Home | History | Annotate | Line # | Download | only in netinet
in.c revision 1.152
      1 /*	$NetBSD: in.c,v 1.152 2015/05/02 14:41:32 roy 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 1.152 2015/05/02 14:41:32 roy Exp $");
     95 
     96 #include "opt_inet.h"
     97 #include "opt_inet_conf.h"
     98 #include "opt_mrouting.h"
     99 
    100 #include <sys/param.h>
    101 #include <sys/ioctl.h>
    102 #include <sys/errno.h>
    103 #include <sys/kernel.h>
    104 #include <sys/malloc.h>
    105 #include <sys/socket.h>
    106 #include <sys/socketvar.h>
    107 #include <sys/sysctl.h>
    108 #include <sys/systm.h>
    109 #include <sys/proc.h>
    110 #include <sys/syslog.h>
    111 #include <sys/kauth.h>
    112 
    113 #include <sys/cprng.h>
    114 
    115 #include <net/if.h>
    116 #include <net/route.h>
    117 #include <net/pfil.h>
    118 
    119 #include <net/if_ether.h>
    120 
    121 #include <netinet/in_systm.h>
    122 #include <netinet/in.h>
    123 #include <netinet/in_var.h>
    124 #include <netinet/ip.h>
    125 #include <netinet/ip_var.h>
    126 #include <netinet/in_ifattach.h>
    127 #include <netinet/in_pcb.h>
    128 #include <netinet/if_inarp.h>
    129 #include <netinet/ip_mroute.h>
    130 #include <netinet/igmp_var.h>
    131 
    132 #ifdef IPSELSRC
    133 #include <netinet/in_selsrc.h>
    134 #endif
    135 
    136 static u_int	in_mask2len(struct in_addr *);
    137 static void	in_len2mask(struct in_addr *, u_int);
    138 static int	in_lifaddr_ioctl(struct socket *, u_long, void *,
    139 	struct ifnet *);
    140 
    141 static int	in_addprefix(struct in_ifaddr *, int);
    142 static int	in_scrubprefix(struct in_ifaddr *);
    143 static void	in_sysctl_init(struct sysctllog **);
    144 
    145 #ifndef SUBNETSARELOCAL
    146 #define	SUBNETSARELOCAL	1
    147 #endif
    148 
    149 #ifndef HOSTZEROBROADCAST
    150 #define HOSTZEROBROADCAST 1
    151 #endif
    152 
    153 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */
    154 #ifndef IN_MULTI_HASH_SIZE
    155 #define IN_MULTI_HASH_SIZE	509
    156 #endif
    157 
    158 static int			subnetsarelocal = SUBNETSARELOCAL;
    159 static int			hostzeroisbroadcast = HOSTZEROBROADCAST;
    160 
    161 /*
    162  * This list is used to keep track of in_multi chains which belong to
    163  * deleted interface addresses.  We use in_ifaddr so that a chain head
    164  * won't be deallocated until all multicast address record are deleted.
    165  */
    166 
    167 LIST_HEAD(in_multihashhead, in_multi);		/* Type of the hash head */
    168 
    169 static struct pool		inmulti_pool;
    170 static u_int			in_multientries;
    171 static struct in_multihashhead *in_multihashtbl;
    172 static u_long			in_multihash;
    173 static krwlock_t		in_multilock;
    174 
    175 #define IN_MULTI_HASH(x, ifp) \
    176     (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE])
    177 
    178 struct in_ifaddrhashhead *	in_ifaddrhashtbl;
    179 u_long				in_ifaddrhash;
    180 struct in_ifaddrhead		in_ifaddrhead;
    181 
    182 void
    183 in_init(void)
    184 {
    185 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
    186 	    NULL, IPL_SOFTNET);
    187 	TAILQ_INIT(&in_ifaddrhead);
    188 
    189 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    190 	    &in_ifaddrhash);
    191 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
    192 	    &in_multihash);
    193 	rw_init(&in_multilock);
    194 
    195 	in_sysctl_init(NULL);
    196 }
    197 
    198 /*
    199  * Return 1 if an internet address is for a ``local'' host
    200  * (one to which we have a connection).  If subnetsarelocal
    201  * is true, this includes other subnets of the local net.
    202  * Otherwise, it includes only the directly-connected (sub)nets.
    203  */
    204 int
    205 in_localaddr(struct in_addr in)
    206 {
    207 	struct in_ifaddr *ia;
    208 
    209 	if (subnetsarelocal) {
    210 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    211 			if ((in.s_addr & ia->ia_netmask) == ia->ia_net)
    212 				return (1);
    213 	} else {
    214 		TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list)
    215 			if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet)
    216 				return (1);
    217 	}
    218 	return (0);
    219 }
    220 
    221 /*
    222  * Determine whether an IP address is in a reserved set of addresses
    223  * that may not be forwarded, or whether datagrams to that destination
    224  * may be forwarded.
    225  */
    226 int
    227 in_canforward(struct in_addr in)
    228 {
    229 	u_int32_t net;
    230 
    231 	if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr))
    232 		return (0);
    233 	if (IN_CLASSA(in.s_addr)) {
    234 		net = in.s_addr & IN_CLASSA_NET;
    235 		if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))
    236 			return (0);
    237 	}
    238 	return (1);
    239 }
    240 
    241 /*
    242  * Trim a mask in a sockaddr
    243  */
    244 void
    245 in_socktrim(struct sockaddr_in *ap)
    246 {
    247 	char *cplim = (char *) &ap->sin_addr;
    248 	char *cp = (char *) (&ap->sin_addr + 1);
    249 
    250 	ap->sin_len = 0;
    251 	while (--cp >= cplim)
    252 		if (*cp) {
    253 			(ap)->sin_len = cp - (char *) (ap) + 1;
    254 			break;
    255 		}
    256 }
    257 
    258 /*
    259  *  Routine to take an Internet address and convert into a
    260  *  "dotted quad" representation for printing.
    261  */
    262 const char *
    263 in_fmtaddr(struct in_addr addr)
    264 {
    265 	static char buf[sizeof("123.456.789.123")];
    266 
    267 	addr.s_addr = ntohl(addr.s_addr);
    268 
    269 	snprintf(buf, sizeof(buf), "%d.%d.%d.%d",
    270 		(addr.s_addr >> 24) & 0xFF,
    271 		(addr.s_addr >> 16) & 0xFF,
    272 		(addr.s_addr >>  8) & 0xFF,
    273 		(addr.s_addr >>  0) & 0xFF);
    274 	return buf;
    275 }
    276 
    277 /*
    278  * Maintain the "in_maxmtu" variable, which is the largest
    279  * mtu for non-local interfaces with AF_INET addresses assigned
    280  * to them that are up.
