in.c revision 1.189
1 /* $NetBSD: in.c,v 1.189 2016/12/06 07:01:47 knakahara 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 1.189 2016/12/06 07:01:47 knakahara 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 * Determine whether an IP address is in a reserved set of addresses 262 * that may not be forwarded, or whether datagrams to that destination 263 * may be forwarded. 264 */ 265 int 266 in_canforward(struct in_addr in) 267 { 268 u_int32_t net; 269 270 if (IN_EXPERIMENTAL(in.s_addr) || IN_MULTICAST(in.s_addr)) 271 return (0); 272 if (IN_CLASSA(in.s_addr)) { 273 net = in.s_addr & IN_CLASSA_NET; 274 if (net == 0 || net == htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT)) 275 return (0); 276 } 277 return (1); 278 } 279 280 /* 281 * Trim a mask in a sockaddr 282 */ 283 void 284 in_socktrim(struct sockaddr_in *ap) 285 { 286 char *cplim = (char *) &ap->sin_addr; 287 char *cp = (char *) (&ap->sin_addr + 1); 288 289 ap->sin_len = 0; 290 while (--cp >= cplim) 291 if (*cp) { 292 (ap)->sin_len = cp - (char *) (ap) + 1; 293 break; 294 } 295 } 296 297 /* 298 * Routine to take an Internet address and convert into a 299 * "dotted quad" representation for printing. 300 */ 301 const char * 302 in_fmtaddr(struct in_addr addr) 303 { 304 static char buf[sizeof("123.456.789.123")]; 305 306 addr.s_addr = ntohl(addr.s_addr); 307 308 snprintf(buf, sizeof(buf), "%d.%d.%d.%d", 309 (addr.s_addr >> 24) & 0xFF, 310 (addr.s_addr >> 16) & 0xFF, 311 (addr.s_addr >> 8) & 0xFF, 312 (addr.s_addr >> 0) & 0xFF); 313 return buf; 314 } 315 316 /* 317 * Maintain the "in_maxmtu" variable, which is the largest 318 * mtu for non-local interfaces with AF_INET addresses assigned 319 * to them that are up. 320 */ 321 unsigned long in_maxmtu; 322 323 void 324 in_setmaxmtu(void) 325 { 326 struct in_ifaddr *ia; 327 struct ifnet *ifp; 328 unsigned long maxmtu = 0; 329 int s = pserialize_read_enter(); 330 331 IN_ADDRLIST_READER_FOREACH(ia) { 332 if ((ifp = ia->ia_ifp) == 0) 333 continue; 334 if ((ifp->if_flags & (IFF_UP|IFF_LOOPBACK)) != IFF_UP) 335 continue; 336 if (ifp->if_mtu > maxmtu) 337 maxmtu = ifp->if_mtu; 338 } 339 if (maxmtu) 340 in_maxmtu = maxmtu; 341 pserialize_read_exit(s); 342 } 343 344 static u_int 345 in_mask2len(struct in_addr *mask) 346 { 347 u_int x, y; 348 u_char *p; 349 350 p = (u_char *)mask; 351 for (x = 0; x < sizeof(*mask); x++) { 352 if (p[x] != 0xff) 353 break; 354 } 355 y = 0; 356 if (x < sizeof(*mask)) { 357 for (y = 0; y < NBBY; y++) { 358 if ((p[x] & (0x80 >> y)) == 0) 359 break; 360 } 361 } 362 return x * NBBY + y; 363 } 364 365 static void 366 in_len2mask(struct in_addr *mask, u_int len) 367 { 368 u_int i; 369 u_char *p; 370 371 p = (u_char *)mask; 372 memset(mask, 0, sizeof(*mask)); 373 for (i = 0; i < len / NBBY; i++) 374 p[i] = 0xff; 375 if (len % NBBY) 376 p[i] = (0xff00 >> (len % NBBY)) & 0xff; 377 } 378 379 /* 380 * Generic internet control operations (ioctl's). 381 * Ifp is 0 if not an interface-specific ioctl. 382 */ 383 /* ARGSUSED */ 384 static int 385 in_control0(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 386 { 387 struct ifreq *ifr = (struct ifreq *)data; 388 struct in_ifaddr *ia = NULL; 389 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 390 struct sockaddr_in oldaddr, *new_dstaddr; 391 int error, hostIsNew, maskIsNew; 392 int newifaddr = 0; 393 bool run_hook = false; 394 bool need_reinsert = false; 395 struct psref psref; 396 int bound; 397 398 switch (cmd) { 399 case SIOCALIFADDR: 400 case SIOCDLIFADDR: 401 case SIOCGLIFADDR: 402 if (ifp == NULL) 403 return EINVAL; 404 return in_lifaddr_ioctl(so, cmd, data, ifp); 405 case SIOCGIFADDRPREF: 406 case SIOCSIFADDRPREF: 407 if (ifp == NULL) 408 return EINVAL; 409 return ifaddrpref_ioctl(so, cmd, data, ifp); 410 } 411 412 bound = curlwp_bind(); 413 /* 414 * Find address for this interface, if it exists. 415 */ 416 if (ifp != NULL) 417 ia = in_get_ia_from_ifp_psref(ifp, &psref); 418 419 hostIsNew = 1; /* moved here to appease gcc */ 420 switch (cmd) { 421 case SIOCAIFADDR: 422 case SIOCDIFADDR: 423 case SIOCGIFALIAS: 424 case SIOCGIFAFLAG_IN: 425 if (ifra->ifra_addr.sin_family == AF_INET) { 426 int s; 427 428 if (ia != NULL) 429 ia4_release(ia, &psref); 430 s = pserialize_read_enter(); 431 IN_ADDRHASH_READER_FOREACH(ia, 432 ifra->ifra_addr.sin_addr.s_addr) { 433 if (ia->ia_ifp == ifp && 434 in_hosteq(ia->ia_addr.sin_addr, 435 ifra->ifra_addr.sin_addr)) 436 break; 437 } 438 if (ia != NULL) 439 ia4_acquire(ia, &psref); 440 pserialize_read_exit(s); 441 } 442 if ((cmd == SIOCDIFADDR || 443 cmd == SIOCGIFALIAS || 444 cmd == SIOCGIFAFLAG_IN) && 445 ia == NULL) { 446 error = EADDRNOTAVAIL; 447 goto out; 448 } 449 450 if (cmd == SIOCDIFADDR && 451 ifra->ifra_addr.sin_family == AF_UNSPEC) { 452 ifra->ifra_addr.sin_family = AF_INET; 453 } 454 /* FALLTHROUGH */ 455 case SIOCSIFADDR: 456 if (ia == NULL || ia->ia_addr.sin_family != AF_INET) 457 ; 458 else if (ifra->ifra_addr.sin_len == 0) { 459 ifra->ifra_addr = ia->ia_addr; 460 hostIsNew = 0; 461 } else if (in_hosteq(ia->ia_addr.sin_addr, 462 ifra->ifra_addr.sin_addr)) 463 hostIsNew = 0; 464 /* FALLTHROUGH */ 465 case SIOCSIFDSTADDR: 466 if (ifra->ifra_addr.sin_family != AF_INET) { 467 error = EAFNOSUPPORT; 468 goto out; 469 } 470 /* FALLTHROUGH */ 471 case SIOCSIFNETMASK: 472 if (ifp == NULL) 473 panic("in_control"); 474 475 if (cmd == SIOCGIFALIAS || cmd == SIOCGIFAFLAG_IN) 476 break; 477 478 if (ia == NULL && 479 (cmd == SIOCSIFNETMASK || cmd == SIOCSIFDSTADDR)) { 480 error = EADDRNOTAVAIL; 481 goto out; 482 } 483 484 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 485 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 486 NULL) != 0) { 487 error = EPERM; 488 goto out; 489 } 490 491 if (ia == NULL) { 492 ia = malloc(sizeof(*ia), M_IFADDR, M_WAITOK|M_ZERO); 493 if (ia == NULL) { 494 error = ENOBUFS; 495 goto out; 496 } 497 ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr); 498 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 499 ia->ia_ifa.ifa_netmask = sintosa(&ia->ia_sockmask); 500 #ifdef IPSELSRC 501 ia->ia_ifa.ifa_getifa = in_getifa; 502 #else /* IPSELSRC */ 503 ia->ia_ifa.ifa_getifa = NULL; 504 #endif /* IPSELSRC */ 505 ia->ia_sockmask.sin_len = 8; 506 ia->ia_sockmask.sin_family = AF_INET; 507 if (ifp->if_flags & IFF_BROADCAST) { 508 ia->ia_broadaddr.sin_len = sizeof(ia->ia_addr); 509 ia->ia_broadaddr.sin_family = AF_INET; 510 } 511 ia->ia_ifp = ifp; 512 ia->ia_idsalt = cprng_fast32() % 65535; 513 LIST_INIT(&ia->ia_multiaddrs); 514 IN_ADDRHASH_ENTRY_INIT(ia); 515 IN_ADDRLIST_ENTRY_INIT(ia); 516 ifa_psref_init(&ia->ia_ifa); 517 518 newifaddr = 1; 519 } 520 break; 521 522 case SIOCSIFBRDADDR: 523 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE, 524 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 525 NULL) != 0) { 526 error = EPERM; 527 goto out; 528 } 529 /* FALLTHROUGH */ 530 531 case SIOCGIFADDR: 532 case SIOCGIFNETMASK: 533 case SIOCGIFDSTADDR: 534 case SIOCGIFBRDADDR: 535 if (ia == NULL) { 536 error = EADDRNOTAVAIL; 537 goto out; 538 } 539 break; 540 } 541 error = 0; 542 switch (cmd) { 543 544 case SIOCGIFADDR: 545 ifreq_setaddr(cmd, ifr, sintocsa(&ia->ia_addr)); 546 break; 547 548 case SIOCGIFBRDADDR: 549 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 550 error = EINVAL; 551 goto out; 552 } 553 ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_broadaddr)); 554 break; 555 556 case SIOCGIFDSTADDR: 557 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 558 error = EINVAL; 559 goto out; 560 } 561 ifreq_setdstaddr(cmd, ifr, sintocsa(&ia->ia_dstaddr)); 562 break; 563 564 case SIOCGIFNETMASK: 565 /* 566 * We keep the number of trailing zero bytes the sin_len field 567 * of ia_sockmask, so we fix this before we pass it back to 568 * userland. 