1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * Copyright (C) 2001 WIDE Project. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)in.c 8.4 (Berkeley) 1/9/95 33 */ 34 35 #include <sys/cdefs.h> 36 __FBSDID("$FreeBSD$"); 37 38 #include "opt_inet.h" 39 40 #define IN_HISTORICAL_NETS /* include class masks */ 41 42 #include <sys/param.h> 43 #include <sys/eventhandler.h> 44 #include <sys/systm.h> 45 #include <sys/sockio.h> 46 #include <sys/malloc.h> 47 #include <sys/priv.h> 48 #include <sys/socket.h> 49 #include <sys/jail.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/proc.h> 53 #include <sys/sysctl.h> 54 #include <sys/syslog.h> 55 #include <sys/sx.h> 56 57 #include <net/if.h> 58 #include <net/if_var.h> 59 #include <net/if_arp.h> 60 #include <net/if_dl.h> 61 #include <net/if_llatbl.h> 62 #include <net/if_private.h> 63 #include <net/if_types.h> 64 #include <net/route.h> 65 #include <net/route/nhop.h> 66 #include <net/route/route_ctl.h> 67 #include <net/vnet.h> 68 69 #include <netinet/if_ether.h> 70 #include <netinet/in.h> 71 #include <netinet/in_fib.h> 72 #include <netinet/in_var.h> 73 #include <netinet/in_pcb.h> 74 #include <netinet/ip_var.h> 75 #include <netinet/ip_carp.h> 76 #include <netinet/igmp_var.h> 77 #include <netinet/udp.h> 78 #include <netinet/udp_var.h> 79 80 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 81 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 82 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 83 84 static void in_socktrim(struct sockaddr_in *); 85 static void in_purgemaddrs(struct ifnet *); 86 87 static bool ia_need_loopback_route(const struct in_ifaddr *); 88 89 VNET_DEFINE_STATIC(int, nosameprefix); 90 #define V_nosameprefix VNET(nosameprefix) 91 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW, 92 &VNET_NAME(nosameprefix), 0, 93 "Refuse to create same prefixes on different interfaces"); 94 95 VNET_DEFINE_STATIC(bool, broadcast_lowest); 96 #define V_broadcast_lowest VNET(broadcast_lowest) 97 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW, 98 &VNET_NAME(broadcast_lowest), 0, 99 "Treat lowest address on a subnet (host 0) as broadcast"); 100 101 VNET_DEFINE(bool, ip_allow_net240) = false; 102 #define V_ip_allow_net240 VNET(ip_allow_net240) 103 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240, 104 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0, 105 "Allow use of Experimental addresses, aka Class E (240/4)"); 106 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */ 107 108 VNET_DEFINE(bool, ip_allow_net0) = false; 109 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0, 110 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0, 111 "Allow use of addresses in network 0/8"); 112 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */ 113 114 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT; 115 #define V_in_loopback_mask VNET(in_loopback_mask) 116 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS); 117 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen, 118 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 119 NULL, 0, sysctl_loopback_prefixlen, "I", 120 "Prefix length of address space reserved for loopback"); 121 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */ 122 123 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 124 #define V_ripcbinfo VNET(ripcbinfo) 125 126 static struct sx in_control_sx; 127 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control"); 128 129 /* 130 * Return 1 if an internet address is for a ``local'' host 131 * (one to which we have a connection). 132 */ 133 int 134 in_localaddr(struct in_addr in) 135 { 136 u_long i = ntohl(in.s_addr); 137 struct in_ifaddr *ia; 138 139 NET_EPOCH_ASSERT(); 140 141 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 142 if ((i & ia->ia_subnetmask) == ia->ia_subnet) 143 return (1); 144 } 145 146 return (0); 147 } 148 149 /* 150 * Return 1 if an internet address is for the local host and configured 151 * on one of its interfaces. 152 */ 153 bool 154 in_localip(struct in_addr in) 155 { 156 struct in_ifaddr *ia; 157 158 NET_EPOCH_ASSERT(); 159 160 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) 161 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) 162 return (true); 163 164 return (false); 165 } 166 167 /* 168 * Like in_localip(), but FIB-aware. 169 */ 170 bool 171 in_localip_fib(struct in_addr in, uint16_t fib) 172 { 173 struct in_ifaddr *ia; 174 175 NET_EPOCH_ASSERT(); 176 177 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) 178 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr && 179 ia->ia_ifa.ifa_ifp->if_fib == fib) 180 return (true); 181 182 return (false); 183 } 184 185 /* 186 * Return 1 if an internet address is configured on an interface. 187 */ 188 int 189 in_ifhasaddr(struct ifnet *ifp, struct in_addr in) 190 { 191 struct ifaddr *ifa; 192 struct in_ifaddr *ia; 193 194 NET_EPOCH_ASSERT(); 195 196 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 197 if (ifa->ifa_addr->sa_family != AF_INET) 198 continue; 199 ia = (struct in_ifaddr *)ifa; 200 if (ia->ia_addr.sin_addr.s_addr == in.s_addr) 201 return (1); 202 } 203 204 return (0); 205 } 206 207 /* 208 * Return a reference to the interface address which is different to 209 * the supplied one but with same IP address value. 210 */ 211 static struct in_ifaddr * 212 in_localip_more(struct in_ifaddr *original_ia) 213 { 214 struct epoch_tracker et; 215 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr; 216 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib; 217 struct in_ifaddr *ia; 218 219 NET_EPOCH_ENTER(et); 220 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) { 221 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr; 222 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib; 223 if (!V_rt_add_addr_allfibs && (original_fib != fib)) 224 continue; 225 if ((original_ia != ia) && (original_addr == addr)) { 226 ifa_ref(&ia->ia_ifa); 227 NET_EPOCH_EXIT(et); 228 return (ia); 229 } 230 } 231 NET_EPOCH_EXIT(et); 232 233 return (NULL); 234 } 235 236 /* 237 * Tries to find first IPv4 address in the provided fib. 238 * Prefers non-loopback addresses and return loopback IFF 239 * @loopback_ok is set. 240 * 241 * Returns ifa or NULL. 242 */ 243 struct in_ifaddr * 244 in_findlocal(uint32_t fibnum, bool loopback_ok) 245 { 246 struct in_ifaddr *ia = NULL, *ia_lo = NULL; 247 248 NET_EPOCH_ASSERT(); 249 250 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 251 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib; 252 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib)) 253 continue; 254 255 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr))) 256 break; 257 if (loopback_ok) 258 ia_lo = ia; 259 } 260 261 if (ia == NULL) 262 ia = ia_lo; 263 264 return (ia); 265 } 266 267 /* 268 * Determine whether an IP address is in a reserved set of addresses 269 * that may not be forwarded, or whether datagrams to that destination 270 * may be forwarded. 