    281  */
    282 unsigned long in_maxmtu;
    283 
    284 void
    285 in_setmaxmtu(void)
    286 {
    287 	struct in_ifaddr *ia;
    288 	struct ifnet *ifp;
    289 	unsigned long maxmtu = 0;
    290 
    291 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
    292 		if ((ifp = ia->ia_ifp) == 0)
    293 			continue;
    294 		if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP)
    295 			continue;
    296 		if (ifp->if_mtu > maxmtu)
    297 			maxmtu = ifp->if_mtu;
    298 	}
    299 	if (maxmtu)
    300 		in_maxmtu = maxmtu;
    301 }
    302 
    303 static u_int
    304 in_mask2len(struct in_addr *mask)
    305 {
    306 	u_int x, y;
    307 	u_char *p;
    308 
    309 	p = (u_char *)mask;
    310 	for (x = 0; x < sizeof(*mask); x++) {
    311 		if (p[x] != 0xff)
    312 			break;
    313 	}
    314 	y = 0;
    315 	if (x < sizeof(*mask)) {
    316 		for (y = 0; y < NBBY; y++) {
    317 			if ((p[x] & (0x80 >> y)) == 0)
    318 				break;
    319 		}
    320 	}
    321 	return x * NBBY + y;
    322 }
    323 
    324 static void
    325 in_len2mask(struct in_addr *mask, u_int len)
    326 {
    327 	u_int i;
    328 	u_char *p;
    329 
    330 	p = (u_char *)mask;
    331 	memset(mask, 0, sizeof(*mask));
    332 	for (i = 0; i < len / NBBY; i++)
    333 		p[i] = 0xff;
    334 	if (len % NBBY)
    335 		p[i] = (0xff00 >> (len % NBBY)) & 0xff;
    336 }
    337 
    338 /*
    339  * Generic internet control operations (ioctl's).
    340  * Ifp is 0 if not an interface-specific ioctl.
    341  */
    342 /* ARGSUSED */
    343 int
    344 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
    345 {
    346 	struct ifreq *ifr = (struct ifreq *)data;
    347 	struct in_ifaddr *ia = NULL;
    348 	struct in_aliasreq *ifra = (struct in_aliasreq *)data;
    349 	struct sockaddr_in oldaddr;
    350 	int error, hostIsNew, maskIsNew;
    351 	int newifaddr = 0;
    352 
    353 	switch (cmd) {
    354 	case SIOCALIFADDR:
    355 	case SIOCDLIFADDR:
    356 	case SIOCGLIFADDR:
    357 		if (ifp == NULL)
    358 			return EINVAL;
    359 		return in_lifaddr_ioctl(so, cmd, data, ifp);
    360 	case SIOCGIFADDRPREF:
    361 	case SIOCSIFADDRPREF:
    362 		if (ifp == NULL)
    363 			return EINVAL;
    364 		return ifaddrpref_ioctl(so, cmd, data, ifp);
    365 	}
    366 
    367 	/*
    368 	 * Find address for this interface, if it exists.
    369 	 */
    370 	if (ifp != NULL)
    371 		IFP_TO_IA(ifp, ia);
    372 
    373 	switch (cmd) {
    374 	case SIOCAIFADDR:
    375 	case SIOCDIFADDR:
    376 	case SIOCGIFALIAS:
    377 	case SIOCGIFAFLAG_IN:
    378 		if (ifra->ifra_addr.sin_family == AF_INET)
    379 			LIST_FOREACH(ia,
    380 			    &IN_IFADDR_HASH(ifra->ifra_addr.sin_addr.s_addr),
    381 			    ia_hash) {
    382 				if (ia->ia_ifp == ifp &&
    383 				    in_hosteq(ia->ia_addr.sin_addr,
    384 				    ifra->ifra_addr.sin_addr))
    385 					break;
    386 			}
    387 		if ((cmd == SIOCDIFADDR ||
    388 		    cmd == SIOCGIFALIAS ||
    389 		    cmd == SIOCGIFAFLAG_IN) &&
    390 		    ia == NULL)
    391 			return (EADDRNOTAVAIL);
    392 
    393 		if (cmd == SIOCDIFADDR &&
    394 		    ifra->ifra_addr.sin_family == AF_UNSPEC) {
    395 			ifra->ifra_addr.sin_family = AF_INET;
    396 		}
    397 		/* FALLTHROUGH */
    398 	case SIOCSIFADDR:
    399 		hostIsNew = 1;
    400 		if (ia == NULL || ia->ia_addr.sin_family != AF_INET)
    401 			;
    402 		else if (ifra->ifra_addr.sin_len == 0) {
    403 			ifra->ifra_addr = ia->ia_addr;
    404 			hostIsNew = 0;
    405 		} else if (in_hosteq(ia->ia_addr.sin_addr,
    406 		           ifra->ifra_addr.sin_addr))
    407 			hostIsNew = 0;
    408 		/* FALLTHROUGH */
    409 	case SIOCSIFDSTADDR:
    410 		if (ifra->ifra_addr.sin_family != AF_INET)
    411 			return (EAFNOSUPPORT);
    412 		/* FALLTHROUGH */
    413 	case SIOCSIFNETMASK:
    414 		if (ifp == NULL)
    415 			panic("in_control");
    416 
    417 		if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN)
    418 			break;
    419 
    420 		if (ia == NULL &&
    421 		    (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR))
    422 			return (EADDRNOTAVAIL);
    423 
    424 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    425 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    426 		    NULL) != 0)
    427 			return (EPERM);
    428 
    429 		if (ia == NULL) {
    430 			ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO);
    431 			if (ia == NULL)
    432 				return (ENOBUFS);
    433 			TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list);
    434 			ifaref(&ia->ia_ifa);
    435 			ifa_insert(ifp, &ia->ia_ifa);
    436 			ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    437 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    438 			ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask);
    439 #ifdef IPSELSRC
    440 			ia->ia_ifa.ifa_getifa = in_getifa;
    441 #else /* IPSELSRC */
    442 			ia->ia_ifa.ifa_getifa = NULL;
    443 #endif /* IPSELSRC */
    444 			ia->ia_sockmask.sin_len = 8;
    445 			ia->ia_sockmask.sin_family = AF_INET;
    446 			if (ifp->if_flags & IFF_BROADCAST) {
    447 				ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr);
    448 				ia->ia_broadaddr.sin_family = AF_INET;
    449 			}
    450 			ia->ia_ifp = ifp;
    451 			ia->ia_idsalt = cprng_fast32() % 65535;
    452 			LIST_INIT(&ia->ia_multiaddrs);
    453 			newifaddr = 1;
    454 		}
    455 		break;
    456 
    457 	case SIOCSIFBRDADDR:
    458 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
    459 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
    460 		    NULL) != 0)
    461 			return (EPERM);
    462 		/* FALLTHROUGH */
    463 
    464 	case SIOCGIFADDR:
    465 	case SIOCGIFNETMASK:
    466 	case SIOCGIFDSTADDR:
    467 	case SIOCGIFBRDADDR:
    468 		if (ia == NULL)
    469 			return (EADDRNOTAVAIL);
    470 		break;
    471 	}
    472 	error = 0;
    473 	switch (cmd) {
    474 
    475 	case SIOCGIFADDR:
    476 		ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr));
    477 		break;
    478 
    479 	case SIOCGIFBRDADDR:
    480 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    481 			return (EINVAL);
    482 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr));
    483 		break;
    484 
    485 	case SIOCGIFDSTADDR:
    486 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    487 			return (EINVAL);
    488 		ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr));
    489 		break;
    490 
    491 	case SIOCGIFNETMASK:
    492 		/*
    493 		 * We keep the number of trailing zero bytes the sin_len field
    494 		 * of ia_sockmask, so we fix this before we pass it back to
    495 		 * userland.
