1 /* $NetBSD: in.c,v 1.228 2018/04/08 13:52:22 christos 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 1.228 2018/04/08 13:52:22 christos Exp $"); 95 96 #include "arp.h" 97 98 #ifdef _KERNEL_OPT 99 #include "opt_inet.h" 100 #include "opt_inet_conf.h" 101 #include "opt_mrouting.h" 102 #include "opt_net_mpsafe.h" 103 #endif 104 105 #include <sys/param.h> 106 #include <sys/ioctl.h> 107 #include <sys/errno.h> 108 #include <sys/kernel.h> 109 #include <sys/malloc.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/sysctl.h> 113 #include <sys/systm.h> 114 #include <sys/proc.h> 115 #include <sys/syslog.h> 116 #include <sys/kauth.h> 117 #include <sys/kmem.h> 118 119 #include <sys/cprng.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <net/if_arp.h> 126 #include <net/if_ether.h> 127 #include <net/if_types.h> 128 #include <net/if_llatbl.h> 129 #include <net/if_dl.h> 130 131 #include <netinet/in_systm.h> 132 #include <netinet/in.h> 133 #include <netinet/in_var.h> 134 #include <netinet/ip.h> 135 #include <netinet/ip_var.h> 136 #include <netinet/in_ifattach.h> 137 #include <netinet/in_pcb.h> 138 #include <netinet/in_selsrc.h> 139 #include <netinet/if_inarp.h> 140 #include <netinet/ip_mroute.h> 141 #include <netinet/igmp_var.h> 142 143 #ifdef IPSELSRC 144 #include <netinet/in_selsrc.h> 145 #endif 146 147 static u_int in_mask2len(struct in_addr *); 148 static void in_len2mask(struct in_addr *, u_int); 149 static int in_lifaddr_ioctl(struct socket *, u_long, void *, 150 struct ifnet *); 151 152 static void in_addrhash_insert_locked(struct in_ifaddr *); 153 static void in_addrhash_remove_locked(struct in_ifaddr *); 154 155 static int in_addprefix(struct in_ifaddr *, int); 156 static void in_scrubaddr(struct in_ifaddr *); 157 static int in_scrubprefix(struct in_ifaddr *); 158 static void in_sysctl_init(struct sysctllog **); 159 160 #ifndef SUBNETSARELOCAL 161 #define SUBNETSARELOCAL 1 162 #endif 163 164 #ifndef HOSTZEROBROADCAST 165 #define HOSTZEROBROADCAST 0 166 #endif 167 168 /* Note: 61, 127, 251, 509, 1021, 2039 are good. */ 169 #ifndef IN_MULTI_HASH_SIZE 170 #define IN_MULTI_HASH_SIZE 509 171 #endif 172 173 static int subnetsarelocal = SUBNETSARELOCAL; 174 static int hostzeroisbroadcast = HOSTZEROBROADCAST; 175 176 /* 177 * This list is used to keep track of in_multi chains which belong to 178 * deleted interface addresses. We use in_ifaddr so that a chain head 179 * won't be deallocated until all multicast address record are deleted. 180 */ 181 182 LIST_HEAD(in_multihashhead, in_multi); /* Type of the hash head */ 183 184 static struct pool inmulti_pool; 185 static u_int in_multientries; 186 static struct in_multihashhead *in_multihashtbl; 187 static u_long in_multihash; 188 static krwlock_t in_multilock; 189 190 #define IN_MULTI_HASH(x, ifp) \ 191 (in_multihashtbl[(u_long)((x) ^ (ifp->if_index)) % IN_MULTI_HASH_SIZE]) 192 193 /* XXX DEPRECATED. Keep them to avoid breaking kvm(3) users. */ 194 struct in_ifaddrhashhead * in_ifaddrhashtbl; 195 u_long in_ifaddrhash; 196 struct in_ifaddrhead in_ifaddrhead; 197 static kmutex_t in_ifaddr_lock; 198 199 pserialize_t in_ifaddrhash_psz; 200 struct pslist_head * in_ifaddrhashtbl_pslist; 201 u_long in_ifaddrhash_pslist; 202 struct pslist_head in_ifaddrhead_pslist; 203 204 void 205 in_init(void) 206 { 207 pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", 208 NULL, IPL_SOFTNET); 209 TAILQ_INIT(&in_ifaddrhead); 210 PSLIST_INIT(&in_ifaddrhead_pslist); 211 212 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 213 &in_ifaddrhash); 214 215 in_ifaddrhash_psz = pserialize_create(); 216 in_ifaddrhashtbl_pslist = hashinit(IN_IFADDR_HASH_SIZE, HASH_PSLIST, 217 true, &in_ifaddrhash_pslist); 218 mutex_init(&in_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE); 219 220 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true, 221 &in_multihash); 222 rw_init(&in_multilock); 223 224 in_sysctl_init(NULL); 225 } 226 227 /* 228 * Return 1 if an internet address is for a ``local'' host 229 * (one to which we have a connection). If subnetsarelocal 230 * is true, this includes other subnets of the local net. 231 * Otherwise, it includes only the directly-connected (sub)nets. 232 */ 233 int 234 in_localaddr(struct in_addr in) 235 { 236 struct in_ifaddr *ia; 237 int localaddr = 0; 238 int s = pserialize_read_enter(); 239 240 if (subnetsarelocal) { 241 IN_ADDRLIST_READER_FOREACH(ia) { 242 if ((in.s_addr & ia->ia_netmask) == ia->ia_net) { 243 localaddr = 1; 244 break; 245 } 246 } 247 } else { 248 IN_ADDRLIST_READER_FOREACH(ia) { 249 if ((in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) { 250 localaddr = 1; 251 break; 252 } 253 } 254 } 255 pserialize_read_exit(s); 256 257 return localaddr; 258 } 259 260 /* 261 * like in_localaddr() but can specify ifp. 262 */ 263 int 264 in_direct(struct in_addr in, struct ifnet *ifp) 265 { 266 struct ifaddr *ifa; 267 int localaddr = 0; 268 int s; 269 270 KASSERT(ifp != NULL); 271 272 #define ia (ifatoia(ifa)) 273 s = pserialize_read_enter(); 274 if (subnetsarelocal) { 275 IFADDR_READER_FOREACH(ifa, ifp) { 276 if (ifa->ifa_addr->sa_family == AF_INET && 277 ((in.s_addr & ia->ia_netmask) == ia->ia_net)) { 278 localaddr = 1; 279 break; 280 } 281 } 282 } else { 283 IFADDR_READER_FOREACH(ifa, ifp) { 284 if (ifa->ifa_addr->sa_family == AF_INET && 285 (in.s_addr & ia->ia_subnetmask) == ia->ia_subnet) { 286 localaddr = 1; 287 break; 288 } 289 } 290 } 291 pserialize_read_exit(s); 292 293 return localaddr; 294 #undef ia 295 } 296 297 /* 298 * Determine whether an IP address is in a reserved set of addresses 299 * that may not be forwarded, or whether datagrams to that destination 300 * may be forwarded. 301 */ 302 int 303 in_canforward(struct in_addr in) 304 { 305 u_int32_t net; 306 307 if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr)) 308 return (0); 309 if (IN_CLASSA(in.s_addr)) { 310 net = in.s_addr & IN_CLASSA_NET; 311 if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 312 return (0); 313 } 314 return (1); 315 } 316 317 /* 318 * Trim a mask in a sockaddr 319 */ 320 void 321 in_socktrim(struct sockaddr_in *ap) 322 { 323 char *cplim = (char *) &ap->sin_addr; 324 char *cp = (char *) (&ap->sin_addr + 1); 325 326 ap->sin_len = 0; 327 while (--cp >= cplim) 328 if (*cp) { 329 (ap)->sin_len = cp - (char *) (ap) + 1; 330 break; 331 } 332 } 333 334 /* 335 * Maintain the "in_maxmtu" variable, which is the largest 336 * mtu for non-local interfaces with AF_INET addresses assigned 337 * to them that are up. 338 */ 339 unsigned long in_maxmtu; 340 341 void 342 in_setmaxmtu(void) 343 { 344 struct in_ifaddr *ia; 345 struct ifnet *ifp; 346 unsigned long maxmtu = 0; 347 int s = pserialize_read_enter(); 348 349 IN_ADDRLIST_READER_FOREACH(ia) { 350 if ((ifp = ia->ia_ifp) == 0) 351 continue; 352 if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP) 353 continue; 354 if (ifp->if_mtu > maxmtu) 355 maxmtu = ifp->if_mtu; 356 } 357 if (maxmtu) 358 in_maxmtu = maxmtu; 359 pserialize_read_exit(s); 360 } 361 362 static u_int 363 in_mask2len(struct in_addr *mask) 364 { 365 u_int x, y; 366 u_char *p; 367 368 p = (u_char *)mask; 369 for (x = 0; x < sizeof(*mask); x++) { 370 if (p[x] != 0xff) 371 break; 372 } 373 y = 0; 374 if (x < sizeof(*mask)) { 375 for (y = 0; y < NBBY; y++) { 376 if ((p[x] & (0x80 >> y)) == 0) 377 break; 378 } 379 } 380 return x * NBBY + y; 381 } 382 383 static void 384 in_len2mask(struct in_addr *mask, u_int len) 385 { 386 u_int i; 387 u_char *p; 388 389 p = (u_char *)mask; 390 memset(mask, 0, sizeof(*mask)); 391 for (i = 0; i < len / NBBY; i++) 392 p[i] = 0xff; 393 if (len % NBBY) 394 p[i] = (0xff00 >> (len % NBBY)) & 0xff; 395 } 396 397 /* 398 * Generic internet control operations (ioctl's). 