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 33 #include <sys/cdefs.h> 34 #include "opt_inet.h" 35 36 #define IN_HISTORICAL_NETS /* include class masks */ 37 38 #include <sys/param.h> 39 #include <sys/eventhandler.h> 40 #include <sys/systm.h> 41 #include <sys/sockio.h> 42 #include <sys/malloc.h> 43 #include <sys/priv.h> 44 #include <sys/socket.h> 45 #include <sys/jail.h> 46 #include <sys/kernel.h> 47 #include <sys/lock.h> 48 #include <sys/proc.h> 49 #include <sys/sysctl.h> 50 #include <sys/syslog.h> 51 #include <sys/sx.h> 52 53 #include <net/if.h> 54 #include <net/if_var.h> 55 #include <net/if_arp.h> 56 #include <net/if_dl.h> 57 #include <net/if_llatbl.h> 58 #include <net/if_private.h> 59 #include <net/if_types.h> 60 #include <net/route.h> 61 #include <net/route/nhop.h> 62 #include <net/route/route_ctl.h> 63 #include <net/vnet.h> 64 65 #include <netinet/if_ether.h> 66 #include <netinet/in.h> 67 #include <netinet/in_fib.h> 68 #include <netinet/in_var.h> 69 #include <netinet/in_pcb.h> 70 #include <netinet/ip_var.h> 71 #include <netinet/ip_carp.h> 72 #include <netinet/igmp_var.h> 73 #include <netinet/udp.h> 74 #include <netinet/udp_var.h> 75 76 #ifdef MAC 77 #include <security/mac/mac_framework.h> 78 #endif 79 80 static int in_aifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 81 static int in_difaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 82 static int in_gifaddr_ioctl(u_long, caddr_t, struct ifnet *, struct ucred *); 83 84 static void in_socktrim(struct sockaddr_in *); 85 static void in_purgemaddrs(struct ifnet *); 86 87 static bool ia_need_loopback_route(const struct in_ifaddr *); 88 89 VNET_DEFINE_STATIC(int, nosameprefix); 90 #define V_nosameprefix VNET(nosameprefix) 91 SYSCTL_INT(_net_inet_ip, OID_AUTO, no_same_prefix, CTLFLAG_VNET | CTLFLAG_RW, 92 &VNET_NAME(nosameprefix), 0, 93 "Refuse to create same prefixes on different interfaces"); 94 95 VNET_DEFINE_STATIC(bool, broadcast_lowest); 96 #define V_broadcast_lowest VNET(broadcast_lowest) 97 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, broadcast_lowest, CTLFLAG_VNET | CTLFLAG_RW, 98 &VNET_NAME(broadcast_lowest), 0, 99 "Treat lowest address on a subnet (host 0) as broadcast"); 100 101 VNET_DEFINE(bool, ip_allow_net240) = false; 102 #define V_ip_allow_net240 VNET(ip_allow_net240) 103 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net240, 104 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net240), 0, 105 "Allow use of Experimental addresses, aka Class E (240/4)"); 106 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-240 */ 107 108 VNET_DEFINE(bool, ip_allow_net0) = false; 109 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, allow_net0, 110 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_allow_net0), 0, 111 "Allow use of addresses in network 0/8"); 112 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-0 */ 113 114 VNET_DEFINE(uint32_t, in_loopback_mask) = IN_LOOPBACK_MASK_DFLT; 115 #define V_in_loopback_mask VNET(in_loopback_mask) 116 static int sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS); 117 SYSCTL_PROC(_net_inet_ip, OID_AUTO, loopback_prefixlen, 118 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 119 NULL, 0, sysctl_loopback_prefixlen, "I", 120 "Prefix length of address space reserved for loopback"); 121 /* see https://datatracker.ietf.org/doc/draft-schoen-intarea-unicast-127 */ 122 123 VNET_DECLARE(struct inpcbinfo, ripcbinfo); 124 #define V_ripcbinfo VNET(ripcbinfo) 125 126 static struct sx in_control_sx; 127 SX_SYSINIT(in_control_sx, &in_control_sx, "in_control"); 128 129 /* 130 * Return 1 if an internet address is for a ``local'' host 131 * (one to which we have a connection). 132 */ 133 int 134 in_localaddr(struct in_addr in) 135 { 136 u_long i = ntohl(in.s_addr); 137 struct in_ifaddr *ia; 138 139 NET_EPOCH_ASSERT(); 140 141 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 142 if ((i & ia->ia_subnetmask) == ia->ia_subnet) 143 return (1); 144 } 145 146 return (0); 147 } 148 149 /* 150 * Return 1 if an internet address is for the local host and configured 151 * on one of its interfaces. 152 */ 153 bool 154 in_localip(struct in_addr in) 155 { 156 struct in_ifaddr *ia; 157 158 NET_EPOCH_ASSERT(); 159 160 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) 161 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr) 162 return (true); 163 164 return (false); 165 } 166 167 /* 168 * Like in_localip(), but FIB-aware. 169 */ 170 bool 171 in_localip_fib(struct in_addr in, uint16_t fib) 172 { 173 struct in_ifaddr *ia; 174 175 NET_EPOCH_ASSERT(); 176 177 CK_LIST_FOREACH(ia, INADDR_HASH(in.s_addr), ia_hash) 178 if (IA_SIN(ia)->sin_addr.s_addr == in.s_addr && 179 ia->ia_ifa.ifa_ifp->if_fib == fib) 180 return (true); 181 182 return (false); 183 } 184 185 /* 186 * Return 1 if an internet address is configured on an interface. 187 */ 188 int 189 in_ifhasaddr(struct ifnet *ifp, struct in_addr in) 190 { 191 struct ifaddr *ifa; 192 struct in_ifaddr *ia; 193 194 NET_EPOCH_ASSERT(); 195 196 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 197 if (ifa->ifa_addr->sa_family != AF_INET) 198 continue; 199 ia = (struct in_ifaddr *)ifa; 200 if (ia->ia_addr.sin_addr.s_addr == in.s_addr) 201 return (1); 202 } 203 204 return (0); 205 } 206 207 /* 208 * Return a reference to the interface address which is different to 209 * the supplied one but with same IP address value. 210 */ 211 static struct in_ifaddr * 212 in_localip_more(struct in_ifaddr *original_ia) 213 { 214 struct epoch_tracker et; 215 in_addr_t original_addr = IA_SIN(original_ia)->sin_addr.s_addr; 216 uint32_t original_fib = original_ia->ia_ifa.ifa_ifp->if_fib; 217 struct in_ifaddr *ia; 218 219 NET_EPOCH_ENTER(et); 220 CK_LIST_FOREACH(ia, INADDR_HASH(original_addr), ia_hash) { 221 in_addr_t addr = IA_SIN(ia)->sin_addr.s_addr; 222 uint32_t fib = ia->ia_ifa.ifa_ifp->if_fib; 223 if (!V_rt_add_addr_allfibs && (original_fib != fib)) 224 continue; 225 if ((original_ia != ia) && (original_addr == addr)) { 226 ifa_ref(&ia->ia_ifa); 227 NET_EPOCH_EXIT(et); 228 return (ia); 229 } 230 } 231 NET_EPOCH_EXIT(et); 232 233 return (NULL); 234 } 235 236 /* 237 * Tries to find first IPv4 address in the provided fib. 238 * Prefers non-loopback addresses and return loopback IFF 239 * @loopback_ok is set. 240 * 241 * Returns ifa or NULL. 242 */ 243 struct in_ifaddr * 244 in_findlocal(uint32_t fibnum, bool loopback_ok) 245 { 246 struct in_ifaddr *ia = NULL, *ia_lo = NULL; 247 248 NET_EPOCH_ASSERT(); 249 250 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 251 uint32_t ia_fib = ia->ia_ifa.ifa_ifp->if_fib; 252 if (!V_rt_add_addr_allfibs && (fibnum != ia_fib)) 253 continue; 254 255 if (!IN_LOOPBACK(ntohl(IA_SIN(ia)->sin_addr.s_addr))) 256 break; 257 if (loopback_ok) 258 ia_lo = ia; 259 } 260 261 if (ia == NULL) 262 ia = ia_lo; 263 264 return (ia); 265 } 266 267 /* 268 * Determine whether an IP address is in a reserved set of addresses 269 * that may not be forwarded, or whether datagrams to that destination 270 * may be forwarded. 