    496 		 */
    497 		oldaddr = ia->ia_sockmask;
    498 		oldaddr.sin_len = sizeof(struct sockaddr_in);
    499 		ifreq_setaddr(cmd, ifr, (const void *)&oldaddr);
    500 		break;
    501 
    502 	case SIOCSIFDSTADDR:
    503 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    504 			return (EINVAL);
    505 		oldaddr = ia->ia_dstaddr;
    506 		ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr));
    507 		if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) {
    508 			ia->ia_dstaddr = oldaddr;
    509 			return error;
    510 		}
    511 		if (ia->ia_flags & IFA_ROUTE) {
    512 			ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr);
    513 			rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST);
    514 			ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr);
    515 			rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP);
    516 		}
    517 		break;
    518 
    519 	case SIOCSIFBRDADDR:
    520 		if ((ifp->if_flags & IFF_BROADCAST) == 0)
    521 			return EINVAL;
    522 		ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr));
    523 		break;
    524 
    525 	case SIOCSIFADDR:
    526 		error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)),
    527 		    1, hostIsNew);
    528 		if (error == 0) {
    529 			(void)pfil_run_hooks(if_pfil,
    530 			    (struct mbuf **)SIOCSIFADDR, ifp, PFIL_IFADDR);
    531 		}
    532 		break;
    533 
    534 	case SIOCSIFNETMASK:
    535 		in_ifscrub(ifp, ia);
    536 		ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr));
    537 		ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    538 		error = in_ifinit(ifp, ia, NULL, 0, 0);
    539 		break;
    540 
    541 	case SIOCAIFADDR:
    542 		maskIsNew = 0;
    543 		if (ifra->ifra_mask.sin_len) {
    544 			/* Only scrub if we control the prefix route,
    545 			 * otherwise userland gets a bogus message */
    546 			if ((ia->ia_flags & IFA_ROUTE))
    547 				in_ifscrub(ifp, ia);
    548 			ia->ia_sockmask = ifra->ifra_mask;
    549 			ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr;
    550 			maskIsNew = 1;
    551 		}
    552 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    553 		    (ifra->ifra_dstaddr.sin_family == AF_INET)) {
    554 			/* Only scrub if we control the prefix route,
    555 			 * otherwise userland gets a bogus message */
    556 			if ((ia->ia_flags & IFA_ROUTE))
    557 				in_ifscrub(ifp, ia);
    558 			ia->ia_dstaddr = ifra->ifra_dstaddr;
    559 			maskIsNew  = 1; /* We lie; but the effect's the same */
    560 		}
    561 		if (ifra->ifra_addr.sin_family == AF_INET &&
    562 		    (hostIsNew || maskIsNew)) {
    563 			error = in_ifinit(ifp, ia, &ifra->ifra_addr, 0,
    564 			    hostIsNew);
    565 		}
    566 		if ((ifp->if_flags & IFF_BROADCAST) &&
    567 		    (ifra->ifra_broadaddr.sin_family == AF_INET))
    568 			ia->ia_broadaddr = ifra->ifra_broadaddr;
    569 		if (error == 0)
    570 			(void)pfil_run_hooks(if_pfil,
    571 			    (struct mbuf **)SIOCAIFADDR, ifp, PFIL_IFADDR);
    572 		break;
    573 
    574 	case SIOCGIFALIAS:
    575 		ifra->ifra_mask = ia->ia_sockmask;
    576 		if ((ifp->if_flags & IFF_POINTOPOINT) &&
    577 		    (ia->ia_dstaddr.sin_family == AF_INET))
    578 			ifra->ifra_dstaddr = ia->ia_dstaddr;
    579 		else if ((ifp->if_flags & IFF_BROADCAST) &&
    580 		    (ia->ia_broadaddr.sin_family == AF_INET))
    581 			ifra->ifra_broadaddr = ia->ia_broadaddr;
    582 		else
    583 			memset(&ifra->ifra_broadaddr, 0,
    584 			      sizeof(ifra->ifra_broadaddr));
    585 		break;
    586 
    587 	case SIOCGIFAFLAG_IN:
    588 		ifr->ifr_addrflags = ia->ia4_flags;
    589 		break;
    590 
    591 	case SIOCDIFADDR:
    592 		in_purgeaddr(&ia->ia_ifa);
    593 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR,
    594 		    ifp, PFIL_IFADDR);
    595 		break;
    596 
    597 #ifdef MROUTING
    598 	case SIOCGETVIFCNT:
    599 	case SIOCGETSGCNT:
    600 		error = mrt_ioctl(so, cmd, data);
    601 		break;
    602 #endif /* MROUTING */
    603 
    604 	default:
    605 		return ENOTTY;
    606 	}
    607 
    608 	if (error != 0 && newifaddr) {
    609 		KASSERT(ia != NULL);
    610 		in_purgeaddr(&ia->ia_ifa);
    611 	}
    612 
    613 	return error;
    614 }
    615 
    616 /* Add ownaddr as loopback rtentry. */
    617 static void
    618 in_ifaddlocal(struct ifaddr *ifa)
    619 {
    620 	struct in_ifaddr *ia;
    621 
    622 	ia = (struct in_ifaddr *)ifa;
    623 	if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY ||
    624 	    (ia->ia_ifp->if_flags & IFF_POINTOPOINT &&
    625 	    in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr)))
    626 	{
    627 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
    628 		return;
    629 	}
    630 
    631 	rt_ifa_addlocal(ifa);
    632 }
    633 
    634 /* Rempve loopback entry of ownaddr */
    635 static void
    636 in_ifremlocal(struct ifaddr *ifa)
    637 {
    638 	struct in_ifaddr *ia, *p;
    639 	struct ifaddr *alt_ifa = NULL;
    640 	int ia_count = 0;
    641 
    642 	ia = (struct in_ifaddr *)ifa;
    643 	/* Delete the entry if exactly one ifaddr matches the
    644 	 * address, ifa->ifa_addr. */
    645 	TAILQ_FOREACH(p, &in_ifaddrhead, ia_list) {
    646 		if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr))
    647 			continue;
    648 		if (p->ia_ifp != ia->ia_ifp)
    649 			alt_ifa = &p->ia_ifa;
    650 		if (++ia_count > 1 && alt_ifa != NULL)
    651 			break;
    652 	}
    653 
    654 	if (ia_count == 0)
    655 		return;
    656 
    657 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
    658 }
    659 
    660 void
    661 in_purgeaddr(struct ifaddr *ifa)
    662 {
    663 	struct ifnet *ifp = ifa->ifa_ifp;
    664 	struct in_ifaddr *ia = (void *) ifa;
    665 
    666         /* stop DAD processing */
    667 	arp_dad_stop(ifa);
    668 
    669 	in_ifscrub(ifp, ia);
    670 	in_ifremlocal(ifa);
    671 	LIST_REMOVE(ia, ia_hash);
    672 	ifa_remove(ifp, &ia->ia_ifa);
    673 	TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list);
    674 	if (ia->ia_allhosts != NULL)
    675 		in_delmulti(ia->ia_allhosts);
    676 	ifafree(&ia->ia_ifa);
    677 	in_setmaxmtu();
    678 }
    679 
    680 void
    681 in_purgeif(struct ifnet *ifp)		/* MUST be called at splsoftnet() */
    682 {
    683 	if_purgeaddrs(ifp, AF_INET, in_purgeaddr);
    684 	igmp_purgeif(ifp);		/* manipulates pools */
    685 #ifdef MROUTING
    686 	ip_mrouter_detach(ifp);
    687 #endif
    688 }
    689 
    690 /*
    691  * SIOC[GAD]LIFADDR.
