in.c revision 1.216
1 /* $NetBSD: in.c,v 1.216 2018/01/19 08:01:05 ozaki-r Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1991, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)in.c 8.4 (Berkeley) 1/9/95 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: in.c,v 1.216 2018/01/19 08:01:05 ozaki-r Exp $"); 95 96 #include "arp.h" 97 98 #ifdef _KERNEL_OPT 99 #include "opt_inet.h" 100 #include "opt_inet_conf.h" 101 #include "opt_mrouting.h" 102 #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 SOFTNET_LOCK_UNLESS_NET_MPSAFE(); 755 error = in_control0(so, cmd, data, ifp); 756 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE(); 757 758 return error; 759 } 760 761 /* Add ownaddr as loopback rtentry. */ 762 static void 763 in_ifaddlocal(struct ifaddr *ifa) 764 { 765 struct in_ifaddr *ia; 766 767 ia = (struct in_ifaddr *)ifa; 768 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY || 769 (ia->ia_ifp->if_flags & IFF_POINTOPOINT && 770 in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr))) 771 { 772 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 773 return; 774 } 775 776 rt_ifa_addlocal(ifa); 777 } 778 779 /* Remove loopback entry of ownaddr */ 780 static void 781 in_ifremlocal(struct ifaddr *ifa) 782 { 783 struct in_ifaddr *ia, *p; 784 struct ifaddr *alt_ifa = NULL; 785 int ia_count = 0; 786 int s; 787 struct psref psref; 788 int bound = curlwp_bind(); 789 790 ia = (struct in_ifaddr *)ifa; 791 /* Delete the entry if exactly one ifaddr matches the 792 * address, ifa->ifa_addr. */ 793 s = pserialize_read_enter(); 794 IN_ADDRLIST_READER_FOREACH(p) { 795 if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr)) 796 continue; 797 if (p->ia_ifp != ia->ia_ifp) 798 alt_ifa = &p->ia_ifa; 799 if (++ia_count > 1 && alt_ifa != NULL) 800 break; 801 } 802 if (alt_ifa != NULL && ia_count > 1) 803 ifa_acquire(alt_ifa, &psref); 804 pserialize_read_exit(s); 805 806 if (ia_count == 0) 807 goto out; 808 809 rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa); 810 if (alt_ifa != NULL && ia_count > 1) 811 ifa_release(alt_ifa, &psref); 812 out: 813 curlwp_bindx(bound); 814 } 815 816 static void 817 in_scrubaddr(struct in_ifaddr *ia) 818 { 819 820 /* stop DAD processing */ 821 if (ia->ia_dad_stop != NULL) 822 ia->ia_dad_stop(&ia->ia_ifa); 823 824 in_scrubprefix(ia); 825 in_ifremlocal(&ia->ia_ifa); 826 827 mutex_enter(&in_ifaddr_lock); 828 if (ia->ia_allhosts != NULL) { 829 in_delmulti(ia->ia_allhosts); 830 ia->ia_allhosts = NULL; 831 } 832 mutex_exit(&in_ifaddr_lock); 833 } 834 835 /* 836 * Depends on it isn't called in concurrent. It should be guaranteed 837 * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control 838 * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl 839 * and the latter is called via ifa->ifa_ifp's if_detach. The functions 840 * never be executed in concurrent. 841 */ 842 void 843 in_purgeaddr(struct ifaddr *ifa) 844 { 845 struct in_ifaddr *ia = (void *) ifa; 846 struct ifnet *ifp = ifa->ifa_ifp; 847 848 /* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */ 849 850 ifa->ifa_flags |= IFA_DESTROYING; 851 in_scrubaddr(ia); 852 853 mutex_enter(&in_ifaddr_lock); 854 in_addrhash_remove_locked(ia); 855 TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list); 856 IN_ADDRLIST_WRITER_REMOVE(ia); 857 ifa_remove(ifp, &ia->ia_ifa); 858 /* Assume ifa_remove called pserialize_perform and psref_destroy */ 859 mutex_exit(&in_ifaddr_lock); 860 IN_ADDRHASH_ENTRY_DESTROY(ia); 861 IN_ADDRLIST_ENTRY_DESTROY(ia); 862 ifafree(&ia->ia_ifa); 863 in_setmaxmtu(); 864 } 865 866 static void 867 in_addrhash_insert_locked(struct in_ifaddr *ia) 868 { 869 870 KASSERT(mutex_owned(&in_ifaddr_lock)); 871 872 LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 873 ia_hash); 874 IN_ADDRHASH_ENTRY_INIT(ia); 875 IN_ADDRHASH_WRITER_INSERT_HEAD(ia); 876 } 877 878 void 879 in_addrhash_insert(struct in_ifaddr *ia) 880 { 881 882 mutex_enter(&in_ifaddr_lock); 883 in_addrhash_insert_locked(ia); 884 mutex_exit(&in_ifaddr_lock); 885 } 886 887 static void 888 in_addrhash_remove_locked(struct in_ifaddr *ia) 889 { 890 891 KASSERT(mutex_owned(&in_ifaddr_lock)); 892 893 LIST_REMOVE(ia, ia_hash); 894 IN_ADDRHASH_WRITER_REMOVE(ia); 895 } 896 897 void 898 in_addrhash_remove(struct in_ifaddr *ia) 899 { 900 901 mutex_enter(&in_ifaddr_lock); 902 in_addrhash_remove_locked(ia); 903 #ifdef NET_MPSAFE 904 pserialize_perform(in_ifaddrhash_psz); 905 #endif 906 mutex_exit(&in_ifaddr_lock); 907 IN_ADDRHASH_ENTRY_DESTROY(ia); 908 } 909 910 void 911 in_purgeif(struct ifnet *ifp) /* MUST be called at splsoftnet() */ 912 { 913 914 IFNET_LOCK(ifp); 915 if_purgeaddrs(ifp, AF_INET, in_purgeaddr); 916 igmp_purgeif(ifp); /* manipulates pools */ 917 #ifdef MROUTING 918 ip_mrouter_detach(ifp); 919 #endif 920 IFNET_UNLOCK(ifp); 921 } 922 923 /* 924 * SIOC[GAD]LIFADDR. 925 * SIOCGLIFADDR: get first address. (???) 926 * SIOCGLIFADDR with IFLR_PREFIX: 927 * get first address that matches the specified prefix. 928 * SIOCALIFADDR: add the specified address. 929 * SIOCALIFADDR with IFLR_PREFIX: 930 * EINVAL since we can't deduce hostid part of the address. 931 * SIOCDLIFADDR: delete the specified address. 932 * SIOCDLIFADDR with IFLR_PREFIX: 933 * delete the first address that matches the specified prefix. 