271 */ 272 int 273 in_canforward(struct in_addr in) 274 { 275 u_long i = ntohl(in.s_addr); 276 277 if (IN_MULTICAST(i) || IN_LINKLOCAL(i) || IN_LOOPBACK(i)) 278 return (0); 279 if (IN_EXPERIMENTAL(i) && !V_ip_allow_net240) 280 return (0); 281 if (IN_ZERONET(i) && !V_ip_allow_net0) 282 return (0); 283 return (1); 284 } 285 286 /* 287 * Sysctl to manage prefix of reserved loopback network; translate 288 * to/from mask. The mask is always contiguous high-order 1 bits 289 * followed by all 0 bits. 290 */ 291 static int 292 sysctl_loopback_prefixlen(SYSCTL_HANDLER_ARGS) 293 { 294 int error, preflen; 295 296 /* ffs is 1-based; compensate. */ 297 preflen = 33 - ffs(V_in_loopback_mask); 298 error = sysctl_handle_int(oidp, &preflen, 0, req); 299 if (error || !req->newptr) 300 return (error); 301 if (preflen < 8 || preflen > 31) 302 return (EINVAL); 303 V_in_loopback_mask = 0xffffffff << (32 - preflen); 304 return (0); 305 } 306 307 /* 308 * Trim a mask in a sockaddr 309 */ 310 static void 311 in_socktrim(struct sockaddr_in *ap) 312 { 313 char *cplim = (char *) &ap->sin_addr; 314 char *cp = (char *) (&ap->sin_addr + 1); 315 316 ap->sin_len = 0; 317 while (--cp >= cplim) 318 if (*cp) { 319 (ap)->sin_len = cp - (char *) (ap) + 1; 320 break; 321 } 322 } 323 324 /* 325 * Generic internet control operations (ioctl's). 326 */ 327 int 328 in_control_ioctl(u_long cmd, void *data, struct ifnet *ifp, 329 struct ucred *cred) 330 { 331 struct ifreq *ifr = (struct ifreq *)data; 332 struct sockaddr_in *addr = (struct sockaddr_in *)&ifr->ifr_addr; 333 struct epoch_tracker et; 334 struct ifaddr *ifa; 335 struct in_ifaddr *ia; 336 int error; 337 338 if (ifp == NULL) 339 return (EADDRNOTAVAIL); 340 341 /* 342 * Filter out 4 ioctls we implement directly. Forward the rest 343 * to specific functions and ifp->if_ioctl(). 344 */ 345 switch (cmd) { 346 case SIOCGIFADDR: 347 case SIOCGIFBRDADDR: 348 case SIOCGIFDSTADDR: 349 case SIOCGIFNETMASK: 350 break; 351 case SIOCGIFALIAS: 352 sx_xlock(&in_control_sx); 353 error = in_gifaddr_ioctl(cmd, data, ifp, cred); 354 sx_xunlock(&in_control_sx); 355 return (error); 356 case SIOCDIFADDR: 357 sx_xlock(&in_control_sx); 358 error = in_difaddr_ioctl(cmd, data, ifp, cred); 359 sx_xunlock(&in_control_sx); 360 return (error); 361 case OSIOCAIFADDR: /* 9.x compat */ 362 case SIOCAIFADDR: 363 sx_xlock(&in_control_sx); 364 error = in_aifaddr_ioctl(cmd, data, ifp, cred); 365 sx_xunlock(&in_control_sx); 366 return (error); 367 case SIOCSIFADDR: 368 case SIOCSIFBRDADDR: 369 case SIOCSIFDSTADDR: 370 case SIOCSIFNETMASK: 371 /* We no longer support that old commands. */ 372 return (EINVAL); 373 default: 374 if (ifp->if_ioctl == NULL) 375 return (EOPNOTSUPP); 376 return ((*ifp->if_ioctl)(ifp, cmd, data)); 377 } 378 379 if (addr->sin_addr.s_addr != INADDR_ANY && 380 prison_check_ip4(cred, &addr->sin_addr) != 0) 381 return (EADDRNOTAVAIL); 382 383 /* 384 * Find address for this interface, if it exists. If an 385 * address was specified, find that one instead of the 386 * first one on the interface, if possible. 387 */ 388 NET_EPOCH_ENTER(et); 389 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 390 if (ifa->ifa_addr->sa_family != AF_INET) 391 continue; 392 ia = (struct in_ifaddr *)ifa; 393 if (ia->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr) 394 break; 395 } 396 if (ifa == NULL) 397 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 398 if (ifa->ifa_addr->sa_family == AF_INET) { 399 ia = (struct in_ifaddr *)ifa; 400 if (prison_check_ip4(cred, 401 &ia->ia_addr.sin_addr) == 0) 402 break; 403 } 404 405 if (ifa == NULL) { 406 NET_EPOCH_EXIT(et); 407 return (EADDRNOTAVAIL); 408 } 409 410 error = 0; 411 switch (cmd) { 412 case SIOCGIFADDR: 413 *addr = ia->ia_addr; 414 break; 415 416 case SIOCGIFBRDADDR: 417 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 418 error = EINVAL; 419 break; 420 } 421 *addr = ia->ia_broadaddr; 422 break; 423 424 case SIOCGIFDSTADDR: 425 if ((ifp->if_flags & IFF_POINTOPOINT) == 0) { 426 error = EINVAL; 427 break; 428 } 429 *addr = ia->ia_dstaddr; 430 break; 431 432 case SIOCGIFNETMASK: 433 *addr = ia->ia_sockmask; 434 break; 435 } 436 437 NET_EPOCH_EXIT(et); 438 439 return (error); 440 } 441 442 int 443 in_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp, 444 struct thread *td) 445 { 446 return (in_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL)); 447 } 448 449 static int 450 in_aifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 451 { 452 const struct in_aliasreq *ifra = (struct in_aliasreq *)data; 453 const struct sockaddr_in *addr = &ifra->ifra_addr; 454 const struct sockaddr_in *broadaddr = &ifra->ifra_broadaddr; 455 const struct sockaddr_in *mask = &ifra->ifra_mask; 456 const struct sockaddr_in *dstaddr = &ifra->ifra_dstaddr; 457 const int vhid = (cmd == SIOCAIFADDR) ? ifra->ifra_vhid : 0; 458 struct epoch_tracker et; 459 struct ifaddr *ifa; 460 struct in_ifaddr *ia; 461 bool iaIsFirst; 462 int error = 0; 463 464 error = priv_check_cred(cred, PRIV_NET_ADDIFADDR); 465 if (error) 466 return (error); 467 468 /* 469 * ifra_addr must be present and be of INET family. 470 * ifra_broadaddr/ifra_dstaddr and ifra_mask are optional. 471 */ 472 if (addr->sin_len != sizeof(struct sockaddr_in) || 473 addr->sin_family != AF_INET) 474 return (EINVAL); 475 if (broadaddr->sin_len != 0 && 476 (broadaddr->sin_len != sizeof(struct sockaddr_in) || 477 broadaddr->sin_family != AF_INET)) 478 return (EINVAL); 479 if (mask->sin_len != 0 && 480 (mask->sin_len != sizeof(struct sockaddr_in) || 481 mask->sin_family != AF_INET)) 482 return (EINVAL); 483 if ((ifp->if_flags & IFF_POINTOPOINT) && 484 (dstaddr->sin_len != sizeof(struct sockaddr_in) || 485 dstaddr->sin_addr.s_addr == INADDR_ANY)) 486 return (EDESTADDRREQ); 487 if (vhid != 0 && carp_attach_p == NULL) 488 return (EPROTONOSUPPORT); 489 490 #ifdef MAC 491 /* Check if a MAC policy disallows setting the IPv4 address. */ 492 error = mac_inet_check_add_addr(cred, &addr->sin_addr, ifp); 493 if (error != 0) 494 return (error); 495 #endif 496 497 /* 498 * See whether address already exist. 