569 */ 570 oldaddr = ia->ia_sockmask; 571 oldaddr.sin_len = sizeof(struct sockaddr_in); 572 ifreq_setaddr(cmd, ifr, (const void *)&oldaddr); 573 break; 574 575 case SIOCSIFDSTADDR: 576 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 577 error = EINVAL; 578 goto out; 579 } 580 oldaddr = ia->ia_dstaddr; 581 ia->ia_dstaddr = *satocsin(ifreq_getdstaddr(cmd, ifr)); 582 if ((error = if_addr_init(ifp, &ia->ia_ifa, false)) != 0) { 583 ia->ia_dstaddr = oldaddr; 584 goto out; 585 } 586 if (ia->ia_flags & IFA_ROUTE) { 587 ia->ia_ifa.ifa_dstaddr = sintosa(&oldaddr); 588 rtinit(&ia->ia_ifa, RTM_DELETE, RTF_HOST); 589 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 590 rtinit(&ia->ia_ifa, RTM_ADD, RTF_HOST|RTF_UP); 591 } 592 break; 593 594 case SIOCSIFBRDADDR: 595 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 596 error = EINVAL; 597 goto out; 598 } 599 ia->ia_broadaddr = *satocsin(ifreq_getbroadaddr(cmd, ifr)); 600 break; 601 602 case SIOCSIFADDR: 603 if (!newifaddr) { 604 in_addrhash_remove(ia); 605 need_reinsert = true; 606 } 607 error = in_ifinit(ifp, ia, satocsin(ifreq_getaddr(cmd, ifr)), 608 NULL, 1); 609 610 run_hook = true; 611 break; 612 613 case SIOCSIFNETMASK: 614 in_scrubprefix(ia); 615 ia->ia_sockmask = *satocsin(ifreq_getaddr(cmd, ifr)); 616 ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr; 617 if (!newifaddr) { 618 in_addrhash_remove(ia); 619 need_reinsert = true; 620 } 621 error = in_ifinit(ifp, ia, NULL, NULL, 0); 622 break; 623 624 case SIOCAIFADDR: 625 maskIsNew = 0; 626 if (ifra->ifra_mask.sin_len) { 627 in_scrubprefix(ia); 628 ia->ia_sockmask = ifra->ifra_mask; 629 ia->ia_subnetmask = ia->ia_sockmask.sin_addr.s_addr; 630 maskIsNew = 1; 631 } 632 if ((ifp->if_flags & IFF_POINTOPOINT) && 633 (ifra->ifra_dstaddr.sin_family == AF_INET)) { 634 new_dstaddr = &ifra->ifra_dstaddr; 635 maskIsNew = 1; /* We lie; but the effect's the same */ 636 } else 637 new_dstaddr = NULL; 638 if (ifra->ifra_addr.sin_family == AF_INET && 639 (hostIsNew || maskIsNew)) { 640 if (!newifaddr) { 641 in_addrhash_remove(ia); 642 need_reinsert = true; 643 } 644 error = in_ifinit(ifp, ia, &ifra->ifra_addr, 645 new_dstaddr, 0); 646 } 647 if ((ifp->if_flags & IFF_BROADCAST) && 648 (ifra->ifra_broadaddr.sin_family == AF_INET)) 649 ia->ia_broadaddr = ifra->ifra_broadaddr; 650 run_hook = true; 651 break; 652 653 case SIOCGIFALIAS: 654 ifra->ifra_mask = ia->ia_sockmask; 655 if ((ifp->if_flags & IFF_POINTOPOINT) && 656 (ia->ia_dstaddr.sin_family == AF_INET)) 657 ifra->ifra_dstaddr = ia->ia_dstaddr; 658 else if ((ifp->if_flags & IFF_BROADCAST) && 659 (ia->ia_broadaddr.sin_family == AF_INET)) 660 ifra->ifra_broadaddr = ia->ia_broadaddr; 661 else 662 memset(&ifra->ifra_broadaddr, 0, 663 sizeof(ifra->ifra_broadaddr)); 664 break; 665 666 case SIOCGIFAFLAG_IN: 667 ifr->ifr_addrflags = ia->ia4_flags; 668 break; 669 670 case SIOCDIFADDR: 671 ia4_release(ia, &psref); 672 in_purgeaddr(&ia->ia_ifa); 673 ia = NULL; 674 run_hook = true; 675 break; 676 677 #ifdef MROUTING 678 case SIOCGETVIFCNT: 679 case SIOCGETSGCNT: 680 error = mrt_ioctl(so, cmd, data); 681 break; 682 #endif /* MROUTING */ 683 684 default: 685 error = ENOTTY; 686 goto out; 687 } 688 689 /* 690 * XXX insert regardless of error to make in_purgeaddr below work. 691 * Need to improve. 692 */ 693 if (newifaddr) { 694 ifaref(&ia->ia_ifa); 695 ifa_insert(ifp, &ia->ia_ifa); 696 697 mutex_enter(&in_ifaddr_lock); 698 TAILQ_INSERT_TAIL(&in_ifaddrhead, ia, ia_list); 699 IN_ADDRLIST_WRITER_INSERT_TAIL(ia); 700 in_addrhash_insert_locked(ia); 701 mutex_exit(&in_ifaddr_lock); 702 } else if (need_reinsert) { 703 in_addrhash_insert(ia); 704 } 705 706 if (error == 0) { 707 if (run_hook) 708 (void)pfil_run_hooks(if_pfil, 709 (struct mbuf **)cmd, ifp, PFIL_IFADDR); 710 } else if (newifaddr) { 711 KASSERT(ia != NULL); 712 in_purgeaddr(&ia->ia_ifa); 713 ia = NULL; 714 } 715 716 out: 717 if (!newifaddr && ia != NULL) 718 ia4_release(ia, &psref); 719 curlwp_bindx(bound); 720 return error; 721 } 722 723 int 724 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp) 725 { 726 int error; 727 728 #ifndef NET_MPSAFE 729 mutex_enter(softnet_lock); 730 #endif 731 error = in_control0(so, cmd, data, ifp); 732 #ifndef NET_MPSAFE 733 mutex_exit(softnet_lock); 734 #endif 735 736 return error; 737 } 738 739 /* Add ownaddr as loopback rtentry. */ 740 static void 741 in_ifaddlocal(struct ifaddr *ifa) 742 { 743 struct in_ifaddr *ia; 744 745 ia = (struct in_ifaddr *)ifa; 746 if (ia->ia_addr.sin_addr.s_addr == INADDR_ANY || 747 (ia->ia_ifp->if_flags & IFF_POINTOPOINT && 748 in_hosteq(ia->ia_dstaddr.sin_addr, ia->ia_addr.sin_addr))) 749 { 750 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 751 return; 752 } 753 754 rt_ifa_addlocal(ifa); 755 } 756 757 /* Remove loopback entry of ownaddr */ 758 static void 759 in_ifremlocal(struct ifaddr *ifa) 760 { 761 struct in_ifaddr *ia, *p; 762 struct ifaddr *alt_ifa = NULL; 763 int ia_count = 0; 764 int s; 765 struct psref psref; 766 int bound = curlwp_bind(); 767 768 ia = (struct in_ifaddr *)ifa; 769 /* Delete the entry if exactly one ifaddr matches the 770 * address, ifa->ifa_addr. */ 771 s = pserialize_read_enter(); 772 IN_ADDRLIST_READER_FOREACH(p) { 773 if (!in_hosteq(p->ia_addr.sin_addr, ia->ia_addr.sin_addr)) 774 continue; 775 if (p->ia_ifp != ia->ia_ifp) 776 alt_ifa = &p->ia_ifa; 777 if (++ia_count > 1 && alt_ifa != NULL) 778 break; 779 } 780 if (alt_ifa != NULL && ia_count > 1) 781 ifa_acquire(alt_ifa, &psref); 782 pserialize_read_exit(s); 783 784 if (ia_count == 0) 785 goto out; 786 787 rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa); 788 if (alt_ifa != NULL && ia_count > 1) 789 ifa_release(alt_ifa, &psref); 790 out: 791 curlwp_bindx(bound); 792 } 793 794 static void 795 in_scrubaddr(struct in_ifaddr *ia) 796 { 797 798 /* stop DAD processing */ 799 if (ia->ia_dad_stop != NULL) 800 ia->ia_dad_stop(&ia->ia_ifa); 801 802 in_scrubprefix(ia); 803 in_ifremlocal(&ia->ia_ifa); 804 805 if (ia->ia_allhosts != NULL) { 806 in_delmulti(ia->ia_allhosts); 807 ia->ia_allhosts = NULL; 808 } 809 } 810 811 /* 812 * Depends on it isn't called in concurrent. It should be guaranteed 813 * by ifa->ifa_ifp's ioctl lock. The possible callers are in_control 814 * and if_purgeaddrs; the former is called iva ifa->ifa_ifp's ioctl 815 * and the latter is called via ifa->ifa_ifp's if_detach. The functions 816 * never be executed in concurrent. 