934 * return values: 935 * EINVAL on invalid parameters 936 * EADDRNOTAVAIL on prefix match failed/specified address not found 937 * other values may be returned from in_ioctl() 938 */ 939 static int 940 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data, 941 struct ifnet *ifp) 942 { 943 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 944 struct ifaddr *ifa; 945 struct sockaddr *sa; 946 947 /* sanity checks */ 948 if (data == NULL || ifp == NULL) { 949 panic("invalid argument to in_lifaddr_ioctl"); 950 /*NOTRECHED*/ 951 } 952 953 switch (cmd) { 954 case SIOCGLIFADDR: 955 /* address must be specified on GET with IFLR_PREFIX */ 956 if ((iflr->flags & IFLR_PREFIX) == 0) 957 break; 958 /*FALLTHROUGH*/ 959 case SIOCALIFADDR: 960 case SIOCDLIFADDR: 961 /* address must be specified on ADD and DELETE */ 962 sa = (struct sockaddr *)&iflr->addr; 963 if (sa->sa_family != AF_INET) 964 return EINVAL; 965 if (sa->sa_len != sizeof(struct sockaddr_in)) 966 return EINVAL; 967 /* XXX need improvement */ 968 sa = (struct sockaddr *)&iflr->dstaddr; 969 if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET) 970 return EINVAL; 971 if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in)) 972 return EINVAL; 973 break; 974 default: /*shouldn't happen*/ 975 #if 0 976 panic("invalid cmd to in_lifaddr_ioctl"); 977 /*NOTREACHED*/ 978 #else 979 return EOPNOTSUPP; 980 #endif 981 } 982 if (sizeof(struct in_addr) * NBBY < iflr->prefixlen) 983 return EINVAL; 984 985 switch (cmd) { 986 case SIOCALIFADDR: 987 { 988 struct in_aliasreq ifra; 989 990 if (iflr->flags & IFLR_PREFIX) 991 return EINVAL; 992 993 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */ 994 memset(&ifra, 0, sizeof(ifra)); 995 memcpy(ifra.ifra_name, iflr->iflr_name, 996 sizeof(ifra.ifra_name)); 997 998 memcpy(&ifra.ifra_addr, &iflr->addr, 999 ((struct sockaddr *)&iflr->addr)->sa_len); 1000 1001 if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /*XXX*/ 1002 memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr, 1003 ((struct sockaddr *)&iflr->dstaddr)->sa_len); 1004 } 1005 1006 ifra.ifra_mask.sin_family = AF_INET; 1007 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 1008 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 1009 1010 return in_control(so, SIOCAIFADDR, &ifra, ifp); 1011 } 1012 case SIOCGLIFADDR: 1013 case SIOCDLIFADDR: 1014 { 1015 struct in_ifaddr *ia; 1016 struct in_addr mask, candidate, match; 1017 struct sockaddr_in *sin; 1018 int cmp, s; 1019 1020 memset(&mask, 0, sizeof(mask)); 1021 memset(&match, 0, sizeof(match)); /* XXX gcc */ 1022 if (iflr->flags & IFLR_PREFIX) { 1023 /* lookup a prefix rather than address. */ 1024 in_len2mask(&mask, iflr->prefixlen); 1025 1026 sin = (struct sockaddr_in *)&iflr->addr; 1027 match.s_addr = sin->sin_addr.s_addr; 1028 match.s_addr &= mask.s_addr; 1029 1030 /* if you set extra bits, that's wrong */ 1031 if (match.s_addr != sin->sin_addr.s_addr) 1032 return EINVAL; 1033 1034 cmp = 1; 1035 } else { 1036 if (cmd == SIOCGLIFADDR) { 1037 /* on getting an address, take the 1st match */ 1038 cmp = 0; /*XXX*/ 1039 } else { 1040 /* on deleting an address, do exact match */ 1041 in_len2mask(&mask, 32); 1042 sin = (struct sockaddr_in *)&iflr->addr; 1043 match.s_addr = sin->sin_addr.s_addr; 1044 1045 cmp = 1; 1046 } 1047 } 1048 1049 s = pserialize_read_enter(); 1050 IFADDR_READER_FOREACH(ifa, ifp) { 1051 if (ifa->ifa_addr->sa_family != AF_INET) 1052 continue; 1053 if (cmp == 0) 1054 break; 1055 candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr; 1056 candidate.s_addr &= mask.s_addr; 1057 if (candidate.s_addr == match.s_addr) 1058 break; 1059 } 1060 if (ifa == NULL) { 1061 pserialize_read_exit(s); 1062 return EADDRNOTAVAIL; 1063 } 1064 ia = (struct in_ifaddr *)ifa; 1065 1066 if (cmd == SIOCGLIFADDR) { 1067 /* fill in the if_laddrreq structure */ 1068 memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len); 1069 1070 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1071 memcpy(&iflr->dstaddr, &ia->ia_dstaddr, 1072 ia->ia_dstaddr.sin_len); 1073 } else 1074 memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr)); 1075 1076 iflr->prefixlen = 1077 in_mask2len(&ia->ia_sockmask.sin_addr); 1078 1079 iflr->flags = 0; /*XXX*/ 1080 pserialize_read_exit(s); 1081 1082 return 0; 1083 } else { 1084 struct in_aliasreq ifra; 1085 1086 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */ 1087 memset(&ifra, 0, sizeof(ifra)); 1088 memcpy(ifra.ifra_name, iflr->iflr_name, 1089 sizeof(ifra.ifra_name)); 1090 1091 memcpy(&ifra.ifra_addr, &ia->ia_addr, 1092 ia->ia_addr.sin_len); 1093 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1094 memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr, 1095 ia->ia_dstaddr.sin_len); 1096 } 1097 memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask, 1098 ia->ia_sockmask.sin_len); 1099 pserialize_read_exit(s); 1100 1101 return in_control(so, SIOCDIFADDR, &ifra, ifp); 1102 } 1103 } 1104 } 1105 1106 return EOPNOTSUPP; /*just for safety*/ 1107 } 1108 1109 /* 1110 * Initialize an interface's internet address 1111 * and routing table entry. 1112 */ 1113 int 1114 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, 1115 const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub) 1116 { 1117 u_int32_t i; 1118 struct sockaddr_in oldaddr, olddst; 1119 int s, oldflags, flags = RTF_UP, error, hostIsNew; 1120 1121 if (sin == NULL) 1122 sin = &ia->ia_addr; 1123 if (dst == NULL) 1124 dst = &ia->ia_dstaddr; 1125 1126 /* 1127 * Set up new addresses. 1128 */ 1129 oldaddr = ia->ia_addr; 1130 olddst = ia->ia_dstaddr; 1131 oldflags = ia->ia4_flags; 1132 ia->ia_addr = *sin; 1133 ia->ia_dstaddr = *dst; 1134 hostIsNew = oldaddr.sin_family != AF_INET || 1135 !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr); 1136 if (!scrub) 1137 scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family || 1138 !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr); 1139 1140 /* 1141 * Configure address flags. 1142 * We need to do this early because they maybe adjusted 1143 * by if_addr_init depending on the address. 