499 */ 500 iaIsFirst = true; 501 ia = NULL; 502 NET_EPOCH_ENTER(et); 503 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 504 struct in_ifaddr *it; 505 506 if (ifa->ifa_addr->sa_family != AF_INET) 507 continue; 508 509 it = (struct in_ifaddr *)ifa; 510 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 511 prison_check_ip4(cred, &addr->sin_addr) == 0) 512 ia = it; 513 else 514 iaIsFirst = false; 515 } 516 NET_EPOCH_EXIT(et); 517 518 if (ia != NULL) 519 (void )in_difaddr_ioctl(cmd, data, ifp, cred); 520 521 ifa = ifa_alloc(sizeof(struct in_ifaddr), M_WAITOK); 522 ia = (struct in_ifaddr *)ifa; 523 ifa->ifa_addr = (struct sockaddr *)&ia->ia_addr; 524 ifa->ifa_dstaddr = (struct sockaddr *)&ia->ia_dstaddr; 525 ifa->ifa_netmask = (struct sockaddr *)&ia->ia_sockmask; 526 callout_init_rw(&ia->ia_garp_timer, &ifp->if_addr_lock, 527 CALLOUT_RETURNUNLOCKED); 528 529 ia->ia_ifp = ifp; 530 ia->ia_addr = *addr; 531 if (mask->sin_len != 0) { 532 ia->ia_sockmask = *mask; 533 ia->ia_subnetmask = ntohl(ia->ia_sockmask.sin_addr.s_addr); 534 } else { 535 in_addr_t i = ntohl(addr->sin_addr.s_addr); 536 537 /* 538 * If netmask isn't supplied, use historical default. 539 * This is deprecated for interfaces other than loopback 540 * or point-to-point; warn in other cases. In the future 541 * we should return an error rather than warning. 542 */ 543 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) 544 printf("%s: set address: WARNING: network mask " 545 "should be specified; using historical default\n", 546 ifp->if_xname); 547 if (IN_CLASSA(i)) 548 ia->ia_subnetmask = IN_CLASSA_NET; 549 else if (IN_CLASSB(i)) 550 ia->ia_subnetmask = IN_CLASSB_NET; 551 else 552 ia->ia_subnetmask = IN_CLASSC_NET; 553 ia->ia_sockmask.sin_addr.s_addr = htonl(ia->ia_subnetmask); 554 } 555 ia->ia_subnet = ntohl(addr->sin_addr.s_addr) & ia->ia_subnetmask; 556 in_socktrim(&ia->ia_sockmask); 557 558 if (ifp->if_flags & IFF_BROADCAST) { 559 if (broadaddr->sin_len != 0) { 560 ia->ia_broadaddr = *broadaddr; 561 } else if (ia->ia_subnetmask == IN_RFC3021_MASK) { 562 ia->ia_broadaddr.sin_addr.s_addr = INADDR_BROADCAST; 563 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 564 ia->ia_broadaddr.sin_family = AF_INET; 565 } else { 566 ia->ia_broadaddr.sin_addr.s_addr = 567 htonl(ia->ia_subnet | ~ia->ia_subnetmask); 568 ia->ia_broadaddr.sin_len = sizeof(struct sockaddr_in); 569 ia->ia_broadaddr.sin_family = AF_INET; 570 } 571 } 572 573 if (ifp->if_flags & IFF_POINTOPOINT) 574 ia->ia_dstaddr = *dstaddr; 575 576 if (vhid != 0) { 577 error = (*carp_attach_p)(&ia->ia_ifa, vhid); 578 if (error) 579 return (error); 580 } 581 582 /* if_addrhead is already referenced by ifa_alloc() */ 583 IF_ADDR_WLOCK(ifp); 584 CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, ifa, ifa_link); 585 IF_ADDR_WUNLOCK(ifp); 586 587 ifa_ref(ifa); /* in_ifaddrhead */ 588 sx_assert(&in_control_sx, SA_XLOCKED); 589 CK_STAILQ_INSERT_TAIL(&V_in_ifaddrhead, ia, ia_link); 590 CK_LIST_INSERT_HEAD(INADDR_HASH(ia->ia_addr.sin_addr.s_addr), ia, 591 ia_hash); 592 593 /* 594 * Give the interface a chance to initialize 595 * if this is its first address, 596 * and to validate the address if necessary. 597 */ 598 if (ifp->if_ioctl != NULL) { 599 error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia); 600 if (error) 601 goto fail1; 602 } 603 604 /* 605 * Add route for the network. 606 */ 607 if (vhid == 0) { 608 error = in_addprefix(ia); 609 if (error) 610 goto fail1; 611 } 612 613 /* 614 * Add a loopback route to self. 615 */ 616 if (vhid == 0 && ia_need_loopback_route(ia)) { 617 struct in_ifaddr *eia; 618 619 eia = in_localip_more(ia); 620 621 if (eia == NULL) { 622 error = ifa_add_loopback_route((struct ifaddr *)ia, 623 (struct sockaddr *)&ia->ia_addr); 624 if (error) 625 goto fail2; 626 } else 627 ifa_free(&eia->ia_ifa); 628 } 629 630 if (iaIsFirst && (ifp->if_flags & IFF_MULTICAST)) { 631 struct in_addr allhosts_addr; 632 struct in_ifinfo *ii; 633 634 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 635 allhosts_addr.s_addr = htonl(INADDR_ALLHOSTS_GROUP); 636 637 error = in_joingroup(ifp, &allhosts_addr, NULL, 638 &ii->ii_allhosts); 639 } 640 641 /* 642 * Note: we don't need extra reference for ifa, since we called 643 * with sx lock held, and ifaddr can not be deleted in concurrent 644 * thread. 645 */ 646 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, ifa, IFADDR_EVENT_ADD); 647 648 return (error); 649 650 fail2: 651 if (vhid == 0) 652 (void )in_scrubprefix(ia, LLE_STATIC); 653 654 fail1: 655 if (ia->ia_ifa.ifa_carp) 656 (*carp_detach_p)(&ia->ia_ifa, false); 657 658 IF_ADDR_WLOCK(ifp); 659 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 660 IF_ADDR_WUNLOCK(ifp); 661 ifa_free(&ia->ia_ifa); /* if_addrhead */ 662 663 sx_assert(&in_control_sx, SA_XLOCKED); 664 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 665 CK_LIST_REMOVE(ia, ia_hash); 666 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 667 668 return (error); 669 } 670 671 static int 672 in_difaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 673 { 674 const struct ifreq *ifr = (struct ifreq *)data; 675 const struct sockaddr_in *addr = (const struct sockaddr_in *) 676 &ifr->ifr_addr; 677 struct ifaddr *ifa; 678 struct in_ifaddr *ia; 679 bool deleteAny, iaIsLast; 680 int error; 681 682 if (cred != NULL) { 683 error = priv_check_cred(cred, PRIV_NET_DELIFADDR); 684 if (error) 685 return (error); 686 } 687 688 if (addr->sin_len != sizeof(struct sockaddr_in) || 689 addr->sin_family != AF_INET) 690 deleteAny = true; 691 else 692 deleteAny = false; 693 694 iaIsLast = true; 695 ia = NULL; 696 IF_ADDR_WLOCK(ifp); 697 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 698 struct in_ifaddr *it; 699 700 if (ifa->ifa_addr->sa_family != AF_INET) 701 continue; 702 703 it = (struct in_ifaddr *)ifa; 704 if (deleteAny && ia == NULL && (cred == NULL || 705 prison_check_ip4(cred, &it->ia_addr.sin_addr) == 0)) 706 ia = it; 707 708 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 709 (cred == NULL || prison_check_ip4(cred, 710 &addr->sin_addr) == 0)) 711 ia = it; 712 713 if (it != ia) 714 iaIsLast = false; 715 } 716 717 if (ia == NULL) { 718 IF_ADDR_WUNLOCK(ifp); 719 return (EADDRNOTAVAIL); 720 } 721 722 CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link); 723 IF_ADDR_WUNLOCK(ifp); 724 ifa_free(&ia->ia_ifa); /* if_addrhead */ 725 726 sx_assert(&in_control_sx, SA_XLOCKED); 727 CK_STAILQ_REMOVE(&V_in_ifaddrhead, ia, in_ifaddr, ia_link); 728 CK_LIST_REMOVE(ia, ia_hash); 729 730 /* 731 * in_scrubprefix() kills the interface route. 