817 */ 818 void 819 in_purgeaddr(struct ifaddr *ifa) 820 { 821 struct in_ifaddr *ia = (void *) ifa; 822 struct ifnet *ifp = ifa->ifa_ifp; 823 824 KASSERT(!ifa_held(ifa)); 825 826 in_scrubaddr(ia); 827 828 mutex_enter(&in_ifaddr_lock); 829 in_addrhash_remove_locked(ia); 830 TAILQ_REMOVE(&in_ifaddrhead, ia, ia_list); 831 IN_ADDRLIST_WRITER_REMOVE(ia); 832 ifa_remove(ifp, &ia->ia_ifa); 833 mutex_exit(&in_ifaddr_lock); 834 835 #ifdef NET_MPSAFE 836 pserialize_perform(in_ifaddrhash_psz); 837 #endif 838 IN_ADDRHASH_ENTRY_DESTROY(ia); 839 IN_ADDRLIST_ENTRY_DESTROY(ia); 840 ifafree(&ia->ia_ifa); 841 in_setmaxmtu(); 842 } 843 844 static void 845 in_addrhash_insert_locked(struct in_ifaddr *ia) 846 { 847 848 KASSERT(mutex_owned(&in_ifaddr_lock)); 849 850 LIST_INSERT_HEAD(&IN_IFADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 851 ia_hash); 852 IN_ADDRHASH_ENTRY_INIT(ia); 853 IN_ADDRHASH_WRITER_INSERT_HEAD(ia); 854 } 855 856 void 857 in_addrhash_insert(struct in_ifaddr *ia) 858 { 859 860 mutex_enter(&in_ifaddr_lock); 861 in_addrhash_insert_locked(ia); 862 mutex_exit(&in_ifaddr_lock); 863 } 864 865 static void 866 in_addrhash_remove_locked(struct in_ifaddr *ia) 867 { 868 869 KASSERT(mutex_owned(&in_ifaddr_lock)); 870 871 LIST_REMOVE(ia, ia_hash); 872 IN_ADDRHASH_WRITER_REMOVE(ia); 873 } 874 875 void 876 in_addrhash_remove(struct in_ifaddr *ia) 877 { 878 879 mutex_enter(&in_ifaddr_lock); 880 in_addrhash_remove_locked(ia); 881 mutex_exit(&in_ifaddr_lock); 882 #ifdef NET_MPSAFE 883 pserialize_perform(in_ifaddrhash_psz); 884 #endif 885 IN_ADDRHASH_ENTRY_DESTROY(ia); 886 } 887 888 void 889 in_purgeif(struct ifnet *ifp) /* MUST be called at splsoftnet() */ 890 { 891 if_purgeaddrs(ifp, AF_INET, in_purgeaddr); 892 igmp_purgeif(ifp); /* manipulates pools */ 893 #ifdef MROUTING 894 ip_mrouter_detach(ifp); 895 #endif 896 } 897 898 /* 899 * SIOC[GAD]LIFADDR. 900 * SIOCGLIFADDR: get first address. (???) 901 * SIOCGLIFADDR with IFLR_PREFIX: 902 * get first address that matches the specified prefix. 903 * SIOCALIFADDR: add the specified address. 904 * SIOCALIFADDR with IFLR_PREFIX: 905 * EINVAL since we can't deduce hostid part of the address. 906 * SIOCDLIFADDR: delete the specified address. 907 * SIOCDLIFADDR with IFLR_PREFIX: 908 * delete the first address that matches the specified prefix. 909 * return values: 910 * EINVAL on invalid parameters 911 * EADDRNOTAVAIL on prefix match failed/specified address not found 912 * other values may be returned from in_ioctl() 913 */ 914 static int 915 in_lifaddr_ioctl(struct socket *so, u_long cmd, void *data, 916 struct ifnet *ifp) 917 { 918 struct if_laddrreq *iflr = (struct if_laddrreq *)data; 919 struct ifaddr *ifa; 920 struct sockaddr *sa; 921 922 /* sanity checks */ 923 if (data == NULL || ifp == NULL) { 924 panic("invalid argument to in_lifaddr_ioctl"); 925 /*NOTRECHED*/ 926 } 927 928 switch (cmd) { 929 case SIOCGLIFADDR: 930 /* address must be specified on GET with IFLR_PREFIX */ 931 if ((iflr->flags & IFLR_PREFIX) == 0) 932 break; 933 /*FALLTHROUGH*/ 934 case SIOCALIFADDR: 935 case SIOCDLIFADDR: 936 /* address must be specified on ADD and DELETE */ 937 sa = (struct sockaddr *)&iflr->addr; 938 if (sa->sa_family != AF_INET) 939 return EINVAL; 940 if (sa->sa_len != sizeof(struct sockaddr_in)) 941 return EINVAL; 942 /* XXX need improvement */ 943 sa = (struct sockaddr *)&iflr->dstaddr; 944 if (sa->sa_family != AF_UNSPEC && sa->sa_family != AF_INET) 945 return EINVAL; 946 if (sa->sa_len != 0 && sa->sa_len != sizeof(struct sockaddr_in)) 947 return EINVAL; 948 break; 949 default: /*shouldn't happen*/ 950 #if 0 951 panic("invalid cmd to in_lifaddr_ioctl"); 952 /*NOTREACHED*/ 953 #else 954 return EOPNOTSUPP; 955 #endif 956 } 957 if (sizeof(struct in_addr) * NBBY < iflr->prefixlen) 958 return EINVAL; 959 960 switch (cmd) { 961 case SIOCALIFADDR: 962 { 963 struct in_aliasreq ifra; 964 965 if (iflr->flags & IFLR_PREFIX) 966 return EINVAL; 967 968 /* copy args to in_aliasreq, perform ioctl(SIOCAIFADDR). */ 969 memset(&ifra, 0, sizeof(ifra)); 970 memcpy(ifra.ifra_name, iflr->iflr_name, 971 sizeof(ifra.ifra_name)); 972 973 memcpy(&ifra.ifra_addr, &iflr->addr, 974 ((struct sockaddr *)&iflr->addr)->sa_len); 975 976 if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /*XXX*/ 977 memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr, 978 ((struct sockaddr *)&iflr->dstaddr)->sa_len); 979 } 980 981 ifra.ifra_mask.sin_family = AF_INET; 982 ifra.ifra_mask.sin_len = sizeof(struct sockaddr_in); 983 in_len2mask(&ifra.ifra_mask.sin_addr, iflr->prefixlen); 984 985 return in_control(so, SIOCAIFADDR, &ifra, ifp); 986 } 987 case SIOCGLIFADDR: 988 case SIOCDLIFADDR: 989 { 990 struct in_ifaddr *ia; 991 struct in_addr mask, candidate, match; 992 struct sockaddr_in *sin; 993 int cmp, s; 994 995 memset(&mask, 0, sizeof(mask)); 996 memset(&match, 0, sizeof(match)); /* XXX gcc */ 997 if (iflr->flags & IFLR_PREFIX) { 998 /* lookup a prefix rather than address. */ 999 in_len2mask(&mask, iflr->prefixlen); 1000 1001 sin = (struct sockaddr_in *)&iflr->addr; 1002 match.s_addr = sin->sin_addr.s_addr; 1003 match.s_addr &= mask.s_addr; 1004 1005 /* if you set extra bits, that's wrong */ 1006 if (match.s_addr != sin->sin_addr.s_addr) 1007 return EINVAL; 1008 1009 cmp = 1; 1010 } else { 1011 if (cmd == SIOCGLIFADDR) { 1012 /* on getting an address, take the 1st match */ 1013 cmp = 0; /*XXX*/ 1014 } else { 1015 /* on deleting an address, do exact match */ 1016 in_len2mask(&mask, 32); 1017 sin = (struct sockaddr_in *)&iflr->addr; 1018 match.s_addr = sin->sin_addr.s_addr; 1019 1020 cmp = 1; 1021 } 1022 } 1023 1024 s = pserialize_read_enter(); 1025 IFADDR_READER_FOREACH(ifa, ifp) { 1026 if (ifa->ifa_addr->sa_family != AF_INET) 1027 continue; 1028 if (cmp == 0) 1029 break; 1030 candidate.s_addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr.s_addr; 1031 candidate.s_addr &= mask.s_addr; 1032 if (candidate.s_addr == match.s_addr) 1033 break; 1034 } 1035 if (ifa == NULL) { 1036 pserialize_read_exit(s); 1037 return EADDRNOTAVAIL; 1038 } 1039 ia = (struct in_ifaddr *)ifa; 1040 1041 if (cmd == SIOCGLIFADDR) { 1042 /* fill in the if_laddrreq structure */ 1043 memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin_len); 1044 1045 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1046 memcpy(&iflr->dstaddr, &ia->ia_dstaddr, 1047 ia->ia_dstaddr.sin_len); 1048 } else 1049 memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr)); 1050 1051 iflr->prefixlen = 1052 in_mask2len(&ia->ia_sockmask.sin_addr); 1053 1054 iflr->flags = 0; /*XXX*/ 1055 pserialize_read_exit(s); 1056 1057 return 0; 1058 } else { 1059 struct in_aliasreq ifra; 1060 1061 /* fill in_aliasreq and do ioctl(SIOCDIFADDR) */ 1062 memset(&ifra, 0, sizeof(ifra)); 1063 memcpy(ifra.ifra_name, iflr->iflr_name, 1064 sizeof(ifra.ifra_name)); 1065 1066 memcpy(&ifra.ifra_addr, &ia->ia_addr, 1067 ia->ia_addr.sin_len); 1068 if ((ifp->if_flags & IFF_POINTOPOINT) != 0) { 1069 memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr, 1070 ia->ia_dstaddr.sin_len); 1071 } 1072 memcpy(&ifra.ifra_dstaddr, &ia->ia_sockmask, 1073 ia->ia_sockmask.sin_len); 1074 pserialize_read_exit(s); 1075 1076 return in_control(so, SIOCDIFADDR, &ifra, ifp); 1077 } 1078 } 1079 } 1080 1081 return EOPNOTSUPP; /*just for safety*/ 1082 } 1083 1084 /* 1085 * Initialize an interface's internet address 1086 * and routing table entry. 