271 */ 272 int 273 in_canforward(struct in_addr in) 274 { 275 u_long i = ntohl(in.s_addr); 276 277 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i)) 278 return (0); 279 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240) 280 return (0); 281 if (IN_ZERONET(i) && !V_ip_allow_net0) 282 return (0); 283 return (1); 284 } 285 286 /* 287 * Sysctl to manage prefix of reserved loopback network; translate 288 * to/from mask. The mask is always contiguous high-order 1 bits 289 * followed by all 0 bits. 290 */ 291 static int 292 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS) 293 { 294 int error, preflen; 295 296 /* ffs is 1-based; compensate. */ 297 preflen = 33 - ffs(V_in_loopback_mask); 298 error = sysctl_handle_int(oidp, &preflen, 0, req); 299 if (error || !req->newptr) 300 return (error); 301 if (preflen < 8 || preflen > 31) 302 return (EINVAL); 303 V_in_loopback_mask = 0xffffffff << (32 - preflen); 304 return (0); 305 } 306 307 /* 308 * Trim a mask in a sockaddr 309 */ 310 static void 311 in_socktrim(struct sockaddr_in *ap) 312 { 313 char *cplim = (char *) &ap->sin_addr; 314 char *cp = (char *) (&ap->sin_addr + 1); 315 316 ap->sin_len = 0; 317 while (--cp >= cplim) 318 if (*cp) { 319 (ap)->sin_len = cp - (char *) (ap) + 1; 320 break; 321 } 322 } 323 324 /* 325 * Generic internet control operations (ioctl's). 326 */ 327 int 328 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp, 329 struct ucred *cred) 330 { 331 struct ifreq *ifr = (struct ifreq *)data; 332 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 333 struct epoch_tracker et; 334 struct ifaddr *ifa; 335 struct in_ifaddr *ia; 336 int error; 337 338 if (ifp == NULL) 339 return (EADDRNOTAVAIL); 340 341 /* 342 * Filter out 4 ioctls we implement directly. Forward the rest 343 * to specific functions and ifp->if_ioctl(). 344 */ 345 switch (cmd) { 346 case SIOCGIFADDR: 347 case SIOCGIFBRDADDR: 348 case SIOCGIFDSTADDR: 349 case SIOCGIFNETMASK: 350 break; 351 case SIOCGIFALIAS: 352 sx_xlock(&in_control_sx); 353 error = in_gifaddr_ioctl(cmd, data, ifp, cred); 354 sx_xunlock(&in_control_sx); 355 return (error); 356 case SIOCDIFADDR: 357 sx_xlock(&in_control_sx); 358 error = in_difaddr_ioctl(cmd, data, ifp, cred); 359 sx_xunlock(&in_control_sx); 360 return (error); 361 case OSIOCAIFADDR: /* 9.x compat */ 362 case SIOCAIFADDR: 363 sx_xlock(&in_control_sx); 364 error = in_aifaddr_ioctl(cmd, data, ifp, cred); 365 sx_xunlock(&in_control_sx); 366 return (error); 367 case SIOCSIFADDR: 368 case SIOCSIFBRDADDR: 369 case SIOCSIFDSTADDR: 370 case SIOCSIFNETMASK: 371 /* We no longer support that old commands. */ 372 return (EINVAL); 373 default: 374 if (ifp->if_ioctl == NULL) 375 return (EOPNOTSUPP); 376 return ((*ifp->if_ioctl)(ifp, cmd, data)); 377 } 378 379 if (addr->sin_addr.s_addr != INADDR_ANY && 380 prison_check_ip4(cred, &addr->sin_addr) != 0) 381 return (EADDRNOTAVAIL); 382 383 /* 384 * Find address for this interface, if it exists. If an 385 * address was specified, find that one instead of the 386 * first one on the interface, if possible. 387 */ 388 NET_EPOCH_ENTER(et); 389 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 390 if (ifa->ifa_addr->sa_family != AF_INET) 391 continue; 392 ia = (struct in_ifaddr *)ifa; 393 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr) 394 break; 395 } 396 if (ifa == NULL) 397 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 398 if (ifa->ifa_addr->sa_family == AF_INET) { 399 ia = (struct in_ifaddr *)ifa; 400 if (prison_check_ip4(cred, 401 &ia->ia_addr.sin_addr) == 0) 402 break; 403 } 404 405 if (ifa == NULL) { 406 NET_EPOCH_EXIT(et); 407 return (EADDRNOTAVAIL); 408 } 409 410 error = 0; 411 switch (cmd) { 412 case SIOCGIFADDR: 413 *addr = ia->ia_addr; 414 break; 415 416 case SIOCGIFBRDADDR: 417 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 418 error = EINVAL; 419 break; 420 } 421 *addr = ia->ia_broadaddr; 422 break; 423 424 case SIOCGIFDSTADDR: 425 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 426 error = EINVAL; 427 break; 428 } 429 *addr = ia->ia_dstaddr; 430 break; 431 432 case SIOCGIFNETMASK: 433 *addr = ia->ia_sockmask; 434 break; 435 } 436 437 NET_EPOCH_EXIT(et); 438 439 return (error); 440 } 441 442 int 443 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp, 444 struct thread *td) 445 { 446 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL)); 447 } 448 449 static int 450 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 451 { 452 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 453 const struct sockaddr_in *addr = &ifra->ifra_addr; 454 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 455 const struct sockaddr_in *mask = &ifra->ifra_mask; 456 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 457 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 458 struct epoch_tracker et; 459 struct ifaddr *ifa; 460 struct in_ifaddr *ia; 461 bool iaIsFirst; 462 int error = 0; 463 464 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR); 465 if (error) 466 return (error); 467 468 /* 469 * ifra_addr must be present and be of INET family. 470 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 471 */ 472 if (addr->sin_len != sizeof(struct sockaddr_in) || 473 addr->sin_family != AF_INET) 474 return (EINVAL); 475 if (broadaddr->sin_len != 0 && 476 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 477 broadaddr->sin_family != AF_INET)) 478 return (EINVAL); 479 if (mask->sin_len != 0 && 480 (mask->sin_len != sizeof(struct sockaddr_in) || 481 mask->sin_family != AF_INET)) 482 return (EINVAL); 483 if ((ifp->if_flags & IFF_POINTOPOINT) && 484 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 485 dstaddr->sin_addr.s_addr == INADDR_ANY)) 486 return (EDESTADDRREQ); 487 if (vhid != 0 && carp_attach_p == NULL) 488 return (EPROTONOSUPPORT); 489 490 /* 491 * See whether address already exist. 