732 */ 733 in_scrubprefix(ia, LLE_STATIC); 734 735 /* 736 * in_ifadown gets rid of all the rest of 737 * the routes. This is not quite the right 738 * thing to do, but at least if we are running 739 * a routing process they will come back. 740 */ 741 in_ifadown(&ia->ia_ifa, 1); 742 743 if (ia->ia_ifa.ifa_carp) 744 (*carp_detach_p)(&ia->ia_ifa, cmd == SIOCAIFADDR); 745 746 /* 747 * If this is the last IPv4 address configured on this 748 * interface, leave the all-hosts group. 749 * No state-change report need be transmitted. 750 */ 751 if (iaIsLast && (ifp->if_flags & IFF_MULTICAST)) { 752 struct in_ifinfo *ii; 753 754 ii = ((struct in_ifinfo *)ifp->if_afdata[AF_INET]); 755 if (ii->ii_allhosts) { 756 (void)in_leavegroup(ii->ii_allhosts, NULL); 757 ii->ii_allhosts = NULL; 758 } 759 } 760 761 IF_ADDR_WLOCK(ifp); 762 if (callout_stop(&ia->ia_garp_timer) == 1) { 763 ifa_free(&ia->ia_ifa); 764 } 765 IF_ADDR_WUNLOCK(ifp); 766 767 EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa, 768 IFADDR_EVENT_DEL); 769 ifa_free(&ia->ia_ifa); /* in_ifaddrhead */ 770 771 return (0); 772 } 773 774 static int 775 in_gifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, struct ucred *cred) 776 { 777 struct in_aliasreq *ifra = (struct in_aliasreq *)data; 778 const struct sockaddr_in *addr = &ifra->ifra_addr; 779 struct epoch_tracker et; 780 struct ifaddr *ifa; 781 struct in_ifaddr *ia; 782 783 /* 784 * ifra_addr must be present and be of INET family. 785 */ 786 if (addr->sin_len != sizeof(struct sockaddr_in) || 787 addr->sin_family != AF_INET) 788 return (EINVAL); 789 790 /* 791 * See whether address exist. 792 */ 793 ia = NULL; 794 NET_EPOCH_ENTER(et); 795 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 796 struct in_ifaddr *it; 797 798 if (ifa->ifa_addr->sa_family != AF_INET) 799 continue; 800 801 it = (struct in_ifaddr *)ifa; 802 if (it->ia_addr.sin_addr.s_addr == addr->sin_addr.s_addr && 803 prison_check_ip4(cred, &addr->sin_addr) == 0) { 804 ia = it; 805 break; 806 } 807 } 808 if (ia == NULL) { 809 NET_EPOCH_EXIT(et); 810 return (EADDRNOTAVAIL); 811 } 812 813 ifra->ifra_mask = ia->ia_sockmask; 814 if ((ifp->if_flags & IFF_POINTOPOINT) && 815 ia->ia_dstaddr.sin_family == AF_INET) 816 ifra->ifra_dstaddr = ia->ia_dstaddr; 817 else if ((ifp->if_flags & IFF_BROADCAST) && 818 ia->ia_broadaddr.sin_family == AF_INET) 819 ifra->ifra_broadaddr = ia->ia_broadaddr; 820 else 821 memset(&ifra->ifra_broadaddr, 0, 822 sizeof(ifra->ifra_broadaddr)); 823 824 NET_EPOCH_EXIT(et); 825 return (0); 826 } 827 828 static int 829 in_match_ifaddr(const struct rtentry *rt, const struct nhop_object *nh, void *arg) 830 { 831 832 if (nh->nh_ifa == (struct ifaddr *)arg) 833 return (1); 834 835 return (0); 836 } 837 838 static int 839 in_handle_prefix_route(uint32_t fibnum, int cmd, 840 struct sockaddr_in *dst, struct sockaddr_in *netmask, struct ifaddr *ifa, 841 struct ifnet *ifp) 842 { 843 844 NET_EPOCH_ASSERT(); 845 846 /* Prepare gateway */ 847 struct sockaddr_dl_short sdl = { 848 .sdl_family = AF_LINK, 849 .sdl_len = sizeof(struct sockaddr_dl_short), 850 .sdl_type = ifa->ifa_ifp->if_type, 851 .sdl_index = ifa->ifa_ifp->if_index, 852 }; 853 854 struct rt_addrinfo info = { 855 .rti_ifa = ifa, 856 .rti_ifp = ifp, 857 .rti_flags = RTF_PINNED | ((netmask != NULL) ? 0 : RTF_HOST), 858 .rti_info = { 859 [RTAX_DST] = (struct sockaddr *)dst, 860 [RTAX_NETMASK] = (struct sockaddr *)netmask, 861 [RTAX_GATEWAY] = (struct sockaddr *)&sdl, 862 }, 863 /* Ensure we delete the prefix IFF prefix ifa matches */ 864 .rti_filter = in_match_ifaddr, 865 .rti_filterdata = ifa, 866 }; 867 868 return (rib_handle_ifaddr_info(fibnum, cmd, &info)); 869 } 870 871 /* 872 * Routing table interaction with interface addresses. 873 * 874 * In general, two types of routes needs to be installed: 875 * a) "interface" or "prefix" route, telling user that the addresses 876 * behind the ifa prefix are reached directly. 877 * b) "loopback" route installed for the ifa address, telling user that 878 * the address belongs to local system. 879 * 880 * Handling for (a) and (b) differs in multi-fib aspects, hence they 881 * are implemented in different functions below. 882 * 883 * The cases above may intersect - /32 interface aliases results in 884 * the same prefix produced by (a) and (b). This blurs the definition 885 * of the "loopback" route and complicate interactions. The interaction 886 * table is defined below. The case numbers are used in the multiple 887 * functions below to refer to the particular test case. 888 * 889 * There can be multiple options: 890 * 1) Adding address with prefix on non-p2p/non-loopback interface. 891 * Example: 192.0.2.1/24. Action: 892 * * add "prefix" route towards 192.0.2.0/24 via @ia interface, 893 * using @ia as an address source. 894 * * add "loopback" route towards 192.0.2.1 via V_loif, saving 895 * @ia ifp in the gateway and using @ia as an address source. 896 * 897 * 2) Adding address with /32 mask to non-p2p/non-loopback interface. 898 * Example: 192.0.2.2/32. Action: 899 * * add "prefix" host route via V_loif, using @ia as an address source. 900 * 901 * 3) Adding address with or without prefix to p2p interface. 902 * Example: 10.0.0.1/24->10.0.0.2. Action: 903 * * add "prefix" host route towards 10.0.0.2 via this interface, using @ia 904 * as an address source. Note: no sense in installing full /24 as the interface 905 * is point-to-point. 906 * * add "loopback" route towards 10.0.9.1 via V_loif, saving 907 * @ia ifp in the gateway and using @ia as an address source. 908 * 909 * 4) Adding address with or without prefix to loopback interface. 910 * Example: 192.0.2.1/24. Action: 911 * * add "prefix" host route via @ia interface, using @ia as an address source. 912 * Note: Skip installing /24 prefix as it would introduce TTL loop 913 * for the traffic destined to these addresses. 914 */ 915 916 /* 917 * Checks if @ia needs to install loopback route to @ia address via 918 * ifa_maintain_loopback_route(). 919 * 920 * Return true on success. 921 */ 922 static bool 923 ia_need_loopback_route(const struct in_ifaddr *ia) 924 { 925 struct ifnet *ifp = ia->ia_ifp; 926 927 /* Case 4: Skip loopback interfaces */ 928 if ((ifp->if_flags & IFF_LOOPBACK) || 929 (ia->ia_addr.sin_addr.s_addr == INADDR_ANY)) 930 return (false); 931 932 /* Clash avoidance: Skip p2p interfaces with both addresses are equal */ 933 if ((ifp->if_flags & IFF_POINTOPOINT) && 934 ia->ia_dstaddr.sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr) 935 return (false); 936 937 /* Case 2: skip /32 prefixes */ 938 if (!