->ia_sockmask.sin_addr;
   1092 		prefix.s_addr &= mask.s_addr;
   1093 	}
   1094 
   1095 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1096 		if (rtinitflags(ia))
   1097 			p = ia->ia_dstaddr.sin_addr;
   1098 		else {
   1099 			p = ia->ia_addr.sin_addr;
   1100 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1101 		}
   1102 
   1103 		if (prefix.s_addr != p.s_addr)
   1104 			continue;
   1105 
   1106 		/*
   1107 		 * if we got a matching prefix route, move IFA_ROUTE to him
   1108 		 */
   1109 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
   1110 			rtinit(&target->ia_ifa, RTM_DELETE,
   1111 			    rtinitflags(target));
   1112 			target->ia_flags &= ~IFA_ROUTE;
   1113 
   1114 			error = rtinit(&ia->ia_ifa, RTM_ADD,
   1115 			    rtinitflags(ia) | RTF_UP);
   1116 			if (error == 0)
   1117 				ia->ia_flags |= IFA_ROUTE;
   1118 			return error;
   1119 		}
   1120 	}
   1121 
   1122 	/*
   1123 	 * noone seem to have prefix route.  remove it.
   1124 	 */
   1125 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
   1126 	target->ia_flags &= ~IFA_ROUTE;
   1127 	return 0;
   1128 }
   1129 
   1130 #undef rtinitflags
   1131 
   1132 /*
   1133  * Return 1 if the address might be a local broadcast address.
   1134  */
   1135 int
   1136 in_broadcast(struct in_addr in, struct ifnet *ifp)
   1137 {
   1138 	struct ifaddr *ifa;
   1139 
   1140 	if (in.s_addr == INADDR_BROADCAST ||
   1141 	    in_nullhost(in))
   1142 		return 1;
   1143 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
   1144 		return 0;
   1145 	/*
   1146 	 * Look through the list of addresses for a match
   1147 	 * with a broadcast address.
   1148 	 */
   1149 #define ia (ifatoia(ifa))
   1150 	IFADDR_FOREACH(ifa, ifp)
   1151 		if (ifa->ifa_addr->sa_family == AF_INET &&
   1152 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
   1153 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
   1154 		     in_hosteq(in, ia->ia_netbroadcast) ||
   1155 		     (hostzeroisbroadcast &&
   1156 		      /*
   1157 		       * Check for old-style (host 0) broadcast.
   1158 		       */
   1159 		      (in.s_addr == ia->ia_subnet ||
   1160 		       in.s_addr == ia->ia_net))))
   1161 			return 1;
   1162 	return (0);
   1163 #undef ia
   1164 }
   1165 
   1166 /*
   1167  * perform DAD when interface becomes IFF_UP.
   1168  */
   1169 void
   1170 in_if_link_up(struct ifnet *ifp)
   1171 {
   1172 	struct ifaddr *ifa;
   1173 	struct in_ifaddr *ia;
   1174 
   1175 	/* Ensure it's sane to run DAD */
   1176 	if (ifp->if_link_state == LINK_STATE_DOWN)
   1177 		return;
   1178 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
   1179 		return;
   1180 
   1181 	IFADDR_FOREACH(ifa, ifp) {
   1182 		if (ifa->ifa_addr->sa_family != AF_INET)
   1183 			continue;
   1184 		ia = (struct in_ifaddr *)ifa;
   1185 
   1186 		/* If detached then mark as tentative */
   1187 		if (ia->ia4_flags & IN_IFF_DETACHED) {
   1188 			ia->ia4_flags &= ~IN_IFF_DETACHED;
   1189 			if (if_do_dad(ifp) && ia->ia_dad_start != NULL)
   1190 				ia->ia4_flags |= IN_IFF_TENTATIVE;
   1191 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
   1192 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1193 		}
   1194 
   1195 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
   1196 			/* Clear the duplicated flag as we're starting DAD. */
   1197 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
   1198 			ia->ia_dad_start(ifa);
   1199 		}
   1200 	}
   1201 }
   1202 
   1203 void
   1204 in_if_up(struct ifnet *ifp)
   1205 {
   1206 
   1207 	/* interface may not support link state, so bring it up also */
   1208 	in_if_link_up(ifp);
   1209 }
   1210 
   1211 /*
   1212  * Mark all addresses as detached.
