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