1144 */ 1145 if (ia->ia4_flags & IN_IFF_DUPLICATED) { 1146 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1147 hostIsNew = 1; 1148 } 1149 if (ifp->if_link_state == LINK_STATE_DOWN) { 1150 ia->ia4_flags |= IN_IFF_DETACHED; 1151 ia->ia4_flags &= ~IN_IFF_TENTATIVE; 1152 } else if (hostIsNew && if_do_dad(ifp)) 1153 ia->ia4_flags |= IN_IFF_TRYTENTATIVE; 1154 1155 /* 1156 * Give the interface a chance to initialize 1157 * if this is its first address, 1158 * and to validate the address if necessary. 1159 */ 1160 s = splsoftnet(); 1161 error = if_addr_init(ifp, &ia->ia_ifa, true); 1162 splx(s); 1163 /* Now clear the try tentative flag, its job is done. */ 1164 ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE; 1165 if (error != 0) { 1166 ia->ia_addr = oldaddr; 1167 ia->ia_dstaddr = olddst; 1168 ia->ia4_flags = oldflags; 1169 return error; 1170 } 1171 1172 if (scrub || hostIsNew) { 1173 int newflags = ia->ia4_flags; 1174 1175 ia->ia_ifa.ifa_addr = sintosa(&oldaddr); 1176 ia->ia_ifa.ifa_dstaddr = sintosa(&olddst); 1177 ia->ia4_flags = oldflags; 1178 if (hostIsNew) 1179 in_scrubaddr(ia); 1180 else if (scrub) 1181 in_scrubprefix(ia); 1182 ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr); 1183 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 1184 ia->ia4_flags = newflags; 1185 } 1186 1187 i = ia->ia_addr.sin_addr.s_addr; 1188 if (ifp->if_flags & IFF_POINTOPOINT) 1189 ia->ia_netmask = INADDR_BROADCAST; /* default to /32 */ 1190 else if (IN_CLASSA(i)) 1191 ia->ia_netmask = IN_CLASSA_NET; 1192 else if (IN_CLASSB(i)) 1193 ia->ia_netmask = IN_CLASSB_NET; 1194 else 1195 ia->ia_netmask = IN_CLASSC_NET; 1196 /* 1197 * The subnet mask usually includes at least the standard network part, 1198 * but may may be smaller in the case of supernetting. 1199 * If it is set, we believe it. 1200 */ 1201 if (ia->ia_subnetmask == 0) { 1202 ia->ia_subnetmask = ia->ia_netmask; 1203 ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask; 1204 } else 1205 ia->ia_netmask &= ia->ia_subnetmask; 1206 1207 ia->ia_net = i & ia->ia_netmask; 1208 ia->ia_subnet = i & ia->ia_subnetmask; 1209 in_socktrim(&ia->ia_sockmask); 1210 1211 /* re-calculate the "in_maxmtu" value */ 1212 in_setmaxmtu(); 1213 1214 ia->ia_ifa.ifa_metric = ifp->if_metric; 1215 if (ifp->if_flags & IFF_BROADCAST) { 1216 ia->ia_broadaddr.sin_addr.s_addr = 1217 ia->ia_subnet | ~ia->ia_subnetmask; 1218 ia->ia_netbroadcast.s_addr = 1219 ia->ia_net | ~ia->ia_netmask; 1220 } else if (ifp->if_flags & IFF_LOOPBACK) { 1221 ia->ia_dstaddr = ia->ia_addr; 1222 flags |= RTF_HOST; 1223 } else if (ifp->if_flags & IFF_POINTOPOINT) { 1224 if (ia->ia_dstaddr.sin_family != AF_INET) 1225 return (0); 1226 flags |= RTF_HOST; 1227 } 1228 1229 /* Add the local route to the address */ 1230 in_ifaddlocal(&ia->ia_ifa); 1231 1232 /* Add the prefix route for the address */ 1233 error = in_addprefix(ia, flags); 1234 1235 /* 1236 * If the interface supports multicast, join the "all hosts" 1237 * multicast group on that interface. 1238 */ 1239 mutex_enter(&in_ifaddr_lock); 1240 if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) { 1241 struct in_addr addr; 1242 1243 addr.s_addr = INADDR_ALLHOSTS_GROUP; 1244 ia->ia_allhosts = in_addmulti(&addr, ifp); 1245 } 1246 mutex_exit(&in_ifaddr_lock); 1247 1248 if (hostIsNew && 1249 ia->ia4_flags & IN_IFF_TENTATIVE && 1250 if_do_dad(ifp)) 1251 ia->ia_dad_start((struct ifaddr *)ia); 1252 1253 return error; 1254 } 1255 1256 #define rtinitflags(x) \ 1257 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 1258 ? RTF_HOST : 0) 1259 1260 /* 1261 * add a route to prefix ("connected route" in cisco terminology). 1262 * does nothing if there's some interface address with the same prefix already. 1263 */ 1264 static int 1265 in_addprefix(struct in_ifaddr *target, int flags) 1266 { 1267 struct in_ifaddr *ia; 1268 struct in_addr prefix, mask, p; 1269 int error; 1270 int s; 1271 1272 if ((flags & RTF_HOST) != 0) 1273 prefix = target->ia_dstaddr.sin_addr; 1274 else { 1275 prefix = target->ia_addr.sin_addr; 1276 mask = target->ia_sockmask.sin_addr; 1277 prefix.s_addr &= mask.s_addr; 1278 } 1279 1280 s = pserialize_read_enter(); 1281 IN_ADDRLIST_READER_FOREACH(ia) { 1282 if (rtinitflags(ia)) 1283 p = ia->ia_dstaddr.sin_addr; 1284 else { 1285 p = ia->ia_addr.sin_addr; 1286 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1287 } 1288 1289 if (prefix.s_addr != p.s_addr) 1290 continue; 1291 1292 /* 1293 * if we got a matching prefix route inserted by other 1294 * interface address, we don't need to bother 1295 * 1296 * XXX RADIX_MPATH implications here? -dyoung 1297 */ 1298 if (ia->ia_flags & IFA_ROUTE) { 1299 pserialize_read_exit(s); 1300 return 0; 1301 } 1302 } 1303 pserialize_read_exit(s); 1304 1305 /* 1306 * noone seem to have prefix route. insert it. 1307 */ 1308 error = rtinit(&target->ia_ifa, RTM_ADD, flags); 1309 if (error == 0) 1310 target->ia_flags |= IFA_ROUTE; 1311 else if (error == EEXIST) { 1312 /* 1313 * the fact the route already exists is not an error. 1314 */ 1315 error = 0; 1316 } 1317 return error; 1318 } 1319 1320 /* 1321 * remove a route to prefix ("connected route" in cisco terminology). 1322 * re-installs the route by using another interface address, if there's one 1323 * with the same prefix (otherwise we lose the route mistakenly). 