   1213  */
   1214 void
   1215 in_if_link_down(struct ifnet *ifp)
   1216 {
   1217 	struct ifaddr *ifa;
   1218 	struct in_ifaddr *ia;
   1219 
   1220 	IFADDR_FOREACH(ifa, ifp) {
   1221 		if (ifa->ifa_addr->sa_family != AF_INET)
   1222 			continue;
   1223 		ia = (struct in_ifaddr *)ifa;
   1224 
   1225 		/* Stop DAD processing */
   1226 		if (ia->ia_dad_stop != NULL)
   1227 			ia->ia_dad_stop(ifa);
   1228 
   1229 		/*
   1230 		 * Mark the address as detached.
   1231 		 */
   1232 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
   1233 			ia->ia4_flags |= IN_IFF_DETACHED;
   1234 			ia->ia4_flags &=
   1235 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
   1236 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1237 		}
   1238 	}
   1239 }
   1240 
   1241 void
   1242 in_if_down(struct ifnet *ifp)
   1243 {
   1244 
   1245 	in_if_link_down(ifp);
   1246 }
   1247 
   1248 void
   1249 in_if_link_state_change(struct ifnet *ifp, int link_state)
   1250 {
   1251 
   1252 	switch (link_state) {
   1253 	case LINK_STATE_DOWN:
   1254 		in_if_link_down(ifp);
   1255 		break;
   1256 	case LINK_STATE_UP:
   1257 		in_if_link_up(ifp);
   1258 		break;
   1259 	}
   1260 }
   1261 
   1262 /*
   1263  * in_lookup_multi: look up the in_multi record for a given IP
   1264  * multicast address on a given interface.  If no matching record is
   1265  * found, return NULL.
   1266  */
   1267 struct in_multi *
   1268 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
   1269 {
   1270 	struct in_multi *inm;
   1271 
   1272 	KASSERT(rw_lock_held(&in_multilock));
   1273 
   1274 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
   1275 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
   1276 			break;
   1277 	}
   1278 	return inm;
   1279 }
   1280 
   1281 /*
   1282  * in_multi_group: check whether the address belongs to an IP multicast
   1283  * group we are joined on this interface.  Returns true or false.
   1284  */
   1285 bool
   1286 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
   1287 {
   1288 	bool ingroup;
   1289 
   1290 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
   1291 		rw_enter(&in_multilock, RW_READER);
   1292 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1293 		rw_exit(&in_multilock);
   1294 	} else {
   1295 		/* XXX Recursive call from ip_output(). */
   1296 		KASSERT(rw_lock_held(&in_multilock));
   1297 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1298 	}
   1299 	return ingroup;
   1300 }
   1301 
   1302 /*
   1303  * Add an address to the list of IP multicast addresses for a given interface.
   1304  */
   1305 struct in_multi *
   1306 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
   1307 {
   1308 	struct sockaddr_in sin;
   1309 	struct in_multi *inm;
   1310 
   1311 	/*
   1312 	 * See if address already in list.
   1313 	 */
   1314 	rw_enter(&in_multilock, RW_WRITER);
   1315 	inm = in_lookup_multi(*ap, ifp);
   1316 	if (inm != NULL) {
   1317 		/*
   1318 		 * Found it; just increment the reference count.
   1319 		 */
   1320 		inm->inm_refcount++;
   1321 		rw_exit(&in_multilock);
   1322 		return inm;
   1323 	}
   1324 
   1325 	/*
   1326 	 * New address; allocate a new multicast record.
   1327 	 */
   1328 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
   1329 	if (inm == NULL) {
   1330 		rw_exit(&in_multilock);
   1331 		return NULL;
   1332 	}
   1333 	inm->inm_addr = *ap;
   1334 	inm->inm_ifp = ifp;
   1335 	inm->inm_refcount = 1;
   1336 
   1337 	/*
   1338 	 * Ask the network driver to update its multicast reception
   1339 	 * filter appropriately for the new address.
   1340 	 */
   1341 	sockaddr_in_init(&sin, ap, 0);
   1342 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
   1343 		rw_exit(&in_multilock);
   1344 		pool_put(&inmulti_pool, inm);
   1345 		return NULL;
   1346 	}
   1347 
   1348 	/*
   1349 	 * Let IGMP know that we have joined a new IP multicast group.
   1350 	 */
   1351 	if (igmp_joingroup(inm) != 0) {
   1352 		rw_exit(&in_multilock);
   1353 		pool_put(&inmulti_pool, inm);
   1354 		return NULL;
   1355 	}
   1356 	LIST_INSERT_HEAD(
   1357 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
   1358 	    inm, inm_list);
   1359 	in_multientries++;
   1360 	rw_exit(&in_multilock);
   1361 
   1362 	return inm;
   1363 }
   1364 
   1365 /*
   1366  * Delete a multicast address record.