    692  *	SIOCGLIFADDR: get first address. (???)
    693  *	SIOCGLIFADDR with IFLR_PREFIX:
    694  *		get first address that matches the specified prefix.
    695  *	SIOCALIFADDR: add the specified address.
    696  *	SIOCALIFADDR with IFLR_PREFIX:
    697  *		EINVAL since we can't deduce hostid part of the address.
    698  *	SIOCDLIFADDR: delete the specified address.
    699  *	SIOCDLIFADDR with IFLR_PREFIX:
    700  *		delete the first address that matches the specified prefix.
    701  * return values:
    702  *	EINVAL on invalid parameters
    703  *	EADDRNOTAVAIL on prefix match failed/specified address not found
    704  *	other values may be returned from in_ioctl()
    705  */
    706 static int
    707 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
    708     struct ifnet *ifp)
    709 {
    710 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
    711 	struct ifaddr *ifa;
    712 	struct sockaddr *sa;
    713 
    714 	/* sanity checks */
    715 	if (data == NULL || ifp == NULL) {
    716 		panic("invalid argument to in_lifaddr_ioctl");
    717 		/*NOTRECHED*/
    718 	}
    719 
    720 	switch (cmd) {
    721 	case SIOCGLIFADDR:
    722 		/* address must be specified on GET with IFLR_PREFIX */
    723 		if ((iflr->flags & IFLR_PREFIX) == 0)
    724 			break;
    725 		/*FALLTHROUGH*/
    726 	case SIOCALIFADDR:
    727 	case SIOCDLIFADDR:
    728 		/* address must be specified on ADD and DELETE */
    729 		sa = (struct sockaddr *)&iflr->addr;
    730 		if (sa->sa_family != AF_INET)
    731 			return EINVAL;
    732 		if (sa->sa_len != sizeof(struct sockaddr_in))
    733 			return EINVAL;
    734 		/* XXX need improvement */
    735 		sa = (struct sockaddr *)&iflr->dstaddr;
    736 		if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET)
    737 			return EINVAL;
    738 		if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in))
    739 			return EINVAL;
    740 		break;
    741 	default: /*shouldn't happen*/
    742 #if 0
    743 		panic("invalid cmd to in_lifaddr_ioctl");
    744 		/*NOTREACHED*/
    745 #else
    746 		return EOPNOTSUPP;
    747 #endif
    748 	}
    749 	if (sizeof(struct in_addr) * NBBY < iflr->prefixlen)
    750 		return EINVAL;
    751 
    752 	switch (cmd) {
    753 	case SIOCALIFADDR:
    754 	    {
    755 		struct in_aliasreq ifra;
    756 
    757 		if (iflr->flags & IFLR_PREFIX)
    758 			return EINVAL;
    759 
    760 		/* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */
    761 		memset(&ifra, 0, sizeof(ifra));
    762 		memcpy(ifra.ifra_name, iflr->iflr_name,
    763 			sizeof(ifra.ifra_name));
    764 
    765 		memcpy(&ifra.ifra_addr, &iflr->addr,
    766 			((struct sockaddr *)&iflr->addr)->sa_len);
    767 
    768 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
    769 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
    770 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
    771 		}
    772 
    773 		ifra.ifra_mask.sin_family = AF_INET;
    774 		ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in);
    775 		in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen);
    776 
    777 		return in_control(so, SIOCAIFADDR, &ifra, ifp);
    778 	    }
    779 	case SIOCGLIFADDR:
    780 	case SIOCDLIFADDR:
    781 	    {
    782 		struct in_ifaddr *ia;
    783 		struct in_addr mask, candidate, match;
    784 		struct sockaddr_in *sin;
    785 		int cmp;
    786 
    787 		memset(&mask, 0, sizeof(mask));
    788 		memset(&match, 0, sizeof(match));	/* XXX gcc */
    789 		if (iflr->flags & IFLR_PREFIX) {
    790 			/* lookup a prefix rather than address. */
    791 			in_len2mask(&mask, iflr->prefixlen);
    792 
    793 			sin = (struct sockaddr_in *)&iflr->addr;
    794 			match.s_addr = sin->sin_addr.s_addr;
    795 			match.s_addr &= mask.s_addr;
    796 
    797 			/* if you set extra bits, that's wrong */
    798 			if (match.s_addr != sin->sin_addr.s_addr)
    799 				return EINVAL;
    800 
    801 			cmp = 1;
    802 		} else {
    803 			if (cmd == SIOCGLIFADDR) {
    804 				/* on getting an address, take the 1st match */
    805 				cmp = 0;	/*XXX*/
    806 			} else {
    807 				/* on deleting an address, do exact match */
    808 				in_len2mask(&mask, 32);
    809 				sin = (struct sockaddr_in *)&iflr->addr;
    810 				match.s_addr = sin->sin_addr.s_addr;
    811 
    812 				cmp = 1;
    813 			}
    814 		}
    815 
    816 		IFADDR_FOREACH(ifa, ifp) {
    817 			if (ifa->ifa_addr->sa_family != AF_INET)
    818 				continue;
    819 			if (cmp == 0)
    820 				break;
    821 			candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr;
    822 			candidate.s_addr &= mask.s_addr;
    823 			if (candidate.s_addr == match.s_addr)
    824 				break;
    825 		}
    826 		if (ifa == NULL)
    827 			return EADDRNOTAVAIL;
    828 		ia = (struct in_ifaddr *)ifa;
    829 
    830 		if (cmd == SIOCGLIFADDR) {
    831 			/* fill in the if_laddrreq structure */
    832 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len);
    833 
    834 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    835 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
    836 					ia->ia_dstaddr.sin_len);
    837 			} else
    838 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
    839 
    840 			iflr->prefixlen =
    841 				in_mask2len(&ia->ia_sockmask.sin_addr);
    842 
    843 			iflr->flags = 0;	/*XXX*/
    844 
    845 			return 0;
    846 		} else {
    847 			struct in_aliasreq ifra;
    848 
    849 			/* fill in_aliasreq and do ioctl(SIOCDIFADDR) */
    850 			memset(&ifra, 0, sizeof(ifra));
    851 			memcpy(ifra.ifra_name, iflr->iflr_name,
    852 				sizeof(ifra.ifra_name));
    853 
    854 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
    855 				ia->ia_addr.sin_len);
    856 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
    857 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
    858 					ia->ia_dstaddr.sin_len);
    859 			}
    860 			memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask,
    861 				ia->ia_sockmask.sin_len);
    862 
    863 			return in_control(so, SIOCDIFADDR, &ifra, ifp);
    864 		}
    865 	    }
    866 	}
    867 
    868 	return EOPNOTSUPP;	/*just for safety*/
    869 }
    870 
    871 /*
    872  * Delete any existing route for an interface.