(ifp->if_flags & IFF_POINTOPOINT) && 939 (ia->ia_sockmask.sin_addr.s_addr == INADDR_BROADCAST)) 940 return (false); 941 942 return (true); 943 } 944 945 /* 946 * Calculate "prefix" route corresponding to @ia. 947 */ 948 static void 949 ia_getrtprefix(const struct in_ifaddr *ia, struct in_addr *prefix, struct in_addr *mask) 950 { 951 952 if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) { 953 /* Case 3: return host route for dstaddr */ 954 *prefix = ia->ia_dstaddr.sin_addr; 955 mask->s_addr = INADDR_BROADCAST; 956 } else if (ia->ia_ifp->if_flags & IFF_LOOPBACK) { 957 /* Case 4: return host route for ifaddr */ 958 *prefix = ia->ia_addr.sin_addr; 959 mask->s_addr = INADDR_BROADCAST; 960 } else { 961 /* Cases 1,2: return actual ia prefix */ 962 *prefix = ia->ia_addr.sin_addr; 963 *mask = ia->ia_sockmask.sin_addr; 964 prefix->s_addr &= mask->s_addr; 965 } 966 } 967 968 /* 969 * Adds or delete interface "prefix" route corresponding to @ifa. 970 * Returns 0 on success or errno. 971 */ 972 static int 973 in_handle_ifaddr_route(int cmd, struct in_ifaddr *ia) 974 { 975 struct ifaddr *ifa = &ia->ia_ifa; 976 struct in_addr daddr, maddr; 977 struct sockaddr_in *pmask; 978 struct epoch_tracker et; 979 int error; 980 981 ia_getrtprefix(ia, &daddr, &maddr); 982 983 struct sockaddr_in mask = { 984 .sin_family = AF_INET, 985 .sin_len = sizeof(struct sockaddr_in), 986 .sin_addr = maddr, 987 }; 988 989 pmask = (maddr.s_addr != INADDR_BROADCAST) ? &mask : NULL; 990 991 struct sockaddr_in dst = { 992 .sin_family = AF_INET, 993 .sin_len = sizeof(struct sockaddr_in), 994 .sin_addr.s_addr = daddr.s_addr & maddr.s_addr, 995 }; 996 997 struct ifnet *ifp = ia->ia_ifp; 998 999 if ((maddr.s_addr == INADDR_BROADCAST) && 1000 (!(ia->ia_ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)))) { 1001 /* Case 2: host route on broadcast interface */ 1002 ifp = V_loif; 1003 } 1004 1005 uint32_t fibnum = ifa->ifa_ifp->if_fib; 1006 NET_EPOCH_ENTER(et); 1007 error = in_handle_prefix_route(fibnum, cmd, &dst, pmask, ifa, ifp); 1008 NET_EPOCH_EXIT(et); 1009 1010 return (error); 1011 } 1012 1013 /* 1014 * Check if we have a route for the given prefix already. 1015 */ 1016 static bool 1017 in_hasrtprefix(struct in_ifaddr *target) 1018 { 1019 struct epoch_tracker et; 1020 struct in_ifaddr *ia; 1021 struct in_addr prefix, mask, p, m; 1022 bool result = false; 1023 1024 ia_getrtprefix(target, &prefix, &mask); 1025 1026 /* Look for an existing address with the same prefix, mask, and fib */ 1027 NET_EPOCH_ENTER(et); 1028 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1029 ia_getrtprefix(ia, &p, &m); 1030 1031 if (prefix.s_addr != p.s_addr || 1032 mask.s_addr != m.s_addr) 1033 continue; 1034 1035 if (target->ia_ifp->if_fib != ia->ia_ifp->if_fib) 1036 continue; 1037 1038 /* 1039 * If we got a matching prefix route inserted by other 1040 * interface address, we are done here. 1041 */ 1042 if (ia->ia_flags & IFA_ROUTE) { 1043 result = true; 1044 break; 1045 } 1046 } 1047 NET_EPOCH_EXIT(et); 1048 1049 return (result); 1050 } 1051 1052 int 1053 in_addprefix(struct in_ifaddr *target) 1054 { 1055 int error; 1056 1057 if (in_hasrtprefix(target)) { 1058 if (V_nosameprefix) 1059 return (EEXIST); 1060 else { 1061 rt_addrmsg(RTM_ADD, &target->ia_ifa, 1062 target->ia_ifp->if_fib); 1063 return (0); 1064 } 1065 } 1066 1067 /* 1068 * No-one seem to have this prefix route, so we try to insert it. 1069 */ 1070 rt_addrmsg(RTM_ADD, &target->ia_ifa, target->ia_ifp->if_fib); 1071 error = in_handle_ifaddr_route(RTM_ADD, target); 1072 if (!error) 1073 target->ia_flags |= IFA_ROUTE; 1074 return (error); 1075 } 1076 1077 /* 1078 * Removes either all lle entries for given @ia, or lle 1079 * corresponding to @ia address. 1080 */ 1081 static void 1082 in_scrubprefixlle(struct in_ifaddr *ia, int all, u_int flags) 1083 { 1084 struct sockaddr_in addr, mask; 1085 struct sockaddr *saddr, *smask; 1086 struct ifnet *ifp; 1087 1088 saddr = (struct sockaddr *)&addr; 1089 bzero(&addr, sizeof(addr)); 1090 addr.sin_len = sizeof(addr); 1091 addr.sin_family = AF_INET; 1092 smask = (struct sockaddr *)&mask; 1093 bzero(&mask, sizeof(mask)); 1094 mask.sin_len = sizeof(mask); 1095 mask.sin_family = AF_INET; 1096 mask.sin_addr.s_addr = ia->ia_subnetmask; 1097 ifp = ia->ia_ifp; 1098 1099 if (all) { 1100 /* 1101 * Remove all L2 entries matching given prefix. 1102 * Convert address to host representation to avoid 1103 * doing this on every callback. ia_subnetmask is already 1104 * stored in host representation. 1105 */ 1106 addr.sin_addr.s_addr = ntohl(ia->ia_addr.sin_addr.s_addr); 1107 lltable_prefix_free(AF_INET, saddr, smask, flags); 1108 } else { 1109 /* Remove interface address only */ 1110 addr.sin_addr.s_addr = ia->ia_addr.sin_addr.s_addr; 1111 lltable_delete_addr(LLTABLE(ifp), LLE_IFADDR, saddr); 1112 } 1113 } 1114 1115 /* 1116 * If there is no other address in the system that can serve a route to the 1117 * same prefix, remove the route. Hand over the route to the new address 1118 * otherwise. 1119 */ 1120 int 1121 in_scrubprefix(struct in_ifaddr *target, u_int flags) 1122 { 1123 struct epoch_tracker et; 1124 struct in_ifaddr *ia; 1125 struct in_addr prefix, mask, p, m; 1126 int error = 0; 1127 1128 /* 1129 * Remove the loopback route to the interface address. 1130 */ 1131 if (ia_need_loopback_route(target) && (flags & LLE_STATIC)) { 1132 struct in_ifaddr *eia; 1133 1134 eia = in_localip_more(target); 1135 1136 if (eia != NULL) { 1137 error = ifa_switch_loopback_route((struct ifaddr *)eia, 1138 (struct sockaddr *)&target->ia_addr); 1139 ifa_free(&eia->ia_ifa); 1140 } else { 1141 error = ifa_del_loopback_route((struct ifaddr *)target, 1142 (struct sockaddr *)&target->ia_addr); 1143 } 1144 } 1145 1146 ia_getrtprefix(target, &prefix, &mask); 1147 1148 if ((target->ia_flags & IFA_ROUTE) == 0) { 1149 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1150 1151 /* 1152 * Removing address from !IFF_UP interface or 1153 * prefix which exists on other interface (along with route). 