1324 */ 1325 static int 1326 in_scrubprefix(struct in_ifaddr *target) 1327 { 1328 struct in_ifaddr *ia; 1329 struct in_addr prefix, mask, p; 1330 int error; 1331 int s; 1332 1333 /* If we don't have IFA_ROUTE we have nothing to do */ 1334 if ((target->ia_flags & IFA_ROUTE) == 0) 1335 return 0; 1336 1337 if (rtinitflags(target)) 1338 prefix = target->ia_dstaddr.sin_addr; 1339 else { 1340 prefix = target->ia_addr.sin_addr; 1341 mask = target->ia_sockmask.sin_addr; 1342 prefix.s_addr &= mask.s_addr; 1343 } 1344 1345 s = pserialize_read_enter(); 1346 IN_ADDRLIST_READER_FOREACH(ia) { 1347 if (rtinitflags(ia)) 1348 p = ia->ia_dstaddr.sin_addr; 1349 else { 1350 p = ia->ia_addr.sin_addr; 1351 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1352 } 1353 1354 if (prefix.s_addr != p.s_addr) 1355 continue; 1356 1357 /* 1358 * if we got a matching prefix route, move IFA_ROUTE to him 1359 */ 1360 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1361 struct psref psref; 1362 int bound = curlwp_bind(); 1363 1364 ia4_acquire(ia, &psref); 1365 pserialize_read_exit(s); 1366 1367 rtinit(&target->ia_ifa, RTM_DELETE, 1368 rtinitflags(target)); 1369 target->ia_flags &= ~IFA_ROUTE; 1370 1371 error = rtinit(&ia->ia_ifa, RTM_ADD, 1372 rtinitflags(ia) | RTF_UP); 1373 if (error == 0) 1374 ia->ia_flags |= IFA_ROUTE; 1375 1376 ia4_release(ia, &psref); 1377 curlwp_bindx(bound); 1378 1379 return error; 1380 } 1381 } 1382 pserialize_read_exit(s); 1383 1384 /* 1385 * noone seem to have prefix route. remove it. 1386 */ 1387 rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target)); 1388 target->ia_flags &= ~IFA_ROUTE; 1389 return 0; 1390 } 1391 1392 #undef rtinitflags 1393 1394 /* 1395 * Return 1 if the address might be a local broadcast address. 1396 */ 1397 int 1398 in_broadcast(struct in_addr in, struct ifnet *ifp) 1399 { 1400 struct ifaddr *ifa; 1401 int s; 1402 1403 KASSERT(ifp != NULL); 1404 1405 if (in.s_addr == INADDR_BROADCAST || 1406 in_nullhost(in)) 1407 return 1; 1408 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1409 return 0; 1410 /* 1411 * Look through the list of addresses for a match 1412 * with a broadcast address. 1413 */ 1414 #define ia (ifatoia(ifa)) 1415 s = pserialize_read_enter(); 1416 IFADDR_READER_FOREACH(ifa, ifp) { 1417 if (ifa->ifa_addr->sa_family == AF_INET && 1418 !in_hosteq(in, ia->ia_addr.sin_addr) && 1419 (in_hosteq(in, ia->ia_broadaddr.sin_addr) || 1420 in_hosteq(in, ia->ia_netbroadcast) || 1421 (hostzeroisbroadcast && 1422 /* 1423 * Check for old-style (host 0) broadcast. 1424 */ 1425 (in.s_addr == ia->ia_subnet || 1426 in.s_addr == ia->ia_net)))) { 1427 pserialize_read_exit(s); 1428 return 1; 1429 } 1430 } 1431 pserialize_read_exit(s); 1432 return (0); 1433 #undef ia 1434 } 1435 1436 /* 1437 * perform DAD when interface becomes IFF_UP. 1438 */ 1439 void 1440 in_if_link_up(struct ifnet *ifp) 1441 { 1442 struct ifaddr *ifa; 1443 struct in_ifaddr *ia; 1444 int s, bound; 1445 1446 /* Ensure it's sane to run DAD */ 1447 if (ifp->if_link_state == LINK_STATE_DOWN) 1448 return; 1449 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) 1450 return; 1451 1452 bound = curlwp_bind(); 1453 s = pserialize_read_enter(); 1454 IFADDR_READER_FOREACH(ifa, ifp) { 1455 struct psref psref; 1456 1457 if (ifa->ifa_addr->sa_family != AF_INET) 1458 continue; 1459 ifa_acquire(ifa, &psref); 1460 pserialize_read_exit(s); 1461 1462 ia = (struct in_ifaddr *)ifa; 1463 1464 /* If detached then mark as tentative */ 1465 if (ia->ia4_flags & IN_IFF_DETACHED) { 1466 ia->ia4_flags &= ~IN_IFF_DETACHED; 1467 if (if_do_dad(ifp) && ia->ia_dad_start != NULL) 1468 ia->ia4_flags |= IN_IFF_TENTATIVE; 1469 else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0) 1470 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1471 } 1472 1473 if (ia->ia4_flags & IN_IFF_TENTATIVE) { 1474 /* Clear the duplicated flag as we're starting DAD. */ 1475 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1476 ia->ia_dad_start(ifa); 1477 } 1478 1479 s = pserialize_read_enter(); 1480 ifa_release(ifa, &psref); 1481 } 1482 pserialize_read_exit(s); 1483 curlwp_bindx(bound); 1484 } 1485 1486 void 1487 in_if_up(struct ifnet *ifp) 1488 { 1489 1490 /* interface may not support link state, so bring it up also */ 1491 in_if_link_up(ifp); 1492 } 1493 1494 /* 1495 * Mark all addresses as detached. 1496 */ 1497 void 1498 in_if_link_down(struct ifnet *ifp) 1499 { 1500 struct ifaddr *ifa; 1501 struct in_ifaddr *ia; 1502 int s, bound; 1503 1504 bound = curlwp_bind(); 1505 s = pserialize_read_enter(); 1506 IFADDR_READER_FOREACH(ifa, ifp) { 1507 struct psref psref; 1508 1509 if (ifa->ifa_addr->sa_family != AF_INET) 1510 continue; 1511 ifa_acquire(ifa, &psref); 1512 pserialize_read_exit(s); 1513 1514 ia = (struct in_ifaddr *)ifa; 1515 1516 /* Stop DAD processing */ 1517 if (ia->ia_dad_stop != NULL) 1518 ia->ia_dad_stop(ifa); 1519 1520 /* 1521 * Mark the address as detached. 1522 */ 1523 if (!(ia->ia4_flags & IN_IFF_DETACHED)) { 1524 ia->ia4_flags |= IN_IFF_DETACHED; 1525 ia->ia4_flags &= 1526 ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED); 1527 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1528 } 1529 1530 s = pserialize_read_enter(); 1531 ifa_release(ifa, &psref); 1532 } 1533 pserialize_read_exit(s); 1534 curlwp_bindx(bound); 1535 } 1536 1537 void 1538 in_if_down(struct ifnet *ifp) 1539 { 1540 1541 in_if_link_down(ifp); 1542 lltable_purge_entries(LLTABLE(ifp)); 1543 } 1544 1545 void 1546 in_if_link_state_change(struct ifnet *ifp, int link_state) 1547 { 1548 1549 switch (link_state) { 1550 case LINK_STATE_DOWN: 1551 in_if_link_down(ifp); 1552 break; 1553 case LINK_STATE_UP: 1554 in_if_link_up(ifp); 1555 break; 1556 } 1557 } 1558 1559 /* 1560 * in_lookup_multi: look up the in_multi record for a given IP 1561 * multicast address on a given interface. If no matching record is 1562 * found, return NULL. 1563 */ 1564 struct in_multi * 1565 in_lookup_multi(struct in_addr addr, ifnet_t *ifp) 1566 { 1567 struct in_multi *inm; 1568 1569 KASSERT(rw_lock_held(&in_multilock)); 1570 1571 LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) { 1572 if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp) 1573 break; 1574 } 1575 return inm; 1576 } 1577 1578 /* 1579 * in_multi_group: check whether the address belongs to an IP multicast 1580 * group we are joined on this interface. Returns true or false. 