    873  */
    874 void
    875 in_ifscrub(struct ifnet *ifp, struct in_ifaddr *ia)
    876 {
    877 
    878 	in_scrubprefix(ia);
    879 }
    880 
    881 /*
    882  * Initialize an interface's internet address
    883  * and routing table entry.
    884  */
    885 int
    886 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia,
    887     const struct sockaddr_in *sin, int scrub, int hostIsNew)
    888 {
    889 	u_int32_t i;
    890 	struct sockaddr_in oldaddr;
    891 	int s = splnet(), flags = RTF_UP, error;
    892 
    893 	if (sin == NULL)
    894 		sin = &ia->ia_addr;
    895 
    896 	/*
    897 	 * Set up new addresses.
    898 	 */
    899 	oldaddr = ia->ia_addr;
    900 	if (ia->ia_addr.sin_family == AF_INET)
    901 		LIST_REMOVE(ia, ia_hash);
    902 	ia->ia_addr = *sin;
    903 	LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, ia_hash);
    904 
    905 	/* Set IN_IFF flags early for arp_ifinit() */
    906 	if (hostIsNew && if_do_dad(ifp) && !in_nullhost(ia->ia_addr.sin_addr)) {
    907 		if (ifp->if_link_state == LINK_STATE_DOWN)
    908 			ia->ia4_flags |= IN_IFF_DETACHED;
    909 		else
    910 			ia->ia4_flags |= IN_IFF_TENTATIVE;
    911 	}
    912 
    913 	/*
    914 	 * Give the interface a chance to initialize
    915 	 * if this is its first address,
    916 	 * and to validate the address if necessary.
    917 	 */
    918 	if ((error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0)
    919 		goto bad;
    920 	splx(s);
    921 	if (scrub) {
    922 		ia->ia_ifa.ifa_addr = sintosa(&oldaddr);
    923 		in_ifscrub(ifp, ia);
    924 		ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr);
    925 	}
    926 
    927 	/* Add the local route to the address */
    928 	in_ifaddlocal(&ia->ia_ifa);
    929 
    930 	i = ia->ia_addr.sin_addr.s_addr;
    931 	if (IN_CLASSA(i))
    932 		ia->ia_netmask = IN_CLASSA_NET;
    933 	else if (IN_CLASSB(i))
    934 		ia->ia_netmask = IN_CLASSB_NET;
    935 	else
    936 		ia->ia_netmask = IN_CLASSC_NET;
    937 	/*
    938 	 * The subnet mask usually includes at least the standard network part,
    939 	 * but may may be smaller in the case of supernetting.
    940 	 * If it is set, we believe it.
    941 	 */
    942 	if (ia->ia_subnetmask == 0) {
    943 		ia->ia_subnetmask = ia->ia_netmask;
    944 		ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask;
    945 	} else
    946 		ia->ia_netmask &= ia->ia_subnetmask;
    947 
    948 	ia->ia_net = i & ia->ia_netmask;
    949 	ia->ia_subnet = i & ia->ia_subnetmask;
    950 	in_socktrim(&ia->ia_sockmask);
    951 	/* re-calculate the "in_maxmtu" value */
    952 	in_setmaxmtu();
    953 	/*
    954 	 * Add route for the network.
    955 	 */
    956 	ia->ia_ifa.ifa_metric = ifp->if_metric;
    957 	if (ifp->if_flags & IFF_BROADCAST) {
    958 		ia->ia_broadaddr.sin_addr.s_addr =
    959 			ia->ia_subnet | ~ia->ia_subnetmask;
    960 		ia->ia_netbroadcast.s_addr =
    961 			ia->ia_net | ~ia->ia_netmask;
    962 	} else if (ifp->if_flags & IFF_LOOPBACK) {
    963 		ia->ia_dstaddr = ia->ia_addr;
    964 		flags |= RTF_HOST;
    965 	} else if (ifp->if_flags & IFF_POINTOPOINT) {
    966 		if (ia->ia_dstaddr.sin_family != AF_INET)
    967 			return (0);
    968 		flags |= RTF_HOST;
    969 	}
    970 	error = in_addprefix(ia, flags);
    971 	/*
    972 	 * If the interface supports multicast, join the "all hosts"
    973 	 * multicast group on that interface.
    974 	 */
    975 	if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) {
    976 		struct in_addr addr;
    977 
    978 		addr.s_addr = INADDR_ALLHOSTS_GROUP;
    979 		ia->ia_allhosts = in_addmulti(&addr, ifp);
    980 	}
    981 
    982 	if (hostIsNew && if_do_dad(ifp) &&
    983 	    !in_nullhost(ia->ia_addr.sin_addr) &&
    984 	    ia->ia4_flags & IN_IFF_TENTATIVE)
    985 		arp_dad_start((struct ifaddr *)ia);
    986 
    987 	return (error);
    988 bad:
    989 	splx(s);
    990 	LIST_REMOVE(ia, ia_hash);
    991 	ia->ia_addr = oldaddr;
    992 	if (ia->ia_addr.sin_family == AF_INET)
    993 		LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr),
    994 		    ia, ia_hash);
    995 	return (error);
    996 }
    997 
    998 #define rtinitflags(x) \
    999 	((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \
   1000 	    ? RTF_HOST : 0)
   1001 
   1002 /*
   1003  * add a route to prefix ("connected route" in cisco terminology).