399 * Ifp is 0 if not an interface-specific ioctl. 400 */ 401 /* ARGSUSED */ 402 static int 403 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 404 { 405 struct ifreq *ifr = (struct ifreq *)data; 406 struct in_ifaddr *ia = NULL; 407 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 408 struct sockaddr_in oldaddr, *new_dstaddr; 409 int error, hostIsNew, maskIsNew; 410 int newifaddr = 0; 411 bool run_hook = false; 412 bool need_reinsert = false; 413 struct psref psref; 414 int bound; 415 416 switch (cmd) { 417 case SIOCALIFADDR: 418 case SIOCDLIFADDR: 419 case SIOCGLIFADDR: 420 if (ifp == NULL) 421 return EINVAL; 422 return in_lifaddr_ioctl(so, cmd, data, ifp); 423 case SIOCGIFADDRPREF: 424 case SIOCSIFADDRPREF: 425 if (ifp == NULL) 426 return EINVAL; 427 return ifaddrpref_ioctl(so, cmd, data, ifp); 428 } 429 430 bound = curlwp_bind(); 431 /* 432 * Find address for this interface, if it exists. 433 */ 434 if (ifp != NULL) 435 ia = in_get_ia_from_ifp_psref(ifp, &psref); 436 437 hostIsNew = 1; /* moved here to appease gcc */ 438 switch (cmd) { 439 case SIOCAIFADDR: 440 case SIOCDIFADDR: 441 case SIOCGIFALIAS: 442 case SIOCGIFAFLAG_IN: 443 if (ifra->ifra_addr.sin_family == AF_INET) { 444 int s; 445 446 if (ia != NULL) 447 ia4_release(ia, &psref); 448 s = pserialize_read_enter(); 449 IN_ADDRHASH_READER_FOREACH(ia, 450 ifra->ifra_addr.sin_addr.s_addr) { 451 if (ia->ia_ifp == ifp && 452 in_hosteq(ia->ia_addr.sin_addr, 453 ifra->ifra_addr.sin_addr)) 454 break; 455 } 456 if (ia != NULL) 457 ia4_acquire(ia, &psref); 458 pserialize_read_exit(s); 459 } 460 if ((cmd == SIOCDIFADDR || 461 cmd == SIOCGIFALIAS || 462 cmd == SIOCGIFAFLAG_IN) && 463 ia == NULL) { 464 error = EADDRNOTAVAIL; 465 goto out; 466 } 467 468 if (cmd == SIOCDIFADDR && 469 ifra->ifra_addr.sin_family == AF_UNSPEC) { 470 ifra->ifra_addr.sin_family = AF_INET; 471 } 472 /* FALLTHROUGH */ 473 case SIOCSIFADDR: 474 if (ia == NULL || ia->ia_addr.sin_family != AF_INET) 475 ; 476 else if (ifra->ifra_addr.sin_len == 0) { 477 ifra->ifra_addr = ia->ia_addr; 478 hostIsNew = 0; 479 } else if (in_hosteq(ia->ia_addr.sin_addr, 480 ifra->ifra_addr.sin_addr)) 481 hostIsNew = 0; 482 /* FALLTHROUGH */ 483 case SIOCSIFDSTADDR: 484 if (ifra->ifra_addr.sin_family != AF_INET) { 485 error = EAFNOSUPPORT; 486 goto out; 487 } 488 /* FALLTHROUGH */ 489 case SIOCSIFNETMASK: 490 if (ifp == NULL) 491 panic("in_control"); 492 493 if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN) 494 break; 495 496 if (ia == NULL && 497 (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) { 498 error = EADDRNOTAVAIL; 499 goto out; 500 } 501 502 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 503 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 504 NULL) != 0) { 505 error = EPERM; 506 goto out; 507 } 508 509 if (ia == NULL) { 510 ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO); 511 if (ia == NULL) { 512 error = ENOBUFS; 513 goto out; 514 } 515 ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr); 516 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 517 ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask); 518 #ifdef IPSELSRC 519 ia->ia_ifa.ifa_getifa = in_getifa; 520 #else /* IPSELSRC */ 521 ia->ia_ifa.ifa_getifa = NULL; 522 #endif /* IPSELSRC */ 523 ia->ia_sockmask.sin_len = 8; 524 ia->ia_sockmask.sin_family = AF_INET; 525 if (ifp->if_flags & IFF_BROADCAST) { 526 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr); 527 ia->ia_broadaddr.sin_family = AF_INET; 528 } 529 ia->ia_ifp = ifp; 530 ia->ia_idsalt = cprng_fast32() % 65535; 531 LIST_INIT(&ia->ia_multiaddrs); 532 IN_ADDRHASH_ENTRY_INIT(ia); 533 IN_ADDRLIST_ENTRY_INIT(ia); 534 ifa_psref_init(&ia->ia_ifa); 535 /* 536 * We need a reference to make ia survive over in_ifinit 537 * that does ifaref and ifafree. 538 */ 539 ifaref(&ia->ia_ifa); 540 541 newifaddr = 1; 542 } 543 break; 544 545 case SIOCSIFBRDADDR: 546 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 547 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 548 NULL) != 0) { 549 error = EPERM; 550 goto out; 551 } 552 /* FALLTHROUGH */ 553 554 case SIOCGIFADDR: 555 case SIOCGIFNETMASK: 556 case SIOCGIFDSTADDR: 557 case SIOCGIFBRDADDR: 558 if (ia == NULL) { 559 error = EADDRNOTAVAIL; 560 goto out; 561 } 562 break; 563 } 564 error = 0; 565 switch (cmd) { 566 567 case SIOCGIFADDR: 568 ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr)); 569 break; 570 571 case SIOCGIFBRDADDR: 572 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 573 error = EINVAL; 574 goto out; 575 } 576 ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr)); 577 break; 578 579 case SIOCGIFDSTADDR: 580 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 581 error = EINVAL; 582 goto out; 583 } 584 ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr)); 585 break; 586 587 case SIOCGIFNETMASK: 588 /* 589 * We keep the number of trailing zero bytes the sin_len field 590 * of ia_sockmask, so we fix this before we pass it back to 591 * userland. 592 */ 593 oldaddr = ia->ia_sockmask; 594 oldaddr.sin_len = sizeof(struct sockaddr_in); 595 ifreq_setaddr(cmd, ifr, (const void *)&oldaddr); 596 break; 597 598 case SIOCSIFDSTADDR: 599 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 600 error = EINVAL; 601 goto out; 602 } 603 oldaddr = ia->ia_dstaddr; 604 ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr)); 605 if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) { 606 ia->ia_dstaddr = oldaddr; 607 goto out; 608 } 609 if (ia->ia_flags & IFA_ROUTE) { 610 ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr); 611 rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST); 612 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 613 rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP); 614 } 615 break; 616 617 case SIOCSIFBRDADDR: 618 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 619 error = EINVAL; 620 goto out; 621 } 622 ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr)); 623 break; 624 625 case SIOCSIFADDR: 626 if (!newifaddr) { 627 in_addrhash_remove(ia); 628 need_reinsert = true; 629 } 630 error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)), 631 NULL, 1); 632 633 run_hook = true; 634 break; 635 636 case SIOCSIFNETMASK: 637 in_scrubprefix(ia); 638 ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr)); 639 ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr; 640 if (!newifaddr) { 641 in_addrhash_remove(ia); 642 need_reinsert = true; 643 } 644 error = in_ifinit(ifp, ia, NULL, NULL, 0); 645 break; 646 647 case SIOCAIFADDR: 648 maskIsNew = 0; 649 if (ifra->ifra_mask.sin_len) { 650 in_scrubprefix(ia); 651 ia->ia_sockmask = ifra->ifra_mask; 652 ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr; 653 maskIsNew = 1; 654 } 655 if ((ifp->if_flags & IFF_POINTOPOINT) && 656 (ifra->ifra_dstaddr.sin_family == AF_INET)) { 657 new_dstaddr = &ifra->ifra_dstaddr; 658 maskIsNew = 1; /* We lie; but the effect's the same */ 659 } else 660 new_dstaddr = NULL; 661 if (ifra->ifra_addr.sin_family == AF_INET && 662 (hostIsNew || maskIsNew)) { 663 if (!newifaddr) { 664 in_addrhash_remove(ia); 665 need_reinsert = true; 666 } 667 error = in_ifinit(ifp, ia, &ifra->ifra_addr, 668 new_dstaddr, 0); 669 } 670 if ((ifp->if_flags & IFF_BROADCAST) && 671 (ifra->ifra_broadaddr.sin_family == AF_INET)) 672 ia->ia_broadaddr = ifra->ifra_broadaddr; 673 run_hook = true; 674 break; 675 676 case SIOCGIFALIAS: 677 ifra->ifra_mask = ia->ia_sockmask; 678 if ((ifp->if_flags & IFF_POINTOPOINT) && 679 (ia->ia_dstaddr.sin_family == AF_INET)) 680 ifra->ifra_dstaddr = ia->ia_dstaddr; 681 else if ((ifp->if_flags & IFF_BROADCAST) && 682 (ia->ia_broadaddr.sin_family == AF_INET)) 683 ifra->ifra_broadaddr = ia->ia_broadaddr; 684 else 685 memset(&ifra->ifra_broadaddr, 0, 686 sizeof(ifra->ifra_broadaddr)); 687 break; 688 689 case SIOCGIFAFLAG_IN: 690 ifr->ifr_addrflags = ia->ia4_flags; 691 break; 692 693 case SIOCDIFADDR: 694 ia4_release(ia, &psref); 695 ifaref(&ia->ia_ifa); 696 in_purgeaddr(&ia->ia_ifa); 697 pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa); 698 ifafree(&ia->ia_ifa); 699 ia = NULL; 700 break; 701 702 #ifdef MROUTING 703 case SIOCGETVIFCNT: 704 case SIOCGETSGCNT: 705 error = mrt_ioctl(so, cmd, data); 706 break; 707 #endif /* MROUTING */ 708 709 default: 710 error = ENOTTY; 711 goto out; 712 } 713 714 /* 715 * XXX insert regardless of error to make in_purgeaddr below work. 716 * Need to improve. 