1581 */ 1582 bool 1583 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags) 1584 { 1585 bool ingroup; 1586 1587 if (__predict_true(flags & IP_IGMP_MCAST) == 0) { 1588 rw_enter(&in_multilock, RW_READER); 1589 ingroup = in_lookup_multi(addr, ifp) != NULL; 1590 rw_exit(&in_multilock); 1591 } else { 1592 /* XXX Recursive call from ip_output(). */ 1593 KASSERT(rw_lock_held(&in_multilock)); 1594 ingroup = in_lookup_multi(addr, ifp) != NULL; 1595 } 1596 return ingroup; 1597 } 1598 1599 /* 1600 * Add an address to the list of IP multicast addresses for a given interface. 1601 */ 1602 struct in_multi * 1603 in_addmulti(struct in_addr *ap, ifnet_t *ifp) 1604 { 1605 struct sockaddr_in sin; 1606 struct in_multi *inm; 1607 1608 /* 1609 * See if address already in list. 1610 */ 1611 rw_enter(&in_multilock, RW_WRITER); 1612 inm = in_lookup_multi(*ap, ifp); 1613 if (inm != NULL) { 1614 /* 1615 * Found it; just increment the reference count. 1616 */ 1617 inm->inm_refcount++; 1618 rw_exit(&in_multilock); 1619 return inm; 1620 } 1621 1622 /* 1623 * New address; allocate a new multicast record. 1624 */ 1625 inm = pool_get(&inmulti_pool, PR_NOWAIT); 1626 if (inm == NULL) { 1627 rw_exit(&in_multilock); 1628 return NULL; 1629 } 1630 inm->inm_addr = *ap; 1631 inm->inm_ifp = ifp; 1632 inm->inm_refcount = 1; 1633 1634 /* 1635 * Ask the network driver to update its multicast reception 1636 * filter appropriately for the new address. 1637 */ 1638 sockaddr_in_init(&sin, ap, 0); 1639 if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) { 1640 rw_exit(&in_multilock); 1641 pool_put(&inmulti_pool, inm); 1642 return NULL; 1643 } 1644 1645 /* 1646 * Let IGMP know that we have joined a new IP multicast group. 1647 */ 1648 if (igmp_joingroup(inm) != 0) { 1649 rw_exit(&in_multilock); 1650 pool_put(&inmulti_pool, inm); 1651 return NULL; 1652 } 1653 LIST_INSERT_HEAD( 1654 &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp), 1655 inm, inm_list); 1656 in_multientries++; 1657 rw_exit(&in_multilock); 1658 1659 return inm; 1660 } 1661 1662 /* 1663 * Delete a multicast address record. 1664 */ 1665 void 1666 in_delmulti(struct in_multi *inm) 1667 { 1668 struct sockaddr_in sin; 1669 1670 rw_enter(&in_multilock, RW_WRITER); 1671 if (--inm->inm_refcount > 0) { 1672 rw_exit(&in_multilock); 1673 return; 1674 } 1675 1676 /* 1677 * No remaining claims to this record; let IGMP know that 1678 * we are leaving the multicast group. 1679 */ 1680 igmp_leavegroup(inm); 1681 1682 /* 1683 * Notify the network driver to update its multicast reception 1684 * filter. 1685 */ 1686 sockaddr_in_init(&sin, &inm->inm_addr, 0); 1687 if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin)); 1688 1689 /* 1690 * Unlink from list. 1691 */ 1692 LIST_REMOVE(inm, inm_list); 1693 in_multientries--; 1694 rw_exit(&in_multilock); 1695 1696 pool_put(&inmulti_pool, inm); 1697 } 1698 1699 /* 1700 * in_next_multi: step through all of the in_multi records, one at a time. 1701 * The current position is remembered in "step", which the caller must 1702 * provide. in_first_multi(), below, must be called to initialize "step" 1703 * and get the first record. Both macros return a NULL "inm" when there 1704 * are no remaining records. 1705 */ 1706 struct in_multi * 1707 in_next_multi(struct in_multistep *step) 1708 { 1709 struct in_multi *inm; 1710 1711 KASSERT(rw_lock_held(&in_multilock)); 1712 1713 while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) { 1714 step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]); 1715 } 1716 if ((inm = step->i_inm) != NULL) { 1717 step->i_inm = LIST_NEXT(inm, inm_list); 1718 } 1719 return inm; 1720 } 1721 1722 struct in_multi * 1723 in_first_multi(struct in_multistep *step) 1724 { 1725 KASSERT(rw_lock_held(&in_multilock)); 1726 1727 step->i_n = 0; 1728 step->i_inm = LIST_FIRST(&in_multihashtbl[0]); 1729 return in_next_multi(step); 1730 } 1731 1732 void 1733 in_multi_lock(int op) 1734 { 1735 rw_enter(&in_multilock, op); 1736 } 1737 1738 void 1739 in_multi_unlock(void) 1740 { 1741 rw_exit(&in_multilock); 1742 } 1743 1744 int 1745 in_multi_lock_held(void) 1746 { 1747 return rw_lock_held(&in_multilock); 1748 } 1749 1750 struct in_ifaddr * 1751 in_selectsrc(struct sockaddr_in *sin, struct route *ro, 1752 int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref) 1753 { 1754 struct rtentry *rt = NULL; 1755 struct in_ifaddr *ia = NULL; 1756 1757 KASSERT(ISSET(curlwp->l_pflag, LP_BOUND)); 1758 /* 1759 * If route is known or can be allocated now, take the 1760 * source address from the interface. Otherwise, punt. 1761 */ 1762 if ((soopts & SO_DONTROUTE) != 0) 1763 rtcache_free(ro); 1764 else { 1765 union { 1766 struct sockaddr dst; 1767 struct sockaddr_in dst4; 1768 } u; 1769 1770 sockaddr_in_init(&u.dst4, &sin->sin_addr, 0); 1771 rt = rtcache_lookup(ro, &u.dst); 1772 } 1773 /* 1774 * If we found a route, use the address 1775 * corresponding to the outgoing interface 1776 * unless it is the loopback (in case a route 1777 * to our address on another net goes to loopback). 1778 * 1779 * XXX Is this still true? Do we care? 1780 */ 1781 if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 1782 int s; 1783 struct ifaddr *ifa; 1784 /* 1785 * Just in case. May not need to do this workaround. 1786 * Revisit when working on rtentry MP-ification. 