1087 */ 1088 int 1089 in_ifinit(struct ifnet *ifp, struct in_ifaddr *ia, 1090 const struct sockaddr_in *sin, const struct sockaddr_in *dst, int scrub) 1091 { 1092 u_int32_t i; 1093 struct sockaddr_in oldaddr, olddst; 1094 int s, oldflags, flags = RTF_UP, error, hostIsNew; 1095 1096 if (sin == NULL) 1097 sin = &ia->ia_addr; 1098 if (dst == NULL) 1099 dst = &ia->ia_dstaddr; 1100 1101 /* 1102 * Set up new addresses. 1103 */ 1104 oldaddr = ia->ia_addr; 1105 olddst = ia->ia_dstaddr; 1106 oldflags = ia->ia4_flags; 1107 ia->ia_addr = *sin; 1108 ia->ia_dstaddr = *dst; 1109 hostIsNew = oldaddr.sin_family != AF_INET || 1110 !in_hosteq(ia->ia_addr.sin_addr, oldaddr.sin_addr); 1111 if (!scrub) 1112 scrub = oldaddr.sin_family != ia->ia_dstaddr.sin_family || 1113 !in_hosteq(ia->ia_dstaddr.sin_addr, olddst.sin_addr); 1114 1115 /* 1116 * Configure address flags. 1117 * We need to do this early because they maybe adjusted 1118 * by if_addr_init depending on the address. 1119 */ 1120 if (ia->ia4_flags & IN_IFF_DUPLICATED) { 1121 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1122 hostIsNew = 1; 1123 } 1124 if (ifp->if_link_state == LINK_STATE_DOWN) { 1125 ia->ia4_flags |= IN_IFF_DETACHED; 1126 ia->ia4_flags &= ~IN_IFF_TENTATIVE; 1127 } else if (hostIsNew && if_do_dad(ifp)) 1128 ia->ia4_flags |= IN_IFF_TRYTENTATIVE; 1129 1130 /* 1131 * Give the interface a chance to initialize 1132 * if this is its first address, 1133 * and to validate the address if necessary. 1134 */ 1135 s = splnet(); 1136 error = if_addr_init(ifp, &ia->ia_ifa, true); 1137 splx(s); 1138 /* Now clear the try tentative flag, it's job is done. */ 1139 ia->ia4_flags &= ~IN_IFF_TRYTENTATIVE; 1140 if (error != 0) { 1141 ia->ia_addr = oldaddr; 1142 ia->ia_dstaddr = olddst; 1143 ia->ia4_flags = oldflags; 1144 return error; 1145 } 1146 1147 if (scrub || hostIsNew) { 1148 int newflags = ia->ia4_flags; 1149 1150 ia->ia_ifa.ifa_addr = sintosa(&oldaddr); 1151 ia->ia_ifa.ifa_dstaddr = sintosa(&olddst); 1152 ia->ia4_flags = oldflags; 1153 if (hostIsNew) 1154 in_scrubaddr(ia); 1155 else if (scrub) 1156 in_scrubprefix(ia); 1157 ia->ia_ifa.ifa_addr = sintosa(&ia->ia_addr); 1158 ia->ia_ifa.ifa_dstaddr = sintosa(&ia->ia_dstaddr); 1159 ia->ia4_flags = newflags; 1160 } 1161 1162 /* Add the local route to the address */ 1163 in_ifaddlocal(&ia->ia_ifa); 1164 1165 i = ia->ia_addr.sin_addr.s_addr; 1166 if (ifp->if_flags & IFF_POINTOPOINT) 1167 ia->ia_netmask = INADDR_BROADCAST; /* default to /32 */ 1168 else if (IN_CLASSA(i)) 1169 ia->ia_netmask = IN_CLASSA_NET; 1170 else if (IN_CLASSB(i)) 1171 ia->ia_netmask = IN_CLASSB_NET; 1172 else 1173 ia->ia_netmask = IN_CLASSC_NET; 1174 /* 1175 * The subnet mask usually includes at least the standard network part, 1176 * but may may be smaller in the case of supernetting. 1177 * If it is set, we believe it. 1178 */ 1179 if (ia->ia_subnetmask == 0) { 1180 ia->ia_subnetmask = ia->ia_netmask; 1181 ia->ia_sockmask.sin_addr.s_addr = ia->ia_subnetmask; 1182 } else 1183 ia->ia_netmask &= ia->ia_subnetmask; 1184 1185 ia->ia_net = i & ia->ia_netmask; 1186 ia->ia_subnet = i & ia->ia_subnetmask; 1187 in_socktrim(&ia->ia_sockmask); 1188 /* re-calculate the "in_maxmtu" value */ 1189 in_setmaxmtu(); 1190 /* 1191 * Add route for the network. 1192 */ 1193 ia->ia_ifa.ifa_metric = ifp->if_metric; 1194 if (ifp->if_flags & IFF_BROADCAST) { 1195 ia->ia_broadaddr.sin_addr.s_addr = 1196 ia->ia_subnet | ~ia->ia_subnetmask; 1197 ia->ia_netbroadcast.s_addr = 1198 ia->ia_net | ~ia->ia_netmask; 1199 } else if (ifp->if_flags & IFF_LOOPBACK) { 1200 ia->ia_dstaddr = ia->ia_addr; 1201 flags |= RTF_HOST; 1202 } else if (ifp->if_flags & IFF_POINTOPOINT) { 1203 if (ia->ia_dstaddr.sin_family != AF_INET) 1204 return (0); 1205 flags |= RTF_HOST; 1206 } 1207 error = in_addprefix(ia, flags); 1208 /* 1209 * If the interface supports multicast, join the "all hosts" 1210 * multicast group on that interface. 1211 */ 1212 if ((ifp->if_flags & IFF_MULTICAST) != 0 && ia->ia_allhosts == NULL) { 1213 struct in_addr addr; 1214 1215 addr.s_addr = INADDR_ALLHOSTS_GROUP; 1216 ia->ia_allhosts = in_addmulti(&addr, ifp); 1217 } 1218 1219 if (hostIsNew && 1220 ia->ia4_flags & IN_IFF_TENTATIVE && 1221 if_do_dad(ifp)) 1222 ia->ia_dad_start((struct ifaddr *)ia); 1223 1224 return error; 1225 } 1226 1227 #define rtinitflags(x) \ 1228 ((((x)->ia_ifp->if_flags & (IFF_LOOPBACK | IFF_POINTOPOINT)) != 0) \ 1229 ? RTF_HOST : 0) 1230 1231 /* 1232 * add a route to prefix ("connected route" in cisco terminology). 1233 * does nothing if there's some interface address with the same prefix already. 1234 */ 1235 static int 1236 in_addprefix(struct in_ifaddr *target, int flags) 1237 { 1238 struct in_ifaddr *ia; 1239 struct in_addr prefix, mask, p; 1240 int error; 1241 int s; 1242 1243 if ((flags & RTF_HOST) != 0) 1244 prefix = target->ia_dstaddr.sin_addr; 1245 else { 1246 prefix = target->ia_addr.sin_addr; 1247 mask = target->ia_sockmask.sin_addr; 1248 prefix.s_addr &= mask.s_addr; 1249 } 1250 1251 s = pserialize_read_enter(); 1252 IN_ADDRLIST_READER_FOREACH(ia) { 1253 if (rtinitflags(ia)) 1254 p = ia->ia_dstaddr.sin_addr; 1255 else { 1256 p = ia->ia_addr.sin_addr; 1257 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1258 } 1259 1260 if (prefix.s_addr != p.s_addr) 1261 continue; 1262 1263 /* 1264 * if we got a matching prefix route inserted by other 1265 * interface address, we don't need to bother 1266 * 1267 * XXX RADIX_MPATH implications here? -dyoung 1268 */ 1269 if (ia->ia_flags & IFA_ROUTE) { 1270 pserialize_read_exit(s); 1271 return 0; 1272 } 1273 } 1274 pserialize_read_exit(s); 1275 1276 /* 1277 * noone seem to have prefix route. insert it. 1278 */ 1279 error = rtinit(&target->ia_ifa, RTM_ADD, flags); 1280 if (error == 0) 1281 target->ia_flags |= IFA_ROUTE; 1282 else if (error == EEXIST) { 1283 /* 1284 * the fact the route already exists is not an error. 1285 */ 1286 error = 0; 1287 } 1288 return error; 1289 } 1290 1291 /* 1292 * remove a route to prefix ("connected route" in cisco terminology). 1293 * re-installs the route by using another interface address, if there's one 1294 * with the same prefix (otherwise we lose the route mistakenly). 