717 */ 718 if (newifaddr) { 719 ifaref(&ia->ia_ifa); 720 ifa_insert(ifp, &ia->ia_ifa); 721 722 mutex_enter(&in_ifaddr_lock); 723 TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list); 724 IN_ADDRLIST_WRITER_INSERT_TAIL(ia); 725 in_addrhash_insert_locked(ia); 726 /* Release a reference that is held just after creation. */ 727 ifafree(&ia->ia_ifa); 728 mutex_exit(&in_ifaddr_lock); 729 } else if (need_reinsert) { 730 in_addrhash_insert(ia); 731 } 732 733 if (error == 0) { 734 if (run_hook) 735 pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa); 736 } else if (newifaddr) { 737 KASSERT(ia != NULL); 738 in_purgeaddr(&ia->ia_ifa); 739 ia = NULL; 740 } 741 742 out: 743 if (!newifaddr && ia != NULL) 744 ia4_release(ia, &psref); 745 curlwp_bindx(bound); 746 return error; 747 } 748 749 int 750 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 751 { 752 int error; 753 754 #ifndef NET_MPSAFE 755 KASSERT(KERNEL_LOCKED_P()); 756 #endif 757 error = in_control0(so, cmd, data, ifp); 758 759 return error; 760 } 761 762 /* Add ownaddr as loopback rtentry. */ 763 static void 764 in_ifaddlocal(struct ifaddr *ifa) 765 { 766 struct in_ifaddr *ia; 767 768 ia = (struct in_ifaddr *)ifa; 769 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY || 770 (ia->ia_ifp->if_flags & IFF_POINTOPOINT && 771 in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr))) 772 { 773 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 774 return; 775 } 776 777 rt_ifa_addlocal(ifa); 778 } 779 780 /* Remove loopback entry of ownaddr */ 781 static void 782 in_ifremlocal(struct ifaddr *ifa) 783 { 784 struct in_ifaddr *ia, *p; 785 struct ifaddr *alt_ifa = NULL; 786 int ia_count = 0; 787 int s; 788 struct psref psref; 789 int bound = curlwp_bind(); 790 791 ia = (struct in_ifaddr *)ifa; 792 /* Delete the entry if exactly one ifaddr matches the 793 * address, ifa->ifa_addr. */ 794 s = pserialize_read_enter(); 795 IN_ADDRLIST_READER_FOREACH(p) { 796 if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr)) 797 continue; 798 if (p->ia_ifp != ia->ia_ifp) 799 alt_ifa = &p->ia_ifa; 800 if (++ia_count > 1 && alt_ifa != NULL) 801 break; 802 } 803 if (alt_ifa != NULL && ia_count > 1) 804 ifa_acquire(alt_ifa, &psref); 805 pserialize_read_exit(s); 806 807 if (ia_count == 0) 808 goto out; 809 810 rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa); 811 if (alt_ifa != NULL && ia_count > 1) 812 ifa_release(alt_ifa, &psref); 813 out: 814 curlwp_bindx(bound); 815 } 816 817 static void 818 in_scrubaddr(struct in_ifaddr *ia) 819 { 820 821 /* stop DAD processing */ 822 if (ia->ia_dad_stop != NULL) 823 ia->ia_dad_stop(&ia->ia_ifa); 824 825 in_scrubprefix(ia); 826 in_ifremlocal(&ia->ia_ifa); 827 828 mutex_enter(&in_ifaddr_lock); 829 if (ia->ia_allhosts != NULL) { 830 in_delmulti(ia->ia_allhosts); 831 ia->ia_allhosts = NULL; 832 } 833 mutex_exit(&in_ifaddr_lock); 834 } 835 836 /* 837 * Depends on it isn't called in concurrent. It should be guaranteed 838 * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control 839 * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl 840 * and the latter is called via ifa->ifa_ifp's if_detach. The functions 841 * never be executed in concurrent. 842 */ 843 void 844 in_purgeaddr(struct ifaddr *ifa) 845 { 846 struct in_ifaddr *ia = (void *) ifa; 847 struct ifnet *ifp = ifa->ifa_ifp; 848 849 /* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */ 850 851 ifa->ifa_flags |= IFA_DESTROYING; 852 in_scrubaddr(ia); 853 854 mutex_enter(&in_ifaddr_lock); 855 in_addrhash_remove_locked(ia); 856 TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list); 857 IN_ADDRLIST_WRITER_REMOVE(ia); 858 ifa_remove(ifp, &ia->ia_ifa); 859 /* Assume ifa_remove called pserialize_perform and psref_destroy */ 860 mutex_exit(&in_ifaddr_lock); 861 IN_ADDRHASH_ENTRY_DESTROY(ia); 862 IN_ADDRLIST_ENTRY_DESTROY(ia); 863 ifafree(&ia->ia_ifa); 864 in_setmaxmtu(); 865 } 866 867 static void 868 in_addrhash_insert_locked(struct in_ifaddr *ia) 869 { 870 871 KASSERT(mutex_owned(&in_ifaddr_lock)); 872 873 LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 874 ia_hash); 875 IN_ADDRHASH_ENTRY_INIT(ia); 876 IN_ADDRHASH_WRITER_INSERT_HEAD(ia); 877 } 878 879 void 880 in_addrhash_insert(struct in_ifaddr *ia) 881 { 882 883 mutex_enter(&in_ifaddr_lock); 884 in_addrhash_insert_locked(ia); 885 mutex_exit(&in_ifaddr_lock); 886 } 887 888 static void 889 in_addrhash_remove_locked(struct in_ifaddr *ia) 890 { 891 892 KASSERT(mutex_owned(&in_ifaddr_lock)); 893 894 LIST_REMOVE(ia, ia_hash); 895 IN_ADDRHASH_WRITER_REMOVE(ia); 896 } 897 898 void 899 in_addrhash_remove(struct in_ifaddr *ia) 900 { 901 902 mutex_enter(&in_ifaddr_lock); 903 in_addrhash_remove_locked(ia); 904 #ifdef NET_MPSAFE 905 pserialize_perform(in_ifaddrhash_psz); 906 #endif 907 mutex_exit(&in_ifaddr_lock); 908 IN_ADDRHASH_ENTRY_DESTROY(ia); 909 } 910 911 void 912 in_purgeif(struct ifnet *ifp) /* MUST be called at splsoftnet() */ 913 { 914 915 IFNET_LOCK(ifp); 916 if_purgeaddrs(ifp, AF_INET, in_purgeaddr); 917 igmp_purgeif(ifp); /* manipulates pools */ 918 #ifdef MROUTING 919 ip_mrouter_detach(ifp); 920 #endif 921 IFNET_UNLOCK(ifp); 922 } 923 924 /* 925 * SIOC[GAD]LIFADDR. 926 * SIOCGLIFADDR: get first address. (???) 927 * SIOCGLIFADDR with IFLR_PREFIX: 928 * get first address that matches the specified prefix. 929 * SIOCALIFADDR: add the specified address. 930 * SIOCALIFADDR with IFLR_PREFIX: 931 * EINVAL since we can't deduce hostid part of the address. 932 * SIOCDLIFADDR: delete the specified address. 933 * SIOCDLIFADDR with IFLR_PREFIX: 934 * delete the first address that matches the specified prefix. 935 * return values: 936 * EINVAL on invalid parameters 937 * EADDRNOTAVAIL on prefix match failed/specified address not found 938 * other values may be returned from in_ioctl() 939 */ 940 static int 941 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data, 942 struct ifnet *ifp) 943 { 944 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 945 struct ifaddr *ifa; 946 struct sockaddr *sa; 947 948 /* sanity checks */ 949 if (data == NULL || ifp == NULL) { 950 panic("invalid argument to in_lifaddr_ioctl"); 951 /*NOTRECHED*/ 952 } 953 954 switch (cmd) { 955 case SIOCGLIFADDR: 956 /* address must be specified on GET with IFLR_PREFIX */ 957 if ((iflr->flags & IFLR_PREFIX) == 0) 958 break; 959 /*FALLTHROUGH*/ 960 case SIOCALIFADDR: 961 case SIOCDLIFADDR: 962 /* address must be specified on ADD and DELETE */ 963 sa = (struct sockaddr *)&iflr->addr; 964 if (sa->sa_family != AF_INET) 965 return EINVAL; 966 if (sa->sa_len != sizeof(struct sockaddr_in)) 967 return EINVAL; 968 /* XXX need improvement */ 969 sa = (struct sockaddr *)&iflr->dstaddr; 970 if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET) 971 return EINVAL; 972 if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in)) 973 return EINVAL; 974 break; 975 default: /*shouldn't happen*/ 976 #if 0 977 panic("invalid