1787 */ 1788 s = pserialize_read_enter(); 1789 IFADDR_READER_FOREACH(ifa, rt->rt_ifp) { 1790 if (ifa == rt->rt_ifa) 1791 break; 1792 } 1793 if (ifa != NULL) 1794 ifa_acquire(ifa, psref); 1795 pserialize_read_exit(s); 1796 1797 ia = ifatoia(ifa); 1798 } 1799 if (ia == NULL) { 1800 u_int16_t fport = sin->sin_port; 1801 struct ifaddr *ifa; 1802 int s; 1803 1804 sin->sin_port = 0; 1805 ifa = ifa_ifwithladdr_psref(sintosa(sin), psref); 1806 sin->sin_port = fport; 1807 if (ifa == NULL) { 1808 /* Find 1st non-loopback AF_INET address */ 1809 s = pserialize_read_enter(); 1810 IN_ADDRLIST_READER_FOREACH(ia) { 1811 if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK)) 1812 break; 1813 } 1814 if (ia != NULL) 1815 ia4_acquire(ia, psref); 1816 pserialize_read_exit(s); 1817 } else { 1818 /* ia is already referenced by psref */ 1819 ia = ifatoia(ifa); 1820 } 1821 if (ia == NULL) { 1822 *errorp = EADDRNOTAVAIL; 1823 goto out; 1824 } 1825 } 1826 /* 1827 * If the destination address is multicast and an outgoing 1828 * interface has been set as a multicast option, use the 1829 * address of that interface as our source address. 1830 */ 1831 if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) { 1832 struct ip_moptions *imo; 1833 1834 imo = mopts; 1835 if (imo->imo_multicast_if_index != 0) { 1836 struct ifnet *ifp; 1837 int s; 1838 1839 if (ia != NULL) 1840 ia4_release(ia, psref); 1841 s = pserialize_read_enter(); 1842 ifp = if_byindex(imo->imo_multicast_if_index); 1843 if (ifp != NULL) { 1844 /* XXX */ 1845 ia = in_get_ia_from_ifp_psref(ifp, psref); 1846 } else 1847 ia = NULL; 1848 if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) { 1849 pserialize_read_exit(s); 1850 if (ia != NULL) 1851 ia4_release(ia, psref); 1852 *errorp = EADDRNOTAVAIL; 1853 ia = NULL; 1854 goto out; 1855 } 1856 pserialize_read_exit(s); 1857 } 1858 } 1859 if (ia->ia_ifa.ifa_getifa != NULL) { 1860 ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa, 1861 sintosa(sin))); 1862 if (ia == NULL) { 1863 *errorp = EADDRNOTAVAIL; 1864 goto out; 1865 } 1866 /* FIXME NOMPSAFE */ 1867 ia4_acquire(ia, psref); 1868 } 1869 #ifdef GETIFA_DEBUG 1870 else 1871 printf("%s: missing ifa_getifa\n", __func__); 1872 #endif 1873 out: 1874 rtcache_unref(rt, ro); 1875 return ia; 1876 } 1877 1878 int 1879 in_tunnel_validate(const struct ip *ip, struct in_addr src, struct in_addr dst) 1880 { 1881 struct in_ifaddr *ia4; 1882 int s; 1883 1884 /* check for address match */ 1885 if (src.s_addr != ip->ip_dst.s_addr || 1886 dst.s_addr != ip->ip_src.s_addr) 1887 return 0; 1888 1889 /* martian filters on outer source - NOT done in ip_input! */ 1890 if (IN_MULTICAST(ip->ip_src.s_addr)) 1891 return 0; 1892 switch ((ntohl(ip->ip_src.s_addr) & 0xff000000) >> 24) { 1893 case 0: 1894 case 127: 1895 case 255: 1896 return 0; 1897 } 1898 /* reject packets with broadcast on source */ 1899 s = pserialize_read_enter(); 1900 IN_ADDRLIST_READER_FOREACH(ia4) { 1901 if ((ia4->ia_ifa.ifa_ifp->if_flags & IFF_BROADCAST) == 0) 1902 continue; 1903 if (ip->ip_src.s_addr == ia4->ia_broadaddr.sin_addr.s_addr) { 1904 pserialize_read_exit(s); 1905 return 0; 1906 } 1907 } 1908 pserialize_read_exit(s); 1909 1910 /* NOTE: packet may dropped by uRPF */ 1911 1912 /* return valid bytes length */ 1913 return sizeof(src) + sizeof(dst); 1914 } 1915 1916 #if NARP > 0 1917 1918 struct in_llentry { 1919 struct llentry base; 1920 }; 1921 1922 #define IN_LLTBL_DEFAULT_HSIZE 32 1923 #define IN_LLTBL_HASH(k, h) \ 1924 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1925 1926 /* 1927 * Do actual deallocation of @lle. 1928 * Called by LLE_FREE_LOCKED when number of references 1929 * drops to zero. 1930 */ 1931 static void 1932 in_lltable_destroy_lle(struct llentry *lle) 1933 { 1934 1935 LLE_WUNLOCK(lle); 1936 LLE_LOCK_DESTROY(lle); 1937 kmem_intr_free(lle, sizeof(*lle)); 1938 } 1939 1940 static struct llentry * 1941 in_lltable_new(struct in_addr addr4, u_int flags) 1942 { 1943 struct in_llentry *lle; 1944 1945 lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP); 1946 if (lle == NULL) /* NB: caller generates msg */ 1947 return NULL; 1948 1949 /* 1950 * For IPv4 this will trigger "arpresolve" to generate 1951 * an ARP request. 1952 */ 1953 lle->base.la_expire = time_uptime; /* mark expired */ 1954 lle->base.r_l3addr.addr4 = addr4; 1955 lle->base.lle_refcnt = 1; 1956 lle->base.lle_free = in_lltable_destroy_lle; 1957 LLE_LOCK_INIT(&lle->base); 1958 callout_init(&lle->base.la_timer, CALLOUT_MPSAFE); 1959 1960 return (&lle->base); 1961 } 1962 1963 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1964 (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1965 1966 static int 1967 in_lltable_match_prefix(const struct sockaddr *prefix, 1968 const struct sockaddr *mask, u_int flags, struct llentry *lle) 1969 { 1970 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1971 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1972 struct in_addr lle_addr; 1973 1974 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1975 1976 /* 1977 * (flags & LLE_STATIC) means deleting all entries 1978 * including static ARP entries. 1979 */ 1980 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, pfx, msk) && 1981 ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) 1982 return (1); 1983 1984 return (0); 1985 } 1986 1987 static void 1988 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1989 { 1990 struct ifnet *ifp __diagused; 1991 size_t pkts_dropped; 1992 bool locked = false; 1993 1994 LLE_WLOCK_ASSERT(lle); 1995 KASSERT(llt != NULL); 1996 1997 /* Unlink entry from table if not already */ 1998 if ((lle->la_flags & LLE_LINKED) != 0) { 1999 ifp = llt->llt_ifp; 2000 IF_AFDATA_WLOCK_ASSERT(ifp); 2001 lltable_unlink_entry(llt, lle); 2002 locked = true; 2003 } 2004 2005 /* 2006 * We need to release the lock here to lle_timer proceeds; 2007 * lle_timer should stop immediately if LLE_LINKED isn't set. 2008 * Note that we cannot pass lle->lle_lock to callout_halt 2009 * because it's a rwlock. 