492 */ 493 iaIsFirst = true; 494 ia = NULL; 495 NET_EPOCH_ENTER(et); 496 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 497 struct in_ifaddr *it; 498 499 if (ifa->ifa_addr->sa_family != AF_INET) 500 continue; 501 502 it = (struct in_ifaddr *)ifa; 503 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 504 prison_check_ip4(cred, &addr->sin_addr) == 0) 505 ia = it; 506 else 507 iaIsFirst = false; 508 } 509 NET_EPOCH_EXIT(et); 510 511 if (ia != NULL) 512 (void )in_difaddr_ioctl(cmd, data, ifp, cred); 513 514 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 515 ia = (struct in_ifaddr *)ifa; 516 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 517 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 518 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 519 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock, 520 CALLOUT_RETURNUNLOCKED); 521 522 ia->ia_ifp = ifp; 523 ia->ia_addr = *addr; 524 if (mask->sin_len != 0) { 525 ia->ia_sockmask = *mask; 526 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 527 } else { 528 in_addr_t i = ntohl(addr->sin_addr.s_addr); 529 530 /* 531 * If netmask isn't supplied, use historical default. 532 * This is deprecated for interfaces other than loopback 533 * or point-to-point; warn in other cases. In the future 534 * we should return an error rather than warning. 535 */ 536 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) 537 printf("%s: set address: WARNING: network mask " 538 "should be specified; using historical default\n", 539 ifp->if_xname); 540 if (IN_CLASSA(i)) 541 ia->ia_subnetmask = IN_CLASSA_NET; 542 else if (IN_CLASSB(i)) 543 ia->ia_subnetmask = IN_CLASSB_NET; 544 else 545 ia->ia_subnetmask = IN_CLASSC_NET; 546 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 547 } 548 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 549 in_socktrim(&ia->ia_sockmask); 550 551 if (ifp->if_flags & IFF_BROADCAST) { 552 if (broadaddr->sin_len != 0) { 553 ia->ia_broadaddr = *broadaddr; 554 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 555 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 556 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 557 ia->ia_broadaddr.sin_family = AF_INET; 558 } else { 559 ia->ia_broadaddr.sin_addr.s_addr = 560 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 561 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 562 ia->ia_broadaddr.sin_family = AF_INET; 563 } 564 } 565 566 if (ifp->if_flags & IFF_POINTOPOINT) 567 ia->ia_dstaddr = *dstaddr; 568 569 if (vhid != 0) { 570 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 571 if (error) 572 return (error); 573 } 574 575 /* if_addrhead is already referenced by ifa_alloc() */ 576 IF_ADDR_WLOCK(ifp); 577 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 578 IF_ADDR_WUNLOCK(ifp); 579 580 ifa_ref(ifa); /* in_ifaddrhead */ 581 sx_assert(&in_control_sx, SA_XLOCKED); 582 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 583 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 584 ia_hash); 585 586 /* 587 * Give the interface a chance to initialize 588 * if this is its first address, 589 * and to validate the address if necessary. 590 */ 591 if (ifp->if_ioctl != NULL) { 592 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 593 if (error) 594 goto fail1; 595 } 596 597 /* 598 * Add route for the network. 599 */ 600 if (vhid == 0) { 601 error = in_addprefix(ia); 602 if (error) 603 goto fail1; 604 } 605 606 /* 607 * Add a loopback route to self. 608 */ 609 if (vhid == 0 && ia_need_loopback_route(ia)) { 610 struct in_ifaddr *eia; 611 612 eia = in_localip_more(ia); 613 614 if (eia == NULL) { 615 error = ifa_add_loopback_route((struct ifaddr *)ia, 616 (struct sockaddr *)&ia->ia_addr); 617 if (error) 618 goto fail2; 619 } else 620 ifa_free(&eia->ia_ifa); 621 } 622 623 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 624 struct in_addr allhosts_addr; 625 struct in_ifinfo *ii; 626 627 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 628 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 629 630 error = in_joingroup(ifp, &allhosts_addr, NULL, 631 &ii->ii_allhosts); 632 } 633 634 /* 635 * Note: we don't need extra reference for ifa, since we called 636 * with sx lock held, and ifaddr can not be deleted in concurrent 637 * thread. 638 */ 639 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD); 640 641 return (error); 642 643 fail2: 644 if (vhid == 0) 645 (void )in_scrubprefix(ia, LLE_STATIC); 646 647 fail1: 648 if (ia->ia_ifa.ifa_carp) 649 (*carp_detach_p)(&ia->ia_ifa, false); 650 651 IF_ADDR_WLOCK(ifp); 652 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 653 IF_ADDR_WUNLOCK(ifp); 654 ifa_free(&ia->ia_ifa); /* if_addrhead */ 655 656 sx_assert(&in_control_sx, SA_XLOCKED); 657 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 658 CK_LIST_REMOVE(ia, ia_hash); 659 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 660 661 return (error); 662 } 663 664 static int 665 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 666 { 667 const struct ifreq *ifr = (struct ifreq *)data; 668 const struct sockaddr_in *addr = (const struct sockaddr_in *) 669 &ifr->ifr_addr; 670 struct ifaddr *ifa; 671 struct in_ifaddr *ia; 672 bool deleteAny, iaIsLast; 673 int error; 674 675 if (cred != NULL) { 676 error = priv_check_cred(cred, PRIV_NET_DELIFADDR); 677 if (error) 678 return (error); 679 } 680 681 if (addr->sin_len != sizeof(struct sockaddr_in) || 682 addr->sin_family != AF_INET) 683 deleteAny = true; 684 else 685 deleteAny = false; 686 687 iaIsLast = true; 688 ia = NULL; 689 IF_ADDR_WLOCK(ifp); 690 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 691 struct in_ifaddr *it; 692 693 if (ifa->ifa_addr->sa_family != AF_INET) 694 continue; 695 696 it = (struct in_ifaddr *)ifa; 697 if (deleteAny && ia == NULL && (cred == NULL || 698 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0)) 699 ia = it; 700 701 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 702 (cred == NULL || prison_check_ip4(cred, 703 &addr->sin_addr) == 0)) 704 ia = it; 705 706 if (it != ia) 707 iaIsLast = false; 708 } 709 710 if (ia == NULL) { 711 IF_ADDR_WUNLOCK(ifp); 712 return (EADDRNOTAVAIL); 713 } 714 715 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 716 IF_ADDR_WUNLOCK(ifp); 717 ifa_free(&ia->ia_ifa); /* if_addrhead */ 718 719 sx_assert(&in_control_sx, SA_XLOCKED); 720 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 721 CK_LIST_REMOVE(ia, ia_hash); 722 723 /* 724 * in_scrubprefix() kills the interface route. 