   1004  * does nothing if there's some interface address with the same prefix already.
   1005  */
   1006 static int
   1007 in_addprefix(struct in_ifaddr *target, int flags)
   1008 {
   1009 	struct in_ifaddr *ia;
   1010 	struct in_addr prefix, mask, p;
   1011 	int error;
   1012 
   1013 	if ((flags & RTF_HOST) != 0)
   1014 		prefix = target->ia_dstaddr.sin_addr;
   1015 	else {
   1016 		prefix = target->ia_addr.sin_addr;
   1017 		mask = target->ia_sockmask.sin_addr;
   1018 		prefix.s_addr &= mask.s_addr;
   1019 	}
   1020 
   1021 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1022 		if (rtinitflags(ia))
   1023 			p = ia->ia_dstaddr.sin_addr;
   1024 		else {
   1025 			p = ia->ia_addr.sin_addr;
   1026 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1027 		}
   1028 
   1029 		if (prefix.s_addr != p.s_addr)
   1030 			continue;
   1031 
   1032 		/*
   1033 		 * if we got a matching prefix route inserted by other
   1034 		 * interface address, we don't need to bother
   1035 		 *
   1036 		 * XXX RADIX_MPATH implications here? -dyoung
   1037 		 */
   1038 		if (ia->ia_flags & IFA_ROUTE)
   1039 			return 0;
   1040 	}
   1041 
   1042 	/*
   1043 	 * noone seem to have prefix route.  insert it.
   1044 	 */
   1045 	error = rtinit(&target->ia_ifa, RTM_ADD, flags);
   1046 	if (error == 0)
   1047 		target->ia_flags |= IFA_ROUTE;
   1048 	else if (error == EEXIST) {
   1049 		/*
   1050 		 * the fact the route already exists is not an error.
   1051 		 */
   1052 		error = 0;
   1053 	}
   1054 	return error;
   1055 }
   1056 
   1057 /*
   1058  * remove a route to prefix ("connected route" in cisco terminology).
   1059  * re-installs the route by using another interface address, if there's one
   1060  * with the same prefix (otherwise we lose the route mistakenly).
   1061  */
   1062 static int
   1063 in_scrubprefix(struct in_ifaddr *target)
   1064 {
   1065 	struct in_ifaddr *ia;
   1066 	struct in_addr prefix, mask, p;
   1067 	int error;
   1068 
   1069 	/* If we don't have IFA_ROUTE we should still inform userland */
   1070 	if ((target->ia_flags & IFA_ROUTE) == 0)
   1071 		return 0;
   1072 
   1073 	if (rtinitflags(target))
   1074 		prefix = target->ia_dstaddr.sin_addr;
   1075 	else {
   1076 		prefix = target->ia_addr.sin_addr;
   1077 		mask = target->ia_sockmask.sin_addr;
   1078 		prefix.s_addr &= mask.s_addr;
   1079 	}
   1080 
   1081 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1082 		if (rtinitflags(ia))
   1083 			p = ia->ia_dstaddr.sin_addr;
   1084 		else {
   1085 			p = ia->ia_addr.sin_addr;
   1086 			p.s_addr &= ia->ia_sockmask.sin_addr.s_addr;
   1087 		}
   1088 
   1089 		if (prefix.s_addr != p.s_addr)
   1090 			continue;
   1091 
   1092 		/*
   1093 		 * if we got a matching prefix route, move IFA_ROUTE to him
   1094 		 */
   1095 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
   1096 			rtinit(&target->ia_ifa, RTM_DELETE,
   1097 			    rtinitflags(target));
   1098 			target->ia_flags &= ~IFA_ROUTE;
   1099 
   1100 			error = rtinit(&ia->ia_ifa, RTM_ADD,
   1101 			    rtinitflags(ia) | RTF_UP);
   1102 			if (error == 0)
   1103 				ia->ia_flags |= IFA_ROUTE;
   1104 			return error;
   1105 		}
   1106 	}
   1107 
   1108 	/*
   1109 	 * noone seem to have prefix route.  remove it.
   1110 	 */
   1111 	rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target));
   1112 	target->ia_flags &= ~IFA_ROUTE;
   1113 	return 0;
   1114 }
   1115 
   1116 #undef rtinitflags
   1117 
   1118 /*
   1119  * Return 1 if the address might be a local broadcast address.
   1120  */
   1121 int
   1122 in_broadcast(struct in_addr in, struct ifnet *ifp)
   1123 {
   1124 	struct ifaddr *ifa;
   1125 
   1126 	if (in.s_addr == INADDR_BROADCAST ||
   1127 	    in_nullhost(in))
   1128 		return 1;
   1129 	if ((ifp->if_flags & IFF_BROADCAST) == 0)
   1130 		return 0;
   1131 	/*
   1132 	 * Look through the list of addresses for a match
   1133 	 * with a broadcast address.
   1134 	 */
   1135 #define ia (ifatoia(ifa))
   1136 	IFADDR_FOREACH(ifa, ifp)
   1137 		if (ifa->ifa_addr->sa_family == AF_INET &&
   1138 		    !in_hosteq(in, ia->ia_addr.sin_addr) &&
   1139 		    (in_hosteq(in, ia->ia_broadaddr.sin_addr) ||
   1140 		     in_hosteq(in, ia->ia_netbroadcast) ||
   1141 		     (hostzeroisbroadcast &&
   1142 		      /*
   1143 		       * Check for old-style (host 0) broadcast.
   1144 		       */
   1145 		      (in.s_addr == ia->ia_subnet ||
   1146 		       in.s_addr == ia->ia_net))))
   1147 			return 1;
   1148 	return (0);
   1149 #undef ia
   1150 }
   1151 
   1152 /*
   1153  * perform DAD when interface becomes IFF_UP.