1295 */ 1296 static int 1297 in_scrubprefix(struct in_ifaddr *target) 1298 { 1299 struct in_ifaddr *ia; 1300 struct in_addr prefix, mask, p; 1301 int error; 1302 int s; 1303 1304 /* If we don't have IFA_ROUTE we have nothing to do */ 1305 if ((target->ia_flags & IFA_ROUTE) == 0) 1306 return 0; 1307 1308 if (rtinitflags(target)) 1309 prefix = target->ia_dstaddr.sin_addr; 1310 else { 1311 prefix = target->ia_addr.sin_addr; 1312 mask = target->ia_sockmask.sin_addr; 1313 prefix.s_addr &= mask.s_addr; 1314 } 1315 1316 s = pserialize_read_enter(); 1317 IN_ADDRLIST_READER_FOREACH(ia) { 1318 if (rtinitflags(ia)) 1319 p = ia->ia_dstaddr.sin_addr; 1320 else { 1321 p = ia->ia_addr.sin_addr; 1322 p.s_addr &= ia->ia_sockmask.sin_addr.s_addr; 1323 } 1324 1325 if (prefix.s_addr != p.s_addr) 1326 continue; 1327 1328 /* 1329 * if we got a matching prefix route, move IFA_ROUTE to him 1330 */ 1331 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1332 struct psref psref; 1333 int bound = curlwp_bind(); 1334 1335 ia4_acquire(ia, &psref); 1336 pserialize_read_exit(s); 1337 1338 rtinit(&target->ia_ifa, RTM_DELETE, 1339 rtinitflags(target)); 1340 target->ia_flags &= ~IFA_ROUTE; 1341 1342 error = rtinit(&ia->ia_ifa, RTM_ADD, 1343 rtinitflags(ia) | RTF_UP); 1344 if (error == 0) 1345 ia->ia_flags |= IFA_ROUTE; 1346 1347 ia4_release(ia, &psref); 1348 curlwp_bindx(bound); 1349 1350 return error; 1351 } 1352 } 1353 pserialize_read_exit(s); 1354 1355 /* 1356 * noone seem to have prefix route. remove it. 1357 */ 1358 rtinit(&target->ia_ifa, RTM_DELETE, rtinitflags(target)); 1359 target->ia_flags &= ~IFA_ROUTE; 1360 return 0; 1361 } 1362 1363 #undef rtinitflags 1364 1365 /* 1366 * Return 1 if the address might be a local broadcast address. 1367 */ 1368 int 1369 in_broadcast(struct in_addr in, struct ifnet *ifp) 1370 { 1371 struct ifaddr *ifa; 1372 int s; 1373 1374 KASSERT(ifp != NULL); 1375 1376 if (in.s_addr == INADDR_BROADCAST || 1377 in_nullhost(in)) 1378 return 1; 1379 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1380 return 0; 1381 /* 1382 * Look through the list of addresses for a match 1383 * with a broadcast address. 1384 */ 1385 #define ia (ifatoia(ifa)) 1386 s = pserialize_read_enter(); 1387 IFADDR_READER_FOREACH(ifa, ifp) { 1388 if (ifa->ifa_addr->sa_family == AF_INET && 1389 !in_hosteq(in, ia->ia_addr.sin_addr) && 1390 (in_hosteq(in, ia->ia_broadaddr.sin_addr) || 1391 in_hosteq(in, ia->ia_netbroadcast) || 1392 (hostzeroisbroadcast && 1393 /* 1394 * Check for old-style (host 0) broadcast. 1395 */ 1396 (in.s_addr == ia->ia_subnet || 1397 in.s_addr == ia->ia_net)))) { 1398 pserialize_read_exit(s); 1399 return 1; 1400 } 1401 } 1402 pserialize_read_exit(s); 1403 return (0); 1404 #undef ia 1405 } 1406 1407 /* 1408 * perform DAD when interface becomes IFF_UP. 1409 */ 1410 void 1411 in_if_link_up(struct ifnet *ifp) 1412 { 1413 struct ifaddr *ifa; 1414 struct in_ifaddr *ia; 1415 int s, bound; 1416 1417 /* Ensure it's sane to run DAD */ 1418 if (ifp->if_link_state == LINK_STATE_DOWN) 1419 return; 1420 if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING)) 1421 return; 1422 1423 bound = curlwp_bind(); 1424 s = pserialize_read_enter(); 1425 IFADDR_READER_FOREACH(ifa, ifp) { 1426 struct psref psref; 1427 1428 if (ifa->ifa_addr->sa_family != AF_INET) 1429 continue; 1430 ifa_acquire(ifa, &psref); 1431 pserialize_read_exit(s); 1432 1433 ia = (struct in_ifaddr *)ifa; 1434 1435 /* If detached then mark as tentative */ 1436 if (ia->ia4_flags & IN_IFF_DETACHED) { 1437 ia->ia4_flags &= ~IN_IFF_DETACHED; 1438 if (if_do_dad(ifp) && ia->ia_dad_start != NULL) 1439 ia->ia4_flags |= IN_IFF_TENTATIVE; 1440 else if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0) 1441 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1442 } 1443 1444 if (ia->ia4_flags & IN_IFF_TENTATIVE) { 1445 /* Clear the duplicated flag as we're starting DAD. */ 1446 ia->ia4_flags &= ~IN_IFF_DUPLICATED; 1447 ia->ia_dad_start(ifa); 1448 } 1449 1450 s = pserialize_read_enter(); 1451 ifa_release(ifa, &psref); 1452 } 1453 pserialize_read_exit(s); 1454 curlwp_bindx(bound); 1455 } 1456 1457 void 1458 in_if_up(struct ifnet *ifp) 1459 { 1460 1461 /* interface may not support link state, so bring it up also */ 1462 in_if_link_up(ifp); 1463 } 1464 1465 /* 1466 * Mark all addresses as detached. 1467 */ 1468 void 1469 in_if_link_down(struct ifnet *ifp) 1470 { 1471 struct ifaddr *ifa; 1472 struct in_ifaddr *ia; 1473 int s, bound; 1474 1475 bound = curlwp_bind(); 1476 s = pserialize_read_enter(); 1477 IFADDR_READER_FOREACH(ifa, ifp) { 1478 struct psref psref; 1479 1480 if (ifa->ifa_addr->sa_family != AF_INET) 1481 continue; 1482 ifa_acquire(ifa, &psref); 1483 pserialize_read_exit(s); 1484 1485 ia = (struct in_ifaddr *)ifa; 1486 1487 /* Stop DAD processing */ 1488 if (ia->ia_dad_stop != NULL) 1489 ia->ia_dad_stop(ifa); 1490 1491 /* 1492 * Mark the address as detached. 1493 */ 1494 if (!(ia->ia4_flags & IN_IFF_DETACHED)) { 1495 ia->ia4_flags |= IN_IFF_DETACHED; 1496 ia->ia4_flags &= 1497 ~(IN_IFF_TENTATIVE | IN_IFF_DUPLICATED); 1498 rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL); 1499 } 1500 1501 s = pserialize_read_enter(); 1502 ifa_release(ifa, &psref); 1503 } 1504 pserialize_read_exit(s); 1505 curlwp_bindx(bound); 1506 } 1507 1508 void 1509 in_if_down(struct ifnet *ifp) 1510 { 1511 1512 in_if_link_down(ifp); 1513 } 1514 1515 void 1516 in_if_link_state_change(struct ifnet *ifp, int link_state) 1517 { 1518 1519 switch (link_state) { 1520 case LINK_STATE_DOWN: 1521 in_if_link_down(ifp); 1522 break; 1523 case LINK_STATE_UP: 1524 in_if_link_up(ifp); 1525 break; 1526 } 1527 } 1528 1529 /* 1530 * in_lookup_multi: look up the in_multi record for a given IP 1531 * multicast address on a given interface. If no matching record is 1532 * found, return NULL. 1533 */ 1534 struct in_multi * 1535 in_lookup_multi(struct in_addr addr, ifnet_t *ifp) 1536 { 1537 struct in_multi *inm; 1538 1539 KASSERT(rw_lock_held(&in_multilock)); 1540 1541 LIST_FOREACH(inm, &IN_MULTI_HASH(addr.s_addr, ifp), inm_list) { 1542 if (in_hosteq(inm->inm_addr, addr) && inm->inm_ifp == ifp) 1543 break; 1544 } 1545 return inm; 1546 } 1547 1548 /* 1549 * in_multi_group: check whether the address belongs to an IP multicast 1550 * group we are joined on this interface. Returns true or false. 1551 */ 1552 bool 1553 in_multi_group(struct in_addr addr, ifnet_t *ifp, int flags) 1554 { 1555 bool ingroup; 1556 1557 if (__predict_true(flags & IP_IGMP_MCAST) == 0) { 1558 rw_enter(&in_multilock, RW_READER); 1559 ingroup = in_lookup_multi(addr, ifp) != NULL; 1560 rw_exit(&in_multilock); 1561 } else { 1562 /* XXX Recursive call from ip_output(). */ 1563 KASSERT(rw_lock_held(&in_multilock)); 1564 ingroup = in_lookup_multi(addr, ifp) != NULL; 1565 } 1566 return ingroup; 1567 } 1568 1569 /* 1570 * Add an address to the list of IP multicast addresses for a given interface. 