725 */ 726 in_scrubprefix(ia, LLE_STATIC); 727 728 /* 729 * in_ifadown gets rid of all the rest of 730 * the routes. This is not quite the right 731 * thing to do, but at least if we are running 732 * a routing process they will come back. 733 */ 734 in_ifadown(&ia->ia_ifa, 1); 735 736 if (ia->ia_ifa.ifa_carp) 737 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR); 738 739 /* 740 * If this is the last IPv4 address configured on this 741 * interface, leave the all-hosts group. 742 * No state-change report need be transmitted. 743 */ 744 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 745 struct in_ifinfo *ii; 746 747 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 748 if (ii->ii_allhosts) { 749 (void)in_leavegroup(ii->ii_allhosts, NULL); 750 ii->ii_allhosts = NULL; 751 } 752 } 753 754 IF_ADDR_WLOCK(ifp); 755 if (callout_stop(&ia->ia_garp_timer) == 1) { 756 ifa_free(&ia->ia_ifa); 757 } 758 IF_ADDR_WUNLOCK(ifp); 759 760 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa, 761 IFADDR_EVENT_DEL); 762 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 763 764 return (0); 765 } 766 767 static int 768 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 769 { 770 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 771 const struct sockaddr_in *addr = &ifra->ifra_addr; 772 struct epoch_tracker et; 773 struct ifaddr *ifa; 774 struct in_ifaddr *ia; 775 776 /* 777 * ifra_addr must be present and be of INET family. 778 */ 779 if (addr->sin_len != sizeof(struct sockaddr_in) || 780 addr->sin_family != AF_INET) 781 return (EINVAL); 782 783 /* 784 * See whether address exist. 785 */ 786 ia = NULL; 787 NET_EPOCH_ENTER(et); 788 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 789 struct in_ifaddr *it; 790 791 if (ifa->ifa_addr->sa_family != AF_INET) 792 continue; 793 794 it = (struct in_ifaddr *)ifa; 795 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 796 prison_check_ip4(cred, &addr->sin_addr) == 0) { 797 ia = it; 798 break; 799 } 800 } 801 if (ia == NULL) { 802 NET_EPOCH_EXIT(et); 803 return (EADDRNOTAVAIL); 804 } 805 806 ifra->ifra_mask = ia->ia_sockmask; 807 if ((ifp->if_flags & IFF_POINTOPOINT) && 808 ia->ia_dstaddr.sin_family == AF_INET) 809 ifra->ifra_dstaddr = ia->ia_dstaddr; 810 else if ((ifp->if_flags & IFF_BROADCAST) && 811 ia->ia_broadaddr.sin_family == AF_INET) 812 ifra->ifra_broadaddr = ia->ia_broadaddr; 813 else 814 memset(&ifra->ifra_broadaddr, 0, 815 sizeof(ifra->ifra_broadaddr)); 816 817 NET_EPOCH_EXIT(et); 818 return (0); 819 } 820 821 static int 822 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg) 823 { 824 825 if (nh->nh_ifa == (struct ifaddr *)arg) 826 return (1); 827 828 return (0); 829 } 830 831 static int 832 in_handle_prefix_route(uint32_t fibnum, int cmd, 833 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa, 834 struct ifnet *ifp) 835 { 836 837 NET_EPOCH_ASSERT(); 838 839 /* Prepare gateway */ 840 struct sockaddr_dl_short sdl = { 841 .sdl_family = AF_LINK, 842 .sdl_len = sizeof(struct sockaddr_dl_short), 843 .sdl_type = ifa->ifa_ifp->if_type, 844 .sdl_index = ifa->ifa_ifp->if_index, 845 }; 846 847 struct rt_addrinfo info = { 848 .rti_ifa = ifa, 849 .rti_ifp = ifp, 850 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST), 851 .rti_info = { 852 [RTAX_DST] = (struct sockaddr *)dst, 853 [RTAX_NETMASK] = (struct sockaddr *)netmask, 854 [RTAX_GATEWAY] = (struct sockaddr *)&sdl, 855 }, 856 /* Ensure we delete the prefix IFF prefix ifa matches */ 857 .rti_filter = in_match_ifaddr, 858 .rti_filterdata = ifa, 859 }; 860 861 return (rib_handle_ifaddr_info(fibnum, cmd, &info)); 862 } 863 864 /* 865 * Routing table interaction with interface addresses. 866 * 867 * In general, two types of routes needs to be installed: 868 * a) "interface" or "prefix" route, telling user that the addresses 869 * behind the ifa prefix are reached directly. 870 * b) "loopback" route installed for the ifa address, telling user that 871 * the address belongs to local system. 872 * 873 * Handling for (a) and (b) differs in multi-fib aspects, hence they 874 * are implemented in different functions below. 875 * 876 * The cases above may intersect - /32 interface aliases results in 877 * the same prefix produced by (a) and (b). This blurs the definition 878 * of the "loopback" route and complicate interactions. The interaction 879 * table is defined below. The case numbers are used in the multiple 880 * functions below to refer to the particular test case. 881 * 882 * There can be multiple options: 883 * 1) Adding address with prefix on non-p2p/non-loopback interface. 884 * Example: 192.0.2.1/24. Action: 885 * * add "prefix" route towards 192.0.2.0/24 via @ia interface, 886 * using @ia as an address source. 887 * * add "loopback" route towards 192.0.2.1 via V_loif, saving 888 * @ia ifp in the gateway and using @ia as an address source. 889 * 890 * 2) Adding address with /32 mask to non-p2p/non-loopback interface. 891 * Example: 192.0.2.2/32. Action: 892 * * add "prefix" host route via V_loif, using @ia as an address source. 893 * 894 * 3) Adding address with or without prefix to p2p interface. 895 * Example: 10.0.0.1/24->10.0.0.2. Action: 896 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia 897 * as an address source. Note: no sense in installing full /24 as the interface 898 * is point-to-point. 899 * * add "loopback" route towards 10.0.9.1 via V_loif, saving 900 * @ia ifp in the gateway and using @ia as an address source. 901 * 902 * 4) Adding address with or without prefix to loopback interface. 903 * Example: 192.0.2.1/24. Action: 904 * * add "prefix" host route via @ia interface, using @ia as an address source. 905 * Note: Skip installing /24 prefix as it would introduce TTL loop 906 * for the traffic destined to these addresses. 907 */ 908 909 /* 910 * Checks if @ia needs to install loopback route to @ia address via 911 * ifa_maintain_loopback_route(). 912 * 913 * Return true on success. 914 */ 915 static bool 916 ia_need_loopback_route(const struct in_ifaddr *ia) 917 { 918 struct ifnet *ifp = ia->ia_ifp; 919 920 /* Case 4: Skip loopback interfaces */ 921 if ((ifp->if_flags & IFF_LOOPBACK) || 922 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)) 923 return (false); 924 925 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */ 926 if ((ifp->if_flags & IFF_POINTOPOINT) && 927 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 928 return (false); 929 930 /* Case 2: skip /32 prefixes */ 931 if (!