   1154  */
   1155 void
   1156 in_if_link_up(struct ifnet *ifp)
   1157 {
   1158 	struct ifaddr *ifa;
   1159 	struct in_ifaddr *ia;
   1160 
   1161 	/* Ensure it's sane to run DAD */
   1162 	if (ifp->if_link_state == LINK_STATE_DOWN)
   1163 		return;
   1164 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
   1165 		return;
   1166 
   1167 	IFADDR_FOREACH(ifa, ifp) {
   1168 		if (ifa->ifa_addr->sa_family != AF_INET)
   1169 			continue;
   1170 		ia = (struct in_ifaddr *)ifa;
   1171 
   1172 		/* If detached then mark as tentative */
   1173 		if (ia->ia4_flags & IN_IFF_DETACHED) {
   1174 			ia->ia4_flags &= ~IN_IFF_DETACHED;
   1175 			if (if_do_dad(ifp))
   1176 				ia->ia4_flags |= IN_IFF_TENTATIVE;
   1177 			else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0)
   1178 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1179 		}
   1180 
   1181 		if (ia->ia4_flags & IN_IFF_TENTATIVE) {
   1182 			/* Clear the duplicated flag as we're starting DAD. */
   1183 			ia->ia4_flags &= ~IN_IFF_DUPLICATED;
   1184 			arp_dad_start(ifa);
   1185 		}
   1186 	}
   1187 }
   1188 
   1189 void
   1190 in_if_up(struct ifnet *ifp)
   1191 {
   1192 
   1193 	/* interface may not support link state, so bring it up also */
   1194 	in_if_link_up(ifp);
   1195 }
   1196 
   1197 /*
   1198  * Mark all addresses as detached.
   1199  */
   1200 void
   1201 in_if_link_down(struct ifnet *ifp)
   1202 {
   1203 	struct ifaddr *ifa;
   1204 	struct in_ifaddr *ia;
   1205 
   1206 	IFADDR_FOREACH(ifa, ifp) {
   1207 		if (ifa->ifa_addr->sa_family != AF_INET)
   1208 			continue;
   1209 		ia = (struct in_ifaddr *)ifa;
   1210 
   1211 		/* Stop DAD processing */
   1212 		arp_dad_stop(ifa);
   1213 
   1214 		/*
   1215 		 * Mark the address as detached.
   1216 		 */
   1217 		if (!(ia->ia4_flags & IN_IFF_DETACHED)) {
   1218 			ia->ia4_flags |= IN_IFF_DETACHED;
   1219 			ia->ia4_flags &=
   1220 			    ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED);
   1221 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1222 		}
   1223 	}
   1224 }
   1225 
   1226 void
   1227 in_if_down(struct ifnet *ifp)
   1228 {
   1229 
   1230 	in_if_link_down(ifp);
   1231 }
   1232 
   1233 void
   1234 in_if_link_state_change(struct ifnet *ifp, int link_state)
   1235 {
   1236 
   1237 	switch (link_state) {
   1238 	case LINK_STATE_DOWN:
   1239 		in_if_link_down(ifp);
   1240 		break;
   1241 	case LINK_STATE_UP:
   1242 		in_if_link_up(ifp);
   1243 		break;
   1244 	}
   1245 }
   1246 
   1247 /*
   1248  * in_lookup_multi: look up the in_multi record for a given IP
   1249  * multicast address on a given interface.  If no matching record is
   1250  * found, return NULL.
   1251  */
   1252 struct in_multi *
   1253 in_lookup_multi(struct in_addr addr, ifnet_t *ifp)
   1254 {
   1255 	struct in_multi *inm;
   1256 
   1257 	KASSERT(rw_lock_held(&in_multilock));
   1258 
   1259 	LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) {
   1260 		if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp)
   1261 			break;
   1262 	}
   1263 	return inm;
   1264 }
   1265 
   1266 /*
   1267  * in_multi_group: check whether the address belongs to an IP multicast
   1268  * group we are joined on this interface.  Returns true or false.
   1269  */
   1270 bool
   1271 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags)
   1272 {
   1273 	bool ingroup;
   1274 
   1275 	if (__predict_true(flags & IP_IGMP_MCAST) == 0) {
   1276 		rw_enter(&in_multilock, RW_READER);
   1277 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1278 		rw_exit(&in_multilock);
   1279 	} else {
   1280 		/* XXX Recursive call from ip_output(). */
   1281 		KASSERT(rw_lock_held(&in_multilock));
   1282 		ingroup = in_lookup_multi(addr, ifp) != NULL;
   1283 	}
   1284 	return ingroup;
   1285 }
   1286 
   1287 /*
   1288  * Add an address to the list of IP multicast addresses for a given interface.
   1289  */
   1290 struct in_multi *
   1291 in_addmulti(struct in_addr *ap, ifnet_t *ifp)
   1292 {
   1293 	struct sockaddr_in sin;
   1294 	struct in_multi *inm;
   1295 
   1296 	/*
   1297 	 * See if address already in list.
   1298 	 */
   1299 	rw_enter(&in_multilock, RW_WRITER);
   1300 	inm = in_lookup_multi(*ap, ifp);
   1301 	if (inm != NULL) {
   1302 		/*
   1303 		 * Found it; just increment the reference count.
   1304 		 */
   1305 		inm->inm_refcount++;
   1306 		rw_exit(&in_multilock);
   1307 		return inm;
   1308 	}
   1309 
   1310 	/*
   1311 	 * New address; allocate a new multicast record.
   1312 	 */
   1313 	inm = pool_get(&inmulti_pool, PR_NOWAIT);
   1314 	if (inm == NULL) {
   1315 		rw_exit(&in_multilock);
   1316 		return NULL;
   1317 	}
   1318 	inm->inm_addr = *ap;
   1319 	inm->inm_ifp = ifp;
   1320 	inm->inm_refcount = 1;
   1321 
   1322 	/*
   1323 	 * Ask the network driver to update its multicast reception
   1324 	 * filter appropriately for the new address.
   1325 	 */
   1326 	sockaddr_in_init(&sin, ap, 0);
   1327 	if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) {
   1328 		rw_exit(&in_multilock);
   1329 		pool_put(&inmulti_pool, inm);
   1330 		return NULL;
   1331 	}
   1332 
   1333 	/*
   1334 	 * Let IGMP know that we have joined a new IP multicast group.
   1335 	 */
   1336 	if (igmp_joingroup(inm) != 0) {
   1337 		rw_exit(&in_multilock);
   1338 		pool_put(&inmulti_pool, inm);
   1339 		return NULL;
   1340 	}
   1341 	LIST_INSERT_HEAD(
   1342 	    &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp),
   1343 	    inm, inm_list);
   1344 	in_multientries++;
   1345 	rw_exit(&in_multilock);
   1346 
   1347 	return inm;
   1348 }
   1349 
   1350 /*
   1351  * Delete a multicast address record.
   1352  */
   1353 void
   1354 in_delmulti(struct in_multi *inm)
   1355 {
   1356 	struct sockaddr_in sin;
   1357 
   1358 	rw_enter(&in_multilock, RW_WRITER);
   1359 	if (--inm->inm_refcount > 0) {
   1360 		rw_exit(&in_multilock);
   1361 		return;
   1362 	}
   1363 
   1364 	/*
   1365 	 * No remaining claims to this record; let IGMP know that
   1366 	 * we are leaving the multicast group.
   1367 	 */
   1368 	igmp_leavegroup(inm);
   1369 
   1370 	/*
   1371 	 * Notify the network driver to update its multicast reception
   1372 	 * filter.