1571 */ 1572 struct in_multi * 1573 in_addmulti(struct in_addr *ap, ifnet_t *ifp) 1574 { 1575 struct sockaddr_in sin; 1576 struct in_multi *inm; 1577 1578 /* 1579 * See if address already in list. 1580 */ 1581 rw_enter(&in_multilock, RW_WRITER); 1582 inm = in_lookup_multi(*ap, ifp); 1583 if (inm != NULL) { 1584 /* 1585 * Found it; just increment the reference count. 1586 */ 1587 inm->inm_refcount++; 1588 rw_exit(&in_multilock); 1589 return inm; 1590 } 1591 1592 /* 1593 * New address; allocate a new multicast record. 1594 */ 1595 inm = pool_get(&inmulti_pool, PR_NOWAIT); 1596 if (inm == NULL) { 1597 rw_exit(&in_multilock); 1598 return NULL; 1599 } 1600 inm->inm_addr = *ap; 1601 inm->inm_ifp = ifp; 1602 inm->inm_refcount = 1; 1603 1604 /* 1605 * Ask the network driver to update its multicast reception 1606 * filter appropriately for the new address. 1607 */ 1608 sockaddr_in_init(&sin, ap, 0); 1609 if (if_mcast_op(ifp, SIOCADDMULTI, sintosa(&sin)) != 0) { 1610 rw_exit(&in_multilock); 1611 pool_put(&inmulti_pool, inm); 1612 return NULL; 1613 } 1614 1615 /* 1616 * Let IGMP know that we have joined a new IP multicast group. 1617 */ 1618 if (igmp_joingroup(inm) != 0) { 1619 rw_exit(&in_multilock); 1620 pool_put(&inmulti_pool, inm); 1621 return NULL; 1622 } 1623 LIST_INSERT_HEAD( 1624 &IN_MULTI_HASH(inm->inm_addr.s_addr, ifp), 1625 inm, inm_list); 1626 in_multientries++; 1627 rw_exit(&in_multilock); 1628 1629 return inm; 1630 } 1631 1632 /* 1633 * Delete a multicast address record. 1634 */ 1635 void 1636 in_delmulti(struct in_multi *inm) 1637 { 1638 struct sockaddr_in sin; 1639 1640 rw_enter(&in_multilock, RW_WRITER); 1641 if (--inm->inm_refcount > 0) { 1642 rw_exit(&in_multilock); 1643 return; 1644 } 1645 1646 /* 1647 * No remaining claims to this record; let IGMP know that 1648 * we are leaving the multicast group. 1649 */ 1650 igmp_leavegroup(inm); 1651 1652 /* 1653 * Notify the network driver to update its multicast reception 1654 * filter. 1655 */ 1656 sockaddr_in_init(&sin, &inm->inm_addr, 0); 1657 if_mcast_op(inm->inm_ifp, SIOCDELMULTI, sintosa(&sin)); 1658 1659 /* 1660 * Unlink from list. 1661 */ 1662 LIST_REMOVE(inm, inm_list); 1663 in_multientries--; 1664 rw_exit(&in_multilock); 1665 1666 pool_put(&inmulti_pool, inm); 1667 } 1668 1669 /* 1670 * in_next_multi: step through all of the in_multi records, one at a time. 1671 * The current position is remembered in "step", which the caller must 1672 * provide. in_first_multi(), below, must be called to initialize "step" 1673 * and get the first record. Both macros return a NULL "inm" when there 1674 * are no remaining records. 1675 */ 1676 struct in_multi * 1677 in_next_multi(struct in_multistep *step) 1678 { 1679 struct in_multi *inm; 1680 1681 KASSERT(rw_lock_held(&in_multilock)); 1682 1683 while (step->i_inm == NULL && step->i_n < IN_MULTI_HASH_SIZE) { 1684 step->i_inm = LIST_FIRST(&in_multihashtbl[++step->i_n]); 1685 } 1686 if ((inm = step->i_inm) != NULL) { 1687 step->i_inm = LIST_NEXT(inm, inm_list); 1688 } 1689 return inm; 1690 } 1691 1692 struct in_multi * 1693 in_first_multi(struct in_multistep *step) 1694 { 1695 KASSERT(rw_lock_held(&in_multilock)); 1696 1697 step->i_n = 0; 1698 step->i_inm = LIST_FIRST(&in_multihashtbl[0]); 1699 return in_next_multi(step); 1700 } 1701 1702 void 1703 in_multi_lock(int op) 1704 { 1705 rw_enter(&in_multilock, op); 1706 } 1707 1708 void 1709 in_multi_unlock(void) 1710 { 1711 rw_exit(&in_multilock); 1712 } 1713 1714 int 1715 in_multi_lock_held(void) 1716 { 1717 return rw_lock_held(&in_multilock); 1718 } 1719 1720 struct in_ifaddr * 1721 in_selectsrc(struct sockaddr_in *sin, struct route *ro, 1722 int soopts, struct ip_moptions *mopts, int *errorp, struct psref *psref) 1723 { 1724 struct rtentry *rt = NULL; 1725 struct in_ifaddr *ia = NULL; 1726 1727 KASSERT(ISSET(curlwp->l_pflag, LP_BOUND)); 1728 /* 1729 * If route is known or can be allocated now, take the 1730 * source address from the interface. Otherwise, punt. 1731 */ 1732 if ((soopts & SO_DONTROUTE) != 0) 1733 rtcache_free(ro); 1734 else { 1735 union { 1736 struct sockaddr dst; 1737 struct sockaddr_in dst4; 1738 } u; 1739 1740 sockaddr_in_init(&u.dst4, &sin->sin_addr, 0); 1741 rt = rtcache_lookup(ro, &u.dst); 1742 } 1743 /* 1744 * If we found a route, use the address 1745 * corresponding to the outgoing interface 1746 * unless it is the loopback (in case a route 1747 * to our address on another net goes to loopback). 1748 * 1749 * XXX Is this still true? Do we care? 1750 */ 1751 if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 1752 int s; 1753 struct ifaddr *ifa; 1754 /* 1755 * Just in case. May not need to do this workaround. 1756 * Revisit when working on rtentry MP-ification. 1757 */ 1758 s = pserialize_read_enter(); 1759 IFADDR_READER_FOREACH(ifa, rt->rt_ifp) { 1760 if (ifa == rt->rt_ifa) 1761 break; 1762 } 1763 if (ifa != NULL) 1764 ifa_acquire(ifa, psref); 1765 pserialize_read_exit(s); 1766 1767 ia = ifatoia(ifa); 1768 } 1769 if (ia == NULL) { 1770 u_int16_t fport = sin->sin_port; 1771 struct ifaddr *ifa; 1772 int s; 1773 1774 sin->sin_port = 0; 1775 ifa = ifa_ifwithladdr_psref(sintosa(sin), psref); 1776 sin->sin_port = fport; 1777 if (ifa == NULL) { 1778 /* Find 1st non-loopback AF_INET address */ 1779 s = pserialize_read_enter(); 1780 IN_ADDRLIST_READER_FOREACH(ia) { 1781 if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK)) 1782 break; 1783 } 1784 if (ia != NULL) 1785 ia4_acquire(ia, psref); 1786 pserialize_read_exit(s); 1787 } else { 1788 /* ia is already referenced by psref */ 1789 ia = ifatoia(ifa); 1790 } 1791 if (ia == NULL) { 1792 *errorp = EADDRNOTAVAIL; 1793 return NULL; 1794 } 1795 } 1796 /* 1797 * If the destination address is multicast and an outgoing 1798 * interface has been set as a multicast option, use the 1799 * address of that interface as our source address. 1800 */ 1801 if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) { 1802 struct ip_moptions *imo; 1803 1804 imo = mopts; 1805 if (imo->imo_multicast_if_index != 0) { 1806 struct ifnet *ifp; 1807 int s; 1808 1809 if (ia != NULL) 1810 ia4_release(ia, psref); 1811 s = pserialize_read_enter(); 1812 ifp = if_byindex(imo->imo_multicast_if_index); 1813 if (ifp != NULL) { 1814 /* XXX */ 1815 ia = in_get_ia_from_ifp_psref(ifp, psref); 1816 } else 1817 ia = NULL; 1818 if (ia == NULL || ia->ia4_flags & IN_IFF_NOTREADY) { 1819 pserialize_read_exit(s); 1820 if (ia != NULL) 1821 ia4_release(ia, psref); 1822 *errorp = EADDRNOTAVAIL; 1823 return NULL; 1824 } 1825 pserialize_read_exit(s); 1826 } 1827 } 1828 if (ia->ia_ifa.ifa_getifa != NULL) { 1829 ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa, 1830 sintosa(sin))); 1831 if (ia == NULL) { 1832 *errorp = EADDRNOTAVAIL; 1833 return NULL; 1834 } 1835 /* FIXME NOMPSAFE */ 1836 ia4_acquire(ia, psref); 1837 } 1838 #ifdef GETIFA_DEBUG 1839 else 1840 printf("%s: missing ifa_getifa\n", __func__); 1841 #endif 1842 return ia; 1843 } 1844 1845 #if NARP > 0 1846 1847 struct in_llentry { 1848 struct llentry base; 1849 }; 1850 1851 #define IN_LLTBL_DEFAULT_HSIZE 32 1852 #define IN_LLTBL_HASH(k, h) \ 1853 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1854 1855 /* 1856 * Do actual deallocation of @lle. 1857 * Called by LLE_FREE_LOCKED when number of references 1858 * drops to zero. 1859 */ 1860 static void 1861 in_lltable_destroy_lle(struct llentry *lle) 1862 { 1863 1864 LLE_WUNLOCK(lle); 1865 LLE_LOCK_DESTROY(lle); 1866 kmem_intr_free(lle, sizeof(*lle)); 1867 } 1868 1869 static struct llentry * 1870 in_lltable_new(struct in_addr addr4, u_int flags) 1871 { 1872 struct in_llentry *lle; 1873 1874 lle = kmem_intr_zalloc(sizeof(*lle), KM_NOSLEEP); 1875 if (lle == NULL) /* NB: caller generates msg */ 1876 return NULL; 1877 1878 /* 1879 * For IPv4 this will trigger "arpresolve" to generate 1880 * an ARP request. 