(ifp->if_flags & IFF_POINTOPOINT) && 932 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST)) 933 return (false); 934 935 return (true); 936 } 937 938 /* 939 * Calculate "prefix" route corresponding to @ia. 940 */ 941 static void 942 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask) 943 { 944 945 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) { 946 /* Case 3: return host route for dstaddr */ 947 *prefix = ia->ia_dstaddr.sin_addr; 948 mask->s_addr = INADDR_BROADCAST; 949 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) { 950 /* Case 4: return host route for ifaddr */ 951 *prefix = ia->ia_addr.sin_addr; 952 mask->s_addr = INADDR_BROADCAST; 953 } else { 954 /* Cases 1,2: return actual ia prefix */ 955 *prefix = ia->ia_addr.sin_addr; 956 *mask = ia->ia_sockmask.sin_addr; 957 prefix->s_addr &= mask->s_addr; 958 } 959 } 960 961 /* 962 * Adds or delete interface "prefix" route corresponding to @ifa. 963 * Returns 0 on success or errno. 964 */ 965 static int 966 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia) 967 { 968 struct ifaddr *ifa = &ia->ia_ifa; 969 struct in_addr daddr, maddr; 970 struct sockaddr_in *pmask; 971 struct epoch_tracker et; 972 int error; 973 974 ia_getrtprefix(ia, &daddr, &maddr); 975 976 struct sockaddr_in mask = { 977 .sin_family = AF_INET, 978 .sin_len = sizeof(struct sockaddr_in), 979 .sin_addr = maddr, 980 }; 981 982 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL; 983 984 struct sockaddr_in dst = { 985 .sin_family = AF_INET, 986 .sin_len = sizeof(struct sockaddr_in), 987 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr, 988 }; 989 990 struct ifnet *ifp = ia->ia_ifp; 991 992 if ((maddr.s_addr == INADDR_BROADCAST) && 993 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) { 994 /* Case 2: host route on broadcast interface */ 995 ifp = V_loif; 996 } 997 998 uint32_t fibnum = ifa->ifa_ifp->if_fib; 999 NET_EPOCH_ENTER(et); 1000 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp); 1001 NET_EPOCH_EXIT(et); 1002 1003 return (error); 1004 } 1005 1006 /* 1007 * Check if we have a route for the given prefix already. 1008 */ 1009 static bool 1010 in_hasrtprefix(struct in_ifaddr *target) 1011 { 1012 struct epoch_tracker et; 1013 struct in_ifaddr *ia; 1014 struct in_addr prefix, mask, p, m; 1015 bool result = false; 1016 1017 ia_getrtprefix(target, &prefix, &mask); 1018 1019 /* Look for an existing address with the same prefix, mask, and fib */ 1020 NET_EPOCH_ENTER(et); 1021 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1022 ia_getrtprefix(ia, &p, &m); 1023 1024 if (prefix.s_addr != p.s_addr || 1025 mask.s_addr != m.s_addr) 1026 continue; 1027 1028 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 1029 continue; 1030 1031 /* 1032 * If we got a matching prefix route inserted by other 1033 * interface address, we are done here. 1034 */ 1035 if (ia->ia_flags & IFA_ROUTE) { 1036 result = true; 1037 break; 1038 } 1039 } 1040 NET_EPOCH_EXIT(et); 1041 1042 return (result); 1043 } 1044 1045 int 1046 in_addprefix(struct in_ifaddr *target) 1047 { 1048 int error; 1049 1050 if (in_hasrtprefix(target)) { 1051 if (V_nosameprefix) 1052 return (EEXIST); 1053 else { 1054 rt_addrmsg(RTM_ADD, &target->ia_ifa, 1055 target->ia_ifp->if_fib); 1056 return (0); 1057 } 1058 } 1059 1060 /* 1061 * No-one seem to have this prefix route, so we try to insert it. 1062 */ 1063 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib); 1064 error = in_handle_ifaddr_route(RTM_ADD, target); 1065 if (!error) 1066 target->ia_flags |= IFA_ROUTE; 1067 return (error); 1068 } 1069 1070 /* 1071 * Removes either all lle entries for given @ia, or lle 1072 * corresponding to @ia address. 1073 */ 1074 static void 1075 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 1076 { 1077 struct sockaddr_in addr, mask; 1078 struct sockaddr *saddr, *smask; 1079 struct ifnet *ifp; 1080 1081 saddr = (struct sockaddr *)&addr; 1082 bzero(&addr, sizeof(addr)); 1083 addr.sin_len = sizeof(addr); 1084 addr.sin_family = AF_INET; 1085 smask = (struct sockaddr *)&mask; 1086 bzero(&mask, sizeof(mask)); 1087 mask.sin_len = sizeof(mask); 1088 mask.sin_family = AF_INET; 1089 mask.sin_addr.s_addr = ia->ia_subnetmask; 1090 ifp = ia->ia_ifp; 1091 1092 if (all) { 1093 /* 1094 * Remove all L2 entries matching given prefix. 1095 * Convert address to host representation to avoid 1096 * doing this on every callback. ia_subnetmask is already 1097 * stored in host representation. 1098 */ 1099 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 1100 lltable_prefix_free(AF_INET, saddr, smask, flags); 1101 } else { 1102 /* Remove interface address only */ 1103 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 1104 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 1105 } 1106 } 1107 1108 /* 1109 * If there is no other address in the system that can serve a route to the 1110 * same prefix, remove the route. Hand over the route to the new address 1111 * otherwise. 1112 */ 1113 int 1114 in_scrubprefix(struct in_ifaddr *target, u_int flags) 1115 { 1116 struct epoch_tracker et; 1117 struct in_ifaddr *ia; 1118 struct in_addr prefix, mask, p, m; 1119 int error = 0; 1120 1121 /* 1122 * Remove the loopback route to the interface address. 1123 */ 1124 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) { 1125 struct in_ifaddr *eia; 1126 1127 eia = in_localip_more(target); 1128 1129 if (eia != NULL) { 1130 error = ifa_switch_loopback_route((struct ifaddr *)eia, 1131 (struct sockaddr *)&target->ia_addr); 1132 ifa_free(&eia->ia_ifa); 1133 } else { 1134 error = ifa_del_loopback_route((struct ifaddr *)target, 1135 (struct sockaddr *)&target->ia_addr); 1136 } 1137 } 1138 1139 ia_getrtprefix(target, &prefix, &mask); 1140 1141 if ((target->ia_flags & IFA_ROUTE) == 0) { 1142 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1143 1144 /* 1145 * Removing address from !IFF_UP interface or 1146 * prefix which exists on other interface (along with route). 