   1373 	 */
   1374 	sockaddr_in_init(&sin, &inm->inm_addr, 0);
   1375 	if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin));
   1376 
   1377 	/*
   1378 	 * Unlink from list.
   1379 	 */
   1380 	LIST_REMOVE(inm, inm_list);
   1381 	in_multientries--;
   1382 	rw_exit(&in_multilock);
   1383 
   1384 	pool_put(&inmulti_pool, inm);
   1385 }
   1386 
   1387 /*
   1388  * in_next_multi: step through all of the in_multi records, one at a time.
   1389  * The current position is remembered in "step", which the caller must
   1390  * provide.  in_first_multi(), below, must be called to initialize "step"
   1391  * and get the first record.  Both macros return a NULL "inm" when there
   1392  * are no remaining records.
   1393  */
   1394 struct in_multi *
   1395 in_next_multi(struct in_multistep *step)
   1396 {
   1397 	struct in_multi *inm;
   1398 
   1399 	KASSERT(rw_lock_held(&in_multilock));
   1400 
   1401 	while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) {
   1402 		step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]);
   1403 	}
   1404 	if ((inm = step->i_inm) != NULL) {
   1405 		step->i_inm = LIST_NEXT(inm, inm_list);
   1406 	}
   1407 	return inm;
   1408 }
   1409 
   1410 struct in_multi *
   1411 in_first_multi(struct in_multistep *step)
   1412 {
   1413 	KASSERT(rw_lock_held(&in_multilock));
   1414 
   1415 	step->i_n = 0;
   1416 	step->i_inm = LIST_FIRST(&in_multihashtbl[0]);
   1417 	return in_next_multi(step);
   1418 }
   1419 
   1420 void
   1421 in_multi_lock(int op)
   1422 {
   1423 	rw_enter(&in_multilock, op);
   1424 }
   1425 
   1426 void
   1427 in_multi_unlock(void)
   1428 {
   1429 	rw_exit(&in_multilock);
   1430 }
   1431 
   1432 int
   1433 in_multi_lock_held(void)
   1434 {
   1435 	return rw_lock_held(&in_multilock);
   1436 }
   1437 
   1438 struct sockaddr_in *
   1439 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
   1440     int soopts, struct ip_moptions *mopts, int *errorp)
   1441 {
   1442 	struct rtentry *rt = NULL;
   1443 	struct in_ifaddr *ia = NULL;
   1444 
   1445 	/*
   1446          * If route is known or can be allocated now, take the
   1447          * source address from the interface.  Otherwise, punt.
   1448 	 */
   1449 	if ((soopts & SO_DONTROUTE) != 0)
   1450 		rtcache_free(ro);
   1451 	else {
   1452 		union {
   1453 			struct sockaddr		dst;
   1454 			struct sockaddr_in	dst4;
   1455 		} u;
   1456 
   1457 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
   1458 		rt = rtcache_lookup(ro, &u.dst);
   1459 	}
   1460 	/*
   1461 	 * If we found a route, use the address
   1462 	 * corresponding to the outgoing interface
   1463 	 * unless it is the loopback (in case a route
   1464 	 * to our address on another net goes to loopback).
   1465 	 *
   1466 	 * XXX Is this still true?  Do we care?
   1467 	 */
   1468 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
   1469 		ia = ifatoia(rt->rt_ifa);
   1470 	if (ia == NULL) {
   1471 		u_int16_t fport = sin->sin_port;
   1472 
   1473 		sin->sin_port = 0;
   1474 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
   1475 		sin->sin_port = fport;
   1476 		if (ia == NULL) {
   1477 			/* Find 1st non-loopback AF_INET address */
   1478 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
   1479 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
   1480 					break;
   1481 			}
   1482 		}
   1483 		if (ia == NULL) {
   1484 			*errorp = EADDRNOTAVAIL;
   1485 			return NULL;
   1486 		}
   1487 	}
   1488 	/*
   1489 	 * If the destination address is multicast and an outgoing
   1490 	 * interface has been set as a multicast option, use the
   1491 	 * address of that interface as our source address.
   1492 	 */
   1493 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
   1494 		struct ip_moptions *imo;
   1495 		struct ifnet *ifp;
   1496 
   1497 		imo = mopts;
   1498 		if (imo->imo_multicast_ifp != NULL) {
   1499 			ifp = imo->imo_multicast_ifp;
   1500 			IFP_TO_IA(ifp, ia);		/* XXX */
   1501 			if (ia == 0 || ia->ia4_flags & IN_IFF_NOTREADY) {
   1502 				*errorp = EADDRNOTAVAIL;
   1503 				return NULL;
   1504 			}
   1505 		}
   1506 	}
   1507 	if (ia->ia_ifa.ifa_getifa != NULL) {
   1508 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
   1509 		                                      sintosa(sin)));
   1510 		if (ia == NULL) {
   1511 			*errorp = EADDRNOTAVAIL;
   1512 			return NULL;
   1513 		}
   1514 	}
   1515 #ifdef GETIFA_DEBUG
   1516 	else
   1517 		printf("%s: missing ifa_getifa\n", __func__);
   1518 #endif
   1519 	return satosin(&ia->ia_addr);
   1520 }
   1521 
   1522 static void
   1523 in_sysctl_init(struct sysctllog **clog)
   1524 {
   1525 	sysctl_createv(clog, 0, NULL, NULL,
   1526 		       CTLFLAG_PERMANENT,
   1527 		       CTLTYPE_NODE, "inet",
   1528 		       SYSCTL_DESCR("PF_INET related settings"),
   1529 		       NULL, 0, NULL, 0,
   1530 		       CTL_NET, PF_INET, CTL_EOL);
   1531 	sysctl_createv(clog, 0, NULL, NULL,
   1532 		       CTLFLAG_PERMANENT,
   1533 		       CTLTYPE_NODE, "ip",
   1534 		       SYSCTL_DESCR("IPv4 related settings"),
   1535 		       NULL, 0, NULL, 0,
   1536 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
   1537 
   1538 	sysctl_createv(clog, 0, NULL, NULL,
   1539 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1540 		       CTLTYPE_INT, "subnetsarelocal",
   1541 		       SYSCTL_DESCR("Whether logical subnets are considered "
   1542 				    "local"),
   1543 		       NULL, 0, &subnetsarelocal, 0,
   1544 		       CTL_NET, PF_INET, IPPROTO_IP,
   1545 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
   1546 	sysctl_createv(clog, 0, NULL, NULL,
   1547 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1548 		       CTLTYPE_INT, "hostzerobroadcast",
   1549 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
   1550 		       NULL, 0, &hostzeroisbroadcast, 0,
   1551 		       CTL_NET, PF_INET, IPPROTO_IP,
   1552 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
   1553 }
   1554