2010 */ 2011 LLE_ADDREF(lle); 2012 LLE_WUNLOCK(lle); 2013 if (locked) 2014 IF_AFDATA_WUNLOCK(ifp); 2015 2016 /* cancel timer */ 2017 callout_halt(&lle->lle_timer, NULL); 2018 2019 LLE_WLOCK(lle); 2020 LLE_REMREF(lle); 2021 2022 /* Drop hold queue */ 2023 pkts_dropped = llentry_free(lle); 2024 arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped); 2025 2026 if (locked) 2027 IF_AFDATA_WLOCK(ifp); 2028 } 2029 2030 static int 2031 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr, 2032 const struct rtentry *rt) 2033 { 2034 int error = EINVAL; 2035 2036 if (rt == NULL) 2037 return error; 2038 2039 /* 2040 * If the gateway for an existing host route matches the target L3 2041 * address, which is a special route inserted by some implementation 2042 * such as MANET, and the interface is of the correct type, then 2043 * allow for ARP to proceed. 2044 */ 2045 if (rt->rt_flags & RTF_GATEWAY) { 2046 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 2047 rt->rt_ifp->if_type != IFT_ETHER || 2048 #ifdef __FreeBSD__ 2049 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 2050 #else 2051 (rt->rt_ifp->if_flags & IFF_NOARP) != 0 || 2052 #endif 2053 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 2054 sizeof(in_addr_t)) != 0) { 2055 goto error; 2056 } 2057 } 2058 2059 /* 2060 * Make sure that at least the destination address is covered 2061 * by the route. This is for handling the case where 2 or more 2062 * interfaces have the same prefix. An incoming packet arrives 2063 * on one interface and the corresponding outgoing packet leaves 2064 * another interface. 2065 */ 2066 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 2067 const char *sa, *mask, *addr, *lim; 2068 int len; 2069 2070 mask = (const char *)rt_mask(rt); 2071 /* 2072 * Just being extra cautious to avoid some custom 2073 * code getting into trouble. 2074 */ 2075 if (mask == NULL) 2076 goto error; 2077 2078 sa = (const char *)rt_getkey(rt); 2079 addr = (const char *)l3addr; 2080 len = ((const struct sockaddr_in *)l3addr)->sin_len; 2081 lim = addr + len; 2082 2083 for ( ; addr < lim; sa++, mask++, addr++) { 2084 if ((*sa ^ *addr) & *mask) { 2085 #ifdef DIAGNOSTIC 2086 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 2087 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 2088 #endif 2089 goto error; 2090 } 2091 } 2092 } 2093 2094 error = 0; 2095 error: 2096 return error; 2097 } 2098 2099 static inline uint32_t 2100 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 2101 { 2102 2103 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 2104 } 2105 2106 static uint32_t 2107 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 2108 { 2109 2110 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 2111 } 2112 2113 static void 2114 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 2115 { 2116 struct sockaddr_in *sin; 2117 2118 sin = (struct sockaddr_in *)sa; 2119 memset(sin, 0, sizeof(*sin)); 2120 sin->sin_family = AF_INET; 2121 sin->sin_len = sizeof(*sin); 2122 sin->sin_addr = lle->r_l3addr.addr4; 2123 } 2124 2125 static inline struct llentry * 2126 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 2127 { 2128 struct llentry *lle; 2129 struct llentries *lleh; 2130 u_int hashidx; 2131 2132 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 2133 lleh = &llt->lle_head[hashidx]; 2134 LIST_FOREACH(lle, lleh, lle_next) { 2135 if (lle->la_flags & LLE_DELETED) 2136 continue; 2137 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 2138 break; 2139 } 2140 2141 return (lle); 2142 } 2143 2144 static int 2145 in_lltable_delete(struct lltable *llt, u_int flags, 2146 const struct sockaddr *l3addr) 2147 { 2148 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2149 struct ifnet *ifp __diagused = llt->llt_ifp; 2150 struct llentry *lle; 2151 2152 IF_AFDATA_WLOCK_ASSERT(ifp); 2153 KASSERTMSG(l3addr->sa_family == AF_INET, 2154 "sin_family %d", l3addr->sa_family); 2155 2156 lle = in_lltable_find_dst(llt, sin->sin_addr); 2157 if (lle == NULL) { 2158 #ifdef LLTABLE_DEBUG 2159 char buf[64]; 2160 sockaddr_format(l3addr, buf, sizeof(buf)); 2161 log(LOG_INFO, "%s: cache for %s is not found\n", 2162 __func__, buf); 2163 #endif 2164 return (ENOENT); 2165 } 2166 2167 LLE_WLOCK(lle); 2168 lle->la_flags |= LLE_DELETED; 2169 #ifdef LLTABLE_DEBUG 2170 { 2171 char buf[64]; 2172 sockaddr_format(l3addr, buf, sizeof(buf)); 2173 log(LOG_INFO, "%s: cache for %s (%p) is deleted\n", 2174 __func__, buf, lle); 2175 } 2176 #endif 2177 if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 2178 llentry_free(lle); 2179 else 2180 LLE_WUNLOCK(lle); 2181 2182 return (0); 2183 } 2184 2185 static struct llentry * 2186 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr, 2187 const struct rtentry *rt) 2188 { 2189 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2190 struct ifnet *ifp = llt->llt_ifp; 2191 struct llentry *lle; 2192 2193 IF_AFDATA_WLOCK_ASSERT(ifp); 2194 KASSERTMSG(l3addr->sa_family == AF_INET, 2195 "sin_family %d", l3addr->sa_family); 2196 2197 lle = in_lltable_find_dst(llt, sin->sin_addr); 2198 2199 if (lle != NULL) { 2200 LLE_WLOCK(lle); 2201 return (lle); 2202 } 2203 2204 /* no existing record, we need to create new one */ 2205 2206 /* 2207 * A route that covers the given address must have 2208 * been installed 1st because we are doing a resolution, 2209 * verify this. 2210 */ 2211 if (!(flags & LLE_IFADDR) && 2212 in_lltable_rtcheck(ifp, flags, l3addr, rt) != 0) 2213 return (NULL); 2214 2215 lle = in_lltable_new(sin->sin_addr, flags); 2216 if (lle == NULL) { 2217 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 2218 return (NULL); 2219 } 2220 lle->la_flags = flags; 2221 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 2222 memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen); 2223 lle->la_flags |= (LLE_VALID | LLE_STATIC); 2224 } 2225 2226 lltable_link_entry(llt, lle); 2227 LLE_WLOCK(lle); 2228 2229 return (lle); 2230 } 2231 2232 /* 2233 * Return NULL if not found or marked for deletion. 2234 * If found return lle read locked. 