1154 * No entries should exist here except target addr. 1155 * Given that, delete this entry only. 1156 */ 1157 in_scrubprefixlle(target, 0, flags); 1158 return (0); 1159 } 1160 1161 NET_EPOCH_ENTER(et); 1162 CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1163 ia_getrtprefix(ia, &p, &m); 1164 1165 if (prefix.s_addr != p.s_addr || 1166 mask.s_addr != m.s_addr) 1167 continue; 1168 1169 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) 1170 continue; 1171 1172 /* 1173 * If we got a matching prefix address, move IFA_ROUTE and 1174 * the route itself to it. Make sure that routing daemons 1175 * get a heads-up. 1176 */ 1177 if ((ia->ia_flags & IFA_ROUTE) == 0) { 1178 ifa_ref(&ia->ia_ifa); 1179 NET_EPOCH_EXIT(et); 1180 error = in_handle_ifaddr_route(RTM_DELETE, target); 1181 if (error == 0) 1182 target->ia_flags &= ~IFA_ROUTE; 1183 else 1184 log(LOG_INFO, "in_scrubprefix: err=%d, old prefix delete failed\n", 1185 error); 1186 /* Scrub all entries IFF interface is different */ 1187 in_scrubprefixlle(target, target->ia_ifp != ia->ia_ifp, 1188 flags); 1189 error = in_handle_ifaddr_route(RTM_ADD, ia); 1190 if (error == 0) 1191 ia->ia_flags |= IFA_ROUTE; 1192 else 1193 log(LOG_INFO, "in_scrubprefix: err=%d, new prefix add failed\n", 1194 error); 1195 ifa_free(&ia->ia_ifa); 1196 return (error); 1197 } 1198 } 1199 NET_EPOCH_EXIT(et); 1200 1201 /* 1202 * remove all L2 entries on the given prefix 1203 */ 1204 in_scrubprefixlle(target, 1, flags); 1205 1206 /* 1207 * As no-one seem to have this prefix, we can remove the route. 1208 */ 1209 rt_addrmsg(RTM_DELETE, &target->ia_ifa, target->ia_ifp->if_fib); 1210 error = in_handle_ifaddr_route(RTM_DELETE, target); 1211 if (error == 0) 1212 target->ia_flags &= ~IFA_ROUTE; 1213 else 1214 log(LOG_INFO, "in_scrubprefix: err=%d, prefix delete failed\n", error); 1215 return (error); 1216 } 1217 1218 void 1219 in_ifscrub_all(void) 1220 { 1221 struct ifnet *ifp; 1222 struct ifaddr *ifa, *nifa; 1223 struct ifaliasreq ifr; 1224 1225 IFNET_RLOCK(); 1226 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1227 /* Cannot lock here - lock recursion. */ 1228 /* NET_EPOCH_ENTER(et); */ 1229 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) { 1230 if (ifa->ifa_addr->sa_family != AF_INET) 1231 continue; 1232 1233 /* 1234 * This is ugly but the only way for legacy IP to 1235 * cleanly remove addresses and everything attached. 1236 */ 1237 bzero(&ifr, sizeof(ifr)); 1238 ifr.ifra_addr = *ifa->ifa_addr; 1239 if (ifa->ifa_dstaddr) 1240 ifr.ifra_broadaddr = *ifa->ifa_dstaddr; 1241 (void)in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, 1242 ifp, NULL); 1243 } 1244 /* NET_EPOCH_EXIT(et); */ 1245 in_purgemaddrs(ifp); 1246 igmp_domifdetach(ifp); 1247 } 1248 IFNET_RUNLOCK(); 1249 } 1250 1251 int 1252 in_ifaddr_broadcast(struct in_addr in, struct in_ifaddr *ia) 1253 { 1254 1255 return ((in.s_addr == ia->ia_broadaddr.sin_addr.s_addr || 1256 /* 1257 * Optionally check for old-style (host 0) broadcast, but 1258 * taking into account that RFC 3021 obsoletes it. 1259 */ 1260 (V_broadcast_lowest && ia->ia_subnetmask != IN_RFC3021_MASK && 1261 ntohl(in.s_addr) == ia->ia_subnet)) && 1262 /* 1263 * Check for an all one subnetmask. These 1264 * only exist when an interface gets a secondary 1265 * address. 1266 */ 1267 ia->ia_subnetmask != (u_long)0xffffffff); 1268 } 1269 1270 /* 1271 * Return 1 if the address might be a local broadcast address. 1272 */ 1273 int 1274 in_broadcast(struct in_addr in, struct ifnet *ifp) 1275 { 1276 struct ifaddr *ifa; 1277 int found; 1278 1279 NET_EPOCH_ASSERT(); 1280 1281 if (in.s_addr == INADDR_BROADCAST || 1282 in.s_addr == INADDR_ANY) 1283 return (1); 1284 if ((ifp->if_flags & IFF_BROADCAST) == 0) 1285 return (0); 1286 found = 0; 1287 /* 1288 * Look through the list of addresses for a match 1289 * with a broadcast address. 1290 */ 1291 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 1292 if (ifa->ifa_addr->sa_family == AF_INET && 1293 in_ifaddr_broadcast(in, (struct in_ifaddr *)ifa)) { 1294 found = 1; 1295 break; 1296 } 1297 return (found); 1298 } 1299 1300 /* 1301 * On interface removal, clean up IPv4 data structures hung off of the ifnet. 1302 */ 1303 void 1304 in_ifdetach(struct ifnet *ifp) 1305 { 1306 IN_MULTI_LOCK(); 1307 in_pcbpurgeif0(&V_ripcbinfo, ifp); 1308 in_pcbpurgeif0(&V_udbinfo, ifp); 1309 in_pcbpurgeif0(&V_ulitecbinfo, ifp); 1310 in_purgemaddrs(ifp); 1311 IN_MULTI_UNLOCK(); 1312 1313 /* 1314 * Make sure all multicast deletions invoking if_ioctl() are 1315 * completed before returning. Else we risk accessing a freed 1316 * ifnet structure pointer. 1317 */ 1318 inm_release_wait(NULL); 1319 } 1320 1321 static void 1322 in_ifnet_event(void *arg __unused, struct ifnet *ifp, int event) 1323 { 1324 struct epoch_tracker et; 1325 struct ifaddr *ifa; 1326 struct in_ifaddr *ia; 1327 int error; 1328 1329 NET_EPOCH_ENTER(et); 1330 switch (event) { 1331 case IFNET_EVENT_DOWN: 1332 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1333 if (ifa->ifa_addr->sa_family != AF_INET) 1334 continue; 1335 ia = (struct in_ifaddr *)ifa; 1336 if ((ia->ia_flags & IFA_ROUTE) == 0) 1337 continue; 1338 ifa_ref(ifa); 1339 /* 1340 * in_scrubprefix() kills the interface route. 1341 */ 1342 in_scrubprefix(ia, 0); 1343 /* 1344 * in_ifadown gets rid of all the rest of the 1345 * routes. This is not quite the right thing 1346 * to do, but at least if we are running a 1347 * routing process they will come back. 1348 */ 1349 in_ifadown(ifa, 0); 1350 ifa_free(ifa); 1351 } 1352 break; 1353 1354 case IFNET_EVENT_UP: 1355 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1356 if (ifa->ifa_addr->sa_family != AF_INET) 1357 continue; 1358 ia = (struct in_ifaddr *)ifa; 1359 if (ia->ia_flags & IFA_ROUTE) 1360 continue; 1361 ifa_ref(ifa); 1362 error = ifa_del_loopback_route(ifa, ifa->ifa_addr); 1363 rt_addrmsg(RTM_ADD, ifa, ifa->ifa_ifp->if_fib); 1364 error = in_handle_ifaddr_route(RTM_ADD, ia); 1365 if (error == 0) 1366 ia->ia_flags |= IFA_ROUTE; 1367 error = ifa_add_loopback_route(ifa, ifa->ifa_addr); 1368 ifa_free(ifa); 1369 } 1370 break; 1371 } 1372 NET_EPOCH_EXIT(et); 1373 } 1374 EVENTHANDLER_DEFINE(ifnet_event, in_ifnet_event, NULL, EVENTHANDLER_PRI_ANY); 1375 1376 /* 1377 * Delete all IPv4 multicast address records, and associated link-layer 1378 * multicast address records, associated with ifp. 1379 * XXX It looks like domifdetach runs AFTER the link layer cleanup. 1380 * XXX This should not race with ifma_protospec being set during 1381 * a new allocation, if it does, we have bigger problems. 1382 */ 1383 static void 1384 in_purgemaddrs(struct ifnet *ifp) 1385 { 1386 struct epoch_tracker et; 1387 struct in_multi_head purgeinms; 1388 struct in_multi *inm; 1389 struct ifmultiaddr *ifma; 1390 1391 SLIST_INIT(&purgeinms); 1392 IN_MULTI_LIST_LOCK(); 1393 1394 /* 1395 * Extract list of in_multi associated with the detaching ifp 1396 * which the PF_INET layer is about to release. 