cmd to in_lifaddr_ioctl"); 978 /*NOTREACHED*/ 979 #else 980 return EOPNOTSUPP; 981 #endif 982 } 983 if (sizeof(struct in_addr) * NBBY < iflr->prefixlen) 984 return EINVAL; 985 986 switch (cmd) { 987 case SIOCALIFADDR: 988 { 989 struct in_aliasreq ifra; 990 991 if (iflr->flags & IFLR_PREFIX) 992 return EINVAL; 993 994 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */ 995 memset(&ifra, 0, sizeof(ifra)); 996 memcpy(ifra.ifra_name, iflr->iflr_name, 997 sizeof(ifra.ifra_name)); 998 999 memcpy(&ifra.ifra_addr, &iflr->addr, 1000 ((struct sockaddr *)&iflr->addr)->sa_len); 1001 1002 if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /*XXX*/ 1003 memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr, 1004 ((struct sockaddr *)&iflr->dstaddr)->sa_len); 1005 } 1006 1007 ifra.ifra_mask.sin_family = AF_INET; 1008 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 1009 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 1010 1011 return in_control(so, SIOCAIFADDR, &ifra, ifp); 1012 } 1013 case SIOCGLIFADDR: 1014 case SIOCDLIFADDR: 1015 { 1016 struct in_ifaddr *ia; 1017 struct in_addr mask, candidate, match; 1018 struct sockaddr_in *sin; 1019 int cmp, s; 1020 1021 memset(&mask, 0, sizeof(mask)); 1022 memset(&match, 0, sizeof(match)); /* XXX gcc */ 1023 if (iflr->flags & IFLR_PREFIX) { 1024 /* lookup a prefix rather than address. */ 1025 in_len2mask(&mask, iflr->prefixlen); 1026 1027 sin = (struct sockaddr_in *)&iflr->addr; 1028 match.s_addr = sin->sin_addr.s_addr; 1029 match.s_addr &= mask.s_addr; 1030 1031 /* if you set extra bits, that's wrong */ 1032 if (match.s_addr != sin->sin_addr.s_addr) 1033 return EINVAL; 1034 1035 cmp = 1; 1036 } else { 1037 if (cmd == SIOCGLIFADDR) { 1038 /* on getting an address, take the 1st match */ 1039 cmp = 0; /*XXX*/ 1040 } else { 1041 /* on deleting an address, do exact match */ 1042 in_len2mask(&mask, 32); 1043 sin = (struct sockaddr_in *)&iflr->addr; 1044 match.s_addr = sin->sin_addr.s_addr; 1045 1046 cmp = 1; 1047 } 1048 } 1049 1050 s = pserialize_read_enter(); 1051 IFADDR_READER_FOREACH(ifa, ifp) { 1052 if (ifa->ifa_addr->sa_family != AF_INET) 1053 continue; 1054 if (cmp == 0) 1055 break; 1056 candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr; 1057 candidate.s_addr &= mask.s_addr; 1058 if (candidate.s_addr == match.s_addr) 1059 break; 1060 } 1061 if (ifa == NULL) { 1062 pserialize_read_exit(s); 1063 return EADDRNOTAVAIL; 1064 } 1065 ia = (struct in_ifaddr *)ifa; 1066 1067 if (cmd == SIOCGLIFADDR) { 1068 /* fill in the if_laddrreq structure */ 1069 memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len); 1070 1071 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1072 memcpy(&iflr->dstaddr, &ia->ia_dstaddr, 1073 ia->ia_dstaddr.sin_len); 1074 } else 1075 memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr)); 1076 1077 iflr->prefixlen = 1078 in_mask2len(&ia->ia_sockmask.sin_addr); 1079 1080 iflr->flags = 0; /*XXX*/ 1081 pserialize_read_exit(s); 1082 1083 return 0; 1084 } else { 1085 struct in_aliasreq ifra; 1086 1087 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */ 1088 memset(&ifra, 0, sizeof(ifra)); 1089 memcpy(ifra.ifra_name, iflr->iflr_name, 1090 sizeof(ifra.ifra_name)); 1091 1092 memcpy(&ifra.ifra_addr, &ia->ia_addr, 1093 ia->ia_addr.sin_len); 1094 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1095 memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr, 1096 ia->ia_dstaddr.sin_len); 1097 } 1098 memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask, 1099 ia->ia_sockmask.sin_len); 1100 pserialize_read_exit(s); 1101 1102 return in_control(so, SIOCDIFADDR, &ifra, ifp); 1103 } 1104 } 1105 } 1106 1107 return EOPNOTSUPP; /*just for safety*/ 1108 } 1109 1110 /* 1111 * Initialize an interface's internet address 1112 * and routing table entry. 1113 */ 1114 int 1115 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, 1116 const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub) 1117 { 1118 u_int32_t i; 1119 struct sockaddr_in oldaddr, olddst; 1120 int s, oldflags, flags = RTF_UP, error, hostIsNew; 1121 1122 if (sin == NULL) 1123 sin = &ia->ia_addr; 1124 if (dst == NULL) 1125 dst = &ia->ia_dstaddr; 1126 1127 /* 1128 * Set up new addresses. 1129 */ 1130 oldaddr = ia->ia_addr; 1131 olddst = ia->ia_dstaddr; 1132 oldflags = ia->ia4_flags; 1133 ia->ia_addr = *sin; 1134 ia->ia_dstaddr = *dst; 1135 hostIsNew = oldaddr.sin_family != AF_INET || 1136 !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr); 1137 if (!scrub) 1138 scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family || 1139 !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr); 1140 1141 /* 1142 * Configure address flags. 1143 * We need to do this early because they maybe adjusted 1144 * by if_addr_init depending on the address. 1145 */ 1146 if (ia->ia4_flags & IN_IFF_DUPLICATED) { 1147 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1148 hostIsNew = 1; 1149 } 1150 if (ifp->if_link_state == LINK_STATE_DOWN) { 1151 ia->ia4_flags |= IN_IFF_DETACHED; 1152 ia->ia4_flags &= ~IN_IFF_TENTATIVE; 1153 } else if (hostIsNew && if_do_dad(ifp) 1154 #if NARP > 0 1155 && ip_dad_count > 0 1156 #endif 1157 ) 1158 ia->ia4_flags |= IN_IFF_TRYTENTATIVE; 1159 1160 /* 1161 * Give the interface a chance to initialize 1162 * if this is its first address, 1163 * and to validate the address if necessary. 1164 */ 1165 s = splsoftnet(); 1166 error = if_addr_init(ifp, &ia->ia_ifa, true); 1167 splx(s); 1168 /* Now clear the try tentative flag, its job is done. */ 1169 ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE; 1170 if (error != 0) { 1171 ia->ia_addr = oldaddr; 1172 ia->ia_dstaddr = olddst; 1173 ia->ia4_flags = oldflags; 1174 return error; 1175 } 1176 1177 if (scrub || hostIsNew) { 1178 int newflags = ia->ia4_flags; 1179 1180 ia->ia_ifa.ifa_addr = sintosa(&oldaddr); 1181 ia->ia_ifa.ifa_dstaddr = sintosa(&olddst); 1182 ia->ia4_flags = oldflags; 1183 if (hostIsNew) 1184 in_scrubaddr(ia); 1185 else if (scrub) 1186 in_scrubprefix(ia); 1187 ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr); 1188 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 1189 ia->ia4_flags = newflags; 1190 } 1191 1192 i = ia->ia_addr.sin_addr.s_addr; 1193 if (ifp->if_flags & IFF_POINTOPOINT) 1194 ia->ia_netmask = INADDR_BROADCAST; /* default to /32 */ 1195 else if (IN_CLASSA(i)) 1196 ia->ia_netmask = IN_CLASSA_NET; 1197 else if (IN_CLASSB(i)) 1198 ia->ia_netmask = IN_CLASSB_NET; 1199 else 1200 ia->ia_netmask = IN_CLASSC_NET; 1201 /* 1202 * The subnet mask usually includes at least the standard network part, 1203 * but may may be smaller in the case of supernetting. 1204 * If it is set, we believe it. 1205 */ 1206 if (ia->ia_subnetmask == 0) { 1207 ia->ia_subnetmask = ia->ia_netmask; 1208 ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask; 1209 } else 1210 ia->ia_netmask &= ia->ia_subnetmask; 1211 1212 ia->ia_net = i & ia->ia_netmask; 1213 ia->ia_subnet = i & ia->ia_subnetmask; 1214 in_socktrim(&ia->ia_sockmask); 1215 1216 /* re-calculate the "in_maxmtu" value */ 1217 in_setmaxmtu(); 1218 1219 ia->ia_ifa.ifa_metric = ifp->if_metric; 1220 if (ifp->if_flags & IFF_BROADCAST) { 1221 ia->ia_broadaddr.sin_addr.s_addr = 1222 ia->ia_subnet | ~ia->ia_subnetmask; 1223 ia->ia_netbroadcast.s_addr = 1224 ia->ia_net | ~ia->ia_netmask; 1225 } else if (ifp->if_flags & IFF_LOOPBACK) { 1226 ia->ia_dstaddr = ia->ia_addr; 1227 flags |= RTF_HOST; 1228 } else if (ifp->if_flags & IFF_POINTOPOINT) { 1229 if (ia->ia_dstaddr.sin_family != AF_INET) 1230 return (0); 1231 flags |= RTF_HOST; 1232 } 1233 1234 /* Add the local route to the address */ 1235 in_ifaddlocal(&ia->ia_ifa); 1236 1237 /* Add the prefix route for the address */ 1238 error = in_addprefix(ia, flags); 1239 1240 /* 1241 * If the interface supports multicast, join the "all hosts" 1242 * multicast group on that interface. 