2235 */ 2236 static struct llentry * 2237 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 2238 { 2239 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2240 struct llentry *lle; 2241 2242 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 2243 KASSERTMSG(l3addr->sa_family == AF_INET, 2244 "sin_family %d", l3addr->sa_family); 2245 2246 lle = in_lltable_find_dst(llt, sin->sin_addr); 2247 2248 if (lle == NULL) 2249 return NULL; 2250 2251 if (flags & LLE_EXCLUSIVE) 2252 LLE_WLOCK(lle); 2253 else 2254 LLE_RLOCK(lle); 2255 2256 return lle; 2257 } 2258 2259 static int 2260 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 2261 struct rt_walkarg *w) 2262 { 2263 struct sockaddr_in sin; 2264 2265 LLTABLE_LOCK_ASSERT(); 2266 2267 /* skip deleted entries */ 2268 if (lle->la_flags & LLE_DELETED) 2269 return 0; 2270 2271 sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0); 2272 2273 return lltable_dump_entry(llt, lle, w, sintosa(&sin)); 2274 } 2275 2276 #endif /* NARP > 0 */ 2277 2278 static int 2279 in_multicast_sysctl(SYSCTLFN_ARGS) 2280 { 2281 struct ifnet *ifp; 2282 struct ifaddr *ifa; 2283 struct in_ifaddr *ifa4; 2284 struct in_multi *inm; 2285 uint32_t tmp; 2286 int error; 2287 size_t written; 2288 struct psref psref; 2289 int bound; 2290 2291 if (namelen != 1) 2292 return EINVAL; 2293 2294 bound = curlwp_bind(); 2295 ifp = if_get_byindex(name[0], &psref); 2296 if (ifp == NULL) { 2297 curlwp_bindx(bound); 2298 return ENODEV; 2299 } 2300 2301 if (oldp == NULL) { 2302 *oldlenp = 0; 2303 IFADDR_FOREACH(ifa, ifp) { 2304 if (ifa->ifa_addr->sa_family != AF_INET) 2305 continue; 2306 ifa4 = (void *)ifa; 2307 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2308 *oldlenp += 2 * sizeof(struct in_addr) + 2309 sizeof(uint32_t); 2310 } 2311 } 2312 if_put(ifp, &psref); 2313 curlwp_bindx(bound); 2314 return 0; 2315 } 2316 2317 error = 0; 2318 written = 0; 2319 IFADDR_FOREACH(ifa, ifp) { 2320 if (ifa->ifa_addr->sa_family != AF_INET) 2321 continue; 2322 ifa4 = (void *)ifa; 2323 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2324 if (written + 2 * sizeof(struct in_addr) + 2325 sizeof(uint32_t) > *oldlenp) 2326 goto done; 2327 error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr, 2328 oldp, sizeof(struct in_addr)); 2329 if (error) 2330 goto done; 2331 oldp = (char *)oldp + sizeof(struct in_addr); 2332 written += sizeof(struct in_addr); 2333 error = sysctl_copyout(l, &inm->inm_addr, 2334 oldp, sizeof(struct in_addr)); 2335 if (error) 2336 goto done; 2337 oldp = (char *)oldp + sizeof(struct in_addr); 2338 written += sizeof(struct in_addr); 2339 tmp = inm->inm_refcount; 2340 error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp)); 2341 if (error) 2342 goto done; 2343 oldp = (char *)oldp + sizeof(tmp); 2344 written += sizeof(tmp); 2345 } 2346 } 2347 done: 2348 if_put(ifp, &psref); 2349 curlwp_bindx(bound); 2350 *oldlenp = written; 2351 return error; 2352 } 2353 2354 static void 2355 in_sysctl_init(struct sysctllog **clog) 2356 { 2357 sysctl_createv(clog, 0, NULL, NULL, 2358 CTLFLAG_PERMANENT, 2359 CTLTYPE_NODE, "inet", 2360 SYSCTL_DESCR("PF_INET related settings"), 2361 NULL, 0, NULL, 0, 2362 CTL_NET, PF_INET, CTL_EOL); 2363 sysctl_createv(clog, 0, NULL, NULL, 2364 CTLFLAG_PERMANENT, 2365 CTLTYPE_NODE, "multicast", 2366 SYSCTL_DESCR("Multicast information"), 2367 in_multicast_sysctl, 0, NULL, 0, 2368 CTL_NET, PF_INET, CTL_CREATE, CTL_EOL); 2369 sysctl_createv(clog, 0, NULL, NULL, 2370 CTLFLAG_PERMANENT, 2371 CTLTYPE_NODE, "ip", 2372 SYSCTL_DESCR("IPv4 related settings"), 2373 NULL, 0, NULL, 0, 2374 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2375 2376 sysctl_createv(clog, 0, NULL, NULL, 2377 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2378 CTLTYPE_INT, "subnetsarelocal", 2379 SYSCTL_DESCR("Whether logical subnets are considered " 2380 "local"), 2381 NULL, 0, &subnetsarelocal, 0, 2382 CTL_NET, PF_INET, IPPROTO_IP, 2383 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2384 sysctl_createv(clog, 0, NULL, NULL, 2385 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2386 CTLTYPE_INT, "hostzerobroadcast", 2387 SYSCTL_DESCR("All zeroes address is broadcast address"), 2388 NULL, 0, &hostzeroisbroadcast, 0, 2389 CTL_NET, PF_INET, IPPROTO_IP, 2390 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2391 } 2392 2393 #if NARP > 0 2394 2395 static struct lltable * 2396 in_lltattach(struct ifnet *ifp) 2397 { 2398 struct lltable *llt; 2399 2400 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 2401 llt->llt_af = AF_INET; 2402 llt->llt_ifp = ifp; 2403 2404 llt->llt_lookup = in_lltable_lookup; 2405 llt->llt_create = in_lltable_create; 2406 llt->llt_delete = in_lltable_delete; 2407 llt->llt_dump_entry = in_lltable_dump_entry; 2408 llt->llt_hash = in_lltable_hash; 2409 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 2410 llt->llt_free_entry = in_lltable_free_entry; 2411 llt->llt_match_prefix = in_lltable_match_prefix; 2412 lltable_link(llt); 2413 2414 return (llt); 2415 } 2416 2417 #endif /* NARP > 0 */ 2418 2419 void * 2420 in_domifattach(struct ifnet *ifp) 2421 { 2422 struct in_ifinfo *ii; 2423 2424 ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP); 2425 2426 #if NARP > 0 2427 ii->ii_llt = in_lltattach(ifp); 2428 #endif 2429 2430 #ifdef IPSELSRC 2431 ii->ii_selsrc = in_selsrc_domifattach(ifp); 2432 KASSERT(ii->ii_selsrc != NULL); 2433 #endif 2434 2435 return ii; 2436 } 2437 2438 void 2439 in_domifdetach(struct ifnet *ifp, void *aux) 2440 { 2441 struct in_ifinfo *ii = aux; 2442 2443 #ifdef IPSELSRC 2444 in_selsrc_domifdetach(ifp, ii->ii_selsrc); 2445 #endif 2446 #if NARP > 0 2447 lltable_free(ii->ii_llt); 2448 #endif 2449 kmem_free(ii, sizeof(struct in_ifinfo)); 2450 } 2451
Indexes created Sun Sep 21 20:09:37 GMT 2025