1397 * We need to do this as IF_ADDR_LOCK() may be re-acquired 1398 * by code further down. 1399 */ 1400 IF_ADDR_WLOCK(ifp); 1401 NET_EPOCH_ENTER(et); 1402 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1403 inm = inm_ifmultiaddr_get_inm(ifma); 1404 if (inm == NULL) 1405 continue; 1406 inm_rele_locked(&purgeinms, inm); 1407 } 1408 NET_EPOCH_EXIT(et); 1409 IF_ADDR_WUNLOCK(ifp); 1410 1411 inm_release_list_deferred(&purgeinms); 1412 igmp_ifdetach(ifp); 1413 IN_MULTI_LIST_UNLOCK(); 1414 } 1415 1416 struct in_llentry { 1417 struct llentry base; 1418 }; 1419 1420 #define IN_LLTBL_DEFAULT_HSIZE 32 1421 #define IN_LLTBL_HASH(k, h) \ 1422 (((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1)) 1423 1424 /* 1425 * Do actual deallocation of @lle. 1426 */ 1427 static void 1428 in_lltable_destroy_lle_unlocked(epoch_context_t ctx) 1429 { 1430 struct llentry *lle; 1431 1432 lle = __containerof(ctx, struct llentry, lle_epoch_ctx); 1433 LLE_LOCK_DESTROY(lle); 1434 LLE_REQ_DESTROY(lle); 1435 free(lle, M_LLTABLE); 1436 } 1437 1438 /* 1439 * Called by LLE_FREE_LOCKED when number of references 1440 * drops to zero. 1441 */ 1442 static void 1443 in_lltable_destroy_lle(struct llentry *lle) 1444 { 1445 1446 LLE_WUNLOCK(lle); 1447 NET_EPOCH_CALL(in_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx); 1448 } 1449 1450 static struct llentry * 1451 in_lltable_new(struct in_addr addr4, u_int flags) 1452 { 1453 struct in_llentry *lle; 1454 1455 lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_NOWAIT | M_ZERO); 1456 if (lle == NULL) /* NB: caller generates msg */ 1457 return NULL; 1458 1459 /* 1460 * For IPv4 this will trigger "arpresolve" to generate 1461 * an ARP request. 1462 */ 1463 lle->base.la_expire = time_uptime; /* mark expired */ 1464 lle->base.r_l3addr.addr4 = addr4; 1465 lle->base.lle_refcnt = 1; 1466 lle->base.lle_free = in_lltable_destroy_lle; 1467 LLE_LOCK_INIT(&lle->base); 1468 LLE_REQ_INIT(&lle->base); 1469 callout_init(&lle->base.lle_timer, 1); 1470 1471 return (&lle->base); 1472 } 1473 1474 #define IN_ARE_MASKED_ADDR_EQUAL(d, a, m) ( \ 1475 ((((d).s_addr ^ (a).s_addr) & (m).s_addr)) == 0 ) 1476 1477 static int 1478 in_lltable_match_prefix(const struct sockaddr *saddr, 1479 const struct sockaddr *smask, u_int flags, struct llentry *lle) 1480 { 1481 struct in_addr addr, mask, lle_addr; 1482 1483 addr = ((const struct sockaddr_in *)saddr)->sin_addr; 1484 mask = ((const struct sockaddr_in *)smask)->sin_addr; 1485 lle_addr.s_addr = ntohl(lle->r_l3addr.addr4.s_addr); 1486 1487 if (IN_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0) 1488 return (0); 1489 1490 if (lle->la_flags & LLE_IFADDR) { 1491 /* 1492 * Delete LLE_IFADDR records IFF address & flag matches. 1493 * Note that addr is the interface address within prefix 1494 * being matched. 1495 * Note also we should handle 'ifdown' cases without removing 1496 * ifaddr macs. 1497 */ 1498 if (addr.s_addr == lle_addr.s_addr && (flags & LLE_STATIC) != 0) 1499 return (1); 1500 return (0); 1501 } 1502 1503 /* flags & LLE_STATIC means deleting both dynamic and static entries */ 1504 if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)) 1505 return (1); 1506 1507 return (0); 1508 } 1509 1510 static void 1511 in_lltable_free_entry(struct lltable *llt, struct llentry *lle) 1512 { 1513 size_t pkts_dropped; 1514 1515 LLE_WLOCK_ASSERT(lle); 1516 KASSERT(llt != NULL, ("lltable is NULL")); 1517 1518 /* Unlink entry from table if not already */ 1519 if ((lle->la_flags & LLE_LINKED) != 0) { 1520 IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp); 1521 lltable_unlink_entry(llt, lle); 1522 } 1523 1524 /* Drop hold queue */ 1525 pkts_dropped = llentry_free(lle); 1526 ARPSTAT_ADD(dropped, pkts_dropped); 1527 } 1528 1529 static int 1530 in_lltable_rtcheck(struct ifnet *ifp, u_int flags, const struct sockaddr *l3addr) 1531 { 1532 struct nhop_object *nh; 1533 struct in_addr addr; 1534 1535 KASSERT(l3addr->sa_family == AF_INET, 1536 ("sin_family %d", l3addr->sa_family)); 1537 1538 addr = ((const struct sockaddr_in *)l3addr)->sin_addr; 1539 1540 nh = fib4_lookup(ifp->if_fib, addr, 0, NHR_NONE, 0); 1541 if (nh == NULL) 1542 return (EINVAL); 1543 1544 /* 1545 * If the gateway for an existing host route matches the target L3 1546 * address, which is a special route inserted by some implementation 1547 * such as MANET, and the interface is of the correct type, then 1548 * allow for ARP to proceed. 1549 */ 1550 if (nh->nh_flags & NHF_GATEWAY) { 1551 if (!(nh->nh_flags & NHF_HOST) || nh->nh_ifp->if_type != IFT_ETHER || 1552 (nh->nh_ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) != 0 || 1553 memcmp(nh->gw_sa.sa_data, l3addr->sa_data, 1554 sizeof(in_addr_t)) != 0) { 1555 return (EINVAL); 1556 } 1557 } 1558 1559 /* 1560 * Make sure that at least the destination address is covered 1561 * by the route. This is for handling the case where 2 or more 1562 * interfaces have the same prefix. An incoming packet arrives 1563 * on one interface and the corresponding outgoing packet leaves 1564 * another interface. 1565 */ 1566 if ((nh->nh_ifp != ifp) && (nh->nh_flags & NHF_HOST) == 0) { 1567 struct in_ifaddr *ia = (struct in_ifaddr *)ifaof_ifpforaddr(l3addr, ifp); 1568 struct in_addr dst_addr, mask_addr; 1569 1570 if (ia == NULL) 1571 return (EINVAL); 1572 1573 /* 1574 * ifaof_ifpforaddr() returns _best matching_ IFA. 1575 * It is possible that ifa prefix does not cover our address. 1576 * Explicitly verify and fail if that's the case. 1577 */ 1578 dst_addr = IA_SIN(ia)->sin_addr; 1579 mask_addr.s_addr = htonl(ia->ia_subnetmask); 1580 1581 if (!IN_ARE_MASKED_ADDR_EQUAL(dst_addr, addr, mask_addr)) 1582 return (EINVAL); 1583 } 1584 1585 return (0); 1586 } 1587 1588 static inline uint32_t 1589 in_lltable_hash_dst(const struct in_addr dst, uint32_t hsize) 1590 { 1591 1592 return (IN_LLTBL_HASH(dst.s_addr, hsize)); 1593 } 1594 1595 static uint32_t 1596 in_lltable_hash(const struct llentry *lle, uint32_t hsize) 1597 { 1598 1599 return (in_lltable_hash_dst(lle->r_l3addr.addr4, hsize)); 1600 } 1601 1602 static void 1603 in_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa) 1604 { 1605 struct sockaddr_in *sin; 1606 1607 sin = (struct sockaddr_in *)sa; 1608 bzero(sin, sizeof(*sin)); 1609 sin->sin_family = AF_INET; 1610 sin->sin_len = sizeof(*sin); 1611 sin->sin_addr = lle->r_l3addr.addr4; 1612 } 1613 1614 static inline struct llentry * 1615 in_lltable_find_dst(struct lltable *llt, struct in_addr dst) 1616 { 1617 struct llentry *lle; 1618 struct llentries *lleh; 1619 u_int hashidx; 1620 1621 hashidx = in_lltable_hash_dst(dst, llt->llt_hsize); 1622 lleh = &llt->lle_head[hashidx]; 1623 CK_LIST_FOREACH(lle, lleh, lle_next) { 1624 if (lle->la_flags & LLE_DELETED) 1625 continue; 1626 if (lle->r_l3addr.addr4.s_addr == dst.s_addr) 1627 break; 1628 } 1629 1630 return (lle); 1631 } 1632 1633 static void 1634 in_lltable_delete_entry(struct lltable *llt, struct llentry *lle) 1635 { 1636 1637 lle->la_flags |= LLE_DELETED; 1638 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED); 1639 #ifdef DIAGNOSTIC 1640 log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); 1641 #endif 1642 llentry_free(lle); 1643 } 1644 1645 static struct llentry * 1646 in_lltable_alloc(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1647 { 1648 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1649 struct ifnet *ifp = llt->llt_ifp; 1650 struct llentry *lle; 1651 char linkhdr[LLE_MAX_LINKHDR]; 1652 size_t linkhdrsize; 1653 int lladdr_off; 1654 1655 KASSERT(l3addr->sa_family == AF_INET, 1656 ("sin_family %d", l3addr->sa_family)); 1657 1658 /* 1659 * A route that covers the given address must have 1660 * been installed 1st because we are doing a resolution, 1661 * verify this. 1662 */ 1663 if (!