1147 * No entries should exist here except target addr. 1148 * Given that, delete this entry only. 1149 */ 1150 in_scrubprefixlle(target, 0, flags); 1151 return (0); 1152 } 1153 1154 NET_EPOCH_ENTER(et); 1155 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1156 ia_getrtprefix(ia, &p, &m); 1157 1158 if (prefix.s_addr != p.s_addr || 1159 mask.s_addr != m.s_addr) 1160 continue; 1161 1162 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1163 continue; 1164 1165 /* 1166 * If we got a matching prefix address, move IFA_ROUTE and 1167 * the route itself to it. Make sure that routing daemons 1168 * get a heads-up. 1169 */ 1170 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1171 ifa_ref(&ia->ia_ifa); 1172 NET_EPOCH_EXIT(et); 1173 error = in_handle_ifaddr_route(RTM_DELETE, target); 1174 if (error == 0) 1175 target->ia_flags &= ~IFA_ROUTE; 1176 else 1177 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1178 error); 1179 /* Scrub all entries IFF interface is different */ 1180 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 1181 flags); 1182 error = in_handle_ifaddr_route(RTM_ADD, ia); 1183 if (error == 0) 1184 ia->ia_flags |= IFA_ROUTE; 1185 else 1186 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1187 error); 1188 ifa_free(&ia->ia_ifa); 1189 return (error); 1190 } 1191 } 1192 NET_EPOCH_EXIT(et); 1193 1194 /* 1195 * remove all L2 entries on the given prefix 1196 */ 1197 in_scrubprefixlle(target, 1, flags); 1198 1199 /* 1200 * As no-one seem to have this prefix, we can remove the route. 1201 */ 1202 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1203 error = in_handle_ifaddr_route(RTM_DELETE, target); 1204 if (error == 0) 1205 target->ia_flags &= ~IFA_ROUTE; 1206 else 1207 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1208 return (error); 1209 } 1210 1211 void 1212 in_ifscrub_all(void) 1213 { 1214 struct ifnet *ifp; 1215 struct ifaddr *ifa, *nifa; 1216 struct ifaliasreq ifr; 1217 1218 IFNET_RLOCK(); 1219 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1220 /* Cannot lock here - lock recursion. */ 1221 /* NET_EPOCH_ENTER(et); */ 1222 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 1223 if (ifa->ifa_addr->sa_family != AF_INET) 1224 continue; 1225 1226 /* 1227 * This is ugly but the only way for legacy IP to 1228 * cleanly remove addresses and everything attached. 1229 */ 1230 bzero(&ifr, sizeof(ifr)); 1231 ifr.ifra_addr = *ifa->ifa_addr; 1232 if (ifa->ifa_dstaddr) 1233 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1234 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 1235 ifp, NULL); 1236 } 1237 /* NET_EPOCH_EXIT(et); */ 1238 in_purgemaddrs(ifp); 1239 igmp_domifdetach(ifp); 1240 } 1241 IFNET_RUNLOCK(); 1242 } 1243 1244 int 1245 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 1246 { 1247 1248 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1249 /* 1250 * Optionally check for old-style (host 0) broadcast, but 1251 * taking into account that RFC 3021 obsoletes it. 1252 */ 1253 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK && 1254 ntohl(in.s_addr) == ia->ia_subnet)) && 1255 /* 1256 * Check for an all one subnetmask. These 1257 * only exist when an interface gets a secondary 1258 * address. 1259 */ 1260 ia->ia_subnetmask != (u_long)0xffffffff); 1261 } 1262 1263 /* 1264 * Return 1 if the address might be a local broadcast address. 1265 */ 1266 int 1267 in_broadcast(struct in_addr in, struct ifnet *ifp) 1268 { 1269 struct ifaddr *ifa; 1270 int found; 1271 1272 NET_EPOCH_ASSERT(); 1273 1274 if (in.s_addr == INADDR_BROADCAST || 1275 in.s_addr == INADDR_ANY) 1276 return (1); 1277 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1278 return (0); 1279 found = 0; 1280 /* 1281 * Look through the list of addresses for a match 1282 * with a broadcast address. 1283 */ 1284 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1285 if (ifa->ifa_addr->sa_family == AF_INET && 1286 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 1287 found = 1; 1288 break; 1289 } 1290 return (found); 1291 } 1292 1293 /* 1294 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1295 */ 1296 void 1297 in_ifdetach(struct ifnet *ifp) 1298 { 1299 IN_MULTI_LOCK(); 1300 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1301 in_pcbpurgeif0(&V_udbinfo, ifp); 1302 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 1303 in_purgemaddrs(ifp); 1304 IN_MULTI_UNLOCK(); 1305 1306 /* 1307 * Make sure all multicast deletions invoking if_ioctl() are 1308 * completed before returning. Else we risk accessing a freed 1309 * ifnet structure pointer. 1310 */ 1311 inm_release_wait(NULL); 1312 } 1313 1314 static void 1315 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event) 1316 { 1317 struct epoch_tracker et; 1318 struct ifaddr *ifa; 1319 struct in_ifaddr *ia; 1320 int error; 1321 1322 NET_EPOCH_ENTER(et); 1323 switch (event) { 1324 case IFNET_EVENT_DOWN: 1325 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1326 if (ifa->ifa_addr->sa_family != AF_INET) 1327 continue; 1328 ia = (struct in_ifaddr *)ifa; 1329 if ((ia->ia_flags & IFA_ROUTE) == 0) 1330 continue; 1331 ifa_ref(ifa); 1332 /* 1333 * in_scrubprefix() kills the interface route. 1334 */ 1335 in_scrubprefix(ia, 0); 1336 /* 1337 * in_ifadown gets rid of all the rest of the 1338 * routes. This is not quite the right thing 1339 * to do, but at least if we are running a 1340 * routing process they will come back. 1341 */ 1342 in_ifadown(ifa, 0); 1343 ifa_free(ifa); 1344 } 1345 break; 1346 1347 case IFNET_EVENT_UP: 1348 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1349 if (ifa->ifa_addr->sa_family != AF_INET) 1350 continue; 1351 ia = (struct in_ifaddr *)ifa; 1352 if (ia->ia_flags & IFA_ROUTE) 1353 continue; 1354 ifa_ref(ifa); 1355 error = ifa_del_loopback_route(ifa, ifa->ifa_addr); 1356 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib); 1357 error = in_handle_ifaddr_route(RTM_ADD, ia); 1358 if (error == 0) 1359 ia->ia_flags |= IFA_ROUTE; 1360 error = ifa_add_loopback_route(ifa, ifa->ifa_addr); 1361 ifa_free(ifa); 1362 } 1363 break; 1364 } 1365 NET_EPOCH_EXIT(et); 1366 } 1367 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY); 1368 1369 /* 1370 * Delete all IPv4 multicast address records, and associated link-layer 1371 * multicast address records, associated with ifp. 1372 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1373 * XXX This should not race with ifma_protospec being set during 1374 * a new allocation, if it does, we have bigger problems. 1375 */ 1376 static void 1377 in_purgemaddrs(struct ifnet *ifp) 1378 { 1379 struct epoch_tracker et; 1380 struct in_multi_head purgeinms; 1381 struct in_multi *inm; 1382 struct ifmultiaddr *ifma; 1383 1384 SLIST_INIT(&purgeinms); 1385 IN_MULTI_LIST_LOCK(); 1386 1387 /* 1388 * Extract list of in_multi associated with the detaching ifp 1389 * which the PF_INET layer is about to release. 