1881 */ 1882 lle->base.la_expire = time_uptime; /* mark expired */ 1883 lle->base.r_l3addr.addr4 = addr4; 1884 lle->base.lle_refcnt = 1; 1885 lle->base.lle_free = in_lltable_destroy_lle; 1886 LLE_LOCK_INIT(&lle->base); 1887 callout_init(&lle->base.la_timer, CALLOUT_MPSAFE); 1888 1889 return (&lle->base); 1890 } 1891 1892 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1893 (((ntohl((d).s_addr) ^ (a)->sin_addr.s_addr) & (m)->sin_addr.s_addr)) == 0 ) 1894 1895 static int 1896 in_lltable_match_prefix(const struct sockaddr *prefix, 1897 const struct sockaddr *mask, u_int flags, struct llentry *lle) 1898 { 1899 const struct sockaddr_in *pfx = (const struct sockaddr_in *)prefix; 1900 const struct sockaddr_in *msk = (const struct sockaddr_in *)mask; 1901 1902 /* 1903 * (flags & LLE_STATIC) means deleting all entries 1904 * including static ARP entries. 1905 */ 1906 if (IN_ARE_MASKED_ADDR_EQUAL(lle->r_l3addr.addr4, pfx, msk) && 1907 ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) 1908 return (1); 1909 1910 return (0); 1911 } 1912 1913 static void 1914 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1915 { 1916 struct ifnet *ifp __diagused; 1917 size_t pkts_dropped; 1918 1919 LLE_WLOCK_ASSERT(lle); 1920 KASSERT(llt != NULL); 1921 1922 /* Unlink entry from table if not already */ 1923 if ((lle->la_flags & LLE_LINKED) != 0) { 1924 ifp = llt->llt_ifp; 1925 IF_AFDATA_WLOCK_ASSERT(ifp); 1926 lltable_unlink_entry(llt, lle); 1927 } 1928 1929 /* cancel timer */ 1930 if (callout_halt(&lle->lle_timer, &lle->lle_lock)) 1931 LLE_REMREF(lle); 1932 1933 /* Drop hold queue */ 1934 pkts_dropped = llentry_free(lle); 1935 arp_stat_add(ARP_STAT_DFRDROPPED, (uint64_t)pkts_dropped); 1936 } 1937 1938 static int 1939 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1940 { 1941 struct rtentry *rt; 1942 int error = EINVAL; 1943 1944 KASSERTMSG(l3addr->sa_family == AF_INET, 1945 "sin_family %d", l3addr->sa_family); 1946 1947 rt = rtalloc1(l3addr, 0); 1948 if (rt == NULL) 1949 return error; 1950 1951 /* 1952 * If the gateway for an existing host route matches the target L3 1953 * address, which is a special route inserted by some implementation 1954 * such as MANET, and the interface is of the correct type, then 1955 * allow for ARP to proceed. 1956 */ 1957 if (rt->rt_flags & RTF_GATEWAY) { 1958 if (!(rt->rt_flags & RTF_HOST) || !rt->rt_ifp || 1959 rt->rt_ifp->if_type != IFT_ETHER || 1960 #ifdef __FreeBSD__ 1961 (rt->rt_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1962 #else 1963 (rt->rt_ifp->if_flags & IFF_NOARP) != 0 || 1964 #endif 1965 memcmp(rt->rt_gateway->sa_data, l3addr->sa_data, 1966 sizeof(in_addr_t)) != 0) { 1967 goto error; 1968 } 1969 } 1970 1971 /* 1972 * Make sure that at least the destination address is covered 1973 * by the route. This is for handling the case where 2 or more 1974 * interfaces have the same prefix. An incoming packet arrives 1975 * on one interface and the corresponding outgoing packet leaves 1976 * another interface. 1977 */ 1978 if (!(rt->rt_flags & RTF_HOST) && rt->rt_ifp != ifp) { 1979 const char *sa, *mask, *addr, *lim; 1980 int len; 1981 1982 mask = (const char *)rt_mask(rt); 1983 /* 1984 * Just being extra cautious to avoid some custom 1985 * code getting into trouble. 1986 */ 1987 if (mask == NULL) 1988 goto error; 1989 1990 sa = (const char *)rt_getkey(rt); 1991 addr = (const char *)l3addr; 1992 len = ((const struct sockaddr_in *)l3addr)->sin_len; 1993 lim = addr + len; 1994 1995 for ( ; addr < lim; sa++, mask++, addr++) { 1996 if ((*sa ^ *addr) & *mask) { 1997 #ifdef DIAGNOSTIC 1998 log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", 1999 inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); 2000 #endif 2001 goto error; 2002 } 2003 } 2004 } 2005 2006 error = 0; 2007 error: 2008 rtfree(rt); 2009 return error; 2010 } 2011 2012 static inline uint32_t 2013 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 2014 { 2015 2016 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 2017 } 2018 2019 static uint32_t 2020 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 2021 { 2022 2023 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 2024 } 2025 2026 static void 2027 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 2028 { 2029 struct sockaddr_in *sin; 2030 2031 sin = (struct sockaddr_in *)sa; 2032 memset(sin, 0, sizeof(*sin)); 2033 sin->sin_family = AF_INET; 2034 sin->sin_len = sizeof(*sin); 2035 sin->sin_addr = lle->r_l3addr.addr4; 2036 } 2037 2038 static inline struct llentry * 2039 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 2040 { 2041 struct llentry *lle; 2042 struct llentries *lleh; 2043 u_int hashidx; 2044 2045 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 2046 lleh = &llt->lle_head[hashidx]; 2047 LIST_FOREACH(lle, lleh, lle_next) { 2048 if (lle->la_flags & LLE_DELETED) 2049 continue; 2050 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 2051 break; 2052 } 2053 2054 return (lle); 2055 } 2056 2057 static int 2058 in_lltable_delete(struct lltable *llt, u_int flags, 2059 const struct sockaddr *l3addr) 2060 { 2061 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2062 struct ifnet *ifp __diagused = llt->llt_ifp; 2063 struct llentry *lle; 2064 2065 IF_AFDATA_WLOCK_ASSERT(ifp); 2066 KASSERTMSG(l3addr->sa_family == AF_INET, 2067 "sin_family %d", l3addr->sa_family); 2068 2069 lle = in_lltable_find_dst(llt, sin->sin_addr); 2070 if (lle == NULL) { 2071 #ifdef DIAGNOSTIC 2072 log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); 2073 #endif 2074 return (ENOENT); 2075 } 2076 2077 LLE_WLOCK(lle); 2078 lle->la_flags |= LLE_DELETED; 2079 #ifdef DIAGNOSTIC 2080 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 2081 #endif 2082 if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC) 2083 llentry_free(lle); 2084 else 2085 LLE_WUNLOCK(lle); 2086 2087 return (0); 2088 } 2089 2090 static struct llentry * 2091 in_lltable_create(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 2092 { 2093 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2094 struct ifnet *ifp = llt->llt_ifp; 2095 struct llentry *lle; 2096 2097 IF_AFDATA_WLOCK_ASSERT(ifp); 2098 KASSERTMSG(l3addr->sa_family == AF_INET, 2099 "sin_family %d", l3addr->sa_family); 2100 2101 lle = in_lltable_find_dst(llt, sin->sin_addr); 2102 2103 if (lle != NULL) { 2104 LLE_WLOCK(lle); 2105 return (lle); 2106 } 2107 2108 /* no existing record, we need to create new one */ 2109 2110 /* 2111 * A route that covers the given address must have 2112 * been installed 1st because we are doing a resolution, 2113 * verify this. 2114 */ 2115 if (!