   1367  */
   1368 void
   1369 in_delmulti(struct in_multi *inm)
   1370 {
   1371 	struct sockaddr_in sin;
   1372 
   1373 	rw_enter(&in_multilock, RW_WRITER);
   1374 	if (--inm->inm_refcount > 0) {
   1375 		rw_exit(&in_multilock);
   1376 		return;
   1377 	}
   1378 
   1379 	/*
   1380 	 * No remaining claims to this record; let IGMP know that
   1381 	 * we are leaving the multicast group.
   1382 	 */
   1383 	igmp_leavegroup(inm);
   1384 
   1385 	/*
   1386 	 * Notify the network driver to update its multicast reception
   1387 	 * filter.
   1388 	 */
   1389 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
   1390 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
   1391 
   1392 	/*
   1393 	 * Unlink from list.
   1394 	 */
   1395 	LIST_REMOVE(inm, inm_list);
   1396 	in_multientries--;
   1397 	rw_exit(&in_multilock);
   1398 
   1399 	pool_put(&inmulti_pool, inm);
   1400 }
   1401 
   1402 /*
   1403  * in_next_multi: step through all of the in_multi records, one at a time.
   1404  * The current position is remembered in "step", which the caller must
   1405  * provide.  in_first_multi(), below, must be called to initialize "step"
   1406  * and get the first record.  Both macros return a NULL "inm" when there
   1407  * are no remaining records.
   1408  */
   1409 struct in_multi *
   1410 in_next_multi(struct in_multistep *step)
   1411 {
   1412 	struct in_multi *inm;
   1413 
   1414 	KASSERT(rw_lock_held(&in_multilock));
   1415 
   1416 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
   1417 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
   1418 	}
   1419 	if ((inm = step->i_inm) != NULL) {
   1420 		step->i_inm = LIST_NEXT(inm, inm_list);
   1421 	}
   1422 	return inm;
   1423 }
   1424 
   1425 struct in_multi *
   1426 in_first_multi(struct in_multistep *step)
   1427 {
   1428 	KASSERT(rw_lock_held(&in_multilock));
   1429 
   1430 	step->i_n = 0;
   1431 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
   1432 	return in_next_multi(step);
   1433 }
   1434 
   1435 void
   1436 in_multi_lock(int op)
   1437 {
   1438 	rw_enter(&in_multilock, op);
   1439 }
   1440 
   1441 void
   1442 in_multi_unlock(void)
   1443 {
   1444 	rw_exit(&in_multilock);
   1445 }
   1446 
   1447 int
   1448 in_multi_lock_held(void)
   1449 {
   1450 	return rw_lock_held(&in_multilock);
   1451 }
   1452 
   1453 struct sockaddr_in *
   1454 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
   1455     int soopts, struct ip_moptions *mopts, int *errorp)
   1456 {
   1457 	struct rtentry *rt = NULL;
   1458 	struct in_ifaddr *ia = NULL;
   1459 
   1460 	/*
   1461          * If route is known or can be allocated now, take the
   1462          * source address from the interface.  Otherwise, punt.
   1463 	 */
   1464 	if ((soopts & SO_DONTROUTE) != 0)
   1465 		rtcache_free(ro);
   1466 	else {
   1467 		union {
   1468 			struct sockaddr		dst;
   1469 			struct sockaddr_in	dst4;
   1470 		} u;
   1471 
   1472 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
   1473 		rt = rtcache_lookup(ro, &u.dst);
   1474 	}
   1475 	/*
   1476 	 * If we found a route, use the address
   1477 	 * corresponding to the outgoing interface
   1478 	 * unless it is the loopback (in case a route
   1479 	 * to our address on another net goes to loopback).
   1480 	 *
   1481 	 * XXX Is this still true?  Do we care?
   1482 	 */
   1483 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1484 		ia = ifatoia(rt->rt_ifa);
   1485 	if (ia == NULL) {
   1486 		u_int16_t fport = sin->sin_port;
   1487 
   1488 		sin->sin_port = 0;
   1489 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
   1490 		sin->sin_port = fport;
   1491 		if (ia == NULL) {
   1492 			/* Find 1st non-loopback AF_INET address */
   1493 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1494 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
   1495 					break;
   1496 			}
   1497 		}
   1498 		if (ia == NULL) {
   1499 			*errorp = EADDRNOTAVAIL;
   1500 			return NULL;
   1501 		}
   1502 	}
   1503 	/*
   1504 	 * If the destination address is multicast and an outgoing
   1505 	 * interface has been set as a multicast option, use the
   1506 	 * address of that interface as our source address.