1390 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1391 * by code further down. 1392 */ 1393 IF_ADDR_WLOCK(ifp); 1394 NET_EPOCH_ENTER(et); 1395 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1396 inm = inm_ifmultiaddr_get_inm(ifma); 1397 if (inm == NULL) 1398 continue; 1399 inm_rele_locked(&purgeinms, inm); 1400 } 1401 NET_EPOCH_EXIT(et); 1402 IF_ADDR_WUNLOCK(ifp); 1403 1404 inm_release_list_deferred(&purgeinms); 1405 igmp_ifdetach(ifp); 1406 IN_MULTI_LIST_UNLOCK(); 1407 } 1408 1409 struct in_llentry { 1410 struct llentry base; 1411 }; 1412 1413 #define IN_LLTBL_DEFAULT_HSIZE 32 1414 #define IN_LLTBL_HASH(k, h) \ 1415 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1416 1417 /* 1418 * Do actual deallocation of @lle. 1419 */ 1420 static void 1421 in_lltable_destroy_lle_unlocked(epoch_context_t ctx) 1422 { 1423 struct llentry *lle; 1424 1425 lle = __containerof(ctx, struct llentry, lle_epoch_ctx); 1426 LLE_LOCK_DESTROY(lle); 1427 LLE_REQ_DESTROY(lle); 1428 free(lle, M_LLTABLE); 1429 } 1430 1431 /* 1432 * Called by LLE_FREE_LOCKED when number of references 1433 * drops to zero. 1434 */ 1435 static void 1436 in_lltable_destroy_lle(struct llentry *lle) 1437 { 1438 1439 LLE_WUNLOCK(lle); 1440 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx); 1441 } 1442 1443 static struct llentry * 1444 in_lltable_new(struct in_addr addr4, u_int flags) 1445 { 1446 struct in_llentry *lle; 1447 1448 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1449 if (lle == NULL) /* NB: caller generates msg */ 1450 return NULL; 1451 1452 /* 1453 * For IPv4 this will trigger "arpresolve" to generate 1454 * an ARP request. 1455 */ 1456 lle->base.la_expire = time_uptime; /* mark expired */ 1457 lle->base.r_l3addr.addr4 = addr4; 1458 lle->base.lle_refcnt = 1; 1459 lle->base.lle_free = in_lltable_destroy_lle; 1460 LLE_LOCK_INIT(&lle->base); 1461 LLE_REQ_INIT(&lle->base); 1462 callout_init(&lle->base.lle_timer, 1); 1463 1464 return (&lle->base); 1465 } 1466 1467 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1468 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 ) 1469 1470 static int 1471 in_lltable_match_prefix(const struct sockaddr *saddr, 1472 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1473 { 1474 struct in_addr addr, mask, lle_addr; 1475 1476 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1477 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1478 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1479 1480 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1481 return (0); 1482 1483 if (lle->la_flags & LLE_IFADDR) { 1484 /* 1485 * Delete LLE_IFADDR records IFF address & flag matches. 1486 * Note that addr is the interface address within prefix 1487 * being matched. 1488 * Note also we should handle 'ifdown' cases without removing 1489 * ifaddr macs. 1490 */ 1491 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1492 return (1); 1493 return (0); 1494 } 1495 1496 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1497 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1498 return (1); 1499 1500 return (0); 1501 } 1502 1503 static void 1504 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1505 { 1506 size_t pkts_dropped; 1507 1508 LLE_WLOCK_ASSERT(lle); 1509 KASSERT(llt != NULL, ("lltable is NULL")); 1510 1511 /* Unlink entry from table if not already */ 1512 if ((lle->la_flags & LLE_LINKED) != 0) { 1513 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 1514 lltable_unlink_entry(llt, lle); 1515 } 1516 1517 /* Drop hold queue */ 1518 pkts_dropped = llentry_free(lle); 1519 ARPSTAT_ADD(dropped, pkts_dropped); 1520 } 1521 1522 static int 1523 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1524 { 1525 struct nhop_object *nh; 1526 struct in_addr addr; 1527 1528 KASSERT(l3addr->sa_family == AF_INET, 1529 ("sin_family %d", l3addr->sa_family)); 1530 1531 addr = ((const struct sockaddr_in *)l3addr)->sin_addr; 1532 1533 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0); 1534 if (nh == NULL) 1535 return (EINVAL); 1536 1537 /* 1538 * If the gateway for an existing host route matches the target L3 1539 * address, which is a special route inserted by some implementation 1540 * such as MANET, and the interface is of the correct type, then 1541 * allow for ARP to proceed. 1542 */ 1543 if (nh->nh_flags & NHF_GATEWAY) { 1544 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER || 1545 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1546 memcmp(nh->gw_sa.sa_data, l3addr->sa_data, 1547 sizeof(in_addr_t)) != 0) { 1548 return (EINVAL); 1549 } 1550 } 1551 1552 /* 1553 * Make sure that at least the destination address is covered 1554 * by the route. This is for handling the case where 2 or more 1555 * interfaces have the same prefix. An incoming packet arrives 1556 * on one interface and the corresponding outgoing packet leaves 1557 * another interface. 1558 */ 1559 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) { 1560 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp); 1561 struct in_addr dst_addr, mask_addr; 1562 1563 if (ia == NULL) 1564 return (EINVAL); 1565 1566 /* 1567 * ifaof_ifpforaddr() returns _best matching_ IFA. 1568 * It is possible that ifa prefix does not cover our address. 1569 * Explicitly verify and fail if that's the case. 1570 */ 1571 dst_addr = IA_SIN(ia)->sin_addr; 1572 mask_addr.s_addr = htonl(ia->ia_subnetmask); 1573 1574 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr)) 1575 return (EINVAL); 1576 } 1577 1578 return (0); 1579 } 1580 1581 static inline uint32_t 1582 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1583 { 1584 1585 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1586 } 1587 1588 static uint32_t 1589 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1590 { 1591 1592 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1593 } 1594 1595 static void 1596 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1597 { 1598 struct sockaddr_in *sin; 1599 1600 sin = (struct sockaddr_in *)sa; 1601 bzero(sin, sizeof(*sin)); 1602 sin->sin_family = AF_INET; 1603 sin->sin_len = sizeof(*sin); 1604 sin->sin_addr = lle->r_l3addr.addr4; 1605 } 1606 1607 static inline struct llentry * 1608 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1609 { 1610 struct llentry *lle; 1611 struct llentries *lleh; 1612 u_int hashidx; 1613 1614 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1615 lleh = &llt->lle_head[hashidx]; 1616 CK_LIST_FOREACH(lle, lleh, lle_next) { 1617 if (lle->la_flags & LLE_DELETED) 1618 continue; 1619 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1620 break; 1621 } 1622 1623 return (lle); 1624 } 1625 1626 static void 1627 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1628 { 1629 1630 lle->la_flags |= LLE_DELETED; 1631 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1632 #ifdef DIAGNOSTIC 1633 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1634 #endif 1635 llentry_free(lle); 1636 } 1637 1638 static struct llentry * 1639 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1640 { 1641 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1642 struct ifnet *ifp = llt->llt_ifp; 1643 struct llentry *lle; 1644 char linkhdr[LLE_MAX_LINKHDR]; 1645 size_t linkhdrsize; 1646 int lladdr_off; 1647 1648 KASSERT(l3addr->sa_family == AF_INET, 1649 ("sin_family %d", l3addr->sa_family)); 1650 1651 /* 1652 * A route that covers the given address must have 1653 * been installed 1st because we are doing a resolution, 1654 * verify this. 1655 */ 1656 if (!