1243 */ 1244 mutex_enter(&in_ifaddr_lock); 1245 if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) { 1246 struct in_addr addr; 1247 1248 addr.s_addr = INADDR_ALLHOSTS_GROUP; 1249 ia->ia_allhosts = in_addmulti(&addr, ifp); 1250 } 1251 mutex_exit(&in_ifaddr_lock); 1252 1253 if (hostIsNew && 1254 ia->ia4_flags & IN_IFF_TENTATIVE && 1255 if_do_dad(ifp)) 1256 ia->ia_dad_start((struct ifaddr *)ia); 1257 1258 return error; 1259 } 1260 1261 #define rtinitflags(x) \ 1262 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 1263 ? RTF_HOST : 0) 1264 1265 /* 1266 * add a route to prefix ("connected route" in cisco terminology). 1267 * does nothing if there's some interface address with the same prefix already. 1268 */ 1269 static int 1270 in_addprefix(struct in_ifaddr *target, int flags) 1271 { 1272 struct in_ifaddr *ia; 1273 struct in_addr prefix, mask, p; 1274 int error; 1275 int s; 1276 1277 if ((flags & RTF_HOST) != 0) 1278 prefix = target->ia_dstaddr.sin_addr; 1279 else { 1280 prefix = target->ia_addr.sin_addr; 1281 mask = target->ia_sockmask.sin_addr; 1282 prefix.s_addr &= mask.s_addr; 1283 } 1284 1285 s = pserialize_read_enter(); 1286 IN_ADDRLIST_READER_FOREACH(ia) { 1287 if (rtinitflags(ia)) 1288 p = ia->ia_dstaddr.sin_addr; 1289 else { 1290 p = ia->ia_addr.sin_addr; 1291 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1292 } 1293 1294 if (prefix.s_addr != p.s_addr) 1295 continue; 1296 1297 /* 1298 * if we got a matching prefix route inserted by other 1299 * interface address, we don't need to bother 1300 * 1301 * XXX RADIX_MPATH implications here? -dyoung 1302 */ 1303 if (ia->ia_flags & IFA_ROUTE) { 1304 pserialize_read_exit(s); 1305 return 0; 1306 } 1307 } 1308 pserialize_read_exit(s); 1309 1310 /* 1311 * noone seem to have prefix route. insert it. 1312 */ 1313 error = rtinit(&target->ia_ifa, RTM_ADD, flags); 1314 if (error == 0) 1315 target->ia_flags |= IFA_ROUTE; 1316 else if (error == EEXIST) { 1317 /* 1318 * the fact the route already exists is not an error. 1319 */ 1320 error = 0; 1321 } 1322 return error; 1323 } 1324 1325 /* 1326 * remove a route to prefix ("connected route" in cisco terminology). 1327 * re-installs the route by using another interface address, if there's one 1328 * with the same prefix (otherwise we lose the route mistakenly). 1329 */ 1330 static int 1331 in_scrubprefix(struct in_ifaddr *target) 1332 { 1333 struct in_ifaddr *ia; 1334 struct in_addr prefix, mask, p; 1335 int error; 1336 int s; 1337 1338 /* If we don't have IFA_ROUTE we have nothing to do */ 1339 if ((target->ia_flags & IFA_ROUTE) == 0) 1340 return 0; 1341 1342 if (rtinitflags(target)) 1343 prefix = target->ia_dstaddr.sin_addr; 1344 else { 1345 prefix = target->ia_addr.sin_addr; 1346 mask = target->ia_sockmask.sin_addr; 1347 prefix.s_addr &= mask.s_addr; 1348 } 1349 1350 s = pserialize_read_enter(); 1351 IN_ADDRLIST_READER_FOREACH(ia) { 1352 if (rtinitflags(ia)) 1353 p = ia->ia_dstaddr.sin_addr; 1354 else { 1355 p = ia->ia_addr.sin_addr; 1356 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1357 } 1358 1359 if (prefix.s_addr != p.s_addr) 1360 continue; 1361 1362 /* 1363 * if we got a matching prefix route, move IFA_ROUTE to him 1364 */ 1365 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1366 struct psref psref; 1367 int bound = curlwp_bind(); 1368 1369 ia4_acquire(ia, &psref); 1370 pserialize_read_exit(s); 1371 1372 rtinit(&target->ia_ifa, RTM_DELETE, 1373 rtinitflags(target)); 1374 target->ia_flags &= ~IFA_ROUTE; 1375 1376 error = rtinit(&ia->ia_ifa, RTM_ADD, 1377 rtinitflags(ia) | RTF_UP); 1378 if (error == 0) 1379 ia->ia_flags |= IFA_ROUTE; 1380 1381 ia4_release(ia, &psref); 1382 curlwp_bindx(bound); 1383 1384 return error; 1385 } 1386 } 1387 pserialize_read_exit(s); 1388 1389 /* 1390 * noone seem to have prefix route. remove it. 1391 */ 1392 rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target)); 1393 target->ia_flags &= ~IFA_ROUTE; 1394 return 0; 1395 } 1396 1397 #undef rtinitflags 1398 1399 /* 1400 * Return 1 if the address might be a local broadcast address. 1401 */ 1402 int 1403 in_broadcast(struct in_addr in, struct ifnet *ifp) 1404 { 1405 struct ifaddr *ifa; 1406 int s; 1407 1408 KASSERT(ifp != NULL); 1409 1410 if (in.s_addr == INADDR_BROADCAST || 1411 in_nullhost(in)) 1412 return 1; 1413 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1414 return 0; 1415 /* 1416 * Look through the list of addresses for a match 1417 * with a broadcast address. 1418 */ 1419 #define ia (ifatoia(ifa)) 1420 s = pserialize_read_enter(); 1421 IFADDR_READER_FOREACH(ifa, ifp) { 1422 if (ifa->ifa_addr->sa_family == AF_INET && 1423 !in_hosteq(in, ia->ia_addr.sin_addr) && 1424 (in_hosteq(in, ia->ia_broadaddr.sin_addr) || 1425 in_hosteq(in, ia->ia_netbroadcast) || 1426 (hostzeroisbroadcast && 1427 /* 1428 * Check for old-style (host 0) broadcast. 1429 */ 1430 (in.s_addr == ia->ia_subnet || 1431 in.s_addr == ia->ia_net)))) { 1432 pserialize_read_exit(s); 1433 return 1; 1434 } 1435 } 1436 pserialize_read_exit(s); 1437 return (0); 1438 #undef ia 1439 } 1440 1441 /* 1442 * perform DAD when interface becomes IFF_UP. 1443 */ 1444 void 1445 in_if_link_up(struct ifnet *ifp) 1446 { 1447 struct ifaddr *ifa; 1448 struct in_ifaddr *ia; 1449 int s, bound; 1450 1451 /* Ensure it's sane to run DAD */ 1452 if (ifp->if_link_state == LINK_STATE_DOWN) 1453 return; 1454 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) 1455 return; 1456 1457 bound = curlwp_bind(); 1458 s = pserialize_read_enter(); 1459 IFADDR_READER_FOREACH(ifa, ifp) { 1460 struct psref psref; 1461 1462 if (ifa->ifa_addr->sa_family != AF_INET) 1463 continue; 1464 ifa_acquire(ifa, &psref); 1465 pserialize_read_exit(s); 1466 1467 ia = (struct in_ifaddr *)ifa; 1468 1469 /* If detached then mark as tentative */ 1470 if (ia->ia4_flags & IN_IFF_DETACHED) { 1471 ia->ia4_flags &= ~IN_IFF_DETACHED; 1472 if (if_do_dad(ifp) && ia->ia_dad_start != NULL) 1473 ia->ia4_flags |= IN_IFF_TENTATIVE; 1474 else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0) 1475 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1476 } 1477 1478 if (ia->ia4_flags & IN_IFF_TENTATIVE) { 1479 /* Clear the duplicated flag as we're starting DAD. */ 1480 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1481 ia->ia_dad_start(ifa); 1482 } 1483 1484 s = pserialize_read_enter(); 1485 ifa_release(ifa, &psref); 1486 } 1487 pserialize_read_exit(s); 1488 curlwp_bindx(bound); 1489 } 1490 1491 void 1492 in_if_up(struct ifnet *ifp) 1493 { 1494 1495 /* interface may not support link state, so bring it up also */ 1496 in_if_link_up(ifp); 1497 } 1498 1499 /* 1500 * Mark all addresses as detached. 1501 */ 1502 void 1503 in_if_link_down(struct ifnet *ifp) 1504 { 1505 struct ifaddr *ifa; 1506 struct in_ifaddr *ia; 1507 int s, bound; 1508 1509 bound = curlwp_bind(); 1510 s = pserialize_read_enter(); 1511 IFADDR_READER_FOREACH(ifa, ifp) { 1512 struct psref psref; 1513 1514 if (ifa->ifa_addr->sa_family != AF_INET) 1515 continue; 1516 ifa_acquire(ifa, &psref); 1517 pserialize_read_exit(s); 1518 1519 ia = (struct in_ifaddr *)ifa; 1520 1521 /* Stop DAD processing */ 1522 if (ia->ia_dad_stop != NULL) 1523 ia->ia_dad_stop(ifa); 1524 1525 /* 1526 * Mark the address as detached. 1527 */ 1528 if (!