   1507 	 */
   1508 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
   1509 		struct ip_moptions *imo;
   1510 		struct ifnet *ifp;
   1511 
   1512 		imo = mopts;
   1513 		if (imo->imo_multicast_ifp != NULL) {
   1514 			ifp = imo->imo_multicast_ifp;
   1515 			IFP_TO_IA(ifp, ia);		/* XXX */
   1516 			if (ia == 0 || ia->ia4_flags & IN_IFF_NOTREADY) {
   1517 				*errorp = EADDRNOTAVAIL;
   1518 				return NULL;
   1519 			}
   1520 		}
   1521 	}
   1522 	if (ia->ia_ifa.ifa_getifa != NULL) {
   1523 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
   1524 		                                      sintosa(sin)));
   1525 		if (ia == NULL) {
   1526 			*errorp = EADDRNOTAVAIL;
   1527 			return NULL;
   1528 		}
   1529 	}
   1530 #ifdef GETIFA_DEBUG
   1531 	else
   1532 		printf("%s: missing ifa_getifa\n", __func__);
   1533 #endif
   1534 	return satosin(&ia->ia_addr);
   1535 }
   1536 
   1537 struct in_llentry {
   1538 	struct llentry		base;
   1539 };
   1540 
   1541 #define	IN_LLTBL_DEFAULT_HSIZE	32
   1542 #define	IN_LLTBL_HASH(k, h) \
   1543 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
   1544 
   1545 /*
   1546  * Do actual deallocation of @lle.
   1547  * Called by LLE_FREE_LOCKED when number of references
   1548  * drops to zero.
   1549  */
   1550 static void
   1551 in_lltable_destroy_lle(struct llentry *lle)
   1552 {
   1553 
   1554 	LLE_WUNLOCK(lle);
   1555 	LLE_LOCK_DESTROY(lle);
   1556 	kmem_intr_free(lle, sizeof(*lle));
   1557 }
   1558 
   1559 static struct llentry *
   1560 in_lltable_new(struct in_addr addr4, u_int flags)
   1561 {
   1562 	struct in_llentry *lle;
   1563 
   1564 	lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP);
   1565 	if (lle == NULL)		/* NB: caller generates msg */
   1566 		return NULL;
   1567 
   1568 	/*
   1569 	 * For IPv4 this will trigger "arpresolve" to generate
   1570 	 * an ARP request.
   1571 	 */
   1572 	lle->base.la_expire = time_uptime; /* mark expired */
   1573 	lle->base.r_l3addr.addr4 = addr4;
   1574 	lle->base.lle_refcnt = 1;
   1575 	lle->base.lle_free = in_lltable_destroy_lle;
   1576 	LLE_LOCK_INIT(&lle->base);
   1577 	callout_init(&lle->base.la_timer, 1);
   1578 
   1579 	return (&lle->base);
   1580 }
   1581 
   1582 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m)	(			\
   1583 	    (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 )
   1584 
   1585 static int
   1586 in_lltable_match_prefix(const struct sockaddr *prefix,
   1587     const struct sockaddr *mask, u_int flags, struct llentry *lle)
   1588 {
   1589 	const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix;
   1590 	const struct sockaddr_in *msk = (const struct sockaddr_in *)mask;
   1591 
   1592 	/*
   1593 	 * (flags & LLE_STATIC) means deleting all entries
   1594 	 * including static ARP entries.