(flags & LLE_IFADDR) && 2116 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 2117 return (NULL); 2118 2119 lle = in_lltable_new(sin->sin_addr, flags); 2120 if (lle == NULL) { 2121 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 2122 return (NULL); 2123 } 2124 lle->la_flags = flags; 2125 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 2126 memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen); 2127 lle->la_flags |= (LLE_VALID | LLE_STATIC); 2128 } 2129 2130 lltable_link_entry(llt, lle); 2131 LLE_WLOCK(lle); 2132 2133 return (lle); 2134 } 2135 2136 /* 2137 * Return NULL if not found or marked for deletion. 2138 * If found return lle read locked. 2139 */ 2140 static struct llentry * 2141 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 2142 { 2143 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 2144 struct llentry *lle; 2145 2146 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 2147 KASSERTMSG(l3addr->sa_family == AF_INET, 2148 "sin_family %d", l3addr->sa_family); 2149 2150 lle = in_lltable_find_dst(llt, sin->sin_addr); 2151 2152 if (lle == NULL) 2153 return NULL; 2154 2155 if (flags & LLE_EXCLUSIVE) 2156 LLE_WLOCK(lle); 2157 else 2158 LLE_RLOCK(lle); 2159 2160 return lle; 2161 } 2162 2163 static int 2164 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 2165 struct rt_walkarg *w) 2166 { 2167 struct sockaddr_in sin; 2168 2169 LLTABLE_LOCK_ASSERT(); 2170 2171 /* skip deleted entries */ 2172 if (lle->la_flags & LLE_DELETED) 2173 return 0; 2174 2175 sockaddr_in_init(&sin, &lle->r_l3addr.addr4, 0); 2176 2177 return lltable_dump_entry(llt, lle, w, sintosa(&sin)); 2178 } 2179 2180 #endif /* NARP > 0 */ 2181 2182 static int 2183 in_multicast_sysctl(SYSCTLFN_ARGS) 2184 { 2185 struct ifnet *ifp; 2186 struct ifaddr *ifa; 2187 struct in_ifaddr *ifa4; 2188 struct in_multi *inm; 2189 uint32_t tmp; 2190 int error; 2191 size_t written; 2192 struct psref psref; 2193 int bound; 2194 2195 if (namelen != 1) 2196 return EINVAL; 2197 2198 bound = curlwp_bind(); 2199 ifp = if_get_byindex(name[0], &psref); 2200 if (ifp == NULL) { 2201 curlwp_bindx(bound); 2202 return ENODEV; 2203 } 2204 2205 if (oldp == NULL) { 2206 *oldlenp = 0; 2207 IFADDR_FOREACH(ifa, ifp) { 2208 if (ifa->ifa_addr->sa_family != AF_INET) 2209 continue; 2210 ifa4 = (void *)ifa; 2211 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2212 *oldlenp += 2 * sizeof(struct in_addr) + 2213 sizeof(uint32_t); 2214 } 2215 } 2216 if_put(ifp, &psref); 2217 curlwp_bindx(bound); 2218 return 0; 2219 } 2220 2221 error = 0; 2222 written = 0; 2223 IFADDR_FOREACH(ifa, ifp) { 2224 if (ifa->ifa_addr->sa_family != AF_INET) 2225 continue; 2226 ifa4 = (void *)ifa; 2227 LIST_FOREACH(inm, &ifa4->ia_multiaddrs, inm_list) { 2228 if (written + 2 * sizeof(struct in_addr) + 2229 sizeof(uint32_t) > *oldlenp) 2230 goto done; 2231 error = sysctl_copyout(l, &ifa4->ia_addr.sin_addr, 2232 oldp, sizeof(struct in_addr)); 2233 if (error) 2234 goto done; 2235 oldp = (char *)oldp + sizeof(struct in_addr); 2236 written += sizeof(struct in_addr); 2237 error = sysctl_copyout(l, &inm->inm_addr, 2238 oldp, sizeof(struct in_addr)); 2239 if (error) 2240 goto done; 2241 oldp = (char *)oldp + sizeof(struct in_addr); 2242 written += sizeof(struct in_addr); 2243 tmp = inm->inm_refcount; 2244 error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp)); 2245 if (error) 2246 goto done; 2247 oldp = (char *)oldp + sizeof(tmp); 2248 written += sizeof(tmp); 2249 } 2250 } 2251 done: 2252 if_put(ifp, &psref); 2253 curlwp_bindx(bound); 2254 *oldlenp = written; 2255 return error; 2256 } 2257 2258 static void 2259 in_sysctl_init(struct sysctllog **clog) 2260 { 2261 sysctl_createv(clog, 0, NULL, NULL, 2262 CTLFLAG_PERMANENT, 2263 CTLTYPE_NODE, "inet", 2264 SYSCTL_DESCR("PF_INET related settings"), 2265 NULL, 0, NULL, 0, 2266 CTL_NET, PF_INET, CTL_EOL); 2267 sysctl_createv(clog, 0, NULL, NULL, 2268 CTLFLAG_PERMANENT, 2269 CTLTYPE_NODE, "multicast", 2270 SYSCTL_DESCR("Multicast information"), 2271 in_multicast_sysctl, 0, NULL, 0, 2272 CTL_NET, PF_INET, CTL_CREATE, CTL_EOL); 2273 sysctl_createv(clog, 0, NULL, NULL, 2274 CTLFLAG_PERMANENT, 2275 CTLTYPE_NODE, "ip", 2276 SYSCTL_DESCR("IPv4 related settings"), 2277 NULL, 0, NULL, 0, 2278 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL); 2279 2280 sysctl_createv(clog, 0, NULL, NULL, 2281 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2282 CTLTYPE_INT, "subnetsarelocal", 2283 SYSCTL_DESCR("Whether logical subnets are considered " 2284 "local"), 2285 NULL, 0, &subnetsarelocal, 0, 2286 CTL_NET, PF_INET, IPPROTO_IP, 2287 IPCTL_SUBNETSARELOCAL, CTL_EOL); 2288 sysctl_createv(clog, 0, NULL, NULL, 2289 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2290 CTLTYPE_INT, "hostzerobroadcast", 2291 SYSCTL_DESCR("All zeroes address is broadcast address"), 2292 NULL, 0, &hostzeroisbroadcast, 0, 2293 CTL_NET, PF_INET, IPPROTO_IP, 2294 IPCTL_HOSTZEROBROADCAST, CTL_EOL); 2295 } 2296 2297 #if NARP > 0 2298 2299 static struct lltable * 2300 in_lltattach(struct ifnet *ifp) 2301 { 2302 struct lltable *llt; 2303 2304 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 2305 llt->llt_af = AF_INET; 2306 llt->llt_ifp = ifp; 2307 2308 llt->llt_lookup = in_lltable_lookup; 2309 llt->llt_create = in_lltable_create; 2310 llt->llt_delete = in_lltable_delete; 2311 llt->llt_dump_entry = in_lltable_dump_entry; 2312 llt->llt_hash = in_lltable_hash; 2313 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 2314 llt->llt_free_entry = in_lltable_free_entry; 2315 llt->llt_match_prefix = in_lltable_match_prefix; 2316 lltable_link(llt); 2317 2318 return (llt); 2319 } 2320 2321 #endif /* NARP > 0 */ 2322 2323 void * 2324 in_domifattach(struct ifnet *ifp) 2325 { 2326 struct in_ifinfo *ii; 2327 2328 ii = kmem_zalloc(sizeof(struct in_ifinfo), KM_SLEEP); 2329 KASSERT(ii != NULL); 2330 2331 #if NARP > 0 2332 ii->ii_llt = in_lltattach(ifp); 2333 #endif 2334 2335 #ifdef IPSELSRC 2336 ii->ii_selsrc = in_selsrc_domifattach(ifp); 2337 KASSERT(ii->ii_selsrc != NULL); 2338 #endif 2339 2340 return ii; 2341 } 2342 2343 void 2344 in_domifdetach(struct ifnet *ifp, void *aux) 2345 { 2346 struct in_ifinfo *ii = aux; 2347 2348 #ifdef IPSELSRC 2349 in_selsrc_domifdetach(ifp, ii->ii_selsrc); 2350 #endif 2351 #if NARP > 0 2352 lltable_free(ii->ii_llt); 2353 #endif 2354 kmem_free(ii, sizeof(struct in_ifinfo)); 2355 } 2356
Indexes created Sun Sep 21 20:09:37 GMT 2025