(ia->ia4_flags & IN_IFF_DETACHED)) { 1529 ia->ia4_flags |= IN_IFF_DETACHED; 1530 ia->ia4_flags &= 1531 ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED); 1532 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1533 } 1534 1535 s = pserialize_read_enter(); 1536 ifa_release(ifa, &psref); 1537 } 1538 pserialize_read_exit(s); 1539 curlwp_bindx(bound); 1540 } 1541 1542 void 1543 in_if_down(struct ifnet *ifp) 1544 { 1545 1546 in_if_link_down(ifp); 1547 #if NARP > 0 1548 lltable_purge_entries(LLTABLE(ifp)); 1549 #endif 1550 } 1551 1552 void 1553 in_if_link_state_change(struct ifnet *ifp, int link_state) 1554 { 1555 1556 switch (link_state) { 1557 case LINK_STATE_DOWN: 1558 in_if_link_down(ifp); 1559 break; 1560 case LINK_STATE_UP: 1561 in_if_link_up(ifp); 1562 break; 1563 } 1564 } 1565 1566 /* 1567 * in_lookup_multi: look up the in_multi record for a given IP 1568 * multicast address on a given interface. If no matching record is 1569 * found, return NULL. 1570 */ 1571 struct in_multi * 1572 in_lookup_multi(struct in_addr addr, ifnet_t *ifp) 1573 { 1574 struct in_multi *inm; 1575 1576 KASSERT(rw_lock_held(&in_multilock)); 1577 1578 LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) { 1579 if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp) 1580 break; 1581 } 1582 return inm; 1583 } 1584 1585 /* 1586 * in_multi_group: check whether the address belongs to an IP multicast 1587 * group we are joined on this interface. Returns true or false. 1588 */ 1589 bool 1590 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags) 1591 { 1592 bool ingroup; 1593 1594 if (__predict_true(flags & IP_IGMP_MCAST) == 0) { 1595 rw_enter(&in_multilock, RW_READER); 1596 ingroup = in_lookup_multi(addr, ifp) != NULL; 1597 rw_exit(&in_multilock); 1598 } else { 1599 /* XXX Recursive call from ip_output(). */ 1600 KASSERT(rw_lock_held(&in_multilock)); 1601 ingroup = in_lookup_multi(addr, ifp) != NULL; 1602 } 1603 return ingroup; 1604 } 1605 1606 /* 1607 * Add an address to the list of IP multicast addresses for a given interface. 1608 */ 1609 struct in_multi * 1610 in_addmulti(struct in_addr *ap, ifnet_t *ifp) 1611 { 1612 struct sockaddr_in sin; 1613 struct in_multi *inm; 1614 1615 /* 1616 * See if address already in list. 1617 */ 1618 rw_enter(&in_multilock, RW_WRITER); 1619 inm = in_lookup_multi(*ap, ifp); 1620 if (inm != NULL) { 1621 /* 1622 * Found it; just increment the reference count. 1623 */ 1624 inm->inm_refcount++; 1625 rw_exit(&in_multilock); 1626 return inm; 1627 } 1628 1629 /* 1630 * New address; allocate a new multicast record. 1631 */ 1632 inm = pool_get(&inmulti_pool, PR_NOWAIT); 1633 if (inm == NULL) { 1634 rw_exit(&in_multilock); 1635 return NULL; 1636 } 1637 inm->inm_addr = *ap; 1638 inm->inm_ifp = ifp; 1639 inm->inm_refcount = 1; 1640 1641 /* 1642 * Ask the network driver to update its multicast reception 1643 * filter appropriately for the new address. 1644 */ 1645 sockaddr_in_init(&sin, ap, 0); 1646 if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) { 1647 rw_exit(&in_multilock); 1648 pool_put(&inmulti_pool, inm); 1649 return NULL; 1650 } 1651 1652 /* 1653 * Let IGMP know that we have joined a new IP multicast group. 1654 */ 1655 if (igmp_joingroup(inm) != 0) { 1656 rw_exit(&in_multilock); 1657 pool_put(&inmulti_pool, inm); 1658 return NULL; 1659 } 1660 LIST_INSERT_HEAD( 1661 &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp), 1662 inm, inm_list); 1663 in_multientries++; 1664 rw_exit(&in_multilock); 1665 1666 return inm; 1667 } 1668 1669 /* 1670 * Delete a multicast address record. 1671 */ 1672 void 1673 in_delmulti(struct in_multi *inm) 1674 { 1675 struct sockaddr_in sin; 1676 1677 rw_enter(&in_multilock, RW_WRITER); 1678 if (--inm->inm_refcount > 0) { 1679 rw_exit(&in_multilock); 1680 return; 1681 } 1682 1683 /* 1684 * No remaining claims to this record; let IGMP know that 1685 * we are leaving the multicast group. 1686 */ 1687 igmp_leavegroup(inm); 1688 1689 /* 1690 * Notify the network driver to update its multicast reception 1691 * filter. 1692 */ 1693 sockaddr_in_init(&sin, &inm->inm_addr, 0); 1694 if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin)); 1695 1696 /* 1697 * Unlink from list. 1698 */ 1699 LIST_REMOVE(inm, inm_list); 1700 in_multientries--; 1701 rw_exit(&in_multilock); 1702 1703 pool_put(&inmulti_pool, inm); 1704 } 1705 1706 /* 1707 * in_next_multi: step through all of the in_multi records, one at a time. 1708 * The current position is remembered in "step", which the caller must 1709 * provide. in_first_multi(), below, must be called to initialize "step" 1710 * and get the first record. Both macros return a NULL "inm" when there 1711 * are no remaining records. 1712 */ 1713 struct in_multi * 1714 in_next_multi(struct in_multistep *step) 1715 { 1716 struct in_multi *inm; 1717 1718 KASSERT(rw_lock_held(&in_multilock)); 1719 1720 while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) { 1721 step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]); 1722 } 1723 if ((inm = step->i_inm) != NULL) { 1724 step->i_inm = LIST_NEXT(inm, inm_list); 1725 } 1726 return inm; 1727 } 1728 1729 struct in_multi * 1730 in_first_multi(struct in_multistep *step) 1731 { 1732 KASSERT(rw_lock_held(&in_multilock)); 1733 1734 step->i_n = 0; 1735 step->i_inm = LIST_FIRST(&in_multihashtbl[0]); 1736 return in_next_multi(step); 1737 } 1738 1739 void 1740 in_multi_lock(int op) 1741 { 1742 rw_enter(&in_multilock, op); 1743 } 1744 1745 void 1746 in_multi_unlock(void) 1747 { 1748 rw_exit(&in_multilock); 1749 } 1750 1751 int 1752 in_multi_lock_held(void) 1753 { 1754 return rw_lock_held(&in_multilock); 1755 } 1756 1757 struct in_ifaddr * 1758 in_selectsrc(struct sockaddr_in *sin, struct route *ro, 1759 int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref) 1760 { 1761 struct rtentry *rt = NULL; 1762 struct in_ifaddr *ia = NULL; 1763 1764 KASSERT(ISSET(curlwp->l_pflag, LP_BOUND)); 1765 /* 1766 * If route is known or can be allocated now, take the 1767 * source address from the interface. Otherwise, punt. 1768 */ 1769 if ((soopts & SO_DONTROUTE) != 0) 1770 rtcache_free(ro); 1771 else { 1772 union { 1773 struct sockaddr dst; 1774 struct sockaddr_in dst4; 1775 } u; 1776 1777 sockaddr_in_init(&u.dst4, &sin->sin_addr, 0); 1778 rt = rtcache_lookup(ro, &u.dst); 1779 } 1780 /* 1781 * If we found a route, use the address 1782 * corresponding to the outgoing interface 1783 * unless it is the loopback (in case a route 1784 * to our address on another net goes to loopback). 1785 * 1786 * XXX Is this still true? Do we care? 1787 */ 1788 if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 1789 int s; 1790 struct ifaddr *ifa; 1791 /* 1792 * Just in case. May not need to do this workaround. 1793 * Revisit when working on rtentry MP-ification. 1794 */ 1795 s = pserialize_read_enter(); 1796 IFADDR_READER_FOREACH(ifa, rt->rt_ifp) { 1797 if (ifa == rt->rt_ifa) 1798 break; 1799 } 1800 if (ifa != NULL) 1801 ifa_acquire(ifa, psref); 1802 pserialize_read_exit(s); 1803 1804 ia = ifatoia(ifa); 1805 } 1806 if (ia == NULL) { 1807 u_int16_t fport = sin->sin_port; 1808 struct ifaddr *ifa; 1809 int s; 1810 1811 sin->sin_port = 0; 1812 ifa = ifa_ifwithladdr_psref(sintosa(sin), psref); 1813 sin->sin_port = fport; 1814 if (ifa == NULL) { 1815 /* Find 1st non-loopback AF_INET address */ 1816 s = pserialize_read_enter(); 1817 IN_ADDRLIST_READER_FOREACH(ia) { 1818 if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK)) 1819 break; 1820 } 1821 if (ia != NULL) 1822 ia4_acquire(ia, psref); 1823 pserialize_read_exit(s); 1824 } else { 1825 /* ia is already referenced by psref */ 1826 ia = ifatoia(ifa); 1827 } 1828 if (ia == NULL) { 1829 *errorp = EADDRNOTAVAIL; 1830 goto out; 1831 } 1832 } 1833 /* 1834 * If the destination address is multicast and an outgoing 1835 * interface has been set as a multicast option, use the 1836 * address of that interface as our source address. 