   1595 	 */
   1596 	if (IN_ARE_MASKED_ADDR_EQUAL(lle->r_l3addr.addr4, pfx, msk) &&
   1597 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
   1598 		return (1);
   1599 
   1600 	return (0);
   1601 }
   1602 
   1603 static void
   1604 in_lltable_free_entry(struct lltable *llt, struct llentry *lle)
   1605 {
   1606 	struct ifnet *ifp;
   1607 	size_t pkts_dropped;
   1608 
   1609 	LLE_WLOCK_ASSERT(lle);
   1610 	KASSERT(llt != NULL);
   1611 
   1612 	/* Unlink entry from table if not already */
   1613 	if ((lle->la_flags & LLE_LINKED) != 0) {
   1614 		ifp = llt->llt_ifp;
   1615 		IF_AFDATA_WLOCK_ASSERT(ifp);
   1616 		lltable_unlink_entry(llt, lle);
   1617 	}
   1618 
   1619 	/* cancel timer */
   1620 	if (callout_stop(&lle->lle_timer))
   1621 		LLE_REMREF(lle);
   1622 
   1623 	/* Drop hold queue */
   1624 	pkts_dropped = llentry_free(lle);
   1625 #ifdef __FreeBSD__
   1626 	ARPSTAT_ADD(dropped, pkts_dropped);
   1627 #else
   1628 	(void) pkts_dropped; /* FIXME */
   1629 #endif
   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 /* XXX */
   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 #ifdef __FreeBSD__
   1703 	RTFREE_LOCKED(rt);
   1704 #endif
   1705 	return error;
   1706 }
   1707 
   1708 static inline uint32_t
   1709 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize)
   1710 {
   1711 
   1712 	return (IN_LLTBL_HASH(dst.s_addr, hsize));
   1713 }
   1714 
   1715 static uint32_t
   1716 in_lltable_hash(const struct llentry *lle, uint32_t hsize)
   1717 {
   1718 
   1719 	return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize));
   1720 }
   1721 
   1722 static void
   1723 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
   1724 {
   1725 	struct sockaddr_in *sin;
   1726 
   1727 	sin = (struct sockaddr_in *)sa;
   1728 	memset(sin, 0, sizeof(*sin));
   1729 	sin->sin_family = AF_INET;
   1730 	sin->sin_len = sizeof(*sin);
   1731 	sin->sin_addr = lle->r_l3addr.addr4;
   1732 }
   1733 
   1734 static inline struct llentry *
   1735 in_lltable_find_dst(struct lltable *llt, struct in_addr dst)
   1736 {
   1737 	struct llentry *lle;
   1738 	struct llentries *lleh;
   1739 	u_int hashidx;
   1740 
   1741 	hashidx = in_lltable_hash_dst(dst, llt->llt_hsize);
   1742 	lleh = &llt->lle_head[hashidx];
   1743 	LIST_FOREACH(lle, lleh, lle_next) {
   1744 		if (lle->la_flags & LLE_DELETED)
   1745 			continue;
   1746 		if (lle->r_l3addr.addr4.s_addr == dst.s_addr)
   1747 			break;
   1748 	}
   1749 
   1750 	return (lle);
   1751 }
   1752 
   1753 static int
   1754 in_lltable_delete(struct lltable *llt, u_int flags,
   1755     const struct sockaddr *l3addr)
   1756 {
   1757 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1758 	struct ifnet *ifp = llt->llt_ifp;
   1759 	struct llentry *lle;
   1760 
   1761 	IF_AFDATA_WLOCK_ASSERT(ifp);
   1762 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1763 	    "sin_family %d", l3addr->sa_family);
   1764 
   1765 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1766 	if (lle == NULL) {
   1767 #ifdef DIAGNOSTIC
   1768 		log(LOG_INFO, "interface address is missing from cache = %p  in delete\n", lle);
   1769 #endif
   1770 		return (ENOENT);
   1771 	}
   1772 
   1773 	if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
   1774 		LLE_WLOCK(lle);
   1775 		lle->la_flags |= LLE_DELETED;
   1776 #ifdef __FreeBSD__
   1777 		EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
   1778 #endif
   1779 #ifdef DIAGNOSTIC
   1780 		log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
   1781 #endif
   1782 		if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
   1783 			llentry_free(lle);
   1784 		else
   1785 			LLE_WUNLOCK(lle);
   1786 	}
   1787 
   1788 	return (0);
   1789 }
   1790 
   1791 static struct llentry *
   1792 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
   1793 {
   1794 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1795 	struct ifnet *ifp = llt->llt_ifp;
   1796 	struct llentry *lle;
   1797 
   1798 	IF_AFDATA_WLOCK_ASSERT(ifp);
   1799 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1800 	    "sin_family %d", l3addr->sa_family);
   1801 
   1802 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1803 
   1804 	if (lle != NULL) {
   1805 		LLE_WLOCK(lle);
   1806 		return (lle);
   1807 	}
   1808 
   1809 	/* no existing record, we need to create new one */
   1810 
   1811 	/*
   1812 	 * A route that covers the given address must have
   1813 	 * been installed 1st because we are doing a resolution,
   1814 	 * verify this.