(flags & LLE_IFADDR) && 1664 in_lltable_rtcheck(ifp, flags, l3addr) != 0) 1665 return (NULL); 1666 1667 lle = in_lltable_new(sin->sin_addr, flags); 1668 if (lle == NULL) { 1669 log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); 1670 return (NULL); 1671 } 1672 lle->la_flags = flags; 1673 if (flags & LLE_STATIC) 1674 lle->r_flags |= RLLE_VALID; 1675 if ((flags & LLE_IFADDR) == LLE_IFADDR) { 1676 linkhdrsize = LLE_MAX_LINKHDR; 1677 if (lltable_calc_llheader(ifp, AF_INET, IF_LLADDR(ifp), 1678 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1679 in_lltable_free_entry(llt, lle); 1680 return (NULL); 1681 } 1682 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, 1683 lladdr_off); 1684 lle->la_flags |= LLE_STATIC; 1685 lle->r_flags |= (RLLE_VALID | RLLE_IFADDR); 1686 lle->la_expire = 0; 1687 } 1688 1689 return (lle); 1690 } 1691 1692 /* 1693 * Return NULL if not found or marked for deletion. 1694 * If found return lle read locked. 1695 */ 1696 static struct llentry * 1697 in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) 1698 { 1699 const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; 1700 struct llentry *lle; 1701 1702 IF_AFDATA_LOCK_ASSERT(llt->llt_ifp); 1703 KASSERT(l3addr->sa_family == AF_INET, 1704 ("sin_family %d", l3addr->sa_family)); 1705 KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) != 1706 (LLE_UNLOCKED | LLE_EXCLUSIVE), 1707 ("wrong lle request flags: %#x", flags)); 1708 1709 lle = in_lltable_find_dst(llt, sin->sin_addr); 1710 if (lle == NULL) 1711 return (NULL); 1712 if (flags & LLE_UNLOCKED) 1713 return (lle); 1714 1715 if (flags & LLE_EXCLUSIVE) 1716 LLE_WLOCK(lle); 1717 else 1718 LLE_RLOCK(lle); 1719 1720 /* 1721 * If the afdata lock is not held, the LLE may have been unlinked while 1722 * we were blocked on the LLE lock. Check for this case. 1723 */ 1724 if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) { 1725 if (flags & LLE_EXCLUSIVE) 1726 LLE_WUNLOCK(lle); 1727 else 1728 LLE_RUNLOCK(lle); 1729 return (NULL); 1730 } 1731 return (lle); 1732 } 1733 1734 static int 1735 in_lltable_dump_entry(struct lltable *llt, struct llentry *lle, 1736 struct sysctl_req *wr) 1737 { 1738 struct ifnet *ifp = llt->llt_ifp; 1739 /* XXX stack use */ 1740 struct { 1741 struct rt_msghdr rtm; 1742 struct sockaddr_in sin; 1743 struct sockaddr_dl sdl; 1744 } arpc; 1745 struct sockaddr_dl *sdl; 1746 int error; 1747 1748 bzero(&arpc, sizeof(arpc)); 1749 /* skip deleted entries */ 1750 if ((lle->la_flags & LLE_DELETED) == LLE_DELETED) 1751 return (0); 1752 /* Skip if jailed and not a valid IP of the prison. */ 1753 lltable_fill_sa_entry(lle,(struct sockaddr *)&arpc.sin); 1754 if (prison_if(wr->td->td_ucred, (struct sockaddr *)&arpc.sin) != 0) 1755 return (0); 1756 /* 1757 * produce a msg made of: 1758 * struct rt_msghdr; 1759 * struct sockaddr_in; (IPv4) 1760 * struct sockaddr_dl; 1761 */ 1762 arpc.rtm.rtm_msglen = sizeof(arpc); 1763 arpc.rtm.rtm_version = RTM_VERSION; 1764 arpc.rtm.rtm_type = RTM_GET; 1765 arpc.rtm.rtm_flags = RTF_UP; 1766 arpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY; 1767 1768 /* publish */ 1769 if (lle->la_flags & LLE_PUB) 1770 arpc.rtm.rtm_flags |= RTF_ANNOUNCE; 1771 1772 sdl = &arpc.sdl; 1773 sdl->sdl_family = AF_LINK; 1774 sdl->sdl_len = sizeof(*sdl); 1775 sdl->sdl_index = ifp->if_index; 1776 sdl->sdl_type = ifp->if_type; 1777 if ((lle->la_flags & LLE_VALID) == LLE_VALID) { 1778 sdl->sdl_alen = ifp->if_addrlen; 1779 bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); 1780 } else { 1781 sdl->sdl_alen = 0; 1782 bzero(LLADDR(sdl), ifp->if_addrlen); 1783 } 1784 1785 arpc.rtm.rtm_rmx.rmx_expire = 1786 lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; 1787 arpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA); 1788 if (lle->la_flags & LLE_STATIC) 1789 arpc.rtm.rtm_flags |= RTF_STATIC; 1790 if (lle->la_flags & LLE_IFADDR) 1791 arpc.rtm.rtm_flags |= RTF_PINNED; 1792 arpc.rtm.rtm_index = ifp->if_index; 1793 error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); 1794 1795 return (error); 1796 } 1797 1798 static void 1799 in_lltable_post_resolved(struct lltable *llt, struct llentry *lle) 1800 { 1801 struct ifnet *ifp = llt->llt_ifp; 1802 1803 /* gratuitous ARP */ 1804 if ((lle->la_flags & LLE_PUB) != 0) 1805 arprequest(ifp, &lle->r_l3addr.addr4, &lle->r_l3addr.addr4, 1806 lle->ll_addr); 1807 } 1808 1809 static struct lltable * 1810 in_lltattach(struct ifnet *ifp) 1811 { 1812 struct lltable *llt; 1813 1814 llt = lltable_allocate_htbl(IN_LLTBL_DEFAULT_HSIZE); 1815 llt->llt_af = AF_INET; 1816 llt->llt_ifp = ifp; 1817 1818 llt->llt_lookup = in_lltable_lookup; 1819 llt->llt_alloc_entry = in_lltable_alloc; 1820 llt->llt_delete_entry = in_lltable_delete_entry; 1821 llt->llt_dump_entry = in_lltable_dump_entry; 1822 llt->llt_hash = in_lltable_hash; 1823 llt->llt_fill_sa_entry = in_lltable_fill_sa_entry; 1824 llt->llt_free_entry = in_lltable_free_entry; 1825 llt->llt_match_prefix = in_lltable_match_prefix; 1826 llt->llt_mark_used = llentry_mark_used; 1827 llt->llt_post_resolved = in_lltable_post_resolved; 1828 lltable_link(llt); 1829 1830 return (llt); 1831 } 1832 1833 struct lltable * 1834 in_lltable_get(struct ifnet *ifp) 1835 { 1836 struct lltable *llt = NULL; 1837 1838 void *afdata_ptr = ifp->if_afdata[AF_INET]; 1839 if (afdata_ptr != NULL) 1840 llt = ((struct in_ifinfo *)afdata_ptr)->ii_llt; 1841 return (llt); 1842 } 1843 1844 void * 1845 in_domifattach(struct ifnet *ifp) 1846 { 1847 struct in_ifinfo *ii; 1848 1849 ii = malloc(sizeof(struct in_ifinfo), M_IFADDR, M_WAITOK|M_ZERO); 1850 1851 ii->ii_llt = in_lltattach(ifp); 1852 ii->ii_igmp = igmp_domifattach(ifp); 1853 1854 return (ii); 1855 } 1856 1857 void 1858 in_domifdetach(struct ifnet *ifp, void *aux) 1859 { 1860 struct in_ifinfo *ii = (struct in_ifinfo *)aux; 1861 1862 igmp_domifdetach(ifp); 1863 lltable_free(ii->ii_llt); 1864 free(ii, M_IFADDR); 1865 } 1866