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