1837 */ 1838 if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) { 1839 struct ip_moptions *imo; 1840 1841 imo = mopts; 1842 if (imo->imo_multicast_if_index != 0) { 1843 struct ifnet *ifp; 1844 int s; 1845 1846 if (ia != NULL) 1847 ia4_release(ia, psref); 1848 s = pserialize_read_enter(); 1849 ifp = if_byindex(imo->imo_multicast_if_index); 1850 if (ifp != NULL) { 1851 /* XXX */ 1852 ia = in_get_ia_from_ifp_psref(ifp, psref); 1853 } else 1854 ia = NULL; 1855 if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) { 1856 pserialize_read_exit(s); 1857 if (ia != NULL) 1858 ia4_release(ia, psref); 1859 *errorp = EADDRNOTAVAIL; 1860 ia = NULL; 1861 goto out; 1862 } 1863 pserialize_read_exit(s); 1864 } 1865 } 1866 if (ia->ia_ifa.ifa_getifa != NULL) { 1867 ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa, 1868 sintosa(sin))); 1869 if (ia == NULL) { 1870 *errorp = EADDRNOTAVAIL; 1871 goto out; 1872 } 1873 /* FIXME NOMPSAFE */ 1874 ia4_acquire(ia, psref); 1875 } 1876 #ifdef GETIFA_DEBUG 1877 else 1878 printf("%s: missing ifa_getifa\n", __func__); 1879 #endif 1880 out: 1881 rtcache_unref(rt, ro); 1882 return ia; 1883 } 1884 1885 int 1886 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst) 1887 { 1888 struct in_ifaddr *ia4; 1889 int s; 1890 1891 /* check for address match */ 1892 if (src.s_addr != ip->ip_dst.s_addr || 1893 dst.s_addr != ip->ip_src.s_addr) 1894 return 0; 1895 1896 /* martian filters on outer source - NOT done in ip_input! */ 1897 if (IN_MULTICAST(ip->ip_src.s_addr)) 1898 return 0; 1899 switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) { 1900 case 0: 1901 case 127: 1902 case 255: 1903 return 0; 1904 } 1905 /* reject packets with broadcast on source */ 1906 s = pserialize_read_enter(); 1907 IN_ADDRLIST_READER_FOREACH(ia4) { 1908 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 1909 continue; 1910 if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 1911 pserialize_read_exit(s); 1912 return 0; 1913 } 1914 } 1915 pserialize_read_exit(s); 1916 1917 /* NOTE: packet may dropped by uRPF */ 1918 1919 /* return valid bytes length */ 1920 return sizeof(src) + sizeof(dst); 1921 } 1922 1923 #if NARP > 0 1924 1925 #define IN_LLTBL_DEFAULT_HSIZE 32 1926 #define IN_LLTBL_HASH(k, h) \ 1927 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1928 1929 /* 1930 * Do actual deallocation of @lle. 1931 * Called by LLE_FREE_LOCKED when number of references 1932 * drops to zero. 1933 */ 1934 static void 1935 in_lltable_destroy_lle(struct llentry *lle) 1936 { 1937 1938 KASSERT(lle->la_numheld == 0); 1939 1940 LLE_WUNLOCK(lle); 1941 LLE_LOCK_DESTROY(lle); 1942 llentry_pool_put(lle); 1943 } 1944 1945 static struct llentry * 1946 in_lltable_new(struct in_addr addr4, u_int flags) 1947 { 1948 struct llentry *lle; 1949 1950 lle = llentry_pool_get(PR_NOWAIT); 1951 if (lle == NULL) /* NB: caller generates msg */ 1952 return NULL; 1953 1954 /* 1955 * For IPv4 this will trigger "arpresolve" to generate 1956 * an ARP request. 1957 */ 1958 lle->la_expire = time_uptime; /* mark expired */ 1959 lle->r_l3addr.addr4 = addr4; 1960 lle->lle_refcnt = 1; 1961 lle->lle_free = in_lltable_destroy_lle; 1962 LLE_LOCK_INIT(lle); 1963 callout_init(&lle->la_timer, CALLOUT_MPSAFE); 1964 1965 return lle; 1966 } 1967 1968 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1969 (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1970 1971 static int 1972 in_lltable_match_prefix(const struct sockaddr *prefix, 1973 const struct sockaddr *mask, u_int flags, struct llentry *lle) 1974 { 1975 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1976 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1977 struct in_addr lle_addr; 1978 1979 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1980 1981 /* 1982 * (flags & LLE_STATIC) means deleting all entries 1983 * including static ARP entries. 1984 */ 1985 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) && 1986 ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) 1987 return (1); 1988 1989 return (0); 1990 } 1991 1992 static void 1993 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1994 { 1995 size_t pkts_dropped; 1996 1997 LLE_WLOCK_ASSERT(lle); 1998 KASSERT(llt != NULL); 1999 2000 pkts_dropped = llentry_free(lle); 2001 arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped); 2002 } 2003 2004 static int 2005 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr, 2006 const struct rtentry *rt) 2007 { 2008 int error = EINVAL; 2009 2010 if (rt == NULL) 2011 return error; 2012 2013 /* 2014 * If the gateway for an existing host route matches the target L3 2015 * address, which is a special route inserted by some implementation 2016 * such as MANET, and the interface is of the correct type, then 2017 * allow for ARP to proceed. 2018 */ 2019 if (rt->rt_flags & RTF_GATEWAY) { 2020 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 2021 rt->rt_ifp->if_type != IFT_ETHER || 2022 (rt->rt_ifp->if_flags & IFF_NOARP) != 0 || 2023 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 2024 sizeof(in_addr_t)) != 0) { 2025 goto error; 2026 } 2027 } 2028 2029 /* 2030 * Make sure that at least the destination address is covered 2031 * by the route. This is for handling the case where 2 or more 2032 * interfaces have the same prefix. An incoming packet arrives 2033 * on one interface and the corresponding outgoing packet leaves 2034 * another interface. 2035 */ 2036 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 2037 const char *sa, *mask, *addr, *lim; 2038 int len; 2039 2040 mask = (const char *)rt_mask(rt); 2041 /* 2042 * Just being extra cautious to avoid some custom 2043 * code getting into trouble. 2044 */ 2045 if (mask == NULL) 2046 goto error; 2047 2048 sa = (const char *)rt_getkey(rt); 2049 addr = (const char *)l3addr; 2050 len = ((const struct sockaddr_in *)l3addr)->sin_len; 2051 lim = addr + len; 2052 2053 for ( ; addr < lim; sa++, mask++, addr++) { 2054 if ((*sa ^ *addr) & *mask) { 2055 #ifdef DIAGNOSTIC 2056 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 2057 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 2058 #endif 2059 goto error; 2060 } 2061 } 2062 } 2063 2064 error = 0; 2065 error: 2066 return error; 2067 } 2068 2069 static inline uint32_t 2070 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 2071 { 2072 2073 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 2074 } 2075 2076 static uint32_t 2077 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 2078 { 2079 2080 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 2081 } 2082 2083 static void 2084 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 2085 { 2086 struct sockaddr_in *sin; 2087 2088 sin = (struct sockaddr_in *)sa; 2089 memset(sin, 0, sizeof(*sin)); 2090 sin->sin_family = AF_INET; 2091 sin->sin_len = sizeof(*sin); 2092 sin->sin_addr = lle->r_l3addr.addr4; 2093 } 2094 2095 static inline struct llentry * 2096 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 2097 { 2098 struct llentry *lle; 2099 struct llentries *lleh; 2100 u_int hashidx; 2101 2102 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 2103 lleh = &llt->lle_head[hashidx]; 2104 LIST_FOREACH(lle, lleh, lle_next) { 2105 if (lle->la_flags & LLE_DELETED) 2106 continue; 2107 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 2108 break; 2109 } 2110 2111 return (lle); 2112 } 2113 2114 static int 2115 in_lltable_delete(struct lltable *llt, u_int flags, 2116 const struct sockaddr *l3addr) 2117 { 2118 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2119 struct ifnet *ifp __diagused = llt->llt_ifp; 2120 struct llentry *lle; 2121 2122 IF_AFDATA_WLOCK_ASSERT(ifp); 2123 KASSERTMSG(l3addr->sa_family == AF_INET, 2124 "sin_family %d", l3addr->sa_family); 2125 2126 lle = in_lltable_find_dst(llt, sin->sin_addr); 2127 if (lle == NULL) { 2128 #ifdef LLTABLE_DEBUG 2129 char buf[64]; 2130 sockaddr_format(l3addr, buf, sizeof(buf)); 2131 log(LOG_INFO, "%s: cache for %s is not found\n", 2132 __func__, buf); 2133 #endif 2134 return (ENOENT); 2135 } 2136 2137 LLE_WLOCK(lle); 2138 #ifdef LLTABLE_DEBUG 2139 { 2140 char buf[64]; 2141 sockaddr_format(l3addr, buf, sizeof(buf)); 2142 log(LOG_INFO, "%s: cache for %s (%p) is deleted\n", 2143 __func__, buf, lle); 2144 } 2145 #endif 2146 llentry_free(lle); 2147 2148 return (0); 2149 } 2150 2151 static struct llentry * 2152 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr, 2153 const struct rtentry *rt) 2154 { 2155 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2156 struct ifnet *ifp = llt->llt_ifp; 2157 struct llentry *lle; 2158 2159 IF_AFDATA_WLOCK_ASSERT(ifp); 2160 KASSERTMSG(l3addr->sa_family == AF_INET, 2161 "sin_family %d", l3addr->sa_family); 2162 2163 lle = in_lltable_find_dst(llt, sin->sin_addr); 2164 2165 if (lle != NULL) { 2166 LLE_WLOCK(lle); 2167 return (lle); 2168 } 2169 2170 /* no existing record, we need to create new one */ 2171 2172 /* 2173 * A route that covers the given address must have 2174 * been installed 1st because we are doing a resolution, 2175 * verify this. 2176 */ 2177 if (!(flags & LLE_IFADDR) && 2178 in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0) 2179 return (NULL); 2180 2181 lle = in_lltable_new(sin->sin_addr, flags); 2182 if (lle == NULL) { 2183 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 2184 return (NULL); 2185 } 2186 lle->la_flags = flags; 2187 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 2188 memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen); 2189 lle->la_flags |= (LLE_VALID | LLE_STATIC); 2190 } 2191 2192 lltable_link_entry(llt, lle); 2193 LLE_WLOCK(lle); 2194 2195 return (lle); 2196 } 2197 2198 /* 2199 * Return NULL if not found or marked for deletion. 2200 * If found return lle read locked. 2201 */ 2202 static struct llentry * 2203 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 2204 { 2205 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2206 struct llentry *lle; 2207 2208 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 2209 KASSERTMSG(l3addr->sa_family == AF_INET, 2210 "sin_family %d", l3addr->sa_family); 2211 2212 lle = in_lltable_find_dst(llt, sin->sin_addr); 2213 2214 if (lle == NULL) 2215 return NULL; 2216 2217 if (flags & LLE_EXCLUSIVE) 2218 LLE_WLOCK(lle); 2219 else 2220 LLE_RLOCK(lle); 2221 2222 return lle; 2223 } 2224 2225 static int 2226 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 2227 struct rt_walkarg *w) 2228 { 2229 struct sockaddr_in sin; 2230 2231 LLTABLE_LOCK_ASSERT(); 2232 2233 /* skip deleted entries */ 2234 if (lle->la_flags & LLE_DELETED) 2235 return 0; 2236 2237 sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0); 2238 2239 return lltable_dump_entry(llt, lle, w, sintosa(&sin)); 2240 } 2241 2242 #endif /* NARP > 0 */ 2243 2244 static int 2245 in_multicast_sysctl(SYSCTLFN_ARGS) 2246 { 2247 struct ifnet *ifp; 2248 struct ifaddr *ifa; 2249 struct in_ifaddr *ifa4; 2250 struct in_multi *inm; 2251 uint32_t tmp; 2252 int error; 2253 size_t written; 2254 struct psref psref; 2255 int bound; 2256 2257 if (namelen != 1) 2258 return EINVAL; 2259 2260 bound = curlwp_bind(); 2261 ifp = if_get_byindex(name[0], &psref); 2262 if (ifp == NULL) { 2263 curlwp_bindx(bound); 2264 return ENODEV; 2265 } 2266 2267 if (oldp == NULL) { 2268 *oldlenp = 0; 2269 IFADDR_FOREACH(ifa, ifp) { 2270 if (ifa->ifa_addr->sa_family != AF_INET) 2271 continue; 2272 ifa4 = (void *)ifa; 2273 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2274 *oldlenp += 2 * sizeof(struct in_addr) + 2275 sizeof(uint32_t); 2276 } 2277 } 2278 if_put(ifp, &psref); 2279 curlwp_bindx(bound); 2280 return 0; 2281 } 2282 2283 error = 0; 2284 written = 0; 2285 IFADDR_FOREACH(ifa, ifp) { 2286 if (ifa->ifa_addr->sa_family != AF_INET) 2287 continue; 2288 ifa4 = (void *)ifa; 2289 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2290 if (written + 2 * sizeof(struct in_addr) + 2291 sizeof(uint32_t) > *oldlenp) 2292 goto done; 2293 error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr, 2294 oldp, sizeof(struct in_addr)); 2295 if (error) 2296 goto done; 2297 oldp = (char *)oldp + sizeof(struct in_addr); 2298 written += sizeof(struct in_addr); 2299 error = sysctl_copyout(l, &inm->inm_addr, 2300 oldp, sizeof(struct in_addr)); 2301 if (error) 2302 goto done; 2303 oldp = (char *)oldp + sizeof(struct in_addr); 2304 written += sizeof(struct in_addr); 2305 tmp = inm->inm_refcount; 2306 error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp)); 2307 if (error) 2308 goto done; 2309 oldp = (char *)oldp + sizeof(tmp); 2310 written += sizeof(tmp); 2311 } 2312 } 2313 done: 2314 if_put(ifp, &psref); 2315 curlwp_bindx(bound); 2316 *oldlenp = written; 2317 return error; 2318 } 2319 2320 static void 2321 in_sysctl_init(struct sysctllog **clog) 2322 { 2323 sysctl_createv(clog, 0, NULL, NULL, 2324 CTLFLAG_PERMANENT, 2325 CTLTYPE_NODE, "inet", 2326 SYSCTL_DESCR("PF_INET related settings"), 2327 NULL, 0, NULL, 0, 2328 CTL_NET, PF_INET, CTL_EOL); 2329 sysctl_createv(clog, 0, NULL, NULL, 2330 CTLFLAG_PERMANENT, 2331 CTLTYPE_NODE, "multicast", 2332 SYSCTL_DESCR("Multicast information"), 2333 in_multicast_sysctl, 0, NULL, 0, 2334 CTL_NET, PF_INET, CTL_CREATE, CTL_EOL); 2335 sysctl_createv(clog, 0, NULL, NULL, 2336 CTLFLAG_PERMANENT, 2337 CTLTYPE_NODE, "ip", 2338 SYSCTL_DESCR("IPv4 related settings"), 2339 NULL, 0, NULL, 0, 2340 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2341 2342 sysctl_createv(clog, 0, NULL, NULL, 2343 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2344 CTLTYPE_INT, "subnetsarelocal", 2345 SYSCTL_DESCR("Whether logical subnets are considered " 2346 "local"), 2347 NULL, 0, &subnetsarelocal, 0, 2348 CTL_NET, PF_INET, IPPROTO_IP, 2349 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2350 sysctl_createv(clog, 0, NULL, NULL, 2351 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2352 CTLTYPE_INT, "hostzerobroadcast", 2353 SYSCTL_DESCR("All zeroes address is broadcast address"), 2354 NULL, 0, &hostzeroisbroadcast, 0, 2355 CTL_NET, PF_INET, IPPROTO_IP, 2356 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2357 } 2358 2359 #if NARP > 0 2360 2361 static struct lltable * 2362 in_lltattach(struct ifnet *ifp) 2363 { 2364 struct lltable *llt; 2365 2366 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 2367 llt->llt_af = AF_INET; 2368 llt->llt_ifp = ifp; 2369 2370 llt->llt_lookup = in_lltable_lookup; 2371 llt->llt_create = in_lltable_create; 2372 llt->llt_delete = in_lltable_delete; 2373 llt->llt_dump_entry = in_lltable_dump_entry; 2374 llt->llt_hash = in_lltable_hash; 2375 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 2376 llt->llt_free_entry = in_lltable_free_entry; 2377 llt->llt_match_prefix = in_lltable_match_prefix; 2378 lltable_link(llt); 2379 2380 return (llt); 2381 } 2382 2383 #endif /* NARP > 0 */ 2384 2385 void * 2386 in_domifattach(struct ifnet *ifp) 2387 { 2388 struct in_ifinfo *ii; 2389 2390 ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP); 2391 2392 #if NARP > 0 2393 ii->ii_llt = in_lltattach(ifp); 2394 #endif 2395 2396 #ifdef IPSELSRC 2397 ii->ii_selsrc = in_selsrc_domifattach(ifp); 2398 KASSERT(ii->ii_selsrc != NULL); 2399 #endif 2400 2401 return ii; 2402 } 2403 2404 void 2405 in_domifdetach(struct ifnet *ifp, void *aux) 2406 { 2407 struct in_ifinfo *ii = aux; 2408 2409 #ifdef IPSELSRC 2410 in_selsrc_domifdetach(ifp, ii->ii_selsrc); 2411 #endif 2412 #if NARP > 0 2413 lltable_free(ii->ii_llt); 2414 #endif 2415 kmem_free(ii, sizeof(struct in_ifinfo)); 2416 } 2417