   1815 	 */
   1816 	if (!(flags & LLE_IFADDR) &&
   1817 	    in_lltable_rtcheck(ifp, flags, l3addr) != 0)
   1818 		return (NULL);
   1819 
   1820 	lle = in_lltable_new(sin->sin_addr, flags);
   1821 	if (lle == NULL) {
   1822 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
   1823 		return (NULL);
   1824 	}
   1825 	lle->la_flags = flags;
   1826 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
   1827 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
   1828 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
   1829 	}
   1830 
   1831 	lltable_link_entry(llt, lle);
   1832 	LLE_WLOCK(lle);
   1833 
   1834 	return (lle);
   1835 }
   1836 
   1837 /*
   1838  * Return NULL if not found or marked for deletion.
   1839  * If found return lle read locked.
   1840  */
   1841 static struct llentry *
   1842 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr)
   1843 {
   1844 	const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr;
   1845 	struct llentry *lle;
   1846 
   1847 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
   1848 	KASSERTMSG(l3addr->sa_family == AF_INET,
   1849 	    "sin_family %d", l3addr->sa_family);
   1850 
   1851 	lle = in_lltable_find_dst(llt, sin->sin_addr);
   1852 
   1853 	if (lle == NULL)
   1854 		return NULL;
   1855 
   1856 	if (flags & LLE_EXCLUSIVE)
   1857 		LLE_WLOCK(lle);
   1858 	else
   1859 		LLE_RLOCK(lle);
   1860 
   1861 	return lle;
   1862 }
   1863 
   1864 static void
   1865 in_sysctl_init(struct sysctllog **clog)
   1866 {
   1867 	sysctl_createv(clog, 0, NULL, NULL,
   1868 		       CTLFLAG_PERMANENT,
   1869 		       CTLTYPE_NODE, "inet",
   1870 		       SYSCTL_DESCR("PF_INET related settings"),
   1871 		       NULL, 0, NULL, 0,
   1872 		       CTL_NET, PF_INET, CTL_EOL);
   1873 	sysctl_createv(clog, 0, NULL, NULL,
   1874 		       CTLFLAG_PERMANENT,
   1875 		       CTLTYPE_NODE, "ip",
   1876 		       SYSCTL_DESCR("IPv4 related settings"),
   1877 		       NULL, 0, NULL, 0,
   1878 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
   1879 
   1880 	sysctl_createv(clog, 0, NULL, NULL,
   1881 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1882 		       CTLTYPE_INT, "subnetsarelocal",
   1883 		       SYSCTL_DESCR("Whether logical subnets are considered "
   1884 				    "local"),
   1885 		       NULL, 0, &subnetsarelocal, 0,
   1886 		       CTL_NET, PF_INET, IPPROTO_IP,
   1887 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
   1888 	sysctl_createv(clog, 0, NULL, NULL,
   1889 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1890 		       CTLTYPE_INT, "hostzerobroadcast",
   1891 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
   1892 		       NULL, 0, &hostzeroisbroadcast, 0,
   1893 		       CTL_NET, PF_INET, IPPROTO_IP,
   1894 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
   1895 }
   1896 
   1897 static struct lltable *
   1898 in_lltattach(struct ifnet *ifp)
   1899 {
   1900 	struct lltable *llt;
   1901 
   1902 	llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE);
   1903 	llt->llt_af = AF_INET;
   1904 	llt->llt_ifp = ifp;
   1905 
   1906 	llt->llt_lookup = in_lltable_lookup;
   1907 	llt->llt_create = in_lltable_create;
   1908 	llt->llt_delete = in_lltable_delete;
   1909 #if 0
   1910 	llt->llt_dump_entry = in_lltable_dump_entry;
   1911 #endif
   1912 	llt->llt_hash = in_lltable_hash;
   1913 	llt->llt_fill_sa_entry = in_lltable_fill_sa_entry;
   1914 	llt->llt_free_entry = in_lltable_free_entry;
   1915 	llt->llt_match_prefix = in_lltable_match_prefix;
   1916 	lltable_link(llt);
   1917 
   1918 	return (llt);
   1919 }
   1920 
   1921 void *
   1922 in_domifattach(struct ifnet *ifp)
   1923 {
   1924 	struct in_ifinfo *ii;
   1925 
   1926 	ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP);
   1927 	KASSERT(ii != NULL);
   1928 
   1929 	ii->ii_llt = in_lltattach(ifp);
   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 	lltable_free(ii->ii_llt);
   1948 	kmem_free(ii, sizeof(struct in_ifinfo));
   1949 }
   1950