(flags & LLE_IFADDR) && 1657 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1658 return (NULL); 1659 1660 lle = in_lltable_new(sin->sin_addr, flags); 1661 if (lle == NULL) { 1662 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1663 return (NULL); 1664 } 1665 lle->la_flags = flags; 1666 if (flags & LLE_STATIC) 1667 lle->r_flags |= RLLE_VALID; 1668 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1669 linkhdrsize = LLE_MAX_LINKHDR; 1670 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1671 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1672 in_lltable_free_entry(llt, lle); 1673 return (NULL); 1674 } 1675 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1676 lladdr_off); 1677 lle->la_flags |= LLE_STATIC; 1678 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1679 lle->la_expire = 0; 1680 } 1681 1682 return (lle); 1683 } 1684 1685 /* 1686 * Return NULL if not found or marked for deletion. 1687 * If found return lle read locked. 1688 */ 1689 static struct llentry * 1690 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1691 { 1692 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1693 struct llentry *lle; 1694 1695 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1696 KASSERT(l3addr->sa_family == AF_INET, 1697 ("sin_family %d", l3addr->sa_family)); 1698 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) != 1699 (LLE_UNLOCKED | LLE_EXCLUSIVE), 1700 ("wrong lle request flags: %#x", flags)); 1701 1702 lle = in_lltable_find_dst(llt, sin->sin_addr); 1703 if (lle == NULL) 1704 return (NULL); 1705 if (flags & LLE_UNLOCKED) 1706 return (lle); 1707 1708 if (flags & LLE_EXCLUSIVE) 1709 LLE_WLOCK(lle); 1710 else 1711 LLE_RLOCK(lle); 1712 1713 /* 1714 * If the afdata lock is not held, the LLE may have been unlinked while 1715 * we were blocked on the LLE lock. Check for this case. 1716 */ 1717 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) { 1718 if (flags & LLE_EXCLUSIVE) 1719 LLE_WUNLOCK(lle); 1720 else 1721 LLE_RUNLOCK(lle); 1722 return (NULL); 1723 } 1724 return (lle); 1725 } 1726 1727 static int 1728 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1729 struct sysctl_req *wr) 1730 { 1731 struct ifnet *ifp = llt->llt_ifp; 1732 /* XXX stack use */ 1733 struct { 1734 struct rt_msghdr rtm; 1735 struct sockaddr_in sin; 1736 struct sockaddr_dl sdl; 1737 } arpc; 1738 struct sockaddr_dl *sdl; 1739 int error; 1740 1741 bzero(&arpc, sizeof(arpc)); 1742 /* skip deleted entries */ 1743 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1744 return (0); 1745 /* Skip if jailed and not a valid IP of the prison. */ 1746 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1747 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0) 1748 return (0); 1749 /* 1750 * produce a msg made of: 1751 * struct rt_msghdr; 1752 * struct sockaddr_in; (IPv4) 1753 * struct sockaddr_dl; 1754 */ 1755 arpc.rtm.rtm_msglen = sizeof(arpc); 1756 arpc.rtm.rtm_version = RTM_VERSION; 1757 arpc.rtm.rtm_type = RTM_GET; 1758 arpc.rtm.rtm_flags = RTF_UP; 1759 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1760 1761 /* publish */ 1762 if (lle->la_flags & LLE_PUB) 1763 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1764 1765 sdl = &arpc.sdl; 1766 sdl->sdl_family = AF_LINK; 1767 sdl->sdl_len = sizeof(*sdl); 1768 sdl->sdl_index = ifp->if_index; 1769 sdl->sdl_type = ifp->if_type; 1770 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1771 sdl->sdl_alen = ifp->if_addrlen; 1772 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1773 } else { 1774 sdl->sdl_alen = 0; 1775 bzero(LLADDR(sdl), ifp->if_addrlen); 1776 } 1777 1778 arpc.rtm.rtm_rmx.rmx_expire = 1779 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1780 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1781 if (lle->la_flags & LLE_STATIC) 1782 arpc.rtm.rtm_flags |= RTF_STATIC; 1783 if (lle->la_flags & LLE_IFADDR) 1784 arpc.rtm.rtm_flags |= RTF_PINNED; 1785 arpc.rtm.rtm_index = ifp->if_index; 1786 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1787 1788 return (error); 1789 } 1790 1791 static void 1792 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle) 1793 { 1794 struct ifnet *ifp = llt->llt_ifp; 1795 1796 /* gratuitous ARP */ 1797 if ((lle->la_flags & LLE_PUB) != 0) 1798 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4, 1799 lle->ll_addr); 1800 } 1801 1802 static struct lltable * 1803 in_lltattach(struct ifnet *ifp) 1804 { 1805 struct lltable *llt; 1806 1807 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1808 llt->llt_af = AF_INET; 1809 llt->llt_ifp = ifp; 1810 1811 llt->llt_lookup = in_lltable_lookup; 1812 llt->llt_alloc_entry = in_lltable_alloc; 1813 llt->llt_delete_entry = in_lltable_delete_entry; 1814 llt->llt_dump_entry = in_lltable_dump_entry; 1815 llt->llt_hash = in_lltable_hash; 1816 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1817 llt->llt_free_entry = in_lltable_free_entry; 1818 llt->llt_match_prefix = in_lltable_match_prefix; 1819 llt->llt_mark_used = llentry_mark_used; 1820 llt->llt_post_resolved = in_lltable_post_resolved; 1821 lltable_link(llt); 1822 1823 return (llt); 1824 } 1825 1826 struct lltable * 1827 in_lltable_get(struct ifnet *ifp) 1828 { 1829 struct lltable *llt = NULL; 1830 1831 void *afdata_ptr = ifp->if_afdata[AF_INET]; 1832 if (afdata_ptr != NULL) 1833 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt; 1834 return (llt); 1835 } 1836 1837 void * 1838 in_domifattach(struct ifnet *ifp) 1839 { 1840 struct in_ifinfo *ii; 1841 1842 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1843 1844 ii->ii_llt = in_lltattach(ifp); 1845 ii->ii_igmp = igmp_domifattach(ifp); 1846 1847 return (ii); 1848 } 1849 1850 void 1851 in_domifdetach(struct ifnet *ifp, void *aux) 1852 { 1853 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1854 1855 igmp_domifdetach(ifp); 1856 lltable_free(ii->ii_llt); 1857 free(ii, M_IFADDR); 1858 } 1859